Datasets:

CShorten

/

CORD-19-prototype

like 0

Prognosis of hospitalized patients with 2009 H1N1 influenza in Spain: influence of neuraminidase inhibitors

BACKGROUND: The H1N1 influenza pandemic strain has been associated with a poor prognosis in hospitalized patients. The present report evaluates the factors influencing prognosis. METHODS: A total of 813 patients hospitalized with H1N1 influenza in 36 hospitals (nationwide) in Spain were analysed. Detailed histories of variables preceding hospital admission were obtained by interview, validating data on medications and vaccine with their attending physicians. Data on treatment and complications during hospital stay were recorded. As definition of poor outcome, the endpoints of death and admission to intensive care were combined; and as a further outcome, length of stay was used. RESULTS: The mean age was 38.5 years (SD 22.8 years). There were 10 deaths and 79 admissions to intensive care (combined, 88). The use of neuraminidase inhibitors was reported by 495 patients (60.9%). The variables significantly associated with a poor outcome were diabetes (OR = 2.21, 95% CI = 1.21–4.02), corticosteroid therapy (OR = 3.37, 95% CI = 1.39–8.20) and use of histamine-2 receptor antagonists (OR = 2.68, 95% CI = 1.14–6.36), while the use of neuraminidase inhibitors (OR = 0.57, 95% CI = 0.34–0.94) was protective. Neuraminidase inhibitors within the first 2 days after the influenza onset reduced hospital stay by a mean of 1.9 days (95% CI = 4.7–6.6). CONCLUSIONS: The use of neuraminidase inhibitors decreases the length of hospital stay and admission to intensive care and/or death.

Influenza A pandemic H1N1 2009 virus infections began to spread in Spain during spring 2009. Reports suggested high mortality in children and adults associated with the new virus in Mexico 1,2 and Argentina, 3 as well as in previously healthy young people. Analysis of cases hospitalized in the USA showed a mortality rate of 7%, with 25% of patients being admitted to the intensive care unit (ICU). 4 A study of 32 patients infected with the pandemic virus strain admitted to Spanish ICUs found a mortality rate of 25%, somewhat lower than in Latin American countries. 5 These findings suggest that the H1N1 virus is more virulent than previous strains. As there was no specific targeted vaccine giving protection against the H1N1 influenza virus available at the beginning of the outbreak, health authorities began to recommend administration of neuraminidase inhibitors to reduce transmission and/or complications. Various studies have suggested that these drugs are also effective in reducing the severity of the infection. 4, 6, 7 We reviewed nationwide Spanish data on hospitalized patients with 2009 H1N1 influenza A in order to: (i) evaluate the frequency of adverse outcomes during hospitalization; and (ii) identify the factors influencing poor/good outcome, including the use of neuraminidase inhibitors shortly after the onset of symptoms. We carried out a multicentre study in 36 hospitals from seven Spanish regions (Andalusia, Catalonia, Castile and Leon, Madrid, Navarre, the Basque Country and Valencia). Between July 2009 and February 2010 we selected hospitalized patients with influenza syndrome, acute respiratory infection, septic shock or multiple organ failure in whom influenza virus A (H1N1) 2009 infection was confirmed by real-time reversetranscription PCR (RT-PCR) from nasopharyngeal swabs; haemagglutinin (HA) sequencing was performed. We excluded patients who had nosocomial infection, defined as pandemic virus infection in a patient that appears ≥48 h after admission for another cause. All information collected was treated as confidential, in strict observance of legislation for observational studies. The study was approved by the Ethics Committees of the hospitals involved, following the Declaration of Helsinki principles. Written informed consent was obtained from all patients included in the study. During the pandemic flu all patients suspected of having the disease, either in outpatient clinics or hospitals, were diagnosed by RT-PCR of samples from nasopharyngeal swabs. Within the next 48 h, hospitalized patients were interviewed at the centre. Of these, 23 rejected participation and 12 were excluded because flu had been acquired after hospital admission. The following demographic variables and pre-existing medical conditions were recorded for all study participants: age, sex, ethnicity, educational level, smoking, alcoholism, pregnancy in women aged 15 -49 years, history of pneumonia in the previous two years, chronic obstructive pulmonary disease (COPD), asthma, cardiovascular disease, renal failure, diabetes, HIV infection, disabling neurological disease, cancer, transplantation, morbid obesity (body mass index ≥40), use of neuraminidase inhibitors before hospital admission (and their timing relative to the onset of symptoms, verified after contacting the prescribing general practitioner), use of other medications in the 90 days before hospital admission (corticosteroids, antibiotics etc.) and treatment received during hospitalization (medications, catheters and mechanical ventilation). For each vaccine, a case was considered vaccinated if the vaccine had been received ≥14 days before the onset of symptoms. Data were collected during hospital admission and the clinical chart was also reviewed after discharge. The outcome variables were admission to an ICU, in-hospital death and length of hospital stay (in days). Given that the number of deaths was very low, a combined endpoint was classified as 'poor outcome': ICU admission and/or in-hospital death. Bivariate comparisons were made using Pearson's x 2 test for categorical variables and Student's t-test for continuous variables. As a measure of association, the relative risk (OR) and 95% CI were calculated. Logistic regression was applied in the multivariate analysis for dichotomous adverse outcomes. To determine the variables to be included in the multivariate analysis, the procedure described by Sun et al. 8 was followed. Intermediate variables were discarded. We ran two stepwise models, one backward and another forward, including variables with P, 0.2. 9, 10 We constructed a list of predictors of mortality identified in other studies. Using information from stepwise models and the list of predictors, a saturated model was built, and by using a heuristic approach, variables that did not change the coefficient of the bundles by more than 10% were discarded, in order to construct a parsimonious model retaining all important confounders. To analyse the impact of different variables on the length of hospital stay, patients who died were excluded from these analyses. Given that hospital stay did not follow the normal curve, natural logarithms were used. Firstly, to select potential variables related to length of stay, we used Cox regression in the same fashion as described above for the logistic regression analysis. The variables selected by this model were tested by including other potential candidates according to the logistic regression analyses. Secondly, an analysis of covariance was applied to estimate the adjusted means of hospital length of stay. All analyses were made using the Stata 10/SE package (College Station, TX, USA). There were a total of 813 patients [410 (50.4%) were female, of which 51 (12%) were pregnant]. The mean age was 38.5 years (SD 22.8) and 24% were aged ,18 years. The use of neuraminidase inhibitors was reported by 495 patients (60.9%), with oseltamivir being administered in all cases but two (zanamivir). During hospitalization, 79 patients (9.7%) were admitted to the ICU and 10 died (1.2%), of whom 9 were not receiving intensive care. No death occurred in pregnant women, of whom only one was admitted to the ICU. The timings of the use of neuraminidase inhibitors before hospital admission were: 332 patients in the first 24 h after the onset of symptoms, 97 between 24 -48 h and 66 after 48 h ( Table 1) . The relationship between study variables and ICU admission/ in-hospital death is shown in Table 2 . In the univariate analyses, age, most comorbidities (COPD, diabetes, liver failure and cardiovascular disease), ex-smoking, corticosteroid therapy and histamine-2 receptor antagonists were associated with an adverse outcome during hospitalization. In the multivariate models, the variables significantly associated with a poor outcome were diabetes (OR¼ 2.21, 95% CI¼ 1.21 -4.02), corticosteroid therapy (OR¼ 3.37, 95% CI ¼ 1.39-8.20) and use of histamine-2 receptor antagonists (OR ¼ 2.68, 95% CI ¼ 1.14 -6.36). Use of neuraminidase inhibitors was protective (OR ¼0.57, 95% CI ¼ 0.34 -0.94). Pneumonia at admission, COPD, ex-smoking and liver failure showed a trend to association. Delgado-Rodríguez et al. The trend analysis for age in the multivariable analysis yielded a P value of 0.11, with advanced age associated with a higher risk of adverse outcome. When the timing of treatment with neuraminidase inhibitors after the onset of influenza was analysed, the benefit was confined to administration within the first 48 h after the onset of symptoms. Table 3 shows the variables associated with length of hospital stay. The use of neuraminidase inhibitors within the first 2 days after the onset of influenza reduced hospital stay by a mean of 1.9 days (from 6.6 to 4.7, P,0.001), whereas delayed administration was associated with an increase in hospital stay. Pneumonia diagnosed at admission was clearly associated with longer hospital stay, as were comorbidities (COPD, neurological impairment and cardiovascular disease) and some therapies (proton pump inhibitors). We found that traditional risk factors associated with hospitalization in patients with influenza (COPD and corticosteroid therapy before admission) were also found in our patients. Likewise, the use of neuraminidase inhibitors reduced the probability of adverse outcomes during hospital stay and significantly shortened the length of stay. This study is observational and can be affected by several limitations. Kumar 11 has recently highlighted the drawbacks of observational studies in estimating the benefits of early viral treatment in the prognosis of flu. We agree that selection bias is difficult to avoid. Immortal time bias or survival-durationrelated selection bias imply that the late use of antivirals may be related to a better prognosis, whereas in fact our results suggest the opposite. Our results show no benefit of late neuraminidase treatment. In Israel, a retrospective cohort study documented a higher rate of complications after admission. 12 Severe complications (excluding hypoxia and uncomplicated pneumonia) occurred more frequently with late oseltamivir. In the same way, a Spanish study of ICU patients showed that ICU length of stay, days of mechanical ventilation and mortality were reduced in patients receiving early treatment versus late treatment with oseltamivir. 13 These reports do not give comparisons with flu patients without antiviral treatment. The mortality rate in our study (1.2%) was low in comparison with other studies. This may be due to the fact that our patients were not all admitted to the ICU, 5,14 and did not all have pneumonia at hospital admission. 15 Even so, the mortality rate was clearly lower than that found in the USA at the beginning of the pandemic (7%) 4 or the 4.9% reported in Canada. 16 Likewise, the rate of ICU admission (9.7%) was lower than that found in the USA (25%) 4 and Canada (16%), 16 although it was similar to the 8% reported in New Zealand Maoris. 17 Some form of selection bias cannot be completely ruled out as our study patients had to be interviewed to collect data on the use of medications before admission and other risk factors related to disease severity. In a study carried out in Catalonia (north-east Spain), of 773 cases hospitalized, 37.9% were admitted to the ICU. 18 In contrast, in Andalusia (southern Spain), 28 out of 311 hospitalized cases (9%) were admitted to an ICU. 19 In another Spanish study of patients admitted to the ICU, the mortality rate was 22%. 20 Taken together, these data suggest that patients who died shortly after admission were not picked up by our study. The predisposing factors for a higher probability of adverse outcome during hospitalization were broadly similar to those found in other studies. 16, 21 In one international series of patients with community-acquired pneumonia, male sex and obesity were predictors of mortality, although we did not find similar results. 15 We found a significant association between reductions in ICU admission/death and the administration of neuraminidase inhibitors within the first 48 h after the onset of symptoms, similarly to the findings of Jain et al. 4 and other studies. 6, 7 In these reports none of the pregnant women who died had taken neuraminidase inhibitors within the first two days after the onset of illness. Early use of neuraminidase inhibitors was associated with shorter hospital stay. Other reports have found no relationship between antiviral treatment and hospital stay. 22 In summary, we found that early treatment with neuraminidase inhibitors had a beneficial effect on outcomes during

Veillonella montpellierensis Endocarditis

Veillonella spp. rarely cause infections in humans. We report a case of Veillonella endocarditis documented by isolating a slow-growing, gram-negative microbe in blood cultures. This microbe was identified as the newly recognized species Veillonella montpellierensis (100% homology) by 16S RNA gene sequence analysis.

V eillonella are anaerobic, gram-negative cocci, part of the normal flora of the mouth, gastrointestinal tract, and vaginal tract. Veillonella dispar, V. atypica, and V. parvula have been cultured from human specimens. They are infrequently isolated in human infections. Rarely, Veillonella species have been the only etiologic agents identified in serious infections such as meningitis, osteomyelitis, prosthetic joint infection, pleuropulmonary infection, endocarditis, and bacteremia. A new species, V. montpellierensis, has recently been isolated from the gastric fluid of a newborn and from the amniotic fluid of 2 women (1). Its pathogenic role is still debated. A 75-year-old woman was admitted to the intensive care unit with septic shock. She had a history of diabetes mellitus. A cardiac murmur had been noted 8 years earlier but was not investigated further. On physical examination, the patient had aortic and mitral murmur. Reagent strip for urinalysis detected leukocytes and nitrites. After 3 blood cultures and urinalysis, the patient was treated for septic shock secondary to upper urinary tract infection with ceftriaxone, 2 g/day intravenously. The patient's condition rapidly improved with antimicrobial drugs and dopamine. Three days after admission, she was afebrile and hemodynamically stable; she was transferred to the urology department for acute pyelonephritis, which had not been confirmed by computed tomographic (CT) scan. Urine culture yielded Gardnerella vaginalis. Chest radiograph showed a patchy density of the right inferior pulmonary lobe confirmed by chest CT scan that suggested either pneumonia or neoplasia. On day 6, a transesophageal echocardiograph, performed because of the cardiac murmur, showed oscillating intracardiac masses on the aortic and mitral valves. Because the blood cultures were still negative, we determined that the patient had culture-negative endocarditis and replaced ceftriaxone with amoxicillin, 12 g/day for 6 weeks, in addition to gentamicin, 3 mg/kg/day for 3 weeks. On day 26, another transesophageal echocardiograph was performed and showed that the vegetation on the aortic valve had disappeared and the mitral vegetation was greatly reduced. The patient was discharged after 42 days of antimicrobial drug treatment, and follow-up was not possible. On day 14 after sampling, 2 of 3 anaerobic blood cultures (automated blood culture BACTEC 9240 system (Becton Dickinson, Le Pont de Claix, France) yielded a slow-growing, gram-negative microbe. Blood was subcultured onto Columbia agar with 5% sheep blood (Mérieux, Marcy l'Etoile, France) under 5% CO 2 and anaerobic atmosphere and resulted in small colonies. This slowgrowing microbe was lost after 2 subcultures, and no isolate is available for further description. The isolate retrieved in the blood culture was identified by 16S rRNA gene sequence analysis. The template DNA was prepared from a few colonies that were isolated on the blood agar incubated anaerobically. DNA was extracted by using Fastprep DNA extraction kit (Ozyme, St Quentin en Yvelines, France) according to the manufacturer's recommendations and was subjected to polymerase chain reaction (PCR) targeting the 16S rRNA gene as previously described (2) . Sequencing the PCR product (2) showed a 1,531-nucleotide sequence. This sequence shared 100% homology with that of V. montpellierensis (GenBank accession no. AY244769) and was already reported (GenBank accession no. AY244769) in a previous article (3) . In this article, the isolated Veillonella strain (that was isolated from our patient) was first identified as "candidatus V. atypica" since the sequencing of the amplicon disclosed 94% sequence similarity with that of V. atypica (3). V. montpellierensis had not yet been described. PCR contamination was unlikely since this organism had never been amplified in our laboratory and negative controls remained negative. According to the modified Duke criteria (4), our patient had definite endocarditis. Anaerobic microbes do not commonly cause endocarditis (5) . Most cases of anaerobic endocarditis are caused by anaerobic cocci, Propionibacterium acnes, and Bacteroides fragilis group (5) . We describe the seventh reported case of well-documented infectious endocarditis in which a Veillonella species was the sole pathogen and the first due to V. montpellierensis. Characteristics of the 7 Veillonella endocarditis patients are summarized in the Table. Five of them fulfilled the Duke modified criteria for definite endocardi-tis; the 2 others were possible endocarditis. All previously reported cases of Veillonella endocarditis were due to either V. dispar (9,10), V. parvula (11) , or V. alcalescens (6) (7) (8) , currently considered V. parvula (12) . One patient had no history of fever (7), and 1 patient had no preexisting valvular disease (8) . Five patients had an infected mitral valve; 4 of the 5 had prosthetic valves. Our patient had mitral and aortic endocarditis. All patients had positive blood culture except 2, for whom the diagnosis was made by culturing the valve (6, 11) . Veillonella spp. had also been isolated from intravenous drug users with polymicrobial endocarditis (13); V. parvula was isolated from a lung abscess in a patient with echocardiographic vegetations, but blood cultures were negative (14) . We could not test the susceptibility of the organism because the bacterium was lost on subculture. In treating infections with Veillonella species, penicillin has been the antimicrobial agent of choice (10) . However, recent studies found a notably high resistance to penicillin G (MIC >2µg/mL) (15) . These penicillin G-resistant isolates showed generally reduced susceptibility to ampicillin or amoxicillin but remained susceptible to amoxicillin and clavulanate (15) . We treated our patient for culture-negative endocarditis with amoxicillin. As the clinical state of our patient improved, we did not change antimicrobial agents. Our isolate has recently been compared with 3 other isolates and classified as a new Veillonella species named V. montpellierensis by Jumas-Bilak et al. (1) . We demonstrate here that V. montpellierensis is pathogenic for humans and may be included as a new agent of endocarditis caused by fastidious pathogens. We report here the seventh case of endocarditis due to Veillonella spp. identified by PCR amplification and sequencing of 16S rDNA gene and the first case of endocarditis due to V. montpellierensis. This case reemphasizes the usefulness of molecular methods in identifying fastidious microorganisms and in describing new clinical entities (3) .

Primate-to-Human Retroviral Transmission in Asia

We describe the first reported transmission to a human of simian foamy virus (SFV) from a free-ranging population of nonhuman primates in Asia. The transmission of an exogenous retrovirus, SFV, from macaques (Macaca fascicularis) to a human at a monkey temple in Bali, Indonesia, was investigated with molecular and serologic techniques. Antibodies to SFV were detected by Western blotting of serum from 1 of 82 humans tested. SFV DNA was detected by nested polymerase chain reaction (PCR) from the blood of the same person. Cloning and sequencing of PCR products confirmed the virus's close phylogenetic relationship to SFV isolated from macaques at the same temple. This study raises concerns that persons who work at or live around monkey temples are at risk for infection with SFV.

R ecent epidemics such as HIV and severe acute respiratory syndrome (SARS) have changed the way we view emerging infectious diseases; these epidemics show that animal reservoirs are important sources of new infectious threats to humans. Contact between humans and animals is a crucial rate-limiting step in this process, although data describing the variables that influence animal-tohuman transmission are relatively scarce. Nonhuman primates, by virtue of their genetic, physiologic, and sometimes social similarities to humans, are particularly likely sources of infectious agents that pose a threat to humans (1, 2) . Data on simian immunodeficiency virus (SIV)/HIV dramatize this point; scientists now theorize that SIVs were transmitted from primates to humans on several occasions (3, 4) . As a result, concern is increasing that other infectious agents enzootic in primate populations may endanger humans (5) . The family of SIV is 1 of 4 primateborne retroviruses known to infect humans (6) . Simian T-cell lymphotropic viruses, enzootic in both Asian and African Old World monkeys and apes, may have repeatedly crossed the species barrier (7, 8) . The resulting human form of the virus, HTLV, is the etiologic agent of 2 human diseases, adult T-cell leukemia and tropical spastic paresis (9) . Serologic studies have demonstrated evidence of primateto-human transmission of simian retrovirus (SRV), a retrovirus enzootic among Old World monkeys, in laboratory workers exposed to captive primates (10) . To date, no disease has been linked to human infection with SRV. Finally, in the past decade, evidence of infection with simian foamy virus (SFV) has been identified in 1% to 4% of persons who come into frequent contact with primates in zoos and primate laboratories and among 1% of bushmeat hunters in Central Africa (11) (12) (13) (14) (15) . SFVs are exogenous retroviruses enzootic in both New and Old World primates (16) (17) (18) . Phylogenetic analyses of SFVs indicate a species-specific distribution of virus strains not unlike that of SIV among some African primate species (19) (20) (21) (22) (23) . Among captive primate populations, seroprevalence of antibodies to SFV may reach 100% in adults, with many animals seroconverting before the onset of sexual maturity (19, 24 ; J. Allan, unpub. data). Fewer data are available on the seroprevalence of antibodies to SFV among free-ranging populations of primates. SFV is present in highest concentrations in the saliva of infected laboratory macaques (Macaca mulatta and M. fascicularis) and African green monkeys (Cercopithecus aethiops), which suggests that bites, scratches, and mucosal splashes with saliva from primates are likely mechanisms of transmission (25) (26) (27) . Because SFV is not known to occur naturally in humans, detecting serologic or molecular evidence, or both, of infection in a human, along with a history of close contact with primates, constitutes strong evidence for primate-to-human transmission, i.e., a marker for crossspecies transmission (28) . Existing serologic and molecular techniques can sensitively and specifically detect human SFV infections (29) . Primates and humans come into contact in a variety of contexts in Asia, including owning primate pets, observing performance monkeys, participating in ecotourism activities, hunting primates for bushmeat, and visiting monkey temples (30, 31) . Monkey temples are religious sites that have, over time, become associated with populations of free-ranging primates. Monkey temples are common throughout southern and Southeast Asia, where primates play an important role in culture. Asia's monkey temples annually bring millions of people, including hundreds of thousands of tourists, from around the world into close proximity with free-ranging primates (32) (33) (34) . Worldwide, monkey temples may account for more human-primate contact than any other context. In Bali, ≈45 temple sites contain substantial populations of free-ranging macaques (33) . Those who spend the most time in or around monkey temples include workers who maintain the temples; nuns, monks, and others who live on or around temple grounds; merchants who sell a variety of goods to tourists; and farmers whose fields are raided by macaques. Worshippers and tourists may also come into contact with temple macaques. These temples are thus an important context in which to investigate crossspecies transmission of infectious agents. Serologic data on free-ranging Southeast Asian macaques, though incomplete, suggest that SFV is enzootic among these macaque populations and that corresponding rates of SFV infection are high (80%-100%) (L. Jones-Engel, unpub. data). We hypothesized that humans who come into contact with these macaques might similarly show evidence of infection with SFV and investigated this proposition among a group of persons who worked at or around the Sangeh monkey temple in Bali, Indonesia. The Sangeh monkey temple is in central Bali, Indonesia, ≈20 km north of Denpasar, Bali's most populous city. The 17th-century Hindu temple complex at Sangeh serves the surrounding community and is a popular domestic and international tourist destination. Approximately 200 free-ranging M. fascicularis roam throughout the 6-hectare temple complex and into the surrounding rice fields and farms. Most of the macaques' caloric intake is from daily provisions provided by temple workers and food given to them by visitors. In July 2000, as part of a larger study on the epidemiology of exposure to primateborne viruses at the Sangeh monkey temple, 82 workers from Sangeh agreed to participate in the present study (32) . After informed consent was obtained, a questionnaire designed to elicit demographic data as well as data describing the frequency and type of exposure to Sangeh's macaques was administered in Bahasa Indonesia, the national language of Indonesia. Subsequently, 10 mL of blood was withdrawn from each participants antecubital vein, 6 mL was centrifuged to extract serum, and the remainder was mixed with EDTA. Serum specimens and whole blood were then stored at -20°C. In July 2000, 38 macaques within the Sangeh monkey temple area and surrounding forest were darted opportunistically and sedated with 3 mg/kg of Telazol (tiletamine HCl/zolazepam HCl). Following universal precautions, researchers withdrew 10 mL of blood from each macaque's femoral vein. The macaques were closely monitored during anesthesia and recovery. Six milliliters of blood was placed in a serum separator tube and centrifuged in the field to extract the serum. The remaining blood was placed in a tube containing EDTA. Sera and whole blood were frozen and stored at -20°C. Western blot immunoassays were performed with a few modifications (21) . Briefly, human foreskin fibroblast cells were infected with SFVbab1 (an isolate from a baboon) and maintained until notable cytopathologic changes were observed (19) . Culture supernatant fluid containing virus was harvested, and SFV was purified through a 20% sucrose cushion, separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis, and the antigens were blotted onto nitrocellulose sheets. The nitrocellulose paper was blocked with 3% bovine serum albumin and subsequently incubated with serum at a dilution of 1:40. Viral proteins were detected with the streptavidin-biotin system (Amersham Inc., Arlington Heights, IL, USA) by using diaminobenzidine as the substrate for color development. The criterion used for a positive sample was antibody reactivity to both p70 and p74 gag-related proteins (19) . DNA was purified from whole blood by using the QIAamp DNA Blood Mini Kit (Qiagen, Inc., Valencia, CA, USA). Briefly, 20 µL protease and 200 µL buffer AL were combined with 200 µL whole blood and incubated at 56°C for 10 min. After incubation, 200 µL ethanol was added, and the entire mixture was applied to a QIAamp spin column. The purified DNA was eluted from the column with 70 µL nuclease-free water, and concentration was determined spectrophotometrically at optical density 260 nm. The presence of SFV DNA was determined by using nested PCR. Five hundred ng purified DNA was combined with a PCR reaction mixture with a final concentration of 10 mmol/L Tris (pH 9.0), 50 mmol/L KCl, 0.1% Triton X-100, 2 mmol/L MgCl 2 , 200 µmol/L each dNTP, 0.15 mg/mL BSA, 1 µm Taq polymerase, and 400 nmol/L of each primer in a total volume of 50 µL. The following primer pairs were used: first round, forward 5′ CAG TGA ATT CCA GAA TCT CTT C 3′, reverse 5′ CAC TTA TCC CAC TAG ATG GTT C 3′; and second round, forward 5′ CCA GAA TCT CTT CAT ACT AAC TA 3′, reverse 5′ GAT GGT TCC CTA AGC AAG GC 3′ (29) . "Touchdown PCR" was used for both rounds with reaction conditions of initial denaturation at 94°C for 2 min, followed by 7 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 45 s, with a 2°C decrease in annealing temperature per cycle to 48°C, followed by 33 cycles of 94°C for 30 s, 48°C for 30 s, and 72°C for 45 seconds with a final extension at 72°C for 2 min. Second-round conditions were the same, except 19 cycles were used instead of 33. Twenty µL of the PCR reaction underwent electrophoresis on a 1% low melting point agarose gel. DNA bands (365-bp product) were excised from the gel and purified by using Wizard PCR Preps DNA Purification System (Promega Corp., Madison, WI, USA). The PCR product was ligated into the pCR 2.1-TOPO vector by using the TOPO TA Cloning Kit (Invitrogen Corp., Carlsbad, CA, USA.). An SFV DNA plasmid (pSFV-1Lgp), representing long terminal repeat, gag, and pol of SFV-1, was included as a positive PCR control and for determining sensitivity of detection by serial dilution (provided by A. Mergia). TOP10 cells were transformed with the ligation reaction, plated onto Luria broth agar plates containing 50 µg/mL kanamycin, and incubated overnight at 37°C. Miniscreen DNA was purified by using Wizard Plus Minipreps DNA Purification System (Promega). Samples were sequenced with the ABI 373 automated fluorescent sequencer using BigDye Terminator cycle sequencing chemistry (Applied Biosystems, Foster City, CA, USA). Five hundred ng purified DNA from whole blood was combined in a PCR reaction mixture with a final concentration of 10 mmol/L Tris (pH 9.0), 50 mmol/L KCl, 0.1% Triton X-100, 2.5 mmol/L MgCl 2 , 200 µmol/L each dNTP, 0.15 mg/mL BSA, 1 U Taq polymerase, and 400 nmol/L of each primer in a total volume of 50 µL. The following primers were used: forward, 12SA, 5´ CTG GGA TTA GAT ACC CAC TAT 3´, and reverse, 12SO, 5´ GTC GAT TAT AGG ACA GGT TCC TCT A 3´ (35) . Cycling conditions were the following: initial denaturation at 94°C for 5 min, followed by 35 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 5 min, with a final extension at 72°C for 5 min. The 101-bp product underwent electrophoresis and was processed for DNA sequencing essentially as described for SFV. The alignments were made in Bioedit (http://www. mbio.ncsu.edu/BioEdit/bioedit.html) and ClustalX 1.81 (ftp://ftp-igbmc.u-strasbg.fr/pub/ClustalX/). Columns in the alignment in which gaps had been inserted in regions with insertions, and deletions were stripped before the analyses. DNA trees were created with the neighbor-joining method by using the Phylip program (DNAdist; Neighbor), and the output was generated with Treeview (http://taxonomy. zoology.gla.ac.uk/rod/treeview.html). The GenBank accession numbers for the SFV and mitochondrial DNA sequences reported here are AY628152-69 and AY633510-39, respectively. The seroprevalence of SFV among the Sangeh macaques is presented in Table 1 . Thirty-eight macaques (29 males and 9 females; 4 juveniles, 6 subadults, 28 adults) were sampled. Thirty-four (89.5%) of the 38 macaques were seropositive for SFV by Western blot; 2 (50%) of 4 juveniles, all 6 (100%) subadults, and 26 (93%) of the adults were antibody positive. All 9 females were SFV seropositive. SFV seroprevalence in this free-ranging macaque population was consistent with seroprevalence studies done in captive (25) and other free-ranging macaque populations (L. Jones-Engel, unpub. data). Nearly two thirds of the study sample was male (62.2%), and the mean age of the study participants was 35 years (Table 2) . Of participants who reported being exposed, 23 (28.0%) of 82 persons reported having been bitten by the temple's macaques ( Table 3) . Five of the 23 reported being bitten more than once. Thirty-one (37.8%) of 82 reported having been scratched, including 9 persons who had been scratched more than once. In total, 37 (45.1%) of 82 reported having been either bitten or scratched. Of the 82 persons whose serum was tested for antibodies to SFV, 81 had negative results by Western blot. Serum from a 47-year-old farmer (BH66) was reactive to SFV gag proteins, p70/p74 (Figure 1 ). This man reported that he visited the monkey temple every day and had previously been bitten once and scratched on more than one occasion by macaques there. He reported that the bite and scratches, which occurred on his hands and toes, had bled and that he had washed the wounds and applied traditional medicines. He denied owning or coming into contact with a pet primate. He also denied hunting primates or consuming primate meat. To determine whether SFV was present in humans and macaques, we performed nested PCR amplification of SFV by using conserved primers designed to detect macaque SFV (29) . SFV was detected in the macaques tested, as well as in the 1 human participant (BH66), whose serum contained antibodies to SFV. Blood samples from all 81 human participants that were seronegative for antibodies to SFV were also negative for SFV by PCR. The limits of sensitivity for nested PCR were 1-10 SFV DNA copies per 500 ng cellular DNA, as determined by dilution of a positive control plasmid. Quantitative PCR products of BH66, 5 of the Sangeh macaques (M. fascicularis), and 13 pet macaques (M. tonkeana, M. maura, and M. fascicularis) from Sulawesi, Indonesia, were cloned at least twice, sequenced, and compared with published sequences from a rhesus macaque (M. mulatta) (SFV-1mac, M55279), an African green monkey (Cercopithecus aethiops) (SFV-3agm, M74895), and a chimpanzee (Pan troglodytes) (SFVcpz, U04327). As shown in Figure 2 , SFV from BH66 was most closely related to an SFV sequence amplified from 1 of the macaques at the Sangeh Monkey Temple (BP6). To verify that the BH66 sample was of human origin and not a mislabeled sample from an SFV-infected monkey, we amplified a small fragment from the 12S ribosomal mitochondrial DNA from 2 healthy unexposed humans, BH66, several macaque species, and African green monkeys; cloned the products; and derived DNA sequences. Phylogenetic analysis (Figure 3) showed that the human DNA sample grouped with mitochondrial DNA sequences from humans, confirming that the BH66 sample was from a human. Lymphocytes from BH66 were not available for isolation of SFV directly. This report documents the first case of SFV infection in a person with known exposure to free-ranging Asian primates. Antibodies to SFV were detected in serum, and SFV genomic segments were amplified from blood of the infected person. Because PCR is prone to contamination, which leads to false-positive results, PCR products were sequenced and compared to SFV from other macaque species. Phylogenetic analysis showed that SFV amplified from the infected human was most similar to SFV from M. fascicularis at the Sangeh monkey temple. Although ascertaining exactly how or when the person acquired his infection was not possible, he did report having been bitten once and scratched on a number of occasions by macaques at Sangeh. He denied other past contacts with primates. Because only 1 infected person was detected and our sample size was limited (82 persons), we cannot estimate the prevalence of SFV infection in this human sample. Previous research on naturally acquired SFV infection among bushmeat hunters in Central Africa found 10 who seroconverted in a population of 1,099 (1%), of which 3 also had positive results by PCR. Surveillance of larger numbers of persons exposed to primates at monkey temples is necessary to estimate the risk of SFV infection in this context. Human and macaque sympatry in Southeast Asia dates back as far as 25,000 years, as evidenced by remains at Niah Cave in Borneo (37) . Hindu and Buddhist temples are a relatively recent addition to the landscape, first appearing 1,000 to 4,000 years ago. Contemporary human-macaque commensalism at each monkey temple is shaped by the behavioral characteristics of the particular monkey population as well as the community's unique geographic, social, religious, and economic factors. Human-macaque contact differs at the various monkey temples. At Sangeh, where tourism has become an important economic resource, macaques depend on visitor feeding for most of their nutrition and have learned to climb on visitors' heads and shoulders to obtain food. Local photographers, who make a living photographing visitors with monkeys, encourage this behavior. Macaques, sometimes provoked by visitors, can become aggressive and will bite or scratch people. The Sangeh Temple Committee, which manages the Sangeh Monkey Forest, estimates ≈250 persons work in and around the monkey forest. Of the 82 persons who participated in the present study, 28% had been bitten by a macaque, and 6% had been bitten more than once. In comparison, 700,000 tourists visit the 4 main monkey temples on the island of Bali (Padangtegal/Ubud, Sangeh, Alas Kedaton, and Uluwatu) annually and as many as 5% of these visitors are bitten by macaques (A. Fuentes, unpub. data). These data suggest that visitors receive the preponderance of macaque bites. The literature describes ≈40 cases of human infection with SFV. Long-term follow-up data are unavailable for many of these cases. However, no disease has been linked to SFV among infected humans. Furthermore, no cases of human-to-human transmission have been described, even among those receiving transfusions of blood products from a worker at a primate center who was later shown to be infected with SFV (38) . Notwithstanding, more data are needed before SFV can be proclaimed a "virus without a disease." First, SFV infection has not been extensively studied in immunocompromised persons, so whether SFV has a more aggressive course in an immunologically "permissive" environment is unknown. Two persons died shortly after receiving baboon Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 11, No. 7, July 2005 ) , and SFV-3agm is a published sequence from an African green monkey (Cercopithecus aethiops). SFVcpz is a published sequence from a chimpanzee (Pan troglodytes) and was used an outgroup for this tree. The SFV human strain (BH66) clustered with an SFV sequence amplified from BP6 one of the macaques at the Sangeh monkey temple. The SFV DNA tree was created with the neighbor-joining method by using the PHYLIP program (DNAdist; Neighbor). Bootstrap replicates were 1,000. Bootstrap values were calculated by using Seqboot, DNAdist, Neighbor, and Consense (PHYLIP programs). Bootstrap values 60% are shown. The SFV tree was plotted in Treeview. liver transplants in 1994, and SFV was detected in the blood and tissues of both at autopsy (39) . More research in this area is needed. Moreover, little is known about the epizootiology of SFV among wild primates. Although SFV infection in humans has not culminated in any observable symptoms, SFV strains may differ in their capacity to infect or cause disease in humans. Finally, whether SFV can adapt in humans after transmission and potentially lead to disease needs to be examined. The recent SARS epidemic vividly demonstrates how the economic infrastructure and dense population of Asia facilitate the rapid international spread of disease. The combination of large primate reservoirs, prevalent humanprimate contact, a growing immunocompromised population, and advanced infrastructure in Asia increases the likelihood of a primateborne zoonosis emerging on this continent. The demonstration of SFV transmission in the context of a monkey temple in Bali points to a broad public health concern: other enzootic primate infectious agents may cross the species barrier and cause significant morbidity and mortality in human populations. Given our lack of knowledge of the effects of SFV, as well as the poorly defined risk of other primateborne zoonoses, steps should be taken to decrease the risks of cross-species transmission among the many persons who visit these sites. Previous data on both worker and visitor exposure to macaques at monkey temples suggest that human behaviors, especially the practice of feeding macaques, is a risk factor for being bitten or scratched (33, 34) . We have recommended that monkeys be fed only by specially trained personnel who minimize physical contact with monkeys. Such restrictions have been successfully employed at other monkey forests in Asia. For example, Singapore's Bukit Timah Nature Reserve has nearly 200 free-ranging macaques (M. fascicularis), and in 2002, an estimated 380,000 visitors made use of the reserve's trail system, yet contact between monkeys and visitors at Bukit Timah is rare. This finding may be ascribed to the park's policy, enforced by stiff fines, of prohibiting visitors from feeding macaques. This study reports the first case of human SFV infection in Asia and also, for the first time, links natural transmission of SFV to a person to a specific population of primates. Our findings suggest that workers in and around monkey temples can become infected with SFV. By implication, visitors to monkey temples, especially those who are bitten by a macaque, may also be at risk for SFV infection. Because many visitors to monkey temples are international travelers, these findings have ramifications for the potential global spread of primateborne infectious agents. Demonstrating natural cross-species transmission in a context that does not involve hunting for bushmeat implies that other contexts of primate-human contact may also facilitate the transmission of simian pathogens. These data point to the need for further research into SFV transmission in other contexts, including pet ownership and performance animals, as well as in the diverse geographic areas where humans and primates come into contact. Such research will help describe the overall picture of the Bootstrap replicates were 1,000. Bootstrap values were calculated by using Seqboot, DNAdist, Neighbor, and Consense (PHYLIP programs). Bootstrap values >60% are shown. The mtDNA tree was plotted in Treeview. This analysis suggests that BH66 was of human origin. Although the phylogenetic tree constructed with mtDNA from a variety of monkey samples can be used to distinguish human from monkey mtDNA, a large number of nuclear mtDNA sequences, have evolved as pseudogenes (36) . These sequences can be highly divergent from mtDNA and resulted in some ambiguity as mtDNA amplified from several monkeys did not group with other members of the same species. Because of the nature and variability of these sequences, definitive conclusions about mtDNA phylogenies could not be determined; however, mtDNA trees were still useful for determining the origin of mtDNA material. emergence of primateborne pathogens such as HIV/SIV and will form a scientific basis for guiding policies and programs to deter the spread of emerging zoonotic pathogens.

Application of Consensus Scoring and Principal Component Analysis for Virtual Screening against β-Secretase (BACE-1)

BACKGROUND: In order to identify novel chemical classes of β-secretase (BACE-1) inhibitors, an alternative scoring protocol, Principal Component Analysis (PCA), was proposed to summarize most of the information from the original scoring functions and re-rank the results from the virtual screening against BACE-1. METHOD: Given a training set (50 BACE-1 inhibitors and 9950 inactive diverse compounds), three rank-based virtual screening methods, individual scoring, conventional consensus scoring and PCA, were judged by the hit number in the top 1% of the ranked list. The docking poses were generated by Surflex, five scoring functions (Surflex_Score, D_Score, G_Score, ChemScore, and PMF_Score) were used for pose extraction. For each pose group, twelve scoring functions (Surflex_Score, D_Score, G_Score, ChemScore, PMF_Score, LigScore1, LigScore2, PLP1, PLP2, jain, Ludi_1, and Ludi_2) were used for the pose rank. For a test set, 113,228 chemical compounds (Sigma-Aldrich® corporate chemical directory) were docked by Surflex, then ranked by the same three ranking methods motioned above to select the potential active compounds for experimental test. RESULTS: For the training set, the PCA approach yielded consistently superior rankings compared to conventional consensus scoring and single scoring. For the test set, the top 20 compounds according to conventional consensus scoring were experimentally tested, no inhibitor was found. Then, we relied on PCA scoring protocol to test another different top 20 compounds and two low micromolar inhibitors (S450588 and 276065) were emerged through the BACE-1 fluorescence resonance energy transfer (FRET) assay. CONCLUSION: The PCA method extends the conventional consensus scoring in a quantitative statistical manner and would appear to have considerable potential for chemical screening applications.

Molecular docking-based virtual screening is widely used to discover novel ligands in the early stages of drug development [1, 2, 3, 4] . Various docking programs, such as DOCK [5] , AutoDock [6] , Surflex [7] , FlexX [8] , GOLD [9] , and Glide [10, 11] , have been developed. As an essential component of these programs, the scoring function is able to evaluate the fitness between the ligand and receptor guiding the conformational and orientational search of ligand-binding poses. Since the 1990s, several dozens of scoring functions have been reported in the literature [12, 13] . Current scoring functions can be roughly classified as force-field-based methods [5, 14, 15] , empirical scoring functions [16, 17] , and knowledge-based statistical potentials [18] . The existing limitations in current docking and scoring include a lack of protein flexibility, inadequate treatment of solvation, and the simplistic nature of the energy function employed [19, 20, 21, 22] . In particular, the major weakness of docking programs lies in the scoring functions [12, 13] . Considering the computational cost and time required for virtual screening, all of the current scoring functions use various approximations resulting in inaccuracy in the score and rank of the ligand-binding poses [19] as well as in false positives mixed in with the top scorers in the ranking list when virtual screening was performed with only a single scoring function. Some studies focus on calculating proteinligand free binding energy, free energy perturbation (FEP), thermodynamic integration (TI) [23, 24, 25] , MM-PB/SA, MM-GB/SA [26, 27, 28] and linear interaction energy (LIE) [29, 30, 31] , which were used to perform post-docking processing. Although these methods are reported to be significantly more robust and more accurate than scoring functions, the accuracy is less than that usually required in typical lead optimization applications to differentiate highly similar compounds. Attempts have been made to reduce the weakness of a single scoring function. In 1999, Charifson et al. introduced a consensus scoring method [20] . Many studies have suggested that employing consensus-scoring approaches can improve the performance by compensating for the deficiencies of the scoring functions with each other [19, 20, 21, 22] . Although the rationale for consensus scoring is still a subject of study, it has become a popular practice. Compared with the calculation of free binding energy mentioned above, the combination of three or four individual functions to perform consensus scoring is a relatively cheap computational method. Wang et al. carried out an idealized computer experiment with three different ranking strategies (''rank-by-number'', ''rankby-rank'', and ''rank-by-vote'') to explore why the consensus scoring method performs better than the single scoring function [32] . However, the application of consensus scoring approaches is not always practical under ideal conditions because many obstacles prevent us from obtaining satisfied enrichment rates. These obstacles are as follows: (1) the binding scores calculated by the different scoring functions are typically given in different units and signs; (2) the scoring functions employed in consensus scoring often come from different categories; and (3) the linear relationship between many scoring functions (i.e., one scoring function can be expressed linearly by one or some other scoring functions). In addition to the three ranking strategies introduced by Wang et al., several groups employed another consensus scoring method involving the linear combination of several scoring functions. In the study by Guo et al., five commercially available scoring function were weighted and summed to build a consensus score [33] by training with a 53-molecule set. Verdonk et al. also employed a linear combination of three scoring functions to rerank the compounds [34] . Although an improvement was found for this consensus scoring method, the correlation between the scoring function and the experimental binding affinity is relatively poor. For a quantitative linear combination of the original scoring functions, the method for determining the appropriate weighting factors (correlation coefficients) for each scoring function is a complex problem. In this study, we present an alternative method, principal component analysis (PCA) [35, 36, 37] , for performing a linear combination of multiple scoring functions, formulating a modified ranking score and PCscore, and re-scoring and re-ranking the compounds after virtual screening. PCA is a powerful tool for pattern recognition, classification, modeling, and other aspects of data evaluation [36] . In addition, PCA is a linear transformation technique used to simplify a data set by reducing the dimensionality of multivariate data while preserving as much of the relevant information as possible. The principal components (PCs) are linear combinations of the original variables. The linear coefficients of the inverse relationships of linear combinations are called the component loadings. It represents the correlation coefficients between the original variables and the PC. In the present study, the first principal component (PC1) accounts for the maximum variance (eigenvalue) in the original dataset. The second principal component (PC2) is orthogonal (uncorrelated) to the first one, and it accounts for most of the remaining variance. This procedure is continued until the total variance is accounted for. The method of PCA makes use of intercorrelations that originate from the covariance matrix of the variables. This work was performed as part of a project aimed at identifying strong, selective inhibitors of b-secretase (BACE-1) to overcome the shortcomings of the existing drugs to treat Alzheimer's disease (AD) [38, 39, 40] . It is generally accepted that Alzheimer's disease is caused by extracellular senile plaque deposition and that the intracellular formation of neurofibrillary tangles in the brain. b-amyloid peptides, which form the senile plaques, are formed by the action of the b-secretase and csecretase enzyme on the amyloid precursor protein (APP) [41, 42, 43] . The design of a lead compound that can inhibit APP binding to the active site of BACE-1 will prevent the cleavage of APP from the b-amyloid peptide and thus eventually prevent senile plaque formation [44, 45] . In the present study, the training set is composed of 50 confirmed BACE-1 inhibitors and 9950 inactive compounds [46, 47, 48] . Three rank-based virtual screening methods, individual scoring, conventional consensus scoring and PCA scoring were examined to identify BACE-1 inhibitors. To validate the efficacy of PCA ranking method, after virtual screening of 113,228 compounds (Sigma-AldrichH corporate chemical directory) [49] and the BACE-1 fluorescence resonance energy transfer (FRET) assay, we found two drug-like and lowmicromolar inhibitors. For the training set, in order to reduce artificial enrichment [34] , a subset of WDI (World Drug Index) was specifically designated as inactive molecules. Firstly, WDI was filtered to eliminate compounds whose molecular weight was either less than 200 or greater than 800. In addition, the compounds whose log P is larger than 7 and the number of rotatable bonds is more than 15 should be abandoned. Secondly, the remaining 37,843 WDI compounds were subjected to diverse selection based on 2D UNITY fingerprints. The dissimilarity selection was performed by the Selector module in SYBYL, which resulted in 9950 compounds with a maximum Tanimoto index of 0.69. The active set was compiled from a diverse selection of 50 BACE-1 inhibitors from the total compounds available in the Prous Integrity Drugs & Biologics database [50] . This library of 10,000 compounds as a training set has an active content of 0.5%, which mimics real-life screening situations. In order to extend the application of the present study, a total of 113,228 compounds (Sigma-AldrichH corporate chemical directory, Z272000, 1997) [49] were used as the test set. Both the training set and test set compounds were stored as a SYBYL SLN list and converted to SYBYL mol2 format using Concord [51]. The ligand-bound (1W51) structure of BACE-1 was used [52] . The procedure used to prepare the structure was as follows: hydrogen was added, the protonation states were assigned, and a highly limited optimization was performed to reduce bad contacts and the overall strain energy in the protein structure. The aspartate located on the active site was adjusted to an ideal protonation state, the Asp32 was protonated, the Asp228 was ionized [46] . Virtual screening experiments were performed using the Surflex docking program [7, 53, 54] with an empirical scoring function (based on the Hammerhead docking system). The empirical scoring function has been updated and re-parameterized with additional negative training data along with a search engine that relies on a surface-based molecular similarity method. Standard parameters were used as implemented in the SYBYL software (version 8.1) [51]. The search strategy of Surflex employs an idealized ligand (called protomol), which utilizes various molecular fragments. Molecular fragments were tessellated in the active site and optimized based on the scoring function. The search algorithm utilized the morphological similarity function, which is evaluated between the protomol and the putative ligands. For the docking algorithms, a post-dock minimization procedure was applied using the BFGS quasi-Newton method and an internal Dreiding force field. For each compound, the top 30 ranked poses were saved. Five scoring functions in SYBYL, including D_Score [55] , G_Score [9] , ChemScore [56] , Surflex_Score [7, 53] , and PMF_Score [57, 58] , were applied to extract the stored poses. Next, five pose groups were produced, with each pose group containing 10,000 compounds. For pose ranking, we use 12 scoring functions including the five scoring functions (D_Score, G_Score, ChemScore, Surflex_Score, and PMF_Score) from the SYBYL software and the seven scoring functions (LigScore1 [17] , LigScore2 [17] , PLP1 [59] , PLP2 [59] , jain [60] , Ludi_1, and Ludi_2 [61, 62] ) from the Discovery Studio software (version 2.1) [63] . Although the five pose groups generated by Surflex have been post-minimized using the internal Dreiding force field, these five pose groups were further minimized in the protein environment using the CFF force field [64] when the seven scoring functions were used for scoring by the Discovery Studio software. All high-throughput docking calculations were performed on a Linux cluster using the CentOS 5.4 operating system. In this study, we adopted the ''rank-by-number'' strategy in the consensus scoring to combine the results of multiple scoring functions. The ''rank-by-number'' strategy was previously found to outperform the ''rank-by-rank'' and ''rank-by-vote'' strategy because it can summarize most of the information [32] . For the ''rank-by-number'' strategy, the consensus score of each binding pose is an average of the values determined by each of the individual scoring functions in a given consensus scoring scheme. With this strategy, a moderate number of scoring functions (i.e., three or four) have been proposed to be sufficient for significantly improving the results. Therefore, we chose 4 of the 12 scoring functions (D_score, jain, and Ludi_1, Surflex_Score) to perform consensus scoring in the present study. Because the binding scores calculated by the different scoring functions are typically given in different units, it is almost impossible to compute consensus scores simply by summing up the binding scores determined by each of the individual scoring functions. Therefore, we scaled the binding scores of each scoring function to unit variance and centered (i.e., the mean value is zero, the standard deviation is one). The Z-scaled scoring function values (ZScore) are computed by where f i is the scoring value of a certain scoring function, m is the mean value and s is the standard deviation of this scoring function observed for the entire test set. The consensus score in a certain pose group is the average of ZScore by 4 individual scoring functions mentioned above in the given consensus-scoring scheme. We describe PCA mathematically as described below. Consider p random variables X 1 , X 2 , …, X p , the original system can be rotated to form a new coordinate. Let S be the covariance matrix associated with the random vector X9 = [X 1 , X 2 , …, X p ]. The corresponding eigenvalue-eigenvector pairs are (l 1 , e 1 ), (l 2 , e 2 ), …, (l p , e p ), and the ith principal component is given by: PC i~e 0 i X~e i1 X 1 ze i2 X 2 z:::ze ip X p , i~1,2,:::,p ð1Þ Then Var(PC i )~e 0 i Se i~li i~1,2,:::,p ð2Þ Thus, the principal components are uncorrelated, and their variances are equal to the eigenvalues of S. Another property of the principal components is: Var(X 1 )z:::zVar(X p )~l 1 zl 2 z:::zl p~V ar(PC 1 ) Then the proportion of the total population variance due to the kth PC is:~l k =(l 1 zl 2 z:::zl p ) k~1,2,:::,p: Consequently, if most of the total population variance for large p can be attributed to the first two or three components, then these first two or three components could serve as a substitute for the original variables with a minimal loss of information. Moreover, if the weight of the last PCs occupied a highly trivial part of the total population variance, then the last PCs can be neglected (i.e., set to zero). In the present study, there were five extracted pose groups, and we used only eight scoring functions, which included LigScore1, PLP1, jain, Ludi_1, D_Score, G_Score, ChemScore, and Sur-flex_Score, to perform the PCA for each group (The details are mentioned in the results section). Thus, for the training set, the eight scoring functions were used as the variables (i.e., columns of the matrix) and 10000 compounds were arranged in the rows of the matrix. Then the 1000068 correlation matrix was established. Because the scoring functions in our test produce binding scores with different units and signs, the signs of the binding scores produced by LigScore1, PLP1, jain, Ludi_1, D_Score, G_Score, and ChemScore were reversed to ensure that positive binding scores always indicated higher binding affinities. All of the binding scores were scaled to unit variance and centered. Thus, each column of data had an average of zero and a standard deviation of one. For each of the five scoring function extracted poses, we have calculated the eigenvalues and cumulative contribution rate. The first three principal components were extracted. Each principal component is a linear combination of eight Z-scaled scoring functions, which formulate a modified ranking score function, PCscore. PCscore is set to re-score and re-rank the extracted poses from each of the five scoring functions. PCscore can be written as follows: WipÃZScore p i~1,2,:::,p Where the n terms, ZScore p , are the Z-scaled scoring function, and the coefficients, w ip , are the loading values (i.e., the elements of p principal component eigenvector e p ). For example, for PC1score The linear coefficient values (loading values) for w 11 , w 12 … w 18 were the elements of the first principal component eigenvector, e 1 . In the present study, an SPSS version 16.0 statistical analysis package (SPSS Inc.) was used to normalize and calculate the principal components for all of the scoring data. After virtual screening of the test set, 40 chemical compounds (Sigma-Aldrich Co. LLC) were purchased for experimental test through fluorescence resonance energy transfer (FRET) assay based on the results of conventional consensus scoring and PCA scoring. According to the previously described method [65] , during the assay, compounds were diluted to 8 different concentrations and incubated 60 minutes at room temperature. Additional measurements were performed in the presence of detergent or with an incubation time of only 3 min to check for nonspecific effects (e.g., compound aggregation [66, 67] ). Briefly, fluorescence progress curves of 30 mL reaction volumes were measured on a Gen5 TM ELISA reader (BioTekH Instruments, Inc.) upon excitation at 545 nm and emission at 580 nm in 384well microtiter plates (Corning, 3654). Linear regression analysis was calculated with the SPSS 16.0 software. Upon docking 10,000 compounds of the training set with Surflex, every compound yields 30 poses in the active pocket of the target (1W51), and no solution was found on the outside of the active pocket. After docking, we used five scoring functions to extract the pose and twelve scoring functions to rank the extracted poses resulting in 60 different scoring combinations. The top 1% of the ranked database was set as the threshold value, that is, the evaluation of the effectiveness of the scoring protocols involved numbering the actives for the top 100 candidates. The enrichment rates of the scoring protocols are presented in Table 1 . The best-scored pose was always used to represent the plausible binding mode of a particular compound. In many cases, the pose varied from one to another as dictated by the separate scoring functions. Inspection of each scoring function in Table 1 indicated that the quality of the extracted poses is similar. No single scoring function outperforms the others with respect to the extraction. The Surflex_Score provided reliable poses that were ranked best by D_Score and jain. The D_Score and ChemScore provided reliable poses that were ranked best by Ludi_1. For the pose ranking, it appears that the ranking by Ludi_1 retrieved more actives than the other scoring functions. Ludi_1 retrieved 20 inhibitors with D_Score and ChemScore pose extraction and 18 inhibitors with Surflex_Score and G_Score pose extraction. Ludi_1 was derived by empirically fitting a set of protein-ligand complexes with experimentally measured binding affinities. It is a sum of the five contributions including hydrogen bonds, perturbed ionic interactions, lipophilic interactions, the freezing of internal degrees of freedom of the ligand, and the loss of translational and rotational entropy of the ligand. At the same time, D_Score also performs well for ranking the docking poses. It retrieved 20 inhibitors with Surflex_Score and 19 inhibitors with PMF_Score pose extraction. This good performance can be attributed to D_Score providing the most accurate approximation of the binding energy where both the electrostatic and hydrophobic contributions to the binding energy are counted. In addition, a distance-dependent dielectric attenuates the chargecharge, and other polar interactions were considered. PMF_Score also provided reliable poses that were ranked best by D_Score. It yields 19 actives in the top 1% of the ranked list. However, it failed to rank any sensible docking poses regardless of what poses were extracted by itself or by the other scoring functions. Thus, for the BACE-1 target, PMF_Score appears to be more capable of accurately docking and correctly identifying the true binding mode, but the disadvantage of PMF_Score is the enrichment of active compounds. Inspection of Table 1 demonstrates that two paired scoring with LigScore1 & LigScore2 and PLP1 & PLP2 retrieved an equal number of active compounds. It is not surprising that both Ligscore1 & Ligscore and PLP1 & PLP2 use the same scoring functions with only slightly different algorithms and parameters sets [68] . There are three different versions of Ludi (i.e., Ludi_1, Ludi_2, Ludi_3) [61, 62, 69] . According to the Discovery Studio user manual, only the weight factors employed by Ludi_2 for each term are derived by fitting to experimentally determined binding affinities. In fact, all three versions were tested for enrichment rates of virtual screening against BACE-1 in our study, and we found that Ludi_1 outperforms the other two versions. Prior to re-ranking the results from the virtual screening using consensus scoring and PCA, the intercorrelations between the scoring functions mentioned above were investigated. The original data of each scoring function were scaled to unit variance and centered. The correlations between the binding scores computed by the 12 scoring functions are summarized in Table 2 . For the four scoring functions (i.e., Ligscore1, Ligscore2, PLP1 and PLP2), Table 2 exhibited a high correlation between any two of them. The correlation was higher for LigScore and PLP (R = 0.97,0.98) because they belong to the empirical scoring function category and the sum of the pairwise linear potentials between the ligand and the protein heavy atoms with parameters is dependent on the interaction type. In addition, Ludi_1 and Ludi_2 also exhibited a very high correlation (R = 0.911). However, the correlation coefficients between Ludi and the other functions, such as PLP and LigScore, were smaller because the master equations that describe the binding free energy used in Ludi are different from those used in the PLP and LigScore functions. In addition, the algorithms vary for the same term in the master equation, such as hydrogen bonding and hydrophobic effect. Furthermore, there was a higher correlation between G_Score and D_Score (R = 0.771). We were not surprised by this result because both functions are in the same category, both of their algorithms adopted force-field-based methods that estimate the enthalpic contribution upon binding, and both of them use a very similar treatment of the energy terms. As depicted in Table 2 , moderate correlation was exhibited by Surflex_Score and either the D_Score or the G_Score function; between D_Score and the ChemScore, LigScore1, LigScore2, PLP1, or PLP2 function; and between Jain and the LigScore1, LigScore2, or PLP2 function. This is consistent with virtually all of the scoring functions being designed to reflect the basic features in protein-ligand interactions including hydrogen bonds and hydrophobic contacts. Moreover, the binding scores computed by these scoring functions are all correlated, to some extent, to the known binding constants. Therefore, some intercorrelation between them is natural. PMF shared the least with all of the other scoring functions. Its unique knowledge-based algorithm parameterized using crystal complexes is different from the rest of the scoring functions being considered [57, 58] . The hit-rates observed among the top 1% of the screening set using the ''rank-by-number'' strategy are shown in Table 3 . By comparing the performance of all of the consensus ranking schemes tested, it appears that the consensus ranking does statistically outperform the best of the individual scoring function. The Surflex_Score pose extracted group produced 24 hits in the top 1% of the screening set when the quadruple-scoring scheme was applied. The improvements are not trivial. The best individual scoring function, jainScore, produces only 20 hits in the top 1% of the screening set. Our results are in agreement with the previous study, which suggested that, in theory, combining multiple scoring functions should always provide improved performance over individual scoring functions in simulated virtual screening experiments [32] . According to the present results, we cannot definitively conclude that more scoring functions result in a better performance. For example, application of double-scoring schemes (e.g., Surflex_S-core&D_Score) could also obtain 24 hits in the top 1% of the screening set, which is the same result obtained using the quadruple-scoring scheme. However, it is important to note that double-scoring schemes do not outperform the best individual scoring function in all cases. For example, Surflex_Score&jain could obtain only 19 hits, which is slightly less than the 20 hits obtained from the single scoring function, jain. Therefore, it is largely unpredictable which combinations of scoring functions would produce the optimal results. In practice, it is better to test all possible combinations of scoring functions on the appropriate samples. Some studies have shown that consensus ranking does not outperform the best individual scoring function [70, 71] . They argued that if one knew in advance which scoring functions worked best for a given target, the better performance could be achieved using this scoring function alone and by concentrating on only the highest ranking compounds. Given the contradiction between their arguments and our results, we explained as followed: Firstly, the three scoring functions (D_Score, jain, Ludi_1) that we chose performed the best in single scoring. It is important to consider which scoring functions should be chosen to perform consensus scoring. We used an additional four scoring functions to perform consensus scoring, but the performance was not as good as the four functions that we chose. Due to the variation in the performance of the different scoring functions, blindly choosing scoring functions to perform consensus scoring will decrease the enrichment rates. Secondly, the four scoring functions that we chose were independent of each other. It is reasonable to expect that an effective consensus scoring scheme would combine complementary scoring functions rather than highly correlated ones. As indicated in Table 2 , if the consensus scoring schemes contained Ligscore1 and Ligscore2 as well as PLP1 and PLP2, it would perform poorly compared to the other schemes. Thirdly, Verdonk et al. performed a computational experiment on the simulated effect of consensus ranking with an increasing number of scoring functions using the rank-by-number protocol [34] . They noted that if the first scoring function performs well (standard deviation = 1.0), then adding additional scoring functions (standard deviation = 3.0) to perform consensus ranking can reduce the enrichment rates compared to the most accurate single scoring function. The main reason for this phenomenon was that noise was added to the protocol. However, if all of the scoring functions have a standard deviation of 2.0, then adding extra scoring functions to the consensus ranking protocol always improves the enrichment rates. In our study, all of the binding scores were scaled to unit variance and centered, which was consistent with the results reported by Verdonk et al [34] . In summary, our present results suggest that application of triple-scoring and quadruple-scoring schemes are more robust and accurate than any single scoring procedure. Principal component analysis (PCA) can extract information from large-scale scoring data and decompose multiple scoring functions into one or two scoring functions, which can be used to re-score and re-rank the compound binding poses. As mentioned above, the PLP1&PLP2, ligscore1&ligscore2 and Ludi_1&Ludi_2 have high relative between each other. In addition, PMF_Score failed to rank any sensible docking poses. Thus, we do not use these four scoring functions in the following study. We constructed five 1000068 matrices using the remaining eight scoring functions as the matrix column and the 10000 compounds as the raws. Then the matrix was transformed such that each column of data had an average of zero and a standard deviation of one. As observed from Table 4 , we can derive eight uncorrelated descriptors (the principal components) from each scoring matrix. The weight of each principal component was determined based on their contribution rate to the variance (eigenvalues, l). We found that the first three principal components (PC1, PC2, PC3) account for .80% of the total variance for each pose group. The PC4, PC5, PC6, PC7, and PC8 could be omitted in further studies due to their trivial contribution to the total variance. This result is in agreement with the aim of introducing PCA to significantly minimize the number of variables and to omit the principle components with low variance that will not affect the total variance. These principal components may lack physical meaning by themselves because they may act as statistical descriptors. Nevertheless, we could still assess the physical meaning of each PC according to the energy terms of each scoring function. To the best of our knowledge, the physical meaning of PC1 could be attributed to van der Waals interactions, the physical meaning of PC2 could be attributed to electrostatic interactions, and the physical meaning of PC3 could be attributed to the hydrophobic interactions between the protein and ligands. The loadings express how well the new abstract principal components correlate with the old variables. Loading values (i.e., correlation coefficients) .0.7 are marked in boldface type in Table 5 . PC3 accounts for approximately 11% of the total variance. It is interesting to note that PC3 negatively correlates with Surflex_-Score, D_Score, G_Score, and ChemScore (SYBYL software) but positively correlates with PLP1, LigScore1, jain, and Ludi_1 (Discovery Studio software). The first two PC loadings against each other are shown in Figure 1 . Because the PCA is invariant to the mirroring through the origin, the data shown here indicate that there is a significant correlation between LigScore1 and PLP1. Likewise, the correlation between G_Score and D_Score in relation to the data is unambiguous and significant. There is no need to measure and evaluate all of the variables to achieve the same characterization in further studies. It is sufficient to measure one variable per group. The present results show that Jain contains nearly the same information as D_Score and has low loading on PC2. Because PC2 could be attributed to electrostatic interactions between the protein and the ligands, Jain has no significant influence on the electrostatic interactions between the protein and the ligands. Among the eight loads in PC2, Ludi_1 exhibits the maximum Table 2 . The Correlation Matrix of each scoring function (poses extracted by Surflex_Score). value (Figure 1 ), which means Ludi_1 plays a significant role in the description of PC2. According to Eq 5, the PC is a linear combination of multiple original variables. Therefore, we can formulate the first three PCscore scoring functions for each pose group from the data in Table 5 . For example, for the poses extracted by Surflex_Score, the first PC scoring function was: Next, the docked compounds are re-scored and re-ranked using the PCscore scoring functions as mentioned above. The enrichment rates are also determined by noting the numbers of active compounds retrieved in the top 1% of the ranked database (Table 5) . When comparing the enrichment rates from PC1score to the results obtained from a single scoring function or conventional consensus scoring functions, PC1score exhibited better performance for the enrichment rates regardless of the scoring function employed to extract the compound pose. For example, PC1score yields 26 active compounds in the Surflex_Score and PMF_Score pose group, which outperforms both the single scoring function with 20 or less active compounds and the consensus scoring method with 24 active compounds. As indicated in Table 5 , application of PC1score results in more active compounds than the application of PC2score and PC3score for each of the five pose group due to the descriptiveness of the first principal component, which shares the maximum amount of the whole variance followed by the decreasing descriptiveness of the other PCs. We did not obtain any active compounds in the top 1% of the ranked database using PC3score because the values of the eigenvalue of PC3 were ,1. The PCA can illustrate the relationship between the different compounds and the different scoring functions. The compounds can be plotted in the space defined by two PCs (score plot, Figure 2 ), which identifies active compounds as a function of inactive compounds. The values of the scores can be understood as the values of the compound in the new variable space, i.e., the principal component space. In Figure 2 , active compounds are depicted as red circles, and inactive compounds are depicted as black squares. The results showed that most of the data were scattered along the PC1 axis. The scattering variation along the PC1 axis is larger than that along the PC2 axis, which corresponds to the values of eigenvalue and reflects the descriptive power of first two PCs scores. Because the new variable space is normalized with zero mean, the most active compounds, which are farther from the origin, have values significantly different from the mean and can be considered outliers. Moreover, we found that the scattering positions of the true BACE-1 inhibitors are located on the right side of the PC1 axis indicating that PC1 plays a significant discriminating role among active and inactive compounds. As for the PC2 axis, all of the data were scattered in a narrow area from -3 to 3, and the discriminating power among active and inactive compounds was weaker. To further investigate the validity of docking based virtual screening, after virtual screening of 113,228 compounds against BACE-1 by Surflex, we employed conventional consensus scoring and PCA scoring protocol to select compounds for experiment test against BACE-1. Standing the view of economic point, the number of compounds to be tested in computational docking studies should be restricted in a smaller and reasonable range, therefore, we used several filters to the select the final compounds in Surflex_score extracted pose for experiment. In an initial attempt, we employed conventional consensus scoring protocol to select the potential inhibitors. Firstly, we selected the top 300 compounds according to the ranking results of the conventional consensus scoring protocol, e.g., quadruplescoring scheme (Surflex_Score&D_Score&jain&Ludi_1); Secondly, visual inspection has been given to all individual complexes for the top 300 compounds, Surflex provides interaction information between the protein and ligand for each docking experiment, only those compounds with interactions to the catalytic residues (Asp32 and Asp228) and other relevant residues are extracted; Thirdly, to remove unsuitable compounds that would not reach and pass the clinical trials due to undesired and toxic properties, the so-called Lipinski ''Rule-of-five'' [72] , a very popular method was used to evaluate the drug likeness of a candidate structure. Finally, 20 drug-like compounds were selected to purchase from Sigma-Aldrich Co. LLC. By the BACE-1 fluorescence resonance energy transfer (FRET) assay, disappointingly, no inhibitor was found among the compounds selected by the conventional consensus scoring protocol. Based on the theory that PCA can summarize most of the information from the original scoring functions, we employed PC1score to re-rank the 113,228 Surflex_score extracted poses, as mentioned above, PC1score is a linear combination of eight scoring functions (Surflex_Score, G_Score, D_Score, ChemScore, LigScore1, PLP1, Jain and Ludi_1). By the same filter protocol as the conventional consensus scoring, another 20 drug-like compounds were select for purchase among the top 300 compounds. Excitingly, this time two compounds (S450588 and 276065), with a remarkable 10% hit rate, emerged as the BACE-1 inhibitors in the low-micromolar range, showing IC 50 values of 51.6 and 85.3 mM, respectively ( Table 6 ). The chemical structures of these two compounds were showed in Table S1 . As depicted in Figure 3A , after compound 1 docking into 1W51 structure, the protonated 2-NH 3 group of the lysine moiety form hydrogen bond with Asp228, Gly230 and Thr231, respectively, the 6-NH group form hydrogen bond with Tyr198. The benzyl ring (P1) fills the S1 pocket shaped by the Tyr71, Phe108, and Trp115 residues, while carbobenzyloxy moiety (P29) fills the S29 pocket shaped by the Tyr71, R128, and Y198 residues, so as to allow the carbonyl group to form hydrogen bond with the Thr72 residue, the benzyl group to establish a cation-p interaction with the guanidine group of Arg128. As depicted in Figure 3B , in the catalytic site, for the small size and symmetric overall shape of compound 2, one of the hydroxyl group of the tartaric diamide core are involved in hydrogen bonds with the side chain of the catalytic Asp32 and Asp228, respectively, whereas the other hydroxyl group form hydrogen bond with Thr72, and one of the amide group form hydrogen bond with Gly230. Both sides of compound 2 are benzyl groups, one of the benzyl group occupy the S2 pocket shaped by the Asn233, Arg235 and Ser325 residues, the other benzyl group occupy the S29 pocket, establish a cation-p interaction with the guanidine group of Arg128. BACE-1 is one of the major Alzheimer's disease target [38, 39, 40] . To find novel BACE-1 inhibitors, a lot of academic research centres and pharmaceutical industries are quite active in this field. Merck research group performed in vitro highthroughput screening (HTS) and found a single molecule (a 1,3,5 trisubstituted benzene) as a hit from a multi-million compound library [73] . Johnson and Johnson also reported a novel cyclic guanidine screening lead, the initial screening lead had an IC 50 value of 900 nM [45] . Astex Therapeutics has taken a fragment-based lead generation approach [74] . After the virtual screening of a fragment library, a small number of potential structures were soaked with BACE-1 crystals in anticipation of obtaining a co-crystal with the enzyme. Huang et al. performed in silico Screening of 180,000 small chemicals, they found 10 diacylurea inhibitors showed an IC 50 value lower than 100 mM in a enzymatic assay and four of them were cell penetrant (EC 50 ,20 mM) [75] . Despite the availability of many reliable in silico approaches and robust in vitro commercially available assays, discovering BACE-1 inhibitors still remains a challenging task. In the present study, based on the virtual screening of 10,000 compounds of training set, the PCA approach yielded consistently superior rankings compared to conventional consensus scoring and single scoring. By virtual screening of 113,228 compounds, and application of PCA approach to re-rank the score list, two drug like BACE-1 inhibitors were emerged as an effective low-micromolar inhibitors. It suggested that the application of PCA provides a more robust strategy for ranking compounds. The advantages of PCA are as follows. First, PCA is efficient. For each five pose group, the application of PCA can result in superior enrichment of known inhibitors compared to either the conventional consensus scoring or the best individual scoring. In addition, the application of PCA for postprocessing of the scoring data from virtual screening was not timeconsuming. Second, PCA is reliable. PCA is mainly useful when there is limited knowledge about the target and its inhibitors. If we have no idea which scoring function would return the best enrichment rates (i.e., several known active compounds are required to determine the best scoring functions), then adopting PCA to formulate a new scoring function can provide better performance than blindly using a scoring function or some combination scoring functions when performing virtual screening 50 , and training with a data set were not necessary. When a training set was available, there are several other groups that perform a different type of post-docking processing using statistical methods and data mining. Wilton et al. discussed the use of several rank-based virtual screening methods, such as binary kernel discrimination, similarity searching, sub-structural analysis, support vector machine (SVM), and trend vector analysis, for prioritizing compounds in lead-discovery programs [76, 77] . Jacobsson et al. employed three different multivariate statistical methods including PLS discriminant analysis, rule-based methods, and Bayesian classification to analyze multidimensional scoring data from four different target proteins (i.e., the estrogen receptor R (ERR), matrix metalloprotease 3 (MMP3), factor Xa (fXa), and acetylcholine esterase (AChE)). The classifiers that they built showed that the precision is approximately 90% for three of the targets and approximately 25% for acetylcholine esterase for correctly predicting an active compound [78] . The difference between our work and their's is that we do not need a training set because PCA is a form of unsupervised learning and relies entirely on the input data itself. In addition, PCA is simpler than the methods mentioned above (SVM, trend vector analysis, PLS discriminant analysis, rule-based methods, and Bayesian classification). With eight different scoring functions, Terp et al. docked a set of known inhibitors to three different matrix metalloproteases. They obtained scores analyzed using PCA and partial least-squares methods (PLS) [79] . The regression model they built has a good q 2 for predicting the activity of active compounds. The major difference between the present work and the work performed by Terp et al. is that we performed structure-based virtual screening on both active and inactive compounds. Terp et al. included only known inhibitors to quantitatively predict the binding affinity. They did not discuss whether the docking scores have been calculated from a docking mode of an inactive compound that does not actually bind. In the virtual screening process, more attention is focused on how to identify promiscuous active compounds in a database of mainly inactive compounds rather than on how to rank a set of known binders (i.e., predicting the binding affinity of the different active compounds). It should be emphasized the necessity of experimental validation for potential researchers, because no ranking method may help if not associated with verification in the experiment. In an initial attempt, we applied the conventional consensus scoring method to re-rank the score list and experimental test through BACE-1 FRET assay, no inhibitor was found. However, when we applied the PCA scoring method and experimental test through BACE-1 FRET assay, a remarkable 10% hit rate was achieved. On this basis, we summed up some experience as followed: when virtual screening of a new chemical database, the potential researchers usually do not know which kind of individual scoring function work best for the target protein, furthermore, for consensus scoring protocol, they are uncertain which kind of scoring functions should be used to combine for getting the best enrichment rates. Once trapped in this dilemma, the researchers could use PCA scoring protocol to re-rank the results from the virtual screening, a prominent advantage of application of PCA scoring protocol can summarize most of the information from the original scoring functions and improve the enrichment rate, which has been proved to be robust and reliable in the present study. In conclusion, although the PCA approach is not intended to improve all aspects of virtual screening, such as generating more accurate binding poses, it extends conventional consensus scoring in a quantitative statistical manner, therefore, it has great potential for use in the virtual screening process. Future experiments are needed to further analyze the performance of PCA for other receptor binding sites. We believe that the two low-micromolar inhibitors described here may represent a starting point for finding potent and selective molecules capable of preventing BACE-1 activity for the treatment of Alzheimer's disease.

2,500-year Evolution of the Term Epidemic

The term epidemic (from the Greek epi [on] plus demos [people]), first used by Homer, took its medical meaning when Hippocrates used it as the title of one of his famous treatises. At that time, epidemic was the name given to a collection of clinical syndromes, such as coughs or diarrheas, occurring and propagating in a given period at a given location. Over centuries, the form and meaning of the term have changed. Successive epidemics of plague in the Middle Ages contributed to the definition of an epidemic as the propagation of a single, well-defined disease. The meaning of the term continued to evolve in the 19th-century era of microbiology. Its most recent semantic evolution dates from the last quarter of the 20th century, and this evolution is likely to continue in the future.

A t the start of the 21st century, epidemics of infectious diseases continue to be a threat to humanity. Severe acute respiratory syndrome, avian influenza, and HIV/AIDS have, in recent years, supported the reality of this threat. Civil wars and natural catastrophes are sometimes followed by epidemics. Climate change, tourism, the concentration of populations in refugee camps, the emergence of new human pathogens, and ecologic changes, which often accompany economic development, contribute to the emergence of infectious diseases and epidemics (1) . Epidemics, however, have occurred throughout human history and have influenced that history. The term epidemic is ≈2,500 years old, but where does it come from? When works that put forward new ideas are translated, determining the original terminology (in Ancient Greek in this case) is not easy. In 430 BC, when Hippocrates was collecting the clinical observations he would publish in Epidemics, his treatise that forms the foundation of modern medicine, at least 3 terms were used in Ancient Greece to describe situations that resembled those described by Hippocrates: nosos, phtoros, and loimos (2) . Nosos, meaning disease, was used by Plato in the 4th century BC and clearly had the same meaning 2 centuries earlier in the works of Homer and Aeschylus. Nosos encompasses disease of the mind, body, and soul: physical, including epilepsy, and moral (i.e., psychological and psychiatric). Phtoros or phthoros means ruin, destruction, deterioration, damage, unhappiness, and loss, after war for example. The word was frequently used by Aeschylus and Aristophanes, was known in the 8th century BC, and was later used by Plato and Thucydides. Its meaning has remained general. Bailly translates loimos as plague or contagious scourge. Used by Esiodus in the 7th century BC and later by Sophocles and Herodotus, this term is ancient. Its translation as plague should be interpreted in the sense of a scourge rather than as the disease plague. In the Septuagint, a translation of the Old Testament into Greek by 70 Greek Jews from Alexandria, this word is used in the book of Kings to describe the 10 plagues of Egypt. But the term epidemic already existed in 430 BC. The Greek word epidemios is constructed by combining the preposition epi (on) with the noun demos (people), but demos originally meant "the country" (inhabited by its people) before taking the connotation "the people" in classical Greek. Indeed, the word epidemios was used by Homer, 2 centuries before Hippocrates, in the Odyssey (canto I, verses 194 and 230), where it was used to mean "who is back home" and "who is in his country" in contrast to a voyager who is not: , "because someone said that your father was back (home)" (canto I, verse 194). In this context, epidemios means indigenous or endemic. In the Iliad, Homer confirmed this meaning (canto XXIV, verse 262), by using 2,500-year Evolution of the Term Epidemic also polemos epidemios to mean civil war: , "this one who liked passionately the frightening civil war" (canto IX, verse 64). Later, Plato and Xenophon (400 BC) used the word to describe a stay in a country or the arrival of a person: , "a Parian who, I learned, was in town" (Plato, Apology, chapter I, paragraph 38). The verb epidemeo was used by Thucydides (460 BC-395 BC) to mean "to stay in one's own country," in contrast to apodemeo, "to be absent from one's country, to travel." For Plato, epidemeo meant "to return home after a voyage, to be in town." Later, the orators Demosthenes (384 BC-322 BC) and Eschines (390 BC-314 BC) used this word to refer to a stranger who came to a town with the intention of living there, and the verb epidemeo was used to mean "to reside." Typical of Greek semantics, epidemeo takes its meaning from the result of the action, rather than from the action itself. It relates to something that has already happened, with the implication that it had previously happened elsewhere. Authors before Hippocrates used epidemios for almost everything (persons, rain, rumors, war), except diseases. Hippocrates was the first to adapt this word as a medical term. Written in the 5th century BC, Hippocrates' Corpus Hippocraticum contains 7 books, titled Epidemics (3). Hippocrates used the adjective epidemios (on the people) to mean "which circulates or propagates in a country" (4). This adjective gave rise to the noun in Greek, epidemia. We do not know why Hippocrates chose epidemios to title his books instead of nosos, a well-established term meaning disease. Examining the meaning of the term before, during, and after his time may help us understand his choice. Schematically speaking, epidemios (or epidemeo) was used successively to mean "being at homeland" (Homer), "arriving in a country" or "going back to homeland" (Plato), and later "stranger coming in a city" (Demosthenes). Sophocles (495 BC-406 BC) used the adjective in Oedipus Tyrannos to refer to something (a rumor, noise, fame, or reputation) spreading in a country: ???' ???????? ??? ?'?? ???????? ?'??? , "I shall go (to make war) to Oedipus, against his fame which spread (in the country)" (verse 494). Oedipus Tyrannos was written at approximately the same time as Corpus Hippocraticum; consequently, we can infer that during Hippocrates' time, epidemios acquired a dynamic meaning, probably more adapted to describing a group of physical syndromes that circulate and propagate seasonally in a human population (i.e., on the people) than nosos, a term used to describe diseases at the individual level. How epidemios, meaning "on the people," became adapted to mean "that which circulates or propagates in a country" is a crucial question. This evolution occurred during the second half of the 5th century (450 BC-400 BC), a period of intense activity in Greek literature, particularly with the prolificacy of Sophocles. But while nosos or loimos were frequently used, epidemios was not. In the Perseus Digital Library (www.perseus.tufts.edu), a database that does not yet include Hippocrates' works, the adjective epidemios was used only 9 times, including 4 times in Homer, in the 489 major referenced Greek texts (≈4.8 millions words, 0.02 occurrences per 10,000 words). Its Doric variants epidamos and epidemos were used 3 times. In comparison, nosos (disease) was used 712 times in the Perseus database (1.47 occurrences per 10,000 words). The verb epidemeo was used 144 times, primarily during the 4th century and always meaning "to live in or return to one's own country." This lack of material makes accurately exploring the reasons for the semantic evolution across centuries difficult. In Oedipus Tyrannos, Sophocles qualified the sense of epidemios as it referred to reputation or fame; fame naturally spreads in a country. , "It is a fact that the disease was propagating in the country" (Epidemics, book I, chapter 3). Although Sophocles used epidemios once in that new sense, Hippocrates established a medical meaning for the term. In Epidemics, books I and III constitute lists of diseases describing clinical cases. Hippocrates compared these cases and grouped them to generate series of similar cases. He adopted a classification approach, initially seeking clinical similarities between cases, thereby discovering, in addition to the notion of epidemic, the more fundamental concepts of symptom and syndrome. However, Hippocrates believed that prognosis was a major aspect of medicine. This belief led him to consider disease a dynamic process with its own progression, a temporal dimension, that represents a first nosologic evolution: syndromic groupings become diseases. Another of the books written by the physician from Kos-Airs, Waters, and Placesdeals with the relationships between diseases and the environment, focusing particularly on the habitat of the patients and the season in which disease occurs. Hippocrates tried to determine the effect of environmental factors on what could be described as the distribution of diseases. He was, thus, more concerned about grouping together winter diseases or autumn diseases or diseases that occurred in a particular place or in persons whose way of life had changed than in identifying a large number of cases of the same disease in winter or autumn, at a particular place, or in association with a particular way of life. For Hippocrates, whose nosologic approach already contained a major element of preoccupation with the environment, the first meaning of epidemic was groups of cases 2,500-year Evolution of the Term Epidemic resembling each other clinically and the second meaning was groups of different diseases occurring at the same place or in the same season and sometimes spreading "on the people." Thus, Hippocrates applied the word epidemios to groupings of syndromes or diseases, with reference to atmospheric characteristics, seasons or geography, and sometimes propagation of a given syndrome in the human population. Semantic confusion caused the great Emile Littré, who translated Hippocrates' works into French in the first half of the 19th century, to make a nosologic error. Hippocrates described what is known today, since the work of Littré, as the Cough of Perinthus. This account can be found in Epidemics book VI. Hippocrates described coughs that started toward the winter solstice and were accompanied by many symptoms: sore throat, leg paralysis, peripneumonia, problems with night vision, voice problems, difficulty swallowing, difficulty breathing, and aches. When Littré published his translation and commentaries on Epidemics in 1846, he mistakenly considered the Cough of Perinthus to be a single disease (5) . This error made retrospectively diagnosing the diseases of Perinthus difficult, if not impossible. Moreover, as Littré saw this collection of illnesses as a single disease, he essentially turned it into an epidemic, probably because he had the modern sense of the term in mind and thought that Hippocrates had observed and described an epidemic illness unknown to modern medicine (5). According to Grmek, "Littré took chapter VI, 7.1 as a general description of an epidemic in the sense of this word in the medical language of the 19th century rather than in the sense intrinsic to the works of Hippocrates. In the Corpus Hippocraticum, the noun 'epidemic' designates a collection of diseases observed at a given place, during a given period. A disease described as epidemic, such as epidemic cough, is a condition occurring from time to time in a given place, the appearance of which is closely linked to changes in season and climatic variations from year to year" (5). Historians of medicine and philologists have over the years attributed the Cough of Perinthus to diphtheria, influenza, epidemic encephalitis, dengue fever, acute poliomyelitis, and many other diseases. However, a French physician named Chamseru, who practiced in the 18th century, almost a century before Littré, finally got to the bottom of what may be meant by the Cough of Perinthus, probably because the term epidemic had not yet taken on the meaning it had in Littré's time. According to Chamseru, the Cough of Perinthus could have encompassed several diseases, among them diphtheria, influenza, and whooping cough (5) . Thucydides (460 BC-395 BC) interrupted his account of the Peloponnesian War to describe the famous Plague of Athens, which occurred at the start of the summer in 430 BC. This description was long considered among the first descriptions of an epidemic. Indeed, whereas Thucydides used nosos, the term plague, which is used by all the translators of his work, is used in the sense of the Latin term pestis, a term with no clear etymology (4), meaning contagious disease, epidemic, or scourge. The description of the Plague of Athens, like that of the Cough of Perinthus by Hippocrates, is an essential text in the philologic and semantic study of epidemics (5) . We must therefore consider, as for Littré's translation, the meaning that translators have assigned to the original description by Thucydides. Thucydides never used the term epidemic that Hippocrates was in the process of establishing. Under the term nosos, Thucydides described a series of clinical signs, which originated in the south of Ethiopia and propagated throughout Egypt, Libya, and then Greece. Thucydides used the words nosos, kakos (evil), ponos (pain), phtoros (ruin, destruction), and loimos (scourge) to describe what his translators call plagues. In her translation of Thucydides' works (6) , published in 1991, in the chapter titled Second Invasion of Attica: the Plague of Athens (the original Greek work had no title), Jacqueline de Romilly translated nosos as disease or epidemic. Similarly, she translated loimos, kakos, and phtoros as disease or epidemic and the list of clinical signs (the original Greek meant "following these things") as symptoms. de Romilly rendered the text more elegant and accessible to 20th-century readers by this translation but gave the words used by Thucydides their 20th-century meaning rather than the meanings they had in the 5th century BC. Herein lies the principal problem of translation. But was the Plague of Athens a true epidemic, in the modern sense? The death rate for the disease was extremely high, reaching up to 25% in 1 group of soldiers, and Pericles died of it. Historians have tried to understand the origin of this plague, and various diseases have been suggested, e.g., typhus, measles, smallpox, bubonic plague, ergotism, or an unknown disease. Thucydides wrote that all preexisting diseases were transformed into a plague and that persons in good health were affected in the absence of a predisposing cause (7) . The large number of symptoms and of possible and probable causes rules out the possibility of an epidemic in the modern sense of the term. Instead, the Plague of Athens seems to have been the appearance of a large number of diseases that affected the population at the same time. Plague therefore has the same meaning here as epidemic in the works of Hippocrates. These 2 terms have been used in association or confused throughout history. However, epidemic existed at this time, even if the notion of epidemic as we mean it in modern times had not yet emerged. After the nonmedical use of the term epidemic by Homer, Sophocles, Plato, and Xenophon, Hippocrates gave it its medical meaning. However, the term has since undergone a long evolution. The adjective epidemios gave rise to the Greek noun epidemia. The Greek term epidemia in turn gave rise to the Latin term epidimia or epidemia. The term ypidime in Medieval French has its origins in these Latin words and went on to become épydime in the 14th century, epidimie in the 17th century, and then epidémie in the 18th century. Not until 22 centuries after Hippocrates, in the second half of the 19th century, were the terms épidémiologie (1855), épidémiologique (1878), and épidémiologiste (1896) coined in French and notions attached to them developed. Of course, at approximately the same time, corresponding terms appeared in the English language. The term epidemic and the terms linked to it therefore required an extremely long time to be constructed. This evolution is representative of the evolution of science and medicine over the centuries and reflects the semantic evolution of the term. In parallel with the evolution of the term epidemic itself, its meaning also changed over time. If we limit ourselves to the meaning that epidemic has acquired with respect to infectious diseases, we can identify 4 major steps in its semantic evolution in the medical sense. For Hippocrates, an epidemic meant a collection of syndromes occurring at a given place over a given period, e.g., winter coughs on the island of Kos or summer diarrheas on other islands. Much later, in the Middle Ages, the long and dra-matic succession of waves of The Plague enabled physicians of the time to identify this disease with increasing precision and certainty; they began to recognize epidemics of the same, well-characterized disease. Then, with the historic contributions of Louis Pasteur and Robert Koch, epidemics of a characteristic disease could be attributed to the same microbe, which belonged to a given genus and species. The last stage in the semantic evolution of the term epidemic was the progressive acquisition of the notion that most epidemics were due to the expansion of a clone or clonal complex of bacteria or viruses known as the epidemic strain (8) . More recently, microevolution of a clone of a bacterium (the epidemic strain) was shown to occur during an epidemic with person-to-person transmission (9) . The Table summarizes these 4 major stages in the semantic evolution of the term epidemic. In the second half of the 20th century, epidemic was also applied to noninfectious diseases, as in cancer epidemic or epidemic of obesity. The extension of the meaning to noninfectious causes refers to a disease that affects a large number of people, with a recent and substantial increase in the number of cases. This semantic extension of epidemic also concerns nonmedical events; the term is used by journalists to qualify anything that adversely affects a large number of persons or objects and propagates like a disease, such as crack cocaine or computer viruses. What can we gain from investigating the origin and meaning of the word epidemic or from studying its semantic evolution? Beyond simply satisfying our curiosity, the slow evolution of the form and meaning of the term suggests that we still have much to learn about the concept of epidemic. Dr Martin is a bacteriologist and chef de laboratoire at the Pasteur Institute and director of the Pasteur Institute of New Caledonia. His research interests focus on epidemics. Ms Martin-Granel is a professeur certifiée de lettres classiques and teaches French literature, Latin, and Ancient Greek in secondary school in the south of France. She has recently translated a text by Petrarch into French.

The Acute Environment, Rather than T Cell Subset Pre-Commitment, Regulates Expression of the Human T Cell Cytokine Amphiregulin

Cytokine expression patterns of T cells can be regulated by pre-commitment to stable effector phenotypes, further modification of moderately stable phenotypes, and quantitative changes in cytokine production in response to acute signals. We showed previously that the epidermal growth factor family member Amphiregulin is expressed by T cell receptor-activated mouse CD4 T cells, particularly Th2 cells, and helps eliminate helminth infection. Here we report a detailed analysis of the regulation of Amphiregulin expression by human T cell subsets. Signaling through the T cell receptor induced Amphiregulin expression by most or all T cell subsets in human peripheral blood, including naive and memory CD4 and CD8 T cells, Th1 and Th2 in vitro T cell lines, and subsets of memory CD4 T cells expressing several different chemokine receptors and cytokines. In these different T cell types, Amphiregulin synthesis was inhibited by an antagonist of protein kinase A, a downstream component of the cAMP signaling pathway, and enhanced by ligands that increased cAMP or directly activated protein kinase A. Prostaglandin E2 and adenosine, natural ligands that stimulate adenylyl cyclase activity, also enhanced Amphiregulin synthesis while reducing synthesis of most other cytokines. Thus, in contrast to mouse T cells, Amphiregulin synthesis by human T cells is regulated more by acute signals than pre-commitment of T cells to a particular cytokine pattern. This may be appropriate for a cytokine more involved in repair than attack functions during most inflammatory responses.

Different functional subsets of CD4 T cells are crucially involved in immune defense against diverse pathogens. At least four effector subsets are derived by differentiation from naïve CD4 T cells, and each expresses a characteristic combination of transcription factors, soluble mediators and surface molecules [1, 2] . Th1 cells predominantly produce interferon-c (IFNc) and protect against intracellular pathogens; Th2 cells produce interleukin (IL)-4, IL-5 and IL-13 and help to eliminate extracellular parasites; Th17 cells produce IL-17a and IL-17f and are crucial in fighting against extracellular bacteria and fungi; whereas induced T regulatory cells (iTreg) produce IL-10 and Transforming Growth Factor b (TGFb), and suppress T and B cell effector responses. Although the initial Th1 and Th2 subsets are relatively stable, recent studies have demonstrated some flexibility and plasticity, particularly in other CD4 T cell subsets. Cytokines and other soluble mediators in the lymph node and inflamed tissue can further affect the cytokine expression profile of effector CD4 T cells [3] , in some cases by changing the differentiation status of the effector CD4 T cells. Primed precursor CD4 T (Thpp) cells, that produce mainly IL-2 and chemokines when stimulated, remain uncommitted with respect to their effector cytokine pattern and can later differentiate into either Th1 or Th2 cells [4] [5] [6] . Suppressive Treg cells, expressing the forkhead transcription factor Foxp3, can lose the expression of Foxp3 and acquire the ability to produce pro-inflammatory cytokines during autoimmunity [7] . Th17 cells can acquire the ability to produce IFNc in Th1 polarizing conditions [8, 9] . Adoptively transferred IL-4-producing Th2 effector cells can produce IFNc during viral challenge infections [10] . Th9 cells develop from Th2 populations in the presence of TGFb [11, 12] and T follicular helper (Tfh) cells may represent a further differentiation step from several of the other subsets [13] . Acute modifications of cytokine patterns can also occur. IL-12+ IL-18 enhance the secretion of IFNc by Th1 cells [14, 15] , and IL-2 enhances cytokine production [16, 17] . In contrast, IL-10, TGFb, prostaglandin E2 (PGE2) and adenosine inhibit inflammatory cytokine production [18] [19] [20] . Mouse Th2 cells, but not naive or Th1 cells, express Amphiregulin (AR), a member of Epidermal Growth Factor (EGF) family. Like other EGF members, AR is expressed as a transmembrane precursor protein and released by proteolytic cleavage [21, 22] . Soluble AR binds to EGF receptors and promotes proliferation and differentiation of epithelial cells, fibroblasts and keratinocytes [23] [24] [25] . AR-deficient mice [26] showed slower kinetics of clearance [27] of the helminth parasite, Trichuris muris, that is cleared most effectively by Th2-biased responses. AR production is also induced in human mast cells by IgE cross-linking [28, 29] , in human eosinophils by IL-5 [30] , and in human basophils by IL-3 [31] . Thus AR is induced by activation of at least four cell types contributing to Type 2 inflammation, suggesting a role for AR during an allergic immune response. In addition, production of AR by immune cells is potentially important for tissue remodeling and repair [21, 26] during and after damaging immune responses. Very little is known about the regulation of AR gene expression in human T cells. We examined the regulation of AR synthesis by human T cells, and found that in contrast to mice, many subsets of human T cells, including CD4 and CD8, naive and memory, Th1 and Th2, all express AR in response to TCR stimulation. Factors that elevate cAMP levels synergized with TCR stimulation to enhance AR expression, while inhibiting expression of most inflammatory cytokines. Thus in human T cells, AR production is regulated strongly by the environmental context during stimulation, but not restricted to particular precommitted effector subsets of T cells. Biotinylated goat anti-human AR and biotinylated goat normal IgG (isotype control), and APC conjugated anti-human CCR4 (205410) were obtained from R&D Systems (Minneapolis, MN). LEAF TM purified anti-human CD3e (OKT3), APC-Cy7 conjugated anti-human CD4 (RPA-T4), Pacific Blue or APC-Cy7 conjugated anti-human CD69 (FN50), PE-Cy5 conjugated antihuman CD154 (24-31), PerCP-Cy5.5 conjugated anti-human CD27 (O323), Alexa Fluor 700 conjugated anti-human CD62L (DREG-56), Pacific Blue conjugated anti-human CXCR3 (TG1/ CXCR3), PE-Cy7 or PE-Cy5 conjugated anti-human CD123 (6H6), Alexa Fluor 700 conjugated anti-human IL-2 (MQ1-17H12), FITC conjugated anti-human IL-4 (MP4-25D2), and PerCP-Cy5.5 conjugated anti-human IL-17A (BL168) were purchased from BioLegend (San Diego, CA). Functional grade purified anti-human CD28 (CD28.2), PE conjugated anti-human CD45RA (HI100), FITC conjugated anti-human CD45RO (UCHL1), PE-Cy5 conjugated anti-human CD19 (HIB19), PE-Cy7 conjugated anti-human IFNc (4S.B3), and APC-conjugated streptavidin were obtained from eBioscience (San Diego, CA). Alexa Fluor 488 conjugated anti-human CXCR5 (RF8B2) and PE-Cy7 conjugated anti-human CCR7 (3D12) were purchased from BD Bioscience (San Jose, CA). QdotH 605 conjugated antihuman CD3 (UCHT1), PE-Texas Red and QdotH 705 conjugated anti-human CD8a (3B5), PE-Texas Red conjugated anti-human CD4 (S3.5), TRI-COLOR and QdotH 800 conjugated antihuman CD14 (TüK4), QdotH 655 conjugated anti-human CD45RA (MEM-56), Pacific Blue conjugated anti-human TNFa (MP9-20A4), and LIVE?DEAD Fixable Yellow Dead Cell Stain Kit were obtained from Invitrogen (Carlsbad, CA). 7-Aminoactinomycin D (7-AAD) and TAPI-1 was obtained from Calbiochem (Gibbstown, NJ). cAMP agonist (8-CPT-cAMP) and cAMP antagonist (Rp-8-Br-cAMP) were purchased from BioLog (Bremen, Germany). Phorbol 12-myristate 13-acetate (PMA), ionomycin, monensin, PGE2, forskolin and 3-Isobutyl-1methylxanthine (IBMX), adenosine were obtained from Sigma (St.Louis, MO). Heparinized blood was obtained from healthy donors under a protocol approved by the University of Rochester Medical Center Research Subjects Review Board. Written, informed consent was obtained from all subjects. PBMC were isolated by Ficoll-Hypaque (Cellgro, Herndon, VA) density gradient centrifugation. Cells were suspended in complete RPMI-8 (RPMI-1640 medium containing 100U penicillin/streptomycin (Invitrogen) supplemented with 8% heat-inactivated fetal calf serum (FCS, HyClone, Logan, UT)). In the experiment treating cells with adenosine, serum-free medium X-VIVO TM 20 (Lonza, Walkersville, MD) was used. To purify human naïve and memory CD4 T cells from PBMC, fresh PBMC were stained with antibodies specific for cell surface markers and CD4+CD8-CD14-CD123-CD45RA+CD45RO-(naïve CD4 T cells) and CD4+CD8-CD14-CD123-CD45RA-CD45RO+ (memory CD4 T cells) were sorted on a FACSAria (BD Bioscience, San Jose, CA). Purified human naïve CD4 T cells were stimulated with irradiated (100Gy) allogeneic Epstein-Barr virus (EBV) -transformed B cells (1:1 ratio) in complete RPMI-8 medium at 10 5 cells/mL in round-bottom 96-well plate. Th1-biased cultures contained recombinant human IL-2 (5 ng/mL, PeproTech), recombinant human IL-12 (20 ng/mL, PeproTech) and anti-IL-4 (5 mg/ml, R&D Systems). Th2-biased cultures contained recombinant human IL-2 (5 ng/mL), recombinant human IL-4 (20 ng/mL, R&D Systems), anti-IL-12 (5 mg/ml, ebioscience) and anti-IFNc (5 mg/ml, R&D Systems). Fresh medium containing 5 ng/mL IL-2 was added if necessary to cultures showing strong proliferation. The cultures were restimulated and expanded every seven days. To enrich for cells with the Th1 or Th2 phenotypes, after 14 days priming, Th1 and Th2 cells were stimulated with platebound anti-CD3+ anti-CD28 for 8 hours. IFNc+ Th1 cells and IL-5+ Th2 cells were stained and sorted by the MACS cytokine secretion assay (Miltenyi Biotec, Auburn, CA) according to the manufacturer's instructions. The enriched IFNc+ Th1 cells and IL-5+ Th2 cells were expanded as previously for 14 days. For intracellular staining (ICS) of AR, PBMC (10 6 per well) were stimulated with medium alone, anti-CD3 (5 mg/ml) + anti-CD28 (1 mg/ml), Staphylococcal enterotoxin B (SEB, 1 mg/ml), PMA (10 ng/mL) + ionomycin (500 ng/mL), influenza H1N1 peptides (H1N1 [New Caledonia/New York], 20 ng/mL/peptide), Fel d1 (50 mg/mL, INDOOR, Charlottesville, VA), Der p1 (50 mg/mL, INDOOR) or Tetanus peptides (3 mg/mL/peptide) in round-bottom 96-well plate (Costar, Corning Inc., Corning, NY). Th1 or Th2 cultures were treated with medium alone or PMA + ionomycin. After 10 hours stimulation (with 2 mM monensin present for the last 8 hours), the cells were first stained with LIVE?DEAD Fixable Yellow Dead Cell Stain Kit, and then stained for cell surface markers CD4, CD8, CD14, CD123, and CD45RA. After cells were fixed and permeabilized using Fix-Perm (BD Bioscience), the cells were stained with anti-AR, anti-IFNc, anti-IL-2, anti-IL-4, anti-IL-17A, anti-TNFa, anti-CD3 and anti-CD69 (or anti-CD154) intracellularly. For cell surface staining of AR, PBMC were stimulated with medium alone or 1 mg/ml SEB in the presence or absence of 50 mM TAPI-1, an ADAM17 protease inhibitor [32] , for 6, 12, and 24 hours. After stimulation, the cells were stained with antibodies against AR, CD3, CD4, CD8, CD14, CD123, CD69 and 7-AAD. Data were acquired using an LSR II flow cytometer (BD Bioscience), and analyzed with FlowJo software (Tree Star Inc., Ashland, OR). Total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer's instructions. cDNA was prepared by reverse transcription from total RNA using MultiScribe TM Reverse Transcriptase (Applied Biosystems, Foster City, CA) with random hexamer primers (Applied Biosystems). Quantitative real-time PCR (RT-PCR) was performed using the Applied Biosystems 7900HT Sequence Detection System. Primers and probes specific for AR, Heparin-binding EGF-like Growth Factor (HB-EGF), IL-2, IFNc, IL-3, IL-4, IL-5, IL-10, IL-13, CD3d, EGF, Neuregulin (NRG) 1-4, epiregulin (EREG), betacellulin (BTC) and TGFa were all obtained from TaqMan Gene Expression Assays (Applied Biosystems). CD3d gene expression was used as an endogenous control for normalizing mRNA amounts. All samples were run in duplicate and data were analyzed using SDS software (Applied Biosystems). Purified CD4 T cells were treated with medium alone or CD3/ CD28 beads (cells:beads 2:1) in the presence or absence of 50 mM TAPI-1. After 24 hours, the supernatants were collected and AR was measured using the human Amphiregulin DuoSet ELISA Development kit (R&D Systems). The detection limit of the assay was 7.8 pg/mL. Because the antiserum for the ELISA was produced by immunization with bacterial recombinant human AR, and the standard is also non-glycosylated AR, this ELISA probably underestimates the concentration of normal human glycosylated AR. Mouse Th2 cells produce AR in response to TCR-mediated activation [27] , and the expression of AR by hemopoietic cells contributes to the clearance of a helminth parasite. However, our recent studies showed that basophils were the major human PBMC type that produced AR in response to anti-CD3/CD28 stimulation [31] , whereas production of AR by T cells was much lower. Therefore we examined human T cells in more detail, to determine whether human T cells could produce AR, and if so, whether this was produced preferentially by human Th2 cells. Human PBMCs were stimulated with soluble anti-CD3+ anti-CD28, SEB, or PMA + ionomycin for 10 hours (protein secretion inhibitors were added during the last 8 hours). CD69 staining increased on almost all anti-CD3/CD28-and P+I-stimulated cells, and a subset of SEB-stimulated cells ( Figure 1A ). AR staining was increased, only in the CD69+ population, and this increase was most obvious in the P+I-stimulated cells. For all three stimulation conditions, the staining intensity for AR increased for the whole CD69+ population, i.e. separate positive and negative populations were not resolved, and so the percentage of cells in the AR+ gate may be an underestimate of the total number of cells expressing AR. The specificity of AR staining was demonstrated by using a control goat antiserum (right column). Similar results were obtained with CD8 T cells ( Figure 1A) . To independently confirm AR expression by human T cells, and to test whether T cells produced AR as a direct result of TCR stimulation, human CD4 and CD8 T cells were purified by sorting, and stimulated with beads coated with anti-CD3+ anti-CD28 antibodies. At different times, RNA was extracted from the cells, and levels of AR and IL-2 mRNA measured by RT-PCR. AR mRNA levels increased rapidly after stimulation, and returned to low levels after ten hours, whereas IL-2 showed slower kinetics ( Figure 1B) . The kinetics of AR production were similar in CD4 and CD8 T cells. Thus human T cells directly express AR in response to polyclonal TCR stimulation. As demonstrated by other studies [33] , HB-EGF mRNA was also upregulated in activated human CD4 T cells (Figure 2A) , although the levels were lower than AR and peaked at a later time ( Figure 2B ). TGFa and EREG mRNA were also detected in resting CD4 T cells, but not increased during TCR activation. Other EGF members were undetectable. In our previous mouse experiments [27] , AR and HB-EGF were also the only EGF family members induced by TCR stimulation (data not shown). Expression of HB-EGF protein was confirmed by cell surface and intracellular staining (data not shown). EGF family members (including AR) are initially expressed as transmembrane proteins and released into the extracellular region after cleavage by metalloproteases, particularly ADAM17 [22] . To determine whether T cells also initially expressed surface AR and then released the soluble cleavage product, surface AR was stained during TCR activation in the presence or absence of the ADAM17/TACE inhibitor TAPI-1 [32] . TAPI-1 increased AR expression on the surface of both CD4 and CD8 T cells measured by frequency ( Figure 3A ) or fluorescence intensity (data not shown). Conversely, TAPI-1 decreased soluble AR in the supernatant ( Figure 3B ). In the absence of TAPI-1, AR expression on T cells gradually decreased and was barely detectable after 24 hours. As ADAM17 mRNA was detected by RT-PCR in resting human T cells and upregulated on activation (data not shown), these results suggested that AR was first synthesized as a membrane protein on human T cells and then released by ADAM17 cleavage, as in other cell types [34, 35] . In mice, AR was expressed selectively in TCR-activated Th2 cells [27] but not Th1 ( Figure S1 ) or naive CD4 T cells. This was a pre-committed, intrinsic property of the Th2 cells, as Th2 but not Th1 cells expressed AR even when in vitro-derived mouse Th1 and Th2 cell lines were activated together in the same culture (data not shown). However, most human CD4 (and CD8) T cells expressed AR in response to PMA plus ionomycin stimulation ( Figure 1A ). We therefore examined in more detail which human T cell subsets were responsible for AR production. Naive and memory CD4 and CD8 T cells produce AR. To examine the ability of naive and memory T cell subpopulations to express AR, we stimulated PBMC with an allogeneic EBV-transformed B cell line, which would be expected to activate a small fraction of both memory and naive CD4 and CD8 T cells. Alloantigens stimulated a fraction of both CD4 and CD8 T cells to produce AR ( Figure 4A ), relative to the unstimulated control. The specificity of staining was confirmed by isotype control antibodies. The cells producing AR (and other cytokines) were included in the CD69+ population. AR was induced by allogeneic stimulation in both CD45RAand CD45RA+ subsets of CD4 and CD8 T cells at frequencies ranging from 0.028% to 0.35%. These levels were comparable to the frequencies of CD4+ CD45RA+ or CD45RA-T cells producing IL-2, or CD4+ CD45RA-memory T cells producing IFNc. As expected, IL-2 was produced by both memory (CD45RA-) and naive (CD45RA+) CD4 T cells, whereas IFNc (and IL-4 at low levels) were produced mainly by memory cells ( Figure 4B ). The expression of AR by both CD45RA-and CD45RA+ subsets of CD4 T cells was tested at the mRNA level in SEBstimulated cells sorted according to AR and CD45RA expression ( Figure 4C ). Confirming the specificity of the anti-AR antibody AR is produced by memory CD4 T cell subsets expressing different cytokine phenotypes. Although naive CD4 T cells are relatively homogeneous, the memory population includes a wide range of differentiated effector subsets. As AR is expressed selectively by mouse Th2 cells, we examined whether AR production by human CD4 memory T cells was preferentially associated with expression of a particular cytokine or surface marker pattern. Th1-and Th2-biased human CD4 T cell populations were induced by stimulation of sorted naive human CD4 T cells with an allogeneic B cell line in Th1-or Th2-biasing cytokine conditions. The populations were further enriched by using the Cytokine Secretion Assay to sort IFNc-or IL-5producing cells, respectively. The resulting populations were strongly polarized, but unlike mouse T cells, both Th1 and Th2 human cell lines expressed AR ( Figure 5A ). These results were confirmed using ex vivo human CD4 T cell populations. Human PBMC were stimulated with SEB, and AR and other cytokines measured by intracellular staining. Naive cells (CD45RA+) expressed high levels of IL-2 and AR, but very low levels of either IFNc or IL-4 (data not shown). Memory cells produced all cytokines tested, at varying frequencies. To determine whether AR expression was associated positively or negatively with subset-specific cytokines, the frequencies of cells expressing AR plus each of the other cytokines were measured from the ICS results. These values were then compared with the double-producing frequencies predicted for random association of each cytokine pair, by multiplying the individual frequencies for each cytokine. Figure 5B shows that AR was expressed in association with TNFa, IL-2, IFNc, IL-4 and IL-17 at slightly higher frequencies than predicted by random association. Similarly, TNFa and IL-2 showed positive associations with all other cytokines. In contrast, the subset-specific cytokines IFNc, IL-4 and IL-17 showed mostly negative associations between each other, as expected. These results were confirmed at the RNA level by sorting SEB-stimulated human PBMC according to surface AR expression. Both AR+ and AR-memory CD4 T cell populations expressed similar levels of IL-4 and IFNc as measured by RT-PCR ( Figure 5C ). IL-2 mRNA levels were higher in AR+ T cells, in both CD45RA-and CD45RA+ cells. AR is produced in response to antigen stimulation. We next tested whether human CD4 T cells expressed AR during antigen/APC stimulation in response to influenza peptides, allergens or tetanus antigens to stimulate Type 1, Type 2 and Thpp-biased recall responses, respectively [6] . PBMCs were stimulated with antigens for 10 hours, and AR and other cytokines measured by ICS. Although these three antigens induced in vivo recall responses with characteristically different levels of IL-2, IFNc and IL-4, all three antigens induced substantial production of AR in the activated (CD154+) cells ( Figure 5D ). Similar results were obtained with cells from multiple subjects, although the magnitudes of the antigen responses were variable for all cytokines. Thus AR can be expressed by all the conventional defined subsets of T cells that we have tested, including CD4 and CD8, naïve and memory, Thpp, Th1 and Th2. To confirm the protein results, influenza-specific CD69+ IFNc+ cells were sorted from two subjects (results from one subject are shown in Figure 5E) , and RT-PCR demonstrated that AR mRNA levels were strongly elevated in the IFNc+ influenza-specific cells AR is produced by T cell subsets expressing different chemokine receptors and surface markers. Chemokine receptors expressed selectively by T cell subsets lead to different homing and chemotactic properties. Expression patterns of chemokine receptors are partly but not entirely related to cytokine commitment patterns [36] [37] [38] [39] . Additional surface markers, including CD27 and the homing receptor CD62L are also expressed heterogeneously on human CD4 T cells. We therefore examined AR expression within subsets of memory CD4 T cells defined by the expression of these proteins. AR was produced at approximately similar frequencies by CD4 T cells positive or negative for the chemokine receptors CCR4, CCR7, CXCR3 and CXCR5, as well as CD62L and CD27 ( Figure 6 ). However, expression of the activation-induced protein CD69 was strongly correlated with AR expression, as seen in previous figures. Taken together with the data described above, AR expression appears to be a general ability of most or all subtypes of human T cells after TCR activation. As our results had demonstrated that AR production was not limited to a pre-committed subset of T cells, we then tested whether AR production was regulated by acute signals in the immediate milieu during TCR stimulation. In many cell types AR is strongly regulated by the cAMP-PKA-CREB signaling pathway. AR expression was significantly up-regulated by cAMP-elevating agents in both resting and anti-CD3 stimulated human PBMC populations enriched for human T cells [40] . However, in that study the negatively-selected T cell population would also have contained basophils, and we have shown that basophils express AR rapidly in response to IL-3 and cAMP agonist ( [31] and unpublished data). Thus anti-CD3 stimulation of the CD4 T cell + basophil population could have induced IL-3 production by T cells, indirectly resulting in AR production by basophils. We have now re-examined the effect of cAMP elevation on the expression of AR by different T cell subsets. TCR and cAMP signals synergize to induce AR expression. TCR activation alone (which transiently elevates cAMP [41] ) induced transient AR mRNA expression (Figure 7A) , and a strong cAMP agonist (PKA activator 8-CPT-cAMP) also induced low levels of AR in the absence of other signals. However, TCR and PKA signaling synergized to induce higher and more sustained levels of AR and HB-EGF mRNA ( Figure 7A ), as well as high levels of AR protein (supernatant plus cell-associated, Figure 7B ). This strong synergy contrasts with a previous study [40] , possibly due to the presence of basophils in the responding population in that study. As both naive and memory CD4 T cells produce AR (Figure 4 ), we tested whether PKA activation would enhance AR expression in both populations. Figure 7C shows the response of purified Figure 5 . Several human CD4 T cell subsets can produce AR. (A) Allogeneic Th1 and Th2 cell lines from three subjects were stimulated with PMA + ionomycin for 6 hours. The percentage of cells expressing IFNc, IL-4, and AR was analyzed by ICS. (B) The expression of AR and other cytokines was measured in SEB-stimulated PBMC from four subjects by ICS, calculating the frequencies of single cytokine producers, and all possible combinations of double-producers, among the CD154+ CD4+ T cells. The figure shows the ratio between the observed frequencies of doubleproducing T cells for each cytokine pair, and the expected frequencies (calculated as the product of the individual frequencies for each cytokine). Values represent the ratios for the double-producer combination defined by the row and column labels. Ratios above or below 1 are indicated by solid or open symbols, respectively. (C) IL-4, IFNc and IL-2 mRNA levels were measured by RT-PCR in the sorted populations described in Figure 4C . (D) PBMC were treated with influenza H1N1 peptides or tetanus (five subjects each), or the allergens Fel d1 (solid symbols) or Der p1 (open symbols)(three subjects each). The numbers of memory CD4 T cells expressing AR and other cytokines were measured by ICS. The backgrounds (no antigen) have been subtracted. Each symbol represents one individual and the filled bar is the mean of all tested subjects. (E) CD69+ CD4+ T cells (Control_CD69+) were sorted from PBMC incubated in medium alone. CD69+IFNc+ and CD69+IFNc-CD4 T cells were sorted from influenza peptidetreated PBMC using the cytokine secretion assay. The mRNA levels of IFNc and AR were measured by RT-PCR. Results in (A-C) are representative of at least three experiments, (D) represents two experiments using a total of 5 independent subjects, and (E) represents two experiments. doi:10.1371/journal.pone.0039072.g005 AR expression is modified by natural and synthetic modulators of cAMP signaling. In the experiments described above, cAMP signaling was altered by an agonist (8-CPT-cAMP) that directly targeted PKA to mimic the increase of intracellular cAMP levels. To further confirm that the cAMP-PKA-CREB signaling pathway regulates AR expression, we tested natural and pharmacological agents that increase the intracellular levels of cAMP by acting at two additional steps: PGE2 and adenosine are natural ligands for G-protein coupled receptors that activate adenylyl cyclase [20, [42] [43] [44] ; forskolin activates adenylyl cyclase directly [20] ; and IBMX is a broad inhibitor of cAMP-degrading phosphodiesterases [45] . Consistently, all four cAMP elevating agents upregulated AR mRNA and protein expression in anti-CD3-stimulated T cells. In each case, the elevated signal was blocked by the cAMP antagonist ( Figure 7D ). The enhancement of AR by the PDE inhibitor suggested that PDE reduced the moderate levels of cAMP induced by TCR activation in CD4 T cells [41] . AR and other cytokines are regulated reciprocally by cAMP signals. In contrast to the enhancement of AR expression, the cAMP agonist inhibited expression of many other cytokines ( Figure 7E ), and all four PKA-activating agents described in Figure 7 inhibited expression of IL-2 and IFNc (data not shown). These results are consistent with previous studies with cAMP agonists and natural cAMP elevating agents, such as PGE2 and adenosine [19, 20] . Thus AR expression in T cells is enhanced under conditions that suppress the production of many other cytokines. In contrast to the preferential expression of AR by mouse Th2 cells, we have now shown that synthesis of human AR is not restricted to a particular human T cell subset. AR can be produced by activated naive and memory CD4 and CD8 T cells, including Th1 and Th2 phenotypes. Our results suggest that AR is not a specific product of certain pre-committed effector subsets of human CD4 T cells, but instead is regulated mainly by additional signals present during T cell activation, particularly signals influencing the cAMP signaling pathway. The lack of precommitment suggests that, in contrast to the memory of effector functions carried by T cells committed to Th1, Th2, Th17 etc phenotypes, the amount of AR produced in a particular immune response is regulated by the local environment during that response, but is less influenced by previous immune priming. The discrepancy we have identified between mouse and human T cell regulation highlights the importance of performing cross-species comparisons of effector T cell phenotypes. AR production was not restricted to a defined T cell effector subset, but AR and IL-2 levels were moderately correlated in both naïve and memory CD4 T cells. Although this could indicate the existence of a previously-unrecognized subset, it is possible that the correlation could be the result of shared transcriptional or mRNA stability regulatory factors, or to similar activation thresholds for IL-2 and AR. Expression of AR also showed moderate correlation with the expression of TNFa. High levels of AR mRNA and protein were induced by synergy between TCR signals and signals that elevated cAMP or activated PKA. This contrasts with a previous report suggesting that both resting and anti-CD3 stimulated T cells significantly up-regulated AR in response to a cAMP agonist [40] . However, the enriched T cell population used in that study was purified by negative selection and very likely included basophils, which we have shown are potent producers of AR in response to IL-3 [31] . AR expression is also strongly enhanced by cAMP agonists in basophils (Y. Qi and T.R. Mosmann, unpublished data) and so it is possible that basophils may have produced the AR in response to the cAMP agonist without TCR stimulation. In contrast to the induction of AR by cAMP elevating agents, these mediators suppress inflammatory responses by inhibiting cytokine expression and T cell proliferation. Synthesis of several pro-inflammatory or Type 1 cytokines is inhibited by cAMP ( Figure 7E and [19, 20, 46, 47] , whereas cAMP can either inhibit or enhance production of Type 2 cytokines such as IL-4, IL-5 and IL-13 ( Figure 7E and [47] ) depending on the stimulation conditions [48, 49] . Natural mediators that elevate the cAMP pathway and lead to PKA activation include PGE2 (mainly via the G protein-coupled receptors E2 and E4 on T cells) and adenosine (mainly via the A 2A receptor on T cells). Both mediators are produced at sites of immune inflammation, adenosine by degradation of ATP from dying cells, and PGE2 by activated macrophages. PKA activation signals also synergized with TCR signals to induce HB-EGF mRNA and protein expression in human CD4 T cells (data not shown). Thus during the progression of an inflammatory response, there may be a switch from pro-inflammatory cytokine production to AR (and HB-EGF) production. Our findings allow us to construct a model of the role of T cell derived AR in adaptive immunity. During an immune response, initial immune attack mechanisms that destroy the pathogen are superseded at later times by suppression that reduces immunopathology, and tissue repair that restores normal structure and function. T lymphocytes are major cellular contributors to all three phases, and are thought to play a role in repair by producing HB-EGF and bFGF [33] . AR and HB-EGF, as members of the EGF family, promote the proliferation of fibroblasts, epithelial cells, and smooth muscle cells, which are major cell types repaired or remodeled at local tissue sites during an inflammatory response. Tissues with chronic inflammation show extensive cell proliferation, tissue thickening and reduced elasticity. Regulation of the balance between attack and repair cytokines produced by T cells is thus crucial to the successful outcome of the response. . TCR and cAMP synergize to induce AR production in human CD4 T cells. Purified CD4 T cells were incubated with or without TCR stimulation (anti-CD3/CD28 beads) and the cAMP agonist. (A) AR and HB-EGF mRNA expression was measured by RT-PCR. (B) The concentrations of AR in the supernatant and cell lysates were measured by ELISA. (C) Enriched CD45RA+CD45RO-(naïve) and CD45RA-CD45RO+ (memory) CD4 T cells were treated with medium alone, or anti-CD3/CD28 beads in the presence or absence of cAMP agonist (1 , 1000 mM) . The concentration of AR in the supernatant at 24 hours was measured by ELISA. (D) Purified CD4 T cells were treated with medium alone, or anti-CD3/CD28 beads in the presence or absence of the cAMP-modifying agents shown. RNA was extracted at 4 hours, and AR mRNA was measured by RT-PCR. The concentration of AR in the 24-hour supernatant was measured by ELISA. (E) PBMC were treated with anti-CD3+ anti-CD28 antibodies in the presence or absence of cAMP agonist or antagonist for 8 hours. CD4 T cells were purified by cell sorting and RNA was extracted. The mRNA levels of AR and other cytokines were measured by RT-PCR. All results are representative of at least three experiments. doi:10.1371/journal.pone.0039072.g007 In this model, AR derived from human T cells would be expressed mainly in response to tissue injury, consistent with the importance of the local environmental signals for AR regulation. This contrasts with the requirement for specific effector mechanisms to combat different pathogens, in which pre-commitment to cytokine effector phenotypes (thus linking antigen and effector specificities) may be more effective for regulating clearance functions. Collectively, the coordinate inhibition of pro-inflammatory cytokines and induction of tissue-remodeling cytokines of the EGF family may represent a switch from pathogen clearance to tissue repair mechanisms by effector human T cells. Figure S1 Mouse Th2 but not Th1 cells express AR in response to TCR activation. In vitro induced allogeneic Th1 and Th2 cell lines [50] from B6PL or AR 2/2 mice were stimulated with plate-coated anti-CD3 (2 mg/mL) + anti-CD28 (1 mg/mL) antibodies for 6 hours. Expression of AR, IFNc and IL-4 in CD4 T cells was analyzed by ICS. Biotinylated goat antimouse AR antibodies were obtained from R&D Systems. LEAF TM purified anti-mouse CD3e (145-2C11) and LEAF TM purified antimouse CD28 (37.51) were purchased from BioLegend. APC-Cy7 conjugated anti-mouse CD3 (17A2), Alexa Fluor 700 conjugated anti-mouse CD4 (GK1.5), Pacific Blue conjugated anti-mouse CD44 (IM7), PerCP-Cy5.5 conjugated anti-mouse CD69 (H1.2F3), APC conjugated anti-mouse IL-2 (JES6-5H4), PE-Cy7 conjugated anti-mouse IL-4 (BVD6-24G2), PE conjugated antimouse IL-5 (TRFK5), and FITC-conjugated streptavidin were obtained from eBioscience. PE-Alexa Fluor 610 conjugated antimouse IFNc (XMG1.2) was obtained from Invitrogen. Similar results were obtained in at least three experiments. (TIF)

Allo-SCT for multiple myeloma: a review of outcomes at a single transplant center

Allogeneic stem cell transplant for multiple myeloma (MM) is one treatment associated with long-term disease-free survival. The high incidence of treatment-related mortality and relapses, however, are important reasons for controversy about the role of allografting in the management of MM. We reviewed our results of allografting for MM spanning a period of 34 years in order to better define long-term outcomes and identify areas of progress as well as areas requiring improvement. A total of 278 patients received allogeneic marrow or PBSCs after high-dose myeloablative (N=144) or reduced intensity, non-myeloablative (N=134) regimens. In multivariable analysis, adjusting for differences in patient groups, reduced intensity/non-myeloablative transplants were associated with significantly less acute GVHD, lower transplant mortality, better PFS and overall survival. There were no significant differences in relapse, progression or chronic GVHD, when adjusted. In multivariable analysis of patients receiving only non-myeloablative transplants, decreased overall survival and PFS were associated with relapse after a prior autograft and a β2 microglobulin >4.0. Transplant mortality was reduced and only influenced by a prior tandem autograft.

The survival of patients with multiple myeloma (MM) has improved over the last decade as a result of melphalan-based high-dose therapy followed by auto-SCT, the introduction of novel anti-myeloma agents with increased efficacy in relapsed and refractory MM, and improvements in supportive care. 1 --6 Registry data indicate an improvement in median survival from 3 to 5 years, primarily among younger patients, as a result of these treatment innovations. 7 Despite these new developments, MM remains an incurable disease for the large majority, as all but a few patients will relapse. Allogeneic hematopoietic cell transplantation is currently one treatment with a potential for long-term disease control although its curative potential is debated. This is in part due to the graft-vs-myeloma effect, mediated by immune competent donor lymphocytes, best illustrated by the induction of sustained (molecular) remissions following donor lymphocyte infusions, 8 but could also be due in part to absence of contaminating myeloma cells in the donor graft and documented lower levels of residual disease. 9, 10 The role of allo-SCT in MM, however, is controversial due to the high mortality and morbidity associated with conventional myeloablative regimens and because convincing evidence for a survival benefit is lacking. 11 --13 In the last decade, non-myeloablative allo-SCT has gained in popularity due to significantly reduced TRM. 14, 15 Among four reports comparing auto-SCT with allo-SCT, two have shown survival advantages for the nonmyeloablative approach when compared with tandem autologous transplantation. 16 --19 A recently reported US clinical trial prospectively comparing tandem autologous transplant to autologous followed by non-myeloablative allo-SCTs found no differences in PFS or OS at 3 years. 20 In contrast, a European multicenter trial found than tandem autologous, non-myeloablative allo-SCT resulted in superior OS compared with single or tandem auto-SCT. 19 Furthermore, at least one registry report comparing conventional ablative with non-myeloabalative/reduced intensity allo-SCTs have shown similar survival outcomes with lower TRM for patients receiving non-ablative transplants yet higher rates of relapse and PFS inferior to ablative allo-SCT. 21 We reviewed our results of allo-SCT for patients with MM beginning in 1975 with the aim of identifying factors associated with improvements in disease-free survival and OS as preparative regimens have changed from ablative to non-myeloablative. Beginning in 1975, patients with MM were referred to the University of Washington, Fred Hutchinson Cancer Research Center or the Seattle Veterans Hospital for consideration of allo-SCT. Patients were evaluated for suitability for transplant based on treatment protocols in effect at the time. Patient records, laboratory, X-rays and marrow aspirates were reviewed to confirm the diagnosis of MM. To be considered for marrow transplantation, patients had to meet the established criteria for active, symptomatic MM according to Durie and Salmon 22 and had to have received at least one cycle of conventional dose chemotherapy. Patients with a Karnofsky score of o50, a pulmonary diffusion capacity of o50% of predicted and symptomatic heart failure were excluded. Non-ablative transplant candidates were allowed to enroll with a diffusion capacity as low as 30%. Standard hematologic and chemistry studies were used to evaluate organ function. A suitable marrow donor was required, which included HLA identical relatives, HLA haplo-identical relatives or an unrelated donor who was phenotypically HLA identical, or single allele or Ag HLA-mismatched at class I with the patient. Transplants occurred between January 1975 and September 2008. The date of last follow-up was August 2011. Initially, patients who had achieved a complete response (CR) to firstline therapy and were without any evidence of disease were excluded from transplantation. This policy changed, however, as non-myeloablative allo-SCT regimens were adopted. Ablative allo-SCT were utilized as standalone therapy. In contrast, non-abaltive allo-SCT were performed in the majority of patients, 2 --4 months following recovery from a standard auto-SCT utilizing high-dose melphalan. The auto-SCT was utilized to provide cyto-reduction before the non-ablative Allo-SCT, yet allow the patient time to recover from the effects of high-dose therapy used for auto-SCT. Maintenance therapies were not used following allo-SCT. For purposes of this analysis, patients with at least a 50% reduction in monoclonal proteins in the blood or a 75% reduction in 24 h quantitative Bence Jones protein, to their most recent chemotherapy before allo-SCT or auto-SCT, in the case of tandem transplants, were categorized as having sensitive disease, whereas all other patients were judged to have chemotherapy-resistant disease. Responses were categorized according to the IMWG criteria. 23 If certain data were missing that were required for response categorization, for example immunofixation for CR, the patient was classified as responding in the next lower category. An analysis of OS, PFS, TRM, relapse or progression, acute and chronic GVHD was undertaken. The initial analysis compared outcomes using non-myeloablative conditioning for the allogeneic transplant vs those with myeloablative conditioning. In the analysis of relapse or progression, time-dependent competing risks of treatment failure such as death from TRM were included. Cox regression models for these outcomes were adjusted for patient age (continuous), donor sex, chemotherapy responsive vs resistant disease, related vs unrelated donor, time from diagnosis to transplant (o2.5 years vs 42.5 years), prior radiation, prior number of chemotherapy regimens (continuous), b-2 microglobulin 44.0 either at diagnosis or transplant, and abnormal cytogenetics or FISH either at diagnosis or transplant. Abnormal cytogenetics included multiple abnormalities or any abnormality by conventional cytogenetics other than hyperdiploidy. Abnormal FISH included deletion 13, deletion 17, translocation 4;14, 14;16, or 14;20. Because data were missing for some patients, data available for abnormalities were compared with patients who had no abnormalities and patients with missing data. Subsequent multivariate analyses of risk factors for the same outcomes among patients receiving non-myeloablative conditioning included the factors noted above, plus single allo-SCT vs tandem autologous-allo SCT, and progression after a prior autologous SCT used as stand-alone treatment. Patient characteristics are shown in Table 1 . Patients receiving non-myeloablative allo-SCTs were older by a median of 8 years. There were no important differences in the percentages of patients with advanced Durie Salmon staging, IgG or IgA subtypes, number of prior regimens, or total cycles of chemotherapy. Availability of data on beta-2 microglobulin levels, albumin and cytogenetic data were limited. A higher percentage of patients receiving ablative regimens had been given local radiation therapy, 50% compared with patients receiving non-ablative regimens, 33%. One third of the patients receiving non-ablative conditioning had progressed after an autologous transplant, while only four patients receiving ablative conditioning had progressed after an autologous transplant. A higher percentage of patients receiving ablative regimens were judged to have refractory disease, 77%, (based on less than a partial response to their last salvage chemotherapy), compared with 52% of patients who received non-ablative regimens. Relatively few patients were in remission before allografting; two patients undergoing myeloa- blative allografts were in 2 nd CR, whereas among the nonmyeloablative group, three were in first CR and four in 2 nd CR. The regimens used for transplant differed significantly by the time periods during which patients were transplanted with almost all ablative allo-SCTs occurring between 1975 and 2000, whereas the non-ablative approach was utilized from 1998 to 2008. (Table 2 ) The conditioning regimens given to ablative allo-SCT recipients consisted mostly of fractionated TBI 9 --12 Gy, plus CY, and/or BU. BU and CY without TBI were utilized for 69 patients. The non-ablative regimens were primarily TBI 2 Gy with or without fludarabine, whereas 14 patients received additional melphalan 100 mg/m 2 . Most donors were HLA-matched siblings (n ¼ 198) for both ablative and non-ablative transplants, however, a greater percentage of non-ablative transplants were performed from unrelated donors. Marrow was the primary stem cell source for most of the patients receiving ablative conditioning whereas PBSCs were used almost exclusively for non-ablative recipients. The majority of regimens for GVHD prophylaxis in ablative recipients consisted of a calcineurin inhibitor with MTX or steroids. Almost all recipients of non-ablative regimens received a calcineurin inhibitor and mycophenolic acid for GVHD prophylaxis. Response to transplant Among the 144 ablative transplant recipients, 33 (23%) achieved CRs, 33 (23%) a partial response, 12 (8%) did not respond and 67 (46%) were not evaluable owing to early death. Of 134 patients receiving non-ablative transplants, 51 (38%) achieved a CR, 48 (36%) a partial response, 31 (23%) did not respond, whereas 4 (3%) were not evaluable owing to early death. Patients who achieved a CR (n ¼ 84) had 5 and 10 year survivals of 62 and 53% compared with patients who did not achieve a CR (n ¼ 132) and excluding patients who were not evaluable owing to early death, who had 5 and 10 year survivals of 28% and 17%, respectively. Among patients who received ablative conditioning, 104 developed acute GVHD; 7 grade 1, 44 grade 2, 34 grade 3 and 19 grade 4. Of patients who received non-ablative conditioning acute GVHD occurred in 90; grade 1 in 6, grade 2 in 72, grade 3 in 8 and grade 4 in 4. The cumulative incidences of chronic extensive GVHD were 27% and 66% for patients receiving ablative and non-ablative conditioning regimens, respectively. Causes of death varied significantly between patients receiving ablative and non-ablative transplants. (Table 3 ) Among patients who died after receiving ablative conditioning, major causes included fungal infections (n ¼ 20), respiratory failure from diffuse alveolar damage or acute respiratory distress syndrome (n ¼ 8), acute GVHD (n ¼ 18), multi-organ failure (n ¼ 16), viral infections (n ¼ 13) and progressive disease (n ¼ 39). In contrast, only three patients receiving non-ablative transplants died of any of these causes. The major causes of death among recipients of nonablative transplants were mostly chronic GVHD (n ¼ 12) and progressive disease (n ¼ 50). At the time of last follow-up, August 2011, among 144 patients receiving ablative conditioning, 14 were alive a median of 15.1 years (3.6 --23.5) post transplant, of whom 6 had relapsed. Among 134 patients receiving non-ablative conditioning, 56 were alive a median of 7.1 years (2.9 --12.9) post transplant, of whom 25 had relapsed. At 2 years, the probabilities of non-relapse mortality were 18% and 55% for non-ablative and ablative regimens, respectively. At 6 years, the probabilities of relapse or disease progression were 55% and 34% for non-ablative and ablative regimens, respectively. For patients undergoing ablative allo-SCTs, the probabilities of OS and PFS are 11 and 8% at 15 years. For patients undergoing non-ablative transplants, the probabilities of OS and PFS are 39 and 16% at 10 years. (Figure 1 ) The best outcomes were found among 88 patients who received an autologous transplant, followed by a non-myeloablative allograft within 4 months of the autologous transplant and who had not progressed after a prior autologous transplant. (Figure 2 ) Their 10-year OS was 49% and PFS was 27%. Cox regression analysis of overall mortality, PFS, TRM, relapse or progression and acute or chronic GVHD between non-myeloablative and ablative conditioning regimens are shown in Table 4 . When adjusted for patient and donor factors, non-myeloablative conditioning resulted in significantly lower overall mortality HR 0.40 (0.3 --0.6), improved PFS HR 0.55 (0.4 --0.8) and much lower TRM HR 0.22 (0.1 --0.4). The risks of acute GVHD grades 2 --4 were also significantly lower with non-myeloablative regimens HR 0.41 (0.3 --0.6). The risks of relapse or progression and chronic GVHD when adjusted for competing risks of death and patient and donor factors, were not significantly different between ablative and non-ablative conditioning, despite the almost exclusive use of PBSC for the non ablative recipients. In a separate multivariable analysis, outcomes of only patients undergoing non-ablative allogeneic transplants were considered. (Table 5 ) The most important predictors of survival, PFS and In order to discern any association between chronic GVHD and disease progression, we examined this association and its effects on PFS, in a time-dependent fashion among recipients of nonablative transplants. We found only a weak association between patients with clinical extensive chronic GVHD and reduced rates of progression or relapse HR ¼ 0.74 (0.4 --1.3), P ¼ 0.32. This resulted in no net benefit on PFS HR ¼ 0.89 (0.5 --1.5), P ¼ 0.65. In this retrospective review of allo-SCT for MM going back 34 years, significant improvements were observed in the TRM associated with the introduction of non-myeloablative conditioning. Mortality censored for relapse was 55% among the 144 patients receiving ablative transplants compared with only 18% in the non-myeloablative group. As a result, the survival at 10 years from transplant was significantly superior for non-ablative transplants, 35% compared with 15%. As these two groups were not prospectively studied and were not treated contemporaneously, it is likely that other factors including better antiinfectious prophylaxis and treatment, and the use of PBSCs may Bold numerals refer to number of patients for each heading. Allo-SCT for multiple myeloma W Bensinger et al have contributed in part to these improvements. Indeed, there were almost no deaths due to viral or fungal pathogens among non-myeloablative recipients; a major cause of mortality among ablative transplant recipients. In addition, there were major differences between the groups in patient age, relapse after prior autologous transplant, and proportion of patients resistant to their last chemotherapy regimen just before transplant. Although the perception is that patients with MM tolerate allografting more poorly than patients with other hematologic malignancies, a recent analysis from the European Group for Blood and Marrow Transplantation suggested that when adjusted for risk factors including age, disease stage, interval from diagnosis to transplant and donor factors, outcomes for patients with MM were similar. 24 Additionally, there are now newer drugs available to treat relapse that were not available previously which would certainly affect survival after relapsed disease. In univariate analysis, non-myeloablative transplants were associated with an apparent greater risk of disease progression or relapse, 55% at 6 years for non-myeloablative compared with 34% for ablative conditioning. When adjusted for competing risks of death due to higher TRM associated with ablative transplants, however, these differences were not statistically significant. Although this result does not appear to agree with the analysis of others such as the EBMT registry data, the study was only a univariate analysis and did not account for competing causes of death, as ours did. 25 Nevertheless, the amount of residual disease present at transplant, provides a greater challenge for clearance by the allogeneic donor graft when a non-myeloablative regimen is utilized and is still the primary cause of treatment failure. When comparing the incidences of chronic GVHD, 27% of the ablative recipients developed CGVHD compared with 66% for non-ablative recipients. As the risk of CGVHD is time-dependent, and more nonmyeloablative patients survived the early phases of transplant, this did not prove to be significantly higher when adjusted for competing causes of death. In an attempt to overcome this limitation, many groups have employed a tandem autologous, non-myeloablative allogeneic transplant with the aim of providing major cytoreduction, but an opportunity for the patient to recover from high-dose chemotherapy before the Allo-SCT. 14, 16, 26 In multivariable analysis, patients receiving a tandem autologous, non-myeloablative allogeneic transplant had reduced non-relapse mortality, but did not independently affect other outcomes. This analysis also indicated that relapse after a prior autologous transplant is associated with inferior survival as well as other outcome measures. As seen in prior studies, a b-2 microglobulin 44 was also independently associated with increased risk of progression or relapse as well as inferior survival. Female donors were associated with a significantly reduced risk of relapse or progression, consistent with other analyses that have shown more of a graft-vs disease effect from female to male transplants. These analyses agree with other studies showing prior autograft failure to be one of the major risk factors for disease progression after non-myeloablative allo-SCT. 27 The observation that prior autograft failures do poorly with an allo-SCT argues against the recommendation some have made to delay an allo-SCT until disease progression after initial treatment or autologous transplant. 7 In some retrospective analyses, a non-myeloablative allograft was able to overcome certain high-risk FISH characteristics such as the 4;14 translocation. 28 Our patient population contained too few patients with 4;14 to analyze this separately, however, in the multivariable analysis only high B2 microglobulin and not adverse cytogenetics were associated with inferior outcomes. This does not mean that cytogenetics are not important but merely reflect a limited number of observations in our database to directly address that question. It is clear that reduced intensity allo-SCT regimens can result in reliable donor engraftment with a relatively low mortality compared with high-dose regimens. The immunologic effect of the allograft is, however, relatively modest requiring a prior autologous transplant for cytoreduction. Even with the tandem auto-non-myeloablative allo-SCT approach, relapses beyond 3--5 year continue to occur, making disease recurrence the primary cause of treatment failure after tandem auto, non-myeloablative allo-SCT. Future studies of allo-SCT in MM should focus on regimens that are less toxic but able to preserve anti-tumor effects such as radioisotopes linked to antibodies that target myeloma cells or other marrow-based cells. It should be relatively easy to combine targeted radiotherapy with a non-myeloablative regimen to create a more tolerable cytoreductive protocol. It is also worth reconsidering more myeloablative regimens, as supportive care has improved greatly in the past 20 years. As previously noted, when younger patients are transplanted earlier from initial diagnosis, TRM is reduced. Another strategy to make non-myeloablative regimens more effective would be to combine the donor graft with infusions of allogeneic donor lymphocytes or subsets of lymphocytes in the form of 'engineered grafts', for example CD4 lymphocytes, which may have a graft vs myeloma effect without increasing GVHD. 29 It may also be possible to exploit killer-Ig-like mismatching between donor and recipient, which has been shown to result in improved PFS due to a reduced rate of relapse. 30, 31 Maintenance strategies, which have been shown to delay disease progression after auto-SCT may also be effective after allo-SCT. 32, 33 Finally, it may be worthwhile to exploit monoclonal antibodies targeting myeloma cells such as the CD40 Ag or CS-1 Ag, in order to increase the ability of donor allogeneic cells to eliminate residual host disease. 34 In any case, due to the substantial morbidity and mortality associated with allografting as well as the uncertain benefits, future approaches to allografting for myeloma should only be performed within well-designed clinical trials. The authors declare no conflict of interest.

UNC93B1 Mediates Innate Inflammation and Antiviral Defense in the Liver during Acute Murine Cytomegalovirus Infection

Antiviral defense in the liver during acute infection with the hepatotropic virus murine cytomegalovirus (MCMV) involves complex cytokine and cellular interactions. However, the mechanism of viral sensing in the liver that promotes these cytokine and cellular responses has remained unclear. Studies here were undertaken to investigate the role of nucleic acid-sensing Toll-like receptors (TLRs) in initiating antiviral immunity in the liver during infection with MCMV. We examined the host response of UNC93B1 mutant mice, which do not signal properly through TLR3, TLR7 and TLR9, to acute MCMV infection to determine whether liver antiviral defense depends on signaling through these molecules. Infection of UNC93B1 mutant mice revealed reduced production of systemic and liver proinflammatory cytokines including IFN-α, IFN-γ, IL-12 and TNF-α when compared to wild-type. UNC93B1 deficiency also contributed to a transient hepatitis later in acute infection, evidenced by augmented liver pathology and elevated systemic alanine aminotransferase levels. Moreover, viral clearance was impaired in UNC93B1 mutant mice, despite intact virus-specific CD8+ T cell responses in the liver. Altogether, these results suggest a combined role for nucleic acid-sensing TLRs in promoting early liver antiviral defense during MCMV infection.

Initiation of inflammation following infection requires recognition of the invading microbe by innate immune pattern recognition receptors (PRRs) that signal in response to pathogen-associated molecular patterns (PAMPs). PRRs recognize selfand microbe-associated molecules [1] [2] [3] [4] . Members of the Toll-like receptor (TLR) family of PRRs are transmembrane receptors that are expressed either on the cell surface or within the endosomal compartment and respond to a variety of PAMPs [1] . Murine TLR3, TLR7 and TLR9 are expressed in the endolysosome and are implicated in recognition of viral dsRNA, ssRNA and dsDNA, respectively [1, [5] [6] [7] [8] [9] [10] [11] . Ligation of the nucleic acid-sensing TLRs results in transcription of antiviral genes including type I IFNs (IFN-a/b) and proinflammatory cytokines [1] . TLR3 responses require signaling through the adaptor molecule Toll/IL-1R domain-containing adapter-inducing interferon-b (TRIF), while TLR7 and TLR9 are dependent on the adaptor molecule myeloid differentiation primary response gene 88 (MyD88) to activate transcription factors and induce gene transcription [1, [12] [13] [14] . Murine cytomegalovirus (MCMV) is a betaherpesvirus that can establish acute infection in multiple organs including the liver. Acute MCMV infection induces an early systemic proinflammatory cytokine response including high levels of type I IFNs, IFN-c, IL-12 and TNF-a [15] [16] [17] [18] [19] . Infection in the liver induces early production of IFN-a, predominantly by plasmacytoid dendritic cells (pDCs), by 40 h post-infection [20, 21] . Type I IFN production mediates downstream responses including chemokine and cytokine production as well as monocyte/macrophage, natural killer (NK) cell and T cell recruitment [20] [21] [22] [23] . Early type I IFN signaling is necessary for NK cell recruitment to the liver, where they deliver the antiviral cytokine IFN-c within the first 48 h post-MCMV infection [23] . The NK cell IFN-c response is an important early step in the control of liver infection [24, 25] . This response induces IFN-c-dependent chemokines, which contribute to the recruitment of CD8+ T cells to the liver [26] . Liver CD8+ T cell responses occur by days 5 and 7 post-MCMV infection and are an important source of cytokines late in acute infection that contribute to resistance against MCMV [26] [27] [28] [29] [30] . While early responses to MCMV infection in the liver are well understood, it remains unclear how the virus is sensed in this compartment. This is in contrast with other sites, namely the spleen, in which studies by our group and others have definitively shown a role for TLR9 and MyD88 signaling in IFN-a, proinflammatory cytokine and cellular responses in addition to restriction of virus replication [6, 20, 31, 32] . Although TLR7 alone does not appear to have a strong role in MCMV recognition, TLR7 and TLR9 combined deficiency was shown to severely impair pDC responses against MCMV in the spleen [33] . A significant but minor role for TLR3 signaling in the spleen has also been suggested in response to MCMV infection [6] . In the liver, however, studies by our group have demonstrated that early innate responses are TLR9-independent but MyD88-dependent [20] . Liver pDCs from mice genetically deficient in TLR9 produce wild-type (WT) levels of IFN-a at 40 h post-MCMV infection, with intact downstream cellular and proinflammatory cytokine responses. Further, TLR9-deficient mice do not exhibit elevated liver viral titers. Conversely, MyD88-deficient mice have severely impaired liver cytokine and cellular responses, and are unable to control virus replication in this compartment [20, 32] . MyD88 is a common adaptor molecule for TLR9 and TLR7 signaling; however, evaluation of TLR7-deficient mice also demonstrated that TLR7 signals alone were not required to initiate liver antiviral defense [20] . These TLR-independent but MyD88-dependent antiviral responses suggested possible redundancies among TLR signals in the liver compartment in response to MCMV infection [20, 32] . To investigate this possibility, we utilized mice containing an H412R missense mutation in the endoplasmic reticulum protein UNC93B1 to address the combined function of nucleic acidsensing TLRs in the liver during acute MCMV infection. The UNC93B1 mutation (known as 'triple d' or '3d') impairs signaling through TLR3, TLR7 and TLR9 due to improper trafficking of these receptors to the endosomal compartment, and has been shown to affect exogenous antigen presentation [34, 35] . Our studies show that proinflammatory cytokine production after early infection with MCMV is dependent on UNC93B1. Further, UNC93B1 deficiency exacerbates liver disease and increases viral burden, although MCMV-specific CD8+ T cell responses are not impaired. Collectively, these results suggest a level of redundancy within the liver to promote viral recognition by demonstrating that a combination of nucleic acid-sensing TLRs contributes to innate inflammatory responses during MCMV infection. Considering the potential of endosomal TLR signals to induce proinflammatory cytokine expression, UNC93B1 deficient 3d mice were first assessed for systemic IFN-a, IFN-c, IL-12 and TNF-a production during early infection with a moderate (5610 4 PFU) dose of MCMV. C57BL/6 (WT) and 3d mice were uninfected or MCMV-infected for 40 h or 48 h. Serum was collected at indicated time points and IFN-a, IFN-c, IL-12p70 and TNF-a were measured by enzyme-linked immunosorbent assay (ELISA). In WT mice, maximal production of IFN-a, IFN-c and IL-12p70 was detected at 40 h post-MCMV infection before declining by 48 h post-infection (Fig. 1 , A-C). In contrast, 3d mice exhibited lower serum levels of these cytokines in response to MCMV infection. Specifically, while serum IFN-a reached 13006420 pg/mL at 40 h post-MCMV infection in WT mice, IFN-a production was reduced by three-fold at this infection time point in 3d mice (4506300 pg/mL), with comparable levels maintained at 48 h post-infection (Fig. 1A) . Likewise, while average IFN-c concentrations in WT mice reached maximal levels of 5306245 pg/mL at 40 h post-MCMV infection, 3d mice failed to induce detectable levels of this cytokine (Fig. 1B) . IL-12p70 production similarly peaked at 40 h post-MCMV infection in WT mice, with levels reaching 10006600 pg/mL. 3d mice, however, produced 12-fold less IL-12p70 at this infection time point (Fig. 1C) . Serum TNF-a levels were elevated in response to MCMV infection at both 40 h and 48 h post-infection in WT mice (69610 and 63617 pg/mL, respectively, Fig. 1D ). In 3d mice, average concentrations of TNF-a were 6-fold lower than WT at 40 h and 2.5-fold lower than WT at 48 h after infection. These results indicate a requirement for endosomal TLR signals for early systemic proinflammatory cytokine production, and concur with previous reports [35] . Having observed a reduction in the level of proinflammatory cytokines in the serum of 3d mice, cytokine responses in liver cells from 3d mice infected with MCMV were evaluated to address the impact of endosomal TLR signaling in a localized tissue site of infection. The best characterized liver cytokine responses are IFNa and IFN-c [20, 21, 36] . Therefore, to determine whether combined endosomal TLR signaling contributes to the production of these cytokines during MCMV infection, IFN-a and IFN-c in liver homogenates and in individual cell populations were measured in WT and 3d mice uninfected or infected with MCMV for 40 h and 48 h. As shown in Fig. 2A , WT mice displayed fourfold higher levels of IFN-a than 3d mice at 40 h post-infection, with increased levels still evident at 48 h following infection. Since pDCs expressing the marker PDCA-1 have been shown to produce the majority of liver IFN-a at 40 h post-MCMV infection [20] , this cell type was examined in WT and 3d livers during early infection. There was evidence of pDC accumulation in the livers of both WT and 3d mice (Fig. 2B ). However, liver pDCs from 3d mice were impaired in their ability to express IFN-a. There were 4-fold fewer PDCA-1+ pDCs expressing intracellular IFN-a at 40 h, and 3-fold fewer at 48 h post-infection, in 3d mice as compared to WT (Fig. 2C ). This trend was also reflected in the proportion of PDCA-1+ IFN-a+ pDCs at 40 h and 48 h after infection in 3d mice (0.8%60.4% and 2%61%) compared with WT (2%61% and 6%61%). 3d mice similarly demonstrated a defect in liver IFN-c production in response to MCMV infection. In WT mice, IFN-c reached maximal levels of 350061200 pg/g liver at 40 h before contracting by approximately half at 48 h post-infection. In contrast, at 40 h post-infection, 3d mice induced 5-fold less IFN-c than WT (Fig. 2D ). NK cells are an important early source of IFNc in the liver during MCMV infection [24, 25] , and accumulated at this site in both WT and 3d mice (Fig. 2E ). Using intracellular cytokine staining, results shown in Fig. 2F demonstrate a 7-fold reduction in the absolute numbers of NK1.1+ TCRb-liver cells expressing IFN-c in 3d mice at 40 h post-infection as compared to WT. There were also fewer IFN-c-expressing liver NK cells in 3d mice by proportion (0.8%60.3% and 1%60.4%) when compared to WT (6%62% and 4%60.4%) at 40 h and 48 h, respectively. Together, these results demonstrate that, in addition to an effect on systemic cytokine production, combined endosomal TLR signals can affect the expression of critical proinflammatory cytokines in the liver during MCMV infection. The innate immune response is important both in establishing early control of virus replication and in coordinating downstream adaptive responses. Following MCMV infection, virus-specific CD8+ T cells are recruited to the liver within 5 days and control viral replication at this site through release of cytotoxic molecules and production of cytokines such as IFN-c and TNF-a [26, 27] . Given the abated liver cytokine responses observed in 3d mice, the effect of endosomal TLR signaling on liver CD8+ T cell responses was examined at late time points during acute MCMV infection. The results shown in Fig. 3A demonstrate comparable absolute numbers of CD8+ T cells in WT and 3d mice at days 5 and 7 post-MCMV infection, a trend that was also reflected in proportion (data not shown). To determine whether CD8+ T cells in 3d mice were properly activated against MCMV infection, intracellular expression of IFN-c and TNF-a in CD8+ T cells was examined following ex vivo restimulation with H-2D b M45 viral peptide, an immunodominant epitope of MCMV [29] . CD8+ T cells from 3d mice expressed these two cytokines at day 5 and day 7 post-MCMV infection by proportion and absolute numbers at levels that were comparable or slightly increased over WT (Fig. 3 , B-E). As an indication of degranulation, surface expression of CD107a was also examined on liver CD8+ T cells from mice infected with MCMV; however, no differences in CD107a expression were detected between 3d and WT mice (data not shown). These results suggest that MCMV-specific CD8+ T cell responses in the liver are not compromised in the absence of endosomal TLR signals. Previous studies have demonstrated resolution of virus-induced liver disease after 5 days of MCMV infection in WT mice [22, 24, 27, 37] . Therefore, given the impaired inflammatory responses observed in the absence of endosomal TLR signaling, liver sections prepared from 3d and WT mice that were uninfected or infected for 3, 5, or 7 days were hematoxylin and eosin (H&E) stained to evaluate pathology. The histological appearance of liver sections from uninfected 3d and WT mice appeared comparable (Fig. 4, A and B ). By day 3 post-MCMV infection, clusters of infiltrating cells or inflammatory foci, which have been shown to coincide with sites of MCMV antigen expression [23, 37] , were present in WT mice and persisted through day 5 before inflammation was resolved by day 7 post-infection (Table 1 , Fig. 4 , C, E and G). While the inflammatory foci per area of liver were equally apparent in liver sections from 3d mice infected for 3 days (Table 1) , there was an increased presence of cytomegalic inclusion bodies characteristic of MCMV-infected cells that were not readily apparent in WT mice (Fig. 4D) . By day 5 postinfection, livers from 3d mice were characterized by widespread areas of inflammation compared to the more punctate foci in WT livers (Fig. 4, E and F) . Moreover, the inflammatory foci per area of liver in 3d mice at day 5 post-infection were significantly more numerous and contained a greater number of nucleated cells compared to WT (Table 1) . However, by day 7 post-infection, inflammation in 3d mice showed signs of resolution that were similar to WT (Table 1, Fig. 4 , G and H). To further evaluate the effects of endosomal TLR responses on overall liver function, the liver enzyme alanine aminotransferase (ALT) was measured in serum samples from WT and 3d mice that were uninfected or infected with MCMV for 3, 5 or 7 days. Uninfected mice had comparable baseline levels of systemic ALT (Fig. 5 ). By day 3 of infection, similar elevations in ALT levels were detected in both groups. In contrast, by day 5 post-infection, the levels of systemic ALT in 3d mice reached values of 9006400 U/ L and were three-fold higher than the ALT levels measured in WT mice (2706100 U/L; Fig. 5 ), indicating augmented liver disease. In contrast, by day 7 post-infection, 3d mice demonstrated a sharp decline in systemic ALT levels to values of 117663 U/L, which were comparable to ALT levels exhibited in WT mice (108685 U/L). Thus, at this time point of acute infection, both groups of mice exhibited ALT levels near baseline levels, consistent with the resolution of virus-induced liver pathology observed in 3d and WT mice (Fig. 4, G and H; Fig. 5 ). Taken together, these results suggest that in the absence of endosomal TLR signals, there is pronounced liver inflammation accompanied by a transient increase in liver damage during acute MCMV infection. Given the diminished early cytokine responses and enhanced liver disease observed in 3d mice, the contribution of endosomal TLR responses to control of virus replication in the liver was assessed in WT and 3d mice infected with MCMV. Compared with those in WT mice, viral titers were elevated by 1 log on day 3 of infection in 3d mice, and remained significantly higher at day 5 post-infection when compared to WT (Fig. 6) . Ultimately, while WT livers showed evidence of viral clearance at days 7 and 9 postinfection, 3d liver virus titers remained approximately 2 logs higher at these times of infection as compared to WT mice. Thus, endosomal TLR signaling contributes to the control of MCMV replication in the liver during acute infection. The aim of these studies was to identify the TLR signaling pathways required for the innate recognition of virus infection in the liver, a common target organ of many viruses that significantly contributes to innate immune defenses [38] [39] [40] . Moreover, the liver contains various innate immune cells that express TLRs [41, 42] ; however, the role of TLRs in host defense against infection at this site remains largely unclear. Because responses in the liver do not appear dependent on individual TLRs [6, 20, 32] , we utilized 3d mice, which lack endosomal TLR3, TLR7 and TLR9 signaling due to a mutation in the endoplasmic reticulumresident protein UNC93B1 [34, 35] , to address the contribution of endosomal TLRs in liver antiviral defenses against acute MCMV infection. The results demonstrated impaired production of proinflammatory cytokines by NK cells and pDCs in livers from MCMV-infected 3d mice. Additionally, 3d mice had elevated viral titers in the liver that coincided with transient but exacerbated liver disease, although virus-specific CD8+ T cell responses were not affected. Interestingly, a previous study demonstrated that TLR3 was not required for the generation of adaptive antiviral responses to MCMV [43] , although there is evidence that TLR3 signaling contributes in part to the early control of MCMV infection by the systemic induction of type I IFN [6] . Other studies have implicated a synergistic role for TLR7 and TLR9 in promoting MCMV recognition and immune defense in the spleen [33] . The impaired liver responses observed in 3d mice that were not apparent in TLR9 or TLR7-deficient mice [20] suggests that a level of redundancy unique to innate immunity is in place within infected tissue sites to rapidly respond to viral infection. Overall, these studies advance our understanding of the process of viral recognition in the complex liver environment and suggest that UNC93B1 is a critical intermediate factor in innate virus sensing activated by MCMV. Previous studies have demonstrated impaired serum cytokine production and increased susceptibility to infection in 3d mice infected with a high lethal dose of MCMV [35] ; however, our study is the first report to document the contribution of the 3d mutation to impaired MCMV defense in the liver. In agreement with the study by Tabeta et al. [35] , our results, using a moderate dose of MCMV, demonstrated a diminished serum cytokine response (Fig. 1) , reduced splenic IFN-a production and elevated viral titers in the spleen (data not shown). These results were not unexpected given the known role for the nucleic acid-sensing TLRs in MCMV recognition and the subsequent production of proinflammatory cytokines and type I IFN by splenic pDCs [19, 31, 33, 44] . Production of type I IFN is a critical early step in antiviral defense, and we further demonstrated diminished levels of liver IFN-a in MCMV-infected 3d mice. This defect was in part due to an impaired ability of liver pDCs to express this cytokine and is consistent with previous reports identifying pDCs as an important early source of IFN-a in response to TLR7 and TLR9 ligands [10, 20, 31, 32, 33, [45] [46] [47] . Altogether, these results concur with previous studies indicating that pDCs are the predominant leukocyte producer of type I IFN in the liver during early MCMV infection [20] , and demonstrate that production of these cytokines by pDCs is influenced by the 3d mutation. Despite impairments in liver IFN-a production in 3d mice, it is notable that this response was not totally abrogated. This suggests the presence of alternative pathways to type I IFN production in the liver, and may be the reason that IFNa/bR-deficient mice die by day 5 in response to infection with a moderate dose of MCMV, as reported previously [36] , while 3d mice do not. Studies have also demonstrated production of type I interferon from cells other than pDCs at 44-48 hours following MCMV infection [32, 48, 49] . In addition, the liver contains a variety of parenchymal and nonparenchymal cells that express TLRs and are capable of type I IFN production [38, [40] [41] [42] . There may also be TLR-independent pathways leading to the production of these cytokines in response to MCMV, including cytoplasmic RNA-and DNA-sensing receptors, as have been demonstrated with other virus models but have yet to be examined in the context of MCMV infection [1] [2] [3] [4] [50] [51] [52] [53] [54] [55] . It has been established that NK cell inflammatory responses and production of IFN-c are essential to defense against MCMV in the liver [23] [24] [25] . In 3d mice, NK cell production of IFN-c in the liver was severely impaired during early MCMV infection and likely contributed to increased viral burden and liver pathology. The reduced levels of IFN-c in NK cells may reflect the deficiency of serum IL-12 seen in 3d mice. These results concur with previous reports that type I IFN regulates IL-12 production by conventional DCs and consequently the production of IFN-c by NK cells [19, 48, 56] . The defect in liver cytokine production in 3d mice is reminiscent of that reported for MyD88-deficient mice. Notably, however, MyD88-deficient mice exhibited more severe liver pathology and greater elevations in viral titers when compared to WT than we observed in 3d mice [20] . Interestingly, mice deficient in TLR7 and TLR9 exhibited decreased levels of systemic IFN-a/b and increased susceptibility to MCMV infection Data presented represent the mean 6 SD of two liver sections unless otherwise indicated. b Number of foci per 8650 mm 2 areas in day 5 infected 3d livers was significantly higher than WT at the same infection time point, p#0.04. C Nucleated cells from large areas of inflammation contained .60 nucleated cells in less defined foci, and do not include a calculation of SD. doi:10.1371/journal.pone.0039161.t001 Figure 5 . Assessment of virus-induced liver disease in 3d mice. Serum samples were collected from C57BL/6 (WT) or 3d mice that were either uninfected or infected for 3, 5, or 7 days with MCMV. Circulating levels of ALT were measured as described in Methods. Data are the combined results from three independent experiments and show the mean 6 SD (n = 6-8 mice per group). Asterisks denote p values#0.003. doi:10.1371/journal.pone.0039161.g005 [33] . Taken together, these observations further support the notion that the liver possesses compensatory mechanisms to combat viral infection in the combined absence of endosomal TLR signals. Accordingly, despite the early effects of endosomal TLR deficiency, 3d mice were able to mount robust CD8+ T cell cytokine responses. It should be noted that the 3d defect has previously been shown to impair exogenous antigen presentation, including cross presentation, which has a reported role in priming CD8+ T cell responses during MCMV infection [35] . However, we detected no overt defect in CD8+ T cell responses within the first seven days of MCMV infection in the liver. Further, examination of activation markers suggested that a similar proportion of CD8+ T cells from 3d mice were more highly activated when compared to WT (data not shown). Several studies have demonstrated the contribution of activated virus-specific CD8+ T cells to effective hepatic immunity against MCMV infection [26] [27] [28] . In addition, the normal development of adaptive responses despite impaired innate responses is well documented during MCMV infection. Studies have shown that reduced levels of type I IFN do not affect the accumulation or activation of antigenspecific CD8+ T cells in response to low or moderate MCMV inoculums [48] . Likewise, while IL-12 is critical in inducing NK cell IFN-c expression, T cell responses can occur in an IL-12independent manner [17, 18, [57] [58] [59] . NK cells have the potential to negatively regulate CD8+ T cell responses during MCMV infection [60] [61] [62] ; thus, it is probable that impaired NK cell function in the absence of endosomal TLR signals contributes to inflated T cell responses. The presence of increased virus in the liver may also contribute to the robust T cell recruitment and cytokine production observed at late infection time points in 3d mice. Interestingly, although the severity of viral liver pathology was more pronounced in 3d mice, inflammation and liver injury subsided late in infection. These observations suggest that CD8+ T cells in the liver can respond to limited amounts of type I interferon for activation in the presence of compromised innate responses, emphasizing the prevalence of compensatory mechanisms in place within the liver to deal with infection and promote adaptive immunity. In contrast, IFN-a/bR deficiency negatively impacts innate inflammatory responses and resistance to MCMV infection in the liver [21, 24, 36] . In conclusion, this study indicates that UNC93B1, which is essential for combined endosomal TLR signaling, contributes to development of effective innate immune responses to an acute Figure 6 . Effect of the 3d mutation on viral clearance. Livers were harvested from C57BL/6 (WT) or 3d mice that were either uninfected or infected for 3, 5, 7 or 9 days with MCMV (5610 4 PFU). Viral titers were determined using a standard plaque assay. The level of detection of the plaque assay is 2 log PFU/g liver (dashed line). Each data point represents an individual WT (filled diamonds) or 3d mouse (open circles). Data from days 0, 3, 5, and 7 are the combined results of three independent experiments (n = 4-7 mice per group). In data from day 9, n = 5 mice per group. Asterisks denote a significant difference between WT and 3d mean PFU/g liver (p values#0.04). doi:10.1371/journal.pone.0039161.g006 virus infection in the liver. Our results show that this contribution involves modulation of early innate proinflammatory cytokine production from liver pDCs and NK cells and subsequent control of MCMV replication and pathology before activation of an adaptive immune response. Altogether, these results highlight a process of virus recognition with multiple pathways in place to promote host resistance to infection in the liver microenvironment. Pathogen-free C57BL/6J mice were obtained from the Jackson Laboratory (Bar Harbor, ME). C57BL/6 Unc93b1 3d/3d mice were a kind gift from Dr. Bruce Beutler (The Scripps Research Institute, La Jolla, CA) and were generated as described [35] . C57BL/6J mice were housed and UNC93b1 3d/3d mice were bred in pathogenfree mouse facilities at Brown University. Male and female mice aged 8-10 weeks were used in experiments. 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. All animal work was approved by the Brown University Institutional Animal Care and Use Committee (Protocol Number: 0909082 and 0903035). MCMV Smith strain was used in all experiments. This strain was prepared as a salivary gland-passaged stock from CD1 mice. Moderate dose infection (5610 4 PFU per mouse) was initiated on day 0 by intraperitoneal injection. In vivo responses were measured at indicated time points. For infectious viral titer quantification, organs were weighed, homogenized in cold supplemented DMEM (Invitrogen Life Technologies) and supernatants were collected following centrifugation. Serially diluted samples were used to inoculate confluent monolayers of bone marrow stromal cells (ATCC M2-10B4) in 24-well tissue culture plates and incubated for one hour at 37uC, 5% CO 2 . Following incubation, inoculums were removed and monolayers were overlaid with a 16DMEM/0.5% low-melt agarose solution. Cells were incubated for 7 days at 37uC, 5% CO 2 , then fixed in 10% buffered formalin and stained with crystal violet. Plaques were counted to determine viral titer as previously described [22, 24, 26] . Liver leukocytes were prepared as previously described [23] [24] [25] . Briefly, following mechanical dissociation of tissue, red blood cells were removed by lysis with ammonium chloride and leukocytes separated by Percoll density gradient. To generate homogenates for cytokine analysis, the liver caudate lobe was homogenized in RPMI 1640 (Invitrogen Life Technologies) and supernatants collected following centrifugation. Serum was isolated from whole blood by centrifugation in the presence of heparin and stored at 280uC until further use in cytokine analyses or ALT assays. Liver homogenates and serum were tested for cytokines by standard sandwich ELISA. IFN-c, IL-12p70 and TNF-a were assayed by DuoSet (R&D Systems). IFN-a was measured by VeriKine mouse ELISA kit (R&D Systems). Limits of detection were 15 pg/mL for DuoSets and 12.5 pg/mL for VeriKine ELISA kits. The following fluorochrome-conjugated mAbs were used in flow cytometric analyses: NK1.1-PE and TCRb-APC to distinguish NK cells; CD8a-PECy7 and TCRb-FITC to distinguish CD8+ T cells; and PDCA-1-APC (Miltenyi Biotec) to identify pDCs. Prior to surface staining, cells were incubated with anti-CD16/CD32 mAb to block nonspecific binding of Abs to Fcc III/ II receptors (clone 2.4G2). Unless otherwise noted, antibodies were obtained from BD Biosciences or eBioscience. For intracellular staining of cytokines, cells were treated with Brefeldin A (eBioscience) for 4 hours at 37uC, 5% CO 2 and permeabilized prior to labeling with IFN-a-FITC (R&D Systems), IFN-c-PE, or TNF-a-APC (BD Biosciences). When indicated, leukocytes were stimulated for 5 hours with 100 ng/mL MCMV M45 peptide in addition to Brefeldin A treatment. Samples were acquired using a FACSCalibur and analyzed with BD Cell Quest software. For analysis, viable cells were gated by FSC and SSC. Isotype controls were used to set positive analysis gates. Portions of the median liver lobes were isolated, fixed in 10% neutral buffered formalin, and paraffin embedded. Tissue sections (5 mm) were stained with H&E and analyzed microscopically. Images shown were photographed at the indicated magnification with a DP70 digital camera and software (Optical Analysis Corporation). Inflammatory foci, defined as discrete clusters containing between 6-60 nucleated cells, were quantitated as described previously [23] . In brief, inflammatory foci were identified, at a magnification of 200, as clusters of cells in totals of 8650 mm 2 areas per representative tissue. In some cases, liver sections had larger areas of inflammation with .60 cells with less defined foci, and are indicated as such. Numbers of nucleated cells per inflammatory foci were counted in 20 individual foci per representative tissue at a magnification of 400. Liver alanine aminotransferase (ALT) was measured in serum samples by Marshfield Labs (Marshfield, WI). Student's t test was used to determine statistical significance of experimental results when indicated (p#0.05).

Bioinformatics analysis of rabbit haemorrhagic disease virus genome

BACKGROUND: Rabbit haemorrhagic disease virus (RHDV), as the pathogeny of Rabbit haemorrhagic disease, can cause a highly infectious and often fatal disease only affecting wild and domestic rabbits. Recent researches revealed that it, as one number of the Caliciviridae, has some specialties in its genome, its reproduction and so on. RESULTS: In this report, we firstly analyzed its genome and two open reading frameworks (ORFs) from this aspect of codon usage bias. Our researches indicated that mutation pressure rather than natural is the most important determinant in RHDV with high codon bias, and the codon usage bias is nearly contrary between ORF1 and ORF2, which is maybe one of factors regulating the expression of VP60 (encoding by ORF1) and VP10 (encoding by ORF2). Furthermore, negative selective constraints on the RHDV whole genome implied that VP10 played an important role in RHDV lifecycle. CONCLUSIONS: We conjectured that VP10 might be beneficial for the replication, release or both of virus by inducing infected cell apoptosis initiate by RHDV. According to the results of the principal component analysis for ORF2 of RSCU, we firstly separated 30 RHDV into two genotypes, and the ENC values indicated ORF1 and ORF2 were independent among the evolution of RHDV.

Synonymous codons are not used randomly [1] . The variation of codon usage among ORFs in different organisms is accounted by mutational pressure and translational selection as two main factors [2, 3] . Levels and causes of codon usage bias are available to understand viral evolution and the interplay between viruses and the immune response [4] . Thus, many organisms such as bacteria, yeast, Drosophila, and mammals, have been studied in great detail up on codon usage bias and nucleotide composition [5] . However, same researches in viruses, especially in animal viruses, have been less studied. It has been observed that codon usage bias in human RNA viruses is related to mutational pressure, G +C content, the segmented nature of the genome and the route of transmission of the virus [6] . For some vertebrate DNA viruses, genome-wide mutational pressure is regarded as the main determinant of codon usage rather than natural selection for specific coding triplets [4] . Analysis of the bovine papillomavirus type 1 (BPV1) late genes has revealed a relationship between codon usage and tRNA availability [7] . In the mammalian papillomaviruses, it has been proposed that differences from the average codon usage frequencies in the host genome strongly influence both viral replication and gene expression [8] . Codon usage may play a key role in regulating latent versus productive infection in Epstein-Barr virus [9] . Recently, it was reported that codon usage is an important driving force in the evolution of astroviruses and small DNA viruses [10, 11] . Clearly, studies of synonymous codon usage in viruses can reveal much about the molecular evolution of viruses or individual genes. Such information would be relevant in understanding the regulation of viral gene expression. Up to now, little codon usage analysis has been performed on Rabbit haemorrhagic disease virus (RHDV), which is the pathogen causing Rabbit haemorrhagic disease (RHD), also known as rabbit calicivirus disease (RCD) or viral haemorrhagic disease (VHD), a highly infectious and often fatal disease that affects wild and domestic rabbits. Although the virus infects only rabbits, RHD continues to cause serious problems in different parts of the world. RHDV is a single positive stranded RNA virus without envelope, which contains two open reading frames (ORFs) separately encoding a predicted polyprotein and a minor structural protein named VP10 [12] . After the hydrolysis of self-coding 3C-like cysteinase, the polyprotein was finally hydrolyzed into 8 cleavage products including 7 nonstructural proteins and 1 structural protein named as VP60 [13, 14] . Studies on the phylogenetic relationship of RHDVs showed only one serotype had been isolated, and no genotyping for RHDV was reported. It reported that the VP10 was translated with an efficiency of 20% of the preceding ORF1 [15] . In order to better understand the characteristics of the RHDV genome and to reveal more information about the viral genome, we have analyzed the codon usage and dinucleotide composition. In this report, we sought to address the following issues concerning codon usage in RHDV: (i) the extent and causes of codon bias in RHDV; (ii) A possible genotyping of RHDV; (iii) Codon usage bias as a factor reducing the expression of VP10 and (iiii) the evolution of the ORFs. The 30 available complete RNA sequences of RHDV were obtained from GenBank randomly in January 2011. The serial number (SN), collection dates, isolated areas and GenBank accession numbers are listed in Table 1 . To investigate the characteristics of synonymous codon usage without the influence of amino acid composition, RSCU values of each codon in a ORF of RHDV were calculated according to previous reports (2 Sharp, Tuohy et al. 1986 ) as the followed formula: Where g ij is the observed number of the ith codon for jth amino acid which has n i type of synonymous codons. The codons with RSCU value higher than 1.0 have positive codon usage bias, while codons with value lower than 1.0 has relative negative codon usage bias. As RSCU values of some codons are nearly equal to 1.0, it means that these codons are chosen equally and randomly. The index GC3s means the fraction of the nucleotides G+C at the synonymous third codon position, excluding Met, Trp, and the termination codons. The ENC, as the best estimator of absolute synonymous codon usage bias [16] , was calculated for the quantification of the codon usage bias of each ORF [17] . The predicted values of ENC were calculated as ENC = 2 + s + 29 where s represents the given (G+C) 3 % value. The values of ENC can also be obtained by EMBOSS CHIPS program [18] . Analyses were conducted with the Nei-Gojobori model [19] , involving 30 nucleotide sequences. All positions containing gaps and missing data were eliminated. The values of dn, ds and ω (dn/ds) were calculated in MEGA4.0 [20] . Multivariate statistical analysis can be used to explore the relationships between variables and samples. In this study, correspondence analysis was used to investigate the major trend in codon usage variation among ORFs. In this study, the complete coding region of each ORF was represented as a 59 dimensional vector, and each dimension corresponds to the RSCU value of one sense codon (excluding Met, Trp, and the termination codons) [21] . Correlation analysis was used to identify the relationship between nucleotide composition and synonymous codon usage pattern [22] . This analysis was implemented based on the Spearman's rank correlation analysis way. All statistical processes were carried out by with statistical software SPSS 17.0 for windows. The values of nucleotide contents in complete coding region of all 30 RHDV genomes were analyzed and listed in Table 2 and Table 3 . Evidently, (C+G)% content of the ORF1 fluctuated from 50.889 to 51.557 with a mean value of 51.14557, and (C+G)% content of the ORF2 were ranged from 35.593 to 40.113 with a mean value of 37.6624, which were indicating that nucleotides A and U were the major elements of ORF2 against ORF1. Comparing the values of A 3 %, U 3 %, C 3 % and G 3 %, it is clear that C 3 % was distinctly high and A 3 % was the lowest of all in ORF1 of RHDV, while U 3 % was distinctly high and C 3 % was the lowest of all in ORF2 of Table 2 Identified nucleotide contents in complete coding region (length > 250 bps) in the ORF1 of RHDV (30 isolates) genome Table 4 . Most preferentially used codons in ORF1 were C-ended or G-ended codons except Ala, Pro and Ser, however, A-ended or G-ended codons were preferred as the content of ORF2. In addition, the dn, ds and ω(dN/dS) values of ORF1 were separately 0.014, 0.338 and 0.041, and the values of ORF2 were 0.034, 0.103 and 0.034, respectively. The ω values of two ORFs in RHDV genome are generally low, indicating that the RHDV whole genome is subject to relatively strong selective constraints. COA was used to investigate the major trend in codon usage variation between two ORFs of all 30 RHDV selected for this study. After COA for RHDV Genome, one major trend in the first axis (f' 1 ) which accounted for 42.967% of the total variation, and another major trend in the second axis (f' 2 ) which accounted for 3.632% of the total variation. The coordinate of the complete coding region of each ORF was plotted in Figure 1 defining by the first and second principal axes. It is clear that coordinate of each ORF is relatively isolated. Interestingly, we found that relatively isolated spots from ORF2 tend to cluster into two groups: the ordinate value of one group (marked as Group 1) is To estimate whether the evolution of RHDV genome on codon usage was regulated by mutation pressure or natural selection, the A%, U%, C%, G% and (C+G)% were compared with A 3 %, U 3 %, C 3 %, G 3 % and (C 3 +G 3 )%, respectively (Table 5 ). There is a complex correlation among nucleotide compositions. In detail, A 3 %, U 3 %, C 3 % and G 3 % have a significant negative correlation with G%, C%, U% and A% and positive correlation with A%, U%, C% and G%, respectively. It suggests that nucleotide constraint may influence synonymous codon usage patterns. However, A 3 % has non-correlation with U% and C%, and U 3 % has noncorrelation with A% and G%, respectively, which haven't indicated any peculiarity about synonymous codon usage. Furthermore, C 3 % and G 3 % have non-correlation with A%, G% and U%, C%, respectively, indicating these data don't reflect the true feature of synonymous codon usage as well. Therefore, linear regression analysis was implemented to analyze the correlation between synonymous codon usage bias and nucleotide compositions. Details of correlation analysis between the first two principle axes (f' 1 and f' 2 ) of each RHDV genome in COA and nucleotide contents were listed in Table 6 . In surprise, only f2 values are closely related to base nucleotide A and G content on the third codon position only, suggesting that nucleotide A and G is a factor influencing the synonymous codon usage pattern of RHDV genome. However, f' 1 value has non-correlation with base nucleotide contents on the third codon position; it is observably suggest that codon usage patterns in RHDV were probably influenced by other factors, such as the second structure of viral genome and limits of host. In spite of that, compositional constraint is a factor shaping the pattern of synonymous codon usage in RHDV genome. Figure 1 A plot of value of the first and second axis of RHDV genome in COA. The first axis (f' 1 ) accounts for 42.967% of the total variation, and the second axis (f' 2 ) accounts for 3.632% of the total variation. Table 5 Summary of correlation analysis between the A, U, C, G contents and A 3 , U 3 , C 3 , G 3 contents in all selected samples There have been more and more features that are unique to RHDV within the family Caliciviridae, including its single host tropism, its genome and its VP10 as a structural protein with unknown function. After we analyzed synonymous codon usage in RHDV (Table 2) , we obtained several conclusions and conjectures as followed. 4.1 Mutational bias as a main factor leading to synonymous codon usage variation ENC-plot, as a general strategy, was utilized to investigate patterns of synonymous codon usage. The ENC-plots of ORFs constrained only by a C 3 +G 3 composition will lie on or just below the curve of the predicted values [18] . ENC values of RHDV genomes were plotted against its corresponding (C 3 +G 3 ) %. All of the spots lie below the curve of the predicted values, as shown in Figure 2 , suggesting that the codon usage bias in all these 30 RHDV genomes is principally influenced by the mutational bias. As we know, the efficiency of gene expression is influenced by regulator sequences or elements and codon usage bias. It reported that the RNA sequence of the 3terminal 84 nucleotides of ORF1were found to be crucial for VP10 expression instead of the encoded peptide. VP10 coding by ORF2 has been reported as a low expressive structural protein against VP60 coding by ORF1 [5] . And its efficiency of translation is only 20% of VP60. According to results showed by Table 4 , it revealed the differences in codon usage patterns of two ORFs, which is a possible factor reducing the expression of VP10. Although VP10 encoded by ORF2, as a minor structural protein with unknown functions, has been described by LIU as a nonessential protein for virus infectivity, the ω Figure 2 Effective number of codons used in each ORF plotted against the GC3s. The continuous curve plots the relationship between GC3s and ENC in the absence of selection. All of spots lie below the expected curve. value of ORF2 suggests VP10 plays an important role in the certain stage of whole RHDV lifecycle. After combining with low expression and ω value of VP10, we conjectured that VP10 might be beneficial for the replication, release or both of virus by inducing infected cell apoptosis initiate by RHDV. This mechanism has been confirmed in various positive-chain RNA viruses, including coxsackievirus, dengue virus, equine arterivirus, footand-mouth disease virus, hepatitis C virus, poliovirus, rhinovirus, and severe acute respiratory syndrome [23] [24] [25] [26] [27] [28] [29] , although the details remain elusive. As preceding description, ENC reflects the evolution of codon usage variation and nucleotide composition to some degree. After the correlation analysis of ENC values between ORF1 and ORF2 (Table 7) , the related coefficient of ENC values of two ORFs is 0.230, and p value is 0.222 more than 0.05. These data revealed that no correlation existed in ENC values of two ORFs, indicating that codon usage patterns and evolution of two ORFs are separated each other. Further, this information maybe helps us well understand why RSCU and ENC between two ORFs are quite different. Interestingly, we found that relatively isolated spots from ORF2 tend to cluster into two groups: the ordinate value of one group (marked as Group 1) is positive value and the other one (marked as Group 2) is negative value. And all of those strains isolated before 2000 belonged to Group 2, including Italy-90, RHDV-V351, RHDV-FRG, BS89, RHDV-SD and M67473.1. Although RHDV has been reported as only one type, this may be a reference on dividing into two genotypes. In this report, we firstly analyzed its genome and two open reading frameworks (ORFs) from this aspect of codon usage bias. Our researches indicated that mutation pressure rather than natural is the most important determinant in RHDV with high codon bias, and the codon usage bias is nearly contrary between ORF1 and ORF2, which is maybe one of factors regulating the expression of VP60 (encoding by ORF1) and VP10 (encoding by ORF2). Furthermore, negative selective constraints on the RHDV whole genome implied that VP10 played an important role in RHDV lifecycle. We conjectured that VP10 might be beneficial for the replication, release or both of virus by inducing infected cell apoptosis initiate by RHDV. According to the results of the principal component analysis for ORF2 of RSCU, we firstly separated 30 RHDV into two genotypes, and the ENC values indicated ORF1 and ORF2 were independent among the evolution of RHDV. All the results will guide the next researches on the RHDV as a reference.

Oligonucleotide Based Magnetic Bead Capture of Onchocerca volvulus DNA for PCR Pool Screening of Vector Black Flies

BACKGROUND: Entomological surveys of Simulium vectors are an important component in the criteria used to determine if Onchocerca volvulus transmission has been interrupted and if focal elimination of the parasite has been achieved. However, because infection in the vector population is quite rare in areas where control has succeeded, large numbers of flies need to be examined to certify transmission interruption. Currently, this is accomplished through PCR pool screening of large numbers of flies. The efficiency of this process is limited by the size of the pools that may be screened, which is in turn determined by the constraints imposed by the biochemistry of the assay. The current method of DNA purification from pools of vector black flies relies upon silica adsorption. This method can be applied to screen pools containing a maximum of 50 individuals (from the Latin American vectors) or 100 individuals (from the African vectors). METHODOLOGY/PRINCIPAL FINDINGS: We have evaluated an alternative method of DNA purification for pool screening of black flies which relies upon oligonucleotide capture of Onchocerca volvulus genomic DNA from homogenates prepared from pools of Latin American and African vectors. The oligonucleotide capture assay was shown to reliably detect one O. volvulus infective larva in pools containing 200 African or Latin American flies, representing a two-four fold improvement over the conventional assay. The capture assay requires an equivalent amount of technical time to conduct as the conventional assay, resulting in a two-four fold reduction in labor costs per insect assayed and reduces reagent costs to $3.81 per pool of 200 flies, or less than $0.02 per insect assayed. CONCLUSIONS/SIGNIFICANCE: The oligonucleotide capture assay represents a substantial improvement in the procedure used to detect parasite prevalence in the vector population, a major metric employed in the process of certifying the elimination of onchocerciasis.

Onchocerciasis, or river blindness has historically represented one of the most important neglected tropical diseases in the developing world as measured as a cause of socio-economic disruption [1] . It is also considered a candidate for elimination by the international community [2, 3] . As a result of these factors, the international community is currently supporting several programs whose goals are either to eliminate the disease as a public health problem, or to locally eliminate the causative agent of the disease, Onchocerca volvulus. These include the Onchocerciasis Elimination of the Americas (OEPA), the African Programme for Onchocerciasis Control (APOC), and the Ugandan Onchocerciasis Elimination Program (UOEP). Entomological criteria play an important role in the elimination criteria recommended by the World Health Organization (WHO) [4] and those currently utilized by OEPA [5] . Entomological data play an especially important role in the certification of elimination following the cessation of treatment, as the prevalence of infective stages of the parasite in the fly population is the timeliest measure of transmission in a given area. However, demonstrating that transmission is interrupted requires that large numbers of flies be tested. For example, current OEPA guidelines require that the prevalence of flies carrying infective larvae (L3) be less than 1/2000 in every sentinel community for transmission to be interrupted [5] . In order to be able to state with a 95% confidence that the prevalence of infective flies is less that 1/2000 requires examining approximately 6000 flies from each sentinel community. Examining such large numbers of insects using conventional methods (dissection) is impractical. For this reason, the current guidelines recommend the use of pool screening PCR based methods to conduct the entomological studies necessary to document transmission interruption [4] . Currently, the accepted method for pool screening vector black flies to detect O. volvulus relies upon screening DNA prepared from fly pools with a PCR assay targeting a repeated sequence family (the O-150 repeat [6] ) specific for parasites of the genus Onchocerca. Algorithms have been developed that permit one to derive a point estimate of the prevalence of infection in the fly population (and an associated confidence interval) from the number of PCR positive pools and the number of flies contained in each pool [7] . Furthermore, because the infective stage of the parasite is the only form found in the black fly head capsule, separated pools of heads and bodies may be screened to obtain estimates of the prevalence of infective flies (flies with infective stages in their head) and the prevalence of infected flies (flies with immature larval stages in their bodies). This approach has been used to monitor transmission of O. volvulus in many foci of Latin America and Africa [8] [9] [10] [11] , as well as to certify the interruption of transmission in foci on both continents [5, [12] [13] [14] . Previous modeling studies have shown that increasing pool sizes has relatively little effect on the accuracy of the estimate of prevalence of infection obtained, so long as the proportion of positive pools remains less than the majority of pools screened [15] . Thus, in situations where pool screening is used to certify transmission interruption (where infective flies are extremely rare or non-existent) the pool size is only limited by the biochemical constraints of the assay. The current method of DNA extraction for the O-150 PCR assay is based upon adsorption to a silica matrix [16] . This preparation results in DNA samples that still contain inhibitors of the PCR, limiting the number of flies that may be included in each pool. Currently, pool sizes are limited to 50 individual heads or bodies (in the case of flies from Latin America) [9] or 100 individual heads or bodies (in the case of flies from Africa) [7] . Developing alternative methods to prepare DNA that would permit an increase in the maximum number of heads or bodies in each pool would decrease the cost and effort necessary to screen the requisite large numbers of flies necessary to certify transmission interruption. Magnetic bead based purification protocols have been developed for many different pathogens. Most of these involve direct capture of the pathogen using beads coated with pathogen-specific antibodies. This method, known as immunomagnetic separation (IMS), has been successfully used to purify and concentrate viruses [17] , bacteria [18, 19] and fungal [20] pathogens. Similarly, methods have been developed which use oligonucleotides to magnetically purify pathogen genomic DNA [21] . Here, we describe a magnetic bead capture method to isolate O. volvulus DNA from homogenates prepared from pools of Latin American and African Simulium vector black flies. This method is shown to be an improvement upon the current DNA purification method utilizing organic extraction and silica adsorption. Simulium ochraceum s.l. females were collected in public areas of the community of José María Morelos y Pavón, Chiapas, México between the hours of 0700 and 1000. Previous studies have demonstrated that the majority of flies captured during this period were nulliparous, and the risk of infection was therefore minimal. Simulium damnosum s.l. were obtained from breeding sites on public land located near the communities of Bodajugu and Sakora. These communities are located in the Region des Cascades in Southwestern Burkina Faso. This region is located within the area of the former Onchocerciasis Control Programme in West Africa, where onchocerciasis has been eliminated as a public health problem. O. volvulus L3 were obtained from experimentally infected Simulium damnosum s.l. flies 7 days after infection with skin microfilariae. The flies were kept at 25uC and 80% relative humidity to allow the microfilariae to develop into L3. Larvae were isolated from the flies by dissection into dissecting medium (IMDM+10% FCS+2x Penicillin-Streptomycin-Fungizone) using a dissecting microscope. The cleaned larvae were frozen in 9% DMSO, 4 mM PVP, 10% FCS in Grace medium using Bio-Coll (freezing to 240uC at 1uC/minute followed by 30 minutes at 240uC) and then transferred to liquid nitrogen for long-term storage. The parasite material was prepared in the Tropical Medicine Research Station, Kumba, Cameroon, and is being stored at the New York Blood Center. Pools containing varying numbers of black flies were prepared and the heads and bodies separated by freezing and agitation, as previously described [7] . A single O. volvulus L3 was added to each pool. Head and body pools were placed in a 1.5 ml microcentrifuge tube and purified using magnetic silica coated beads (Machery-Nagel GmbH & Co, Bethlehem, PA, USA) following the instructions provided by the manufacturer. In brief, the pools were homogenized in 200 ml of T1 buffer, 25 ml of proteinase K solution provided in the kit was added and the homogenates were incubated at 56uC for 30 minutes. The homogenates were subjected to centrifugation at 13,4006g for 5 minutes at room temperature. A total of 225 ml of the supernatant was transferred to a fresh tube containing 24 ml B-beads (Machery-Nagel) and 360 ml MB2 buffer (Machery-Nagel), and the tube shaken for 5 minutes at room temperature. The magnetic beads were isolated by placing the tubes in a six tube magnetic separator (Dynal MPC-S; Invitrogen). The supernatants were removed and discarded, and 600 ml of MB3 wash buffer was added to each sample. The bead/ DNA complexes were washed by shaking for 5 min at room temperature. The beads were collected in the magnetic separator The absence of infective larvae of Onchocerca volvulus in the black fly vector of this parasite is a major criterion used to certify that transmission has been eliminated in a focus. This process requires screening large numbers of flies. Currently, this is accomplished by screening pools of flies using a PCR-based assay. The number of flies that may be included in each pool is currently limited by the DNA purification process to 50 flies for Latin American vectors and 100 flies for African vectors. Here, we describe a new method for DNA purification that relies upon a specific oligonucleotide to capture and immobilize the parasite DNA on a magnetic bead. This method permits the reliable detection of a single infective larva of O. volvulus in pools containing up to 200 individual flies. The method described here will dramatically improve the efficiency of pool screening of vector black flies, making the process of elimination certification easier and less expensive to implement. as before, and the washing procedure repeated with successive washes with 600 ml of MB4 and MB5 wash buffers (Machery-Nagel). Following the wash in the MP5 buffer, the beads were exposed to air for one minute to permit the traces of ethanol to evaporate. DNA was eluted from the beads by the addition of 100 ml of elution buffer. The beads were shaken for 5 minutes at room temperature to elute the DNA, and the beads removed by placing the tubes in the magnetic separator. The supernatant containing the purified DNA was then transferred to a fresh tube. Oligonucleotide capture of O. volvulus DNA Pools of spiked heads and bodies were prepared as described above. The head and body pools were homogenized in 500 ml of 10 mM Tris-HCl (pH 8.0) 1 mM EDTA, and proteinase K added to a final concentration of 2 mg/ml. The homogenates were incubated at 56uC for 2 hours, and dithiothreitol added to a final concentration of 20 mM. The samples were heated to 100uC for 30 minutes and subjected to three freeze-thaw cycles. The homogenates were subjected to centrifugation at 13,4006g for 5 minutes and the supernatant placed into a new tube. The solutions were brought to a final concentration of 100 mM Tris-HCl (pH 7.5) 100 mM NaCl. A total of 5 ml of a 0.5 mM solution of OVS2-biotin primer (59B-AATCTCAAAAAACGGGTA-CATA-39, where B = biotin) was added to each sample. The samples were then heated to 95uC for three minutes and allowed to cool slowly to room temperature. While the probe was annealing to the DNA in the solution, 10 ml of Dynal M-280 strepavidin coated beads (Invitrogen) were placed in a single well of a 96 well tissue culture plate. The plate was placed on a magnetic capture unit (Dynal MPC-96, Invitrogen) and the beads collected for 2 minutes. The beads were then washed five times with 200 ml binding buffer (100 mM Tris-HCl (pH 7.5) 100 mM NaCl) per wash, resuspended in 10 ul and added to the sample. The samples were incubated at 4uC overnight on a roller to permit the oligonucleotide-DNA hybrids to bind to the beads. The samples were placed in the magnetic separator for two minutes to capture the beads and the supernatant discarded. The beads were resuspended in 150 ml of binding buffer by pipetting, and the beads captured by placing the tubes in the magnetic separator for two minutes. The wash step was repeated five times. The beads were then resuspended in 20 ml of sterile water, heated to 80uC for 2 minutes and cooled rapidly on ice for two minutes. The beads were removed by placing the tubes in the magnetic capture apparatus, and the supernatant containing the purified DNA transferred to a new tube. A total of 2.5 ml of the purified genomic DNA was used as a template for the PCR amplifications carried out in a total volume of 50 mL containing 0.5 mmol/L of O-150 primer (59-GAT-TYTTCCGRCGAANARCGC-39) and 0.5 mmol/L of biotinylated O-150 primer (59-B-GCNRTRTAAATNTGNAAATTC-39, where B = biotin; N = A, G, C, or T; Y = C or T; and R = A or G). Reaction mixtures also contained 60 mM Tris-HCl, (pH 9.0), 15 mM (NH4) 2 SO4, 2 mM MgCl 2 , 0.2 mM each of dATP, dCTP, dGTP and dTTP, and 2.5 units of Taq polymerase (Invitrogen). Cycling conditions consisted of five cycles of one minute at 94uC, two minutes at 37uC, and 30 seconds at 72uC, followed by 35 cycles of 30 seconds each at 94uC , 37uC , and 72uC. The reaction was completed by incubating at 72uC for six minutes. Amplification products were detected by PCR enzyme-linked immunosorbent assay (ELISA), essentially as previously described [10] . Briefly, 5 ml of each PCR reaction was bound to a streptavidin-coated ELISA plate, and the DNA strands denatured by treatment with alkali. The bound PCR fragments were then hybridized to a fluorescein-labeled O. volvulus-specific oligonucleotide probe (OVS2: 59-AATCTCAAAAAACGGGTACATA-FL-39), and the bound probe detected with an alkaline phosphataselabeled anti-fluorescein antibody (fragment FA; Roche Diagnostics). Bound antibody was detected using the ELISA amplification reagent (BluePhos) kit from KPL (Gaithersburg, USA) following the manufacturer's instructions. Color development was stopped by the addition of 100 ml AP stop solution, and the plates read in an ELISA plate reader set at 630 nm. Samples were scored positive if their optical density exceeded either the mean plus three standard deviations of ten negative control wells run in parallel or 0.1, whichever was greater. An initial series of experiments were carried out with pools of heads and bodies of S. ochraceum s.l. (a major Latin American vector of onchocerciasis). Pools containing varying numbers of heads of bodies were spiked with a single O. volvulus L3, and DNA prepared from the pools using either the conventional method of organic extractions followed by adsorption to a silica matrix, or by oligonucleotide capture of O. volvulus genomic DNA followed by magnetic purification of the captured oligonucleotide-DNA complexes. The conventional method consistently produced a positive signal in pools containing up to 50 heads (Table 1) . Pools containing greater than 50 heads were not positive in the assay. All (Table 1 ). In contrast, positive signals were obtained in all pools containing up to 200 heads or bodies in the assays performed on the oligonucleotide capture purified DNA samples ( Table 1) . The preliminary experiments suggested that the oligonucleotide capture method was capable of detecting one L3 in pools of up to 200 heads or bodies. To further explore the sensitivity of the assay, the experiment was repeated using 10 separate pools containing 200 heads or bodies spiked with a single L3. All pools were found to be positive, suggesting that the oligonucleotide capture assay was capable of consistently detecting a single L3 in pools of up to 200 heads or bodies of S. ochraceum s.l. (Table 2) . Previous studies had demonstrated that the conventional silica adsorption method was capable of detecting a single infected S. damnosum s.l. (the major African vector of onchocerciasis) fly in a pool containing up to 99 uninfected flies [7] . To determine if the performance of the oligonucleotide capture assay was similar when applied to S. damnosum s.l., the spiking experiments were repeated employing pools containing 200 S. damnosum s.l. heads or bodies. All spiked pools were found to be positive ( Table 2 ), suggesting that the capture assay preformed equally well on both African and Latin American vectors of onchocerciasis. For the capture assay to be cost effective, it should be competitive with the cost of the conventional silica adsorption assay. The two assays require roughly the same amount of technical time, so labor costs may be assumed to be equivalent per sample for the two assays. However, because it will be possible to increase the size of the pools 2-4 fold when using the capture assay, a reduction in labor costs of between 50% and 75% would be realized when costs are considered on a per-fly-tested basis. Similarly, the per-sample cost of carrying out the conventional assay is roughly $2.22 per pool, while the cost of the magnetic bead assay is $3.81 per pool. However, because the magnetic bead permits more flies to be tested per pool, cost savings in reagents are realized when the costs are amortized on a per fly basis (Table 3) . Previous studies have demonstrated that the algorithms used to predict the prevalence of infection in a population from data derived from screening pools of samples are relatively insensitive to the size of the pool, so long as the proportion of positive pools does not represent a substantial majority of the samples screened [15] . Thus, the size of pools used in a pool screening protocol is more likely to be limited by the biochemistry of the detection assay than by the underlying statistical uncertainties associated with screening pools of samples. This is particularly true when positive samples are extremely rare, as in the case when monitoring for transmission interruption. The data presented above suggest that the oligonucleotide capture method of purifying O. volvulus DNA is superior to the conventional silica adsorption method. Mixing experiments have demonstrated that PCR inhibitors carried through the silica adsorption process limits the size of the pools that may be screened to 50 individual flies for S. ochraceum and 100 for S. damnosum s.l (data not shown). The oligonucleotide capture method appears to result in DNA preparations that are freer of PCR inhibitors than are those prepared using silica adsorption. The practical result of this improvement is that it permits a 2 to 4fold increase in the number of black fly heads or bodies that can be included in a single pool. This increase in pool size results in a dramatic cost savings in the per-unit cost of the O-150 pool screen assay. Using the oligonucleotide capture assay, labor costs are reduced by 50-75%, while the overall cost of reagents needed is $3.81 per pool of 200 flies or less than $0.02 per individual fly. The decrease in cost and corresponding increase in the efficiency of the assay will make it more practical to screen the large numbers of flies necessary to demonstrate transmission interruption and to certify elimination of onchocerciasis. Because the oligonucleotide capture assay is a modification of the conventional silica adsorption assay, the equipment required to carry out both assays is quite similar. The only additional equipment necessary when replacing the conventional assay with the oligonucleotide assay is the magnetic capture apparatus. This unit is relatively inexpensive, costing less than $579 (Invitrogen's list price). This would be recovered in reagent costs alone after screening just 114 pools of flies. The O-150 PCR is based upon the amplification of a genus specific tandemly repeated DNA sequence present in the genome of Onchocerca parasites [22] . Thus, the standard O-150 PCR will amplify sequences present in all Onchocerca species, including Onchocerca ochengi, a cattle parasite that is sympatric with O. volvulus in sub-Saharan Africa. Currently, the O-150 PCR is made species or strain specific by modifying the amplification conditions to limit amplification to species specific members of the O-150 repeat family [23] or by the use of species or strain specific probes to detect the resulting amplicons [8] . However the oligonucleotide used in the capture assay (OVS2) has previously Table 3 . Cost analysis of silica adsorption and oligonucleotide capture assays. Step

Prediction and Identification of T Cell Epitopes in the H5N1 Influenza Virus Nucleoprotein in Chicken

T cell epitopes can be used for the accurate monitoring of avian influenza virus (AIV) immune responses and the rational design of vaccines. No T cell epitopes have been previously identified in the H5N1 AIV virus nucleoprotein (NP) in chickens. For the first time, this study used homology modelling techniques to construct three-dimensional structures of the peptide-binding domains of chicken MHC class Ι molecules for four commonly encountered unique haplotypes, i.e., B4, B12, B15, and B19. H5N1 AIV NP was computationally parsed into octapeptides or nonapeptides according to the peptide-binding motifs of MHC class I molecules of the B4, B12, B15 and B19 haplotypes. Seventy-five peptide sequences were modelled and their MHC class I molecule-binding abilities were analysed by molecular docking. Twenty-five peptides (Ten for B4, six for B12, two for B15, and seven for B19) were predicted to be potential T cell epitopes in chicken. Nine of these peptides and one unrelated peptide were manually synthesized and their T cell responses were tested in vitro. Spleen lymphocytes were collected from SPF chickens that had been immunised with a NP-expression plasmid, pCAGGS-NP, and they were stimulated using the synthesized peptides. The secretion of chicken IFN-γ and the proliferation of CD8(+) T cells were tested using an ELISA kit and flow cytometry, respectively. The significant secretion of chicken IFN-γ and proliferation of CD8(+) T lymphocytes increased by 13.7% and 11.9% were monitored in cells stimulated with peptides NP(89–97) and NP(198–206), respectively. The results indicate that peptides NP(89–97) (PKKTGGPIY) and NP(198–206) (KRGINDRNF) are NP T cell epitopes in chicken of certain haplotypes. The method used in this investigation is applicable to predicting T cell epitopes for other antigens in chicken, while this study also extends our understanding of the mechanisms of the immune response to AIV in chicken.

The introduction into the human population of animal-derived influenza A viruses with a novel haemagglutinin (HA), or a novel HA and neuraminidase (NA), and their subsequent spread could result in global influenza pandemics [1] . Since 2003, the highly pathogenic H5N1 avian influenza virus (AIV) has caused numerous cases of severe disease and death in humans [2] . An influenza pandemic could ensue if this virus developed the capacity to spread easily among humans [3] [4] [5] . Migratory birds constitute the natural reservoir for AIVs, but chickens may play a key role in the transmission to humans [6] . Epitopes can be used for accurately monitoring immune responses to AIV and for the rational design of protective vaccines. However, only two epitopes from the H5N1 avian influenza A/Vietnam/1194/2004 virus are included in the Immune Epitope Database and Analysis Resources (IEDB). The majority of T cell and B cell epitopes have been identified in mouse, human, or rabbit hosts. Few epitopes have been described in chicken [7] . The structural basis of peptide binding to mammalian major histocompatibility complex (MHC) class I molecules is well understood [8] . The peptide-binding groove is formed by the a1 and a2 domains. Each domain contributes four strands to an eight -stranded anti-parallel b-sheet. Two long interrupted helices, one from each domain, pack against the side of this sheet in an orientation directed away from the cell membrane. There is a series of pockets (A-F) along the peptide-binding groove where highly polymorphic amino acids mediate recognition via haplotype-specific associations with antigens and T cell receptors. However, highly conserved residues are found at both ends of the peptide-binding groove that form a network of hydrogen bonds, which directly interact with hydrogen bonds at the peptide's Nterminus and C-terminus [9] [10] [11] . Peptides that bind to MHC class I molecules are usually octamers or nonamers, where only one or a few residues can interact with polymorphic residues in the groove. These residues are known as anchor residues when they are found in an anchoring position [12] [13] [14] [15] . Several public databases and prediction services are available for MHC molecular ligands and peptide motifs, including SYFPEITHI and RANKPEP [16, 17] . Unlike mammalian studies, the majority of investigations of MHC class I molecules in chicken remain limited to primary sequences and the structure of MHC class I molecule from the B21 haplotype was solved only recently [18, 19] . The chicken MHC B system is located on chicken Chromosome 16 (Chr16) and is composed of tightly linked polymorphic regions: BF (class I) and BL (class IIb) and a large family of polymorphic Ig-superfamily (IgSF) genes called BG-the latter sharing sequence similarities with mammalian MHC butyrophilin, myelin oligodendrocyte glycoprotein, and TRIM genes [20, 21] . The MHC B-F molecules are structurally and functionally similar to mammalian MHC class I molecules. The two class I genes are known as BF1 and BF2, although the BF2 gene is mainly expressed. The MHC class I (B-F) molecules present antigen peptides to the CD8 + T lymphocytes, which have a central role in the immune system. The MHC class I molecules of different haplotypes have specific peptide-binding tropisms, and B-F-associated peptide-binding motifs have been determined for several haplotypes, including B4, B12, B15, and B19 [22] . Thirteen peptides derived from the v-src gene of the Rous sarcoma virus (RSV) Prague strain were predicted to fit the peptide-binding motif of the MHC class I molecules of B12 haplotype and all 13 synthetic peptides actually bound to the BF12 class I molecule in subsequent binding tests [23] . This demonstrates that peptide-binding motifs can be used to predict the antigen peptides presented by chicken MHC class I molecules. However, no T cell epitope of H5N1 AIV NP has yet been identified in the chicken [7] , while no information is available on the structure of chicken MHC class I molecules belonging to the B4, B12, B15, and B19 haplotypes. For the first time, the current study reports the homology modelling structures of the peptidebinding domains of chicken MHC class I molecules for the B4, B12, B15, and B19 haplotypes. Potential T cell epitopes were predicted by molecular docking of 25 peptides in the H5N1 AIV NP in chicken. NP 89-97 and NP 198-206 were shown to be T cell epitopes of H5 AIV NP by analysing the CD8 + T cell proliferation and the interferon (IFN-c) expression. To our knowledge, this is the first report to describe the structure of chicken MHC class I molecules from the B4, B12, B15 and B19 haplotypes and the T cell epitopes of H5N1 AIV NP in chicken. All computations were conducted using the Discovery Studio 2.5 (DS2.5) program developed by Accelrys Software Inc and the SYBYL 8.1 program developed by Tripos Inc on an SGI Fuel workstation running Red Hat Enterprise 5.3 and a Dell server running the Red Hat Enterprise 5.2 Linux operating system. SPF chickens were housed in HEPA-filtered isolators. Chicken lymphocytes collected from immunized chickens, which were approved by Harbin Veterinary Research Institute, Chinese Academy of Agricultural Sciences and performed in accordance with animal ethics guidelines and approved protocols. The Animal Ethics Committee approval number is Heilongjiang-SYXK-2006-032. DNA vaccine pCAGGS-NP was constructed and assessed according to Ma and Jiang [24, 25] . The synthetic gene for NP of the isolate A/Goose/Gongdong/1/96 (H5N1) with codons optimized for chicken usage was synthesized by PCR assembly of long single-strand DNA templates (100 bases in length) (the oligonucleotide sequences are available upon request). The synthetic optiNP was cloned into the plasmid vector pCAGGS under the control of the chicken b-actin promoter. The plasmid was named pCAGG-NP. The expression of the NP protein from the plasmid was confirmed by indirect immunofluorescence assay and Western blotting of plasmid transfected CEF cells. Nine of the predicted peptides (Table 1) were selected to test their T cell responses in vitro. They were synthesized to .90% purity by the GenScript Corporation (Nanjing, China): NP [7] [8] [9] [10] [11] [12] [13] [14] (KRSYEQME), Amino sequence of H5N1 AIV(GD/1/96) NP was parsed into octamers or nonamers according to the peptide-binding motifs of MHC class I molecules belonging to the B4, B12, B15, and B19 haplotypes. The 3D structures of those peptides and the MHC class I molecules were predicted using homology modelling method and then the binding affinity between each peptide and MHC class I molecule was analysed using molecular docking. Those peptides, which have correct binding conformation and high binding affinity, were predicted to be potential T cell epitopes in chicken. Predicted peptides 9 of 25 were synthesized and used to stimulate splenic lymphocytes collected from NP DNA vaccineimmunized SPF chickens. MHC class I molecule-restricted potential T-cell epitopes were confirmed by detecting of CD8 + T cell proliferation and interferon (IFN-c) expression. The amino acid sequences of chicken MHC class I molecules belonging to the B4, B12, B15, and B19 haplotypes were obtained from GenBank (B4, GenBankID: CAK54654.1; B12, GenBankID: BAG69386.1; B15, GenBankID: BAG69413.1; and B19, Gen-BankID: BAG69441.1). BLASTP was performed to search for homologous proteins in the Protein Data Bank (PDB) using the BlOSUM62 Scoring Matrix optimized with a gap penalty of 11 and a gap extension penalty of 1. The MODELER [27] program was then executed in DS2.5 to construct the 3D structures of chicken MHC class I molecules for the four haplotypes. The quality of the 3D models was evaluated with a Ramachandran Plot using the PROCHECK [28] program and the Verify Protein (Profiles-3D) program [29] in DS2.5. To improve the quality of the models, unsatisfied loop regions in each model were refined using the loop-refinement protocol based on the MODELER energy in DS 2.5. All structures were then minimized using the CHARMm force field [30] and an explicit solvent model (TIP3P water) with a steepest descent method for 8000 steps and a conjugated gradient minimization for a further 8000 steps. The cavity depth, lipophilic potential (LP), flexibility (FX), and electrostatic potential (EP) of the four structures was analysed using the MOLCAD program in SYBYL8.1. During this process, the Gasteiger-Hückel charges were assigned to all atoms, while surface maps were generated and visualized using SYBYL8.1. The NP protein sequence of H5N1 isolate A/Goose/Gongdong/1/96 (H5N1) was downloaded from the UniProt database (UniProtID: NCAP_I96A0) and automatically parsed as octapep-tides or nonapeptides using a computer program, which was developed in our laboratory, based on the peptide-binding motifs of chicken MHC class I molecules belonging to the B4, B12, B15, and B19 haplotypes [22] . The motifs were as follows: The structures of the octapeptides and nonapeptides were modelled using DS2.5. In total, seventy-five peptide structures were prepared for docking analysis. Surflex-Dock [31] is used to dock ligands into a protein-binding site and it is particularly successful at eliminating false positive results [32] while offering unparalleled enrichment in virtual highthroughput screening combined with state-of-the-art speed, accuracy, and usability [33, 34] . To test the feasibility of using Surflex-Dock for docking MHC molecules and peptides, 50 peptide-MHC complexes retrieved from the PDB database were separated and re-docked using the Surflex-Dock program. The RMSD was calculated for each modelled pair and crystal structure pair to evaluate the docking program. Surflex-Dock was used to dock the octapeptides and nonapeptides to their corresponding MHC class I molecule receptors, where the parameters of the threshold and bloat values of the program were optimized to 0.51 and 1. The interaction modes and binding energies of the docked complexes were analysed using DS2.5. Peptides with a higher docking score and rational conformation were predicted as candidate T cell epitopes for each haplotype. For the DNA vaccine immunizations, 13 three-week-old SPF chickens were separated into two groups, eight were immunized twice with 100 mg of pCAGGS-NP in their leg muscle at threeweek intervals, while five chickens were injected with the same volume of PBS as controls. Sera were collected weekly for detecting of NP antibody. Serum antibodies to H5N1 AIV NP were detected using an indirect ELISA method as described by [24] , which used prokaryotically-expressed NP as the antigen. The testing steps as follows, after washing of the plates, 50 ml of the test serum mixed in the test wells with 50 ml of antigen diluted 1:10 in ELISA buffer. After incubation at 37uC for 1 h, 100 ml horseradish peroxidase conjugate, diluted 1:1000 in ELISA buffer, was added to each well and plates were further incubated at 37uC for 1 h. After two washing steps, 100 ml of TMB substrate was added and incubated at room temperature for 10 min. The reaction was stopped by adding 100 ml H 2 SO 4 2 M. The extinctions were measured at 490 nm with a micro ELISA reader (Bio-Rad). To prepare splenic lymphocytes, all eight pCAGGS-NP immunized chickens were killed by cardiac puncture blood collection. Sterile spleens were collected and meshed through a sieve screen using a syringe plunger to obtain a single-cell suspension in tissue culture medium (RPMI 1640, Gibco BRL NY, USA). Cell suspensions were overlaid onto Histopaque 1077 density gradient medium and centrifuged at 1800 rpm for 20 min at 18uC. Lymphocytes were collected from the interface and washed three times in RPMI, before cells were counted using a trypan blue dye exclusion assay. Splenic lymphocytes collected from the 8 immunized chickens were stained with 5 mM CFSE (Invitrogen) in pre-warmed PBS for 10 min, washed three times, and suspended in RPMI 1640 containing 2 mM L-glutamine, 100 IU mL -1 penicillin, 100 IU mL -1 streptomycin, and 10% foetal bovine serum (R10 medium). Cells were plated at 10 6 well -1 in 24-well plates with RPMI 1640 medium, before stimulation for 5d with 100 mg mL -1 of the nine peptides generated from NP and the unrelated peptide N 71-78 . Cells were then washed with PBS and stained using anti-CD8-PE before flow cytometric analysis, which was performed on Cytomics FC500 MCL(Beckman) and analysed with its embedded software CXP. Two repeats were performed simultaneously for each peptide during the flow cytometric analysis. Splenic lymphocytes collected form the 8 vaccinated chickens were plated at 10 6 well -1 in 24-well plates with RPMI 1640 medium, before stimulation for 48 h using 100 mg mL -1 of the nine peptides generated from NP and the unrelated peptide N 71-78 . Cells were then collected and centrifuged at 1000 rpm for 5 min. The cell culture supernatants were then analysed to determine the chicken IFN-c using chicken IFN-c CytoSets TM (Invitrogen), according to the manufacturer's instructions [35] . The extinctions were measured at 450 nm with a micro ELISA reader (Bio-Rad). For each peptide, two copies of splenic lymphocytes from one chicken were treated and analysed simultaneously. Results were expressed as mean 6 S.E.M. for two replicates. Statistical analyses were performed using SPSS 16.0 for Windows. Significant differences (P,0.05) between means were tested by one-way ANOVA, followed by Tukey's Honestly Significant Difference test. The four haplotypes selected in this study belonged to 11 commonly encountered unique haplotypes (B2, B4, B5, B6, B7, B12, B13, B14, B15, B19, and B21) and the motifs of their binding peptides were previously reported [22, 23] . The peptide-binding groove of the chicken MHC class I molecule contains a1 and a2 domains [18] , so only those two domains were used for modelling and analysis in this study. BLASTP search results showed that the chicken MHC class I molecule 3BEV (PDB accession number) belonging to the B21 haplotype had the highest sequences identity and similarity with the four target sequences. The sequence identity and similarity between 3BEV and B4, B12, B15, and B19 were 84.3%, 84.3%, 87.1%, and 89.9%, and 89.3%, 89.9%, 91%, and 93.3%, respectively. The sequence alignments of the four target sequences and 3BEV are shown in Fig. 1. 3BEV has been crystallized to a high resolution of 2.1 Å , so it was selected as the template for modelling the four protein structures. The initial crude homology models were built using MOD-ELLER and refined using modules for loop refinement and CHARMM minimization in DS2.5. Thus, the final structures were of high quality. The Profiles-3D scores of all amino acid residues in the four structures were greater than zero (Fig. 2) , while an evaluation of the stereochemical quality of the models using PROCHECK showed that no residues were in the disallowed regions of the Ramachandran plots (Fig. 3) . This indicated that the backbone dihedral angles, phi and psi, of the four structure models were reasonably accurate. The four models are similar to the previously reported B21 haplotype. The peptide-binding domain is formed by two helices at the top and an eight-stranded sheet at the bottom (Fig. 4A) . The pairwise root mean-square differences (RMSD) between the Ca positions of 3BEV and the structures of the B4, B12, B15, and B19 haplotypes are very small (0.23, 0.20, 0.24, and 0.30 respectively). However, there are some variations around the peptide-binding grooves and these differences may determine differences in their peptide-binding properties. Specific residues in the binding groove have an essential role in peptide binding. Figure 4B shows the key residues comprising the binding site, when all the models and the template were superimposed. Peptide binding to a given class I MHC molecule requires the presence of anchor residues that complement the physicochemical characteristics of the specificity pockets [36] . Anchor residues are usually found at peptide position 2 (P2), where they interact with pocket B, but sometimes at position 5 or 6 (P5 or P6), where they interact with pocket C or E. A C-terminal residue (PC) also typically binds in pocket F [8] . The anchor residues Glu or Asp (P2) in B4 have an electrostatically favourable interaction with the side-chain of the pocket B residue Aspa9, which points up from the b-sheet. Similarly, anchor residues P2 Arg in B15 and B19 provide an electrostatically favourable interaction with the negatively-charged residues Aspa34 and Glua62, the side-chains of which point towards the binding groove from the a-helix. The presence of Glu8 or Asp8 as major residues in the C-terminal position of the B4 motif is unprecedented, and there are no previous mammalian examples of negatively-charged residues in the C-terminal anchor position [16] . It is possible that the positively-charged Arga80 residue of the F pocket is the key to the binding of anchor residues P8Glu or P8Asp. There is an anchor residue in the central cavity region corresponding to pocket C in B12 and B4. This is very similar to the mammalian MHC class I molecules H-2 Kb and H-2 Db, and it may be related to the Glya68(69) residue without a side-chain located in the a-helix (the residue number in parentheses is derived from mammals, whereas that without parentheses is from the chicken). The volumes of the binding grooves in B4 and B12 are 662.45 Å 3 and 637.677 Å 3 , respectively, which are greater than those of B15 (582.595 Å 3 ) and B19 (581.676 Å 3 ) (Fig. 5) . This may explain why the former have anchor residues in the middle region of their binding peptides. In general, anchor residues point downwards into the binding groove, which allows them to fill the pocket and maintain the stability of the peptide-MHC complex. The electrostatic potential of the peptide-binding groove of the B4 haplotype is highly positively-charged, whereas those of the B15 and B19 haplotypes are negatively-charged and that of B12 is neutral (Fig. 6) . The peptide-binding grooves are highly lipophilic in all models (Fig. 7) . The B pockets of the peptide-binding grooves . Verification score plots of the models (plots generated using DS2.5). The verification scores of all residues in B15 were greater than zero, whereas one residue in B12 and B19 and three residues in B4 were less than zero. doi:10.1371/journal.pone.0039344.g002 are more flexible than the F pockets in all structures (Fig. 8) , which suggests that the variation of the anchor residues in the B pocket is greater than in the F pocket. As described in the Materials and Methods, 50 peptide-MHC complexes were selected from the PDB database to validate the docking process. All the peptides were extracted from the complexes, before they were re-docked to the corresponding MHC molecules using Surflex-Dock. When compared with the corresponding initial complexes, the average RMSD was 1.8 and the docking scores were all greater than 8.0 (Table 2) . Docking was also successfully performed with v-src C-tail peptide 517-524 (LPACVLEV) and an identified T cell epitope [23] to the modelled B12 MHC class I molecule, where the docking score was 10.91 and the docking complex conformation was rational. Thus, it was appropriate to use Surflex-Dock for screening T cell epitopes based on their homology-modelled structures. Based on the assessment results, the criteria for T cell epitope prediction using Surflex-Dock were defined as follows: a) the peptide made close contact with the groove and it docked in the correct direction; b) the anchor residues bound to the anchor site in a rational conformation; and c) the docking score of the complex was greater than 8.0. A total of 25 potential T cell epitope peptides were predicted (eight for B4, six for B12, two for B15, and seven for B19). Nine peptides, which marked in bold in table 1, were selected to test their T cell responses in vitro. The binding energies of those complexes were also calculated. The results are shown in Table 1 . Peptide stimulation experiments were conducted using splenic lymphocytes derived from NP DNA vaccine-immunized chickens to verify some of the predicted potential T cell epitopes. To assess the immune effects of the vaccine, serum NP antibody was detected using the ELISA method. Compared to the control group, a significant increase (P,0.01) in blood NP antibody was observed in the immunized group two weeks after the first vaccination. After the boost, the blood NP antibody level of the immunized group increased further and it remained at a high level throughout the duration of the experiment (Fig. 9 ). This indicated that the immune systems of the chickens were activated by the vaccine immunization and that the splenic lymphocytes of chickens were sensitized. Activation of Lymphocytes Using Peptides NP To verify the predicted T cell epitopes in NP, ten synthetic peptides (Top nine of the 25 predicted peptides from NP and one unrelated peptide) were incubated with sensitized splenic lymphocytes for 5 d. Flow cytometry analysis showed that the proliferation of CD8 + T lymphocytes increased by 13.7% and 11.9% in cells stimulated with the peptides NP 89-97 and NP 198-206, respectively (Fig. 10) . Chicken IFN-c concentration in cells stimulated using peptides NP 89-97 and NP 198-206 were significantly higher than the control and unrelated peptide-stimulated cells (Fig. 11) . These results demonstrate that the peptides NP 89-97 and NP 198-206 are NP T cell epitopes in chickens of certain haplotypes. The objective of this study was to predict and verify T cell epitopes in the H5N1 AIV NP in chicken. Using a motif combined with a structure-based method, 25 potential T cell epitope peptides were predicted in the H5N1 AIV NP in chickens of B4, B12, B15, and B19 haplotypes. NP 89-97 and NP 198-206 were found to induce a significant proliferation of CD8 + T lymphocytes and they increased the secretion of chicken IFN-c in sensitized splenic lymphocytes. These data suggest that peptides NP 89-97 and NP 198-206 are NP T cell epitopes in chickens of certain haplotypes. This study is important for the following two reasons. First, this is the first study to determine the structural characteristics of the peptide-binding domains of chicken MHC class I molecules belonging to the B4, B12, B15, and B19 haplotypes using a combined motif-structure method to predict T cell epitopes in chickens. Second, NP 89-97 and NP 198-206 are the first two T cell epitopes to be identified in AIV NP in chickens of certain haplotypes. Homology modelling is widely used in many areas of structure-based analysis and study [37, 38] . However, there are few studies of chicken MHC class I molecules compared with those of human or mouse. The only chicken MHC class I molecule structure that has been solved is the B21 haplotype [18] . Thus, there is little information available on the structure and function of chicken MHC class I molecules. Therefore, the current study used homology modelling to investigate the peptide-binding domains of chicken MHC class I molecules belonging to the B4, B12, B15 and B19 haplotypes. To the best of our knowledge, this is the first attempt to understand the peptide-binding properties of these molecules based on their structures. The assessment indicated that the models were of high quality in terms of their folding and they were suitable for structure-based T cell epitope prediction. The structural characteristics of the peptide-binding properties of these MHC class I molecules were described in the results section. Only four haplotypes with known motifs were selected in this study. However, it is possible to use this solution to predict T cell epitopes of MHC class I molecules belonging to different haplotypes for other antigens in chickens, because the only difference would be the increased number of peptides required for molecular docking. This study found that 25 out of 75 peptides were potential T cell epitopes in the H5 AIV NP in chickens of the four haplotypes. An analysis of previous results showed that some of those peptides have been identified as T cell epitopes in humans. There is evidence that NP 91-99 (KTGGPIYKR), NP 361-375 (RGVQIAS-NENMETME), and NP 174-184 (RRSGAAGAAVK), which are derived from the H3N2, H3N2, and H1N1, respectively, are T cell epitopes, with MHC restriction alleles of HLA-A68, H-2Db, and HLA-B27, respectively [39] . Of these peptides, NP 89-97 (PKKTGGPIY) and NP 362-369 (GVQIASNE) are completely conserved in all influenza virus strains, while NP 179-186 (AGAAVKGV) is relatively conserved. Some potential epitopes were also shared by several haplotypes. Chickens used for meat and egg production are always heterozygotes, so the identification of these shared T cell epitopes will facilitate the development of broad-spectrum protective vaccines for chickens. Furthermore, epitopes shared by birds and humans are of great importance in the design of rational vaccines for protecting humans and birds from AIV infection. This study used a DNA vaccine plasmid expressing the NP from A/Goose/Gong Dong/1/96 (H5N1). A DNA vaccine expressing HA from A/Goose/Gong Dong/1/96 (H5N1) has been found to elicit antibody responses and protect chickens against challenge with HPAI virus [25] . Methods used for producing the HA-expression DNA vaccine were adopted when preparing the DNA expression plasmid pCAGG-NP. NP antibody responses were detected in immunized chickens, suggesting that this vaccine elicits successful immune responses to the NP antigen in chickens. The T cell responses were not detected directly in this study, but it was inferred that a successful antibody response would have been accompanied by a successful T cell response based on our knowledge of the DNA vaccine. Nine of the peptides were synthesized and used to stimulate sensitized splenic lymphocytes to verify that they were epitopes. Increases in the proliferation of CD8 + T lymphocytes and the secretion of chicken IFN-c demonstrated the antigenicity of these peptides. NP 89-97 (PKKTGGPIY) and NP 198-206 (KRGINDRNF) induced significant T cell responses in splenic lymphocytes. An analysis of the prediction results showed that NP 89-97 was predicted to be a T cell epitope for both B15 and B19, while NP 198-206 also belonged to the B19 haplotype. Thus, it is suggested that the major haplotype of the experimental SPF chickens might be B19. NP 89-97 also overlapped with a previously identified HLA-A68 restriction T cell epitope NP 91-99 (KTGGPIYKR) of the H3N2 AIV [7] . This further verifies the significance of this epitope, which could be used in human and chicken vaccines to provide protection against different influenza virus subtypes. Using in silico and in vitro approaches, this study identified two novel T cell epitopes NP 89-97 (PKKTGGPIY) and NP [198] [199] [200] [201] [202] [203] [204] [205] [206] (KRGINDRNF) in the H5N1 AIV NP in chickens of certain haplotypes. The method used in this investigation is applicable to predicting T cell epitopes for other antigens in chicken, while this study also extends our understanding of the mechanisms of the immune response to AIV in chickens.

Novel Paramyxoviruses in Free-Ranging European Bats

The zoonotic potential of paramyxoviruses is particularly demonstrated by their broad host range like the highly pathogenic Hendra and Nipah viruses originating from bats. But while so far all bat-borne paramyxoviruses have been identified in fruit bats across Africa, Australia, South America, and Asia, we describe the detection and characterization of the first paramyxoviruses in free-ranging European bats. Moreover, we examined the possible impact of paramyxovirus infection on individual animals by comparing histo-pathological findings and virological results. Organs from deceased insectivorous bats of various species were sampled in Germany and tested for paramyxovirus RNA in parallel to a histo-pathological examination. Nucleic acids of three novel paramyxoviruses were detected, two viruses in phylogenetic relationship to the recently proposed genus Jeilongvirus and one closely related to the genus Rubulavirus. Two infected animals revealed subclinical pathological changes within their kidneys, suggestive of a similar pathogenesis as the one described in fruit bats experimentally infected with Hendra virus. Our findings indicate the presence of bat-born paramyxoviruses in geographic areas free of fruit bat species and therefore emphasize a possible virus–host co-evolution in European bats. Since these novel viruses are related to the very distinct genera Rubulavirus and Jeilongvirus, a similarly broad genetic diversity among paramyxoviruses in other Microchiroptera compared to Megachiroptera can be assumed. Given that the infected bats were either found in close proximity to heavily populated human habitation or areas of intensive agricultural use, a potential risk of the emergence of zoonotic paramyxoviruses in Europe needs to be considered.

Members of the virus family Paramyxoviridae are divided into two subfamilies, Paramyxovirinae and Pneumovirinae, comprising a vast variety of animal-and human-pathogenic viruses [1] . Within the subfamily Paramyxovirinae, five genera have been classified, Respiro-, Morbilli-, Rubula-, Avula-, and Henipavirus, as well as a fastgrowing group of unclassified viruses. The increased molecular characterization of recently isolated paramyxoviruses indicates a much greater genetic diversity within the subfamily Paramyxovirinae than previously assumed. Furthermore, the detection of highly human-pathogenic paramyxoviruses has also influenced the attention drawn to paramyxovirus research and to the isolation of further novel paramyxoviruses from hosts that are suggested as likely species to transmit newly emerging viruses. Bats are among this highly suspected group of animals [2] . They belong to the most successful and diverse mammals on earth and comprise approximately 1,200 chiropteran species distributed worldwide. In the last two decades important zoonotic viruses including Ebola, Marburg, and SARS virus, but also paramyxoviruses such as Hendra and Nipah virus have been identified in various Pteropus spp. (flying foxes) fruit bats [3] [4] [5] [6] [7] [8] [9] [10] [11] . For Menangle virus, another paramyxovirus isolated from fruit bats, less pathogenic courses of disease in humans have been described [12] . For other bat paramyxoviruses isolated, infections in humans have yet to be associated, e.g. Tioman virus from flying fox [13] , bat parainfluenza virus from flying fox [14] , Tuhoko virus from flying fox [6] , Mapuera virus from non-pteropid fruit bat [15] , and Henipalike viruses also from non-pteropid fruit bat [4] . All viruses of the family Paramyxoviridae so far detected in bat species have been identified in fruit bats across Africa, Australia, South America, Asia, and Madagascar [3] [4] [5] [6] 10] . Only a few studies attempting the isolation of paramyxoviruses in bats concerned insectivorous bat species, and all of them turned out with negative results [16, 17] . The only indication of paramyxoviruses in this group of bat species was the detection of Nipah virus antibodies in lesser Asiatic yellow bats (Scotophilus kuhlii) [16] . The present study aimed to detect and isolate novel paramyxoviruses in free-ranging European insectivorous bats and to estimate a possible impact of paramyxovirus infection on infected individual animals by comparing histo-pathological findings and virological results. As part of a study to investigate diseases in free-ranging bats in Germany [18] , 120 deceased bats from 2009 of 15 different European vespertilionid species (Eptesicus nilssoni, E. serotinus, Myotis bechsteini, M. daubentonii, M. mystacinus, M. nattereri, Nyctalus leisleri, N. noctula, Pipistrellus kuhli, P. nathusii, P. pipistrellus, P. pygmaeus, Plecotus auritas, P. austriacus, Vespertilio murinus) were examined. The bat carcasses originated from 4 different geographic regions in Germany, i.e. Berlin greater metropolitan area (n = 83), Bavaria (n = 30), Brandenburg (n = 5), and Baden-Wuerttemberg (n = 2). Bat carcasses were stored at -20uC for transportation before performing a full necropsy. For histo-pathological examination, a small piece of tissue from all organs was fixed in buffered 4% formalin, processed routinely and embedded in liquid paraffin. Paraffin blocks were cut at 2-5 mm thickness and stained with hematoxylin-eosin [19] . Immuno-histochemistry was performed on all organs of PCR positive bats using rabbit immune sera against Beilong virus, J-virus, Menangle virus, Tioman virus and Nipah virus as described previously [20] . Samples of lung, liver, heart, and kidney, and conspicuous tissues (e.g. enlarged spleen) from each bat were homogenized in buffer and transferred to RNAlater (1:1). Pooled organ tissue from each bat was used for RNA/DNA extraction (PureLink TM Viral RNA/DNA Mini Kit, Invitrogen, Germany) and further cDNA synthesis according to the manufacturer's instructions (TaqManH Reverse Transcription Reagents, Applied Biosystems, Germany). Broadly reactive paramyxovirusspecific RT-PCR assays were applied [21] , yielding amplicons of 538 base pairs (PAR primers) and 486 base pairs (RES-MOR-HEN primers) located across domains I and II of the RNA polymerase (L)-coding sequence, a region of the genome suitable for phylogenetic analyses [22] . Since cDNA was readily available, PCR conditions were modified using the optimization method by Taguchi [23] . For this, based on the use of orthogonal arrays representing individual reactions with components at different concentration levels, a minimal number of experiments is allowed. To increase PCR sensitivity, the product yield for each reaction is used to calculate the optimal concentration of each reaction component. By using this method, a novel PCR reaction mixture was determined and henceforth used. For first-round PCR in the seminested assay, PCR mixtures contained 3 pmol each of forward and reverse primers, 16PlatinumH Taq buffer (Invitrogen), 250 nmol MgCl 2 (Invitrogen), 2.5 pmol desoxynucleoside triphosphates (Invitrogen), 2 ml of cDNA, and 1.25 U of PlatinumH Taq polymerase (Invitrogen). Water was then added to a final volume of 25 ml. The PCR mixture was sequentially incubated at 94uC for 2 min for denaturation, and then 40 cycles at 94uC for 15 s, 50uC for 30 s, 72uC for 30 s, and a final extension at 72uC for 7 min. For the second amplification in the seminested PCR assay, 16PlatinumH Taq buffer, 25 nmol MgCl 2 , 2.5 pmol desoxynucleoside triphosphates, 3 pmol each of forward and reverse primers, 1.25 U PlatinumH Taq, 1 ml PCR product from the first reaction, adding water to a final volume of 25 ml. The cycling conditions were identical to the ones of the first round. PCR products were run on an 1.5% agarose gel containing ethidium bromide. Images were captured on E.A.S.Y. RH-3 gel documentation system (Herolab, Germany). Amplicons from the PCR reaction were purified using the MSBH Spin PCRapace kit (Invitek, Germany). Both strands of the amplicons were sequenced with a BigDye Terminator v 3.1 Cycle Sequencing kit on an ABI Genetic Analyzer 3500 6l D6 automated sequencer (Applied Biosystems, Germany) using the corresponding PCR primers. Remaining reaction conditions were performed in accordance to the manufacturer's protocol. On the basis of newly acquired sequence information, specific qPCR assays were designed (Table 1) to screen pooled organ tissues of all 120 bats. Cycler conditions for all qPCR assays were as follows: predenaturation (95uC for 10 min), 45 amplification cycles (95uC for 30 s, 60uC for 30 s, 72uC for 30 s), and final extension (72uC for 10 min). Additional primers were designed using conserved regions between Jeilongviruses and Henipaviruses to extend the sequence obtained by PAR primers (primers and protocol are available on request). Bayesian reconstruction of phylogenetic trees was performed in concordance with the current proposals of Paramyxoviridae taxonomy using MrBayes, version 3.1.2 [24, 25] . The underlying alignment by ClustalW was based on a 529 base pair fragment (PAR Primer) and 1,593 base pairs amplicon (long fragment) from PCR reactions. The evolutionary history was inferred using the bayesian MCMC method. First, a model selection for these calculations was performed with jModelTest [26] and model GTR+I+G (invariable sites, gamma distribution) was selected for the PAR and the RES-MOR-HEN fragment, and GTR+G to study the long fragment alignment. The calculation parameters were as follows: number of runs: four, number of generations: 1,000,000, sample frequency: 100, burn in: 25%. The results were finally visualized by the FigTree v1.2.1 program, a graphical viewer of phylogenetic trees. Based on the GTR substitution model the estimated transversion ratio, proportion of invariable sites and gamma distribution parameters were estimated automatically. For confirmation of virus isolation and determination of the infected organs, RNA/DNA extraction and PCR analysis including sequencing was performed on all individual organs from infected bats. For two isolates, a second RT-PCR with primers RES-MOR-HEN [21] and the same PCR conditions as described above was conducted to acquire fragments comparable to previously isolated novel Henipa-like viruses from African fruit bats [4] . Bat species confirmation was achieved by sequencing and analyzing the mitochondrial DNA as described [27] . The modified PCR protocol (PAR primers) resulted in a 10fold increase of sensitivity compared to the published protocol which was applied as a two-step PCR (Figure 1 ). With this optimization, three out of 120 pooled samples were PCR positive for paramyxoviruses using PAR primers ( Table 2 ). The identified viruses were termed after the infected bat species: BatPV/ Myo.mys/E20/09 (Accession number JN086950), BatPV/Pip.pip/E95/09 (Accession number JN086951), and BatPV/Nyc.noc/E155/09 (Accession number JN086952). Fragments of 529 bp length were aligned with homologous fragments of the partial polymerase gene of other members of the family Paramyxoviridae from GenBank ( Figure 2 ). Phylogenetic analysis confirmed three distinct isolates within the subfamily Paramyxovirinae. BatPV/Nyc.noc/E155/09 was in basal association to other members of the genus Rubulavirus. For both BatPV/ Myo.mys/E20/09 and BatPV/Pip.pip/E95/09, the closest association was observed to J-virus and Beilong virus (unclassified viruses) [28, 29] . A longer sequence of 1,593 bp was generated for BatPV/Pip.pip/E95/09 and used for an extended phylogenetic analysis ( Figure 3 ). The highest similarity was revealed for BatPV/Myo.mys/E20/09 to J-Virus with 66.4%, for BatPV/ Pip.pip/E95/09 also to J-Virus with 64.1%, and for BatPV/ Nyc.noc/E155/09 to Rubulavirus with 62.1% (Table 3) . Within the subfamily Paramyxovirinae the extent of minimal nucleotide homology for the partial polymerase gene between different viruses in the same genus ranges from 64.1% (Rubulavirus) to 76.8% (Henipavirus), whereas the extent of nucleotide similarity between viruses from different genera is between 40.3% (Morbillivirus) and 58.1% (Henipavirus). Analysing the nucleotide homology of the partial polymerase gene of the new insectivorous bat paramyxoviruses, no definite correlation to one of the other paramyxovirus genera could be obtained. The comparison of sequences of BatPV/Myo.mys/E20/09 (Accession number JN086953) and BatPV/Pip.pip/E95/09 (Accession number JN086954), obtained from the PCR assay with RES-MOR-HEN primers (Table 2) , confirmed the results of the above-mentioned phylogenetic analysis (data not shown). The first paramyxovirus termed BatPV/Myo.mys/E20/09 was detected in pooled organs and was subsequently confirmed in the kidney only of one adult male whiskered bat (Myotis mystacinus) found in Bavaria. Histological examination of the internal organs revealed multifocal mild interstitial nephritis with lymphoplasmacytic infiltrates and occasional neutrophiles. Lungs had mild nonsuppurative interstitial pneumonia and marked leucocytostasis in most blood vessels. Additionally, there was distinct activation of the lymphoreticular tissue of the spleen with moderate follicular hyperplasia and sparse irregularly distributed small foci of lymphocytes and plasma cell aggregations within the liver. The second virus (BatPV/Pip.pip/E95/09) was detected in the pooled organs of an adult female common pipistrelle bat (Pipistrellus pipistrellus) also found in Bavaria. No specific infected organ could subsequently be determined due to sample size limitations. Histologically the animal had multifocal moderate interstitial nephritis with segmental infiltrates of lymphocytes, plasma cells, and occasional single neutrophiles (Figure 4 ). There was mild generalized interstitial pneumonia and moderate follicular hyperplasia of the spleen. There were mild intrasinusoidal infiltrates of neutrophiles, lymphocytes, and plasma cells within the liver. The third virus (BatPV/Nyc.noc/E155/09) was detected in pooled organs and was subsequently confirmed in the lung of only one adult female noctule bat (Nyctalus noctula) found in Berlin. Histologically the bat revealed marked follicular hyperplasia of the spleen without further inflammatory organ lesions. The lung was severely congested, and oedematous fluid was present in the lung parenchyma. Using immune sera against Beilong virus, J-virus, Menangle virus, Tioman virus and Nipah virus in immuno-histochemistry, no stained antigens were visualized in any of the paramyxoviruspositive bats although all immune sera worked well against their homologous virus in corresponding positive controls. After screening all 120 bats of 15 species with three virusspecific qPCR assays, an identical paramyxovirus to BatPV/ Myo.mys/E20/09 was detected in the spleen of one additional Myotis mystacinus (E120/09). During the past decade, bats have increasingly been recognized as members of the animal group with the highest relative risk to harbour novel emerging zoonotic pathogens [2] . The emergence of Hendra and Nipah virus provided the first evidence of a zoonotic paramyxovirus originating from bats with a broad host range including humans. Interestingly, despite the enormously diverse chiropteran animal order, so far, with the exception of rabies, only fruit bats have been implicated as a reservoir of a number of new and emerging zoonotic viruses [3] [4] [5] [6] [7] [8] [9] [10] . With this study, we were able to describe the detection and characterization of the first three paramyxoviruses in insectivorous bats. The genetic distance between these three novel paramyxoviruses and the closest related member known is higher than that of members within other paramyxovirus genera, suggesting that all three viruses might be considered as unassigned paramyxoviruses. Thus the two viruses BatPV/Myo.mys/E20/09 and BatPV/ Figure 1 . Improved detection sensitivity after Taguchi optimization of the Paramyxovirinae subfamily-specific PCR [21] . Gel electrophoresis of amplification products of the second round seminested PCR using 10-fold serial dilutions (10 0 to 10 24 ) of a cDNAsample (kidney of sample E20/09). (A) PCR protocol adopted for twostep PCR as previously published [21] using the pan-PAR-F1/PAR-R primer pair (1st run) and the pan-PAR-F2/PAR-R primer pair (2nd run). (B) Optimized protocol using the pan-PAR-F1/PAR-R primer pair (1st run) and the pan-PAR-F2/PAR-R primer pair (2nd run). doi:10.1371/journal.pone.0038688.g001 Figure 2 . Phylogenetic analysis of the partial L-gene sequence obtained from PCR fragments after Pan-Paramyxovirinae-PCR with PAR primers (529 bp) [21] . The revealed gap-free alignment was used to generate a phylogenetic tree of the novel bat paramyxoviruses (red) concordant with representatives from all known genera of paramyxoviruses with MrBayes. Posterior probability rates are given next to the tree nodes. PLoS ONE | www.plosone.org Pip.pip/E95/09 might even be considered as members of a new putative genus, as they contain an amino acid identity of 79.5% of the partial L-gene, the highest conserved region of the paramyxovirus genome. Further precise genetic analyses will have to prove whether they ought to be integrated into the proposed new genus Jeilongvirus [29] , comprising J-virus and Beilong virus, with amino acid identities between 69.9% and 74%, respectively, although their viral antigens in PCR positive organs are not immunologically cross-reactive. The third novel paramyxovirus, BatPV/Nyc.noc/E155/09, has a basal association with the genus Rubulavirus and could therefore become a member of this genus, although nucleic acid identity was slightly lower than that between already classified members within this genus and no crossreactivity of viral antigens with immune sera of closely related rubulaviruses was obtained. Besides virus detection, our study allowed a direct correlation of virology and histo-pathology results. In previous studies in which bats were examined for paramyxovirus infections, no overt clinical disease was noted [5, 30] , despite occasional high prevalences of antibodies against various paramyxoviruses (e.g. Hendra and Nipah virus) and the detection of paramyxovirus RNA in different bat organs [5, [30] [31] [32] . For Hendra virus infections of pteropid bats, the subclinical course has been confirmed by experimental infection. Kidneys are the only site of pathological lesions after Hendra virus infection of pteropid bats, with mild interstitial perivascular infiltrates by mononuclear inflammatory cells, while virus nucleic acids were also detected in lung, spleen, gastrointestinal tract, and urine [33] . In contrast, Nipah virus was only detected in kidneys (male) and uteri (female) of pteropid bats after experimental infection [34] . In all Hendra virus and Nipah virus experimental infection studies the amount of virus recovered reached the limit of detection level, a similar situation encountered in our study. In infected insectivorous bats, low band intensities of PCR products of organs tested positive for paramyxovirus infection indicated a low virus load. Likewise, the kidneys were the only organ infected in the male whiskered bat (BatPV/ Myo.mys/E20/09) and presumably in the female common pipistrelle. Interestingly, both animals had mononuclear inflammatory interstitial infiltrates similar to the reported experimental Hendra virus infections. Unfortunately, attempts to prove the evidence by specific immunohistochemistry with antibodies directed against Beilong virus, J-virus, Menangle virus, Tioman virus and Nipah virus were not successful. A number of reasons could account for this result. Either the noted inflammatory changes GenBank Accession numbers of novel paramyxoviruses PAR fragment: JN086950 (BatPV/Myo.mys/E20/09), JN086951 (BatPV/Pip.pip/E95/09), JN086952 (BatPV/Nyc.noc/E155/09). RSV = respiratory syncytial virus. doi:10.1371/journal.pone.0038688.g002 were indeed unrelated to paramyxovirus infection or the immunogenic epitopes of paramyxoviruses in European insectivorous bats differ significantly from their australo-asian relatives, hence prohibiting bonding between the reagents or, taking into account that molecular investigations indicated a low virus load, the number of infectious particles was too low to be picked up by immunohistochemistry. Although unequivocal evidence of the causative association between paramyxovirus infection and renal inflammation remains open it is important to note, that nephritis is a rare finding in insectivorous bats. Out of 500 examined deceased bats only 3% had inflammatory changes within their kidneys, while 20% of these respective cases were clearly associated to bacterial disease [19] . Considering this background information together with findings from experimental paramyovirus infections in fruit bats a possible link between the renal mononuclear infiltrates and the detected novel paramyxovirus nucleic acids seems feasible. With this, a possible transmission route via urine like in Hendra virus infection could also be assumed for these viruses. In contrast, the detection of BatPV/Nyc.noc/E155/09 limited to the lungs of the female noctule bat is suggestive of an oronasal and/or salivary route of transmission. Our findings confirm a promising approach for the ultra-sensitive detection of paramyxoviruses by applying a modified PCR protocol as a powerful tool, including non-invasive sampling (oral swabs, urine, faeces). However, it should be emphasized that substantial insights regarding the estimation of possible spill-over events can only be achived by a combination of virology, histo-pathology, and bat ecology investigations. Emerging paramyxoviruses from fruit bats in spill-over hosts have regularly been associated with ecosystem and land-use changes resulting in an increased overlap of bats, domestic animals, and human ecologies and thereby increased opportunities for bat-borne zoonotic diseases to emerge [35] . As demonstrated in this study, paramyxoviruses basally related to Henipaviruses also exist in geographic areas distant to the distribution range of fruit bats, the suspected natural hosts for Henipaviruses, indicating a possible virus-host co-evolution beyond this animal group. Since infected bats were found in close proximity to heavily populated human habitations as well as intensive agricultural use, a potential risk for the emergence of zoonotic paramyxoviruses in Europe should be further elucidated. Although the three novel paramyxoviruses detected in three distinct European bat species cannot be readily assigned to any previously described paramyxovirus genus, they are associated to two very distinct genera (Rubulavirus and Jeilongvirus) [22] , indicating a similarly broad genetic diversity among paramyxoviruses in insectivorous bats compared to fruit bats. Given the much larger diversity amongst insectivorous bats with over 1,000 bat species in contrast to 186 fruit bat species [36] , we predict a far higher diversity of paramyxoviruses in insectivorous bats. Extensive phylogenetic studies of insectivorous bat-born paramyxoviruses will provide further insight into the suggested co-evolution of paramyxoviruses and bats [35] . In addition to Africa, Australia, South America, and Asia, the detection of novel paramyxoviruses in European bats extends the possible geographic overlap with other susceptible spill-over hosts. Is there a possibility for paramyxoviruses of insectivorous bats to emerge as zoonotic pathogens? Before any answer to this question can be attempted, further research on paramyxovirus diversity and distribution combined with the understanding of dynamics of pathogen cycles within bat populations will be needed as well as investigations into pathogenicity factors of these viruses, like receptors for host invasion. Particularly as so far transmission of bat related paramyxoviruses did not occur directly between bats and humans but depended on a secondary host species like horses or pigs.

Lipid Rafts and Alzheimer’s Disease: Protein-Lipid Interactions and Perturbation of Signaling

Lipid rafts are membrane domains, more ordered than the bulk membrane and enriched in cholesterol and sphingolipids. They represent a platform for protein-lipid and protein–protein interactions and for cellular signaling events. In addition to their normal functions, including membrane trafficking, ligand binding (including viruses), axonal development and maintenance of synaptic integrity, rafts have also been implicated in the pathogenesis of several neurodegenerative diseases including Alzheimer’s disease (AD). Lipid rafts promote interaction of the amyloid precursor protein (APP) with the secretase (BACE-1) responsible for generation of the amyloid β peptide, Aβ. Rafts also regulate cholinergic signaling as well as acetylcholinesterase and Aβ interaction. In addition, such major lipid raft components as cholesterol and GM1 ganglioside have been directly implicated in pathogenesis of the disease. Perturbation of lipid raft integrity can also affect various signaling pathways leading to cellular death and AD. In this review, we discuss modulation of APP cleavage by lipid rafts and their components, while also looking at more recent findings on the role of lipid rafts in signaling events.

It has traditionally been difficult to reach any kind of consensus concerning the definition of lipid rafts. However, in 2006 at the Keystone Symposium on Lipid Rafts and Cell Function in Colorado it was agreed that "membrane rafts are small (10-200 nm), heterogeneous, highly dynamic, sterol, and sphingolipid-enriched domains that compartmentalize cellular processes. Small rafts can sometimes be stabilized to form larger platforms through proteinprotein and protein-lipid interactions" (Pike, 2006) . The long, saturated acyl chains of sphingolipids allow tight packing hence their juxtaposition with the kinked, unsaturated acyl chains of bulk membrane phospholipids leads to phase separation (Brown and London, 2000) . The cholesterol molecules can act as "spacers," filling any gaps in sphingolipid packing (Simons and Ikonen, 1997) . The notion of lipid rafts, while not new, has never been far from controversy (Pike, 2009) , their existence frequently questioned (Munro, 2003) . It is almost 30 years since Karnovsky et al. (1982) suggested that protein diffusion in membranes is not free, but somehow constrained. They went on to detail the now familiar concept of the organization of lipids in domains, which may have functional significance. The differential detergent (Triton X-100) solubility of glycosylphosphatidylinositol (GPI)-anchored proteins and transmembrane proteins was first shown by Hooper and Turner (1988) . Later work showed the importance of lipid rafts in protein sorting and segregation, with GPI anchored proteins being preferentially localized in lipid rafts (Brown and Rose, 1992; Brown and London, 2000; Pike, 2009 ). Other lipid modifications of proteins have also been described, such as palmitoylation and myristoylation which may influence raft localization (Brown and London, 2000; Smotrys and Linder, 2004; Pike, 2009 ). In describing membrane lipid clusters as moving platforms, or rafts, which are enriched in both sphingolipids and cholesterol, perhaps the most important finding was that proteins could be segregated being selectively included or excluded from the rafts (Simons and Ikonen, 1997) . In this way, raft localization can serve to facilitate or obstruct protein interactions (Brown and London, 2000; Lingwood and Simons, 2009) or act as a protein scaffold while allowing diffusion (Maxfield and Tabas, 2005) . Despite significant efforts in developing methodology and techniques for lipid raft research, there are still some shortcomings in precise determination of their size, structure, and composition which were recently critically discussed by Simons et al. (2010) . The most important issue which still induces debate is related to application of detergents for isolation of lipid rafts and methylβ-cyclodextrin for extraction of cholesterol from cell membranes which could lead to formation of artificial complexes not existing in the natural environment. It has now been established that detergent-resistant membranes (DRMs) are not identical to lipid rafts and the data on proteins detected within DRMs from various types of cells should be treated with caution and not always considered to be functionally raft localized (Lichtenberg et al., 2005; Chen et al., 2009a) . Development of a new technique that purifies nano-meso scale DRMs at 37˚C in an ionic buffer that preserves the lamellar phase of the metastable inner leaflet lipids could be a significant step toward purifying individual physiologically relevant rafts (Morris et al., 2011) . Although Triton X-100 is still the standard detergent in preparation of lipid rafts some groups claim that use of Brij 96 produces more reliable results although comparing properties of both detergents in the same set of experiments www.frontiersin.org have not produced significantly different results. Moreover it has been suggested that preparation of the rafts in a buffer mimicking the cytoplasmic environment will preserve the structure of the rafts under physiological conditions by retaining the proteins associated with the intracellular part of the membrane (Chen et al., 2009b) . Another source of controversy in lipid raft research which still has to be resolved is determined by the lack of suitable detection techniques in living cells. This resulted in estimation of the putative size of lipid rafts in the range of 5-700 nm. In the last decade a number of new techniques assisting in estimation of raft size and helping in their visualization has been developed and applied (for review see Simons et al., 2010) . They include single molecule spectroscopy and super-resolution microscopy such as fluorescence recovery after photobleaching (FRAP), stimulated emission depletion (STED), Förster resonance energy transfer (FRET), total internal reflection fluorescence (TIRF), and fluorescence correlation spectroscopy (FCS) techniques. However, even using these techniques the subwavelength lipid domains have never been directly visualized but their size was predicted to be <20 nm (Eggeling et al., 2009) . Despite the limitations these new techniques confirmed the existence of nanoscale cholesterol-based assemblies of lipids and proteins in the membranes of living cells and allowed to characterize further their dynamics and properties. When comparing the data on the size and precise protein and lipid composition of lipid rafts it is important to bear in mind that they are determined not only by the type of detergent used, the conditions of the experiments and resolution of techniques applied, but also by the type of tissue and cells used for their isolation and visualization. This has to be taken into account when comparing the data and making conclusions about the physiological relevance of proteins detected in lipid rafts. From this point of view a systematic meta-analysis of existing data considering various experimental conditions, tissue specificity, and resolution of the techniques applied might provide a useful tool for understanding raft heterogeneity and for optimizing further research into this intriguing subject. The best-characterized raft proteins include lipid-modified proteins containing saturated acyl chains, such as GPI-anchored proteins, and doubly acylated proteins, such as Src family kinases and the α subunits of heterotrimeric G proteins (Hooper, 1999; Liang et al., 2001; Oh et al., 2001) . Many other physiologically important proteins have been investigated for their possible raft localizations, including key proteins involved in Alzheimer's disease (AD), such as amyloid precursor protein (APP; Parkin et al., 1999) , β-site APP cleaving enzyme (BACE-1; Ehehalt et al., 2003; Kalvodova et al., 2005) , the γ-secretase complex (Hur et al., 2008) , a disentegrin and metalloprotease ADAM10 (Harris et al., 2009), acetylcholinesterase (AChE; Xie et al., 2010b) , angiotensinconverting enzyme (ACE; Parkin et al., 2003) , a ligand for the Notch receptor (Jagged1; Parr-Sturgess et al., 2010) , and most recently an amyloid β (Aβ)-degrading enzyme, neprilysin (NEP; Sato et al., 2012) . However the full list of raft-associated proteins is still far from completion. The progress in understanding structural and functional diversity of lipid rafts led to the concept of caveolae as a sub-type of lipid rafts, appearing as 50-80 nm pits in the plasma membrane. Although caveolae are sometimes considered to be synonymous with lipid rafts, it is now clear that they represent only a subset of rafts whose properties are defined by their major protein components belonging to the caveolin family and denoted Cav-1, Cav-2, and Cav-3 (Parton and Simons, 1995, 2007) . The characterization of PTRF (polymerase I and transcript release factor, originally identified as an RNA Pol I transcription factor, also called Cav-P60 and now termed cavin-1) and subsequently of other cavin family members as important constituents in caveolae formation has revealed new levels of complexity in the biogenesis of these plasma membrane invaginations (Briand et al., 2011) . Like the rafts themselves, the caveolae are enriched in cholesterol, glycosphingolipids, and SM. They are the site of several important protein-protein interactions, for example, the neurotrophin receptors, TrkA and p75(NTR), whose respective interactions with caveolin regulates neurotrophin signaling in the brain (Bilderback et al., 1999) . Caveolins also regulate G-proteins, MAPK, PI3K, and Src tyrosine kinases (Marin, 2011) . Another group of proteins which was suggested as markers of caveolae are flotillins (also known as reggies) which were discovered as proteins involved in nerve regeneration (Schulte et al., 1997) . The flotillins are palmitoylated and myristoylated proteins, anchored to the plasma membrane and are considered to be exclusively raft localized (Schneider et al., 2008) . Flotillins belong to a larger class of integral membrane proteins that have an evolutionarily conserved domain called the prohibitin homology domain (PHB) which determines the affinity of proteins carrying it to the lipid rafts (Morrow and Parton, 2005) . In the context of this review it is important to mention another cholesterol and ganglioside enriched membrane domain, which share some similarity but are physically and functionally distinct from the lipid rafts, the tetraspanin-enriched microdomains (TEMs). Tetraspanins are a large family of small membranespanning proteins (Yanez-Mo et al., 2011) , numbering at least 32 family members in mammals. They are involved in a whole range of cellular processes, from cell morphology and motility to signaling pathways (Hemler, 2005) . In general, lipid rafts can be considered as signaling platforms that bring together various ingredients of the biological membranes determining specificity of the cells and their functioning. They include receptors, channels, recognition molecules, coupling factors and enzymes, facilitating their interaction and supporting signaling. Lipid rafts have been implicated in a plethora of both physiological and pathological processes. Among beneficial processes are axonal growth and branching (Kamiguchi, 2006; Grider et al., 2009; Munderloh et al., 2009 ) and hence raft disruption impedes axonogenesis (Petro and Schengrund, 2009) . Rafts are also involved in the stabilization of synapses (Willmann et al., 2006) . Raft components are also involved in cholera toxin entry via GM1 ganglioside (Holmgren et al., 1973) , HIV-1 entry (gp120), and conversion of prion protein (PrP c ) to its infectious form (PrP Sc ; Fantini et al., 2002; Vieira et al., 2010) . Lipid rafts play a critical role in entry, replication, assembly, and budding of various types of viruses (Suzuki and Suzuki, 2006) . In a more general sense, lipid rafts have been suggested to be involved in cardiovascular disease, carcinogenesis, and immune system diseases (Michel and Bakovic, 2007) . However, this review will focus on the role of lipid rafts in the pathogenesis of AD, a condition with which lipid rafts have been demonstrated to have multifarious links (Cordy et al., 2006; Cheng et al., 2007; Vetrivel and Thinakaran, 2010) . In the last two decades the amyloid cascade hypothesis of AD (Hardy and Higgins, 1992) has prevailed leading to accumulation of a significant amount of data on the molecular basis of the disease which have recently been reviewed by those who originally proposed the concept (Selkoe, 2001; Hardy, 2009) . From the clinical point of view, the disease is characterized by global cognitive decline, associated with brain pathology involving accumulation of extracellular amyloid aggregates (also known as senile plaques) of amyloid β peptide (Aβ) and intracellular neurofibrillary tangles of hyper-phosphorylated tau protein (Bothwell and Giniger, 2000) . According to the amyloid cascade hypothesis, it is the Aβ which is principally responsible for many of the pathological features of the disease (Hardy and Higgins, 1992; Sakono and Zako, 2010) with Aβ oligomers representing the most toxic species (Haass and Selkoe, 2007; Walsh and Selkoe, 2007; Sakono and Zako, 2010) . Accumulation of amyloid plaques is accompanied by astrogliosis and microgliosis (Grilli et al., 2003) and the most affected brain areas are the neocortex and hippocampus (Sisodia and Gallagher, 1998) . Although there are strong genetic links, including APP and presenilin mutations (Bothwell and Giniger, 2000) , as well as the apolipoprotein ε4 allele (Saunders et al., 1993; Deane et al., 2008; Huang, 2010) , sporadic AD is the dominant form. From this point of view predominance of AD research based on the mechanisms of early onset disease versus the broader spectrum of the factors leading to the sporadic form might be one of the reasons for the failure of the majority of therapeutic trials and lack of any preventive measures 20 years since the amyloid hypothesis has been proposed. APP is a type I integral membrane protein. It exists in three isoforms (APP 695 , APP 751 , and APP 770 ), generated from differential splicing of exons 7 and 8 (Sandbrink et al., 1996) . Exon 7 is homologous to protease inhibitors of the Kunitz type (KPI domain), while exon 8 is related to the MRC OX-2 antigen in thymocytes (Kitaguchi et al., 1988; Sandbrink et al., 1996) . APP 695 lacks both KPI and OX-2 domains, while APP 751 only lacks the OX-2 domain (Henriques et al., 2007) . In terms of distribution, APP mRNA is expressed in almost every tissue, where only the isoform ratio differs (Araki et al., 1991) . It is APP 695 that predominates in neurons (Gralle and Ferreira, 2007) . There are two proteolytic pathways of APP processing (Figure 1) . Amyloidogenic processing involves sequential cleavage of APP by βand γ-secretases (for review see Zhang et al., 2012) . This process ultimately releases Aβ peptide, responsible in large part for the pathogenesis of AD and a soluble ectodomain, sAPPβ. The second, non-amyloidogenic, pathway involves α-secretase cleavage of APP. This cleavage occurs between Lys16 and Leu17, within the Aβ region (Allinson et al., 2003) and precludes formation of Aβ. There is also the release of the large, soluble ectodomain, the neuroprotective sAPPα, which is shed from the cell surface (Allinson et al., 2004) and has been found in the CSF (Palmert et al., 1989) . At the time when the terms were introduced, α-, β-, and γsecretases were activities known to cleave APP proteolytically, but not defined as specific enzymes performing the cleavage. It was later found that BACE-1 was responsible for β-secretase activity (Vassar et al., 1999) and that a complex of presenilin, nicastrin, Aph-1, and Pen-2 had γ-secretase activity Hooper, 2005) . The α-secretase cleavage was found to be mediated by zinc metalloproteases of the disintegrin and metalloprotease (ADAM) family, specifically ADAM10 and ADAM17 (also known as TNFα converting enzyme, TACE). ADAM10 was found to be the dominant enzyme of APP processing in SH-SY5Y cells, given a lesser effect of ADAM17 knockdown (Allinson et al., 2004) . Recent investigations have suggested that ADAM10 is the principal APP secretase in primary neurons, with ADAM17 playing more of an auxiliary role (Kuhn et al., 2010; Lichtenthaler, 2010) . Allinson et al. (2004) suggested a model whereby the group of metalloproteases would each contribute to greater or lesser extents to APP cleavage in different cells and under different conditions. The ADAMs have a wide number of substrates, including angiotensinconverting enzyme (ACE; Allinson et al., 2004) , ACE2 (Lambert et al., 2005) , the prion protein (Vincent et al., 2001) as well as each other ). The ADAMs have been reviewed in greater detail elsewhere (Allinson et al., 2003; Edwards et al., 2008; van Goor et al., 2009) . The idea that lipid rafts may somehow modulate APP cleavage and hence affect the progression of AD has been around for over 10 years and previously reviewed (Cordy et al., 2006; Cheng et al., 2007; Vetrivel and Thinakaran, 2010) . APP itself is not generally a raft protein, although a small proportion of APP is localized in lipid rafts (Parkin et al., 1999) . Regulation of APP raft localization has been suggested to involve an interaction between the C-terminus of APP and flotillin-1 (Chen et al., 2006) . A more recent study suggested that flotillin-2 may also act as a scaffolding protein, clustering APP in lipid rafts. Schneider et al. (2008) postulated that a transient interaction between APP and flotillin-2 may regulate endocytosis of APP, which is important for its processing. It has also been suggested that another raft component, cholesterol, has a role in binding APP promoting its raft localization (Beel et al., 2010) although it was earlier proposed that one of the physiological functions of APP and Aβ is to control cholesterol transport (Yao and Papadopoulos, 2002) . There is also evidence that the adaptor protein Disabled1 (Dab1) interacts with APP regulating its processing and that the glycoprotein reelin promotes interaction of APP and Dab1 and their localization to lipid rafts involving phosphorylation by Fyn kinase (Hoe et al., 2009; Minami et al., 2011) . APP trafficking to the lipid rafts was shown to be dependent on the low-density lipoprotein receptor-related protein (LRP) which also promotes BACE1-APP interaction (Yoon et al., 2007) . On the other hand it was also demonstrated that ApoER2, a member of the low density lipoprotein receptor (LDL-R), negatively affects APP internalization and its expression stimulates Aβ production by shifting the proportion of APP from the non-raft region to the raft membrane domains (Fuentealba et al., 2007) . www.frontiersin.org FIGURE 1 | Schematic representation of APP processing and role of its products in AD pathology. The proteolytic processing of the large, transmembrane, amyloid precursor protein (APP) occurs in two distinct amyloidogenic and non-amyloidogenic pathways. The amyloidogenic pathway involves the sequential cleavage of APP by an aspartic proteinase, β-secretase, which releases a soluble ectodomain (sAPPβ) and the C-terminal fragment CTF99. This, in turn, is cleaved by another aspartic proteinase, γ-secretase, generating the transcriptional regulator APP intracellular domain (AICD), and releasing the 39-42 amino acid amyloid-β peptide (Aβ). Due to its very high ability to aggregation, Aβ forms dimers, trimers, and higher level oligomers which are toxic to cells and cause neuronal death. Formation of amyloid plaques from Aβ aggregates in complex with other proteins is a hallmark of AD but is considered as a scavenging process. In the non-amyloidogenic pathway APP molecules are cleaved at the α-secretase site within the Aβ-domain releasing a soluble ectodomain sAPPα and the C-terminal fragment CTF83. Proteolytic cleavage of CTF83 by γ-secretase releases AICD and p3 fragment whose functions are still unknown. The AICD fragment produced in the amyloidogenic pathway binds to a stabilizing factor Fe65 and in a complex with other factors (the histone acetyl transferase, Tip60, and a Mediator complex subunit Med12) can act as transcription factor regulating expression of a variety of genes, including an Aβ-degrading enzyme neprilysin. This process was found to be specific to the neuronal APP 695 isoform. AICD produced in the non-amyloidogenic pathway and from other APP isoforms (APP 751 and APP 770 ) is most likely to be degraded (e.g., by some intracellular proteases, e.g., insulin-degrading enzyme). Soluble APP ectodomains, sAPPα, and sAPPβ, have been shown to have neuroprotective properties. In search of an explanation as to how APP could be cleaved in two distinct (amyloidogenic and non-amyloidogenic) pathways, it was suggested that APP was present in two cellular pools (Ehehalt et al., 2003) as schematically presented in Figure 2 . Amyloidogenic processing was suggested to be linked to lipid rafts as their integrity was critical for Aβ formation (Simons et al., 1998; Cordy et al., 2003; Ehehalt et al., 2003; Rushworth and Hooper, 2011) and Aβ production was also indicated to be raft-localized (Lee et al., 1998) . Furthermore, the precise protein/lipid composition of the rafts was shown to influence Aβ release (Lemkul and Bevan, 2011) and aggregation (Ikeda et al., 2011) . Tetraspanins were also shown to regulate APP processing (Yanez-Mo et al., 2011) and perturbation of tetraspanin enriched domains (TEMs) can affect cell signaling pathways (Hemler, 2005) . For example, the α-secretase ADAM10 is regulated by tetraspanins and in this way, tetraspanins can affect the non-amyloidogenic processing of APP (Arduise et al., 2008) . In addition to this, tetraspanins also regulate the amyloidogenic pathway of APP processing since γ-secretase was shown to be associated with TEMs (Wakabayashi et al., 2009 ). The lipid raft is shown as part of the plasma membrane. The phospholipid domain (light blue) is separate from the lipid raft. The latter is enriched in glycosphingolipids and sphingomyelin (red) on the exofacial leaflet and glycerolipids (e.g., phosphatidylserine and phosphatidylethanolamine; green) on the cytofacial leaflet. Cholesterol (black) is enriched in both leaflets. The acyl chains in lipid rafts are more able to pack together. APP (Aβ region in maroon) is localized in raft and non-raft fractions, but predominates outside rafts. The α-secretase is not raft-associated, while the βand γ-secretases predominate in rafts. The cell surface is shown for clarity, although β-cleavage predominantly occurs in endosomes. BACE-1 was found to be palmitoylated at three residues, which indicated possible raft localization and interaction with raftresident lipids (Kalvodova et al., 2005; Hattori et al., 2006) . When BACE-1 was targeted to lipid rafts via GPI-anchoring, production of Aβ was increased, indicating upregulation of amyloidogenic APP processing (Cordy et al., 2003) . A more recent study suggests that GPI-anchorage increases preferential cleavage at the β-site of APP, producing the full length Aβ, whereas wtBACE-1 cleaves APP at two sites (β and β' sites) producing full length and N-terminally truncated Aβ (Vetrivel et al., 2011) . Another study has reported the effects of GPI-anchorage of ADAM10, the principal α-secretase, and showed that no wtADAM10 was raft localized while all GPI-ADAM10 was in lipid rafts. It was associated with a reduction in Aβ as GPI-ADAM10 competed with BACE-1 for the APP substrate . In addition to BACE1, it was shown that the subunits of the γ-secretase complex are also enriched in the lipid rafts (Lee et al., 1998; Hur et al., 2008) and that S-palmitoylation plays a role in localization and stability of nicastrin and Aph-1 within the rafts although not affecting the γ-secretase processing of APP (Cheng et al., 2009) . While caveolin-1 was shown to be an important regulator of γ-secretase spatial distribution and activity (Kapoor et al., 2010) , the proteins of the γ-secretase complex, e.g., PS1, in turn, can induce lipid raft formation and decrease the membrane fluidity . A subset of proteins, in particular voltage-dependent anion channel 1 and contact in associated protein 1, are also associated with γ-secretase in lipid rafts and affect APP processing (Hur et al., 2012) . Another important lipid-raft associated protein which was shown to play an important role in APP processing is the GPIanchored prion protein (PrP; Naslavsky et al., 1997) . It was demonstrated that PrP c regulates APP processing by inhibiting BACE1 activity and that the effect of PrP c on the β-secretase cleavage of APP requires the localization of PrP c to cholesterol-rich lipid rafts and is mediated by the N-terminal polybasic region of PrP C via interaction with glycosaminoglycans (Parkin et al., 2007) . This interaction decreased BACE1 at the cell surface and in endosomes where it preferentially cleaves wild type APP but increased it in the Golgi where it preferentially cleaves APP with the Swedish mutation (APP Swe ). Although deletion of PrP c in transgenic mice expressing human mutated APP Swe , Ind had no effect on APP processing and Aβ levels, deletion of PrP c in HEK293 cells expression wild type reduced APP cleavage by BACE1. Because there was no effect of PrP c deletion on processing of APP Swe in HEK cells the authors suggested that PrP c may be a key protective player against sporadic Alzheimer disease versus its less common familiar form (Griffiths et al., 2011) . Because there is no detected decrease of PrP c content in the AD brain (Saijo et al., 2011) it is possible to suggest that age-and disease-dependent disruption of lipid rafts might be a cause of decreased ability of PrP c to control BACE1 activity resulting in accumulation of Aβ peptide in the case of sporadic AD. The amyloid-degrading enzyme, NEP, has been shown to be partially associated with lipid rafts in human synoviocytes in "ectopeptidase-rich membrane microdomains" (Riemann et al., 2001) , and also in pre-B and B cell lines (Nalm-6 and Raji; Angelisová et al., 1999) . By targeting NEP to different intracellular www.frontiersin.org compartments of neurons, including the lipid rafts, Hama and colleagues demonstrated that the endogenous targeting signal in wild-type NEP is well optimized for the overall neuronal clearance of Aβ (Hama et al., 2004) . Only the mature, fully glycosylated form of NEP, preferentially in its dimerized form, can be found in lipid rafts in direct association with phosphatidylserine (Hama et al., 2004) . These authors also suggested that the localization of NEP in different intracellular compartments may be involved in the metabolism of distinct pools of Aβ and that the endogenous targeting signal in wild-type NEP is well optimized for the overall neuronal clearance of Aβ. The partitioning of NEP into lipid rafts has also recently been confirmed by Sato and colleagues . Although another Aβ-degrading enzyme, insulysin (insulin-degrading enzyme, IDE), is primarily a cytosolic protein it is also, in part, associated with lipid rafts where it may facilitate Aβ clearance (Bulloj et al., 2008) . Apart from compartmentalization within the lipid rafts, the amyloidogenic processing pathway was shown to be dependent on endocytosis (Ehehalt et al., 2003) . In addition to flotillin, caveolin-1 is responsible for partitioning of γ-secretase between the plasma membrane and endosomes and cells depleted of caveolin-1 had more γ-secretase localized within the clathrin-coated noncaveolar endocytic vesicles, although different caveolins have been shown to have different effects on APP processing (Nishiyama et al., 1999; Kapoor et al., 2010) . Lipid rafts affect APP processing not only through favoring interactions between APP and BACE-1 but also by promoting endocytosis of APP. This process was shown to be APP isoform-dependent with the neuronal APP 695 isoform to be mostly processed via the β-secretase pathway whereas APP 751 and APP 770 mainly undergo α-secretase cleavage (Belyaev et al., 2010) . Some of the earliest work about subcellular localization of APP processing indicated that sAPPβ could be detected in neuronal NT2N cell lysates with the absence of sAPPα or p3 fragments. This suggested an intracellular β-secretase pathway (Chyung et al., 1997; Hartmann et al., 1997) . It is worth noting that the β-secretase pathway operates differently with wild-type APP (wtAPP) and APP carrying the Swedish mutation (APP Sw ; Haass et al., 1995; Griffiths et al., 2011 ). An alternative trafficking pathway has been recently reported, whereby APP can bypass endosomes and be trafficked directly to the lysosomes. This, though, does not occur with either APP Sw or APP London variants suggesting that these mutations affect APP transit (Lorenzen et al., 2010) . Although there is quite a substantial amount of data on the physiological role of N-terminal soluble ectodomains of APP -sAPPα and sAPPβ, for review see Chasseigneaux and Allinquant (2012) , the C-terminal products of APP proteolytic cleavage have only recently started to attract special attention. It is now rather well documented (although not without some controversy) that the C-terminal fragment of APP, AICD, can act as a transcription factor (Cao and Sudhof, 2001; Leissring et al., 2002; Schettini et al., 2010) . The controversy of the AICD research has been underpinned by utilization of different cell types expressing different APP isoforms and by the difficulties in detecting AICD due to its very short half life (Cupers et al., 2001) . However, it was repeatedly demonstrated that functional AICD, which translocates to the cell nucleus and up-regulates expression of a reporter NEP gene, is produced only via the amyloidogenic pathway and only from the APP 695 isoform (Goodger et al., 2009; Belyaev et al., 2010) . This process was also shown to be neuronal cell specific and lipid raft dependent (Belyaev et al., 2010) . In an earlier work by Cao and Sudhof (2001) , a yeast 2-hybrid (Y2H) screen was used to identify binding partners of the C-terminal domain of APP which revealed the role of Fe65 and the histone acetyltransferase (HAT) Tip60 in formation of functionally active AICD. AICD regulates the transcription of several target genes, some better characterized than others (Beckett et al., 2012; Pardossi-Piquard and Checler, 2012) . The most well documented gene upregulated by AICD is of the amyloid-degrading enzyme neprilysin (Pardossi-Piquard et al., 2005; Belyaev et al., 2009) . However, there is also evidence that APP itself (von Rotz et al., 2004) , BACE1 (von Rotz et al., 2004) , GSK-3β (Kim et al., 2003) and aquaporin-1 (Huysseune et al., 2009) can be regulated by AICD. In addition to APP, regulation of the GSK-3β can be considered as a link between AICD and AD pathology especially taking into account the data on elevated levels of AICD in the brain of AD patients (Ghosal et al., 2009) . Moreover, the ability of AICD to regulate expression of APP and BACE1 suggests a feedback mechanism of its own regulation by proteolytic processing of its precursor (Grimm et al., 2012a) . AICD also has a direct link to lipid metabolism as it has been found to suppress the expression of the major lipoprotein receptor LRP1 and as such affect apoE/cholesterol metabolism (Liu et al., 2007) . On the other hand AICD controls expression of the alkyldihydroxyacetonephosphate-synthase which regulates plasmalogen synthesis in the cells (Grimm et al., 2011b) and reduced levels of these brain-specific lipids are characteristic of the AD brain (Han et al., 2001; Rothhaar et al., 2012) . Reduced plasmalogen levels in the AD brain might have direct effect on production of Aβ since they were shown to inhibit activity of γ-secretase (Rothhaar et al., 2012) . There is also evidence that AICD regulates sphingolipid synthesis via serine-palmitoyl transferase (Grimm et al., 2011a) , and as such may control composition of lipid rafts and APP processing. The wide range of putative AICD target genes highlights the role of APP signaling in normal brain functioning and in AD pathology. SPHINGOMYELIN The major component of lipid rafts, sphingomyelin (SM), is characteristic only for eukaryotic cells where it comprises about 10-15% of total phospholipids and even more in the brain and peripheral nervous tissue. SM and its metabolites play an important role as second messengers in signal transduction events during development, differentiation and immune response of the organisms (Nalivaeva et al., 2000; Hannun et al., 2001) . SM is essential for the activity of some types of receptors, including the α7 nicotinic receptor (Colón-Sáez and Yakel, 2011), NMDA receptors (Wheeler et al., 2009) , neurotrophic tyrosine kinase receptor type 2 (Trovò et al., 2011) , serotonin 1A receptor (Jafurulla et al., 2008) and the urokinase receptor (uPAR; Sahores et al., 2008) . It was also found Frontiers in Physiology | Membrane Physiology and Biophysics that some disease-related membrane proteins (APP, gp120, and PrP) have a common SM-recognition site which underscores the role of lipid rafts in AD, HIV, and prion diseases . Investigation of lipid raft biology was enhanced by the discovery of SM-specific probes, e.g., lysenin, which serve as powerful tools to study the organization and biological function of this lipid in biological membranes (Hullin-Matsuda and Kobayashi, 2007; Shogomori and Kobayashi, 2008) . These studies have demonstrated functional and structural diversity of lipid rafts and characterized in the plasma membrane of Jurkat T cells the SM-rich domains which had spatial and functional specificity compared to the GM1-rich domains (Kiyokawa et al., 2005) . Formation of SM clusters in the membranes of neuronal cells was shown to depend on localization of SM synthase (SMS) isoforms in various cell compartments and that the activity of SMS is the rate-limiting step in SM cluster formation. In accordance with this it was also demonstrated that SM clusters were formed only in the vicinity of SM synthase proteins. In particular, it was also found that the SMS2 isoform is specific for the dendrites of hippocampal neurons (Kidani et al., 2012) . The pattern of individual species of sphingomeylin was also found to be different in various types of cells and even in cell compartments (e.g., in lipid rafts versus detergent-soluble fractions) which testifies to the role of these lipids in determining cell-specific membrane properties (Valsecchi et al., 2007) . In the aging brain and especially in AD pathology, lipid metabolism, and in particular that of SM, undergoes significant changes. By using a shotgun lipidomics approach Han and colleagues have compared levels of over 800 lipid species in control and AD brains and demonstrated a significant decrease in SM and increases in ceramide levels in the affected brain. They suggested a model in which an AD-related increase in SMase activity results in faster SM hydrolysis (and increased ceramide production) which would lead to alterations in lipid raft formation (Han et al., 2011) . These authors also hypothesized that this would impair functions of GLUT4, which was previously shown to be functionally linked to lipid rafts (Michel and Bakovic, 2007) and lead to the dysfunctions in energy homeostasis characteristic of the disease. Moreover it was also shown that accumulation of the amyloid peptide in neuronal cells leads to SMase activation and SM depletion which, in turn, affects cellular trafficking and abnormal APP processing (Soreghan et al., 2003) . Further work demonstrated that the most toxic form of amyloid peptide Aβ 42 can also activate SMase (Grimm et al., 2005; Grosgen et al., 2010) . Since excessive SMasemediated cleavage of SM occurs early in AD it is likely to disrupt a range of protein-lipid interactions and hence downstream signaling pathways (Haughey et al., 2010) . In addition to AD-linked SMase over-activity, there have been reports of deficiencies of the enzymes responsible for sphingolipid synthesis (Piccinini et al., 2010) . Gangliosides are glycosphingolipids with one or more sialic acid residues. They are essential components of all animal cell membranes where they are anchored in the external leaflet by the hydrophobic ceramide part of their molecule while the oligosaccharide chain protrudes into the extracellular medium. Gangliosides are particularly abundant in the plasma membranes of neuronal cells have been implicated in various cellular functions since the high heterogeneity of their oligosaccharide structures allows specific interactions with various proteins and oligosaccharide chains of other molecules at the cell membrane surface. Gangliosides to a great extent determine the fine structure of membranes, e.g., lateral diffusion of its components, as well as the organization of lipid rafts (Cantù et al., 2011) . Disruption of ganglioside metabolism leads to various neurological diseases (for review see Walkley, 2003) . Studies of ganglioside-knockout mice have demonstrated that depletion of various classes of gangliosides results in development of neurodegeneration and pathology similar to Alzheimer's or Parkinson's diseases Wu et al., 2011) . In the aging brain, and especially in AD patients, ganglioside content significantly decreases in various brain structures, especially in the areas of the brain related to the pathogenesis of the disease (Kracun et al., 1991) . The link of gangliosides with AD pathology has been mostly related to their localization in lipid rafts which have been suggested to act as a platform for GM1-induced aggregation of Aβ peptide (Parton, 1994; Zha et al., 2004; Ariga et al., 2008) . Labeled Aβ shows high affinity for GM1 containing membranes, suggesting that GM1 acts as an Aβ binding molecule (Kakio et al., 2004) . It has been suggested that the N-terminal region of Aβ interacts with GM1 clusters through hydrogen bonding and electrostatic interactions (Lemkul and Bevan, 2011) . It was also suggested that cholesterol may facilitate GM1 clustering (Kakio et al., 2001) . Extraction of cortical lipid rafts from human AD brains showed an increase in both GM1 and GM2 (Molander-Melin et al., 2005) despite an overall reduction of gangliosides (Ariga et al., 2008) . There are data indicating that gangliosides accumulate in senile plaques and that they may be involved in the conversion of Aβ to a neurotoxic oligomeric form (Molander-Melin et al., 2005; Okada et al., 2008) . Various synaptosomal and liposomal studies have suggested that GM1 accelerates formation of amyloid fibrils (Yamamoto et al., 2004; Zha et al., 2004) . Further analysis of the role of lipid rafts in Aβ aggregation have confirmed that raft proteins do not play a significant role in this process as proteinase K or SDS treatment did not affect the aggregation-promoting capacity of the rafts. Moreover, the cholesterol-depleting agent methyl-βcyclodextrin (MβCD) also did not have an effect on Aβ oligomerization (Kim et al., 2006) . Further cell studies reinforce the role of gangliosides in Aβ aggregation. Rafts taken from ganglioside rich cells (C 2 C 12 ) were able to induce Aβ aggregation more potently than ganglioside-poor cells (HeLa and SK-N-MC). Similarly, rafts isolated from brain tissue rich in gangliosides were able to increase Aβ aggregation compared to rafts isolated from the liver with lower ganglioside content. Also, CHO-K1 cells genetically deficient in the synthesis of complex gangliosides had lower levels of Aβ aggregation than the wild type cells (Kim et al., 2006) . An examination of the effects of GM1 on APP processing in SH-SY5Y and COS7 cells showed an inhibition of α-cleavage (Zha et al., 2004) . On the other hand, Aβ 1-40 oligomers were found to stimulate the amyloidogenic processing of APP by reducing www.frontiersin.org membrane fluidity and complexing with GM1 ganglioside (Peters et al., 2009 ). Although precise mechanisms of GM1 changes with age are still unknown it was demonstrated that GM1 content in neuronal membranes, particularly in DRM microdomains, increases with age and this increase was more pronounced in the brain of apoE4 knock-in mice compared to apoE3 knock-in animals (Yamamoto et al., 2004) . Recently two novel mechanisms linking gangliosides with AD have been described by Grimm et al. (2012b) . They discovered that Aβ binds to GD3-synthase (GD3S, the key-enzyme in converting a-series gangliosides to major brain specific b-series) and inhibits its activity. On the other hand, the APP intracellular domain AICD, together with Fe65, was found to down-regulate expression of the GD3S gene. This provides an explanation for age-dependent and AD-related changes in brain ganglioside patterns and supports an essential role of APP in ganglioside homeostasis. Another link between gangliosides and AD lies in their ability to alter APP processing. As observed by the authors, GM3 was able to decrease production of Aβ while GD3 was shown to increase its levels in COS7 cells. This might be related to the changes in the properties of lipid rafts containing different amounts of GM3 or GD3 ganglioside species and their ability to regulate activity of βand γ-secretases. Taking into account the important role of gangliosides in normal brain function, their involvement in the pathogenesis of AD should be treated with caution since different species of gangliosides will have a different impact on the integrity and properties of neuronal membranes. Recently, by using G3 synthase KO mice (lacking only b-series gangliosides) and GM2/GD2 synthase KO mice (which lack almost all gangliosides except GM3 and GD3) it was found that these ganglioside species are important for neuroprotection and anti-inflammatory response via maintenance of lipid rafts . Intraventricular treatment of AD patients with GM1 was shown to stop the progression of cognitive deterioration and improve motor performance and neuropsychological assessments (Svennerholm et al., 2002) . Peripheral utilization of GM1 for prevention of Aβ aggregation in the brain via establishing peripheral/brain dynamics of Aβ was suggested as a possible therapeutic approach for AD since in the PS/APP mice this was shown to reduce Aβ accumulation in the brain (Matsuoka et al., 2003) . However, due to the antigenic properties of gangliosides this might provoke some side-effects (López-Requena et al., 2007) . Another important component of cellular membranes and of lipid rafts, cholesterol, has been recently extensively reviewed in relation to its role in AD (Marzolo and Bu, 2009; Burns and Rebeck, 2010; Mathew et al., 2011) . The picture is by no means a simple one with cholesterol being ascribed both positive and negative roles. Indeed, other diseases are also linked to cholesterol, while it remains unclear which are specifically raft-associated. Such pathologies include Smith-Lemli-Opitz syndrome, Huntington's disease and Niemann-Pick Type C disease (Korade and Kenworthy, 2008) . In terms of normal aging, the data seem to show that the changes in cholesterol are highly dependent on the brain region and cell types used in the studies . Cholesterol levels have been shown to be elevated in AD patients as well as the levels of cholesterol precursors in the mevalonate pathway, farnesylpyrophosphate, and geranylgeranylpyrophosphate (Hooff et al., 2010; Kolsch et al., 2010) . However, this view is by no means unanimous and other groups have found lower levels of cholesterol in AD brains (Kolsch et al., 2010; Leduc et al., 2010) , in addition to its precursors lanosterol and lathosterol (Kolsch et al., 2010) , with lower levels of its synthesizing enzyme, HMG CoA reductase (Leduc et al., 2010) . However, quantification of global cholesterol levels are not necessarily reflective of the number or distribution of lipid rafts (Leduc et al., 2010) . As cholesterol is integral to ordered lipid rafts, the consequences of cholesterol depletion are widely regarded as effects of raft disruption (Hao et al., 2001; Mondal et al., 2009) . In this context the cholesterol content in the membranes has an inversely proportional relationship with the membrane-perturbing effects of Aβ oligomers (Cecchi et al., 2009 ). As such, if cholesterol increases do elevate lipid raft abundance, then it would increase Aβ formation (Simons et al., 1998; Ehehalt et al., 2003) and, on the contrary, low cholesterol levels will lead to up-regulation of the activity of the α-secretase, ADAM10 (Kojro et al., 2001) . Cholesterol depletion in cells by MβCD was shown to affect APP-processing and formation of functionally active AICD resulting in reduced levels of expression of the amyloid-degrading enzyme, neprilysin (Belyaev et al., 2010) . On the other hand, cholesterol has been shown to bind C99, which promotes amyloidogenic processing (Beel et al., 2010) and the increase of Aβ levels, in turn, can cause changes in cholesterol homeostasis in the Golgi and plasma membrane ). Formation of the Aβ "seed" and initiation of Aβ aggregation was also shown to be cholesterol dependent (Mizuno et al., 1999; Kakio et al., 2001; Simons et al., 2001) . In a more biophysical sense, raised cholesterol has been implicated in facilitating the insertion of Aβ into the plasma membrane. In so doing, Aβ then destroys the cells' membrane integrity (Ji et al., 2002) . Just as cholesterol and lipid rafts have been shown to affect APP processing, APP-derived species have been shown to impact on cholesterol and lipid homeostasis. Aβ peptides modulate the metabolism of cholesterol, in particular its esterification rate, and of phospholipids in hepatocytes, neuronal cells, and in the entire brain Koudinova et al., 1996 Koudinova et al., , 2000 . It was also found that Aβ peptides alter vesicle trafficking and cholesterol homeostasis (Liu et al., 1998) . On the other hand, it was shown that cholesterol binds to APP at the α-secretase cleavage sites and Aβ itself can bind cholesterol and prevent its interaction with low-density lipoprotein (Yao and Papadopoulos, 2002) . This confirmed that Aβ might act as a component of lipoprotein complexes and affect reverse cholesterol transport from neuronal tissue to the periphery in addition to its role in cholesterol synthesis and intracellular dynamics (Koudinov et al., 2001; Michikawa et al., 2001) . In a later study Aβ 40 has been shown to inhibit a key enzyme in the biosynthesis of cholesterol, HMG CoA reductase (Grimm et al., 2005) . Further to this, APP intracellular domain AICD was found to regulate cholesterol levels via LRP1 (Grosgen et al., 2010) . However, there also are data reporting that cholesterol in physiological concentrations can protect neuronal cells Frontiers in Physiology | Membrane Physiology and Biophysics against Aβ-induced toxicity and slow down the process of formation of toxic aggregates of Aβ with metal ions, in particular with aluminum (Granzotto et al., 2011) . This correlates with the data suggesting that cholesterol may have a protective effect against membrane disruption by amyloid species (in this case, Aβ-derived diffusible ligands: ADDLs). Cholesterol supplemented SH-SY5Y cells were shown to display reduced binding of ADDLs to the plasma membrane, while oligomers increased in membrane presence after treatment with the cholesterol-depleting agent MβCD (Cecchi et al., 2009) . Statins are a class of drugs which inhibit HMG CoA reductase, a key enzyme in the biosynthesis of cholesterol. Given the links between cholesterol and AD, it is not surprising that there are multiple investigations on the effects of statins in AD pathology. The current state of the field and problems have been recently discussed and the role of statins has been reviewed Wood et al., 2010) . The authors point out that the effects of statins are not only related to APP processing nor specific for the AD pathology. Furthermore, the physiological effects of statins are not solely due to inhibition of cholesterol biosynthesis but include perturbation of other mevalonate-dependent pathways such as protein prenylation. A number of studies have shown positive effects of statins in AD (Jick et al., 2000; Buxbaum et al., 2001 Buxbaum et al., , 2002 . This is in agreement with the previous studies demonstrating that depletion of cholesterol reduces Aβ in cultured neurons (Simons et al., 1998) . One recent study has suggested that fluvastatin is able to modify the trafficking of APP. Use of this drug was stated to increase lysosomal degradation of APP C-terminal fragments (CTFs) and hence facilitate Aβ clearance (Shinohara et al., 2010) . However, the reduction in Aβ levels is not necessarily linked with the cognitive benefits. A recent randomized, double-blinded, placebo-controlled trial of simvastatin showed reduced Aβ in treated patients, but no corresponding improvement in cognitive performance (ADAS-Cog score; Sano et al., 2011) . The biochemical studies are fairly uniform in that cholesterol depletion results in reductions in key ADrelated markers. However, the results of epidemiological research are unequivocal. The Cochrane Dementia and Cognitive Improvement Group study reported that statin treatment had no effect on the prevention or treatment of dementia (McGuinness et al., 2009a,b) . However, taking into account the highly variable relationship between the initiation of statin therapy and the time and severity of the AD, it is very difficult to get a conclusive assessment of the accumulated data on the beneficial effect of statins and any such study should use a defined set of criteria during epidemiological meta-analysis (Shepardson et al., 2011a,b) . Lipid rafts have been implicated as the sites for a great number of signaling pathways (Allen et al., 2007) . Perturbation of, or changes, in lipid rafts could therefore affect neuronal signaling, including cholinergic transmission. As cholinergic hypofunction is key to the pathogenesis of AD (Schliebs, 2005; Schliebs and Arendt, 2006) , there are strong links between lipid rafts, neuronal signaling pathways, and AD. Various lipids, such as eicosanoids, docosanoids, and cannabinoids, can act as signaling mediators and their abnormal metabolism has implications in AD (Farooqui, 2011) . These lipid mediators can modulate the metabolism of sphingolipids through activation of SMases. Furthermore, sphingolipid-derived molecules such as ceramide and ceramide 1-phosphate can act as lipid mediators and accumulate in the AD brain. This accumulation of sphingolipid derivatives can activate cytosolic phospholipase A 2 (cPLA 2 ), which leads to changes in membrane fluidity and permeability with concomitant alterations in ion homeostasis. Furthermore, the degradation products of cPLA 2 metabolism are often pro-inflammatory (Frisardi et al., 2011) . This work shows how the lipid raft constituents, in this case sphingolipids, can affect AD processes in an APP-independent manner. Here, accumulation and action of these lipid mediators promotes inflammation, a process characteristic of AD (Akiyama et al., 2000) . Numerous neurotransmitter signaling systems, especially receptor function, are influenced by lipid rafts. The range of receptors involved is all-encompassing, including ionotropic receptors such as α-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid (AMPA receptors), γ-aminobutyric acid (GABA) receptors, Nmethyl-d-aspartate (NMDA), and nicotinic acetylcholine (nACh) receptors. The influence of lipid rafts also extends to metabotropic G-protein coupled receptors, such as the muscarinic acetylcholine receptor (mAChR; Allen et al., 2007; Rushworth and Hooper, 2011) . Although all these signaling systems can be linked to AD, it is the cholinergic system which is of central importance, as its hypofunction is considered one of the hallmarks of the disease (Coyle et al., 1983; Auld et al., 2002; Schliebs and Arendt, 2006) and its involvement is therefore principally considered here. Lipid rafts have been implicated in regulating the clustering of nAChRs via the myristoylated peripheral membrane protein, rapsyn, which is constitutively lipid raft localized. Perturbation of the rafts impedes the interaction between rapsyn and nAChRs, which reduces clustering of the receptors and hence their function (Zhu et al., 2006) . Protein reconstitution studies using lipid vesicles imply that the partitioning of the nAChR into raft domains is not solely due to the intrinsic biophysical properties of the receptor but requires a signaling event to translocate the protein into specific membrane domains (Bermúdez et al., 2010) . Furthermore, cholesterol affects the properties, both structural and functional, of some acetylcholine receptors . Agrin signaling represents one pathway which can enhance the translocation of the nAChR into lipid rafts (Campagna and Fallon, 2006; Zhu et al., 2006) . In particular, one of the most prominent examples of a lipid raft-linked signaling protein is the α7 nicotinic acetylcholine receptor (nAChR), a Ca 2+ channel (Albuquerque et al., 2009 ), which has a close involvement with Aβ and cognitive decline (Albuquerque et al., 2009; Jurgensen and Ferreira, 2009; Hernandez et al., 2010) . On balance, the α7 nAChR seems to have a www.frontiersin.org neuroprotective role in model systems. Significant amounts of the α7 nAChR are associated with lipid rafts and these rafts are essential for maintenance of function (Bruses et al., 2001) . Lipid rafts appear to assist the interaction between soluble Aβ and the receptor (Khan et al., 2010) . Although α7 KO mice show no apparent cognitive defects, Aβ 42 is enriched and Aβ oligomers are enhanced in hippocampi from these animals (Hernandez et al., 2010) . The α7 nAChR regulates, in part, the pleiotropic intracellular signaling effects of Ca 2+ . In addition to this, the receptor is closely linked to the signaling of cAMP (via adenylyl cyclase) as well as the kinases Fyn and PI3K (Oshikawa et al., 2003) . Disruption of lipid rafts with MβCD or SMase has significant effects on receptor desensitization kinetics (Colón-Sáez and Yakel, 2011) and hence the many important downstream signaling events. Additional interactions of the α7 nAChR have been assessed using a proteomic screen, which identified several proteins involved in neurite outgrowth and maintenance, namely α-catenin 2, BASP1/NAP-22, gelsolin, homer 1, and neuromodulin (Paulo et al., 2009 ). Further to this, the AD therapeutic donepezil protects against glutamate toxicity in part through stimulation of the α7 nAChR (Shen et al., 2010) . Given the central role of the α7 nAChR, it has been suggested promoting its activation via novel agonists may represent a new therapeutic approach in AD (Wang, 2010) . Lipid raft localization has recently been linked to another protein of the cholinergic system, namely acetylcholinesterase (AChE), although the functional implications of this are as yet unclear (Xie et al., 2010b; Hicks et al., 2011) . AChE inhibition by compounds such as rivastigmine or galantamine represents the major therapeutic option for treating the cognitive impairment seen in the early stages of AD yet the relationship between AChE and AD represents something of a paradox. AChE exists in a number of different molecular forms (G 1 , G 2 , G 4 ) of which the tetrameric G 4 form is predominant in brain. In AD, brain G 4 AChE levels are seen to fall as the disease progresses, while G 1 and G 2 levels rise somewhat, as compared to normal brains (Atack et al., 1983; Garcia-Ayllon et al., 2010) . AChE is found associated with amyloid plaques, leading to the suggestion that AChE may promote Aβ aggregation (Moran et al., 1993; Inestrosa et al., 1996) . In some brain regions with AD pathology, virtually all of the AChE is localized in these complexes (Mesulam et al., 1987) . A direct interaction between Aβ and AChE has been proposed, with binding occurring at the peripheral anionic site (PAS) of the enzyme. Those AChE inhibitors which occupy the PAS (e.g., propidium) show the most significant reductions in fibril formation (Bartolini et al., 2003) since the catalytic site is not required for interaction with Aβ (Inestrosa et al., 1996) . Furthermore, monoclonal antibodies directed against the PAS inhibit fibril formation (Reyes et al., 1997) , which has led to the development of PAS blockers, such as the DUO compounds, that also occupy the active site. They show inhibitory activity on AChE as well as inhibition of Aβ 40 fibril formation (Alptuzun et al., 2010) and have been suggested as potential novel AD therapeutics targeting two facets of the disease. The related enzyme butyrylcholinesterase (BuChE), as well as a synthetic peptide derived from the BuChE C terminus (BSP41), have also been shown to reduce amyloid fibril formation (Diamant et al., 2006) . The corresponding AChE synthetic peptide, however, did not significantly affect Aβ fibril formation. AChE is not a transmembrane protein, rather it is anchored to the plasma membrane by the proline rich membrane anchor (PRiMA) which is a 20-kDa type I transmembrane protein which can be acylated (Henderson et al., 2010; Xie et al., 2010b) . PRiMA contains a CRAC (cholesterol recognition amino acid consensus) motif which sequesters PRiMA into lipid rafts and hence AChE is also partly associated with rafts although the functional significance of this interaction is unclear (Xie et al., 2010a) . It has been suggested that the lipid raft localization and shedding of AChE may have a certain role in the pathology of AD (Hicks et al., 2011) . PrP is another component of lipid rafts which was shown to act as a cellular receptor and change Ca 2+ signaling upon activation by some ligands, e.g., by antibody cross-linking in T-lymphocytes (Stuermer et al., 2005) . PrP c was also shown to be a receptor for Aβ oligomers at nanomolar concentrations and binding of Aβ oligomers to PrP c results in the blockage of hippocampal LTP. Incubation of hippocampal slices with PrP antibody abolishes the effect of Aβ oligomers and rescues synaptic plasticity (Lauren et al., 2009) . Most recently it was demonstrated PrP c also interacts with the NMDA receptor complex in a copper-dependent manner to allosterically reduce glycine affinity for the receptor and that Aβ(1-42), copper chelators or PrP c inactivation all enhance the activity of glycine at the receptor resulting in steady-state NMDAR currents and neurotoxicity. It was suggested that the physiological role of PrP c might be related to limitation of excessive NMDAR activity which might cause neuronal damage (You et al., 2012) . Together with the data on BACE1 regulation by PrP c this provides a unifying molecular mechanism explaining the interplay between toxic Aβ species, NMDA receptor-mediated toxicity and copper homeostasis in pathogenesis of AD (Rushworth and Hooper, 2011) . It is possible for lipid rafts to modulate signaling in a general way by affecting the activities of signaling molecules involved in multiple pathways. These include, for example, the pleiotropic src kinases (Arcaro et al., 2007) with effects on the PI3K-Akt signaling pathway. Other raft-localized signaling proteins include the epidermal growth factor receptor (EGFR), which associates with caveolins (Couet et al., 1997) and is involved in diverse processes including cell cycle regulation, endocytosis, and the MAPK cascade (Oda et al., 2005) . This has led to the concept of the signalosome, containing interacting components of signaling pathways (e.g., EGFR) within lipid rafts along with other scaffolding proteins, such as caveolins, and sequestering the complex from other interacting proteins that may disrupt the signaling process. A prime example is the CD40 signalosome associated with cell growth in B cell lymphomas (Pham et al., 2002) . Similar signaling platforms operate in neuronal systems, such as that involving estrogen receptor (ER) interactions (reviewed in Marin, 2011) . This signaling mechanism is linked to neurogenesis, neuronal differentiation, synaptic plasticity, and neuroprotection (including against Aβ). Recent research indicates that lipid rafts are the site of formation of a complex between the ER, insulin growth factor 1 receptor (IGF-1R), Cav-1, and a voltage gated anion channel, VDAC. The formation of this signaling complex is neuroprotective but also lipid raft dependent (Marin, 2011) . The "signalosome" paradigm was developed further by Chadwick et al. (2011) who isolated lipid rafts in control and 3xTg AD mice and compared their respective proteomes by mass spectrometry. Proteins so identified were then clustered into specific signaling pathways, which allowed an appraisal of which lipid raft signaling pathways may be altered in AD, rather than changes in individual proteins. This systems biology approach indicated that, in lipid rafts, wild-type mice had higher activation of prosurvival pathways such as PTEN and Wnt/β-catenin, whereas 3xTg mice showed activation of p53 and JNK signaling pathways. In addition, 3xTg mice had a deficit in growth factor signaling, neurodevelopmental signaling and signaling through the sonic hedgehog pathway (Chadwick et al., 2011) . Another proteomic analysis extracted post-synaptic lipid rafts and used LC-MS/MS to analyze their protein content. They found an enrichment of cell adhesion molecules, channels/transporters and G-protein related species. Their data linked lipid rafts to cell adhesion with cell-cell contact regions and cell adhesion points being enriched in rafts. Further to this, H + -ATPase and Na + -K + ATPase are enriched in lipid rafts, being responsible for maintaining ionic gradients and modulating neuronal excitability. This study also found the postsynaptic density to be associated with lipid rafts (Suzuki et al., 2011) . Among channels localized in lipid rafts and involved in maintenance of neuronal cell homeostasis, in particular of astrocytes, are K + -buffering inwardly rectifying Kir4.1 channels and the water channel AQP4 (Hibino and Kurachi, 2007) . Residence in the lipid rafts was also shown to be important for the activity of a member of the chloride channel family, ClC-2, which is widely expressed in the brain and other organs (Cornejo et al., 2009) . Ever since the initial descriptions and characterizations of lipid rafts, the field has been beset by controversy (Pike, 2009 ) from their very existence down to the very specific aspects of how to isolate these structures in the laboratory. However, current consensus is that lipid rafts do represent dynamic structural components of cellular membranes integrating signaling events and regulating cell functioning and that their dysregulation can lead to disease. This review has largely concentrated on the links between lipid rafts and AD pathology since processing of AD-related APP and production of Aβ peptide are clearly affected by lipid rafts. Also Aβ signaling involves interactions of proteins resident to lipid rafts. The mechanism underlying these effects has been examined in some detail although numerous gaps in knowledge still remain. Although the attempts to modulate lipid rafts and hence amyloidogenic processing have failed to translate into successful drugs, some epidemiological studies still indicate that inhibition of cholesterol synthesis through statin treatment might be beneficial if applied early (Solomon and Kivipelto, 2009; Shepardson et al., 2011a,b) . The complexity of AD pathology and etiology dictates to consider involvement of lipid rafts in its pathogenesis in a more generic way, not only as a simplistic link between cholesterol levels, amyloid burden and cognition. More important for normal brain functioning is to maintain lipid metabolism in the aging brain at its normal rate and integrity. In terms of future progress, lipid raft research might open new avenues in regulation of the proteolytic and signaling processes involved in AD pathology. The most recent discovery of the role of lipid raft disruption in decreased production of the functionally active transcriptional regulator AICD which might lead to an aberrant expression of its target genes including amyloid-degrading enzymes neprilysin (Howell et al., 1995; Carson and Turner, 2002; Belyaev et al., 2009; Liu et al., 2011) , suggests that any therapeutics aimed at manipulation of lipid raft composition should be treated with caution. The role of the lipid components in cell membrane functioning and their structural variability and adaptive potential is extremely important for normal functioning of cells and organisms and much of the recent work in this area is both novel and revealing. However, it is important to note that these discoveries should be recognized as important advances in cell science and not seen as stepping stones to a therapeutic panacea. of (benzylidene-hydrazono)-1,4dihydropyridines with β-amyloid, acetylcholine, and butyrylcholine esterases. Bioorg. Med. Chem. 18, 2049 -2059 . Angelisová, P., Drbal, K., Horejsí, V., and Cerný, J. (1999 . Association of CD10/neutral endopeptidase 24.11 with membrane microdomains rich in glycosylphosphatidylinositolanchored proteins and Lyn kinase. Blood 93, 1437-1439. Araki, W., Kitaguchi, N., Tokushima, Y., Ishii, K., Aratake, H., Shimohama, S., Nakamura, S., and Kimura, J. (1991) . Trophic effect of β-amyloid www.frontiersin.org

West Nile Virus Infection in Killer Whale, Texas, USA, 2007

In 2007, nonsuppurative encephalitis was identified in a killer whale at a Texas, USA, marine park. Panviral DNA microarray of brain tissue suggested West Nile virus (WNV); WNV was confirmed by reverse transcription PCR and sequencing. Immunohistochemistry demonstrated WNV antigen within neurons. WNV should be considered in cases of encephalitis in cetaceans.

virus of the genus Flavivirus that is transmitted by mosquitoes. In humans and animals, WNV has been associated with a spectrum of clinical conditions from asymptomatic infections to sudden death. These have been identifi ed in a variety of animal species. Among marine mammals, WNV infection has been reported in a harbor seal (Phoca vitulina) (1) . We describe WNV infection in a killer whale (Orcinus orca) and seroprevalence in conspecifi c cohort and noncohort groups. In 2007, a 14-year-old male killer whale at a marine park in San Antonio, Texas, USA, died suddenly without notable premonitory signs. On gross examination, mild multifocal meningeal hyperemia and petechial parenchymal hemorrhage were noted in the right cerebrum and cerebellum. The left hemisphere of the brain appeared normal. Focally extensive tan discoloration and fi brosis were present in the right accessory lung lobe with associated hemorrhage and congestion. Both lung lobes were mildly and diffusely heavy and wet. All thoracic and abdominal lymph nodes were moderately enlarged and edematous. The second gastric chamber displayed numerous chronic and active ulcerations of 1.5-2 cm. Fresh and buffered 10% formalin-fi xed specimens were collected. Fresh tissues were stored at -80°C. Tissues fi xed in 10% buffered formalin were processed routinely and stained with hematoxylin and eosin for histologic examination. Histologic review demonstrated moderate multifocal subacute vasculitis and nonsuppurative encephalitis. Infl ammatory lesions of the central nervous system were focused in gray matter of the medulla oblongata, pons, mesencephalan, and cerebellum. Lesions were bilateral but more severe on the right side. Meninges demonstrated moderate focally extensive and multifocal areas of acute meningeal congestion and hemorrhage. Mild multifocal lymphocytic infi ltrates expanded the leptomeninges. Blood vessels demonstrated mild to moderate acute necrosis and lymphocytic and contained plasmacytic and neutrophilic infi ltrates within vascular walls. Encephalitis was characterized by perivascular lymphocytes and fewer plasma cells expanding the Virchow-Robbins spaces. Small, scattered, perivascular ring hemorrhages were noted. A few multifocal loosely arranged glial nodules were within cerebral white matter. Predominant lesions in the lungs were areas of chronic and active abscessation amid a focally extensive area of mixed infl ammation and fi brosis. There was moderate diffuse acute pulmonary edema and congestion. Gastric ulcerations were present in the fi rst gastric chamber and were chronic and active. They were characterized by central ulcerations with necrosis and a mixed infl ammatory infi ltrate surrounded by variable fi brosis and a rim of epithelial hyperplasia. Changes in spleen, lymph node, and kidney included acute edema, congestion, and vascular dilation. Conventional diagnostic assays were performed for aerobic, anaerobic, and fungal microbes in liver, lung, kidney, cerebrospinal fl uid, and brain. All yielded minimal growth of Escherichia coli. The fi nal diagnosis was fulminant peracute bacteremia and septicemia secondary to a primary viral infection associated with nonsuppurative encephalitis. Published etiologic considerations for cetacean nonsuppurative encephalitis include morbillivirus and protozoal infections (2) . A DNA microarray with highly conserved sequences from >1,000 viruses was selected to screen for known and novel viruses (3). Total RNA was extracted from brain tissue and hybridized to a microarray as described (4) . Analysis of the resulting hybridization pattern demonstrated a strong hybridization signal to many oligonucleotide probes on the microarray from the family Flaviviridae, in particular to WNV. Consensus reverse transcription PCR primers (5) targeting WNV were used to confi rm the microarray results. Sequencing of the 261-bp amplicon (GenBank accession no. HQ610502) yielded a sequence with 99% nt identity and 100% aa acid identity to WNV strain OK03 (GenBank accession no. EU155484.1), a strain originally identifi ed in Oklahoma, USA. To further support a WNV diagnosis, we performed immunohistochemical staining on brain tissue. The immunoperoxidase stain used was a commercial rabbit polyclonal antibody (BioRelience Corp., Rockville, MD, USA) with peroxidase-tagged goat antirabbit immunoglobulin G (DakoCytomation, Carpinteria, CA, USA) bridge and 3-amino-9-ethylcarbazole (DakoCytomation) as the chromogen. This staining demonstrated abundant WNV antigen within the cytoplasm of a small number of neurons and glial cells and in fewer macrophages in the brain tissue (Figure) . We evaluated WNV exposure within the same cohort, as well as a geographically distant cohort of whales by using serologic testing. All testing was performed at the same laboratory by using a standard plaque-reduction neutralization test. In this assay, a 90% neutralization cutoff was used (6) . A 90% plaque-reduction titer >10 was considered positive. Serum from the affected whale and 5 cohort killer whales from the same marine park in San Antonio as well as 5 whales housed at another facility in Orlando, Florida, USA, were evaluated. In each facility, the animals have regular contact with each other. The facilities are geographically separated so the animals do not have exposure to those in the other park. All 6 animals from Texas had 90% plaque-reduction titers >10, ranging from 40 to 80. The 5 whales housed together in Orlando had no measurable titer. We demonstrate that WNV can infect and cause disease in killer whales. These fi ndings broaden the known host tropism of WNV to include cetaceans in addition to previously known pinnipeds. Although we cannot defi nitively attribute the cause of death of this whale to WNV, the observed lesions are consistent with those caused by WNV in other animals. The serologic results demonstrate that subclinical infections can occur and that exposure can be variable. We did not determine specifi c dates of exposure for these populations. Both Bexar County, Texas, and Orange County, Florida, have had WNV in wildlife since 2002. We continue annual serology on previously negative animals to document seroconversion. Mosquito management practices are similar in both facilities and have been expanded since this diagnosis. Differences in WNV prevalence or mosquito numbers may have played a role in the different serologic results. Health evaluations of free-ranging and captive cetaceans should include WNV serology to assess exposure rates. This report focuses on killer whales, but the "loafi ng" behavior (stationary positioning at the water's surface) is commonly seen in many coastal dolphins, thereby increasing the likelihood of mosquito bites and exposure to WNV. Serologic screening of bottlenose dolphins (Tursiops truncatus) from the Indian River Lagoon demonstrated WNV titers (7) . WNV-associated disease in these animals has not been reported. Active screening for WNV may enhance diagnostic investigations. As with many species of birds and mammals, WNV infection carries a risk for zoonotic transmission. Until the implications of this infection in marine mammals are better understood, biologists and veterinarians working with cetaceans should consider this possibility. Potential viral shedding can occur through the oropharygeal cavity and feces as well as through blood and organs during necropsies. Finally, our study demonstrates the broad applicability of using panviral microarray-based diagnostics. Even though PCR diagnostics are well developed for WNV, the agent was not initially considered as a potential pathogen in this species. Panviral microarray can be used not only to identify novel viruses but also to detect unsuspected agents. Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 17, No. 8, August 2011

Differential Seroprevalence of Human Bocavirus Species 1-4 in Beijing, China

BACKGROUND: Four species of human bocaviruses (HBoV1-4) have been identified based on phylogenetic analysis since its first report in 2005. HBoV1 has been associated with respiratory disease, whereas HBoV2-4 are mainly detected in enteric infections. Although the prevalence of HBoVs in humans has been studied in some regions, it has not been well addressed globally. METHODOLOGY/PRINCIPAL FINDINGS: Cross-reactivity of anti-VP2 antibodies was detected between HBoV1, 2, 3, and 4 in mouse and human serum. The prevalence of specific anti-VP2 IgG antibodies against HBoV1-4 was determined in different age groups of healthy individuals aged 0-70 years old in Beijing, China, using a competition ELISA assay based on virus-like particles of HBoV1-4. The seroprevalence of HBoV1-4 was 50%, 36.9%, 28.7%, and 0.8%, respectively, in children aged 0-14 years (n = 244); whereas the seroprevalence of HBoV1-4 was 66.9%, 49.3%, 38.7% and 1.4%, respectively, in healthy adults (≥15 years old; n = 142). The seropositive rate of HBoV1 was higher than that of HBoV2, HBoV3, and HBoV4 in individuals older than 0.5 years. Furthermore, IgG seroconversion of HBoV1 (10/31, 32.3%), HBoV2 (8/31, 25.8%), and HBoV3 (2/31, 6.5%) was found in paired sera collected from children with respiratory tract infections who were positive for HBoV1 according to PCR analysis. CONCLUSIONS/SIGNIFICANCE: Our data indicate that HBoV1 is more prevalent than HBoV2, HBoV3, and HBoV4 in the population we sampled in Beijing, China, suggesting that HBoV species may play differential roles in disease.

Human bocavirus (HBoV), a member of the Parvoviridae family, is a potential etiologic agent of respiratory disease and of acute gastroenteritis [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] . Based on phylogenetic analysis of viral genomes, four species of HBoVs (HBoV1-4) have been identified [1, [7] [8] [9] . HBoV1 is associated with respiratory tract diseases [1, 2, [12] [13] [14] . HBoV2 and 3 have been detected in the respiratory tract, but are associated mainly with stool samples [8] [9] [10] [11] 15, 16] . HBoV4 has been detected in enteric infections [9] . However, as HBoVs are frequently co-detected with other viral infections in patients with respiratory or enteric infections, the exact roles of HBoVs in pathogenicity are unclear. HBoVs are small, non-enveloped viruses with a linear singlestranded DNA genome of approximately 5 kb in length. The genome consists of four open reading frames (ORFs), encoding two nonstructural proteins (NS1 and NP1) and two overlapping capsid proteins (VP1 and VP2) [1] . The lack of a well-established cell culture system or animal model to propagate HBoVs has hampered understanding of the infection and pathogenicity of HBoVs. Studies have shown that the VP2 protein harbors the major antigen of HBoV and can form the empty virus-like particles (VLPs) which mimic HBoV virions morphologically and antigenically. The VP2 VLPs have been successfully used as antigens for detecting antibodies against HBoVs [17] [18] [19] [20] [21] [22] . Currently, detection of HBoVs nucleic acid is primarily used to estimate the prevalence of HBoV species in clinical samples. The prevalence of HBoV1, which is mainly detectable in children under two years old [23] , is 2-19% in patients suffering from acute respiratory tract infections (ARTIs) worldwide as detected by PCR analysis [1, 2, 10, 12, 13, 23, 24] . The detection rate of HBoV2, HBoV3, and HBoV4 DNA in stool samples have been reported as 1-26%, 0.4-5%, and 0-2%, respectively [8, 9, 23, 25] . The prevalence of HBoV2-4 is higher in children than in adults according to some, but not all studies [8, 9, 11, 16, 25] . However, the data collected from patients may not represent HBoV infection in the general population as subclinical infections can occur and HBoV persists in the nasopharynx [26] [27] [28] [29] . Seroepidemilogical investigations of healthy populations may be more useful than patient studies in assessing the prevalence, spread, and exposure distribution of HBoVs in the population. The seroprevalence data also allow a comparison between the frequency of natural infection and the frequency of this virus in individuals with infections [30] . However, previous seroepidemiological studies of HBoVs have mainly focused on HBoV1. HBoV1 specific IgG antibodies were frequently detected in children, with a seropositive rate ranging from 40.7%-60% for children , four years old and up to .85% for those $ four years old [19] . The seropositive rate of HBoV1 VP2-specific IgG antibodies is about 94% in healthy adults [18] . However, Kantola et al. recently showed cross-reactivity between the VP2 VLPs of HBoV1-4, which can largely affect the seropositive data of HBoV species. They showed that after depletion of cross-reactive antibodies, the approximate seroprevalences of HBoV1-4 in adults were 59%, 34%, 15%, and 2%, respectively [22] . However, the samples used in that study were only collected from healthy young people with a narrow range of age, including 115 subjects aged 21-32 years from Finland and 80 subjects aged 18-20 years from Pakistan. Given that the prevalence of a virus infection may vary by age as well as geographically, there remains to be a need to estimate the seroprevalence of HBoVs based on the detection of antibodies against HBoV1-4 from data of a more complete age range and from other global, geographic regions. In the present study, we used a competition ELISA (cELISA) assay to estimate the seroprevalence of HBoV1-4 in healthy Chinese individuals ranging in age from 0 to 70 years in Beijing, China. We also compared the seroconversion of anti-HBoV IgG antibodies in 31 paired serum samples from ARTI children who were positive for HBoV1 by PCR. Our findings provide informative data for evaluating the prevalence and pathologic roles of HBoVs. To produce antigens that can be used to evaluate the seroprevalence of HBoVs, the VP2 genes of HBoV1-4 were expressed in baculovirus to generate VLPs. The VP2 genes of HBoV1, HBoV2, and HBoV3 were amplified from stool samples by PCR. The sequences of these genes were verified by phylogenetic analysis ( Figure 1A ). The VP2 genes from respiratory and stool specimens were very similar in sequences. The HBoV1 strain 111-BJ07 used in this study clustered with the reference HBoV1 strain ST2 [1] , with homologies of 99.8%. The HBoV2 strain 211-BJ07 has 97.9% identity with the reference HBoV2 strain PK-2255 [7] . The HBoV3 strain 46-BJ07 has 99.6% identity with reference HBoV3 strain W471 [8] . As we did not find any HBoV4 positive samples, the VP2 gene of HBoV4 was synthesized according to the sequence of HBoV4 strain NI-385 [9] . The VP2 proteins of HBoV1-4 were expressed in insect cells using a baculovirus expression system. The generation of HBoV VLPs was verified by ultracentrifuge-purification and electron microscopy (EM). Typical parvovirus-like particles of 22-24 nm in size, similar to the particles of infectious parvovirus virions, were visualized under EM (data not shown). The purified VLPs of HBoV1, 2, 3 and 4 were also reactive to murine antisera specific for the VP2 protein of the respective HBoV species in Western blot assays ( Figure 1B ). Sequence alignment showed a high amino acid identity of VP2 between HBoV species, with 77.5% identity between HBoV1 and HBoV2, 77.7% identity between HBoV1 and HBoV3, 77.5% identity between HBoV1 and HBoV4, 89.4% identity between HBoV2 and HBoV3, 88.5% identity between HBoV2 and HBoV4, and 90.9% identity between HBoV3 and HBoV4 (data not shown), indicating possible cross-reactivity between HBoV species. To evaluate the potential cross-reactivities, we examined the reactivity between the HBoV1-4 VLPs and purified mouse antisera against HBoV1, 2, 3, and 4 VP2 using Western blot and ELISA assays. Western blot analysis showed that the antisera against HBoV1, 2, 3, and 4 reacted with 400 ng of VLPs of each HBoV species, as indicated by detection of specific bands of 60 kD in size ( Figure 2A) . Similar cross-reactivities were also detected by ELISA assays ( Figure 2B ). Mouse sera against HBoV1, 2, 3, and 4 reacted strongly with the homologous VLPs. Moreover, all four antisera reacted with the heterologous HBoV VLPs when the concentration of the mice antiserum was high (.0.25 mg/mL). No significant reactivity was observed with the human parvovirus B19 VP2 or influenza virus H5 hemaglutinin (H5) antiserum ( Figure 2B ). These results are consistent with those reported by Kantola et al [22] . To develop ELISA assays that can detect antibodies against HBoV species, we first sought to identify HBoV-positive andnegative sera samples using Western blot analysis. To obtain negative sera samples, we tested the reactivity of sera samples collected from infants aged 0-6 months who visited Beijing Children's Hospital for regular health check-ups against HBoV VLPs. Only the samples that were negative for HBoV1, 2, 3, and 4 VLPs simultaneously were selected as the negative samples. The negativity of those samples was confirmed using a competition ELISA (cELISA) assay, in which the absorbance value at 450 nm did not change significantly with or without competition assays ( Figure 3 ). Serum samples identified by cELISA as positive for individual HBoV species were obtained from 3 children and 4 adults and used as positive controls. However, we did not obtain a serum sample that was positive only for HBoV4 based on the assay; we used a sample from an adult that was IgG positive for HBoV1, HBoV2 and HBoV4. To determine the concentration of VLPs required for exhaustive antibody competition of HBoV1-4, the serum samples were monocompeted with concentrations of HBoV1, 2, 3, or 4 VLPs ranging from 0-32 mg/mL ( Figure 3 ). We found that HBoV VLPs concentration of 16 mg/mL was the effective concentration to perform the VLP-based competition assays. To determine the seroprevalence of HBoV1, 2, 3, and 4 in humans, we developed an ELISA protocol for detecting IgG antibodies against HBoV1-4 using VLPs as coating antigens. The concentration of the coated VLPs (0.125 mg/mL) for this ELISA assay was optimized using chessboard titration tests. The positive sera and negative sera showed different results at 1:200 serum dilutions. As the results obtained using lower serum dilutions were similar to those obtained at 1:200 (data not shown), the 1:200 dilution was used in subsequent ELISA analysis of serum samples. To determine the cut-off values for the ELISA, we determined the mean values and standard deviation of negative sera for HBoV1, 2, 3, and 4 using HBoV VLP ELISA at dilutions of 1:200. We used the mean absorbance at 450 nm of the negative sera plus threefolds the standard deviation as the cut-off values, as previously described [19, 31] . For HBoV1, 2, 3 and 4, the cut-off values were 0.344, 0.304, 0.321, and 0.31, respectively. A sample was considered positive for HBoV1, 2, 3, or 4 if its absorbance at 450 nm was above the cut-off value of the respective species in ELISA. As there are cross-reactivities between HBoV species in ELISA assay, we developed a cELISA assay to evaluate the seroprevalence of each HBoV species. We evaluated the specificity of this protocol using heterologous competition assays with parvovirus B19 and human parvovirus 4 (PARV4) VP2. Neither human parvovirus B19 nor PARV4 VP2 inhibited the reactivity of IgG against HBoV1, 2, 3, or 4 in the serum from adults or children ( Figure 3 ). These results suggest that there is no antigenic crossreactivity between HBoVs VP2 and human parvovirus B19 VP2 or between HBoVs VP2 and PARV4 VP2. To confirm the specificity of this ELISA protocol, we tested sera from 15 adults that were positive for human parvovirus B19. In the HBoV VLP ELISA, we did not detect false-positive signals for HBoVs (data not shown). To determine the seroprevalence of HBoV1-4 in adults, we used the HBoV VLPs ELISA method to detect IgG antibodies against HBoV1-4 in 142 serum samples collected from healthy individuals 15-70 years old. Without competition, more than 90% samples were positive for HBoV1, HBoV2, and HBoV3; 73 (51.4%) samples were positive for HBoV4 (Table 1) . However, these IgG seroprevalences decreased with competition by VLPs of heterologous HBoV species (Figure 4 ). The cELISA resulted in seroprevalences of 66.9%, 49.3%, 38.7%, and 1.4%, for HBoV1, 2, 3, and 4 IgG, respectively. The seropositive rate of HBoV1 was higher than that of HBoV2, HBoV3, and HBoV4 (x 2 = 23,P,0.01) (Table 1 and Figure 5 ). Twenty-five of the 142 (17.6%) adult samples that were positive for IgG antibody against HBoV1, 2, and 3, and 17 (12%) that were positive for HBoV4 in the ELISA were negative for IgG against all 4 HBoVs in the cELISA. Among positive serum samples against single HBoV species, there were 26 for HBoV1, eight for HBoV2, and two for HBoV3 based on the cELISA results. The seropositive rates of HBoV1, 2, and 3 in children increased with age according to results from the ELISA without competition. The rates for HBoV1, 2, and 3 reached to 77.9%, 61.1%, and 58.9%, respectively, in the age group 5-14 years. The seroprevalence for HBoV4 among children aged 0-0.5 year and 5-14 years increased from 10.3% to 31.6% (Table 1) . However, according to results of the cELISA, the seroprevalences of HBoV1, 2, 3, and 4 IgG decreased about 10-27% (Table 1, Figure 5 ). The seroprevalence among children aged 0-0.5 years and children aged 5-14 years increased from 6.9% to 60%, 6.9% to 49.5%, and 6.9% to 38.9% for HBoV1, HBoV2, and HBoV3, respectively, in the cELISA. Of note, among 29 infants aged 0-6 months, two (6.9%) were positive for HBoV1, 2, and 3 IgG, and one (3.4%) was positive for HBoV4. The detection of HBoVs IgG antibodies among children aged 0-6 months may be due to the presence of maternal antibodies. However, the seropositive rate of HBoV1 was higher than that of HBoV2, 3, or 4 in the age groups of 0.5-2 years, 2-5 years, and 5-14 years (x 2 = 10.1,P,0.01; x 2 = 9.3, P,0.01; x 2 = 8.42,P,0.05, respectively). Based on the results of the cELISA, there were 38 (15.6%) individuals positive for HBoV1, nine (3.7%) for HBoV2, and two (0.8%) for HBoV3 among positive serum samples against single HBoV species. Notably, the two samples positive for HBoV4 IgG were also positive for HBoV1, 2, and 3 IgG. To characterize the antibody response after HBoV infection further, we measured the IgG antibody in 31 pairs of sera samples that were collected from children with ARTIs. All these children were positive for HBoV1 according to PCR analysis. Based on cELISA, the seroprevalences of HBoV1, 2, 3, and 4 in acute-phase sera were 29%, 25.8%, 16.1%, and 0%, respectively. Of the 31 samples, ten (32.3%) showed an IgG seroconversion for HBoV1, eight (25.8%) for HBoV2, two (6.5%) for HBoV3, and none for HBoV4 (Table2). Two pairs of sera showed concurrent seroconversions to HBoV1 and HBoV2, and one pair showed concurrent seroconversions to HBoV1, HBoV2, and HBoV3. In addition, absorbance at 450 nm increasing in the convalescent-phase serum was observed in seven children (22.6%) who were positive for HBoV1 IgG and eight (25.8%) children who were positive for HBoV2 IgG positive at the acute-phase. In this study, we evaluated the cross-reactivity of mouse antisera against HBoV1, 2, 3, and 4 VP2 with VP2 VLPs for four species of human bocaviruses. Considerable cross-reactivity was found among the four HBoV species. However, human IgG antibodies against HBoV did not cross react with human parvovirus B19 and PARV4. These findings are consistent with those of a previous report [22] , suggesting that to obtain accurate data of the seroprevalences of the different HBoV species, it is necessary to correct for cross-reactivity between the four HBoVs. Comparison of the VP2 protein sequences of HBoV1-4 revealed a high degree of similarity. These similarities may account for the cross-reactivity of these viruses, and should be confirmed through epitope analysis. Human parvovirus B19 and PARV4 have no antigen crossreactivity with HBoVs, suggesting that human parvovirus B19 and PARV4 antibodies do not interfere with the HBoV seroprevalence results. The specificity of human bocavirus VLP-based ELISA with human sera has also been demonstrated with human parvovirus B19 VP2 and PARV 4 VP2 by Kantola et al [22] . We used a VLP-based ELISA to assess the seroprevalence of HBoV1-4 in healthy individuals in Beijing, China, ranging from 0 to 70 years old. Our results show that the seroprevalences of HBoV1, 2, and 3 range from 40.3-67.8% in children 0.5-2 years old and are up to 100% in adults. To eliminate the interference of antibody cross-reactivity, the results were corrected with a cE-LISA. The results of our cELISA suggest that the seroprevalences of HBoVs are significantly lower than those obtained without competition, especially in adults. These findings further indicate a high degree of antigenic cross-reactivity between HBoV1-4. Our findings are consistent with those of a recent study of Finish and Pakistani individuals, in which the seroprevalences of HBoVs decreased after depletion of heterologous HBoV reactive antibodies [22] . Our results from the cELISA are lower than those previously reported for HBoV1 in adults and children [18, 19, 32] , where HBoV1 antibodies against HBoV1 VP2 were detected by ELISA. Hence, the seroprevalence of HBoVs may be overestimated due to the serological cross-reactivity among the four HBoV species in previous studies. However, according to our cELISA results, the seroprevalence of HBoV1-3 in Beijing adults is somewhat higher than that reported in Finnish and Pakistani adults [22] . This disparity may be attributed to the difference of geographical location [22] . Our age stratification data indicates that HBoVs circulate widely in the human population, and the primary infection with HBoV1-3 occurs in children aged six months and older after the maternal antibodies have waned. The seropositive rates of IgG antibodies against HBoV2, 3 and 4 were lower than that of HBoV1 in individuals of $0.5 years old. These data indicate that HBoV1 is the predominant circulating species of HBoV in Beijing, whereas HBoV2-4 do not seem to have a major impact on HBoV infections. This finding agrees with reports on HBoVs prevalence obtained from most parts of the world using DNA analysis [10, 14, 16] . Of note, our finding that the seroprevalence of HBoV4 is much lower than that of HBoV1-3 is consistent with the rare detection of HBoV4 DNA in clinical samples [9] . Overall, the seroprevalence of HBoVs indicated by results of the cELISA decreased less in children than in adults in comparison to the results without competition, indicating that HBoV infections are more specific in children, especially for HBoV1 [22] . These results suggest that HBoV species may play differential roles in disease. Our results show that 17.6% of the adults who were positive for IgG against HBoV1, 2, 3, or 4 based on the ELISA results were negative for IgG against all four HBoVs based on the cELISA results. This finding may be due to antibody waning in the individuals with low IgG levels [22] or to low-affinity antibodies [20, 33] . To assess the antibody response of HBoV1-4 IgG in children with acute respiratory tract infections, we used cELISA to determine the seroconversion rate of 31 paired sera from children who were positive for HBoV1 according to PCR analysis. We found that the HBoV1-specific seroprevalence was higher in convalescent sera than in acute sera. However, heterotypic seroconversion against HBoV2 and HBoV3 was also observed. Moreover, for some patients who were positive for HBoV1 and HBoV2 IgG at the acute phase, the absorbance value was higher in convalescent sera. These data may suggest the concurrent production of antibodies against HBoV2 and 3 during the infection of HBoV1, as it has been shown that B-cell memory can be boosted either by the homologous virus or by heterologous, yet immunological related virus [20, 34, 35] . Overall, our findings suggest a differential prevalence of HBoV species in healthy individuals aged 0-70 years old. HBoV1 appears to be the dominant species responsible for HBoV infections among HBoV1, 2, 3, and 4 in Beijing, China. The seroprevalence of HBoV1-3 increased with age in children. Our study provides a basis for future evaluation of the epidemiology, genotype distribution, and pathogenesis of HBoVs worldwide. Serum specimens were collected from 386 healthy individuals aged 0 to 70 years in 2008; 244 specimens were from infants and children who visited Beijing Children's Hospital for regular health check-ups. The 142 specimens from adults were provided by the Beijing Blood Center. Exclusion criteria for all subjects included pregnancy, any abnormalities in renal and liver function tests, HIV/AIDS, sexual transmitted diseases, tumor, recurrent or acute infection, and medication. None of the subjects had any respiratory infection for at least three months prior to the blood samples being taken. In addition, paired acute-phase (at the time of admission) and convalescent -phase (2 weeks after disease onset) serum samples were collected from 31 children (median age 17 months; range of 1 month to 9 years) with acute lower respiratory tract infections (ALRTIs) when they were hospitalized at the Beijing Children's Hospital. The DNA of HBoV1, but not HBoV2-4, was detected in the nasopharyngeal aspirates of these 31 patients at the time of admission by nested PCR and sequence analysis using primers targeting the viral proteins (VP) 1/2 region [9] . All serum samples were stored at 280uC prior to use. Written informed consent was obtained from all participants or guardians on behalf of children. This study was approved by the ethical review committee of the Institute of Pathogen Biology, Chinese Academy of Medical Sciences. The full-length VP2 genes of HBoV1-4 were used for viruslike particle (VLP) production in this study. HBoV1-3 VP2 genes were amplified from HBoV-positive stool specimens (111-BJ07, 211-BJ07, and 46-BJ07; GenBank accession numbers: JQ240469, JQ240470 and HM132056) [15] . HBoV1 is 1,629 bp (nt 3,373-5,001 according to HBoV1 strain ST1, GenBank accession number DQ000495), HBoV2 is 1,617 bp (nt 3,306-4,922 according to HBoV2 strain PK2255, GenBank accession number FJ170279), and HBoV3 is 1,620 bp (nt 3,410-5,029 according to HBoV3 strain W471, GenBank accession number EU918736) in length. The VP2 genes of HBoV4 (1626 bp in length, nt3331-4956 based on HBoV4 strain NI-385, GenBank accession number NC012729) were synthesized by Sangon Biotech (Shanghai, China). These genes were verified by phylogenetic analysis using the Clustal W and MegAlign programs in the MEGA 4.0 software package [36] . The phylogenetic tree with 1,000 bootstrap replicates was generated based on the complete sequences of the VP2 genes used in this study and reference sequences from GenBank. HBoV1 strains ST1, ST2, and TW2888_06; HBoV2 strains W153, PK-2255, LZ55602, and 277-BJ07; HBoV3 strains W471 and W855; and HBoV4 strains NI-385 were used as reference sequences (GenBank accession numbers DQ000495, DQ000496, EU984237, EU082213, FJ170279, GU301645, JQ240471, EU918736, FJ948861, NC012729, respectively). The full-length VP2 genes of HBoV1, 2, 3, and 4 were cloned into the baculovirus expression vector pFastbac1 and expressed using Bac-to-BacH Baculovirus Expression System (Invitrogen, Carlsbad, CA), according to the manufacturer's protocol. VLPs were obtained as previously described [17, 37] . High Five cells (Invitrogen) were infected with recombinant baculoviruses at a multiplicity of infection (MOI) of five and harvested after three days. Cells were suspended in 25 mM NaHCO 3 solution at 2610 7 cells/mL and kept on ice for 30 min. After centrifugation at 18,000 rpm for 10 min at 4uC, (NH4) 2 SO 4 was added to the supernatants at a final concentration of 20% (w/v). Precipitants were harvested by centrifugation and dissolved in CsCl solutions with densities of 1.4 g/mL in Tris-EDTA buffer (10 mmol/L Tris, pH 8.7, 1 mmol/L EDTA, and 0.5% Triton X-100). After centrifugation at 35,000 rpm and 18uC for 39 h in a SW41 rotor centrifuge (Beckman Coulter, Fullerton, CA), the fractions were analyzed by SDS-PAGE and Western blot using mouse sera against HBoV VP2 proteins. A Tecnai12 transmission electron microscope (FEI, Hillsboro, OR) at 80 kV was used to verify the morphology of the VLPs. SDS-PAGE and Western blot analysis were used to identify the VLPs [18] . Recombinant VP2 proteins of HBoV1, 2, 3 and 4 were also expressed in E. coli Rosetta (DE3) (Novagen, Madison,WI,) cells and purified as previously described [37] . The genes used in prokaryotic cloning were the same as those used in a baculovirus expression system. The recombinant antigens were used to immunize mice to produce antibodies (see below). As controls, human parvovirus B19 VP2 was expressed, as described previously [38] , using the construct pFastBac1B19VP2 provided by Dr. Xiaohui Zou at the National Institute for Viral Disease Control and Prevention, Chinese Center for Disease control and Prevention. Additionally, the PARV4 VP2 gene (1,659 bp in length, nt 3,464-5,122 based on the NC_007018 reference sequence) was synthesized by Sangon Biotech (Shanghai, China) and cloned into baculovirus vector pFBGP67-His [39] . The recombinant baculoviruses were generated in Sf9 cells using the Bac-to-BacH Baculovirus Expression System (Invitrogen) protocol provided by the manufacturer. High Five cells were infected with recombinant baculovirus expressing PARV4 VP2 gene at a MOI of five. The infected cells were collected three days post-infection and purified using a HisTrap HP 1 ml column (GE Healthcare, Waukesha, WI). The concentrations of all purified protein were determined using the Pierce BCA Protein Assay Kit (Thermo Scientific, Rockford, IL) and stored at 280uC prior to use. ELISA was used to determine the anti-HBoV antibodies, as described elsewhere [40] . The purified HBoV VLPs were used as coating antigen (0.125 mg/mL). The absorbance of each serum sample was read at 450 nm and the mean values of the duplicate samples were calculated. Sera pooled from ten samples showing HBoV-specific IgG responses were used as the internal references in all experiments. To minimize false positive results of the ELISA assay due to impurities in immunizing and coating antigens, the coating antigen and the protein used to prepare mouse antibodies were derived from a baculovirus expression system and a prokaryotic expression system. BALB/c mice were injected subcutaneously with the purified proteins obtained from E. coli. This study was carried out in accordance with the animal experiment regulations of the Chinese government. All animal experiments were performed in the facilities of the Institute of Laboratory Animal Sciences (ILAS), Chinese Academy of Medical Sciences (CAMS). All experimental procedures were approved (license number SCXKJ2009-0017) and supervised by the Animal Protection and Usage Committee of ILAS, CAMS. Sera collected from the treated mice were purified using protein-G sepharose columns (GE Healthcare, Waukesha, WI) and purified IgG antibody were quantified using Pierce BCA Protein Assay Kit (Thermo Scientific). The purified mouse IgG antibodies were serially diluted from 8 mg/mL to 0.008 mg/mL. Sera from pre-immune mice served as the negative control. Rabbit antiserum against human parvovirus B19 (a gift from Dr. Xia Xiao of National Institute for Viral Disease Control and Prevention, Chinese Center for Disease control and Prevention) and mice antiserum against influenza virus H5 hemaglutinin (HA) [39] were used as unrelated controls. The cross-reactivity between HBoV1, 2, 3, and 4 was tested using Western blot analysis and ELISA. To measure antibodies specific to VP2 antigen of an individual HBoV species (HBoV1, 2, 3, or 4), antibodies in serum samples were absorbed with heterologous VLPs of the other three HBoVs, as previously described [22] . Briefly, HBoV VLPs were serially diluted from 32 mg/mL to 0.5 mg/mL to determine the concentration needed for effective competition between cross-reactive antibodies. For detection of specific HBoV antibodies, three heterologous HBoV VLPs of 16 mg/mL were added to a 1:200 dilution of plasma [17] and incubated for 2 hr at 4uC prior to performing the ELISA assay. In parallel with the heterologous competition, human sera samples were used to compete with VLP that was homologous to the immobilized antigen. The absorbance at 450 nm was residual absorbance value. Net absorbance values were calculated by subtracting the residual absorbance value from the raw absorbance value read at 450 nm [22] . Seropositive rates were evaluated using x 2 tests. A P value #0.05 was considered significant.

Clarithromycin Suppresses Human Respiratory Syncytial Virus Infection-Induced Streptococcus pneumoniae Adhesion and Cytokine Production in a Pulmonary Epithelial Cell Line

Human respiratory syncytial virus (RSV) sometimes causes acute and severe lower respiratory tract illness in infants and young children. RSV strongly upregulates proinflammatory cytokines and the platelet-activating factor (PAF) receptor, which is a receptor for Streptococcus pneumoniae, in the pulmonary epithelial cell line A549. Clarithromycin (CAM), which is an antimicrobial agent and is also known as an immunomodulator, significantly suppressed RSV-induced production of interleukin-6, interleukin-8, and regulated on activation, normal T-cell expressed and secreted (RANTES). CAM also suppressed RSV-induced PAF receptor expression and adhesion of fluorescein-labeled S. pneumoniae cells to A549 cells. The RSV-induced S. pneumoniae adhesion was thought to be mediated by the host cell's PAF receptor. CAM, which exhibits antimicrobial and immunomodulatory activities, was found in this study to suppress the RSV-induced adhesion of respiratory disease-causing bacteria, S. pneumoniae, to host cells. Thus, CAM might suppress immunological disorders and prevent secondary bacterial infections during RSV infection.

Human respiratory syncytial virus (RSV) is one of the most important infectious agents causing acute lower respiratory tract illness, such as bronchiolitis and pneumonia, in infants and young children [1, 2] . Viral RNA generated during RSV replication is recognized by host pattern recognition molecules, such as Toll-like receptor 3 (TLR3) and retinoic acid inducible gene-I (RIG-I), and it induces type I and type III interferon [3, 4] . Transcriptional induction of proinflammatory cytokines, chemokines, and interferons is mediated by NF-κB and interferon regulatory factors (IRFs) [5, 6] . These mediators are believed to contribute to the pathophysiology of RSV infection, such as mucous hypersecretion, swelling of submucous, and infiltration of lymphocytes, neutrophils, eosinophils, and macrophages [7] . Frequently, there are coinfections with respiratory viruses, including RSV, and bacteria that cause communityacquired respiratory diseases, such as Streptococcus pneumoniae and Haemophilus influenzae. There is evidence for a positive correlation between infections with S. pneumoniae and RSV in the pathogenesis of otitis media, pneumonia, and meningitis [8] [9] [10] [11] . S. pneumoniae and H. influenzae colonize to the host respiratory epithelium via host cell surface receptors, such as the platelet-activating factor (PAF) receptor [12] [13] [14] . These bacteria interact with the PAF receptor via phosphorylcholine, which is a component of the bacterial cell surface. Both live and heat-killed S. pneumoniae cells show an increased adhesion to human epithelial cells infected with RSV [15] . The upregulation of PAF receptor expression that is induced by respiratory virus infections, including those caused by RSV, results in the enhanced adhesion of S. pneumoniae and H. influenzae to respiratory epithelial cells [15] [16] [17] . PAF receptor expression and S. pneumoniae cell adhesion are also upregulated by exposure to acid, which causes tissue injury and an inflammatory response [18] . Clarithromycin (CAM) is 14-membered ring macrolide antibiotic that also acts as a biological reaction modifier with anti-inflammatory properties. In Japan, CAM is applied to diffuse panbronchiolitis, chronic bronchiolitis, otitis media, and chronic sinusitis as an immunomodulator [19] [20] [21] . The anti-inflammatory mechanism of CAM has not yet been completely clarified, but one of the important mechanisms for its anti-inflammatory action is considered to be the suppression of NF-κB [22] [23] [24] . Recently, we reported that fosfomycin, which is an antibiotic, suppressed RSV-induced interleukin (IL)-8, regulated on activation, normal T-cell expressed and secreted (RANTES), and the PAF receptor by suppressing NF-κB activity [25, 26] . On the other hand, Wang et al. report that CAM suppressed rhinovirus-induced Staphylococcus aureus and H. influenzae adhesions to nasal epithelial cells [27] . So we anticipate that CAM suppresses RSV-induced bacterial adhesion to epithelial cells, because expression of PAF receptor is controlled by NF-κB [28, 29] In the present study, we examined the effect of CAM on cytokine production, PAF receptor expression, and RSV infection-induced S. pneumoniae adhesion to respiratory epithelial cells. Lines, Bacteria, and Reagents. RSV strain Long, human type II pulmonary epithelial cell line A549 and S. pneumoniae strain R6 were obtained from the American Type Culture Collection (ATCC, Manassas, VA). RSV was grown in HEp-2 cells. The virus titer of RSV was determined using a plaque-forming assay with HEp-2 cells as the indicator cells [25] . RSV infection to A549 cells was performed at multiplicity of infection (MOI) of 1. CAM was donated by Abbott Japan (Tokyo, Japan). A PAF receptor antagonist, 1-O-hexadecyl-2-acetyl-sn-glycero-3-phospho(N,N,N,trimethyl)-hexanolamine, was purchased from Calbiochem-Merck (Darmstadt, Germany). An NF-κB inhibitor, pyrrolidine dithiocarbamate (PDTC), was purchased from Sigma-Aldrich (St. Louis, MO). Production. A549 cells were infected with RSV at MOI of 1. After 24-hour infection, culture supernatants of RSV-infected and -uninfected cells were collected. The amounts of IL-6, IL-8, and RANTES in the culture supernatants were determined by enzyme-linked immunosorbent assay (ELISA) (DuoSet ELISA development kit, R&D systems, Minneapolis, MN). (RT-PCR). Semiquantitative RT-PCR was carried out as described previously [4, 30] . The cell surface expression of the PAF receptor was examined by flow cytometry as previously described [26] . The cells were harvested from culture flasks using a cell scraper and then incubated with 2.5 μg/mL of mouse anti-PAF receptor monoclonal antibody (11A4 (clone 21); Cayman Chemical, Ann Arbor, MI) or mouse IgG2a, κ isotype control antibody (eBioscience, San Diego, CA). After incubation at 4 • C for 30 min, cells were collected by centrifugation and washed once with Dulbecco's phosphate-buffered saline (PBS (−)). Cell suspensions were incubated with a phycoerythrin-conjugated goat anti-mouse IgG F(ab) 2 fragment antibody (1 : 100 dilution) (Abcam, Cambridge, UK) at 4 • C for 30 min, and the stained cells were assessed with FACSCalibur (BD Bioscience, San Jose, CA). Assay. S. pneumoniae adhesion was assayed using fluorescein-isothiocyanate-(FITC-) labeled S. pneumoniae as previously described [26] . Briefly, a bacterial suspension in 0.1 M NaCl-50 mM sodium carbonate buffer (pH9.5) at 1 × 10 8 CFU/mL was prepared. FITC isomer-I (Dojindo Laboratories, Kumamoto, Japan) was added at a concentration of 1 mg/mL, and the mixture was incubated at 4 • C for 1 h. The cells were washed three times with PBS (−). CAM was added to monolayers of A549 cells 1 h prior to RSV infection. The A549 cells infected with RSV at an MOI of 1 for 24 h and uninfected A549 cells were incubated with FITC-labeled S. pneumoniae cells at MOI of 10 for 30 min at 37 • C. For the control experiments, either 20 μg/mL of the PAF receptor antagonist or 10 μg/mL of the mouse anti-PAF receptor monoclonal antibody (11A4(clone 21)) was added to the A549 cells 1 h prior to the addition of the FITC-labeled bacteria. The cell monolayer was gently washed three times with PBS (−) and observed by fluorescence microscopy. Alternatively, the cells were harvested with cell scraper and then assessed by flow cytometry as previously described [26] . First, we examined the effect of CAM on RSV replication in A549 cells. RSV infection to A549 cells was performed at MOI of 1. After 24 and 36 h of infection, significant alterations of the RSV titers or expression levels of G mRNA were not observed by the addition of CAM even at a concentration of 100 μg/mL (Figure 1 ). When A549 cells were infected with RSV at MOI of 1, RANTES, IL-8, and IL-6 were markedly induced. These cytokine inductions were significantly suppressed in the presence of CAM in a dose-dependent manner ( Figure 2 ). The degree of suppression by CAM was less than that by an NF-κB inhibitor, PDTC. PAF receptor expression on the cell surface is upregulated during RSV infection in A549 cells [26] . The RSV-induced upregulation of the PAF receptor was significantly suppressed by CAM and PDTC in a dose-dependent manner ( Figure 3 ). The degree of suppression by CAM was slightly less than that by PDTC. Suppression of the PAF receptor expression was also observed when A549 cells were posttreated with CAM (4 or 12 h after RSV infection) (data not shown). We examined the adhesion of FITC-labeled S. pneumoniae cells to A549 cells by fluorescence microscopy (Figure 4 ) and flow cytometry ( Figure 5 ). RSV infection significantly enhanced the adhesion of S. pneumoniae to A549 cells, and this enhancement was suppressed by adding a PAF receptor antagonist (Figures 4 and 5) or anti-PAF receptor monoclonal antibody (data not shown). This result indicated that the RSV-induced S. pneumoniae adhesion occurs via the PAF receptor on A549 cells. The bacterial adhesion was significantly suppressed by CAM, as well as PDTC. These lines of evidence confirmed that the expression of the PAF receptor was induced by RSV infection and indicated that this induction, and subsequent RSV-induced S. pneumoniae adhesion, can be suppressed by CAM treatment. Macrolides, with the exception of the 16-membered ring type, have both anti-inflammatory and antibacterial functions [20, 21] . One of the important mechanisms of antiinflammatory action is the suppression of NF-κB activation [22] [23] [24] . Our recent studies show that RSV upregulates proinflammatory cytokines, such as IL-6, and chemokines, such as IL-8 and RANTES, in the respiratory epithelial cell line A549. Furthermore, the induction of chemokines by RSV is significantly suppressed by an antibiotic, fosfomycin, via suppression of NF-κB activation [25] . In the present study, CAM was shown to suppress IL-6, IL-8, and RANTES, which are induced by RSV infection, at concentrations of 10 and 100 μg/mL. Patel et al. reported that the concentration of CAM in fluid of the bronchopulmonary epithelial lining was 34.2±5.16 μg/mL at 4 h, 23.01±11.9 μg/mL at 12 h in healthy adults orally administered CAM 500 mg [31] . We observed that CAM did not affect RSV replication even at a concentration of 100 μg/mL. However, it is reported that respiratory virus, such as RSV [32] , rhinovirus [33, 34] , and influenza virus [35] , replication is suppressed by 14-membered ring macrolides, including CAM. The reasons of contradictory results between the report of Asada et al. [32] and our present study have been unclear. These two studies used different types of epithelial cells and different experimental conditions of RSV infection. Asada et al. used primary human tracheal epithelial cells, and in contrast we used A549 cell line. Asada et al. carry out infection at a lower titer of RSV (10 −3 TCID 50 /cell) and measuring virus titer at a longer period (3-5 days) after infection. Our results indicated that suppression of the RSV-induced cytokines by CAM was not caused by the amount of replicated RSV. In other words, CAM was suggested to have suppressive activity of cytokine production independent of viral replication. Both IL-8 and RANTES, which are strongly upregulated during RSV infection, play important roles in pathogenesis [36, 37] . IL-8 primarily activates neutrophils and promotes their migration. RANTES is secreted from respiratory epithelial cells and promotes migration of eosinophils, basophils, monocytes, and neutrophils. In particular, RANTES is an efficient eosinophil chemoattractant involved in the pathogenesis of asthma [38] . CAM has been suggested to suppress the inflammatory disorders induced by RSV. In the present study, we also observed that CAM suppressed enhanced S. pneumoniae adhesion by RSV infection in A549 cells. The RSV-induced S. pneumoniae adhesion was mainly mediated by host PAF receptor, as indicated by that suppressed by the PAF receptor antagonist and anti-PAF receptor monoclonal antibody. The PAF receptor acts as a receptor for S. pneumoniae and H. influenzae [12] [13] [14] . Transcription of the PAF receptor gene is controlled by NF-κB [28, 29] . We confirmed it by that the RSV-induced PAF receptor expression and S. pneumoniae adhesion were suppressed by an NF-κB inhibitor, PDTC. We revealed that CAM also suppressed PAF receptor expression induced by RSV infection and S. pneumoniae adhesion to RSV-infected A549 cells. It should be caused by the suppression of by inhibiting PAF receptor expression. CAM showed more potent suppression of RSV-induced S. pneumoniae adhesion and production of proinflammatory cytokines and chemokines than fosfomycin, as we reported previously [25, 26] . Notably, CAM significantly suppressed RSV-induced IL-6 production, whereas fosfomycin did not significantly [25] . This finding may be caused by that CAM is more potent than fosfomycin; however, the actual reason for this disparity is not clear. The upregulation of PAF receptor expression and the enhanced adhesion of pathogenic bacteria, such as S. pneumoniae, to respiratory epithelial cells are considered to be a major risk factor for secondary bacterial infections after primary respiratory viral infections. CAM may suppress both secondary bacterial infections and immunological disorders induced by RSV, without suppressing viral replication. Infection with other respiratory viruses, such as human parainfluenza virus 3 [16] and rhinovirus [17] , also upregulates known receptors for the pathogenic bacteria, including PAF receptor and S. pneumoniae adhesion. On the other hand, influenza virus does not upregulate the known receptors for bacteria, whereas bacterial adhesion is increased by the infection [16] . McCullers [39] reported that influenza-induced bacterial adhesion to A549 cells was not inhibited by PAF receptor antagonist, and the PAF receptor knock-out mice did not show lower susceptibility to experimental secondary pneumonia caused by S. pneunimoae following influenza infection compared to the parent mice. Lines of evidence suggest that adherent inducing mechanisms of S. pneumoniae to host respiratory epithelial cells are varied among viruses. So CAM may not always suppress virus-induced pathogenic bacteria adhesion. We proposed that clarithromycin efficiently suppressed PAF receptor-mediated Streptococcus pneumoniae adhesion to respiratory epithelial cells as well as RSV-induced proinflammatory cytokine and chemokine production. Clarithromycin may suppress secondary bacterial infections and immunological disorders during RSV infection.

Post-Transcriptional Control of Type I Interferon Induction by Porcine Reproductive and Respiratory Syndrome Virus in Its Natural Host Cells

Porcine reproductive and respiratory syndrome virus (PRRSV) is not only a poor inducer of type I interferon but also inhibits the efficient induction of type I interferon by porcine transmissible gastroenteritis virus (TGEV) and synthetic dsRNA molecules, Poly I:C. However, the mechanistic basis by which PRRSV interferes with the induction of type I interferon in its natural host cells remains less well defined. The purposes of this review are to summarize the key findings in supporting the post-transcriptional control of type I interferon in its natural host cells and to propose the possible role of translational control in the regulation of type I interferon induction by PRRSV.

Porcine reproductive and respiratory syndrome virus (PRRSV) is a single-stranded, positive-sense RNA virus with a genome size of approximately 15 kb. PRRSV belongs to the Arteriviridae in the order of Nidovirale. PRRSV causes acute respiratory disease in neonatal and young piglets and reproductive failure in pregnant sows. PRRSV primarily infects and destroys alveolar macrophages during acute infection of swine. In addition to macrophages, PRRSV has also been identified by immunohistochemistry in dendritic-like cells in tonsils and lymph nodes [1] , suggesting that dendritic cells may either be susceptible to PRRSV or capture PRRSV by taking up apoptotic infected cells. tumor necrosis factor- (TNF-α), are absolutely dependent on activation of PI3K [9] . Although the nuclear factor kappa B (NF-B) pathway has been shown to contribute to the transcriptional activation of type I interferon by PRRSV [15] , one study shows that NF-B is more likely related to induction of inflammatory cytokines such as TNF- and IL-6, rather than type I interferon, after influenza A virus infection and CpG ODN stimulation [9] . This discrepancy may be due to the cell types and viruses used in different studies since different cell types and different viruses exhibit distinct features in the induction of type I interferon pathway. Virus replication and viral infectivity are usually not essential to the induction of type I interferon since both UV-inactivated and heat-inactivated influenza A viruses induce more abundant interferon- than their live virus counterparts [16] . In 1998, Albina et al. first reported the failure of PRRSV in inducing the production of interferon- protein in the lung secretions of infected pigs and in the supernatants of PRRSV-infected alveolar macrophages and peripheral blood mononuclear cells [17] . Furthermore, they observed that PRRSV was also capable of blocking the production of interferon- protein in macrophages by a well-characterized and potent interferon- inducer, swine transmissible gastroenteritis virus (TGEV), a member of the coronaviridae family. Virus replication and infectivity is essential to the inhibition of interferon- production since UV-inactivated PRRSV fails to block the induction of interferon- by TGEV. Interestingly, a recent study reported that PRRSV infectivity is not essential to the inhibition of type I interferon in plasmacytoid dendritic cells, which are resistant to PRRSV infection [4] . The observation that PRRSV is a poor inducer of interferon- is further confirmed by Buddaert et al. [18] . In 2004, Chung et al. reported that PRRSV activated the transcription of interferon- and Mx1 in lung tissues at day 1 and peaked at day 7 after infection in acutely infected animals, suggesting that PRRSV does activate the transcription of interferon- and interferon induced genes such as Mx1 [19] . Others have further reported that different PRRSV isolates exhibit different capacities in inducing the production of interferon- in alveolar macrophages [20] . The activation of interferon- transcription in PRRSV-infected monocyte-derived dendritic cells, peripheral blood mononuclear cells, alveolar macrophages and in porcine alveolar macrophages of PRRSV-infected animals has also been reported [3, 21, 22, 23] . Taken together, the existing evidence clearly indicates that PRRSV activates the transcription of type I interferon in its natural susceptible cells. In MARC-145 cells, however, Miller reported that PRRSV not only failed to activate the transcription of interferon- and , but also suppressed the transcription of interferon- activated by Poly I:C [24] . This suggests a cell type dependent difference in activating the transcription of type I interferon by PRRSV. Marked differences in type I interferon induced antiviral activity against PRRSV between porcine alveolar macrophages and MARC-145 cells have also been reported recently [25] . Since MARC-145 cells are not the natural susceptible cells for PRRSV, the implication of the results in understanding the pathogenesis of PRRSV is debatable. Overall, the existing evidence clearly suggests that PRRSV may have intrinsic properties to inhibit or reduce the induction of interferon- and . To further elucidate the molecular mechanisms by which PRRSV inhibits the induction of interferon- transcription in MARC-145 cells, Luo et al. examined the role of several key molecules in mediating the transcriptional activation of interferon- [26] . They reported that PRRSV interferes with the nuclear translocalization of IRF-3 by inactivating IPS-1, a downstream molecule of the RIG-I pathway [26] . Studies by Beura et al., suggested that the nonstructural proteins of PRRSV, including Nsp1, Nsp2, Nsp11 and Nsp4, when over-expressed in cell culture alone have the ability to antagonize the nuclear translocation of IRF-3 and the promoter activity of interferon- activated by Poly I:C and Sendai virus [27] . Similar studies have to some degree confirmed that over-expressed Nsp1 and Nsp2 of PRRSV in cell culture antagonize the promoter activity of interferon- as determined by using the luciferase reporter gene expression system [15, 28, 29, 30] . Sun et al. provided a detailed review on the role of PRRSV nonstructural and structural proteins in modulating the transcriptional activation of type I interferon in MARC-145 cells and human cell culture system including HeLa cells and 293T cells [12] . Despite the ease and convenience of using the luciferase reporter system to dissect the role of over-expressed viral proteins in regulating the transcription of interferon-, such results are often contradictory to the authentic virus infection of natural susceptible cells in vitro and in vivo [10, 16, 31] . Therefore, the implication of such results in natural virus infection is uncertain. In 2004, Lee et al. first described the discrepancy between interferon- mRNA level and interferon- protein production in PRRSV-infected porcine alveolar macrophages [20] . They suggested that a post-transcriptional regulatory mechanism contributed to the inhibition of interferon- production. We have observed the same phenomenon in PRRSV-infected porcine monocyte-derived dendritic cells [32] . Despite the transient and abundant interferon- and  mRNA molecules, very little or no interferon- protein was detected in either cell lysates or supernatants of PRRSV-infected cells at different time points after virus infection, indicating a post-transcriptional control of type I interferon induction. Currently, very little is known about the translational control of type I interferon production by PRRSV. PRRSV inhibits the PI3K-dependent Akt (PI3K/Akt) pathway during late infection [33] , which makes the translation repressor, 4E-BP1, hyperactive and reduces global protein synthesis. Interestingly, a recent study has demonstrated that PI3K/Akt inhibition can also lead to the phosphorylation of eIF-2α to inhibit cellular translation [34] . We have indeed observed increased phosphorylation of eIF-2α in PRRSV-infected porcine Mo-DC during late infection (unpublished observation). Thus, it is possible that the translation of IFN-α is reduced partly by PI3K/Akt inhibition. However, it is more likely that PRRSV has employed multiple strategies to inhibit cellular protein synthesis as a simple way to evade the host defenses, which may also contribute to the inhibited type I interferon induction. In response to some virus infections, host dsRNA-activated protein kinase (PKR) phosphorylates eIF-2α to inhibit both cellular and viral translation [35, 36, 37] . PRRSV replication is known to trigger the stress-activated proteins kinases to modulate cytokine production in porcine alveolar macrophages [38] . It has also been demonstrated that cleavage of the eukaryotic translation initiation factor 4G (eIF4G) by either viral proteases or cellular proteases such as caspase 3 activated during apoptosis can result in rapid block of cellular protein synthesis [39, 40, 41, 42] . Furthermore, poly(A) tail elongation mediated by viruses may repress the efficient translation of type I interferon [43] . Both UV-inactivated and heat-inactivated influenza A viruses are more potent than wild-type viruses in inducing the production of type I interferon [16] . Cytopathogenecity has been associated with translational shut-down of host genes including interferon [44] . PRRSV is highly pathogenic in alveolar macrophages and monocyte-derived dendritic cells and rapidly destroys these target cells by apoptosis and necrosis [2, 45, 46] . Studies have clearly shown that interferon mRNA has a short life span after induction [43, 47] . We speculate that the combination of high cytopathogenicity of virus and short half-life of interferon mRNAs may at least partially contribute to the low interferon proteins detected in virus-infected cells. It is also possible that the variability of different PRRSV isolates in inducing type I interferon is related to their varied cytopathogenicity [20, 23, 48] . For influenza A virus, different isolates can vary in their ability to induce interferon by up to 100-fold [16] . Finally, it remains to be determined whether PRRSV induces the shutoff of host protein synthesis to favor its own protein synthesis. More research efforts should be directed to the understanding of the translational control of type I interferon by PRRSV in its natural host cells. PRRSV activates the transcription of type I interferon in porcine alveolar macrophages, peripheral blood mononuclear cells, and monocyte-derived dendritic cells. However, PRRSV interferes with the translation of type I interferon in these cells partly through cytopathogenicity since UV and heat-inactivated viruses lose their ability to interfere with the induction of type I interferon by porcine transmissible gastroenteritis virus or Poly I:C. Further studies are needed to delineate the exact mechanisms by which PRRSV interferes with the translation of type I interferon in its natural host cells.

Clinical review: Special populations - critical illness and pregnancy

Critical illness is an uncommon but potentially devastating complication of pregnancy. The majority of pregnancy-related critical care admissions occur postpartum. Antenatally, the pregnant patient is more likely to be admitted with diseases non-specific to pregnancy, such as pneumonia. Pregnancy-specific diseases resulting in ICU admission include obstetric hemorrhage, pre-eclampsia/eclampsia, HELLP (hemolysis, elevated liver enzymes, and low platelet count) syndrome, amniotic fluid embolus syndrome, acute fatty liver of pregnancy, and peripartum cardiomyopathy. Alternatively, critical illness may result from pregnancy-induced worsening of pre-existing diseases (for example, valvular heart disease, myasthenia gravis, and kidney disease). Pregnancy can also predispose women to diseases seen in the non-pregnant population, such as acute respiratory distress syndrome (for example, pneumonia and aspiration), sepsis (for example, chorioamnionitis and pyelonephritis) or pulmonary embolism. The pregnant patient may also develop conditions co-incidental to pregnancy such as trauma or appendicitis. Hemorrhage, particularly postpartum, and hypertensive disorders of pregnancy remain the most frequent indications for ICU admission. This review focuses on pregnancy-specific causes of critical illness. Management of the critically ill mother poses special challenges. The physiologic changes in pregnancy and the presence of a second, dependent, patient may necessitate adjustments to therapeutic and supportive strategies. The fetus is generally robust despite maternal illness, and therapeutically what is good for the mother is generally good for the fetus. For pregnancy-induced critical illnesses, delivery of the fetus helps resolve the disease process. Prognosis following pregnancy-related critical illness is generally better than for age-matched non-pregnant critically ill patients.

an abnormally adherent placenta that implants in the uterine wall, usually in scar tissue following previous CS. With increasing severity there is placenta incretainvasion of the myometrium and placenta percreta -in which the placenta invades the extra-uterine pelvic tissues. Uterine rupture during labor is another potential complication and is infrequently associated with previous CS. Postpartum hemorrhage (PPH) involves blood loss of greater than 500 mL within 24 hours regardless of the mode of birth. However, there is no universally accepted defi nition, and consideration should be given to physiologic response in addition to absolute blood loss. PPH is the most frequent indication for ICU admission. In 60% to 70% of cases, the cause of PPH is failure of uterine contraction following delivery. Th is uterine atony results in continuous bleeding that is often painless. Placental retention is the second most common cause of PPH (20% to 30% of cases). Genital trauma results in approximately 10% of cases of PPH and usually is associated with laceration of the vagina/cervix following instrumental delivery. Coagulation disorders may also result in PPH. Th ese may be congenital -hemophilia or von Willebrand disease -or acquired: sepsis, amniotic fl uid embolus (AFE) syndrome, acute fatty liver of pregnancy, preeclampsia, or HELLP (hemolysis, elevated liver enzymes, and low platelet count) syndrome. An aggressive coordinated multidisciplinary approach between obstetricians, midwives, anesthesiologists, the labora tory, and blood bank is required (Figure 1 ). Initial management depends on the cause of the hemorrhage and whether delivery of the fetus has occurred. Where the hemorrhage occurs postpartum, uterine atony with or without retained products should be suspected. Oxytocin should be administered, the bladder should be emptied, and the uterus massaged. Intravenous access is obtained, blood is sent for crossmatching, and if crossmatched blood is unavailable, O negative blood is obtained. Th e obstetrician should examine the genital tract for evidence of trauma. If bleeding persists, prostaglandin therapy -either intravenous prostaglandin E 2 or 15-methyl prostaglandin F 2α -is administered. Consideration should be given to uterine tamponade, by packing or balloon compression. Continued bleeding should result in surgical intervention: arterial ligation, B-Lynch suture, Cesarean hyste rectomy, or uterine artery embolization [4] . It is important that these interventions occur without delay; with each advancing minute, continued blood loss leads to loss of clotting factors, coagulopathy, and the potential for exsanguination. It is imperative that the patient be kept warm and that blood products and fl uids be warmed through a rapid transfusion device. An arterial line should be placed to measure blood pressure and pulse pressure variability and to take blood samples to track hematocrit and acid-base status. Ideally, quali tative tests of clot quality (for example, thrombo elastography) would be combined with coagulation studies to guide blood component resuscitation. In the event of uncontrolled non-surgical bleeding, consideration should be given to the administration of recombinant factor VIIa (90 μg/kg, repeated at 20 minutes if there is no response) [5] [6] [7] . Pre-eclampsia toxemia (PET) is a multisystem disease characterized by impaired organ perfusion resulting from vasospasm and activation of the coagulation system. PET is defi ned as hypertension and proteinuria appearing after the 20th week of gestation and resolving within 6 to 12 weeks of delivery. It occurs in 2% to 3% of all pregnancies and is more common in primigravida or the fi rst pregnancy with a particular partner. Other risk factors include a positive family history, pre-existing hyper tension, diabetes mellitus, multiple pregnancy, increasing maternal age, and obesity. Th e pathogenesis of PET appears to result from abnormal placenta formation (termed 'placentation'). Failure of the second phase of trophoblast invasion results in the lack of destruction of the muscularis layer of the spiral arterioles, impairing vasodilation in response to increases in blood fl ow, resulting in placental ischemia. Th e ischemic placenta releases vasoactive substances, leading to systemic endothelial injury and systemic organ dysfunction. As pregnancy progresses, placenta ischemia worsens, and the mother becomes hypovolemic and hypertensive and may develop renal dysfunction. Th ere is disordered prostaglandin metabolism, with an imbalance between vasoconstrictive thromboxane and vasodilatory prostacyclin, resulting in platelet dysfunction, endothelial damage, and further vasoconstriction. PET is classifi ed as mild, moderate, or severe. Severe pre-eclampsia is defi ned as one of the following: (a) severe hypertension (systolic blood pressure of greater than 160 mm Hg or diastolic blood pressure of greater than 110 mm Hg), (b) proteinuria of greater than 5 g per 24 hours, (c) oliguria of less than 400 mL per 24 hours, (d) cerebral irritability, (e) epigastric or right upper quadrant pain (liver capsule distension), or (f ) pulmonary edema. PET is associated with signifi cant morbidity and mortality, for the mother and fetus. PET usually resolves following delivery of the fetus but may manifest postpartum. A variety of antihypertensive agents, including hydralazine, labetalol, sodium nitroprusside, alpha blockers, calcium channel blockers, and methyl dopa, have been advocated in PET. Hydralazine and labetalol are the most widely used of these in the critical care setting. Hydralazine is administered by slow intravenous injection: fi rst with 5 mg intravenously and repeated at 20-minute intervals (5 to 10 mg depending on the response). Once blood pressure control is achieved, hydrala zine can be repeated as needed (usually about every 3 hours). Labetolol is started at 20 mg intravenously as a bolus, and the dose is doubled every 10 minutes until blood pressure is under control. Th e maximum dose is 220 mg. Magnesium is usually co-administered to provide vaso dila tation and prevent seizures. Care should be taken with fl uid resuscitation because of the risk of pulmonary edema. Eclampsia is an extreme complication of PET and is defi ned by the occurrence of seizures in the absence of other neurologic disorders. Up to 40% of seizures occur following delivery. Convulsions are believed to result from severe intracranial vasospasm, local ischemia, intracranial hypertension, and endothelial dysfunction associated with vasogenic and cytotoxic edema. Seizures tend to be self-limiting, and status epilepticus is unusual. Magnesium sulphate (MgSO 4 ) is superior to phenytoin and benzodiazepines in the prevention of recurrent eclamptic seizures [8] . Magnesium toxicity is rare in the absence of renal failure. Respiratory arrest caused by magnesium toxicity can be reversed with calcium. While traditionally the mortality from eclampsia has been high, death is now uncommon, and much of the mortality is attributable to hepatic complications, including hepatic failure, hemorrhage, or infarction [9] . HELLP syndrome is a constellation of fi ndings that include hemolysis with a microangiopathic hemolytic anemia (hemolysis), elevated liver function tests, and thrombo cytopenia (low platelets). HELLP syndrome complicates up to 3 per 1,000 pregnancies, may present as a severe manifestation of PET, and occurs in up to 20% of patients. Th e majority of patients are diagnosed before 37 weeks of gestation. Th ere is a clear overlap between PET and HELLP syndrome, and it is unclear whether the latter is a primary or secondary disease process. Th e syndrome is believed to be due to generalized endothelial and microvascular injury from activation of the complement and coagulation cascades, increased vascular tone, and platelet aggregation. Th is results in areas of hemorrhage and necrosis within the liver and may evolve to large hematomas, capsular tears, and intraperitoneal bleeding. Laboratory criteria for this syndrome include microangiopathic hemolysis with schistocytes present on the peripheral blood smear, platelet count of fewer than 50,000/mm 3 , serum total bilirubin of more than 20 μmol/L (1.2 mg/100 mL), serum lactate dehydrogenase of more than 600 U/L, and serum aspartate transaminase of more than 70 U/L. Th e diff erential diagnosis includes thrombotic thrombo cytopenic purpura and hemolytic uremic syndrome, cold agglutinins, and acute fatty liver of pregnancy. In regard to distinguishing these conditions, it is likely that the patient with HELLP syndrome will have a more severe liver dysfunction, potentially complicated by hepatic infarction or subcapsular hematoma. HELLP syndrome is a medical emergency and the mother's blood pressure and coagulation status must be stabilized rapidly. Corticosteroids should be administered to the mother to advance the lung maturity of the fetus if the gestation age is below 34 weeks. In the majority of cases, liver, renal, and hematopoietic function normalizes after 5 days. In about 30% of cases, HELLP syndrome will develop in the postpartum period. Dexamethasone does not benefi t the mother with HELLP syndrome [10] . Delivery of the fetus can signifi cantly abrogate HELLP syndrome. Th e timing of delivery depends on fetal maturity and the severity of illness of the mother. Th e major life-threatening complications of HELLP syndrome are hepatic hemorrhage, subcapsular hematoma, liver rupture, and multi-organ failure. Liver hemorrhage is managed conservatively where possible with aggressive blood product resuscitation to reverse the coagulopathy and ensure adequate oxygen-carrying capacity. Th e development of a subcapsular hematoma may lead to hepatic rupture, which is potentially life-threatening for the mother and fetus, and 50% maternal and up to 60% fetal mortality rates have been reported. Manage ment of threatened or actual hepatic rupture involves drainage of the hematoma, packing, oversewing of lacera tions, or partial hepatectomy. Consideration should be given to hepatic arterial embolization, either in the high-risk patient or in the period following operative stabilization. Acute fatty liver disease of pregnancy (AFLP) occurs in about 1 per 10,000 pregnancies characterized by hepatic microvesicular steatosis and manifests in the third trimester. Without early diagnosis and treatment (fetal delivery), the patient may develop acute liver failure and hepatic encephalopathy. It is more common in primiparas, in twin pregnancies, and in patients who have preeclampsia. AFLP is a mitochondrial disorder [11] related to inherited mutations that cause a defi ciency of the long-chain 3-hydroxyacyl coenzyme A dehydrogenase (LCHAD), a fatty acid beta-oxidation enzyme. When a heterozygous mother has a fetus that is homozygous for these mutations, the fetus is unable to metabolize longchain fatty acids; these acids accumulate in the fetus and spill over into the materna l circulation [12] . Th is mutant gene and defi cient coenzyme product leads to accumulation of long-chain fatty acid metabolites that are hepatotoxic [13] . Th e patient usually presents with vague symptoms, vomiting, or abdominal pain and may develop preeclampsia. Often there are no specifi c clinical signs, except tenderness in the right upper quadrant. Serum aminotransferase and bilirubin levels are signifi cantly elevated, and in later stages there is a coagulopathy evidenced by low fi brinogen and a prolonged prothrombin time. Compared with HELLP, thrombocytopenia and hypertension are unusual. Urate levels may be extremely high and there may be signifi cant hypoglycemia. Conclusive diagnosis requires liver biopsy, although owing to coagulopathy thi s is rarely possible or practical. Th e diff erential diagnosis includes HELLP syndrome, pre-eclampsia, and acute hepatitis due to alcohol or a virus [14] . Th e treatment of choice is urgent delivery of the fetus. Th is stops the overload of the mother's fatty acid oxidation system from fetal production and leakage into the maternal circulation [13] . AFE syndrome is a devastating complication that usually occurs within 24 hours of delivery. It manifests with acute severe hypoxic respiratory failure, associated with shock, disseminated intravascular coagulopathy (DIC), confusion, and seizures (Table 3 and Figure 2) . Th e incidence of AFE syndrome is unclear, and reports in the literature vary between 1:8,000 and 1:80,000 deliveries [15] . Th e disease is likely under-reported because of the absence of clear diagnostic criteria. Similarly, the mortality rate with AFE syndrome has been reported to be as high as 85%, and the majority of survivors suff er chronic neurologic defi cit [16] . Th e pathophysiology of AFE syndrome is unclear. Previously, this syndrome was believed to result from embolization of amniotic fl uid into the pulmonary circulation. However, an anaphylactoid or hyper sensitivity reaction to the contents of this fl uid is more likely [17] . Th e presence of amniotic fl uid in the pulmonary circulation is neither sensitive nor specifi c, as this has been identifi ed in mothers who do not develop AFE syndrome. Patients may present with seizure-type activity or acute respiratory distress. Acute lung injury results in profound hypoxemia, intense hypoxic pulmonary vasoconstriction, and acute right heart failure (Table 4) , result ing in hemodynamic collapse. Bowing of the right ventricle into the left results in acute diastolic and then acute systolic failure of the left ventricle. Th ere is simultaneous DIC that may manifest with bleeding from the placental bed. Nausea, vomiting, headache, confusion, and seizures commonly follow. Death from AFE syndrome results from multi-organ failure, exsanguinations, or cardiac arrest. Neurologic injury is common in survivors. Th ere is no diagnostic test for AFE syndrome. In a high-risk peripartum patient (Table 4 ) or one who has recently undergone termination of pregnancy (with hypertonic saline), the combination of coagulopathy, ARDS, and shock should be considered AFE syndrome until otherwise proven. Th e diff erential diagnosis includes sepsis, particularly due to chorioamniitis, thromboembolic pulmonary embolism, and aspiration pneumonitis. Th e fetus should be delivered emergently to avoid fetal demise. CS may be complicated by excess bleeding, requiring ligation of the uterine arteries and perhaps hysterectomy. Th ere is no specifi c treatment, although both aprotinin and activated protein C, compounds that modulate infl ammation and coagulation, may have some utility [18] . Critical care management should be directed at maintaining oxygen delivery and supporting the heart and circulation with inotropes and vasopressors. Early echocardiography is extremely useful to determine the nature of the cardiac injury (right versus left ventricular failure). Right ventricular failure can be worsened by high levels of positive end-expiratory pressure and vaso pressors and can be managed with milrinone, enoxamone, dobutamine, inhaled nitric oxide, and nebulized prostacyclin. Very large quantities of blood products may be required to control the coagulopathy, and this can result in signifi cant fl uid overload. Early consideration should be given to continuous renal replacement therapy. Although extracorporeal membrane oxygenation may appear to be an ideal approach to cardiopulmonary failure, excess bleeding may limit its application [19] . Peripart um cardiomyopathy is defi ned as a dilated cardiomyopathy of unknown cause associated with pregnancy. It occurs in the last gestational month or in fi rst 5 months postpartum and is associated with no other cardiac disease [20] . Although this condition is relatively rare, its exact incidence is unclear (the reported incidence varies between 1 in 1,500 and 1 in 15,000 pregnancies), and the condition is more common among Africans and Haitians. It is associated with older maternal age, obesity, multiparity, multiple pregnancies, and pregnancy-induced hypertension. Th e patient typically presents with symptoms of congestive heart failure: dyspnea, orthopnea, shortness of breath on exertion, upper abdominal pain, and so on. Echocardiography demonstrates systolic dysfunction. Proposed pathogenic hypotheses for peripartum cardiomyopathy include viral myocarditis, auto-immunemediated injury, and prolonged tocolysis. Current evidence strongly suggests that the disease is triggered by a by-product of prolactin metabolism, resulting in unbalanced peri-/postpartum oxidative stress [21] . Th e hormone is proteolytically cleaved, and an antiangio-genic, proapoptotic, and proinfl ammatory 16-kDa byproduct appears to attack the myocardium [22] . Medical therapy is eff ective in the majority of patients; treatment is commenced with a loop diuretic, and if the patient is postpartum, an angiotensin-converting enzyme inhibitor or antiotensin receptor blocker is added. In severe cases, placement of an intra-aortic balloon counter pulsation device or extracorporeal membrane oxygenation [23] may be necessary. Th ere is a very high incidence of thromboembolic disease associated with peripartum cardiomyopathy, and anticoagulation is essential. Early experience with prolactin inhibitors, such as bromocriptine, appears to be positive, and this may become the mainstay of treatment in the future [24] . Mortality appears to be high; in the US, mortality occurs in 25% to 50% of cases, usually within 3 months of diagnosis [25] . Availability of quality health care for the vulnerable population may be a component of this. Approximately 50% of women recover their ventricular function within 6 months of delivery. In some cases, cardiac trans plantation is necessary. Despite recovery, cardiomyopathy may recur in subsequent pregnancies. Pregnancy predisposes women to four specifi c infectious complications: pyelonephritis, chorio amnionitis (including septic abortion), endometritis (often following Cesarean delivery), and pneumonia. Pyelonephritis results from colonization of the kidney with Gram-negative bacteria secondary to loss of ureteral sphincter tone associated with progesterone. Pneumonia results, at least in part, from aspiration of gastric contents as a consequence of loss of lower esophageal sphincter tone and diaphragmatic elevation. Patients are also at elevated risk for viral and fungal pneumonia due to pregnancy-induced immunosup pres sion. Chorioamnionitis results from altera tions in the pH and increased glycogen content of the vagina, resulting in loss of the barrier for bacterial entry. It may complicate chorionic villus sampling, amnio centesis, or attempted instrumental (septic) abortion. Bacteremia in pregnancy is relatively common (reportedly occurring in 8% to 9% of pregnancies), whereas progression to severe sepsis and septic shock is relatively rare [26] [27] [28] [29] ; the rate of sepsis ranges from 1 in 7,654 to 1 in 8,338 deliveries reported [30] . Kankuri and colleagues [31] reported that only 1 of 43,483 mothers developed septic shock during the peripartum period. Despite these low progression rates, the mortality from sepsis in pregnancy remains signifi cant. Infections may be Gramnegative, Gram-positive, or rarely anerobic in etiology. Th e most commonly isolated organisms are Escherichia coli, enterococci, and beta hemolytic streptococci. Th e majority of infections occur postpartum; 'puerpural sepsis' or 'puerperal fever' is an umbrella term for a variety of infections that occur in the puerperium. Th e leading risk factor for puerpural sepsis is Cesarean delivery. Other signifi cant risk factors include retained products of conception, episiotomy, and prolonged rupture of the amniotic membranes. Infection may involve endometritis, parametritis (spread through the uterine wall), peritonitis, or thrombophlebitis of the pelvic veins. Endometritis is most commonly associated with group A streptococcal (GAS) infection, although Staphylococus, coliform, and anerobes may also be present. Hand washing and disinfectants dramatically decreased the incidence of puerperal fever. Pre-emptive antibiotics are administered if prolonged rupture of the membranes has occurred, to treat amnionitis, or if the woman has a fever and a foul-smelling vaginal discharge. Th e most recent maternal mortality report from the UK revealed that sepsis was the leading cause of maternal death between 2006 and 2008 [32] . Sepsis resulted in 29 maternal deaths. Fifty percent of deaths occurred following Cesarean delivery, whereas 7 occurred after vaginal delivery. One third of these deaths occurred before 24 weeks of gestation. Th ere was a marked seasonal pattern; most deaths occurred between December and April. Associated factors included minority ethnic origin and the presence of sickle cell disease or trait. Interestingly, obesity was not a risk factor. Infant mortality in aff ected pregnancies was 45%. GAS was the principal pathogen, causing nearly 50% of deaths. All mothers who died from GAS either worked with or had children, and most of the mothers had a direct or family history of sore throat or respiratory infection. In the developing world, the mortality rate from puerperal sepsis remains extremely high (greater than 70%) and tetanus is a common cause of infection. Th e clinical manifestations of sepsis include the hallmarks of systemic infl ammation that may be followed by coagulopathy, vasoplegia, and evolving multiple organ failure. Th e patient typically has a temperature of 38°C during the period from the end of the fi rst to the end of the 10th day after childbirth or abortion. Purpura fulminans may be associated with GAS infection. Physicians must have a high index of suspicion for sepsis in any peripartum patient who presents with fever and evidence of organ dysfunction: confusion, oliguria, tachycardia, and so on. Treatment includes fl uid resuscitation, empiric antibiotic therapy, and source control. Early aggressive volume resuscitation is essential and is followed with vasopressor therapy if necessary. Th ere is no evidence to suggest that norepinephrine has an adverse eff ect on fetal well-being. Th ere are no data on the use of vasopressin during pregnancy. Considering current data and the specifi c exclusion of pregnant (but not postpartum) patients from the PROWESS (Recombinant Human Activated Protein C Worldwide Evaluation in Severe Sepsis) trial and other trials, we cannot recommend the use of activated protein C during pregnancy. However, this agent may be of value in the postpartum period and clinicians should weigh up its value as a potent anti-infl ammatory/anticoagulant agent versus the risk of bleeding complications. Goal-directed resuscitation of the pregnant patient is recommended. One must be aware of physiologic changes and adjust the goals of the resuscitation accordingly. Although central venous pressure remains essentially unchanged during pregnancy, venous oxygen saturation (SvO 2 ) progressively decreases in the later stages. Hence, achieving goals of 70% to 75% may not be possible -or necessary -in this setting. Antimicrobials are administered on the basis of the 'best guess' source of infection. In general, tetracycline and quinolone antibiotics should be avoided in early pregnancy. Penicillins, macrolides, and cephalosporins appear relatively safe within normal dosage range. When cardiac arrest occurs in late pregnancy, cardiopulmonary resuscitation (CPR) is diffi cult [33] . Th e airway should be secured without delay, and the patient should be positioned to ensure left lateral displacement of the uterus. Unfortunately, this results in less eff ective CPR [33] . Drugs and defi brillation regimens should follow standard Advanced Cardiovascular Life Support guidelines. A decision regarding Cesarean delivery should be made immediately. Current data suggest that this should be performed within 5 minutes, to ensure viability of the mother [34] . Extraction of the fetus results in an increase in maternal blood volume and release of aortocaval compression [35] . Th is should be undertaken from the second trimester onwards. If the decision to perform Cesarean delivery is made, there should be no delay and CPR should be continued throughout surgery. Maternal brain death, due, for example, to an intracranial bleed, raises diffi cult ethical and legal issues [36, 37] . Depending on the gestational age, it may be possible, if the fetus is near viability, to maintain maternal organ support to allow the fetus to reach viability [38] . Although it may appear expensive to sustain the vegetative state of the brain-dead mother, this cost may be signifi cantly less than the long-term cost of care of a severely premature neonate, which would require neonatal ICU and a variety of health-care interventions. Critical illness is an uncommon but potentially devastating complication of pregnancy. Th e fi rst priority for the intensivist is to stabilize the mother, with the understanding that what is good for the mother is good for the fetus. Th e intensivist must be aware of the physiologic changes associated with pregnancy and their timing. Critical care interventions are similar to those for the non-pregnant patient; however, adjustment of physiologic targets for metabolic, pulmonary, and hemodynamic control may be necessary. Th e majority of acquired pregnancy-related diseases, such as PET, AFLP, and cardiomyopathy, are abrogated by delivery. However, timing and the state of fetal maturation are crucial to decision making.

PaCO(2 )and alveolar dead space are more relevant than PaO(2)/FiO(2 )ratio in monitoring the respiratory response to prone position in ARDS patients: a physiological study

INTRODUCTION: Our aims in this study were to report changes in the ratio of alveolar dead space to tidal volume (VD(alv)/V(T)) in the prone position (PP) and to test whether changes in partial pressure of arterial CO(2 )(PaCO(2)) may be more relevant than changes in the ratio of partial pressure of arterial O(2 )to fraction of inspired O(2 )(PaO(2)/FiO(2)) in defining the respiratory response to PP. We also aimed to validate a recently proposed method of estimation of the physiological dead space (VD(physiol)/V(T)) without measurement of expired CO(2). METHODS: Thirteen patients with a PaO(2)/FiO(2 )ratio < 100 mmHg were included in the study. Plateau pressure (Pplat), positive end-expiratory pressure (PEEP), blood gas analysis and expiratory CO(2 )were recorded with patients in the supine position and after 3, 6, 9, 12 and 15 hours in the PP. Responders to PP were defined after 15 hours of PP either by an increase in PaO(2)/FiO(2 )ratio > 20 mmHg or by a decrease in PaCO(2 )> 2 mmHg. Estimated and measured VD(physiol)/V(T )ratios were compared. RESULTS: PP induced a decrease in Pplat, PaCO(2 )and VD(alv)/V(T )ratio and increases in PaO(2)/FiO(2 )ratios and compliance of the respiratory system (Crs). Maximal changes were observed after six to nine hours. Changes in VD(alv)/V(T )were correlated with changes in Crs, but not with changes in PaO(2)/FiO(2 )ratios. When the response was defined by PaO(2)/FiO(2 )ratio, no significant differences in Pplat, PaCO(2 )or VD(alv)/V(T )alterations between responders (n = 7) and nonresponders (n = 6) were observed. When the response was defined by PaCO(2), four patients were differently classified, and responders (n = 7) had a greater decrease in VD(alv)/V(T )ratio and in Pplat and a greater increase in PaO(2)/FiO(2 )ratio and in Crs than nonresponders (n = 6). Estimated VD(physiol)/V(T )ratios significantly underestimated measured VD(physiol)/V(T )ratios (concordance correlation coefficient 0.19 (interquartile ranges 0.091 to 0.28)), whereas changes during PP were more reliable (concordance correlation coefficient 0.51 (0.32 to 0.66)). CONCLUSIONS: PP induced a decrease in VD(alv)/V(T )ratio and an improvement in respiratory mechanics. The respiratory response to PP appeared more relevant when PaCO(2 )rather than the PaO(2)/FiO(2 )ratio was used. Estimated VD(physiol)/V(T )ratios systematically underestimated measured VD(physiol)/V(T )ratios.

Since its first description in 1967 [1] , it has been accepted that acute respiratory distress syndrome (ARDS) includes a number of lung injuries of various origins whose consequences are decreased lung capacity available for ventilation, leading to the concept of "baby lung" [2] . Considerable progress has been made over the past decade in the ventilatory management of patients with ARDS. In particular, a strict limitation of tidal volume (V T ) and plateau pressure (Pplat) below 30 cmH 2 O reduces mortality [3] . The application of positive end-expiratory pressure (PEEP) is recognized to recruit the lung and to restore functional residual capacity [4] , but its optimum level is still widely debated [5] . The prone position (PP) may also be part of the ventilatory strategy. This method was proposed more than 30 years ago, initially in pathophysiological studies [6, 7] . Recently, Sud et al. [8] suggested, on the basis of pooled data from randomized, controlled trials, that PP may improve survival in the subgroup of patients with the most severe ARDS, that is, those with a ratio of partial pressure of arterial O 2 to fraction of inspired O 2 (PaO 2 / FiO 2 ) < 100 mmHg. Many questions remain unresolved. In particular, response to PP is usually defined according to changes in PaO 2 , with responders being those in whom the PaO 2 /FiO 2 ratio increases > 20 mmHg after one to six hours in the PP [9] [10] [11] . However, we have previously reported that PP allows recruitment of a slow compartment previously excluded from ventilation [12] . This was associated with a decrease in partial pressure of arterial CO 2 (PaCO 2 ), an indirect reflection of the reduction of the alveolar dead space (VD alv ) [12] . Gattinoni et al. [10] also reported that the prognosis is improved in patients in whom PaCO 2 declines after an initial PP session. Finally, VD alv appears to be an independent risk factor for mortality in patients with ARDS [13] . In a recent study, Siddiki et al. [14] proposed evaluating the physiological dead space fraction (VD physiol / V T ) by using a rearranged alveolar gas equation for PaCO 2 without any expired CO 2 measurement. In this context, we conducted a prospective physiological study to evaluate the impact of PP on ventilatory mechanics, gas exchange and VD alv . Our main objective was to validate our hypothesis that changes in PaCO 2 and VD alv might be more relevant than changes in PaO 2 in defining the respiratory response to PP. Our second objective was to validate the method of evaluation of the VD physiol /V T proposed by Siddiki et al. [14] . In our unit, patients with a PaO 2 /FiO 2 ratio < 100 mmHg after 24 to 48 hours of mechanical ventilation are systematically turned to PP when hemodynamically stable [15] . Our study was approved by the Ethics Committee of the "Société de Réanimation de Langue Française" (SRLF-CE 07-213). After obtaining informed consent from the patients' relatives, 15 patients were included in the study between January 2008 and March 2010. Inclusion criteria were (1) the presence of ARDS according to the definition of the Acute Respiratory Distress Syndrome Network [3] ; (2) persistence of severe hypoxemia after 48 hours of mechanical ventilation, defined as a PaO 2 /FiO 2 ratio < 100 mmHg; and (3) hemodynamic stability, defined as systolic blood pressure > 90 mmHg with norepinephrine infusion at a rate < 0.5 μg/kg/minute. Patients with chronic obstructive pulmonary disease were excluded. All patients were ventilated in volume-controlled mode (Servo-i; Maquet SA, Ardon, France), sedated and paralyzed by infusion of atracurium. The heat and moisture exchanger was routinely removed and replaced by a heated humidifier to reduce instrumental dead space as previously reported [16] . The ventilator settings included a "moderately restricted" V T of 6 to 8 mL/kg measured body weight, a respiratory rate allowing us to limit hypercapnia without generating intrinsic PEEP and an inspiration/expiration ratio of 1:2 with an end inspiratory pause of 0.5 seconds. Pplat was strictly limited < 30 cmH 2 O, and the PEEP selected was that which corrected the intrinsic PEEP, if any [17] . Ventilator settings were kept constant throughout the study. A recruitment maneuver was never used, and suction was not systematically performed. All patients were continuously monitored in terms of blood pressure with an arterial catheter, heart rate and O 2 saturation by pulse oximetry. The study was conducted during the first session of PP. Our sessions routinely last 15 to 18 hours per day. Blood gas analysis, Pplat, total PEEP, end-tidal CO 2 (P etCO2 ) and mixed expired CO 2 (P ECO2 ) were recorded with the patient in the supine position, just before turning the patient to the PP, and every 3 hours in the PP until 15 hours had elapsed. Expired CO 2 was measured by a sensor positioned between the proximal end of the endotracheal tube and the Y piece of the ventilator circuit (COSMO; Novametrix, Wallingford, CT, USA). The ratio of VD/V T was calculated using the simplified Bohr equation [18] as follows: (1) VD alv /V T = 1 -P etCO2 / PaCO 2 and (2) VD physiol /V T = 1 -P ECO2 /PaCO 2 . The estimated VD physiol /V T ratio was calculated as 1 -[(0.86 × VCO 2est )/(VE × PaCO 2 )], where VCO 2est is the estimated CO 2 production calculated using the Harris-Benedict equation [19] and VE is the expired minute ventilation. Intrinsic PEEP was measured during a four-second end-expiratory occlusion period. Pplat was measured during a 0.5-second end-inspiratory pause. Respiratory system compliance (Crs) was calculated as Crs = V T / (Pplat -PEEP total ). Responders to PP were defined in two different ways: (1) an increase in PaO 2 /FiO 2 ratio > 20 mmHg after 15 hours of PP or (2) a decrease in PaCO 2 > 2 mmHg after 15 hours of PP. Statistical analysis was performed using StatView 5 software (SAS Institute Inc., Cary, NC, USA). The continuous variables were expressed as medians (1st to 3rd interquartile range). Analysis of variance for repeated measurements was used for each parameter, and P < 0.05 was considered statistically significant. Measured VD physiol /V T and estimated VD physiol /V T were compared according to Bland-Altman analysis, together with the concordance correlation coefficient in 78 paired data. The same method was used to compare variations of measured and estimated VD physiol /V T every three hours while the patient was in PP. Two patients were excluded from the study because of a history of severe chronic obstructive pulmonary disease, which left a study population of 13 patients. The patients' median age was 53 years (1st to 3rd interquartile range, 48 to 59 years), their median Simplified Acute Physiology Score II score was 62 (1st to 3rd interquartile range, 35 to 71) and their median Sequential Organ Failure Assessment score was 11 (1st to 3rd interquartile range, [8] [9] [10] [11] [12] [13] . All patients except one had ARDS of pulmonary origin. Eight patients had pneumonia, with six cases related to streptococcus pneumonia and two due to influenza (H1N1 virus). Two patients had aspiration, one had toxic shock syndrome and two had ARDS due to miscellaneous causes. No patient had abdominal hypertension or traumatic lung injury. Eleven patients required norepinephrine infusion. Respiratory parameters and blood gas analysis at the time of inclusion are reported in Table 1 . A significant increase in PaO 2 /FiO 2 ratio occurred after 15 hours of PP, from 70 mmHg (51 to 77) in the supine position to 99 mmHg in the prone (83 to 139) (P < 0.0001) ( Table 2) . A significant decrease in PaCO 2 was also observed, from 58 mmHg (52 to 60) to 52 mmHg (47 to 56) (P = 0.04) ( Table 2) , with the lowest value occurring after nine hours of PP. As noted in Table 2 , Pplat was significantly reduced (P = 0.0004) and Crs improved (from 16 mL/cmH 2 O (13 to 30) to 18 mL/ cmH 2 O (15 to 30); P = 0.02). Finally, the VD alv /V T ratio was significantly reduced from 0.42 (0.35 to 0.47) to 0.40 (0.26 to 0.45), with the lowest value occurring after three hours in PP (hour 3) (0.31) ( Table 2) . Seven patients were classified as "PaO 2 responders" and six were classified as "PaO 2 nonresponders" according to PaO 2 /FiO 2 ratio changes. No differences in VD alv /V T ratios or PaCO 2 or Pplat alterations during PP were observed between groups (Table 3 and Figure 1 ), whereas Crs increased more in the responders (Table 3) . Seven patients were also classified as "PaCO 2 responders" and six as "PaCO 2 nonresponders" according to the PaCO 2 changes. However, when compared with the PaO 2 /FiO 2 classification, four patients were classified differently. As shown in Table 4 and Figure 2 , VD alv /V T , PaO 2 /FiO 2 , PaCO 2 , Pplat and Crs were significantly more altered in responders than in nonresponders. As shown in Figure 3 , we found no correlation between changes in VD alv /V T and changes in PaO 2 /FiO 2 (P = 0.95), whereas we found a negative correlation between changes in VD alv /V T and changes in Crs (r = 0.29, P = 0.03). As shown in Figure 4 , estimated VD physiol /V T systematically underestimated measured VD physiol /V T , with a poor concordance correlation coefficient of 0.19 (95% confidence interval (95% CI) 0.091 to 0.28), a bias of 0.16 and an agreement between -0.05 and 0.37. Concerning changes in VD physiol /V T during PP, estimated VD physiol /V T had a concordance correlation coefficient of 0.51 (95% CI 0.32 to 0.66) (Figure 4 ). One of the objectives of our study was to describe alterations in VD alv induced by PP. ARDS is characterized by a heterogeneous lung with the existence of a slow compartment [18, 20] , defined as areas available for, but partially or totally excluded from, ventilation due in part to a bronchiolar collapse [12, 21] . In a previous study, we reported that PP may induce recruitment of this slow compartment, as suggested by its ability to counteract intrinsic PEEP and to decrease the expiratory time constant [12] . In the same study, we also reported that PP leads to a decrease in PaCO 2 , suggesting diminution of VD alv (alveolar dead space) [12] . Our present study demonstrates that PP may induce a decrease in VD alv . It occurred from the third hour and was maintained throughout the PP session. VD alv may be the consequence of nonperfused or poorly perfused lung areas in ventilated anterior areas, but also of a slow compartment partially excluded from ventilation. Our results suggest that PP induces functional lung recruitment, especially since decreases in VD alv related to PP were associated with a decrease in Pplat and strongly correlated with improvement in compliance. did not report a decrease in Pplat in PP, as we found, but after returning patients to the supine position [22] . This could be explained by the fact that they used roll under the upper part of the chest wall, leading to a significant impairment in chest wall compliance [22] , whereas we did not. The most beneficial reported effect of PP is oxygenation improvement [24, 25] . However, this better oxygenation can be due to (1) lung recruitment related to restoration of functional residual capacity [7] and improvement of the diaphragmatic movement in the posterior part [26] [27] [28] or (2) simply to an improvement in the ventilation/perfusion ratio due to a decreased hydrostatic gradient between the anterior and posterior parts of the lung [26, 29] . Whereas the first mechanism is crucial, one can say that the second mechanism is less important. This is why the second objective of our study was to test whether the response to PP in terms of PaCO 2 was physiologically more relevant than in terms of PaO 2 /FiO 2 ratio. Gattinoni et al. [10] reported that an increase in PaO 2 /FiO 2 ratio > 20 mmHg after six hours of PP is not predictive of the patient's prognosis, whereas a decline in PaCO 2 ≥1 mmHg is. In our present study, 7 of 13 patients were PaO 2 responders (increased PaO 2 /FiO 2 ratio > 20 mmHg after 15 hours of PP). However, changes in Pplat, PaCO 2 and VD alv did not differ between PaO 2 responders and PaO 2 nonresponders. On the other hand, 7 of 13 patients were PaCO 2 responders (decreased PaCO 2 > 2 mmHg after 15 hours of PP). PaCO 2 responders had a significant decrease in Pplat and VD alv , as well as a significant increase in oxygenation and compliance, compared with nonresponders. Our results are in accordance with a recent study of 32 ARDS patients [23] , in which the investigators reported that PaCO 2 variation induced by PP, and not PaO 2 /FiO 2 variation, is associated with lung recruitability. Interestingly, in our study, changes in VD alv were not correlated with changes in oxygenation but were strongly correlated with changes in compliance of the respiratory system. An unexpected result of our work concerns the change over time of respiratory mechanics, blood gas analysis and VD alv . For many years, our PP protocol has been to turn patients to PP for up to 15 to 18 hours per day for 3 days [15] . In the study by Mancebo et al. [30] , which concluded that PP may reduce mortality in patients with severe ARDS, PP sessions lasted 20 hours/ day. In a recent study, we demonstrated that PP sessions that lasted 18 hours/day were independently associated with survival [31] . In the present study, the maximum effect of PP for VD alv , PaCO 2 and Pplat occurred six to nine hours after turning patients to PP. Later the effect seemed to be a decline. How this affects the effect of PP on patient prognosis remains to be elucidated. The second objective of our study was to validate a recently proposed method to evaluate the VD physiol /V T ratio [14] . The method is based on CO 2 production calculated from the Harris-Benedict equation [19] and on the expired minute ventilation. Siddiki et al. [14] reported that it was associated with mortality in acute lung injury patients in a dose-response manner and proposed its routine use to estimate VD physiol /V T . However, they did not report any comparison with measured VD physiol /V T . In the present study, we have demonstrated that this method significantly underestimates VD physiol /V T , rendering it not accurate enough to assess the degree of lung injury. Interestingly, changes in estimated VD physiol /V T during PP appeared better correlated with changes in measured VD physiol /V T and could be proposed in the future in this field. Siddiki et al. [14] proposed the method in the context of a much larger series than ours and in patients with less severe ARDS, rendering it difficult to draw any definitive conclusions. Our work is limited by the small number of patients included. This is a consequence of our routine protocol, which strictly restricts PP to patients with the most severe ARDS, that is, those with a PaO 2 /FiO 2 ratio < 100 mmHg after 48 hours of ventilation. This also explains why it is not possible to link our results to outcomes. However, despite this limitation, we consider our results relevant from a physiological point of view. In conclusion, our study demonstrates that PP induces a decrease in PaCO 2 and VD alv . This is related to an improvement in respiratory mechanics, with a decrease in Pplat and an increase in compliance. Testing the response to PP appeared to be physiologically more relevant using PaCO 2 changes than PaO 2 /FiO 2 changes. How this may affect management at the bedside remains to be studied. Estimated VD physiol /V T ratios systematically underestimated measured VD physiol /V T ratios. • PP induced a decrease in VD alv /V T , which was correlated with an improvement in respiratory mechanics. • Defining the respiratory response to PP appeared more relevant when using PaCO 2 changes rather than PaO 2 /FiO 2 changes. • Estimated VD physiol /V T using the Harris-Benedict equation systematically underestimated measured VD physiol /V T . Abbreviations ARDS: acute respiratory distress syndrome; P ECO2 : mixed expired PCO 2 ; PEEP: positive end-expiratory pressure; P etCO2 : end-tidal PCO 2 ; PP: prone position; Pplat: plateau pressure; VD alv : alveolar dead space; VD physiol : physiological dead space.

Lung Function and Organ Dysfunctions in 178 Patients Requiring Mechanical Ventilation During The 2009 Influenza A (H1N1) Pandemic

INTRODUCTION: Most cases of the 2009 influenza A (H1N1) infection are self-limited, but occasionally the disease evolves to a severe condition needing hospitalization. Here we describe the evolution of the respiratory compromise, ventilatory management and laboratory variables of patients with diffuse viral pneumonitis caused by pandemic 2009 influenza A (H1N1) admitted to the ICU. METHOD: This was a multicenter, prospective inception cohort study including adult patients with acute respiratory failure requiring mechanical ventilation (MV) admitted to 20 ICUs in Argentina between June and September of 2009 during the influenza A (H1N1) pandemic. In a standard case-report form, we collected epidemiological characteristics, results of real-time reverse-transcriptase--polymerase-chain-reaction viral diagnostic tests, oxygenation variables, acid-base status, respiratory mechanics, ventilation management and laboratory tests. Variables were recorded on ICU admission and at days 3, 7 and 10. RESULTS: During the study period 178 patients with diffuse viral pneumonitis requiring MV were admitted. They were 44 ± 15 years of age, with Acute Physiology And Chronic Health Evaluation II (APACHE II) scores of 18 ± 7, and most frequent comorbidities were obesity (26%), previous respiratory disease (24%) and immunosuppression (16%). Non-invasive ventilation (NIV) was applied in 49 (28%) patients on admission, but 94% were later intubated. Acute respiratory distress syndrome (ARDS) was present throughout the entire ICU stay in the whole group (mean PaO(2)/FIO(2 )170 ± 25). Tidal-volumes used were 7.8 to 8.1 ml/kg (ideal body weight), plateau pressures always remained < 30 cmH(2)O, without differences between survivors and non-survivors; and mean positive end-expiratory pressure (PEEP) levels used were between 8 to 12 cm H(2)O. Rescue therapies, like recruitment maneuvers (8 to 35%), prone positioning (12 to 24%) and tracheal gas insufflation (3%) were frequently applied. At all time points, pH, platelet count, lactate dehydrogenase assay (LDH) and Sequential Organ Failure Assessment (SOFA) differed significantly between survivors and non-survivors. Lack of recovery of platelet count and persistence of leukocytosis were characteristic of non-survivors. Mortality was high (46%); and length of MV was 10 (6 to 17) days. CONCLUSIONS: These patients had severe, hypoxemic respiratory failure compatible with ARDS that persisted over time, frequently requiring rescue therapies to support oxygenation. NIV use is not warranted, given its high failure rate. Death and evolution to prolonged mechanical ventilation were common outcomes. Persistence of thrombocytopenia, acidosis and leukocytosis, and high LDH levels found in non-survivors during the course of the disease might be novel prognostic findings.

On April 2009, a novel influenza A (H1N1) virus emerged in Mexico and spread rapidly across the world [1, 2] . As of 17 June 2010, more than 214 countries had reported confirmed cases of infection with pandemic 2009 influenza A (H1N1) virus, including at least 18,156 deaths [3] . Unlike seasonal influenza, in which hospitalizations occur among patients younger than 2 and older than 65 years, or in those with underlying diseases [4] , this novel virus affected otherwise healthy young and middle-aged adults and obese individuals [2, 5] . Patients with previous respiratory disease, immunocompromised hosts and pregnant women were affected as frequently as with seasonal influenza [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] . Although a mild form of the disease was prevalent, it soon became evident that the 2009 influenza A (H1N1) virus could also provoke severe, acute respiratory failure requiring admission to the intensive care unit (ICU) for mechanical ventilation [16] , which was reflected in the severe pathological injury found at autopsy [17] . The Argentinian population was greatly affected during the pandemic, with a total of 1,390,566 cases of influenza-like illness requiring 14,034 hospitalizations. Of the 11,746 confirmed cases of patients infected with the new strain, 617 died [18] . This represents a death rate per infection of 4.3% in hospitalized cases; an intermediate figure compared to 3.6% in Brazil, 1.2% in Chile, and approximately 6% in Uruguay, Colombia and Venezuela [19] . It should be noted that these numbers reflect great uncertainty, particularly with regard to case diagnosis. Lack of testing of mild disease and difficulties due to laboratory overload have also been well described [15, 20] . These general problems have been acknowledged by experts [21] . The severity of disease was rapidly perceived by health authorities and scientific societies. Hence, a committee of experts of the Argentinian Society of Intensive Care Medicine decided to focus on the most acutely ill patients: those presenting with diffuse viral pneumonitis requiring mechanical ventilation. They designed an epidemiological study, recently-published, to determine risk factors and outcomes [15] ; this is one of many series up to the present that have described epidemiological and clinical aspects of the 2009 influenza A (H1N1) pandemic [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] . There remains, however, a paucity of data published on physiological evolution during ICU stay [22] . This present study, concurrently planned with the first by the same committee of experts, thus aims to provide such information. Our objectives were: first, to characterize alterations of oxygenation, respiratory mechanics and the use of mechanical ventilation; second, to explore compliance with protective lung ventilation; and, finally, to assess the evolution of laboratory findings and organ dysfunctions throughout the course of the disease. This was a multicenter, inception cohort study that included patients aged > 15 years admitted to the ICU with a previous history of influenza-like illness, evolving to acute respiratory failure that required mechanical ventilation during the 2009 winter in the Southern Hemisphere. These patients had confirmed or probable disease caused by the 2009 influenza A (H1N1) virus and were included in the Registry of Cases of the Argentinian Society of Intensive Care Medicine (SATI), created to characterize local aspects of the pandemic. On 27 June 2009, a form to collect online epidemiological data was posted on the official SATI website. A detailed description and analysis of this information was recently published [14] . There was also an optional, more comprehensive casereport form to complete, developed by experts of the SATI's Respiratory Committee for recording certain prespecified variables throughout ICU stay, which included mechanical ventilation (MV), respiratory mechanics, oxygenation, blood chemistry and organ failure variables. This information was collected over 10 days and is analyzed in the present study. Patients were characterized as confirmed, probable or possible cases of 2009 influenza A (H1N1) [20] according to the findings in the respiratory samples collected on admission. Some specimens, however, were not analyzed because laboratories soon became overloaded, especially at the beginning of the pandemic. As of 25 September 2009, the weekly update of the Ministry of Health reported that in patients ≥5 years with influenzalike illness, the 2009 influenza A (H1N1) virus had displaced other respiratory viruses in 93.4% of the samples processed [23, 24] . As a result of this, probable and suspected cases were considered as caused by the novel virus and were so included in the study. We collected dates of hospital and ICU admission, and of MV onset; demographics; risk factors for influenza A; actual weight; height; severity of illness (Acute Physiology And Chronic Health Evaluation II, APACHE II), organ failures (Sequential Organ Failure Assessment, SOFA); type of MV used, as noninvasive (NIV) and invasive; and date of intubation. Ideal body weight (IBW, ml/kg) and body mass index (BMI) were calculated; obesity was defined as a BMI > 30. At MV onset (Day 0) and on Days 3, 7 and 10, until death or discharge, whichever occurred first, we recorded: (1) MV-related variables. (2) MV modes: volume-controlled ventilation (VCV); pressure-controlled ventilation (PCV); bilevel mode; pressure support ventilation (PSV); other. (3) Tidal volume (Vt, in ml/kg of IBW) (4) Pressures: peak, plateau pressures, total positive end-expiratory pressure (PEEP) and driving pressure (plateau pressure -PEEP), in cmH2O. The main outcome measure was hospital mortality; secondary outcomes were length of MV, of ICU (LOSICU) and of hospital (LOSHOSP) stays. In case of missing observations, local study coordinators were contacted to provide the corresponding values. Proportions were calculated as percentages of existing data. No assumptions for missing data were made. Statistical analysis was performed with SPSS 17.0 (SPSS Inc., Chicago, IL, USA). Data were analyzed for the entire population; for the subgroups of survivors vs. non-survivors; and for patients receiving NIV on admission vs. those who did not. Descriptive statistics used were: mean ± standard deviations (SD) and median and 25-75% interquartile ranges (IQR) for continuous data of normal and non-normal distribution, respectively; and percentages for categorical data. Differences between subgroups were analyzed with unpaired t test, Mann-Whitney U test, and Chi-square tests, as appropriate. A P-value of <.05 was considered statistically significant. A Kaplan-Meier curve was constructed to evaluate survival over the follow-up period. Over time, normally distributed data were analyzed with two-way repeated measures of ANOVA. At the pre-specified time points, differences within the entire group and subgroups, and between subgroups, were tested using paired and unpaired t tests, respectively. In non-normally distributed data, differences over time within the entire group and the subgroups were analyzed with Friedman's and Wilcoxon tests. Comparisons between subgroups at the pre-specified time points were tested with Mann-Whitney U test. The Bonferroni correction was used to adjustments for multiple comparisons. The local Institutional Review Boards waived the need for informed consent, given the general lack of knowledge on the clinical and outcome characteristics of the ongoing pandemic and to the non-interventional study design. General characteristics (Table 1) Between 6 June and 28 August 2009, the SATI's online Registry included 337 patients admitted to 35 ICUs with confirmed/probable/possible diffuse viral pneumonitis caused by influenza A (H1N1), with acute respiratory failure requiring MV (14) . Of these, 178 consecutive patients admitted to 20 ICUs were followed over time, and are presented in this study. To address any potential concern that unconfirmed cases could belong to a different population of patients, we performed a sensitivity analysis of clinical and outcome characteristics data after exclusion of these patients. The results of this analysis did not differ from those of the primary assessments, so the 178 patients are considered for evaluation. Briefly, patients were middle-aged, with no gender preponderance; they had a history of symptoms of nearly one-week duration and were ventilated at 1 [0 to -2] day after hospital admission. Pre-existent respiratory diseases, obesity, and diseases causing immunosuppression were the most frequent comorbid conditions; and prevalence of pregnancy was higher than in the general population, as expected [25] . Non-survivors were sicker on admission; duration of previous symptoms was longer; and organ failures were more severe. Obesity and immunosuppression were significantly more frequent as predisposing conditions. Ninety-three patients survived (52%) (See Figure 1 ). (Table 2) During the study period, the entire group had Vt values between 7.8 to 8.1 ml/kg of IBW, with plateau pressures remaining always < 30 cmH 2 O. Non-survivors displayed a trend towards lower Vt and higher plateau pressures, which differed significantly from survivors only at Day 7. Intermediate PEEP levels were used, and decreased in survivors from Day 3 onwards. Driving pressures were similar over time in all patients; only at admission did non-survivors exhibit higher values. PaO 2 /FIO 2 increased significantly over time in all patients and in survivors. It remained, however, < 200 in the whole group throughout the entire ICU stay due to non-survivor values. Non-survivors displayed significantly lower PaO 2 /FIO 2 at all time points. Lung infiltrates (in quadrants) peaked at day 3 (3.1 ± 1.0 vs. 2.9 ± 1 at Day 0, P < 0.01) and then decreased during the study in the entire group, especially at Day 10 (2.8 ± 1.1, P < 0.83 vs. Day 0), which reflected the improvement in survivors (3.1 ± 1.0 at Day 3 vs. 2.9 ± 1.0 at Day 10, P < 0.01). In Figure 2 , the utilization of ventilation modes and rescue therapies in the entire group are shown. Briefly, PCV use equaled VCV at Day 10, preceded by deterioration in oxygenation and respiratory mechanics: PaO 2 / FIO 2 78 ± 24 vs. 128 ± 33, (P = 0.03); PaCO 2 44 ± 4 vs. 35 ± 3 mmHg (P = 0.04); pH 7.29 ± 0.03 vs. 7.39 ± 0.05 (P = 0.05), and plateau pressures of 30 ± 2 vs. 25 ± 3 cmH 2 O (P = 0.03). Recruitment maneuvers became significantly more common in non-survivors at Day 3 (46%, vs. 29% in survivors; P = 0.03), as did prone positioning (24%, vs. 14%; P = 0.001). After that, only prone positioning remained significantly more used in nonsurvivors (at Day 7: 38%; vs. 14%, P = 0.004; and at Day 10: 25%; vs. 5%, P = 0.02). Six patients received tracheal gas insufflation; only one survived. Neuromuscular blockers were prescribed in 18% of patients on admission; and their use was subsequently more frequent in non-survivors (Day 3: 14% vs. 8%, P = 0.02; and Day 7: 14% vs. 8%, P = 0.04). The main causes of death were refractory hypoxemia (64%); followed by multiorgan dysfunction syndrome (15%) and shock (10%). Prolonged mechanical ventilation and long ICU and hospital stays were frequent (Table 1) . Tracheostomy was performed in 29 patients (16%) at Day 14 [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] . Acid-base variables and fluid balance (Table 3) Arterial pH increased over time in the whole cohort and in both subgroups, perhaps secondary to general resuscitation measures. Despite this, non-survivors displayed significantly lower pH at all time points, owing to changes in base excess on Days 0 and 3, and to pCO 2 elevations thereafter. Respiratory rates remained unchanged, only increasing at Day 10 in non-survivors; nevertheless, this corresponded to the highest pCO 2 values, indicating the more severe respiratory compromise. Bicarbonate paralleled pH behavior. Changes in fluid balance did not show clear trends: only at Day 10 they decreased significantly, expressing survivors' behavior. Forty-nine patients (28%) underwent a trial of NIV on admission; they were significantly less ill and had a lower incidence of immunosuppression. Oxygenation and outcome variables were similar to those of patients not receiving NIV. Sixty-one percent of patients (n = 30) receiving NIV survived; duration of NIV was of 8 (2 to 18) hours. There were no differences between survivors and nonsurvivors in the duration of the procedure, or in the type of interface or respirator used. Of note, most patients on NIV (46 out of 49; 94%) had to be intubated and ventilated invasively for hypoxemic failure. Characteristics associated to NIV success/failure are shown in Table 5 . NIV was also used for treating post-extubation respiratory failure in 12 of 178 patients (7%), with success (reintubation not needed) in 8 cases (66%). The most consistent changes over time were found in platelet count, which increased significantly in the whole cohort (P < 0.000 for Days 3, 7 and 10 vs. Day 0), secondary to elevations in survivors. At all time points, platelets differed between survivors and non-survivors. Conversely, white blood cell count showed a progressive Creatine-kinase and markers of liver injury (alanine/ aspartate aminotransferases, serum bilirubin; not shown) were mildly elevated and displayed no substantial changes. On the contrary, lactate-dehydrogenase levels were significantly higher in non-survivors throughout the study. Creatinine levels were stable over the period, but were significantly higher in non-survivors on Days 0 and 3. Finally, SOFA score diminished over time in all patients (P < 0.000 for Days 7 and 10 vs. Day 0), as a result of the decrease in survivors. SOFA was significantly lower in survivors throughout the study. In Figure 3 , the differences between survivors and non-survivors are displayed. We report on a large, prospective cohort of 2009 influenza A (H1N1) patients that were mechanically ventilated for acute respiratory failure due to diffuse pneumonitis during the pandemic in Argentina. Though most were middle-aged, previously healthy adults, patients with preexistent lung disease, immunosuppression, obesity and pregnancy were also affected. Mortality was high and evolution to chronic critical illness was common, as shown by prolonged mechanical ventilation, high needs of tracheostomy, and lengthened ICU and hospital stays. Patients had characteristically a history of protracted symptoms and displayed severe compromise of oxygenation compatible with ARDS throughout the study period, which only improved in survivors. At all time points, PaO 2 /FIO 2 differed significantly between survivors and non-survivors, requiring higher FIO 2 and PEEP in this last subgroup. Yet the levels of applied PEEP were only in the intermediate range, similar to mean values of 8.7 cmH 2 O of PEEP in an international study on mechanical ventilation [26] , which may explain the relatively high FIO 2 used in our study. Driving pressures were similar in both subgroups most of the time, suggesting an intention to limit alveolar excursion as part of a protective strategy. It is striking that, as has been described in similar studies on mechanical ventilation performed during the 2009 influenza A (H1N1) pandemic [6, 7] , tidal volumes used were between 7.5 and 8.3 ml/kg IBW, certainly higher than the 6 ml/kg demonstrated as being lungprotective [27] . Indeed, barriers to implementing lowtidal volume have been identified and might explain physician behavior [28] . Despite this, plateau pressures did remain below 30 cmH 2 O [29] , indicating that lung compliance might have been preserved. Perhaps clinicians focused on plateau pressures rather than on tidal volumes [30] since it still remains unclear which should be limited to avoid ventilator-induced lung injury [31] . We, like others [6, 7, 32, 33] , could not find differences in utilized tidal volumes between survivors and non-survivors. Even so, non-survivors tended to display lower values, probably reflecting physician efforts to intensify protective ventilation strategies in the most severely compromised. Some researchers [34, 35] have suggested that allowing higher tidal volumes in a population of young and previously healthy patients with strong ventilatory drive might reveal an attempt to restrain heavy sedation and neuromuscular blocker use. Notwithstanding this, we believe that these findings may also represent clinicians' inadequate prescription, as described in other scenarios [36] . Not unexpectedly, VCV was the most common ventilator mode used. PCV use increased throughout the study period, peaking at Day 10. This is in contrast with the recently identified trend towards decreased PCV utilization. Transition to PCV mode was associated with preceding physiological worsening, so clinicians might have perceived PCV utilization as part of a global lungprotective strategy [37] . Refractory hypoxemia was the main cause of death. As in other studies [6, 7, 11] , rescue therapies were frequently applied, with utilization highest 72 hours after admission. Recruitment maneuvers and prone positioning were the primary adjuvants utilized; ECMO and HFOV are currently not available in Argentina. A Table 3 Oxygenation and acid-base variables, and fluid balance in all patients, and in survivors and non-survivors. prolonged mechanical ventilation course was frequent as reported elsewhere [6] . NIV was the first ventilation approach in 28% of cases, with 94% later requiring invasive ventilation, as has been documented in other studies [6, 7, 11] . These common experiences should caution against delaying proper ventilatory support in this group, given that rapid deterioration is common. A recent meta-analysis suggests that NIV does not decrease the need for intubation, so evidence to support its use in severe ARDS is questionable [38] . In our study, improved outcomes with NIV could be due to milder disease, evidenced by APACHE II. The small number of patients that were not intubated precludes a statistical analysis; however, they were younger, with less severe disease and better oxygenation. Significant changes in fluid balance were late and reflected changes in survivors. Negative fluid balances could never be obtained, perhaps suggesting a continuing need for hemodynamic support: 72% of patients presented with shock [14] . On the whole, fluid balances remained between those achieved by "liberal" and "conservative" strategies of the fluids and catheters treatment trial, depending on the day evaluated [39] . Thus far, it is not clear whether the negative fluid balance has a causal role in improving outcome in ALI/ARDS, or if it simply expresses the global recovery of patients. Another important finding was that arterial pH consistently and significantly differed between survivors and non-survivors, as described elsewhere [40, 41] . During the first 72 hours acidosis had a major metabolic component, likely as a sign of hemodynamic impairment. After the first week, respiratory acidosis ensued, indicating either the effects of protective ventilation, or merely deterioration due to progressive shunt, profound ventilation/perfusion mismatch and increased deadspace. With respect to blood chemistry, the usual findings of thrombocytopenia, leukocytosis and mildly elevated creatine-kinase blood levels were present [21, 42] . Regrettably, the lymphocyte count was not recorded. In viral infections, thrombocytopenia occurred frequently. Although the mechanisms by which the 2009 influenza A (H1N1) virus causes thrombocytopenia are unknown, its lack of resolution is a marker of poor prognosis. Both leukocytosis and leucopenia have been found in hospitalized patients with 2009 influenza A (H1N1) [2, 43] ; in our study, persistent leukocytosis was associated with increased mortality. LDH elevations have been previously described in fatal cases [2] , which corresponded to our finding of higher LDH levels in non-survivors at all time points. Such elevations have also been reported in seasonal influenza [44] . In experimental studies, increased LDH is a marker of human fetal membrane cell apoptosis induced by influenza virus [45] . Finally, multiorgan failure was frequent, and predictably more severe in non-survivors. This study has several strengths: first, the clinical characteristics and time course of pandemic 2009 influenza A (H1N1) are thoroughly described and analyzed. Second, data were collected prospectively in consecutive patients and with a standardized casereporting form, representing a large, nationwide cohort. Third, temporal patterns of mechanical ventilation use, acid-base and blood chemistry variables, as well as fluid balance and organ failures, are carefully analyzed. Prognostic implications are highlighted. Finally, we present the largest experience with NIV use during the pandemic. Study limitations include the focus on mechanically ventilated patients, excluding less severe cases also admitted to the ICU. Many cases could not be confirmed because laboratories were overwhelmed with clinical samples, which is also described elsewhere [7, 14] . Data about transmission to healthcare workers were not recorded, especially regarding NIV. Currently, most information about its use during an epidemic relies upon expert opinion [46] . In 178 patients with diffuse viral pneumonitis caused by the 2009 influenza A (H1N1) virus admitted to the ICU and followed over time, ARDS was the rule, requiring high ventilation support and frequent use of rescue therapies. Death, organ failures, and evolution to prolonged mechanical ventilation were common. In most cases, noninvasive ventilation failed to prevent endotracheal intubation. Finally, elevated LDH levels, lack of recovery of platelet count and persistent acidosis and leukocytosis in non-survivors behaved as prognostic findings. • In 2009 influenza A (H1N1) patients, hospital admission with prompt indication of mechanical ventilation -a marker of severe disease -was associated with a history of symptoms of nearly one-week duration. • An initial NIV trial was not effective to avoid intubation in most patients; thus, this ventilation approach should likely be discarded in this setting. • Mortality and morbidity were frequent: death was common and was mainly caused by persistent, refractory hypoxemia. Prolonged mechanical ventilation and ICU and hospital stays were typical. • pH, platelet count, LDH and SOFA differed significantly between survivors and non-survivors over time. Lack of recovery of platelet count and persistence of leukocytosis might be markers of poor prognosis. • Every effort should be done to increase adherence to protective ventilation in the real world. Abbreviations ALI: acute lung injury; ARDS: acute respiratory distress syndrome; BMI: body mass index; CXR: plain chest X-ray film; IBW: ideal body weight; ICU: Intensive Care Unit; LDH: lactate dehydrogenase assay; LOS: length of stay; MV: mechanical ventilation; NIV: non-invasive ventilation; PaO2/FIO2: relation between patient arterial pO 2 and inspired oxygen fraction used; PCV: pressure-controlled ventilation; PEEP: positive end-expiratory pressure; PSV: pressure support ventilation; RR: respiratory rate; RT-PCR: real-time reversetranscriptase-polymerase-chain-reaction; SATI: Argentinian Society of Intensive Care; SOFA: Sequential Organ Failure Assessment; VCV: volumecontrolled ventilation; Vt: tidal volume.

Comparative analysis of mycobacterium and related actinomycetes yields insight into the evolution of mycobacterium tuberculosis pathogenesis

BACKGROUND: The sequence of the pathogen Mycobacterium tuberculosis (Mtb) strain H37Rv has been available for over a decade, but the biology of the pathogen remains poorly understood. Genome sequences from other Mtb strains and closely related bacteria present an opportunity to apply the power of comparative genomics to understand the evolution of Mtb pathogenesis. We conducted a comparative analysis using 31 genomes from the Tuberculosis Database (TBDB.org), including 8 strains of Mtb and M. bovis, 11 additional Mycobacteria, 4 Corynebacteria, 2 Streptomyces, Rhodococcus jostii RHA1, Nocardia farcinia, Acidothermus cellulolyticus, Rhodobacter sphaeroides, Propionibacterium acnes, and Bifidobacterium longum. RESULTS: Our results highlight the functional importance of lipid metabolism and its regulation, and reveal variation between the evolutionary profiles of genes implicated in saturated and unsaturated fatty acid metabolism. It also suggests that DNA repair and molybdopterin cofactors are important in pathogenic Mycobacteria. By analyzing sequence conservation and gene expression data, we identify nearly 400 conserved noncoding regions. These include 37 predicted promoter regulatory motifs, of which 14 correspond to previously validated motifs, as well as 50 potential noncoding RNAs, of which we experimentally confirm the expression of four. CONCLUSIONS: Our analysis of protein evolution highlights gene families that are associated with the adaptation of environmental Mycobacteria to obligate pathogenesis. These families include fatty acid metabolism, DNA repair, and molybdopterin biosynthesis. Our analysis reinforces recent findings suggesting that small noncoding RNAs are more common in Mycobacteria than previously expected. Our data provide a foundation for understanding the genome and biology of Mtb in a comparative context, and are available online and through TBDB.org.

Tuberculosis is still a major killer worldwide, causing an estimated 2-3 million deaths per year [1] . The sequence of the pathogen Mycobacterium tuberculosis (Mtb) strain H37Rv has been available for a decade [2, 3] , but the biology of the pathogen remains poorly understood. Available genome sequences from Mtb strains and other closely related Mycobacteria present an opportunity to bring the power of comparative genomics to the study of Mtb. We report here the results of a comparative analysis of 31 publicly available genomes (http://www.tbdb.org, Figure 1 , Table 1 ). These include eight closely related members of the Mtb complex that can cause tuberculosis disease, (two M. bovis strains and six Mtb strains). To gain further insight into the Mycobacterium cluster, we also included a related Rhodococcus (also involved in bioremediation), a pathogenic Nocardia, four Corynebacteria (two pathogens and two that are commercially useful in amino acid production), two Streptomyces (antibiotic-producing soil bacteria), Acidothermus cellulolyticus (a thermophilic actinobacteria from the hot springs of Yellowstone), Propionibacterium acnes (causative agent of common acne), and Bifidobacterium longum (a digestive track commensal often found Mycobacterium tuberculosis F11 (ExPEC) Y Y Causes TB; isolated from TB patient in S. Africa [5] Mycobacterium bovis BCG str. Pasteur 1173P2 Y Y Causes bovine TB; attenuated vaccine strain [6] Mycobacterium bovis AF2122/97 Y Y Causes bovine TB [7] Mycobacterium tuberculosis Haarlem Y Y Causes TB; MDR strain [5] Mycobacterium tuberculosis C Y Y Causes TB; isolated in NY City [5] Mycobacterium tuberculosis CDC1551 Y Y Causes TB; highly contagious & virulent strain [8] Mycobacterium ulcerans AGY99 Y Causes Buruli ulcer [9] Mycobacterium marinum Y From fish; Skin lesions in human [10] Mycobacterium leprae TN Y Y Causes leprosy [11] Mycobacterium avium 104 Y Opportunistic pathogen; can causeTB-type pulmonary infection [12] Mycobacterium avium subsp. Paratuberculosis K-10 Y Y Causes paratuberculosis; obligate pathogen of cattle [13] Mycobacterium sp. MCS Soil bacteria; degrades PAH [14] Mycobacterium sp. KMS Soil bacteria; degrades PAH [14] Mycobacterium smegmatis MC2155 Y Widely used model for Mtb isolated from human smegma; causes soft tissue lesions [12] Mycobacterium vanbaalenii PYR-1 Soil bacteria; degrades PAH [14] Mycobacterium gilvum Soil bacteria; Degrades PAH + wide variety of organic compounds [14] Mycobacterium abscessus Y Skin & soft tissue infections [15] Rhodococcus jostii RHA1 Soil bacteria important for biofuels research and bioremediation; degrades PCB + wide variety of organic compounds [16] Nocardia farcinica IFM 10152 Y Causes nocardiosis [17] Corynebacterium glutamicum ATCC 13032 Produces amino acids (Glu) [18] Corynebacterium efficiens YS-314 Produces amino acids (Glu) [19] Corynebacterium diphtheriae NCTC13129 Y Causes diphtheria [20] Corynebacterium jeikeium K411 Y Causes nocosomial infections [21] Streptomyces avermitilis MA-4680 Soil bacteria;antibiotic-producing [22] Streptomyces coelicolor A3(2) Soil bacteria;antibiotic producing [23] Acidothermus cellulolyticus 11B Hot springs of Yellowstone [24] Rhodobacter sphaeroides Gram -, motile; photosyn.; fixes N 2 [14] in yogurt). We extend this comparative analysis to other more distantly related Actinobacteria to yield additional insight into evolutionary trends. We examined protein evolution across these 31 organisms, both at the nucleotide level and at the level of protein families, including studying gene families associated with the transition from nonpathogenic soil-dwelling bacteria to obligate pathogens. Our results highlight the importance of lipid metabolism and its regulation, and reveal differences in the evolutionary profiles for genes implicated in saturated and unsaturated fatty acid metabolism. Our analysis also suggests that DNA repair and molybdopterin cofactors are expanded in pathogenic Mycobacteria and Mtb. We also identified highly conserved elements within noncoding regions using whole-genome multiple alignments and gene expression data. These conserved elements include 37 predicted conserved promoter regulatory motifs, of which 14 correspond to previously reported motifs. They also include approximately 50 predicted novel noncoding RNAs. Guided by our computational analysis, we tested and experimentally confirmed the expression of 4 novel small RNAs in Mtb. The results of our analyses are available on our website, and provide a foundation for understanding the genome and biology of Mtb in a comparative context. We used SYNERGY [27, 28] to reconstruct the phylogeny of proteins across all 31 organisms, define sets of orthologs ("orthogroups"), and construct a phylogenetic tree of the genomes (Figure 1 ). An orthogroup is defined as the set of genes descended from a single common ancestral gene in the last common ancestor of the species under consideration [28] , containing both orthologs and possibly paralogs (Methods). At each node in the phylogenetic tree, we tabulated orthogroup appearances, duplications, and losses ( Figure 2 ). Figure 2 gives an overview of the evolution of gene families within these species. Full listings of the events tabulated in Figure 2 , as well as additional information about each orthogroup, can be found on the Supplementary Information website: To examine the evolution of entire pathways and gene families, we categorized orthogroups according to GO (Gene Ontology) and GO Slim terms [29] , PFAM domains [30] , metabolic pathways, predicted regulons (sets of genes predicted to be regulated by a common regulatory protein), and groups of genes upregulated under certain lipids (Methods). We also looked for orthogroups undergoing positive selection by calculating the ratio of nonsynonymous to synonymous mutations (the d N /d S ratio). Figure 3 shows several examples of pathway or gene family profiles and the predicted evolutionary events associated with the gene family. The sort of graphic presented in Figure 3 is browsable for every pathway, PFAM, and GO term in our Supplementary Information website. Tables 2 and 3 show the PFAM and GO categories most expanded (with the most orthogroup members) in the Mtb clade relative to the non-pathogenic Mycobacteria, and Tables 4 and 5 show those most expanded in the Mycobacteria relative to the non-Mycobacteria. Despite the smaller genome sizes present in the pathogenic Mycobacteria, and the resulting background of orthogroup loss in the evolution towards pathogens, we observe significant expansions in certain gene families in the pathogenic Mycobacteria and the Mtb complex relative to non-pathogenic relatives. We also observe evidence for selection in certain families on branches leading to the pathogenic Mycobacteria, the Mtb complex, and the soildwelling Mycobacteria. As expected, many genes known to be related to pathogenicity or antigenic variability are among the groups most expanded in the Mtb clade relative to soil dwelling Mycobacteria as well as being among the categories with the most variability in copy number in their category-level profiles overall, including toxin-antitoxin genes, genes containing PE (Pro-Glu) and PPE (Pro-Pro-Glu) domains, MCE (Mammalian Cell Entry) genes, genes involved in the synthesis of the mycolic acid coat, Esx genes, and gene involved in antibiotic resistance. Complete results for all groupings are available on our Supplementary Information website. Below we focus on specific additional families showing noteworthy expansions and trends. The single most significant trend in our analysis of protein family evolution is that genes related to lipid metabolism are greatly expanded across all Mycobacteria and related organisms, consistent with previous observations [2, 31] (Table 5 ). Our analysis extends these previous observations by identifying the emergence of this expansion in lipid metabolism genes as occurring at the root node of the Mycobacteria and Rhodococcus (Figure 3 ). Genes predicted to be involved in the metabolism of saturated fatty acids are more expanded than those involved in the metabolism of unsaturated fatty acids. Using a compendium of microarray expression experiments (Methods), we compiled a list of genes upregulated in the presence of different fatty acid sources. We found that genes upregulated under unsaturated conditions have more uniform phylogenetic profiles, while those upregulated under saturated conditions, cholesterol or ceramide have expanded through duplications in pathogenic Mycobacteria ( Figure 4 ). Saturated fatty acids and cholesterol are more prevalent in an animal host than in the soil, which contains mostly unsaturated fatty acid from plant inputs. Since it is believed that Mtb uses cholesterol as a carbon source within the host [32] , this could reflect an Figure 2 Summary of SYNERGY results: Number of gains, losses, and duplication at each node. For each node, the node number is marked in black; the total number of genes present at each node is indicated in red, and the numbers of gains, losses, and duplications are indicated in parenthesis in blue http://www.broadinstitute.org/ftp/pub/seq/msc/pub/SYNERGY/index.html. adaptation to the host environment. Consistent with our observations in host-adapted Mycobacteria, Desulfovibrio desulfuricans intestinal strains contain a higher ratio of saturated to unsaturated fatty acids than soil strains of Desulfovibrio desulfuricans [33] . Our analysis also reveals differences in evolutionary profiles between genes predicted to be involved in catabolism and anabolism of lipids. Both sets of genes are expanded in soil-dwelling and pathogenic Mycobacteria, but lipid synthesis genes are additionally expanded in pathogens relative to soil dwellers. General lipid synthesis genes are expanded across the Mycobacteria, but certain groups of lipid synthesis genes, including those related to cell wall synthesis, are further expanded in the Mtb complex (see Supplementary Information) . In pathogenic Mycobacteria, the waxy mycolic acid coat helps evade the host immune system [34] . Consistent with this, we see categories related to mycolic acid synthesis showing up among the most non-uniform categories, highly expanded in the Mtb complex (see Supplementary Information) . In contrast, some lipid degradation gene families are more expanded in the soil-dwelling Mycobacteria than in the pathogens (Supplementary Data). The soil-dwellers have the unusual ability to degrade a vast array of compounds, including diverse lipids. In addition to gene family expansions, we observe evidence for selection on the coding sequence of lipid metabolism genes. In our d N /d S calculations, we observe enrichment for positive selection in lipid degradation genes on the branch leading to the pathogenic Mycobacteria (Additional file 1: Table S2 ). For example, Rv2524c, the multifunctional FAS-I polypeptide utilized during de novo fatty synthesis [35] , has the second highest d N /d S value on this branch. Additional lipid metabolism genes with elevated d N /d S values include 15 genes predicted to be involved in the β-oxidation pathway of fatty acid degradation: seven fadE (acyl coA dehydrogenase) genes, three fadD (fatty acid CoA ligase) genes, two fadB (NADPH quinone oxidoreductase/3-hydroxybutyryl-CoA dehydrogenase) genes, one fadA (acetyl-CoA acyltransferase) gene, and two echA (enoyl-CoA hydratase) genes. Hence, we observe expansions of lipid biosynthesis genes, as well as observing evidence for positive selection acting on genes within the β-oxidation pathway. Both the lipid biosynthesis and lipid degradation pathways are specialized within the pathogenic Mycobacteria. This expansion could possibly benefit the pathogen in a manner to accommodate production and modification of cell wall lipids involved in manipulation of host immune response. The lipid degradation is particularly beneficial for the long term survival of the pathogen metabolizing host lipids encountered during infection. KstR is a transcription factor known to be involved in lipid and cholesterol degradation [36, 37] . It has been recently shown that Mtb uses cholesterol as a carbon source within the host [32] . Strikingly, KstR exhibits an evolutionary history that parallels the expansion of lipid metabolism genes in the Mycobacteria, and displays a singular conservation in its regulatory binding sites. KstR appears to have evolved at the last common ancestor of the Mycobacteria and Rhodococcus. In all Figure S1 ). Remarkably, these paralogs of KstR are all absent in the pathogenic Mycobacteria. Thus, coincident with the expansion in lipid metabolism genes described above, the KstR gene appears to have emerged through gene duplication within the existing gene family of tetR-like transcriptional regulators at the last common ancestor of Mycobacteria and Rhodococcus. All other members of this gene family were subsequently lost in the Mtb complex, while the KstR protein was maintained and underwent limited sequence divergence. There is another homolog to KstR found in Mtb H37Rv (Rv3557c) that has previously been reported to also be involved in cholesterol metabolism, named KstR2 [38] . However, KstR is much more similar to the other members of the Mycobacterial tetR family discussed above than it is to KstR2. KstR2 is categorized into a separate orthogroup (orthogroup 32655) and is more distantly related to KstR. The high sequence conservation of the KstR transcription factor is mirrored in the conservation of KstR binding sites across numerous promoters. KstR binding sites are known to be highly conserved across the Mycobacteria, out to Rhodococcus and Nocardia [36] . These sites are conserved in both sequence and position within their respective promoters. In our analysis, both in searches using known transcription factor binding motifs, as well as in our de novo motif searches, a subset of KstR binding sites are the most conserved transcription factor motifs observed. They are also among the most conserved of any noncoding sequence we identified. The conservation of the KstR gene and binding sites, the emergence of KstR at the ancestor of Rhodococcus and the Mycobacteria, and the loss of KstR paralogs within the pathogenic Mycobacteria, suggests that this transcription factor and its evolving regulon have played an important role in the expansion of lipid metabolism and its adaptation to pathogenicity in Mtb. Mtb, as well as non-tuberculous Mycobacteria, differ from other bacteria in several key respects of DNA repair [ [39] [40] [41] [42] . Within the host, Mtb must combat damage to its DNA from macrophage-generated reactive oxygen and nitrogen intermediates. The mechanisms by which this is accomplished are not fully understood [43, 44] . Although genes implicated in DNA repair have not expanded in the Mtb lineage, we note that the set of genes showing positive selection on the Mtb lineage in our d N /d S analysis is enriched for genes involved in the COG category for DNA replication, recombination, and repair (Additional file 1: Table S2 ). Several of the genes in this set with highest d N /d S values are known DNA repair genes (including recA, recB, and dnaE2), and several additional genes are helicases (dnaB, helZ, and gyrB). Interestingly, we observe that recA has the highest d N / d S score of all the genes in Mtb on the branch leading to the Mtb complex, and recB also has a very high score. Mycobacteria lack a mutSL-based mismatch repair (MMR) system [42] , and it is believed that recA may be involved in compensating pathways. dnaE2 (DNA polymerase III) also has one of the highest d N /d S values on the branch leading to Mtb, and both dnaE1 (DNA polymerase III) and dnaE2 show evidence of selection on the branch leading to the pathogenic Mycobacteria. In Mtb, damage-induced base-substitution mutagenesis is dependent on dnaE2. Loss of dnaE2 activity renders Mtb hypersensitive to DNA damage, eliminates induced mutagenesis, attenuates virulence, and reduces the frequency of drug resistance in vivo [39, 45] . dnaE1 provides essential, high-fidelity replicative polymerase function [39] , and is expressed in response to DNA damage, along with dnaE2 and recA [39, 45] . We also observe positive selection for dinX (DNA polymerase IV) on the branch leading to the pathogenic Mycobacteria (branch-site model) in our d N /d S analysis (see Supplementary Information website). Most organisms use specialized DNA polymerases that are able to catalyze translesion synthesis (TLS) across sites of damage, including the dinB group of Y family polymerases. There are two dinB-family polymerases in Mtb (dinX and dinP). Unlike in other bacteria, dinX and dinP expression are not dependent on recA, the SOS response, or the presence of DNA damage, and could therefore serve a novel yet uncharacterized role in Mtb [46] [47] [48] [49] . Genes involved in the first steps of pterin cofactor (a component of the molybdenum cofactor) biosynthesis are known to be expanded in the Mtb complex [50] . Molybdenum cofactor-requiring enzymes (such as xanthine oxidase and aldehyde oxidase) could have physiological functions in the metabolism of reactive oxygen species during stress response [51] . Molybdenum cofactor is an efficient catalyst in oxygen-transfer reactions, can be used in anaerobic respiration, and can catalyze redox reactions in carbon, nitrogen, and sulfur metabolism. Recently, genes related to molybdenum cofactor protein synthesis have been shown to be upregulated under conditions of stress in Mtb [52] . Molybdenum cofactor biosynthesis has been previously linked to pathogenesis. The regulator of the moa1 locus, MoaR1, was identified as having a SNP in M. bovis BCG, but not in virulent M. bovis or Mtb [53] . In addition, moa3 is present with varying frequency in the RD1 region, which is absent in M. bovis BCG, of pathogenic strains [54] . In agreement with previous observations of expansions of molybdopterin biosynthesis genes, we observe five protein domains related to pterin cofactor biosynthesis among the top protein domains expanded in the Mtb complex compared to the non-pathogenic Mycobacteria ( Table 2 , -"d"). Among the top GO terms expanded in the Mtb clade relative to the soil dwellers (Table 3) , there are also several groups involved in pterin and molybdopterin biosynthesis. Some of these gene copies (the moa1 locus) are believed to have been acquired by lateral gene transfer on the branch leading to the Mtb complex [10, 50] . We also observe evidence for selection on molybdenumrelated genes in our d N /d S data. On the branch leading to the pathogenic Mycobacteria, several orthogroups with high log likelihood scores when testing for selection are related to molybdenum (see Supplementary Information website). The orthogroup containing BisC (biotin sulfoxide reductase, a molybdoenzyme), as well as the orthogroup containing ModA (an ABC-family molybdate transporter), are among those with the highest d N /d S values on the branch leading to the pathogens. MoaB2 is one of the highest-scoring genes on all three branches tested. There are also many categories of unknown function that are greatly expanded in the Mtb clade relative to the nonpathogenic Mycobacteria (Tables 2 and 3 , red). For example, Rv0918 (in the Pfam group of unknown function PF08681) was found in a genetic screen that facilitates isolation of mutants defective in arresting the maturation of phagosomes [55] , helping Mtb to survive within host cells. PF07161 contains four lipoproteins (LprF, LprG, LprA, LppX). LprG and LppX were found to be in vivo essential genes by TraSH analysis [56] . Sequence conservation -or phylogenetic footprintingprovides a powerful approach for identifying potential functional noncoding sequences, and has been used in a variety of eukaryotic and prokaryotic organisms to identify protein coding genes, noncoding RNAs, and regulatory elements [57, 58] . For optimal power, the organisms being analyzed must be sufficiently distant such that non-functional elements have diverged, but not so distant such that functional elements have evolved or re-arranged. Organisms within the Mtb complex are all highly similar at the sequence level, and thus by themselves do not allow for effective phylogenetic footprinting. By leveraging the evolutionary similarity of the most distantly related Mycobacteria and Actinomycetes, we gained additional power to allow us to detect functional sequences under purifying selection, albeit only those shared by at least a majority of Mycobacteria. We used this approach to predict two classes of conserved noncoding sequences: small noncoding RNAs and transcription factor binding motifs. Small noncoding RNAs (sRNAs) have been shown to play a role in regulating gene expression in numerous bacterial species [59] , including Streptococcus [60, 61] . Yet only recently were sRNAs reported in Mycobacteria [60, 62] . Using a combination of direct isolation of small RNAs, and validation by Northern blotting and 5' and 3' RACE transcript mapping, Arnvig and Young [62] first described nine sRNAs in Mtb. Subsequently, DiChiara et al. [63] describe 34 small RNAs in M. bovis BCG, of which many were conserved in both Mtb and M. smegmatis. To build on these results, we used a combination of comparative genomics, RNA-seq, and experimental validation by Northern blotting to identify additional sRNAs conserved among the Mycobacteria (Methods). Our computational results provide evidence for 50 conserved small RNAs in Mtb that have not been previously reported. It is likely that additional conserved regions are expressed under other diverse conditions. Figure 5a shows the expression and conservation map for one of our predicted RNAs in the GenomeView Browser [64] . Table 6 shows a listing of the top 12 candidate RNAs. To verify a subset of these candidate small RNAs, we used Northern blot analysis on four of the top predicted regions (Methods). The results (Figure 5b) show signals corresponding to small RNAs from each of four candidates (Table 6 , labeled 1, 2, 3, and 9). All transcripts were near the expected size, or slightly larger. Full-length gels are provided in Additional file 3: Figure S2 . Consistent with previous work, the majority of small RNAs were seen as more than one size transcript [62] . This suggests that small RNAs might be generated by processing of larger transcripts. In the RNAseq data, there are longer "tails" extending outside of the main peak that corresponds to the RNA prediction-different length RNAs could be responsible for the additional bands of higher mass. Few transcription factor binding motifs have been identified in Mtb. Transcription factors for which binding motifs have been identified include KstR [36] , DosR [67] , IdeR [68] , ZurB [69] , Crp [70] , CsoR [71] , FurA [72] , MprAB [73] , and Acr [74] . Because of the limited knowledge of transcriptional regulation in Mtb, we searched for additional motifs computationally. We combined comparative sequence analysis with microarray data to identify a large number of motifs conserved in Mycobacteria. We clustered microarray data contained in the TB database [75] and searched for upstream regulatory motifs shared in the upstream regions of the resulting clusters using AlignACE (Methods). Because of significant noise in the results, we used a set of stringent filters, including a requirement that candidate motifs be highly conserved. 37 motif instances passed our stringent filters ( Table 7 , Methods). 14 of the top 37 (38%) motif instances correspond to cases of known Mtb motifs (several known Mtb motifs were found more than once, in different clusters, or in clusters with different size parameters). In contrast, none of the top motifs showed similarity only to known E. coli or Corynebacteria motifs. Within these top motifs, we were able to identify four of the nine known Mtb motifs (DosR, IdeR, KstR, and ZurB). As described above, the KstR motif shows a much stronger signal, in terms of both conservation and information content, than any of the other motifs (top of the ranked conservation list, Table 7 ). Based on the distribution of highly conserved predicted motif instances for KstR across the genome, we predict a more general role for KstR in lipid metabolism. We see KstR motif instances near many other lipid genes not related to cholesterol degradation, in support of the view that KstR is a more general lipid regulator controlling a large regulon [36] . One of the most interesting new motif candidates that shows up in our analysis is a conserved palindromic motif, consisting of a highly conserved TAC... GTA separated by 6 bp of less well conserved sequence (marked with an X in Table 7 ) that is found in clusters of 2-3 closely spaced sites upstream of several genes related to fatty acid metabolism ( Figure 6 ). There is a cluster of 3 evenly spaced sites upstream of Rv3229c (linoeyl-coA desaturase), a cluster of 2 sites upstream of the adjacent Rv3230c (oxidoreductase), and a cluster of 3 sites upstream of Rv2524c (fatty acid synthase). This is the second highest-scoring new motif identified (Table 7) . This motif shows up as one of the top motifs associated with the clusters of genes upregulated under saturated fatty acid conditions (specifically palmitate). To better understand Mtb, we performed a comparative analysis of 31 organisms from the Tuberculosis Database. We studied the evolution of protein families and metabolic pathways, looked for proteins with evidence Figure 5 New predicted RNAs. a) An example of a new predicted RNA. This is the RNA2 in Table 6 . This figure shows a screenshot from the GenomeView Browser [64] . The light blue bars show the coding regions (Rv1230c and Rv1231); the tan bar shows the conserved region predicted by Gumby [65] ; and the green bar shows the region predicted to fold by Evofold [66] . The yellow and green plots in the center show the RNA-seq data. Green signifies reads from the negative strand, and yellow shows the total reads (positive and negative strands). The multiple alignment is shown on the bottom (darker grey signifies a higher degree of conservation; red signifies no alignment at that position). You can see that this predicted RNA region is conserved through M. avium. The rulers at the top show the gene structure. Small red squares show where stop codons are present all six reading frames, indicating that this intergenic region is unlikely to be a protein-coding region missed in the annotation. b) Northern blots validating four of the new, predicted small RNAs (RNA1, RNA2, RNA3, and RNA9 in Table 6 ). 1 Conserved intergenic regions determined by Gumby. 2 Indicates whether this region is predicted to fold by Evofold. 3 Region in M. smegmatis that aligns with the conserved region in Mtb, and its corresponding RPKM value. 4 Tested experimentally 5 Orientation relative to neighboring genes. The first and last characters give the strands of the flanking genes; the middle character gives the strand for the predicted RNA. of selection, and searched for new noncoding RNAs and transcription factor binding site motifs. The most striking features of our analysis are related to lipid metabolism and its regulation. In addition to observing a general expansion of lipid metabolism genes in the Mycobacteria and Rhodococcus, we observe increased expansions of genes related to saturated fatty acid metabolism in the pathogenic Mycobacteria compared to the soil-dwelling Mycobacteria. We also note differences in evolutionary profiles for catabolic and anabolic lipid metabolism genes, and evidence for positive selection in lipid metabolism genes. The cis-regulatory elements bound by the KstR protein, a known regulator of lipid/cholesterol metabolism, are among the strongest, most highly conserved noncoding signals across the Mycobacteria. Both KstR and its binding sites are highly conserved, appearing at the last common ancestor between Rhodococcus and the Mycobacteria. Within our set of organisms, we examine the evolution of pathogenicity, moving from the soil-dwelling Mycobacteria up to the intracellular parasites of the Mtb complex. We see expansions of many known gene families related to pathogenicity (PE/PPE genes, antibiotic resistance genes, genes involved in the synthesis of the mycolic acid coat, MCE genes, and Esx genes). By similarity of phylogenetic profiles, we can predict likely candidates for novel gene families related to pathogenicity. For example, we see similar expansions in gene families related to biosynthesis of molybdopterin. We further observe evidence of positive selection on molybdenum-related genes, providing further support for the importance of molybdenum in these pathogens. On the branch leading to the pathogenic Mycobacteria, we also observe evidence for positive selection in genes related to replication, recombination, and repair. It is possible that these DNA repair-related processes give the pathogenic Mycobacteria an advantage when dealing with the assault on its DNA by macrophage-generated reactive oxygen and nitrogen intermediates. Our whole-genome alignments, coupled with RNA-seq and microarray data, allowed us to predict novel noncoding features, including small RNAs (four of which we have validated experimentally), and potential transcription factor binding sites. The main forces driving genome evolution in prokaryotes include gene genesis, lateral gene transfer, and gene loss. Our analysis of protein evolution using SYNERGY does not examine whether orthogroups appearing have arisen by lateral gene transfer or by gene genesis involving duplication and divergence from other orthogroups. A detailed comparison to categorize these orthogroup appearances according to lateral or vertical gene transfer is beyond the scope of this study, but other studies indicate that lateral gene transfer has played a significant role in Mycobacterial evolution and the evolution of pathogenesis [79] [80] [81] [82] [83] . A recent paper suggests that the Mycobacterial genome has been shaped by a biphasic process involving gene acquisition (including lateral gene transfer) and duplications followed by gene loss [79] . Other studies report numerous genes, including a large number involved in lipid metabolism, that have been acquired by horizontal gene transfer at different phylogenetic strata and have led to the emergence of pathogenesis in Mtb [80, 81] . Previous studies indicate a possible more ancient lateral gene transfer of fatty acid biosynthesis genes from α-proteobacteria to actinobacteria [84] . However, genetic studies show that the Mtb complex and pathogenic Mycobacteria do not exchange genetic material frequently [85, 86] , so there is limited lateral gene transfer within the Mtb complex. We are currently performing high-throughput Chromatin Immunoprecipitation (ChIP)-Seq experiments in several different Mycobacteria, including Mtb, M. smegmatis, and M. vanbaalenii [87] . We plan to integrate the information obtained from our comparative analysis with data coming from these high-throughput experiments, as well as other 'omic datasets, using a systems biology approach. This will enable construction of gene regulatory networks for Mtb, and examination of their evolution across species. The 31 organisms used in our analysis are described in Table 1 . These genome sequences are all contained in Table 7 Motifs passing our set of stringent filters, ranked by their degree of conservation (Continued) 50 k indicates the value of k in the k-means clustering process (50, 100, 200, or 250) b MAP score indicates the AlignACE MAP score [76] c Specificity score [77] d CompareACE score ≥ 0.7 to the alignment for this known motif e CompareACE score to its reverse complement f number of ScanACE hits in the genome that are conserved in ≥ 8 genomes g sequence logo [78] the TB database (TBDB) [75] . The three unpublished sequences generated at the Broad Institute (M. tuberculosis F11, M. tuberculosis Haarlem, and M. tuberculosis C) are high-quality genome sequences. M. tuberculosis F11 and M. tuberculosis Haarlem are finished, and M. tuberculosis C has 6.7× coverage and 4 scaffolds. The Broad Institute sequencing read pipeline interacts with the sample management system to ensure the read is associated with the correct sample. Vector identification, length checks and quality clipping were performed on all reads. Contamination checks and organism checks were also performed using a kmer-based algorithm that can compare sequence to a profile from any organism. The SYNERGY algorithm [27, 28] was applied to the 31 genomes in Table 1 . SYNERGY organizes groups of genes across organisms into orthogroups, or groups of orthologs and paralogs, which consist of all the genes descended from a single ancestral gene in the species' last common ancestor. SYNERGY also associates orthogroups with a gene tree, from which we can derive an "extended phylogenetic profile", showing the gene copy number in each extant organism and at each ancestral node. Importantly, by reconciling an organism tree with each gene tree, SYNERGY provides an evolutionary scenario for each gene tree predicting where all losses, gains, and duplications occurred in its evolution. These lists of losses, gains, and duplications contain actual evolutionary events, as well as artifacts caused by genes that could not be properly categorized by SYNERGY. However, we observe that SYNERGY is effective at properly categorizing genes into orthogroups, and the SYNERGY orthogroups were very useful in our analysis. Analysis of the 31 genomes resulted in a total of 32,505 orthogroups, including those containing single genes from only a single genome (below). There were 177 "uniform" (1:1:1:1...) orthogroups representative of some of the most conserved and indispensible housekeeping genes. Additional file 4: Figure S3 summarizes the SYNERGY orthogroups. We started running SYNERGY using an initial phylogenetic tree generated using orthologs based on bidirectional best BLAST hits. The list of uniform orthogroups from the first SYNERGY run was used to construct a refined phylogenetic tree. SYNERGY was then re-run using the refined phylogenetic trees. To generate our final phylogenetic tree, the final set of 177 31-way orthologs (31-way uniform orthogroups from the SYNERGY analysis) were aligned according to their nucleotide sequences with CLUSTALW [88] and concatenated, distances were computed with Phylip's DNADIST algorithm [89] , and Phylip's FITCH algorithm was used to create the tree. Because of the similarity of the genomes within the Mtb complex, we were not able to resolve the phylogeny using only these 177 proteins that are uniform across all 31 organisms. In order to better resolve the tree within the Mtb cluster, we computed a separate tree using 1747 orthogroups that are uniform across the Mtb cluster and M. ulcerans, which we used as an outgroup. Using this expanded gene set, we were able to resolve the tree for the Mtb cluster. Bootstrap analysis was performed to validate tree topologies. Phylip's SEQBOOT was used to create 1000 bootstrap input replicates for each tree. Phylip's CON-SENSE was used to obtain a bootstrap tree (Additional file 5: Figure S4 ) Metabolic pathways and functional groups EFICAZ [90] was used to assign EC numbers for proteins in all 31 organisms. Metabolic pathways were constructed in Biocyc [91, 92] . An orthogroup was considered to be part of a metabolic pathway if any of its component genes had been identified as part of that pathway using this pipeline. We obtained the Gene Ontology (GO) [29] and GO Slim terms for each of the 31 organisms using BLAS-T2GO [93] . PFAM assignments [30] were taken from http://www.tbdb.org [75] . An orthogroup was associated with a GO, GO Slim, or PFAM descriptor if greater than half of its protein members were associated with that descriptor. For each node in the phylogenetic tree, we tabulated orthogroups lost, gained, or duplicated. Using GO terms, GO Slim terms, and PFAM domain groupings with less than 500 members, we calculated over-representations within losses, gains, and duplications each of these groupings at each node using the hypergeometric test. A complete summary of gains, losses, and duplications for all nodes in the phylogenetic tree is available on our supplementary information website. Extended phylogenetic profiles for each category (metabolic pathways, GO terms, GO Slim terms and PFAM categories) were obtained from SYNERGY output by summing the phylogenetic profiles from their component orthogroups. We define a category-level phylogenetic profile as the sum of its component orthogrouplevel phylogenetic profiles. The evolution of each of these categories can be quickly visualized on our website. Since genes with the same phylogenetic profile can be linked functionally [94] , the webpage for each category contains a link to other categories with similar phylogenetic profiles (Methods). Categories with the most similar profiles were obtained by calculating Euclidean distances to all other profiles. Instances of expanded or missing pathways across the 31 organisms will have non-uniform pathway-level phylogenetic profiles. Thus we tabulated the number of genes in each genome for each category, and automatically searched for gene categories whose copy number (normalized for genome size) had the most non-uniform distribution across the 31 organisms in order to identify the most significant examples of expansions or losses. To identify categories with bimodal properties (such as a categories with a loss or a large expansion on only certain branches of the phylogenetic tree), we clustered each profile into two groups and looked for the pathways with the greatest separation between the two clusters. We used k-means (k = 2) to cluster the profile vectors, and compared the intra-and inter-cluster point-to-centroid distances to find the clusters with the greatest separation. We ranked categories by this separation to find bimodal categories. We further select those that have at least five organisms in the smallest of the two clusters, and an average of at least five genes per genome. P-values are calculated from a T-test between the values for the two groups, with Bonferroni correction applied. In our Supplementary Information website we list those categories with p < 0.05, ranked by the difference between their inter-to intra-centroid distances. When we select the metabolic pathways, PFAM domains, and GO terms with the most non-uniform category-level phylogenetic profiles overall, we find that many of the top categories are lipid metabolism-related categories expanded in the Mycobacteria. We also measured the similarity between evolutionary profiles to find the PFAM categories and GO terms with the biggest difference between pre-defined sets of organisms. For example, we compared both the Mtb complex and a group consisting of other pathogenic Mycobacteria to the set of soil-dwelling Mycobacteria in order to examine the evolution of soil-dwelling, free-living Mycobacteria into more pathogenic Mycobacteria that require a host to survive. We used the following categories: 1. All Mycobacteria (excluding M. leprae because of its massive gene loss). 2. All non-Mycobacteria in our set (excluding Nocardia and Rhodococcus because of their similarity to Mycobacteria 6. R. jostii RHA1 and N. farcinia We calculated differences between two sets of organisms exactly as we calculated distances between clusters (above). However, rather than using different clusters of organisms determined by k-means clustering, we used these pre-defined clusters of organisms. We looked at distances between the following sets of organisms: 1-2, 3-4, 3-5, 3-6, 4-5, 4-6, 5-6. For each PFAM domain or GO term represented in at least two organisms in these pairings, we calculated p-values for the differences between the profile values by T-test (Bonferroni-corrected by the number of PFAM domains represented in that set of organisms) and computed inter-and intra-centroid distances (as described in the above paragraph). We compiled lists of those that are most expanded and a list of those most contracted across these pairings. On our website we have included complete lists of PFAM categories, including those that do not make the strict Bonferroni-corrected p-value cutoff. Many potentially interesting expansions do not make the overly conservative Bonferroni-corrected p-value cutoff [95, 96] . Using a compendium of 946 microarray experiments from the TB database [75] , we used several different clustering methods to generate predicted regulons. We searched the upstream regions of these regulons for shared transcriptional regulatory motifs. We clustered microarray data by hierarchical and k-means clustering. Because real regulons can be of varying sizes, we performed k-means with k = 50, 100, 200, and 250, then used all the resulting clusters for further analysis. We found that the clusters obtained from hierarchical clustering were not very useful because their size distribution did not approximate that of real regulons as well as those from k-means; therefore we did not analyze clusters from hierarchical clustering further. We used AlignACE [97] to search the upstream regions of the genes in these clusters for motifs. We used the methods for operon prediction, selecting upstream regions, and applying AlignACE to prokaryotic genomes as described in McGuire et al. [77] . Briefly, because of the presence of operons in prokaryotes, we must choose the upstream region of the operon head rather than the region immediately upstream of the gene of interest. Since it is more important to include the correct region than to erroneously include extra incorrect regions, we use a loose operon definition and include sequences for several different possibilities if there is any ambiguity. We look upstream of our gene of interest and select all intergenic sequences until we encounter either a divergent intergenic region or an intergenic region longer than 300 bp. Motifs of interest were selected by applying a set of filters: specificity score [77] , quality of alignment (AlignACE MAP score) [97] , palindromicity [77] , and conservation. To determine the degree of conservation, a search matrix was constructed for each motif. Each of the other genomes was searched with this search matrix using CompareACE, and N-way conserved sites were identified. N-way conserved hits are hits identified upstream of orthologous genes in N genomes, where orthology is defined by membership in the same SYNERGY orthogroup. To select interesting motifs we required specificity score < 1e-10, palindromicity > 0.7, MAP score > 5, and at least 8 sites conserved in 8 genomes. Motifs were compared to a library of search matrices for 9 known Mtb motifs (Acr, Crp, CsoR, DosR, FurA, IdeR, KstR, MprAB, and ZurB), as well as a library of 55 E. coli motifs [98] and 22 Corynebacterial motifs [99] . Comparison of motifs was done using CompareACE [76] . Within each experiment, we extracted a list of genes upregulated 1.5 and 2 standard deviations above the mean. For each category, we considered a gene to be upregulated if it was upregulated in more than 50% of the experiments making up that category. We then searched for genes that were only upregulated under certain conditions or sets of conditions. We looked at the evolution of these sets of Mtb H37Rv genes by taking the other members of their orthogroups across all 31 other organisms. Evolution of these groups can be visualized in our supplementary information http://www.broadinstitute.org/ftp/pub/seq/ msc/pub/SYNERGY/index.html. We used PAML to calculate d N /d S values according to several different evolutionary models [100, 101] . Since orthogroups contain paralogs as well as orthologs, we used the gene trees output from SYNERGY when running PAML. Some orthogroups may contain single-copy orthologs in only two closely related organisms, whereas others could contain paralogs in all 31 organisms. For the basic model, we used the following parameters: model = 0 and getSE = 1 (to calculate standard errors). This simple evolutionary model gives one value of d N /d S for each orthogroup, averaged over all lineages as well as all positions in the gene [102] . While this model does not reflect the evolutionary history that has taken place, it is nevertheless a very blunt yet efficient tool for observing selection. To gain insight into the evolution of the three major clades of the phylogenetic tree we also used a "branch model" where a different d N /d S value is allowed on a "foreground" branch (but d N /d S is averaged along positions in the protein) [103, 104] . This was done in PAML by using "Model = 2". We compared this model to the basic model using a log-likelihood χ 2 test with d.o.f. = 1. For each of the three foreground branches, we used a Bonferroni correction equal to the number of orthogroups present at the branch. We ran this separately for three different "foreground" branches on the phylogenetic tree (labeled in Figure 1 ): A) The branch leading to the Mtb complex; B) The branch leading to pathogenic Mycobacteria; and C) The branch leading to soil-dwelling, non-pathogenic Mycobacteria. The loglikelihood model that we use here compares this branch model to the simple model with a single value of d N /d S described above, and tests whether the model allowing d N /d S to differ on the foreground branch fits the data better than the basic model. Branch-site models allow d N /d S to vary across branches of the tree and among sites in the protein. We also used the branch-site model of Zhang and Nielsen [100] using Model = 2, NSsites = 2, and fix_blength = 2. We used the model = 0 calculations to determine branch lengths for the branch-site model calculations to save computational time. We compared the results for a subset of the orthogroups with and without fixed tree lengths and determined there was little difference in the results). We chose the same three sets of branches (A-C) that we used for the branch model described above. We compared this model to the corresponding null model using a log-likelihood χ 2 test with d.o.f. = 1 [100] . For each of the three foreground branches, we used a Bonferroni correction equal to the number of orthogroups present at the branch. The branch-site model was the most informative. We calculated the functional group over-representations separately for each functional group dataset. These datasets included 21 COG categories, 168 KEGG categories, 749 metabolic pathways, and 7 additional Mycobacteria-specific groupings (PE genes, PPE genes, toxinantitoxin genes, DosR regulon, esx genes, Rv0474 regulon, and the KstR regulon). We multiplied the hypergeometric p-values by a Bonferroni correction equal to the number of categories or tests performed. We constructed whole-genome alignments of all 31 organisms, as well as subsets including only Mycobacterial organisms, and organisms within the Mtb complex. These alignments can be downloaded from our website. Our whole genome multiple alignments reveal unannotated stretches of conservation in noncoding regions including transcription factor binding sites in promoter regions, noncoding RNAs, and mis-annotated proteins. To generate whole-genome multiple alignments, we first aligned the reference genome to each target genome in a pairwise manner. The process of pairwise whole-genome alignment consists of using Pattern-Hunter [105] to identify anchors of local alignment, grouping collinear anchors separated by a limited distance into chains, filtering out smaller chains that shadow larger ones, and finally using LAGAN [106] to globally align the entire chain. Once all genomes have been aligned to the reference, we then identified intervals of the reference that map tightly to a single interval of some or all of the target genomes, and we consider these the endpoints of blocks of multiple alignment. These blocks are generally smaller than any precursor pairwise alignment, because a rearrangement or loss of detectable similarity in any genome will truncate the block for all member genomes. We then ran the multiple aligner MLAGAN on each block. Finally, to facilitate searches for constrained regions of the reference, we projected the blocks onto the reference genome, effectively unwinding all genome rearrangements in the target genomes relative to the reference. We visualized the alignments in the Geno-meView browser [64] . We used Gumby [65] to select conserved regions in our multiple alignments using a value of p < 0.5. In a multiple alignment of all 19 Mycobacterial genomes, we identified 4697 regions of conservation overlapping coding genes in the reference annotation, and 394 regions in intergenic regions. We also used the method of Ruzzo and Tompa [107] to identify conserved regions. Scores were normalized to the background inferred from the 3 rd-base frequencies. For all H37Rv coding sequences, all bases in the third position were extracted from the 31-way multiple alignment. These were concatenated in a new multiple alignment only containing third bases. From this new multiple alignment we calculated the baseline conservation which is used to normalize the conservation scores for the regular alignment. Both sets of highly conserved regions can be viewed as alignment tracks for the Geno-meView browser [64] , downloadable on our website. We predicted regions likely to form RNAs within the conserved intergenic regions of our multiple alignment of 19 Mycobacteria, using Evofold [66] . We divided the intergenic region into 240-bp segments, tiled by 80 bp, to run Evofold. Looking within intergenic regions, we identified 536 regions with Evofold (regions greater than 5 bp in length with length-normalized folding potential score > 0.2). We examined these 536 regions, as well as the 394 conserved intergenic regions found by Gumby, to see if any of these showed significant expression in our logphase Mtb RNA-seq data. We calculated RPKM [108] values for each of these regions. We examined the regions with RPKM value ≥200 and a number of RNAseq reads ≥ 20. We eliminated an additional 35 regions which corresponded to known RNAs from the Mtb annotation, or RNAs similar to those found in M. bovis and Streptococcus [60] [61] [62] [63] , including 26 tRNAs, 2 riboswitches, and 3 found in other organisms. To select intergenic regions with high levels of expression that do not correspond to UTRs, we also calculated RPKM values for the 100 bp regions of the flanking genes closest to the intergenic regions. We selected those intergenic regions with the highest ratio of the RPKM value of the region of interest (within the intergenic region) to the RPKM of the start/stop of the flanking genes. We also looked for regions with a gap in expression between the gene and the region of interest. This will eliminate many regions that merely correspond to UTRs, and select for regions that are disproportionately expressed within the intergenic region only. We found this method to be most useful for selecting regions of interest, and successfully enriched our top hits for previously known small RNAs. The top 50 predicted RNAs can be viewed as a track in the GenomeView browser (see Supplementary Information) . We further examined log-phase RNA-seq data from M. smegmatis to confirm that many of the orthologous regions also show expression in M. smegmatis. Mycobacterium tuberculosis H37Rv and M. smegmatis were grown at 37°c in 7H9 media supplemented with 10% ADC (Becton Dickinson), 0.2% glycerol and 0.05% Tween 80. For log phase, cells were grown to OD 540 0.2. Roller bottles were used for culturing M. tuberculosis, and shaker flasks for M. smegmatis. Bacterial pellet from log-phase cultures of M. tuberculosis and M. smegmatis were resuspended in TRIzol reagent (Invitrogen) and immediately transferred to 2 ml screw-cap tubes containing 0.1 mm zirconia/silica beads (BioSpec Products). M. tuberculosis cells were lysed using a FastPrep-24 bead-beater (MP Biomedicals) 3 times for 30 seconds each at speed 6. M. smegmatis cells were lysed using MagNalyser (Roche). Samples were kept on ice for 1 min between pulses. The TRIzol extracted RNA was treated twice with DNAse and further purified using RNAeasy kit (Qiagen). We generated mRNA-seq libraries for sequencing on Illumina's GA Sequencer (San Diego, CA). 2 μg purified RNA was depleted of ribosomal RNA using Ambion's MICROBExpress Kit (Austin, TX) as per manufacture's recommended protocol. The enriched mRNA was used to prepare libraries using Illumina's Directional mRNAseq Library Prep v1.0 protocol. Briefly, 100 ng mRNA was fragmented with cations and heat, end-repaired, adapted by sequential ligation of unique 5-prime and 3prime adapters, reverse transcribed, PCR amplified, and purified using Agencourt's AMPure Beads (Beverly, MA). The libraries were visualized on an Agilent 2100 Bioanalyzer (Santa Clara, CA) and found to have the expected average fragment length of~250 bp. Total RNA was isolated from Mtb as described previously [109] with minor modifications. Briefly, logphase cells were pelleted, resuspended in TRIzol (Invitrogen), and transferred to Lysing Matrix B tube (QBiogene). The cells were lysed using MagNalyser (Roche), and RNA extracted with Trizol reagent as instructed by the manufacturer. RNA was treated with Turbo DNase (Ambion) for 30 minutes at 37°C twice and purified further using TRIzol solution and 100% Ethanol. Total RNA was separated on 10% TBE-Urea acrylamide gels (Bio-Rad) and electroblotted onto Hybond N + membranes (GE Healthcare). After UV cross-linking the membranes were pre-hybridized and hybridized with labeled probes at 48°C as per the DIG manual (Roche). Probe sequences are CGATGGTCGAAAAGGAACTCGA-TACGGCTATGCGGTTCT (RNA1), AGTTCACGA

Proteasome-Dependent Disruption of the E3 Ubiquitin Ligase Anaphase-Promoting Complex by HCMV Protein pUL21a

The anaphase-promoting complex (APC) is an E3 ubiquitin ligase which controls ubiquitination and degradation of multiple cell cycle regulatory proteins. During infection, human cytomegalovirus (HCMV), a widespread pathogen, not only phosphorylates the APC coactivator Cdh1 via the multifunctional viral kinase pUL97, it also promotes degradation of APC subunits via an unknown mechanism. Using a proteomics approach, we found that a recently identified HCMV protein, pUL21a, interacted with the APC. Importantly, we determined that expression of pUL21a was necessary and sufficient for proteasome-dependent degradation of APC subunits APC4 and APC5. This resulted in APC disruption and required pUL21a binding to the APC. We have identified the proline-arginine amino acid pair at residues 109–110 in pUL21a to be critical for its ability to bind and regulate the APC. A point mutant virus in which proline-arginine were mutated to alanines (PR-AA) grew at wild-type levels. However, a double mutant virus in which the viral ability to regulate the APC was abrogated by both PR-AA point mutation and UL97 deletion was markedly more attenuated compared to the UL97 deletion virus alone. This suggests that these mutations are synthetically lethal, and that HCMV exploits two viral factors to ensure successful disruption of the APC to overcome its restriction on virus infection. This study reveals the HCMV protein pUL21a as a novel APC regulator and uncovers a unique viral mechanism to subvert APC activity.

Regulation of protein degradation plays a key role in many cellular processes ranging from cell cycle progression, innate immunity, and antigen presentation to the turnover of misfolded or oxidized proteins. Most degradation is carried out by the ubiquitin-proteasome system (UPS). Ubiquitin is added to proteins by a cascade of ubiquitin conjugating enzymes, resulting in a polyubiquitinated protein which is subsequently degraded by the 26S proteasome. As a means to regulate protein function, it is no surprise that many viruses have co-opted the UPS for their own benefit. Viruses can promote proteasome degradation of antiviral host proteins either by encoding their own E3 ubiquitin ligase, targeting proteins to a cellular E3 ligase, or even inducing ubiquitin-independent degradation of targets. Examples of viral E3 ligases include the herpes simplex virus-1 protein ICP0 [1] and Kaposi's sarcoma-associated herpesvirus proteins K3 and K5 (for a review, see [2] ). Viral proteins that can hijack a cellular E3 ligase include human immunodeficiency virus-1 vpr and vif (for a review, see [3] ), paramyxovirus V [4] , and human papillomavirus E6 and E7 (for a review, see [5] ). Finally, the human cytomegalovirus (HCMV) protein pp71 uses a ubiquitin-independent mechanism to target the Rb and hDaxx proteins [6, 7] . In fact, pharmacological inhibition of the proteasome blocks multiple stages of the viral life cycle, suggesting that viruses rely on activities of the UPS for their replication [8] [9] [10] [11] [12] . On the other hand, viruses must also modulate cellular E3 ligase activity in order to replicate because ubiquitination regulates many important cellular processes central to virus infection. The SV40 large T antigen inhibits the SCF fbw7 ubiquitin ligase to increase cyclin E levels [13] , and influenza virus NS1 inhibits TRIM 25-mediated ubiquitination of RIG-I, thereby attenuating interferon production [14] . The anaphase-promoting complex (APC) or cyclosome is a macromolecular complex that contains cullin-ring E3 ubiquitin ligase activity and is conserved across all eukaryotes (for a review, see [15] ). It has at least eleven subunits and two co-activator proteins (CDC20 (cell-division cycle protein 20) and Cdh1 (CDC20 homologue 1)), which are separated into three subcomplexes. These include the cullin-ring ligase domain (composed of APC2, 10, and 11), the specificity arm (composed of APC3, 6, 7, and 8), and the bridge (composed of APC1, 4, and 5). Cdh1 and CDC20 activate APC activity to prevent premature entry into S phase and to promote progression through mitosis, respectively. The APC complex ubiquitinates more than 40 proteins, including A-and B-type cyclins, to regulate their stability. It also regulates degradation of its own coactivator proteins, Cdh1 and CDC20, as a form of feedback regulation. Due to its central role in cell cycle progression, the APC is also a promising target for anti-cancer therapeutics [16] . Many viruses modulate the host cell cycle to establish optimal conditions for their replication. Several viral proteins have been reported to target the APC, possibly to force the cell into an S phaselike biochemical environment to promote efficient viral replication. Proteins from adenovirus, chicken anemia virus, human papillomavirus, human T-lymphotropic virus, hepatitis B virus, parapoxvirus, and HCMV have been reported to regulate the function of the APC [17] [18] [19] [20] [21] [22] [23] [24] . However, the mechanisms used by these viruses during infection to subvert the APC are largely unknown. HCMV is a globally important opportunistic pathogen that causes severe diseases in immunocompromised individuals and is the leading viral cause of congenital diseases. This virus stimulates cell cycle progression of quiescent cells into an S phase-like environment but concurrently blocks host DNA synthesis [25] . HCMV promotes cell cycle progression likely in part by inactivating Rb [6, 26] and regulating the APC [24, 27, 28] . It has been reported that HCMV has two means to regulate the APC. The multifunctional viral kinase pUL97 phosphorylates the APC coactivator Cdh1, thus likely inhibiting its activity [24] . Nonetheless, abrogation of UL97 alone only results in a modest increase in APC activity during infection [24] . Independent of UL97mediated Cdh1 regulation, HCMV also induces degradation of two APC subunits, APC4 and APC5, leading to the dissociation of the complex during infection [24] . The viral factor and associated mechanism responsible for regulating degradation of the APC subunits have not been identified. In this study, we demonstrate that the HCMV protein pUL21a interacts with the APC, resulting in proteasome-dependent degradation of APC4 and APC5. Expression of pUL21a dissociates the APC cullin-ring ligase subcomplex from its specificity arm. This regulation alters APC activity and increases levels of a subset of APC-regulated cell cycle proteins. We have identified residues proline-arginine (PR109-110) in pUL21a to be critical for its ability to bind and regulate the APC. A mutant virus in which the viral ability to regulate the APC is abrogated by both alanine substitution of proline-arginine residues in pUL21a and UL97 deletion is markedly more defective compared to the pUL97 deletion virus alone. This suggests that HCMV has evolved an invasive strategy of using both viral factors to regulate the APC to facilitate its infection. Our study has identified the HCMV protein pUL21a as a novel APC regulator and elucidated a unique mechanism to subvert APC activity. pUL21a Interacts with the Anaphase-Promoting Complex (APC) HCMV pUL21a is a 15 kDa, highly unstable protein that is expressed with early kinetics [29] . One identified function of this protein is to facilitate efficient viral DNA synthesis [30] . However, this protein shares no significant homology with any known protein. To provide mechanistic insight into its activity, we used a proteomics approach to identify interacting partners of pUL21a during infection. We created a recombinant virus (ADgfpUL21a) in which the UL21a coding sequence was tagged with the green fluorescent protein (GFP) coding sequence. This virus grew with wild-type kinetics, and the tagged protein was fortuitously much more stable than native pUL21a [29] . A GFP tag can stabilize certain fusion proteins [31] , and made it possible to detect interacting proteins in our study. We infected fibroblasts with either ADgfpUL21a or control HCMV (ADgfp) that expressed free GFP only. At 48 hours post infection (hpi), we isolated the protein complexes from infected cells by a rapid one-step immunoaffinity purification on magnetic beads coated with GFP antibody-coupled protein A. Electrophoresis analysis revealed multiple protein bands that were specific to the pUL21a-containing sample ( Figure 1A ). We analyzed pUL21a-specific protein bands by mass spectrometry and identified the proteins depicted with arrows as APC specificity arm subunits, APC3, APC7, and APC8 (Table S1) . We validated these interactions in HCMV infected cells by coimmunoprecipitation followed by immunoblot analysis. Here we used APC3 and APC8 as the marker for the APC complex. Pulldown of pGFP-UL21a, but not the GFP control, isolated both APC3 and APC8 ( Figure 1B ). The lower band detected by APC8 antibody was nonspecific as it neither co-immunoprecipitated with APC3 antibody ( Figure 1B ) nor was affected by shRNA knockdown of APC8 ( Figure S1 ). In the reciprocal experiment, APC3 antibody co-immunoprecipitated APC8 and pGFP-UL21a but not GFP. Neither GFP nor APC3 antibody co-immunoprecipitated cellular PCNA ( Figure 1B) , and an antibody against HA did not co-immunoprecipitate any of the proteins detected here (data not shown), thus providing additional evidence for the specificity of these interactions. As pGFP-UL21a is co-immunoprecipitated with multiple APC subunits, we interpret the result to suggest that pUL21a binds to the APC complex, even though the precise subunit where pUL21a directly interacts with remains unknown. To determine if this interaction also occurred with native pUL21a, we performed co-immunoprecipitation assays on lysates from cells infected with wild-type virus (ADgfp), and we also included lysate from cells infected with UL21a deletion virus (ADsubUL21a) as a negative control ( Figure 1C ). Infected cells were treated with proteasome inhibitor, MG132, as pUL21a was highly unstable and otherwise could not accumulate to levels allowing reproducible detection of this interaction [29] . In the In this study, we report an intriguing mechanism used by human cytomegalovirus (HCMV) to regulate a cellular E3 ubiquitin ligase, the anaphase promoting complex (APC). The ability to hijack the ubiquitin-proteasome system for regulating protein degradation and to manipulate the cell cycle for viral genome synthesis is critical in many viral infections. The APC is a master cell cycle modulator that targets a number of regulatory proteins for proteasomal degradation. It can prevent cells from entry into S-phase, thus creating a hindrance for viruses needing to coerce cells into a cellular environment favorable for viral DNA synthesis. We have identified an HCMV protein, pUL21a, which uses a seemingly counterintuitive mechanism to regulate the APC. It interacts with the APC to target the subunits of this ubiquitin ligase for proteasomal degradation. This causes disruption of the complex and reduces its activity. Furthermore, a virus lacking pUL21a and pUL97, which is another HCMV-encoded APC regulator, was highly attenuated when compared to loss of UL97 alone, suggesting that HCMV uses two proteins to fully disarm the APC. This study identifies a herpesviral protein that uses a unique, proteasome-dependent mechanism to regulate the activity of this prominent cellular E3 ubiquitin ligase. presence of MG132, the level of native pUL21a was markedly increased and could be co-immunoprecipitated with APC3 antibody. This interaction was specific as the antibody did not co-immunoprecipitate PCNA or the viral DNA polymerase accessory factor UL44. To test if pUL21a was able to bind to the APC in the absence of other HCMV proteins, we performed co-immunoprecipitation assay on lysates from 293T cells transfected with constructs expressing the GFP-amino terminal tagged UL21a (gfpUL21a wt ) or UL21a carrying two stop codons at its amino terminus to abrogate pUL21a expression (gfpUL21a stop ) ( Figure 1D ). Both gfpUL21a wt and gfpUL21a stop were expressed at equal levels but only gfpUL21a wt associated with APC3 or APC8. Additionally, APC3 antibody co-immunoprecipitated gfpUL21a wt but not gfpUL21a stop . We conclude that pUL21a interacts with the APC and does not require other HCMV proteins for this interaction to occur. at an MOI of 5, collected at 48 hpi, and were immunoprecipitated with GFP antibody. Eluted proteins were run on an SDS-containing polyacrylamide gel and silver stained. The bands indicated with an arrow were identified by mass spectrometry as APC3 (100 kDa), APC7, and APC8 (both at 65 kDa). (B) GFP-tagged pUL21a interacts with the APC in HCMV infection. MRC-5 cells were infected as described in panel A, and lysates were subjected to coimmunoprecipitation with GFP or APC3 mouse monoclonal antibodies. Cell lysates and eluted proteins were analyzed by immunoblotting with indicated antibodies. GFP blots were cropped to save space but were from the same lane and exposed film. Non-specific cross-reacting bands are indicated by asterisk (see text). Partial proteolysis was often seen with the GFP-tagged UL21a protein, particularly in cell lysate samples. (C) Native pUL21a interacts with the APC in HCMV infection. MRC-5 cells were infected with ADgfp or ADsubUL21a (as described in Materials and Methods) in the presence (+) or absence (2) of 10 mM MG132. Cell lysates were prepared at 24 hpi and immunoprecipitated with APC3 antibody. Cell lysates and eluted proteins were analyzed by immunoblotting. (D) Interaction of pUL21a with the APC does not require other viral proteins. 293T cells were transfected with plasmids expressing gfpUL21a wt or gfpUL21a stop . Cells were collected 72 hours later and cell lysates were immunoprecipitated as in panel B. Cell lysates and eluted proteins were analyzed by immunoblotting. PCNA and pUL44 were used as cellular and viral negative controls, respectively. doi:10.1371/journal.ppat.1002789.g001 The Carboxyl-Terminus of pUL21a Contains the APC Binding Site To begin understanding the nature of this interaction, we identified the APC-binding domain of pUL21a. Sequence alignment of pUL21a with its homologues in chimpanzee CMV (CCMV) and Rhesus CMV (RhCMV) revealed a highly conserved N-terminus (residues 1-47), divergent middle region (residues 48-83), and C-terminus that contained several conserved residues (residues 84-123), including a proline-arginine (PR) pair at residues 109-110 ( Figure 2A ). We created a series of truncation mutations targeting each region in the GFP-tagged pUL21a, and tested the ability of mutant UL21a proteins to interact with the APC in 293T cells ( Figure 2B ). All mutants were expressed at similar levels and were efficiently immunoprecipitated by the GFP antibody ( Figure 2C , and data not shown). As expected, full-length gfpUL21a wt co-immunoprecipitated both APC3 and APC8 while the gfpUL21a stop mutant did not. Importantly, while the carboxylterminal fragment of pUL21a consistently co-immunoprecipitated APC3 and APC8, the amino-terminal and middle fragments were unable to do so. Thus the carboxyl-terminus of pUL21a contains the APC binding domain. To define the precise sequence of the APC binding site, we made gfpUL21a mutants in which each of five conserved residue clusters within its carboxyl terminus were individually substituted with alanine residues (Figure 2A ). As a control, we also made alanine substitutions for the non-conserved proline-histidine pair at residues 111-112 (PH111-112AA) ( Figure 2A ). All mutants were stable and were efficiently pulled down by the GFP antibody ( Figure 2D , and data not shown). Among them, only the PR109-110AA mutant lost the ability to bind to the APC. Substitutions of the adjoining non-conserved residues (PH111-112AA) had no effect on APC binding. To validate the result in the context of infection, we constructed recombinant HCMV viruses expressing GFP-tagged or native forms of PR109-110AA or PH111-112AA pUL21a variants (ADgfpUL21a PR-AA , ADgfpUL21a PH-AA , AD-pmUL21a PR-AA , and ADpmUL21a PH-AA ). During infection, a reciprocal interaction between gfpUL21a PH-AA and APC3 could be detected while gfpUL21a PR-AA and APC3 did not interact ( Figure 2E ). Furthermore, untagged pUL21a PH-AA , but not pUL21a PR-AA , was co-immunoprecipitated with APC3 when stabilized by MG132 ( Figure S2 ). Together, these results indicate that the carboxyl terminus of pUL21a contains the APC binding domain and the residues PR 109-110 are critical for this binding. It has recently been reported that the APC bridge subunits APC4 and APC5 are degraded during HCMV infection and the complex dissociates [24] . To test if pUL21a was required for these events, we first examined APC subunit accumulation during infection with or without pUL21a. Levels of APC4 and APC5 proteins were markedly reduced during wild-type infection relative to mock-infected cells at 24 hpi ( Figure 3A ). However, no reduction was observed in APC4 and APC5 levels during infection with the UL21a-deletion virus. The pUL21a-deficient virus fails to express late viral genes due to a defect in viral DNA synthesis [30] . To rule out any role of late genes in APC4 and APC5 degradation, we treated infected cells with phosphonoacetic acid (PAA) to block viral DNA synthesis and late gene expression. APC4 and APC5 levels were reduced during infection with wild-type virus but remained elevated during infection with the UL21a-deletion virus, even following PAA treatment. Furthermore, there was no appreciable difference in APC4 and APC5 transcript levels between wild-type and deletion virus infections ( Figure 3B ). These data suggest that the changes in APC4 and APC5 protein levels occur at the level of protein stability. Consistent with this hypothesis, MG132 enhanced APC4 and APC5 protein levels during infection with wild-type but not deletion virus ( Figure 3C ). Thus, pUL21a-mediated loss of APC4 and APC5 was due to proteasomal degradation. Moreover, the APC binding mutant virus ADpmUL21a PR-AA was unable to degrade APC4 and APC5 while the ADpmUL21a PH-AA virus was as efficient as the wild-type control virus. These data support the conclusion that pUL21a binding to the APC promotes proteasomal degradation of APC4 and APC5. We next tested if the APC binding ability of pUL21a was also required for APC dissociation during infection. In this experiment, we used APC3 and APC10 as the marker for the specificity arm and cullin-ring ligase subcomplex of the APC, respectively. These two subcomplexes sit on opposite sides of the APC. APC10 has been proposed to bind APC substrates along with coactivator proteins, including Cdh1 [32] . APC10 associates with APC2 and APC11 of the ligase subcomplex, but its location in the inner cavity of the APC allows for contact with APC3 and APC6 of the specificity arm. In cells infected with ADpmUL21a PR-AA , total levels of APC3 and APC10 were similar to those in cells infected with ADpmUL21a PH-AA , allowing for a direct analysis of the efficiency of their association with the complex ( Figure 3D ). APC3 could not co-immunoprecipitate APC10 in ADpmUL21a PH-AAinfected cells, consistent with dissociation of the complex in the presence of functional pUL21a. In cells infected with ADpmU-L21a PR-AA , APC3 was able to pull down APC10 efficiently, indicating that the two subcomplexes remained associated. Finally, the integrity of the APC during ADpmUL21a PH-AA infection was largely restored upon addition of MG132, even though total protein levels were reduced likely due to MG132-induced cell death ( Figure 3D , and data not shown). These data were recapitulated during infection of wild-type and UL21a deletion viruses ( Figure S3 ). Our data provides strong evidence supporting the model that binding of pUL21a to the APC induces degradation of the APC bridge arm resulting in complex dissociation. As APC8 was co-immunoprecipitated with pUL21a in our original screen, it raised the possibility that pUL21a might require APC8 to target APC4 and APC5. For instance, pUL21a might bind to APC8 to disrupt the structure of the APC leading to APC4 and APC5 degradation, or it might use APC8 as a docking site for recruiting protein degradation enzymes to target APC4 and APC5. To test this, we depleted APC8 in these cells by shRNA knockdown ( Figure S4 ). Following shRNA depletion of APC8, the APC4 and APC5 levels remained reduced in cells infected with wild-type virus compared to those with UL21a-deletion virus, even though APC knockdown did seem to affect the overall stability of APC4 and APC5 in pUL21a-independent manner ( Figure S4 ). This suggests that pUL21a-mediated degradation of APC4 and APC5 is independent of APC8. To determine the functional consequence of pUL21a-dependent APC dissociation, we first analyzed the accumulation of APC substrates during wild-type or UL21a-deletion virus infection. The protein levels of APC substrates Cdh1 (that is also an APC coactivator) and geminin were markedly increased in wild-type virus infection as previously reported [27, 33] ( Figure 4A ). However, their levels were reduced during infection with the UL21a-deletion virus, suggesting increased APC activity. The geminin transcript accumulated to wild-type levels even without pUL21a, providing evidence that the difference in protein accumulation was not due to transcriptional regulation ( Figure 4B ). PAA treatment had no effect on substrate accumulation, ruling out pUL21a-mediated late gene expression as the source of the observed phenotype ( Figure S5A ). MG132 largely restored substrate levels during UL21a deletion viral infection, indicating that the difference is likely due to increased proteasome degradation ( Figure 4C ). These results were also recapitulated during infection of APC binding mutant virus ADpmUL21a PR-AA and its control virus ADpmUL21a PH-AA ( Figure 4C ). To confirm that decreased APC substrate accumulation during mutant virus infection was due to APC activity, we used shRNAs to knock down APC8 or the coactivator Cdh1 to deplete APC activity. Both APC8 and Cdh1 shRNAs efficiently reduced expression of their respective targets ( Figures 4D and S5B) . Importantly, APC8 knockdown restored geminin and Cdh1 levels in cells infected with ADpmUL21a PR-AA or ADsubUL21a virus to those with ADpmUL21a PH-AA or ADgfp virus. Likewise, Cdh1 knockdown restored geminin levels in cells infected with the pUL21a-deficient viruses. Thus, our results indicate that pUL21a association with the APC allows it to target APC4 and APC5 subunits for degradation to alter APC activity during infection. It is noteworthy that not all APC substrates were subjected to pUL21a-mediated regulation. We did not observe significant difference in Cdc6 or a drastic reduction in thymidine kinase protein levels in the UL21a mutant relative to wild-type viral infection (data not shown). It is possible that these APC substrates are regulated by multiple mechanisms, including APC-independent viral regulation, pUL21a-mediated alteration in APC substrate specificity, and pUL97-mediated phosphorylation of the APC coactivator Cdh1. In fact, Cdh1 from both wild type and UL21a mutant virus infected cells migrated slower in an SDS-PAGE gel compared to that from mock cells, which was previously shown to be due to phosphorylation ( Figure 4E ) [28] . Therefore, virus-induced, Cdh1 phosphorylation-mediated APC regulation appears intact even without pUL21a during HCMV infection. As the APC prevents the premature entry of the cell cycle into S phase, we predicted that increased APC activity in the absence of pUL21a would not compromise the ability of HCMV to arrest infected cells at G1/S phase boundary. Consistent with this hypothesis, cells infected with wild type, ADpmUL21a PH-AA , or ADpmUL21a PR-AA virus showed indistinguishable cell cycle profiles throughout infection, with the majority of cells phenotypically arrested in G1 phase ( Figure S6 ). To test if pUL21a was sufficient to alter APC activity, we first analyzed 293T cells that over-expressed pUL21a by transient transfection. Expression of pUL21a alone was sufficient to markedly reduce the levels of APC4 and APC5 ( Figure S7A) , and as expected, geminin and Cdh1 levels were elevated in these cells. These pUL21a-expressing cells were largely arrested in G2/ M phase ( Figure S7B ), failed to multiply, and ultimately died ( Figure S7C ). The biological characteristics of pUL21a-expressing cells are therefore consistent with reduced APC activity. To more precisely test if pUL21a was able to regulate the APC in the absence of other HCMV proteins, we developed an inducible pUL21a expression system. We constructed a HeLa cell line stably expressing a GFP-tagged TetR (tetracycline repressor) gene. We then transduced this cell line with lentiviruses expressing pUL21a stop , pUL21a PH-AA , or pUL21a PR-AA under a CMV-TetO (tetracycline operator) promoter. pUL21a protein accumulation was only detected in the presence of tetracycline, suggesting tight regulation of pUL21a expression ( Figure 5A ), although its levels were significantly lower than those expressed in transiently transfected cells ( Figure S7A) . Importantly, the addition of tetracycline significantly reduced APC4 and APC5 protein levels in cells expressing pUL21a PH-AA , but not pUL21a stop or pUL21a PR-AA . To assess the consequence of pUL21a on APC activity, we synchronized cells expressing pUL21a PH-AA (i.e. wild-type pUL21a) in mitosis with nocodazole and then assayed their ability to progress out of mitosis after release from nocodazole treatment. In the absence of tetracycline and pUL21a, cells readily progressed through the mitotic phase following release. In the experiment shown in Figure 5B , 26% and 48% of cells entered the next G1 phase by 2 and 4 hours, respectively. In the presence of tetracycline where pUL21a was expressed, progression through the mitotic phase was clearly delayed. As the result, only 5% and 24% of cells reached G1 by 2 and 4 hours, even though by 8 hours most of pUL21a-expressing cells were able to enter G1, likely due to low expression of pUL21a in these cells as compared to those in transiently transfected cells. Additionally, following nocodazole withdrawal, APC substrates geminin and cyclin B1 remained elevated in the presence of tetracycline while their levels were reduced in its absence ( Figure 5C ). Our results provide strong evidence that pUL21a expression alone is sufficient to regulate APC activity. In the final experiments, we tested the consequence of pUL21amediated APC regulation on HCMV replication in fibroblasts. We first tested if the ability of pUL21a to regulate the APC would be responsible for its previously reported role in promoting viral DNA replication [30] . We compared the growth of ADpmUL21a PR-AA mutant virus (i.e. pUL21a APC-binding deficient) to that of wildtype, ADpmUL21a PH-AA (i.e. pUL21a APC-binding competent), or UL21a deletion viruses in multi-step growth curve analysis. We found that ADpmUL21a PR-AA grew indistinguishably from wildtype and ADpmUL21a PH-AA viruses in both cycling and G0synchronized fibroblasts, whereas the UL21a deletion virus had a 100-fold defect ( Figure 6A ) [29] . Furthermore, knockdown of Cdh1 and APC8 was unable to enhance UL21a-deletion virus replication (data not shown). This suggests that pUL21a has at least two independent activities. One is to facilitate viral DNA replication via an unknown mechanism and is responsible for the growth defect of UL21a deletion virus. The second activity is to regulate the APC, whose impact on virus replication is not apparent under the aforementioned experimental conditions. As two HCMV proteins, pUL97 and pUL21a, are capable of regulating the APC, we hypothesized that one of these two proteins acted to compensate for the loss of the other during infection. Consistent with this hypothesis, HCMV appeared to retain the ability, at least to some extent, to regulate the APC even when pUL21a or pUL97 is absent ( Figure 4E , and data not shown) [24] . To test this hypothesis more directly, we created recombinant HCMV viruses ADpmUL21a PH-AA /subUL97 and ADpmU-L21a PR-AA /subUL97. The two viruses were derived from AD-pmUL21a PH-AA and ADpmUL21a PR-AA , respectively, and both contained an additional deletion in UL97. Both recombinant viruses grew slower than wild-type virus due to lack of the multifunctional pUL97 protein ( Figure 6C ). However, reconstitution of ADpmUL21a PR-AA /subUL97 that lacked pUL21a APCbinding activity following BAC transfection was markedly slower than that of ADpmUL21a PH-AA /subUL97 ( Figure 6B ). At day 25 post transfection, while cells transfected with the BAC clone of ADpmUL21a PH-AA /subUL97 showed nearly 100% of CPE indicated by virus-driven GFP expression, GFP-positive foci in cells transfected with the BAC clone of ADpmUL21a PR-AA /subUL97 were distinctly smaller. Furthermore, multi-step growth curve analysis showed that titers of ADpmUL21a PR-AA /subUL97 were 13-and 14-fold lower than that of ADpmUL21a PH-AA /subUL97 at 14 and 21 days post infection (dpi), respectively ( Figure 6C ). As a control to show that this phenotype was not due to general viral attenuation resulting from the UL97 deletion, we also constructed double mutant viruses ADpmUL21a PH-AA /subUL117 and AD-pmUL21a PR-AA /subUL117. These two viruses were derived similarly from ADpmUL21a PH-AA and ADpmUL21a PR-AA , but also contained a deletion in viral gene UL117. We chose UL117 as the control because its mutation attenuated virus growth but not viral early or early-late gene expression so UL97 expression was unlikely affected [34] . BAC transfection reconstituted both mutant viruses at similar efficiency and produced viruses with similar titers (data not shown). Multi-step growth analysis demonstrated that ADpmUL21a PH-AA /subUL117 and ADpmUL21a PR-AA /subUL117 Figure 6 . Abrogation of both pUL21a APC regulatory activity and pUL97 results in a more severe attenuation in HCMV growth than pUL97 deletion alone. (A) Abrogation of pUL21a-APC binding alone is not sufficient to alter HCMV replication. MRC-5 cells in serum-containing (cycling condition) or serum-free (G0 condition) media were infected with ADgfp, ADsubUL21a, ADpmUL21a PH-AA , or ADpmUL21a PR-AA at an MOI of 0.01. Production of cell-free virus at indicated times was determined by plaque assay. (B) Abrogation of both UL97 and the pUL21a-APC binding site markedly reduced the efficiency of HCMV reconstitution as compared to abrogation of UL97 alone. To reconstitute ADpmUL21a PR-AA /subUL97 and pADpmUL21a PH-AA /subUL97 viruses, MRC-5 fibroblasts were transfected with their corresponding BAC clones. For each recombinant virus, three independent clones were tested. Shown are representative images of virus spread indicated by virus-driven GFP expression at indicated days post transfection of two of the three clones. Images were taken under a Leica fluorescent microscope. (C) Abrogation of both UL97 and the pUL21a-APC binding site markedly reduced HCMV replication as compared to abrogation of UL97 alone. MRC-5 cells were infected with indicated recombinant viruses at an input genome number equivalent to that of 0.03 infectious units of wild type virus/cell. Production of cell-free virion DNA at indicated times was determined by qPCR analysis and normalized to input levels of ADpmUL21a PH-AA , which was set to 1. (D) Multi-step growth analysis of double mutant viruses that carried the UL117 deletion and point mutation in the UL21a-APC binding site. Cells were infected with indicated recombinant viruses and analyzed as described in panel C. The input value of ADgfp was set to 1. doi:10.1371/journal.ppat.1002789.g006 replicated at similar kinetics ( Figure 6D ). At 14 dpi, the titer of ADpmUL21a PR-AA /subUL117 was slightly lower (e.g. 1.5-fold) than that of ADpmUL21a PH-AA /subUL117. However, growth of mutant virus carrying only the UL117 deletion tracked with ADpmUL21a PR-AA /subUL117, suggesting that the difference between the PH and PR mutants at 14 dpi, if any, is minimal. Together, our data provide evidence that disruptions of both pUL97 and the APC regulatory activity of pUL21a are synthetically lethal to HCMV replication. This is consistent with a working model that these two functions enable HCMV to cope with APC activity to promote virus replication (Figure 7) . In sum, we have shown that the HCMV protein pUL21a antagonizes the APC by promoting proteasome-mediated disruption of this prominent cellular E3 ubiquitin ligase. HCMV has been shown to have two different means to regulate the anaphase-promoting complex (APC) during infection [24, 27, 28] . It can induce phosphorylation of APC co-activator Cdh1, and it induces dissociation of the complex by promoting proteasomal degradation of two components of the bridge subcomplex, APC4 and APC5. The viral protein pUL97 appears to be responsible for Cdh1 phosphorylation [24] . However, pUL97 is an HCMV-encoded kinase that has many reported roles [26, 35] . How this particular pUL97 activity impacts HCMV infection remains elusive. Importantly, the viral factor or precise molecular mechanism mediating APC4 and APC5 degradation has not been identified, and how APC disruption contributes to HCMV replication is not known. Here, we have identified the HCMV protein pUL21a as the viral factor that mediates APC disruption. It does so by interacting with the APC and inducing proteasome-dependent degradation of APC4 and APC5, which results in complex dissociation. This is the first identified viral protein that modulates the APC in this manner. We also show, for the first time, the impact of viral modulation of the APC, particularly by pUL21a, on HCMV replication. Loss of pUL21amediated APC regulation has minimal impact on virus replication but the combined loss of both pUL97-and pUL21a-mediated regulation markedly attenuates growth of the virus relative to single loss of pUL21a-or pUL97-mediated regulation. Our studies support a working model in which HCMV uses pUL97mediated Cdh1 phosphorylation and pUL21a-mediated complex disruption to control APC activity for efficient virus infection (Figure 7 ). Why has HCMV developed these two distinct mechanisms that seemingly lead to a similar biological consequence? It is possible that these two mechanisms have differential roles in HCMV infection under different conditions or in particular cell types, even though either one seems sufficient and can compensate for loss of the other in fibroblasts. Alternatively, it is possible that these two mechanisms serve as the fallback for one another or act synergistically to maximize the ability of the virus to acquire a complete control of the APC during infection. In any event, the fact that HCMV uses multiple means to subvert the APC underlines its critical role in HCMV infection. This is particularly true for large DNA viruses such as HCMV, which often encode multiple viral factors to regulate the same or related cellular targets central to their infection [36] . However, it is often challenging to dissect these intertwined viral mechanisms during infection because of the presence of other factors targeting the same process. The regulation of the APC represents one such critical but complex viral regulatory strategy, and our studies shed light into its role and mechanism during HCMV infection. Several viral factors from different viral families have been reported to use diverse mechanisms to regulate the APC. For instance, the human papillomavirus E2 protein binds to and inhibits the Cdh1 activator protein [20] , while the parapoxvirus virus protein PACR (poxviral APC regulator) functions as an enzymatically inactive APC11 mimic [23, 37] . The chicken anemia virus (CAV) protein apoptin can bind to the APC at the bridge and cause its dissociation using an unknown mechanism [19] . The fact that proteins from both HCMV and CAV target the APC bridge subcomplex suggests that viruses have evolved regulatory strategies converging on this sub-complex as an efficient means to disable APC activity. It is intriguing to speculate that modulating the APC complex by dissolving the bridge may allow viruses to alter substrate specificity of the APC instead of completely abolishing its activity, as the enzymatic portion of APC is known to have activity in vitro [23, 38] . HCMV does not appear to directly destroy the enzymatic subcomplex of APC, so it is of interest to determine if the APC retains some activity or is directed to target different substrates during virus infection. Several viral proteins have now been reported to regulate the APC in overexpression, and evidence correlating the role of these factors and viral replication is emerging. Deletion of the parapoxvirus PACR or CAV protein apoptin markedly attenuated virus growth in tissue culture even though their ability and role in inhibiting the APC during infection has not been clarified [23, 39] . Recently, the UL97 kinase of HCMV has been shown to phosphorylate Cdh1 and partially inhibit the APC during infection but with unknown consequences for viral replication [24] . Our study elucidates the mechanism by which pUL21a regulates APC in the context of virus infection and indicates a role of this pUL21a activity in viral replication. Mutation abolishing the APC binding activity of pUL21a had no impact on viral growth in tissue culture, but the loss of both pUL21a-APC binding and pUL97 markedly attenuated viral replication relative to the loss of pUL97 alone. Our data suggest that HCMV has evolved a sophisticated strategy by encoding both pUL97 and pUL21a to overcome APC activity. However, further experiments are needed to unequivocally demonstrate the vital role of APC regulation in HCMV replication and provide mechanistic insight into how this regulation impacts its biology. How does pUL21a target APC4 and APC5 for proteasome degradation? pUL21a does not contain a sequence domain that would suggest it as an E3 ligase, thus likely ruling out this possibility. Currently, we also do not know which subunit of the APC complex that pUL21a directly binds to so the precise mechanism that it uses to degrade APC4 and APC5 remains elusive. It is certainly possible that pUL21a may bind to a subunit neighboring to APC4 and APC5 so it can disrupt the APC structure leading to APC4 and APC5 degradation, or recruit a protein degradation enzyme (e.g. E3 ubiquitin ligase) to destabilize the subunits. However, knockdown of APC8 does not abrogate the ability of pUL21a to degrade APC4 and APC5, suggesting that APC8 is not involved and the presence of the entire complex is not required. Intriguingly, pUL21a itself is a highly unstable protein and likely degraded in a ubiquitin-independent manner [29, 40] . It is tempting to speculate that pUL21a may directly bind APC4 and APC5 and target them for degradation in a ubiquitin-independent manner. One focus of future work is to identify the APC component that pUL21a directly binds to and elucidate the mechanism of how pUL21a targets APC4 and APC5 to the proteasome. What would be the benefit for the virus to alter APC activity? The APC may restrict HCMV replication via several mechanisms. The APC not only promotes cell cycle progression through M phase, it also prevents cells from prematurely entering S phase. Thus virus-mediated APC regulation may help HCMV maintain an S phase-like cellular environment for viral replication. The APC targets more than 40 proteins for degradation, so it may deplete host factors critical to viral replication. Consequently, viruses may need to alter the substrate specificity of the APC or allow accumulation of APC substrates critical for viral replication. Interestingly, the only viruses within the poxvirus and herpesvirus families that are known to modulate the APC (e.g. parapoxviruses and HCMV) are those that do not encode viral thymidine kinase (TK) and ribonucleotide reductase subunit M2 (RRM2). Both enzymes are APC substrates and critical for the production of deoxyribonucleotides. It is tempting to speculate that this viral regulation of the APC may provide viruses a means to produce sufficient nucleotides to replicate their genome [23, 27] . Nonetheless, the APC also targets proteins involved in cellular DNA synthesis, glycolysis and glutaminolysis, and cell cycle progression, all of which could impact viral replication [41] . Moreover, the APC may also promote ubiquitination and degradation of viral proteins to restrict infection [42] . Several HCMV proteins contain a putative destruction Box (D-box) motif, an APC recognition signal commonly found in its substrates [24] . Future work is needed to differentiate these possibilities and unravel the APC substrates that may be critical for viral replication. Insight into the mechanism of pUL21a-mediated APC regulation may also have broad impact on cancer and neuronal disease. Due to its essential role in cell cycle progression, the APC is a promising target for novel anti-cancer therapeutics [16, 43] . In fact, we found in this study that overexpression of pUL21a essentially prevented the proliferation of a transformed cell line ( Figure S7 ), suggesting that pUL21a regulation of the APC could inhibit cancer cell growth. Furthermore, several recent studies have also highlighted a vital role of the APC in neuronal development (for a review, see [44] ). HCMV infects neuronal cells and congenital HCMV infection leads to neuronal disease and severe complications such as blindness, hearing loss, and mental retardation. It is reasonable to speculate that inhibition of the APC by pUL21a may play a role in promoting neuronal disease in congenitally infected infants. Therefore, an understanding of pUL21a-APC interaction may reveal novel mechanisms of APC assembly and regulation, give further impetus to target the APC for anti-cancer therapies, and uncover new insights into the molecular basis of HCMV pathogenesis. Primary embryonic lung fibroblasts (MRC-5), human newborn foreskin fibroblasts (HFFs), 293T, and Hela cells were propagated in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum, non-essential amino acids, and penicillin-streptomycin. Transient transfection of expression constructs were carried out using lipofectamine according to the manufacturers' instructions. pYD-C235 is a pLPCX-derived retroviral vector (Clontech) that expresses a DsRed gene driven by an internal ribosome entry site 2 (IRES2) [45] . pYD-C474 was created by PCR amplifying the coding sequence of the pGFP-UL21a fusion protein from pADgfpUL21a (see below) and ligating it into the multiple cloning site of pYD-C235. pYD-C580 was created by replacing the coding sequence of wild-type UL21a in pYD-C474 with that of mutant UL21a carrying two stop-codon mutations at the N-terminus (i.e. UL21a stop ) [30] . Vectors expressing pGFP-UL21a truncation mutants were derived from pYD-C235 while vectors expressing point mutants were derived from pYD-C474. Truncation mutants were made by PCR amplifying the targeted UL21a coding sequences and point mutants were created using a QuickChange XL kit (Stratagene). Primers used to create these mutants are listed in Table S2 . pYD-C160, pYD-C175, and pYD-C682 are pRetro-EBNA derived retroviral expression vectors that expressed GFP, UL21a, and UL21a stop , respectively. pYD-C648 and pYD-C649 are pLKO-based lentiviral vectors expressing GFP-TetR and carrying the CMV-TetO 2 promoter, respectively (generous gifts from Roger Everett, University of Glasgow Centre for Viral Research) [46] . YD-C665, YD-C667, and YD-C669 are lentiviral expression vectors created by cloning the UL21a stop , UL21a PH-AA , and UL21a PR-AA sequences into the multiple cloning site of YD-C649. To produce pLKO-based lentiviruses, 293T cells were transfected with corresponding pLKO vectors along with packaging plasmids. Lentivirus was collected at 48 and 72 hours and used to transduce MRC-5 cells. To create GFP-TetR expressing stable cells, Hela cells were transduced with pYD-C648 derived lentivirus and sorted for GFP expression 48 hours later. GFP-positive cells were collected, grown in the presence of G418 (500 mg/ml), and frozen as cells stably expressing GFP-TetR. These stable cells were then transduced with lentivirus derived from YD-C665, YD-C667, and YD-C669, selected with puromycin (2 mg/ml), and tested for tetracycline (1 mg/ml)-regulated expression of targeted genes. For shRNA knockdown, MRC-5 cells were transduced with pLKO-based lentivirus expressing shRNA against the targeted gene for 48 hours. The shRNA sequence for Cdh1 knockdown was 59CCAGTCAGAACCGGAAAGCCA39 and the shRNA sequence for APC8 knockdown was 59GCAGGAGGTAA-TATGCTATAA39. All pLKO-based shRNA lentiviral vectors were purchased from the Washington University Children's Discovery Institute/Genome Center. The primary antibodies used in this study included anti-b actin (AC-15, Abcam); anti-HA (HA.11, Covance); anti-GFP (3E6 and A6455, Invitrogen); anti-APC3 (AF3.1, Santa Cruz and 610454, BD); anti-APC8 (6114, Biolegend); anti-APC4 (A301-176A, Bethyl laboratories); anti-APC5 (A301-026A, Bethyl laboratories); antigeminin (sc-13015, Santa Cruz); anti-Cdh1 (DH01, Calbiochem); anti-cyclin B1 (ms868 P1, Thermo-Scientific); anti-UL21a [29] ; anti-IE2 (mAB8140, Chemicon); and anti-IE1 and anti-pp28 (generous gifts from Thomas Shenk, Princeton University) [45] . Phosphonoacetic acid (PAA), MG132, tetracycline, gancyclovir (GCV), and propidium iodide (PI) were purchased from Sigma-Aldrich. Lipofectamine 2000 and Protein A-conjugated Dynabeads were purchased from Invitrogen. Recombinant HCMV AD169 viruses were reconstituted from transfection of corresponding BAC-HCMV clones as previously described [34] . Viral stocks were prepared by ultra-centrifugation of infected culture supernatant through 20% D-sorbitol cushion and re-suspending pelleted virus in serum-free medium. The following BAC-HCMV clones were used in the present study, and were constructed using PCR-based linear recombination as previously reported [29] , unless indicated otherwise. pAD-GFP, which carried the GFP-tagged genome of the HCMV AD169 strain, was used to produce wild-type virus ADgfp [45] . pADgfpUL21a, which carried an N-terminally GFP-tagged version of pUL21a, was used to produce ADgfpUL21a virus [29] . pADsubUL21a, which carried a GalK/kanamycin dual mutagenic cassette in place of the UL21a coding sequence, was used to produce UL21a-deletion virus ADsubUL21a [29] . pADgfpU-L21a PR-AA , pADgfpUL21a PH-AA , pADpmUL21a PR-AA , or pADp-mUL21a PH-AA carried point mutation PR109-110AA or PH111-112AA in the GFP tagged or native UL21a gene, respectively. These recombinant BAC clones were used to produce corresponding point mutant viruses. pADpmUL21a PH-AA /subUL97 and pADpmUL21a PR-AA /subUL97 carried the GalK/kanamycin mutagenic cassette in place of UL97 on the background of pADpmUL21a PR-AA and pADpmUL21a PH-AA BAC clones. Similarly, pADpmUL21a PH-AA /subUL117, pADpmUL21a PR-AA / subUL117, and pADsubUL117 carried the GalK/kanamycin mutagenic cassette in place of UL117 on the background of pADpmUL21a PR-AA , pADpmUL21a PH-AA , and pAD-GFP BAC clones, respectively. All BACs were confirmed by restriction digestion, PCR, and sequencing. HCMV virus titers were determined in duplicate in HFFs by tissue culture infectious dose 50 (TCID 50 ) assay or plaque assay. Relative viral genome numbers were determined by real-time quantitative PCR (qPCR) as described previously [29] . For most infections, subconfluent MRC-5 cells in serumcontaining medium were inoculated with recombinant HCMV virus at an input genome number equivalent to that of 3-5 infectious units of wild type virus/cell for 1 hour, unless otherwise indicated. Inoculum was removed and fresh medium was replenished. For infection of G0-synchronized cells, MRC-5 cells were incubated in serum-free medium for 72 hours, infected as described above, and maintained in serum-free media throughout the infection. For shRNA knockdown experiments, subconfluent MRC-5 cells were transduced with lentivirus for 24 hours, incubated in fresh medium for additional 48 hours, and infected as described above. When necessary, PAA (100 mg/ml) was added immediately following infection, and MG132 (10 mM) was added 12-14 hours prior to harvest. For viral growth analysis, virus production in the media of infected cultures was determined by TCID 50 , plaque assay, or qPCR. For qPCR analysis, virion DNA was prepared as previously described [29] . Briefly, cell-free supernatants were treated with DNase I to remove contaminating DNA, and virions were lysed with proteinase K and SDS. DNA was extracted with phenol/chloroform/isoamyl alcohol and precipitated with ethanol. The DNA was subjected to qPCR using primers and a taqman probe specific for UL54. For immunoprecipitation, frozen cell pellets were lysed in lysis buffer (0.5% NP-40, 50 mM Tris-Cl pH 8.0, 125 mM NaCl, supplemented with protease and phosphatase inhibitors) using an end-over-end rotator at 4uC for 30 minutes. Cell extracts were cleared by centrifugation at 16,0006 g for 15 minutes. Supernatants were incubated with protein A-coated Dynabeads that were coupled to 1 mg anti-HA (HA.11, Covance), 1 mg anti-GFP (3E6, Invitrogen) or 2 mg anti-APC3 (AF3.1, Santa Cruz) mouse monoclonal antibodies at 4uC for 1-2 hours. Beads were washed with PBS and immunoprecipitated protein complexes were eluted by boiling beads in reducing sample buffer for 5 minutes. Cell extracts (pre-IP) were also collected and boiled in reducing sample buffer. For mass spectrometry analysis, protein complexes were resolved by SDS-polyacrylamide gel electrophoresis (Invitrogen) followed by staining with a silver stain kit (Sigma-Aldrich). Protein bands specific to immunoprecipitated pUL21a complex were excised for identification by MS/MS mass spectrometry [47] . For immunoblotting, total cell or pre-IP extracts were lysed in sample buffer containing SDS and protease and phosphatase inhibitors. Proteins were resolved on a SDS polyacrylamide gel, transferred to a polyvinylidene difluoride (PVDF) membrane, hybridized with a primary antibody, reacted with the horseradish peroxidase-conjugated secondary antibody, and visualized using chemiluminescent substrate (Thermo Scientific). Total RNA was extracted with TRIzol (Invitrogen) and treated with Turbo DNA-free reagent (Ambion) to remove genomic DNA contaminants. cDNA was reverse transcribed from total RNA with random hexamer primers using the High Capacity cDNA reverse transcription kit (Applied Biosystems). cDNA was quantified using SYBR Advantage qPCR Premix (Clontech) and primers for the cellular genes geminin, APC4, APC5, and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) as an internal control (see below). cDNA from infected cells was used to generate a standard curve for each gene examined. The standard curve was then used to calculate the relative amount of specific RNA present in a sample. Primers used for RT-qPCR are as follows: geminin, forward 59GCCTTCTGCATCTGGATCTCTT39 and reverse 59CGAT GTTTCCTTTTGGACAAGC39 [24] ; APC4, forward 59ATT CTCGTCCTTGGAGGAAGCTCT39 and reverse 59TTCTG GCCATCCGAGTTACTTCAG39 [24] ; APC5, forward 59GTG CCATGTTCTTAGTGGCCAAGT39 and reverse 59GATGCG CTCTTTGCAGTCAACCTT-39 [24] ; GAPDH, forward 59CT GTTGCTGTAGCCAAATTCGT39 and reverse 59ACCCACT CCTCCACCTTTGAC39 [30] . To determine cellular DNA content, cells were trypsinized, collected by low-speed centrifugation, fixed, and permeabilized in ice-cold 70% ethanol overnight. Cells were stained with propidium iodide only, or double-stained with propidium iodide and anti-pUL44 antibody to identify HCMV-infected cells. Total or pUL44positive cells were determined for their DNA content by cell-cycle analysis with flow-cytometry. Percentages of cells in each cell cycle compartment were calculated using CellQuest or FlowJo software. Figure S1 APC8 knockdown by shRNA. MRC-5 cells were transduced with lentivirus expressing shRNA targeting either Luc (negative control) or APC8. Forty-eight hours post transduction, cells were infected with mock, wild-type (ADgfp), or UL21adeletion virus (ADsubUL21a). Cell lysates were collected at 72 hpi and analyzed by immunoblotting. Note that the asterisk-marked bottom band that reacted with the APC8 antibody was nonspecific as it was not affected by the APC8-targting shRNA. (TIF) Figure S2 Amino acid residues PR 109-110 of pUL21a are required for its APC binding during HCMV infection. Cells were infected with indicated virus and MG132 was added to the final concentration of 10 mM at 6 hpi. Cells were collected at 20 hpi and lysates were immunoprecipitated with APC3 antibody. Cell lysates and eluted proteins were analyzed by immunoblotting. (TIF) Figure S3 pUL21a dissociates the APC by promoting degradation of the bridge subcomplex. MRC-5 cells were infected with ADgfp or ADsubUL21a, and MG132 was added to the final concentration of 10 mM at 6 hpi. Cells were collected at 20 hpi and lysates were immunoprecipitated with APC3 antibody. Both cell lysates and eluted proteins were analyzed by immunoblotting. (TIF) Figure S4 APC8 is not required for pUL21a-mediated degradation of APC4 and APC5. Knockdown and subsequent immunoblot were performed as described in the legend to Figure S1 .

Deciphering Novel Host–Herpesvirus Interactions by Virion Proteomics

Over the years, a vast array of information concerning the interactions of viruses with their hosts has been collected. However, recent advances in proteomics and other system biology techniques suggest these interactions are far more complex than anticipated. One particularly interesting and novel aspect is the analysis of cellular proteins incorporated into mature virions. Though sometimes considered purification contaminants in the past, their repeated detection by different laboratories suggests that a number of these proteins are bona fide viral components, some of which likely contribute to the viral life cycles. The present mini review focuses on cellular proteins detected in herpesviruses. It highlights the common cellular functions of these proteins, their potential implications for host–pathogen interactions, discusses technical limitations, the need for complementing methods and probes potential future research avenues.

Over the last decades, many host-pathogen interactions have been characterized using genetics, biochemical, and microscopy approaches. These discoveries relied on mutants, pharmacological reagents, immunoprecipitations, immunofluorescence, electron microscopy, cell fractionation, and Western blotting to name a few of the methods employed. These approaches provided much precious information but, given the typical focus of these approaches on individual molecules, likely only revealed a small portion of the proteins involved. Other methods such as high throughput two-hybrid and genetic screens, nucleic acid arrays, RNA interference, and proteomics are now proving essential tools to tackle the complexity of these interactions. The main advantages of mass spectrometry, for instance, are that it is a fast, sensitive and potentially a quantitative approach to identify putative novel players, particularly when coupled to efficient purification schemes. Already, proteomics revealed how viruses modulate the expression of host proteins (Rassmann et al., 2006; Sun et al., 2008; Tong et al., 2008; Antrobus et al., 2009; Pastorino et al., 2009; Thanthrige-Don et al., 2009; Zandi et al., 2009; Zhang et al., 2009 Zhang et al., , 2010 Coombs et al., 2010; Emmott et al., 2010; Lu et al., 2010 Lu et al., , 2012 Munday et al., 2010; Bartel et al., 2011; Lietzen et al., 2011; Ramirez-Boo et al., 2011; Chou et al., 2012) . A relatively new and interesting field is the characterization of host-pathogen interactions within mature purified virions. As reviewed on several occasions, several studies reported the presence of individual cellular proteins in viral particles (Bernhard et al., 2005; Maxwell and Frappier, 2007; Viswanathan and Fruh, 2007; Friedel and Haas, 2011; Zheng et al., 2011) . This includes vaccinia virus (Krauss et al., 2002) , influenza virus (Shaw et al., 2008) , HIV (Gurer et al., 2002; Cantin et al., 2005; Ott, 2008) , vesicular stomatitis virus (Moerdyk-Schauwecker et al., 2009) , and several herpesviruses (see below). Though these cellular components have often been considered purification contaminants, the presence of similar proteins in both related and unrelated viruses suggests that some of them may be biologically relevant. The identification of virion-associated host proteins could thus lead to the discovery of novel therapeutic tools against viruses. The present review focuses on their identification and putative roles with respect to the proteomics of herpesviruses. Thus far, the protein composition of eight different herpesvirions has been studied by mass spectrometry. These studies include the alphaherpesvirinae herpes simplex virus type 1 (HSV-1) and pseudorabies virus (PRV; Loret et al., 2008; Kramer et al., 2011) , the betaherpesvirinae human and murine cytomegaloviruses (HCMV and MCMV, respectively; Kattenhorn et al., 2004; Varnum et al., 2004) and the gammaherpesvirinae Kaposi sarcoma herpesvirus (KSHV), gamma herpesvirus 68 (γHV68), Epstein-Barr virus (EBV), and Alcelaphine (Bortz et al., 2003; Johannsen et al., 2004; Bechtel et al., 2005; Zhu et al., 2005; Dry et al., 2008) . Interestingly, host proteins were detected in all herpesvirions analyzed so far, as summarized in Table 1 . For instance, our laboratory previously reported the protein composition of mature extracellular HSV-1 viral particles and identified as many as 49 cellular proteins (Loret et al., 2008) . Similarly, studies focusing on PRV and EBV reported up to 48 and 43 cellular proteins, respectively (Johannsen et al., 2004; Kramer et al., 2011) . Meanwhile, Varnum et al. (2004) found as many as 70 different host proteins in extracellular HCMV virions. While fewer cellular proteins were reported for other viral particles, it is clear that herpesviruses can potentially incorporate many proteins from its host. Moreover, of the 173 different proteins detected in herpesvirions, nine protein groups are present in at least four distinct herpesvirions. This includes 14-3-3, actin, annexins, cofilin, translation factors, GAPDH, heat shock proteins, pyruvate kinase M2, and various Rab GTPases. These results indicate that, first of all, it is common for herpesviruses to incorporate cellular proteins into their viral particles and, secondly, that www.frontiersin.org ? different viruses share similar host proteins. Most excitingly, it also suggests that these host proteins may play common roles throughout the herpesviral family. This defines an interesting and novel set of host-pathogen interactions taking place within the virus itself, rather than the cell. It is tempting to speculate that some viruses might have a higher capacity to steal cellular proteins because of their size and symmetry. Herpesviruses are indeed large viruses containing a layer called the tegument between their capsids and envelopes that could accommodate non-viral proteins. Though some host proteins may randomly be incorporated into virions, others may rather be selected to insure the optimal replication of the viruses that carry them. Bioinformatics databases such as the KEGG, Gene Ontology, or DAVID are useful tools to get an overview of the functional interplay of proteins (Ashburner et al., 2000; Huang Da et al., 2009; Kanehisa et al., 2010) . As pointed out by Friedel and Haas (2011) , complex statistical tools are available to quantitatively evaluate the implication of proteins in various processes but these are beyond the scope of the present review. Here an analysis of the proteins identified in herpesvirions was instead performed with the Ingenuity Pathways Analysis database (Ingenuity ® Systems), which contains all the known physical and functional links among cellular proteins and defines their most significant functions. That analysis indicates that many of the cellular proteins found in herpesvirions normally modulate trafficking, cell proliferation, cell death, cell migration, cell metabolism, or the cytoskeleton (Figure 1 , upper pie chart). Though subtle differences between family members are noticeable when looking at individual viruses, similar functions are found (Figure 1, other charts) . Immune-related molecules are also important constituents for several viruses, including HSV-1, KSHV, γHV68, Alcelaphine, and MCMV. Altogether, this provides an overall picture whereby herpesviruses, not surprisingly, modulate all of the important aspects of the cell but where each virus might deploy its energies slightly differently. The main surprise is that so many cellular proteins are detected within assembled viral particles, which raises an important question as to their biological significance and mode of action. The overall picture that several important cellular functions might be modulated by the host proteins incorporated into viral particles is intriguing. This clever strategy is consistent with the parasitic nature of all viruses, including herpesviruses, which would presumably gain some replication advantage from stealing cellular modulators rather than coding for them in their own genomes. The most critical question is the benefit for the viruses to incorporate these cellular proteins in their assembled particles, particularly since these proteins also exist in the cells. While this is open to discussion, one possibility is that some of the incorporated cellular proteins may be remnants of the final capsid envelopment process. Alternatively, this may allow the prompt action of some of these proteins immediately upon viral entry. This could jumpstart the expression and/or duplication of the viral genome, as it is the case for the herpesviral VHS, VP16, ICP0, and ICP4 proteins that are present in virions (Lam et al., 1996; Everett, 2000; Halford and Schaffer, 2001; Ellison et al., 2005; Hancock et al., 2006; Loret et al., 2008; Sarma et al., 2008; Loret and Lippe, 2012) . Other early potential sites of action are the process of viral entry itself, intracellular capsid transport, import of the viral genome through the nuclear pore or immune modulation, all common steps among herpesviruses. Whatever the case might be, the question remains as to why the cellular pool of these proteins would not suffice. Several options may be considered. First, it may be that the virions incorporate specific isoforms, splice variants or post-translationally modified proteins that could have properties or functions distinct than their cellular counterparts. Second, the incorporation of a host protein from one cell type might permit the infection of a different cell type that does not express such protein. For example, alpha herpesviruses initially infect mucosal cells and could acquire host proteins that are beneficial to infect dormant neuronal cells. Finally, the host proteins might be in complex with viral proteins and it is those complexes that are active to promote the infection. These possibilities are of course speculative at this point and need to be explored. One aspect where the incorporation of host proteins in mature virions might be beneficial is molecules involved in intracellular trafficking. Work by numerous laboratories demonstrated that the transport machinery used to move cellular proteins is also employed by viruses (Simons and Warren, 1984; Lodish et al., 2000; Sollner, 2004; Greber and Way, 2006; Mercer et al., 2010) . This is essential for their proteins and particles to reach their final destination, for example, the site of viral replication, assembly, and/or envelopment. Along with SNARES proteins, Rab and Arf GTPases are master regulators of molecular trafficking throughout the cell (Sollner and Rothman, 1996; Zerial and McBride, 2001; Mizuno-Yamasaki et al., 2012) . So far,VAMP3, a SNARE, was identified in PRV virions (Kramer et al., 2011) but it may only be a matter of time until other SNARES are discovered in other members of the herpes family. This is relevant as another SNARE was reported to facilitate the envelopment of MCMV capsids (Cepeda and Fraile-Ramos, 2011) . In contrast, a great number of Rab proteins have been identified in herpesvirions, particularly HSV-1 and PRV (Table 1) . One stimulating option is that these proteins regulate the displacement of viral capsids in the cell, which could justify their incorporation in the viral particles. As Rab and Arf proteins collectively modulate several intracellular transport steps within the cell, it is anticipated they may be involved in various stages of the infection. For instance, Rab1, which is present in HSV-1 extracellular virions (Loret et al., 2008) , and Rab43 were recently demonstrated to modulate the final envelopment of the virus (Zenner et al., 2011) . Similarly, Rab6, found in HSV-1 and PRV (Loret et al., 2008; Kramer et al., 2011) , is also necessary for the efficient assembly of the related HCMV (Indran and Britt, 2011) . It will now be of interest to determine if the virion-associated pool of these GTPases actively participates in the viral life cycle. Interestingly, several Rab proteins have been implicated in autophagosome formation and maturation (Chua et al., 2011) . While it is difficult to consider how virion-incorporated Rab proteins play a role at that stage, they might rather be incorporated into the virions as MCMV FIGURE 1 | The proteins from Table 1 were analyzed with the Ingenuity database to define their putative functions in the context of an infection. To this end, the protein accession numbers (or GI numbers) were queried from the Ingenuity database. For the purpose of this figure, all known functions associated with these proteins were exported to Microsoft Excel and regrouped. In the top pie chart, the cellular proteins found in all the herpesvirions were analyzed collectively, while the other pie charts depict the host proteins incorporated into each virus. Since each protein can be associated with multiples functions in the database, the results of those analyses are expressed as relative values instead of raw numbers, which consequently exceeds the original number of proteins analyzed. The percentages therefore represent the number of proteins falling into a given category with the total of each pie chart being 100%. A graphical legend of the categories is provided at the bottom right corner of the figure. a consequence of their involvement in autophagosome formation and concomitant viral envelopment. Given the vast impact of Rab proteins on the cell, it will be a major challenge to decipher all their roles in the life cycle of herpesviruses, particularly for the pool present in mature virions. Molecular trafficking is not only dependent on SNARES, Rab, and Arf proteins, it is also intimately linked to the cytoskeleton. It is thus not surprising that herpesviruses devote some of their resources toward regulating this central cellular machinery. For instance, herpesviruses significantly reorganize both cellular and nuclear actin as well as microtubules (Norrild et al., 1986; Avitabile et al., 1995; Sharma-Walia et al., 2004; Simpson-Holley et al., 2005; De Regge et al., 2006; Saksena et al., 2006) . They also travel along microtubules during both entry and egress and interact with several cellular molecular motors (Sodeik et al., 1997; Smith et al., 2001; Dohner et al., 2002; Marozin et al., 2004; Lee et al., 2006; Wolfstein et al., 2006; Radtke et al., 2010) as well as cortical and nuclear actin filaments (Forest et al., 2005; Feierbach et al., 2006; Roberts and Baines, 2011) . Furthermore, some members incorporate in their viral particles tubulin or actin-related components ( Table 1 ; Wong and Chen, 1998; Grunewald et al., 2003) . Actin has been reported to compensate the loss of various viral tegument proteins in PRV (del Rio et al., 2005; Michael et al., 2006) and may thus act as an abundant filling agent, so its significance in herpesviral particles remains enigmatic. Similarly, the relevance of intermediate filament components vimentin and keratins in some herpes virions ( Table 1) is difficult to assess given these filaments are not as well characterized as other cytoskeletal elements. It may nevertheless be important for herpesviruses, particularly since they are not all associated with the common skin or hair contaminants often detected in mass spectrometry (Hertel, 2011) . Viruses tend to monopolize for their own purpose their host expression apparatus, including protein translation (Bushell and Sarnow, 2002) . For example, the prototypic HSV-1 ICP27 viral protein regulates all aspect of mRNAs including transcription, splicing, nuclear export, and translation for the benefit of the virus (Rice and Knipe, 1988; Sekulovich et al., 1988; Sandri-Goldin and Mendoza, 1992; Smith et al., 1992; Hardwicke and Sandri-Goldin, 1994; Hardy and Sandri-Goldin, 1994; Brown et al., 1995; Soliman et al., 1997; Chen et al., 2002; Lindberg and Kreivi, 2002; Ellison et al., 2005; Larralde et al., 2006; Fontaine-Rodriguez and Knipe, 2008) . As these cellular functions are highly regulated, the inclusion of DDX3X, a multifunctional RNA helicase that also regulates transcription, nuclear export, and translation that is used by several viruses (Schroder, 2010 (Schroder, , 2011 ) may be relevant. Its incorporation into mature virions could thus accelerate viral gene expression in the early stages of the infection. Similarly, the presence of translation initiation or elongation factors in virions (Table 1 ) may also jumpstart gene expression in favor of the viruses. Interestingly, HSV-1 does not require cells to be in the S-phase and even arrests the cell cycle at the G1/S transition step (Shadan et al., 1994; Song et al., 2000) , which partly explains why it can grow in non-dividing neurons. While the precise mechanism of this arrest is unclear, it is known that the viral ICP0 protein and the VP16 cellular partner HCF modulate the cell cycle (Hobbs and DeLuca, 1999; Lomonte and Everett, 1999; Piluso et al., 2002) . Moreover, ICP0 interacts with the host cyclin D3 (Kawaguchi et al., 1997) . However, it was recently reported that stress, rather than the cell cycle per se, may be a critical feature (Bringhurst and Schaffer, 2006) . Clearly, the interaction of herpesviruses with the cell proliferation apparatus is complex and likely involves several host and viral proteins. Identifying novel players that might be incorporated into mature virions may thus be very useful to clarify this process. An interesting scenario is the possible regulation of apoptosis by host proteins loaded onto viral particles. Apoptosis is regulated both negatively and positively by several viruses (Teodoro and Branton, 1997; Goodkin et al., 2004) , presumably to insure their survival at the early stages of the infection but their efficient release later on. Conceptually, the presence of anti-apoptotic proteins in herpes particles might thus provide a mean to quickly evade death upon entry, while the presence of pro-apoptotic proteins on newly assembled/enveloped viral particles may trigger or stimulate their extracellular release. Only further work will resolve this open question. Several factors generally contribute to variation among proteomic studies. Hence, the preparation of the samples (e.g., in-gel trypsin digestion versus liquid digestion and chromatography) may lead to the detection of different populations of tryptic peptides. Moreover, the sensitivity of the mass spectrometers and the abundance of the proteins in the samples also impact peptide detection. The relative abundance of a peptide is itself influenced by the complexity of the samples, where some proteins may evade identification. Finally, each protein differs in its properties (ionization, resolution in SDS-PAGE gels), which will be reflected in their detection. This includes SNARES, which are transmembrane proteins resistant to SDS extraction (Yang et al., 1999; Kubista et al., 2004) . It is thus likely that some of the proteins in Table 1 are present in more viral particles than reported and that additional proteins are indeed incorporated in herpesvirions. More specific aspects regarding herpesviruses includes the purification schemes employed to enrich the viral particles, which will directly influence the purity of the samples and hence the potential detection of contaminants. One important caveat is that some host proteins may simply stick to the large viral particles. Another one is common contaminants such as some hair/skin associated keratins or as mentioned above actin, which may simply fill the virions. However, even potential contaminants cannot simply be discarded since actin and even some keratins may indeed participate in viral life cycles. Moreover, the relative abundance of all the cellular proteins within the cell is unknown, so it is not possible to rule out potential contaminants on the sole basis of abundance. It is thus critical to orthogonally validate all proteomics hits. Various tools are available to define the biological relevance of host proteins identified in viral particles, including Western blotting, immuno-electron microscopy or functional screens. One powerful method is RNA interference. However, given the dual presence of the host proteins within the viral particles and the cell itself, this becomes a challenging task. RNA interference also has its own caveats (false positives and negatives). Another common step is the expression of dominant positive or negative mutants. In all cases, one major difficulty is that the host proteins may be essential for the cells and their depletion may lead to cytotoxicity, thus proper controls are needed. In addition, the host proteins might be essential for the virus within the cells but only accessory within the virions. Consequently, depletion of a protein may have limited impact on the virus since complemented by the other pool of that protein in the virus or the cell. Small reduction or stimulation in viral yields may thus result. It such cases, it may be necessary to produce the virus on cells that lack these proteins to see if this makes a difference. One should also consider animal models since tissue culture based screens may miss important players, for instance modulators of the immune system or virulence factors. Clearly, multiple experimental strategies are needed to ultimately insure the biological significance of the host proteins found in viral particles. The identification and functions of host proteins in viral particles is an important step toward the elucidation of novel host-pathogen interactions. In the case of herpesvirions, this is well under way with eight different family members analyzed so far. One main aspect is to sort biologically relevant cellular proteins from sticky contaminants. The orthogonal validation of the host proteins found in herpesvirions using biologically relevant assays is thus critical. As pointed out above, it will necessary to analyze all these proteins in the background of two pools, one cellular and one virion-associated, which are likely to complement one another. An interesting possibility is that some isoforms or specific posttranslationally modified host proteins may be loaded into the capsids. Thus a detailed analysis of the host proteins present in viral particles will be important and a potential way to distinguish them from their cell-associated counterparts. Another issue is the expected variation among cell types. In that respect, it would be most interesting to examine the cellular protein content of HSV-1 produced in neurons in opposition to the virions produced on other cell types. Finally, the mechanisms by which all these host proteins are recruited to the viral particles will also need to be explored. Thus the proteomics of viral particles is only the beginning of the adventure, which should prove most exciting yet challenging. I am indebted to the Canadian Institutes of Health Research (grant # MOP82921) for funding our proteomics research. I also wish to thank Kerstin Radtke for excellent suggestions and Daniel Henaff for critical reading of the manuscript.

Naturally-Occurring Genetic Variants in Human DC-SIGN Increase HIV-1 Capture, Cell-Transfer and Risk of Mother-To-Child Transmission

BACKGROUND: Mother-to-child transmission (MTCT) is the main cause of HIV-1 infection in children worldwide. Dendritic cell–specific ICAM-3 grabbing-nonintegrin (DC-SIGN, also known as CD209) is an HIV-1 receptor that enhances its transmission to T cells and is expressed on placental macrophages. METHODS AND FINDINGS: We have investigated the association between DC-SIGN genetic variants and risk of MTCT of HIV-1 among Zimbabwean infants and characterized the impact of the associated mutations on DC-SIGN expression and interaction with HIV-1. DC-SIGN promoter (p-336C and p-201A) and exon 4 (198Q and 242V) variants were all significantly associated with increased risk of intrauterine (IU) HIV-1 infection. Promoter variants decreased DC-SIGN expression both in vitro and in placental CD163(+) macrophages (Hofbauer cells) of HIV-1 unexposed infants but not of HIV-1 exposed infants. The exon 4 protein-modifying mutations increased HIV-1 capture and transmission to T cells in vitro. CONCLUSION: This study provides compelling evidence to support an important role of DC-SIGN in IU HIV-1 infection.

In 2010, UNAIDS estimates that 390,000 children acquired HIV-1-infection worldwide mostly through mother-to-child transmission (MTCT) [1] . Overall transmission rates in the absence of any intervention vary from 12 to 42%. Although antiretroviral therapy (ART) can reduce MTCT to as low as 2% [2] , limited access to timely diagnostics and drugs in resource-poor settings blunts the potential impact of this strategy. A better understanding of the mechanisms acting in MTCT of HIV-1 is crucial for the design of interventions other than ART for transmission prevention. MTCT of HIV-1 can occur during pregnancy (in utero, IU), at delivery (intrapartum, IP) and via breastfeeding (postpartum, PP). HIV-1 can cross the placental barrier in utero either by microtransfusion or by transcytosis across the trophoblast cell layer [2] . IP transmission may occur through direct contact between infant mucosa and HIV-1 infected maternal blood and/ or cervico-vaginal secretions [2] . Finally, HIV-1 in breast milk may result in PP infection of the newborn through mucosal exposure [2] . High maternal viral loads in serum and breast milk and low CD4 cell count as well as obstetric factors such as preterm delivery, vaginal delivery, and prolonged membrane rupture have been correlated with increased risk of MTCT of HIV-1 [2, 3] . Genetic variations in HIV-1 co-receptors and determinants of immunity have been shown to influence the outcome of MTCT of HIV-1 [2, 4] . Variants that result in either increased CCR5 expression or a non-functional receptor (32 base-pair deletion variant) influenced risk of vertical transmission [5, 6] . The CCR5 32 base-pair deletion is absent in African populations [7] . Genetic polymorphism of innate immunity determinants such as toll-like receptor 9 and mannose-binding protein also increased the risk of MTCT [8] [9] [10] . Discordance at the human leucocyte antigen (HLA) class I loci between mother and child or specific HLA alleles also protect against MTCT [11, 12] . Dendritic cell-specific ICAM-3 grabbing-nonintegrin (DC-SIGN, encoded by CD209) is a C-type lectin that binds to many pathogens including HIV-1 [13] . This interaction with HIV-1 leads to viral capture and subsequent transmission to adjacent T cells [14, 15] . DC-SIGN is expressed on the cell surface of myeloid dendritic cells and some macrophage subsets including Hofbauer cells present in the placenta [13, 16] . In the context of HIV-1, DC-SIGN may not only promote trans-infection of T cells but signalling initiated by HIV-1 binding may also influence immune responses and enhance productive infection of the dendritic cells themselves [17] [18] [19] . Given the presence of DC-SIGN in the placenta and its known interaction with HIV-1, we hypothesized that polymorphism affecting its expression or function might influence the risk of MTCT of HIV-1. Here, we report significant associations between DC-SIGN genetic variants that modulate DC-SIGN expression in placental macrophages, promote HIV-1 capture and transmission to T cells and increase risk of MTCT among Zimbabwean infants. We studied a subgroup of 197 infants born to ART-naive HIV-1-infected mothers recruited in the ZVITAMBO study, which enrolled 14,000 mother-baby pairs between November 1997 and January 2000 in Harare, Zimbabwe [20] . ART prophylaxis for HIV-1-positive antenatal women was not available in the Harare public-sector during ZVITAMBO patient recruitment. The samples were consecutively drawn from two groups: 97 HIV-1positive mother/HIV-1-positive child pairs and 100 HIV-1positive mother/HIV-1-negative child pairs. Modes of infant HIV-1 transmission were determined using definitions adapted from Bryson and colleagues [21] and were described elsewhere [22] . Full methods for recruitment, baseline characteristics collection, laboratory procedures have been described elsewhere [20] . MTCT of HIV in the whole ZVITAMBO cohort occurred during the IU (22,9%), IP (48%) and PP periods (29.1%) [20] . Haplotype reconstruction was performed as previously described [23] . Haplotype-tagged single nucleotide polymorphisms (htSNPs) were determined using the HaploBlockFinder software with a minor allele frequency over 5% [24] and numbers were redefined compared to our previous publication [23] for their frequency in the present study population. Ten htSNPs were selected corresponding to the 10 major haplotypes from the 20 SNPs (rs number in Table S1 ) found in the Zimbabwean population as we previously described [23] . These htSNPs along with the 3 others exon 4 mutations were genotyped in the 197 infants by direct PCR sequencing analysis as previously described [23] . Putative transcription factors binding sites in promoter region were analysed with TESS interface (http//:www.cbil. upenn.edu/tess) using the TRANSFAC database. Genomic DNA from homozygous patients with or without mutation was amplified in the promoter region from nucleotide 2507 to 21 and cloned between the Bgl II and Hind III multiple cloning sites in the pGL2-Basic vectors (Invitrogen, Canada inc, Burlington, Canada). All recombinants clones were verified by DNA sequencing. Luciferase assay was performed as previously described [22, 25] . Firefly luciferase reporter vector was cotransfected with constitutive expressor of Renilla luciferase, phRL-CMV (Promega, Madison, WI, USA). Firefly luciferase activity was normalized to Renilla luciferase activity. Site-directed mutagenesis was carried out using pcDNA3-DC-SIGN vectors obtained from Drs. S. Pöhlmann, F. Baribaud, F. Kirchhoff and R.W. Doms [26] through the AIDS Research and Reference Reagent Program, NIAID, NIH: pcDNA3-DC-SIGN exon 4 was amplified from genomic DNA of variants carriers and replaced between PspI and EspI (Fermantas, Burlington, Canada) restriction sites. All recombinants clones were verified by DNA sequencing for the presence of mutations of interest and conservation of the coding frame. Stably transfected cell lines were generated from Raji cells (ATCC, Manassas, USA) by nucleofection (Cell line Nucleofector Kit V, Amaxa, Walkersville, USA) and maintained in RPMI 1640 10% FBS containing 1 mg/ ml of G418 (Invitrogen). DC-SIGN-expressing cells were sorted (sorter BD ARIA, BD Biosciences, Mississauga, Canada) and limiting dilutions were performed. Cell lines were grown from a single clone. 3610 5 cells/well of each Raji transfectants were plated in 96-well plates and pre-incubated with mannan (200 mg/ ml, Sigma-Aldrich, St-Louis, USA), DC-SIGN blocking antibody (clone AZN-D1, R&D systems, Minneapolis, USA) (20 mg/ml), matching isotype control or medium for 30 minutes at 4uC before adding 50 ng of p24 equivalent of virus (HIV-1 HxBru-ADA or HIV-1 JRCSF ). Cells were incubated 2 h at 37uC and washed with PBS (Invitrogen). Cells were lysed and assayed for p24 Ag using ELISA. In co-culture experiment, Raji transfectants were pulsed as above and 3610 5 phytohemagglutinin-L activated human primary CD4 + T-lymphocytes (ratio 1:1) isolated from peripheral blood mononuclear cells were added. Cells were cultivated in RPMI 10% FBS 100 U/ml rIL-2 for 5 days. Supernatants were collected and p24 was measured using ELISA. CD4 + T-lymphocytes were isolated from healthy donors and activated as previously described [27] . HIV-1 stocks were generated by transient transfection of HEK293T cells with HxBRU ADA encoding the R5-tropic HIV-1ADA Env [28] or JRCSF proviral construct using the standard calcium-phosphate method. Viral stock was titrated using ELISA p24 Ag (BioChain, Hayward, USA). DC-SIGN expression was monitored by flow cytometry analysis using FITC-labelled anti-DC-SIGN antibodies clones DCN46 (BD Biosciences, Missisaugua, Canada) and 5D7 (Santa Cruz Biotechnology, Santa Cruz, USA). The cells were also incubated with isotype-matched control antibodies. Flow cytometry was performed using a BD FACS-Scan (BD Biosciences). Full-term placentas were obtained following non-complicated pregnancies and deliveries at Hôpital St-Luc and Hôpital Bethesda in Cotonou, Benin. All infants were delivered vaginally except for two who were delivered by caesarean section and none of the mothers presented with signs of sexually transmitted infections or placental malaria infection. Placentas with signs of inflammation (chorioamnionitis) were excluded. These HIV-1-infected mothers received a combination of three antiviral drugs (full regimen) during pregnancy and delivery. Infants received a full regimen or a single dose nevirapine at delivery and none of them were HIVinfected. A small piece of each placenta was collected and processed within 3 hours following the delivery and washed extensively with PBS to remove blood and maternal cells. Mononuclear cells were mechanically isolated from placental tissue using a Medimachine (BD Biosciences) and purified on Histopaque gradient (Sigma-Aldrich, Oakville, Canada). Placental mononuclear cells were cryopreserved until flow cytometry analysis. Cells from 4 wild-type (WT) p-336T/p-201C and 11 homozygote or heterozygote p-336C/p-201A infants for the promoter variants born to HIV-1-negative mothers were analysed. Placental cells from 3 WT and 6 homozygote or heterozygote infants born to HIV-1-positive mothers were also analysed. Placental cells were analysed by flow cytometry using CD3-, CD19-(eBioscineces, San Diego, USA), and CD56-PerCPCy5.5 (BD Biosciences), CD14-Alexa700 (BioLegend, San Diego, USA), CD163-APC (R&D System, Minneapolis, USA), DC-SIGN-FITC, CD68-PE and HLA-DR-PECy7 (BD Biosciences) antibodies and isotype-matched controls. Dead cells stained with Live/ Death Aqua (Invitrogen) and lineage cells stained with CD3-, CD19-and CD56-PerCPCy5.5 antibodies were excluded. Placental macrophages were initially gated for CD14 expression and high granularity. Geometric mean fluorescence intensity (MFI) was calculated in DC-SIGN + CD163 + and DC-SIGN + CD163 2 subsets to assess the level of DC-SIGN expression and their level of maturation. Change geometric MFI represents the difference between specific marker expression and its FMO (fluorescence minus one). Flow cytometry was performed using a BD LSR-Fortessa (BD Biosciences). Statistical analysis was performed using GraphPad PRISM 5.0 for Windows (GraphPad Software inc. San Diego, CA). In order to assess the association between each of the DC-SIGN hapotype ( Table 1) or htSNP (Table 2 ) alleles with MTCT of HIV-1, those subjects who were heterozygous and homozygous for the haplotype or htSNP alleles were compared separately with subjects who tested negatively for that allele (reference category). The association between each of the putative haplotype or htSNP alleles and risks of MTCT of HIV-1 was investigated using crude and adjusted multivariate logistic regression to derive odds ratio (OR) and 95% confidence interval (CI) as estimates of relative risks. Specifically, the models were adjusted for the maternal viral load. The analyses were restricted to those haplotypes found at a frequency above 5% in the study population (Table 1 ) or to those SNPs found only in H4 or H6 (Table 2) . Differences in frequencies of haplotypes and htSNPs were compared between groups using Fisher's exact test. All SNPs were in Hardy-Weinberg Equilibrium [23] . For luciferase DC-SIGN/HLA-DR/CD68 expression, capture and transmission assays comparisons between WT and variants were assessed with the unpaired two-tailed Student's t test. Written informed consent was obtained from all mothers who participated in the study. We carried out an association study of DC-SIGN polymorphism in 197 infants born to ART-naive HIV-1-infected mothers recruited in Harare, Zimbabwe [20] . Among them, 97 were HIV-1-infected and 100 were uninfected. Of the 97 HIV-1-infected infants, 57 were infected IU, 11 IP, and 17 PP. Timing of infection could not be determined for 12 HIV-1-infected infants as specimens were not available at some time points. Baseline characteristics of mothers and infants were reported previously [22] . Briefly, maternal age and CD4 + T cell count, child sex, mode of delivery, duration of membrane rupture and gestational age were similar among all groups. Maternal viral load .29 000 copies/ml was associated with increased risk of both IU and PP HIV-1 transmission, OR: 3.64, 95% CI: 1.82-7.31, P = 0.0002 and OR: 4.45, 95% CI: 1.50-13.2, P = 0.0045, respectively. Ten htSNPs from the 20 SNPs ( Figure 1A ) corresponding to the 10 major DC-SIGN haplotypes ( Figure 1B) previously described among Zimbabweans [23] , were genotyped in the study samples. Haplotypes with frequencies above 5% in the study population were analysed for their potential association with MTCT of HIV-1. Infants carrying H4 or H6 haplotypes had increased risk of IU HIV-1-infection, whereas H2 haplotype carriers were less likely to be infected during pregnancy compared to infant noncarriers ( Table 1) . None of the haplotypes were significantly associated with altered risks of IP or PP infections. The H4 and H6 hapotypes remained significantly associated with IU HIV-1 infection (OR: 4.98, 95% CI: 1.32-18.8, P = 0.0168 and OR: 2.93, 95% CI: 1.27-6.76, P = 0.0113, respectively) after adjustment was made for maternal viral load. H2 haplotype remained significantly associated with protection against IU infection after adjustment for maternal viral load (OR: 0.23, 95% CI: 0.10-0.51, P = 0.0003). To identify the causal SNPs associated with increased IU transmission of HIV-1, we determined the association between IU HIV-1 infection and each of H4 and H6 signature SNPs. Promoter p-201A (rs11465366) and exon 4 198Q (rs41374747) variants are found exclusively in H6 while exon 4 242V (rs11465380) variant tag H4 ( Figure 1B) . Both H4 and H6 haplotypes harbour promoter variant p-336C (rs4804803) that is known to influence DC-SIGN promoter activity [25] and increased risk of HIV-1 parenteral infection [29] . These variants were all associated with increased risk of IU HIV-1 infection after adjustment for maternal viral load (Table 2 ). In a step-wise logistic regression analysis including all DC-SIGN associated SNPs and maternal viral load, DC-SIGN 242V variant (OR: 4.87, 95% CI: 1.19-19.9, P = 0.0261) and maternal viral load (OR: 3.30, 95% CI: 1.48-7.37, P = 0.0033) remained independent predictors of HIV-1 IU acquisition. Maternal DC-SIGN haplotypes were not associated with MTCT of HIV-1 (Table S2) . We have previously investigated the association between DC-SIGN-related (DC-SIGNR, encoded by CD209L) genetic variants and MTCT of HIV-1 in the same subset of infants [22] . DC-SIGNR is a DC-SIGN homologue expressed at the cell-surface of endothelial cells of placental capillaries [30] . DC-SIGNR promoter p-198A and intron 2 180A variants were significantly associated with increased risk of MTCT. When adjustment was made for all the significant DC-SIGN and DC-SIGNR associations in logistic regression analysis, DC-SIGN exon 4 242V (OR: 5.03, 95% CI: 1.18-21.4, P = 0.0275) and DC-SIGNR intron 2 180A (OR: 6.93, 1.51-31.7, P = 0.0119) variants remained associated with increased risk of IU transmission, suggesting that DC-SIGN and DC-SIGNR are independent predictors of IU of HIV-1 among Zimbabweans. We next investigated the impact of the HIV-1 associated promoter variants on both DC-SIGN transcriptional activity in vitro and expression in fetal macrophages (Hofbauer cells). Variant p-336C decreased the transcriptional activity of Sp1 binding site [25, 31] . Transcription factor binding site analysis predicted that variant p-201A would create a c-myc binding site. To test the effect of these promoter variants on transcription, we transiently transfected HeLa cells with a luciferase reporter gene under the control of DC-SIGN promoter region -507 to -1 containing AP-1, Sp1, Ets-1 and NF-KB transcription factors that are essential for promoter activity [32] and harbouring promoter WT p-336T/p-201C or variant p-336C/p-201A sequences ( Figure 2A ). As previously reported [25, 31] , the luciferase activity of the p-336C/p-201C variant construct was lower than that of WT p-336T/p-201C ( Figure 2B ) but the decreased did not reach significance in our assay. The p-201A variant either alone or in combination with p-336C significantly reduced DC-SIGN transcriptional activity in vitro (Ratio p-336T/p-201C/p-336C/p-201A = 3.13, P = 0.0039). Uninfected infants harboured more frequently H1 and H2 haplotypes reaching significance for H2 (Table 1) . H2 carries two promoter variants, p-939T and p-139C, that differ from WT H1 haplotype ( Figure 1B) . The promoter variants p-939T and p-139C did not show any influence on DC-SIGN transcriptional activity in vitro ( Figure S1 ). However, HeLa cells derived from cervical carcinoma might not represent the best model to study the impact of promoter variants on DC-SIGN expression in macrophages. To address this issue, we further determine the net impact of susceptibilityassociated promoter mutations on DC-SIGN expression by measuring total DC-SIGN protein expression in Hofbauer cells. These cells are found within the chorionic villi beneath the syncytiotrophoblast layer at the maternal-fetal interface [33] . Term placentas contain a distinct population of Hofbauer cells that co-express DC-SIGN, CD163, CD14, CD68 and HLA-DR, a phenotype similar to alternatively activated macrophages (M2) known for their immunosuppressive properties [16, 33, 34] . Hofbauer cells were analysed by flow cytometry after isolation of mononuclear cells from term placentas of promoter WT p-336T/ p-201C and variants p-336C/p-201A carriers. CD14 + cells of high granularity that were negative for T, B and NK cell markers (CD3, CD19 and CD56) were identified as Hofbauer cells (CD14 + population) and two subsets of DC-SIGN + cells were observed (CD163 + and CD163 2 , Figure 2C ). CD163 + cells expressed significantly higher levels of DC-SIGN, HLA-DR and CD68 compared to CD163 2 cells ( Figure 2C ). We then compared levels of DC-SIGN expression in CD163 + and CD163 2 cells between infants carrying or not carrying promoter variants and born from HIV-1-negative or HIV-1-positive mothers ( Figure 2D ). In infants born to HIV-1-negative mothers, levels of DC-SIGN expression were reduced 1.9-fold (P = 0.0091) in CD163 + cells and 1.8-fold (P = 0.0305) in CD163 2 cells from infants carrying the promoter variants compared to infants harbouring the WT promoter sequence. Interestingly, DC-SIGN expression varied according to the mothers' HIV-1 status. In infants harbouring the WT sequence, levels of DC-SIGN expression were reduced 3.2-fold (P = 0.0402) by CD163 + cells and 2.2-fold (P = 0.0378) by CD163 2 cells in infants born from HIV-1-positive mothers compared to infants born from HIV-1-negative mothers. However, it remained unchanged in infants carrying the promoter variants. Hence, p-336C and p-201A altered DC-SIGN expression in placental Hofbauer cells and but their impact vary according maternal HIV-1 status. Protein-modifying Variants Increase Viral Capture and Transfer to T cells DC-SIGN molecules on the cell surface enhance HIV-1 infection by capturing virions and transmitting them to CD4 + T-lymphocytes [14, 15] . The neck region, encoded by exon 4, is important for efficient binding to HIV-1 [35] . We hypothesized that the exon 4 protein-modifying variants associated with IU HIV-1 infection could affect the interaction between DC-SIGN and HIV-1. To assess viral capture, exon 4 from the DC-SIGN expression vector was replaced by exon 4 from infants carrying WT, 242V (designated as L242V) or 198Q, 214D and 221Q (designated as R198Q) variants ( Figure 3A ). Raji cells do not express endogenous DC-SIGN and allowed us to investigate the net impact of exon 4 mutations on DC-SIGN HIV-1 affinity. Raji cells ( Figure 3B) were stably transfected and cell lines grown from a single clone. Since viral capture is influenced by cell-surface expression of DC-SIGN [35] , we selected cell lines with similar baseline DC-SIGN surface expression ( Figure 3B ). The stable Raji transfectants were pulsed with equal amount of R5 tropic HIV-1 HXBRU-ADA or HIV-1 JRCSF strains extensively washed to remove the unbound virus, and then lysed. The parental Raji cells were used as controls. The number of virions used was not saturating since capture increased in a dosedependent manner ( Figure S2A ). Interestingly, DC-SIGN L242V and R198Q variants were more efficient at capturing viral particles than WT ( Figure 3C ). HIV-1 capture on the Raji transfectants was stable over time ( Figure S2B ) and dependent on DC-SIGN interaction since the capture was reduced to background levels following incubation with DC-SIGN antibody (AZN-D1) or mannan ( Figure 3C ). Similar results were obtained when cells were pulsed with HIV-1 JRCSF strain ( Figure S2C ). To investigate whether DC-SIGN exon 4 mutations could also enhance cell transmission of HIV-1, we co-cultivated activated primary human CD4 + T lymphocytes with HIV-1 pulsed Raji transfectants. Transmission was quantified by measuring HIV-1 p24 in the supernatants after 5 days. The DC-SIGN variants significantly increased viral transmission to CD4 + T-lymphocytes ( Figure 3D ). Transmission was dependent on DC-SIGN expression since Raji cells or transfectants pre-incubated with DC-SIGN antibody failed to transmit HIV-1 to CD4 + T lymphocytes. Moreover, cell infection was not due to viral particles shed into the supernatant since virus was undetectable in the absence of CD4 + T lymphocytes ( Figure 3D) . Thus, the DC-SIGN neck region variants associated with IU HIV-1 infection enhanced both the capture of HIV-1 by DC-SIGN and its subsequent transmission to the CD4 + T lymphocytes. In vitro studies have shown that the interaction between DC-SIGN and HIV-1 can enhance short-term viral transfer to other susceptible cell types such as T lymphocytes [14, 15, 36] . Based on these findings, a Trojan horse model has been proposed whereby was compared in CD163+ and CD1632 subsets from infants bearing or not promoter variants and born from HIV-1-negative mothers (HIV-1 Unexposed; p-336T/p-201C group n = 4; p-336C or p-336C/p-201A group n = 11) or from HIV-1-positive mothers (HIV-1 Exposed; p-336T/p-201C group n = 3; p-336C or p-336C/p-201A group n = 6). Results in C and D are mean 6 S.E.M. values of MFI and difference between subsets or variants was calculated with Student's t test. doi:10.1371/journal.pone.0040706.g002 HIV-1 may subvert DC-SIGN-expressing submucosal dendritic cells to promote dissemination from the periphery to the lymphoid tissues [13] . To date, relatively few studies have assessed the potential impact of DC-SIGN polymorphism in adult HIV-1 infection and the findings have not been consistent. While some [29, 37, 38] have found a significant association, others have not [39] [40] [41] . Little is currently known about the mechanisms underlying HIV-1 passage across the placenta. We and others [16, 33, 34] have shown that DC-SIGN is expressed by placental Hofbauer cells. Moreover, the identification of natural and functional DC-SIGN genetic variants associated with an increased risk of IU HIV-1 infection further support the implication of DC-SIGN in HIV-1 dissemination across the placenta. DC-SIGN polymorphism was not associated with IP and PP HIV-1 infections. However, the relatively small number of subjects analysed in the IP (n = 11) and PP (n = 17) groups may have limited the power of the present study to detect any association and therefore we cannot rule out the possibility that DC-SIGN could also contribute to IP and PP transmission of HIV-1. DC-SIGN promoter p-336C and exon 4 242V variants are observed on the H4 haplotype while p-336C, p-201A and 198Q variants are found on the H6 haplotype. Thus exon 4 variants are always transmitted with one or both promoter variants. These variants were all associated with IU HIV-1 infection and yet the promoter variants reduced DC-SIGN expression in Hofbauer cells whereas the exon 4 mutations enhanced capture and transmission of HIV-1 to CD4 + T lymphocytes. Compensatory mutations frequently evolve to dampen the effect of other mutations. Since the DC-SIGN gene has been under strong evolutionary pressure to conserve its function [42] it is not surprising that mutations increasing the affinity of DC-SIGN for pathogens have appeared that can compensate for mutations that reduce its expression. Interestingly, HIV-1 itself can also affect DC-SIGN expression. Indeed, HIV-or antibody-stimulated DC-SIGN signalling in monocytes-derived dendritic cells (MDDC) reduced DC-SIGN expression and prevented cell maturation [19, 43] . In infants harbouring the WT sequence, levels of DC-SIGN expression were significantly lower in infants born from HIV-1-positive mothers than those born from HIV-1-negative mothers ( Figure 2D ). However, the impact of HIV-1 was negligible or not noticeable in infants carrying the promoter variants since baseline DC-SIGN expression levels were already low in these subjects. Although we cannot exclude that ART may also modulate DC-SIGN expression, it is reasonable to believe that HIV affects DC-SIGN expression in the tissue since in vitro experiments support this hypothesis [19, 43] . In the present study, Zimbabwean infants were born from ART-naive HIV-1-positive mothers. Given the fact that they were all exposed to HIV during their intra-uterine life, they may have harboured similar levels of DC-SIGN expression ( Figure 2D) . Hence, the positive association observed between IU HIV-1 infection and DC-SIGN H4 and H6 haplotypes may thus result from exon 4 protein-modifying mutations found within these haplotypes that enhanced capture of HIV-1 by Hofbauer cells within the chorionic villi in close proximity to maternal infected cells and facilitate short-term transfer of the virus to the infant's T lymphocytes [16] . On the other hand, we cannot rule out the possibility that DC-SIGN variants might promote HIV-1 infection of Hofbauer cells and subsequently IU transmission of HIV-1. Hofbauer cells express both the HIV-1 CD4 receptor and the CCR5 co-receptor [16, 44] and HIV-1 genomic materials have been detected in placental macrophages [45] . Kumar et al observed compartmentalized HIV-1 replication within the placenta during IU transmission [46] and proposed that viral selection during IU transmission could be the manifestation of HIV-1 placental adaptation to the unique repertoire of cellular targets and increased adherence to C-type lectins which further support the implication of DC-SIGN in IU transmission of HIV-1. However, the net impact of this phenomenon on MTCT of HIV-1 remains to be determined since it has also been shown that placental macrophages can restrict HIV-1 replication [47] . In addition to the enhancement of HIV-1 capture and transmission to target cells, DC-SIGN genetic variants may also contribute to a local immunological environment that promotes viral replication and dissemination of HIV-1 across the placenta [17, 19] . HIV-1 or HIV-1-derived products activated fetal macrophages and T lymphocytes and promoted the establishment of a productive infection within the placenta [46, [48] [49] [50] . In response to dengue infection, MDCCs from DC-SIGN p-336CT heterozygous individuals produced higher levels of pro-inflammatory factors such as TNF-alpha, IL-12 and IP-10 than those from WT p-336TT homozygotes [51] . Moreover, TNF-alpha enhanced HIV-1 replication and transcytosis within the placenta and TNFalpha level correlated with the amount of HIV-1 transcripts [52] [53] [54] . It is tempting to speculate that infants harbouring the p-336TT genotype, such as H1 and H2 carriers, may produce less TNF-alpha to reduce or thwart HIV-1 replication in the placenta. In this study, we demonstrate for the first time, the impact of DC-SIGN natural polymorphisms on its expression in placental cells and interaction with HIV-1 and provide compelling evidence to support an important role of DC-SIGN in IU HIV-1 infection. These findings raise the possibility that similar mechanisms may operate with other human pathogens known to interact with DC-SIGN and warrant further investigation. p24 Ag was measured by ELISA. Where indicated, cells were pre-incubated with anti-DC-SIGN (AZN-D1) or with mannan to inhibit DC-SIGN interaction with HIV-1. HIV-1 capture is shown relative to WT (WT = 100%). (D) HIV-1 transfer to T lymphocytes by DC-SIGN variants. Raji-transfectants were pulsed as in (C) and subsequently co-cultivated with activated human primary CD4 + T lymphocytes from two donors for 5 days. Virus release into the supernatant was measured by ELISA p24. Where indicated, cells were pre-incubated with AZN-D1. HIV-1 transmission is shown relative to WT (WT = 100%). Results are mean 6 SD of duplicates for each donor (D) or three independent experiments (C). Student's t test was used to calculate differences in % capture and transmission among Raji DC-SIGN transfectants L242V, R198Q and WT. doi:10.1371/journal.pone.0040706.g003 pGL2-Basic, the parental vector without a promoter. Expression of the DC-SIGN promoter constructs was calculated relative to the value of pGL2-Basic, which was arbitrarily set as 1. Data are mean 6 SD values of 3 independent experiments performed in triplicates and there were no significant differences in the relative expression between variants and wild-type (WT) as determined with Student's t test. (DOCX) Figure S2 HIV-1 capture by Raji transfectants (a) Dosedependent HIV-1 capture by Raji transfectants. Raji and Raji-H1 transfectants were incubated with 25, 50 and 100 ng of p24equivalent of HIV-1 HXBru-ADA for 2 h at 37uC, washed with cold PBS 1X and lysed in 0,5% Triton X-100. Cell-associated p24 contents were measured by ELISA. (b) Residual cell-associated HIV-1 over time. 3610 5 Raji-transfectants were exposed to 150 ng of p24-equivalent of HIV-1 HXBru-ADA for 2 h at 37uC, washed and incubated in fresh medium at 37uC for different time points (0, 1 h, 2 h, 4 h and 18 h). Cell-associated p24-contents were measured by ELISA after lysis in 0,5% Triton X-100. (c) Capture assay with HIV-1 JR-CSF . 3610 5 cells were incubated with 50 ng of p24-equivalent of HIV-1 JR-CSF for 2 h at 37uC, washed extensively with cold PBS 1X and lysed in 0,5% Triton X-100. Cell-associated p24 contents were measured by ELISA. Where indicated, cells were pre-incubated 30 min at 4uC with 20 mg/ml of anti-DC-SIGN (AZND1) or with mannan (200 mg/ml) to inhibit DC-SIGN interaction with HIV-1 before pulsing with HIV-1 JR-CSF . HIV-1 capture is shown relative to wild-type (WT = 100%). Data are mean 6 SD of 2 independent experiments performed in duplicates. Student's t test was used to calculate differences in % capture between the Raji DC-SIGN transfectants L242V, R198Q and WT. (DOCX)

A New Model for Hendra Virus Encephalitis in the Mouse

Hendra virus (HeV) infection in humans is characterized by an influenza like illness, which may progress to pneumonia or encephalitis and lead to death. The pathogenesis of HeV infection is poorly understood, and the lack of a mouse model has limited the opportunities for pathogenetic research. In this project we reassessed the role of mice as an animal model for HeV infection and found that mice are susceptible to HeV infection after intranasal exposure, with aged mice reliably developing encephalitic disease. We propose an anterograde route of neuroinvasion to the brain, possibly along olfactory nerves. This is supported by evidence for the development of encephalitis in the absence of viremia and the sequential distribution of viral antigen along pathways of olfaction in the brain of intranasally challenged animals. In our studies mice developed transient lower respiratory tract infection without progressing to viremia and systemic vasculitis that is common to other animal models. These studies report a new animal model of HeV encephalitis that will allow more detailed studies of the neuropathogenesis of HeV infection, particularly the mode of viral spread and possible sequestration within the central nervous system; investigation of mechanisms that moderate the development of viremia and systemic disease; and inform the development of improved treatment options for human patients.

Hendra virus (HeV) causes serious systemic infection with pneumonia and encephalitis in humans, horses and various laboratory animals [1, 2, 3, 4] . It is a single-stranded, negative-sense RNA virus belonging to the family Paramyxoviridae and is classified within the genus Henipavirus which it shares with one other virus, Nipah virus (NiV). HeV first emerged in the Brisbane suburb of Hendra in 1994, where it caused the deaths of one human and fourteen horses [5] . Since then a further thirty four HeV outbreaks have been identified along the mid to north-eastern coast of Australia with infection of five more humans (of whom three died) and numerous horses [6, 7, 8, 9] . Pteropid bats have been identified as the reservoir host [10] , however epidemiological evidence does not support direct bat to human transmission. Horses have been an intermediate host in the transmission of disease to humans in all cases. There are as yet no readily available effective therapies or prophylaxis for HeV infection, either for use in humans or other susceptible animals. Of necessity, HeV pathogenesis studies and evaluation of vaccine and therapeutic candidates must be carried out in animal infection models under Biosafety Level 4 (BSL4 conditions). Several species have been used for this purpose including: ferrets, hamsters, guinea pigs, pigs, cats, horses, and African green monkeys [3, 11, 12, 13, 14, 15, 16] . With bats, the list comprises species from six orders including; Rodentia, Primates, Chiroptera, Cetartiodactyla, Perrisodactyla and Carnivora. The broad species susceptibility is unusual for a member of the family Paramyxoviridae and is attributed largely to the highly conserved nature [17] of the host receptors for the virus, Ephrin B2 and B3 [18, 19] . Despite the possession of relevant receptors [19] , the laboratory mouse, a most useful host on account of their small size, ease of handling, and vast library of available reagents, is reported to be resistant to HeV infection and disease [20] . Westbury et al in 1995 reported resistance of mice to HeV infection in a study that was designed to identify a suitable laboratory animal model of HeV disease. Juvenile BALB/c mice were inoculated with 5000 median tissue culture infective doses (TCID 50 ) of virus by a parenteral route and observed for clinical signs of infection. Mice remained clinically well throughout the 21 day study period and, after euthanasia, there was no evidence of infection by gross or histological examination, virus isolation or serology. Similar results were reported by Wong et al in 2003, who investigated the susceptibility of mice to the closely related Nipah virus [21] by inoculating juvenile Swiss brown mice by either parenteral or intranasal routes. An understanding of the mechanisms of resistance of mice to HeV may provide novel targets for therapeutic and preventative intervention of human infections. Furthermore, circumvention of such mechanisms may induce a useful mouse model of HeV disease. Therefore, in view of the limited previous work, we decided to re-evaluate the apparent resistance of mice to HeV infection by investigating the outcome of HeV exposure by various routes to inbred mice of different ages and strains. Additionally, quantitative real-time polymerase chain reaction (qPCR), a technique not available at the time of the initial studies, would be used for detecting evidence of viral replication. We found that mice are susceptible to HeV infection when exposed via the intranasal route, but resist infection when challenged by a parenteral route. Infection manifested as acute, transient, and asymptomatic virus replication in the upper and lower respiratory tracts, together with clinically significant encephalitis that has a longer incubation period than is reported for other models of fulminating HeV disease. The pattern of central nervous system involvement (CNS) supports neuroinvasion by the anterograde route (spread from the neuron cell body toward the axon terminus) and, importantly, transneuronal spread within the CNS. Over all, the study demonstrated that mice are susceptible to HeV infection and has provided a new and important model for HeV induced encephalitis. Ten juvenile (8 weeks) and ten aged (12 months) C57BL/6 mice were each divided into two groups and challenged with 50,000 TCID 50 HeV using either an intranasal or subcutaneous route of exposure and monitored daily for 21 days post infection (DPI). Weight loss and temperature changes were not observed for any animals beyond expected minor daily fluctuations. Aged and juvenile mice exposed by the subcutaneous route remained clinically well during the period of observation and at euthanasia there was no evidence of HeV infection by histology, immunohistochemistry or qPCR (Table 1) . Specific binding antibody to the soluble form of the HeV G glycoprotein (HeV sG) was detected in sera of two juvenile animals only, using a Luminex microsphere assay [22] . Low levels of serum neutralizing antibody was also detected in one of these animals by serum neutralisation test (SNT) ( Table 2) . By contrast, aged animals exposed intranasally were either unexpectedly found dead (mice 14 and 15, 12 and 11 DPI) or developed peracute neurological disease during the study period (mice 11-13) necessitating euthanasia on 16, 20 and 21 DPI. Affected animals showed ataxia, muscle tremors and hypersensitivity. One juvenile mouse (mouse 5) with intranasal exposure also developed a neurological illness requiring euthanasia. Mouse 5 displayed a slower disease onset compared to the aged mice, with a three day waxing and waning depressive illness culminating in severe depression, lack of response to stimuli and hypothermia. In contrast to the systemic vasculitis and multi-organ involvement that are features of HeV infection in other animal species, inflammatory lesions and HeV antigen were only identified in the brain tissue of clinically affected mice and not in heart, lungs, spleen, liver, kidney, ovary, uterus, thymus, pharynx, or mesenteric lymph nodes of any mouse. Encephalitis was confirmed in all five clinically affected mice (Table 1) , with meningitis in four of these (mice 5 and 11-13); vasculitis was not identified. Encephalitic lesions were characterised by neuronal degeneration, microglial activation, glial reaction, perivascular cuffing and, where present, non-suppurative meningitis ( Figures 1A and 1B) . Viral antigen and lesions were detected in the olfactory tract, cortex by the olfactory tract and piriform lobe in each case (Table 3 ). In mouse 12 lesions were more extensive and included the olfactory bulb, amygdala, thalamus, hippocampus and pons. Except for the pons, all neuroanatomical sites involved are associated with the afferent pathways of olfaction in the brain. In addition, we analysed tissues and whole blood from each animal for the presence of viral genome using qPCR. All mice that developed clinical disease (excluding mouse 15 for which samples were not available) had high levels of viral genome present in brain tissue ranging from 10 6 . 3 to 10 11 . 8 HeV copies/10 12 18S copies ( Figure 2 ). All other mice were negative for viral genome in brain tissue. All tissues examined by qPCR including; heart, lung, thymus, pharynx, spleen, kidney, ovary, uterus and mesenteric lymph nodes, were negative for viral genome except in two cases; heart tissue of mouse 5 and lung tissue of mouse 14. In both cases C T values were high at 38.8 and 38.3 respectively, indicating low levels of genome and nearing the cut off for a positive sample set at a C T of 39.6. It is of interest to note that mouse 14 died comparatively early in the course of the study (day 12) and the viral RNA detected in the lungs was consistent with the results of mice that were euthanased in the earlier phases of the subsequent time course trial (described below). Low levels of viral genome (C T 39.2) were also detected in blood samples from one juvenile intranasally exposed mouse, mouse 3. Whether viral genome detected in the above three cases reflected replication at the time of euthanasia or residual genome from earlier transient infection could not be determined. Virus reisolation was attempted on all tissues positive for viral RNA by qPCR with a C T value less than 39.6. Virus isolation was negative for all tissues assayed including brain tissue where high levels of viral genome, antigen, and lesions were detected with and without neutralizing antibody. All animals that developed clinical disease were positive for a binding antibody response to HeV sG by Luminex assay. Neutralising antibody was also detected in two of these animals by SNT, however titres were evidently insufficient to control intracranial infection. In general, neutralising antibody titers were considerably lower than those reported in convalescent horses following field HeV infection [23] . We wished to ascertain whether the aforementioned outcomes of exposure were restricted to the C57BL/6 mouse strain. Accordingly, the intranasal component of the above study was repeated in the widely used BALB/c strain. Five juvenile and five adult BALB/c mice were challenged with 50,000 TCID 50 HeV via the intranasal route and monitored daily for 21 days. Clinical outcomes observed for BALB/c strain mice exposed to Hendra were generally similar to those observed for C57BL/6 mice. As with C57BL/6 mice, weight loss and temperature changes were not observed. All five juvenile mice remained healthy throughout the trial period, whereas three of the five aged mice (mouse 26, 27 and 29) developed clinical neurological disease necessitating euthanasia on 11, 17 and 18 DPI ( Table 1 ). Examination of tissues by histology and immuno-histochemistry revealed similarities and also differences to findings for C57BL/6 mice. In contrast to C57BL/6 mice, lesions and viral antigen were detected in nervous tissues of both symptomatic and asymptomatic animals ( Table 1 ). Two of the three animals that developed clinical disease had encephalitis and viral antigen in the brain: in the third animal (mouse 26) the olfactory bulb was not sampled, and only half the brain was available for histological assessment (the rest was used for qPCR and virus isolation). Four asymptomatic mice (one aged and three juvenile) had encephalitis and viral antigen in brain, with a further two asymptomatic, mice having encephalitis only. Encephalitic lesions and viral antigen deposition in BALB/c mice were also largely confined to neuro-anatomical sites associated with afferent pathways of olfaction ( Table 3 ). The lesions were generally more extensive in aged animals and, while this likely accounts for the differences in clinical signs observed here between old and young mice, it raises the possibility that following longer periods of observation illness may develop in young mice. Viral genome was detected in brain tissue of all young and aged BALB/c mice, apart from one juvenile mouse in which neither antigen nor lesions were identified ( Figure 2 ). Viral genome loads in brain ranged from 10 5.1 -10 13.1 copies HeV/ 10 12 18S; viral genome was also present in lung tissue of four of the five aged animals and one of the five juvenile animals (Mouse 23, Figure 2 ), albeit at lower levels than seen in brain tissue. Genome was also detected in pharyngeal tissue of one aged mouse (#29) at very low levels (C T 38.9), without lesions or HeV antigen by histology and immunohistochemistry. As with the C57BL/6 mice, virus could not be isolated from any samples including brain where high levels of viral genome, antigen, and lesions were detected and often in the absence of neutralizing antibody. Specific antibodies to HeV sG were detected in all mice by Luminex assay and neutralising antibody was detected in five of the ten animals (see Table 2 ). Again, SNT titres were low l (1:5 to 1:20) and did not correlate with clinical outcome. The above results suggested that aged mice were more susceptible to clinical disease following intranasal HeV exposure than juvenile animals. Using logistic regression analysis to determine if this trend was significant we found that 80% of aged animals developed clinical disease (s.e. = 12.7%) and only 10% of juvenile animals (s.e. = 9.5%) suggesting a real difference (70%, s.e. = 15.8%) in susceptibility to the development of disease between the aged and juvenile animals (p,0.01). The study investigating the susceptibility of BALB/c strain mice to HeV infection described above revealed that four out of five aged mice had viral RNA in lung tissue as late as 21 DPI in the absence of clinically apparent respiratory disease or detectable pulmonary pathology at euthanasia. In light of these findings, a time-course study was performed to evaluate whether infection of the upper and lower respiratory tracts developed in aged mice following HeV exposure. Aged BALB/c mice exposed to HeV by the intranasal route were euthanased at each time point of the following schedule; two mice every 48 hours from 1-14 DPI, two mice every 72 hours from 15 to 23 DPI, and three mice at 28 days PI, or on development of clinical disease according to predetermined humane endpoints defined from the previous study. Whole brain, lung, nasal turbinates and other tissue samples were examined by histology and immunohistochemistry. Lung tissue was further analysed for presence of viral genome and live infectious virus by qPCR and virus isolation, respectively (Table 4) . Mice sampled between days 6 and 14 pi were positive for HeV viral antigen in bronchoalveolar tissue, but antigen deposits were both dense and focal and it was not possible to distinguish on morphologic grounds whether alveolar lining cells, alveolar interstitium, or endothelial cells were involved. Live virus was reisolated from the lungs of mice sampled between 4 and 10 DPI (Table 4 ) consistent with transient infection of the lower respiratory tract after intranasal exposure to HeV. The viral titres detected were comparatively low with a maximum titre (10 2.3 TCID 50) recovered from a mouse on day 6. Interestingly, HeV antigen deposition in lung ( Figure 1D ) was not associated with detectable bronchoalveolitis and pulmonary vasculitis was not identified. In the upper respiratory tract, immunopositive cellssometimes accompanied by small focal necrotising inflammatory lesions (Figure1E and 1F), were detected in olfactory mucosa of mice between 6 to 17 DPI. It was not determined whether cells involved were supporting cells, basal cells or olfactory sensory neurons. Vasculitis and subsequent multi-organ infection is a feature of all reported animal models of HeV. In the initial observational study above there was little evidence of HeV virus spread to tissues other than brain apart from low levels of genome recovered from lung in some animals. To assess the extent of systemic involvement outside the respiratory tract, all other tissues for each animal collected in the time-course study were analysed for lesions and viral antigen/genome. All tissues examined apart from the aforementioned nasal mucosa, lung, and brain were negative for both lesions and antigen. In brain, immunopositive cells were first detected 6 DPI (Table 4) , and from 8 DPI lesions characterised by neuronal degeneration, microglial activation, glial reaction, perivascular cuffing and non-suppurative meningitis were also identified. Vasculitis was not detected in any animal. All other tissues excluding brain, which was fixed for histology and immuno-histochemistry in its entirety, were assessed for viral genome by qPCR and virus isolation was attempted from any positive samples. Some tissues, notably thymus that likely incorporated cranial mediastinal lymph node, were positive for viral genome (Table 5) , although negative for viral antigen and lesions. On occasion the relative genetic load was comparable to that seen in lung tissues although virus was not reisolated from any PCR positive tissues other than lung. For assessment of transient viremia, blood samples were also collected at euthanasia from each animal in the time-course study and RNA extracted using a Ribopure blood extraction kit. Previous work showed this extraction method to be the most sensitive for qPCR detection of HeV RNA in experimentally infected horses (A. Foord, personal communication). In mice, a low level (C T of 39.2) of viral RNA was detected in blood from only one animal (#111); this level of genome was very close to our cut off of C T of 39.6 (equivalent to 1 copy RNA from the standard curve). HeV Viral Antigen and Lesions in the Brain were Largely Confined to Neuroanatomical Sites Associated with the Afferent Olfactory Pathway As viremia and systemic spread was not an important feature of HeV in mice, neuroinvasion was unlikely to have been mediated via the haematogenous route. To better characterise the route of neuroinvasion, the distribution of viral antigen and lesions within the brain were examined in more detail. Brains collected at each point of the time-course study were transversely sectioned at 2 mm intervals; paraffin embedded sections were stained with haematoxylin and eosin for histology and with polyclonal anti-HeV N antibody for detection of HeV antigen (Table 6) . HeV antigen was first detected in clinically healthy mice euthanased on day 6 and was located in the olfactory bulb involving periglomerular cells, mitral cells, granule cells and associated cell processes ( Figure 1C ). By day 8 there was histological evidence of an inflammatory response closely associated with immunopositive cells. On 9 and 10 DPI three mice showing clinical signs of disease were electively euthanased. In addition to antigen in the olfactory bulb of the brain, two of these mice had antigen detectable in the piriform lobe, olfactory tubercle and amygdala which are all components of the primary olfactory cortex and are connected with each other via synapses. Clinically healthy mice euthanased between days 12 and 20 post infection were uniformly positive for antigen and also inflammatory lesions in the olfactory bulb and piriform lobes. Occasionally, lesions and antigen were seen in the amygdala and more caudally in the hippocampus. The latter feature was especially seen in those mice euthanased in the latter part of the time-course study (day 17 onwards); both of these anatomic structures are associated with the afferent pathways of olfaction in the brain. Three mice developed clinical disease between days 12 and 20 pi, one each on days 16, 17 and 18. As seen in Table 6 , lesions and antigen detected in the brains of these mice were more extensive than for the animals that did not develop clinical disease. In diseased mice, lesions and antigen were not only detected in the common structures described above, but also in higher processing structures of the cortex such as the thalamus and hypothalamus. Interestingly, two of these mice, mouse 116 and 120, were positive for antigen and lesions in the medulla and pons, and in mouse 116 the lesions and antigen were found in the vestibular nuclei proximal to the vestibulocochlear nerve. Mice 122 and 123, euthanased 23 DPI and 28 DPI respectively, were clinically healthy. In these mice, lesions and antigen were confined to the olfactory bulb. Table 3 . Distribution of histologic lesions and viral antigen in the brain of intranasally HeV challenged mice. Viral Antigen is Largely Restricted to Neuronal Cells within the Brain As encephalitis was a major feature of Hendra infection in mice, co-localisation studies were performed using confocal imaging to identify affected cell types within the brain. Perfused brains were collected from three aged BALB/c mice, on DPI 9, 10, and 11 and stained for HeV antigen. Antigen was seen in brain sections from all three mice and with a similar distribution to that observed by immunohistochemistry. Antigen appeared to be present in cells with neuronal morphology only ( Figure 3A) . In order to confirm this observation, sections were labeled with antibodies recognizing neurons (NeuN and NFP1), astrocytes (GFAP) microglia (IBA1) and oligodendrocytes (MBP). Dual labeling of sections of infected mice with these markers and with HeV antibodies indicated that there was no co-localisation of Hendra antigen with GFAP or MBP ( Figure 3B and 3C) . On rare occasions there was evidence of low intensity staining for Hendra viral antigen and microglial markers in single cells ( Figure 3D) . The Hendra protein labelling appeared to be discrete and circumscribed which would be consistent with its presence within a cellular compartment such as lysosomes. Compared with uninfected control tissue, there was reduced labeling of neurons with neuronal markers (NeuN and NFP1) in infected tissue. Therefore, we could not use colocalisation studies with these antibodies to confirm HeV antigen in neurons. In order to explore whether viral proteins could be identified within endothelial cells of capillaries or larger vessels, differential interference contrast (DIC) images were taken of areas of Hendra replication. The capillary lumen and endothelial cell cytoplasm were clearly identified and in no case was Hendra antigen detected in the cytoplasm of endothelial cells ( Figure 3E ). Taken together, this data suggest that HeV replication is restricted to neurons during infection of the brain. This paper reports the successful infection of mice following intranasal exposure to HeV, with reliable induction of viral encephalitis in aged mice of two strains. A previous investigation by Westbury et al [20] reported that juvenile immunocompetent BALB/c mice exposed to HeV by a parenteral route did not develop clinical or pathological signs of infection although intracranial inoculation of suckling mice had been uniformly lethal (G. Crameri, personal communication). Infection studies were not pursued further and up to now a mouse model of HeV infection has not been available. We have now confirmed progressive involvement of the mouse CNS by HeV which is most likely established via a nonhaematogenous route of neuroinvasion, as has been established or proposed for other paramyxoviruses including Sendai virus in mice [24] , Canine distemper virus in ferrets [25] , and the closely related Nipah virus in pigs [26] . Our data are consistent with anterograde HeV entry into the brain, possibly via olfactory sensory neurons (OSN). Viral antigen was detected in both the olfactory mucosa and the olfactory bulbs of intranasally challenged mice as early as day 6 post infection and these two structures are connected by olfactory sensory neurons (OSN). OSNs are located within the olfactory mucosa of the nasal cavity and project cilia out into the lumen of the nasal cavity to detect odorants [27] . Their axons project to the olfactory bulb of the brain [28] and thereby provide pathogens with a potential route of direct transmission from the nasal cavity to the brain. Within the olfactory bulbs OSNs synapse with mitral, tufted (projection neurons) and peri/juxtaglomerular cells (interneurons) in the glomeruli [29] , and secondary dendrites of the mitral and tufted cells synapse with dendrites of granule cells, which reside largely in the granule cell layer [30] . The trigeminal nerve may provide another possible route of neuroinvasion from the nasal cavity to the glomeruli of the olfactory bulbs. Peripheral peptidergic fibres of the trigeminal nerve innervate the olfactory epithelium and at least some of these fibres will send collaterals to the olfactory bulb (fibres terminate in the glomeruli) [31] en route to the trigeminal ganglion and the contribution of these fibres to viral invasion of the olfactory bulbs requires closer investigation. Viral antigen and lesions were also identified in the pons and medulla in four mice. We suggest that in these cases brain infection occurred via another peripheral neural pathway that enters the CNS at the brainstem such as the trigeminal [27] or the vestibulocochlear nerve [32] and it is of interest that in one animal antigen was localised to the vestibular nuclei of the medulla, which would be consistent with neuroinvasion of the vestibulocochlear nerve [32] after infection arising in the inner ear. The confinement of HeV antigen to neurons of mice, and the pattern and time course of viral antigen and lesions detected in their olfactory glomeruli, mitral and granule cell layers of the olfactory bulbs, as well as deeper structures in the olfactory pathway (such as elements of the olfactory cortex including the olfactory tubercles, piriform lobe and amygdala [33] , and then onto the hippocampus, thalamus and hypothalamus [27] ) suggests that HeV uses a transneuronal mode of spread within the murine brain that is mediated via synaptic connections. Transneuronal spread via synapses has been described for both H5 influenza virus [34] and Herpes simplex virus type 1 [35] . Transneuronal spread via synapses using cell contact dependent processes without budding of infectious virions may explain how HeV could not be reisolated from infected mouse brains, despite the presence of encephalitis, viral antigen and recovery of substantial viral genome. Transneuronal virus movement within the brain using cell contact dependent processes at the synapse without budding of infectious virions or syncytial cell formation has been proposed for another member of the family Paramyxoviridae, measles virus [36] . In humans, encephalitis represents a significant complication of HeV infection. Of the seven persons known to have contracted Hendra infection three have died from encephalitis and a fourth developed encephalitis and recovered [1, 4, 37] . Of the three patients that succumbed to encephalitis one case was a recrudescent infection, occurring 13 months after the acute disease [38] . Meningoencephalitis is also an important feature of the natural infection in horses, where it may persist weeks into convalescence, and is also a feature of other experimental infection models [2, 3, 39] . The pathogenesis of HeV infection of the central nervous system (CNS), including the means of spread throughout the brain and the mechanism of recrudescence, is poorly understood. The mouse model of HeV encephalitis lends itself well to pathogenesis studies of this type as the animals do not succumb to fulminating systemic infection prior to the establishment of significant neuropathology, the CNS lesions develop over a sufficiently long period of time that their progression can be monitored in considerable detail, and high levels of genome persist in brain in spite of virus neutralizing antibody, albeit that it occurs at low level. A particular advantage is that mice are small and comparatively easy to handle at BSL4 and there is an established library of available reagents for neuropathogenetic studies. HeV infection has also been established in two strains of mice, BALB/c and C57BL/6, providing investigators with the opportunity to employ transgenic mouse systems on two genetic backgrounds to examine disease processes, particularly with regard to the immune response in CNS infection. Although at present there is insufficient data available from human HeV cases to conclude whether neuroinvasion by the olfactory sensory neuron pathway is of primary pathogenetic importance in this species, we do note that the most likely route of human infection is contact droplet infection of the nasopharynx, and so exposure by this route is plausible. More importantly, the pattern of spread of HeV through the brain of mice is best explained by an ability of the virus to employ direct trans-neuronal transmission without concomitant generation of an infectious virion: these observations are of key pathogenetic significance to the human disease. Not only do they provide a possible explanation for failure to reisolate HeV from both sub-acute (I. Smith, unpublished data) and recrudescent [38] human cases of HeV using conventional techniques and in whom there was ample evidence of its ongoing replication, they raise important challenges to optimizing therapeutic interventions. A significant difference in response to infection is seen between aged and juvenile mice where aged mice appear to have a greater propensity to develop clinical disease compared with juvenile animals. A similar trend has been noted for other viral infections of mice, most notably SARS infection. In this case aged mice were susceptible to clinical disease development after challenge with SARS while their juvenile counterparts were not [40] . This observed difference in response to HeV infection between juvenile and aged mice can be used to model and investigate the mechanisms whereby age may affect clinical outcome. This model may be of particular value in studying aspects of recrudescence. In most other animal models of HeV infection, exposure rapidly leads to acute and fatal disease. In juvenile mice, we have observed that clinical disease may not develop (up to at least day 21 post exposure) despite strong evidence of viral replication in brain that is also continuing in the face of circulating antibody to G protein. Furthermore evidence of encephalitis in these animals was observed in the absence of vasculitis. Some of these characteristics are a feature of both recrudescent human HeV [1] and also NiV [41, 42] infection; the mouse HeV model could be employed to explore the mechanisms by which the potential for virus replication may be sustained in convalescent individuals. The HeV mouse infection model also provides an opportunity to study mechanisms of viral suppression. Our studies show that although intranasally exposed mice develop transient respiratory infection, the low viral titres detected suggest that infection is controlled in its early phase and cleared. Furthermore, this respiratory infection does not progress to a significant systemic disease as occurs in other animal species [2, 3, 13, 15] . Suppression of infection in the absence of evidence for a robust neutralising adaptive immune response suggests an important role for the mouse innate immune system in this process. HeV has been shown to employ various strategies to evade clearance and control by the innate immune system [43] and we suggest that in the mouse such strategies are rendered ineffective. It will be important in future research to elucidate the mechanism by which mice eliminate respiratory infection and resist systemic disease in order to guide the development of therapeutics that might mimic the process. In summary, this paper reports three novel observations. Firstly, mice are susceptible to HeV infection after intranasal exposure with aged animals reliably developing encephalitic disease. Their small size and ease of handling, as well as the vast range of biological reagents available for use in the species, render them highly valuable infection models for HeV encephalitis. Secondly, our data strongly support a role for not only anterograde neuroinvasion along sensory neurones but also transneuronal viral spread of HeV within the brain. This finding is of particular importance as infection may be established within the nervous system before current systemic treatments are able to be delivered across the blood-brain barrier. Lastly, mice are resistant to systemic vasculitis and fulminating HeV disease, the mechanisms of which can be explored with therapeutic potential. These findings provide a significant contribution to the body of knowledge in the field and have opened up new areas of investigation, from which significant understanding and further research can arise of benefit to humans as well as animals. For all studies reported in this paper, the animal husbandry and experimental design were approved by the CSIRO Australian Animal Health Laboratory's Animal Ethics Committee. All animal experimentation was conducted following the Australian National Health and Medical Research Council's Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. Ten aged (12-13 months) and ten juvenile (7-8 weeks) C57BL/6 mice and thirty-one aged (10-13 months) and five juvenile (7-8 weeks) BALB/c mice were used for challenge experiments. Mice were housed in groups of four to five according to age, in cages in a room at BSL4. Animals were fed once daily with complete mouse chow and provided with water ad libitum. Animals were allowed 7 days to acclimatise before challenge and were implanted with a subcutaneous temperature chip 5 days prior to challenge. For this procedure and all others requiring restraint for manipulations, an intraperitoneal injection of ketamine (75mg/ kg; Ketamil; Ilium, Smithfield, Australia) and medetomidine (10mg/kg; Domitor; Novartis, Pendle Hill, Australia) was used to induce anaesthesia. Anaesthesia was reversed by intraperitoneal administration of atipemazole (1mg/kg; Antisedan; Novartis). Animals were monitored daily at which time temperature, weight and clinical data were collected and recorded. Staff wore fully encapsulating suits with an external air supply and all work with live virus was carried out at BSL4. Mice were anaesthetised and exposed to a low passage clinical isolate of HeV (Hendra virus/Australia/Horse/2008/Redlands) [44] by either the intranasal or subcutaneous route. For the intranasal route, five aged (12 months) and five juvenile (8 weeks) Table 6 . Distribution of histologic lesions and viral antigen in the brain of intranasally HeV challenged mice. mice were anaesthetised, placed in dorsal recumbency and exposed to 50,000 TCID 50 in 50 ml saline by slow drop application to the nares. For the subcutaneous route, five aged and five juvenile mice were anaesthetised and exposed to 50,000 TCID 50 virus in 200 ml saline by subcutaneous injection. Animals were monitored daily thereafter and euthanased when reaching a previously determined end-point or at 21 days post infection (pi). The humane end-point was defined as a constant weight loss recorded over 3 days or reaching a 20% loss of pre-infection weight, and/or clinical signs consistent with neurological involvement including ataxia, tremors, depression and behavioural changes. As clinical signs of Hendra infection have not been observed in mice previously, the pre-determined endpoint was modelled on infection outcomes recorded for guinea pigs. The intranasal component of the above study was repeated in five aged (12 months) and five juvenile (8 weeks) BALB/c strain mice. Aged (10-13 months) BALB/c strain mice were challenged with 50,000 TCID 50 of HeV via the intranasal route. Mice were euthanased and samples collected every second day from days 2-14 pi, every 3 days from days 17-23 pi and, lastly, on day 28 pi; at least one replicate was used for each time point in case mice were lost from the study for reasons not associated with HeV, with three mice being sampled on day 28. If mice reached a pre-determined humane endpoint prior to the scheduled sampling day they were euthanased and samples collected and processed as for animals euthanased according to the schedule. Three aged (12-13 months) BALB/c strain mice were exposed to 50,000 TCID 50 of HeV using the intranasal route described above. Deeply anesthetised mice were perfused with paraformaldehyde on 9, 10, and 11 DPI for confocal analysis of the brain. At the end of the study, animals were anaesthetised and blood was collected via cardiac puncture, placed into EDTA and serum separator tubes. Following terminal exsanguination brain, heart, lungs, liver, kidney, pharynx, mesenteric lymph nodes, ovaries, uterus, spleen and thymus tissues were collected. For virus isolation and molecular studies samples were placed into 750 ml viral transport media [PBS with 1% bovine serum albumin and 1x antibiotics (Anti-anti, Invitrogen)] with 250 ml aluminium silicate beads (Biospec Products Inc., Bartlesville, OK, USA). The remaining tissues were fixed in 10% neutral buffered formalin for 48 hours prior to routine processing for histology and immunohistochemistry, the latter using a rabbit polyclonal antibody raised against the NiV N protein [45] . Nasal turbinate and oral swab samples were also collected at euthanasia, swabs were placed into 750 ml viral transport media. Tissue samples were homogenised and aliquots removed for RNA extraction and the remainder was stored at 280uC for subsequent processing. For the time course study, the brains of all animals were fixed in their entirety and not sampled for virus isolation and molecular studies. Blood was collected via cardiac puncture and 500 ml whole blood placed into 1.3 ml RNAlater for Ribopure RNA extraction. Otherwise sampling was as described above. For HeV antigen localisation in brain cell studies, animals were deeply anesthetised and euthanased by perfusion with paraformaldehyde to achieve optimal fixation of tissues for confocal microscopic studies. The thoracic cavity was opened and the right auricle of the heart removed. Three ml saline was injected into the left ventricle of the heart to achieve exsanguination. Following saline injection, 11 ml of 4% paraformaldehyde (volume/volume) was slowly injected into the left ventricle of the heart, perfusing the entire animal. Following perfusion, mice were placed under 4% paraformaldehyde soaked tissues and allowed to fix for 1 hour. After initial fixation, the brain was removed and immersion fixed in 4% paraformaldehyde (volume/volume) for a further 24 hours after which it was placed into PBSA for processing. For RNA extraction, homogenised tissue samples or swab samples from studies 1-3 were centrifuged at 16 000 g for 2 minutes to pellet debris and 100 ml supernatant was mixed with 265 ml MagMAX Lysis/Binding solution (Ambion, Victoria, Australia) and removed from BSL4. RNA was extracted using the MagMax-96 viral RNA isolation kit (Ambion). For the susceptibility studies, 100 ml of EDTA blood was mixed with 265 ml MagMAX Lysis/Binding solution (Ambion, Victoria, Australia) and RNA extracted as described above. For the time course study, RNA was extracted from whole blood mixed in RNAlater using the Mouse RiboPure-Blood RNA Isolation Kit (Ambion). TaqMan qPCR was performed using the AgPath-ID one-step reverse transcription-PCR kit (Applied Biosystems, Victoria, Australia), targeting the N gene of HeV as previously described [46] . Positive results were defined by a cycle threshold (C T ) value of ,39.6 based on a standard curve where this C T represented one copy of target RNA. All samples were normalised against the housekeeping gene 18S and expressed as copy numbers HeV/10 12 copy numbers of 18S. Virus isolation was performed on all HeV RNA positive samples detected in the HeV Taqman assay using the standard Vero cell line maintained in our institute and originally obtained from the American Tissue Culture Collection (ATCC) [10, 39] . Supernatants from the homogenised samples were inoculated onto Vero cell monolayers and scored positive if syncytia were present after 4 and 5 days. Sera collected both prior to challenge and at euthanasia were analysed for binding to HeV sG protein using a Luminex microsphere assay. Assays were performed on a Bio-Plex Protein Array System (Bio-RadLaboratories, Inc., CA, USA) as previously described [22] . Bio-Plex Manager Software (v 4.1) (Bio-RadLaboratories, Inc., CA, USA) was used for data acquisition and analysis. All samples were assayed simultaneously for each mouse strain. A strong positive control was assayed alongside all samples as was a negative control. A positive result was defined as samples occasionally seen in microglia, identified with labelling for Iba1 (green) in olfactory bulb at day 9 PI. HeV antigen labeling (arrow) was discrete and circumscribed, consistent with its presence within a cellular compartment such as a lysosome. (E) Section of olfactory bulb showing a capillary amongst HeV infected cells at day 10 PI. HeV antigen (red) and GFAP (green) are not colocalised. The endothelial cell cytoplasm (arrow) is negative for HeV antigen. Scale bars = A) 50 mm, B) 10 mm, C) 15 mm, D) 10 mm and E) 10 mm. doi:10.1371/journal.pone.0040308.g003 with a median fluorescence intensity (MFI) greater than the mean of all pre-challenge samples plus 3 x the standard deviation, MFI .406. Sera collected at euthanasia were gamma-irradiated to inactivate virus. Sera were serially-doubly diluted in a 96 well plate (final volume 50ul/well) to which 200 TCID 50 HeV was added and incubated for 1 hr at 37uC. Following incubation, 2x10 4 Vero cells/well were added and the assay read after 3 days incubation at 37uC with 5% CO 2 . Whole brains, collected and fixed in 4% paraformaldehyde were immunolabelled as previously described [47] . In brief, they were sectioned at 50 mm thickness with a Leica Vibrating Microtome and stored in PBS at 4uC. Sections were treated with 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 1 hour and blocked with 0.5% Bovine serum albumin (Sigma-Aldrich) in PBS (PBS/BSA) overnight at 4uC. Primary antibodies were diluted in 0.5% PBS/BSA and incubated on sections for 2 hours at 37uC. Antibodies and dilutions were: rabbit anti-IBA1-microglia marker PBS. Mouse Ig Blocking Reagent (Vector Laboratories) was used with mouse primary antibodies: anti-Human Neurofilament protein (DakoCytomation) 1:50, anti-NeuN, (Millipore) 1:50, and a mouse monoclonal antibody raised against whole HeV (AAHL). Bound primary antibody was detected with species-specific secondary antibodies conjugated to Alexa 488 or 568 (Invitrogen) diluted 1:200 in PBS/BSA for 2 hours at 37uC. Sections were washed 3 times for 5 minutes with PBS and nuclei labelled with DAPI diluted 1:1000 in dH 2 O for 30 minutes. Sections were rinsed twice with dH 2 O, mounted with Vectashield Mounting Medium (Vector Laboratories) and coverslips sealed with nail varnish.

Ipomoelin, a Jacalin-Related Lectin with a Compact Tetrameric Association and Versatile Carbohydrate Binding Properties Regulated by Its N Terminus

Many proteins are induced in the plant defense response to biotic stress or mechanical wounding. One group is lectins. Ipomoelin (IPO) is one of the wound-inducible proteins of sweet potato (Ipomoea batatas cv. Tainung 57) and is a Jacalin-related lectin (JRL). In this study, we resolved the crystal structures of IPO in its apo form and in complex with carbohydrates such as methyl α-D-mannopyranoside (Me-Man), methyl α-D-glucopyranoside (Me-Glc), and methyl α-D-galactopyranoside (Me-Gal) in different space groups. The packing diagrams indicated that IPO might represent a compact tetrameric association in the JRL family. The protomer of IPO showed a canonical β-prism fold with 12 strands of β-sheets but with 2 additional short β-strands at the N terminus. A truncated IPO (ΔN10IPO) by removing the 2 short β-strands of the N terminus was used to reveal its role in a tetrameric association. Gel filtration chromatography confirmed IPO as a tetrameric form in solution. Isothermal titration calorimetry determined the binding constants (K(A)) of IPO and ΔN10IPO against various carbohydrates. IPO could bind to Me-Man, Me-Glc, and Me-Gal with similar binding constants. In contrast, ΔN10IPO showed high binding ability to Me-Man and Me-Glc but could not bind to Me-Gal. Our structural and functional analysis of IPO revealed that its compact tetrameric association and carbohydrate binding polyspecificity could be regulated by the 2 additional N-terminal β-strands. The versatile carbohydrate binding properties of IPO might play a role in plant defense.

Plant defense is a complicated mechanism in response to mechanical wounding, herbivore and microorganism attack. Many proteins, namely wound-inducible proteins, are expressed to prevent pathogen infection, inhibit digestion by insects, and repair injured tissues [1, 2] . One group of wound-inducible proteins is lectin, the carbohydrate binding protein [3, 4] . Plant lectins are involved in the plant defense mechanism because of carbohydrate binding properties [5] [6] [7] [8] [9] . The toxicity of lectins was also confirmed in animal experiments [10, 11] . Plant lectins show resistance to digestive enzymes and can bind selectively to the carbohydrate moieties of gut epithelial cells to interfere in nutrient digestion and absorption [12] , so they could be a natural insecticide. In addition, plant lectins have been used for blood typing and immunological assay. The lectin concanavalin A is commercially used in affinity chromatography for purifying glycoproteins. Plant lectins have long been reported as potential inhibitors of viruses [13] [14] [15] [16] [17] . Most plant lectins were originally isolated from seeds and vegetative storage tissues. Accumulating data have revealed that plants ubiquitously synthesize lectins in response to abiotic and biotic stresses. These inducible lectins are synthesized and then exported to vacuoles by signal peptides or reside in the cytoplasm [18, 19] . The physiological function of plant lectins for subcellular localization remains obscure. However, the major assumption is that lectins are involved in defense and may also have a role in signal transduction for response to stress [20] . Structure analysis of plant lectins demonstrated a diverse group of proteins that can be classified into 6 different groups (http:// www.cermav.cnrs.fr/lectines/): monocot lectin, hevein domain lectins, b-prism lectins, b-trefoil lectins, cyanovirin-N homologs, and legume lectin. Jacalin-related lectins (JRL) have a b-prism fold. In 1996, the structure of Jacalin from seed of jackfruit (Artocarpus integrifolia) was first reported to have a tetrameric association for binding to galactose [21] . Later, Maclura pomifera seed agglutinin was reported to have the same tetrameric structure as Jacalin [22] . The other lectin, Artocarpin, from seed of jackfruit (Artocarpus heterophyllus) shares the same tetrameric association for binding to mannose [23] . Moringa M from black mulberry (Morus nigra) forms a tetrameric association like that of Jacalin [24] . JRLs were once thought to be confined to the Moraceae. However, increasing structural evidence reveals that the lectins with a bprism fold exist universally in plants and animals [25] but with different quaternary association. Heltuba is a plant tuber lectin from Helianthus tuberosus (Jerusalem artichoke) that has a donut shape with an octahedral assembly by the b-prism building block [26] . Caselpa is a rhizome lectin from Calystegia sepium (Hedge bindweed) that has a dimeric form [27] . PPL is a plant seed lectin from Parkia platycephala that contains 3 repetitive bprism domains and forms a dimeric form with hexahedral assembly [28] . Ipomoelin (IPO), expressed in the leaves of sweet potato (Ipomoea batatas cv. Tainung 57), was found easily inducible by wounding and methyl jasmonate [29, 30] . Previous study showed that IPO can agglutinate human blood and bind to different carbohydrates, such as methyl a-D-mannopyranoside (Me-Man), methyl a-Dglucopyranoside (Me-Glc), mannose, glucose and galactose [10] . In this study, we resolved the crystal structures of IPO in the apo form and in complex with Me-Man, Me-Glc and methyl a-Dgalactopyranoside (Me-Gal) to reveal the different quaternary associations of IPO and its binding pocket for carbohydrates. A The packing diagram of apo IPO with 5 molecules in green. Four of 5 molecules form a tetramer and the 5th molecule can form another tetramer by the red one, the yellow one and the blue one in the center. The molecules in red, yellow and blue are generated by symmetric operations (-X, Y, -Z), (X, -Y, -Z), and (-X-1, -Y, Z). (B) The resolved IPO-Me-Man complex in green is 2 molecules in an asymmetric unit. However, the other 2 molecules in red are generated by the symmetric operation (X, -Y, -Z) to form a tetramer in the center. (C) The resolved IPO-Me-Glc complex is a tetrameric form, and (D) the IPO-Me-Gal complex is also a tetramer. All packing diagrams reveal its tetrameric nature. doi:10.1371/journal.pone.0040618.g001 The carbohydrate binding pockets are indicated by green mesh. Tetrameric IPO is represented by monomer A in blue, monomer B in purple, monomer C in light blue, and monomer D in pink. The symmetric axis is represented by a black ellipse in the center of tetramer. (C) Structure-based multiple sequence alignment of Jacalin family. Five homologs were selected for sequence comparison from resolved protein structures: Ipomoelin-tetramer from Ipomoea batatas (PDB: 3R52); Calsepa-dimer from Calystegia sepium (PDB: 1OUW); Banlec-dimer from Musa acuminate (PDB: 2BMZ); Jacalin-tetramer from Artocarpus hirsutus (PDB: 1TOQ); Parkia-hexamer from Parkia platycephala (PDB: 1ZGS); and Heltuba-octomer from Helianthus tuberosus (PDB: 1C3K). Positions of identical conserved residues are shown in white on dark grey background, and regions of similarly conserved residues in light grey are boxed. Representation of secondary structure elements and numbering above the alignment is based on the IPO structure. The secondary structure elements below the alignment are based on the Heltuba truncated IPO (DN10IPO) was prepared to reveal its role in tetrameric association in solution by gel filtration chromatography. In addition, the carbohydrate binding constants of IPO and DN10IPO were determined by isothermal titration calorimetry (ITC). DN10IPO showed a recovered mannose/glucose-specific lectin. Structural and functional analysis identified IPO as a member of the JRL family but with a different tetrameric association. The N-terminus of IPO plays a critical role in regulating broad carbohydrate binding. The apo IPO showed an orthorhombic space group of I222. A reasonable volume of the unit cell (Vm) for the Matthew coefficient was estimated at 2.19 Å 3 /Da and 44% solvent content by 8 IPO molecules. However, only 5 IPO molecules in an asymmetric unit could be built after molecular replacement. We structure. The carbohydrates Me-Man, Me-Glc, and Me-Gal share 9 hydrogen-bonding interactions with Gly21, Tyr97, Gly141, Trp142, Tyr143 and Asp145 of IPO (blue triangle). The residues of IPO located at the interface are boxed in red. The two short b strands at the N terminus are also involved in the interface. The underlined Jacalin-tetramer representing the sequence is extracted from the C terminus of Jacalin (chain B). doi:10.1371/journal.pone.0040618.g002 obtained a higher Matthew coefficient with 3.51 Å 3 /Da and 65% solvent content. In the packing diagram for apo IPO, we observed a tetrameric association with an additional monomer in an asymmetric unit ( Figure 1A ). The additional monomer could form a tetrameric association with the other 3 neighboring molecules, which were generated by symmetric operations (-X, Y, -Z), (X, -Y, -Z), and (-X-1, -Y, Z). So 4 IPO molecules could form a tetramer. To determine the carbohydrate binding pocket of IPO, carbohydrates such as Me-Man, Me-Glc and Me-Gal were used to co-crystallize with the IPO protein. The crystals of IPOcarbohydrate complexes were determined in different space groups. IPO-Me-Man belongs to an orthorhombic space group C222 1 . The Matthew coefficient and solvent content for IPO-Me-Man had a reasonable value of 2.21 Å 3 /Da and 44.4% for 2 molecules in an asymmetric unit. Although only 2 IPO molecules were built in the IPO-Me-Man complex, the other 2 IPO molecules could be generated by symmetric operation (X, -Y, -Z) and resulted in a tetrameric association ( Figure 1B ). The crystal of IPO-Me-Glc was determined to be a monoclinic space group P2 1 . The Matthews coefficient and solvent content was 2.26 Å 3 /Da and 45.5% for 4 molecules. The packing results for IPO-Me-Man and IPO-Me-Glc indicated that the carbohydrates binding to IPO might result in a compact packing as compared with that of apo IPO. In addition, the resolved structure of IPO-Me-Glc formed a tetrameric association ( Figure 1C ). IPO-Me-Gal belongs to an orthorhombic space group P2 1 2 1 2 1 . The Matthews coefficient and solvent content were 2.25 Å 3 /Da and 45.2%, respectively, for 4 molecules in an asymmetric unit. The 4 IPO-Me-Gal molecules shown in Figure 1D form the same tetrameric association as that of IPO-Me-Glc. On the basis of crystal packings of apo IPO and IPO-carbohydrate complexes, IPO would form a tetrameric association. The monomeric IPO from residues 1 to 154 shows a typical bprism fold found in the JRL family, with 12 b-sheets (b3-b14) and 2 additional short, extended, N-terminal b-strands (b1-b2) (Figure 2A and 2C) . Each b-prism fold comprises 3 Greek-key motifs forming 3 planes by 3 four-stranded b-sheets: plane 1 by b3 to b4 and b13 to b14; plane 2 by b5 to b8; plane 3 by b9 to b12. Furthermore, the structure of these b-sheets comprises b1 from residues Gln4 to Leu5, b2 from residues His8 to Ser9, b3 from residues Ala11 to Gly17, b4 from residues Gln22 to Arg27, b5 from residues Lys34 to Gly41, b6 from residues Leu47 to Ser55, b7 from residues Ile61 to Gly65, b8 from residues Tyr74 to Asn79, b9 from residues Ile84 to Tyr94, b10 from residues Tyr97 to Thr107, b11 from residues Glu111 to Gly116, b12 from residues Thr121 to Lys126, b13 from residues Asn131 to Ser140, and b14 from residues Val144 to Ala153 (Figure 2A and 2C). Four IPO protomers form a compact tetrameric association by swapping their extended N termini from residues 1 to 10. We analyzed the tetrameric association of IPO-Me-Glc. As shown in Figure 2B , the 2 extended N termini from monomer A in blue and monomer B in purple swap with each other. The interacting interface between the four IPO protomers is formed by the extended N termini. Consequently, a larger buried interface between monomers A and B is 1,522 Å 2 . The residues located at the interface are 2-10, 12, 15-30, 59-67, 91-92, 98, 121, 134, 137, 139-140, 146, 150, and 152 in monomer A (as shown in red box in Figure 2C ). In total, 13 hydrogen bonds are formed by the residues Leu5, His8, Asn19, Gln22, Ser25, Arg27, Asp60, Ile61, Thr63, Thr121, Asn139 and Tyr150 in the interface between monomers A and B. The buried interface between monomer C and monomer D is 1,554 Å 2 . Furthermore, the buried interface between monomers A and C is 755 Å 2 , which is mainly contributed by the interacting residues of N-terminal residues 4 to 17 and C-terminal residues 91, 121-126, 128, and 151. In addition, the interface between monomers D and B is 731 Å 2 . The carbohydrate binding pocket of IPO was confirmed at loops b13 and b14 by the structures of IPO-Me-Glc, IPO-Me-Man and IPO-Me-Gal (as shown in Figure 2B with green mesh). In the chain A of IPO-Me-Glc, 9 hydrogen bonds are formed by the residues Gly21, Tyr97, Gly141, Trp142, Tyr143 and Asp145 of IPO and the atoms O1, O3, O4, O5, and O6 of Me-Glc ( Figure 3A and Table 1 ). The atom C7 of Me-Glc is involved in the methyl carbon (Me)…p interaction with Trp142 of IPO. The hydrogen bonds are slightly different between chain A and chains B to D. The hydrogen bonds of chains B to D are formed between the same residues of chain A and Me-Man, except for Table 1 ). The differences might result from the binding of cadmium ion (Cd 2+ ). In chains B to D, the Cd 2+ atom forms 5 coordinates by the O atom of the carbonyl group of Asn19, OG atom of Ser18, and 3 water molecules. One of the 3 water molecules forms a hydrogen bond with Asp145 ( Figure 3B ). In the structure of IPO-Me-Man, two IPO protomers were built, and only one Me-Man molecule could be observed in chain A. The temperature factor of Me-Man in the structure of IPO-Me-Man is 66.5 Å 2 , which is higher than that of Me-Glc, with 34.5 Å 2 (Table 1) . This phenomenon might indicate that only a few Me-Man molecules bound to IPO proteins in IPO-Me-Man, which resulted in a higher temperature factor. Nine hydrogen bonds are formed by the residues Gly21, Tyr97, Gly141, Trp142, Tyr143, and Asp145 of IPO and the atoms O1, O3, O4, O5 and O6 of Me-Man ( Figure 3C and Table 1 ). The atom C7 of Me-Man is also involved in the Me…p interaction with Trp142 of IPO. The binding orientation of Me-Man is similar to that of Me-Glc. In the structure IPO-Me-Gal, 10 hydrogen bonds are formed by the same residues Gly21, Tyr97, Gly141, Trp142, Tyr143, and Asp145 of IPO ( Figure 3D and Table 1 ). The atom C7 of Me-Gal is shown in the Me…p interaction with Trp142 of IPO. This revealed the importance of the methyl group of carbohydrates for binding to IPO. To validate that the quaternary association of IPO is also a tetrameric form in solution, purified IPO was used in gel filtration experiments. The molecular mass of IPO could be calculated according to the linear regression equation of the standard protein markers purchased from BioRad ( Figure 4C ). In the preliminary study, IPO protein was dissolved in running buffer (27 mM Tris-HCl pH 7.0, 2 M NaCl) without additional carbohydrates. We obtained a retarded result, with corresponding molecular mass 4.0 kDa (Peak 3 in Figure 4A ). Thus, IPO has the binding ability of dextran in the matrix of the Superdex 200 column. To eliminate the binding effect of IPO to dextran, running buffer was prepared with an additional 0.2 M Me-Glc, and a shift of the IPO peak could be observed, with corresponding molecular mass of 53.3 kDa (Peak 2 in Figure 4A ). Consequently, running buffer with an additional 1 M glucose was prepared to totally eliminate the binding effect of IPO. The corresponding molecular mass of IPO in solution was 64.7 kDa (Peak 1 in Figure 4A ). The molecular mass of recombinant IPO with a His tag was 17.3 kDa for a monomer and 69.2 kDa for a tetramer. The results from gel filtration experiments demonstrated that IPO shows a tetrameric association in solution. To further identify the role of the N terminus in the tetramerization of IPO, we prepared a truncated IPO (DN10IPO) by removing residues 1 to 10 to monitor the change in quaternary association. The native IPO protein or the truncated IPO protein was dissolved in the running buffer with 1 M glucose. Peak 1 in Figure 4B represents the native IPO, with molecular mass 63.2 kDa, which is a tetrameric size. Peak 2 in Figure 4B represents the DN10IPO, with molecular mass 21.9 kDa, which is near the truncated monomer size (16.3 kDa). The results further confirmed that IPO has a tetrameric association and its N terminus plays an important role in forming a tetramer. To determine the binding constants of IPO to Me-Man, Me-Glc and Me-Gal, 1 mM IPO solution was titrated with 25 mM carbohydrate solution. The interaction of IPO and carbohydrate was an exothermal reaction. The optimal curves and thermodynamics parameters could be fitting and calculated by Microcal Origin 7.0. The K A of IPO to Me-Man was the highest, 7.04610 3 M 21 . The K A values for Me-Gal and Me-Glc were 4.09610 3 M 21 and 2.01610 3 M 21 , respectively (Table 2 and Figure 5 ). Subsequently, carbohydrates without the methyl group were used to determine the binding affinity of IPO. From preliminary study, 1 mM IPO titrated with 25 mM Man, Glc, and Gal revealed no obvious exothermal reaction. After increasing the concentration with 3 mM IPO titrated with 75 mM Man, Glc, and Gal, the exothermal curves could be observed and calculated. The K A values for IPO binding to Man, Gal and Glc were 1.05610 2 M 21 , 0.57610 2 M 21 , and 0.32610 2 M 21 (Table 2 and Figure 5 ). Thus, the interactions between IPO and carbohydrates were stronger with than without the methyl group. Figure 6 ). Thus, the N-terminus of IPO is involved in tetramerization in regulating the binding affinity to carbohydrates. We submitted the coordinates of a monomer of apo IPO (e.g., chain A; Figure S1C ) to the web service Matras for 3-D protein structure comparison [31] . We found the highest Z-score, 124.5, for the template structure, a dimeric form of Calsepa from Calydyrgia sepium (PDB: 1OUW; Figure S1D ) [27] , in our molecular replacement procedure. The following structures were PPL from Parkia platycephala with a hexahedral ring (PDB: 1ZGR; Figure S1E ) [28] , Heltuba from Helianthus tuberosus with an octahedral ring Figure S1F ) [26] , Banlec from banana with an another kind of dimeric form (PDB: 2BMZ; Figure S1B ) [32] , and Jacalin from jackfruit seeds with a tetrameric form (PDB:1UGW; Figure S1A ) [33] . These data indicate the various quaternary structures in the JRL family, despite the same b-prism fold of protomer. The various quaternary associations in the JRL family exhibited different contacts between protomers. A previous report indicated that the buried interface of the Calsepa dimer is 1,327 Å 2 by a probe with 1.6 Å radius [27] . Here, we analyzed the buried interface of the selected structures from the above comparison by using the PDBe PISA service with 1.4 Å radius [34] . The buried interface area from tetrameric IPO encompasses 1,539 Å 2 , which is larger than that of Calsepa (1,202 Å 2 ), PPL (1,294 Å 2 ), Banlec (750 Å 2 ), Heltuba (736 Å 2 ), and Jacalin (1023 Å 2 ). The N terminus of the protomer in the JRL family has an important role in the quaternary association by swapping in the interface and then forming a dimer, tetramer, hexamer, and octomer. To compare the difference between the tetrameric Jacalin ( Figure S1A ) and the tetrameric IPO ( Figure S1C ), the tetramer of Jacalin showed a looser interface than that of IPO. Therefore, IPO formed a different compact tetramer. In this study, we resolved the crystal structures of IPO-Me-Man, IPO-Me-Glc and IPO-Me-Gal complexes. These monosaccharides showed similar orientation to bind to IPO. The binding pocket of IPO contains 6 residues such as Gly21, Tyr97, Gly141, Trp142, Tyr143 and Asp145, to form hydrogen bonds with different monosaccharides (Figure 3 Table 2 ). The carbohydrate binding manner of IPO is not confined as is the mannose-glucose-specific binding lectin. In addition to determining monosaccharides with the methyl group, we used monosaccharides without a methyl group, such as mannose (Man), glucose (Glc), and galactose (Gal), to determine their binding constant to IPO. Since the lower binding affinity of IPO titrated with Man, Glc or Gal couldn't get the best fitting for the titration curves, the n value was consequently fixed at 1.0 for fitting the curves ( [36] . The results show no differences with or without the methyl group of monosaccharides for binding properties in Artocarpin and Banlec possibly because of no aromatic side chain of residues in Artocarpin and Banlec like the residue Trp142 in IPO ( Figure 7A and 7B) . Interestingly, IPO shared similar binding properties to Jacalin for its Tyr122, which [37] . To examine the binding mode of Me-Man for Artocarpin, Banlec, Jacalin, and IPO, the binding position of Me-Man with IPO showed a distant binding site as compared with that for Artocarpin, Banlec, and Jacalin ( Figure 7D ). DN10IPO could be recovered as the mannose/glucose specific lectin if DN10IPO represented the monomeric IPO and wild-type IPO represented the tetrameric IPO. The monomeric IPO showed 5 times and 6 times binding affinity to Me-Man and Me-Glc, respectively, as compared with those of tetrameric IPO. Therefore, the N terminus of IPO is involved in the carbohydrate recognition, which results in the carbohydrate binding polyspecificity of tetrameric IPO. From the tetrameric IPO structure, the residue Leu5 and His8 in the N terminus of monomer B (chain B) forms 3 hydrogen bonds with the residue Asn19 in the loop between b3 and b4 of monomer A (chain A) (Figure 8 ). The hydrogen bonds might pull out the loop of b3-b4 and form a larger binding cavity for different carbohydrates in monomer A. However, in DN10IPO, the hydrogen bonds would disappear and might relocate the b3-b4 loop to cause a smaller binding cavity. The axial O4 of Me-Gal would not easily enter into the smaller binding cavity. The results might be confirmed by the crystal structure of DN10IPO-Me-Man in further study. In conclusion, we resolved the structures of apo IPO and IPO in complex with Me-Man, Me-Glc and Me-Gal. IPO is proposed to have a tetrameric association by 4 protomers of the b-prism with an additional N terminus, which shows a compact tetrameric association in the JRL family. From gel filtration experiments, we confirmed the tetrameric association of IPO in solution. The N terminus of IPO plays an important role in forming a tetramer. In addition, the binding pocket of IPO was identified and found to bind to Me-Glc, Me-Man, and Me-Gal with similar hydrogen bond networks. Furthermore, the binding constants of IPO were determined by ITC. The IPO structures further extend the diverse quaternary structures of the JRL family of plants and show versatile carbohydrate binding properties regulated by the N terminus. Thus, the wound-inducible protein IPO from sweet potato has versatile carbohydrate binding properties and might play a role in plant defense. The expression vector pTZ18UH containing the IPO gene [GenBank: D89823.1] (pTZ18UH-IPO) of sweet potato (I. batatas cv. Tainung 57) was constructed previously [10] . A truncated form of IPO by removing 10 residues of N terminus (pTZ18UH- The pTZ18UH-IPO and pTZ18UH-DN10IPO vectors were transformed into Escherichia coli BL21 (DE3) cells (Novagen). A single colony was cultured in 5 ml LB medium containing 100 mg/ ml ampicillin (LB/Amp) at 37uC overnight. The medium was further transferred into 600 ml LB/Amp to an A 600 of about 0.5 to 0.7 and then induced with 0.1 mM isopropyl-b-D-thiogalactopyranoside (IPTG) at 25uC for 6 hr. Cells harvested by centrifugation were resuspended in a loading buffer (20 mM sodium phosphate, pH 7.4, 0.5 M sodium chloride, 20 mM imidazole). After breaking cells by use of an ultrasonicator (Sonicator 3000, Misonix), the supernatant of the crude cell lysate was loaded onto a Histrap FF column (GE Healthcare) with use of an Ä kta Prime fast protein liquid chromatography (FPLC) system (GE Healthcare). After washing the Histrap FF column with 3x column volume of loading buffer (1x phosphate buffered saline, 5 mM adenosine triphosphate, 10 mM MgSO 4 ), the IPO protein was eluted by use of elution buffer (50 mM sodium phosphate, pH 7.4, Table 3 . Crystallography statistics for apo ipomoelin (IPO) and IPO in complex with carbohydrates methyl a-D-mannopyranoside (Me-Man), methyl a-D-glucopyranoside (Me-Glc) and methyl a-D-galactopyranoside (Me-Gal). sodium formate, 40% w/v polyethylene glycol 3,350. The crystals of IPO-Me-Gal appeared in 7 days. A mixture of the reservoir solution with 100% glycerol in a 4:1 volume ratio was used as cryo-protectant for data collection. The diffraction data were collected at 100K and detected by a Quantum 315 or Quantum 210 CCD detector at the BL13B1 or BL13C1 beamlines of NSRRC (Hsinchu, Taiwan). All diffraction data were processed and scaled with use of the HKL2000 program [38] . The diffraction statistics are in Table 3 . We used a blastp search for the amino acid sequence of IPO [GenBank: BAA14024.1] against the algorithm of the National Center for Biotechnology Information (NCBI) protein databank database for searching structural templates. The amino acid sequence of Calystegia sepium agglutinin (Calsepa), a JRL (PDB: 1OUW), showed 53% sequence identity to that of IPO. The monomeric structure of Calsepa was further used in a search to determine the structure of apo IPO by molecular replacement with use of the program CNS [39] . After cross-rotation and translation of molecular replacement, 4 values were obtained. Initial rigid body refinement for the 4 monomeric structures gave a 48.8% R-factor. Clear continuous electron density could be observed after calculation of Fourier maps, and the 5 th molecule of apo IPO was further built accordingly. Because of different space groups for the structures of the IPO-Me-Man, IPO-Me-Glc and IPO-Me-Gal complexes, the resolved monomeric apo IPO was used as a search template in the following molecular replacement method. The solutions of cross-rotation and translation could be obtained with 2 molecules for the IPO-Me-Man complex, 4 molecules for IPO-Me-Glc and 4 molecules for IPO-Me-Gal. Those solutions were further applied to initial rigid body refinement, and reasonable values were obtained (e.g., 36 .7% Rfactor for IPO-Me-Man, 35.4% for IPO-Me-Glc, and 32.9% for IPO-Me-Gal). Manual model rebuilding involved use of Coot [40] , alternating refinement by the CNS program, with 5% or 10% of the observed reflections randomly selected and set aside for calculation of the R free value. The final refined statistics are in Table 3 . For the protein interface of the tetrameric form, IPO-Me-Glc was used as a representative for analysis by the web service PDBe PISA [34] . All molecular representations were prepared with use of Deep-View [41] and PyMOL [42] . The coordinates of monomers of apo IPO (e.g., chain A) were subjected to the web service Matras for structure comparison [31] . The quantification of protein solution for binding assay was determined by the UV absorption method. The purified IPO protein was dialyzed against 20 mM Tris-HCl, 150 mM NaCl (pH 7.0) at 4uC overnight. The concentration of IPO and DN10IPO was determined by UV absorption spectroscopy at 280 nm with the specific extinction coefficient e of 22,920 M 21 cm 21 , which was determined from the prediction of IPO primary sequence. From Beer-Lambert law, A = e6b6C where A is the absorbance of the sample at 280 nm, b is the pathlength in 1 cm, and C is the protein concentration (M). The protein concentration C could be calculated from the equation. Carbohydrates were prepared by weighting the amount on a microbalance before dissolving in dialysis buffer (20 mM Tris-HCl, 150 mM NaCl pH 7.0). ITC measurements involved use of a MicroCal iTC200 microcalorimeter (GE Healthcare) at 25uC. In individual titration, 1-2 ml carbohydrate solution was added at 180-s intervals by use of a computer-controlled 40 ml syringe to a cell containing 280 ml IPO protein solution under constant stirring at 1,000 rpm. The concentration of IPO protein was 1-3 mM and that of Me-Man, Me-Glc, Me-Gal, Man, Glc and Gal 25-75 mM. The titration of carbohydrate solution in this range of concentration to the dialysis buffer was used as a control. Measurements of the heat change determined from the binding constant (K A ), reaction stoichiometry (n), and enthalpy (DH). The 18 experimental data were fitted for a 1:1 binding model (one-site of fitting) with Microcal Origin 7.0 software. Free energy (DG) and binding entropy (DS) were calculated by the equations DG = -RTlnK A and DG = DH -TDS. R is the gas constant and T the absolute temperature. The optimal c-value in ITC calculation varied between 1 and 10. However, for titrations with Man, Glc and Gal, the c-values were ,1. The atomic coordinates and structure factors of apo IPO and IPO-carbohydrate structures have been deposited in the RCSB Protein Data bank, with 3R50 for apo IPO, 3R51 for IPO-Me-Man complex, 3R52 for IPO-Me-Glc complex and 4DDN for IPO-Me-Gal complex.

Retroviral Env Glycoprotein Trafficking and Incorporation into Virions

Together with the Gag protein, the Env glycoprotein is a major retroviral structural protein and is essential for forming infectious virus particles. Env is synthesized, processed, and transported to certain microdomains at the plasma membrane and takes advantage of the same host machinery for its trafficking as that used by cellular glycoproteins. Incorporation of Env into progeny virions is probably mediated by the interaction between Env and Gag, in some cases with the additional involvement of certain host factors. Although several general models have been proposed to explain the incorporation of retroviral Env glycoproteins into virions, the actual mechanism for this process is still unclear, partly because structural data on the Env protein cytoplasmic tail is lacking. This paper presents the current understanding of the synthesis, trafficking, and virion incorporation of retroviral Env proteins.

All replication-competent retroviruses encode genes for three major proteins: Gag, Pol, and Env. Complex retroviruses, such as human immunodeficiency virus type 1 (HIV-1), encode additional regulatory and accessory proteins required for efficient replication in host cell or the infected host organism. Gag, an essential retroviral protein, is necessary and sufficient for the assembly, budding, and release of virus-like particles (VLPs) in all types of retroviruses except the spumaviruses. Gag is synthesized on cytosolic ribosomes and is assembled as a polyprotein precursor. During and/or shortly after budding and release, the polyprotein is cleaved into several domains by the viral protease ( Figure 1 ) as reviewed in [1] [2] [3] . The major domains of the precursor Gag are the matrix (MA), capsid (CA), and nucleocapsid (NC). The primary role of the N-terminal MA domain is targeting of the Gag precursor protein to the site of assembly, typically the plasma membrane (PM). In general, electrostatic interactions between basic amino acid residues in MA and the acidic inner leaflet of the PM are important for Gag-membrane targeting [4, 5] . In the case of HIV-1, the N-terminal myristate group and a cluster of basic residues in the MA domain of the HIV-1 Gag that interacts with phosphatidylinositol-4,5-bisphosphate (PI(4,5)P 2 ) together target the Gag precursor Pr55 Gag to the PM [6, 7] . Although the Gag-membrane targeting of both murine leukemia virus (MLV) and Mason-Pfizer monkey virus (MPMV) is also affected by PI(4,5)P 2 modulation [8, 9] , it has been reported that the membrane targeting of Rous sarcoma virus (RSV) and human T-lymphotropic virus type 1 (HTLV-1) is largely independent of PI(4,5)P 2 [10, 11] . The MA domain also plays a role in the incorporation of the Env glycoprotein into virions. The CA domain is important for Gag-Gag interactions during virus assembly and constitutes the outer part of the viral core after Gag processing by the viral protease [12] [13] [14] . NC is the primary nucleic acid binding domain of Gag. This small, basic domain is responsible for the binding and incorporation of the viral RNA genome into virions, which is mediated by Gag interactions with genomic RNA. Gag proteins are synthesized and transported to the PM. Many studies demonstrate that the major site of HIV-1 assembly is the PM [15] [16] [17] [18] , although late endosomes could be a platform for virus assembly under specific conditions [19] . In primary macrophages, HIV-1 has been shown to assemble in endosomal vesicles. However, studies have recently suggested that the above vesicles are not late endosomes but rather membrane invaginations connected to the PM [20] [21] [22] . In addition to Gag, the other major structural retroviral protein is the Env glycoprotein. Env proteins are required for virus entry into target cells and are thus essential for forming infectious retroviral particles. In this paper, we discuss current knowledge about the biosynthesis, intracellular trafficking, and virion incorporation of retroviral Env proteins, as well as the membrane microdomains involved in virus assembly and/or transfer. Most of this information was obtained from studies on HIV-1. Retroviral Env glycoproteins are synthesized from a spliced form of the viral genomic RNA as reviewed in [23] [24] [25] (Figure 1 ). Translation of the Env protein occurs on ribosomes bound to the endoplasmic reticulum (ER) and starts with the leader sequence, which contains a small, N-terminal hydrophobic signal peptide. The Env protein is cotranslationally inserted into the lumen of the rough ER. In the ER, the leader sequence is removed by cellular signal peptidases. In addition, Env polypeptides are N-and O-glycosylated and subsequently trimmed [26, 27] . The number and location of glycosylated residues varies broadly among retroviruses. The hydrophobic transmembrane (TM) domain prevents Env proteins from being fully released into the lumen of the ER [28, 29]. The amino acid sequence following the TM is referred to as the cytoplasmic tail (CT), which varies from 30 to around 150 residues, depending on the virus. Env proteins are folded and assembled into oligomers in the RER. Retroviral Env proteins form trimers [30] [31] [32] [33] . The HIV-1 accessory protein Vpu binds to the CD4 receptor through its cytoplasmic domain and downregulates the receptor by transporting it to the proteasome for degradation, thereby preventing premature interactions between Env and its receptor [34] [35] [36] . In the Golgi, cleavage of the retroviral Env precursor occurs at a polybasic (e.g., K/R-X-K/R-R) motif by cellular proteases such as furin or closely related enzymes probably within or near the trans-Golgi network (TGN) [37] [38] [39] [40] [41] [42] [43] . For HIV-1, the surface glycoprotein gp120 and the TM glycoprotein gp41, which bind together noncovalently, are both formed from the same precursor protein, gp160. Gp160 processing is essential for the activation of Env fusogenicity and virus infectivity [38, 42, [44] [45] [46] . Similarly, cleavage of Env is also essential for membrane fusion and virus infectivity in MLV [39, [47] [48] [49] [50] , in RSV [51, 52] , and in mouse mammary tumor virus (MMTV) [53] . A recent report showed that cleavage of MLV Env by furin also plays an important role in Env intracellular trafficking and incorporation [54] . Although most retroviral Env proteins including that of HIV-1 are associated with intracellular membranes [55] [56] [57] , at least part of the gp120/gp41 trimer complex traffics through the secretory pathway to the PM. It has been suggested that AP-1, one of adaptor proteins for clathrin-coated vesicle formation, is involved in the correct sorting of HIV-1 Env from the TGN to the PM, [58, 59] . It has been reported that intracellular CTLA-4-containing secretory granules are involved in the trafficking of HIV-1 Env to the PM although the subsequent trafficking of Env after the Golgi is not well understood [60] . After reaching the PM, like those of other lentiviruses, HIV-1 Env undergoes rapid endocytosis, which is mediated by the interaction between the μ2 subunit of the clathrin adaptor AP-2 and a membrane-proximal, Tyr-based motif (YxxL) in the gp41 CT [58, 61, 62]. Although some of the endocytosed Env is recycled back to the PM, most retroviral Env is associated with intracellular membranes [63, 64] . The level of gp120-gp41 oligomers on HIV-1 virions is relatively low [33] . Maintaining low levels of Env at the cell surface allows the infected cells to evade the host immune response and to avoid induction of Env-mediated apoptosis. Gammaretroviruses such as MLV and MPMV also have dileucine-and Tyr-based motifs in their Env CT. These motifs are important to regulate intracellular trafficking of Env of both retroviruses via interactions with clathrin adaptors [65, 66]. As for pseudotyping of gammaretroviruses, it has been reported that the feline endogenous retrovirus RD114 Env does not allow pseudotyping with viral cores from lentiviruses such as SIV, whereas the RD114 Env is incorporated into MLV virions [67-69]. Intracellular trafficking of Gag and Env was examined using a set of chimeric viruses between MLV and RD114 [57]. Interestingly, it was found that the RD114 Env was mainly localized along the secretory pathway, whereas the MLV Env was mostly localized in endosomes, and that intracellular localization was dependent on specific motifs in the Env CT [57]. In addition, subsequent work revealed that an acidic cluster in the RD114 Env CT regulates assembly of not only the RD114 Env but also the MLV Env through the interaction with a host factor, phosphofurin acidic-cluster-sorting protein 1 [66]. Several models have been proposed for the incorporation of retroviral Env glycoprotein into virions as reviewed in [23, 70] (Figure 2 ). Passive incorporation is the simplest model for the incorporation of Env proteins into virus particles (Figure 2(a) ). There are several lines of evidence supporting this model. First, viral pseudotyping with a foreign glycoprotein can occur easily in many cases although there are some exceptions, one of which is the exclusion of HIV-1 or SIV Env with the long CT from most retrovirus cores [70] . With respect to HIV-1, the virus can be pseudotyped with Env glycoproteins not only from several other retroviruses but also with those from other virus families such as ortho (para) myxoviruses and flaviviruses [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] . Second, retroviruses allow passive incorporation of host membrane proteins into virus particles [85] [86] [87] . Most cellular proteins are incorporated into the retrovirus envelope without significant sorting [88, 89] . Finally, in the case of HIV-1, several studies have demonstrated that the gp41 CT can be removed without affecting incorporation of the Env into virions, although this has been shown to occur only for some laboratory cell lines such as HeLa or 293T [90] [91] [92] [93] [94] . Although several lines of evidence support the passive incorporation model for retroviral Env, there is much evidence indicating that Env incorporation into virions is regulated by direct interactions between Gag and Env proteins (Figure 2(b) ). Although removal of the gp41 CT sequence of HIV-1 has little effect on Env incorporation in some cell types, as described above, smaller deletions in CT regions cause severe defects in Env incorporation [95] [96] [97] [98] [99] [100] . The MA domain of Gag has been shown to be important for Env incorporation into virions [91, 92, 101, 102] . The defect in Env incorporation caused by deletion of the gp41 CT is reversed by several MA mutations, indicating that an interaction between Env and the MA domain of Gag is required for incorporation of full-length Env into virions, at least in the case of HIV-1 [93, 98] . More evidence for direct Gag-Env interaction comes from the finding that HIV-1 Env directs Gag budding to the basolateral surface of polarized epithelial Madin-Darby canine kidney (MDCK) cells through the CT of HIV-1 Env, whereas Gag alone buds in a nonpolarized fashion [103] [104] [105] [106] . The Tyr-based motif in the gp41 CT is also utilized in polarized budding of HIV-1 in lymphocytes [107] . Surprisingly, the polarized budding of HIV-1 in MDCK cells could also be promoted by MLV and HTLV-1 Env through their CT [108] . It also has been reported that coexpression of Pr55 Gag inhibits endocytosis of HIV-1 Env through its interaction with the gp41 CT [63]. Another example of the specific Gag-Env interactions was demonstrated using Gag and Env proteins of MLV and HIV-1 in rat neurons [109] . Similarly, MLV Env is preferentially recruited onto MLV Gag through its CT domain in the presence of both MLV and HIV-1 cores although the authors also show an alternative mechanism by which the recruitment to HIV-1 budding sites is independent of the CT domain of MLV Env [110] . Furthermore, RSV Env is exclusively recruited to RSV budding sites through its CT, suggesting that the interaction between Env and Gag is direct in the case of this avian retrovirus [111] . In addition to the circumstantial evidence discussed above, some biochemical data suggest a direct interaction between Gag and Env. In vitro binding between MA and a gp41 CT-GST fusion protein has been reported for both HIV-1 and SIV [112, 113] . Peptides corresponding to a large central domain of gp41 CT inhibited the capture of membrane-free Pr55 Gag with an anti-p24 antibody [114] . In addition, a stable, detergent-resistant gp41-Pr55 Gag interaction was detected in immature HIV-1 virions. The retention of gp41 in detergent-treated virions is dependent on the CT region, suggesting a direct or indirect interaction between Pr55 Gag and gp41 [115, 116] . In the third model, it is assumed that host cellular factors (mostly proteins) play a role in bridging Gag and Env in virus-infected cells (Figure 2(c) ). Several host factors have been reported to bind to Gag and/or Env of HIV-1 or SIV however, only a couple of host factors were shown to be required for Env incorporation and/or viral replication. The 47-kDa tail-interacting protein (TIP47) has been reported to bridge Gag and Env, allowing efficient Env incorporation in HIV-1 [117, 118] . The same group also showed that both the WE motif near the N-terminus of the MA domain and the YW motif in the gp41 CT domain are important for interactions between Gag or Env and TIP47 [118] . In a subsequent paper, the same group showed that mutations in either the WE motif of MA or the YW motif in the gp41 CT caused defects in virus replication in primary monocyte-derived macrophages [119] . Although this finding of an important role for TIP47 in Env incorporation in HIV-1 has received much attention from retrovirologists, no confirmatory data have been published by other researchers in this field. Human discs large protein (hDlg1) has been reported to interact with the CT of HTLV-1 Env and to colocalize with both Env and Gag in virus-infected cells [120] . Subsequent work demonstrated that Dlg1 also binds HIV-1 Gag and that the expression level of Dlg1 is inversely correlated with HIV-1 Env expression and incorporation levels of the Env proteins, although the mechanism behind this phenomena needs to be investigated [121] . Prenylated Rab acceptor 1 (PRA1), which was identified as a Rab regulatory protein, was reported to be a binding partner for the SIV gp41 CT in a mammalian yeast twohybrid (Y2H) assay [122] . Although colocalization of PRA1 and SIV Env was observed, changes in the endogenous levels of PRA1 did not affect virus production, Env incorporation, or infectivity of SIV or HIV-1 [123] . A Prohibitin 1/Prohibitin 2 (Phb1/Phb2) heterodimer was identified as a binding partner of the gp41 CT of HIV-1 using human T-cell lines and tandem affinity chromatography [124] . Phb1 and Phb2 are members of the prohibitin superfamily of proteins, which are localized to several cellular compartments such as the mitochondria, nucleus, and the PM [125, 126] . Gp41 CT mutants, in which binding to Phb1/Phb2 is disrupted, could replicate well in permissive cell types such as MT-4, but could not replicate efficiently in nonpermissive H9 cells [124] . Further analysis is necessary to elucidate the mechanism by which these proteins regulate virus replication through interactions with Env. Luman, a transcription factor that is mainly localized to the ER, was found to interact with the gp41 CT of HIV-1 in a Y2H screen using a cDNA library from human peripheral blood lymphocytes (PBL) [127] . Overexpression of a constitutively active form of this protein reduced the intracellular levels of Gag and Env, leading to a decrease in virus release. The mechanism for this negative effect on virus assembly involves Luman binding to Tat, which decreases Tat-medicated transcription [127] . By using a Y2H screen with human cDNA libraries, p115-RhoGEF, an activator of Rho GTPase, was found to interact with the gp41 CT through its C-terminal regulatory domain [128] . The gp41 mutants that lost the ability to bind p115 showed impaired replication kinetics in T-cell lines such as SupT1, H9, and Jurkat, suggesting that the gp41 CT could modulate the activity of p115-RhoGEF to support virus replication [128] . In addition to the host factors described above, calmodulin [129] [130] [131] [132] and α-catenin [133] [134] [135] have been reported to interact with HIV-1 and/or SIV. However, their roles in virus replication, especially with respect to the Env functions of both proteins, have not been clearly elucidated. Regardless of whether direct or indirect interactions between retroviral Gag and Env proteins are required for Env incorporation into virions, a great deal of experimental evidence suggests that retroviruses assemble and bud from "membrane microdomains." The most well-known microdomains are "lipid raft(s)," which are enriched in cholesterol and sphingolipids [136, 137] . Lipid rafts are widely thought to function as a platform for the assembly of protein complexes and to allow various biological processes such as cellular transport and signal transduction to proceed efficiently [138, 139] . Lipid rafts are reportedly used as assembly platforms or entry scaffolds in the replication of enveloped viruses such as retroviruses [140] [141] [142] [143] [144] [145] [146] . The association of Gag/Env with lipid rafts is important for the regulation of Env incorporation and pseudotyping [143, 144, 147, 148] . Evidence that both the HIV-1 Pr55 Gag and Env proteins are preferentially localized to lipid rafts comes from biochemical studies as well as direct observations by microscopy [142, 149, 150] . Another membrane microdomain for retrovirus assembly is the "tetraspanin-enriched microdomain (TEM)" [151] [152] [153] [154] . Tetraspanins are a superfamily of cell surface proteins that are ubiquitously expressed in mammalian cells. TEMs also act as platforms for signal transduction and immune responses. TEMs have been reported to be involved in the assembly and release of not only HIV-1, but also HTLV-1 and HCV [155] . When both HIV-1 and influenza virus were produced in the same cell, only HIV-1 colocalized with the TEM marker, and its release was inhibited by an anti-CD9 Ab, which led to extensive aggregation of tetraspanins [156] . Analysis of dynamics of both lipid rafts and TEMs by quantitative microscopy has revealed that components of both lipid rafts and TEMs are recruited during viral assembly to create a new microdomain that is different from preexisting membrane microdomains [153, 157] . There have been three recent reports in which both pseudotyping and microdomain issues were discussed. In the first paper, the authors examined HIV-1 assembly under conditions where the Env proteins of HIV-1 and Ebola virus were coexpressed with HIV-1 Gag in the same cell [158] . They found that infectious HIV-1 virions were released with both types of Env proteins. Interestingly, however, the virions contained either HIV-1 Env or Ebola virus glycoprotein (GP), but not both Env proteins within a single virion. These results suggest that HIV-1 Env and Ebola virus GP localized to distinct microdomains on the surface of the same cell [158] . In the second paper, the subcellular localization of Gag and Env proteins was investigated using a combination of three different retroviral Env proteins (RSV Env, MLV Env, or vesicular stomatitis virus (VSV) G) and two different Gag proteins (RSV or HIV-1) [111] . Both VSV-G and MLV Env were redistributed to the virus budding sites when coexpressed with HIV-1 or RSV Gag. In contrast, RSV Env was mostly transported to RSV budding sites. A subsequent paper from the same group showed that the CT of MLV is not required for recruitment of MLV Env to HIV-1 budding sites, suggesting that there are no specific interactions between MLV Env and HIV-1 Gag [110] . Collectively, these results also suggest that retroviral Env glycoproteins are not recruited to preexisting membrane platforms but rather that they are actively recruited to newly formed microdomains on the cell surface [111] . Human retroviruses such as HIV-1 and HTLV-1 spread more efficiently between target T cells by cell-cell infection than by cell-free infection [159, 160] . Sattentau et al. proposed, in analogy to the "immunological synapse", the "virological synapse (VS)" as a point of contact between virusinfected cells and uninfected cells [161, 162] . The molecular mechanisms of retroviral VS formation are as follows. (1) With respect to HIV-1 T-cell VS, initial contact between virus-infected cells and uninfected cells occurs through gp120-CD4 binding. Subsequent interactions between integrins and ICAMs enforce and maintain the stability of these junctions. (2) The gp120-CD4 interaction recruits CD4, coreceptors such as CXCR or CCR5, adhesion molecules, and filamentous actin into the synaptic area. (3) The cellular secretory machinery and microtubule organizing centers (MTOC) are polarized towards the HIV-1 assembly sites at the PM to form the VS. It has been reported that a socalled microsynapse formed by nanotubes between virusinfected cells and uninfected cells is also involved in cell-cell infection of HIV-1 [84, 163] . In cell-cell transfer of HTLV-1-infected cells, an extracellular matrix structure referred to as the "viral biofilm" was proposed as an alternative to the VS [164] . In addition to HIV-1 and HTLV-1, the spread of MLV between fibroblasts also occurs via the VS [165, 166] . It is noteworthy that assembly of MLV is directed towards cellcell contact sites through the interaction of the CT of MLV Env with Gag [167, 168] . Although the concept of cell-cell infection through the VS is now well appreciated, the detailed molecular mechanism of VS assembly and its relevance to viral spread in vivo will require further elucidation through the use of more advanced techniques. Incorporation of Env glycoproteins into virions is crucial for producing infectious retroviral particles. Although this paper has introduced several experimental models for retroviral Env trafficking and/or incorporation, the correct mechanism for this process is still unclear. The following questions must be clearly addressed to not only gain a better understanding of this complex biological process, but also to develop new antiretroviral compounds that target Env incorporation. (1) What are the structures of the CTs of retroviral Env proteins? The answers for this question will give useful information on elucidating a role of the Env CTs in the Env trafficking and/or incorporation in virus-infected cells. (2) What host factor(s) are necessary for the retroviral Env trafficking and/or incorporation into virions? (3) Where and how Env and Gag proteins of retroviruses are recruited to the assembly sites in order to form infectious virus particles? [

Production, Characterization and Applications for Toxoplasma gondii-Specific Polyclonal Chicken Egg Yolk Immunoglobulins

BACKGROUND: Toxoplasma gondii may cause abortions, ocular and neurological disorders in warm-blood hosts. Immunized mammals are a wide source of hyperimmune sera used in different approaches, including diagnosis and the study of host-parasite interactions. Unfortunately, mammalian antibodies present limitations for its production, such as the necessity for animal bleeding, low yield, interference with rheumatoid factor, complement activation and affinity to Fc mammalian receptors. IgY antibodies avoid those limitations; therefore they could be an alternative to be applied in T. gondii model. METHODOLOGY/PRINCIPAL FINDINGS: In this study we immunized hens with soluble tachyzoite antigens of T. gondii (STAg) and purified egg yolk antibodies (IgY) by an inexpensive and simple method, with high yield and purity degree. IgY anti-STAg antibodies presented high avidity and were able to recognize a broad range of parasite antigens, although some marked differences were observed in reactivity profile between antibodies produced in immunized hens and mice. Interestingly, IgY antibodies against Neospora caninum and Eimeria spp. did not react to STAg. We also show that IgY antibodies were suitable to detect T. gondii forms in paraffin-embedded sections and culture cell monolayers. CONCLUSIONS/SIGNIFICANCE: Due to its cost-effectiveness, high production yield and varied range of possible applications, polyclonal IgY antibodies are useful tools for studies involving T. gondii.

Toxoplasma gondii is a well-adapted parasite of warm-blooded hosts, including human and production animals. The parasite has distinct developmental stages, such as the fast-replicating tachyzoites that are present in the acute phase of infection, while the slow-replicating bradyzoites form tissue cysts in muscular and nervous tissues during the chronic phase of infection [1] [2] . Infection usually takes place after accidental ingestion of raw or undercooked meat containing tissue cysts, oocyst-contaminated food or water, and by transplacental passage of tachyzoites in active infection during the gestational period [3] [4] . Congenital infection may lead to neurological disorders, ocular disease and/or fetal death [5] . T. gondii is also a major cause of infectious abortion and induces clinical disease in production animals, especially goat, sheep and swine, elevating production costs and resulting in considerable impact on the livestock industry as well as the public health [6] . Immunized hens transfer circulating antibodies from blood to egg yolk by specific receptors on the surface of the ovarian follicle [7] . Immunized hens may continue to produce antibody-rich eggs for up to two years -also known as IgY, which shares great similarity to mammalian IgG, and is a reasonably stable bivalent protein with estimated molecular weight around 180 kDa [7] [8] . Egg yolk antibodies have been shown to be effective in immunohistochemical assays and for recognition of specific epitopes [9] . Polyclonal IgY have been previously described for the purification of polyepitope vaccine candidates against Plasmodium falciparum [10] , passive immunization against Eimeria acervulina [11] , immunoprophylactic protocols against influenza virus [12] , and analysis of tumoral biomarkers [13] , among others. Specific polyclonal antibodies generated in mammals against T. gondii have been extremely useful in the last decades to unravel several aspects related to the parasite biology as host-parasite interactions, parasite invasion and replication mechanisms, candidate epitopes for diagnosis, characterization of T. gondii virulence factors and candidate protein for immunoprophylactic assays [14] [15] [16] [17] . Here, we hypothesized whether polyclonal IgY could be a useful tool in the T. gondii model. In that sense, we produced, characterized and standardized distinct protocols for IgY application, using polyclonal IgY antibodies raised against T. gondii soluble tachyzoite antigen (STAg). The first step consisted of the purification of IgY antibodies from egg yolks in order to obtain the highest yield and purity. Antibodies were separated from other constituents of the egg yolk by the addition of acid water and later precipitated by salting-out (19% sodium sulfate) (Fig. 1A) . Although enriched in IgY antibodies, the water soluble fraction (S1) still presented undesirable low molecular weight proteins after acid water precipitation of pooled egg yolk (Fig. 1B , lane S1) that were eliminated after salting-out step (Fig. 1B, lane P2) , presenting an average protein concentration of 4 mg/mL. We recovered 115 eggs from four STAg-immunized chickens during a 49-day interval. Each egg presented an average of 12 mL of yolk and the mean yield was 48 mg of polyclonal antibody per egg. Precipitated antibodies contained in P2 fraction were further submitted to purification by exclusion-size chromatography (Fig. 1C ), resulting in a protein peak between the 13th and 19th fractions, although elevated purity was detected mainly in the 14th fraction (F14), as it may be observed in its SDS-PAGE profile (Fig. 1D) . Thus, the used protocol was able to retrieve a mean of 4 mg of total IgY from 1 ml of crude egg yolk. Slot-blot assays confirmed IgY in fractions from salting-out and gel filtration purified fractions until the concentration of 110 ng of protein (Fig. 1E ). After purification of an enriched IgY fraction, kinetics and avidity of the purified antibodies were evaluated by indirect ELISA. We verified that immunized hens presented detectable levels of specific IgY starting on day 21 after the first immunization (p.i.). The production of IgY anti-STAg increased significantly along the next weeks, and developed a progressive and sustained response, reaching its peak on day 42 p.i. ( Fig. 2A) . High avidity of IgY anti-STAg antibodies was already observed in the first reactive samples, at 21 days p.i., as demonstrated by the lack of significant differences in EI values between 6 M urea-treated and untreated samples (Fig. 2B) . The antigen recognition repertoire and avidity maturation of IgY anti-STAg were also verified by immunoblot. Similar to ELISA, reactivity was first observed on day 21 after immunization, with strong staining of a 30 kDa protein (Fig. 2C) . From the 35 th day onwards, IgY anti-STAg antibodies recognized a broad range of T. gondii antigens. The 6 M urea treatment presented a reduction in staining intensity of the described antigens between the 21 st and 35 th days p.i., with normal reactivity being observed after that period (Fig. 2C) . Mice immunized by the same route present a similar antibody seroconversion kinetics, with reactivity to STAg being detected on 14 th day p.i., preceded by initial p30 recognition on 7 th day p.i. in serum samples (Fig. S1 ). Once we were able to obtain pure, high avidity IgY anti-STAg antibodies, we aimed to verify potential applications of those antibodies in T. gondii model. The standardized immunohistochemical assay using IgY anti-STAg antibodies was able to detect antigenic distinct structures in brain tissue sections, as follows: (i) parasitophorous vacuoles, demonstrated as rounded and compartmentalized structures (Figs. 3A and 3B); (ii) tissue cysts, represented by a wall bright fluorescence (Figs. 3C and 3D); and (iii) free tachyzoites (Fig. 3C) . In cell culture experiments, the antibodies were used in immunocytochemical assays, where monolayers of HeLa cells infected with T. gondii were incubated with IgY anti-STAg. In these experiments, we observed isolated or grouped intracellular tachyzoites closely to cell nucleus, detected by an intense fluorescence on the whole tachyzoite surface (Fig. 4) . Overlay of fluorescence emitted by the stained parasites and cell nucleus with phase contrast images of the fibroblast monolayer provided a valuable measure of the parasitism of the infected monolayers. We next observed the repertoire of the polyclonal antibodies raised in avian and mammalian models. Although p30 was evenly recognized by antibodies from the distinct species by onedimensional immunoblotts (WB1D), differential protein recognition pattern was noted (Fig. 5A ). Only chicken-derived antibodies recognized proteins with approximate molecular weights of 40 kDa, while IgG anti-STAg from mice reacted strongly to ,50 kDa antigen. These differences were maintained independently of the route used to immunize the mice (i.m. or s.c., Fig. 5A ). In order to further observe those differences in recognition, we performed 2D-immunoblots (WB2D) using the polyclonal antibodies obtained from immunized hens and mice. The differences in recognition were clearly noted in WB2D assay, with IgY antibodies recognizing a sequence of five antigenic spots at ,40 kDa, however with distinct isoelectric points -from neutral to basic pH (Fig. 5B ). On the other hand, mouse IgG reacted strongly to two acidic antigens at ,50 kDa ( Fig. 5C ), while IgY reacted faintly to the same spots and a basic protein with similar molecular weight (Fig. 5B ). Such differences were also noted in the distinct recognition patterns in antigens with 60-80 kDa, as well as the absence of immunodominant p22 antigen by chicken IgY. Additionally, STAg immobilized in nitrocellulose strips were probed to anti-Neospora caninum e anti-Eimeria spp. IgY in order to check the specificity of the antibodies. The assay revealed that IgY antibodies against the closely related protozoa did not cross-react with Toxoplasma antigens ( Figure 5D ). The use of specific antibodies produced in mammals against antigenic targets and cell markers are essential tools for the study of host-parasite relationships. Unfortunately, these antibodies may be painstaking to obtain, and depend on animal bleeding in order to retrieve the desired immunoglobulins [18] . Chickens present antibody production kinetics and avidity maturation similar to mammals, and IgY antibodies appear in serum approximately 4 to 7 days after inoculation of antigen [9] . In the present study, we produced and purified polyclonal anti-T. gondii antibodies from STAg-immunized hens. These anti-STAg antibodies were obtained from the egg yolk, in a protocol which animal bleeding turned out to be unnecessary [19] . Generally, yolk IgY arise around 7 th to 10 th day after the sera antibodies, in a sera concentration-dependent manner [20] . In addition, the large amounts of polyclonal antibodies recovered from the egg yolk in this process, the high avidity of the produced IgY antibodies, and their potential applications are itself the great features of the protocol herein described. Egg yolks contain large amounts of IgY passively transferred by hens, with mass ranging between 2-8 mg per mL after purification, and it has been estimated that 2-10% of total egg yolk antibodies are antigen-specific [21, 9] . To obtain STAg-specific IgY we used simple protocols, which are easily reproduced without the need of refined equipment or reagents. Additionally, installations required for IgY production are smaller and easier to handle than those designed for small ruminants, commonly used for large scale antisera production [22] . The production of specific IgY has been previously described for the recognition of a broad range of targets, including Plasmodium falciparum [23] , hepatitis A virus [24] and tumor cells [25] [26] . Avidity maturation of an antibody represents an increasing affinity interaction between antibodies and antigens. However, this maturation is slowly induced in mammals [27] . We demonstrated that STAg-immunized hens produced large amounts of high avidity polyclonal IgY anti-STAg antibodies after the first booster, with early recognition of p30 proteins. It is known that SAG1 protein (p30) triggers the human IgG maturation in latter stages of T. gondii infection [28] . Additionally, IgY antibodies recognized a broad range of proteins, with distinct isoelectric points determined by 2D-immunoblot. Previous studies have shown that serum samples from naturally infected humans react to a broad spectrum of T. gondii proteins [29] and experimentally infected mice produce specific antibodies against high and low molecular weight antigens of tachyzoites [30] . Consistently, our results demonstrated that IgY anti-STAg presented an antigen recognition profile similar to mouse serum, although some antigens were more immunogenic for hens than mice, probably due to the phylogenetic distance among these animals. In addition, IgY antibodies raised against closely related parasites did not recognize STAg proteins, which denotes an interesting feature of the IgY system, since crossreactivity is a normal feature of mammalian IgG. Hassl and colleagues (1987) [31] , in a preliminary study, demonstrated some of characteristics of anti-T. gondii IgY and also compared protocols for antibody production and purification, however it was not demonstrated the possible applications of these antibodies to study the parasite. We herein demonstrated for the first time distinct applications for specific IgY antibodies in the T. gondii model. As previously stated, this antibody producing system offers high yield of specific antibodies [32] . We believe that immunization of hens with selected T. gondii targets could improve the range of applications for IgY antibodies. For the P. falciparum model, it has been demonstrated that monospecific and polyclonal IgY antibodies were advantageous [10] . In that sense, immunizations using single antigens of T. gondii could be employed to better understand the role of the selected molecules, as tools to improve the diagnosis of acute toxoplasmosis [33] or the knowledge of hostparasite interaction mechanisms [34] . Additionally, the identification of conserved surface antigens among apicomplexan organisms is usually performed by mammal specific IgG antibodies and contributes to our understanding of parasite evolution [35] [36] [37] . As chickens are phylogenetically distant from mammals and have been shown to be refractory to T. gondii infection [38] , and differential recognition profile by antibodies from both classes of animals has already been described [39] . The comparative analysis of antigen recognition may be a useful tool to determine different virulence factors of the parasite, thorough proteomic studies for vaccine or diagnostic development. Moreover, the role of T. gondii antigens involved in parasite virulence has been largely addressed over the last decade [40] [41] . The most common protocols used for that purpose are the incubation of the parasite with specific antisera or the generation of strains with targeted depletions, preventing molecular interactions between host and parasite proteins [16] . As shown in this study, IgY antibodies may also be used in this context and further studies are necessary to determine whether these specific antibodies could be employed in therapeutical protocols during acute toxoplasmosis, especially in immunizations performed with single antigens. In conclusion, our results demonstrated that polyclonal IgY anti-STAg antibodies are a promising complementary tool for studies of the T. gondii infection model. Laying hens of the Isa Brown lineage, 25 weeks of age, were used for immunization protocols. Hens were kept in individual cages and received commercial rations and water ad libitum. C57BL/6 mice, 6-8 weeks old, were infected with T. gondii ME-49 strain to obtain parasite positive brain tissue sections [42] and immunized with STAg to obtain polyclonal IgG antibodies anti-T. gondii [43] . All animal procedures were approved by the institutional ethics committee in animal experimentation (Comissão de É tica no Uso de Animais da Universidade Federal de Uberlândia -Protocol No. 107/11), and were performed based on the Ethical Principles in Animal Research adopted by the Brazilian College of Animal Experimentation and instructions disposed in the 2000 Report of the AVMA Panel on Euthanasia [44] . Soluble tachyzoite antigen (STAg) was obtained using a previously described protocol [45] . Briefly, tachyzoites of T. gondii (RH strain) were maintained by serial passage in BALB/c mice and obtained from peritoneal exudates from mice infected 48 h earlier. Only exudates containing 100% of free-ranging tachyzoites were used for antigen preparation. Parasite suspension was washed in phosphate buffered saline (PBS, pH 7.2), added to a protease inhibitor cocktail (Complete Ultra tablets, Roche, USA), processed through lysis by repeated freezing and thawing cycles, sonicated, and centrifuged at 14,0006g for 30 min at 4uC. After supernatant recovery, total protein was estimated by the Bicinchoninic acid kit (BCA, Sigma, St. Louis, MO, USA) and aliquots were stored at 280uC until use. The same protocol was performed with cells extracted from the peritoneal cavity of mice, used as experimental controls. Hens (n = 4) were immunized by muscular route with emulsion composed of STAg and Freund's adjuvant, according to previously described protocol [46] . Primary immunization was performed with 100 mg of STAg in 250 mL of PBS and equal volume of Freund's complete adjuvant (Sigma). Two boosters were performed at 15 day intervals, with 100 mg of STAg plus Freund's incomplete adjuvant. In parallel, Neospora caninum (n = 4) and PBS (n = 2) -immunized hens were maintained for parasite-specific and Figure 1 . Purification of egg yolk antibodies. (A) Separation of the water soluble fraction (S1) from a lipid-rich precipitate (P1), after the incubation of crude egg yolk with acid water (pH 5.0-5.2); S1 precipitation by salting-out (19% Na 2 SO 4 ) produced an enriched IgY pellet (P2) and a supernatant with contaminants (S2). (B): Purity degree of IgY samples determined by SDS-PAGE (8%), demonstrating that salting-out protocol produced high purity IgY antibodies. (C) Size-exclusion chromatography of the P2 fraction, with peak of IgY between 13th and 19th fractions, where (D) the 14th fraction presented the highest degree of purity. (E): IgY enrichment conferred by Slot-blot assay, IgY was detected until the concentration of 0.01 mg of protein (box). Bovine sera albumin (BSA) was used as negative control. doi:10.1371/journal.pone.0040391.g001 irrelevant IgY purification, respectively. Hens were monitored daily for adverse effects and individual laid eggs were daily collected and stored at 4uC until further processing. In addition, C57BL/6 mice were immunized by intramuscular and subcutaneous routes with STAg, for comparative purposes. Primary immunization was performed with 25 mg of STAg in 50 mL of PBS and equal volume of Freund's complete adjuvant (Sigma). Two boosters were performed at 15 day intervals, with 25 mg of STAg with Freund's incomplete adjuvant. Serum samples were collected weekly and stored at 220uC until use. IgY was purified by the water-dilution method as previously described [19] . In order to obtain a representative sampling of antibody production along the weeks after immunization, eggs laid weekly from each hen were pooled prior to IgY extraction. Then, egg white was removed and egg yolk was diluted 10-fold in deionized ultrapure water adjusted to pH 5.0-5.2 with 1N HCl and homogenized thoroughly. After centrifugation at 10.0006g for 25 min at 4uC, the supernatant was collected, consisting of lipid-free fraction (S1). S1 was then precipitated with the addition of 19% sodium sulphate (w/v). After centrifugation (10.0006g, 25 min, 4uC), the pellet was retrieved and represented the IgYenriched fraction (P2). P2 samples were resuspended and dialyzed against PBS to eliminate residual salt. Additionally, P2 samples were submitted to exclusion-size gel chromatography using Sephacryl S-300 column (GE Healthcare, Uppsala, Sweden), at a flow rate of 3 mL/min, and the IgY-enriched protein fraction was determined by 280 nm readings. Actual protein concentration was measured by BCA kit (Sigma) and samples were stored at 220uC until use. This protocol was also applied to extract IgY antibodies from the eggs of hens immunized in house with N. caninum soluble antigen, as well as from commercially available egg yolk powder containing anti-Eimeria spp.(E. acervulia, E. maxima e E. tenella) IgY antibodies (Supracox, Investigación Aplicada, Sociedad Anónima de Capital Variable, Puebla, Mexico). Possible cross- Figure 3 . Immunohistochemistry for T. gondii detection using polyclonal IgY anti-STAg antibodies. Paraffin-embedded brain sections of mice chronically infected with ME-49 strain were incubated with IgY anti-STAg and rabbit IgG anti-IgY conjugated to fluorescein isothiocyanate (FITC). (A and B) , segmented parasitophorus vacuoles were detected into host cell cytoplasm around the DAPI-stained nucleus (blue). (C and D), IgY anti-STAg antibodies strongly recognized antigens on outer walls of tissue cysts and free tachyzoites (C, arrow). doi:10.1371/journal.pone.0040391.g003 reactivity of the purified IgY antibodies to mouse proteins were ruled out by immunoblots (Fig. S2) . The quality of the purification protocols was analyzed in polyacrylamide gel electrophoresis with sodium dodecyl sulphate (SDS-PAGE) at 8% in non-reducing conditions. IgY enrichment was assessed by slot dot assays, as follows: proteins samples obtained during IgY extraction were sequentially diluted 10-fold (10 mg to 10 24 mg) and transferred to nitrocellulose membrane by a vacuum apparatus (Bio-Dot SF; Bio-Rad, EUA). The presence of detectable chicken antibodies was verified by an rabbit anti-IgY antibody conjugated to peroxidase (Sigma) as secondary antibody, revealed by H 2 O 2 and DAB (Sigma). The kinetics of chicken IgY anti-STAg production and avidity maturation were evaluated by an indirect ELISA. The optimal conditions for ELISA were obtained through block titration of the reagents. Briefly, high affinity microtiter plates (Costar Corning Incorporated, USA) were coated with STAg (10 mg/mL) in 0.06 M carbonate buffer (pH 9.6) and incubated overnight at 4uC. Plates were washed 3 times with PBS-Tween 0.05% (PBS-T) and blocked with PBS-T plus 1% bovine serum albumin (PBS-T-BSA) for 1 h at room temperature. P2 samples were adjusted to 2 mg/well in PBS-T-BSA, added to the wells in duplicate and incubated for 1 h at 37uC. After washing, plates were incubated with rabbit anti-IgY antibody labeled with peroxidase, diluted 1:30.000 in PBS-T, for 1 h at 37uC. The reaction was revealed by adding 0.01 M 2,29-azino-bis(3-ethylbenzthiazoline-6-sulphonic acid) (ABTS, Sigma) and 0.03% H 2 O 2 , and optical density (OD) was determined at 405 nm. Two positive quality-controls and six negative controls (irrelevant IgY) were included in each plate in order to calculate the cut off, which was established as the mean OD values for irrelevant IgY plus three standard deviations. The same protocol was applied to mouse IgG anti-STAg obtained after immunization, using a goat IgG anti-mouse IgG conjugated to peroxidase (Sigma), at 1:2000 dilution. Results were expressed as ELISA index (EI) as previously proposed [47] as follows: EI = OD sample /OD cut off , where values of EI .1.2 were considered positive. To measure avidity maturation, plates were coated and blocked as described above. After the addition of P2 samples in quadruplicates, two of the wells were rinsed with 6 M urea in PBS-T, whereas the other duplicate wells were rinsed with PBS-T only for 10 min at room temperature. Next, all wells were washed 3 times in PBS-T and subsequent steps were performed as described for the regular indirect ELISA. To investigate the antigen recognition repertoire of IgY anti-STAg antibodies, immunoblot assays were carried out along with avidity maturation analysis against STAg. In 1D-immunoblot, STAg was separated on 12% SDS-PAGE under non-reducing conditions, and electrotransferred to nitrocellulose membranes. Non-specific interactions were blocked by 5% skim milk-PBS-T incubation, for 2 h at room temperature. Nitrocellulose strips were then incubated with P2 samples from weekly egg yolk pools adjusted to 2 mg of IgY. In parallel, avidity maturation was measured by strip treatment with 6 M urea for 10 min at room temperature. IgY was detected by incubating the secondary antibody, rabbit anti-IgY labelled with peroxidase, diluted at 1:20.000, for 2 h at room temperature. Reaction was revealed by adding 10 mg of 3,39-diaminobenzidine tetrahydrochloride (DAB, Sigma) in 15 mL of Tris buffered saline (TBS) and 12 mL of 30% hydrogen peroxide (Sigma). Reaction was stopped with distilled water. Experiments with similar conditions were performed for IgY anti-N. caninum and anti-Eimeria spp., as well as for the establishment of recognition kinetics of mouse IgG obtained from mice immunized by i.m. and s.c. routes (1:100 sera dilution; 1:2000 anti-mouse IgG conjugated to peroxidase -Sigma). A 2D-immunoblot assay was also carried out to evaluate the antigen recognition profile of immunized hens to STAg. Briefly, 60 mg of STAg dialysed in ultrapure water, was separated by isoelectric focusing (IEF) on 7-cm immobilized pH gradient strips (ReadyStrip TM IPG Strip pH 3-10) overnight at room temperature, following the manufacturer instructions for equipment and reagents (GE, Healthcare, Uppsala, Sweden). After IEF, strips were equilibrated and loaded onto precast 12% polyacrylamide gels. Electrophoresis was performed and 2D-gels were stained with Coomassie brilliant blue G-250H (Sigma) or electrotransferred to nitrocellulose membranes. 2D-immunoblot was performed as described above for 1D-immunoblot. In parallel, to compare the STAg recognition profile between mammalian and avian hosts, immobilized STAg membranes were probed with sera of experimentally T. gondii immunized mice, diluted 1:100 in 1% skim milk in PBS-T. Primary antibodies were detected by goat anti-mouse IgG antibody labeled with peroxidase (1:1000, Sigma) for 2 h at room temperature. Reaction was revealed with DAB and stopped with distilled water. In order to detect T. gondii forms on paraffin-embedded brain tissue sections of mice chronically infected with T. gondii ME-49 strain [42] , we carried out an immunohistochemical assay by using IgY anti-STAg as primary antibodies. Brain sections were deparaffinized and incubated with specific IgY (30 mg/mL) diluted in PBS for 30 min at 37uC. After washing, slides were incubated with rabbit anti-IgY labeled with fluorescein isothiocyanate (FITC, Sigma) diluted 1:300 in addition to DAPI for 60 min at 37uC. Slides were mounted with carbonate-buffered glycerin (pH 9.0) and read at an inverted fluorescence microscopy system (FSX-100, Olympus, Tokyo, Japan). In order to detect intracellular T. gondii replication, we standardized a cell culture based immunocytochemical assay using IgY anti-STAg as primary antibodies. Briefly, HeLa cells (ATCC No. CCL-2) cultured in 24-well plates were incubated with tachyzoites of T. gondii RH strain in a multiplicity of infection (MOI) equal to 1, in RPMI-1640 medium supplemented with 10% fetal calf serum for 24 h at 37uC and 5% CO 2 atmosphere. After incubation, cells were fixed with 4% formaldehyde and permeabilized with 0.1% Triton-X 100, followed by the incubation with the specific IgY antibody (30 mg/mL) for 2 h at 37uC. The next step consisted of the incubation with rabbit anti-IgY antibody labelled with FITC (Sigma), diluted 1:600 in PBS plus 49,6diamidino-2-phenylindole (DAPI, Sigma) for 1 h at 37uC. The reaction was read in an inverted fluorescence microscope system (EVOS, AMG Microscopy Group, USA). Statistical analysis was performed using the GraphPad Prism version 4.0 software (GraphPad Software Inc., La Jolla, CA, USA). The Student t test was used to investigate significant differences between groups. Values of P,0.05 were considered statistically significant.

Preparation of an antitumor and antivirus agent: chemical modification of α-MMC and MAP30 from Momordica Charantia L. with covalent conjugation of polyethyelene glycol

BACKGROUND: Alpha-momorcharin (α-MMC) and momordica anti-HIV protein (MAP30) derived from Momordica charantia L. have been confirmed to possess antitumor and antivirus activities due to their RNA-N-glycosidase activity. However, strong immunogenicity and short plasma half-life limit their clinical application. To solve this problem, the two proteins were modified with (mPEG)(2)-Lys-NHS (20 kDa). METHODOLOGY/PRINCIPAL FINDINGS: In this article, a novel purification strategy for the two main type I ribosome-inactivating proteins (RIPs), α-MMC and MAP30, was successfully developed for laboratory-scale preparation. Using this dramatic method, 200 mg of α-MMC and about 120 mg of MAP30 was obtained in only one purification process from 200 g of Momordica charantia seeds. The homogeneity and some other properties of the two proteins were assessed by gradient SDS-PAGE, electrospray ionization quadruple mass spectrometry, and N-terminal sequence analysis as well as Western blot. Two polyethylene glycol (PEG)ylated proteins were synthesized and purified. Homogeneous mono-, di-, or tri-PEGylated proteins were characterized by matrix-assisted laser desorption ionization-time of flight mass spectrometry. The analysis of antitumor and antivirus activities indicated that the serial PEGylated RIPs preserved moderate activities on JAR choriocarcinoma cells and herpes simplex virus-1. Furthermore, both PEGylated proteins showed about 60%–70% antitumor and antivirus activities, and at the same time decreased 50%–70% immunogenicity when compared with their unmodified counterparts. CONCLUSION/SIGNIFICANCE: α-MMC and MAP30 obtained from this novel purification strategy can meet the requirement of a large amount of samples for research. Their chemical modification can solve the problem of strong immunogenicity and meanwhile preserve moderate activities. All these findings suggest the potential application of PEGylated α-MMC and PEGylated MAP30 as antitumor and antivirus agents. According to these results, PEGylated RIPs can be constructed with nanomaterials to be a targeting drug that can further decrease immunogenicity and side effects. Through nanotechnology we can make them low-release drugs, which can further prolong their half-life period in the human body.

Momordica charantia L. (MC), a Momordica Linn. genus of the family Cucurbitaceae and commonly known as bitter melon, is a traditional medicine plant indigenous to China. 1 The fruit and seed extracts from MC have been used in China for centuries for antivirus, antitumor, and immunopotentiating agent purposes. 2 In recent years, several ribosome-inactivating proteins (RIPs), including momordica anti-HIV protein (MAP30) and α-momorcharin (MMC), β-MMC, and γ-MMC, a group of which belong to the family of single-chain RIPs, were found to have the ability to inhibit protein biosynthesis in tumor cells by catalytic inactivation of the 60S ribosomal subunit. 3, 4 These proteins were also found to be able to inhibit the multiplication of herpes simplex virus-1 (HSV-1), 4, 5 poliovirus I in Hep2 cells, 6 and acquired human immunodeficiency virus type-l (HIV-l). 7 However, the strong immunogenicity, allergic reaction, and short half-life of these proteins have been considered the major barriers for their application as therapeutic agents in vivo. 8, 9 In recent years, researchers have shifted their focus to other technologies. An established technology, polyethylene glycol (PEG) conjugation (PEGylation), can bestow on proteins several benefits, such as increasing plasma half-life, decreasing toxicity, and reducing immunogenicity and antigenicity. 9, 10 The Food and Drug Administration has approved the PEGylated forms of the therapeutic proteins such as uricase, erythropoietin, granulocyte-colony stimulating factor, interferon, adenosine deaminase, asparaginase, and a growth hormone antagonist. Another technology is nanotechnology, which uses nanomaterials for packing potential therapeutic proteins to extend the half-life period or to make them targeting drugs. α-MMC and MAP30 as potential therapeutic proteins possess biological activities such as inhibiting protein biosynthesis (ribosome inactivation), 11 antitumor, antivirus, and, especially, anti-HIV replication. 7, 12, 13 However, as foreign proteins are like other potential therapeutic proteins, poor biocompatibility limits their further development and application. To overcome these problems, in this study we first purified the two main proteins from bitter melon seeds and carried out their PEGylation using a branched 20 kDa (mPEG) 2 -Lys-NHS directed specifically to lysil ε-amino groups. Homogeneous one-mer, two-mer, and three-mer PEG-RIPs were then identified by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF-MS). Not only was their immunogenicity in vivo remarkably decreased but also, importantly, their antitumor and antivirus activities in vitro were moderately influenced when compared with the un-PEGylated counterparts. This work is just the beginning step towards them being used clinically. The application of PEGylation and nanotechnolgy may indicate that the potential application of both α-MMC and MAP30 can be developed for antitumor and antivirus agents in the future. Bitter melon seeds were obtained from the Institute of Agricultural Science and Technique of Sichuan Province, China. Matrices for electrophoresis were products of Sigma-Aldrich (St Louis, MO) and Bio-Rad Laboratories (Hercules, CA). SP-Sepharose FF, Sephacryl S-100, Macro-Cap-SP, and ampholyte were purchased from Amersham Pharmacia Biotech (Piscataway, NJ). (mPEG) 2 -Lys-NHS (20 kDa) was obtained from Shearwater Polymers (Huntsville, AL). Dulbecco's Modified Eagle's Medium (DMEM) and fetal bovine serum used in cell culture were from Gibco BRL (Grand Island, NE). pUC18 DNA was purchased from TAKARA (Dalian, China). Nitrocellulose (NC) membrane was obtained from Bio-Rad Laboratories. Sheep-antimouse Ab-linked to alkaline phosphatase was purchased from Sigma-Aldrich. JAR choriocarcinoma cells were purchased from the Cell Bank of Shanghai Institute of Cell Biology (Shanghai, China). All steps tried either alone or in combination within the process of purification were carried out at 4-6°C unless specifically stated. Firstly, the powder from fresh bitter melon seeds was extracted in 0.15 M NaCl solution and then the pH of the solution was adjusted to 4.0. After simple centrifugation, the supernatant was neutralized and fractionated by 30%-65% ammonium sulfate. The precipitate was dialyzed against the pH 6.3, 0.05 M phosphate buffer. Secondly, the sample was applied onto a SP-Sepharose FF column and eluted with pH 6.3, 0.05 M phosphate buffer containing 0.15 M NaCl. The elution peak containing 30 kDa protein was collected. Thirdly, the portion was loaded onto a Sephacryl S-100 column and the elution peak with 30 kDa protein was pooled. Finally, the sample was applied onto a Macro-Cap-SP column. A linear gradient of 0-0.15 M NaCl in pH 7.0, 20 mM sodium phosphate buffer eluted the column and two peaks with 30 kDa proteins were respectively collected. Briefly, α-MMC was separated on SDS-PAGE and then transferred to an NC membrane. Nonspecific binding was blocked and washed by placing the membrane in a solution of BSA. NC membrane was incubated with a mouse anti-α-MMC McAb and was exposed to a sheep-antimouse Ab linked to alkaline phosphatase. After adding BCIP (5-bromo-4-chloro-3-indolyl phosphate), the colored bands were visualized and photographed. 16 In the analysis of MAP30, NC membrane was incubated with a diluted solution of rat anti-MAP30 PcAb under gentle agitation. Other conditions were the same with α-MMC. In general, PEGylated α-MMC and PEGylated MAP30 can be obtained in the following optimal conditions: 10 mg/mL of α-MMC or MAP30 reacted with (mPEG) 2 -Lys-NHS (mass ratio of PEG:RIP was 2:1) in pH 8.5,100 mM borate buffer at room temperature for 30 minutes. The reaction mixture was applied with Sephacryl S-100 column. The PEGylated conjugates can be collected and assessed by gradient SDS-PAGE and MALDI-TOF-MS. Topological inactivation activity was according to a previously described method. 17 Antitumor activity in vitro JAR choriocarcinoma cells were maintained in an incubator supplied with a humidified atmosphere of 5% CO 2 at 37°C. The culture medium was DMEM containing 20 mM HEPES and 10% fetal bovine serum. The cell viability and proliferation were determined by quantitative 3-(4,5dimethylthiozol-2-yl)-2,5-diphenyltetrazolium bromide (MTT). A total of 3 × 10 4 cells/mL were applied into 96-well microtiter plates at 100 µL per well and exposed to PEGylated α-MMC, PEGylated MAP30, and un-PEGylated counterparts of 0.012, 0.06, 0.3, and 1.5 mg/mL at 24, 48, and 72 hours. Cells without drugs were used as controls. Each concentration was tested in quadruplicate. The determination procedure was carried out according to the previous description. 18 The optical density (OD) at 570 nm was measured using an ELISA plate reader (EL × 800, BIO-TEK, Atlanta, GA). Cell viability and proliferation were observed together with controls. The inhibitory effect of PEgylated rIPs on herpes simplex virus-1 VERO cells were cultured to logarithmic growth phase in 10 mL DMEM containing 10% (v/v) fetal bovine serum, 100 µg/mL penicillin, and 100 U/mL streptomycin. After removing medium, cells were substituted to 5 mL of the above medium and infected with 100 µL of herpes simplex virus-1 (HSV-1) for 1.5 hours. Then, 5 mL maintenance medium containing 2% (v/v) fetal bovine serum, 100 µg/mL penicillin, and 100 U/mL streptomycin was added and cells were continually cultured for about 1-2 days when over 80% cells appeared with cytopathic effect (CPE). The infected supernatant was collected for testing of virus titers. Determination of virus titer VERO cells of 1 × 10 5 cells/mL were applied 100 µL per well to 96-well microtiter plates and cultured for 24 hours. After discarding medium, 100 µL of infected supernatant, which was 10 −1 to 10 −10 diluted with medium, was added to each well (quadruplicate) until CPE did not increase. TCID 50 (50 percent tissue culture infective dose) was calculated according to Reed and Muench. 19, 20 Measurement of cytotoxicity The cytotoxicity of PEGylated α-MMC and PEGylated MAP30 was evaluated by the quantitative MTT test. 21 A total of 1 × 10 5 VERO cells/mL was applied to 96-well microtiter plates at 100 µL per well and was exposed to un-PEGylated proteins and PEGylated-proteins at 3.3 µmol ⋅ L −1 ∼3.3 nmol ⋅ L −1 (100 mg ⋅ L −1 ∼0.1 mg ⋅ L −1 ) and acyclovir (ACV) at 10 µmol ⋅ L −1 ∼0.01 µmol ⋅ L −1 for 48 hours as well as cells without drugs as control. Each concentration was tested in quadruplicate. The determination procedure was carried out according to the previous description. 18 The OD at 490 nm was measured using ELISA: Inhibition of proteins on HSV-1 antigen secretion A total of 1 × 10 5 VERO cells/mL were seeded into 24-well plates at 1 mL/well and then incubated with 5% CO 2 at 37°C for 24 hours. After abandoning the supernatant, 300 µL HSV-1 stock solution with 100 TCID 50 /mL was added to each well and incubated for 1.5 hours. Then, 100 µL of culture medium containing proteins (concentration was mentioned previously in the cytotoxity method) was added to each well and negative and positive cell control established. After incubation for 48 hours, HSV-1 antigen in culture supernatant was tested by human HSV-1 antigen 1 (HSV-1 AG1) enzymelinked immunosorbent assay kit with indirect ELISA: The analysis of virus inactivation VERO cells infected with HSV-1 were added with 20 µL (5 mg/mL) of MTT to each well and incubated at 37°C for 4 hours. Then, 150 µL of DMSO was added to each well. The OD 490 was read by a microplate spectrophotometer. The percentage of inhibition was calculated by the formula described under the heading "Measurement of cytotoxicity." Fifty Sprague Dawley rats were randomly classified into control, α-MMC, PEGylated α-MMC, MAP30, and PEGylated MAP30 groups of ten each. The rats in the control group were subcutaneously injected with saline solution, and the RIP and PEGylated RIP groups were emulsified in Freund's complete adjuvant at a dose of 1.42 mg/kg at the infection frequency of 3 days for 17 days. Blood samples were collected from the capillary vessel in rats' eyes and separated sera were stored at -20°C. Antigen-specific serum IgG levels were measured by ELISA. Briefly, 96-well plates were coated with 100 µL of 30 µg/mL purified antigen (RIPs and PEG-RIPs) in 0.05 M carbonate-coating buffer, SPSS statistical software (SPSS, Inc, Chicago, IL) was used for analysis. P , 0.05 was considered statistically significant. Purification and identification of α-MMC and MAP30 Purification of α-MMC and MAP30 was performed by applying 30%-65% ammonium sulfate precipitation, acidification, SP-Sepharose FF, Sephacryl S-100, and Macro-Cap-SP chromatography. In conclusion, 200 mg of α-MMC and 120 mg of MAP30 with the recoveries of 1.7% and 1.0%, respectively, was obtained from 200 g starting material seeds. The procedure of purification is summarized in Table 1 Through optimized PEGylation reaction, a high total PEGylation ratio can be obtained (about 60%-70%). The gradient SDS-PAGE monitoring (Figure 4) showed that the un-PEGylated counterparts from the reaction mixture were successfully removed by Sephacryl S-100 chromatography or Superdex 75 chromatography. In the MALDI-TOF-MS analysis ( Figure 5 ), one-mer, two-mer, and three-mer PEGylated isomers with 49899.4 Da, 70430.9 Da, and 91057.6 Da (the molar ratios of 1:1, 2:1, and 3:1 of (mPEG) 2 -Lys-NHS:α-MMC) were detected from α-MMC-PEG conjugates. Meanwhile, one-mer and two-mer PEGylated isomers with 50335.6 Da and 70614.2 Da (the molar ratios of 1:1 and 2:1 of (mPEG) 2 -Lys-NHS: MAP30) were found from MAP30-PEG conjugates. To demonstrate their topological inactivation activity, supercoiled DNA (pUC18) was incubated with PEGylated RIPs and un-PEGylated counterparts. In suitable enzymatic digestion conditions, all of the experiment samples cleaved the supercoiled double-stranded DNA to produce nicked circular or linear DNA. As shown in Figure 6 , all of them exhibited DNase-like activity. To investigate the effect of PEGylated α-MMC and PEGylated MAP30 on cell viability and proliferation, JAR cells were seeded on 96-well plates and were treated with increasing concentrations of PEGylated RIPs and un-PEGylated counterparts for 72 hours ( Figure 7A ) and for different times at 1.5 mg/mL ( Figure 7B ). Statistical analysis revealed that the concentrations of 1.5, 0.3, and 0.06 mg/mL started to significantly reduce the proliferation of cells after 48 and 72 hours of incubation. The analyses with the result of Figure 7A not prominent after 24 hours of treatment, but continued incubation for 48 or 72 hours with the proteins enhanced the cytotoxicity on cells. α-MMC and PEGylated α-MMC induced alterations in the morphology of JAR cells (Figure 8 ). After processing for 72 hours, untreated JAR cells extended and flattened ( Figure 8A ), while the treated groups showed fewer cells and abnormal shapes such as shrinkage, blebbing, and loss of membrane asymmetry, indicating the cytotoxic effect of α-MMC and its modifier on JAR cells. In particular, α-MMC-treated groups showed extremely obvious morphological changes ( Figure 8C and D). MAP30 and PEGylated MAP30 also showed a similar cytotoxic effect to JAR cells. The inhibitory effect of PEgylated α-MMC and PEgylated MAP30 on HSV-1 TCID 50 was calculated using the method of Reed and Muench, 19 and the emergence of CPE was used as a positive sign. The CPE of each dilution of HSV-1 cytopathic was tabulated as in Table 2 . Measurement of cytotoxicity was designed to establish appropriate dose range for the research of inhibitory effect of proteins on HSV-1 to VERO cells. The result (Figure 9 ) reflected the effects of different concentrations of test proteins on cell viability of VERO cells and showed that α-MMC/MAP30, PEG-α-MMC/MAP30, and ACV had no significant inhibition on VERO cells in the tested concentration range. The cell viability maintained above 97% from 0.003-0.03 µmol ⋅ L −1 and 85% from 0.3-3.3 µmol ⋅ L −1 . Therefore, the dose range of tested proteins inhibiting VERO cell infection with virus was below 3.3 µmol ⋅ L −1 . The inhibitory effect of α-MMC/MAP30, PEGylated α-MMC/MAP30, and ACV on HSV-1 glycoprotein antigen secretion glycoprotein was tested through the determination of HSV-1 glycoprotein antigen (Table 3) . The results showed a dose-dependent inhibitory effect. IC 50 was 0.67 µmol ⋅ L −1 with α-MMC, 2.94 µmol ⋅ L −1 with PEG-α-MMC, 0.47 µmol ⋅ L −1 with MAP30, 2.39 µmol ⋅ L −1 with PEG-MAP30, and 3.55 µmol ⋅ L −1 with ACV, respectively. Furthermore, IC 50 of α-MMC/MAP30 and PEG-α-MMC/MAP30 was lower than the general drug of ACV. At the same dose, the inhibitory effect on HSV-1 glycoprotein antigen secretion, glycoprotein of MAP30 was higher than that of α-MMC and PEG-MAP30 was higher than PEG-α-MMC. The MTT results showed that the ratio of cell viability can be increased on a dose-dependent feature compared with the control group when protein concentration was higher than 0.03 µmol ⋅ L -1 . But ACV did not display effects on HSV-1 inactivation. At the same dose, the ratio of cell viability with MAP30 was higher than α-MMC and PEG-MAP30 was higher than PEG-α-MMC. This indicated that both α-MMC/MAP30 and PEG-α-MMC/MAP30 had the ability of direct HSV-1 inactivation in a certain range of concentration. Additionally, the effect of inactivation was obvious with the increase of A novel preparative strategy of both α-MMC and MAP30 at one time was successfully established. Some properties of these proteins were assessed by SDS-PAGE, ESI-QUAD-MS, MALDI-TOF-MS, and N-terminal sequence as well as Western blotting. This fast, highly efficient methodology enables us to focus more energy on subsequent research. A branched (mPEG) 2 -Lys-NHS(20 kDa) was used to modify these proteins and the serial PEGylated ones were assessed by gradient SDS-PAGE and MALDI-TOF-MS. The α-MMC modifier possessed one-mer, two-mer, and three-mer PEGbound chains, while MAP30 modifier obtained one-mer and two-mer PEG-bound chains. PEGylated conjugates preserved moderate activities on JAR choriocarcinoma cells and HSV-1. Furthermore, both PEGylated proteins showed about 60%-70% antitumor and antivirus activities as well as a 50%-70% immunogenicity decrease when compared with unmodified counterparts. To sum up, PEGylation of α-MMC and MAP30 may offer a possible way for their clinical application as potential therapeutic agents. Nevertheless, to fulfill the requirements of a useful drug, we should meet the challenges, including the reduction of immunogenicity to the greatest extent, the retention of sufficient activity, the extension of the half-life period, and development of various forms of PEG-RIPs. This work is just the beginning for them to be used clinically. After all, therapeutic proteins must fulfill the requirements described previously. PEGylation can reduce immunogenicity and retain certain biological activity of RIPs, but there is lots of work to do for further decreasing their immunogenicity and extending the half-life period. The International Journal of Nanomedicine is an international, peerreviewed journal focusing on the application of nanotechnology in diagnostics, therapeutics, and drug delivery systems throughout the biomedical field. This journal is indexed on PubMed Central, MedLine, CAS, SciSearch®, Current Contents®/Clinical Medicine, Journal Citation Reports/Science Edition, EMBase, Scopus and the Elsevier Bibliographic databases. The manuscript management system is completely online and includes a very quick and fair peer-review system, which is all easy to use. Visit http://www.dovepress.com/ testimonials.php to read real quotes from published authors. International Journal of Nanomedicine 2012:7 Future work concerns packing RIPs with nanomaterials for targeting drugs and low release.

Evaluation of Four Different Systems for Extraction of RNA from Stool Suspensions Using MS-2 Coliphage as an Exogenous Control for RT-PCR Inhibition

Knowing when, and to what extent co-extracted inhibitors interfere with molecular RNA diagnostic assays is of utmost importance. The QIAamp Viral RNA Mini Kit (A); MagNA Pure LC2.0 Automatic extractor (B); KingFisher (C); and NucliSENS EasyMag (D) RNA extraction systems were evaluated for extraction efficiency and co-purification of inhibitors from stool suspensions. Real-Time Reverse Transcriptase Polymerase Chain Reaction (rRT-PCR) of MS-2 coliphage spiked into each system’s lysis buffer served as an external control for both. Cycle thresholds (Cts) of the MS2 were determined for RNA extracted from stool suspensions containing unknown (n = 93) or varying amounts of inhibitors (n = 92). Stool suspensions from the latter group were also used to determine whether MS-2 and enterovirus rRT-PCR inhibitions were correlated. Specifically 23 RNA extracts from stool suspensions were spiked with enterovirus RNA after extraction and 13 of these stool suspension were spiked with intact enterovirus before extraction. MS2 rRT-PCR inhibition varied for RNAs extracted by the different systems. Inhibition was noted in 12 (13.0%), 26 (28.3%), 7 (7.6%), and 7 (7.6%) of the first 93 RNA extracts, and 58 (63.0%), 55 (59.8%), 37 (40.2%) and 30 (32.6%) of the second 92 extracts for A, B, C, and D, respectively. Furthermore, enterovirus rRT-PCR inhibition correlated with MS2 rRT-PCR inhibition for added enterovirus RNA or virus particles. In conclusion, rRT-PCR for MS-2 RNA is a good predictor of inhibition of enterovirus RNA extracted from stool suspensions. EasyMag performed the best, however all four extraction methods were suitable provided that external controls identified problematic samples.

Diagnosis of enteric viral infections by Real Time reverse transcription -polymerase chain reaction (rRT-PCR) of RNA extracted from stool suspensions is complementing and increasingly replacing diagnosis based on viral isolation and characterization in tissue culture [1, 2] . However, interpretation of results is not always straightforward. The sensitivity of the rRT-PCR is negatively impacted, by compounds present in the clinical sample that may partially or completely inhibit the RT and/or PCR chemistries [3, 4, 5, 6, 7] . Potential inhibitors that might be incompletely removed from stool suspensions during RNA extraction include hemoglobin, immunoglobulins, bilirubin, triglycerides, complex polysaccharides, organic and phenolic compounds, glycogen, fats, and metabolic products especially those from pathological conditions, bacteria, vegetables, medications, anticoagulants, and drugs or alcohol [4, 6, 8, 9, 10, 11] . Adding to the list of endogenous rRT-PCR inhibitors are exogenous inhibitors from extraction protocols such as detergents, chelating compounds and guanidinium HCl [9] . The presence of inhibitory compounds in the extracted RNA and the extent of inhibition can be determined by semiquantitative rRT-PCR of an RNA template that is present in the sample, added to it prior to extraction, or introduced into the rRT-PCR mix. Natural or modified, encapsulated coliphage MS2 RNA is one source of protected RNA that has been used as a noncompetitive external RNA template control [5, 12, 13, 14] . The efficiency of removing inhibitors in patient samples may be related to the intrinsic properties of the method used to extract the RNA [15] . In this study we compared four commercial RNA extraction systems efficiencies for RNA isolation and removal of inhibitors in stool samples. The extraction systems were: QIAamp Viral RNA Mini Kit: manual extraction using silica columns (QIAGEN Inc, Valencia, CA, USA); MagNA Pure LC2.0 Automatic extractor with MagNA Pure LC Total Nucleic Acid Isolation Kit-High Performance: automatic RNA extraction using magnetic beads (Roche Diagnostics, IN, USA); KingFisher (Thermo Electron Corporation, Waltham, MA, USA) semi-automatic extraction using magnetic beads of Ambion MagMAX Viral RNA Isolation kit (Ambion, Inc, Austin, Tx, USA); NucliSENS easyMag (bioMerieux, Marcy l'Etoile, France): semi-automatic extractor using magnetic beads of easyMag extraction kit. In addition, the effectiveness of MS2 as an exogenous template control for measuring inhibitors co-extracted with RNA from stool suspen-sions and its relevance for validation of enterovirus diagnostic rRT-PCR suspensions is discussed. Ninety-three stool suspensions were extracted by the four extraction systems after spiking the lysis buffer of each with MS2 that yielded similar final MS2 concentrations. Any contribution from endogenous MS2 RNA was ruled out since no MS2 RNA was amplified from any of these 93 stool suspensions upon extraction with protocol D when exogenous MS2 was omitted from the lysis buffer. Results for each extraction procedure by assigned levels of inhibition are shown in Table 1 . Individual values are shown in Fig. S1 . The proportion of samples with inhibitors varied among the extraction methods. Of the 93 RNA extracts, 12 (13.0%), 26 (28.3%), 7 (7.6%), and 7 (7.6%) extracted by protocols A, B, C, and D, respectively, contained inhibitors of MS2 rRT-PCR. Median interquartile range (IQR) reduction in Cts for protocols A, B, C and D were 2.4 (0 to 34), 1.9 (0 to 33), 0.58 (0 to 29), and 0.26 (0 to 7), respectively. Finally, the number of samples with an inhibition .6 Ct were 6 (6.5%), 3 (6.4%), 2 (2.2%), and 1 (1.1%) for Extraction Protocols A through D, respectively. The Friedman Test indicated that there was a statistically significant difference in co-extraction of inhibitors of MS2 rRT-PCR among the extraction protocols used, P,0.001. In order to further challenge the performance of the 4 extraction systems to exclude inhibitors of rRT-PCR, 92 stool samples with levels of inhibition ranging from a single cycle to complete inhibition were evaluated. These samples were selected from among archived stool samples previously tested for enterovirus and MS2 after extraction by QIAamp Viral RNA Mini Kit. A sufficient number of samples with high, intermediate, and low levels of inhibitors were chosen for re-analysis to enable comparison between extraction procedures at each of these levels of inhibition. The results of MS2 inhibition for each sample by extraction protocol is shown in Fig. 1 and summarized by levels of inhibition in Table 1 , Experiment 2. The number of samples with inhibitors differed among the four protocols and for some individual samples, the amount of inhibition varied depending on the extraction protocol. Specifically, inhibitors were present in 58 (63.0%), 55 (59.8%), 37 (40.2%) and 30 (32.6%) of the samples prepared by Protocols A, B, C, and D, respectively. Median IQR reduction in Ct levels for protocols A, B, C and D were 11.7 (0 to 34), 7.0 (0 to 30), 4.2 (0 to 29), and 1.3 (0 to 30), respectively. More importantly the number and proportion of samples where inhibition was .6 Ct also varied among the protocols; 44 (47.8%), 29 (31.5%), 14 (15.2%), and 1 (1.1%), for Protocols A, B, C, and D, respectively. The Friedman Test indicated that there was a statistically significant difference in co-extraction of inhibitors of MS2 rRT-PCR among the extraction protocols used, P,0.001. To differentiate between rRT-PCR inhibition due to the presence of inhibitors in the RNA after extraction and variation due to differences in efficiency of extraction, 23 RNA extracts from samples with inhibitors were spiked with two concentrations of purified CoxB3 enterovirus RNA (Fig. 2 ). The first concentration was equivalent to that of MS2 (R 1 , equivalent) and the second contained 64 fold (6 Cts) more enterovirus RNA (R 2 , high). No significant difference between inhibition of MS2 rRT-PCR and enterovirus rRT-PCR was noted when equivalent amounts of enterovirus RNA was added to extracts prepared by protocols A, B and C (Fig. 2. part 2) . However, when the higher amount of enterovirus RNA was added, there was significantly less inhibition of enterovirus rRT-PCR than MS2 rRT-PCR extracts prepared using protocols B, C, and D, but not A. Thus, the decrease in MS2 rRT-PCR can serve as a measure of the amount of inhibitors in the RNA extracted by each of the extraction protocol. Analysis of variance, indicated that there were no significant differences between inhibition of Cts for MS2 and enterovirus RNA for Protocols A, B, and C, but not D, when the amount of enterovirus RNA was equivalent to that of MS2 (P.0.5), whereas there was significantly less inhibition when the 64-fold higher amount of enteroviral RNA was tested for RNA extracted by protocols B, C, and D (P,0.05), but not A (P.0.05). Thirteen stool suspensions with varying amounts of inhibitors were spiked with intact CoxB3 virus at a concentration adjusted to yield a Ct equivalent to that of the MS2. These spiked suspensions were re-extracted in parallel by all four methods. With few exceptions, the extent of inhibition of enterovirus rRT-PCR was similar to that for inhibition of MS2 rRT-PCR (Fig. 3) . The number of samples with $6 Ct inhibition in MS2 rRT-PCR, that also had .6 Ct inhibition of enterovirus rRT-PCR, was 10/10 (100%), 9/9 (100%), 6/7 (85.7%), and 4/4 (100%) for Protocols A, B, C, and D, respectively. Analysis of variance ( Fig. 3 , part 2), indicated that there was no significant difference (paired t-test, P.0.05) between the inhibition of rRT-PCR of the MS2 external control and the added enterovirus (P.0.05) for protocols A, C, and D. In protocol B the inhibition of the rRT-PCR for enterovirus was actually significantly higher than that for MS2 (P = 0.045). Recovery of enterovirus RNA by the four different protocols was equivalent. One of the trends in pathogen identification is the increasing reliance on molecular methods to achieve high throughput, high sensitivity and specificity of results in clinically relevant times [2, 16] . However, low pathogen copy number, inefficient extraction and the incomplete removal of inhibitors, reduce the sensitivity and specificity of the molecular assays. Among clinical samples, stool suspensions contain a wide range of materials that have the potential to inhibit RT and/or PCR chemistries [4, 6, 8, 9, 10, 11] . Thus, it is important to develop a reliable method to determine whether inhibitors are present in nucleic acid prepared directly from stool suspensions. An appropriate internal control does not exist because of wide differences in eukaryotic and prokaryotic composition in fecal material between patients. Four protocols for extracting RNA from stool samples were compared in this study. These four commercial extraction methods are among the most commonly used in the diagnostic laboratories. None of samples 1 to 93 contained MS2 RNA. Similarly Nonove et al (14) did not report any unusual amounts of MS2 RNA that would have indicated prior presence of MS2 in their samples when they added MS2 to 106 stool suspensions at concentrations two to eight-fold less than used in this study. MS2 coliphage was thus used as a non-competitive external control because of its absence from human clinical samples. MS2 is a better external control for rRT-PCR than plasmid or phage DNA since it also measures the efficiency of the RT reaction [5, 12, 13, 14, 17, 18] . Here, the MS2 was added to the lysis buffers to minimize sample manipulation and increase the uniformity between manual and semi-automatic extraction protocols. The presence of rRT-PCR inhibitors was evaluated by extracting equal volumes of stool suspension and eluting with equal volumes of elution buffer. Extractions were performed in parallel to eliminate storage related differences, and RNA was assayed on the same rRT-PCR run to minimize analytical differences [19] . Absence of MS2 in the stool samples 1-93 was confirmed by MS2 rRT-PCR of RNA extracted with the EasyMag RNA extraction system using lysis buffer without exogenous MS2. This extraction protocol, the system that had the least amount of samples with inhibitors, was chosen to increase the chance of finding any endogenous MS2 RNA in the stools. The number of stool suspension RNA extracts with inhibitors of MS2 rRT-PCR varied between RNA extraction protocols for randomly chosen stool suspensions. The amount of inhibitors in RNA extracts for individual stool suspensions also varied according to extraction procedures. This suggests that more than one type of inhibitor may be present and that different procedures do have different proficiencies in excluding them. Stool samples extracted by the KingFisher and the NucliSENS easyMag had fewer extracts with inhibitors than the QIAamp [18, 20, 21, 22] found that procedures similar to that used by EasyMag outperformed manual extraction procedures similar to QIAgen, and that MagNA Pure and KingFisher-like procedures were intermediate. In contrast fewer did not find major differences between some of these systems [23, 24] . Results presented here further confirm that RNA extracted with magnetic bead-based systems contained fewer inhibitors than column-based systems [21] . The number of samples with inhibitors was low in the first group of samples (samples 1-93) that were used to compare RNA extraction procedures. In order to better compare the four extraction procedures, we selected from among previously tested samples, sufficient numbers of archived stool suspensions that had high, intermediate, and low levels of inhibitors after extraction by QIAamp Viral RNA Mini Kit (samples 101-192). The stool suspensions were re-extracted in parallel by all four protocols and evaluated for inhibition of MS2 rRT-PCR. As before, KingFisher and NucliSENS EasyMag semi were better than the QIAamp Viral RNA Mini Kit and MagNA Pure LC2.0 Automatic extractor. Differences in extraction efficiency might be at least as important as inhibitors, as Hata et al. [25] showed by adding separate controls for both extraction and amplification. We performed a similar analysis by adding enterovirus RNA to RNA samples after extraction and intact enterovirus to suspensions before extraction. The inhibition of enterovirus rRT-PCR correlated with MS2 rRT-PCR inhibition. It is important to note that all four extraction protocols yielded equivalent amounts of enterovirus RNA from enterovirus added to suspensions where there was no inhibition of MS2 RNA rRT-PCR in the samples for equal starting and elution volumes. The pattern and amount of enterovirus rRT-PCR inhibition was similar for specific suspensions regardless of whether enterovirus was added before extraction or viral RNA added to the reaction mix after extraction. This implies that the apparent decrease in the amount of MS2 RNA in an RNA extract where MS2 RNA was added to the lysis buffer was primarily due to failure to remove inhibitors present in particular stool suspensions. The standard maximum volumes that could be extracted differ for the procedures: 140 ml for QIAgen; 100 ml for MagNA Pure; 50 ml for KingFisher; and, 200 ml for easyMag. The effect of using these manufacturers' recommended volumes for stool suspensions was not evaluated in the present study, although to have done so would have meant that depending on the procedure used, there would have been a 2 to 4 fold increase in the amount of endogenous inhibitors in the starting volume without a concomitant increase in the extraction, washing and elution buffer volumes. Others have shown that increasing the starting volume increased the co-extraction of inhibitors [21, 25, 26] and that the final rRT-PCR outcome is a balance between specific template and coextracted inhibitors [27] . The majority of rRT-PCR assays have a lower limit of quantitation of 10 target sequences per reaction with a Ct of approximately 35 cycles. A reaction with a three-fold shift upwards in Ct due the presence of inhibitors would still be positive, i.e. a specific signal would still be detectable below the 40 cycle cap recommended for the majority of rRT-PCR assays. A two Ct difference is within the statistical variation that may occur between repeated analyses of the same RNA. In contrast, a positive signal would no longer be detectible if there were a six-fold shift in Ct, which would produce a false negative outcome. In conclusion, inhibition is a complex process. All four extraction methods were suitable provided that an external control was used to identify problematic samples. rRT-PCR of MS2 RNA recovered from MS2 coliphage added to the lysis buffers of RNA extraction systems is a good predictor of inhibition of enteroviral RNA extracted from stool suspensions. More than one inhibitor may be present in the stool suspensions or added during extraction and their efficiency of removal differs between the extraction protocols. The correlation between the extent of MS2 rRT-PCR inhibition and enteroviral rRT-PCR inhibition increases inversely in relation to the amount of enteroviral RNA in the sample. In agreement with Dreier et al [5] , MS2 rRT-PCR inhibition should be tested for each RNA sample from a stool suspensions each time it is tested since there is no way to predict in advance whether inhibitors have been efficiently removed during extraction or remain active after cycles of frozen storage. Viral rRT-PCR results should not be considered as quantitative results when MS2 rRT-PCR is inhibited by more than 3 Ct. Finally, we recommend that any negative viral rRT-PCR result from a sample with an inhibition of .6 Ct for MS2 rRT-PCR should be considered invalid and alternative methods used to re-assay or re-extract the sample. If the RNA is diluted and re-tested only samples with positive results for enterovirus rRT-PCR should be considered as valid. The Ethical Review Board of the Sheba Medical Center, Tel Hashomer approved this study (SMC-8859). The samples and results were stripped of all links to personal details pertaining to, or which could be used to identify individual patients. All data were analyzed anonymously. The Ethical Review Board exempted this study from a requirement for obtaining informed consent. Stool suspensions (N = 185) prepared for routine analysis of clinical stool samples sent to the Central Virology Laboratory (CVL) at Chaim Sheba Medical Center in Israel were used to evaluate the efficiency of four different RNA extraction systems in excluding inhibitors of rRT-PCR. Ninety-three of the stool samples contained unknown amounts of rRT-PCR inhibitors (Table 1; Fig. S1 ). The remaining ninety-two stools suspensions were selected from among 860 stool suspensions archived between 2009 and 2011 that had been sent for routine enterovirus analysis (Table 1 , Figs. 1, 2, 3 and S2 ). The RNA from these archived suspensions had been extracted manually with the QIAamp Viral RNA Mini Kit and MS2 rRT-PCR and inhibition levels were known. MS2 rRT-PCR inhibitors .5 Ct were found in 93 (10.8%) of these samples. A non-random subset of 92 of these 860 samples (samples 101 to 192) with high, intermediate, and low levels of inhibitors were chosen for re-analysis so that would be a sufficient number of samples for comparison at each of these levels of inhibition. Small portions of fecal matter were vortexed for 15 seconds in stool suspension buffer, 2 ml 0.9% saline with glass beads (samples 1 to 93) or 5 ml of M199 containing 15.6 mg of dihydrostreptomycin, 15,625 U of Penicillin G, and 156 U of Mycostatin and 0.1 volume (v/v) of chloroform (samples 101-192). Suspensions were clarified by centrifugation at 2,5006g for 10 min and stored at 220uC until use. A natural E. coli RNA MS2 coliphage [28] (MS2, ATCC 15597-B1) stock was prepared on E. coli Top 10F in NZYCM broth as described by Dreier et al [5] . Aliquots of the stock were frozen at 270uC. The concentrated stock was thawed, serially diluted in dilution buffer [100 mM NaCl, 8 mM MgSO 4 , 50 mM Tris pH 7.5 and 0.01% (w/v) gelatin] and added to the respective extraction lysis buffers used in each of the extraction procedures before use. The amount of external control MS2 template added to the lysis buffers was adjusted to give a Ct of approximately 27 (,10,000 copies/ml) in the rRT-PCR mix upon addition of 5 ml of RNA. RNA was extracted from clarified fecal suspensions using four different commercial protocols. Samples were extracted in parallel to eliminate storage related differences [19] . Extractions were performed according to manufacturers' instructions except that RNA was extracted from 50 ml of stool suspension for all four protocols with addition of respective suspension buffer to reach the recommended aqueous volumes listed above. The RNA was stored at 270uC pending analysis and between analyses (See Fig. S2 ). Samples 1 to 93 were also extracted by protocol D as above, except that exogenous MS2 was omitted from the lysis buffer. The ABI Prism 7500 sequence detection system (Applied Biosystems, Foster City, CA, USA) was used for the amplification and detection of the MS2 and Enterovirus RNA by TaqMan technology as previously described [5, 29, 30] . Briefly, for MS2 rRT-PCR, 5 ml of RNA was added to the AgPath Mastermix (Ambion, Applied Biosystems Inc, Foster City, California), which contained the published concentrations of primers and probes and 5-carboxy-X-rhodamine succinimidyl ester (ROX) as an internal reference dye, whereas 8 ml of RNA was added for all enterovirus rRT-PCR assays. rRT-PCR was performed under the following conditions: 30 min at 48uC, 10 min at 95uC, and 60 cycles of 15 s at 95uC and 1 min at 60uC. RNAs extracted from the same stool suspension by the four procedures were assayed on the same rRT-PCR run to minimize analytical differences [19] . Data were managed and analyzed using Excel (Microsoft) and SPSS (ver. 15) software. For each extraction protocol, the presence and relative amount of rRT-PCR inhibitor(s) was determined by comparing MS2 rRT-PCR Ct results obtained in the absence or presence of stool suspensions. Measured MS2 Cts results were assigned to one of five levels of inhibition: no inhibition, inhibition of 1 to 3 Cts (2 to 8 fold), inhibition of 4 to 6 Cts (16 to 64 fold), inhibition of 7 to 9 Cts (128 to 512 fold) and inhibition of $10 Cts ($1024 fold). Similarly, the presence and relative amount of inhibitors of enterovirus RNA rRT-PCR was determined by comparing enterovirus rRT-PCR Ct results obtained in the absence or presence of stool suspensions for enterovirus added before extraction or enterovirus RNA added after extraction. The inhibition for suspensions that gave Cts for MS2 rRT-PCR below that for MS2 in suspension buffer alone was set to 0. The upper limit for rRT-PCR inhibition was capped at 29 Ct. The non-parametric Friedman Test was used to determine to determine whether there were significant differences between results among the four different extraction protocols. Post-hoc analysis of pairwise differences between protocols was performed using a Wilcoxon Signed-Rank Test (SPSS) with a Bonferroni correction that set the significance level for the pairwise comparisons to P,0.008. Comparison of MS2 rRT-PCR and enteroviral rRT-PCR Cts was by repeated measurements, analysis of variants. Figure S1 MS2 rRT-PCR inhibition in RNA extracted from stool suspensions using four different RNA extraction protocols. Equal amounts of stool suspensions chosen randomly from among samples sent to the laboratory were extracted by four protocols: (A) QIAgen, (B) magNA Pure, (C) KingFisher, and (D) easyMag as described in Methods. MS2 coliphage calculated to give 27 Ct by rRT-PCR was added to the extraction buffer. rRT-PCR values for MS2 in RNA extracted from buffer controls were subtracted from the values for MS2 in RNA extracted from stool suspensions. This difference, the number of Cts of inhibition, is shown in the box to the right of the sample numbers. Negative values were set to 0 and the maximum values for inhibition ''C'' were capped at 29 Cts. Samples with inhibition .10, 7 to 9, 4 to 6, and 1 to 3 Ct are indicated by the colors of the boxes: black, red, tan, and light yellow, respectively. Blank white boxes indicate no inhibition. The numbers of samples in each category and the significance in differences are shown in Table 1 , Experiment 1. (PDF) Figure S2 MS2 rRT-PCR inhibition for RNA extracted from stool after re-freezing and thawing by extraction protocol. 1. Three repeated measurements of MS2 rRT-PCR were performed for the 23 samples by protocols: (A) QIAgen, (B) magNA Pure, (C) KingFisher, and (D) easyMag as described in Methods. The RNA was stored at 270uC between tests. MS2 coliphage was added to the extraction buffer. rRT-PCR values for MS2 in RNA extracted from buffer alone were subtracted from the values for MS2 in RNA extracted from stool suspensions. This difference, the number of Cts of inhibition, is shown in the boxes to the right of the sample numbers. Negative values were set to 0 and the maximum values for inhibition ''C'' were capped at 29 Cts. Samples with inhibition $10, 7 to 9, 4 to 6, and 1 to 3 Ct are indicated by the colors of the boxes: black, red, tan, and light yellow, respectively. Blank white boxes indicate that the samples were not inhibited. Samples were stored at 270uC between measurements. 2. The means for each of the three repeat tests the square of the means (M Sq), the standard deviation of each run (S.D.), and the significance of differences between the means of the repeated measurements by extraction protocol are presented. There were no significant differences between the repeat measurements for each extraction protocol. 3. Friedman test values for comparison among the 4 extraction methods. There were significant differences among the extraction protocols (shaded yellow boxes). 4. The Wilcoxon Singed-Rank Test (SPSS) with Bonferroni correction (Significance if P,0.008) for the pairwise comparison of the means of the three measurements of the rRT-PCR results from the different extraction protocols. Shaded yellow boxes indicate where pairwise differences between extraction protocols were significant. (PDF)

Analysis on the Pathogenesis of Symptomatic Pulmonary Embolism with Human Genomics

BACKGROUND: In the present study, the whole human genome oligo microarray was employed to investigate the gene expression profile in symptomatic pulmonary embolism (PE). METHODS: Twenty patients with PE and 20 age and gender matched patients without PE as controls were enrolled into the present study in the same period. The diagnosis of PE was based on the clinical manifestations and findings on imaging examinations. Acute arterial and/or venous thrombosis was excluded in controls. The whole human genome oligo microarray was employed for detection. Statistical analysis was performed with t test following analysis of very small samples of repeated measurements and Gene Ontology (GO) analysis. RESULTS: Genomic data showed no damage to vascular endothelial cells in PE patients. Genomic data only found increased mRNA expression of a small amount of coagulation factors in PE patients. In the PE group, anticoagulant proteins, Fibrinolytic system and proteins related to platelet functions only played partial roles in the pathogenesis of PE. In addition, the mRNA expressions of a fraction of adhesion molecules were markedly up-regulated. Gene Ontology analysis showed the genes with down-regulated expressions mainly explain the compromised T cell immunity. Symptomatic VTE patients have compromised T cell immunity. CONCLUSION: The damage to vascular endothelial cells is not necessary in the pathogenesis of VTE, and only a fraction of factors involved in the shared coagulation cascade are activated. Genomic results may provide a new clue for clinical diagnosis, treatment and prevention of VTE.

Thrombus which may result in pulmonary embolism (PE) mostly comes from deep venous thrombosis (DVT). Altogether, both PE and DVT are named venous thromboembolism (VTE). Because of the high incidence of misdiagnosis, morbidity and mortality, VTE has become one of the essential medical problems around world (1, 2) . After 1990's, clinical report of incidence of PE are sharply increased yearly in China (3) . After 1990's, clinical reports of PE are sharply increased yearly in China (3) . In other Asia Ivyspring International Publisher regions, such as Japan, Thailand, Singapore, Taiwan, the number of reported PE cases is also increasing (4) (5) (6) (7) (8) . There are nearly 900,000 new PE cases in USA every year (9) , 78,000 new cases in Canada, 100,000-200,000 cases in Europe including France, Italy, Spain, Britain, Germany, and 47,000 cases in Australia every year (10). According to the statistical data, provided by an epidemiological surveillance in US, PE is the third leading cause of death (11, 12) , only lags behind malignancy and myocardial infarction. Virchow's classical triad of abnormalities including blood stasis, hypercoagulability and vessel damage has been regarded as the authentic theory for explaining pathogenesis of VTE since 19's century (13) , however, the triad does not fit all clinical profiles. In 1965, Egeberg et al (14) firstly introduced the concept of thrombophilia based on discovering a family with an autosomal dominant inheritance disease whose members manifested with repeatedly onset of DVT due to depletion of antithrombin III. Since then, many reports have considered that genetic gene mutation was one of the major factors in the pathogenesis of VTE (15) , but there has been a lack of sufficient genetic etiology evidence in many clinical PE patients. In recent years, some scholars have suggested that DVT-PE is a heterogeneous polygenic and multifactor disease involving the interaction of hereditary factors and environment with many risk factors such as trauma, surgery, advanced age, malignancy, pregnancy, heart failure, stasis, oral contraceptive, and so on (16) . In 2004 and 2008, ACCP proposed risk stratifications in patients receiving surgery, and different strategies should be performed for patients with different risks to prevent the occurrence of VTE (17, 18) . However, the annual number of VTE cases is increasing over year actually. Generally, the traditional concept can merely be verified merely in few patients with VTE, the pathogenesis of majority of other patients is still to be elucidated. Gene chip analysis provides the advanced tool for the study of gene function (19) (20) (21) . It can be a reliable approach for the study of differential gene expression between healthy people and patients and for the elucidation of molecular etiology of VTE. Twenty patients enrolled in PE group were those who admitted in hospital during 2007, with 11 males and 9 females, averaged 70(±14) years of age(44-89 years old). ALL patients were diagnosed as PE in accordance with at least 2 of following criteria. 1)Selective pulmonary arteriography showed filling defect or blockage; 2) Pulmonary ventilation perfusion scanning exhibited single or multiple blood flow perfusion defect with normal or abnormal ventilation, mismatched ratio of ventilation/perfusion; 3) Other clinical characteristics including typical manifestation of PE, arterial blood gas analysis, D-dimer test, ultrasoundcardiogram (UCG) and chest computerized tomography (CT) supported the diagnosis and excluded other cardiac and pulmonary disorders. Twenty patients with ischemic heart disease admitted at same period, excluding PE, DVT and other congenital bleeding and thrombosis diseases with comparative clinical presentation (11 males, 9 females, 44-91 years old with mean age 72±14) were enrolled in control group. The study protocol was approved by local ethics committee and informed consent was obtained from all patients in accordance with the declaration of Helsinki. Total RNA isolation 5 ml of peripheral blood samples anti-coagulated with EDTA were drawn from patients suspected with PE immediately after admitting to the hospital and from those patients without PE, respectively. Mononuclear cells were obtained through density gradient centrifugation with Ficoll solution and remaining red blood cells were destroyed with erythrocyte lysis buffer (Qiagen, Hilden, Germany). Total mononuclear cell RNA was extracted with TRIzol(Invitrogen, Carlsbad, USA) and purified with Qiagen RNeasy column (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Isolated total RNA was testified and quantified by means of Nanodrop ND-1000 spectrophotometer (Nanodrop Technology, Cambrige, UK). RNA samples of patients with confirmed diagnosis of PE and controls were labeled using indirect labeling method. Briefly, 1μg of total RNA was reverse transcribed. The cDNA then undergoes second strand synthesis and clean-up to become a template for in vitro transcription (IVT) with T7 RNA Polymerase. During IVT, the modified nucleotide, 5-(3-aminoallyl)-UTP (aaUTP) was incorporate into the cDNA. After that, fluorescent Cy3 was chemically coupled to aaUTP which contains a reactive primary amino group on the C5 position of uracil. The dye incorporation rate was assessed with a Nanodrop ND-1000 spectrophotometer and was found to be between 1.2 and 1.4 pmol/μl. Hybridization was carried out using the Agilent oligonucleotide microarray in situ hybridization plus kit (p/n 5184-3568), fol-lowing the manufacturer's instructions. Briefly, 750 ng of Cy3 labeled sample cDNA was subjected to fragmentation (30 min at 60°C) and then hybridization on 44K Human Whole-Genome 60-mer oligo-chips (G4112F, Agilent Technologies) in a rotary oven (10 rpm, 60°C, 17 h). Slides were disassembled and washed in solutions I and II according to the manufacturer's instructions. Significant differential gene expression analysis Random variance model (RVM) corrected t-test was used for the differential gene expression screening due to the small number of patients (20 cases each group) was far less than the amount of genes and low freedom degree of gene expression signal. P<0.05 is criterion of significantly different genes. Gene Ontology organizes gene function into hierarchical categories based on biological process, molecular function and cellular component (22) (23) (24) . Fisher's exact test was applied for over representation of selected genes in GO biological categories. In order to assess the significance of a particular category by random chance, false discovery rate (FDR) was estimated for all of categories. After 5,000 re-samplings, FDR was defined as FDR=1-Nk/T, where Nk refers to the subtracted number which was from Fisher's test p-values minus ϰ 2 test p-values in random samples. We specified the threshold of significant GO as p-value<0.05, FDR<0.05 and enrichment parameters. Enrichment represents the degree of gene expression significance. The equation of enrichment is as following Re=(nf/n)/(Nf/N) where nf is the number of significant genes within the particular category, n is the total number of genes within the same category, Nf is the number of significant genes in the entire microarray, N is the number of all genes tested. Similar to GO analysis, Fisher's exact test was employed for the study of over-representation of selected genes. According to the traditional theory, the pathogenesis of VTE is associated with abnormalities in blood flow, vessel integrity and blood components. Therefore, we compared gene expressions of these VTE related factors in patients with PE and without PE. Among the expressions of coagulation factor genes, only gene expressions of factor Ⅶ and FIBCD1 were significantly elevated in patients with PE compared with patients without PE. As for genes of other coagulant factors, the expression value of patients with PE was either not significantly different or less than those patients without PE in comparison. There was no significant difference between two groups of patients in gene expression of anticoagulant proteins. And so was in gene expression of fibrinolytic factors except plasminogen Urokinase receptor (PLAUR, Fig.1 ). L-selectin, ITGAL and ICAM-1 are the adhesion molecules originated from different family, mainly distributed on the surface of lymphocytes. ICAM-1 is the ligand of ITGAL which is member of integrin family. Significantly elevated mRNA expression of these adhesion molecules in PE group indicate activated adhesion function of white blood cells. However, mRNA expression of P-selectin (mainly distributed on the surface of ECs and platelets) and E-selection (mainly distributed on the surface of activated ECs) are not elevated in PE group which indicate venous endothelial cells are not impaired in patients with PE (Fig.2) . In expression of platelet aggregation related genes, only 2 genes (GP6 and PAFAH1B2) were significantly elevated in the patients with PE. Compared to patients without PE, the expressions of 3 in 7 genes of platelet adhesion function were significantly increased in patients with PE. As for genes of platelet releasing, the expression of one gene was significantly up and down regulated between 2 groups of patients, respectively (Fig.3 ). To identify the gene categories with differential expression in patients with or without PE, Gene Ontology analysis was carried out on the experiment data. The union of all differential expression genes resulted from data analysis are 2308 genes in patients with PE compared to patients without PE. Among them, 2238 genes are up-regulated and 70 are down-regulated. The main gene ontology categories impacted by these genes involve the up-regulation of 19 functioning categories against 4 down-regulation categories. Up-regulated genes are those genes whose functions are associated with transformation, phosphorylation, cell survival and cell conjugation, et al. While the function of down-regulated genes are relevant to the function of plasma membrane and activity of receptors, especially to the lymphocyte receptor complex and immunological synapse (Fig.4) . Gene ontology analysis exhibited compromised T cell mediated immune function, and t test indicated associated genes were significantly down-regulated in patients with PE than in control groups. Two genes with down-regulated expressions are closely related to the T cell mediated immunity according to GO analysis (with high value of Enrichment). These results reveal that T cell mediated immunity has been declined markedly in PE patients. The declined T cell receptor complex displayed as significanlly diminished mRNA expression of CD3G, CD3D, CD247, ZAP-70, T cell granzyme A and B, which will result in loss of functions of cytotoxic T cells. Additionally, the high mRNA expression of L-selectin, ITGAL and ICAM-1 in PE patients revealed the elevated adhesion of vascular endothelial cells, white blood cells and platelets which indicated that the adhesion molecules play an important role in the pathogenesis of VTE. The functions of platelets are characterized by aggregation, adhesion and release response (25) . Our results showed the mRNA expressions of 3/7 adhesion molecules were markedly up-regulated in PE patients, which suggests the adhesion of platelets plays an important role in the pathogenesis of PE. In 2011, we reported the proportions of CD3+T cells and CD8+T cells were markedly reduced and CD4/CD8 significantly increased in a series of CTEPH patients, which suggests the compromised T cell immunity in CTEPH patients and imbalance between CD3+T cells and CD8+T cells (26) . In addition, we also found that the number of CD3+T cells and CD8+T cells was dramatically reduced in a series of acute PE patients and cytological findings also supported the results from genomics studies on PE patients (27) . The occurrence of PE is related to the deficient or compromised T cell mediated immunity. This deficiency of T cell immunity may occur under the following conditions: 1) viral infection; 2) malignant tumors; 3) medication with immunosuppresants; 4) malnutrition. In the present study, most of subjects were old patients and malnutrition was not clinically obvious. Moreover, malignant tumors were not found in these patients and medication with immunosuppressants was absent. Thus, the compromised T cell immunity might be possibly related to viral infection. In 2010, we reported a patient who died of SARS developed VTE in multiple organs, which implies the viral infection induced systemic VTE and there is correlation between viral infection and occurrence and VTE (28) . Comparisons between PE patients and controls revealed the mRNA expressions of only a few proteins in the coagulation system, anti-coagulation system and fibrinolysis system were markedly up-regulated and only 3 factors or receptors in the shared coagulation cascade were activated, which was inconsistent with traditional theory that coagulation factors are comprehensively activated. We reported in our previous study that the main component of thrombus in acute venous thrombosis was fibrinogen and albumin and cytoskeletal proteins accounted for only a minor fraction of the thrombus (29) . The thrombus composed of fibrinogen is unstable, which makes the delayed thrombolysis feasible and also explain the therapeutic effectiveness of transcatheter thrombolysis not long after APE. Findings support that the formation of thrombus in VTE is not due to the comprehensive activation of coagulation factors as in traditional theory. In the present study, we apply the unitarian theory to explain the pathogenesis of symptomatic VTE: the occurrence of symptomatic VTE is closely related to the compromised immune function as well as the viral infection. This also explains why VTE is frequently found in patients with advanced age, trauma, surgery, malignant tumors, heart failure, immobilization, pregnancy and other risk factors. Our findings provide new knowledge on the etiology and pathophysiology of VTE and novel clue for the clinical diagnosis, treatment and prevention of VTE.

Diversity and roles of (t)RNA ligases

The discovery of discontiguous tRNA genes triggered studies dissecting the process of tRNA splicing. As a result, we have gained detailed mechanistic knowledge on enzymatic removal of tRNA introns catalyzed by endonuclease and ligase proteins. In addition to the elucidation of tRNA processing, these studies facilitated the discovery of additional functions of RNA ligases such as RNA repair and non-conventional mRNA splicing events. Recently, the identification of a new type of RNA ligases in bacteria, archaea, and humans closed a long-standing gap in the field of tRNA processing. This review summarizes past and recent findings in the field of tRNA splicing with a focus on RNA ligation as it preferentially occurs in archaea and humans. In addition to providing an integrated view of the types and phyletic distribution of RNA ligase proteins known to date, this survey also aims at highlighting known and potential accessory biological functions of RNA ligases. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s00018-012-0944-2) contains supplementary material, which is available to authorized users.

More recently, split tRNA (Fig. 1A , scheme c) and tri-split archaeal tRNA genes, which encode for parts of the mature domain, have been discovered [17, 18] . This type of tRNA arises from distinct transcription units that are joined by trans-splicing to yield a functional tRNA. Eukaryal tRNA introns are relatively short and range from 12 to 104 nucleotides in e.g., humans [8] . In eukarya, tRNA introns are less abundant than in archaea-e.g., only 6% of human tRNA genes and 20% of yeast tRNA genes are disrupted by introns [8] -and do not display any conserved structural motifs. Rather, the position of the intron is highly invariable in almost all known eukaryotic pre-tRNA genes (Fig. 1B ). An exception to this rule is exemplified by the non-canonical introns found in the circularly permuted tRNA genes of the red alga Cyanidioschyzon merolae (see scheme accompanying Fig. 1B ) [19] . Several hypotheses for evolutionary origins of tRNA introns have been put forward. An ''intron-early'' scenario proposes the existence of discontiguous primitive tRNA genes interrupted by introns harboring, for example, genes encoding splicing enzymes or aminoacyl transferases. Throughout evolution, these introns could have lost their initial functions and gradually acquired their current features [20] . The related ''split early'' hypothesis speculates that tRNA introns derive from flanking sequences of a priori split tRNA genes [21, 22] . The contrasting ''intron late'' or ''split late'' scenarios assume that tRNA introns arose later in evolution and propagated themselves by gene conversion or transposition [23] . Recent analyses of archaeal genome sequences suggest that tRNA introns in at least some archaeal species have been inserted into contiguous tRNA genes, e.g., by transposition [24] . A following separation of the sequences encoding the mature domains could have given rise to today's split tRNA genes [25, 26] . As their phylogenetic origins, the functions of tRNA introns are still being explored. The presence of introns in some particular pre-tRNAs has been demonstrated to be required for enzymatic modification of nucleotides such as methylation [27] and pseudouridylation [28] [29] [30] . Only recently has the function of tRNA introns in vivo begun to be addressed in genetic experiments carried out with Saccharomyces cerevisiae. Here, removal of all introns of a particular tRNA isodecoder family did not affect growth or translation of the obtained mutants at laboratory conditions [11] . A Secondary structure diagram of end-matured, intron-containing archaeal tRNAs. Schemes a and b represent end-matured tRNAs with introns in the D-arm or the T-arm, respectively. Scheme c depicts an end-matured split pre-tRNA assembled from two separate primary transcripts [25] . B Secondary structure diagram of end-matured, intron-containing eukaryotic pre-tRNAs. The accompanying scheme represents non-canonical introns and end processing sites of a permuted pre-tRNA transcript in the red alga C. merolae. The non-canonical intron in the acceptor arm is assumed to be excised by tRNA end processing enzymes rather than pre-tRNA splicing factors [19] . A, adenosine; C, cytosine; G, guanosine; U, uridine; W, pseudouridine; Y, pyrimidine; R, purine; asterisks indicate positions of additional introns, 5 0 -exonic regions are depicted in blue, 3 0 -exonic regions in red and intronic regions in green. Full grey circles indicate nonconserved nucleotides in regions of variable length, full blue circles indicate the position of the anticodon [31] . It is hoped that similar studies will yield further insights into the relevance of tRNA introns. Yeast mutants accumulating pre-tRNAs [32] initially provided access to substrates suitable for studying tRNA processing in biochemical experiments [33, 34] . These studies led to the conclusion that splicing of pre-tRNAs occurs in two steps. During an endonuclease reaction, the intron is first excised and the resulting tRNA exon halves are then ligated to form a mature tRNA ( Fig. 2A) [35]. The biochemical fractionation of extracts catalyzing pre-tRNA cleavage has led to the identification of tRNA endonuclease proteins in archaea [36] and yeast [37] , facilitating the characterization of the homologous human proteins [38] . In both archaea and eukarya, the tRNA endonuclease catalyses the cleavage of pre-tRNA at the two exon-intron boundaries and generates tRNA exon halves displaying a 2 0 ,3 0 -cyclic phosphate at the 3 0 -end of the 5 0 -exon and a 5 0hydroxyl at the 5 0 -end of the 3 0 -exon (Fig. 2B , box at the top). In addition, the reaction yields a linear intron with 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl termini ( Fig. 2A) . Although being mechanistically and evolutionarily related [39] , it has been shown that archaeal and eukaryal tRNA splicing endonucleases differ with respect to substrate recognition [9] . Studies modifying the substrates used for in vitro processing reactions revealed that splice site recognition by archaeal tRNA splicing endonucleases relies on the presence of the bulge-helix-bulge motif [14] . Archaeal tRNA endonucleases consist of two dimers, each composed of a catalytic and a structural subunit. These dimers combine in an antiparallel fashion and interact with the pseudosymmetric substrate. Each exon-intron boundary is cleaved by one of the diagonally juxtaposed catalytic subunits, which together form the two composite active sites [40] . Despite the conservation of the recognition motif, four types of oligomeric organization of archaeal splicing endonucleases have co-evolved with minor changes of their substrates [41] giving rise to a 4 (homotetramers, found in Euryarchaeota), a 2 (homodimers, found in Euryarchaeota), (ab) 2 (dimers of heterodimers found in some Crenarchaeota such as e.g., Nanoarchaeum equitans), and abcd (heterotetramers, also found in Crenarchaeota) enzymes [39, [42] [43] [44] . In the case of the homotetrameric endonucleases, two of the four identical chains function as structural subunits whereas the two remaining polypeptides serve as catalytic subunits [45] . The monomers of the homodimeric splicing endonucleases might have arisen from an in-frame gene duplication and thus encode for a fusion of a structural and a catalytic domain [42] . The eukaryal tRNA endonucleases are phylogenetically related to their archaeal counterparts [39] . The yeast endonuclease is of the abcd type and consists of two catalytic subunits, ScSEN2 and ScSEN34, and two structural subunits, ScSEN15 and ScSEN54 [37]. All four yeast subunits have homologous human proteins termed HsTSEN2, HsTSEN34, HsTSEN15, and HsTSEN54 [38] . Although it has been shown that at least one eukaryal tRNA endonuclease retained the ability to process archaeal pre-tRNA substrates [46] , eukaryal endonucleases seem to have a different mode of splice site recognition: Eukaryal tRNA endonucleases cleave pre-tRNA substrates at a conserved distance from a structural feature located in the mature domain [47] . Apart from a strictly conserved A:I base pair (Fig. 1B ) [48] , little sequence constraint seems to exist for the intron itself, which can be extensively mutated without disturbing its proper recognition by the endonuclease [49] . Proteins catalyzing the final step of tRNA splicing-ligation of tRNA exon halves [35]-have first been identified by biochemical fractionation approaches in yeast [50, 51] and plants [52] . Initial biochemical experiments already indicated that the ligation step is mechanistically not as conserved among the different archaeal and eukaryal organisms as the endonuclease reaction [53] [54] [55] [56] . Interestingly, although not requiring tRNA splicing enzymes, Escherichia coli has been shown to catalyze ligation of tRNA exon halves [57] . Even earlier, an RNA ligase activity had been identified in bacteriophage T4-infected E. coli [58] . These experiments led to the conclusion that, based on the type of phosphodiester bond established, distinct types of RNA ligase activity exist [10] (Fig. 2B ). Bacteriophages on one hand, and fungi and plants on the other, utilize two related but mechanistically distinct multistep reactions to prepare the 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl termini produced by the upstream endonuclease reaction for ligation (Fig. 2B , left box). Both mechanisms require the hydrolysis of the 2 0 ,3 0 -cyclic phosphate and phosphorylation of the 5 0 -hydroxyl prior to ligation and as a result the phosphate at the newly formed phosphodiester linkage originates from the nucleotide triphosphate cofactor used for the kinase reaction [10] . To indicate that a 5 0 -phosphate is joined to a 3 0 -hydroxyl Diversity and roles of (t)RNA ligases 2659 moiety, these two pathways are hereinafter referred to as 5 0 -3 0 RNA ligase mechanisms. In addition, two mechanisms exist that incorporate the cleavage site-derived phosphate into the splice junction (Fig. 2B , central and right boxes). 3 0 -5 0 RNA ligation converts 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl termini into a 3 0 ,5 0 phosphodiester and can Appp adenosine 5 0 -triphosphate; App adenosine 5 0 -diphosphate; Ap adenosine 5 0 -monophosphate; Gppp guanosine 5 0 -triphosphate; Gpp guanosine 5 0 -diphosphate; Nppp unspecified nucleoside 5 0 -triphosphate; Np unspecified nucleoside 5 0 -monophosphate; pp pyrophosphate; Ap-Lig adenylated ligase protein; NT-domain nucleotidyl transferase domain; Ptase 2 0 -phosphotransferase; NAD ? nicotinamide adenine dinucleotide; Appr[p ADP-ribose-1 00 ,2 00 -cyclic phosphate be detected in archaeal and vertebrate cell extracts. During 2 0 -5 0 RNA ligation, predominantly found in eubacteria and archaea, 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl moieties are joined to give rise to a 2 0 ,5 0 -phosphodiester bond [59, 60] . 5 0 -3 0 RNA ligases in bacteriophages, bacteria, archaea, and kinetoplastids Several bacterial species have evolved to sacrifice individual members of their populations upon phage infection by activating various suicide response pathways [61] . One of these suicide response mechanisms entails the activation of the latent anticodon nuclease PrrC in phage-infected E. coli CTr5x. The nuclease-kept in an inactive state in absence of phage infection-cleaves the host's tRNA Lys in the anticodon loop to shut down protein synthesis and thus impair phage propagation. Cleavage by PrrC yields 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl termini resembling the products of tRNA splicing endonuclease reactions. Bacteriophage T4, however, has evolved to cope with this defense mechanism. The phage pnk and rli genes encode for proteins capable of restoring cleaved tRNA Lys [62] . It has been demonstrated that T4 Pnkp (the protein encoded by the pnk gene) acts both as phosphatase [63] and polynucleotide kinase [64] thus converting the 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl of tRNA halves into 2 0 ,3 0 -cis diol and 5 0 -phosphate termini (Fig. 2B , left box, left branch). T4 Rnl1 (encoded by the rli gene) subsequently joins these tRNA halves by 5 0 -3 0 RNA ligation thus restores a functional tRNA [65] . In addition to T4 Rnl1, another RNA ligase, T4 Rnl2, has been identified in the genome of bacteriophage T4 [66] . T4 Rnl1 and T4 Rnl2 constitute distinct and only distantly related protein families sharing characteristic features (Fig. 3) . Members of the Rnl2 family have been detected and examined in viral, bacterial, and archaeal genomes, suggesting a common phylogenetic origin and possibly a function in RNA repair [66] [67] [68] [69] [70] . Potentially, this type of ligases might also be involved in archaeal tRNA splicing. In trypanosomatids, 5 0 -3 0 RNA ligases of the Rnl2 type are involved in RNA-guided editing of mitochondrial pre-mRNAs by nucleolytic cleavage and ligation [71] . The finding that tRNA splicing in yeast occurs as a twostep reaction, an ATP-independent endonuclease reaction, and an ATP-dependent ligase reaction constituted the first piece of evidence for the existence of a eukaryotic tRNA ligase [35] . Later, an RNA ligase activity was characterized in wheat germ extracts [55, 72] by using a nucleolytic fragment-obtained by RNase T1 digestion of tobacco mosaic virus RNA [73] -as an artificial ligase substrate. A thorough characterization of the splice junction revealed that a 2 0 -phosphomonoester-3 0 ,5 0 -phosphodiester linkage is established in this system [55, 72] . The same has been shown to be true for the yeast tRNA ligase [50] . Yeast (ScTRL1) and plant (exemplified by AtRNL from Arabidopsis thaliana) tRNA ligases-both identified by activity-guided chromatographic purification [50] [51] [52] -are multifunctional enzymes harboring all the activities required to join the 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl termini in three functionally independent domains [74, 75] (Fig. 3) . First, the 2 0 ,3 0 -cyclic phosphate is hydrolyzed by a cyclic phosphodiesterase (CPD) activity to yield a 2 0 -phosphate-3 0hydroxyl terminated 5 0 -exon [50, 76] . In a second step, the 5 0 -hydroxyl at the 3 0 -exon is phosphorylated by a kinase activity [50] . The actual ligation is preceded by ATPdependent adenylation of a lysine located within a KxxG consensus motif in the active site of the nucleotidyl transferase domain [50, 76] . Next, the enzyme transfers the adenyl moiety to the 5 0 -phosphate at the 3 0 -exon [50] . The adenylated RNA is then joined to the 2 0 -phosphate-3 0 -hydroxyl group at the 5 0 -exon with the concomitant release of AMP [51, 52] (Fig. 2B , left box, right branch and Fig. 3 ). Although RNA ligases of bacteriophage T4, fungi, and plants share important mechanistic features and key residues required for catalysis (Fig. 2B , left box and Fig. 3) , their overall sequence similarity is low [52] . However, the presence of the conserved nucleotidyl transferase domain with its ligase motifs in bacteriophage T4, fungal and plant RNA ligases suggests a common evolutionary origin for these enzymes [69] . Despite their relatedness, several differences exist between yeast, plant, and bacteriophage T4 ligases. For example, ligase-and kinase/phosphatase activities reside on separate polypeptides in bacteriophage T4 whereas yeast and plant tRNA ligases are multifunctional proteins harboring all activities required for ligation (Fig. 2B , left box, right branch and Fig. 3 ). Hydrolysis of 2 0 ,3 0 -cyclic phosphates by yeast and plant CPD domains yields 2 0 -phosphate-3 0 -hydroxyl products rather than the 2 0 ,3 0 -cis diol termini generated by T4 Pnkp (Fig. 2B , left box) [50, 55, 72, 76] . This arises from the distinct phosphoesterase domains employed by the bacteriophage T4 (aspartic acid based DxDxT phosphoesterase domain [77] ) and yeast/plant (2H phosphoesterase domain, characterized by two conserved histidines [78] ) ligation pathways [77, 79, 80] (Fig. 3) . In contrast to bacteriophage T4 ligases, the yeast and plant enzymes do not accept 3 0 -phosphate RNA substrates and ligation typically depends on the presence of a 2 0 -phosphate at the substrate termini [79, 81] . Yeast and plant RNA ligases generate splice junctions bearing a 2 0 -phosphate [72] . To convert this noncanonical 2 0 -phosphomonoester-3 0 ,5 0 -phosphodiester linkage into a 3 0 ,5 0 -phosphodiester, an NAD ? -dependent phosphotransferase, termed ScTPT1 in yeast, removes the 2 0phosphate generating ADP-ribose-1 00 ,2 00 -cyclic phosphate (Fig. 2B , left box, right branch) [82] [83] [84] [85] . Yeast and plant tRNA ligases also subtly differ with respect to substrate selection. While yeast tRNA ligase prefers tRNA exon halves over artificial substrates such as, e.g., linear introns [51] , the plant enzyme ligates circular introns as efficiently as tRNA exon halves in direct competition experiments [52] . Although a 5 0 -3 0 RNA ligase activity has been detected in biochemical experiments carried out with HeLa cell extracts [86] , no homologues of known 5 0 -3 0 RNA ligase proteins could be identified in animals [52] . It therefore seems likely that the 5 0 -3 0 RNA ligase in animals has diverged too far from the RNA ligases already known in order to be identified by existing algorithms or represents a completely novel type of enzyme. This view is supported by the recent identification of a 5 0 -3 0 RNA ligase in the cephalochordate Branchiostoma floridae (BfRNL) featuring a stand-alone ligase protein with weak homology to yeast and plant RNA ligases [87] . In this system, the cyclic phosphodiesterase and kinase modules (BfKinase/CPD or BfCLP1) do not reside on the ligase polypeptide (Fig. 2B , left box, right branch and Fig. 3 ). Enzymes catalyzing kinase (HsCLP1) [88] , cyclic phosphodiesterase (HsCNP) [89] [90] [91] , and phosphotransferase (HsTRPT1) [92] activities have been identified in humans. The fact that these enzymes can function in tRNA splicing pathways in vivo [93] [94] [95] has triggered speculation that a yet unidentified 5 0 -3 0 RNA ligase protein might also exist in humans [87] . In support of this assumption, the RNA kinase, HsCLP1, is an integral component of the tRNA endonuclease complex, suggesting the occurrence of coupled endonuclease and kinase reactions in humans [38, 88] . In contrast, tRNA exon half phosphorylation is associated with ligation in S. cerevisiae, where ligase and kinase activities both reside on ScTRL1. However, the relevance of the 5 0 -3 0 RNA ligase mechanism for tRNA splicing in vertebrates is under debate since the 2 0 -phosphotransferase is not essential in mice [96] . To account for the substantial differences in sequence and domain organization, it has been suggested to group 5 0 -3 0 RNA ligases in bacteriophages, yeast, plants, and animals into classes [66, 87] ; however, no uniform nomenclature has become generally accepted to date. Biochemical studies revealed that ligation of tRNA exon halves in vertebrates and archaea is mainly achieved by an alternative mechanism. A 3 0 -5 0 RNA ligase activity resulting in incorporation of the precursor-derived 2 0 ,3 0cyclic phosphate into the splice junction as a 3 0 ,5 0 -phosphodiester was for the first time detected in HeLa cell extracts (Fig. 2B, central box) [53, 56, 97] . The same reaction has later been shown to occur in archaeal cell extracts [54, 98, 99] . Intensive attempts to identify the human tRNA ligase proteins by chromatographic purification were triggered by these initial findings [100, 101] . Recently, chromatographic purification led to the identification of RtcB/HSPC117 proteins as 3 0 -5 0 RNA ligases in the crenarchaeon Pyrobaculum aerophilum and in humans [102, 103] . Concurrently, a candidate approach revealed RtcB in Escherichia coli as a 3 0 -5 0 RNA ligase [104] . The high degree of conservation of HSPC117/RtcB proteins suggested a shared role for this protein family in many organisms already during their initial characterization [102] [103] [104] . RNA ligase activity of the recombinant RtcB proteins from P. aerophilum and E. coli strictly depends on the presence of bivalent metal ions [102, 104] . Studies probing the active sites of archaeal and bacterial RtcB by mutagenesis [102, 104, 105] could in part confirm early predictions concerning essential residues based on structural analyses of homologous RtcB proteins (pdb files 1UC2 and 2EPG) [106, 107] . However, all crystal structures of RtcB proteins available to date represent apo forms of the enzymes and thus the exact functions of individual amino acids lining the presumed active site cannot be assigned unambiguously. Extended structural studies of RtcB proteins in complex with bivalent metal ions, additional cofactors, and RNA substrates will be required to clarify the active site geometry and the distinct metal ion specificity observed for enzymes from different species [102, 104] . Initially, HSPC117/RtcB proteins were assumed to catalyze the direct nucleophilic attack of the 2 0 ,3 0 -cyclic phosphate by a 5 0 -hydroxyl group, which seemed likely for two main reasons. First, 2 0 ,3 0 -cyclic phosphates are energyrich substrates with a favorable leaving group [108] and second, HSPC117/RtcB-catalyzed ligation reactions did not seem to be strictly dependent on the addition of nucleotide triphosphate cofactors [102] [103] [104] 108] . However, only enzyme preparations that were rigorously purified in the presence of chelating agents proved to yield RtcB preparations sufficiently pure to demonstrate the real cofactor requirements of RtcB-catalyzed 3 0 -5 0 RNA ligation [108] . Thus, 3 0 -5 0 RNA ligation by RtcB proteins occurs as a sequential reaction involving the stoichiometric hydrolysis of nucleotide triphosphates rather than direct nucleophilic attack of the 2 0 ,3 0 -cyclic phosphate by a 5 0 -hydroxyl group [108] . One advantage of this somewhat counterintuitive strategy might lie in the suppression of the backward reaction-the cleavage of the newly formed phosphodiester-as it is catalyzed by bacterial 2 0 -5 0 -ligases (see below) [59] . The human tRNA ligase complex 3 0 -5 0 RNA ligation appears to be the prevalent human tRNA splicing pathway [53, 56] and relies on HSPC117 as the essential ligase component as supported by two experimental observations: RNAi-mediated depletion of HSPC117 severely impairs tRNA maturation and mutation of a strictly conserved cysteine residue abolishes ligase activity of the affinity purified protein [103] . Human HSPC117, together with the proteins DDX1, CGI-99, FAM98B, and ASW, forms a stable complex of about 200 kDa (Fig. 4) [103] . Even earlier, the observed co-selection of HSPC117, DDX1, and CGI-99 with cruciform DNA duplexes hinted that these three proteins might interact [109] . In contrast to depletion of HSPC117, RNAimediated silencing of the associated proteins does not severely affect RNA ligase activity in HeLa cell extracts, suggesting further functions of the interacting proteins that remain to be explored [103] . Potential roles of these accessory proteins may include targeting of the complex to appropriate cellular compartments, stabilization of the essential subunit, interaction with substrates or adaptor proteins and prevention of illegitimate, promiscuous ligation. Moreover, individual complex components may mediate a transient association of the ligase complex with the tRNA endonuclease, as it has been suggested for the yeast tRNA endonuclease and ligase [110] . In contrast to its archaeal and bacterial orthologues, recombinant HSPC117 did not act as an RNA ligase in vitro. In addition, HSPC117 per se was incapable of replacing the yeast tRNA ligase TRL1 in complementation experiments (J. Popow, unpublished results). The human tRNA ligase complex components seem to be constitutively and widely expressed in all human and mouse tissues analyzed so far, indicating that these proteins cooperate in a great variety of cellular contexts [103, 111, 112] . DDX1 is a member of the DEAD-box family of putative RNA helicases characterized by the presence of nine conserved sequence motifs [113] . The DEAD-box helicase domain of DDX1 is interrupted by the insertion of a SPRY domain (Fig. 4 ) detectable in numerous proteins and presumably acting as a protein interaction platform [114, 115] . DDX1 has been associated with many molecular functions ranging from mRNA processing [116, 117] to recognition of DNA double-strand breaks [118] and has been demonstrated to exhibit 3 0 -5 0 RNA unwinding activity [117] . The association of DDX1 with HSPC117 suggests an involvement in tRNA splicing. On the other hand, RNAi-mediated depletion of DDX1 only mildly impaired tRNA maturation activity, suggesting that its function might be dispensable for RNA ligation by HSPC117 in HeLa cell extracts [103] . Alternatively, a lack of DDX1 might also have been compensated for by other proteins present in the assayed extracts. Further experiments assessing the effect of mutations within conserved motifs of DDX1 on ligase activity-carried out with purified RNA ligase complex-might successfully address the function of this protein in tRNA splicing. Co-localization studies and extended analysis of proteins associating with DDX1 could answer the question whether DDX1 mediates its functions in conjunction with HSPC117 or whether it is a shared component of multiple protein complexes acting in distinct biological processes. Furthermore, it might be rewarding to test whether the amplification of DDX1 observed in cancerous cell lines [111, 119, 120] , its function in viral replication [121] [122] [123] , and dsRNA recognition in dendritic cells [124] are linked to RNA ligase activity or tRNA processing. Published information concerning the function of CGI-99 is scarce. Although the protein has been found to interact with human [112, 125] and viral [125] proteins in yeast two-hybrid assays, these studies provided little insight into its molecular function. It remains to be determined whether the interaction of CGI-99 with the PA subunit of influenza virus polymerase [125] or its modulation of transcription by RNA polymerase II [126] is in any respect related to RNA ligase activity. The identification of FAM98B as a human tRNA ligase complex component is the first piece of information published concerning its molecular function. Within the context of tRNA splicing, the exact role of FAM98B needs to be established. Since its RNAi-mediated depletion has almost no impact on RNA ligase activity in HeLa cell extracts [103] , it is likely that FAM98B acts as an interaction platform recruiting accessory proteins or RNA substrates. On the other hand, it might also be the case that other cellular proteins compensated for a lack of FAM98B in these experiments. Alternatively, FAM98B may mediate completely unrelated, yet-to-be-discovered functions of HSPC117 complexes. ASW has first been characterized in a genetic screen searching for genes differentially expressed in early neural specification in Xenopus laevis [127] . Although this study could demonstrate that alteration of ASW levels in X. laevis leads to severe developmental defects, it provided limited insights into mechanistic aspects of ASW function. ASW has been speculated to be specific for vertebrates as no homologous genes were detectable in Drosophila melanogaster or Caenorhabditis elegans [127] . Nonetheless, homologues of ASW are detectable in the genomes of at least some arthropods and nematodes (Fig. 5A) . Future studies might answer the question of whether the link between ASW and HSPC117 explains any of the developmental defects observed upon manipulation of ASW levels in X. laevis. Taken together, apart from HSPC117, which appears to be conserved among all domains of life, the additional components of the HSPC117 complex do not show a comparably broad taxonomic coverage. Orthologues of DDX1, FAM98B, and CGI-99 are spread more widely than ASW, which seems to be restricted to fewer species (Fig. 5A ). Another RNA ligase activity leading to the unusual 2 0 ,5 0 -phosphodiester bond has been shown to act in extracts prepared from various bacterial species (Fig. 2B , right box) [57] . Its biochemical purification led to the identification of the ligase gene ligT, conserved not only in bacteria but also in euryarchaeota and crenarchaeota [59, 78] . The apparent dependence of its activity on the presence of modifications in its tRNA substrates suggests that it is involved in repair or processing of tRNA in its host [10, 59] . LigT from E. coli catalyses an ATP-independent equilibrium reaction between 2 0 ,3 0 -cyclic phosphate and 5 0 -hydroxyl termini and 2 0 ,5 0 -phosphodiester bonds [59] . More recently, a homologous archaeal 2 0 -5 0 RNA ligase has been characterized in the euryarchaeon Pyrococcus furiosus confirming the broad phyletic occurrence of LigT proteins [60] . Although the high degree of conservation suggests an evolutionarily important role for 2 0 -5 0 ligase genes, the exact biological function of LigT proteins and 2 0 ,5 0 -phosphodiester bond formation are still unknown [59, 60, 78] . Interestingly, a few eukaryotic genomes also encode homologs of bacterial and archaeal LigT proteins (Figs. 5B, 6) . The activities and biological functions of these proteins have not yet been addressed. Phyletic association of RNA ligases RNA ligase proteins are detectable within all major lines of descent. Considering the different ligation mechanisms and classes of RNA ligase polypeptides, it seems, however, that several species abandoned one or the other mechanism of RNA ligation (Figs. 5B, 6 ). Most representatives of the Basidiomycota and Spermatophyta lineages might have lost RtcB/HSPC117 homologues during the course of evolution. However, there seem to be rare instances (such as the basidiomycote Piriformospora indica, see supplemental information for details) where an RtcB/HSPC117 protein has been retained [52, 102, 103, 128] (Figs. 5B, 6) . None of the 5 0 -3 0 RNA ligases known to date could be allocated to several higher eukaryotes, such as e.g., humans (Fig. 5B) . Nevertheless, this picture might change upon the identification of a currently unknown type of 5 0 -3 0 RNA ligase. The detection of 5 0 -3 0 RNA ligase activity in humans [86] together with the recent discovery of the B. floridae 5 0 -3 0 RNA ligase BfRNL [87] argue in favor of this view. The subcellular localization of tRNA processing events does not seem to be conserved among all eukaryotes. Many localization studies were carried out in the yeast S. cerevisiae which, however, seems to differ in this respect from many other eukaryotes [129] . tRNA splicing is assumed to occur in the cytoplasm in S. cerevisiae, based on the localization of yeast tRNA endonuclease subunits to the outer mitochondrial membrane [130] . Controversial data have been reported concerning the localization of tRNA splicing in plants. On the one hand, some plant tRNA splicing enzymes at least partly localize to the nucleus [131] . On the other hand, impaired nuclear export of tRNA results in accumulation of unspliced pre-tRNA in Arabidopsis thaliana [132] arguing in favor of cytoplasmic tRNA splicing in plants in analogy to the coupling of nuclear export and cytoplasmic tRNA splicing in yeast rna1-1 mutants [32, 133, 134] . In the vertebrate X. laevis, tRNA splicing-preceded by 3 0 -end formation including CCA addition-has been shown to occur in the nucleus by micro-dissection and micro-injection studies carried out with oocytes [135] [136] [137] . In accord with these observations, subunits of the human tRNA endonuclease have been reported to be nuclear proteins [38] . The biochemical characterization of identified tRNA ligase enzymes revealed that they accept not only tRNA exon halves but also a broad range of artificial substrates such as nucleolytic RNA fragments, synthetic RNA duplexes, and linear introns generated by tRNA endonucleases [52, 53, 55, 97, [102] [103] [104] [105] . In addition, human cell extracts have been shown to not only ligate 2 0 ,3 0cyclic phosphate bearing but also 3 0 -phosphate-terminated RNA substrates [97, 138] . Biochemical fractionation of HeLa extracts revealed that RNA 3 0 -phosphate terminal cyclase RTCD1 can convert RNA 3 0 -phosphates into 2 0 ,3 0 -cyclic phosphates in an ATP-dependent reaction [139, 140] . Despite the thorough characterization of bacterial and human RNA 3 0 -phosphate terminal cyclase proteins [140, 141] , the physiological function of this enzyme class is still unknown. The recent mechanistic characterization of RtcB from E. coli suggests that 3 0 -phosphate-terminated RNA is per se a substrate for this enzyme [108] , raising questions about the genuine function of RNA terminal cyclase RtcA in E. coli. The potential to prepare a variety of RNA termini for ligation and the relaxed substrate specificity of the ligase itself in many organisms suggests that tRNA splicing enzymes might also act in other biological contexts. As an example of such a case, ScTRL1 has been shown to be involved in stress-induced, non-canonical splicing of the HAC1 transcript in S. cerevisiae [142, 143] . A related pathway acts in human cells, however, the involved RNA ligase has to date not been identified [144] [145] [146] [147] . Furthermore, RNA ligases have been proposed as host factors for the propagation of viruses, viroids, and viroidlike satellite RNAs in humans and plants [148] . In fact, HSPC117 and DDX1 are involved in RNA processing during replication of the hepatitis delta virus [121] , presumably acting as the host RNA ligase factors assumed to be required for cyclization of the viral RNA genome as previously proposed [149] . Furthermore, components of the HSPC117-complex together with the RNA cyclase, RTCD1, have been shown to interact with kinesin-associated RNA transport granules in mouse brain extracts [150] . The exact functions of these RNA metabolic enzymes in the context of RNA transport await further experiments. In addition to processing of RNA transcripts of various origins, RNA ligases are assumed to be involved in RNA repair pathways. Apart from the wellstudied example of tRNA repair by bacteriophage T4 RNA ligase [62] , further examples of RNA repair have been revealed [67, 79, 151] . These studies showed an amazing potential of RNA ligases to function as safeguards against the deleterious effects of cytotoxic nucleases in yeast and bacteria. Similar efforts might unveil examples of RNA repair in unanticipated physiological contexts not only in bacteria and yeast but also in higher organisms. The high degree of conservation of HSPC117/RtcB proteins and preliminary data indicating that HSPC117 seems to be encoded by an essential gene in mice [152] are highly suggestive of universal and important roles for this protein family. Assigning a biological function to the operon harboring the ligase, rtcB, and the cyclase, rtcA, in some bacteria may help to uncover functions of this type of ligase unrelated to tRNA splicing since E. coli does not encode for intron harboring tRNAs. Finally, the identification of genuine substrates interacting with RNA ligase proteins in vivo may yield fascinating insights into the various processes potentially requiring RNA ligases. Diversity and roles of (t)RNA ligases 2667

Molecular Imaging Reveals a Progressive Pulmonary Inflammation in Lower Airways in Ferrets Infected with 2009 H1N1 Pandemic Influenza Virus

Molecular imaging has gained attention as a possible approach for the study of the progression of inflammation and disease dynamics. Herein we used [(18)F]-2-deoxy-2-fluoro-D-glucose ([(18)F]-FDG) as a radiotracer for PET imaging coupled with CT (FDG-PET/CT) to gain insight into the spatiotemporal progression of the inflammatory response of ferrets infected with a clinical isolate of a pandemic influenza virus, H1N1 (H1N1pdm). The thoracic regions of mock- and H1N1pdm-infected ferrets were imaged prior to infection and at 1, 2, 3 and 6 days post-infection (DPI). On 1 DPI, FDG-PET/CT imaging revealed areas of consolidation in the right caudal lobe which corresponded with elevated [(18)F]-FDG uptake (maximum standardized uptake values (SUVMax), 4.7–7.0). By days 2 and 3, consolidation (CT) and inflammation ([(18)F]-FDG) appeared in the left caudal lobe. By 6 DPI, CT images showed extensive areas of patchy ground-glass opacities (GGO) and consolidations with the largest lesions having high SUVMax (6.0–7.6). Viral shedding and replication were detected in most nasal, throat and rectal swabs and nasal turbinates and lungs on 1, 2 and 3 DPI, but not on day 7, respectively. In conclusion, molecular imaging of infected ferrets revealed a progressive consolidation on CT with corresponding [(18)F]-FDG uptake. Strong positive correlations were measured between SUVMax and bronchiolitis-related pathologic scoring (Spearman’s ρ = 0.75). Importantly, the extensive areas of patchy GGO and consolidation seen on CT in the ferret model at 6 DPI are similar to that reported for human H1N1pdm infections. In summary, these first molecular imaging studies of lower respiratory infection with H1N1pdm show that FDG-PET can give insight into the spatiotemporal progression of the inflammation in real-time.

In March of 2009, an outbreak of a novel variant of H1N1 influenza A virus was reported in cases of influenza illness in Mexico [1] . By June 11, the World Health Organization raised the pandemic alert level to its highest level, declaring the first influenza pandemic in over 40 years [1] . Unlike seasonal influenza viruses, this novel H1N1 pandemic strain (H1N1pdm) tended to affect younger healthier populations and had an increased risk of morbidity and mortality [2] [3] [4] with 12-30% of the population developing clinical influenza, 4% of those requiring hospital admission, and 1 in 5 requiring critical care [5] . In general, however, infection of the H1N1pdm was relatively mild in most persons, although a fatal viral pneumonia with acute respiratory distress syndrome occurred in approximately 18,000 cases. In contrast to seasonal influenza in human cases, H1N1pdm infections showed a tropism for the lung similar to H5N1 [6] . The ability of H1N1pdm viruses to infect the lower respiratory track has been attributed to a broader specificity in the binding of the viral hemagglutinin (HA) to a2-3in addition to a2-6-linked sialic acid (SA) receptors [7, 8] . It is reasonable that the lung tropism of the H1N1pdm contributed to the severity of disease in those individuals with preexisting complications such as asthma and chronic obstructive pulmonary disease (COPD) [6, [9] [10] [11] [12] . Data from limited human autopsies and animal studies of various pandemic strains also suggest contribution of the host innate immune response and the virus in the progression of disease [13] [14] [15] [16] . Molecular imaging can potentially play a strong role in basic infectious disease research and clinical response by providing a noninvasive, spatiotemporal measurement of viral infection and host inflammation [17, 18] . To explore the potential utility of molecular imaging in influenza infection, we chose the ferret (Mustela putorius furo) model. Ferrets have been used as an animal model of influenza infection and pathogenesis since 1934, when they were reported to develop an acute respiratory tract illness when exposed to influenza viruses from humans and swine [19] . In contrast to mice, the ferret can be infected by human isolates without adaptation and display signs and symptoms of infection such as sneezing and nasal secretions that are similar to what is seen in humans [20] [21] [22] [23] . The ferret is an attractive model for imaging influenza pulmonary infections given the ferret's long trachea, large lung capacity, and bronchiolar branching. These anatomical features can potentially bridge imaging with histopathologic evaluation. Finally, the ferret more closely mimics humans in distribution of sialic acid (SA) receptors in the respiratory tract with higher a-2-6 SA in the upper respiratory tract and SA with a-2-3in the lower [24] . Historically, plain film (x-ray) radiography and computed tomography (CT) have been useful for clinical assessments of influenza disease severity in clinical cases [25, 26] . These imaging modalities are limited by characterizing only anatomic changes in the lung parenchyma, such as ground-glass opacity (GGO) and consolidations, which represent different degrees of interstitial and alveolar filling by cells, edema, and inflammatory exudate [27] . In contrast, positron emission tomography (PET) imaging with the radiotracer, [ 18 F]-2-fluoro-2-deoxy-D-glucose ([ 18 F]-FDG), can provide data on metabolic activity of cells by measuring sites of increased glycolysis from leukocyte chemotaxis and accumulation, and provide increased sensitivity in detection of cells during inflammation. [ 18 F]-FDG, an analog of glucose that is moved into cells via facilitated transport, is commonly used in PET imaging as a radiotracer in clinical and basic science research. Recent studies have demonstrated the utility of the [ 18 F]-FDG in assessment of infectious disease burden in animal models of schistosomiasis and tuberculosis [28] [29] [30] . PET/CT has been used in the assessment of one hospitalized H1N1pdm patient and revealed an intense inflammatory response [31] . The uptake of [ 18 F]-FDG in humans and animals suggests the predominant presence of activated neutrophils [32] [33] [34] [35] . In studies of mice infected with influenza A virus, neutrophils play a critical role in protection and recovery from infection, and participate in the process of adaptive immunity to the virus [36] . Coupled with molecular virology and pathology, molecular imaging with important probes of disease has enormous potential to reveal early critical factors that contribute to the clinical progression of illness as well as accelerate screening, the efficacy and mechanistic studies of vaccines and antiviral therapies [17, 18] . To test the hypothesis that FDG-PET/CT imaging could reveal the spatiotemporal nature of H1N1pdm inflammation and disease progression, we chose the ferret model of H1N1pdm influenza infection. Further, for these studies we chose a low passage clinical isolate, A/Kentucky/180/2010 or KY/180, herein, which has a change in the HA1 gene, D222G, which correlates with increased severity of disease in patient cases from several countries [37] [38] [39] [40] . In mice, infection with H1N1pdm engineered to include this change show increased viral titers and pathology, however, in ferrets there do not seem to be any major differences in clinical signs, transmissibility or pathogenicity [41, 42] . Our results show for the first time, the spatiotemporal progression of inflammation with CT and PET using [ 18 F]-FDG in ferrets infected with H1N1pdm in conjunction with histopathology and viral titers over a seven day period. Importantly, the extensive areas of patchy GGO and consolidation seen in the ferret model at 6 days post-infection (DPI) are similar to that reported for human H1N1pdm infections [31] . In vivo imaging with these modalities for anatomic (CT) and molecular (PET) data suggests increased pulmonary inflammation as the amount of circulating virus becomes undetectable. These results suggest that molecular imaging will be a great asset in gaining insight into the temporal and spatial progression of the inflammatory process caused by influenza virus infection. Due to the size of the Siemens Trimodal gantry for PET/CT imaging we chose four month old females rather than male ferrets. The pandemic H1N1 isolate KY/180 employed in these studies was isolated from nasal swab sample provided by the Severe Influenza Pneumonia Surveillance project, a clinical study of hospitalized patients with influenza pneumonia in Kentucky. The patient had a severe course of influenza disease and died after 19 days. Sequencing of the HA1 gene from this isolate revealed the D222G mutation, which has been associated with severe disease in human cases [37] [38] [39] [40] . The second passage of the KY/180 seed was employed in the characterization of infection in the female ferrets. The 50% infectious dose in female ferrets was determined to be 10 0.07 TCID 50 (data not shown). In group 1, six ferrets were mock infected with PBS. In group 2, six ferrets were infected intranasally (i.n.) with KY/180 with a 0.5 ml dose of 0.5610 5.7 TCID 50 per naris. Ferrets were monitored for temperature for 10 DPI, and for body weight and clinical symptoms for 28 DPI. Two animals in each group were euthanized on days 2, 14 and 18 to determine virus and HI titers in blood, lung and several additional organs at 2 DPI. In figure 1A , the body weight changes are shown for mock-and KY/180-infected ferrets. Body weight showed a drop on day 2 where the weight remained for the remainder of the study. Figure 1B shows the average temperatures of the mock-and KY/ 180-infected ferrets for the first 10 DPI. Temperature peaked on days 1 and 5 for KY/180 infected animals with a mean temperature of 103.4uF (SD = 1.71uF) and 103.2uF (SD = 0.52uF), respectively. The average hemagglutinin inhibition (HI) serum antibody titer in the blood on day 14 was 540 (reciprocal dilution) and the average total serum influenza-specific IgG by ELISA was 7610 (reciprocal dilution) (Fig. 1C) . In studies to determine the infectious dose, additional tissues were taken from animals infected at with KY/180 with a 0.5 ml dose of 0.5610 4.7 TCID 50 per naris. On day 2, viral titers were the highest in the nasal turbinates (10 6.25 TCID 50 /g) followed by the caudal lung (10 6.0 TCID 50 /g) and trachea (10 3.75 TCID 50 /g). The lowest levels of virus were observed in the cranial lobe of the lung (10 3.2 TCID 50 /g). Viral titers were measured by TCID 50 in nasal turbinates, trachea, right cranial lobe of the lung, right caudal lobe, brain, liver, spleen, kidney, duodenum, jejunoilieum, colon and rectum. KY/180 was detected in jejunum in one animal (10 2.0 TCID 50 /g), but was not detected in any other tissues (data not shown). Female Fitch ferrets were divided into five groups, with two per group of animals that were mock-infected with PBS (group 1) or intranasally infected with KY/180 (groups 2-5), in a 0.5 ml dose of 0.5610 6.0 TCID 50 per naris, on day 0 (Table 1) . Group 5 was the only group that was imaged each day; while groups 2-4 were imaged and sacrificed on days 1, 2 and 3 post-infection. This study design permitted evaluation of the progression of imaging with infection and pathology in two animals each day as well as continuous imaging of the lungs in one cohort over the seven-day time-period. FDG-PET and CT images of the H1N1pdm-infected and mock-infected ferrets were successfully obtained and fused for two ferrets on days 1, 2, 3 and 6 ( Fig. 2 and 3 ). Volumes of interest (VOI) and corresponding maximum standardized uptake values (SUVMax) were generated for any metabolically active lesions in the lung as well as background activity in the lungs, liver, heart, thymus, and thoracic lymph nodes. Baseline imaging prior to infection showed no focal areas of lung consolidation on CT and background standard uptake values (SUVMax) of the [ 18 F]-FDG levels ranged from 0.7-1.0 for PET ( Fig. 2A and Fig. 3A ). Each figure shows the one two-dimensional coronal plane that were standardized across days to provide a similar orientation and do not necessarily represent the SUVMax as that plane may be out of view. By 1 DPI, an area of consolidation was identified on CT in the right caudal lobe with corresponding radiotracer uptake on PET (Fig. 2B , SUVMax of 4.7). Consolidative areas in the right caudal lobe increased by day 2, with a persistently elevated SUVMax of 3.1(data not shown). By day 3, the consolidation increased in the right caudal lobe ( Fig. 2C and Fig. 3C , SUVMax of 3.7 and 4.4, respectively) and also appeared in the left caudal lobe (SUVMax of 3.2) of ferret 2214 (Fig. 3C ). By 6 DPI, there were widespread areas of patchy consolidation on CT with multiple areas of increased radiotracer uptake in both ferrets in caudal and cranial lobes ( Fig. 2D and Fig. 3D , SUVMax of 6.0 and 7.6 on the right, 4.2 and 4.6 on the left, respectively). These results suggest that inflammation progresses into the lower respiratory airways after infection into the upper part of the lower respiratory system. A ferret from the uninfected cohort was also imaged on day 6, with no focal appearance of consolidation on CT and no evidence of increased [ 18 F]-FDG uptake on PET (image not shown, background SUV of 0.6). To measure viral shedding, each day each ferret was swabbed in the nasal, throat and fecal passages and the viral titer was measured by TCID 50 ( Table 2 ). The highest levels of viral shedding were measured in the throat swabs. Nasal swabs also showed viral shedding for most animals, while the presence of virus in rectal swabs was low although detectable in a few animals. Replication of H1N1pdm in nasal turbinates and lungs were determined post-mortem from the right caudal lobe of the lung taken on euthanasia (Table 3) . Four sections were taken per lobe to provide greater insight into the spread of the virus in the tissue (Fig. 4) . High levels of virus were detected in all nasal turbinate samples at 1, 2, and 3 DPI (95% C.I. = 5.43+/21.00 TCID 50 / mL). Virus was also detected in a majority of lung sections from 1, 2, and 3 DPI. It was absent in the lung sections from one animal on 2 DPI, although it was present in the ferret's nasal turbinates, suggesting that the timing of infection in this animal was slower than the others. Of note, this same animal had a focus of consolidation on CT and radiotracer uptake on PET. This observation also suggests that, while the virus may have been undetectable by TCID 50 , low levels of virus had entered the lower respiratory system. No animals on day 7 post-infection had detectable virus in the nasal turbinates or the sampled lung tissue. Virus was not detected in nasal, throat and fecal passages or the sampled lung tissue from ferrets in any of the controls. Upon necropsy, all but the right caudal lobe of the ferret lung was fixed with paraformaldehyde. Following fixation, sections were taken for histopathology from the right and left cranial lobes, left caudal lobe and the middle accessory lobe. Representative photographs from slides of the left caudal lobe are shown in figure 5 . The ferrets in the control group had intact bronchiolar walls with very minimal infiltration by neutrophils with the exception of the left caudal lobe from control animal 2206 sacrificed on Day 1. Possible causes of this pattern of change may be an underlying systemic vasculopathy which is typically confirmed by evaluation of other organs that were not collected (e.g., kidney, spleen, liver). In general pulmonary lesions associated with influenza infection were roughly comparable at Days 1 and 2 and consisted of variable suppurative or necrosuppurative bronchiolitis and mixed cell alveolitis at minimal to moderate severity levels. By 1 DPI, there were some small foci of inflammation without much infiltration of the bronchi or bronchioles. There was an increased severity of inflammatory findings in lung lobes from infected ferrets on day 3. Specifically, more extensive infiltration of neutrophils can be seen within the bronchiolar lumen, along with necrosupprative bronchiolitis and mixed cell alveolitis. At Day 7, lesions observed in the lung lobes continued to exhibit an increased severity compared to the majority of lung lesions seen at 1 and 2 DPI. Bronchiolar epithelial hyperplasia and cytokaryomegaly were noted in addition to bronchiolitis. To evaluate potential correlations between PET/CT with histopathology, the SUVMax of lesions in the right and left lung of each ferret were compared with the cumulative histopathology scores assigned by the veterinary pathologist (Fig. 6 ). On average, the SUVMax was higher in the right lung than the left lung but the slopes and Spearman's correlation coefficients (r) were similar between the two sides. The highest correlation was seen between the cumulative bronchiolitis score and SUVMax (r of 0.71 and 0.75 on the right and left, respectively). The next highest was between the cumulative bronchitis score and SUVMax (r of 0.69 on the right and 0.67 on the left). A weaker positive correlation was seen between the cumulative alveolitis score and SUVMax (r of 0.47 on the right and 0.57 on the left). Herein, we show for the first time the feasibility of utilizing [ 18 F]-FDG PET coupled with CT imaging of H1N1pdm in ferret to track the progression of pulmonary disease in real-time. We chose a low passage clinical isolate, KY/180, which has a change in the HA1 gene, D222G. The D222G change in H1N1pdm correlates with increased severity of disease in patient cases from several countries [37] [38] [39] [40] . The patient from which we obtained the KY/180 isolate also had a severe course of influenza illness over a period of 19 days that resulted in death. Recently, studies in mice and ferrets infected with pandemic influenza viruses A/California/04/2009 and A/Netherlands/602/2009 engineered with the D222G mutation have shown that the D222G mutation are lethal in mice, but not ferret [41, 42] . The lethality in mice, but not ferrets, has been attributed to the greater abundance of a2-3-SA in the mouse model [42, 43] . All of these viruses have an affinity for a2,6-SAs associated with attachment to and replication in cells of the upper respiratory tract as shown by the high levels of viral replication in the nasal turbinates. Thus, infection of ferrets with these H1N1pdm isolates engineered with D222G and our clinical isolate have not correlated with clinical findings in patients. These results in ferrets are not surprising given that 80% of the fatal cases of H1N1pdm had underlying medical conditions and bacterial infections [6] . Discovery of the molecular components of the host response that may promote pathogenesis will be critical for defining new treatments. Noninvasive imaging can provide real-time in vivo monitoring of the progression of infection, inflammation and disease that may give insight into the mechanisms that modulate disease progression. Recently, Veldhuis Kroeze et al, presented data on the monitoring of pulmonary lesions of H1N1pdm influenza virusinfected ferrets with CT scanning which correlated with disease progression and severity [44] . As those studies demonstrate, CT is a powerful tool, but it will not give the molecular details that can be provided by PET or SPECT imaging of probes that target critical host responses such as neutrophil invasion. In our study we coupled CT scanning with the [ 18 F]-FDG radiotracer and show infection and inflammation of influenza infection in the lower respiratory system with foci of increased [ 18 F]-FDG uptake corresponding to areas of lung opacity on CT, with underlying inflammation on necropsy. In comparison to human CT imaging studies of influenza, the molecular images in the ferret show strong similarity. CT findings in patients with confirmed influenza infection show patchy ground-glass opacities in segmental multifocal distributions, mixed with areas of consolidation in the lung [12, 25, 45] . Moreover, the few case reports of human influenza in which lungs were imaged by [ 18 F]-FDG PET demonstrate areas of high uptake in these ground-glass opacities and consolidation [31] . Our study similarly demonstrates this pattern in the ferret model, also showing patchy opacities on CT with high uptake of radiotracer on PET, with necroscopy-based confirmation of inflammation in the left caudal lobe. Specifically, we show the ferret lung demonstrated progressive consolidation on CT and FDG uptake on PET predominantly in the right caudal lobe, which progressed to the left caudal lobe by day 3 p.i. By day 6, the diffuse metabolically active lesions seen on PET/CT were similar to what has been reported in the human literature during the 2009 H1N1 pandemic [31] . Histopathologic evaluation of the lungs confirmed the progressive nature of the pulmonary lesions and corroborated the radiologic data. Suppurative and necrosuppurative bronchiolitis seen on days 1 and 2 became progressively worse by days 3 and 7 post-infection. The inflammation tended to be patchy or multifocal and an entire lung lobe was never uniformly affected, corresponding to the multiple patchy lesions seen on PET/CT by the end of the study. This also agreed with the analyses of the viral titers in various sections of the right caudal lobe and the PET/CT imaging. Our analyses of viral titers in the four representative sections suggest different levels of infiltration of the lobe. The histopathologic scoring for bronchiolitis correlated the best with the SUVMax of the lesions seen in the right and left lungs on PET. In general, the severity of infection and inflammation on imaging can be represented by (1) the volume of affected lung (i.e. the percentage of diseased lung relative to total lung capacity), and (2) the extent of parenchymal destruction (disruption of pulmonary architecture) and inflammatory cell migration. Our study first aims to correlate FDG uptake measurements with histology, thereby analyzing the extent of parenchymal destruction and cellular infiltrates. It should be considered that there can be variation in matching FDG uptake with histologic severity because more severe architectural distortion can lead to necrosis with more dead cells, therefore showing less uptake of radiotracer among metabolically inactive dead cells and nonviable tissue. Our study, however, shows that progressing inflammatory infiltrates on histology in the studied time period after acute influenza infection corresponds to radiologic trends. Additionally, our study demonstrates spatial progression with increased size and number of abnormal foci in the lung parenchyma during acute infection. Ultimately, utilizing these new imaging tools, we envision a number of future experiments to delineate potential differences in the course of H1N1pdm and H5N1 infection in ferrets. We also plan to explore additional radiotracers that might reveal potential differences in host responses in the immune system and the process of acute injury in the lung. Future studies will assess differences in presentation of those who recover from infection versus those who eventually succumb to infection such as with more lethal isolates such as H5N1. This model should be valuable in rapid assessment of the effect of various treatments on pulmonary inflammation and damage. Finally, these first PET/CT imaging approaches could be extended to a number of other important pulmonary infections caused by pathogens such as hantaviruses, respiratory syncytial virus, and SARS CoV, to gain further insight into the spatiotemporal in vivo dynamics of disease progression [18] . The influenza H1N1pdm virus, A/Kentucky/180/2010, (KY/ 180; GenBank CY99332 and CY99333) was isolated from the nasal swab of a severe hospitalized case (hospitalized in March 2010) provided by the Severe Influenza Pneumonia Surveillance project, an ongoing clinical study of hospitalized patients with influenza pneumonia in Kentucky (courtesy of Dr. Julio Ramirez). The virus was isolated and passaged in the allantoic cavity of tenday-old embryonated hens' eggs at 37uC. The allantoic fluid was harvested 72 h after inoculation, pooled and stored in stored in aliquots at 280C until use. The infectious virus titer of the resulting seed stock was determined by TCID 50 (50% tissue culture infectious dose) and the titer calculated by Reed and Muench [46] and confirmed by plaque assay on MDCK cells. Passage E2 was used for the studies reported herein. Ferret studies were approved by the University of Louisville Institutional Animal Care and Use Committee. University of Louisville has Veterinary Medicine tasked to monitor and support all animal experiments. 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 will be conducted is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. All female Fitch ferrets were obtained from Triple F Farms (Sayre, PA). Ferrets were selected after screening blood samples for the presence of influenza antibodies using a hemagglutination inhibition assay (HI). Ferrets that were seronegative for seasonal and pandemic viruses were shipped directly to the University of Louisville Regional Biocontainment Laboratory and acclimated for seven days prior to initiation of the studies. Animals were fed Teklad Laboratory Diet #2072 (Harlan/Teklad, Madison, WI) and water ad libitum. For the characterization of the progression of infection of the KY/180 clinical isolate we utilized four month old, female ferrets. Prior to infection with virus, ferrets were anesthetized with 0.05 mg/kg atropine, 5.0 mg/kg ketamine, and 0.08 mg/kg dexmedetomidine intramuscularly. Subsequently, six animals were inoculated intransally (i.n.) with 0.5 mL of infectious virus per naris as a bolus, which was diluted to 10 5.7 TCID 50 /mL in phosphate buffered saline (PBS). Six additional animals were inoculated by i.n. with 0.5 mL of infectious virus per naris as a bolus with PBS (mock). Anesthesia was reversed with 0.4 mg/kg atipamezole. Ferrets were monitored daily for temperature and clinical symptoms. On day 2 post-infection, two ferrets were taken to measure viral titers in blood, lung, brain, trachea, nasal turbinates, spleen, kidney, thymus, liver, duodenum, jejuno-ileum, large intestine, and rectum. Gross pathology was defined during necropsy for the lung. At 14 and 28 days post-infection two additional ferrets from each group were analyzed for viral titers and pathology in the lung. For molecular imaging studies, twelve, four-month-old female Fitch ferrets were utilized. Animals were fed food and water ad libitum except 4 h prior to and during CT/PET imaging. On day 0, ferrets were anesthetized prior to inoculation with virus or PBS with ketamine, dexmedetomidine and atropine. Eight animals were inoculated i.n. with 0.5 mL of infectious virus per naris as a bolus, which was diluted to 10 6.0 TCID 50 /mL in phosphate buffered saline (PBS). Four ferrets were inoculated i.n. with 0.5 mL PBS per naris as a mock-infected control. For imaging, on days 0, 1, 2, 3 and 6, anesthetic induction and maintenance were achieved with 1-3% isoflurane. Blood glucose levels were checked prior to administration of the radiolabeled tracer to ensure that they were within normal limits, which typically range from 62-134 mg/dL in ferrets [47] . Glucose was provided to animals to compensate for body fluids lost during imaging. Typically 30 mls of Lactated Ringer's solution (Hospira) was administered subcutaneously (s.q.) following completion of the imaging. Animals were monitored for body temperature and vital signs during imaging. Imaging was performed on 0, 1, 2, 3, and 6 days post-infection (DPI). Each day, four ferrets were imaged with CT and PET on hardware designed for preclinical animal studies, including microCT and microPET, respectively. Two ferrets were euthanized the each day and necropsied to obtain tissue samples for virologic and histopathologic analyses (please see study design Table 1 ). Image acquisition was conducted with a Siemens Inveon Trimodal Scanner (Siemens Preclinical, Knoxville, TN), which is a small animal imaging platform that combines microPET, microCT, and microSPECT modalities within one unit. This combination facilitated co-registration of PET and CT images as the study subject was kept in a uniform position on the scanner bed, minimizing potentially large motion artifacts as a result of repositioning the animal between each scan. The Inveon microCT scanner features a variable-focus tungsten X-ray source with an achievable resolution of 20 mm and a detector with a maximum field of view (FOV) of 8.4 cm65.5 cm. The source-to-object distance was 263.24 mm and the source-to-detector distance was 335.67 mm. The Inveon PET detector provided an axial field of view (FOV) of 12.7 cm with a spatial resolution of 1.44 mm. PET images were reconstructed using a 2D-filtered backprojection algorithm with attenuation correction provided by microCT imaging. For the microCT scan, the following imaging settings were used: two bed positions, 80 kVp, 500 mA, 500 ms exposure time, and 464 binning. After each ferret underwent microCT imaging, the bed position was reset and microPET imaging with 18F-FDG (PETNET, Louisville, KY) began immediately. For each ferret, 2 mCi of 18F-FDG was administered (i.p.) with a 60-90 min uptake period. Radioactive dose was confirmed with an Atomlab 500 Dose Calibrator (Biodex Medical Systems Inc., NY). All imaging data were processed with PMOD software (v3.1; PMOD Technologies Ltd., Zurich, Switzerland). MicroCT data were received from the Inveon platform as DICOM files and PET data as microPET files. Scans were imported into the program's local SQL database with the units for the PET radiotracer in kBq/ cc. PET images were co-registered with the CT images with reslicing done as necessary to facilitate later calculations. For analysis of 18F-FDG levels, the standardized uptake value (SUV) was used. SUV is a widely used semi-quantitative measure that normalizes radiotracer uptake in a given region of interest based on body weight, and calculated for this study as follows: For all calculations, animal weights were expressed in kilograms and FDG activity in megabecquerels. For each image series, SUVs for each voxel were calculated using an external filter in PMOD, with the radionuclide half-life set at 6586.2 sec for 18F-FDG. For each pulmonary lesion, an ellipsoid volume of interest (VOI) was generated that encompassed the structure. Then, automatic isocontour detection was used to refit the VOI by setting a threshold of 50-60% of the difference between the maximum and minimum intensity SUVs in the ellipsoid VOI such that 0.5*(SUV max 2SUV min ). In cases where the automated threshold included contiguous structures in the VOI, manual refitting in conjunction with the co-registered CT scan was used to exclude those surrounding structures. For all VOIs, maximum SUV (SUV Max ) and average and standard deviation of all pixels in the volume (SUV Mean 6 SD) were calculated. For CT analysis, image interpretation was performed by a radiologist (in consultation with the scientific team) having more than ten years of diagnostic experience along with formal certifications by the American Board of Radiology (ABR) and the American Board of Nuclear Medicine (ABNM). Lesions on CT were identified using conventional criteria and terminology; Ground-glass opacity (GGO) is defined in this study as hazy increased lung opacity, with discernible underlying lung architecture such as visible bronchial and vascular structures, representing partial displacement of air in interstitial and alveolar airspaces; Consolidation is defined in this study as high density lung lesions (more dense than GGO) in which vascular and bronchial margins are obscured, representing complete displacement of alveolar air [27] . On days 1, 2, 3 and 6 and prior to euthanasia, swabs were taken from each ferret from nasal, throat and rectal regions. Following scheduled euthanasia, the nasal turbinates and the right caudal lobe of the lung from each ferret, which was divided laterally into four segments, were isolated. All swab and tissue samples were snap-frozen in liquid nitrogen and stored at 280uC until analyzed for virus titer by TCID 50 . Frozen tissues were weighed and diluted 10% weight per volume into cold DMEM with 1% penicillin/ streptomycin and 0.2% BSA before being homogenized and centrifuged to remove debris. Tissue homogenate and swabs were serially diluted 10-fold in DMEM with 2 mg/mL TPCK-Trypsin, 0.2% BSA, 4.5 g/L glucose, 1% penicillin/streptomycin, 2 mM L-glutamine, and 25 mM HEPES. Each sample was analyzed in quadruplicate following incubation in 96-well plates with Madin-Darby canine kidney (MDCK) cell monolayers at 37uC in 5% CO 2 for three days. Supernates were collected from each well were assayed for hemagglutination activity using 0.5% turkey red blood cells as an indicator of infection. Viral titers were expressed as log 10 TCID 50 /mL and were calculated using the Reed-Muench method [48] . The HI test quantitates serum antibody to influenza virus which can prevent agglutination of turkey RBCs (Fitzgerald Industries International Inc., MA). Heat-inactivated serum samples were treated with receptor-destroying enzyme (Sigma-Aldrich) for removing nonspecific inhibitors (followed by RBC adsorption) and were diluted 2-fold serially from initial dilution of 1:10. HA antigen (8 HA units in 25 mL) were added onto each well and incubated for 1 h at RT. Following antigen-antibody reaction, 50 mL of 0.5% turkey RBC were added to each well and incubated for 1 h at RT. HI negative wells were scored based upon a diffuse sheet of agglutinated RBCs covering the bottom. HI positive wells were scored if they showed a well circumscribed button of nonagglutinated RBCs. Lungs were inflated and stored in 10% neutral-buffered formalin. Three lung sections were placed into cassettes per lung section (right cranial, left cranial, left caudal, and right middle lobe) until they were trimmed, paraffin-embedded, and sectioned. Sections were mounted on glass slides and stained with hematoxylin and eosin for microscopic evaluation at Experimental Pathology Laboratories, Inc. by a veterinary pathologist. Sections were examined for the presence of abnormal findings including supprative and necrosupprative inflammation; epithelial hyperplasia and cytokaryomegaly; and fibrinous and exudative changes. Changes were graded with a standardized scale of 0-5, with 0 classified as ''not present'', 1 as ''minimal'', 2 as ''slight/mild'', 3 as ''moderate'', 4 as ''moderately severe'', and 5 as ''severe/high.'' For each ferret, a composite score for pathological changes was generated based on the locations in the respiratory tract (alveoli, bronchioli, bronchi) for statistical evaluation. All statistics were performed using R version 2.13.0 and GraphPad Prism 5. For each image, mean SUVMax and standard deviations were obtained. For each ferret, SUVMax values were correlated with histopathologic scoring using Spearman's Rho (r).

The Dynamics, Causes and Possible Prevention of Hepatitis E Outbreaks

Rapidly spreading infectious diseases are a serious risk to public health. The dynamics and the factors causing outbreaks of these diseases can be better understood using mathematical models, which are fit to data. Here we investigate the dynamics of a Hepatitis E outbreak in the Kitgum region of northern Uganda during 2007 to 2009. First, we use the data to determine that [Image: see text] is approximately 2.25 for the outbreak. Secondly, we use a model to estimate that the critical level of latrine and bore hole coverages needed to eradicate the epidemic is at least [Image: see text] and [Image: see text] respectively. Lastly, we further investigate the relationship between the co-infection factor for malaria and Hepatitis E on the value of [Image: see text] for Hepatitis E. Taken together, these results provide us with a better understanding of the dynamics and possible causes of Hepatitis E outbreaks.

Outbreaks of diseases such as avian influenza, SARS and West Nile Virus have alerted us to the potentially grave public health threat from emerging and re-emerging pathogens [1] [2] [3] . Many important infectious diseases persist on a knife-edge: rapid rates of transmission coupled with brief infectious periods. Such violent epidemic behavior has been observed in plague [4] , cholera [5] , pertussis [6] and more recently Hepatitis E. The recent outbreak of Hepatitis E in northern Uganda, has left many dead and a number of infectives that continue to spread the infection [7] . Hepatitis E is caused by infection with the Hepatitis E virus (HEV) which has a fecal-oral transmission route. It is a self-limiting disease but occasionally develops into an acute severe liver disease. As emerging and re-emerging infectious diseases increase in outbreak frequency, there is a compelling interest in understanding their dynamics [8] [9] [10] . The Kitgum outbreak, which we study here, has been linked to contaminated water or food supplies [11] . An assessment conducted by the Uganda Red Cross and district representatives in Agoro revealed that for a population of about 28,045 with 6,039 households mainly living in camps for internally displaced people in Potika as well as Agoro and Oboko satellite camps, the latrine coverage was as low as 3.7%. This means that there is one latrine for every 27 people. Further, only 23 boreholes were functional implying that the bore hole coverage is 0:38%, or one bore hole per 263 households. Another possible factor that could be implicated in the outbreak of Hepatitis E is its possible relationship with malaria. Malaria has been shown to disarm the immune system and increase susceptibility to viral infections such as HIV [11] . Recently, in a 3-month follow-up study the pattern of co-infection of Plasmodium falciparum malaria and acute Hepatitis A (HAV), in 222 Kenyan children under the age of 5 years was observed [12] . The incidence of HAV infections during P. falciparum malaria was found to be 6.3 times higher than the cumulative incidence of HAV, suggesting that co-infection of the two pathogens may result from changes in host susceptibility. There is also evidence both for [13, 14] and against [15, 16] an association between Hepatitis B viruses and malaria. HEV transmission route is similar to the Hepatitis A virus and thus for HEV it is important to consider possible links to co-infection with malaria. This can be done using mathematical models of multiple pathogens [17] [18] [19] [20] [21] [22] . In this paper, mathematical models are used to study the effects of both environmental conditions and malaria on Hepatitis E infections. The models designed are fit to data from the Kitgum outbreak, to estimate the basic reproduction number and to relate them to the level of contamination of the environment. We assume that the small number of latrines [23] , leads to contamination of environment. This in turn leads to contaminated water. Owing to the few number of bore holes in the region, lack of access to clean water gives rise to the viral infection of Hepatitis E. We formulate two mathematical models: one for Hepatitis Eonly and another for the co-infection with malaria, based on prior work as in [4, 24, 25] . In their framework, an individual is categorized according to their infection status and passes sequentially through the series of non infectious, infectious and recovered classes. A system of ordinary differential equations are then designed, analyzed and later fit to data from the Hepatitis E outbreak in Kitgum district to estimate desired parameters. First we model the epidemiology of Hepatitis E, an environmentally transmitted viral infection. The dynamics of the disease are an SEIR framework, i.e. Susceptible, Exposed, Infectious and Recovered. Hepatitis E virus is mainly spread by the fecal-oral route. This results either from directly touching the contaminated environment and eat without washing hands, or drinking contaminated water. In Kitgum district Uganda, most people live in internally displaced camps. The number of latrines in the area are not enough for the entire population, [11] , and people use the local environment for this purpose. When rain falls, it washes the faeces into water bodies. In the Kitgum region, few people have access to clean bore hole water [11] , and therefore collect water from the contaminated water sources. To model this phenomenon we use l, to denote the proportion of households in Kitgum with access to latrines. Therefore, the rate of change of contamination c of the environment is given by where r is the transmission rate of HEV from the infected human, I, to the environment. In the human population, susceptibles, S, are recruited at a rate m that equals to the per capita natural mortality rate for each group. This assumption is made to keep the population constant, while keeping a turnover of individuals in the population. We assume that a fraction b of the population has access to clean bore hole water and cannot become infected. Susceptible individuals without bore hole access become infected with the Hepatitis E virus at a rate bc, where b is the transmission rate of HEV from the contaminated environment c, to the human. This gives After successful infection, the individual is now exposed to HEV and moves to the exposed class E. The incubation period takes a mean period of 1 s days. The equation for this group is given by At the end of the incubation period 1 s , the individual becomes infectious and moves to group I. At this point, they display signs and symptoms that include fever, fatigue, loss of appetite, nausea, vomiting, abdominal pain, jaundice, dark urine, clay-colored stool and joint pain [26] . The infected individual may recover at a rate c. These dynamics are given by dI dt~s E{(mzc)I: Of the total infected individuals, a fraction p of them die due to the infection, and (1{p) recover to join the immune group R. This implies that Equations (1) to (5) provide a system of equations defining the transmission of HEV between a contaminated environment and humans. Assuming that the dynamics of the environment are fast. This means that the environment reaches a steady-state before the humans. The quasi-stationary-state (QSS) for c can be obtained from equation (1) to give This steady state is then substituted in the human equations to give a reduced system of equations as follows: dR dt~( 1{p)cI{mR: As in [27, 28] , in this system, 1~SzEzIzR. Thus, the last equation in (7) is redundant. This system of equations will be analyzed and fit to data. The endemic stationary state is given by where is the basic reproduction number for HEV. The term s mzs is the proportion of the exposed humans that survive the incubation period. The other fraction, br(1{l)(1{b) mzc is transmission rate of HEV during the infectious period of the human. The disease-free equilibrium point is stable if R 0 v1 (see Supporting Information S1) When R 0 w1 the endemic equilibrium point in equation (8) exists and is stable. This equilibrium is attained via oscillatory dynamics, with period T*2p is the mean age at infection, and G~1 mzs z 1 mzc is the ecological generation length of the infection. Figure 1a is a plot of the data for the outbreak from 2007 through 2009. To estimate model parameters and determine the critical level of control needed to eradicate the epidemic, the model described by the equations in (7) is fit to the data collected during the Kitgum outbreak ( Figure 1 ) During the invasion phase of HEV, the prevalence is approximately Taking the log of both sides of equation (10) and performing linear regression (details in Supporting Information S1) on this equation, (Figure 1b gives an estimated value R 0~2 :15. To determine R 0 when natural mortality is not equal to zero, (i.e. mw0), we use a non-linear differential equation fitting tool, called the PottersWheel Toolbox [29] . In this fitting technique, the c 2 value of the sum of the squares of the differences between the observed and fitted values is minimized by searching through different parameter values. We set m~0:0004, p~0:0169, b~0:0038, l~0:0037, and fit the free parameters, s,c, and the force of transmission br. We repeat this process 50 times to produce a range of best fits. The basic reproduction number for each run is then calculated using the expression in equation (9) The fitted parameters estimate the basic reproduction number R 0 between 2.08-2.39 with average 2.25. This value is similar to that found from the linear regression fitting. Figure 1c is a plot of the model outcome using the parameters generated from the fitting to the Kitgum outbreak. In addition to Hepatitis E, individuals in the Kitgum region were at a risk of acquiring malaria which is endemic to Uganda. To model possible co-infection we adopt the model to include a susceptible group which comprises both those with and without malaria. That is, the total susceptible population S9 = S+M where M is the proportion of individuals infected with malaria. The malaria dynamics will not be modelled in detail here but an assumption is made that malaria continuously invades the population, and individuals move back and forth between infection and recovery from the disease. This implies that The equilibrium state for this model is given by Clearly, this assumption provides a much simplified model when compared to a full model of vector-borne malaria [3, 12, 27] . Our concern here, however, is how background levels of malaria effect transmission dynamics of HEV. In Kitgum, at its lowest point during March 2009, 2,316 cases of malaria were reported out of a total population of 28,045 [7] . Thus 8.3% of the population are infected with malaria at any time, M Ã~0 :083. Recovery rate for malaria is r~0:1429 per week, and thus we set f~0:0129. Under the above assumption, equations (7) are rewritten to incorporate the malaria dynamics in equation (11) where j is a parameter that models change the increase (or decrease) in susceptibility to Hepatitis E of malaria infected individuals [12] . The other parameters are as defined in equations (7) and remain as defined there. We assume here that after exposure to HEV, both the susceptible and malaria infected groups join the exposed, E and subsequently the I group. In other words, individuals that harbor both infections are assumed to develop HEV symptoms at the same speed as those with only HEV. The dynamics of this model for standard parameter values are shown in Figure 2 . Using the next generation method as in van den Driessche and Watmough (2002), [30] , the basic reproduction number for Hepatitis E in presence of malaria is given by where R 0 is as defined in equation (9) When R C v1 infected individuals will have more chances of recovery than of transmitting the disease further hence the epidemic will die out. When R C w1, there exists an endemic equilibrium point as shown in Supporting Information S2 given by p . This implies that the endemic stationary point is attained via damped oscillations. The stability of this point would depend on the sign of the real part, a. If a.0, then the steady state is an unstable spiral, otherwise, it is a stable spiral. If ½j(mzf)z(mzr) 2 4jm(mzfzr) w1, then we have real roots, and stability of this equilibrium state will depend on the signs of these roots. If both are positive, the steady state is an unstable node; if both are negative, it is a stable node. If one of them is positive and the other negative, the steady state is a saddle point. Figure 2a shows the evolution of the malaria infected, M, the exposed, E, and the infected, I with time, while Figure 2b is a phase space portrait in the SI plane. From equation (14) it can be seen that the value for R C is determined by the proportions of susceptibles and malaria infectives in the population. Rearranging and assuming that SzM~1 this equation gives a criteria for an epidemic of This criteria is plotted in Figure 3a . As expected, if jw1 then presence of malaria increases the probability of an outbreak of Hepatitis E, while if jv1 the presence of malaria inhibits Hepatitis E. Assuming that malaria is at equilibrium in the population. (For example week of March 2009, with 2,316 malaria cases), gives M Ã~0 :083, then equation (16) gives a direct relation between j and R 0 . In fitting the model we note that the transmission rates br and j are not independent. Indeed, using PottersWheel to fit the co-infection model shows that the range of values for br is between 1.28-4.69, c between 1.01-1.75 and j values are between 0.02-13.27 (Figure 3b ) All of these values fall on line corresponding to R C between 2.19-2.48. This relationship follows the same curve as the analytical results in Figure 3a . To test potential interaction between Hepatitis E and malaria empirically, we now assume that in the absence of malaria, Hepatitis E has R 0~1 and does not spread. Thus malaria is required for the spread of Hepatitis and jw1. Since R C~2 :3 from the data and M Ã~0 :083, then substituting these values in to equation 14 gives j~16:9. This implies that, under the assumption of co-infection as the factor which promotes Hepatitis E, malaria infected individuals were infected with Hepatitis E up to 16.9 times more than those not infected with malaria. As we gain more information about the role of co-infection, this relationship can be used to improve estimation of br. As in [25, 27] , the criterion under which Hepatitis E will invade the population when malaria is endemic is derived in the Supporting Information S3. Thus, Hepatitis E virus invades if and the co-infection persists if j fzr(1{R 0 ) fR 0 : ð18Þ As in the Hepatitis E-only model, PottersWheel Toolbox is used to investigate the basic reproduction number R C , when natural mortality is not equal to zero, (i.e. mw0) A sequence of parameter estimates are generated, this time setting the fits in sequence to 20. The process is repeated until a set of 50 readings is obtained. The parameters are chosen in such a way that 0:1077vsv0:4777 (i.e. 15v1=sv64 days, the Hepatitis E incubation period, [26, 31, 32] ), and x 2 value is less than 65. The basic reproduction number for each run is calculated using the expression in equation (14) . The Global Burden of Disease (GBD) concept, first published in 1996, constituted the most comprehensive and consistent set of estimates of mortality and morbidity yet produced [33] . A GBD study aims to quantify the burden of premature mortality and disability for major diseases or disease groups, and uses a summary measure of population health, the DALY (Disability-Adjusted Life Years), to combine estimates of the years of life lost and years lived with disabilities. A DALY is defined as an indicator to quantify the burden of the disease and the functional limitation and premature mortality [34] . It can be used across cultures to measure health gaps as opposed to health expectancies, and the difference between a current and an ideal situation where everyone lives up to the age of the standard life expectancy, and in perfect health. In developing the DALY indicator, Murray and Lopez (1996) , [33] Since theDALY combines in one measure the time lived with disability, YLD, and the time lost due to premature mortality, YLL, then The YLL metric essentially corresponds to the number of deaths, P, multiplied by the standard life expectancy, L, at the age at which death occurs. Therefore, To estimate YLD on a population basis, the number of disability cases is multiplied by the average duration of the disease and a weight factor that reflects the severity of the disease on a scale from 0 (perfect health) to 1 (dead) The basic formula (without applying social preferences) for one disabling event is given by where I is the number of incidence cases, DW is the disability weight, and D is the average duration of disability. Since the reported cases are not specified according to age, the estimate will be done on a population basis. Typical symptoms of Hepatitis E include jaundice (yellow discoloration of the skin and sclera of the eyes, dark urine and pale stools), anorexia (loss of appetite), an enlarged, tender liver (hepatomegaly), abdominal pain and tenderness, nausea and vomiting, and fever and the disease may range in severity from sub-clinical to fulminant [35] . To calculate the YLD, we will set the disability weight to that for a diarrhea disease episode, (equal to 0.11, [33] ), in untreated or treated form. In developing the DALY indicator, additional social choices are taken into account. For example, is a year of healthy life gained now worth more to society than a year of healthy life gained sometime in the future? The DALY is an incidence-based measure, rather than a prevalence-based measure. Therefore, to estimate the net present value of years of life lost, a time discount rate to years of life lost in the future is applied, to adjust both costs and health outcomes [36] . Discounting health with time reflects the social preference of a healthy year now, rather than in the future. To do this, the value of a year of life is generally decreased annually by a fixed percentage, d. Therefore, equations (20) and (21) are respectively transformed to YLD~I According the WHO [35] , the life expectancy for a Ugandan male is 51 and 48 for a female. An average of 50 years will be used. The number of latrines and boreholes that would have prevented the Hepatitis E outbreak in Kitgum are calculated using our results in preceding sections. First, it is assumed that if the people had the necessary and sufficient number of latrines in addition to safe drinking, then the outbreak would not have occurred. Then, the costs of constructing the required latrines and boreholes are computed. From the results, the cost of saving one life from Hepatitis E, for one year is determined. The current number of latrines in Kitgum is 1,038 [11] . This implies one latrine per 27 people. According to the rules and regulations of Kampala City Council Authority, Building Inspection department, 1 latrine should be shared by a maximum of 5 people. We can use our estimated values of the basic reproduction number R 0 to determine the level of l and b that make R 0 v1. First, we use equation (9), the parameters in Table 2 , and the value of R 0~2 :11, (linear regression) Assume that contamination is due to insufficient latrines. Then, for R 0~1 , this method estimates that the latrine coverage should be increased to at least 17.1%. This translates into increasing the number of latrines from 1,038 to 4,796 (i.e. 3,756 extra latrines): 1 latrine per 7 people. Similarly, the boreholes should be increased from 23 to 230, that is, 17.7%, or 1 bore hole per 26 households. Similar results are obtained using R 0~2 :25 found from the non-linear fitting tool. In this case of Hepatitis E-only, the latrines should be increased to 16.1% and boreholes must cover 16.6% of the population. From the coinfection model, latrines should be increased to 17.5%: 4,908, (3,870 extra), 1 for 6 people. Boreholes should be increased to 18.1%, a total of 234, or 1 bore hole per 26 households. Our model suggests that to eradicate the epidemic, the minimum number of additional latrines required is 3,47. The average cost of digging and constructing a basic pit latrine is approximately USD 250.00 (quotation from city council official) Therefore, 3,477 would cost a total of USD 869,250.00. Thus, the cost per disability adjusted life year averted in Kitgum, in the case of Hepatitis E is 869,250/7,066 = USD 123.00. In addition to improving hygiene we should consider education. Let us now consider the case of education to the camp dwellers. Taking the simplest and cheapest scenario of hiring twenty (20) guidance and counseling officials to educate the dwellers about hepatitis E for about a month (that is 30) days, moving around the camp. Let us assign each counselor, 10 households per day. We then calculate the total amount in USD that would facilitate such an exercise as shown in Table 3 . From the calculations, it is seen that 104,000 USD would be required. Assuming the success of such an operation this translates into 104,000/7,066 = USD 14.71 cost per disability adjusted life year. The epidemic of HEV in Kitgum lasted a period of over two years [7] . Within this period, 160 individuals have lost their lives. As a result, the disease burden, the functional limitation and premature mortality have equaled a disability adjusted life years equal to 7,066, even allowing for the relatively low life expectancy in this part of Uganda. This paper provides a case study of how a simple epidemic model can be fit to such an outbreak disease. Two fitting methods have been used; the first, an analytical method and the other based on a freely available fitting tool. Using these methods, a reliable estimate of R 0 &2:2 has been provided. We then use the model to find the measures to keep R 0 v1. The necessary levels of latrine and bore hole coverages needed to eradicate the epidemic are both around 16 to 18%. Although the cost of construction of the required number of latrines is a one off cost, the benefits are large. Here we show what the benefits would have been in terms of protection against Hepatitis E. However, other diseases due to poor sanitation that have been reported in Uganda, such as cholera and dysentery, could be prevented in the same way [7] . [35] c Per capita rate of recovery from HEV 0.0238-0.1429/day [11] p The proportion that died during the outbreak 160 9449 [23] b The latent period of HEV 15v 1 s v64/day [26, 32] l The proportion of humans with latrines 3.7% [11] b The proportion of humans with bore hole water 0.38% [11] t

Human Bocaviruses Are Not Significantly Associated with Gastroenteritis: Results of Retesting Archive DNA from a Case Control Study in the UK

Gastroenteritis is a common illness causing considerable morbidity and mortality worldwide. Despite improvements in detection methods, a significant diagnostic gap still remains. Human bocavirus (HBoV)s, which are associated with respiratory infections, have also frequently been detected in stool samples in cases of gastroenteritis, and a tentative association between HBoVs, and in particular type-2 HBoVs, and gastroenteritis has previously been made. The aim of this study was to determine the role of HBoVs in gastroenteritis, using archived DNA samples from the case-control Infectious Intestinal Disease Study (IID). DNA extracted from stool samples from 2,256 cases and 2,124 controls were tested for the presence of HBoV DNA. All samples were screened in a real time PCR pan-HBoV assay, and positive samples were then tested in genotype 1 to 3-specific assays. HBoV was detected in 7.4% but no significantly different prevalence was observed between cases and controls. In the genotype-specific assays 106 of the 324 HBoV-positive samples were genotyped, with HBoV-1 predominantly found in controls whilst HBoV-2 was more frequently associated with cases of gastroenteritis (p<0.01). A significant proportion of HBoV positives could not be typed using the type specific assays, 67% of the total positives, and this was most likely due to low viral loads being present in the samples. However, the distribution of the untyped HBoV strains was no different between cases and controls. In conclusion, HBoVs, including HBoV-2 do not appear to be a significant cause of gastroenteritis in the UK population.

In 2005 a novel parvovirus was discovered in respiratory secretions of young children and was termed Human Bocavirus (HBoV-1) [1] . Other important members of the parvoviridae family include B19 which causes fith disease and human parvovirus 4 (Parv 4) which has not yet been associated with a disease [2] . Parvoviruses in animals are generally associated with systemic disease but also with respiratory and enteric symptoms [3, 4] . Since the discovery of HBoV-1 three other HBoV genotypes have been described, HBoV-2, HBoV-3 and HBoV-4. The association between HBoV-1 and respiratory disease has previously been well established [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19] . Although all HBoVs have also been detected in stool samples with prevalences ranging from ,1% to 20%, only HBoV-2 has been reported to be associated with symptoms of gastroenteritis [20] . Nevertheless, the role of HBoV2 as an aetiological agent of gastroenteritis has not been clearly confirmed, furthermore, to date, no clear association between the presence of HBoV-3 and HBoV-4 and disease has been established [21, 22] . Recent seroepidemiological studies indicate that exposure to HBoVs occurs early in life and 90% of the population are seropositive by the age of 5, although differences were reported in the seroprevalence of type-specific antibodies to the different HBoVs, which suggested that HBoV-1 infections are more prevalent [23] . The Infectious Intestinal Disease Study (IID Study) was a large case control study of gastroenteritis carried out in the UK between 1993-1996 [24] with the aim to determine the burden and aetiology of sporadic cases IID in the UK population. Initially, the use of classical microbiology diagnostic methods and electron microscopy (for virus detection) failed to detect a potential aetiological agent or toxin in 49% of the cases [25] . Retesting of the archived samples from this study using molecular methods for the detection of enteric viruses, bacteria and protozoa revealed viruses to be the most common aetiological agents of gastroenteritis,the diagnostic gap for IID was reduced to 25% from 49% [26] . The aim of the present study was to evaluate the role of HBoVs in IID in the UK population, using archived DNA samples from the matched case-control IID-1 study [26] . In addition, the presence specifically of HBoV-1, 2 or 3 was investigated in order to determine any possible associations between specific HBoV genotypes and IID. A total of 4,380 archived DNA from the IID study [26, 27] were tested for the presence of HBoV DNA. This archive comprised DNA extracted from stool samples from 2,256 cases and 2,124 controls. The qPCR assay targeted the NS1 gene (Ratcliff et al., unpublished method, personal communication) and was performed using an ABI Taqman7500. Oligonucleotide primer and probe sequences and positions are described in table 1. The reaction consisted of 0.1 M DDT (Invitrogen), 1X Platinum Quantitative PCR Supermix-UDG (Invitrogen), Pan-HBoV-F and Pan-HBoV-R primers each at a concentration of 100 mm, Pan-HBoV-NS1 probe at 10 mm concentration, ROX 25 mm (Invitrogen) 2.5 ml of template DNA and RNase free water to a final reaction volume of 25 ml. The amplification consisted of an initial denaturation at 95uC for 10 min, followed by 40 cycles with denaturation at 95uC for 15 sec, annealing at 55uC for 30 sec and extension at 60uC for 45 sec. HBoV-1, 2 and 3-specific primer pair and probes were designed in house through alignment of sequence data available in GenBank. The reaction conditions are as follows; 1X Platinum Quantitative PCR Supermix-UDG (Invitrogen), HBoV-NS1-1F, HBoV-1R, HBoV2-R and HBoV3R primers each at a concentration of 20 mm, the HBoV1,2 and 3 probe at 10 mm concentration, ROX 25 mm (Invitrogen), 2.5 ml of template and RNase free water was added to a final reaction volume of 25 ml. The amplification conditions for the typing assay are the same as those described in the HBoV NS1 detection assay above. Plasmids containing a 1773 bp and a 1737 bp region of the NS1 encoding gene of HBoV-1 and HBoV-3 respectively were used for assay optimisation and as controls. Control material was kindly provided by R. Ratcliff, Adelaide, Australia. The controls were also used in order to generate a standard curve for use with the pan-HBoV assay in order to allow for normalisation of the data generated including the comparison of relative sensitivities of the different assays and for quantitation of DNA present in each of the positive samples. The standard curve was generated using the plasmid containing a genome segment of the HBoV-1 and consisted of a series of 10 fold dilution containing from 300,000 copies/ml down to 3 copies/ml. Inter-and intraassay reproducibility was analysed by performing replicate testing of the standards in a single run (X11) and repeated runs (X2), and the standard curve was also included in each assay run for quality control and normalisation of results. A subset of 17 samples positive in the Pan-HBoV assay but which failed to amplify in the type-specific assays were confirmed using an alternative method published elsewhere [28, 29] , and 6 were further confirmed though direct sequencing of the amplicons obtained after purification either from solution or agarose gels using Agencourt AMPure (Beckman Coulter, USA) and GeneClean Spin kit (QBiogene), respectively, following manufactures protocols. The chi-squared test was used in order to evaluate the significance of differences observed between groups. For comparison of median values (analysis of CT values) the Mann Witney Utest was used. Prevalence Odds Ratio (POR = Pcases/(1-Pcases)/ Pcontrols/(1-Pcontrols)) was calculated in the total cohorts and by age group. Table 1 . HBoV-specific oligonucleotide primers and probes (all located at the NS gene). Sequence ( A total of 7.4% of the samples tested were positive for HBoV. No statistically significant differences were seen in the prevalence of HBoV between cases and asymptomatic controls, POR = 0.79 (Table 2) . Peak HBoV infection was observed in children under the age of 5, both in cases and controls, with significantly higher HBoV incidence in children between 1 and 4 in asymptomatic controls than in the cases of gastroenteritis (POR = 0.6; p,0.02). The number of HBoV positives in older age groups was too small for meaningful statistical analysis. The average CT values were 34.5 and 34.8, and the median CT values were 36.3 and 37.4 in cases and controls respectively (see distribution in Figure 1 ). The majority of HBoV-positives in both cases and controls had copy numbers ranging between 30 and 299 copies/reaction (or between 4.5610 3 and 4.5610 4 copies/ml of feaces). The distribution of HBoV viral loads between cases and controls was comparable and the median CT values between cases and controls were not significantly different (U-test; z = 0.458139, p.0.05). HBoV DNA was found in 149 (46%) samples in the absence of other co-pathogens (Table 3) . No statistically significant differences were observed in the proportion of cases or controls in which HBoV was found as a single organism or in the presence of one or more pathogens in the cohort as a whole, however in the 1-4 years of age group, HBoV in the absence of any other enteric pathogens was found in 29% of the cases, but in 56% of the controls (p,0.05). HBoV infections were detected year round although a peak was observed in the spring/early summer months, between April and June 1994 (Figure 2 ). HBoV DNA was found in 48.8% and 45.7% of female cases and controls, respectively. The distribution of HBoV among females and males was not significantly different from the distribution of females and males in the entire cohort which was 53% and 47%, respectively. A total of 106 (32.7%) HBoV positives were genotyped, whilst 218 (67.3%) remained untyped after testing in the HBoV types 1, 2 or 3 specific assays (Table 4 ). HBoV-1 detection was found predominantly in controls, (p,0.001) and HBoV-2 was predominantly associated with cases (p,0.01). The prevalence of HBoV-3 was not significantly different between cases and controls. HBoV-1 and -3 were predominantly found in children (Table 4) . HBoV-2 in the absence of any other pathogen was detected in 17 (81.3%) of the cases, compared to 9 (47.4%) of the controls. In cases, HBoV-2 was found across the age groups, although more frequently in children ,5, whereas in controls they were found predominantly in children ,5 with only 1 example in an adult ( Table 4 ). The prevalence of HBoV-2 in children ,5 years old was however not significantly different between cases and controls, 2.8% and 2.6%, respectively. A subset of HBoV that were negative in the type 1,2 or 3specific assays were confirmed in an alternative pan-HBoV PCR, and sequencing of a small number confirmed them as types 1, 2 or 3. The majority of the untyped samples (70%) had a CT value of .37 in the screening pan-HBoV PCR, indicative of low viral loads being present in the samples. This represents the largest study to date investigating the role and distribution of HBoVs infections in community acquired sporadic gastroenteritis and in asymptomatic controls. The prevalence of HBoV infection in the UK population was found to be 7.4% across all ages, with a higher percentage of the infections occurring in children ,5 years of age (19%). However, the prevalence of HBoV infections was comparable in cases of gastroenteritis and in age-matched asymptomatic controls. Although the presence of enteric pathogens, eg norovirus or rotavirus, in asymptomatic individuals is well documented, a significantly higher prevalence of the pathogen is seen in cases than in the controls [26] . Therefore, our data suggests that HBoV are not causally associated with gastrointestinal disease in the UK population as a whole, nor in children. The prevalence of detection of HBoV in stool samples in previous studies varies widely (see summary in Table 5 ), but most coincide in reporting the highest prevalence in children. HBoV infections were detected all year round in the UK although a tentative peak was observed in the spring/early summer months in 1994 (between April and June). Different seasonal patterns in the peak prevalence of HBoV have been reported in different countries (see Table 5 ), Of the 324 HBoV positive samples, 106 (32.7%) were genotyped in the type-specific assays. HBoV-1 was found predominantly in controls (p,0.001) and the prevalence of HBoV-3 was similar in cases and controls. Both HBoV-1 and -3 were predominantly found in children. HBoV-2 was predominantly associated with gastroenteritis cases (p,0.01). The overall prevalence in cases was 1.4% and 0.8% in controls, however, in children ,5 year of age, the prevalence in cases and controls was similar, 2.8% and 2.6%, respectively. The prevalence of HBoV-2 in children in the UK was significantly lower than that reported in a study in Australia, in which HBoV-2 was detected in 17.2% and 8.1% of the cases and controls, respectively [22] . The findings of the study in Australia lead to the proposal of HBoV-2 as an important aetiological agent of infantile gastroenteritis. It is noteworthy however, that in the Australian study, the association of HBoV-2 with gastroenteritis was only significant when cases with a bacterial co-pathogen were included in the analysis. Although in the present study HBoV-2 in the absence of other enteric pathogens was found more frequently in cases than in controls, the small numbers found in such large study suggest that the role of these viruses in IID, if any, is likely to be small. A lack of correlation between HBoVs or HBoV-2 and paediatric gastroenteritis was also reported in several smaller studies published elsewhere [30, 31, 32] . A total of 67% of the HBoV-positive samples could not be genotyped using the genotype-specific PCR assays. The majority of these untyped samples (70%) had CT values .37. This suggests that failure to type may be associated with low viral loads and differences in the relative sensitivities of the genotyping assays compared to the detection assay. Although under experimental conditions and using plasmid controls the sensitivities of all assays were comparable, it is likely that when applied to true clinical samples the sensitivity of the type-specific assays was inferior, possibly due to as yet not identified strain variability within genotypes. Also, a HBoV type 4-specific assay was not included in this study, therefore, any possible HBoV4 infections would not have been typed. Of the panel of samples that were tested in an alternative pan-HBoV PCR, the strains typed through sequencing were HBoV-1 (2 samples), HBoV-2 (1 sample) and HBoV-3 (3 samples). Furthermore, the distribution of untyped HBoVs was not significantly different in cases and controls. HBoVs in the absence of other enteric pathogens were seen in 46% of the HBoV-positive samples, and more frequently in the controls, 50.3% vs 40.9% in cases. No significant difference in HBoV load was observed between cases and controls, or between the samples positive for HBoV alone or in the presence of other pathogens. Previous studies have investigated the relationships between viral load and disease severity [33, 34, 35, 36, 37] . In respiratory infections significantly higher HBoV loads were seen in samples collected from children positive for HBoV alone than in those from children with co-infections. In respiratory infections also, viral loads .10 4 were associated with disease, whereas loads ,10 4 were associated with asymptomatic children [33, 38] . This lead to the suggestion that higher viral loads are indicative of a causative role of HBoV in respiratory infections [33, 38] . However, Brieu et al [38] found no significant correlation between viral load and clinical symptoms or disease severity. In conclusion, the results obtained from investigating for the presence of HBoV DNA in archived DNA samples from a large and previously well described case-control study of IID suggest that HBoV, including HBoV-2,do not appear to be a significant cause of gastroenteritis in the UK population, and particularly in the paediatric population. Although HBoVs are relatively frequent across all ages, and in particular in preschool age children, they are found just as frequently among children and adults without symptoms of gastroenteritis.

Structural Origins for the Loss of Catalytic Activities of Bifunctional Human LTA4H Revealed through Molecular Dynamics Simulations

Human leukotriene A4 hydrolase (hLTA4H), which is the final and rate-limiting enzyme of arachidonic acid pathway, converts the unstable epoxide LTA4 to a proinflammatory lipid mediator LTB4 through its hydrolase function. The LTA4H is a bi-functional enzyme that also exhibits aminopeptidase activity with a preference over arginyl tripeptides. Various mutations including E271Q, R563A, and K565A have completely or partially abolished both the functions of this enzyme. The crystal structures with these mutations have not shown any structural changes to address the loss of functions. Molecular dynamics simulations of LTA4 and tripeptide complex structures with functional mutations were performed to investigate the structural and conformation changes that scripts the observed differences in catalytic functions. The observed protein-ligand hydrogen bonds and distances between the important catalytic components have correlated well with the experimental results. This study also confirms based on the structural observation that E271 is very important for both the functions as it holds the catalytic metal ion at its location for the catalysis and it also acts as N-terminal recognition residue during peptide binding. The comparison of binding modes of substrates revealed the structural changes explaining the importance of R563 and K565 residues and the required alignment of substrate at the active site. The results of this study provide valuable information to be utilized in designing potent hLTA4H inhibitors as anti-inflammatory agents.

Leukotriene cascade is associated with the biosynthesis of variety of leukotrienes (LT) from the phospholipids of the nuclear membrane of the leukocytes [1] . The LTs are a group of lipid mediators associated with acute and chronic inflammatory diseases such as asthma, rhinitis, psoriasis, chronic obstructive pulmonary disease, and atherosclerosis [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] . Cytosolic phospholipase A2 (cPLA2) hydrolyzes the ester bond present in sn-2 position of phospholipids and yields lysophospholipids and free fatty acid, arachidonic acid (AA) [1, 14] . This increases the level of free AA available for the synthesis of inflammatory leukotrienes upon the action of more enzymes. The enzyme 5-lipoxygenase (5-LO) assisted by file-lipoxygenase-activating protein (FLAP) converts the AA into the highly unstable allylic epoxide, leukotriene A4 (LTA4) [15] [16] [17] [18] [19] [20] [21] . This unstable intermediate is converted into two different products LTB4 and LTC4 by the action of two different enzymes LTA4 hydrolase (LTA4H) and LTC4 synthase (LTC4S), respectively [1, [22] [23] [24] [25] . The LTC4 is subsequently converted to LTD4 and LTE4 substances by the action of different enzymes. All of these LTB4, LTC4, LTD4, and LTE4 are powerful proinflammatory mediators [1, 26] . The LTA4H, which catalyzes the conversion of LTA4 to the chemotactic agent LTB4, was identified as a bi-functional enzyme capable of processing two highly diverse substrates such as LTA4 (a fatty acid) and peptide through its epoxide hydrolase and aminopeptidase activities [27, 28] . This enzyme was first discovered for its epoxide hydrolase activity and later for its aminopeptidase activity based on the presence of consensus Zn binding motif (HEXXH-X 18 -E), which was found in M1 family of Zn containing aminopeptidases [29] [30] [31] [32] . The natural peptide substrate for this enzyme is still not known but preference is shown over arginyl di-and tripeptide and can selectively be blocked by the mutation of either E296 or Y383 residues [33] [34] [35] [36] . Upon the determination of LTA4H crystal structures it was revealed that this enzyme is composed of three domains, a fully beta N-terminal domain, a mixed alpha/beta catalytic domain, and a fully alpha-helical C-terminal domain ( Figure 1 ) [37] [38] [39] [40] [41] [42] . In terms of the hydrolase activity of the enzyme, D375 from a narrow hydrophobic pocket is specifically required as it is involved in the nucleophilic attack targeting C12 atom of LTA4 [43] . In addition, this residue belongs to the peptide K21 (L365-K385) segment identified by Lys-specific peptide mapping of suicide inactivated LTA4H. The carboxylate moiety of LTA4 was observed to form direct electrostatic interactions with the two positively charged conserved R563 and K565 residues present at the entrance of the active center [28, 44] . These interactions are very much essential in aligning LTA4 along with the catalytic elements of the active site. Based on the mutagenic experiments, E271 residue from another conserved GXMEN motif in the family of zinc peptidases was found to be important for both the functions of the enzyme [14] as the mutagenic replacements abrogated both the activities. A crystal structure of LTA4H with E271Q mutation has revealed only minimal conformational changes and did not explain the loss of enzyme function [14] . It was also suggested that the carboxylate of E271 participates in an acid-induced opening of the epoxide moiety of LTA4 and as N-terminal recognition site in terms of peptide substrates [14, 26, 45] . Some mutagenic experiments have also reported the critical role of R563 residue in epoxide hydrolase reaction by positioning the carboxylate tail along the catalytic elements of the active site [28, 34] . In aminopeptidase reaction, both R563 and K565 residues co-operate each other to ensure the necessary binding strength and productive alignment of the substrate. Altogether, R563 plays important roles in both the functions of the enzyme whereas K565 residue assists R563 in catalyzing the peptide substrate but not in hydrolase reaction, which catalyzes the fatty acid substrate [27] . These residues were reported to be the common carboxylate recognition site for both lipid and peptide substrates in the active site of LTA4H [27] . Mutations of R563 to any other amino acid including a conservative replacement of R by K preserving the positive charge abolished the enzyme function but exhibited a significant residual aminopeptidase activity [27] . Crystal structure with R563A mutant could not reveal any structural changes explaining the complete loss of catalytic activity. It was also reported that esterified LTA4 cannot be the substrate of the enzyme and this phenomenon was explained with the steric hindrance [46] , which also proved that a free carboxylate group of LTA4 is critical for the hydrolase function. The other carboxylate recognition residue K565 located in a way that it can also involve in carboxylate recognition but its mutagenic replacements have not decreased the epoxide hydrolase activity [27] . The difference in observed aminopeptidase activities between the wild type and K565 mutants have suggested that K565 is a carboxylate binding site for peptide substrates also [27] . The information of a binding pocket for its ligand is very important for drug design, particularly for conducting mutagenesis studies [47] . In the literature, the binding pocket of a protein receptor to a ligand is usually defined by those residues that have at least one heavy atom (i.e., an atom other than hydrogen) with a distance from a heavy atom of the ligand. Such a criterion was originally used to define the binding pocket of ATP in the Cdk5-Nck5a* complex [48] that has later proved quite useful in identifying functional domains and stimulating the relevant truncation experiments [49] . The similar approach has also been used to define the binding pockets of many other receptor-ligand interactions important for drug design [50] [51] [52] [53] [54] . This study has focused on structural changes and binding mode differences between wild types and mutant forms of hLTA4H-fatty acid and -tripeptide substrate complexes. Molecular dynamics (MD) simulations of two wild types of the enzyme-substrate complexes and three mutations including E272Q, R563A, and K565A were studied with LTA4 (fatty acid) and Arg-Ala-Arg (RAR, tripeptide) substrate ( Figure 2 ). The binding mode comparison of enzyme-substrate complexes revealed essential structural differences, which were not shown in X-ray structures, explaining the loss of catalytic functions of the enzyme. The overlay of binding modes of LTA4 and RAR substrates has proved the previous assumption of similar but overlapped active site is used in both the functions of the enzyme. The results from this study provide valuable information over the way the two functions of the enzyme are exerted over two substrates of diverse nature. In addition, the structural information obtained from this study can be utilized in structure-based hLTA4H inhibitor design as inhibition of the hydrolase and aminopeptidase functions will lead to the development of anticancer and anti-inflammatory drugs, respectively. As this study was aimed at investigating the binding of fatty acid and peptide substrates of the enzyme, independent hLTA4H-LTA4 (L-LTA4) and hLTA4H-RAR (L-RAR) complexes were prepared. The LTA4 of AA cascade is the natural substrate of the enzyme but no 3D structural information was available for its binding from X-ray studies. Some previous studies have reported the proposed binding of this fatty acid substrate at the active site of the enzyme. In this study, molecular docking methodology using GOLD 5.0.1 program was employed to obtain the reliable binding mode of LTA4. The GOLD program from Cambridge Crystallographic Data Centre, UK uses a genetic algorithm to dock the small molecules into the protein active site [46, 55] . The GOLD allows for a full range of flexibility of the ligands and partial flexibility of the protein. One of the crystal structures of hLTA4H bound with most active inhibitor (PDB ID: 3FUN) solved at high resolution was used in molecular docking. The bound inhibitor of this crystal structure was removed and the apoform of this enzyme was subjected to MD simulations with the parameters discussed below. The representative structure with a RMSD value close to the average structure of last 3 ns of the 5 ns MD simulations was selected and utilized in molecular docking experiments. The active site was defined with a 10 Å radius around the bound inhibitor. The ten top-scoring conformations of every ligand were saved at the end of the calculation. Early termination option was used to skip the genetic optimization calculation when any five conformations of a particular compound were predicted within an RMS deviation value of 1.5 Å . The GOLD fitness score is calculated from the contributions of hydrogen bond and van der Waals interactions between the protein and ligand, intramolecular hydrogen bonds and strains of the ligand [56] . Protein-ligand interactions were analyzed using DS and Molegro Molecule Viewer [57] . The best pose was selected based on the molecular interactions and the distance between epoxy group of LTA4 and metal ion (Zn 2+ ) present in the active site as well as the location of its carboxylate group which interacts with the carboxylate recognition residues R563 and K565. Finally, the enzyme-LTA4 complex was prepared to be used in further steps in this study. In terms of preparing the enzyme-peptide complex to investigate the aminopeptidase function of LTA4H enzyme, 3D coordinates of the bound tripeptide Arg-Ala-Arg (RAR) in a solved X-ray structure of hLTA4H with a mutation E271Q (PDB ID: 3B7T) was utilized. The Superimpose Structures protocol as available in Accelrys Discovery Studio 3.0 (DS) was employed to copy the 3D coordinates of this tripeptide into the representative structure of LTA4H picked from the 5 ns MD simulation by superimposition. This complex was subjected to energy minimization using Energy Minimization protocol of DS before considered further in this study. Many marvelous biological functions in proteins and DNA and their profound dynamic mechanisms, such as switch between active and inactive states [58, 59] , cooperative effects [60] , allosteric transition [61, 62] , intercalation of drugs into DNA [63] , and assembly of microtubules [64] , can be revealed by studying their internal motions [65] . Likewise, to really in-depth understand the action mechanism of receptor-ligand binding, we should consider not only the static structures concerned but also the dynamical information obtained by simulating their internal motions or dynamic process. To realize this, the MD simulation is one of the feasible tools. Initial coordinates for the protein atoms were taken from the wild type (WT) and mutant forms of both L-LTA4, L-RAR complex structures. Mutations were introduced at E271Q, R563A, and K565A of the enzyme based on the previous experimental reports to investigate the single active site that catalyzes two different functions upon diverse substrates [14, 45] . The protonation states of all ionizable residues were set to their normal states at pH 7. Eight MD simulations were performed for systems including WT and mutant forms of L-LTA4 and L-RAR complexes ( Table 1 ). All MD simulations were performed with GROMOS96 forcefield using GROMACS 4.5.3 package running on a high performance linux cluster computer [66, 67] . During the MD simulations, all the protein atoms including divalent metal ion (Zn 2+ ) were surrounded by a cubic water box of SPC3 water molecules that extended 10 Å from the protein and periodic boundary conditions were applied in all directions. The systems were neutralized with Na + and Cl 2 counter ions replacing the water molecules and energy minimization was performed using steepest descent algorithm for 10,000 steps. A 100 ps position restrained MD simulations were performed for every system followed by 5 ns production MD simulations with a time step of 2 fs at constant pressure (1 atm), temperature (300 K). The electrostatic interactions were calculated by the PME algorithm and all bonds were constrained using LINCS algorithm. A twin range cutoff was used for long-range interactions including 0.9 nm for van der Waals and 1.4 nm for electrostatic interactions. The snapshots were collected at every 1 ps and stored for further analyses of MD simulations. The system stability and behavior of the catalytic structural components present in every system were analyzed using the tools available with GROMACS 4.0.5 and PyMol programs. Surprisingly, the LTA4H enzyme catalyzes both hydrolase and aminopeptidase functions over fatty acid and peptide substrates utilizing the same active site. In order to obtain deeper insight upon this unique characteristic of the enzyme, a set of MD simulations were performed with WT and mutated enzymesubstrate complex structures. The natural epoxy substrate LTA4 of arachidonic acid pathway, which is converted to LTB4 upon the action of the enzyme, was selected as fatty acid substrate to investigate the hydrolase function of the enzyme. In the other hand, RAR tripeptide that is reported to be the preferred peptide substrate of the enzyme [14] was selected to investigate the aminopeptidase function of the enzyme. The L-LTA4 complex was prepared through the molecular docking methodology whereas L-RAR complex was prepared by copying the 3D coordinates of RAR from the X-ray crystal structure of LTA4 ( Figure 3 ). The representative structure obtained from the MD simulation of LTA4H-apoform was used in preparing both the complex structures to compare the structural changes effectively with no artifacts. From the reported site-directed mutagenesis experiments, three amino acid residues from the catalytic active site of the enzyme were predicted to be very important for the enzymatic activities of the enzyme. These residues include one negatively charged E271 residue from the central catalytic domain and two positively charged R563 and K565 residues from the Cterminal domain. Studies mentioned that mutation of this negatively charged residue to a neutral glutamine (E271Q) has completely abrogated both the catalytic activities of the enzyme. But the crystal structure solved with this mutation (PDB ID: 1H19) could not report any structural or conformational changes causing this drastic change in the catalytic activity [14] . It was also proposed that metal (Zn 2+ ) ion present close to the epoxide moiety of LTA4 acts as a weak Lewis acid to activate and open the epoxide ring. It continued to explain that the E271 located in proximity to Zn 2+ also participates in this acid-induced opening of the epoxide ring of LTA4. It was also reported that E271 acts as the N-terminal recognition point in stabilizing the peptide substrates in terms of aminopeptidase reaction of the enzyme. The positively charged residues R563 and K565 were predicted as the carboxylate recognition sites for both the substrates of the enzyme. All mutations of R563 including R563K, which preserved the positive charge, have resulted in complete loss of catalytic functions of the enzyme. The R563K mutant has shown a significant aminopeptidase activity. All these mutations of R563 leading to tremendous change in enzyme functions did not reveal any structural changes explaining the loss of activities. Especially in epoxide hydrolase reaction, the role of R563 was presumed to position the carboxylate tail of the substrates along the catalytic components of the active site [45] . In terms of aminopeptidase reaction, both the positively charged (R563 and K565) residues help each other in aligning the substrate with the catalytic elements and maintaining the binding strength. The mutations K565A and K565M lacking the positive charge have reduced the aminopeptidase and revealed that this positively charged residue assists R563 in carboxylate recognition in aminopeptidase reaction [27] . Despite of this information over the importance of E271, R563, and R565, the structural changes explaining the catalytic activities of the enzyme are lacking. The results of MD simulations of WT and mutant enzyme-substrate complexes discussed in this study will provide a deeper insight from the structural perspective. Overall structural stability of the systems The overall stability analyses are considered important to note that the systems did not undergo any unusual changes during the time scale of simulation because of erratic system preparation. In this study, root mean square deviation (rmsd), root mean square fluctuation (rmsf), and intra-molecular hydrogen bonds were used in analyzing and comparing the stability of the systems under study (Figure 4) . In terms of L-LTA4 systems, the calculated average rmsd value of K565A system was 0.158 nm which is lower than the rmsd values of WT (0.189 nm), E272Q (0.186 nm), and R563A (0.202 nm) systems ( Figure 4A ). The R563A system has shown the higher rmsd value and thereby indicating the additional effect of this particular mutation. Despite of these small differences in the rmsd values between the L-LTA4 systems, the systems were stabilized well throughout the timescale of simulation. As another method to investigate the stability of the systems, rmsf values of all systems were calculated during the simulation and plotted. From the plot, it was observed that none of the active site residues were fluctuating higher than 0.2 nm and explained the stable nature of the systems over the time scale of simulation. In addition, the number of intramolecular hydrogen bonds were calculated for all the systems and plotted. The average number of hydrogen bonds revealed that R563A system has formed more number of hydrogen bonds (466.9) compared to WT (452.8), E272Q (451.7) and R565A (450.6) systems which displayed the reduced number of hydrogen bonds. This result also has confirmed the stability of the systems despite of small differences between systems ( Figure 4A ). In L-RAR systems, the average rmsd value of K565A (0.212 nm) was higher than other systems, which is completely contrasted to the rmsd value of equivalent L-LTA4 system whereas the R563A system has shown the average lower rmsd value (0.168 nm). The other two systems, WT and E272Q, have shown the same average rmsd value of 0.189 nm from last 3 ns of the simulation time ( Figure 4B ). In terms of rmsf calculations, the rmsf plot has shown that except D375 and Y378 residues of the active site all other important active site components were stable throughout the simulation. This high fluctuation of these two residues was mainly observed in K565A system. The average number of intramolecular hydrogen bonds during last 3 ns of the simulation was very similar in all the systems. At the end of the simulation time, the R563A and K565A systems started losing their hydrogen bonds and thereby became less intact compared to other systems. All the systems were investigated for the stability and found to be well stabilized during the simulation. Thus the representative structures close to the least rmsd value of each system was obtained and used in structural comparison. The distance between the most important metal (Zn 2+ ) ion and oxygen atom of the epoxy ring in case of L-LTA4 systems is very important for the hydrolase function of the enzyme. The average distance value observed in WT system (0.56 nm) was lower compared to the mutant systems. Among mutant systems R563A system has shown the higher average distance value of 0.78 nm whereas E271Q and K565A systems have shown similar average distance value of 0.69 nm and 0.63 nm, respectively. During the end of the simulations, this distance in E271Q and K565A systems has reduced close to the distance of WT but R563A has maintained higher distance until the end ( Figure 5A ). In terms of L-RAR systems, the distance between the same metal ion and carbonyl oxygen atom of N-terminal peptide bond was measured and compared between the L-RAR WT and mutant systems. As observed in L-LTA4 systems, WT of L-RAR systems has maintained the lower average distance value of 0.59 nm whereas Figure 5B ). The hydrogen bonds formed between the protein and bound ligands were also calculated for all the systems under study to investigate the molecular interactions that are lost during the mutations. In L-LTA4 systems, WT system has formed high number of average hydrogen bonds (5.7) compared to the mutant systems. The K565A system has shown an average number of hydrogen bonds of 5.0 whereas E271Q and R563A systems have formed only 1.7 and 1.8 average number of hydrogen bonds with the bound ligand ( Figure 6A ). This reduced number of hydrogen bonds correlate well with previously reported loss of hydrolase activity of the enzyme in E271Q and R563A mutated systems [14, 45] . In terms of L-RAR systems, the E271Q system has formed more number of average hydrogen bonds (8.7) with the bound RAR. Whereas the WT system has shown an average hydrogen bond value of 8.0, the other mutant systems R563A and K565A have shown the average hydrogen bond values of 4.3 and 3.9, respectively ( Figure 6B ). This result of number of hydrogen bond values observed between protein and RAR has proved the importance of both R564 and K565 to maintain the substrate alignment in the active site and the binding strength of the substrate. But the high number of observed hydrogen bonds in E271Q system does not correlate with the observed loss of catalytic activity due to the mutation of E271. This indicated that other structural disturbances script the loss of catalytic function of the enzyme. L-LTA4 systems. In L-LTA4 systems, the binding mode of LTA4 in WT system has formed hydrogen bonds with Y383 through the oxygen atom of epoxy ring and the distance between this oxygen atom and the metal ion was maintained in a closer distance compared to that of mutant systems. The carboxylate group of LTA4 in WT system was well recognized by both the positively charged residues R563 and K565 which are reported to be the carboxylate recognition sites. Because of this recognition the carboxylate group has formed strong hydrogen bond interactions with R563 and K565. The distance between the carboxylate of E271 and metal (Zn 2+ ) ion was maintained in close vicinity for the reported acid-induced catalytic reaction. The other residue Y378 has also formed a hydrogen bond interaction with the carboxylate group of LTA4 ( Figure 7A ). In E271Q system, the binding mode of LTA4 is so different from that of WT system. Because of the uncharged nature of the mutation (E271Q) the metal ion has slightly moved towards the carboxylate group of LTA4, which has also mutually moved towards the metal ion. This movement of carboxylate group of LTA4 has moved the central epoxy group of LTA4 further down and made it inaccessible by the catalytic metal ion ( Figure 7B ). This change in the E271 and distance between Zn 2+ and epoxy group (0.69 nm) can be directly correlated with the loss of activity ( Figure 5A ). The hydrogen bonds formed with R563 and K565 residues were completely lost because of this mutation. In R563A system, though E271 residue has maintained its hold on the metal ion because of the absence of R563, the important carboxylate recognition site, the LTA4 has moved backwards into the hydrophobic cavity formed by hydrophobic residues such as W311, F362, K364, L365, V366, V367, and V381 ( Figure 7C ). This change has not only brought the distance between Zn 2+ and the epoxy ring of LTA4 higher (0.78 nm) compared to any other L-LTA4 systems but also made hydrogen bonding with K565 impossible. In terms of K565A system, the binding mode of LTA4 was similar to that of WT system. Regardless of mutated K565 the hydrogen bonds were maintained with R563 but still the distance between Zn 2+ and epoxy ring of LTA4 was higher (0.68 nm) in this mutant system as well. The hydrogen bond between LTA4 and Y378 was also maintained as observed in WT system ( Figure 7D ). This observation also correlated the experimental observation that K565A mutation reduces the activity but does not abolish it. L-RAR systems. The binding modes of the tripeptide RAR in all systems were observed and compared to investigate the changes due to the mutated residues. More number of hydrogen bonds was observed between the protein and bound substrate compared to the L-LTA4 systems because of the high number of polar hydrogen in the peptide substrate. In the WT system, strong molecular interactions were observed through the hydrogen bonds and p-cation interactions formed between protein and substrate ( Figure 8A ). The C-terminal carboxylate which is equivalent to the carboxylate of LTA4 has formed strong hydrogen bond interactions with both positively charged residues R563 and K565, the carboxylate recognition sites. These interactions mainly hold the RAR at the active site and improve its binding strength. As reported, the N-terminal amino group interacted with the E271 which is the N-terminal recognition site for the peptide substrates. These interactions altogether brings the carbonyl oxygen atom of N-terminal peptide bond close to the catalytic metal ion. A pcation interaction was formed between the side chain of Cterminal Arg residue and Y378, which was found highly fluctuating in rmsf analysis. This p-cation interaction between the same atoms was observed in R563A and K565A systems as well whereas it was between Y383 and C-terminal Arg residue in E271Q system (Figure 8 ). The binding mode of the substrate in E271Q system was different to that of WT system. The strength of the hydrogen bonds with R563 and K565 has become weak in this system because of the conformational changes of both carboxylate of substrate and R565 residue ( Figure 8B ). The metal ion located in the active center was thrown away from its initial position because of the absence of negatively charged E271. This behavior observed in this system clearly reveals that E271 acts as a hook to hold the Zn 2+ ion in the active site. In terms of R563A system, the observed binding mode of this system is similar to that of E271Q system ( Figure 8C ). The metal ion was hooked by the presence of E271 residue but still the distance between Zn 2+ and the carbonyl oxygen atom of N-terminal peptide bond was high compared to that of WT. The hydrogen bonds were formed between Y378, Y383, and G269 residues. Surprisingly, two hydrogen bonds were formed with K565 residue in absence of R563. This is different compared to that of the equivalent L-LTA4 system where the hydrogen bonds with both the positively charge residues were completely lost. This observation indicates the importance of K565 in assisting R563 in carboxylate recognition during peptide binding. In K565A system, the binding mode of the substrate is folded and completely different to the other systems. The hydrogen bonds were observed only with Y383 and A565 residues along with an additional p-s interaction between Y378 and C- terminal Arg residue ( Figure 8D ). No hydrogen bonds were formed with R563 residue, which is one of the positively charged carboxylate recognition site residues. The molecular interactions observed in R563A and K565A systems revealed that both the residues are important for recognizing the carboxylate group and aligning the peptide substrate along the catalytic elements as reported. Active site structural changes L-LTA4 system. The overlay of each mutant system with the WT system has allowed observing the structural changes occurred because of the mutations (Figure 9 ). In E271Q system, the loop of G269 was fluctuating and moved into the active site when compared to other systems. This change observed in this loop could be because of the newly formed hydrogen bond between G269 and the tilted carboxylate group of LTA4, which was found to be a response to the metal ion that lost the interaction with mutated E271. The carboxylate group of LTA4 and terminal NH 2 of R563 moved away from each other (2.7 to 8.4 Å ) because of the missing E271. A short beta sheet formed by the residues V306-N308 of HEXXH-(X) 18 -E motif that possess two catalytically conserved histidine residues coordinating with Zn 2+ ion disappeared in E271Q system. Another helix followed by this short beta sheet was extended by four amino acids W311-F314 making the important F314 slightly backward from the active center ( Figure 9A ). Two tyrosine residues Y378 and Y383 located opposite to each other in the active site were highly fluctuating in this system to adjust the binding of the LTA4. The K565 residue present in the loop has become a part of the long helix originally formed by T567-A575 residues during the simulation of E271Q system. This change slightly drew back the K565 residue from the active site. The other important positively charged residue did not show any structural changes during the simulation. The R563A system has shown different structural changes compared to E271Q system. The loop containing G269 has shown slight fluctuation only at the location of G269 because of the hydrogen bonds formed between the carboxylate of LTA4 and G269 but the lower part of the loop was stable unlike E271Q system ( Figure 9B ). The short helix formed by V306-N308 residues was maintained in this system and the helix containing F314 was extended but kept for the same length during the simulation. The region (W311-F314) that turned an extended helix was highly fluctuating in this system because of the moving alkyl part of LTA4. This backward movement buried the alkyl part into the hydrophobic pocket, formed by a mixture of aliphatic and aromatic hydrophobic residues (W311, F362, K364, L365, V366, V367, and V381), was observed because of the missing interactions with R563 residue (not shown in figure). Unlike E271Q system, Y378 residue has shown only slight side chain movement as a response to the moving LTA4 whereas Y383 did not show any fluctuations from its initial position. The same helix extension was observed as in E271Q system and thus K565 was included in helix formed by T567-A575 residues. The missing R563 led to the loss of correct alignment of LTA4 along the catalytic elements and severe instability of the binding mode of LTA4. In the final mutant (K565A) system, the G269 loop was completely stable and no hydrogen bond interaction was observed between LTA4 and G269 residue. The short helix of V306-N308 disappeared during the simulation of K565A system as observed in E271Q system ( Figure 9C ). The helix of F314 was extended as seen in R563A system and thus has shown the mixed characteristics of E271Q and R563A systems. The Y378, one of the oppositely located pair of tyrosine residues, has fluctuated highly in this system. The binding mode of LTA4 was quite similar to that of WT except its carboxylate group, which moved back because of the missing hydrogen bonds from K565 residue. But the hydrogen bonds with R563 were maintained and thus kept the alignment of LTA4 along with the catalytic elements. The overlay of active sites of all the systems have made clear about the structural changes where Y378 was observed to be fluctuating differently in each system maintaining a close distance with the substrate ( Figure 9D ). Thus Y378 residue, along with the carboxylate recognition site residues R563 and K565, can play a key role in aligning the substrate at the active site. L-RAR systems. Comparison of active site residues of WT and E271Q systems using the representative structures obtained from the simulations revealed the structural changes led to the differences in the catalytic activity of the enzyme ( Figure 10 ). The overlaid WT and E271Q structures have shown the difference in the locations of catalytically important Zn 2+ ion in the catalytic center ( Figure 10A ). The uncharged nature of the mutant residue Q271 the metal ion has lost the important interaction and moved far away from its original location. The distance between the Zn 2+ ion and the oxygen atom of the N-terminal peptide bond was so high compared to the WT system. Thus the catalytic aminopeptidase reaction becomes impossible in E271Q system. The helix containing Y383 was extended during the simulation E271Q system moving Y383 backward from the active site. This movement of Y383 has formed p-cation and a hydrogen bond interactions with the C-terminal part of RAR whereas in WT system this residue has formed a hydrogen bond interaction with the N-terminal amino group. The other tyrosine residue Y378, which was found to be guiding the substrate along with the carboxylate recognition site residues, has formed hydrogen bond with the carboxylate of RAR. The binding mode of RAR observed in E271Q system has shown only weak hydrogen bonding interactions with R563 as it was moving away from it. The helix formed by K565-A575 residues was shortened slightly leaving K565, one of the carboxylate recognition residues, as a part of loop making it more flexible. But this residue has maintained hydrogen bond interactions with the peptide substrate through its carboxylate group. This change observed in helix containing K565 is different from that of L-LTA4 systems. The K565 is the key residue that assists the other positively charged residue R563 to maintain the proper alignment of RAR in the active site whereas in LTA4 binding K565 is not required to assist R563 residue [27] . Moreover, the absence of negatively charged E271 residue also played a major role in observed loss of catalytic activity. The R563A mutation has caused a high fluctuation of E271 that moved far from the N-terminal amino group of RAR and makes it impossible to act as N-terminal recognition site ( Figure 10B ). The interacting distance between Zn 2+ and E271 was maintained in this mutation. The helix extension was observed near Y383 as displayed in E271Q system and this changed the flexibility of Y383 in the active site. The other tyrosine residue Y378 was highly fluctuating in this mutant system compared to any other systems and formed strong p-cation interaction than it is in WT system. The C-terminal part of RAR substrate has slightly went back as it missed the strong interactions from R563 but the alignment was almost maintained as observed in WT system except the side Cterminal side chain of RAR. Interestingly, K565 has taken the location of R563 in this mutant system to maintain the hydrogen bonds with the carboxylate of RAR and there was no change observed in the K565-A575 helix as observed in E271Q system. The K565A system also has shown some structural changes that were not observed in other L-RAR systems ( Figure 10C ). The short helix (V306-N308) that has shown structural changes in L-LTA4 system disappeared in K565A system and the F314 helix was extended including W311-F314 residues. These changes were observed in other L-RAR systems. The extension of helix has drawn F314 residue back from the active site center. A folded binding mode of RAR was observed in K565 system much different from that of WT and other mutant L-RAR systems. Though R563 is present, the carboxylate group of RAR has moved back from its original position and formed p-interactions with Y378 and completely lost interactions with R563. This observation completely correlates with the observed activity and the reported statement that K565 assists R563 to act as carboxylate recognition site in aligning the substrate along the catalytic elements of the enzyme. The overlay of all L-RAR systems revealed that along with E271, R563, K565 residues, Y378 and Y383 were also important in keeping the peptide substrate aligned within the active site ( Figure 10D ). As observed in L-LTA4 systems, Y378 residue has acted as a baffle to control the binding modes of the substrates. This part of the study has documented various structural changes explaining the differences in activities between WT and mutated forms of the enzyme bound to its two different substrates which were not determined by the X-ray crystallography so far (Table S1 ). The binding modes of fatty acid and peptide substrates that are catalyzed by hydrolase and aminopeptidase functions of the enzyme using same active site were compared. The overlay of two WT systems has given the overview of which parts of the active site were occupied by these two highly diverse substrates. Both the substrates bind perpendicular to each other occupying majorly the different portion of the active site and sharing the carboxylate recognition sites in common ( Figure 11A ). The long alkyl part of LTA4 was snuggly bound into the hydrophobic pocket formed by a mixture of aliphatic and aromatic residues including W311, F314, K364, L365, V367, Y378, and V381. The side chain of Cterminal Arg residue of RAR was fit into the small cavity formed by F356, Y378, S379, M564, K565, and R568 residues. The overlay of active site residues has shown very few structural changes between the WT systems of L-LTA4 and L-RAR systems including the side chain movements of Y378, Y383, and K565 residues ( Figure 11B ). The K565 residue was present in the loop in WT of L-LTA4 system and in the extended helix in WT of L-RAR system and thereby changing the flexibility and interacting behavior of K565. This adds explanation to the importance of K565 residue in assisting R563 residue in aligning the peptide substrate whereas this residue is not required in aligning fatty acid substrate. In this study MD simulation methodology was used to simulate hLTA4H enzyme complexed with its two diverse substrates along with mutated key residues. The aim of this study was to investigate the structural and conformational changes in the bi-functional active site of the enzyme reflecting on the catalytic activity. This was considered very important and necessary to script the reasons for the observed loss of activity due to particular mutations as the solved X-ray structures failed to show the structural changes. Eight systems including two WT, enzyme-LTA4 and enzyme-RAR complexes along with three independent mutations (E271Q, R563A, and K565A) in each complex were simulated in this study. The observed hydrogen bond interaction network and distance between the catalytically important atoms have correlated well with the experimental results. The E271 residue which is considered very important for both functions of the enzyme and E271Q systems have revealed from our study that this residue acts as a hook to hold the catalytic metal ion at its location and also plays a role of N-terminal recognition point for the aminopeptidase function. Both the E271Q systems have lost the expected binding mode of the substrates for the successful catalysis. The other mutant R563A and K565A systems have also revealed the structural changes and binding mode differences explaining the loss of activity in mutant systems. In L-LTA4 systems, the substrate binding mode in R563A system has changed completely that the long alkyl chain of LTA4 was completely buried into the hydrophobic pocket. This difference in binding mode of LTA4 was completely because of the loss of hydrogen bond interaction with R563 residue. In terms of L-RAR systems, the same mutation R563A has affected the binding mode of RAR and N-terminal recognition through E271 residue in peptide binding. Because of this missing N-terminal recognition the catalytic distance between the metal ion and the carbonyl group of the N-terminal peptide bond was high in this system. The K565A systems in both the substrate complexes have shown different structural changes. In L-LTA4 system the binding mode of the substrate was very much similar to the WT explaining the less importance of K565 in LTA4 binding whereas in L-RAR system the binding mode has lost both the N-terminal and C-terminal recognitions leading to the loss of activity. These results obtained from this study can be effectively used in designing future hLTA4H inhibitors as anti-inflammatory and anti-cancer therapeutics. Table S1 The structural changes which were not seen in experimental studies observed through the MD simulation studies.

Molecular and Microscopic Analysis of Bacteria and Viruses in Exhaled Breath Collected Using a Simple Impaction and Condensing Method

Exhaled breath condensate (EBC) is increasingly being used as a non-invasive method for disease diagnosis and environmental exposure assessment. By using hydrophobic surface, ice, and droplet scavenging, a simple impaction and condensing based collection method is reported here. Human subjects were recruited to exhale toward the device for 1, 2, 3, and 4 min. The exhaled breath quickly formed into tiny droplets on the hydrophobic surface, which were subsequently scavenged into a 10 µL rolling deionized water droplet. The collected EBC was further analyzed using culturing, DNA stain, Scanning Electron Microscope (SEM), polymerase chain reaction (PCR) and colorimetry (VITEK 2) for bacteria and viruses. Experimental data revealed that bacteria and viruses in EBC can be rapidly collected using the method developed here, with an observed efficiency of 100 µL EBC within 1 min. Culturing, DNA stain, SEM, and qPCR methods all detected high bacterial concentrations up to 7000 CFU/m(3) in exhaled breath, including both viable and dead cells of various types. Sphingomonas paucimobilis and Kocuria variants were found dominant in EBC samples using VITEK 2 system. SEM images revealed that most bacteria in exhaled breath are detected in the size range of 0.5–1.0 µm, which is able to enable them to remain airborne for a longer time, thus presenting a risk for airborne transmission of potential diseases. Using qPCR, influenza A H3N2 viruses were also detected in one EBC sample. Different from other devices restricted solely to condensation, the developed method can be easily achieved both by impaction and condensation in a laboratory and could impact current practice of EBC collection. Nonetheless, the reported work is a proof-of-concept demonstration, and its performance in non-invasive disease diagnosis such as bacterimia and virus infections needs to be further validated including effects of its influencing matrix.

Bioaerosols are present virtually anywhere in the environment, and their exposure is shown to cause numerous adverse health effects [1] [2] . In addition, there is also a possible release of biowarfare agents in a man-made bio-terror event. A number of studies demonstrated that the respiratory tract can be colonized with disease organisms [3] [4] [5] . Through talking, coughing, sneezing or singing, the potential virulent organisms can be exhaled and spread into the ambient environment [6] , which accordingly causes air contamination. For example, SARS in 2003 and H1N1 in 2009 outbreaks were shown to be attributed to the airborne route of disease transmission [7] [8] [9] [10] . Among many other diseases, respiratory infection accounts for 23.3-42.1% of the total hospital infections [11] , and is listed as the third leading killer [12] . However, present diagnosis procedures using nasal swabs, bronchoalveolar lavages, nasopharyngeal aspirates or sputum samples, appear to cause unpleasant experiences in addition to long detection time. During flu outbreaks, body temperature or isolation procedures are often used to control and prevent further spread, however such methods are lacking scientific evidence and not always effective with those patients infected but in latent period. On another front, exhaled breath condensate (EBC) as a simple and noninvasive method is increasingly being utilized in early disease screening and infectious aerosols measurements, e.g., lung cancer [13, 14] , asthma [15, 16] , and other respiratory problems [17, 18] . In previous studies, human influenza A viruses were detected in exhaled breath using EBC [19, 20] as well as filter [21] , mask [22, 23] and a liquid sampler [24] . In another study, foot-and-mouth disease viruses were also found in the exhaled air from experimentally infected cattle [25] . In addition, high levels of bacterial concentrations in EBC were also observed in other studies [26] [27] [28] [29] . It was recently shown that exhaled breath could be also analyzed for fungal infection by relevant biomarker, e.g., 2-Pentyl furan (2PF) for aspergillosis [30] . Overall, EBC has demonstrated great potential and advantages in early disease screening and diagnosis [31] , opening a new arena for studying airway inflammation and chemistry [32] . Recently, Vereb et al (2011) suggested that exhaled breath can be also used for assessing a variety of environmental exposures [33] . For EBC related studies, the first key step is the collection of exhaled breath. Over the years, a variety of devices (Table S1 , Supporting Information) were developed including Rtube collection system (Respiratory Research, Inc, Charlottesville, VA) and EcoScreenH condenser (Erich Jaeger Gmbh, Wurzbur, Germany). Typically, these devices would be able to collect 1000 ml of EBC samples within about 10 min, however the collection often comes with a lengthy procedure and a higher cost. For example, use of the EcoScreen involves 7 steps: 1) turn on to cool, 2) clean collection tube, 3) clean condensation chamber insert, 4) retrieve cooling sleeve from freezer, 5) sample collection, 6) sample storage and transport, 7) removal of sample (Respiratory Research, Inc, Charlottesville, VA). The RTube eliminates the first 3 steps, but each collection still requires 10 min and costs $23.25 (Respiratory Research, Inc, Charlottesville, VA) compared to 31 min and $47.17 per collection for the EcoScreen. These collection devices are generally expensive, e.g., the EcoScreen costs around $9000. A recent study compared the sampling efficiency of the Rtube (widely used EBC collection device) with that of throat swab method, showing detection rates of 7% and 46.8% for the Rtube and the throat swab method, respectively [20] . It was suggested that the RTube is not applicable for viral detection in exhaled breath [20] . In addition, condenser coatings [34] , sampling temperature [35] and sampling times [36] were shown to affect physical collection efficiencies of available EBC collectors. Among others, the noted problems with these available EBC collection devices are the device availability, reusability and possible cross contamination [37] , which would negatively impact their wide applications. In addition, EBC collection is strictly limited to the method condensation only in most studies [38] . To fully utilize EBC in early disease screening, diagnosis and environmental exposure assessment, simple yet efficient EBC collection device using different methods and biological characterization of the EBC sample are needed. In this study, a novel EBC collection method was developed by using hydrophobic surface, a layer of ice, and a droplet scavenging procedure. The physical collection efficiency (amount of EBC collected per unit of time) of the device was evaluated. In addition, biological analysis and characterization of EBC samples collected from human subjects were conducted using culturing, DNA stain, SEM, qPCR and species identification tool VITEK 2. This work contributes to the effort in applying EBC together with molecular tools as a non-invasive method in rapid disease diagnosis. The collection method and device developed and experiential set up for collecting EBC are shown in Figure 1 and Figure 2 , respectively. As observed in Figure 1 , a simple EBC collection device was developed here. The EBC collection device is composed of four major parts as shown in Figure 1 : collection device cover, collection device base, a layer of ice, and a hydrophobic film (treated by ultralow temperature 270uC). The collection device cover and base were made of Teflon TM polytetrafluoroethylene (PTFE) material, and a parafilm (Parafilm Co. Menasha, WI) used as the hydrophobic surface. The dimensions of the collection device are measured as 80640640 (mm) (length6width6height). In the collection device cover, there is a hole with a diameter of 6 mm as the exhaled breath inlet. The thickness of the collection device cover and base was about 3 mm, and the whole collection device weighs around 105 g. The layer of ice is used to keep the treated hydrophobic film cool. For EBC collection, sterile water (DNA and RNA free) was first added into the collection base of the device up to a depth of 5 mm as observed in Figure 2 . And then, the device base together with the cover was placed in an ultralow temperature (270uC) refrigerator (Thermo Fisher Scientific Co. Marietta, OH) to form a layer of ice. Following this step, a sterile hydrophobic parafilm measured as 8064060.3 (mm)(length6width6thickness) was placed onto the surface of the ice suited in the collection base. To collect EBC samples, a disposable sterile straw with a diameter of 5 mm (16 cm long) is inserted through the exhaled breath inlet shown in Figure 2 , with its end 2 mm above the hydrophobic parafilm. The human subjects are then advised to mouth-breathe without wearing a nose clip through the exhaled breath inlet shown in Figure 2 toward the hydrophobic film for a selected time (1-4 min). Due to the low temperature and hydrophobic nature of the parafilm surface, exhaled breath quickly condenses into tiny liquid droplets on the hydrophobic surface. Assuming an average breathing rate of 12 L/min for an adult, the particle speed from the exhaled breath would be around 10 m/s given the size of the straw (5 mm in diameter). Therefore, during the exhaled breath collection, the bacteria or virus particles would impact onto the hydrophobic surface at a speed of 10 m/s. In addition to condensing used for other EBC collection procedures, the method developed here also rely on the impaction to collect the bacterial and viral particles. Given such a speed, there might be possible particle bounce problems, however the bacterial or viral particles in the exhaled breath usually come with water droplets, which thus minimizes the potential particle bounce problem. After the collection, about 10 ml of DNA and RNA free DI water was pipetted onto the hydrophobic film as observed in Figure 1 . To collect breath samples on the hydrophobic film, one only needs to use the pipette to touch the DI water droplet, and then drag the DI water droplet to scroll over the entire surface. The DI water droplet would move with the pipette without an extra step. The materials collected on the surface would be subsequently scavenged into the water droplet. After this operation, the collected EBC samples in the form of bigger liquid droplet as shown in Figure 1D were transferred to a sterile tube by a pipette for subsequent analysis. The samples collected without the exhaled breath from human subjects are used as the negative controls. The EBC collection efficiency and biological analysis of collected samples were performed as outlined in the experimental procedure shown in Figure S1 (Supporting Information). To investigate the amount of variability in EBC collected by the method developed, six student volunteers were recruited to exhale through the device for 1, 2, 3 and 4 min. The volume of collected EBC was measured by a calibrated pipette (Eppendorf, Hauppauge, NY). The amount of EBC per unit time collected by the device was determined using averages of EBC samples obtained by the volunteers under each of specific collection time tested. For each EBC collection, a different hydrophobic film and a different exhalation straw were used. In addition, the particle size distributions in the exhaled breath through mouth-breathing were also measured in a particle free bio-safety hood using an Optical Particle Counter (OPC) (GRIMM Co. Ltd., Ainring, Germany) at a flow rate of 1.2 L/min. To ensure air stream balance, the OPC was connected to a two-way tubing, which connects to clean air (Biological SafetyHood) and the breathing straw, respectively. In this work, seven patients with onset flu symptoms (their medical information is listed in Table S2 , Supporting Information) were also recruited from the respiratory clinic of Peking University Third Hospital in Beijing. About 40 ml of exhaled breath condensate collected from each of 7 patients was diluted by 10 times and then plated on Trypticase Soy Agar (TSA) (Becton, Dickson and Company, Sparks, MD) plates at 26uC for 2-3 days, and colony forming units (CFUs) were manually counted. The total culturable bacterial aerosol concentration was calculated as CFU/m 3 (exhaled breath) by considering the collection time and an average breathing rate of 12 L/min for an adult. Besides, the culturable bacterial species were identified using VITEKH 2 (BioMérieux, Inc,100 Rodolphe Street, Durham, NC). In addition, molecular detection of bacteria and virus using qPCR and RT-qPCR, respectively, were performed according to the procedures described in Supporting Information S1. To further confirm the bacterial presence DNA stain of EBC sample by Acridine Orange (AO) was also conducted. The differences in collected EBC volumes and culturable bacterial aerosol concentrations obtained by the EBC collection device were analyzed by Analysis of Variance (ANOVA). A pvalue of less than 0.05 indicates a statistically significant difference at a confidence level of 95%. Collection of EBC from human subjects was approved by Peking University Ethnics Committee. Here, a novel EBC collection method and device was developed and evaluated in collecting EBC samples from human subjects using culturing and molecular methods. Compared to those currently available devices shown in Table S1 , our device is lightweight with simplicity, reusability, and lower cost. The developed collection device itself costs less than $10, with about $0.5 for consumables (straw and hydrophobic film) per collection. The time needed for 100 ml EBC including sample collection and removal was around 2 min. The physical collection efficiency of the device is shown in Figure 3 . The data points shown in the figure were averages of the EBC samples collected from six volunteers under each of the collection times (1, 2, 3 and 4 min) tested. In general, the amount of EBC sample collected was observed to increase with increasing collection time were observed among subjects. As also observed in Figure 3 , the method has a good reproducibility (small variations). ANOVA analysis indicated that the collection time had a statistically significant effect on the amount of EBC sample collected per unit of time (p-val- Table S2 ; F and M indicate Female and Male, respectively, 1-7 indicate the subject ID corresponding to those listed in Table S2 ; EBC collection time was 3 min; data points represent averages and standard deviations from at least three replicates. doi:10.1371/journal.pone.0041137.g005 ue = 0.0026). For the 4 min collection, the volume of collected EBC (168.7 mL) was 1.8 times of that (60.0 mL in average) by 1 min. In our study, when no EBC was collected about 1 mL of liquid was obtained from the hydrophobic surface in an environment with a temperature of 17.9-19.3uC and a relative humidity level of 46-52%. In addition, during the breath sample collection, the collection device had a higher air pressure due to the exhaling, thus it is less likely that environmental air would come into the device. This suggests that environmental water vapor had limited impact on the collection method given the total amount of EBC collected. A recent study indicated that the minimum required volume of EBC was 50 mL for follow-up biological and chemical analysis, such as multiplexed cytokine analysis [35] . This on the other hand implies that the EBC device developed in this study can provide adequate amount of EBC sample for rapid analysis. Here, only one type of hydrophobic surface (parafilm) was tested, and in the future different Table S2 ; DI water was used as the negative control. Table S2 ; Bacillus subtilis species was used as the positive control and DI water was used as the negative control; the curves shown here include two duplicates for each EBC sample. doi:10.1371/journal.pone.0041137.g007 hydrophobic materials should be also explored to improve the overall efficiency. As listed in Table S1 , currently available EBC collection devices, e.g., the Rtube and the EcoScreen, are comparable to ours with respect to rate of EBC collection. However, our EBC device has advantages in size, weight, and simplicity. In our study, we used a 16 cm long straw for exhaling toward to the super hydrophobic surface without any control of saliva for the possible contamination. However, our collection time was only 1-4 min, and during such short sampling period the sample contamination by saliva is very limited given the length of the straw. Another advantage of our developed device is the one time use of the hydrophobic parafilm (disposable) and exhalation straw with an easy collection of EBC, which thus prevents the possible cross contamination and facilitates the collection of EBC samples from a large number of subjects. This is particularly useful during an influenza outbreak or a man-made bio-terrorism attack in which a rapid screening of exposed persons needs to be conducted immediately. Here, the EBC samples collected by the developed device from seven human subjects recruited from a respiratory unit of Peking University Third Hospital in Beijing were studied using culturing, DNA stain, SEM and molecular methods. In this study, the particle size distributions trends in a typical exhaled breath were also measured and are shown in Figure 4 . As observed in the figure, the number concentration decreased with increasing particle diameter. For bacterial size ranges (0.65-2.2 mm), a concentration level of 329 to 25819 particles/L was observed, while for larger particles of 2.2-4 mm a concentration level of 60 to 400 particles/L was obtained. In previous studies, similar particle size distribution trend in exhaled breath was also found using the OPC, although the droplet concentrations for respective size ranges were slightly different [21, 39] . Nonetheless, due to its rapid evaporation water droplet itself or those adsorbing on bacterial particles in the exhale breath will certainly affect the results obtained here. The results from OPC indicated that particles of larger than 2.5 mm only accounted for 0.4% of the total particles exhaled. According to ICRP (1994), the total lung deposition efficiency for particles larger than 2 mm is more than 80%, while for smaller particles of less than 1 mm, the deposition efficiency is less than 40%, i.e., 60% exhaled out [44] . In addition, larger particles could stick to the straw wall. Therefore, in the exhaled breath as well as those collected into DI water droplet smaller particles would dominate. Figure 5 shows the concentrations of culturable bacterial aerosols in EBC samples collected from seven human subjects. As shown in the figure, bacterial concentration levels ranged from 693 to 6,293 CFU/m 3 . ANOVA tests indicated that there were statistically significant differences in culturable bacterial aerosol concentrations for EBC samples collected from different subjects (p-value = .0001). In a recent study, human occupants are also identified as the significant contributors for indoor bacteria, i.e., the emission rate is about 37 million gene copies per person per hour, and a distinct indoor air signature of bacteria was demonstrated to be associated with human skin, hair, and nostrils [40] . During human breathing, the bacterial particles from environmental air are continuously inhaled, some of which, i.e., smaller ones, can be exhaled out again by the lung and reside with nostrils. Here, bacterial species Sphingomonas paucimobilis and Kocuria rosea were detected using Vitek2 in six EBC samples as shown in Table S2 . Because of limitation of Vitek 2, certain bacterial species were not identified in our study. Among the subjects, subject #6 had substantially higher culturable bacterial concentrations than other subjects. From his medical conditions shown in Table S2 , it was likely that his fever was caused by the bacterial infections. In his EBC sample, we found Kocuria variants which were thought to cause catheter-related bacteremia [41] . For other human subjects, the culturable bacterial aerosol concentra- tion levels ranged from 700 to 3000 CFU/m 3 and Sphingomonas paucimobilis, a non-fermenting Gram-negative bacillus, were detected. In a previous study, S. paucimobilis was found to cause nosocomia bacteremia outbreak [42] . For negative control samples, we did not observe the bacterial growth, indicating no contamination during the EBC collection. Ideally, bacterial particles in EBC should be collected using a suitable size-selective sampling tool to investigate the bacterial counts for different size range. However, such device is currently not available yet. Compared to the environmental culturable bioaerosol concentrations, those in EBC samples collected had relatively higher levels, thus representing an important source of bioaerosols particularly in a high human occupancy environment. In addition to viruses, Rhodococcus equi, a bacterium causing pyogranulomatous bronchopneumonia, were detected in the exhaled air from foals in a recent study [43] . When pathogenic bacteria are breathed out, they could pose a serious public health threat. Figure 6 shows the qPCR amplification plot from EBC samples collected from seven human subjects in a respiratory clinic. As observed from the figure, bacterial samples were successfully amplified (Ct values were [16] [17] [18] [19] , while the positive sample (B. subtilis) had a Ct value of 15 and the negative control had a value of 28. Based on the DNA standards used, the concentrations of bacterial DNA in the EBC samples (Sample 1-7) were in the range of 0.32 mg/mL-3.15 mg/mL. Detection of the bacterial DNA in EBC samples was also confirmed by the melting curve of qPCR amplification as shown in Figure 7 . As observed in the figure, most EBC samples had a peak at 68uC, the same as that of the positive control B. subtilis. For a few different peaks observed, they might be the possible primer dimer (PD) from the PCR non-specific amplification process. In addition to the qPCR amplification of bacteria in EBC samples collected, DNA stain (AO method) was also performed and the results are shown in Figure S2 . As observed in the figure, both viable (green) and dead (yellow) were found in the EBC samples collected and the positive control B. subtilis samples, while no cells were detected in the negative control. SEM images with different resolutions and agar plate culturing shown in Figure 8 also indicated that EBC samples (cultured) had various types of bacteria based on their morphologies and colony color. From SEM images, it can be estimated that most bacteria are in the range of 0.5-1.0 mm. According to total particle deposition curve developed by ICRP (1994) [44] , more than 60% of bacterial particles of below 1 mm could be exhaled out. These smaller bacterial particles could remain airborne for a prolonged time period, thus playing an important role in airborne transmission of potential diseases. Results shown in Figures 5, 6 , 7, and 8 indicate that high levels of bacterial aerosols were detected in the EBC samples collected, and the results on the other hand also implied that the developed device was efficient in collecting bacterial particles in the exhaled breath. These experimental data further confirm that exhaled breath is an important source of bacterial aerosols in the built environments. In this study, qPCR was also applied to detecting influenza A H3N2 viruses in EBC samples collected by the device. As observed in Figure S3 , H3N2 viruses were detected in the EBC sample collected from subject #3 with a Ct value of 28, while those for subject #1, #2 were shown below the detection limits. In addition, spiking viruses into the samples in general enhanced the overall qPCR signal as observed in Figure S3 . This on the other hand suggests no inhibition or amplification occurred when amplifying H3N2 viruses in EBC samples using qPCR. According to information shown in Table S2 , subject #3 had a fever, but no other information was available at the time of the experiment. In a previous study, it was indicated that use of the RTube for EBC collection had a very low viral detection rate (7%) compared to nasal swabs (46.8%) [20] . Recently, a mask-like sampler was also tested and proved to be useful in detecting viruses using PCR in exhaled breath [23] . It was indicated that airborne virus detection is difficult due to their low concentration and the presence of a wide range of inhibitors, thus optimized molecular biology should be performed to enhance their detection [45] . Although the number of the subjects tested is limited here, the developed method, i.e., EBC collection and qPCR application, was demonstrated successful in detecting viruses from human exhaled breath. This would offer a non-invasive method for diagnosis of respiratory infections by using EBC. In the future, more patients should be tested with the EBC collection device developed here for viral detections. Exhaled breath holds great promise for monitoring human health and for the diagnosis of various lung and systemic diseases, but analysis challenges remain due to the complex matrix of the breath [46, 47] . In this study, different from available devices restricted solely to condensation a simple and low cost EBC collection method using impaction and condensing was developed here for collecting bacteria and virus particles. An important advantage is the reusability of the collection device with a disposable hydrophobic film and an exhalation straw yet with a rapid EBC collection. This would offer the opportunity to collect EBC samples from a large number of subjects, especially during an influenza outbreak or a man-made bioterrorism event, within a shorter time frame. The developed EBC collection method was shown highly successful in detecting bacteria in EBC samples in a clinical setting. The developed EBC collection method was also shown applicable in detecting influenza viruses too. Experimental data here also suggest that exhaled breath, which was shown to contain smaller bacterial particles, could play an important role in airborne transmission of potential diseases. The collection efficiency of other substances including bio-markers (NO,CO, 8isoprostane, hydrogen peroxide, nitrite, volatile organic compounds) using the developed method here is subject to further investigations. In addition, different exhalation modes should be also investigated with the method in collecting EBC. Besides, the dynamics of the air flow, mixing, and effects of temperatures and humidity, condensation, evaporation, growth of particles during the collection as well as the optimal straw length should be also investigated for improving the developed technique. Overall, our developed method here could be easily made available to a laboratory, and have impacts on current practice of EBC collection. Nonetheless, the reported work is a proof-of-concept demonstration, and its performance in non-invasive disease diagnosis such as bacterimia and virus infections needs to be further validated including effects of those influencing factors described. Figure S1 Experimental procedures used in this study include physical characterization and molecular analysis of the EBC collection efficiencies of the device and its pilot application in a respiratory clinic. (TIF) Figure S2 Optical images of EBC samples stained by Acridine Orange (AO): Bacillus subtilis species were used as the positive control and DI water was used as the negative control. (TIF) Figure S3 Detection of H3N2 influenza viruses in EBC samples collected from three human subjects with ID: 1, 2, 3 corresponding to those listed in Table S2 ; In addition, spiked H3N2 virus samples were also amplified; H3N2 viruses were used as the positive control and DI water was used as the negative control. (TIF)

Seasonal influenza vaccination knowledge, risk perception, health beliefs and vaccination behaviours of nurses

The relationship between knowledge, risk perceptions, health belief towards seasonal influenza and vaccination and the vaccination behaviours of nurses was explored. Qualified nurses attending continuing professional education courses at a large London university between 18 April and 18 October 2010 were surveyed (522/672; response rate 77·7%). Of these, 82·6% worked in hospitals; 37·0% reported receiving seasonal influenza vaccination in the previous season and 44·9% reported never being vaccinated during the last 5 years. All respondents were categorized using two-step cluster analyses into never, occasionally, and continuously vaccinated groups. Nurses vaccinated the season before had higher scores of knowledge and risk perception compared to the unvaccinated (P<0·001). Nurses never vaccinated had the lowest scores of knowledge and risk perception compared to other groups (P<0·001). Nurses' seasonal influenza vaccination behaviours are complex. Knowledge and risk perception predict uptake of vaccination in nurses.

Annual epidemics of seasonal influenza result in about 3-5 million cases of severe illness and 250 000-500 000 deaths worldwide [1] . Healthcare workers (HCWs) can be a key source for influenza transmission in communities and hospitals as they are exposed to both infected patients and high-risk groups [2, 3] . Vaccination is the most effective way to prevent infection and severe outcomes [1] and the principal measure to reduce the impact of epidemics, such as hospitalization, mortality and morbidity [2, [3] [4] [5] . Moreover, studies suggest that the vaccination of HCWs has substantial economic benefits as well as health-related benefits, including reduced absenteeism from work and the extra costs of sick leave and staff replacement [4, 6, 7] . For the above reasons, the World Health Organization (WHO), United Kingdom Department of Health (DoH) [8] , United States Centers for Disease Control and Prevention (CDC), other healthcare professional organizations and many countries' government agencies [1, 9, 10] strongly recommend the annual seasonal influenza vaccination of HCWs. However, studies suggest that influenza vaccine uptake in HCWs is often low worldwide [11] [12] [13] [14] . For example, the overall seasonal vaccination rate in England for HCWs was 26 . 4% for the 2009/2010 season [15] . Nurses, as the group having the most patient contact, are more reluctant to be vaccinated than other HCWs [16] [17] [18] [19] [20] [21] [22] [23] . Although predictors influencing nurses' vaccination practices have been identified to some extent regarding knowledge and risk perception [16] [17] [18] [19] [23] [24] [25] [26] [27] , further studies are needed to explore the influences on nurses' attitudes and practices regarding influenza vaccination and to identify the major influencing factors for their vaccination behaviours. This study aimed to examine the relationship between knowledge, risk perceptions, health beliefs towards seasonal influenza and vaccination and the vaccination behaviours of nurses. A cross-sectional survey was conducted of qualified nurses between 18 April and 18 October, 2010. Qualified nurses attending continuing professional education courses at a large university in central London were invited to participate in the study. Potential respondents were given a study information sheet and a questionnaire by the investigator. Completed questionnaires were collected immediately by the investigator or returned by mail to the research team using Freepost addressed envelopes. Questionnaire completion was anonymous so that it was not possible to follow up non-response. Ethical approval was obtained from the University Ethics Committee. The questionnaire collected the following data : (1) knowledge about seasonal influenza and vaccination (22 items requiring true, false or unsure responses) included five dimensions to assess general information, severity of influenza, influenza vaccination, high-risk groups and vaccination-recommended groups; (2) risk perception (12 items with a 4-point Likert scale) towards influenza and pandemic with three dimensions (i.e. personal vulnerability to illness, negative consequences of contracting influenza and severity of influenza) ; (3) health locus of control including internal, chance and powerful others dimensions assessed by the Multidimensional Health Locus of Control (MHLC) scales [28] (18 items) ; (4) vaccination behaviours (nine items) including vaccination status (whether respondents had been vaccinated in the previous season), vaccination intent (whether respondents intended to be vaccinated next season) and vaccination history (how many times respondents had been vaccinated in the last 5 years) ; (5) reasons for accepting or refusing vaccination using two open questions; and (6) demographic characteristics (10 items) including gender, age group, highest educational qualification, place of work, clinical speciality, year of qualification as a nurse and whether or not respondents had direct patient contact. The Cronbach's a-coefficients for the three newly developed scales (sections 1, 2, 4) ranged from 0 . 701 to 0 . 763 and principal components analysis produced a good fit and confirmed the internal design of the instrument. Statistical analysis was performed using SPSS version 15.0 (SPSS Inc., USA). The x 2 test or Fisher's exact test was used to explore the statistical differences between categorical variables. The independentsamples t test was used to compare statistical difference between continuous variables in two groups. The one-way between-groups analysis of variance (ANOVA) was used to explore the differences between more than two groups. Logistic regression was performed to explore the impact of the variables on vaccination status. The two-step cluster analysis procedure was performed to explore the natural groupings (i.e. clusters) within the respondents. The clustering criterion was that the solution had smaller values of Schwarz's Bayesian Information Criterion (BIC), a reasonably large ratio of BIC changes and a large ratio of distance measures. A P value <0 . 05 was considered to denote statistical significance. In total, 672 questionnaires were distributed and 522 were returned representing a response rate of 77 . 7%. The characteristics of the respondents are summarized in Table 1 . Overall 188/508 respondents (37 . 0%) reported receiving a vaccination in the previous season with 44 . 9% never receiving a vaccination during the last 5 years. There was no difference in the demographic characteristics of the vaccinated or unvaccinated respondents in the previous season. The number of years qualified as a nurse for the two groups were 11 . 99¡9 . 085 years and 11 . 89¡8 . 624 years (P=0 . 898), respectively. Comparison of knowledge and risk perception scores and sub-scores of MHLC are summarized in Table 2 . There were significant differences in knowledge scores and risk perception between the vaccinated and unvaccinated nurses and between those with vaccination intent, no intent or unsure. There was no significant difference in the sub-scores of MHLC between the vaccinated and unvaccinated (data not shown in table) but there was a significant difference for the sub-score of powerful others between those groups with different vaccination intent. Direct logistic regression was performed to assess the impact of a number of factors on the likelihood that respondents had been vaccinated in the previous season. The model contained five independent Table 3 , only two of the independent variables made a unique statistically significant contribution to the model (knowledge score and risk perception score). The strongest predictor of vaccination status was the risk perception score, recording an odds ratio of 1 . 76, indicating that respondents who had higher risk perception scores were >1 . 76 times more likely to have been vaccinated in the last 12 months than those with lower scores, controlling for all other factors in the model. Knowledge score with an odds ratio of 1 . 05 indicated that knowledgeable respondents were more likely to be vaccinated than the unknowledgeable, controlling for other factors in the model. The two-step cluster analysis procedure was used to explore the natural groupings within the respondents. First, the auto-clustering exploratory analysis was performed using the categorical variables of vaccination status, vaccination intent, vaccination history and the continuous variables of knowledge score and risk perception score. Of the 522 respondents, 64 were automatically excluded from the analysis due to missing values on one or more of the variables. Of the 458 respondents assigned to clusters, 195 (42 . 6%) were assigned to the first cluster, 143 (31 . Subsequently the analysis was performed using the combined categorical variables of vaccination status in the previous season (=yes) and vaccination history and the continuous variables of knowledge and risk perception scores. The results were auto-clustered into four groups but not explainable. The procedure was repeated with the cluster number fixed to 2 due to the values of BIC, ratio of BIC changes and ratio of distance measures. Of the total 188 vaccinated respondents, 12 were excluded due to missing values. Of the remaining 176 respondents, 107 (60 . 8%) were assigned to cluster 1 and 69 (39 . 2 %) to cluster 2. Vaccinated cluster 1 comprised those vaccinated only in the previous season, i.e. the newly vaccinated group and vaccinated cluster 2 contained those vaccinated in the previous season who had more than one previous vaccination, i.e. the continuously vaccinated group. Then, the same analysis was repeated for the unvaccinated respondents and two clusters emerged, i.e. unvaccinated cluster 1 (never vaccinated) and unvaccinated cluster 2 (used to be vaccinated). The analysis had therefore separated the respondents into reasonable categories. A comparison of variables across all clusters revealed that the never vaccinated had the lowest knowledge score, risk perception score and powerful others sub-score of MHLC compared to the other clusters (P<0 . 001, P<0 . 001, P=0 . 020, respectively) and this difference was statistically significant. For the vaccinated, there were no significant differences across any variable for the newly vaccinated and continuously vaccinated clusters although there was a trend of higher average scores for knowledge and risk perception in the newly vaccinated cluster compared to those of the other clusters (P=0 . 652, P=0 . 288, respectively). For the unvaccinated, there were no statistically significant differences across the variables except for the MHLC 'powerful others ' sub-score (P=0 . 008). Further comparisons were performed to explore whether there were differences across the different items of knowledge and risk perception in the clusters. In the clusters of never vaccinated, other vaccination history and vaccinated with intent, there were significant differences in knowledge related to general information, high-risk groups and vaccination of recommended groups with P values of <0 . 001, <0 . 003 and <0 . 006, respectively. On average those never vaccinated had the lowest score while those vaccinated with intent had the highest scores across all knowledge items. For only one item of risk perception, i.e. personal vulnerability to illness, was there a significant difference between the clusters of never vaccinated and other vaccination history and between never vaccinated and vaccinated with intent (P<0 . 000 respectively). Those never vaccinated had the lowest average score. There was no statistically significant difference in the knowledge and risk perception item scores between the two vaccinated clusters. However, the newly vaccinated usually had higher scores than those of the continuously vaccinated except for one item, i.e. the vaccination of recommended groups. Similarly, for the two unvaccinated clusters there was no difference for knowledge scores, but there was a significant difference in one risk perception item, i.e. personal vulnerability to illness (P=0 . 001). Those never vaccinated had a lower score for this item than those who used to be vaccinated and they were also less knowledgeable compared to the other group. Tables 4 and 5 . In this study, the seasonal influenza vaccination rate in nurses was 37 . 0 % which is higher than previous reports of vaccination coverage ranging from 14 . 3-26 . 4% in HCWs in UK [12, 29, 30] and 16% in nurses reported by Chalmers [27] and similar to O'Reilly et al.'s reported vaccination coverage of nurses in elderly care units [19] . This higher vaccination rate might be explained to some extent by the UK media reports of the risk of seasonal influenza and H1N1 pandemics in 2009 which may have increased the sample nurses' risk perception towards influenza and consequently changed their vaccination decisions as noted in a previous study [31] . This study found that vaccination behaviours in nurses were more complex requiring an analysis of both vaccinated and unvaccinated nurses' behaviours. More levels of vaccination behaviours existed in the sample with the two-step cluster analysis revealing three whole population clusters, i.e. those never vaccinated, those vaccinated this season with intent next year, and those with other vaccination history. Two clusters, the newly vaccinated and continuously vaccinated, were identified for the vaccinated group and another two clusters, never vaccinated and used to be vaccinated, were identified in the unvaccinated group. To improve the influenza vaccination rates in nurses, it may be helpful to develop different strategies which target the nurse groups of the never vaccinated and the occasionally vaccinated. We found that a lack of knowledge about influenza and vaccination was a strong predictor of nurses' vaccination behaviours, especially for those never vaccinated. This cluster had the lowest knowledge score, suggesting that increasing their knowledge might improve their vaccination behaviours. However, it seems there are 'persistent decliners ' who are in the 'habit ' of not having a vaccination. This suggests that future educational campaigns need to be persistent, durative, and intensive if their vaccination behaviours are to be modified. For those who had been vaccinated in the past but not in the current season, knowledge was also a predictor for their vaccination behaviours, which suggests that current vaccination campaigns have failed to address their misgivings about vaccination to maintain their compliance with the annual vaccination recommendation for HCWs. Between those occasionally vaccinated and continuously vaccinated, knowledge levels were not significantly different but the newly vaccinated in 2009 had on average higher knowledge scores than those continuously vaccinated. This may reflect an increase in their risk perceptions towards influenza due to widespread reporting of the risks in the media encouraging them to be vaccinated for the first time in their lives. This suggests that timing may be crucial to the success of vaccination campaigns making behaviour modification easier. Future studies are required to explore the relationship between the content and timings of vaccination campaigns and nurses' first vaccination uptake. This study showed that the perception of personal vulnerability to illness was important in nurses making vaccination decisions. But perceptions of the negative consequences of contracting influenza and severity of influenza were not major factors, a finding which is consistent with findings of previous studies [16] . This suggests that future educational campaigns might be more effective if they focus on the negative personal consequences of contracting influenza and its sequelae rather than nurses' professional duty to protect patients or other vulnerable groups. Additionally, the reasons which nurses gave for having vaccination focused upon their personal health motivation rather than a professional responsibility regardless of whether they were vaccinated or unvaccinated. Concerns about the vaccine's side-effects and effectiveness or safety were the two most frequent reasons for not having a vaccination indicating continuing misconceptions about influenza vaccine in nurses. Future educational campaigns may wish to consider providing targeted information to change these widespread myths in nurses. However, these concerns did not seem to influence vaccination decisions because both vaccinated as well as unvaccinated nurses noted these reasons against vaccination. It may be the case that 2 days of minor discomfort postvaccination is tolerable when set against a year's influenza protection. Unvaccinated nurses reported 'no need ' as their reason not having a vaccination which is consistent with their low-risk perception of contracting influenza. The convenience of the vaccination programme was identified as an organizational reason highlighting the importance of easy access to vaccination to increase its coverage in nurses. Our analysis of health locus of control data found that those never vaccinated had a lowest 'powerful others ' locus of control for their vaccination behaviours, indicating that they did not believe their health was something over which they had no control [32] . This pattern of health beliefs towards influenza vaccination is consistent with their low-risk perception of personal vulnerability to illness and 'no need ' as their reason refusing vaccination and may be an important factor for never vaccinated nurses. Further studies are needed to explore what may influence this pattern of health locus of control in order to modify nurses' vaccination behaviours. Some organizations have recently required mandatory seasonal influenza vaccination for HCWs as a professional and ethical obligation to protect their patients' health [33, 34] . However, ethical issues have been raised with mandatory vaccination because, while promoting the interests of patients and employers, it challenges HCWs' personal autonomy and freedom of choice [35, 36] . Moreover, it has been suggested that vaccination is not the only avenue of influenza prevention and there are several other important measures that healthcare organizations may take to protect both patients and HCWs [37] . Further previous studies have also suggested that not all HCWs support mandatory vaccination [38] . Until mandatory influenza vaccination for HCWs is accepted worldwide, continued efforts to improve nurses' vaccination behaviours will be required. This study has some limitations. First, there is possible selection bias of a convenience sample ; however, the broad range of qualified nurses together with a high response rate strengthen the results. The extent of bias is unknown especially regarding nurses not working in London or in different care settings. Second, the survey relied on self-report vaccination data ; however, Zimmerman et al. [39] found that selfreport data were reliable in comparison with medical records. Third, the three factors explored relating to nurses' vaccination behaviours explained only 8 . 7-11 . 9% of the variance according to the logistic regression analysis (although it was statistically significant) and therefore our results cannot fully explain nurses' vaccination behaviours. Additional predictors will need to be introduced into the model in future studies to fully explain nurses' vaccination behaviours. In conclusion, this study revealed that nurses' influenza vaccination behaviours are complex. Knowledge and risk perception were identified as two predictors influencing nurses' vaccination decisions with the health belief pattern of 'less powerful others ' being an important predictor in the never vaccinated ; however, there are other influential factors which need to be identified in future studies.

IFN-γ Signaling to Astrocytes Protects from Autoimmune Mediated Neurological Disability

Demyelination and axonal degeneration are determinants of progressive neurological disability in patients with multiple sclerosis (MS). Cells resident within the central nervous system (CNS) are active participants in development, progression and subsequent control of autoimmune disease; however, their individual contributions are not well understood. Astrocytes, the most abundant CNS cell type, are highly sensitive to environmental cues and are implicated in both detrimental and protective outcomes during autoimmune demyelination. Experimental autoimmune encephalomyelitis (EAE) was induced in transgenic mice expressing signaling defective dominant-negative interferon gamma (IFN-γ) receptors on astrocytes to determine the influence of inflammation on astrocyte activity. Inhibition of IFN-γ signaling to astrocytes did not influence disease incidence, onset, initial progression of symptoms, blood brain barrier (BBB) integrity or the composition of the acute CNS inflammatory response. Nevertheless, increased demyelination at peak acute disease in the absence of IFN-γ signaling to astrocytes correlated with sustained clinical symptoms. Following peak disease, diminished clinical remission, increased mortality and sustained astrocyte activation within the gray matter demonstrate a critical role of IFN-γ signaling to astrocytes in neuroprotection. Diminished disease remission was associated with escalating demyelination, axonal degeneration and sustained inflammation. The CNS infiltrating leukocyte composition was not altered; however, decreased IL-10 and IL-27 correlated with sustained disease. These data indicate that astrocytes play a critical role in limiting CNS autoimmune disease dependent upon a neuroprotective signaling pathway mediated by engagement of IFN-γ receptors.

CNS resident cells are targets of autoimmune mediated damage but also active participants in disease development, progression and control [1, 2] . However, their contributions to neuroprotection and regulation by inflammatory mediators are not well defined. CNS insults, including autoimmune disease, initiate rapid astrocyte activation characterized by cellular hypertrophy and increased intermediate filament glial fibrillary acidic protein (GFAP) expression [3] [4] [5] . Astrocytes form a physical barrier surrounding areas of inflammation initially limiting bystander tissue damage [6, 7] . However, this barrier subsequently impedes axonal regeneration contributing to sustained disability [3, 5, 8] . Innate and adaptive pro-inflammatory astrocyte functions include production of pro-inflammatory cytokines, reactive oxygen species, chemokines, and matrix metalloproteinases [3] [4] [5] . By contrast, secretion of anti-inflammatory cytokines and scavengers of reactive oxygen species, as well as inhibition of both microglial activation and tumor necrosis factor (TNF) secretion, all support an astrocyte mediated anti-inflammatory function [3] [4] [5] . Therefore, astrocyte activation constitutes a ubiquitous, yet heteroge-neous response associated with both promoting and inhibiting CNS repair [3] [4] [5] . MS and EAE are both associated with T cells secreting IFN-c (Th1 cells) and IL-17 (Th17 cells) which play complex, not fully understood roles in disease initiation and progression [2, 9] . In vivo and in vitro evidence indicates that suppression of encephalitogenic T cell proliferation within the CNS during EAE and activation of anti-inflammatory programs are in part mediated via astrocytes [4, 5] . Similar to astrocytes, IFN-c mediates both proand anti-inflammatory functions during autoimmune disease [10] . Early IFN-c induced effects are pro-inflammatory; IFN-c facilitates inflammatory cell access, shapes their composition, increases major histocompatibility complex (MHC) expression, contributes to macrophage and microglia activation, and initiates oligodendrocyte death [11] . Similarly, IFN-c mediated protection during EAE is also multifaceted [12] [13] [14] [15] . It acts as a negative regulator of neutrophil accumulation, Th17 cell activation, IL-1R signaling, protease secretion, and chemokine activity. It also inhibits proinflammatory cytokine secretion via induction of suppressor of cytokine secretion (SOCS) proteins, facilitates T cell apoptosis and protects oligodendrocytes via inducing endoplasmic reticulum (ER) stress responses [10, 11, 16, 17] . This highlights the critical role of a single mediator in both promoting disease but also limiting inflammatory mediated damage required to initiate repair cascades. Based on the gatekeeper functions of astrocytes and the diverse biological effects of IFN-c, we set out to determine how IFN-c signaling specifically to astrocytes influences CNS autoimmune disease. The results demonstrate that IFN-c signaling to astrocytes had no profound effects on initial disease progression, but played an essential protective role during the transition from acute to chronic disease. Clinical remission induced by IFN-c signaling to astrocytes coincided with reduced demyelination, axonal degeneration, and astrocyte activation. The IFN-c receptor is expressed ubiquitously; however, these data reveal astrocytes as the primary target and mediator of the well established antiinflammatory activity of IFN-c within the CNS. To understand the role of IFN-c signaling to astrocytes during the pathogenesis of CNS autoimmune disease, EAE was induced in transgenic mice expressing a signaling deficient dominant negative IFN-c receptor 1 specifically on astrocytes (GFAPcR1D mice) [18] . Peripheral activation of self reactive T cells in GFAPcR1D mice was similar to wt mice (data not shown). This is consistent with both the CNS restricted transgene expression in the GFAPcR1D mice as well as the similar T cell activation following peripheral immunization with a non-self antigen [18] . Following immunization the incidence of disease, initiation of clinical symptoms, and initial symptom progression were unaltered by the inability of astrocytes to respond to IFN-c ( Fig. 1A ; Table 1 ). In addition, neither immunized group exhibited clinical symptoms of atypical EAE associated with the absence of IFN-c [19] . However, clinical symptoms in GFAPcR1D mice began to diverge from wt mice at , day 12 post immunization (p.i.) prior to the peak of clinical disease. In both groups clinical disease peaked at , day 14 p.i., but severity was increased from a score of 3.1 in wt mice to 4.0 in the GFAPcR1D group ( Fig. 1A ; Table 1 ). GFAPcR1D and wt mice were compared at the peak of acute disease to determine if IFN-c signaling altered astrocyte activation or CNS inflammation. Astrocyte hypertrophy and GFAP expression ( Fig. 1B) were similar in both groups, indicating no overt effects of IFN-c on initial astrocyte reactivity. Furthermore, neither the extent of inflammation nor the anatomic distribution of inflammatory cells was altered (Fig. 1B) . Flow cytometry confirmed no difference in the overall extent of CD45 hi inflammatory cells recruited into the CNS ( Fig. 2A) . Furthermore, percentages of CD4 + T cells ( Fig. 2A) , CD8 + T cells and macrophages within the infiltrates were also similar (Fig. 3) . In contrast to the association between increased EAE severity and neutrophil accumulation in the global absence of IFN-c [14, 15] , only a small percentage of neutrophils were identified in the CNS of both groups by flow cytometry (Fig. 3) ; their low presence was confirmed by the inability to identify cells with the characteristic morphology of neutrophils in the brain by histopathology (Fig. 1B) . The absence of increased neutrophils in the CNS of GFAPcR1D mice during acute disease suggested that clinical disease was aggravated by a mechanism distinct from global IFN-c deficiency. In addition to the similar frequency of CD4 + T cells in the brains of the two groups ( Fig. 2A) , myelin oligodendrocyte glycoprotein (MOG) reactive CD4 + T cells secreting IFN-c were also similar ( Fig. 2B ). By contrast, MOG specific CD4 + T cells secreting IL-17 ( Fig. 2B ) and Foxp3 + regulatory CD4 + T cells (Tregs) were decreased in GFAPcR1D mice compared to wt mice (Fig. 2C) . Reduced Th17 and Tregs may be attributed to increased IFN-c [10, 16] . Alternatively, as IFN-c is protective in EAE [12] [13] [14] [15] , increased disease severity may reflect reduced bioavailable IFN-c due to sequestration of IFN-c binding to the dominant negative receptor. However, measurement of IFN-c in cell free supernatants derived from dissociated brains at the peak of acute disease demonstrated protein levels of 2.160.2 ng/brain in wt mice versus 3.660.5 ng/brain in GFAPcR1D mice (n = 3; p,0.05). As overall frequencies of MOG reactive CD4 + T cells secreting IFN-c were similar, increased IFN-c in the brains of GFAPcR1D mice suggested enhanced secretion at the cellular level. Although a contribution of NK or CD8 T cells could not be excluded, these potential sources of IFN-c were unlikely due to their low frequencies (NK ,5% and CD8 + T cells , 5%, see Fig. 3 ) and their equivalent frequencies in the wt and GFAPcR1D mice. Furthermore, reduced Th17 cell and Treg frequencies were consistent with suppression of these populations due to elevated IFN-c [10, 16, 20] . Inflammation in spinal cords from the GFAPcR1D mice was also similar to wt controls at the peak of clinical symptoms (Fig. 4) . Although numbers and distribution of CD4 + T cells were also similar (4.661.3/mm 2 in GFAPcR1D mice vs. 3.860.1/ mm 2 in wt mice; Fig. S1 ), spinal cords of GFAPcR1D mice exhibited a ,2-fold increase in demyelination (Fig. 4) . Areas of demyelination encompassed 7.361.0% of spinal cord white matter in GFAPcR1D mice versus 3.860.9% in wt mice (p#0.01). Furthermore, the increase in demyelination was associated with a prominent loss of axons in GFAPcR1D mice compared to controls (Fig. 4) . Despite elevated demyelination and axonal loss in the absence of IFN-c signaling to astrocytes, spinal cords showed no evidence of differential astrocyte activation by either immunohistochemistry (Fig. 4 ), or differences in GFAP mRNA expression during the peak of acute disease (Fig. 5 ). Although CCL5, IL-1 and TNF mRNA expression were increased in the spinal cords of the GFAPcR1D mice, no differences in IFN-c, iNOS, CXCL10 or IL-6 mRNA were consistent with similar inflammation (Fig. 5) . These data suggest that the initial disease pathogenesis, reflected by an increased demyelination in spinal cords, but not brain, during ascending paralysis is dampened by IFN-c signaling to astrocytes. Subsequent to peak disease severity the clinical symptoms in wt mice began a modest decline by day 16 p.i. (Fig. 1A ). By contrast, GFAPcR1D mice not only exhibited increased severity of clinical symptoms following day 14 p.i., but the continued disease escalation was associated with increased mortality ( Fig. 1A ; Table 1 ). Sustained morbidity and significant mortality in GFAPcR1D mice after day 30 p.i. implied a critical role for IFN-c induced astrocyte signaling in neuroprotection and limiting disability. A similar absence of clinical remission was found in a preliminary experiment comparing EAE SJL mice carrying the GFAPcR1D gene (data not shown). These data suggest that astrocyte responses to IFN-c are protective, irrespective of genetic background. Escalating disease in GFAPcR1D mice coincided with focal areas of intense inflammation in spinal cords, which contrasted with the more diffuse inflammation in wt mice (Fig. 6 ). Demyelination was also increased with myelin loss encompassing 5.761.8% of spinal cord area in GFAPcR1D mice versus 2.361.4% in wt mice (p#0.05) at day 35 p.i. (Fig. 6 ). The demyelinated areas further exhibited enhanced axonal damage in GFAPcR1D mice (Fig. 6) , supporting a correlation between enhanced tissue damage, sustained morbidity and increased mortality ( Fig. 1A ; Table 1 ). Astrocyte activation associated with areas of myelin loss is a prominent finding in the CNS of both patients with MS and rodents with EAE. Although demyelination was increased in the CNS of GFAPcR1D mice, the extent of astrocyte activation associated with spinal cord white matter lesions was similar in both groups ( Fig. 6 ; ,60 GFAP + cells/mm 2 ). By contrast, the frequency of activated astrocytes in spinal cord grey matter areas that were not associated with demyelinated lesions, was increased in GFAPcR1D mice with 101.5623.0 GFAP + activated astrocytes/mm 2 versus 25.5615.5 in wt mice (p,0.001) (Fig. 7) . A similar increase in astrocyte activation within grey matter distal to white matter lesions was also detected in GFAPcR1D SJL mice during chronic EAE (data not shown). Flow cytometric analysis during chronic disease revealed an ,4fold increase in CD45 hi inflammatory cells confirming increased inflammation in the absence of IFN-c signaling to astrocytes (Fig. 8A ). Similar to the acute disease, relative percentages of neutrophils, macrophages, CD4 + and CD8 + T cells were all similar (Fig. 3) . Furthermore, no differences in expression of activation markers on CNS derived CD4 + T cells, including CD69, CD122, CD127, Fas, FasL, ICOS and CD152 were evident between the groups (data not shown). The percentages of MOG reactive Th1 cells were also not altered in the CNS of GFAPcR1D relative to wt mice (Fig. 8B ) and the decreased percentage of potentially destructive Th17 cells identified during acute disease (Fig 2) , was also sustained during chronic disease (Fig. 8B ). Equivalent expression of IL-7, IL-23 and TGF-b mRNA (Fig. 8C ) suggested that the decrease in Th17 cells was dependent upon an increase in IFN-c and not related to a defect in activation or maintenance signals [20] . Increased inflammation and sustained MHC class II expression on microglia (Fig. 8A ) further suggested that IFN-c signaling to astrocytes down regulates inflammation universally without altering the composition of the CNS infiltrates. This concept is supported by sustained composition of CNS infiltrates during inflammation induced astrocyte apoptosis [21] . The inability to restrain ongoing inflammation during EAE in GFAPcR1D mice was evident at multiple levels. Astrocyte activation was sustained consistent with increased expression of GFAP mRNA. CCL2, CCL5 and CXCL10 mRNA levels were increased (Fig. 8C ). The expression of mRNA encoding potentially destructive immune mediators including IL-1, IL-6, iNOS, and TNF were all increased (Fig. 8C ). IFN-c was ,2-fold higher in the CNS of GFAPcR1D mice (Fig. 9) . Lastly, the level of the antiinflammatory cytokine IL-10 was reduced ( Fig. 9 ), suggesting limited availability of IL-10 may be a key in prolonging astrocyte activation and inflammation [22] . A possible link between reduced IL-10 and the inability of IFN-c signaling to astrocytes is provided by the capacity of IFN-c to induce IL-27 in astrocytes [23] , thereby promoting IL-10 production. Indeed IL-27 in the CNS of the GFAPcR1D mice was significantly reduced compared to wt mice during chronic disease (Fig. 9 ). By contrast, IL-12 ( Fig. 9) , an indirect suppressor of CNS inflammation by promoting IFN-c production [24] , was similar in the CNS of both GFAPcR1D and wt mice. Astrocytes derived from GFAPcR1D and wt mice were stimulated with IFN-c to confirm IFN-c dependent IL-27 secretion by astrocytes [25] . While IFN-c induced IL-27 secretion by astrocytes from wt mice, IL-27 secretion was significantly reduced in cultures derived from GFAPcR1D mice (Fig. 10) . These data support the possibility that IFN-c mediated IL-27 secretion by astrocytes regulates EAE effector T cell function, inflammation and tissue destruction via induction of IL-10; however, this does not exclude the possibility that the inability of astrocytes to respond to IFN-c could directly or indirectly dysregulate a variety of other immunomodulatory functions [4, 5, 17] . In the EAE model of MS, IFN-c functions as a proinflammatory cytokine in the early stages of disease [10, 11, 26 ], yet it also assumes a prominent anti-inflammatory role during the transition to remission [10, [12] [13] [14] [15] . Protection has been attributed to a variety of potentially interrelated mechanisms including limiting neutrophil accumulation, Th17 cell activation, IL-1R signaling, matrix metalloproteinase secretion, pro-inflammatory cytokine secretion and chemokine activity; in addition IFN-c facilitates T cell apoptosis and protects oligodendrocytes from death via an ER stress response [10, 11, 17] . The activities of astrocytes during autoimmunity also range from pro-to antiinflammatory [3] [4] [5] . However, to what extent a direct action of IFN-c on astrocytes contributes to inhibitory mechanism is unclear. The data herein demonstrate that among the multiple Early functions of astrocytes imposed by innate responses are largely pro-inflammatory during EAE [3, 6, 7, 26] . For example, astrocytes contribute to the loss of BBB integrity via secretion of reactive oxygen species, chemokines and pro-inflammatory cytokines [3] [4] [5] . This pro-inflammatory milieu is associated with initial axonal damage prior to accumulation of self reactive T cells [27] , which further promotes immune mediated damage. Blocking the pro-inflammatory transcription factor NF-kB in astrocytes improves recovery during chronic EAE [28, 29] . Supporting an initial pro-inflammatory role of astrocytes dependent on innate, not IFN-c responsiveness, inhibition of IFN-c signaling to astrocytes did not influence EAE onset or incidence, initial disease progression, astrocyte activation, or BBB integrity as indicated by similar entry of inflammatory cells into the brain. By contrast, an inflammation dampening effect of IFN-c became evident during the onset of disease remission. Astrocytes limit ongoing inflammation and pathogenic processes at several levels. They not only facilitate repair by inhibiting inflammatory cell entry into the CNS parenchyma [6, 7, 30] , but also down regulate T cell effector function and proliferation [4, 5, 21] . For example, EAE in GFAP deficient mice results in more severe and widespread inflammation [31] . In addition, T cell-astrocyte interactions as well as astrocyte secretion of an unidentified soluble product, distinct from nitric oxide, prostaglandins, or tryptophan metabolism, suppress T cell proliferation [4, 5] . T cell proliferation was also not inhibited via defective IL-2 release, despite the suggestion that T cell-astrocyte interactions facilitate secretion of anti-inflammatory cytokines [4, 5] . Lastly, astrocytes may limit inflammation by triggering apoptosis in T cells [32] . Nevertheless, the role of IFN-c signaling in these potentially anti-inflammatory, protective mechanisms is not clear. Elevated IFN-c in the CNS of GFAPcR1D mice during both acute and chronic disease excluded limited IFN-c due to sequestration by the transgenic receptor as a mediator of increased clinical severity and mortality. Indeed, enhanced IFN-c production may reflect an attempt to compensate for the loss of IFN-c dependent astrocyte mediated control of inflammation [33] . Sustained expression of pro-inflammatory cytokines, particularly IL-6 and TNF, thus represents a primary mechanism underlying ongoing tissue destruction in GFAPcR1D mice. IL-6 is known to mediate neurological dysfunction [34] and astrocytes are the predominant source of IL-6 during CNS autoimmune disease. Furthermore, its secretion is down regulated by IFN-c induced SOCS activity [17] . On the other hand, sustained TNF implicated dysregulated activation of microglia or macrophages via increased IFN-c or IL-6, as TNF is predominantly secreted by activated CNS macrophages and microglia during EAE [35, 36] . However, the down regulation of microglia activation, including TNF secretion, following interaction with activated astrocytes [37] questions this notion. While both IL-6 and TNF may thus contribute to sustained pathological changes, the source of TNF remains unclear. Similarly, astrocytes are a potential source of the IFN-c induced chemokine CXCL10 during EAE, one of the chemokines controlling T cell recruitment [38] . However, the increased expression of CXCL10 coupled with the inability of the astrocytes in the GFAPcR1D mice to respond to IFN-c suggests altered microglia activation and secretion of CXCL10 [38] or activation of CXCL10 transcription via an independent signaling pathway mediated by TNF or Type 1 interferons [39, 40] . Importantly, pathology further correlated with decreases in both Tregs and IL-10, but not with increased antigen specific Th17 cells, although astrocytes are critical targets of IL-17 [41] . Although it is possible that the increased IFN-c in the CNS influenced peripheral activation of antigen specific Th17 cells, previous data demonstrated no evidence for expression of the transgene in peripheral organs, including lymphoid organs [18] . IL-10, secreted by a variety of cells types including CD4 + T cells, CD8 + T cells and Tregs [42] , inhibits both acute and chronic EAE [43, 44] . The decrease in this anti-inflammatory cytokine suggests that the induction of IL-27 secretion is a critical aspect of astrocyte responses to IFN-c, thus promoting IL-10 secretion by T cells [23, 25] . However, our data is unable to distinguish the contribution of paracrine versus autocrine IFN-c induced signals to astrocytes on IL-10 production, as IL-10 also regulates astrocyte activation [22] , and can itself be secreted by astrocytes to exert autologous functions [45] . In addition to mediating an imbalance of protective versus detrimental cytokines, our results demonstrate that IFN-c signaling to astrocytes limits the extent of astrocyte activation during chronic EAE, specifically in grey matter. Activated astrocytes, especially within and adjacent to areas of demyelination are a prominent feature of white matter plaques associated with both chronic MS and EAE [3] [4] [5] . Astrocytes protect neurons and oligodendroglia, but also form glial scars which inhibit regeneration after neuronal injury [3, 5, 8, 20] . The absence of reactive astrocytes increases axonal regeneration after injury [46] , consistent with the concept that limiting astrogliosis is critical for axonal regeneration after neuronal injury. While the significance of sustained astrocyte activation in grey matter tracks is unclear in our model, it is consistent with increased axonal damage, and suggests possible damage to axons outside the lesions, which may not be manifested by histological analysis. Limiting inflammation following either infection or during an autoimmune attack is a prerequisite for the initiation of repair. This is critically important within the CNS which has limited regenerative capacity. In summary, our data identify astrocytes as prominent targets underlying IFN-c mediated suppression of chronic CNS inflammation [19] . The data further provide a link between sustained inflammatory responses, enhanced demyelination and axonal degeneration associated with loss of neurological function during EAE and chronic progressive MS. Although the inability of astrocytes to respond to IFN-c did not alter disease in the brain, engagement of the IFN-c receptor on astrocytes in spinal cord limits demyelination and functions in a neuroprotective capacity. Current therapies for MS are primarily focused on antiinflammatory and immunomodulatory approaches and have been partially successful in treating acute episodes. The identification of astrocytes as critical responders and mediators of IFN-c signaling in limiting CNS autoimmune disease may provide insights into new approaches to limit long term progression to disability. Brains and spinal cords from perfused mice were homogenized separately as previously described [47, 48] . Homogenates were centrifuged at 4506g for 7 min at 4uC. Supernatants were stored at -80uC for cytokine determination (see below). Cells were resuspended in 30% Percoll (Amersham Biosciences, Piscataway, NJ) and isolated by centrifugation (8006g for 30 min at 4uC) onto 70% Percoll cushions. Non-specific binding was inhibited by incubation with anti-CD16/CD32 (2.4G2; BD Biosciences, San Diego, CA) and a 10% mixture of normal goat, human, mouse and rat serum for 10 min on ice. FITC, PE, PerCP, and APC conjugates with monoclonal antibodies (mAb) used to identify and quantify microglia and inflammatory cells were: CD4 (GK1.5), CD8a (53-6.7), CD45 (30-F11), MHC class II (2G9), Ly6G (1A8) (BD Biosciences), and F4/80 (Serotec, Raleigh, NC). Microglia and inflammatory cells were distinguished based on differential CD45 expression. CD4 + T cells were identified as CD45 hi CD4 + , CD8 + T cells as CD45 hi CD8 + , macrophages as CD45 hi F4/80 + and neutrophils as CD45 hi Ly6G + MHC class II -. MOG specific induction of cytokines was determined by stimulation of 1610 6 CNS cells with 20 mg/ml peptide for 6 h at 37uC with GolgiStop (BD Biosciences) added for the last 4 h. Intracellular cytokines were detected with FITC-anti-IFN-c (clone XMG1.2; BD Biosciences) and PE-anti-IL-17 (clone TC11-18H10; BD Biosciences). Intracellular Foxp3 was detected by staining for cell surface markers, followed by permeabilization with Fixation/ Permeabilization Reagent (eBioscience, San Diego, CA) and incubation with PE-labeled anti-Foxp3 (FJK-16s; eBioscience). Cells were analyzed on a FACSCalibur flow cytometer (BD Biosciences) using CellQuest Pro software (BD Biosciences). Data was analyzed using FlowJo (7.6.1) software (Tree Star Inc., Ashland, OR). Cytokines were determined by ELISA using antibody pairs and recombinant cytokine standards from BD Bioscience. IL-27 was measured using QUANTIKINE Mouse IL-27p28 Immunoassays (R&D Systems Inc., Minneapolis, MN). Following anesthesia, mice were perfused with PBS (pH 7.2). Brains and spinal cords were fixed with Clark's solution (75% ethanol and 25% glacial acetic acid), and embedded in paraffin. Spinal cords were divided into 6 sections prior to embedding, corresponding to cervical, thoracic and lumbar levels. Cross sections (6 mm), were stained with either hematoxylin and eosin (H&E) or luxol fast blue (LFB). Immunoperoxidase staining was used to identify activated astrocytes with anti-GFAP (AbCam, Cambridge, MA) and axonal integrity with SMI-31 and SMI-32 mAb (Sternberger Monoclonals Inc., Lutherville, MD) followed by visualization using Vectastain ABC kit (Vector Laboratories, Burlingame, CA) and 3,39-diaminobenzidine (Sigma-Aldrich). Sections from at least 3 separate experiments containing at least 3 mice per group were reviewed in a blinded manner. Numbers of GFAP + cells in spinal cord were determined in 15 non-overlapping 406 fields (0.2 mm 2 ) in white matter and gray matter. Stained spinal cord sections of all 6 levels on individual glass slides were scanned (406) and digitally imaged at high resolution with an Aperio ScanScope (Vista, CA). Aperio software was used to quantify areas of demyelination within the white matter tracks of each of the 6 sections per individual mouse. For analysis of CD4 + T cells spinal cords were embedded in Tissue-Tek O.C.T. (Andwin Scientific, Tryon, NC), flash frozen in liquid nitrogen and stored at 280uC. Blocks were warmed to 220uC prior to cutting 10 mm sections by cryostat. Following fixation with acetone for 2 min at 4uC, non-specific antibody binding was blocked with Cyto Q Background Buster (Innovex Biosciences, Richmond, CA) for 15 min. Sections were stained with anti-CD4 (L3T4) antibody (Vector Laboratories) diluted in Cyto Q Immuno Diluent (Innovex Biosciences) followed by biotinylated rabbit anti-rat, peroxidase ABC reagent and visualized with NovaRED substrate (all from Vector Laboratories). Sections were counter stained with hematoxylin to visualize the nuclei. Spinal cord sections were scanned (406) and digitally imaged at high resolution with an Aperio ScanScope. Mixed glial cultures (,70% astrocytes) were established from neonatal GFAPcR1D and wt mice as previously described [47] . IL-27 secretion was determined 48 h after rIFN-c (100 ng/ml) stimulation. Frozen tissues were homogenized in TRIzol (Invitrogen, Carlsbad, CA) and cDNA prepared as described [47, 48] . Quantitative real-time PCR was performed using 4 ml of cDNA and SYBR Green Master Mix (Applied Biosystems, Foster City, CA) in duplicate on a 7500 Fast Real-Time PCR System (Applied Biosystems). Expression levels were normalized to ubiquitin or GAPDH using the following formula: Statistical significance was determined by two-tailed Student's t test. A value of p,0.05 was considered statistically significant. Figure S1 CD4 + T cell recruitment into the spinal cord during acute EAE. CD4 + T cells were visualized in 10 mm frozen sections of spinal cords from wt and GFAPcR1D tg mice at day 18 p.i. Immunoperoxidase stain (NovaRED chromogen with hematoxylin counterstain). Scale bars = 50 microns. (TIF)

Influence of Mabs on PrP(Sc) Formation Using In Vitro and Cell-Free Systems

PrP(Sc) is believed to serve as a template for the conversion of PrP(C) to the abnormal isoform. This process requires contact between the two proteins and implies that there may be critical contact sites that are important for conversion. We hypothesized that antibodies binding to either PrP(c)or PrP(Sc) would hinder or prevent the formation of the PrP(C)–PrP(Sc) complex and thus slow down or prevent the conversion process. Two systems were used to analyze the effect of different antibodies on PrP(Sc) formation: (i) neuroblastoma cells persistently infected with the 22L mouse-adapted scrapie stain, and (ii) protein misfolding cyclic amplification (PMCA), which uses PrP(Sc) as a template or seed, and a series of incubations and sonications, to convert PrP(C) to PrP(Sc). The two systems yielded similar results, in most cases, and demonstrate that PrP-specific monoclonal antibodies (Mabs) vary in their ability to inhibit the PrP(C)–PrP(Sc) conversion process. Based on the numerous and varied Mabs analyzed, the inhibitory effect does not appear to be epitope specific, related to PrP(C) conformation, or to cell membrane localization, but is influenced by the targeted PrP region (amino vs carboxy).

Prion diseases are a group of fatal neurodegenerative disorders that are associated with conformational conversion of the cellular prion protein, PrP C , which is mainly a-helical with very few beta sheets, into a b-sheet-rich form, PrP Sc [1] [2] [3] [4] [5] . The mechanism by which PrP C is converted to the abnormal isoform is still not clear, but it is presumed to involve a PrP C -PrP Sc complex, with the latter serving as a conformational template [6] . In this model, PrP Sc serves as a template that binds to PrP C and produces a conformational conversion into the abnormal isoform. This raises the issue of whether there are critical contact sites that mediate conversion. If this is the case, interfering with or blocking complex formation should prevent the PrP C to PrP Sc conversion process. Previous reports have described anti-PrP antibodies that can stop or hinder the conversion process add reference 44 and renumber [7] [8] [9] [10] [11] [12] [13] [14] . Protein misfolding cyclic amplification (PMCA) is an assay that mimics the PrP Sc propagation process under cell-free conditions. In this method PrP Sc is amplified by converting PrP C to a PrP Sc seed during incubation with periodic sonication [15] . PrP Sc generated by PMCA is infectious in wild-type animals [16] and can be indefinitely propagated while preserving the properties of the original PrP Sc strain [16] [17] [18] . Furthermore, PMCA has been quite useful in studying the cofactors that influence PrP conversion [19] [20] [21] [22] [23] [24] [25] [26] , and in detecting PrP Sc from biological samples of humans and animals [17, [27] [28] [29] [30] [31] [32] [33] [34] . We hypothesized that antibodies binding to PrP c and/or PrP Sc might hinder or prevent the formation of the PrP C -PrP Sc complex and thus prevent the conversion process. We compared the effect of individual PrP-specific monoclonal antibodies (Mabs) on the PrP C -PrP Sc conversion process using both an N2a/22L cell culture model and the test-tube PMCA system. Our results demonstrate that the Mabs have a range of inhibitory effects on the PrP C -PrP Sc conversion process. The degree of inhibition is Mab specific and more dependent on the antibody targeting region than on the specific epitope being recognized. Furthermore, since the PMCA-based method is dose-dependent and rapid, it may serve as an ideal screening assay for potential inhibitors of both PrP Sc accumulation and the progression of prion diseases. All procedures involving animals and their care were conducted in accordance with the United States Department of Agriculture Animal Welfare Act and the National Institute of Health policy on Humane Care and Use of Laboratory Animals. Tissue samples from uninfected and prion agent-infected mice and hamsters were obtained using protocols approved by the Institutional Animal Care and Use Committee of the SUNY Downstate Medical Center (protocol #'s 07-250-09 and 07-251-09). A 10% normal hamster brain homogenate (NBH) was prepared in phosphate buffered saline (PBS) containing 1% Triton X-100, 4mM EDTA and 1% protease inhibitor cocktail (Abcam). PrPspecific Mabs were generated against recombinant (murine or hamster) PrP or brain-derived proteinase K (PK)-resistant purified PrP Sc [35] from brains of clinical mice infected with the ME7 mouse-adapted scrapie strain or clinical hamsters infected with the 263K hamster-adapted scrapie strain. The Mabs used in this study were purified (Montage Antibody Purification kit; Millipore, Billerica, CA), isotyped (ELISA Mouse Antibody Isotyping kit; Thermo Fisher, Rockford, IL), and epitope mapped ( Table 1) . The immunoreactivity of all the Mabs were analyzed on western blots against denatured, PK-digested and undigested PrP derived from uninfected and infected brain homogenates as well as by ELISA against recombinant PrP. With the exception of Mab 3F4, each of the individual Mabs had equivalent immunoreactivity against murine and hamster PrP Sc on an immunoglobulin concentration basis. All of the Mabs were highly reactive against both hamster PrP C and PrP Sc isoforms and, for the PMCA studies, were individually added to the 10% NBH at a final concentrations of 50 mg/ml. A 10% 263K brain homogenate was prepared in PBS only and diluted to a final concentration of 10 24 . A 100 ml aliquot of this homogenate was initially combined with 10 ml of 10% NBH (with or without added Mab). Each sample was sonicated (QSONIC at 480W power, 60 Amplitude, 40,000 J energy, 90 sec process time, 3 sec pulse on21 sec pulse off), then incubated at 37uC for 1 hr. This was defined as one cycle of serial PMCA (sPMCA). At the completion of each cycle, an additional 10 ml of 10% NBH (with or without Mab) was added. At the end of every five cycles, 100 ml of the total volume was transferred to a new tube containing an equal volume of 10% NBH (with or without Mab) and the cycling reactions continued. At the completion of 40 cycles (sPMCA 40 ), 500 ml from each sample was PK-treated (100 mg/ml final concentration, 50uC, 30 min), followed by the addition of protease inhibitor cocktail. The sample was heated (100uC, 10 min) and then centrifuged at 16,0006g for 2 min at room temperature. The supernatant was combined with 6X Laemmli sample buffer, and 50 ml was electrophoresed in a 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) followed by transfer to nitrocellulose membrane. The membrane was blocked for 1 hr in PBS containing 0.1% Tween 20 (PBST) with 5% non-fat dry milk and incubated with 2 mg/ml biotinylated Mab 08-6/2F11. The membrane was washed 3 times (10 min each) with PBST, incubated for 60 min in HRP-conjugated streptavidin (Invitrogen) (1:5000 in PBST containing 5% non-fat dry milk) followed by 3 additional PBST washes and detection of proteins with ECL Supersignal West Dura kit (Thermo Fisher). Quantification of PrP Sc was performed by densitometric analysis using NIH Image J software. Cellulose membranes spotted with 99 overlapping 13-mer PrP peptides were produced as previously described [36] . The membranes were blocked with 5% non-fat dry milk/tris-buffered saline containing 0.1% Tween 20 (TBST) probed with antibody diluted 1:5000 in 1% normal goat serum/TBS at 4uC overnight, followed by horseradish peroxidase (HRP)-conjugated goat antimouse secondary (Cappel 55570) for 2 hours at room temperature, and detected using Millipore Immobilon Western chemiluminescent HRP substrate (Cat WBKLS0500). Membranes were regenerated for re-use by shaking with dimethylformamide for 30 minutes, then 8M urea/50mM Tris-HCl pH 8.0/1% b-mercaptoethanol (b-MC)/1%SDS overnight at 37uC, followed by a 30 min wash in the same buffer, and then twice for 30 minutes each in 50% methanol/glacial acetic acid, and finally three times for 5 minutes each in methanol. After air drying membranes were stored in a sealed container at room temperature. Murine neuroblastoma N2a cells (ATCC line CCL 131) were grown in the Minimal Essential Medium supplemented with 10% FBS, penicillin and streptomycin and infected with 2% 22L brain homogenate as described previously [12] . Following infection, the amount of PrP Sc in 200 mg cell lysate aliquots of the N2a/22L cells was determined by PK digestion (1 mg/ml PK for 30 min at 37uC), SDS-PAGE on 12.5% Tris-tricine gels [37] and western blot analysis as previously described [12] . For treatment of cells with Mabs, N2a/22L cells (from the fifth passage after infection and higher) were plated in six-well plates and once the cells were 70-80% confluent, Mabs were added at a final concentration of 10 mg/ml and incubation was continued for 96 hr. Each Mab was tested in three independent experiments using independently infected cell lines. Each experiment included both a positive control (untreated N2a/22L cells) and a negative control (N2a cells), which were subjected to PK digestion. The level of PK-resistant PrP Sc was measured in western blots using HRP-conjugated sheep anti-mouse IgG as the secondary reagent and ECL Supersignal West Dura kit. Membranes were exposed to X-ray film (X-Omat Blue XB-1; Kodak, New Haven, CT,) with a constant exposure time of 30 sec. The films were converted into eight-bit grayscale digital files. Quantification of PrP Sc was performed by densitometric analysis using NIH Image J software v. 1.34. Areas under the curves for three PrP bands representing non-, mono-and diglycosylated isoforms of the protein were summarized from each sample to calculate the total amount of PrP and expressed as percentages of the average value from a positive control (untreated N2a/22L), whereas the optic density of the background was taken from negative control lanes (N2a cells). The PrP-specific Mabs that were evaluated for their ability to prevent PrP C to PrP Sc conversion have linear epitopes that span the entire prion protein from the amino to the carboxy terminus ( Table 1 , Fig. 1 ). We used N2a cells persistently infected with the 22L mouseadapted scrapie strain (N2a/22L) to evaluate the affect of each Mab on PrP Sc formation ( Fig. 2A and 2B) . Treatment with the Mabs did not result in any cytotoxicity to the N2a/22L cells throughout the incubation period. Further, incubation of the. N2a/22L cells with 10 mg/ml purified, irrelevant mouse IgG had no effect on PrP Sc formation compared to untreated N2a/22L cultures (Figs. 2A and 2B). Mab 3F4 did not reduce PrP Sc formation compared to control N2a/22L cultures lacking Mab. Mab 3F4 does not react with mouse prion protein so this was not surprising [38] . The ability of a singly added Mab to inhibit PrP Sc formation was not related to a specific epitope since all of the remaining singly added Mabs inhibited PrP Sc formation to varying degrees. Of the individually added Mabs, 5D6 was the most effective at inhibiting PrP Sc formation (95% inhibition) while 3A2 was the least effective (38% inhibition). Targeting the amino terminus with Mab 7E4 was effective at inhibiting 73% PrP Sc formation as was targeting the octapeptide repeat region using Mab 10E4 which resulted in almost 90% inhibition. Strangely, although their epitopes overlap, Mab 11F12 was less effective than Mab 5D6 at inhibiting PrP Sc formation (53% vs 95% inhibition). This is in contrast to Mabs 8E9 and 1B11, which have overlapping epitopes with 8E9 being more expansive, and resulted in 52% and 42% inhibition, respectively. The combination of 5D6 and 11F12 did not result in an additive inhibitory effect and, in fact, resulted in less inhibition than either one alone. This was confirmed in studies where the addition of 8E9 to 5D6 and 11F12 caused a 45% PrP Sc inhibition, which was slightly better than 8E9 alone, although the predicted additive inhibitory effect of 63% for the three Mab combination (48% for 8E9 plus 15% for the 5D6 and 11F12 combination) was not observed. Studies were performed with PMCA to determine whether a cell-free system can recapitulate the effect of Mabs on PrP C conversion observed in infected cells. This system also allowed us to evaluate whether accessibility of Mab to membrane associated PrP C in the living cells influences the PrP C to PrP Sc conversion process. Mabs (12-50 mg/ml final concentration) were added throughout the sPMCA 40 protocol along with the 10% NBH spiked with a 10 24 dilution of 263K infected brain homogenate as described in the Methods section. This dilution of infected brain homogenate does not result in detectable PK resistant PrP Sc immunostaining (Fig. 3A) and, therefore, did not interfere with the detection of newly formed PrP Sc . At the completion of sPMCA 40 , the samples were digested with PK (100 mg/ml) and analyzed on immunoblots using biotinylated Mab 2F11 which reacts equally with both hamster PrP C and PrP Sc . It is interesting to note that although the 263K-infected brain homogenate displayed the 3 band pattern typical for the multiple glycosylated forms of PrP C and PrP Sc (Fig. 3A) , the sPMCA 40 products in the positive controls and Mab-treated reactions consisted of only a single diglycosylated 30 kDa PrP Sc band observed after PK digestion at the higher levels of inhibition, .50%, but had two bands or a smear when there was less inhibition (Fig. 3B) . PMCA in the presence of Mabs was also used to study the importance of binding site specificity in the PrP C to PrP Sc conversion process (Fig. 3) . We performed sPMCA with different Mab concentrations to determine the minimum amount of Mab necessary to inhibit the conversion process. Using a 10 24 dilution of 263K-infected hamster brain homogenate as the PrP Sc seed and a 10% normal brain homogenate (NBH) as the source of PrP C , we tested the ability of Mabs to inhibit the conversion of PrP C to PrP Sc . For each Mab, final concentrations of 12 mg/ml (lanes 3, 6, 9, 12, 15, 18, 21 and 24), 25 mg/ml (lanes 4, 7, 10, 13, 16, 19, 22 and 25) and 50 mg/ml (lanes 5, 8, 11, 14, 17, 20, 23, and 26) were prepared in hamster NBH and used in the sPMCA reactions. Compared to sPMCA 40 , which contained no Mab (lane 1) and with the exception of 02-3/3A2, the majority of the PrP-specific Mabs inhibited the conversion process in a dose-related manner although some were more effective than others. Mabs 7E4, 10E4, 11F12, 8E11, and 8E9 completely inhibited the conversion process at 50 mg/ml while Mabs 1B11 and 2F7 inhibited the conversion process to a lesser degree. The inhibition caused by the Mabs was a specific response since sPMCA 40 studies replacing Mabs with purified normal mouse IgG (at 12-50 mg/ml) in the 10% NBH did not cause any inhibition of PrP Sc formation (data not shown). It is interesting to note that, with the exception of only 8E9, the epitopes for all the Mabs that caused complete inhibition are located in the amino half of the PrP while those that caused incomplete inhibition are located in the carboxy half of PrP. There was good correlation between the extent of PrP Sc inhibition when 10 mg/ml Mab in cell culture was compared to 12 mg/ml Mab with sPMCA 40 . A separate study using sPMCA 40 demonstrated that Mabs 3F4 and 5D6 caused complete inhibition of PrP Sc formation at 12-50 mg/ml (Fig. 4) . Therefore we extended those studies and evaluated the effects of Mabs 3F4 and 5D6 using a wider range of Mab concentrations (1.5-50 mg/ml). Compared to the other antibodies in this study, Mabs 3F4 and 5D6 had the most pronounced effects on PrP Sc formation as demonstrated by the low concentrations of 3 and 6 mg/ml, respectively, causing complete inhibition (Fig. 4) . The potent inhibitory effect of 5D6 on PrP Sc observed using sPMCA 40 coincides with its dramatic effect in the N2a/22L culture model. Furthermore, the poor PrP Sc inhibition by 3A2 with sPMCA 40 (Fig. 3B) corresponded well with the poor inhibition (only 32% reduction compared to negative control) observed in the cell culture system ( Fig. 2A and 2B ). Currently, there is no effective treatment for prion diseases. To date, hundreds of chemical compounds have been identified that antagonize prion propagation in vitro in cell culture-based assays and/or in vivo in animal studies [39] [40] [41] [42] [43] . Unfortunately, many compounds efficient in in vitro studies were only effective in animal models if treatment was begun before or close to the time of inoculation with the infectious agent [44] . Furthermore, many of the candidate compounds have limited usefulness clinically due to toxicity or their inability to cross the bloodbrain barrier [e.g. Congo red [45] , iododoxorubicin, b-sheet breakers]. Additional therapeutic and/or prophylactic strategies have been and continue to be pursued. Vaccination with recombinant mouse PrP delays the onset of prion disease in mice [46] . Passive immunization with anti-PrP antibodies was shown not only to inhibit formation of PrP Sc in a cell-free system [47] , but was also shown to prevent infection of susceptible N2a cells [7] and to inhibit prion replication in infected cells [8, 47, 48] . The effectiveness of these treatments were also dependent on when they were administered relative to the time of infection. In an initial passive immunization study using wild-type CD1 mice, Mabs 8B4 (to mouse PrP residues 34-52) and 8H4 (to mouse PrP residues 175-185) given immediately after challenge with 139A scrapie by intraperitoneal (IP) injection (50 mg/week), resulted in a significant prolongation of the incubation period with 10% of the 8B4 treated animals remaining disease free in the group challenged with a lower dose of PrP Sc [10] . In another study using higher antibody doses (4000 mg/week IP) of either ICSM 18 (to mouse PrP residues 146-158) or ICSM 35 (to mouse residues 95 to 105), prion infection from a peripheral source was completely prevented if treatment was continued for 7 or 30 days immediately following PrP Sc challenge [9] . Furthermore, a transgenic mouse model that expresses Mab 6H4 is resistant to prion infection via IP injection by a mechanism that involves either perturbation of cellular PrP trafficking/PrP C degradation or disruption of the PrP C -PrP Sc interaction [49] . Previous studies have reported that the 132-140 portion of PrP C [8] or the 132-156 region of PrP [50] [51] [52] [53] are important for the generation of PrP Sc . Rigter et al. [54] found two high affinity binding regions for protein-protein interactions using ovine peptide-arrays: (i) sheep-PrP peptides 43-102, including the amino-terminal octarepeats, and (ii) sheep-PrP peptides 134-177 Figure 2B for representative western blots). PrP Sc western blots were quantitated and the amount of inhibition was determined relative to N2a/22L control cultures. The controls consisted of cells both in the absence of Mab and in the presence of normal mouse IgG. The % PrP Sc inhibition plotted represents the mean 6 SD from three independent experiments as described in Methods. doi:10.1371/journal.pone.0041626.g002 which encompasses most of the scrapie susceptibility-associated polymorphisms in sheep. Moroncini et al. [55] found that residues within the 89-112 and 136-158 segments of PrP C are key components of the PrP C -PrP Sc complex. Beringue et al. [11] reported that antibodies exclusively binding PrP C were relatively inefficient inhibitors of PrP Sc accumulation compared with antibodies that additionally recognize disease-associated PrP isoforms. Féraudet et al. [56] screened 145 anti-PrP Mabs for their capacity to inhibit PrP Sc replication in infected N2a or Rov9 cells. They identified four different linear epitopes that hindered the PrP C to PrP Sc conversion: the amino terminal region 26-35, the octarepeat region 59-89, the intermediate region 97-102, and the central region 130-160. The observation that antibodies that bind to the amino terminus of the prion protein are capable of inhibiting conversion suggests that the Figure 3 . A. Western blot of 263K brain homogenate that was used as the seed for PMCA. Dilutions of the brain homogenate was prepared and either untreated (lanes 1 and 2) or PK-treated (lanes 3 and 4) prior to SDS-PAGE and western blotting. A 1023 dilution prior to and after PK demonstrates the three protein banding pattern typical for 263K brain homogenate whereas no bands are visible at a 1024 dilution of the same homogenate. B. Western blotting of the PMCA products following sPMCA40 in the absence and presence of PrP-Specific Mabs. Fourty cycles of serial PMCA was carried out in the absence or presence of Mabs as described in the text. Each Mab was added at a final concentration of 12, 25, and 50 mg/ ml. Following PK treatment, the PMCA products were subjected to SDS-PAGE, western blotted and immunostained for PrPSc. The protein bands were quantitated and the level of PrPSc inhibition, relative to the no Mab and normal mouse IgG controls, were determined. doi:10.1371/journal.pone.0041626.g003 Figure 4 . Influence of Mabs 3F4 and 5D6 on PrP Sc formation following sPMCA 40 . Mabs 3F4 and 5D6 were added to sPMCA 40 at final concentrations of 0-50 mg/ml. The PMCA products were PK treated and western blotted. The PrP Sc was quantitated and the level of PrP Sc inhibition was determined relative to control reactions. doi:10.1371/journal.pone.0041626.g004 endogenous proteolytic cleavage occurs after the site of conversion. To more completely explore the possible therapeutic effect of anti-PrP antibodies, and to establish another system to analyze the influence of Abs on the conversion process, we screened Mabs produced in our laboratory for their capacity to inhibit PrP Sc formation. This screening was performed using N2a/22L cells and cell-free sPMCA. In N2a/22L cultures, all Mabs that react with mouse PrP reduce PrP Sc formation although with varying efficiency. Thus, similar to previous results [55] , we found that the ability to inhibit PrP C to PrP Sc conversion was not restricted to a single epitope or limited to a specific region of the protein. However, the greatest inhibition was observed with Mabs that targeted epitopes in the amino terminal, unstructured region of the PrP. The greatest inhibition in the N2a/22L cells was with Mab 5D6. This is consistent with a prior study using Mab 6D11 (anti-PrP residues 95-105) which in a screen of multiple Mabs, only one produced the greatest inhibition (,100%) [12] . Mab 6D11 has also been shown to have some efficacy in vivo prolonging the presymptomatic incubation period [57] . The Mab inhibition results obtained using PMCA were similar to that found in the cell culture system. PMCA has the advantages over the cell culture model of being cost-effective, simple, rapid, sensitive, and more amenable to studies of dose dependence. For identification of potential candidate Mabs that might have in vivo activity it is likely that such Mabs would have to produce 90 to 100% inhibition in the much simpler in vitro systems. The interaction of PrP C to PrP Sc is critically dependent on the structural compatibility of the molecules as supported by the existence of a species barrier for prion infection, related to minor differences in the primary sequence of PrP C in different species. Therefore, it is not surprising that antibodies that may alter or mask the critical epitopes on PrP C and/or PrP Sc , involved during the mutual conformational complementarity required in prion propagation, will be inhibitory for prion replication. Although many anti-PrP antibodies targeting different regions of PrP may have some therapeutic effect in vitro, it is not clear how this relates to their efficacy in vivo. On the one hand, it is tempting to speculate that only the antibodies exhibiting near complete inhibition in vitro would be effective in vivo given the obstacle of the blood brain barrier and access to PrP in cells. However, it is also possible that only partial inhibition of conversion is required in vivo allowing the cells to ''recover''. In either case, it would be advantageous for these therapeutic antibodies to have high affinities of binding to PrP C and/or PrP Sc , as well as targeting specific critical PrP domains. One can hypothesize that the simultaneous targeting of more than one critical epitope will lead to greater benefits. However, co-treatment experiments performed with a mixture of two antibodies compatibly binding cell-surface PrP C did not show any benefit with compared to treatment involving a single Mab in our current experiments. In a previous study [58] , we demonstrated synergistic binding with one of our antibody pairs. Synergistic binding of inhibitory Mabs, i.e. reaction with an antibody that increases the binding of the second antibody, would be predicted to enhance the inhibitory effect. Further studies with antibody pairs fitting this description will be required to test this hypothesis. In addition, determining the significance of the Mab's ability to bind both PrP C and PrP Sc may provide further insight into the conversion process. Conceived and designed the experiments: RR TW. Performed the experiments: BC RP TW RR. Analyzed the data: BC RP TW RR. Contributed reagents/materials/analysis tools: RP TW RR. Wrote the paper: RR.

Predictors and outcome of patients with acute respiratory distress syndrome caused by miliary tuberculosis: a retrospective study in Chongqing, China

BACKGROUND: Miliary tuberculosis (TB) is an uncommon cause of acute respiratory distress syndrome (ARDS) with a high mortality. The aim of the present study was to evaluate the clinical characteristics, predictors and outcome of patients with ARDS caused by miliary TB. METHODS: A retrospective study was conducted among patients with a diagnosis of ARDS with miliary TB in four hospitals from 2006 to 2010. Medical records and laboratory examinations of these patients were taken during the first 24 h of admission. RESULTS: Eighty-five patients with miliary TB developed ARDS, 45 of whom survived (52.9%). The median age was 36.6 ± 12.5 years with 38 males (44.7%). Diabetes mellitus (DM) was the most common underlying disease (18.8%).ICU mortality was 47.1%. The time from admission to anti-tuberculosis therapy was 4.5 ± 2.0 days. Mean duration of mechanical ventilation was 8.5 ± 3.0 days in all patients. Duration of time to diagnosis, time from diagnosis to mechanical ventilation, and time to anti-tuberculosis therapy were significantly shorter in survivors than those in non-survivors. Diabetes mellitus (OR 5.431, 95%CI 1.471-20.049; P = 0.005), ALT (70-100U/L, OR 10.029, 95%CI 2.764-36.389; P = 0.001), AST (>94U/L,OR 8.034, 95%CI 2.200-29.341; P = 0.002), D-dimer (>1.6mg/L, OR 3.167, 95%CI 0.896-11.187; P = 0.042), hemoglobin (<90g/L, OR 14.824, 95%CI 3.713-59.179; P = 0.001), albumin (<25g/L, OR 15.896, 95%CI 3.975-63.566; P = 0.001) were independent predictors of ARDS development in the setting of miliary TB. CONCLUSIONS: Accurate diagnosis, early initiation of anti-tuberculosis therapy and mechanical ventilation are important for the outcome of patients with ARDS caused by miliary TB. DM, ALT, AST, D-dimer, hemoglobin, and albumin are independent predictors of ARDS development in patients with miliary TB.

Tuberculosis(TB) remains a major and global health disease [1, 2] . Recent studies have shown the link between acute respiratory distress syndrome (ARDS) and pulmonary TB [3, 4] . Pulmonary TB complicated by ARDS is often found in the setting of miliary TB [3] [4] [5] [6] . Most reports on ARDS caused by miliary TB are small numbers of patients in the case reports. Compared with miliary TB alone, miliary TB with ARDS portends a higher mortality of 33-90% [7] [8] [9] . Duration of miliary TB beyond 20 days tends to markedly increase the risk of ARDS [10] . It is very important for recognition of ARDS caused by miliary TB. Despite being a well-documented entity, miliary TB complicated by ARDS remains a challenging diagnosis due to its variable clinical manifestations and low morbidity. Some predictors such as AST, and ALT [4, 10] in miliary TB with ARDS have drawn our attention. Numerous case reports have mentioned that TB with ARDS is more common than TB complicated by anemia [4, 11] and hypoproteinemia [6, 11] . Thus, identifying the predictors of miliary TB associated with ARDS can play an important role in diagnosis and therapy. The aim of the present study was to determine the predictors and their impact on outcome based on a retrospective analysis of patients with ARDS caused by miliary TB in four hospitals of Chongqing in China, with the hope that this will contribute to a better understanding and improved management of the disease. Over a 5-year period (2006.03. Our research protocol was approved by the institutional review boards of all participating institutions. The original case records including clinical profiles and laboratory parameters at admission were gathered from the registration departments. Age, sex, past medical history, underlying diseases, PaO2/FiO2, time from diagnosis to mechanical ventilation, time from admission to antituberculosis therapy, duration of time to diagnosis, and lengths of stay in the ICU and in the hospital were collected. Acute Physiology and Chronic Health Evaluation (APACHE) III score were calculated on the day of diagnosis with ARDS. The results of acid-fast bacilli(AFB) smears, culture of respiratory specimens including sputum, tracheal aspirate or bronchoalveolarlavage (BAL)fluid and histopathological examination were recorded. Laboratory data including aspartate aminotransferase (AST), alanine aminotransferase (ALT), erythrocyte sedimentation rate (ESR), D-dimer, hemoglobin, and albumin were taken during the first 24 h of admission. The diagnosis of miliary TB was made based on: equality of the size, distribution, density miliary-like nodules bilaterally diffused on chest radiography by at least 2 independent of radiologists. Pulmonary TB was confirmed by at least one of the three following criteria:1)positive AFB smear and/or culture for M. tuberculosis from respiratory specimes;2) histopathological identification of TB granulomain in biopsied tissues of lung, and/or pleura;3)clinical and radiographic improvement after anti-tuberculosis treatment [4] . The diagnosis of ARDS was made in accordance to the diagnostic criteria of European-American Consensus Conference on ARDS [12] :acute in onset with PaO 2 /FiO 2 ≤ 200mmHg, bilateral infiltrates seen on chest radiograph, and pulmonary artery wedge pressure ≤ 18 mm Hg. Survivors were defined as patients who survived to discharge from hospital. Patients with human immunodeficiency virus (HIV), H1N1 Influenza A,and procalcitonin (PCT) positive were excluded. Continuous variables were presented as mean ± standard deviation(SD) or median and compared using an unpaired t-test and the Mann-Whitney U test. Categorical variables were compared using the Chi-squared test. Multivariate logistic regression analysis was performed to determine the predictors. All statistical analysis were performed using SPSS 13.0. P <0.05 was considered significant. Patients' clinical characteristics and outcome are presented in Table 1 . All patients experience typical symptoms of miliary TB. Diabetes mellitus (DM) was the most common underlying disease (18.8%). Thirty-three (38.8%) patients had been initially misdiagnosed with viral pneumonia, hypersensitivity pneumonitis, acute interstitial pneumonia, fungal pneumonia, alveolar cell carcinoma, or meningitis for a median of 7.2 ± 3.4 days from admission. The diagnosis of miliary TB was established by AFB smear and/or culture of respiratory specimens (including sputum,tracheal aspirate or BAL fluid) in 61 patients(71.8%), by histopathological examination of tissue biopsy in 11 patients(12.9%), and by clinicoradiological diagnosis in 13 patients(15.3%). Bacterial isolate drug sensitivity data were available from 43 patients (50.6%), 3(3.5%) demonstrated at least single drug resistance. The time from admission to anti-tuberculosis therapy was 4.5 ± 2.0 days. All 85 patients with ARDS were prescribed anti-tuberculosis medication consisting of isoniazid, rifampicin, ethambutol, and pyrazinamide. Mechanical ventilation was necessary in all 85 patients. Thirty-eight patients(44.7%) required invasive mechanical ventilation while the rest were given non-invasive mechanical ventilation with BiPAP. Mean duration of mechanical ventilation was 8.5 ± 3.0 days with ICU mortality of 47.1%. Thirty-five patients (41.2%) received glucocorticoid therapy (methylprednisolone:80 mg/day) intravenously for a maximum of 5 days when anti-tuberculosis therapy was started. The use of glucocorticoids was associated with a mortality of 22.9% (8/35) compared with 76.0% (38/50) in those who were not treated with glucocorticoids. Comparison between patients with miliary TB developing ARDS and patients with miliary TB alone are shown in Table 2 . Comparison between survivors and non-survivors of ARDS patients are shown in Table 3 . Duration of time to diagnosis, time from diagnosis to mechanical ventilation, and time from admission to anti-tuberculosis therapy were significantly shorter in survivors than non-survivors. Also, DM, ALT, AST, Ddimer, hemoglobin, and albumin showed significant difference between the survivor group and non-survivor group. The 3 pregnant patients underwent termination of pregnancy, one of whom died of respiratory failure. Positive likelihood ratio were performed to analyze sensitivity and specificity of the predictors level(the greater the ratio the greater probability of true positive in a positive result) ( were independent predictors of ARDS development caused by military TB. In the present study,our results demonstrated accurate diagnosis and early therapeutic management were crucial to optimizing outcome and DM, ALT, AST, D-dimer, hemoglobin, and albumin are independent predictors of ARDS development in patients with miliary TB. Tuberculosis remains a major cause of morbidity and mortality around the world, especially in developing countries [13, 14] . According to the 13th annual tuberculosis report of the World Health Organization (WHO), there were an estimated 9.27 million new cases worldwide in 2007, an increase from 9.24 million in 2006 [1] . Further, miliary TB is an uncommon cause of ARDS with a high mortality [4, 15] . The current study was performed in Chongqing, the fourth central municipality of China, which has a prevalence rate of 0.54%, higher than the national average. In our study, the average age of patients with ARDS caused by miliary TB was younger than that previously reported due to the possible reason of host factors including region, race, and environment [3, 4] . Mechanical ventilation is an important treatment for miliary TB complicated by ARDS. Prompt mechanical ventilation from the day of diagnosis can effectively improve outcome, which benefits for the management of ARDS caused by miliary TB. Also, accurate diagnosis and early anti-tuberculosis therapy are crucial to the treatment of miliary TB with ARDS. However, the high overall mortality is attributed to case mix, misdiagnosis, and severity of illness, which ultimately lead to the delay in the initiation of anti-tuberculosis therapy or mechanical ventilation. Albumin plays an important role in regulating plasma osmolality. Hypoproteinemia accelerates fluid exudation, promotes alveolar edema, and contributes to ventilation-perfusion imbalance. Also, in infected mycobacterium tuberculosis, inflammatory cells accumulate in the alveolar spaces, releasing granular enzymes and oxidants which participate in local inflammation and overlapping reactions and damaging the alveolar basement which allows increase in cellular permeability that aggravates oxygen dysfunction and consequently causes ARDS [16, 17] . If the process continues, cytokines activate an inflammatory cascade reaction and lead to other organs dysfunction, resulting in increases in AST and ALT. The changes in AST, ALT, and serum albumin were significantly different between survivor and non-survivor groups. The results showed that AST, ALT, and serum albumin could be independent predictors of ARDS development in miliary TB. A number of studies have documented anemia associated with TB [18] [19] [20] . It is well established that erythropoiesis is suppressed by inflammatory mediators in anemia of chronic infection or inflammation and anemia caused by chronic infections including TB results from the effects of cytokines that mediate the inflammatory response,which may provides the development for ARDS [21] [22] [23] . As we know,the severity of anemia is mainly determined by hemoglobin level. The decrease in hemoglobin is assumed as a reflection of inflammation, which can explain the relationship between ARDS and hemoglobin in our study. This relationship is supported by our finding of hemoglobin being an independent predictor of ARDS development in miliary TB. D-dimer is a fibrin degradation product formed during fibrinolysis, the process of breaking down a blood/fibrin clot. Our study showed that D-dimer was an independent predictor of ARDS development and mortality. Hyperglycemia is known to have a proinflammatory effect, and may be correlated with poor outcome in hospitalized/critically ill patients. Consistent with this correlation, odds ratio analysis after ARDS development identified DM as a risk factor. To further understand the accurate levels of the predictors for predicting ARDS, we used stratified analysis for each index. Positive likelihood ratio was included in the study due to a combination of sensitivity and specificity to reflect the reality of the indicators. ALT (>70-100U/L), AST (> 94U/L), D-dimer (>1.6mg/L), hemoglobin (<90g/ L), and albumin (<25g/L) at the time of admission were independent predictors for ARDS development in the setting of miliary TB. Simple miliary TB can damage organs, inducing mild increases in ALT and AST, although these incremental changes are too small to have any accuracy in predicting the occurrence of ARDS. However, when one develops ARDS, inflammatory mediators will seriously damage the organs, resulting in significantly higher increases in the indexes than those in simple miliary TB, which may translate into higher accuracy for predicting ARDS. Though ESR was significantly elevated in miliary TB, it did not reach statistical significance in association with ARDS. Our findings suggested that ESR has minimal value in predicting ARDS. In addition, for patients with miliary TB who developed ARDS during pregnancy, there are few case reports with favorable outcome. Our data suggested that pregnancy was a risk factor for ARDS, but this clinical observation was limited by the small sample size. The efficacy of systemic corticosteroids is well-documented for several extrapulmonary complications of tuberculosis such as tuberculous meningitis and tuberculous pericarditis [24] [25] [26] [27] [28] , as well as ARDS [29] [30] [31] [32] . The role of glucocorticoids remains controversial in the management of miliary TB complicated by ARDS [33, 34] . In our study, patients treated with glucocorticoids had a lower mortality than those who did not, which might suggest that methylprednisolone at a dose of 80 mg/day was given intravenously at the time when anti-tuberculosis therapy was started might be benefit for ARDS associated with miliary TB. The limitations of this study are studies with large numbers of patients may be required to validate the observations due to a relatively small sample size in the present study. In conclusion, the mortality of ARDS caused by miliary TB remains high. Accurate diagnosis and early therapeutic management of therapy including anti-tuberculosis agents and mechanical ventilation are crucial to optimizing outcome. DM, ALT, AST, D-dimer, hemoglobin, and albumin are independent predictors of ARDS development in patients with miliary TB.

Symptomatic Venous Thromboembolism Is a Disease Related to Infection and Immune Dysfunction

The characteristics of human genomics and cellular immune function between clinically symptomatic venous thromboembolism (VTE) and controls were systematically compared to explore the immunologic pathogenesis of VTE. Microarray assay showed the mRNA expressions of genes related to non-specific cellarer immune and cytokines were significantly down-regulated. Abnormal expressions of CD3+, CD4+, CD8+, NK marker CD16+56+, CD19 and aberrant CD4+/CD8+ ratio were detected in 54 among 56 patients. In PE patients, microarray assay revealed the imbalance in the expressions of genes related to the immune system. The expressions of genes related to non-specific immune cells and cytokines were markedly up-regulated and those associated with cellular immune were dramatically down-regulated. In VTE patients, cytological examination indicated the functions of NK cells were significantly compromised, and the antigen recognition and killing function of T cells markedly decreased. The consistence between genomic and cytological examination suggests the symptomatic VTE is closely associated with the infection and immune dysfunction.

Venous thromboembolism (VTE) including acute pulmonary embolism (APE), chronic thromboembolic pulmonary hypertension (CTEPH) and deep venous thrombosis (DVT) is a global disease. The high morbidity, high misdiagnosis rate and high mortality render PE as a worldwide health problem (1) . In 1965, Egeberg et al first described a parentage with inherited antithrombin deficiency, a member of which repeatedly developed DVT. The antithrombin deficiency is an autosomal dominant genetic disease and they first proposed the concept of thrombophilia (2) . However, evidence on the genetic pathogenesis of VTE is rarely identified in a majority of VTE patients (3) . According to the pathogenesis, VTE is classified into genetic VTE and acquired VTE which is frequently found in clinical practice. The symptomatic VTE is an entity of hereditary VTE and acquired VTE. The American College of Chest Physicians (ACCP) has recommended the guidelines for the diagnosis and prevention of VTE since 1995. A total of 9 issues have been published by the end of 2012 (4) , and the contents have also been continuously renewed. ACCP proposed that Trauma, surgery, old age, malignancies, pregnancy, heart failure and oral administration of contraceptives are the main risk factors of VTE and ACCP also proposed the concept of risk stratification for prophylaxis of VTE by which differ-Ivyspring International Publisher ent managements were used for prevention from VTE in different risk patients (5, 6) . The incidence of symptomatic VTE is not reduced but gradually increased, the reason of which may be related to the unclear pathogenesis of VTE (7) . In 2006, Smeeth et al reported the occurrence of VTE was associated with infection, and VTE was frequently observed within 2 weeks after infection (8) . In 2010, we reported VTE in multiple organs of a patient who died of SARS, suggesting viral infection is a cause of systemic VTE (9) . In addition, in 2010, we detected virus-like microorganisms in the lymphocytes of a young pulmonary hypertension patient with increased D-Dimer, which morphologically confirmed the attack of T cells by virus, and peripheral decreased CD3 + and CD8 + level also indicated virus infection caused significantly compromised function of T cells (10) . In 2011, we reported the decreased CD3 + and CD8 + level with an increased CD4 + /CD8 + ratio in a group of CTEPH patients, suggesting T cellular immune dysfunction and ratio imbalance in CTEPH patients (11) . In the present study, the whole human genome microarray and Gene Ontology (GO) analysis were employed to detect the targeting of symptomatic pulmonary embolism (PE). In addition, flow cytometry was performed to investigate the changes in immune cells in VTE patients, which aimed to validate the results from genome analysis. Based on the findings above, the relationship between immune dysfunction and clinical symptomatic VTE was analyzed. The CPR level in part of VTE inpatients was determined. Genomic study 20 PE inpatients and 20 controls were randomly selected in Cardiology Department, Tongji Hospital of Tongji University. In the PE group, there were 11 males and 9 females, with a mean age of 70±14 years (44~89 years). There were 13 patients with acute PE and 7 with CTEPH. The pulmonary artery pressure was 50-108 mmHg in CTEPH patients. In the control group, 20 patients (11 males and 9 females) with a mean age of 72±14 years (44~91 years) were enrolled during the same period. No significant difference in age was found between PE patients and control patients (P>0.05). Malignancies, use of immunosuppresants or autoimmune diseases were excluded in all patients. A total of 56 clinically proven VTE patients (34 with APE, 9 with DVT and 13 with CTEPH) aged 64±18 years (13~88 years) were recruited from Tongji Hospital, including 26 males and 30 females. The immunological parameters of these patients were all compared with the detection intervals of normal population. The diagnostic criteria of acute PE were the same as the above and the diagnosis of DVT was based on the criteria previously reported (12) . The criteria for CTEPH developed by American AHA were used for the diagnosis of CTEPH (13) . In the present study, 13 patients had the mean pulmonary artery systolic pressure of 54-106 mmHg. Malignancies, use of immunosuppresants or autoimmune diseases were excluded in all patients. All patients were from Shanghai, China. The present study was approved by the Ethics Committee of Tongji Hospital and informed consent was obtained before study. A total of 5 ml of venous EDTA anti-coagulated blood was obtained from patients of both groups and mononuclear cells were isolated by density gradient centrifugation. Red blood cell lysis buffer (Qiagen, Hilden, Germany) was used to isolate mononuclear cells and total RNA was extracted from mononuclear cells with TRIzol (Invitrogen, Carlsbad, USA) followed by purification with RNeasy column (QIAGEN). Treatment with DNase was performed to avoid the influence by genomic DNA. Quantification of extracted RNA was performed with Nanodrop ND-1000 spectrophotometer (Nanodrop Technology, Cambridge, UK). Agilent G4112A Whole Human Genome Oligo Microarrays were purchased from Agilent (USA). A microarray is composed of 44,290 spots including 41675 genes or transcripts, 314 negative control spots, 1924 positive control spots and 359 blank spots. The functions of more than 70% of genes in the microarray have been known. All patients of both groups were subjected to microarray analysis. Indirect approach was applied to mark samples. About 1 μg of total RNA was reversely transcribed into double strand cDNA. After purification, in vitro amplification was performed with Agilent Low RNA Input Linear Amplification Kit (Agilent, Pal alto, USA) and modified UTP [aaUTP, 5-(3-aminoally1)-UTP] was used to replace UTP. The integrated aaUTP can interact with Cy3 NHS ester forming fluorescent products which are then used for hybridization. The integration rate of fluorescence can be determined with a NanodropND-1000 spectrophotometer. Then, hybridization mixture was prepared with Agilent oligonucleotide microarray in situ hybridization plus kit. About 750 ng of fluorescent products were fragmented at 60℃ and hybridization was conducted in Human Whole-Genome 60-mer oligo-chips (G4112F, Agilent Technologies) at 60℃ for 17 h at 10 rpm. After hybridization, the chips were washed with Agilent Gene Expression Wash Buffer according to manufacturer's instructions. Original signals were obtained Agilent scanner and Feature Extraction software. The standardization of original signals was carried out with RMA standardized method and standardized signal values were used for screening of differentially expressed genes. The spots in the microarray were randomly selected and their expressions were confirmed by RT-PCR. Among genes with differential expressions, 3 genes were randomly selected and these genes and house keeping gene (GAPDH) were subjected to RT-PCR. The relative expressions were expressed as the expressions of target genes normalized by that of GAPDH (2 -△△ Ct ). Melting curve and 2 -△△ Ct method were used to compare the difference in the expressions between control group and PE group. Results from RT-PCR were consistent with microarray analysis. Gene Ontology organizes gene function into hierarchical categories based on biological process, molecular function and cellular component. Fisher's exact test was applied for over representation of selected genes in GO biological categories. In order to assess the significance of a particular category by random chance, false discovery rate (FDR) was estimated for all of categories. After 5,000 re-samplings, FDR was defined as FDR=1-Nk/T, where Nk refers to the subtracted number which was from Fisher's test in random samples. We specified the threshold of significant GO as p-value<0.05, FDR<0.05 and enrichment parameters. Enrichment represents the degree of gene expression significance. The equation of enrichment is as following: Re=(nf/n)/(Nf/N) (14) , where nf is the number of significant genes within the particular category, n is the total number of genes within the same category, Nf is the number of significant genes in the entire microarray, N is the number of all genes tested. Agilent Feature extraction software was used to collect original data from microarray followed by analysis with robust multichip average (RMA). Gene intensity data between PE group and control group were compared with t test after calibration with a stochastic variance model. Differentially expressed genes were identified from whole genomes. Independent-Samples T Test was used to compare mRNA levels in samples from PE patients and controls. Statistical tests were performed using SPSS 17.0, and p values <0.05 were considered significant. Before t test, test for equality of variances was performed, if variances were not equal, t test result would be corrected. Sample collection: the fasting venous blood (2 ml) was collected in the morning and added to the ET tube. Flow cytometry was performed to detect the differentiation antigens on immune cells with BECKMANCOULTER EPICS XL-II flow cytometer. In 56 patients, the CD3 + , CD4 + , CD8 + , and CD19 were detected, and the NK cell marker CD16 + +56 + was detected in 50 patients. The CRP level in VTE patients was detected by scintillation turbidimetry. GO analysis targets the compromised immune functions of T cells and the decreased expression of immune receptor complex in PE patients. Among 12 genes related to the activation and chemotaxis of neutrophils, the mRNA expressions of 9 genes were significantly up-regulated in PE group (Figure 1 ). In the PE patients, the mRNA expression of CD14, a mononuclear cell surface antigen, was markedly up-regulated and that of CD74, a macrophage activating factor, was also significantly up-regulated. In addition, the mRNA expressions related to Fc fragment of surface receptors (FCGR2A, FCGR2B, FCGR2C, ITGAL and SCARB1) were largely increased significantly ( Figure 2 ). The mRNA expressions of C1 and C3 remain unchanged. In PE group, the mRNA expressions of C4b, C5, C5b as well as their receptors and complement integrins were markedly up-regulated but those of membrane attack complex component C6, C7, and C9 were significantly down-regulated when compared with the control group ( Figure 3 ). In the PE group, the mRNA expressions of IFN regulatory factors, TNF and IL-10 were markedly up-regulated. IL-2 and IL-23A mRNA expressions were significantly down-regulated when compared with control group (Figure 4 ). In the PE group, the expression of killer lectin-like receptor (KLR) was markedly down-regulated when compared with control group, and the NCR1 mRNA expression was also markedly down-regulated ( Figure 5 ). Only CD86 mRNA expression was significantly up-regulated in PE group ( Figure 6 ). In the PE group, the mRNA expressions of T cell mediated cellular immunity, protein kinases related to transmembrane signal transduction (protein tyro-sine kinase-ZAP70), T cell surface antigens (especially CD3), membrane surface immune receptor complex and T cell granzymes were markedly down-regulated when compared with the control group (Figure 7) . A total of 6 parameters(CD3 + , CD4 + , CD8 + , CD4 + /CD8 + , CD19 and CD16 + +56 + ) were measured in the 56 PE patients, and 53 had abnormalities in one or more parameters: 27 had abnormal CD3 + expression (decreased in 25 and increased in 2) ( Figure 8A ); 18 had aberrant CD4 + expression (decreased in 10 and increased in 8) ( Figure 8B) ; 26 had abnormal CD8 + expression(decreased in 25 and increased in 1) ( Figure 8C ); 30 had aberrant CD4 + /CD8 + ratio (decreased in 7 and increased in 23) ( Figure 8D ); 23 had abnormal NK CD16 + +56 + expression in 50 patients (decreased in 14 and increased in 9) ( Figure 8E ) and 17 had aberrant CD19 expression in 56 patients (decreased in 15 and increased in 2)( Figure 8F ); The CD8 + expression was decreased in 25 out of 56 patients(44.6%), the CD8 + and CD16 + +56 + expressions were decreased in 34 out of 56 patients(60.7%), and the CD8 + , CD16 + +56 + and CD19 expressions were decreased in 39 out of 56 patients(69.8%)( Figure 8G ). In 56 VTE patients, 44 patients received CRP determination. The CRP level in 35 out of 44 patients (79.5%) was higher than normal range (Figure 9 ). The Go analysis of the genomic study targeted the decreased immune function of T cells and immune receptor complex in PE patients, suggesting the occurrence of PE is closely related to the immune dysfunction. Statistical analysis revealed the mRNA expressions of genes associated with innate immunity and cytokines were markedly up-regulated and those related to the cellular immunity of T cells and NK cells significantly down-regulated. In addition, cytological experiment indicated 6 parameters related to immune function were abnormal in 53 of 56 VTE patients. The expressions of CD3 + and CD8 + were markedly reduced and the CD4 + /CD8 + ratio significantly increased. The number of CD16 + +56 + and CD19 cells was reduced. The results from cytological examination and genome analysis were consistent. Among the 56 patients with VTE, 25 (44.6%) had decreased CD3 + T cells, 25 had reduced CD8 + T cells and 23 (41%) had increased CD4 + /CD8 + ratio. These findings indicated the ability of T cells to recognize antigen and transmit activation signals was significantly compromised, and the capability of T cells to kill the pathogen infected cells decreased. The compromised T cell immune function is often identified in patients with malignancy, use of immunosuppressant, viral infection or malnutrition (15, 16) . In the present study, none of patients had malignancies or were treated with immunosuppressants. Therefore, the pathogenesis of VTE might be closely related to viral infection or malnutrition. In our report, the syn-thesis and release of virus-like micro-organisms were noted in the lymphocytes under an electron microscope in a young pulmonary hypertension patient with increased D-Dimer (10) . Among 56 patients with VTE, 14 had decrease of CD16 + +56 + NK cells, which suggests the ability of NK cells to kill intracellular pathogens including virus is impaired. CD19 is only expressed on the B lymphocytes of normal hemopoietic system and the follicular dendritic cells (FDC) of germinal center. The expression of CD19 is detectable back to progenitor B cells and present during the maturation of B lymphocytes. Once the B cells differentiate into plasma cells, the CD19 expression is absent. CD19 involves in the flux of Ca 2+ in the B lymphocytes, and can regulate the activation and proliferation of B cells (17) . Among 56 VTE patients, 15 had decreased CD19 expression, which suggests the activity and proliferation of B cells are compromised. In 53 of 56 VTE patients, the expressions of CD3 + , CD8 + , CD16 + +56 + and CD19 were separately or combinedly down-regulated or the CD4 + /CD8 + ratio was abnormal. These results imply the occurrence of VTE is closely related to the immune function. In the present study, CD16 + +56 + T cells and/or CD8 + T cells were decreased in 34 out of 56 VTE patients (60.7%). Down-regulation of CD16 + +56 + , CD8 + and/or CD19 was found in 39 out of 56 patients (69.64%). These findings indicate the symptomatic VTE is associated with the decrease of innate immunity and adaptive immunity in more than 2/3 of patients. Moreover, 53 VTE patients (94.6%) had one or more immune dysfunctions. Among 56 VTE patients, only 3 had normal immune function. Two patients were a 31-year-old male acute PE patient and 62-year-old female acute PE patient who did not receive genetic testing. The remaining one patient was a 66-year-old female patient who was diagnosed as CTEPH and underwent splenectomy 20 years ago. For patients with abnormal immune function on admission, it was difficult to confirm when the immune function became abnormal. However, we found that more than 10 patients with down-regulation of CD8 + and CD16 + +56 + developed symptomatic VTE sequentially. Among 56 VTE patients, more than 50% of patients had symptoms of respiratory infection or a history of respiratory infection recently. Among 56 VTE patients, CRP was measured in 44 patients and 35 (79.5%) had increase of CRP, which implies inflammation is related to the occurrence of VTE. In response to the invasion of foreign pathogens to human body, instantaneous innate immune response occurs within 0-4 hours after infection, and early innate immune response occurs 4-96 hours after infection (18, 19) . DVT and PE often occurred after 2-10 days postoperatively, which coincided with the infection process of innate and adaptive immune function (20) . During the 3-year follow up period, 21 VTE patients (40%; including those died) were lost to follow up. Among patients receiving follow up, all were treated with warfarin for anti-coagulation. In addition, immunological examination and detection of D-Dimer were also carried at designed time points. Our results showed about 30% of VTE patients receiving follow up did not have increase of D-Dimer level any more at 0.1~1 year after warfarin discontinuation when the immunological examination and detection of D-D dimer showed normal. Of 56 patients with symptomatic VTE, the relationship between VTE and immune dysfunction was found in 53 (94.6%). Nevertheless, patients with immune dysfunction do not develop symptomatic VTE in a short time. The compromised or disorganized immune function may be the internal cause of susceptibility to acquired VTE, and infection acts as a triggering factor of acquired VTE. When the pathogens invade the subjects with immune dysfunction, the pathogens can not be completely removed by the immune system. Thus, patients with compromised or disorganized immune function are susceptible to acquired VTE.

Cedar Virus: A Novel Henipavirus Isolated from Australian Bats

The genus Henipavirus in the family Paramyxoviridae contains two viruses, Hendra virus (HeV) and Nipah virus (NiV) for which pteropid bats act as the main natural reservoir. Each virus also causes serious and commonly lethal infection of people as well as various species of domestic animals, however little is known about the associated mechanisms of pathogenesis. Here, we report the isolation and characterization of a new paramyxovirus from pteropid bats, Cedar virus (CedPV), which shares significant features with the known henipaviruses. The genome size (18,162 nt) and organization of CedPV is very similar to that of HeV and NiV; its nucleocapsid protein displays antigenic cross-reactivity with henipaviruses; and it uses the same receptor molecule (ephrin- B2) for entry during infection. Preliminary challenge studies with CedPV in ferrets and guinea pigs, both susceptible to infection and disease with known henipaviruses, confirmed virus replication and production of neutralizing antibodies although clinical disease was not observed. In this context, it is interesting to note that the major genetic difference between CedPV and HeV or NiV lies within the coding strategy of the P gene, which is known to play an important role in evading the host innate immune system. Unlike HeV, NiV, and almost all known paramyxoviruses, the CedPV P gene lacks both RNA editing and also the coding capacity for the highly conserved V protein. Preliminary study indicated that CedPV infection of human cells induces a more robust IFN-β response than HeV.

Henipaviruses were first discovered in the 1990s following investigation of serious disease outbreaks in horses, pigs and humans in Australia and Malaysia [1, 2] and comprise the only known Biosafety Level 4 (BSL4) agents in the family Paramyxoviridae [3] . Depending upon the geographic locations of outbreaks, and the virus and animal species involved, case mortality is between 40% to 100% in both humans and animals [4, 5] , making them one of the most deadly group of viruses known to infect humans. The genus Henipavirus in the subfamily Paramyxovirinae currently contains two members, Hendra virus (HeV) and Nipah virus (NiV) [6] . Fruit bats in the genus Pteropus, commonly known as flying foxes, have been identified as the main natural reservoir of both viruses although serological evidence suggests that henipaviruses also circulate in non-pteropid bats [7, 8, 9, 10] . The discovery of henipaviruses had a significant impact on our understanding of genetic diversity, virus evolution and host range of paramyxoviruses. Paramyxoviruses, such as measles virus and canine distemper virus, were traditionally considered to have a narrow host range and to be genetically stable with a close to uniform genome size shared by all members of Paramyxovirinae [3] . Henipaviruses shifted this paradigm on both counts having a much wider host range and a significantly larger genome [6] . Identification of bats as the natural reservoir of henipaviruses also played an important role in significantly increasing international scientific attention on bats as an important reservoir of zoonotic viruses, including Ebola, Marburg, SARS and Melaka viruses [11, 12, 13, 14] . Since the discovery of the first henipavirus in 1994, much progress has been made in henipavirus research, from identification of functional cellular receptors to the development of novel diagnostics, vaccine and therapeutics [15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25] . By contrast, there is little understanding of the pathogenesis of these highly lethal viruses. This is due in part to the requirement of a high security BSL4 facility for any live infection studies and in part to the limited range of research tools and reagents for the current small animal models. Research into the mechanisms of henipavirus pathogenesis is also hampered by the lack of related, but non-pathogenic or less pathogenic viruses, thus preventing targeted comparative pathogenetic studies. Early serological investigations in Australia and more recent studies in other regions (e.g., China) indicated the presence of cross-reactive, but not cross-neutralizing, antibodies to henipaviruses in bats of different species [8] . These findings were further supported by the detection of henipavirus-like genomic sequences in African bats [26] . Discovery and isolation of these related viruses will be highly important to our further understanding of henipavirus evolution, mechanism of cross-species transmission, and pathogenesis in different animal species. Here we report the isolation and characterization of a new bat henipavirus which, based on preliminary infection studies, is nonpathogenic in two of the small animal infection models currently used in henipavirus research. We believe that this new virus will provide a powerful tool to facilitate our future study into different aspects of henipaviruses, especially in the less advanced area of pathogenesis. As part of our on-going field studies on HeV genetic diversity and infection dynamics in the Queensland flying fox populations, urine samples were collected on a regular basis for PCR and virus isolation. Since the establishment of the Pteropus alecto primary cell lines in our group [27] , we have intensified our effort to isolate live virus from these urine samples by routinely inoculating separate primary cell lines derived from kidney, spleen, brain, and placenta, as well as Vero cells. Syncytial CPE was observed in kidney cell (PaKi) monolayers 5 days post inoculation (dpi) with two different urine samples (Fig. S1 ) collected in September 2009 from a flying fox colony in Cedar Grove, South East Queensland (see Fig. S2 for map location). No CPE was observed in any of the four other cell lines. Supernatant harvested 6 dpi was used to inoculate fresh PaKi cell monolayers. After two passages in PaKi cells, the virus was able to infect and cause CPE in Vero cells. However, the CPE morphology of CedPV infection in Vero cells was different from that of HeV infection. Further analysis using HeV-specific PCR primers indicated that the new bat virus was not an isolate of HeV. Considering the formation of syncytial CPE by this new virus and the previous success in isolating paramyxoviruses from bat urine [28, 29, 30] , paramyxovirus family-specific and genus-specific primers were used to determine whether this new virus was a member of the family Paramyxoviridae. Positive PCR fragments of the expected sizes were obtained from the Paramyxovirinae and Respirovirus/Morbillivirus/Henipavirus primer sets developed by Tong et al [31] . Sequencing of the PCR products indicated that it was a new paramyxovirus most closely related to HeV and NiV. Based on these preliminary data, the virus was named Cedar virus (CedPV) after the location of the bat colony sampled. Full length genome sequence was determined by a combination of three different approaches, random deep sequencing using 454 technology, sequencing of PCR products obtained using degenerate primers designed based on known henipaviruses, and RACE to determine the precise genome terminal sequences. As shown in Fig. 1 , the genome of CedPV is 18,162 nt in length most similar to that of HeV in the family. The full genome sequence has been deposited to GenBank (Accession No. JQ001776). The genome size is a multiple of 6, hence abiding by the Rule-of-Six observed for all known members of the subfamily Paramyxovirinae [3] . It has a 3-nt intergenic sequence of CTT absolutely conserved at all seven positions and highly conserved gene start and stop signals similar to those present in HeV and NiV (Fig. S3) . Also similar to the HeV genome is the presence of relatively large non-coding regions in the CedPV genome ( Fig. 1 and Table 1 ). The overall proteincoding capacity of the CedPV genome is 87.41% which is significantly lower than the average of 92.00% for other family members but higher than HeV at 82.12%. As the genome size of CedPV and HeV is very similar, the increased coding capacity of CedPV is attributed to an increase in protein sizes for five of the six major proteins, with the L protein being 257-aa larger (Table 1) . At 2,501 aa, the CedPV L protein is the largest, not only in the family Paramyxoviridae but also for all known viruses in the order Mononegavirale. Phylogenetic analysis based on the full length genome sequence and the deduced amino acid sequences of each structural protein confirmed the initial observation that CedPV is most closely related to henipaviruses in the family. A phylogenetic tree based on the deduced sequences of the nucleocapsid protein (N) is presented in Fig. 2 . Phylogenetic tree based on whole genome sequences gave similar results (Fig. S4 ). CedPV is more closely related to HeV and NiV than henipavirus-like sequences detected in African bats [26, 32] as shown in a phylogenetic tree based on the only sequences available of a 550-nt L gene fragment (Fig. S5) . A phosphoprotein (P) gene lacking RNA editing and coding capacity for the V protein First discovered for the parainfluenza virus 5 (PIV5, previously known as simian virus 5), almost all members of Paramyxovirinae have a P gene which produces multiple proteins through an RNA editing mechanism by addition of non-templated G residues leading to production of N-terminal co-linear proteins from different reading frames downstream from the editing site [3, 33] . These multiple gene products are known to play a key role in antagonizing the innate response of susceptible hosts [3] . A search of CedPV for open reading frames (ORF) in the P gene revealed a 737-aa P protein and a 177-aa C protein, but failed to find the highly conserved, cysteine-rich V ORF present in most other paramyxoviruses. The RNA editing site with the sequence of AAAAGGG, which is absolutely conserved in all known HeV and NiV isolates discovered to date, is also missing from the CedPV P gene sequence. To further verify that there are no multiple mRNAs produced from the CedPV P gene, direct sequencing of P gene transcripts was conducted from CedPV-infected Vero cells using multiple sets of primers generating overlapping fragments covering the entire coding region of the P gene. Each produced Hendra and Nipah viruses are 2 highly pathogenic paramyxoviruses that have emerged from bats within the last two decades. Both are capable of causing fatal disease in both humans and many mammal species. Serological and molecular evidence for henipa-like viruses have been reported from numerous locations including Asia and Africa, however, until now no successful isolation of these viruses have been reported. This paper reports the isolation of a novel paramyxovirus, named Cedar virus, from fruit bats in Australia. Full genome sequencing of this virus suggests a close relationship with the henipaviruses. Antibodies to Cedar virus were shown to cross react with, but not cross neutralize Hendra or Nipah virus. Despite this close relationship, when Cedar virus was tested in experimental challenge models in ferrets and guinea pigs, we identified virus replication and generation of neutralizing antibodies, but no clinical disease was observed. As such, this virus provides a useful reference for future reverse genetics experiments to determine the molecular basis of the pathogenicity of the henipaviruses. uniform trace files indicating a lack of RNA editing activities, which is very different from the mixed peaks generated by HeV and NiV immediately after the editing site (Fig. S7 ). To our knowledge, CedPV is the first member of Paramyxovirinae that lacks both RNA editing and any V-related coding sequence in its P gene. Further investigation is required to exclude the possibility that the P-gene editing in CedPV is cell-or tissue-specific and not present or present at an extremely low level in the current viruscell system. The striking similarity in genome size and organization and the presence of highly conserved protein domains among the N, M and L proteins between CedPV and henipaviruses prompted us to investigate the antigenic relatedness of these viruses. Staining of CedPV-infected Vero cells using rabbit anti-henipavirus antibodies indicated the presence of cross-reactivity. This cross-reactivity was further confirmed in reverse by staining of HeV-infected Vero cells using a rabbit serum raised against a recombinant CedPV N protein (Fig. 3) . However, analysis by virus neutralization test using either polyclonal or monoclonal antibodies found that henipavirus-neutralizing antibodies were unable to neutralize CedPV. Similarly, CedPV-neutralizing antibodies obtained in our infection studies (see below) also failed to neutralize either HeV or NiV. It can therefore be concluded that CedPV and henipaviruses share cross-reactive antigenic regions, but not crossneutralizing epitopes. To further investigate the relationship between CedPV and recognized henipaviruses, we investigated the use of the henipavirus receptors, the ephrin-B2 and -B3 host cell proteins, as potential receptors for CedPV infection. Our previous studies have demonstrated that the ephrin-B2 and -B3 expression negative HeLa-USU cell line could support henipavirus infection and formation of syncytial CPE only when either the ephrin-B2 or -B3 gene was transiently expressed in the cells [22, 34] . For CedPV, similar observations were made with respect to the ephrin-B2 receptor. As shown in Fig. 4 , CedPV failed to infect HeLa-USU, but was able to infect and cause syncytial CPE when the human ephrin-B2 gene was expressed. In contrast, when ephrin-B3 molecule was introduced, there was no evidence of infection. Ferrets, guinea pigs, and mice exhibit differing responses to the previously described henipaviruses HeV and NiV, with ferrets and guinea pigs, but not mice developing severe disease characterized by systemic vasculitis [20, 35, 36, 37, 38] . In contrast, ferrets and guinea pigs exposed to CedPV by, respectively, oronasal and intraperitoneal routes remained clinically well although neutralizing antibody was detected in serum between 10 to 21 days pi ( Table 2) . Balb-C mice exposed to CedPV by the oronasal route remained clinically well and did not develop neutralizing antibody in serum by day 21 pi. In ferrets electively euthanized at earlier time-points, there was reactive hyperplasia of tonsillar lymphoid tissue, retropharyngeal and bronchial lymph nodes, accompanied by edema and erythrophagocytosis. CedPV antigen was detected in bronchial lymph node of one animal euthanized on day 6 pi, consistent with viral replication in that tissue; cross-reactive immunostaining against anti-NiV N protein antibodies was also noted (Fig. 5) . No other significant histological lesions were identified. Viral RNA was detected in selected lymphoid tissues of 3 (of 4) ferrets sampled day 6 to 8 pi, including pharynx, spleen, and retropharyngeal and bronchial lymph nodes, as well as the submandibular lymph node of the ferret euthanized on day 20 pi. This pattern of lymphoid involvement suggests that there may be transient replication in the upper and lower respiratory tracts although CedPV genome was not recovered from nasal washes, oral swabs, pharynx or lung tissue of affected animals. Virus isolation was unsuccessful for all PCR positive tissues. As a first step towards the understanding of the pathogenicity difference between CedPV and HeV, we examined the IFN responses in human HeLa cells upon virus infection. As shown in Fig. 6 , while the induction of IFN-a was similar in cells infected with HeV or CedPV, there was a significant difference of IFN-b production upon infection by HeV or CedPV, with CedPVinfected cell producing a much higher level of IFN-b. To investigate the CedPV exposure status of pteropid bats in Queensland and potential co-infection (either concurrent or consecutive) of CedPV with HeV, we tested 100 flying fox sera collected previously for other studies for antibody against the two viruses. Due to the cross-reactivity observed above, virus neutralization tests were conducted to obtain more accurate infection data for each virus. Overall, 23% of the sera were CedPV-positive and 37% HeV-positive (Table S1 ). Co-infection was reflected in 8% of the sera tested. The emergence of bat-borne zoonotic viruses (including HeV, NiV, Ebola, Marburg, and SARS) has had a significant impact on public health and the global economy during the past few decades. With the rapidly expanding knowledge of virus diversity in bat populations around the world, it is predicted that more bat-borne zoonotic viruses are likely to emerge in the future. The discovery of a novel ebolavirus-like filovirus in Spanish microbats demonstrates that the potential for such spill over events is not limited to Africa or Asia [39] . It is therefore important to enhance our preparedness to counter future outbreaks by conducting active pre-emergence research into surveillance, triggers for cross-species transmission, and the science of identification of potential pathogens. Henipaviruses represent one of the most important bat-borne pathogens to be discovered in recent history. Although CedPV displays some differences from existing members of the genus Henipavirus, we propose that CedPV be classified as a new henipavirus based on the following shared features with known henipaviruses: 1) it is antigenically related to current henipaviruses; 2) its genome size and organization is almost identical to those of HeV and NiV; 3) it has a similar prevalence in flying foxes; and 4) it uses ephrin-B2 as the cell entry receptor. The lack of cross-neutralization between CedPV and HeV or NiV was not unexpected from the comparative sequence analysis of all the deduced proteins, especially the G protein (see Table 1 ). It is clear that the genetic relatedness of CedPV with HeV or NiV is much lower than between HeV and NiV. However, the percentage sequence identities of the major viral proteins between CedPV and HeV/NiV are on average at least 10% higher than that between HeV/NiV and any other known paramyxoviruses. Also, there was no antigenic cross-reactivity observed between CedPV and representative viruses of the other paramyxovirus genera in the subfamily Paramyxovirinae (Fig. S6) . Like other paramyxoviruses, the P gene of henipaviruses produces multiple proteins which play a key role in viral evasion of host innate immune responses [4, 40, 41] . One of these is the Cys-rich V protein: all members of the subfamily Paramyxovirinae produce the V protein with the exception of the human parainfluenza virus 1 (hPIV1). Although a putative RNA editing sequence (AAGAGGG) is present at the expected editing site of the P gene, the hPIV1 RNA polymerase does not produce an edited mRNA of the P gene [42] . There are remnants of the V ORF easily detectable in the hPIV1 P gene although the predicted 68-aa ORF region is interrupted by multiple in-frame stop codons. Of the 7 Cys residues conserved between bovine parainfluenza virus 3 and Sendai virus, four are still present in the non-functional V ORF of hPIV1 [42] . In contrast, an extensive ORF and sequence homology search of the CedPV P gene only identified one aa coding region with minimal sequence identity to the V ORFs of HeV and NiV (see Fig. S8 ). In this region, out of the 9 Cys residues conserved between HeV and NiV V proteins, only 2 are present in the CedPV P gene. Furthermore, the sequence (AGATGAG) upstream from this putative ORF V coding region does not match the consensus RNA editing site. It can therefore be concluded that CedPV is the only member of Paramyxovirinae which lacks both the functional V mRNA/protein and the coding capacity for the RNA editing site and ORF V. The evolutionary significance of this finding needs further investigation. Our in vitro study indicated that ephrin B2, but not ephrin B3, was able to restore CedPV infection in the ephrin B2-deficient HeLa cells. While this is highly suggestive that ephrin B2 is the functional entry receptor for CedPV, it should be emphasized that this was not a direct proof that ephrin B2 is the receptor. Further investigation is required to confirm this. In our preliminary studies, it was shown that CedPV was able to replicate in guinea pigs and ferrets, but failed to cause significant clinical diseases, unlike that of the closely related HeV and NiV. These first infection experiments were conducted with a high dose if virus to establish whether the CedPV could replicate in these animals and determine the degree of any clinical disease. A second experiment was then carried out in ferrets to determine the site of replication and tissue tropism in sequentially sacrificed animals. A lower dose was used to gain better comparison with similar infection experiments using HeV and NiV [18, 35] . Although these initial experimental infection studies indicate that CedPV is less or non-pathogenic in these species, it is possible that CedPV may be pathogenic in other hosts, such as horses. We hypothesize that the lack of a V protein may impact on the pathogenicity. In this regard, it was encouraging to observe that infection of human cells by CedPV induced a much more robust IFN-b response than HeV. Further study is required to dissect the exact molecular mechanism of this observed difference. Due to the close relationship between CedPV and HeV, it was important to investigate the possibility of co-infection by these two viruses in the Australian bat population. Based on the detection of neutralizing antibodies at 23% for CedPV, 37% for HeV and 8% for both, it can be concluded that the co-infection rate is very close to the theoretical rate of 8.5% (the product of the two independent infection rates). Based on this limited preliminary analysis, it appears that infection of bats by one henipavirus neither prevents nor enhances the likelihood of infection by the other. In summary, the discovery of another henipavirus in Australian flying foxes highlights the importance of bats as a significant reservoir of potential zoonotic agents and the need to expand our understanding of virus-bat relationships in general. Our future research will be directed at determining whether spill-over of CedPV into other hosts has occurred in the past in Australia, whether CedPV is pathogenic in certain mammalian hosts, and whether CedPV exists in bat populations in geographically diverse regions. All animal studies were approved by the CSIRO Australian Animal Health Laboratory's Animal Ethics Committee and conducted following the Australian National Health and Medical Research Council Code of Practice for the Care and Use of Animals for Scientific Purposes guidelines for housing and care of laboratory animals. Cell lines used this study were Vero (ATCC), HeLa-USU [22] , and the P. alecto primary cell lines derived from kidney (PaKi), brain (PaBr), (spleen) PaSp and placenta (PaPl) recently established in our group [27] . Cells were grown in Dulbecco's Modified Eagle's Medium Nutrient Mixture F-12 Ham supplemented with double strength antibiotic-antimycotic (Invitrogen), 10 mg/ml ciprofloxacin (MP Biomedicals) and 10% fetal calf serum at 37uC in the presence of 5% CO 2 . Urine (approximately 0.5-1 ml) was collected off plastic sheets placed underneath a colony of flying foxes (predominantly Pteropus alecto with some P. Poliocephalus in the mixed population) in Cedar Grove, South East Queensland, Australia and pooled into 2-ml tubes containing 0.5 ml of viral transport medium (SPGA: a mix of sucrose, phosphate, glutamate and albumin plus penicillin, streptomycin and fungizone). The tubes were temporarily stored on ice after collection and transported to a laboratory in Queensland, frozen at 280uC, and then shipped on dry ice to the CSIRO Australian Animal Health Laboratory (AAHL) in Geelong, Victoria for virus isolation. The samples were thawed at 4uC and centrifuged at 16,0006g for 1 min to pellet debris. Urine in the supernatant (approximately 0.5-1 ml) was diluted 1:10 in cell culture media. The diluted urine was then centrifuged at 1,2006g for 5 min and split evenly over Vero, PaKi, PaBr, PaSp and PaPl cell monolayers in 75-cm 2 tissue culture flasks. The flasks were rocked for 2 h at 37uC, 14 ml of fresh cell culture media was added and then incubated for 7 d at 37uC. The flasks were observed daily for toxicity, contamination, or viral cytopathic effect (CPE). Cells showing syncytial CPE were screened using published broadly reactive primers [31] for all known paramyxoviruses and a subset of paramyxoviruses. PCR products were gel extracted and cloned into pGEM T-Easy (Promega) to facilitate sequencing using M13 primers. Sequences were obtained and aligned with known paramyxovirus sequences allowing for initial classification. Whole genome sequence was determined using a combination of 454 sequencing [43] and conventional Sanger sequencing. Virions from tissue culture supernatant were collected by centrifugation at 30,0006g for 60 min and resuspended in 140 ml of PBS and mixed with 560 ml of freshly made AVL for RNA extraction using QIAamp Viral RNA mini kit (Qiagen). Synthesis of cDNA and random amplification was conducted using a modification of a published procedure [44] . Briefly, cDNA synthesis was performed using a random octomer-linked to a 17-mer defined primer sequence: (59-GTTTCCCAGTAGGTCTCNNN NNNNN-39) and SuperScript III Reverse Transcriptase (Invitrogen). 8 ml of ds-cDNA was amplified in 200 ml PCR reactions with hot-start Taq polymerase enzyme (Promega) and 59-A*G*C*A*C TGTAGGTTTCCCAG-TAGGTCTC-39 (where * denotes thiol modifications) as amplification primers for 40 cycles of 95uC/1 min, 48uC/1 min, 72uC/ 1 min after an initial denaturation step of 5 min at 95uC and followed by purification with the QIAquick PCR purification kit (Qiagen). Sample preparation for Roche 454 sequencing (454 Life Sciences Branford, CT, USA) was according to their Titanium series manuals, Rapid Library Preparation and emPCR Lib-L SV. To obtain an accurate CedPV genome sequence, 454 generated data (after removing low quality, ambiguous and adapter sequences) was analysed by both de novo assembly and read mapping of raw reads onto the CedPV draft genome sequence derived from Sanger sequencing. For 454 read mapping, SNPs and DIPs generated with the CLC software were manually assessed for accuracy by visualising the mapped raw reads (random PCR errors are obvious compared to real SNPs and DIPs especially when read coverage is deep). Consensus sequences for both 454 de novo and read mapping assembly methods were then compared to the Sanger sequence with the latter used to resolve conflicts within the low coverage regions as well as to resolve 454 homopolymer errors. Sequences of genome termini were determined by 39-and 59-RACE using a protocol previously published by our group [45] . Briefly, approximately 100 ng of RNA was ligated with adaptor DT88 (see reference for sequence information) using T4 RNA ligase (Promega) followed by cDNA synthesis using the Super-Script III RT kit (Invitrogen) and an adaptor-specific primer, DT89. PCR amplification was then carried out using DT89 and one or more genome-specific primers. PCR products were sequenced directly using either DT89 or genome specific primers by an in-house service group on the ABI Sequencer 3100. The CLC Genomics Workbench v4.5.1 (CLC Inc, Aarhus, Denmark) was used to trim 454 adapter and cDNA/PCR primer sequences, to remove low quality, ambiguous and small reads ,15 bp and to perform de novo and read mapping assemblies all with default parameters. Clone Manager Professional ver 9.11 (Scientific and Educational Software, Cary, NC, USA) was used to join overlapping contigs generated by de novo assembly. Phylogenetic trees were constructed by using the neighbor-joining algorithm with bootstrap values determined by 1,000 replicates in the MEGA4 software package [46] . Quantitative PCR assays (qPCR) were established based on CedPV-specific sequences obtained from the high throughput sequencing. A TaqMan assay on the P gene was developed and used for all subsequent studies. The sequences of the primer/probe are as follows: forward primer, 59-TGCAT TGAGC GAACC CATAT AC; reverse primer, 59-GCACG CTTCT TGACA GAGTT GT; probe, 59-TCCCG AGAAA CCCTC TGTGT TTGA-MGB. The coding region for the CedPV N protein was amplified by PCR with a pair of primers flanked by AscI (59 end) and NotI (39 end) sites for cloning into our previously described GST-fusion expression vector [47] . The expression and purification by gel elution was conducted as previously described [48] . For antibody production, purified protein was injected subcutaneously into 4 different sites of 2 adult (at a dose of 100 mg per animal) New Zealand white female rabbits at days 0 and 27. The CSIRO's triple adjuvant [49] was used for the immunization. Animals were checked for specific antibodies after days 5 and 42 and euthanized at day 69 for the final blood collection. For immunofluorescence antibody test, Vero cell monolayers were prepared in 8-well chamber slides by seeding at a concentration of 30,000 cells/well in 300 ml of cell media and incubating over night at 37uC. The cell monolayers were infected with an MOI of 0.01 of CedPV, HeV or NiV and fixed with 100% ice-cold methanol at 24 h post-infection. The chamber slides were blocked with 100 ml/well of 1%BSA in PBS for 30 min at 37uC before adding 50 ml/well of rabbit sera against CedPV N or NiV N diluted 1:1000. After incubation at 37uC for 30 min, the slides were washed three times in PBS-T and incubated with 50 ml/well of anti-rabbit 488 Alexafluore conjugate (Invitrogen) diluted 1:1000 at 37uC for 30 min. The slides were then washed three times in PBS-T and mounted in 50% glycerol/PBS for observation under a fluorescence microscope. For virus neutralization test, serial two-fold dilutions of sera were prepared in duplicate in a 96-well tissue culture plate in 50 ml cell media (Minimal Essential Medium containing Earle's salts and supplemented with 2 mM glutamine, antibiotic-antimycotic and 10% fetal calf serum). An equal volume containing 200 TCID 50 of target virus was added and the virus-sera mix incubated for 30 min at 37uC in a humidified 5% CO 2 incubator. 100 ml of Vero cell suspension containing 2610 5 cells/ml was added and the plate incubated at 37uC in a humidified 5% CO 2 incubator. After 4 days, the plate was examined for viral CPE. The highest serum dilution generating complete inhibition of CPE is defined as the final neutralizing titer. Human ephrin B2 and B3 genes were cloned into pQCXIH (Clontech) and the resulting plasmids packaged into retrovirus particles in the GP2-293 packaging cell line (Clontech) and pseudotyped with vesicular stomatitis virus G glycoprotein (VSV-G) following the manufacturer's instructions. HeLa-USU cell line [22] was infected with the VSV-G pseudotyped retrovirus particles in the presence of 1 mg/ml polybrene (Sigma). 8 h post infection, the medium was changed and the cells were allowed to recover for 24 h, allowing time for the retroviral insert to be incorporated into the cell genome and for expression of the hygromycin resistance gene. 24 h post infection, cells transformed by the retrovirus were selected for by the addition of 200 mg/ml hygromycin in the media. Stocks of cells that were resistant to hygromycin were prepared and frozen. HeLa-USU and ephrin-expressing HeLa-USU cells were seeded in 6-well tissue culture plates at a density of 250,000 cells/well overnight. The viruses (HeV and CedPV) were diluted to give an MOI of 0.01 and inoculated into the wells. The cell monolayers were examined daily for syncytial CPE. Animal studies were carried out in the BSL4 animal facility at AAHL. Ferrets, guinea pigs and mice were used on the basis of their known and varying responses to exposure to other henipaviruses. Firstly, 2610 6 TCID 50 /ml CedPV passaged twice in bat PaKi cells was administered to 2 male ferrets (1 ml oronasally); 4 female guinea pigs (1 ml intraperitoneally); and 5 female Balb-C mice (50 ml oronasally). Guinea pigs and mice were implanted with temperature sensing microchips (LifeChip Bio-thermo, Destron Fearing) and weighed daily. Ferret rectal temperature and weight was recorded at sampling times. Animals were observed daily for clinical signs of illness and were euthanized at 21 d postinoculation. Sera were collected on days 10, 15 and 21 to test for neutralizing antibody against CedPV. Secondly, on the basis of asymptomatic seroconversion to CedPV noted in ferrets in the first study, 7 further female ferrets were exposed by the oronasal route to a lower dose of 3610 3 TCID 50 . Two animals were euthanized on each of days 6, 8 and 10 post-inoculation and one on day 20. Nasal washes, oral swabs, and rectal swabs were collected on days 2, 4, 6, 8 and 10 and urine was sampled on the day of euthanazia; each specimen was assessed for CedPV genome. A wide range of tissue samples were collected at post mortem examination and assessed by routine histology, immunohistochemistry (using rabbit antibodies raised against recombinant CedPV and NiV N proteins, respectively), qPCR (see above) and virus isolation using reagents and procedures previously established in our group [16] . HeLa cells were infected with Hendra and Cedar viruses at an MOI 0.5 for 24 hours, at which time total cellular RNA was extracted and IFN-a and IFN-b mRNA levels were quantified by real-time PCR using Power SYBR Green RNA-to-CT 1-Step Kit (Applied Biosystems). Primers were as previously described [50] . Sera from 100 flying foxes collected during 2003-2005 from Queensland, Australia were screened for neutralizing antibodies to CedPV. Virus neutralization test was conducted as described above (antibody tests). All serum samples were tested at a dilution of 1:20.

Host Modulators of H1N1 Cytopathogenicity

Influenza A virus infects 5–20% of the population annually, resulting in ∼35,000 deaths and significant morbidity. Current treatments include vaccines and drugs that target viral proteins. However, both of these approaches have limitations, as vaccines require yearly development and the rapid evolution of viral proteins gives rise to drug resistance. In consequence additional intervention strategies, that target host factors required for the viral life cycle, are under investigation. Here we employed arrayed whole-genome siRNA screening strategies to identify cell-autonomous molecular components that are subverted to support H1N1 influenza A virus infection of human bronchial epithelial cells. Integration across relevant public data sets exposed druggable gene products required for epithelial cell infection or required for viral proteins to deflect host cell suicide checkpoint activation. Pharmacological inhibition of representative targets, RGGT and CHEK1, resulted in significant protection against infection of human epithelial cells by the A/WS/33 virus. In addition, chemical inhibition of RGGT partially protected against H5N1 and the 2009 H1N1 pandemic strain. The observations reported here thus contribute to an expanding body of studies directed at decoding vulnerabilities in the command and control networks specified by influenza virulence factors.

The Orthomyxoviridae family member influenza A virus is the causal agent of acute respiratory tract infections suffered annually by 5-20% of the human population. There is a significant impact on morbidity, concentrated in people younger than 20 years, with economic consequences running into the billions of dollars during large epidemics [1] . In addition, viral infections are associated with development of chronic asthma and disease exacerbation in both children and adults. In particular, acute influenza infection can amplify airway inflammation in asthmatic patients and induce alterations in epithelial and stromal cell physiology contributing to allergen sensitization, exaggerated bronchoconstriction, and remodeling of airway epithelia [2] . Mortality rates associated with seasonal flu are low, but the aging population is at risk for development of severe congestive pneumonia which kills ,35,000 people each year in the U.S. [1] . Of continual concern is the threat of emergent high virulence strains such as the Spanish flu (H1N1), Asian flu (H2N2) and Hong Kong flu (H3N2) pandemics which claimed millions of lives world-wide. Current treatments are focused on vaccines and drugs that target viral proteins. However, both of these approaches have limitations as vaccines require yearly development and lag detection of new strains, while viral proteins have a stunning capacity to evolve resistance to targeted agents [3] . The genome of the influenza A virus consists of 8 negative single-strand RNA segments that encode 11 functional peptides necessary for viral replication and virulence [1] . Thus the viral-autonomous repertoire of gene products is extremely limited and influenza A replication is dependent upon hijacking host-cell biological systems to facilitate viral entry, replication, assembly, and budding. The recognition that a suit of human host proteins are required for IVA infection and replication presents additional targeting strategies that may be less prone to deflection by the highly plastic viral genome. Here we have employed the cytopathic effects of H1N1 infection in bronchial epithelial cells as a mechanism to isolate host genes that represent intervention target opportunities by virtue of their contribution to H1N1 infection and replication, or by virtue of their contribution to viral virulence factor-dependent evasion of innate immune responses. A primary whole-genome arrayed siRNA screen identified gene depletions that either deflected or promoted bronchial epithelial cell death upon exposure to the H1N1 A/WSN/33 influenza virus and were not cytotoxic to mock infected cells. Integration with orthogonal data sets, describing host gene function [4] [5] [6] [7] [8] , parsed collective 'targets' into four functional classes. 1) Targets that, when depleted, enhance bronchial epithelial cell survival upon H1N1 exposure, and are required for viral replication. This class presumably represents host factors that facilitate viral infection and/or are required to support viral replication. 2) Targets that, when depleted, reduce bronchial epithelial cell survival upon H1N1 exposure, and are required for viral replication. This important and initially unanticipated class, likely represents proviral host factors that deflect cell death checkpoint responses that would otherwise engage upon detection of viral infection. 3) Targets that, when depleted, reduce bronchial epithelial cell survival upon H1N1 exposure and enhance viral replication relative to controls. Recently discovered innate immune pathway components, such as IFITM3 that are responsive to H1N1 infection, are members of this class, which presumably represent antiviral restriction factors that normally oppose infection. 4) Targets, that when depleted, enhance bronchial epithelial cell survival upon H1N1 exposure and enhance viral replication as compared to controls. These host factors are likely responsible for influenza virus-mediated cytopathic effects. Chemical inhibition of gene products from two classes, RABGGTASE and CHEK1, indicated these targets might be pharmacologically addressable for H1N1 intervention in an epithelial cell autonomous context. Influenza A infection is associated with pathological changes throughout the respiratory tract, however the major site of impact appears to be the respiratory epithelia. Bronchoscopy of patients with uncomplicated influenza infections reveals alterations in the ciliated epithelia of the larynx, trachea, and bronchi that includes vacuolization, loss of cilia, and desquamation of columnar epithelial cells and goblet cells down to the basal cell layer. Importantly, viral antigen is found predominantly in the epithelial cells and mononuclear cells [1] . Therefore, for the studies described here, we employed telomerase-immortalized human bronchial epithelial cells (HBEC30) that retain the capacity to differentiate into a polarized ciliated epithelial sheet [9] . In undifferentiated cell culture, we found that 100% of HBEC30 in culture display viral protein production after a 24-hour exposure to mouse-adapted virus at an MOI of 5 ( Figure 1A , B). This leads to an approximately 50% decrease in cell viability 48-hours post infection ( Figure 1C ). Given these observations, we adopted a whole genome siRNA screening strategy that involved a 48-hour incubation post siRNA transfection, followed by a 48 hour exposure to influenza A/WSN/1933 or carrier, with cell viability as the endpoint assay. Raw viability values were converted to viability Z-scores with a metric that normalized for both position and batch effects ( Figure 1D , see methods). The dynamic range of viability scores observed under screening conditions potentially affords the opportunity to identify both enhancers and resistors of viral pathogenicity ( Table 1 in Supporting Information S1). Considering siRNA pools associated with Z-scores that were equal to or greater than 3 standard deviations above (resistors) or below (sensitizers) the mean of the population, 53 candidate resistors and 182 candidate sensitizers were identified ( Table 2 in Supporting Information S1). A representative sample of targets was further tested for consequences on viral protein accumulation and viral replication. As might be expected, the majority of the siRNA pools that deflect a viral cytopathic response resulted in reduced viral protein accumulation, as detected by quantitation of viral proteins at single cell resolution, and reduced production of infectious particles ( Figure 2A ). Among these, IVNS1ABP and the splicing factor SFPQ directly interact with the viral pathogenicity factor NS1, presumably reflecting a positive role in support of viral corruption of host machinery for viral protein production [10] . Of interest in this class is RRAGD, a small G-protein that supports the amino-acid responsiveness of mTOR as a component of the ''ragulator'' [11] . Several reports have highlighted the importance of viral induction of mTOR for viral replication, but the mechanism is not fully elaborated [6, 12] . Given the participation of endosomes as a viral entry mechanism [13] , it is tempting to speculate that RRAGD is a limiting host factor for viral corruption of mTOR regulation. Additional factors in this group are involved with the host defense response, p53-mediated cell death and vesicle maturation and trafficking. To test for false positives arising from off-target effects of siRNA treatment, we retested 88 siRNA pools as four individual oligos. Approximately 60% of siRNAs retested with two or more oligos reproducing the original phenotype ( Figure S1 ). Among the most potent members of the sensitizer class were the previously described proviral host factor IFITM3 and its homolog IFITM1 ( Table 7 in Supporting Information S1). IFITM3 has been reported to be required for restriction of viral infection and is thought to inhibit viral entry [4, 14] . These gene products are interferon responsive, and depletion was associated with enhanced viral pathogenicity and enhanced viral protein production at limiting MOIs (1 and 0.1) as compared to controls (Figure 2A , B, C). Unexpectedly, cells depleted of IFITM3 produced fewer infection competent viral particles as determined by secondary infection of MDCK cells with cell culture supernatants (Figure 2 A, E). For these assays, HBEC30 cell cultures were infected with an MOI of 5 for 48 hours post transfection with siRNA pools. Supernatants were collected 24 hours post infection and used to infect confluent MDCK cell cultures. Notably, we observed enhanced frequency as well as enhanced amplitude of viral protein accumulation in IFITM3 depleted cells during primary infection. Reduced production of infectious particles, in the face of enhance viral protein production, may therefore be a consequence of either limiting host factors or disruption of viral protein/host factor stoichiometry required for assembly of viable viral particles. Of interest, the viral cytophathic effect was greatly enhanced upon IFITM3 depletion in the presence or absence of the virulence factor NS1, a viral protein known to block many of the innate immunity responses [15] [16] [17] [18] [19] (figure 2F, G). However, deletion of NS1 results in complete failure of infectious particle production even upon IFITM3 depletion ( Figure 2H ). These observations would place IFITM3 function early in the viral life cycle and independent of NS1 function, consistent with reports that indicate IFITM3's antiviral activity is at the level of viral entry [14] . Depletion of the cell cycle/DNA damage checkpoint proteins CDC2 and CHEK1, like IFITM3, appeared to promote viral protein production and cytopathic response, while impairing assembly of infection-competent viral particles. A global comparison of the candidate modulators of H1N1 pathogenicity identified here with two whole-genome siRNA screens for modulators of cell cycle progression revealed a significant intersection ( Figure 3A ). However, CDC2 and CHEK1 depletion show quite distinct consequences on G1 versus G2 arrest suggesting their contribution to H1N1 infection may be independent of cell cycle control. CHEK1 has not been previously isolated in viral pathogenicity or viral replication screens, including those performed with the same siRNA library employed here ( Figure 3B , C, Tables 3 and 4 in Supporting Information S1). To investigate additional biological processes that may be associated with CHEK1 modulation of viral infection, we assembled a context-specific protein-protein interaction sub-network defined by the genomic Z-score distribution of the primary screen ( Figure S7 ). This subnetwork revealed the circadian gene Timeless, recently defined as a master regulator of the host defense response [20] , within the first-degree neighborhood of CHEK1 ( Figure 3D ). Given this association, we investigated the consequence of chemical inhibition of CHEK1 on H1N1 infection. We employed SB218078, an investigational CHEK1 inhibitor similar to one currently in clinical trials as an anti-neoplastic agent, with an in vitro IC50 of 0.015 mM and a K i,app. of 1564 [21, 22] . Pretreatment of cultures with 1 uM or 100 nM SB218078 for 12 hours resulted in significant inhibition of viral protein accumulation together with a marked virus-specific death response by 24 hours (Figure 4A , B). While some viral infection was detected at the 100 nM dose, viral protein production was severely limited at single cell resolution ( Figure 4B , C). These observations suggest that SB218078 is releasing a cell death response to viral detection that would otherwise be suppressed during the viral replication cycle. Viral-induced cell death was also observed upon siRNA-mediated CHEK1 depletion (Figure 2A ). The seemingly contradictory increase in infection frequency upon CHEK1 depletion may therefore be an indirect consequence of infection of low density residual cell populations with hypomorphic CHEK1 activity. Remarkably, SB218078 had no consequence on H1N1 replication in A549 cells, a cancer cell line often employed to test for modulators of viral replication and host responses [5, 23, 24] ( Figure 4E , F). However, a nontransformed, telomerase-immortalized bronchial epithelial cell line, derived from a different patient, HBEC3 [25] , was identical to HBEC30 in its re- Figure 4G ). These observations indicate intervention targets may be available in non-tumorigenic cells that are uncoupled from host regulatory networks in cancer cells, and potentially explain why CHEK1 was not identified in other efforts to date that have universally relied on cancer lines as screen hosts [4] [5] [6] [7] 26] . We next queried the behavior of gene depletions identified here that modulate H1N1 cytopathic effects to those in 4 wholegenome siRNA screens which measured influenza virus replication as the end-point assay [4] [5] [6] [7] . This allowed us to parse collective 'hits' into four functional classes ( Figure 3C , Table 5 in Supporting Information S1). Class 1: genes that, when depleted, enhance bronchial epithelial cell survival upon H1N1 exposure, and are required for viral replication. This class presumably represents host factors that facilitate viral infection and/or are required to support viral replication. Class 2: genes that, when depleted, reduce bronchial epithelial cell survival upon H1N1 exposure, and are required for viral replication. This, initially unanticipated but very intriguing class, likely represents host factors that deflect cell death checkpoint responses that would otherwise engage upon detection of viral infection. Class 3: genes that, when depleted, reduce bronchial epithelial cell survival upon H1N1 exposure and enhance viral replication relative to controls. This class presumably represents antiviral restriction factors that normally oppose infection. Class 4: genes, that when depleted, enhance bronchial epithelial cell survival upon H1N1 exposure and enhance viral replication as compared to controls. Of note, Class 2, which may represent novel intervention target opportunities, includes TRRAP, a large multidomain protein of the phosphoinositide 3-kinase-related kinases (PIKK) family that is a component of many histone acetyltransferase (HAT) complexes. TRRAP was recently identified as a bona fide oncogene in melanoma through cancer genome resequencing efforts, however, its transforming mechanism is unknown [27] . By nature, a challenge to siRNA-screening efforts is false negatives that derive from weak phenotypes due to suboptimal depletion of what are otherwise key factors in the biological process under investigation. One opportunity to help meet this challenge is to employ coherent behavior of gene sets to identify key biological processes supporting a phenotype rather than relying solely on an arbitrary scoring threshold for each individual gene. We employed Netwalk [28] here to facilitate identification of such gene sets based on overrepresentation of functionally coherent subnetworks within the graph (Figures S1, S2, S3, S4, S5, S6, S7, S8 and S9). One such subnetwork implicated prenylation of Rab-family GTPases in support of H1N1 replication ( Figure S4 and Figure 3D ). To test this we employed 3-IPEHPC, a specific inhibitor of the type II Geranylgeranyltransferases (IC50 of 1.27 mM and a K i of 0.211 mM for Rab1a modification [29] ). As such, 3-IPEHPC specifically inhibits modification of Rab-family proteins with a carboxy-terminal CC motif as opposed to the carboxy-terminal CAAX motif [29] . HBEC-30 cells pretreated with 3-IPEHPC for 24 hours were significantly refractory to infection by H1N1 (Figure 5A, B) . Inhibitory activity was observed at concentrations as low as 125 nM ( Figure 5C ). Unlike SB218078, A549 cells were also responsive to 3-IPEHPC ( Figure 5D, E) . While the use of a mouseadapted virus facilitates large-scale screening and allows comparisons with other published screening efforts, the extent to which results translate to seasonal or highly pathogenic strains is not established. Importantly, 3-IPEHPC was protective against infection with the avian strain H5/N1 and the recent pandemic swine flu strain H1/N1 ( Figure 5D, E) . A stark limitation of arrayed siRNA screens is the requirement for ''single gene'' phenotypic penetrance. This can limit sensitivity of detection of relevant molecular entities due to insufficient protein depletion and/or the presence of functionally redundant gene products. As a mechanism to potentially reveal combinatorial contributions of gene function to viral replication and cytopathic effects, we repeated the original screen using a library of 426 human microRNA mimics. These reagents have the advantage of inducing multigenic perturbations, though accurate assignment of target space is a significant challenge. This effort identified a small cohort of miRNA mimics that either enhanced or deflected H1N1induced cell death ( Figure 6A , B, Table 6 in Supporting Information S1). 11 of these were further examined for consequences on H1N1 viral protein production in HBEC30 cells, which identified both sensitizers and resistors that enhanced or repressed viral replication ( Figure 6C ). Of note, a test for ''hits'' that also have activity against the recent pandemic strain Cal/04/ 09 identified two miRNA mimics that impair Cal/04/09 protein production in A549 cells (hsa-miR-495 and hsa-miR-519a, Figure 6D ). To infer biological processes that may be engaged by the miRNAs that can impair H1N1 replication, we examined the intersection of predicted miRNA targets and single-gene perturbations that behaved similarly to the subject miRNA. Candidate miRNA target genes were selected based on seed sequence presence in 39 UTRs as defined by Target Scan context scores. These predictions were intersected with siRNA data from this study and those of the 4 whole-genome siRNA screens that measured influenza virus replication [4] [5] [6] [7] . When considered as a heuristic, this analysis produced three subnetworks that may correspond to the miRNA mode of action, namely the glycosylphosphatidylinositol transamidase, viral and host protein ubiquitylation [30, 31] and alternative mRNA splicing ( Figure 6E ). Here we have focused on isolation of H1N1 pathogenicity response modifiers in human bronchial airway epithelial cells (HBEC). This cell type was selected as tissue culture model that may be enriched for conservation of cell autonomous biological features representative of the viral target tissue. These cells resist plaque formation, but are highly sensitive to single cycle infection. From whole-genome siRNA and miRNA mimic screening, both candidate sensitizer and resistor response modifiers were identified. A key deliverable from this analysis was the identification of gene products that apparently serve to restrain cell death responses that would otherwise engage upon detection of viral infection. Though not required to support cell viability in the absence of viral challenge, depletion of genes in this class enhanced the death response to H1N1 infection concomitant with restraining H1N1 protein production. As such, this class may represent targets for interventions that restrain propagation of multi-cycle infection by facilitating suicide of infected cells prior to production of new infectious particles. A chemically addressable member of this class, CHEK1, showed strong activity in multiple HBEC lines but not in A549, a non-small cell lung tumor derived line commonly employed to model influenza virus infection. This suggests that intervention targets may be available in normal epithelial cells that are uncoupled from host regulatory networks in cancer cells. Racemic 3-IPHPC (2-hydroxy-3-imidazo[1,2-a]pyridine-3-yl-2phosphonopropionic acid) was prepared and characterized as described previously [32, 33] and stored at ,0uC and pH $7 [32, 33] . The purity was $98%by 1 H NMR. The inhibitor was tested in this work as the racemate [32, 33] . It was subsequently demonstrated that the individual enantiomers have markedly different IC 50 and Ki values for inhibition of Rab1a prenylation, thus the racemate value obtained here probably represents an upper limit with respect to the potency of the more active stereoisomer. HBEC30-KT cells were cultured in KSFM (Invitrogen Cat#17005) with 1% Pen/strep antibiotics as previously described [34] . MDCK and A549 cells [from ATCC] were grown in DMEM with 10% FBS. Plaque Assay 5610 5 MDCK [from ATCC] and HBEC-30KT cells [34] were plated in 6 well plates and grown to confluence overnight. Cells were infected with WSN virus at 10 fold dilutions with a starting concentration of 10 8 pfu/ml. Infected cells were allowed to incubate at 37uC with tilting every 10 minutes. After incubation liquid was aspirated and 2 ml of agar solution was added to wells and allowed to solidify for 1 min. Plates were incubated for 48 hrs at 37uC. Following incubation plates were fixed with formaldehyde for 1 hr. Fixative and agar was removed and cells were stained with crystal violet. HBEC30-KT cells [34] were plated in 96 well plates and incubated overnight. Cells were infected with WSN virus at an MOI 5. Whole cell lysates were collected at the indicated time point and separated by 12% SDS-PAGE gel and transferred to a nitrocellulose membrane. Cultures for immunofluorescence were fixed with 4% formaldehyde at indicated time point. Viral protein was detected in both cases with antibodies for pan influenza A (1:200), M2 (1:500) or NP (1:500) proteins{a-tubulin (cellsignalling rabbit mAb, cat#2125S), anti-NS1 [35] , anti-NP (Abcam, cat# ab20343)} followed by detection with either HRP conjugated secondary or staining with Alexa 498 (1:5,000) or Alexa 594 (1:5,000) conjugated secondary antibodies(from Invitrogen). Wells were imaged with a 20x lens on a BD Pathway 855 microscope. Imaged cells were segmented using Hoescht staining and distance from nucleus, aIVA fluorescence intensity was measured, with Attovision software. HBEC30-KT cells were infected with WSN virus and supernatants were collected at 24 hours post infection. Supernatants were then added to MDCK cells at 1% final concentration and MDCKs were fixed 14 hours after supernatant addition and viral production in MDCK cells was detected immunostaining. The siRNA screen was performed using the Dharmacon library targeting 21,125 genes HBEC30 were plated into 96 well plates at 10,000 cells per well and siRNAs were reverse transfected. Each siRNA pool was transfected in two sets of triplicates for a total of 6 wells for each siRNA, three wells for infection with IVA and three wells for mock infection, with a concentration of 50 nM for oligos and 0.1% DharmaFECT 3 reagent. Cells were incubated for 48 hrs after transfection and infected with influenza A/WSN/33 (H1N1) virus at an MOI of 5. Forty-eight hours after infection cell viability was assayed using CellTiter-Glo, 15 ml of Promega's CellTiter-Glo was added to wells on a 96 well plate for a final concentration of 7.5%. Plates were rocked for two minutes followed by 10 minutes incubation. Luciferase activity was measured with a PerkinElmer EnVision reader. The miRNA mimic screen was performed with the Dharmacon miRNA mimic library corresponding to 426 human miRNA's. Screening conditions were identical to those described above with the exception of a 72-hour incubation between transfection and infection. To remove position effects, raw values from each well were normalized to the median well of their respective row using the siMacro found at (http://sourceforge.net). To control for contamination and technical issues the top 5% of outliers with the highest coefficient of variation among triplicates were removed. Outliers were defined as wells with the largest distance among triplicate values. Normalized data was log2 transformed for proper distribution of sensitizers and resistors and a ratio of infected over mock infected was obtained. To control for batch effects, Z-Scores were calculated using batch specific variance where for each siRNA pool i Zi = xi-mbatch/sbatch, where x is the raw data to be normalized, m is the mean of the batch population, and s is the standard deviation of the batch population. Published data sets were obtained from four siRNA screens for influenza A modulators that used viral replication as an end point assay [4] [5] [6] [7] . Candidate hits in our screen were queried for behavior as regards viral replication. Hits that modulated viral replication greater than 1.5 standard deviations were assigned to functional classes. In cases were hits showed multiple phenotypes the strongest phenotype was used for classification. Screening data was compared for overlap with published hit lists for cell cycle regulators [36, 37] , host regulators of HIV infection [38] [39] [40] , and interferon stimulated genes (ISGs) [41] . Published hits that correlated with a change in cell viability greater than two standard deviations were considered as positive hits. Predicted targets of miRNAs based on seed sequence were obtained from TargetScan (http://targetscan.org). Network analysis of predicted hits was completed using Ingenuity IPA (http:// ingenuity.com) and queried for behavior in siRNA screens for regulators of influenza A infection. Z-scores were used as weights for NetWalk analysis [28] . Interactions with 350 highest and 350 lowest Edge Flux values were used to construct the networks with high and low z-scores, respectively. Analyses and graphics were done in the NetWalker desktop application (Komurov et al, manuscript submitted, http://research.cchmc.org/netwalker). HBEC30 or A549 cells were plated on 96 well plates overnight. Media was removed and replaced with media containing SB218078 (1mM, 100nM 10nM) 3-IPEHPC (12.5mM, 1.25mM, 125nM) DMSO (0.06%) or plain media. Cells were incubated overnight and then infected with WSN at an MOI of 5. Cells were fixed with 4% formaldehyde at 8 hours, 12 hours and 24 hours post infection and stained as described previously. SB218078 was purchased from Tocris biosciences cat # 2560 and dissolved in DMSO. 3-IPEHPC was dissolved in PBS. Figure S1 Individual siRNA oligo assays. HBEC30 cells were transfected in triplicate with four individual siRNA oligos and infected with WSN. Cell viability was measured 48 hours post infection and a two-tailed Student's t-test was performed to determine significance. Green boxes are oligos with a p value less than 0.05. (PDF) Figure S2 Network analysis of siRNA screen results. Data from siRNA screen results was used for NetWalk analysis. Nodes are colored based on Z-Score with red for positive and green for negative, edges are colored based on interactions, PPI: protein-protein interaction, TF-Target: gene regulation, GO: GO similarity. Networks analysis was performed with entire data set. (PDF)

Association of Fcγ Receptor IIB Polymorphism with Cryptococcal Meningitis in HIV-Uninfected Chinese Patients

BACKGROUND: As important regulators of the immune system, the human Fcγ receptors (FcγRs) have been demonstrated to play important roles in the pathogenesis of various infectious diseases. The aim of the present study was to identify the association between FCGR polymorphisms and cryptococcal meningitis. METHODOLOGY/PRINCIPAL FINDINGS: In this case control genetic association study, we genotyped four functional polymorphisms in low-affinity FcγRs, including FCGR2A 131H/R, FCGR3A 158F/V, FCGR3B NA1/NA2, and FCGR2B 232I/T, in 117 patients with cryptococcal meningitis and 190 healthy controls by multiplex SNaPshot technology. Among the 117 patients with cryptococcal meningitis, 59 had predisposing factors. In patients with cryptococcal meningitis, the FCGR2B 232I/I genotype was over-presented (OR = 1.652, 95% CI [1.02–2.67]; P = 0.039) and the FCGR2B 232I/T genotype was under-presented (OR = 0.542, 95% CI [0.33–0.90]; P = 0.016) in comparison with control group. In cryptococcal meningitis patients without predisposing factors, FCGR2B 232I/I genotype was also more frequently detected (OR = 1.958, 95% CI [1.05–3.66]; P = 0.033), and the FCGR2B 232I/T genotype was also less frequently detected (OR = 0.467, 95% CI [0.24–0.91]; P = 0.023) than in controls. No significant difference was found among FCGR2A 131H/R, FCGR3A 158F/V, and FCGR3B NA1/NA2 genotype frequencies between patients and controls. CONCLUSION/SIGNIFICANCE: We found for the first time associations between cryptococcal meningitis and FCGR2B 232I/T genotypes, which suggested that FcγRIIB might play an important role in the central nervous system infection by Cryptococcus in HIV-uninfected individuals.

Cryptococcal meningitis is the most common opportunistic fungal infection of the central nervous system in AIDS patients. Among HIV-uninfected patients, several predisposing factors for cryptococcal meningitis such as corticosteroid medication, solid organ transplantation and malignancy, etc, have been indentified. Yet cryptococcal infections in apparently healthy individuals are also increasingly being reported, especially from Asian data [1] [2] [3] . Our previous study has demonstrated an association between mannose-binding lectin (MBL) genetic deficiency and cryptococcal meningitis in HIV-uninfected patients [4] . However, MBL deficiency was present in only 21% of the cases, and for the remaining 79% of patients the underlying mechanism for susceptibility remained unclear. Fc gamma receptors (FccRs) mediate a variety of immune responses after binding to IgG-opsonized pathogens or immune complexes, and therefore act as immune regulators in both autoimmune and infectious diseases [5] [6] [7] [8] [9] . According to their affinity to IgG, FccRs are categorized into high-affinity and lowaffinity receptors. FccRI is the only known high-affinity receptor. Low-affinity FccRs which include FccRIIA, FccRIIB, FccRIIC, FccRIIIA, and FccRIIIB, are encoded by FCGR2A, FCGR2B, FCGR2C, FCGR3A, and FCGR3B genes, respectively. FCGR polymorphisms had been associated with the susceptibility and severity of various infections. FCGR2A 131R/R had been reported to attribute to the susceptibility of meningococcal infection, community-acquired pneumonia (CAP) caused by Haemophilus influenza, and the development of severe malaria [10] [11] [12] . FCGR2A 131H/H was reported to contribute to higher risk of bacteremia in pneumococcal CAP patients [13] . Another study showed that HIV-infected patients with FCGR2A 131R/R genotype progressed to a low CD4 + cell count at a faster rate, but conversely in individuals carried FCGR2A 131H/H there was an increased risk of Pneumocystis jiroveci pneumonia [14] . FCGR3A 158F/V gene polymorphism was not associated with progression of HIV infection, but predicted the risk of Kaposi's sarcoma [14] . A study on infections during induction chemotherapy found that FCGR2A 131H/H was associated with a decreased risk of pneumonia, FCGR3B NA1/NA1 associated with infections, and FCGR3A polymorphisms not associated with infections [15] . Sadki et al. investigated the influence of FCGR3A 158V/F and FCGR2A 131H/R polymorphisms on susceptibility to pulmonary tuberculosis in the Moroccan population but no association was found [16] . A study in East Africa found that the FCGR2B 232T/T genotype provided protectiveness for children against severe malaria [17] . A previous study by Meletiadis et al. investigated FCGR polymorphisms in patients with cryptococcosis, and found that FCGR2A 131R/R and FCGR3A 158V/V were over-presented, and FCGR3B NA2/NA2 was under-presented in patients with cryptococcosis [18] . The purpose of this study was to investigate FCGR polymorphisms in our series of patients to further verify the association between FCGR and cryptococcal meningitis. A total of 117 HIV-uninfected patients with cryptococcal meningitis were included. Subjects from both the patient and control groups were of Chinese Han ethnicity. Clinical information and predisposing factors of the patients are summarized in Table 1 . Of the 190 healthy control subjects, 111 were male (58.4%). The median age of the control subjects was 44 years (range, 12-79 years). Two samples failed genotyping of FCGR3A and 2 samples failed in genotyping of FCGR2B. Allele distributions of the tested FCGR genes in the control group were in Hardy-Weinberg equilibrium. The frequencies of FCGR2A, FCGR3A, FCGR3B and FCGR2B genotypes were shown in Table 2 . An association was found between FCGR2B 232I/T genotypes and cryptococcal meningitis based on dominant and over-dominant model. The FCGR2B 232I/I genotype was over-presented (OR = 1.652, 95% CI [1.02-2.67]; P = 0.039) and the FCGR2B 232I/T genotype was underpresented (OR = 0.542, 95% CI [0.33-0.90]; P = 0.016) in patients with cryptococcal meningitis in comparison with controls. No significant difference was found in the distribution of FCGR2A 131H/R, FCGR3A 158 F/V and FCGR3B NA1/NA2 genotypes. We further compared the genotype distribution of FCGR2A, FCGR3A, FCGR3B and FCGR2B between the 58 patients without predisposing condition and controls. Similar to results from the overall patient group, associations were also found between FCGR2B 232I/T genotypes and cryptococcal meningitis based on dominant and over-dominant model. Specifically, FCGR2B 232I/I genotype was also more frequently detected (OR = 1.958, 95% CI [1.05-3.66]; P = 0.033), and FCGR2B 232I/T genotype was also less frequently detected (OR = 0.467, 95% CI [0.24-0.91]; P = 0.023) in patients without predisposing factor than in controls. For the genotype distribution of other polymorphisms (FCGR2A 131H/R, FCGR3A 158 F/V and FCGR3B NA1/NA2), there was also no significant difference between patients and controls. The distribution of FCGR polymorphisms has been reported to exhibit substantial inter-ethnic variation. According to our population, frequencies of FCGR2A 131R/R, FCGR3B NA2/ NA2, and FCGR2B 232T/T in all subjects were 16%, 11%, and 7% respectively, similar to those reported among other Asian populations (which ranged 9-14%, 11-21%, and 5-11%) [19] [20] [21] [22] [23] [24] [25] [26] . Frequencies of these genotypes in Caucasian population were reported to be 19-34%, 38-43%, and 1-3% [18, 23, [27] [28] [29] [30] , which were different from our data. There seems no marked difference in the distribution of FCGR3A 158F/V genotypes between the Asian and Caucasian populations [18, 21, 23, 31, 32] . The four polymorphisms of low-affinity receptors genotyped in our study have each been demonstrated to affect functions of their encoded receptors. In FCGR2A, the G.A SNP at amino acid position 131 results in a histidine (H) to arginine (R) change (FCGR2A 131H/R), resulting in reduced affinity of the correspondent receptor to IgG2 [33, 34] . The T.G SNP at position 158 of FCGR3A causes a phenylalanine (F) to valine (V) substitution (FCGR23A 131F/V) and FCGR3A 158V/V encoded receptors show higher affinity to IgG1 and IgG3 [35, 36] . In the FCGR3B gene, five nucleotides (141,147,227,277 and 349) are combined to form two main haplotypes termed FCGR3B NA1 and FCGR3B NA2, and receptor encoded by FCG3B NA1 haplotype binds to IgG1 and IgG3 more easily [37] . Finally, FCGR2B 232I/T causes an isoleucine (I) to threonine (T) substitution in the transmembrane domain [22, 38] and receptors encoded by FCGR2B 232T are unable to interact with activating receptors [39] . Although FCGR polymorphisms have been demonstrated to be associated with susceptibility and severity of numerous infections, there has only been one previous genetic association study on the relationship of FCGR genotypes and cryptococcosis [18] . Meletiadis and colleagues genotyped FCGR2A 131H/R, FCGR3A 158F/V and FCGR3B NA1/NA2 in 103 cryptococcosis patients and 395 healthy controls. They found that in patients with cryptococcosis FCGR2A 131R/R and FCGR3A 158V/V were over-presented (Pvalues were 0.04 and 0.04), while FCGR3B NA2/NA2 was underpresented (P-value was 0.04). In our study, we found for the first time that cryptococcal meningitis was associated with the FCGR2B 232I/T genotypes, which was not reported in Metediatis' study. As the only known inhibitory FccR, FccRIIB has an immunoreceptor tyrosine-based inhibitory motif (ITIM) in its cytoplasmic domain, and thus it plays an important role in regulating the immune system [40] . FCGR2B 232I/T is located in the transmembrane domain, and the FccRIIB receptors encoded by FCGR2B 232T are unable to interact with activating receptors and exert inhibitory activity [38] . Published data have suggested the mutation genotype FCGR2B 232T/T to be a susceptible genotype for systemic lupus erythematosus [17, 22, 32] , and this genotype also provided protective effect for severe malaria in East African children [17] . The role of FccRIIB in cryptococcal infection is still not very clear. Like the activatory FccRs, FccRIIB can also recognize the major component of the capsule of C. neoformans, glucuronoxylomannan (GXM). In a previous study by Monari et al., the immunosuppressive effect of GXM was demonstrated to be dependent on FcRcIIB, based on the evidences that FccRIIB blockade inhibits GXM-induced IL-10 production and induces TNF-a secretion, and that the addition of monoclonal antibody to GXM reverses GXM-induced immunosuppression by shifting recognition from FccRIIB to FccRIIA [41] . Another study subsequently demonstrated that GXM triggered NO-induced macrophage apoptosis, which was dependent on FccRII [42] . These data support that FccRIIB plays a critical role in the pathogenesis of cryptococcal infection. In our study, it is the FCGR2B 232I/T heterozygote instead of the minor homozygote 232T/T that is under-presented in patient group. One study on children with idiopathic thrombocytopenia (ITP) also showed a similar pattern, that the FCGR2B 232I/T was less frequently detected in acute ITP in comparison with chronic ITP [27] . The reason for the heterozygotes 232I/T rather than 232T/T under-presenting in our patients and those acute ITP children has not been clarified. Unlike results from Meletiadis' study, no association among FCGR2A 131H/R, FCGR3A 158F/V, FCGR3B NA2/NA2 and cryptococcal meningitis was found in our study. The cause for discrepant results may be multifactorial. One was the ethnic differences between the two studies. Subjects in our study were of Chinese Han ethnicity, while the majority of subjects in Meletiadis' study were Caucasians (60%). As a matter of fact, the FCGR3A 158V allele was significantly increased only in patients who were Caucasian in Meletiadis' study. Secondly, all the cases in our study were diagnosed with cryptococcal meningitis, while some patients from Meletiadis' study were cryptococcosis without central nervous system involvement. Furthermore, both studies had relatively small sample sizes, which could be underpowered to generate positive results. In conclusion, our study suggested that FccRIIB genetic polymorphism may contribute to the susceptibility of cryptococcal meningitis. The overall association is relatively weak, which warrants validation in larger population. This study was reviewed and approved by the Ethic Committee/Institutional Review Board (HIRB) of Huashan Hospital, Fudan University, and informed written consent was obtained from each participant. A total of 200 volunteers and 117 unrelated patients with proven or probably diagnosed cryptococcal meningitis who were referred to Huashan Hospital, Fudan University, China, from 2001 through 2011 were recruited for the present study. Patients who met at least one of the following criteria were considered as proven cryptococcal meningitis: (1) Isolation of C. neoformans from cerebrospinal fluid (CSF) by culture or positive India ink smear, and (2) compatible histopathological findings, which are 5-10 mm encapsulated yeasts observed in brain tissue. Patients who had no microbiological or pathological documentation but present with positive cryptococcal antigen titer ($1:10) in CSF and met at least one of the following criteria were regarded as probable cryptococcal meningitis: (1) abnormal laboratory tests or an increased open pressure ($200 mmH 2 O) of CSF, (2) abnormalities of cranial imaging (Computerized Tomography or Magnetic Resonance Imaging) which could not be explained by other factors, and (3) comorbidities that compromise the host immune system. Cryptococcal antigen was determined using diluted CSF with the Latex-Cryptococcus antigen detection system (Immuno-Mycologics). Patients and volunteers were assessed for predisposing factors as follow, immunocompromising diseases (liver cirrhosis, chronic kidney diseases, autoimmune diseases, malignancies, solid organ transplantation) [2, 3, 43] , and corticosteroid (at prednisone equivalent dose of .0.3 mg/kg/day of for .3 weeks) or immunosuppressive medications (within 90 days before onset of cryptococcal meningitis) [44] , and idiopathic CD4 + T lymphocytopenia (unexplained CD4 + T lymphocytopenia with CD4 + T lymphocyte count ,300 cells/mm 3 ) [45] . Diabetes mellitus was also included, although this common condition is a controversial predisposing factor [3, 46] . Patients without any of the above mentioned predisposing factors were considered as apparently healthy hosts. Ten volunteers were excluded because of disclosed predisposing conditions, and the remaining 190 healthy volunteers were included in the control group. Four functional FCGR polymorphisms including FCGR2A 131H/R, FCGR3A 158F/V, FCGR3B NA1/NA2, and FCGR2B 232I/T were selected for genotyping after literature review of previous studies on association between FCGR polymorphisms and infectious diseases [11] [12] [13] [14] [15] [16] [17] . Venous blood was obtained by venepuncture from each subject. Genomic DNA was extracted using the QIAamp DNA kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. Genotyping of 8 SNPs in FCGRs (Table 3 ) was performed by multiplex SNaPshot technology using an ABI fluorescence-based assay discrimination method (Applied Biosystems, Foster city, CA, USA), which has been described in detail in previous studies [47, 48] . The multiplex SNaPshot detection of single-base extended probe primers was based on fluorescence and extended length detected by capillary electrophoresis on ABI3130XL Sequencer (Applied Biosystems, Foster City, CA, USA). Four pairs of primers for PCR amplification including 5 fragments of 587-2394 bp and 8 primers for SNaPshot extension reactions were designed by Primer3 online software (v.0.4.0) (http://frodo.wi.mit.edu/primer3/) according to the reference sequences from dbSNP (http://www.ncbi.nlm.nih.gov/SNP). There were homologous sequences between FCGRs, the specificity sequences were checked with the sequence databases using BLAST (http://www.ncbi.nlm.nih.gov/blast/blast.cgi). These sequences were also verified by SNPmasker1.1 (http://bioinfo.ebc. ee/snpmasker) to make sure that the different bases were caused by SNP [49] . And each primer pair was tested for potential primer-dimer and hairpin structures using the AutoDimer software (http://www.cstl.nist.gov/biotech/strbase/ AutoDimerHomepage/AutoDimerProgramHomepage.htm). The primers used in this study were listed in Tables 3. The PCR reactions were performed with 1 mL of DNA and 1 mL multiple PCR primers (the concentration was 1 mM) in a total volume of 20 mL containing 16 HotStarTaq buffer, 2.0 mM Mg 2+ , 0.3 mM dNTP, and 1 U HotStarTaq polymerase (Qiagen, Hilden, Germany). The cycling conditions for FCGR2A and FCGR3A were 95uC for 2 min, 35 cycles using 96uC for 20 s, 62uC for 2 min, and 72uC for 3 min, then 72uC for 10 min, and finally kept at 4uC. The cycling conditions for FCGR2B and FCGR3B were 95uC for 2 min, 7 cycles using 96uC for 20 s, 55uC for 2 min, and 72uC for 3 min, then 72uC for 10 min, and finally kept at 4uC. PCR products were then purified (add 1U SAP enzyme to 10 mL PCR products, incubate at 37uC for 1 hour, then, inactivate at 75uC for 15 min). The extension reaction to identify single nucleotide polymorphisms in the PCR products was performed in a total volume of 10 mL containing 2 mL purified PCR product, 1 mL primer (the concentration was 0.8 mM), 5 mL SNaPshot Multiplex Kit (Applied Biosystems, Foster City, CA, USA), and 2 mL ultrapure water. The cycling conditions for extension were 96uC for 1 min, 28 cycles of 96uC for 10 s, 52uC for 5 s, and 60uC for 30 s, and kept at 4uC. Then each extended product was added to 1 U shrimp alkaline phosphatase, incubated at 37uC for 1 hour, and the enzyme inactivated at 75uC for 15 min. Then, 0.5 mL was added to 0.5 mL Liz120 SIZE STANDARD (Applied Biosystems, Foster City, CA, USA), 9 mL Hi-Di (Applied Biosystems, Foster City, CA, USA), and sequenced by ABI3130XL Sequencer (Applied Biosystems, Foster City, CA, USA). Finally, the primary data was analyzed by GeneMapper 4.0 (Applied Biosystems, Foster City, CA, USA). Genotypes were determined by the type of nucleotide presented at SNP site, which was visualized by one or two different color peaks on the figures. For quality control, a random sample of 5% of the cases and controls was genotyped twice by different researchers, with a reproducibility of 100%. The minor allele counts were compared with database (http://www.ncbi.nlm.nih.gov/projects/SNP), and the data were matched well. Genotyping was performed blind to group status. Dominant, over-dominant, recessive and allelic models were applied for the analysis of genotype distribution. Hardy-Weinberg equilibrium, differences in gene polymorphism distributions between patients and controls were analyzed with 262 x 2 tests or Fisher's exact test where appropriate. P-values, odds ratios (ORs) and 95% confidence intervals (CIs) were obtained by SPSS 16.0 for Windows (SPSS, Inc, Chicago, IL). P-values ,0.05 were considered statistically significant.

Non-HIV Pneumocystis pneumonia: do conventional community-acquired pneumonia guidelines under estimate its severity?

BACKGROUND: Non-HIV Pneumocystis pneumonia (PCP) can occur in immunosuppressed patients having malignancy or on immunosuppressive agents. To classify severity, the A-DROP scale proposed by the Japanese Respiratory Society (JRS), the CURB-65 score of the British Respiratory Society (BTS) and the Pneumonia Severity Index (PSI) of the Infectious Diseases Society of America (IDSA) are widely used in patients with community-acquired pneumonia (CAP) in Japan. To evaluate how correctly these conventional prognostic guidelines for CAP reflect the severity of non-HIV PCP, we retrospectively analyzed 21 patients with non-HIV PCP. METHODS: A total of 21 patients were diagnosed by conventional staining and polymerase chain reaction (PCR) for respiratory samples with chest x-ray and computed tomography (CT) findings. We compared the severity of 21 patients with PCP classified by A-DROP, CURB-65, and PSI. Also, patients’ characteristics, clinical pictures, laboratory results at first visit or admission and intervals from diagnosis to start of specific-PCP therapy were evaluated in both survivor and non-survivor groups. RESULTS: Based on A-DROP, 18 patients were classified as mild or moderate; respiratory failure developed in 15 of these 18 (83.3%), and 7/15 (46.7%) died. Based on CURB-65, 19 patients were classified as mild or moderate; respiratory failure developed in 16/19 (84.2%), and 8 of the 16 (50%) died. In contrast, PSI classified 14 as severe or extremely severe; all of the 14 (100%) developed respiratory failure and 8/14 (57.1%) died. There were no significant differences in laboratory results in these groups. The time between the initial visit and diagnosis, and the time between the initial visit and starting of specific-PCP therapy were statistically shorter in the survivor group than in the non-survivor group. CONCLUSIONS: Conventional prognostic guidelines for CAP could underestimate the severity of non-HIV PCP, resulting in a therapeutic delay resulting in high mortality. The most important factor to improve the mortality of non-HIV PCP is early diagnosis and starting of specific-PCP therapy as soon as possible.

Pneumocystis pneumonia (PCP) not related to human immunodeficiency virus (HIV) can occur in immunosuppressed patients having malignancy or on immunosuppressive agents [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] . The mortality of patients with PCP without HIV infections ranges from 0 to 70% [1, [3] [4] [5] [6] [7] [8] [9] [10] , compared to that of HIV-infected PCP patients, which ranges from 10 to 20% [1, 4, 9] . Besides, mortality rates as high as 60-75% have been reported in PCP patients without AIDS who required mechanical ventilation [11, 12] . The higher mortality among non-HIV PCP patients has been attributed to severe lung inflammation [1, 4, 10] , although the exact etiology accounting for these large differences in mortality has not yet been determined. The A-DROP system proposed by the Japanese Respiratory Society (JRS), the CURB-65 score proposed by British Respiratory Society (BTS) and the Pneumonia Severity Index (PSI) proposed by the Infectious of Disease Society of America (IDSA) are widely used in classifying patients with community-acquired pneumonia (CAP) [13] [14] [15] in Japan. For the purpose of evaluating how correctly the conventional prognostic guidelines of CAP reflect the severity of non-HIV PCP, we retrospectively analyzed 21 patients with non-HIV PCP. It has never previously been reported that the severity of non-HIV PCP may be underestimated by these prognostic guidelines. This is the first report focusing on the limit of conventional prognostic guidelines of CAP for non-HIV PCP. From the end of 2009 to the beginning of 2010 we retrospectively reviewed all the cases of PCP diagnosed as CAP in the Kameda Medical Center, Chiba, Japan. All the patients had undergone HIV testing and were negative. Patients with hospital associated pneumonia (HAP) were excluded. PCP was diagnosed based on polymerase chain reaction (PCR) and conventional PCP staining with Grocott methenamine silver stain or Diff-Quick™ staining in respiratory samples such as induced sputum (IS) or bronchoalveolar lavage (BAL) fluid associated with radiographic infiltration on chest X-ray and computed tomography (CT) findings on admission. All radiographic pictures showed infiltration confirmed by the pulmonologist and radiologist in our hospital. The decision to perform IS or BAL examination depended on the patient's general condition at the discretion of attending physicians. PCP diagnosis was based not solely on the positive PCR respiratory specimen but also on the clinical and radiological findings consistent with the diagnosis of PCP as well as complete recovery with anti-Pneumocystis jirovecii treatment alone. No biopsy was performed during this study. We compared the severity of 21 patients with PCP classified by A-DROP, CURB-65, and PSI, and analyzed the background and laboratory data of each. Comparisons of group means were made by unpaired or paired t-tests or Mann-Whitney U-test. Contingency tables were evaluated by Fisher's exact probability test. p values < 0.05 were considered significant. The characteristics of the 21 patients with PCP are shown in Tables 1 and 2. The mean age was 71.5 years (range 57-88). Twelve patients (57.1%) had rheumatic or autoimmune disease which was the most common underlying disease, followed by malignancy (n = 10, 47.6%). Seventeen patients (77.8%) were receiving steroid or immunosuppressants for the underlying disease. Prophylactic therapy consisting of trimethoprim-sulfamethoxazole (TMP/SMX) was used by only 1 patient (4.8%). All the patients received oral or parenteral (TMP/SMX and adjuvant steroid therapy. Eighteen (85.7%) of the 21 patients had acute respiratory failure. Eight patients (38.1%) in the study died. The A-DROP scale was one of the prognostic guidelines for CAP proposed by JRS in 2005. It is a scoring system using age, degree of dehydration (serum blood urea nitrogen (BUN)), SpO 2 < 90%(PaO 2 < 60 mm Hg), orientation, and Table 3) . CURB-65 is also a prognostic guideline for CAP, proposed by BTS in 1996. It seems to be similar to the A-DROP system, the greatest difference being the age value used in the scoring. While JRS defines males > 70 years and females > 75 years as high risk elderly, BTS classifies high risk elderly age as > 65 years for both sexes. In our study 19 patients (90.5%) were classified as mild or moderate (class 0-2); 16 of the 19 patients (84.2%) developed acute respiratory failure, and 8 (42.1%) of the 16 died (Table 3) . The PSI was proposed by the Infectious Diseases Society of America/American Thoracic Society (IDSA/ATS). It is considered to be more complicated and less convenient to use than the other scoring systems. It consists of 19 items such as age, underlying disease, gender, vital signs, etc. In our study, 14 patients (66.7%) were classified as severe or extremely severe by PSI. All the 14 patients developed respiratory failure and 7 patients (50%) died ( Table 3) . The majority of the PCP patients were categorized in risk classes 2-3 by A-DROP and CURB-65, and III/IV by PSI. Of note are the very high patient mortalities in risk classes 2 by A-DROP and CURB-65, and class III/IV by PSI ( Table 3) . The comparison of the prognostic accuracy of each guideline for CAP is shown in Table 4 . It can be seen that the positive predictive values and negative predictive values for mortality in each system were low. The lactate dehydrogenase (LDH) value tended to be higher and Alb/BUN tended to be lower in the non survivor group compared with the survival group, but this was not statistically significant. There were no significant differences in serum β-D-glucan (β-DG), Krebs von den Lungen 6 (KL-6), body mass index (BMI) between the two groups. However, both the time between the initial visit and establishment of a diagnosis and the time between the initial visit and starting PCP therapy were significant much shorter in the survivor than non survivor group (Table 5 ). There are some widely used prognostic guidelines for CAP. These systems appear to be useful in assisting physicians to make more rational decisions regarding the need for admission [13] [14] [15] . Patient mortalities in the risk groups 3-5 on A-DROP and CURB-65, and IV-V on PSI have previously been reported as 11.5-23.3%, 11.6-21.0% and 12.5-29.2%, respectively [16] [17] [18] [19] . A striking fact is that the majority of the PCP patients were categorized as mild to moderate by these guidelines and resulted in respiratory failure, and poor outcomes. We emphasize that mortality prediction in PCP is not correct when these conventional guidelines for CAP are applied, even when PCP develops in the setting of CAP. Also, these guidelines definitely underestimate the severity of PCP as CAP. The important issue is why these guidelines cannot correctly estimate PCP severity. PCP without HIV infection shows quite different clinical pictures compared to PCP with HIV infection. PCP with HIV occurs slowly and gradually [20] . On the other hand, PCP without HIV is typically more acute and severe than when associated with AIDS [10] , often resulting in acute respiratory failure requiring a need for mechanical ventilation. We suppose that this results from the differences of pathologic mechanisms between PCP with and without HIV. It is evident that PCP without HIV is an allergic reaction originating from Pneumocystis jirovecii. Pneumocystis elicits many kinds of immune responses, including those by lymphocytes, macrophages, neutrophils, dendritic cells, and epithelial cells [1, 21] . There is now a considerable body of evidence showing that immune and inflammatory responses to Pneumocystis can have harmful as well as beneficial effects on host lungs. Another reason why the mortality rate of PCP without HIV remains high is presumably that it is difficult to diagnose PCP according to nonspecific signs, symptoms and/or no reliable culture. Bollée et al. documented that the leading symptoms of PCP in HIV-uninfected cancer patients were fever (85.7%), dyspnea (78.6%), cough (57.1%), and all three symptoms (44.6%) on diagnosis [5] , and 14.3% of the patients showed only one symptom. In our study, 4 out of 21 patients (19%) were asymptomatic. In addition, 6/21 patients (28.6%) showed abnormality in chest X-ray on admission. It is possible that steroids and immunosuppressive drugs could mask fever and general fatigue on the initial visit. We strongly believe that clinicians are unable to diagnose non-HIV PCP by clinical picture or chest X-ray alone. The association with P. jirovecii cysts has been reported in HIV-uninfected PCP to be one tenth of that in HIV-PCP [4] . Therefore, the sensitivity of conventional staining methods for diagnosis of HIV-uninfected PCP is lower than that for PCP with HIV. Our study demonstrated the sensitivity of conventional staining to be 23.8%. While Diff-Quik staining is highly sensitive, it requires considerable technical expertise [22] . It is likely that physicians are unable to diagnose PCP without HIV soon enough due to the reasons mentioned above. A clue for making the early diagnosis of PCP is serum β-D-gulcan (β-DG) and chest CT findings. Tasaka et al. reported the β-DG could be a serum indicator for the diagnosis of PCP with the cut-off value of 31 pg/ml [23, 24] . In our study, the sensitivity of β-DG in diagnosing PCP was 10/21 (47.6%) setting the cut-off value at 31 pg/ml. We suggest that testing β-DG is effective for diagnosis of PCP. In testing, BAL is also well known to be more sensitive than IS, as many physicians previously reported [23, 25] . In terms of a radiological approach, high resolution computed tomography (HRCT) should be performed if PCP is suspected. It is commonly known that chest CT shows ground glass appearance with a panlobular pattern or so-called crazy paving appearance in PCP patients [26, 27] . These findings are also found in viral pneumonias, mycoplasmal pneumonia, alveolar hemorrhage, methotrexate pneumonia, and others. However, where the patient's background and characteristics are conducive, the presence of PCP should be suspected. Conventional guidelines for CAP have recommended that clinical outcomes should be evaluated three days after initial therapy has been started [13, [28] [29] [30] [31] [32] [33] . In our study, 12 of the 13 (92.3%) patients who received accurate anti-PCP therapy within 3 days from initial visit were cured. Ten of the 12 (83.3%) patients received empiric therapy for PCP based on patient characteristics, laboratory data and radiological findings on HRCT. On the other hand, 7 of the 8 (87.5%) patients who received anti-PCP therapy that was initiated after day 4 died. PCP without HIV tends to develop acute respiratory failure and results in a more severe, acute form of acute respiratory distress syndrome (ARDS) than PCP with HIV. Thus, only three days of doctor's delay in starting PCP therapy could be fatal as our study showed. The limitation of our study is that it is a retrospective analysis in a very small population. Retrospective studies may be less reliable in terms of the data collected, particularly for data such as physical examination. A prospective study should be carried out and with more cases. In conclusion, we suggest that conventional prognostic guidelines for CAP might underestimate the severity of HIV-uninfected PCP. Physicians should be aware of the possibility that PCP may occur in non-HIV patients having malignancy or rheumatic disease, receiving steroid and/or immunosuppressive therapy. The most important factor for improving the mortality of PCP without HIV could be the time when anti-PCP therapy is started.

Proteomic Investigation of Falciparum and Vivax Malaria for Identification of Surrogate Protein Markers

This study was conducted to analyze alterations in the human serum proteome as a consequence of infection by malaria parasites Plasmodium falciparum and P. vivax to obtain mechanistic insights about disease pathogenesis, host immune response, and identification of potential protein markers. Serum samples from patients diagnosed with falciparum malaria (FM) (n = 20), vivax malaria (VM) (n = 17) and healthy controls (HC) (n = 20) were investigated using multiple proteomic techniques and results were validated by employing immunoassay-based approaches. Specificity of the identified malaria related serum markers was evaluated by means of analysis of leptospirosis as a febrile control (FC). Compared to HC, 30 and 31 differentially expressed and statistically significant (p<0.05) serum proteins were identified in FM and VM respectively, and almost half (46.2%) of these proteins were commonly modulated due to both of the plasmodial infections. 13 proteins were found to be differentially expressed in FM compared to VM. Functional pathway analysis involving the identified proteins revealed the modulation of different vital physiological pathways, including acute phase response signaling, chemokine and cytokine signaling, complement cascades and blood coagulation in malaria. A panel of identified proteins consists of six candidates; serum amyloid A, hemopexin, apolipoprotein E, haptoglobin, retinol-binding protein and apolipoprotein A-I was used to build statistical sample class prediction models. By employing PLS-DA and other classification methods the clinical phenotypic classes (FM, VM, FC and HC) were predicted with over 95% prediction accuracy. Individual performance of three classifier proteins; haptoglobin, apolipoprotein A-I and retinol-binding protein in diagnosis of malaria was analyzed using receiver operating characteristic (ROC) curves. The discrimination of FM, VM, FC and HC groups on the basis of differentially expressed serum proteins demonstrates the potential of this analytical approach for the detection of malaria as well as other human diseases.

The burden of malaria continues to worsen globally with a devastating impact on human health and corresponding impediment to economic improvement [1] . Despite worldwide initiatives, emerging drug resistance in different species of Plasmodium and paucity of information about the exact underlying mechanism of the disease pathogenesis are creating challenges for the management and eradication of the disease. Plasmodium falciparum (Pf) infection represents the major cause of malaria associated morbidity and mortality worldwide. Falciparum malaria (FM) accounts for approximately 247 million cases and one million deaths annually, particularly in sub-Saharan Africa [2] , while outside the African continents, Plasmodium vivax (Pv) is responsible for more than 50% of all malaria cases [3] . In order to survive within the host cells and ensure their reproduction, intracellular parasites like Plasmodium develop versatile mechanisms to exploit their host cells and induce new permeability pathways to permit the uptake of nutrients and the removal of waste products, resulting into activation of multiple host immune cascades and inflammatory responses [4] . Plasmodium infection also affects blood coagulation by diverse pathobiological mechanisms, which results into development of fatal hemorrhagic complication [5, 6] . Investigation of the parasite induced alterations in host proteome and modulation of different vital physiological processes have great clinical relevance in the light of diagnosis and prognosis. Recently, proteomic studies have contributed substantially to our understanding of the clinical proteome of human malaria parasites [7] , profiling humoral immune responses to Plasmodium infection [8] and the malaria parasite infection-induced changes in host erythrocyte membrane proteins [9] . The findings obtained from such studies have provided better understanding of the disease pathogenesis, host-pathogen interactions and host immune response. Analysis of human serum proteome is found to be very useful for the identification of potential disease-related markers, understanding disease pathogenesis and host immune response since various serum proteins exhibit rapid alteration in expression pattern in response to diseased conditions and show direct correlation with disease progression [10] . In recent years, a number of proteomic studies have been carried out to investigate the pathogen induced alterations in human serum/plasma proteome in different infectious diseases including dengue [11] , SARS [12] , leishmaniasis [13] , and leptospirosis [14] . In this study we have investigated the alterations in human serum proteome due to the P. falciparum infection for obtaining mechanistic insight about the disease pathogenesis and host immune response in the most virulent form of human malaria. Additionally, serum proteome changes in FM were compared with vivax malaria (VM); another widely distributed human malaria to study the similarities and differences in host responses against these two major plasmodial infections. To achieve this comparative analysis we have utilized selected dataset of our previous serum proteomics study on VM [15] , while additional proteomics and immune-assay-based experiments were performed using a bigger (compared to our previous report) clinical cohort of VM patients. The comparative study on FM and VM revealed that quite a few serum proteins associated with diverse essential physiological pathways, including acute phase response signaling, cytokine and chemokine signaling, complement cascades and blood coagulation are commonly altered in both of the plasmodial infections, while some uniquely modulated candidates such as calcium binding protein 39, calpain 10, regulator of G-protein signaling 7, serum paraoxonase/arylesterase, transthyretin in FM and ceruloplasmin, vitamin D-binding protein, serum amyloid P, alpha-2-macroglobulin, fibrinogen beta chain precursor in VM, were also identified. Recently, we performed serum proteomic alterations in another clinically relevant infectious disease, leptospirosis [14] . To evaluate the specificity of the identified protein targets and eliminate the generic febrile responses; expression level of the serum proteins differentially expressed in plasmodial infections (compared to the healthy subjects) was analyzed in leptospirosis patients from our previous study [14] . Another major intention of this present study was identification of the characteristics marker proteins, which can readily discriminate malaria patients (FM from VM as well) from healthy population as well as closely related infectious diseases with high accuracy. The potential serum protein biomarkers identified in our study were used to build statistical models, which successfully classified and predicted the clinical phenotypes of controls (healthy and febrile), FM and VM in a blinded study. Stringent inclusion criteria were employed during the selection of malaria patients and controls (HC and FC) to reduce preanalytical variations. Malaria patients (FM and VM) selected for this proteomic analysis were suffering from uncomplicated, nonsevere plasmodial infections with comparable range of parasitemia. Blood samples were collected from the malaria patients before administration of any antimalarial drugs. Majority of the patients were suffering their first episode of malaria, while some of the subjects had a past history of this disease (relapse or recurrent). The average age of the FM and VM patients included in this proteomic analysis was 34.2 years (SD = 10.93; range 20-53; median 35) and 32.9 years (SD = 10.76; range 20-52; median 32), respectively (Table 1) . To maintain uniform population profiles of test (FM and VM) and controls (HC and FC) for differential protein expression analysis, healthy and febrile control (leptospirosis patients) populations with comparable age distribution; mean values 33.4 years (SD = 8.69; range 20-44; median 31.6) and 30.5 years (SD = 8.31; range 23-42; median 26.5), respectively for HC and FC, were selected ( Table 1) . In this proteomics study we have performed two levels of gelbased proteomic analysis using classical two-dimensional gel In classical 2DE analysis, patients suffering their first episode of malaria as well as few patients with a past history of malaria (relapse or recurrent) and higher level of parasitemia (.5000 infected RBCs/mL blood) were included since it was difficult to get bigger cohort of malaria patients with similar parameters. In gel-based proteomic analysis samples were studied individually (n = 63 for classical 2DE and n = 30 for 2D-DIGE) rather than sample pooling to achieve better insights about biological variability from individual samples. In proteomic analysis two major high-abundance serum proteins; albumin and IgG were removed using Albumin & IgG Depletion SpinTrap (GE Healthcare) to reduce the dynamic range of the serum proteome. Depletion of these top two highabundance proteins removes more than 60% of the total protein content in human plasma or serum allowing detection of more proteins by increasing the effective concentration of the lowabundance proteins. Depletion of albumin and IgG effectively increased the overall spot number in 2D gels ( Figure S1A ). The efficiency of albumin and IgG depletion from human serum was evaluated by densitometric analysis of SDS-PAGE gels containing resolved serum proteins before and after depletion ( Figure S1B ). The densitometric analysis revealed around 85% and 80% depletion of albumin and IgG, respectively ( Figure S1C ). Serum proteome analysis of FM patients and healthy controls by 2DE identified 22 statistically significant (p,0.05) differentially expressed (with changes from 24.28 to +78.73-fold) protein spots (Table S1 .1). After staining with GelCode Blue Safe Protein Stain, over 700 protein spots were detected reproducibly in each gel by IMP7 software. Representative 2DE images of serum proteome profile of FM subjects and healthy individuals, and bar-diagrammatic representation of the fold change and 3D views of few selected spots are illustrated in Figure 1A and B. In MS analysis 12 different proteins were identified from the 22 differentially expressed protein spots, since in few cases MS analysis revealed similar identity for multiple protein spots appearing as different entities in 2D gels. The similar identity of multiple spots indicates the possibility of presence of various isoforms of those particular proteins probably due to the complex combinations of posttranslational modifications. Among the 12 identified proteins; 7 proteins were up-regulated (serum amyloid A, hemopexin precursor, apolipoprotein E, a-1-antitrypsin precursor, leucinerich a-2-glycoprotein, a-1-BN glycoprotein and a-1-antichymotrypsin precursor) and 5 proteins were down-regulated (haptoglobin, ficolin 3 precursor, apolipoprotein A-I, clusterin precursor and serum albumin) ( Figure S2 ; Table 2 and S2.1). Interestingly, serum amyloid A (spot U13 and U14) was found to be highly over expressed (.25-fold) in all the FM patients. Around 1300 protein spots were detected on each 2D-DIGE gels in DeCyder 2D software analysis. In 2D-DIGE analysis of FM and HC, total of 121 (around 9.3% of the entire detected spots) differentially expressed spots satisfied the statistical parameters (t-test; p,0.05), among which, 70 protein spots were up-regulated (with changes from 1.02 to 50.9-fold) and the remaining 51 were down-regulated (range from 1.2 to 24-fold) (Table S1 .2). Out of 121 spots, 36 and 9 spots were found to be statistically significant after performing false discovery rate (FDR) correction (Benjamini-Hochberg) and Bonferroni correction, respectively (Table S1 .2). All of the 121 differentially expressed spots (in FM compared to HC) identified in 2D-DIGE analysis were excised and subjected to MALDI-TOF/TOF MS analysis. We obtained reliable MS IDs for 63 protein spots out of 121 (Table S2. 2); while remaining 58 spots remained unidentified and produced virtually empty spectra, most likely owing to the presence of extremely diminutive quantity of proteins as indicated by the retrospective scrutiny of the spot volumes. The 63 protein spots identified by MS represented 30 (14 up-regulated and 16 down-regulated) differentially expressed proteins in FM patients ( Table 2 ; Figure 1C and S3). Proteins identified in 2DE experiment were also obtained in 2D-DIGE; additionally, new candidates were also identified by 2D-DIGE due to the higher sensitivity and reproducibility. 3D views and graphical representation of selected protein spots are shown in Figure 1D . A comprehensive comparative analysis of host responses in FM with that of VM (from the findings of our previous study on VM [15] ) was carried out to categorize the common and unique proteomic alterations in human serum in Pf and Pv infections. Almost half (46.2%) of the total identified proteins were commonly modulated in both plasmodial infections; however, the magnitude of proteomic alteration was different in these two types of malaria (Figure 2A and D). Compared to healthy controls, quite a few serum proteins such as calcium binding protein 39, calpain 10, regulator of G-protein signaling 7, serum paraoxonase/arylesterase and transthyretin precursor were found to be differentially expressed in FM but not VM. In contrary, some proteins like ceruloplasmin, vitamin D-binding protein, serum amyloid P, alpha-2-macroglobulin, fibrinogen beta chain precursor exhibited altered expressions only in VM patients (Table S3 ). Among the 19 proteins, which were differentially expressed (compared to HC) in both of the malaria, only alpha-2-HS-glycoprotein and serotransferrin precursor (transferrin) exhibited opposite trends in Pf and Pv infections. Rest of the 17 proteins exhibited similar trend of differential expression in FM and VM; although, fold-change values were different ( Figure 2D ). Compared to VM, 84 protein spots were found to be differentially expressed in FM (t-test; p,0.05). After FDR (Benjamini-Hochberg) and Bonferroni correction 10 and 3 protein spots (out of 84) remained significant, respectively (Table S1 .3). Out of 84, MS IDs for 43 protein spots were obtained in MALDI-TOF/TOF MS analysis, which indicated differential expressions of 13 proteins in FM compared to VM. Among those 13 differentially expressed proteins, 5 proteins (interleukin-17E precursor, serum amyloid A, ficolin 3 precursor, alpha-1antitrypsin and Ig kappa chain C region) were up-regulated while the remaining 8 proteins (alpha-2-HS-glycoprotein, apolipoprotein E, serotransferrin precursor, alpha-1-antichymotrypsin, leucinerich alpha-2-glycoprotein, AMBP protein, vitamin D-binding protein and haptoglobin) exhibited reduced expression level in Pf infected patients (Table S3 .4). Principal component analysis (PCA) using the extended data analysis (EDA) module of the DeCyder software v7 revealed distinct clustering of the three experimental groups (FM, VM and HC) ( Figure 2B ). Proteins present in at least 80% of the spot maps, which passed the filter of one-way ANOVA (p,0.01) test were included in this multivariate analysis. Additionally, a hierarchical cluster analysis was performed using the same protein selection Figure 2C ). 85 protein spots were found to be significantly differentially expressed in the malaria subjects compared to the healthy individuals. Further comparative analysis was performed keeping leptospiral infection as a febrile control to appraise the specificity of the identified malaria related serum markers. Although, some of the identified proteins exhibited similar trends of differential expressions in malaria and febrile controls, interestingly, expression levels of quite a few candidates including serum amyloid A, haptoglobin precursor, ficolin 3 precursor, hemopexin precursor, interleukin-17E precursor, retinol-binding protein, serotransferrin precursor, and vitronectin precursor were found to be altered in malaria patients (both FM and VM) but not in leptospiral infection (Table S4 ). Altered expression levels of identified serum proteins in falciparum and vivax malaria and leptospirosis (FC) has been illustrated bar-diagrammatically in Figure S4 . Validation of a few differentially expressed proteins was performed using different immunoassay-based methods including immunoturbidimetric assay, ELISA and western blotting to confirm the results of proteomic analysis. Haptoglobin and apolipoprotein A-I (Apo A-I) concentrations were directly quantified turbidimetrically in the serum samples of malaria patients (n = 37), healthy subjects (n = 20) and febrile controls (n = 6). Compared to the healthy controls, both FM and VM patients found to have lower serum level of haptoglobin and Apo A-I (p,0.0001 in a Mann-Whitney test) ( Figure 3A and B). The mean haptoglobin concentration was found to be 0.20860.048 and 0.33360.06 g/L in FM and VM patients respectively, while the healthy and leptospirosis (FC) populations exhibited a mean values of 0.91860.1 g/L and 0.88860.056 g/L (mean 6 SE). Likewise, Apo A-I exhibited more than three times lower mean value in both the malaria patients compared to the healthy subjects (39.3965.43, 43.1964 .96 and 137.0565.33 mg/dL in FM, VM and HC respectively). While the febrile controls shown a mean value of 76.5264.12 mg/dL, which is around 2-fold higher than the malaria patients. Serum retinol-binding protein (RBP) concentration was measured by sandwich ELISA. The serum levels of RBP was found to be significantly (p,0.01) lower in malaria patients ( Figure 3C ). Western blot analyses of four differentially expressed targets proteins; haptoglobin, serum amyloid A, clusterin and retinol-binding protein were performed on a subset of control [HC (n = 12) and FC (n = 6)] and diseased samples [FM and VM (n = 12 each)] ( Figure 3D ). CBB staining of the SDS-PAGE gels and Ponceau staining of the transferred blots containing the resolved proteins indicated equal loading of the samples in each lane ( Figure S5 ). Western blot analysis showed up-regulation of serum amyloid A and downregulation of haptoglobin, clusterin and retinol-binding protein in FM and VM patients (p,0.01) compared to the healthy and febrile controls. These results confirmed our findings from the proteomic analysis, and are illustrated graphically in Figure 3E . Thirty differentially expressed serum proteins identified in FM patients (compared to HC) were subjected to functional pathway analysis using Ingenuity Pathway Analysis (IPA). Out of those 30 candidates, 27 were eligible for network analysis (focus molecule) based on the IPA Knowledge Base criteria. Two overlapping interaction networks were identified where the highest scoring network included 14 out of the 27 focus molecules, while the second network included 10 focus molecules ( Figure S6 ; Table S5 .1). The most significant related functions derived from these overlapping networks included, lipid metabolism (14 proteins, p = 2.92E 209 -5.48E 203 ), inflammatory response (18 proteins, p = 1.07E 211 -5.48E 203 ), molecular transport (15 proteins, p = 2.92E 209 -5.48E 203 ), immune cell trafficking (10 proteins, p = 9.32E 208 -4.12E 203 ) and humoral immune response (7 proteins, p = 1.10E 205 -5.48E 203 ). According to this functional pathway analysis, Pf infection leads to the alteration of multiple serum proteins involved in diverse essential physiological pathways, including acute phase response (Ratio = 0.067, p = 1.11E 218 ) and primary immunodeficiency signaling (ratio = 0.048, p = 5.1E 205 ) (Table S5. 2). Functional analysis of differentially expressed proteins was also performed using Protein ANalysis THrough Evolutionary Relationships (PANTHER) and Database for Annotation, Visualization and Integrated Discovery (DAVID) databases. In PANTHER analysis blood coagulation system (p = 4.88E 205 ) was again identified. Moreover, the heterotrimeric G-protein signaling, interleukin signaling pathway and inflammation mediated by chemokine and cytokine signaling pathways were identified as related physiological pathways with statistical significance (p,0.05) (Table S5. 3). Further, DAVID analysis also confirmed modulation of complement and coagulation cascades (p = 1.28E 204 ) in FM (Table S5. According to the molecular function analysis by GeneSpring, most of the differentially expressed proteins identified in FM are related to binding (59.5%) and enzyme regulation activity (24%). A small fraction is involved in transport (9.5%) and antioxidant activity (7%) ( Figure S7A ; Table S6 ). Majority of the proteins reside in the extracellular region (61%), while some are located in cell (15%), organelle (11%), macromolecular complex (9%), and lumen (4%) as depicted in Figure S7B by cellular component analysis (Table S6 ). Biological processes analysis by GeneSpring indicated that identified proteins are involved in following biological process: response to stimulus (20%), biological regulation (20%), localization (14.5%), cellular process (11.5%), metabolic process (9%), immune system (9%), multi-cellular organismal process (8.5%), biogenesis (3%), signaling and development process (,4%) ( Figure S7C ; Table S6 ). Further comparative analysis with VM [15] indicates that both Pf and Pv infection lead to alteration of multiple serum proteins involved in diverse essential physiological pathways, including acute phase response signaling, chemokine and cytokine signaling, complement cascades, lipid transport and metabolism, and blood coagulation ( Figure 4 ). Table S7 and Figure S8 summarize different biological pathways and physiological functions associated with the differentially expressed serum proteins identified in FM and VM using multiple analytical tools. serum proteome of HC and FM patients. FM and HC samples were labeled with Cy3 and Cy5 respectively, while the protein reference pool (internal standard) was labeled with Cy2. (D) Graphical and 3D fluorescence intensity representations of few selected statistically significant (p,0.05) differentially expressed proteins in FM patients identified in biological variation analysis (BVA) using DeCyder 2D software. Graphs showing the decrease/increase in the standardized log abundance of spot intensity in FM compared to the HC cohort of the study (n = 8 Initially, the fidelity of a potential biomarker subset containing 5 proteins identified in 2DE (Table S8 .1A) was evaluated for discrimination of FM and HC. As shown in Figure S9A , the two groups (FM and HC) could be clearly classified by phenotype, thereby providing an additional, unbiased estimate of class prediction. Secondly, we applied the class prediction model based on initial cohort (n = 10) to independently predict (assign) the phenotypic class to either FM or HC group on an independent blind group of 16 subjects (8 newly recruited FM patients and 8 HC). The model provided accurate phenotypic classification; and 100% of the FM (n = 19) and 94.74% of the HC (n = 19) subjects were accurately classified into correct phenotypes (Table S8 .1C). We achieved 97.37% overall prediction accuracy on independent prediction [HC (n = 19) and FM (n = 19)] using partial least squares discriminant analysis (PLS-DA). For the final validation phase, we compared the performance of the biomarker subset using three well-known machine-learning methods: Decision Trees, Naïve Bayes and support vector machines (SVM). Table S8 .1B summarizes the percentage of samples classified during model training, cross-validation and independent prediction respectively, using the three different classifiers. We achieved, 97.37% overall prediction accuracy with SVM, Decision Trees and Naïve Bayes as well, on blinded prediction using the biomarker dataset for FM and HC (n = 19 each). Further, 7 differentially expressed proteins (Table S8 .2A) identified in 2D-DIGE were implicated as potential classifiers for the discrimination of FM and VM patients and healthy subjects employing similar type of analysis ( Figure S9B ). We achieved 95.83% overall prediction accuracy on blinded prediction (n = 23) using PLS-DA. Table S8 .2B summarizes the percentage of the samples classified during model training, cross-validation and independent prediction respectively. In the final validation phase; we achieved 100 and 95.83% overall prediction accuracy with Decision Trees and SVM respectively, followed by Naïve Bayes (91.66%) on blinded prediction using the biomarker dataset for FM (n = 8), VM (n = 7) and HC (n = 8). Next round of multivariate statistical analysis was performed to evaluate the efficiency of the identified serum proteins to discriminate the FM, VM and FC (patients suffering from leptospiral infection) ( Figure 5A ). 6 differentially expressed proteins (Table S8 .3A) identified in 2D-DIGE were implicated as potential classifiers. Table S8 .3B summarizes the percentage of the samples correctly classified during model training, cross-validation and independent prediction respectively using different classifiers. We achieved, 100% overall prediction accuracy with Decision Trees; 95.83% with SVM and Naïve Bayes and 91.67% with PLS-DA on independent prediction [FM (n = 8), VM (n = 8) and Lep (n = 6)]. Table S8 .1C, S8.2C and S8.3C, provides additional details on the confidence measure obtained on blinded prediction for each subject using a given statistical method. The confidence measure defines the strength of the prediction belonging to the particular class. Receiver operating characteristic (ROC) curve analysis was carried out to evaluate the individual performance of 3 classifier proteins; apolipoprotein A-I, haptoglobin and retinol-binding protein in malaria prediction. These 3 serum proteins were used as potential classifiers (along with other 3 candidates) to build statistical sample class prediction models employing PLS-DA and For proteins with multiple spots in the 2D gels, representative spot detail is provided. Exact values for each spot are provided in (Table S2) . other classification methods for FM, VM, FC and HC discrimination. The area under the ROC curve (AUC) indicates the accuracy of different classifier proteins to distinguish FM, VM and leptospirosis from HC ( Figure 5 ). ROC curves demonstrate apolipoprotein A-I (AUC = 0.957) and haptoglobin (AUC = 0.936) as efficient predictor proteins for falciparum malaria detection. A cut off value .112.1 mg/dL for Apo A-I revealed the specificity and sensitivity of 90% and 95%, respectively; while haptoglobin at a cut off value .0.465 g/L provided 95% specificity and 90% sensitivity in predicting Pf infection. Retinolbinding protein (AUC = 0.879) exhibited moderate sensitivity (66.6%) and specificity (80%) for FM at a cut off value .30.88 mg/mL ( Figure S9C ; Table S9 ). Precondition efficiency of the classifier proteins for vivax malaria was also evaluated and found to be highly appreciable for Apo A-I (AUC = 0.979; 94.12% sensitivity and 95% specificity at a threshold value .96.59 mg/ dL), haptoglobin (AUC = 0.888; 76.47% sensitivity and 95% specificity at a threshold value .0.45 g/L) and retinol-binding protein (AUC = 0.875; 76% sensitivity and 90% specificity at a threshold value .28.61 mg/mL) as well ( Figure S9D ; Table S9 ). Accuracy of classifier serum proteins in prediction of leptospirosis (FC) was also tested ( Figure 5 ). Although Apo A-I (AUC = 0.783; 66.6% sensitivity and 90% specificity at a threshold value .111.1 mg/dL) exhibited fine proficiency in detection of leptospiral infection; performance of haptoglobin (AUC = 0.508; 66.6% sensitivity and 50% specificity at a threshold value .0.845 g/L) and retinol-binding protein (AUC = 0.558; 50% sensitivity and 65% specificity at a threshold value .34.99 mg/ mL) were poor (Table S9 ). ROC analysis revealed that the serum levels of the classifier proteins, particularly Apo A-I and haptoglobin exhibited good correlation with plasmodial infections and could further be investigated as potential surrogate protein markers for malaria. Among the four different species of Plasmodia, which cause malaria in human, Pf and Pv account for over 90% of the total malaria cases worldwide. In this study, we used proteomics to decipher the alteration in human serum proteome due to the Pf infection to gain insight into the disease pathogenesis and host immune response. We also performed a comparative analysis of host response in Pf and Pv infection. The comparative proteomic analysis of plasmodial and leptospiral infection (febrile control) was performed to appraise the generic febrile responses and specify the malaria related serum markers. The ultimate aspiration of this study was to identify potential marker proteins that can distinguish the malaria patients from the healthy or febrile controls as well as discriminate between the Pf and Pv infections with high accuracy. Although a number of earlier proteomic and immunoassay-based studies have demonstrated the alteration of limited set of serum proteins in malaria [16] [17] [18] hitherto, there was no attempt for a comprehensive analysis to establish a panel of classifier proteins that can readily discriminate the FM and VM groups from the controls. Our results indicate that various vital physiological pathways, including acute phase response signaling, cytokine and chemokine signaling, complement cascades, lipid transport and metabolism, and blood coagulation are modulated in Pf and Pv infections (Figure 4) . Alteration of the levels of several acute phase proteins (APPs) and multiple members of serum complement cascade as well as complement regulatory proteins (Table 2) due to the plasmodial infections is consistent with earlier findings [19] [20] [21] . Increased expression levels of circulating acute-phase amyloid proteins like serum amyloid A and P provide non-specific resistance against the pre-erythrocytic stages of Plasmodium, limit tissue damage and promote a rapid return to homeostasis [22, 23] . Interestingly, human serum paraoxonase (PON1) an HDLassociated esterase which protects lipoproteins against oxidation, found to be down-regulated in falciparum malaria patients. Acute inflammatory stimuli lead to reciprocal regulation of SAA and PON-1 [24] . Decreased serum PON-1 activity in context with falciparum malaria may in part be attributable to higher SAA level. In course of the disease progression, malarial parasites growing in the erythrocytes degrade hemoglobin and generate reactive oxygen species (ROS), which lead to increased oxidative stress within the erythrocytes and outside the parasitized cells. As a result, to circumvent the situation, enhanced plasma levels of antioxidant defense associated enzymes/proteins such as superoxide dismutase-1 (SOD-1) are observed in acute malaria patients [25] . In both FM and VM patients we have identified elevated serum level of hemopexin, a heme-binding protein, which provides the second line of defense against hemoglobin-mediated oxidative damage during intravascular hemolysis [26] . Increased production of this acute phase protein by the host defense system could be helpful to circumvent the pro-inflammatory response with oxidative stress generated in patients with Pf or Pv infection. In contrast, haptoglobin (Hp) found to be significantly downregulated in FM and VM patients. Hp removes free hemoglobin (Hb) released during parasite induced hemolysis, and disappears as the Hp-Hb complexes leading to the malaria associated hypo-or ahaptoglobinemia and is a promising inflammatory marker to evaluate the severity of the Plasmodium infection [21] . Earlier reports have demonstrated the possible role of this APP as an epidemiological marker for malaria [21, 27] . Erythrocyte invasion is an essential gateway to malaria disease and a key target for disease intervention. Signaling via the erythrocyte beta 2-adrenergic receptor and heterotrimeric guanine nucleotide-binding protein (Gas) regulates the entry of the human malaria parasite P. falciparum. Disruption of the interaction between the G-alpha-s subunit of the Gs protein and the receptor results in a reduced erythrocyte invasion by the parasite and subsequent low level of parasitemia [28] . Down-regulation in regulator of G-protein signaling in FM patients might be due to some host response to combat this parasitic infection. Conversely, up-regulation of apolipoprotein E was observed in the malaria patients. This apolipoprotein also inhibits Plasmodium invasion, since it shares the cell entry mediators (heparan sulphate proteoglycans and/or low density lipoprotein receptor) with the parasite [29] . The pathway analyses and densely connected networks based on our results provide an insight into the underlying molecular mechanisms of malaria. Early and accurate diagnosis is critical for the effective treatment and management of malaria. In recent years, multivariate projection methods are being successfully applied to analyze biological data obtained through genomic, transcriptomic or proteomic approaches to study various human diseases, with implications for diagnostics and clinical management [30, 31] . A sub-set of the proteins identified in our proteomic analysis was used to build statistical sample class prediction models to identify the classifier marker proteins for FM, VM and HC discrimination. Interestingly, two key classifier proteins: serum amyloid A and haptoglobin differentially expressed consistently in all of the malaria patients (FM and VM) compared to the control subjects (HC and FC) and remained statistically significant after FDR (Benjamini-Hochberg) and Bonferroni correction of the p-values obtained in t-test; indicating very strong correlation between the expression levels of these two serum proteins and plasmodial infections. The recognition ability of the prediction models for FM, VM and HC discrimination and cross-validation was almost 100% (Table S8) . We controlled for the statistical false discovery rate using three distinct, iterative validation steps: (i) k-fold crossvalidation algorithm for the original cohort, (ii) application of the marker subset identified in the original cohort to classify newly recruited patients, and (iii) performance of the marker subset validated using three well-known machine learning methods. In our study, biological replicates were investigated, i.e. each proteome profile was representative of a different human subject, and hence the data-sets are characterized by low homogeneity, conferring to the protocol a very high level of variability and complexity. Indeed the extreme heterogeneity or large biological variations including gender, age, genetic factors, dietary considerations, environmental factors and drug treatment affects the detection, validation and establishment of ''gold standard'' serological biomarkers [10, 32] . Nonetheless, the accurate discrimination among the FM, VM and control groups obtained by various prediction models on the basis of differentially expressed candidate proteins testifies to the excellent potential of this analytical approach for the detection and discrimination of VM and FM ( Figure 5 ; Figure S9 ). It should also be noted that uncomplicated FM and VM patients with diverse range of parasitemia; mainly low and moderate parasitemic (,5000 parasites/mL blood), were used for the validation of the prediction models (Table 1) . Even so, the discrimination accuracy of the study is very appreciable indicating the capability of our analytical approach for the detection of very low-level of parasitemia, which is highly promising from a diagnostic point of view. Although, diagnosis of malaria on the basis of microscopic examination of thin or thick smears of peripheral blood is the most commonly used and well-accepted method, but it requires highly trained personnel for smear interpretation, frequently fails to distinguish mixed-species infections or diagnose patients with ''sub-microscopic'' parasitemia below the detectable limit of blood smears, and in many areas of endemicity the operating characteristics of microscopy are poor [33, 34] . In quest of an early and accurate diagnosis of malaria and discrimination of Pf, Pv or mixed infection, establishment of serum protein markers can be an attractive approach apart from clinical symptoms and conventional microscopic examination of blood smears. To this end, some of the classifier candidate proteins identified in this study; such as serum amyloid A, paraoxonase, apolipoprotein A-I and E, haptoglobin, hemopexin, and complement C4 are very important due to their functional relevance in malaria pathogenesis and could further be investigated as potential surrogate protein markers for clinical implications. Various rapid diagnostic tests (RDTs) are in practice for malaria diagnosis, which diagnose the infection on the basis of detection of parasite proteins/antigens e.g. histidine-rich protein II (HRP-II) or lactate dehydrogenase (LDH) [35] , whereas for the first time we have demonstrated the discrimination between FM and VM patients based on protein expression in human host. Malaria RDTs are used regularly in clinics due to the low cost, sensitivity and less detection time. However, analysis of frozen specimens of blood from parasitaemic patients using existing RDTs is bit challenging. Another limiting factor is the shelf-life of RDTs, since most of the existing RDTs deteriorate rapidly on exposure to moisture (humidity) and high temperature. Moreover, significant variations may appear between technicians in both RDT preparation and result interpretation process depending on experience of the performer, manual proficiency and visual perception [36] . To this end, serum protein markers can be potential candidates for development of an alternative sensitive diagnostic approach for malaria. Development of highly sensitive biosensors for the identified surrogate proteins might be attractive from a diagnostic point of view. In summary, the present study demonstrates the application of diagnostic proteomics to decipher host responses against the human malaria parasites Pf and Pv, and identifies potential candidate biomarkers for these two plasmodial infections. In this comprehensive proteomic analysis we have identified multiple differentially expressed serum proteins with versatile biological functions, indicating the modulation of multiple vital physiological pathways in FM and VM patients. We anticipate that information obtained from this study will provide valuable insight into the underlying molecular mechanisms of malaria and may help to establish early detection surrogates for these infectious parasitic diseases to meet the need for better diagnostics and effective therapy. Some of our identified classifier proteins such as serum amyloid A, apolipoprotein A-I and E and haptoglobin, which successfully discriminated FM from VM might be prognostic host markers for disease severity. To this end, it would be interesting to elucidate the fate of the identified serum proteins in severe malaria patients and could be a future continuation of this study. Diagnostic impact of the identified serum biomarkers in clinics and specificity for malaria prediction can only be established after investigation of the disease patterns in large clinical cohorts. This proteomic analysis was performed with the approval of the institutional ethics committee of Seth GS Medical College and King Edward Memorial hospital, Parel, Mumbai, India. Patients suffering from uncomplicated Pf or Pv infection with asexual parasite count more than 1000 per mL of blood were selected for this study. A total of 37 patients, with uncomplicated FM (n = 20) or VM (n = 17) confirmed through microscopic examination of a thin peripheral blood smear followed by RDT were enrolled for this proteomic study. In addition, blood specimens were collected from age and sex matched leptospirosis patients (n = 6) as febrile controls, and healthy subjects (n = 20) to perform comparative proteomic analysis. Written informed consent was taken from each participant (malaria patients and controls) prior to the sample collection process. Demographic, epidemiological and clinical details of all malaria patients and febrile controls (FC) selected for this proteomic study are provided in Table 1 . Blood samples (5.0 mL) were collected from the antecubital vein of the subjects using serum separation tubes (BD VacutainerH; BD Biosciences). Immediately after blood collection the tubes were kept in ice for 30 mins for clotting. Serum separation was performed as described previously [15] . In brief, after clotting, the samples were centrifuged at 2500 rpm at 20uC for 10 mins and serum was collected carefully from the upper surface. Collected serum was divided into multiple aliquots and stored at 280uC until time of analysis to prevent protein degradation. Prior to proteomic analysis, maximum 2-3 freeze/thaw cycles were allowed for any serum sample to reduce pre-analytical variations. Crude serum was diluted five times with phosphate buffer (pH 7.4) and subjected to mild sonication in a Vibra cell sonicator using the following settings: 6 cycles of 5 sec pulse; 30 sec gap in between; at 20% amplitude. After sonication, the top two highabundance serum proteins (albumin and IgG) were removed using Albumin & IgG Depletion SpinTrap (GE Healthcare) following the manufacturer's instructions. Extraction of protein from depleted serum samples was performed employing TCA/acetone precipitation method as described by Chen et al., with slight modifications [37] . In brief, depleted serum samples were diluted (1:4 ratio) with ice-cold acetone containing 10% (w/v) TCA. Uniform mixing was performed using mild vortexing for 15 sec and the mixture was allowed to incubate at 220uC for 2 hrs for protein precipitation. After completion of the incubation period, tubes were centrifuged at 1000 g for 15 min at 4uC. Supernatants were separated and kept in fresh microcentrifuge tubes, and the pellets were dissolved in rehydration buffer [8 M urea, 2 M thiourea, 4% (w/v) CHAPS, 2% (v/v) IPG buffer (pH 4-7; Linear), 40 mM DTT and traces of bromophenol blue]. In order to precipitate the remaining amount of proteins present in the collected 10% TCA/acetone-containing supernatants, 1 mL icecold acetone was added to each tube and the samples were subjected one additional round of precipitation and extraction process. In all cases, prior to re-suspension in rehydration buffer, the pellet was briefly air-dried. Prior to proteomic analyses, protein concentration in the samples was quantified using the 2D-Quant kit (GE Healthcare) following the manufacturer's instructions. A total of 600 mg of depleted serum protein extract dissolved in 350 mL of rehydration buffer was loaded on 4-7 pH range IPG strips (18 cm) and underwent passive rehydration for 14-16 hrs. Isoelectric focusing (IEF) was performed on an Ettan IPGphore 3 isoelectric focusing unit (GE Healthcare) for overall approximately 78 kVh using the following voltage settings: 200 V for 4 h (step and hold), 500 V for 1 h (step and hold), 1000 V for 1 h (step and hold), 8000 V for 3 h (gradient), and 8000 V for 7:30 h (step and hold). After completion of IEF, the focused IPG strips were stored at 220uC until the second dimensional analysis was performed. Preceding to the second dimensional separation, each strip was equilibrated to reduce and alkylate the proteins (for 15 min each) using equilibration buffer containing 6 M Urea, 75 mM Tris-HCl pH 8.8, 29.3% (v/v) glycerol, 2% (w/v) SDS, and 0.002% (w/v) bromophenol blue. Just prior to use, 1% (w/v) DTT or 2.5% (w/v) IAA was added in the first (reducing) and second (alkylating) equilibration buffer, respectively. The second dimension was performed on 12.5% SDS polyacrylamide gels using an Ettan DALTsix electrophoresis unit (GE Healthcare). After electrophoresis GelCode Blue Safe Protein Stain (Thermo Scientific, USA) was utilized for visualization of the protein spots. Proteins extracted from each of the subjects were run in duplicate to verify the reproducibility and curtail technical artifacts. Each CyDye (Cy3, Cy5 and Cy2) was resuspended in anhydrous N, N-dimethylformamide (DMF) to prepare a stock dye concentration of 1 mM. A working solution of 400 pmol of each CyDye was made by further dilution of the stock with DMF. Samples (test and control) were labeled with Cy3 and Cy5, while a mixture of equal amounts of all samples to be analyzed in the experiment, regarded as internal standard, was labeled with the third fluorescent dye; Cy2 according to the manufacturer's instructions (GE Healthcare). In brief, the pH of each sample was adjusted to 8.5 using 100 mM NaOH. 50 mg of each protein sample [malaria, controls (HC/FC) and internal standard] were separately labeled with 400 pmol of CyDyes. After addition of CyDyes, samples were incubated on ice for 30 in the dark. Labeling reaction was stopped by addition of 10 mM lysine followed by incubation on ice for additional 10 min. Dyeswapping was performed while labeling the test and control samples for eliminating any type of dye effects. After labeling, samples labeled with Cy3, Cy5 and Cy2 were mixed, diluted with the rehydration buffer and loaded on 18 cm, 4-7 pH IPG strips. Subsequent IEF and SDS-PAGE separation were performed following the same protocol as previously described in the 2DE section. Image acquisition and data analysis was performed as described previously [15] . In brief, after staining, the 2D gels were scanned by using LabScan software version 6.0 (GE Healthcare) and analyzed by using ImageMaster 2D Platinum 7.0 software (GE Healthcare). Comparative analysis of FM samples was performed by creating different ''match sets'' and using the HC samples as reference. Spot detection parameters were specified as: Smooth: 7, Saliency: 100 and Min Area: 5. After automatic detection of the spots through IMP7, manual refinement was performed to eliminate any contaminating artifacts, such as streaks or dust particles. Spot quantification was performed in % vol value using ImageMaster algorithm. It provided normalized value that remains relatively independent of variations due to staining or protein loading. The gel analysis tables, histograms and 3D images generated by the software were used for further analysis. 2D-DIGE gels were scanned using Typhoon 9400 variable mode imager (GE Healthcare) at a 100 mm resolution employing suitable excitation/emission wavelengths for each of the CyDye [Cy3 (523/580 nm), Cy5 (633/670 nm) and Cy2 (488/520 nm)]. After scanning, gel images were cropped properly using Im-ageQuant software; version 5.0 (GE Healthcare) prior to importing in DeCyder 2D software; version 7.0 (GE Healthcare) for comparative analysis and relative protein quantification across the FM and control samples. Comparative analysis was performed using two different modules, differential in-gel analysis (DIA) and biological variation analysis (BVA) of the DeCyder software. Preliminary analysis was performed using DIA module to detect spots on a cumulative image derived from merging up to three individual images from an in-gel linked image set (malaria, controls and internal standard). It permits the pair-wise comparisons of each normal and malaria samples to the mixed standard present in each gel and offers spot-wise protein abundance as ratios. Further analysis was performed using BVA module to get the variation in protein expression levels between any of the two experimental groups (FM vs. VM, FM vs. HC and VM vs HC) across all the sets. Statistical significance of the average ratio of expressions was analyzed by Student's t-test. Protein spots exhibiting differential expression with reproducibly and statistical significance (p,0.05) were considered for further analysis. Bonferroni correction (for reducing Type I errors) of the p-values obtained from Student's t-test was performed using standard Bonferroni procedure to recognize those marker proteins which have very strong connection with the diseased state (remains significant after Bonferroni correction). Since Bonferroni correction is extremely conservative; comparatively less stringent false discovery rate (FDR) correction was also performed as detailed in Benjamini and Hochberg [38] . We also performed a comparative analysis of FM data-set obtained in this study with our previously published VM data [15] . Clustering of the three experimental groups (FM, VM and HC) was performed by principal component analysis (PCA) using an algorithm included in the extended data analysis (EDA) module of the DeCyder software. Proteins present in at least 80% of the spot maps and passed the filter of the one-way ANOVA (p,0.01) test were included in this multivariate analysis. Additionally, a hierarchical cluster analysis was performed using the same protein selection criteria. Statistically significant (t-test, p,0.05) differentially expressed proteins spots identified in regular 2DE and 2D-DIGE experiments were selected for further MS analysis to establish protein identity. GelCode Blue stained preparative gels containing much higher amount of protein (1 mg) were used for excision of the spots of interest specified in the 2D-DIGE experiment. Spot excision was performed manually. In-gel digestion of the proteins separated by 2D gel electrophoresis was performed as described by Shevchenko et al., with slight modifications [39] . In short, gel slices were cut into small cubes (,161 mm) and washed with 50 mL of stain removal solution (25 mM ammonium bicarbonate buffer) for removal of CBB stain. After washing, 50 mL of 25 mM ammonium bicarbonate/acetonitrile (1:1 v/v) was added, followed by 5 min incubation with occasional vortexing at room temperature. After incubation, the solutions were removed. These two steps are repeated for three times. Then, 50 mL reduction solution (10 mM DTT in 100 mM ammonium bicarbonate) was added and the gel pieces were incubated for 60 mins at 56uC in an air thermostat. Tubes were allowed to cool to room temperature after incubation, and 50 mL of 25 mM ammonium bicarbonate buffer was added to wash the gel pieces followed by dehydration with 25 mM ammonium bicarbonate/acetonitrile (1:1, v/v). After this step, alkylation solution (50 mM IAA in 100 mM ammonium bicarbonate) was added and the tubes containing the gel pieces were incubated for 30 mins at room temperature in dark. Rehydration and dehydration steps were performed twice and gel pieces were allowed to dry. Once the gel slices were properly dried, trypsin solution (Trypsin Gold; Promega, Madison, Wisconsin, United States) was added to the gel pieces keeping the ratio of trypsin: protein around 1:10 (w/w) and incubated at ice for 30 mins for absorption of the solution. After this step, the tubes were incubated overnight at 37uC. Adequate amount of ammonium bicarbonate buffer was added to cover the gel pieces. Extraction of the digested peptides from the gel matrix was performed using 100 mL of extraction buffer (0.2% formic acid in 66% acetonitrile) after completion of the enzymatic reaction. Extraction step was repeated thrice to ensure maximum recovery of the digested peptides. The collected supernatants were pooled in a single tube and concentrated using speed vac. After extraction trypsin digested samples were further processed using Zip-Tip C18 pipette tips (Millipore, USA) according to the manufacturer's protocol for removal of salts and enrichment of the peptides. Subsequent to enrichment and purification through the Zip-Tip pipette tips, peptide mixtures were dissolved in 0.5 mL of CHCA matrix solution (5 mg/mL CHCA in 50% ACN/0.1% TFA) and spotted onto a freshly cleaned MALDI target plate. Spots were allowed to dry for 30 mins at room temperature. After air drying, the crystallized spots were analysed using a 4800 MALDI-TOF/ TOF mass spectrometer (AB Sciex, Framingham, MA) linked to 4000 series explorer software (version 3.5.3). All mass spectra were recorded in a reflector mode within a mass range from 800 to 4000 Da, using a Nd:YAG 355 nm laser. The acceleration voltage and extraction voltage were kept at 20 kV and 18 kV respectively. Six point calibration of the instrument was automatically performed by a peptide standard Kit (AB Sciex) that included des-Arg1-bradykinin (m/z 904.468), Angiotensin I (m/z 1296.685), Glu1-fibrinopeptide B (m/z 1570.677), ACTH (18-39, m/z 2465.199), ACTH (1-17, m/z 2903.087), and ACTH (7-38, m/z 3657.923). All the MS spectra were obtained from accumulation of 900 shots. MS/MS spectra were acquired for the 15 most abundant precursor ions, with a total accumulation of 1500 laser shots and collision energy of 1 kV. Once the MS survey scans were completed, the data were processed to generate a list of precursor ions for interrogation by MS/MS. The combined MS and MS/ MS peak lists were searched using the GPS TM Explorer software version 3.6 (AB Sciex). Protein identification was performed by MS/MS ion search using MASCOT version 2.1 (http://www. martixscience.com) search engine against the Swiss-Prot database. Searches were carried out with the following parameters; all entries taxonomy, trypsin digestion with one missed cleavage, fixed modifications: carbamidomethylation of cysteine residues, variable modifications: oxidation of methionine residues, mass tolerance 150 ppm for MS and 0.4 Da for MS/MS. Identified proteins having at least two unique matched peptides were selected for further analysis. We have reported only those proteins with a protein identification confidence interval of $95%. Quantitative immunological measurement of two of the differentially expressed proteins identified in this study; haptoglobin and apolipoprotein A-I, in serum samples of healthy controls (n = 20), falciparum malaria (n = 20) and leptospirosis patients (n = 6) were performed using COBAS INTEGRA 400 PLUS system (ROCHE). The serum concentration of those two target proteins in vivax malaria was taken in account for a comparative analysis from our previous report [15] . Crude individual serum samples were subjected directly to immunoturbidimetric quantification using the Tina-quant ver.2 kits (Roche Diagnostics) according to the manufacturer's instructions. Samples and controls were automatically prediluted 1:21 with NaCl solution by the instrument. In this immunological assay the target proteins form precipitates with the specific antiserum which are determined turbidimetrically at 340 nm. Anti-human haptoglobin (rabbit) and Apo A-I (sheep) antibodies were applied for the immunoturbidimetric quantification of haptoglobin and Apo A-I respectively. The instrument was monitored at absorbance measuring mode where the absorbance increase was directly proportional to the concentration of the target proteins. Quantification of another interesting target; retinol-binding protein (RBP) was performed using ELISA. Concentrations of RBP4 in serum samples of HC (n = 20), FC (n = 6), FM (n = 12) and VM (n = 12) patients were measured using AssayMax Human Retinol-Binding Protein-4 (RBP4) ELISA kit (Cat# ER3005-1) from AssayPro (USA) following the manufacturer's instructions. Briefly, quantitative sandwich enzyme assay was employed where RBP4 standard and serum samples (HC, FC, FM and VM) at a dilution of 1:100 were subjected to a microplate pre-coated with a polyclonal antibody specific for RBP4. Samples were sandwiched by the immobilized antibody and biotinylated polyclonal antibody specific for RBP4, which was recognized by a streptavidinperoxidase complex. Color development was performed through the addition of a peroxidase enzyme substrate and optical densities were measured at 450 nm and 570 nm using a SpectraMax M2 e (Molecular Devices, USA). Prior to the western blotting experiment protein concentration in each sample [malaria patients (n = 24), FC (n = 6) and HC (n = 12)] was accurately estimated using the 2D-Quant kit (GE Healthcare) and BCA Protein Assay (Thermo Fisher Scientific). Western blot analysis was performed as described previously [40] . Briefly, serum proteins were separated by 12% SDS-PAGE (50 mg per track) and then transferred onto PVDF membranes under semidry conditions by using ECL semi-dry transfer unit (GE Healthcare). Western blot was performed by using monoclonal/ polyclonal antibody against serum amyloid A (Santacruz Biotechnology, sc-20651), haptoglobin (Santacruz Biotechnology, sc-71207), clusterin (Santacruz biotechnology, sc-8354) and retinolbinding protein (RBP) (Santacruz Biotechnology, sc-69795) and appropriate secondary antibody conjugated with HRP (GeNei (MERCK)-621140380011730 or 621140680011730). Candidate proteins for validation were selected on the basis of fold changes, possible association of the proteins with malaria pathobiology and accessibility of the required antibodies. ImageQuant software; version 5.0 (GE Healthcare) was applied for quantitation of signal intensity of the bands in western blots. Differentially expressed serum proteins in FM were subjected to functional pathway analysis using IPAversion 9.0 (IngenuityH Systems, www.ingenuity.com) to determine association of the identified proteins with various physiological pathways. The significance of association between our dataset and identified networks/pathways was considered on basis of two parameters, ratio and p-values. Differentially expressed proteins in FM patients were also analyzed using PANTHER system; version 7 (http:// www. pantherdb.org) [41] and DAVID database version 6.7 (http://david.abcc.ncifcrf.gov/home.jsp) [42, 43] . The list of Uni-Prot Accession from each dataset was uploaded in tab delimited text format at once, which was mapped against the reference Homo sapiens dataset to extract and summarize functional annotations associated with individual or group of genes and proteins. The gene ontology (GO) categories for 30 proteins was assigned using GeneSpring software package (version 11.5; Agilent Technologies, Santa Clara, USA). Since, GO vocabulary is organized in a hierarchical fashion, the second level of GO terms were presented as a balance between GO term for specificity and maximal coverage. GO terms that were enriched in two or more proteins were considered. In addition, a significance p-value of the enrichment was computed using the hypergeometric probability distribution, which identifies GO categories represented by the 30 proteins relative to their representation on the Biological Genome for Human created using information available at NCBI (ftp://ftp. ncbi.nlm.nih.gov/gene/DATA). To determine the biological pathways with significant enrichment of the input proteins, algorithms in GeneSpring software package performs a standard hyper-geometric calculation to obtain p-value, which signifies the enrichment. Prior to analysis, manually curated biological pathways from Reactome, Biocarta, NCI and PathwayCommons (in Biopax level 2 format) were populated in GeneSpring's database. Pathways with significance pvalue (p,0.05) were chosen for subsequent analysis and interpretation. We used GeneSpring's pathway database to create Shortest Connect network from the selected pathways. The Expand Selection algorithm was performed on the above network to include first and second degree neighbors. Expand Selection uses the GeneSpring pathway database for finding expansion on entities and takes a series of expansions to connect processes/ functions/other biomolecules to the given entities (proteins). This algorithm allows listing all processes and functions in which the given entities participate. We applied proteomics data obtained from 2DE and 2D-DIGE analyses to discriminate among FM, VM, FC and HC groups using multivariate statistical analysis. 5 proteins (haptoglobin, apolipoprotein A-I, hemopexin, apolipoprotein E and serum amyloid A) identified by 2DE (Table S8 .1A) and 7 proteins (haptoglobin, apolipoprotein A-I, hemopexin, apolipoprotein E serum amyloid A, serum amyloid P and serum paraoxonase/ arylesterase 1) identified by 2D-DIGE (Table S8 .2A) were used to develop statistical classifier designed to categorize and predict clinical phenotypes (i.e., FM, VM and HC). Discrimination of malaria (FM and VM) from FC (leptospirosis) and statistical sample class prediction was performed on the basis of differential expression levels of 6 candidate proteins (serum amyloid A, hemopexin, apolipoprotein E, haptoglobin, retinol-binding protein and apolipoprotein A-I) (Table S8 .3A). Selection of the candidate proteins was executed on the basis of their level of differential expression and ability to discriminate between the clinical phenotypes. For multivariate statistical analysis and machine learning, the data were mean centered; scaled and logarithmic transformation was performed in order to lower relatively large differences among the respective spot abundances. 2DE data was additionally normalized using Quantile method to correct for batch difference. 3 levels of validation were used to establish the reliability of identified differentially expressed proteins to detect correct phenotypic classes using Mass Profiler Professional (MPP). We used PLS-DA, SVM, Decision Trees and Naïve Bayes implemented in MPP software package (version 2.2, Agilent Technologies, Santa Clara, USA) for all multivariate and machine learning analysis in this study. Partial least squares is a regression method using the information contained in X data matrix (predictor variables) to predict the behavior of Y data matrix (response variables). PLS method models both X and Y variables simultaneously to find the latent variables in X that will predict the latent variables in Y [44] . The application of PLS as a classification method is indicated as PLS-DA [45, 46] . SVM separates two classes by generating the hyperplane (in a highdimensional feature space) which maximizes the distance from the hyperplane to the closest training examples [47] . In Decision Trees a sample gets classified by following the appropriate path down the decision tree. The Naive Bayesian model is built based on the probability distribution function of the training data along each feature. Since Decision Trees and Naïve Bayes directly handle multi-class problems, we have used the default parameters for these techniques. The SVMs are trained using sequential minimal optimization with a linear kernel. Validation of the obtained predictive models was performed using a standard K-fold cross-validation procedure: observations in input data were randomly divided into three equal parts, two parts were used for model training, and the remaining samples were classified using the constructed model. The whole process was repeated for 10 times. Efficiency of 3 classifier proteins; haptoglobin, apolipoprotein A-I and retinol-binding protein for prediction of malaria (FM and VM) and leptospirosis (febrile control) was analyzed using receiver operating characteristic (ROC) curves [plot of true positives (sensitivity) vs false positives (1-specificity) for each possible cutoff] using GraphPad Prism software package (version 5.02). ROC curve analysis was performed for only those 3 classifier proteins (out of 6) for which absolute serum concentration values (immunoturbidimetric assay/ELISA) were measured. Sensitivity and specificity values for the marker proteins were calculated at different threshold points. Two-sided p-values less than 0.05 were considered statistically significant. Figure S1 Evaluation of the depletion efficiency for albumin and IgG from human serum. Two major highabundance serum proteins; albumin and IgG were removed using Albumin & IgG Depletion SpinTrap (GE Healthcare) to reduce the dynamic range of serum protein concentration. (A) Levels of albumin and IgG in CBB stained 2D gel before and after depletion. 600 mg of total serum proteins were focused on linear pH 4-7 IPG strips (18 cm) and then separated on 12.5% polyacrylamide gels. Depletion of the top two high-abundance proteins (albumin and IgG) introduced nearly two-fold increase in overall spot number in 2D gels. (B) Levels of albumin and IgG in CBB stained 1D-SDS-PAGE gel before and after depletion showing the efficiency of the depletion process. 10 mg of total crude [C] and depleted [D] serum proteins were loaded onto each lane and separated on 10% polyacrylamide gels. (C) Densitometric analysis of the 1D-SDS-PAGE gels revealed around 85% and 80% depletion of albumin and IgG respectively. (PDF) Figure S2 Trends of differentially expressed proteins in falciparum malaria patients visualized in 2DE gels. (A) Representative 2D gels of serum from healthy controls and FM patients. 600 mg of total serum proteins were focused on linear pH 4-7 IPG strips (18 cm) and then separated on 12.5% polyacrylamide gels, which were stained with Gel Code Blue Stain. Protein spots exhibiting significantly altered expression levels are marked on the gels. Down (B) and up (C) -regulation of protein expression levels in FM patients. The 3D images of statistically significant (p,0.05) differentially expressed spots were analyzed using IMP7 software. Data is represented as mean 6 SEM (where n = 20). (PDF) Figure S6 IPA defined interaction networks associated with the differentially expressed proteins in falciparum malaria. Differentially expressed serum proteins identified in FM patients were entered as focus molecules in the analytical software to generate biological processes, pathways and molecular networks associated with the identified proteins. (A) The top-scoring network (score 35); cell signaling, molecular transport, vitamin and mineral metabolism. This network incorporated 14 out of the 27 differentially expressed proteins (focus molecules), (B) The second net-work; lipid metabolism, molecular transport, small molecule biochemistry (score 23). This network incorporated 10 focus molecules. Green and red symbols represent proteins that were down and up-regulated in falciparum malaria, respectively (identified in this study). White symbols represent associated proteins identified in the functional analysis for which the difference in expression level did not achieve statistical significance in our study. (PDF) Figure S7 Gene Ontology (GO) terms for molecular functions, cellular components and biological processes associated with the differentially expressed serum proteins identified in falciparum malaria. A total of 1394 Gene Ontology (GO) terms were identified, of which the distribution of second level of GO terms that were enriched in two or more proteins is shown as molecular functions (A) and cellular components (B) and biological processes (C). (PDF) Figure S8 Biological process regulated by differentially expressed serum proteins identified in falciparum and vivax malaria patients. Regulations were based on Natural Language Processing performed on MEDLINE abstracts as available in GeneSpring software package (version 11.5, Agilent Technologies). Identified process (A) common in both the plasmodial infections (B) specific for P. falciparum (C) specific for P. vivax infection. Red triangles and blue squares represent positive and negative regulations, respectively. (PDF) Figure S9 Discrimination of falciparum and vivax malaria from healthy controls on the basis of differential expressions of selected serum proteins. PLS-DA scores plot for (A) FM (red spheres, n = 10) and HC (green spheres, n = 10) samples, based on 5 differentially expressed proteins (Table S8 .1A) identified using 2DE, (B) FM (red spheres, n = 6), VM (blue spheres, n = 5) and HC (green spheres, n = 5) samples based on 7 differentially expressed proteins (Table S8.

Novel Method for Isolation of Murine Clara Cell Secretory Protein-Expressing Cells with Traces of Stemness

Clara cells are non-ciliated, secretory bronchiolar epithelial cells that serve to detoxify harmful inhaled substances. Clara cells also function as stem/progenitor cells for repair in the bronchioles. Clara cell secretory protein (CCSP) is specifically expressed in pulmonary Clara cells and is widely used as a Clara cell marker. In addition CCSP promoter is commonly used to direct gene expression into the lung in transgenic models. The discovery of CCSP immunoreactivity in plasma membranes of airway lining cells prompted us to explore the possibility of enriching Clara cells by flow cytometry. We established a novel and simple method for the isolation of CCSP-expressing cell Clara cells using a combination of mechanical and enzymatic dissociation followed by flow cytometry sorting technology. We showed that ∼25% of dissociated cells from whole lung expressed CCSP. In the resulting preparation, up to 98% of cells expressed CCSP. Notably, we found that several common stem cell markers including CD44, CD133, Sca-1 and Sox2 were expressed in CCSP(+) cells. Moreover, CCSP(+) cells were able to form spheroid colonies in vitro with 0.97‰ efficiency. Parallel studies in vivo confirmed that a small population of CCSP(−)expressing cells in mouse airways also demonstrates stem cell-like properties such as label retention and harboring rare bronchioalveolar stem cells (BASCs) in terminal bronchioles (TBs). We conclude that CCSP(+) cells exhibit a number of stem cell-like features including stem cell marker expression, bronchosphere colony formation and self-renewal ability. Clara cell isolation by flow cytometry sorting is a useful method for investigating the function of primary Clara cells in stem cell research and mouse models.

Human lungs are composed of three functional and morphological compartments: proximal and distal airways and the alveolar compartment. Proximal airways are lined by a pseudostratified epithelium with a number of cell types with important protective functions such as ciliated cells, goblet cells, and basal cells. More distally, the lining is a simplified columnar epithelium largely made up of non-ciliated secretory cells called Clara cells, and a few ciliated and basal cells. [1, 2] . Further down, the respiratory bronchioles are lined by cuboidal epithelium comprised entirely of ciliated and Clara cells, whereas, the epithelium of the alveolar compartment is comprised of type I and type II cells. In mouse, the pseudostratified epithelium is limited to trachea and extrapulmonary main bronchi while Clara cells make up over 80% of the epithelium, with few interspersed ciliated cells, that line intrapulmonary conducting airways [3] . These features make mouse an excellent tool for studying the functions of Clara cells. Clara cells have several protective properties. They detoxify xenobiotics and oxidant gasses, control inflammation, participate in mucociliary clearance of environmental agents, and proliferate/ differentiate to maintain the ciliated and non-ciliated cell population. Clara cells are a source of cytochrome P450 enzymes that contribute to the metabolism of a variety of substances [4] . In addition to the major Clara cell secretory protein (CCSP), also known as CC10, CC16, Clara cell antigen, secretoglobin 1A1 (SCGB1A1) or uteroglobin, Clara cells also contribute surfactant apoproteins A, B and D, proteases, anti-microbial peptides, several cytokines and chemokines, and mucins in the extracellular fluid lining airspaces. CCSP is the most abundant secretory protein found in the airway surface fluid, expressed exclusively in nonciliated Clara cells and widely used as a marker of the cells [5, 6, 7, 8] .Changes in CCSP levels have a profound impact on not only the composition of airway surface fluid but also the airway epithelial response to environmental stimuli [9, 10] . Another important property of Clara cells is their ability to serve as progenitors for airway lining cells in response to injury. Moreover, subpopulations of CCSP-expressing cells may function as true stem cells of adult airways. Presently it is not known whether the groups overlap or represent distinct cells such as variant Clara cells [11] , type A cells [12] , OCT4-expressing stem cells [13] and bronchioalveolar stem cells (BASCs) [14] . Due to the lack of simple methods for the isolation of primary Clara cells from the lung, the majority of studies have been carried out in vivo or using lung cancer cells for in vitro tests. The major disadvantage of such approaches is the difficulty in performing mechanistic studies in non-neoplastic primary cells. Recently, Wong et al. developed a method for isolating CCSP + cells from bone marrow by flow cytometry sorting [15] . We speculated that this method may also be used to isolate CCSP + (Clara) cells from the lung. In this study we established a simple method for the isolation of CCSP + cells from mouse lung and applied several different means to identify stem cell-like characteristics of CCSP + cell in vitro and in vivo. We propose that this new procedure method for CCSP + cell isolation provides a useful instrument for Clara cell research, for instance in the field of stem cell biology. Mice FVB mice were purchased from the Frederick National Lab, Maryland. Mice were housed under specific pathogen-free conditions under a 12-h light/dark cycle with access to food and water ad libitum. All the procedures used in this study were approved by the NIH Animal Care and Use Committee. The heart, lungs and trachea were removed en bloc from mice following euthanasia by carbon dioxide inhalation. Lungs were separated and lobes minced on ice and incubated with collagenase type I (Invitrogen, Grand Island, NY) at 3 mg/ml in PBS in a volume of 2 ml per lung for 1 hour at 37uC with continuous agitation in an incubator. The suspension was further disaggregated by trituration through a 19 gauge needle (Sherwood Medical Co, St. Louis, MO), diluted in PBS. The crude cell suspension was filtered through a 40 mm cell strainer (BD Biosciences, Sparks, MD) and centrifuged at 700 rpm for 5 min. After discarding supernatant, cells were resuspended in 2 ml of red blood cell lysis buffer (eBioscience, San Diego, CA) for 4 min. Neutralization was performed with 10 ml of Dulbecco's Modified Eagle Medium (DMEM) (Invitrogen) with 10% FBS (Invitrogen) and cells were centrifuged at 700 rpm for 5 min. Cells were resuspended in DMEM/10% FBS with 20 ng/ml gentamycin/0.5 ng/ml amphotericin B (Cascade Biologicas tm , Portland, Oregon), plated in 100 mm dishes and placed to recover in an incubator at 37uC and 5% CO 2 for 18 hours ( Figure 1A ). Recovered cells were trypsinized in 0.05% Trypsin-EDTA (Invitrogen) and resuspended at a concentration of 1610 7 cells in 100 ml PBS with 3% FBS. Two microliters of the rabbit anti-CCSP antibody (Millipore, Billerica, MA) was added, followed by a 30 min incubation on ice. Cells were washed twice in PBS with 3% FBS, then 2 mL goat anti-rabbit-FITC secondary antibody was added and incubated on ice for 30 min. After two washes in PBS with 3% FBS, cells were resuspended in the same but fresh media. Rabbit IgG staining was used as an isotype-matched negative control and CCSP staining with permeabilization of dissociated cells was used as a positive control. CCSP positive (CCSP + ) and negative (CCSP 2 ) fractions were obtained by fluorescence-activated cell sorting (FACS) using Vantage SEH cell sorter (BD, Bedford, MA). They were examined by immunfluorescence, adherent or 2D and sphere cell (3D) cultures and qRT-PCR. Single and dual labeling of cells and tissue sections by immunofluorescence (IF) or immunohistochemistry (IHC) was performed according to previously described methods [16] [17] . The primary antibodies were: goat polyclonal anti-CCSP(T18) (1:50, Santa Cruz Biotech, Santa Cruz, CA), rabbit antiuteroglobin-related protein 1 (UGRP1) (1:100, a kind gift from Dr. Shioko Kimura, NCI/NIH, Bethesda, MD), rabbit anti-pan-cadherin (, 1:50, Abcam, Cambridge, MA), mouse anti-b-catenin (BD). rabbit anti-pan-cytokeratin (1:100, Dako, Carpinteria, CA), rabbit anti-pro-SPC (1:200, Millipore), rat anti-BrdU (1:100, Accurate Chemical & Scientific Corp, Westbury, NY), rabbit anti-sox2(1:2000, Seven Hill, Cincinnati, OH) and rabbit anti-ALDH1(1:500. Abcam). Approximately 1610 5 FACS sorted cells, were washed twice in PBS with 1% FBS and resuspended in 30 ml of Cell Adherence Solution (Crystalgen, Commack, NY). After standing for 2 minutes, 3 ml of the cell mixture was mounted on glass slides, dried for 2 minutes and fixed with 4% paraformaldehyde in PBS for 15 minutes. Both fixed cells and tissue sections were blocked with 1% goat or rabbit normal serum for 1 hour. The blocking solution was removed and 75 ml of primary antibody was added to cells. After 1 hour of incubation at room temperature, slides were washed in PBS three times. Secondary antibodies were added. For dual-labeling IF, additional primary antibodies were added after the third PBS wash, followed by incubation with secondary antibodies conjugated with Alexa fluor 488 or 594 (Invitrogen). All incubations were performed at room temperature and slides were washed in PBS (365 min) between each step and mounted with an anti-fading reagent with 49,6-diamidino-2-phenylindole (DAPI) (Invitrogen). Control slides were included in each analysis in which non-immune serum was substituted for primary antibodies and secondary antibodies individually. All IF images were taken with a Zeiss LSM 510 Meta Mk4 Confocal Microscope (Zeiss, Thornwood, New York). For IHC, signals were developed using 3,39diaminobenzidine (DAB). Total RNA from sorted cells was isolated using an RNeasy minikit (Qiagen, Valencia, CA) by following the manufacturer's protocol. One microgram of RNA was reverse transcribed in a total volume of 20 ml using the QuantiTect RT kit (Qiagen). PCR was performed in triplicate in a MyiQ single color real time PCR detection system (Bio-Rad, Hercules, CA) using SYBR Green PCR kit (Qiagen) according to the manufacturer's protocol. Amplification was confirmed by ethidium bromide staining of the PCR products on an agarose gel. The expression of each target gene was normalized to the expression of 18 S RNA and presented as the ratio of the target gene to 18 S RNA, expressed as 2 2DCt , where Ct is the threshold cycle and DCt = Ct Target For continuous labeling in vivo, BrdU (50 mg/ml) was administered to mice throughout a 7-day period via a subcutaneous miniosmotic pump (Alzet model 2001, Durect Corporation, Cupertino, CA). Alzet pumps were implanted in mice and removed after one week. Mice were sacrificed 4 weeks after removal of the Alzet pumps. Lungs were fixed overnight via tracheal instillation of fresh 4% paraformaldehyde and embedded in paraffin prior to sectioning. Label-retaining cells were identified by BrdU immunofluorescence. BrdU and CCSP double staining was performed and cells exhibiting a nucleus and attachment to basement membrane were counted. Bronchiolioles (BLs) were defined as intrapulmonary airways in which smooth muscle, but neither cartilage nor glands, could be seen. Terminal bronchioles (TBs) contained an intact bronchioalveolar duct junction (BADJ) and visible alveolar duct [17] . In TBs quantification of staining included all cells within 200 mm of the BADJ. A total 49 TBs and 26 BL structures were analyzed in lung sections of five mice. Figure 1 . Schema for purification of primary CCSP positive cells from mouse lung. A) Two month old FVB mice were euthanized by CO 2, lungs removed and lobes collected. After washing in PBS, lobes were minced on ice and incubated in a small cell culture dish with 3 mg/ml collagenase in PBS (total 5 ml) in a shaking platform for 1 hour at 37uC. The suspension was further disaggregated by trituration through a 19 gauge needle, with 5 ml of PBS, filtered through a 40 mm cell strainer and centrifuged at 1000 rpm for 5 min. The supernatant was discarded, cells resuspended in red blood cell lysis buffer for 4 min re-plated into 10 cm culture dishes for recovery overnight (18 hrs). Surviving cells were adhering to the dish. After trypsinization and neutralization by 10% FBS media, cells were resuspended in PBS with 3% FBS, and stained with rabbit anti-CCSP antibody and FITC conjugated anti-rabbit secondary antibody. CCSP + and CCSP 2 cells were sorted with FACS Vantage SE cell sorter. B) Rabbit IgG was used as an isotype -matched negative control; CCSP + population sorted with FACS Vantage SE cell sorter from dissociated lung tissue was 25.37%. doi:10.1371/journal.pone.0043008.g001 It is well established that CCSP is expressed in non-ciliated Clara cells in the airways. CCSP which is widely used as a Clara cell marker is a cytoplasmic secretory protein [3] . Figure 2A revealed intense immunorectivity along the lining of mouse TB. Recently, Wong AP et al. was able to isolate CCSP + cells from bone marrow using flow cytometry [15] . Therefore, we postulated that CCSP may be expressed not only in the cytoplasm, but also in the cell membrane of Clara cells. To obtain evidence for this, we used the well-known cell membrane marker pan-Cadherin [18] . Indeed, in high magnification photomicrographs we were able to demonstrate co-expression with pan-Cadherin in the cell membrane using confocal microscope ( Figure 2B) . These data suggested a possibility for isolating living Clara cells by flow cytometry sorting and lead us to develop the protocol outlined in this study. To test the possibility of CCSP + cell isolation by flow cytometry from mouse lung, we established a simple method to make single cell suspensions from lung tissues. Using a combination of mincing by scissors and incubation in a high concentration of collagenase (3 mg/ml) for digestion, single cells were obtained within 2 hours from euthanasia. After an overnight recovery in DMEM/10%FBS cell culture media, cells adherent to culture dishes were trypsinized and sorted using a flow cytometry sorter. In FVB mice, about 25% of the lung cells sorted from one whole lung single cell suspension were CCSP + (Figure 1 ). Typical yields of sorted cells per mouse were about ,2.5610 5 of CCSP + cells and ,4610 5 of CCSP 2 cells. Sorted cells were plated into 100 mm cell culture dish for overnight in an incubator prior to further studies. Unattached dead cells were removed with the media. Sorted cells were used for RNA isolation and RT-PCR following an overnight recovery. CCSP mRNA expression was detected in CCSP + cell fraction, but not in CCSP 2 cells ( Figure 3A) . We also mounted cells on slides using Cell Adherence Media for immunofluorescence (IF). We found that 98% (225/ 230) of the cells in CCSP + sorted fraction were positive for CCSP IF. We also found that 97% of the cells in the CCSP + fraction revealed the expression of another Clara cell maker UGRP1 by IF. All of the cells in CCSP + sorted fraction expressed pan-keratin. Rare CCSP + sorted cells revealed the presence of pro-SPC ( Figure 3) . These data demonstrated that sorting by flow cytometry is a useful and simple method for harvesting purified CCSPcontaining Clara cells. One of the features of stem/progenitor cells is the expression of stem cell markers. Therefore we performed qRT-PCR for several common stem cell markers including CD44, CD133, Sca-1 and SOX2 in the sorted cells. Interestingly, CD44 was expressed in both CCSP + and CCSP 2 populations at similar levels. In contrast, CD133, Sca-1 and Sox2 demonstrated much lower but detectable levels in CCSP + cells than the levels in CCSP 2 cells. These data indicate that CCSP + cells express stem cell markers, although at low levels ( Figure 4 ). Spheroid culture is a common method to detect stem cell features in vitro [19] . Spheroid colony formation was tested by serial dilution technique. Both CCSP + and CCSP 2 cellular fractions were able to form sphere clones. However, following 10 days of culture ( Figure 5 ) CCSPcells demonstrated a larger colony size and higher efficiency of colony formation than CCSP + cells. Dissociation of spheroid colonies into single cells resulted in reformation of the spheroid colonies, indicating that this phenotype was stable (data not shown). Quiescent or slow-cycling stem cells in adult tissues can retain BrdU over long periods by either segregating chromosomes asymmetrically or dividing slowly. Label-retaining cells can be used to identify populations that contain stem cells [20] . In fact, many such studies have been used to determine putative stem cell locations in mammalian tissues [21, 22] . Using CCSP and BrdU double staining by IF, we found that 1.59% (39/2450) of cells in TBs and only 0.39% (12/4138) of them in BLs were BrdU + / CCSP + (Figure 6 ). The results suggest that the majority of mouse airway CCSP + stem/progenitor cells may reside in TBs. A subpopulation of CCSP + /pro-SPC + cells known as bronchioalveolar stem cells (BASCs) are capable of differentiating into Clara cells and alveolar type II cells and are considered to be adult lung stem cells [14] . In the current study, a rare portion of sorted CCSP + cells were also found to express the type II cell marker pro-SPC ( Figure 3D ). In order to confirm the existence of BASCs in vivo, we performed CCSP/pro-SPC double staining by IF in mouse lungs. Our results showed that 1.1% of TB epithelial cells contained BASCs while no CCSP + /pro-SPC + double positive epitheliums were detected in BLs ( Figure 6 ). In addition, a number of stem cell markers such as CD44, Sox2 and ALDH1 were detected by IF or IHC along the TB epithelium ( Figure 6C-E) . We also found that CD133, CD44, Sca-1 and Sox2 mRNAs were expressed at variable levels in mouse lung tissues ( Figure 6F, 6G) . This provides further evidence for the progenitor role that Clara cells may have in the mouse lung. In this study, we isolated and characterized significantly purified CCSP-expressing cell populations from mouse lung by using high concentrations of collagenase and a flow cytometric sorting method. In addition, we showed that CCSP + cells expressed stem cell markers and form three dimensional spheroid colonies in culture. Furthermore, we confirmed that CCSP + cells may also express stemness characteristic in vivo as evidenced by label retention, the presence of CCSP/pro-SPC double positive BASCs and expression of stem cell markers in the epithelial lining of TBs of mice. Accordingly, the novel method described herein is a significant step in the progress of isolating and characterizing highly purified Clara cells in primary cultures. Based on previous publications, the distribution of CCSP expression in non-ciliated Clara cells is described as cytoplasmic [23, 24, 25] . We made the surprising and novel discovery of CCSP immunoreactivity along cellular membranes of bronchiolar Clara cells. Using pan-cadherin as a cell membrane marker in normal airway epithelium [18, 26, 27] we found CCSP was expressed not only in the cytoplasm, but also in the membrane. These findings gave rise to the possibility that living Clara cells can be isolated by flow cytometry using fluorescing tags. Our successful CCSP + cell sorting further confirmed the distribution of CCSP membranous expression. One explanation is that bronchiolar Clara cells secrete such large quantities of CCSP that part of it remains stuck to the outer surfaces of cell membranes, allowing sorting of CCSPcontaining cells from suspension. Clara cell isolation from rabbit was first reported in the early 1980s by Devereux et al. [28] . After that, several groups were able to isolate pulmonary Clara cells from mouse [29, 30, 31, 32, 33] . The studies have been instrumental in establishing the many functions of Clara cells. However, the majority of the methods are quite complex and rely on protease digestion followed by centrifugal elutriation and/or Percoll density gradient centrifugation. Only one group used FACS for Clara cell isolation from rat based on the reaction of their glutathione content with monochlorobimine to a fluorescent product [34] . The techniques typically resulted in a Clara cell enrichment of 55,90%. A reproducible source of considerably purified Clara cells is necessary for airway stem/ progenitor cell research. Using high concentrations of collagenase for lung tissue digestion followed by flow cytometry sorting, we were able to achieve 98% pure CCSP + (Clara) cell population, providing a very useful and reliable method for Clara cell function and stem cell research. A notable application will be to directly address molecular mechanisms of genes that have been expressed in Clara cells by using CCSP as a lung specific promoter in transgenic mice. To further characterize sorted CCSP + cells, we evaluated the expression of pan-keratin protein in CCSP + cells. All cells expressed pan-keratin indicating that all the CCSP + cells were epithelial. We also found that a few cells expressed pro-SPC. This suggests that CCSP + cells contain rare populations of BASCs (CCSP/SPC double positive cells). In this study, the expression of well documented stem cell markers such as CD44 [35] , CD133 [36, 37] , Sca1 [14, 38] and Sox2 [39, 40] was detectable by qRT-PCR in CCSP + cells. However, the level of CD133, Sca-1 and Sox2 expression was lower in CCSP + cells than that in CCSP 2 cells. One possible explanation is that CCSP is a Clara cell differentiation marker, so a CCSP + population of cells may contain more mature Clara cells, but few stem/progenitor cells, while CCSP 2 cells fraction is a mixture of many cells, such as type I, type II, ciliated cells, basal cells, smooth muscle cells and fibroblast cells and so on. Many of the cells have been shown to have stem cell features [19, 41] . Sphere culture showed that CCSP + cells were able to form spheroid colonies. The sphere colony size and efficiency of colony formation were lower in CCSP + cells compared to CCSP 2 cells. This data further suggests that CCSP + cells do have stem cell features, but stem cell activities are lower than in CCSP 2 cells. Using tissue sections, we found that 1.59% of CCSP positive cells in TBs were label-retaining cells and 1.1% were CCSP/SPC double positive BASCs. These data provide in vivo validation for the in vitro results that the CCSP + cell population contains a small subset of stem cells in the airway. In summary, we discovered that CCSP was not only expressed in the cytoplasm but there was also marked immunoreactivity along in the cell membranes of airway Clara cells. This provided the basis of flow cytometry sorting technology for the isolation of CCSP expressing Clara cells from murine lung. We also found that in vitro, CCSP + cells demonstrated stem cell-like features including stem cell marker expression, bronchosphere colony formation and self-renewal ability. Moreover, a subset of labelretaining cells and BASCs were detectable in the CCSP + population in vivo located in the TBs We conclude that Clara cell isolation by FACS is a useful method for investigating Clara cell function and overall pulmonary stem cell research biology.

Development of Real-Time PCR Array for Simultaneous Detection of Eight Human Blood-Borne Viral Pathogens

BACKGROUND: Real-time PCR array for rapid detection of multiple viral pathogens should be highly useful in cases where the sample volume and the time of testing are limited, i.e. in the eligibility testing of tissue and organ donors. FINDINGS: We developed a real-time PCR array capable of simultaneously detecting eight human viral pathogens: human immunodeficiency virus types 1 and 2 (HIV-1 and -2), hepatitis B virus (HBV), hepatitis C virus (HCV), human T-cell leukemia virus-1 and -2 (HTLV-1 and -2), vaccinia virus (VACV) and West Nile virus (WNV). One hundred twenty (120) primers were designed using a combination of bioinformatics approaches, and, after experimental testing, 24 primer sets targeting eight viral pathogens were selected to set up the array with SYBR Green chemistry. The specificity and sensitivity of the virus-specific primer sets selected for the array were evaluated using analytical panels with known amounts of viruses spiked into human plasma. The array detected: 10 genome equivalents (geq)/ml of HIV-2 and HCV, 50 geq of HIV-1 (subtype B), HBV (genotype A) and WNV. It detected 100–1,000 geq/ml of plasma of HIV-1 subtypes (A – G), group N and CRF (AE and AG) isolates. Further evaluation with a panel consisting of 28 HIV-1 and HIV-2 clinical isolates revealed no cross-reactivity of HIV-1 or HIV-2 specific primers with another type of HIV. All 28 viral isolates were identified with specific primer sets targeting the most conserved genome areas. The PCR array correctly identified viral infections in a panel of 17 previously quantified clinical plasma samples positive for HIV-1, HCV or HBV at as low as several geq per PCR reaction. CONCLUSIONS: The viral array described here demonstrated adequate performance in the testing of donors’ clinical samples. Further improvement in its sensitivity for the broad spectrum of HIV-1 subtypes is under development.

Rapid progress and improvement in molecular technologies have allowed researchers to switch from the traditional approaches of virus detection in clinical samples to multiplexing for simultaneous detection of multiple pathogens in a single assay. A number of different PCR based assays for detection and discovery of multiple pathogens have been developed [1] [2] [3] [4] . Detection microarrays are proven to be useful in the identification and discovery of viruses homologous to known species. They have been used to guide the selection of samples for further analysis by sequencing [1] [2] [3] 5] . However, microarrays based on nucleic acid hybridization are too complex in design and performance for the routine donors testing, and exhibit a comparatively low sensitivity of detection, usually around 10021,000 genome copies of target virus per analyzed sample [1] [2] . Several PCR based assays coupled with oligonucleotide microarray technology (so called resequencing arrays) have been designed to allow simultaneous detection or genotyping of a target group of viruses, such as some critical blood-borne pathogens (3 viruses) [6] , respiratory viruses (16-21 viruses) [7] [8] , and respiratory adenoviruses (6 different serotypes) [9] . Such PCR based approach allows increasing the sensitivity of detection down to 10-100 copies of the target RNA or DNA in a sample. PCR multiplexing should be highly useful when both the volume of the samples and the time of testing are critical, as in the donor eligibility (DE) testing for tissue or organ transplantation [10] . Current regulation requires that DE testing be performed using assays approved and licensed by the U.S. Food and Drug Administration (FDA). However, the automated assay systems that are designed to screen large numbers of samples, without the strict limitation of sample volumes, may not be completely suitable or ideal for the needs of DE testing for tissue or organ transplantation. The main goal of the study presented here was to evaluate the feasibility of developing a sensitive and specific assay for rapid detection and identification of a group of target viral pathogens. The following viral pathogens were included in our array: human immunodeficiency virus types 1 and 2 (HIV-1 and HIV-2), human T-cell leukemia virus-1 and -2 (HTLV-1 and HTLV-2), hepatitis C virus (HCV) and West Nile virus (WNV), all with singlestranded RNA genome; vaccinia virus (VACV) and hepatitis B virus (HBV), both with double-stranded DNA genome, HBV also has single-stranded RNA stage. Some of the listed viruses are included to the required DE testing for tissue transplantation. Besides, historically, some of the targeted viruses have been found to be allograft-transmitted to recipients [11] [12] . In the present study, a real-time PCR array with SYBR-Green chemistry targeting these eight viral pathogens listed above was developed and evaluated with analytical and clinical panels. The array demonstrated acceptable performance in the testing with both analytical panels and donors' clinical samples. The research study conducted at FDA using previously frozen blood samples was reviewed by Department of Health and Human Services, Food and Drug Administration, Research Involving Human Subjects Committee (RISHC Protocol #10-008B entitled ''Detection of Infectious Agents in Previously frozen blood Samples from Patients with Various Illnesses and Healthy Blood Donors''). The 17 clinical plasma samples positive for HBV, HCV or HIV-1 used in this study were existing clinical diagnostic samples kept in NIH Blood Bank. Information of these left over samples had been recorded in such a manner that subjects can not be identified, directly or through identifiers. The written informed consent from the participants was waived under 45 CFR 46.101 (b) (4) . The six plasma samples positive for HIV-1 were estimated to contain 50 to ,90,000 genome copy numbers per ml; six plasma samples positive for HCV were estimated to contain 780 to ,123,000 genome copy number/ml and five plasma samples positive for HBV were estimated to contain ,150 to 16,000 copy number/ml. The copy numbers of viruses in these samples were provided by the NIH Blood Bank. No information about the subtypes or genotypes of these viruses was available. The amount of each clinical sample was sufficient to be tested only once by the PCR array in the study. All positive PCR products obtained in the testing using the PCR array were confirmed for validity by sequencing in the Facility for Biotechnology Resources of FDA/ CBER. We used the ''Insignia'' program (http://insignia.cbcb.umd. edu/query.php), a bioinformatics on line tool developed in the Center for Bioinformatics and Computational Biology, University of Maryland [13] to choose a specific DNA or RNA ''signature'' (a sequence, with customized length and G/C content) for targeted viruses. Comparative sequence analysis of the complete genomes was performed using mVISTA (http://genome.lbl.gov/vista/ mvista/submit.shtml). Multiple nucleotide sequence alignments (NSAs) were then created to visualize the most conserved genome areas using MEGA4 (http://www.megasoftware.net). Specific criteria for the primers and amplicon selection for the SYBR Green based PCR array were: 1) the same range of annealing temperature (T) -57-60uC -for all primers, 2) high G/C content for primers, allowing higher specificity of annealing, and 3) an amplicon size in the range of 100-200 b.p. in order to have a high PCR amplification efficiency and to sufficiently distinguish the products from primer dimers based on melting T peak (Tm). All primers were checked for potential dimer formation using ''Primer Express'' software (version 3.0, Applied Biosystems). After design, all primers were again checked using the National Center for Biotechnology Information (NCBI) Basic Local Alignment Search Tool (BLAST) (http://blast.ncbi.nlm.nih.gov/Blast.cgi) to avoid any cross-reactivity with other species. Newly designed and previously published primer sets adapted for the final version of the real-time PCR array are listed in Table 1 . In addition, primers specific for the human beta-globin gene were included in the array as an internal control for the quality of DNA/RNA preparation as well as for estimation of viral copy number per host cell if needed (the last row of Table 1 ). Total cellular DNA of the chronically infected cell cultures (for HIV-1, HIV-2; HTLV-1, HTLV-2 and VACV), viral genome cDNA copy spiked into human DNA (for HCV and WNV), and DNA isolated from human plasma of the infected individual (for HBV) were used as the positive templates in the initial testing and are listed in the last column of Table S1 . DNA or RNA panels were created by cloning of specific synthetic templates for each virus into the pGEM-T-Easy vector (Promega) by TA cloning, following by in vitro transcription to obtain RNA standards for HCV and WNV. All created plasmids are listed in Table S1 . Nucleotide numbers in Table S1 refer to the location of the partial viral genome cloned into pGEM-T-Easy vector according to the following complete genome sequences available in GenBank: HTLV-1 -L03562.2, HTLV-2 -M10060.1, HIV-1 -K02083.1, HIV-2 -J03654.1, HBV -AF462041.1, HCV -AF271632.1, VACV -AY243312.1, WNV -HQ596519.1. To establish the real-time PCR standard curve the copy number was calculated for each plasmid carrying one copy of the specific viral gene. The size of each plasmid (X b.p.) was used to determine the molecular weight in Daltons (g/mol): W (g/mol) = X b.p. (330 Da 6 2). The copy number of the target viral gene (molecules/ml) was determined from the plasmid concentration (C DNA ) and the molecular weight of each plasmid molecule: Copy number = C DNA (ng/ml) 6 6.02 6 10 23 (Avogadro's number)/W. Knowing the number of plasmid molecules with the target viral gene in a ml, a series of dilutions was made to generate a PCR standard curve. The developed analytical standards were used to calculate the intra and inter-assay reproducibility of quantification for each virus-specific primer set. Mean C(t) values, standard deviation (SD) and coefficient of variation (CV) were calculated from the data obtained in three replicates of each standard dilution for the intraassay reproducibility, and in three real-time PCR assays consisted of three replicates each (nine total) for the inter-assay reproducibility. CV was calculated as SD/Mean C (t) * 100%. Preparation of Viral RNA/DNA for PCR Array Analysis 0.5-1 ml of human plasma was used for the total viral RNA/ DNA extraction using ''QIAamp Viral RNA Mini-Kit'' (Qiagen) and tRNA (SIGMA) was used as a carrier RNA during the preparation. After the final elution step RNA/DNA in 160 ml of buffer AVE was precipitated with 100% ethanol and 3 mM NaCl at 220uC overnight. An RNA/DNA pellet was washed with 70% ethanol, dissolved in 10 ml of DEPC-treated water and then immediately used for cDNA synthesis with SuperScript II RT (Invitrogen) and random hexamers (Invitrogen) in a total reaction volume of 20 ml. The volume of cDNA/DNA sample was then adjusted up to 30 ml with DEPC-treated water and the whole PCR was performed using ''Bio-Rad CFX96 Real-Time System'' with ''Power SYBR Green PCR master mix'' (Applied Biosystems). One reaction (25 ml total) contained: 12.5 ml of PCR master mix, 0.5 mM of each primer and 1.25 ml of DNA/cDNA template. In the single virus testing (sections 3 and 4 of the ''Results'') we used 2.5 ml of DNA/cDNA template from 20 ml of sample after cDNA synthesis. ''Universal'' PCR conditions for all primer sets included to the array were: 95uC for 8 min (one cycle), then 50 cycles of: denaturation at 95uC for 15 s, annealing and extension at 60uC for 1 min, followed by melting curve read from 65uC to 95uC with increment 0.2uC for 5 s. Real-time PCR data were downloaded in 96-well plate format from ''Bio-Rad CFX Manager 2.1'' to MS Excel and analyzed manually. Two types of samples served as the background control for determination of the C(t) cut off. The 1 st type of negative control was 50 ng of human cellular DNA. The 2 nd type of negative control was negative donors' plasma. Data were collected in separate experiments from 8 human cellular DNA controls and 3 negative donors' plasma. To standardized the C(t) cut off for all primers the threshold was set at the PCR machine default setting. Based on a false positive rate of less than 5% the following method [14] was used to estimate the C(t) cut off from the range of C(t) obtained with negative samples for all 24 primer sets in the array: 1. Calculate the margin of error of the confidence interval (CI), W = t* 6 SD/!n, where: n -number of obtained C(t) values, SD-standard deviation, df = n21, and t* (for 95% confidence) is a ''critical value of the T distribution'' [14] . 2. One side CI covers this range: M-W, where M is sample mean. The C(t) cut off calculated from the range (n = 50) of the 1 st type of negative control data was C(t)#41.03. The C(t) cut off calculated from the range (n = 50) of the 2 nd type of negative control data was C(t)#42.7. Even some overlap between the C(t) measurements of truly positive and truly negative samples was detected in PCR array data, the Tm parameter was used to define if the obtained PCR product is specific by comparison to the expected Tm peak range. The analytical sensitivity of each primer set was determined in the single virus testing using FDA/CBER panels (kindly provided by Dr. Stephen Kerby, FDA/CBER) consisting of various amounts of the viruses (0-1,000 genome copies/ml) spiked into the ''normal'' human plasma. Panels for the following viruses were used in testing: HIV-1 (three different panels), HIV-2, HBV (based on genotype A), HCV (genotype 1b), and WNV (based on strain HU2002). These panels have been used for testing of commercially licensed NAT assays and their development was described previously [15] . Three separate HIV-1 RNA panels were used for testing. The first consisted of various amounts of an HIV-1 group M subtype B isolate: 0, 5, 10, 25, 50, 100, and 500 copies/ml. The second consisted of 25 samples representing various concentrations of HIV-1 groups O and N, and group M subtypes A, C, D, E, F and G: 10, 100 and 1,000 copies/ml for each virus. The third panel consisted of HIV-1 circulating recombinant forms (CRFs) AE and AG: 100, 1,000, and 10,000 copies/ml for each CRF. NP24 CAACTTCATCCACGTTTCACC a -to simplify the process of evaluation we used our primer names with sequential numbers, however some of the primers have been designed previously with their original names in the articles listed in the last column of this [25] [26] . HIV-1 groups and subtypes are based on designations reported by the NIH ARRRP. Low passage virus stocks were prepared and median tissue culture infective dose (TCID 50 )/ml of virus containing supernatant determined in fresh human PBMCs isolated as previously described [27] . Infectious unit (IU) of the virus stocks in TCID 50 s were titrated for each isolate and the number of IU used for viral RNA isolation and PCR was calculated based on the dilution factor (1:100) and ranged from ,10 to 2,500 or 1 to 3.4 log 10 TCID 50 per PCR reaction. PCR approach based on SYBR Green chemistry, allowing simultaneous detection of multiple targets, was chosen to be applied for the array performance. Virus-specific primer sets targeting at least three different genomic sites for each viral pathogen were designed for the real-time PCR array. We used the ''Insignia'' program, a bioinformatics tool that helps to choose a specific DNA or RNA ''signature'' for different bacteria and viral pathogens that are included in the pre-built ''Insignia'' database [13] . Sequences of the highly conserved regions, such as coding viral polymerase or structural proteins were selected to design the candidate primers. In addition, we performed nucleotide sequence alignments (NSA) of the complete viral genomes and of the most conserved genome areas for different subtypes/genotypes or different isolates of all targeted viruses. Some previously published primers designed for PCR detection of the target viruses were also adapted and evaluated. Overall, a total of 120 primers were initially designed using specific criteria for the current real-time PCR array (see Materials and Methods) to cover the eight targeted viruses: HIV-1, HIV-2, HBV, HCV, HTLV-1, HTLV -2, WNV, and VACV. The primers were designed and tested for their effectiveness and specificity of amplifying the respective target under uniform PCR conditions (using the same annealing temperature for all primers). Each candidate primer pair was first tested for its specificity and sensitivity of PCR amplification. This was assessed in a real-time PCR cross-testing against human PBMC DNA (50 ng/PCR reaction) from a healthy donor, DNA from human cell cultures infected by various target viruses, or human DNA spiked with a known amount of genome copies of various viruses. It was important to ensure that the selected primers targeting a specific virus would not non-specifically amplify any DNA in a sample. The melting temperature (Tm) peak of the product amplified from the target virus should be clearly differentiable from the Tm peak of primer dimers or any non-specific products produced in PCR. DNA or RNA panels were created from the cloned synthetic templates (listed in Table S1 ) by spiking of 2-10 4 genome copies of each virus into 50 ng of human background DNA. The sensitivity of each candidate primer set for each target virus was assessed using these panels. Example of the experimental testing of HIV-1 specific primer set targeting gag gene (NP3/4) for its sensitivity with DNA analytical standards is shown in Figure S1 . Serial dilutions of a plasmid DNA corresponding to 5-10 4 HIV-1 genome copies spiked into 50 ng of normal human DNA, HIV-1 infected cells (H9/IIIB) (''positive'' control DNA) and uninfected human PBMC (''negative'' control DNA) are used in the experiment. Only a single melting peak Tm = 72.5uC corresponding to HIV-1 specific product was observed and no unspecific amplification was registered. Figure S1B revealed a standard curve showing the correlation between copy number of the target gene and Ct values with a slope = 23.77. The limit of sensitivity was determined in this assay to be 10 viral genome copies/PCR. After evaluation, a total of 24 primer pairs targeting the eight different viruses were chosen for the real-time PCR array based on their specificity and sensitivity (Table 1) . Among them, five of the primer sets were previously published and the other 19 primer sets were originally designed in the current study. Table 1 shows the sequences of the previously published primer sets selected for the real-time PCR array with the reference to the original source. Analytical sensitivity expressed in genome copy/PCR for each primer set (Table 1 ) was estimated using DNA/RNA analytical panels, as described above. Coverage of variants (i.e., different subtypes or genotypes) for each virus (Table 1 ) was estimated using the NSAs. Degenerative nucleotides were introduced into some of the primer sets based on the NSAs performed in-house to obtain a broader coverage. The Tm peak range of the product stated in Table 1 for each primer set was established during further testing of selected primers with analytical and clinical panels. In addition, the intra and inter-assay reproducibility of quantification for all primer sets was evaluated using three replicates of each standard dilution (of DNA or RNA analytical standards) in each of three real-time PCR assay runs. The coefficient of variation (CV) for the C (t) values was #3.3% and #6.7% for intra-and inter-assay, respectively. All the data depicting mean C (t), standard deviation (SD), and CV for each primer set selected for the real-time PCR array with each standard concentration are shown in Table S2 . To further evaluate the sensitivity of the selected primers, we used FDA/CBER panels (kindly provided by Dr. Stephen Kerby, FDA), consisting of various amounts of viruses spiked into ''normal'' human plasma, that are specifically developed and used for the evaluation of commercially licensed NAT assays. Table 2 summarizes the testing results for the selected primers against the target viruses. One out of four primer sets targeting HIV-1, NP3/4, could detect the subtype B HIV-1 RNA at the concentration of 50 copies/ml of human plasma. Two other HIV-1 specific primer sets (NP51/52 and NP170/171) detected the HIV-1 RNA at 100 copies/ml of plasma. The 4th primer set, NP175/174, (targeting the conserved pol region and containing degenerative nucleotides to support broader variant coverage) could detect the virus only at 500 viral genome copies/ml of plasma. HIV-2 RNA was detected with primer set NP86/87 at the concentration of 10 copies/ml and with the primer set NP76/77 at the concentration of 50 copies/ml. Another primer set (NP84/ 85) detected HIV-2 at 100 copies/ml. HBV (genotype A) DNA was detected with the primer set NP11/97 at 50 copies/ml and with NP94/100 at 100 copies/ml. The third HBV specific primer set (NP11/97-mod) detected HBV at 500 copies/ml of plasma. Both HCV specific primer sets targeting the viral 59NTR detected HCV RNA at the concentration of 10 copies/ml of plasma. Two of the WNV-specific primer sets (NP21/22, targeting E protein gene, and NP176/177, targeting NS5) detected WNV at 50 copies/ml and the third primer set (NP178/179, also targeting NS5) gave a positive signal at 100 copies/ml of plasma. The four HIV-1 specific primer sets were further evaluated by testing them against another FDA/CBER analytical panel containing a broad spectrum of HIV-1 subtypes. As summarized in Table 3 , the HIV-1 subtype F (group M) and group O isolates in Table 2 . The results of sensitivity testing of the real-time PCR array primer sets specific for HIV-1, HIV-2, HBV, HCV, and WNV the with FDA/CBER analytical plasma panels. Table 3 . Sensitivity of four HIV-1 specific primer sets selected for the real-time PCR array in testing with FDA/CBER analytical HIV-1 broad spectrum panel. Detection limit (copy number/ml of plasma) was evaluated using FDA/CBER analytical panel, containing pre-set copy number of HIV-1 spiked into 1 ml of ''normal'' human plasma. RNA from 1 ml of plasma was converted to cDNA and divided into eight PCR reactions; two PCR repeats were performed for each primer set. The estimated copy number per PCR reaction is shown in parentheses. the panel could not be successfully amplified by any of the four selected primer sets (detection limit is .1,000 RNA copy/ml of plasma). All other subtypes and CRFs of group M, and the group N isolate could be amplified with at least one out of the four primer sets at 50 to 1,000 RNA copies/ml of plasma (Table 3) . The Tm peaks of the PCR products amplified from different subtypes of HIV-1 varied within 1.5uC ( Table 3 ). All of the four selected HIV-1 specific primer sets amplified HIV-1 subtype B (group M) with the highest sensitivity (50-500 copies/ml of plasma). To examine the specificity and the ability of the array to detect different isolates within subtypes of HIV-1 and different isolates of HIV-2 by the array, we additionally tested our primers with another panel, developed by the Southern Research Institute. This panel contained three different isolates from each subtype (A to G) of HIV-1 group M, three isolates from HIV-1 group O, one isolate from HIV-1 group N and three different isolates of HIV-2. The infectious dose of each HIV isolate in the panel was determined by TCID 50 (median tissue culture infective dose) titration and the dose of virus used in each PCR reaction was calculated. The results of testing of HIV-1 specific primers are shown in Figure 1 . In this experiment we evaluated the coverage of HIV-1 variants and estimated the relative sensitivity of the four HIV-1 specific primer sets based on cycle threshold (C(t)) values obtained with each isolate tested. Two primer sets targeting the conserved pol regions (NP170/171 and NP175/174) were able to detect most of the HIV-1 subtypes with a high sensitivity (C(t) = 15-25). Only one primer set (NP175/174) detected both the group N isolate and all three isolates of group O with Ct = 20-35. We did not detect crossreactivity of HIV-1 or HIV-2 specific primers with the other type of HIV. All three HIV-2 isolates studied (1-3 log 10 TCID 50 /PCR input) were detected with all HIV-2 specific primer sets with a low Ct value: 12-20 (data not shown). After completion of sensitivity testing of primer sets with analytical panels we finalized the expected Tm peaks range for each amplicon and arranged a working array in 96-well plate format ( Figure S2 ). To evaluate the specificity and possible crossreactivity of primers in the array, 20-100 genome copies of each virus were spiked into 50 ng of human DNA to be used as positive templates and the same amount of human DNA was used as a negative control in each experiment. The array was tested with all targeted viruses using DNA standards (listed in Table S1 ). Each positive template was tested in duplicate; one human DNA negative control and one no-template control (NTC) were included in every testing. In these experiments the definition of ''positive'' signal was set up as the threshold cycle cut-off of C(t)#41 (see Materials and Methods) and all Tm peaks are within the expected range ( Figure S2) . No cross-reactivity was detected for any primer sets in the array using DNA standards with this setting (data not shown). Seventeen (17) clinical plasma samples (obtained from NIH Blood Bank) from donors who tested positive for HIV-1, HBV, or HCV were used to evaluate the array sensitivity and specificity. The genome copy of each virus in virus-positive plasma samples was previously determined by the NIH Blood Bank using commercial assays approved for donor screening. Representative results from the testing of three HBV-positive plasma samples (pt.#13-pt.#15) using our array are shown on Figure 2 . The threshold cycle cut off of the assay performed with plasma samples was set up as C(t)#43 (see Materials and Methods). The wells with C(t) lower than the cut-off and the obtained Tm peaks within the expected Tm range indicate positive amplification of HBV target genes and are highlighted with blue circles (Figure 2 , green color code for HBV). For samples #13 and #14 Tm peaks of the products, obtained with only two HBV specific primer sets (wells G2, G3-pt.#13 and wells G5, G6-pt.#14), were within the expected range, with C(t) values of 36.7-42.9 and 36.5-36.6 respectively. For sample #15, all three HBV specific primer sets gave the Tm peaks within the expected range (wells G7, G8 and G9), and the C(t) values range was 33.8-34.2. The genome copy number of HBV in plasma samples #13-15 was 151-518 copies/ ml, corresponding to 6-21 copies/PCR. The wells (white color code, red circles) with the human beta-globin gene specific primer set, serving as the internal control, produced positive signals from all three plasma samples tested, with C(t) values of 25.5-26.3 (wells H3, H6 and H9 - Figure 2 ), which shows that the quality of the RNA/DNA sample preparation was equally good for the plasma of these three patients. The final Tm range for each primer set was adjusted after testing of this clinical panel. Based on the results from the testing of clinical samples and according to NSA performed during the design, two primer sets targeting the S gene region of HBV (NP11/97 and NP11/97-mod) have two different Tm ranges depending on the HBV genotype ( Figure 2 -wells G1 and G3). The lower Tm range (74.5-75.5uC) was detected for genotype A and the higher Tm range (76.8-77.4uC) -for genotypes B, C, and D. The difference in Tm ranges occurred due to single nucleotide polymorphism (SNP) in the amplicon sequence leading to a difference in the G/C nucleotide content of the products. Table 4 summarizes the results from testing the 17 virus-positive clinical samples using the developed array. At least two different specific primer sets could detect a target virus in all cases, with one exception: one HIV-1 positive sample (#4) with ,4 viral genome copies per PCR reaction was amplified with only one primer set (NP175/174). HIV-1 positive clinical samples (#1-#3) with 51-80 viral genome copies/ml of plasma or 2-3 copies/PCR reaction were detected with at least two primer sets. All six HCV-positive samples (#7 to #12) with 16-2,570 viral genome copies/ml of plasma tested positive with both HCV-specific primer sets. Three HBV-positive samples (#15, #16, #17) at 21-66 copies/PCR tested positive by the three HBV-specific primer sets and two HBV-positive samples (#13, #14) with 6-10 copies/PCR tested positive with two out of three HBV-specific primer sets. It is important to note that none of the primer sets showed any crossreactivity with other viruses in the panel of clinical samples tested. In the study presented here we applied a real-time PCR array approach in a 96-well plate platform for detection of a group of target viral pathogens. In contrast to TaqMan PCR, which is commonly used for viral diagnostics, this platform based on PCR with SYBR Green chemistry supports simultaneous detection and identification of 24 different targets corresponding to eight different viruses. PCR based on SYBR Green staining of the double stranded DNA is economically affordable and allows for the detection of mutants with SNPs within the amplicon sequence [34] . The strategy of primer selection for the array included the sequential use of different bioinformatics programs to identify highly-specific primer sets with a maximal variants' coverage of the targeted viruses, while working under ''universal'' PCR conditions. Experimental selection process using a panel of DNA or RNA analytical standards allowed choosing two to four most sensitive and specific primer sets for each targeted virus. The sensitivity (in copy number per PCR reaction) and the intra and inter-assay reproducibility of quantification were characterized for each primer set selected for the array. FDA/CBER analytical plasma panels were used to assess the detection sensitivity of the primer sets selected for the working array. The HIV-2 and HCV RNA was detected at as low as 10 genome copies/ml of human plasma by one out of three and two out two primer sets, respectively. Similarly, WNV RNA and HBV DNA were detected at 50 genome copies/ml of plasma by two out of three and one out of three selected primer sets respectively. All four of the selected HIV-1 specific primer sets detected group M, subtype B, the most common subtype of HIV-1 in the Americas and Western Europe [35] , at 50-00 genome copies/ml of plasma. However, the array was able to amplify all other subtypes, excluding subtype F, of HIV-1 group M with only one primer set (NP170/171), targeting the conserved pol region, at 100-1,000 genome copies/ml of plasma. In addition, the array amplified the group N isolate at 1,000 copies/ml with another primer set (NP175/174) targeting the conserved pol region. All the selected HIV-1 specific primer sets of the array failed to amplify the group O isolate (detection limit is .1,000 copies/ml of plasma). In comparison, the limit of detection (LOD) of recently improved commercial multiplex NAT assays for the broad spectrum of HIV-1 group M isolates is 40 copies/ml, and the LOD for group O is 200 copies/ml [36] [37] [38] . Thus, specific primer sets targeting group O isolates of HIV-1, as well as subtype F (group M) will need to be included in the array to increase the coverage of all existing subtypes/groups of the HIV-1. No crossreactivity was shown for HIV-1 or HIV-2 specific primers with the other type of HIV in a testing with medically relevant levels of viruses in a diversity panel containing 25 different HIV-1 isolates and three natural HIV-2 isolates collected worldwide. There is presently no US Food and Drug Administration (FDA) approved PCR-based NAT testing for blood donors' screening for HTLV-1 and HTLV-2. There is no official FDA (or World Health Organization (WHO)) viral panel released for these two viruses. It is also difficult to obtain HTLV-1 or HTLV-2 positive blood donors' samples in the United States. In the absence of the analytical panels and clinical samples we tested HTLV-1, HTLV-2 and VACV specific primers only with cell culture derived DNA and with DNA standards. The minimum detection limit of HTLV-1 and HTLV-2 specific primer sets estimated with analytical DNA standards was 5-10 genome copies/PCR reaction, which is in the same range as for other primers included to the array. The coverage of the viral variants for the primers targeting these viruses was estimated in silico using multiple sequence alignments performed with complete genome sequences available in Gen Bank. The working array arranged in 96-well plate format was subsequently tested for specificity and potential cross-reactivity with human DNA and with each of the targeted viruses. None of the primer sets selected for a particular target virus in the array produced non-specific cross-reaction toward the other viruses. Comparative performance of the array was also evaluated through the testing of 17 clinical specimens from the United States patient population. All 17 samples were correctly identified in our PCR array with a high sensitivity to contain HIV-1, HCV, or HBV. We found that a combination of several primer sets targeting each virus in the array allows for the detection of different variants of the virus; however, it makes the absolute quantification with uniform DNA/RNA standards a challenge. Quantification by the assay is not always possible when the genetic group of the viral isolate being tested is different than the assay standards. This is one of the reasons why in certain cases the commercial assays underestimate viral loads by up to 1-10 log 10 copies per ml [37] [38] [39] . Nevertheless, relative quantification can be done using all primer sets selected for the array, and the absolute quantification can be performed with DNA/RNA standards using the primers targeting the most conserved genome areas. There are several commercial qualitative multiplex NAT assays now available on the market simultaneously targeting three most important blood-borne viral pathogens (HIV-1, HCV and HBV) [40] . One of them was recently approved by U.S. FDA for screening of blood and organ donors (http://www.fda.gov/BiologicsBlood Vaccines/BloodBloodProducts/ApprovedProducts/LicensedProduc tsBLAs/BloodDonorScreening/InfectiousDisease/ucm306073.htm). We compared the LOD of these multiplex NAT assays [40] with the results of our working PCR array in sensitivity of testing against these important viruses. The LOD for HBV are 38.1-195 geq/ml by ''Procleix Ultrio Tigris'' and 9.2-37. Other PCR-coupled techniques have been developed previously for highly-sensitive pathogens' detection that could reach the sensitivity of the assay up to 1 genome copy per PCR reaction. For example the bioactive amplification with probing (BAP), utilizing a nested PCR and magnetic bead-based hybridization with the specific probe, has been developed for the detection of bovine and avian viruses [41] [42] [43] . In spite of the exceeding sensitivity of such assays targeting a single virus, it may be difficult to adapt the approach or method to meet the main objective of simultaneous detection of multiple target viral pathogens by an array using universal PCR conditions. It is important to note that this array was developed to be adapted by any laboratory. Comparison of experimentally obtained Tm peaks to the range of expected specific Tm peaks allows rapid identification or exclusion of the viral pathogen in a sample. This is an initial study to examine the suitability of using PCR arrays for the detection of a group of target viruses. The current array was developed utilizing five previously published and Table 4 . Tm and C(t) values obtained with primer sets specific for HIV-1, HCV, or HBV in testing of 17 human clinical samples in the format of PCR array targeting eight different viruses. 19 originally designed primers sets. However, it can be expanded to a larger number of targets for the same virus. Targeting of several genome areas increases the detection sensitivity of the target virus and provides an intra-assay confirmation of positive signals. Additionally, any new virus of interest can be added to the list of targeted pathogens. Efforts are underway to test the utility of this real-time PCR array using samples with more diverse biological origin and pathogen content.

TcdC Does Not Significantly Repress Toxin Expression in Clostridium difficile 630ΔErm

In the past decade, Clostridium difficile has emerged as an important gut pathogen. Symptoms of C. difficile infection range from mild diarrhea to pseudomembranous colitis, sometimes resulting in colectomy or death. The main virulence factors of C. difficile are toxin A and toxin B. Besides the genes encoding these toxins (tcdA and tcdB), the pathogenicity locus (PaLoc) also contains genes encoding a sigma factor (tcdR) and a putative anti-sigma factor (tcdC). The important role of TcdR as a sigma factor for toxin expression is undisputed, whereas the role of TcdC as an anti-sigma factor, inhibiting toxin expression, is currently the subject of debate. To clarify the role of TcdC in toxin expression, we generated an isogenic ClosTron-based mutant of tcdC in Clostridium difficile strain 630Δ Erm (CT::tcdC) and determined the transcription levels of the PaLoc genes and the expression levels of the toxins in the wild type strain and the tcdC mutant strain. We found only minor differences in transcription levels of the PaLoc genes between the wild type and CT::tcdC strains and total toxin levels did not significantly differ either. These results suggest that in C. difficile 630Δerm TcdC is not a major regulator of toxin expression under the conditions tested.

Clostridium difficile is an anaerobic, Gram-positive, spore forming rod shaped bacterium that can cause disease with a wide variety of symptoms, ranging from mild diarrhea to severe forms of pseudomembranous colitis [1] [2] [3] . Since 2004, numerous countries have reported outbreaks in health-care facilities caused by hypervirulent C. difficile PCR Ribotype (Type) 027 [1] [2] [3] [4] [5] [6] . Clostridium difficile infection (CDI) caused by Type 027 is associated with a more severe course of the disease and a higher mortality rate than other ribotypes [1, 3, 6] . Recently, increasing numbers of the hypervirulent Type 078 are reported [7] . C. difficile Type 078 is more frequently associated with community acquired CDI and affects a younger population than Type 027 [6] [7] [8] [9] . Furthermore, CDI caused by Type 078 is associated with an increased morbidity compared to other ribotypes [8] . The main virulence factors of the enteropathogenic C. difficile are the two large clostridial toxins, toxin A (TcdA) and toxin B (TcdB). These toxins are glycosyltransferases that inactivate Rho, Rac and Cdc42, thereby disrupting the cytoskeleton and tight junctions of the cells, resulting in apoptosis [10] . This induces an inflammatory response and degradation of the intestinal epithelial cell layer. Besides the genes encoding these toxins (tcdA and tcdB), the pathogenicity locus (PaLoc) also contains genes encoding a sigma factor (tcdR) and a putative anti-sigma factor (tcdC) [11] [12] [13] . In between the toxin genes the tcdE gene is situated, which encodes a putative holin protein [14] . Interestingly, both hypervirulent Types 027 and 078 have been shown to contain mutations in the tcdC gene, encoding the putative negative regulator of toxin gene transcription, and this has been proposed as a possible explanation for their increased virulence [8, 15] . The exponential growth phase of C. difficile has been reported to be associated with a high transcription level of the tcdC gene and low transcription levels of tcdR and the toxin genes, whereas the stationary growth phase is associated with a low transcription level of the tcdC gene and high transcription levels of tcdR and the toxin genes in strain VPI10463 [16] . The synthesis and secretion of the toxins is increased upon entry into the stationary growth phase [16] [17] [18] [19] . The decreasing transcription of tcdC correlates with diminishing TcdC protein levels in stationary growth phase [16, 20] . TcdR is an alternative sigma factor that positively regulates toxin production [11, 12] . The direct interaction of TcdR and the RNA polymerase core enzyme mediates recognition of the toxin promoters and the tcdR promoter [11, 12, 21] . TcdC has been reported to act like an anti-sigma factor for toxin production by destabilizing the TcdR-RNA polymerase core enzyme complex in a way that is not yet fully understood [12] . The reported inverse correlation between the transcription of tcdC and the toxin genes and the expression patterns of the corresponding proteins, together with the biochemical data, has led to the prevailing model that TcdC is an important repressor of toxin expression [12, 16, 17, 20] . This model seems to be supported by the finding that the absence of a functional TcdC caused by a frame shift mutation (D117 bp) in the tcdC gene is linked to a supposed increased toxin production in certain (hyper) virulent strains [15, 22] . Recently, some doubts were raised about the importance of TcdC for regulation of toxin expression on the basis of two findings. First, two studies have found increasing levels of tcdC transcription in time that coincide with increasing transcription of the toxin genes and increasing amounts of toxin production [18, 19] . Second, there is a great variability in toxin expression levels among (hyper) virulent strains, even though these generally carry mutations in tcdC [15, 18, 19] . Therefore, a minor (or modulatory) role for TcdC in the regulation of toxin expression was proposed [18, 19] . Here, we sought to clarify the role of TcdC in regulation of the toxin production by generating an isogenic tcdC mutant (CT::tcdC) using the ClosTron technology. We find only minor differences in transcription levels of the PaLoc genes between the wild type and CT::tcdC strains and the expressed total toxin levels did not significantly differ, suggesting that the role of TcdC in toxin regulation is not of significance under the conditions tested in C. difficile strain 630DErm. The importance of TcdC for regulation of toxin expression was recently challenged by two studies [18, 19] . It was proposed, based on the increasing transcription levels of the PaLoc genes in time and the variability in toxin expression levels among virulent strains, that TcdC has a minor or modulatory role on toxin expression rather than a major role as previously assumed. In this study we sought to clarify the role of TcdC for toxin expression by generating an isogenic tcdC mutant. As toxin gene expression is subject to complex regulation influenced by glucose and cysteine, we performed our experiments in a trypton-yeast (TY) based broth [17, 23] . TY broth does not contain glucose and no cysteine was added. We verified that in TY broth earlier and higher expression of toxins was achieved in comparison to the commonly used Brain Heart Infusion (BHI) broth (data not shown). TcdC consists of three domains: a hydrophobic domain, a proposed dimerization domain and a proposed C-terminal repressor domain ( Figure 1A ) [12] . We successfully disrupted the tcdC gene in the region coding for the repressor domain using ClosTron technology. Disruption of genes using the ClosTron technology results in stable mutants and no or non-functional proteins [24] [25] [26] . The genotype of the disruption was confirmed with conventional PCRs using the tcdC2 primer and the EBS universal primer and with a primer pair (tcdC1 and tcdC2) flanking the ClosTron insertion site ( Figure B ). Sequence analysis confirmed that the disruption was in the proposed repressor domain of the tcdC gene at the expected site (data not shown). In addition, Southern blot analysis using intron-, ermB and tcdCspecific probes clearly confirmed a specific single insertion of the Group II intron in the genome ( Figure 1C ). Western blot analysis, using antibodies against TcdC, confirmed that the isogenic tcdC mutant no longer expressed TcdC ( Figure 1D ). A control blot using antibodies against F 0 F 1 ATPase indicated that the lack of signal in the TcdC Western blot was not a result of lower amounts of proteins loaded in the lanes of PCR ribotype 035 (a PaLoc negative strain) and the tcdC mutant. The growth kinetics of the wild type and CT::tcdC strains showed no significant differences in various media tested ( Figure 1E and data not shown). In TY broth, which does not contain glucose or added cysteine, the wild type strain and the CT::tcdC strains showed an exponential growth phase in the first 8 hours post inoculation and after 12 hours post inoculation both strains entered into the stationary growth phase ( Figure 1E ). Conventional control PCRs confirmed that the disruption of the tcdC gene had remained intact during our growth curves experiments (data not shown). Comparable Relative Transcription Levels of PaLoc Genes in Wild Type and CT::tcdC In order to determine the influence of TcdC on the transcription levels of the PaLoc genes we compared the relative transcription levels of the PaLoc genes of wild type and CT::tcdC strains by reverse transcriptase quantitative real-time PCR (RT-qPCR). We found comparable transcription levels of all PaLoc genes in wild type and CT::tcdC strains. Overall, the logarithmic growth phase was associated with lower transcription levels of the PaLoc genes and by entering into the stationary phase increasing transcription levels of PaLoc genes were found, as previously described for tcdR, tcdE, tcdB and tcdA [16, 18, 19] and tcdC [18, 19] (Figure 2 ). The transcription levels of tcdR in wild type and CT::tcdC strains increased approximately 100-fold between 6 and 24 hours post inoculation ( Figure 2A ). Though the expression of tcdR was, on average, 3-fold higher at the various time points in the CT::tcdC strains compared to the wild type, this difference was not statistically significant ( Figure 2A , all p values $0.088). Similarly, we observed a 10-to 100-fold increase in the transcription levels of tcdB ( Figure 2B ), tcdE ( Figure 2C ), tcdA ( Figure 2D ) and tcdC ( Figure 2E ) when comparing values from the logarithmic growth phase with those observed in the stationary growth phase. The expression levels of tcdB, tcdE, tcdA and tcdC were, on average, 1.5-fold, 2.5-fold, 1.4-fold and 1.7fold higher, respectively, in the CT::tcdC strains compared to the wild type. With one exception, these differences were not found to be significant. The transcription level of tcdB in the CT::tcdC1 strain is significantly (P = 0.046) higher compared to wild type level at 8 hours post inoculation ( Figure 2D ). However, no significant differences are found between the wild type and CT::tcdC strains at any of the other time points. Therefore, we conclude that the disruption of the tcdC gene does not result in a consistently and significantly increased transcription level of the PaLoc genes. Considering the small increase in PaLoc gene expression in the CT::tcdC mutants observed in the RT-qPCR experiments, we were interested to see if this difference translated into higher toxin levels. We determined toxin levels using two independent assays, but found no consistent difference between wild type and mutant cells. First, filter sterilized bacterial supernatants were incubated on a Vero cell (a kind gift of Dr. E.J. Snijder [27] ) monolayer and cytotoxic effects were quantified after 24 hours by determining the end-point titer ( Figure 3A ) [26] . In the exponential growth phase (5 and 8 hours post inoculation) no cytotoxic effects were detectable (data not shown). In the stationary growth phase (12, 24 and 48 hours post inoculation) we observed increasing cytotoxic effects, indicative of the presence of toxin. Importantly, the observed cytotoxic effects were specific for C. difficile toxin A and B, as a pre-incubation of the filter sterilized bacterial supernatants with anti-toxin, a polyclonal antibody against toxin A and toxin B (Techlab), resulted in complete neutralization of cytotoxic effects on the Vero cells at all time points (data not shown). The tcdC mutant strains showed no significant differences in toxin levels compared to the wild type strain (Fig. 3A) . Next, we used an enzyme immunoassay (Ridascreen, Biopharma) for the direct detection and relative quantification of the secreted toxins. In the exponential growth phase (5 and 8 hours post inoculation) no toxins were detectable (data not shown), consistent with the lack of toxicity towards Vero cells described above. In the stationary growth phase (12, 24 and 48 hours post inoculation) increasing toxin levels were detectable. When we compared the toxin levels at various time points, there were equal amounts of toxins in the wild type and tcdC mutant strains. We conclude that the disruption of the tcdC gene does not result in consistently and significantly increased toxin levels. C. difficile infections caused by the (hyper-)virulent Type 027 (NAP1/REA B1) and Type 078 (NAP7/REA BK) are associated with an increased morbidity and severity of disease compared to other types [1, 3] . This increase is suggested to be linked to toxin hyper production [3, 22, 28] . A potential mechanism by which this could occur is through inactivation of a negative regulator of the toxin gene transcription. TcdC has been identified as a negative regulator of toxin production [12] . In the currently prevailing model, a major role for TcdC in the repression of toxin genes has been proposed on the basis of three lines of evidence. First, in C. difficile VPI10463 (a high toxin producing strain that also expresses high levels of TcdC [18, 29] ), an inverse correlation between the transcription of tcdC and the genes encoding the toxins is found [16, 18, 29] . This correlation for TcdC is also observed in protein levels [20] . Second, elegant in vitro experiments have established that heterologously produced and purified TcdC protein can interfere with TcdR-mediated transcription of toxin genes in a way that is not yet fully understood [12] . Finally, a frame shift mutation (D117 bp) in tcdC, that results in a non-functional protein, is associated with increased toxin production in certain (hyper)virulent strains [15, 22] . Recently, it was reported that the introduction of a functional tcdC gene from a high toxin-producing strain that lacks any of the hyper virulence associated tcdC mutations (VPI10463, PCR ribotype 087) into an epidemic strain carrying a non-functional tcdC (M7404, PCR ribotype 027/NAP1/REA B1) can reduce toxin expression levels and moderately attenuate virulence [30] . This data seems consistent with the model discussed above. However, it is unclear how the levels of TcdC in the complemented strain relate to the physiological levels of the protein prior to the inactivation of TcdC in this strain background. The introduced tcdC gene, including its transcription signals, was derived from a different genetic background (VPI10463, Type 087) and was introduced on a multicopy plasmid. In addition, the reintroduction of TcdC in a strain lacking a functional TcdC, may affect processes that are not normally affected. Finally, the experiments were not corrected for the additional copies of the tcdC promoter that could result in the titration of regulators binding to those sequences. In an alternative approach that addresses many of the issues above, the role of tcdC in toxin expression could be addressed by removing it from a background in which it is normally functional. To this end, we generated two independent isogenic ClosTronbased tcdC mutants strain that could be directly compared to its wild type counterpart, in which the TcdC protein was expected to be functional. Our data obtained with these mutant strains show that TcdC does not exert a major or even significant effect on the transcription of the PaLoc genes or the expression levels of the toxins under the conditions tested. Our experiments were performed in a glucose free TY broth medium, since glucose is a known repressor of toxin production [17] . Indeed, we observed earlier and higher levels of toxin production in TY broth than in the commonly used Brain-Heat-Infusion broth (BHIS) based media, that does contain low amounts of glucose (0.2%) and to which frequently cysteine is added. However, also in BHIS we did not observe a significant effect of a tcdC deletion on toxin expression (data not shown). We controlled critical parameters in our experiments by performing conventional PCRs which confirmed that the disruption of tcdC remained intact throughout the growth curve. Western blot analysis with antibodies raised against a TcdC epitope confirmed that the disruption of the tcdC gene resulted in the absence of TcdC protein ( Figure 1D ). The disruption of the tcdC gene did not affect the growth kinetics compared to the wild type strain ( Figure 1E ). In the RT-qPCR experiments, sample to sample variation was corrected by normalizing to the reference gene rpsJ [31] . The rpsJ gene was selected for normalization, since rpsJ was overall the highest ranked reference gene regarding gene expression stability [31] .Reverse transcription was carried out using random hexamers, to prevent gene specific biases [32] . PCR efficiency in the qPCR was determined using a standard curve for each gene, enabling post run correction [33] . To obtain objective data concerning the quantification of the secreted toxins, we used an end point titer assay and an enzyme immunoassay rather than a manual (subjective) cell scoring system [26] . The trends observed in the transcription of the PaLoc genes and the expression of the toxins generally conform to previously reported data [18, 19] . It should be noted that the up-regulation in time of tcdC transcription was not observed in earlier studies on C. difficile VPI10463 [16] but is consistent with more recent reports [18, 19] . We observed an increase in transcription of the PaLoc genes in time, and a concomitant increase in toxicity of culture supernatant in stationary phase that can be attributed to the toxins as it is fully neutralized by anti-toxin against toxin A and B. The disruption of the tcdC gene resulted in an on average 1.7 fold higher transcription level of tcdC in time compared to the wild type strain, although this difference was not found to be statistically significant. It should be noted that we detect these differences because the real time PCR probe detects a region of the gene upstream of the ClosTron insertion site ( Figure 1A ). This finding might indicate some kind of feedback mechanism on TcdC expression. Similar to tcdC gene expression, the disruption of tcdC resulted in a slightly higher transcription level of the other PaLoc genes, although this was generally not significant. Moreover, the increased transcription level of the toxin genes did not result in generated a PCR product of 302 bp for the CT::tcdC strains. Primers tcdC1 and tcdC2 generated a 699 bp PCR product for the wild type and for the CT::tcdC strain a PCR product of circa 2800 bp. (C) Southern blot analysis of EcoRV digested genomic DNA of wild type and CT::tcdC strains with a Group II intron, ermB gene and tcdC specific probes. Note that probing with the ermB probe results in 2 bands for the CT::tcdC strains, since wild type already carries a copy of the ermB gene in the genome [35] . (D) Western blot analysis of TcdC production in wild type and CT::tcdC strain 8 hours post inoculation. The arrow indicates the location of TcdC protein based on MW and absence of the protein in the PaLoc negative Type 035 strain. Note that cross-reaction of TcdC antibody with a protein of similar MW was also observed in Carter et al. [30] . (E) Growth curves of C. difficile 630DErm and C. difficile CT::tcdC mutant strains. The absorbance (OD 600 ) was measured over 48 hrs of growth in TY medium. The error bars indicate the standard error of the mean of six experiments. doi:10.1371/journal.pone.0043247.g001 a detectable increase in toxin levels as measured with two independent assays. Based on the paradigm that TcdC is a major suppressor of toxin production we expected precocious and significantly elevated transcription levels of tcdA, tcdB, tcdE and tcdR in the CT::tcdC strains compared to wild type. However, our data indicate that TcdC exerts a moderate, if any, effect on the transcriptional levels of the PaLoc genes and the expression of toxins in C. difficile 630Derm under the conditions tested. Clostridium difficile strain 630DErm is a derivative of the clinical isolate 630 [34, 35] , a PCR ribotype 012 strain. PCR ribotypes 012 strains constitute 4% of the clinically isolated toxinogenic isolates in Europe [7] . Clostridium difficile 630 (PCR ribotype 012)-derived strains are commonly used to investigate virulence of mutants [26, 36, 37] . An independent study, published during the preparation of this manuscript, reached a similar conclusion with respect to the role of TcdC in toxin regulation in C. difficile 630Derm using an allelic exchange technique [38] . In that paper reintroduction of a single functional copy of tcdC at its native locus did not affect toxin production in strain R20291 either [38] . R20291 is a strain from problematic PCR ribotype 027 (NAP1/REA B1) that was isolated following an outbreak in Stoke Mandeville, UK. Our work and that of Cartman and coworkers [38] seem at odds with the previous reports that clearly demonstrate that TcdC can act as a repressor for toxin gene expression [12, 30] . However, we cannot exclude the possibility that TcdC exerts a more profound effect under specific conditions, or in other strains of C. difficile than 630Derm and R20291. It should be clear though that in vivo relevance of TcdC for toxin regulation in these two strains is limited. In conclusion, we suggest that TcdC might have a modulatory role in regulating toxin expression, and that TcdC functionality is therefore not a major determinant of the (hyper)virulence of C. difficile. This is supported by the lack of correlation between virulence, toxin production and tcdC gene variants that was noted by several other studies [18, 19, 30, 39] . The Clostridium difficile and Escherichia coli (E. coli) strains and plasmids used in this study are described in Table 1 . E. coli strains were grown in Luria Bertani (LB, USB cooperation) medium supplemented with appropriate antibiotics when required. C. difficile strains were grown anaerobically in a microaerobic cabinet (Don Whitley VA1000) at 37uC in pre-reduced 3% Bacto Tryptose (Difco), 2% Yeast extract (Difco) and 0.1% thioglycolate (pH 7.4) medium (TY) or Brain Heart Infusion broth (Oxoid) supplemented with 0.5% yeast extract and 0.01% L-cysteine (Sigma) (BHIS) [40, 41] . When required, the broths were supplemented with appropriated antibiotics. For RNA extraction and toxin quantification, C. difficile 630DErm (wild type) and two independent isogenic tcdC mutant strains (CT::tcdC) were serially diluted and pre-cultured (overnight) in pre-reduced TY broth. Mid-logarithmic growth phase pre-cultures (OD 600 0.4-0.8) were used to inoculate pre-reduced TY broth to a starting OD 600 of 0.05 (60.01). Optical density readings and samples for total toxin quantification were taken at 2, 3, 4, 5, 6 and 8 hours post inoculation in the exponential growth phase and at 12, 24 and 48 hours post inoculation in the stationary phase. Samples for RNA extraction were taken at 6, 8, 12, and 24 hours post inoculation. Samples for Western blot detection of TcdC were taken at 8 hours post inoculation. We routinely monitored the purity of the We generated two independent isogenic tcdC mutants by insertional inactivation of the tcdC gene in the wild type strain 630Derm using ClosTron technology [24, 25] . Briefly, the Perutka algorithm on the ClosTron website (http://www.clostron.com) was used to design primers ( Table 2) for retargeting the Group II intron (Sigma; Targetron). The retargeted intron was cloned using the restriction enzymes BsrGI and HindIII into plasmids pMTL007C-E2 and the constructs were verified by sequencing [25] . The verified plasmid (pDB001) was transformed to E. coli CA434 and transferred to the wild type strain 630Derm via conjugation [34, 41] . The selection of C. difficile transconjugants was done by subculturing on pre-reduced BHIS agar supplemented with thiamphenicol (Sigma; 10 mg/ml) and C. difficile selective supplement (Oxoid). This was followed by several rounds of subculturing on pre-reduced BHIS agar supplemented with lincomycin (Sigma; 20 mg/ml) and C. difficile selective supplement to promote integration of the GroupII intron into the gene of interest. Chromosomal DNA isolated from the transconjugants using a QIAamp blood kit (Qiagen) was used in conventional PCRs and sequence runs to confirm the disruption of tcdC and the nucleotide position of the insertion in the tcdC gene. Primers used for cloning and sequencing are listed in Table 2 . Complementation can be a valuable control for knockout studies. However, as our tcdC mutant strains have no clearly detectable phenotype regarding toxin production, complemented mutant strains are expected to be comparable to wild type and tcdC mutant strains, as also reported recently in an independent study [38] . Therefore, a complementation study would not add to the message of this manuscript. Southern blot analysis was performed to verify a specific single integration into the genome. Genomic DNA was extracted using a Phenol-chloroform extraction [42] . Four mg of genomic DNA was digested with EcoRV enzyme and separated on a 0.8% agarose/0.5xTris-acetate-EDTA gel by electrophoresis. DNA was transferred onto a Hybond N+ filter (Amersham) in 10x saline sodium citrate (SSC) solution. The filter was washed in 2X SSC and baked at 80uC for 2 hours. Prehybridization of the filter was done for 2 hrs at 60uC in 5x SCC, 5x Denhart and 100 mg/ml of yeast tRNA. Probes specific for the group II intron (EBS2-tcdC623as/Sal-R1), ermB gene (oWKS1131/oWKS1132) and tcdC gene (tcdC5-tcdC6) were generated. Primers are listed in Table 2 . The generated probes (100 ng) were radiolabeled ( 32 P dATP) using Klenow enzyme (Roche) and overnight hybridized in 10 ml fresh pre-hybridization buffer at 60uC. The filter was washed for 30 min in 2x SCC, 0.5% SDS, 30 min in 1X SSC, 0.5% SDS and 30 min in 0.5X SSC, 0.5% SDS and analyzed using phosphorimage screen and a Typhoon 9410 scanner (GE healthcare). Antibodies against TcdC were generated by immunizing rabbits with a synthetic peptide (CQLARTPDDYKYKKV) representing a specific TcdC epitope (Genscript). Note that this epitope is located before the Clostron insertion site, and would therefore also be expected to detect truncated TcdC protein, would this be produced. Western blots were performed as follows. C. difficile (2 mL) cultures were harvested by centrifugation (2 min, 11.0006g, 4uC) and washed with Phosphate Buffered Saline (PBS). The bacterial pellets were resuspended in PBS containing protease inhibitor cocktail (Complete, Roche) and lysed by sonification. The bacterial lysates were centrifuged at low speed (3 min, 10006g, 4uC) to remove unbroken bacterial cells [20] . To separate the cytosolic proteins from the membrane associated proteins the bacterial supernatant was centrifuged at 200.0006g, 4uC for 1 hr [20] . The pelleted membrane associated proteins were resuspended in 10 mM Tris-HCl (pH 7.4), 5 mM EDTA with 2% Triton X-100 for 30 min at room temperature. Equal amounts of the resuspended membrane associated proteins were separated on 15% SDS PAGE gel and transferred onto polyvinyl difluoride (PVDF) membranes. Similarly generated membranes with the transferred membrane associated proteins of a Type 035 (PaLoc negative) strain were used for pre incubation of the TcdC antibodies. The membranes were probed with the pre incubated TcdC antibody and an antibody against the b subunit of the E. coli F 0 F 1 ATPase that cross reacts with the homologous protein in C. difficile [20, 43] . The probed membranes were analyzed using secondary anti-mouse horse radish peroxidase conjugated antibodies (Dako), a chemiluminescence detection kit (Amersham) and a Typhoon 9410 scanner (GE Healthcare). Five mL of the C. difficile cultures were 1:1 diluted with ice cold methanol and stored overnight at 280uC. Bacterial pellets, obtained by centrifugation (20 min, 30006g, 4uC), were resuspended into 200 ml lysisbuffer (100 mM EDTA, 200 mM Tris-HCl pH 7.0, 50 mg/ml lysozyme) and incubated for 1 hr at 37uC. Tri-pure reagent (Roche) was used for the extraction of RNA according to the manufacturer's instruction with minor modifications. Briefly, 1 ml Tri-pure was added to the lysed bacterial pellets and incubated for 5 min at room temperature. Per 1 ml Tri-pure, 200 ml chloroform was added and carefully shaken by hand for 3 min, followed by an incubation of 2-5 min at room temperature. The aqueous phase was collected after centrifugation (12,0006g for 15 min at 4uC) and transferred to a fresh tube. RNA was precipitated by mixing the aqueous phase with 500 ml isopropanol, followed by an incubation of 10 min at room temperature. The precipitated RNA was collected by centrifugation (12,0006g, 10 min, 4uC) and resuspended in 100 ml DNase/ RNase free water. The RNA was re-precipitated overnight at 280uC with ammonium acetate (Fluka; 10 mM) and 3 volumes of absolute ethanol. The re-precipitated RNA was washed once with 80% ethanol and dissolved in 50 ml DNase/RNase free water. The RNA was treated twice with a TurboDNase (Ambion) according to the manufacturer's instruction followed by another Tri-pure RNA isolation. The quality and purity of the extracted RNA was assessed using a RNA nano chip on an Agilent Bioanalyzer. A RevertAid TM H Minus Reverse Transcriptase kit (Fermentas) was used to synthesize cDNA according to the manufacturer's instruction. Random hexamers were used to convert 750 ng RNA into cDNA. The synthesized cDNA was treated with RNase (Qiagen) for 1 hour at 37uC and stored at 220uC. The software program Molecular Beacon (Premier Biosoft) was used to design primer pairs and probes (Table 2) for the 2 multiplex quantitative PCRs (qPCR), based on the available genome of C. difficile strain 630 [35] . All primer pairs were first tested by conventional PCR and multiplex PCR to confirm specificity and amplicon sizes. The primer pair and the probe for the amplification of the tcdC gene are in front of the insertion site in the tcdC gene ( Figure 1A ), allowing detection of tcdC transcription levels in wild type and CT::tcdC strains. The real-time multiplex qPCR amplification of the PaLoc genes and the reference gene encoding for a ribosomal protein (rpsJ) was performed on a CFX96 real-time PCR detection system (Biorad) [31] . The amplification efficiencies of the PaLoc and reference genes were determined using serially diluted genomic DNA (standard curve). The manually calculated efficiencies and the reference gene rpsJ were used to normalize the expression levels of the PaLoc genes. The amplification was performed in a 25 ml final volume. The first real-time multiplex qPCR (target genes: tcdA, tcdA and tcdC) contained 25 ml Hotstar mastermix (Qiagen), forward and reverse primers (80 nm each primer), 2.5 mM MgCl 2 , 100 nM of each probe and 2 ml synthesized cDNA. The second multiplex real-time multiplex Q-PCR (target genes: tcdR, tcdE) contained 25 ml Hotstar mastermix (Qiagen), forward and reverse primers (80 nm each primer), 3.5 mM MgCl 2 , 100 nM of each probe and 2 ml synthesized cDNA. The real-time qPCR to quantify the reference gene rpsJ contained 25 ml Hotstar mastermix (Qiagen), forward and reverse primers (80 nm each primer), 2.5 mM MgCl 2 , 0.06% SYBRgreen (Sigma) and 2 ml synthesized cDNA. The real-time qPCR protocol included an enzyme activation step for 15 min at 95uC, followed by 50 cycles of amplification; 95uC for 30 sec, 52uC for 30 sec and 72uC for 30 sec. Total toxin amounts were quantified using 2 assays; a toxin end point titer assay and a commercial available ELISA (Ridascreen, Biopharma). The supernatants of culture samples (1 mL) were collected after centrifugation (30 min, 30006g, 4uC), filter sterilized (0.45 mM cellulose acetate membrane) and stored at 4uC. For the toxin end point titer assay, Vero cells were seeded into a 96 wells plate at a density of 1610 4 cells per well and incubated overnight at 37uC and 5% CO 2 . The filter sterilized supernatants of 5, 8, 12, 24 and 48 hours post inoculation (hpi) were diluted 2, 10 1 , 10 2 , 10 3 , 10 4 and 10 5 fold in cell culture medium (Dulbecco modified Eagle medium (Lonza) supplemented with penicillin 100 u/mL, streptomycin 100 U/mL, fetal calf serum(10%). Fifty ml of the dilutions were added onto the Vero cell monolayers and incubated for 1 hr at 37uC and 5% CO 2 . For the neutralization assay a 2-fold dilution of each tested time point (5, 8, 12, 24 and 48 hpi) was pre-incubated with a 1/100 diluted anti-toxin (Techlab) for 1 hr at 37uC and 5% CO 2 . After the pre-incubation, 50 ml was added onto the Vero cell monolayers. The incubated bacterial supernatants were aspirated off after one hour and replaced with 100 ml cell culture medium. After 24 hrs of incubation the end-point titer was determined of each diluted time point [26] . The end-point titer was defined as the first dilution at which the Vero Cell morphology was indistinguishable from the neutralized 2-fold diluted supernatants [26] . The enzyme immunoassay (Ridascreen, Biopharma) was performed according manufacture's protocol. Statistical analysis was performed using the software package SPSS 18 (IBM). An independent sample t-test was employed to compare the strains at different time points.

Childhood Tuberculosis Presenting with Haemophagocytic Syndrome

Haemophagocytic syndrome is a life threatening complication of systemic infection resulting from an exaggerated immune response to a triggering agent. Prompt recognition and treatment of this disorder can abrogate otherwise high fatality associated with this disorder. A 2 year old girl presented with acute enteritis, developed prolonged fever and organomegaly complicated by multi-organ failure. She fulfilled the diagnostic criteria for haemophagocytic lymphohistiocytosis including bone marrow evidence of haemophagocytosis. In addition she had serological evidence of tubercular infection as well as a positive family history of tuberculosis. She responded rapidly to immunosuppressive therapy and anti-tubercular therapy. Our case illustrates the association of haemophagocytic syndrome with tuberculosis as well as the favourable response obtained with prompt diagnosis and treatment.

Haemophagocytic lymphohistiocytosis (HLH) is a rare disorder caused by unrestrained proliferation and activity of the monocyte-macrophage system with phagocytosis of the mature and immature formed blood cells, release of inflammatory mediators, coagulopathy and often multiorgan failure. It has been described in all age groups, especially in the paediatric-adolescent population. Management usually consists of immunosuppressive agents along with treatment of the underlying condition. The HLH 2004 protocol consists of repeated cycles of cyclosporine-etoposidedexamethasone; however, sustained responses are rare, especially in familial HLH, and most patients eventually relapse [1] . Bone marrow transplant remains the only effective therapy for refractory cases but entails high procedure related mortality. Various studies have reported 5 year survival rates of 50-60% for children with HLH, including familial and acquired forms [2, 3] . The diagnosis of familial HLH is often based on the age of onset, family history including a history of consanguinity, the clinical profile and/or co-existence of inherited immune deficiencies. Frequent relapses are common and these patients are usually candidates for BMT [4] . However, differentiation from early onset acquired HLH can be difficult. Absence of markers of immune deficiency (CHS, GS or XLP) or genetic perforin-granyzyme mutations does not rule out familial HLH. Acquired HLH has been described in association with collagen vascular disease (macrophage activation syndrome), post-transplant, malignancies especially T-cell lymphomas (lymphoma associated HLH) and infections (infection associated HLH). [5] . Both familial and secondary HLH are usually precipitated by an immunological trigger which may be an infectious agent or a drug. Among the infectious agents viruses especially Ebstein-Barr virus and Cytomegalovirus (virus associated HLH) are most commonly implicated, but bacterial, fungal and parasitic infections have also been described [6, 7] . With the possible exception of visceral leishmaniasis, immunomodulation is indicated in most cases [8] . Mycobacterium tuberculosis has been related to haemophagocytic syndrome in case reports from the Indian subcontinent, often with high mortality despite aggressive immunosuppressive therapy [1, [9] [10] [11] . We report a case of haemophagocytic syndrome related to mycobacterial infection which was managed with steroids and IVIG with complete clinical and haematological response. The patient was a 2-year-old female with an unremarkable past, perinatal or family history. She was admitted with fever and diarrhoea of 2 days duration. She was managed with broad spectrum antibiotics, hydration and other supportive measures. High grade fever persisted along with progressive hepatosplenomegaly; on the 10th day of admission she developed ascites, respiratory distress and bilateral ptosis. Chest X-ray revealed bilateral pulmonary infiltrates suggestive of Acute respiratory distress syndrome. Peripheral blood counts revealed anaemia (7.6 gm/dl) and thrombocytopenia (87 9 10 3 /ll). Leucopenia (total leucocyte count 2.4 9 10 3 /ll, absolute neutrophil count 1.1 9 10 3 /ll) developed 4-5 days later. The coagulation profile was deranged with prolonged PT (32 s, INR 3.02) and APTT (39 s) in the absence of overt bleeding. D-dimer was positive. Serum triglycerides were 457 mg/dl, serum ferritin was 1,331 ng/ml and LDH was 1,889 IU/l. Bone marrow aspiration and biopsy revealed prominence of macrophages and histiocytes and phagocytosis of mature myeloid and lymphoid elements (Fig. 1 ). In addition, ELISA (IgM) for M tuberculosis was unequivocally positive at 1.08 U/ml (normal \ 0.90 U/ml) while IgG (0.18 U/ml, normal \ 0.90) and IgA (45.53 U/ml, normal \ 300) were negative, suggestive of acute Tubercular infection. Mantoux test was negative; tests for HBV, HCV and HIV were negative. Transaminases showed a twofold increase (AST 74 IU/l, ALT 87 IU/l) with normal bilirubin levels and normal renal function tests. Based on the fulfilment of 6/8 HLH-2004 criteria, namely fever, splenomegaly, cytopenias, hypertriglyceridemia, hyperferritinemia and bone marrow findings, a diagnosis of Haemophagocytic syndrome was made (Infection Associated HLH) [1] . Immunosuppressive therapy was initiated immediately after bone marrow studies. Methylprednisoslone (30 mg/ kg/day 9 3 days) followed by IVIG (1 gm/kg/day 9 2days) were used initially. HLH protocol was held in abeyance in the event of relapse of cytopenia or persistent fever. The patient was also exhibited anti-tubercular therapy consiting of isoniazid, rifampin, ethambutol and pyrazinamide. With the above treatment the patient responded rapidly; respiratory distress resolved within 24-48 h with resolution of radiological findings on follow-up X-ray chest. High grade fever settled within 24 h, organomegaly resolved over 7-10 days. Cytopenias also resolved over 4-5 days as did biological markers of Haemophagocytic Syndrome. The child was discharged on the 16th day of methylprednisolone and is on regular follow-up with no recurrence of symptoms and normal blood counts. HLH is a distinct clinical entity characterised by fever, pancytopenia, splenomegaly and haemophagocytosis in bone marrow, spleen, liver or lymph nodes. Laboratory investigations usually reveal high triglyceride and ferritin levels, impaired NK and cytotoxic T-cell function and low fibrinogen. It is a syndrome of macrophage activation, usually secondary to an immunological trigger, resulting in phagocytosis of mature and immature red cells, myeloid elements and platelets. In addition there is intense immune system activation causing release of inflammatory mediators IFNc, TNFa, IL-6, IL-10; Th-1 responses and organ system damage. The case described could well have been familial HLH, especially in view of age of onset. However, absence of a history of consanguinity, demonstration of recent mycobacterial infection and prompt response to treatment suggest infection associated HLH. The patient presented with prolonged fever complicated by organomegaly, cytopenias and ARDS, was investigated and treated promptly with good response to treatment. In a study of HLH in children a median age of onset of 17.4 months was described with average duration of fever ranging from 6 to 14 days. Our patient had onset at 24 months of age with fever duration of 10 days before developing symptoms. The patient was diagnosed and treated early at the 11th day of admission as against a median of 19 days described in Western literature [12] . Haemophagocytic syndrome related to childhood tuberculosis has been reported previously, in this patient the diagnosis remained presumptive based on the ELISA, positive family history and rapidity of response to ATT and immunosuppression [10] . Neurological signs described in HLH are encephalopathy, meningism, hypotonia, hemiplegia and seizures [2, 13] . Our patient developed bilateral ptosis which eventually resolved over 2-3 weeks. Phagocytosis, reportedly, most affects the red cells and platelets, however, in our case the majority of the ingested cells were of the myeloid and lymphoid lineages [14] . The classical picture of florid haemophagocytosis is usually not seen in the initial bone marrow and develops over the course of the illness. In this case bone marrow biopsy was not repeated as the parents were unwilling and blood counts rapidly normalised along with signs and symptoms. EBV infection markers were not available at this centre and hence not performed. Our patient had a favourable clinical outcome possibly due to early diagnosis and prompt initiation of specific treatment. A high index of suspicion is required for such cases as it may be an important cause of FUO [12] . Infection associated HLH related to tuberculosis is a treatable disorder with early immunosuppressive therapy.

Activation of the Canonical Bone Morphogenetic Protein (BMP) Pathway during Lung Morphogenesis and Adult Lung Tissue Repair

Signaling by Bone Morphogenetic Proteins (BMP) has been implicated in early lung development, adult lung homeostasis and tissue-injury repair. However, the precise mechanism of action and the spatio-temporal pattern of BMP-signaling during these processes remains inadequately described. To address this, we have utilized a transgenic line harboring a BMP-responsive eGFP-reporter allele (BRE-eGFP) to construct the first detailed spatiotemporal map of canonical BMP-pathway activation during lung development, homeostasis and adult-lung injury repair. We demonstrate that during the pseudoglandular stage, when branching morphogenesis progresses in the developing lung, canonical BMP-pathway is active mainly in the vascular network and the sub-epithelial smooth muscle layer of the proximal airways. Activation of the BMP-pathway becomes evident in epithelial compartments only after embryonic day (E) 14.5 primarily in cells negative for epithelial-lineage markers, located in the proximal portion of the airway-tree, clusters adjacent to neuro-epithelial-bodies (NEBs) and in a substantial portion of alveolar epithelial cells. The pathway becomes activated in isolated E12.5 mesenchyme-free distal epithelial buds cultured in Matrigel suggesting that absence of reporter activity in these regions stems from a dynamic cross-talk between endoderm and mesenchyme. Epithelial cells with activated BMP-pathway are enriched in progenitors capable of forming colonies in three-dimensional Matrigel cultures. As lung morphogenesis approaches completion, eGFP-expression declines and in adult lung its expression is barely detectable. However, upon tissue-injury, either with naphthalene or bleomycin, the canonical BMP-pathways is re-activated, in bronchial or alveolar epithelial cells respectively, in a manner reminiscent to early lung development and in tissue areas where reparatory progenitor cells reside. Our studies illustrate the dynamic activation of canonical BMP-pathway during lung development and adult lung tissue-repair and highlight its involvement in two important processes, namely, the early development of the pulmonary vasculature and the management of epithelial progenitor pools both during lung development and repair of adult lung tissue-injury.

Mammalian lungs are designed to optimize exposure of blood to oxygen. To achieve this, two intertwined and highly branched tree-like tubular systems, one conducting air and the other conducting blood must develop in a coordinated way to generate the millions of functional alveolar gas-exchange units [1, 2, 3] . Lung development in the mouse begins on embryonic day 9.5 (E9.5) when the lung primordium appears as a ventral bud in the primitive foregut [4] . Airway branching begins around E9. [5] [6] [7] [8] [9] [10] [11] [12] and continues through the ''pseudoglandular'' [E12-E16.5] and ''canalicular'' [E16.5-E17.5] stages. Thereafter, during the ''saccular'' stage [E17.5 to postnatal day 4 (P4)] the distal airways form the saccular units which are further subdivided by secondary septae formed during the alveolar stage (P4-P28 in mice) to form mature alveoli. This sequence of events is tightly controlled by the concerted action of growth factors, transcription factors, and mechanical forces [5, 6, 7] . Prominent role in the regulation of lung development and homeostasis is played by members of the Bone Morphogenetic Protein (BMP) family [8] . BMPs, like all other members of the TGFb superfamily, signal via specific membrane receptors that have serine-threonine kinase catalytic activity [9] . Functional BMP receptor units are composed of two Type-I and two Type-II receptor polypeptides. Four different Type-I BMP receptors (ALK2, ALK3/BMPRIa, ALK6/BMPRIb and ALK1), and three Type-II receptors (BMPRII, ActRIIA and ActRIIB) have been identified [10] . Upon ligand binding, the constitutively active Type-II receptors phosphorylate and thus activate their Type-I partners, which in turn phosphorylate their intracellular targets, the receptor-regulated Smad proteins 1, 5 and 8. Phosphorylated Smads form complexes with the ''common'' Smad4 and translocate to the nucleus where they regulate expression of their target genes, synergistically with other transcription factors [8, 11] . BMPs can also signal via Smad-independent intracellular pathways that involve mitogen-activated protein (MAP) kinases [12, 13] . Several studies using transgenic and conventional or conditional knock-out mice have clearly demonstrated the key role played by BMPs during early lung development [14, 15, 16, 17, 18, 19, 20, 21] . Disruption of BMP signaling by ectopically expressing the BMP antagonists noggin or gremlin in the lung epithelium [15, 22] , inactivating BMP receptors [16] or expressing a dominant negative form of the BMP Type-I receptor (dnALK6) result in abnormal distal lung architecture. Remarkably, over-activation of the BMP pathway is also incompatible with normal lung development. Ectopic over-expression of Bmp4 in the epithelium leads to smaller lungs and to substantially reduced epithelial cell proliferation [14] and mice with deletion of the BMP antagonist Follistatin-Like 1 (Fstl1) gene die at birth from respiratory distress and show multiple defects in lung development [23, 24] . Moreover, increasing evidence supports a key role for BMP signaling angiogenesis and vasculogenesis in the lung [25, 26] . A large number of genetically modified mice with lesions in genes encoding either BMP ligands, receptors or antagonists exhibit defective angiogenesis. The importance of BMP signaling for the vascular system is demonstrated by the association of mutations in genes encoding for BMPRII and ALK1 with the development of two genetic vascular diseases, namely, pulmonary arterial hypertension and hereditary hemorrhagic telangiectasia. The BMP signaling pathway has also been implicated in the regulation of adult lung homeostasis and tissue repair following injury [27, 28, 29, 30, 31] . Several BMP ligands have been found upregulated in allergen challenged lungs [27] and notably, ectopic expression of gremlin by adenovirus mediated gene transfer in the lung of adult rats causes severe pulmonary fibrosis [32] illustrating the importance of the BMP pathway for lung homeostasis as well. Consistently, over-expression of gremlin has been observed in humans suffering from Idiopathic Pulmonary Fibrosis (IPF) [28, 29] . Despite the significant progress, the precise mechanism(s) by which BMPs regulate lung development have not been fully delineated [5] . Moreover, characterization of the mechanisms by which BMPs affect adult lung homeostasis and tissue repair is still rudimentary. The current study was based on the premise that identification of the actual cellular targets of BMP mediated signaling during lung development, homeostasis and adult lung tissue repair will facilitate interpretation of earlier genetic studies, guide rational targeting of BMP signaling components in future experiments and contribute to the clarification of the mechanisms of action of this signaling system. Therefore, utilizing a mouse transgenic line carrying the eGFP reporter gene under the control of a canonical BMP pathway responsive regulatory element [33] we have constructed a detailed spatiotemporal map of canonical BMP signaling in the lung and identified putative cellular targets of this signaling pathway during lung development and adult lung tissue-injury repair. Our studies suggest that canonical BMP pathway plays key roles during early development of the pulmonary vasculature and the management of lung epithelial cell progenitors during late lung development and repair of adult lung tissue injury. Animals were housed in individually ventilated cages under specific pathogen free conditions, in full compliance with FELASA (Federation of Laboratory Animal Science Associations) recommendations, at the Animal House Facility of the Foundation for Biomedical Research of the Academy of Athens (Athens, Greece). All procedures for care and treatment of animals were approved by the Institutional Committee on Ethics of Animal Experiments and the Greek Ministry of Agriculture (Permit Number: K/1054). BMP-responsive eGFP expressing (BRE-eGFP) mice [33] kept in the C57BL/6 background maintained on a 12:12 hours light:dark cycle at the Animal Facility of the Foundation for Biomedical Research of the Academy of Athens. Collection of fetal lung tissues was carried out at the embryonic (E11 and E12) pseudoglandular (E13.5 and E14.5), canalicular (E17) and saccular (E19.5) stages. Noon of the day of the vaginal plug was considered embryonic day (E) 0.5. Additionally, tissues were collected on post-natal days 1 (P1), 15 (P15) and 2 months after birth (adult). All embryonic and P1 lungs were removed surgically, cleaned from the surrounding tissues and placed in 4% paraformaldehyde (Merck, 104004) in PBS for 24 hours at 4uC. For analysis of P15 and adult BRE mice the right lung-lobes were used for RNA/Protein isolation, whereas the left lungs were gently perfused with a mixture of 4% PFA:OCT (2:1) using a 20G catheter (Abbocath G717-A01, 4535-20), the bronchus was ligated and placed in 4% PFA for 24 hours at 4uC. Thereafter, the tissues were placed in PBS with 30% sucrose for 24 hours at 4uC, washed with PBS, embedded in OCT (Shandon Cryomatrix) and kept at 280uC until use. Cryostat sections 6-10 mm (Leica CM3050S) were placed on polylysine or SuperFrost Plus slides (Menzel Glaser), kept at room temperature for 3 hours with silica gel (Merck, 101969) and then stored at 280uC. The primary antibodies used for immunostaining were: goat anti-CC10 (Santa Cruz, SC9773), rat anti-PECAM1 (CD31) (BD-Biosciences, 550274), rat anti-VEGFR2 (eBiosiences 14-5821), rabbit anti-SpC (Chemicon, AB3786), rabbit anti-CGRP (Sigma-Aldrich, C8198), mouse anti-a smooth muscle actin-Cy3 (Sigma-Aldrich, C6198), Syrian hamster anti-T1a (DSHB 8.1.1), Sheep anti-BrdU (Fitzgerald, 20-BS17), mouse anti-Foxj1 (generous gift of Dr. S. Brody), mouse biotinylated anti-Muc5ac (Abcam, ab79082), anti-SM22alpha (Abcam, ab14106) and chicken anti-GFP (Abcam, ab13970). The secondary antibodies used were: donkey anti-goat Texas Red (Jackson Immuno-research, 705-076-147), donkey antichicken FITC (Jackson Immuno-research, 703-096-155), goat anti-rat AlexaFluor594 (Molecular Probes, A11007), goat anti-rat AlexaFluor647 (Molecular Probes, A21247), goat anti-Syrian hamster AlexaFluor568 (Molecular Probes, A21112), goat anti-Syrian hamster AlexaFluor647 (Molecular Probes, A21541), donkey anti-rabbit Texas-Red (Jackson Immuno-research, 711-076-152), donkey anti-sheep AlexaFluor568 (Molecular Probes, A21099) and donkey anti-rabbit Cy5 (Jackson Immuno-research, 711-176-152). The slides were mounted using the ProLong Gold Antifade Reagent with DAPI (Molecular Probes, P36931). Images were captured with a Leica DMRA2 fluorescent microscope equipped with a Leica DFC320 and DFC350 FX digital cameras and a Leica TCS-SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany). Image analysis was performed using Adobe Photoshop CS3, ImageJ 1.41 and Volocity LE. pSmad1/5/8 immuno-staining Tissue sections prepared as described in the previous section were treated 1.5% H 2 O 2 in MetOH for 30 min at RT in the dark to inactivate endogenous peroxidise. After blocking with 2% normal donkey serum (Sigma D9663) in TBS-Tween 0.3% for 2 hours at room temperature, the sections were incubated with rabbit anti-pSmad1/5/8 (Chemicon, ab3848) antibody overnight at 4uC in TBS-Tween 0.3% and then incubated for 1 hour at room temperature with the secondary anti-Rabbit HRP (Santa Cruz, SC2301). The slides were developed using DAB chromogen (DAKO K3468) development according to the manufacturer's instructions. Normal rabbit IgG fraction was used as control primary antibody (isotype control). Nuclei were counterstained with Mayer's Hematoxylene. Total RNA from mouse lungs was isolated using the Tri-Reagent protocol (Sigma, St. Louis, MO) and the yield and purity of RNA was determined electrophoretically and spectrophotometrically. After DNase treatment with RQ1 RNase-Free DNase (Promega, M160A), 2 mg of RNA were reverse transcribed into cDNA using M-MLV Reverse Transcriptase (Promega, M170B) and random primers (Invitrogen, 58875) according to the manufacturer's instructions. The primer pairs for real-time PCR were designed using Beacon Designer v7.01 software (Premier Biosoft International, Palo Alto, CA). Sequences for the primer pairs used are given in table S1. The PCR reactions contained SYBRH Green ER TM qPCR Super Mix Universal (Invitrogen, 11762-500), 200 nM of each primer (Invitrogen, Carlsbad), and 0.2 ml of cDNA template in a 20 ml reaction volume. RT-qPCR cycling parameters were initial incubation at 50uC for 2 minutes, denaturation at 95uC for 10 minutes followed by 50 cycles of 15 seconds at 95uC and 40 seconds at 60uC. Data collection was carried out using a Chromo4 Real-Time PCR detector (BioRad) and analyzed with Opticon Monitor 3 expression analysis software (BioRad). Relative levels of mRNA expression were calculated according to the DDCt method [34] . For protein extraction cells were homogenized in lysis buffer containing 50 mM Tris-Cl pH7.6, 150 mM NaCl, 1% (v/v) Triton 6100, 5 mM EDTA pH 8.0, 1 mM PMSF, 1:25 (v/v), Protease Inhibitor (Roche Applied Science, 11836153001), 2 mM Na 3 VO 4 , 5 mM NaF, 2 mM Sodium Pyrophosphate, 13.3 mM b-Glycerophosphate disodium salt. Homogenized cells were centrifuged at 13.000 rpm for 15 min at 4uC and protein samples were stored at 220uC until use. For detection of pSmad1/5/8 separated proteins were transferred to Immobilon-P membrane (Millipore, IPVH00010) and analyzed using rabbit anti-pSmad1/ 5/8 (Chemicon, ab3848, 1:1000) coupled with the anti-rabbit HRP (Santa Cruz SC2301, 1:5.000) and anti-b actin (Sigma-Aldrich, A5316, 1:5.000) combined with the anti-mouse HRP (Sigma-Aldrich, A4416, 1:40.000). Densitometry was done using the ImageJ 1.41 software. In-vivo lung tissue-injury/repair models Female BRE-eGFP animals, between 12-16 weeks of age were treated with naphthalene or bleomycin. For naphthalene treatment, 300 mg/kg naphthalene in corn oil (vehicle) or vehicle alone was administered by intra-peritoneal injection. Lung tissues were collected on days 2, 3, 6, 9 post naphthalene-treatment. For bleomycin treatment, bleomycin (0.033 Units/mouse, Nippon Kayaku co ltd, Bleocin) diluted in PBS or PBS alone (vehicle) was administered intra-tracheally in 30 ml total volume. Lungs were collected on days 2, 4 and 12 post-bleomycin-treatment. Ex-vivo culture of fetal lung explants Transgenic E12.5 lung explants were isolated and cultured, for 8 or 24 hours, in a 1:1 mixture of DMEM: F12 (GiBCO, 11039), supplemented with Insulin-Transferrin-Selenium (Gibco, 51300-044) and antibiotics. Explants were placed on Nuclepore Track-Etched Membranes (Whatman, 110414) and they were incubated in 4-well plates (Thermo Scientific, 144444). LDN193189 (Axon MedChem, Axon 1509) and SB431542 (Sigma-Aldrich, S4317) were used at a concentration of 10 mM. Untreated and vehicle (DMSO, Sigma-Aldrich, D2650) treated lungs were used as controls. In some experiments, E12.5 lung explants were treated with 100 ng/ml BMP4 (Peprotech, 120-05ET), or 250 ng/ml FGF10 (Peprotech, 100-26) or 500 nM Sonic Hedgehog (Shh) agonist-Purmorphamin (Calbiochem, 540220) for 8 hours. Endodermal buds and distal lung mesenchyme were isolated from E12.5 lungs as previously described [35, 36] . Briefly, isolated distal lung tips were treated for approximately 5 min with Trypsin 0.5% Pancreatin 2.5% in HBSS 16 (Gibco, 14170) on ice. Mesenchyme was removed using 26G needles and the epithelial buds as well as the isolated mesenchyme were placed in a (1:1) mixture of Growth Factor Reduced Matrigel (BD Biosciences, 356230) and DMEM: F12 (GiBCO, 11039) 10% FCS (Gibco, 10500), Pen Strep (Gibco, 15070, 1:100) and incubated for 8 hours at 37uC, 5% CO 2 . Adult primary lung cells were isolated as described in [37] with small modifications. Briefly, lungs were isolated, cut in small pieces and digested over-night at 4uC with 0.1% protease type-XIV (Sigma-Aldrich, P5149) in Joklik's MEM (Sigma Aldrich, 8028) containing antibiotics (Gibco, 15070). Lung pieces were transferred in Joklik's MEM supplemented with 10% FBS (Gibco 10500), pipetted gently several times to release cells and passed through 70 mm cell strainer (BD Biosciences, 352350). Cells were washed with MCDB-201 (Sigma-Aldrich, M6770) containing Insulin-Transferrin-Selenium (Gibco, 51300-044) and viability was measured using Trypan Blue. 10 6 cells (10 6 /ml) were cultured in 12-well plates (Nunc, 150628) coated with Bovine collagen (Nutacon, 5409) for 24 hours. Coating was done with 100 mg/cm 2 collagen, for 1 hour at 37uC. Thereafter, cultures were rinsed with MCDB-201 to remove detached cells and 500 ml MCDB-201 with Insulin-Transferrin-Selenium and 1 ng/ml EGF (Peprotech, 315-09) was added. After 1 week, colonies of epithelial cells were formed and infected with recombinant adenoviruses expressing either constitutively-activated ALK3 (caALK3), caALK5, Smad1 or Smad3. After 48 hours incubation the treated cells were used either for cytochemistry or for RNA extraction. P3 mice were anesthetized, the abdominal cavity was opened and ventricular vein and artery were cut. The chest was opened and the right atrium nicked. Lungs were perfused through the right ventricle of the heart with 1 ml ice-cold 16PBS using a 26G syringe, cut in small pieces using a razor blade and digested with the Enzyme mix [Elastase (Sigma-Aldrich, 45124): 3 U/ml, Collagenase: 0.1 U/ml, Dispase: 0.8 U/ml (Roche Applied Science, 11097113001) and DnaseI 50 mg/ml (Sigma-Aldrich, DN25] in PBS, at 37uC for 1 hour with rotation. Equal volume of HBSS ++ [HBSS 16 (Gibco, 14170) , supplemented with 2% FCS (Gibco, 10500), 0.1 M HEPES (Sigma-Aldrich, H0887) and 15% Cell Dissociation Buffer (Gibco, 1315-016) or EGTA 2 mM] was added. Cells were gently agitated several times, passed through a 40 mm cell strainer (BD Biosciences, 352340) and centrifuged 10 minutes at 1200 rpm, at 4uC. The supernatant was discarded and cells were re-suspended in HBSS ++ and counted with a hemocytometer. Specific subpopulations of lung cells were isolated using a BD FACS-Aria IIu cell sorter by staining the isolated lung cells as previously described by Teisanu et al. [38] with biotinylated anti-CD31 (BD Biosciences, 553371), anti-CD45 (Biolegend, 103103) and anti-CD34 (eBioscience, 13-0341-85) combined with Streptavidin APCCy7 (Biolegend, 405208), with anti-Sca1-Alexa Fluor647 (Biolegend, 108118) andanti-EpCam-PECy7 (Biolegend, 118216). DAPI was included for dead cell exclusion. The EpCam pos -CD31 neg -CD45 neg -CD34 neg cells were separated into Sca1 neg and Sca1 low and each of these two populations were separated by sorting into eGFP neg , eGFP low and eGFP high . Isolated cells were used either for further in-vitro cultivation, RNA expression or Western Blot analysis. In vitro cultivation of sorted epithelial cells in Matrigel was done as previously described [38] with minor modifications. Selenium (Gibco, 51300, 1:100), Pen Strep (Gibco, 15070, 1:100), amphotericin B (Gibco, 15290, 1:1000) and 10 mM SB431542 (Sigma-Aldrich, S4317). Culture medium was changed every other day until day 8. SB431542 was then removed and the cultures were maintained for an additional four days before fixing overnight at 4uC with 4% PFA in PBS. The Matrigel was removed by freezing-thawing and rinsing with fresh PBS. The content of the Transwell inserts was embedded in OCT (Shandon Cryomatrix) and 10 mm sections were prepared by cryostat (Leica CM1950). Although epithelial colonies developed in Matrigel even in the absence of SB431542 its presence resulted in higher number of larger colonies and therefore this culture system was preferred. Treatment of the Mlg cells with SB431542 for 24 hours led to substantial increase in FGF10 mRNA synthesis (data not shown) suggesting that stroma cell derived FGF10 may be part of the underlying mechanism. One way analysis of variance (One-way ANOVA) combined with ''Bonferoni's Multiple Comparison Test'', was done using GraphPad Prism. (*/ # ) P,0.05, (**/ ## ) P,0.01, (***/ ### ) P,0.001. Spatiotemporal pattern of expression of the BRE-eGFP reporter during lung development Previous studies utilizing two independent canonical BMP pathway reporter mouse lines, one harboring a BRE-eGFP [33] and another harboring a BRE-lacZ allele [39] have demonstrated expression of the BRE-eGFP and BRE-lacZ reporters respectively in the epithelium of trachea and primary-bronchi already at around E10-E12.5. To extend these findings and map in detail the spatiotemporal pattern of canonical BMP signaling in the lungs, we collected lung tissues from the BRE-eGFP reporter mice at embryonic days E11, E12, E13.5, E14.5, E16, E17.5, E19.5 and postnatal days 1, 15 and 2 months and analyzed tissue sections for eGFP expression. Our analysis demonstrated that the BRE-eGFP reporter was active already at E11, reached maximal levels around birth, thereafter declining to minimal levels in the adult ( Figure 1 and Figure S1 ). Around E11, eGFP expression was confined to the endothelial cells of the large pulmonary vessels, and the developing vascular network in the parenchyma. A very small fraction of sub-epithelial smooth muscle cells (SMCs) expressed very low levels of eGFP at E12 ( Figure S1 ). However, from E13.5 onwards robust eGFP expression was detected in sub-epithelial SMCs located at the proximal regions of the developing airways (Figures 1 and Figures S1 and S2) . From E16 onwards, expression of eGFP was detected in both proximal and distal epithelial compartments.In the proximal epithelium (i.e. the developing airways) eGFP expression exhibited a proximal-distal axis pattern, with high eGFP expression in bronchi and large airways and surprisingly no expression at the distal portion of the growing airway tree. Interestingly, isolated clusters of eGFP pos cells were observed at the tips of airway branch-points. In the distal lung compartment, eGFP expression was detected in cuboidal epithelial cells of the developing saccules ( Figure 1B ). As lung morphogenesis approached completion, the eGFP signal was reduced and in adult mice,(,2 months old), few eGFP pos cells were detected scattered in the epithelium of bronchi and large airways, in the alveolar compartment and in a subpopulation of cardiac cells in the outer layer of the tunica media of the pulmonary veins ( Figure 1B ) [40] . Although the capacity of BRE element to monitor BMP mediated transcriptional activation in-vivo has been demonstrated previously [33, 39, 41] , additional studies were undertaken to further increase confidence with respect to transgene expression in the lung. The kinetics of mRNA expression for eGFP and the BMP target gene Id1 were analyzed by quantitative PCR. EGFP and Id1 mRNA levels followed similar kinetics to eGFP protein expression, reaching maximum levels around birth ( Figure 2A ). P1 lungs were therefore enzymatically digested and lung cells were separated by cell sorting on the basis of eGFP expression into eGFP negative (eGFP neg ), eGFPlow (eGFP low ) and eGFP high (eGFP high ) cells ( Figure 2B ). Western blot analysis of protein extracts from the isolated population demonstrated good correlation between expression of eGFP and levels of pSmad1/5/8 ( Figure 2C , D). Analysis of the mRNA levels for eGFP and BMPregulated target genes such as Smad6, Id1 and Id3 demonstrated very good correlation between the level of eGFP expression and the activation of known BMP targets ( Figure 2E ). Interestingly, expression of Id2 mRNA did not correlate with levels of eGFP expression. Furthermore, treatment of embryonic lung explants exvivo with either BMP-or TGFb-receptor inhibitors resulted in either suppression or augmentation of BRE-eGFP transgene expression, respectively (see below Figure 3B ). Moreover, treatment of embryonic lung explants with FGF10, a known inducer of BMP4 production by distal endodermal cells, or recombinant BMP4 resulted in substantial up-regulation of eGFP expression ( Figure S3 ). Treatment of the lung explants with purmorphamin, a Shh agonist, resulted in redistribution of the eGFP staining pattern that involved a small increase of eGFP expression in a narrow zone of sub-mesothelial cells and a reduction of eGFP expression in the more central parts of the mesenchyme. This pattern of expression could stem from a Shh mediated reduction in FGF10 synthesis by mesenchymal cells [35] that in turn leads to a more confined BMP expression zone. Too further validate the BRE-eGFP reporter animals and assess its sensitivity in comparison to staining tissues for nuclear pSmad1/5/8 by immuno-histochemistry, we stained adjacent tissue sections of P1 lungs, the developmental stage demonstrating maximal BRE-eGFP reporter activity, with anti-eGFP and anti-pSmad1/5/8 antibodies. As shown in Figure S4 , nuclear pSmad1/5/8 was detected in regions with intense eGFP immuno-staining demonstrating consistency between BRE-eGFP reporter activity and pSmad1/5/8 immuno-staining ( figure S4 ). Unfortunately, in our hands the anti-pSmad1,5,8 reagents we have analyzed so far were unable to detect pSmad1/5/8 in lung tissues from embryos before stage E19.5 and thus we were not able to confirm correlation of eGFP and pSmad1/5/8 at earlier developmental stages. A number of studies have indicated that the value of measuring pSmad1/5/8 as the sole indicator for canonical BMP pathway activation is questionable since presence of pSmad1/5/8 does not necessarily lead to transcriptional activity of BMP target genes. Canonical BMP pathway activity is known to be regulated by cell type-specific Smad-interacting transcription factors, co-activators and co-repressors. For example Tbx1 can bind to Smad1, interfere with Smad1/Smad4 interaction and decrease BMP transcriptional activity without affecting the amount of pSmad1/5/8 in the responding cell [42] . Furthermore, Leeuwis et al. [41] demonstrated that in the kidney of BRE-eGFP animals, whereas the expression pattern of BRE-eGFP exhibited excellent correlation with the expression of the BMP7 and the BMP target genes Id1, and Smad6, in contrast pSmad1/5/8 exhibited a brought expression pattern that did not correlate with the expression of BRE-eGFP, BMP7, Id1 or Smad6. Therefore, a safer assessment of canonical BMP pathway should combine analysis of several known BMP target genes and pSmad1/5/8 immuno-stainings when this is technically possible. Collectively, the analysis by Monteiro et al. [33] , Leeuwis et al. [41] and the results described herein indicate that the BRE-eGFP transgenic line reports quite faithfully activation of the canonical BMP pathway in-vivo and ex-vivo and thus it provides an additional method for assessing potential canonical BMP pathway activation. Canonical BMP-pathway is activated in the developing vascular and sub-epithelial smooth muscle networks during early lung development To identify the putative cellular targets of canonical BMP signaling, tissue sections collected from lungs at different developmental stages were stained with antibodies specific for appropriate cellular markers. Double immunostaining for eGFP and CD31 (PECAM-1), a surface marker of mature endothelium, or VEGFR2, a surface marker of vascular progenitor cells ( Figure S1 ), demonstrated that the majority of CD31 pos or VEGFR2 pos cells at E12, E14.5 and E17 were also eGFP pos ( Figure 3A ), confirming that the network of eGFP pos cells in the parenchyma is composed of vascular cells and that the canonical BMP pathway is activated during the establishment of the vascular network in the developing lung. At P1, ,40% of the CD31 pos cells were eGFP pos , from P15 onwards the number of eGFP pos -CD31 pos cells declined and in the adult they were undetectable. EGFP expression with similar kinetics was observed in the endothelial cells lining the large blood vessels ( Figure 3A ). Very few eGFP pos alpha smooth muscle actin (aSMA) double positive cells were found in the walls of the vessels ( Figure 3A , white arrow in P1 vessel). To validate the functional significance of BMP signaling for the development of the vascular network of the developing lung, ex-vivo cultures of E12 lung explants were treated with the respective BMP-or TGFb/Activin-receptor inhibitors LDN193189 and SB431542. Treatment of lung explants for 24 hours with SB431542 led to a substantial increase in eGFP, Id1, and FGF10 mRNA expression, whereas treatment with LDN193189 led to a decrease ineGFP, Id1 and FGF10 mRNA expression and an increase, probably compensatory, in BMP4 and VEGFa mRNA expression ( Figure 3B and D). Remarkably, even 8 hours of treatment with the SB431542 led to an increase in the density of the CD31 pos network and a more pronounced increase in the eGFP pos network, whereas, similar treatment with LDN193189 led to reduction in the density and connectivity of the eGFP pos and CD31 pos cells ( Figure 3C) . Collectively, these findings indicate that during the pseudoglandular stage, in the course of which most of the stereotypical branching morphogenesis is thought to take place, an important target of canonical BMP signaling in the lung is the developing pulmonary vasculature. Strong eGFP expression was detected in the developingsubepithelial SMC layer of the proximal airways from E14.5 to P1 ( Figure 3A and Figures S1 & S2) . The eGFP expression, which interestingly involved only a portion of the sub-epithelial SMCs, was undetectable by P15. It has been proposed that sub-epithelial SMCs originate from a pool of sub-mesothelial, FGF10-expressing, mesenchymal cells. Under the influence of several signalling pathways, such as those mediated by FGF, BMP, Shh, and Wnt, these cells proliferate and passively translocate to surround the proximal portion of the developing airway tree [35, 43, 44, 45, 46] . http://www.sciencedirect.com/science/article/pii/S0012160611 010189 -bb0125To map more precisely the temporal activation of the BRE-eGFP reporter during development of the subepithelial SMC compartment we isolated P11, P12 and P13.5 lungs from the BRE-eGFP reporter animals and analysed coexpression of eGFP and SM22a, a marker of both primitive and mature SMCs [47] , or aSMA a marker of more differentiated SMCs. As shown in Figure S1B , SM22a expression was observed in a sub-epithelial population of cells around the developing proximal airways already at embryonic stage E11, aSMA expression was evident only in the most proximal portion of the SM22a pos zone and eGFP expression was clearly detected in the aSMA pos zone only after embryonic stage E13.5 (very few aSMA pos cells expressed very low levels of eGFP at E12). These findings are consistent with the notion that canonical BMP signalling does not involve primitive SMCs but rather mature aSMA expressing sub-epithelial SMCs. Canonical BMP pathway activation in developing airway epithelial cells Although strong BRE activity was detected in the trachea and primary bronchi as early as E10-E12.5 [33] , consistent with the demonstrated role of BMP signaling in trachea development [48, 49] , BMP-driven eGFP-expression was not detected in the epithelial compartments of the interlobular airways before the canalicular stage of lung development. As shown in Figures 1 and 4A , minimal eGFP expression was detected around E14.5 in the large airways, and only after E17 strong eGFP signal was detected in cartilaginous and large airways following a proximal-distal axis pattern, with more intense signal in the proximal part of the bronchial tree. Immunostaining of BRE-eGFP lung tissue sections with an antibody specific for the Clara cell specific differentiation marker Scgb1a1/CCSP/CC10 (hereafter referred to as CC10), revealed expression of the eGFP reporter from E17.5 and onwards, in luminal cells of the developing airways. Strong CC10 expression was detected around E19.5, approximately two days after the onset of eGFP expression in the airway epithelium ( Figure 4A ) CC10 and eGFP exhibited partially overlapping zones of expression, with eGFP being highly expressed in the proximal portion and CC10 highly expressed in the distal portion of the developing airway tree. A zone of eGFP pos -CC10 pos cells appeared to separate the CC10 pos epithelial population from the eGFP pos airway regions. Postnatal expansion of CC10 pos regions in the conducting airways was accompanied by a gradual reduction of the eGFP pos regions ( Figure 4A ). Interestingly, regions of airways exhibiting enhanced eGFP expression were detected close to the tips of airway branch-points in close proximity to neuro-epithelial bodies (NEBs). Analysis of the eGFP expression in relation to CC10 and CGRP, a differentiation marker of neuro-epithelial cells, revealed that eGFP pos -CC10 low cells formed ''caps'' over clusters of CGRP pos cells. From E19.5, the eGFP pos -CC10 low ''cap'' cells (green-arrows in Figure 4B ) were separated from the eGFP neg -CC10 pos cells by a zone of double positive epithelial cells (yellow-arrows in Figure 4B) . Notably, the zone of eGFP pos -CC10 pos cells coincided with zones richin NEBs ( Figure S5 ). From P15 and beyond, the only NEBassociated eGFP pos cells were CGRP pos neuro-epithelial cells (blue-arrows in Figure 4B ). As the majority of eGFP pos cells did not express the marker for secretory airway epithelium (CC10), their possible relationship to ciliated airway epithelial cells was investigated by staining with antibodies for Foxj1 a ciliated cell specific differentiation marker. As shown in Figure S6 , a remarkably small portion of eGFP pos cells were FoxJ1 pos (yellow-arrows in Figure S6 ). Two alternative models have been proposed regarding the role of BMP signalling during lung branching morphogenesis. One model suggests that mesenchymal derived FGF10 induces Bmp4 expression by the distal bud epithelium to limit, in an autocrine manner, Fgf10-mediated bud outgrowth [45, 50] . The other model, based on the differential response of the distal epithelial buds to BMP4 in the presence or absence of mesenchyme, proposes that Bmp4 produced by the distal epithelium acts in an autocrine manner to limit proliferation of distal epithelial buds, and in a paracrine manner, on the adjacent mesenchyme, to induce production of a mesenchymal signal that enhances proliferation of distal epithelial buds [51] . Thus, according to the latter model the negative or positive effect of BMP signalling on branching is the outcome of a dynamic balance between negative autocrine and positive paracrine mechanism. The absence of eGFP reporter activation in the distal buds of the branching airway tree (Figures 1 and 4) prompted us to investigate whether this is the result of an intrinsic inability of the distal epithelium to activate the canonical BMP pathway or, in accordance to the second model, the result of a dynamic interplay with the adjacent mesenchyme. Therefore, mesenchyme free distal epithelial buds and epithelial free mesenchyme were isolated as previously described [35, 36] and cultured in Matrigel. Remarkably, within 8 hours of incubation, the BRE-eGFP reporter was strongly up-regulated in the mesenchyme-free epithelial buds and conversely it exhibited a tendency for decline in the isolated mesenchyme ( Figure 5 ). These findings are consistent with the model of Bragg et al. [51] and moreover suggest that the elusive, mesenchyme derived, epithelial growth promoting signal could simply act by negatively regulating activation of the canonical BMP pathway in the distal epithelium buds. Collectively, the above findings demonstrate that activation of the canonical BMP pathway in airway epithelial cells coincides with the beginning of the canalicular stage, when the developmental plan of the lung shifts from branching morphogenesis to the development of distinct respiratory epithelial cell compartments [52] . Keeping in mind that gas conducting airways and blood conducting vessels need to develop in a coordinated manner and be properly juxtaposed to support optimal gas-exchange in the adult, it is understandable that the tips of developing airways should reciprocally exchange growth, differentiation or guidance signals with the developing vasculature and other mesenchymal cells to ensure coordinated development and spatial integration of these two branching systems. The finding that disruption of the vascular network in growing lung explants affects branching morphogenesis by reducing primarily ''orthogonal bifurcations'' of the epithelial tubules [53, 54] and the findings presented herein demonstrating the influence of the adjacent mesenchyme on the activation of the canonical BMP pathway reporter ( Figure 5 ) illustrates the importance of the crosstalk between epithelial and vascular cells during early lung development. It should be pointed out that due to our failure to successfully apply anti-pSmad1/5/8 immuno-histochemistry in tissues from embryos earlier than stage E19.5, we cannot conclusively rule out for the moment activation of the canonical BMP pathway at the tips of the early branching endoderm. Further studies will be required to assess whether the BRE-eGFP reporter is not active in the distal endoderm because: a) non-canonical BMP pathways are primarily involved; b) the canonical BMP pathway is active however, does not reach the activation threshold required to activate the BRE-eGFP reporter; or c) that the dynamic balance between the mesenchyme and endoderm which keeps the BRE-eGFP reporter inactive most of the time (Figure 5 ), allows at critical points along the branching process transient spikes of canonical BMP pathway signaling to reach the nucleus and regulate expression of target genes. Over-expression of BMP ligands or inactivation of BMP receptors in the distal part of the developing lung leads to dramatic defects in alveolar development [14, 16, 18, 19] . Since eGFP expression was highly increased in lung parenchyma of the BRE-eGFP mice during the saccular and alveolar stages, the period during which functional alveoli are formed, we characterized the nature of the eGFP pos cells in this part of the developing lung. Tissue sections from different developmental stages were analyzed by co-staining for eGFP and pro-surfactant protein C (pro-SpC), a marker of type-II pneumocytes. As shown in figure 6 and Figure S2 , eGFP staining during the pseudoglandular stage was confined to endothelial cells and peri-bronchial smooth muscle cells of the proximal airways. Pro-SpC pos epithelial cells in the distal airways were uniformly eGFP neg at this developmental stage ( Figure S2 ).In contrast, starting at the canalicular stage (E17) and persisting until approximately P15, a population of eGFP pospro-SpC pos cells emerged in the distal lung compartments ( Figure 6A ).The SpC pos cells at this stage of development are thought to be immature type-II pneumocytes [55] . Morphometric analysis demonstrated that during the saccular and alveolar stages, approximately half of the pro-SpC pos alveolar cells were eGFP pos . Adult tissues, in which alveolarization was complete, contained a very small number of eGFP pos -pro-SpC pos alveolar cells (,0.5-1%) ( Figure 6B ). High resolution confocal microscopy analysis revealed that at all the developmental stages analyzed only a very small portion of type-I pneumocytes were eGFP pos ( Figure S7 ). Interestingly, ectopic over-expression of a constitutive ALK3 in cultures of adult primary pulmonary cells resulted in increased pro-SpC (,6-fold) and Id1 (,15 fold) mRNA levels, in comparison to the untreated or ALK5ca treated groups ( Figure 6C ) further implicating canonical BMP-signaling in the physiology of type-II pneumocytes. The sequential appearance of eGFP neg -SpC pos (immature type-II pneumocytes), eGFP pos -SpC pos and eGFP neg -SpC pos (mature type-II pneumocytes) may suggest that the eGFP pos -SpC pos population is an intermediate step through which immature type-II pneumocytes must pass before developing into mature functionally competent type-II pneumocytes. This conclusion is compatible with the finding that mice with canonical BMP pathway disrupted due to deletion of the Smad1 gene accumulate in their developing lungs high numbers of periodic acid-Schiff (PAS) positive immature type-II pneumocytes [19] . The above findings indicate that during the cannalicular stage, an important target of canonical BMP signaling in the distal lung compartment is the developing type-II pneumocyte population. This could explain the dramatic distal lung phenotype of animals with disrupted BMP signaling in the distal lung epithelium [15, 16, 18, 19] . It is very intriguing that with the beginning of the canalicular stage, when the developmental plan of the lung shifts from branching morphogenesis to the development of distinct respiratory epithelial cell compartments [52] , the BMP-responsive eGFP reporter is activated in two epithelial domains, namely, the proximal airways and the distal developing alveolar sacs. The conclusion that canonical BMP signaling becomes crucial for lung epithelial cell development at the cannalicular stage is consistent with earlier genetic studies demonstrating that deletion of Bmpr1 [16, 18] or Smad1 [19] and ectopic expression of negative regulators of BMP signaling such as XNoggin and dominantnegative ALK6 [15] by SpC-promoter-driven expression, resulted in defects in lung development visible only around E15-E16, i.e. after completion of the branching morphogenesis. The localization of eGFP pos cells in the vicinity of NEBs and the minimal overlap between cells expressing eGFP and markers for specialized secretory (CC10) or ciliated cells (FoxJ1) prompted us to investigate whether active canonical BMP signaling may define cells with an immature progenitor phenotype. Thus, a single two hour pulse of BrdU was given to pregnant mothers at E17 to label cycling epithelial cells which were subsequently detected by staining with anti-BrdU antibodies. BrdU pos cells accounted for ,6.7% of the surface airway epithelium in conducting airways and were uniformly distributed throughout the airway tree ( Figure 7A ). This pattern is compatible with the findings of Giangreco et al. [56] and Rawlins et al. [57] demonstrating that randomly distributed progenitor cells contribute to normal epithelial homeostasis. Of the BrdU pos cells ,77.8% were eGFP neg and ,22.2% were eGFP pos , indicating that both populations contribute to the expansion of the airway epithelial compartment ( Figure 7B ). Despite the dominance of the eGFP neg -BrdU pos cells (in absolute numbers), ,13.4% of the eGFP pos were BrdU pos, compared to ,5.5% for the eGFP neg cells ( Figure 7C ). To further compare the proliferative potential of the eGFP pos and eGFP neg cells different epithelial populations were isolated from dissociated lung tissue by cell sorting and analyzedin-vitro for colony-forming ability. Following a previously described protocol [38] , EpCAM pos , Lineage negative (i.e. CD45 neg -CD31 neg -CD34 neg ) cells from BRE-eGFP transgenic animals were subdivided into Sca1 neg and Sca1 low cells. Finally the Sca1 neg and Sca1 low cells were separated by sorting into eGFP neg , eGFP low and eGFP high cells respectively ( Figure 7D ). Isolated cells were cocultured with MLg feeder cells in a 3-dimensional Matrigel culture system [38] with a modification developed by some of us (HC and BRS) that involved co-culture of the cells in Matrigel for eight days in the presence of the TGFb receptor inhibitor SB431542 (10 mM), followed by an additional four days in the absence of inhibitor. Lin neg -EpCAM pos -Sca1 low cells cultured in this way, in agreement with previous reports [58] , gave rise primarily to large cystic epithelial colonies and Lin neg -EpCAM pos -Sca1 neg cells gave rise almost exclusively to smaller compact epithelial colonies ( Figure 7E ). The clonogenic capacity of both Sca1 low and Sca1 neg cells correlated with the expression levels of the BRE-eGFP reporter with the corresponding eGFP high populations exhibiting highest clonogenic capacity ( Figure 7F ). Previous studies have suggested that cystic colonies derived from Lin neg -EpCAM pos -Sca1 low cells contain progenitors of airway epithelial cells (Muc5Ac pos ), whereas, the compact colonies derived from Lin neg -EpCAM pos -Sca1 neg cells contain primarily SpC pos alveolar progenitors [58] . Consistently, as shown in Figure 7G , whereas all compact colonies derived from the Sca1 neg subpopulations exhibited a similar SpC pos -Muc5Ac neg staining pattern, the large cystic colonies derived from the Sca1 low subpopulations exhibited a more heterogeneous pattern of staining with approximately half of them expressing Muc5Ac and half of them expressing SpC. It is worth noting that adult Lin neg -EpCAM pos -Sca1 neg epithelial cells have very low clonogenic capacity in Matrigel [38] . The development of colonies fromLin neg -EpCAM pos -Sca1 neg cells in the current study is most probably due to the fact that the analysis was done with P3 lungs which were still undergoing intense alveolar development. Previous studies have provided evidence for two types of airway epithelial progenitors, one randomly distributed in the airways that maintains normal epithelial homeostasis and responds to minor injuries of the epithelium, and a second, associated with previously described progenitor cell niches that can renew depleted Clara cells upon severe epithelial injury [38, 56] . The randomly distributed eGFP neg -BrdU pos cells detected in the developing lungs ( Figure 7A ) could represent the former type whereas the eGFP pos epithelial cells could represent the latter type of progenitors. Once lung development is completed, the homeostatic, low turnover, maintenance of the epithelial population can be supported by the distributed Clara cells [38, 56] and thus active BMP signaling may not be required any longer. The decline in eGFP expression in the lungs of the adult BRE-eGFP transgenics may reflect just that. Reactivation of the BRE-eGFP reporter in adult lung after tissue injury suggests an active role of BMP-signaling in adult lung repair after injury The dramatic decline of eGFP expression in the adult lung of the BRE-eGFP reporter mice and the association of eGFP expression with the epithelial progenitor pool prompted us to investigate whether the reporter was reactivated in adult lung during injury and repair. Adult BRE-eGFP animals were analysed in two established models of lung tissue injury and repair, namely the naphthalene and bleomycin induce lung injury models. Naphthalene treatment causes selective depletion of Clara cells in the airways [59] . In contrast, bleomycin causes epithelial cell death, acute inflammation and fibro-proliferative remodeling of the peripheral lung [60] . Immuno-staining for CC10 and CGRP demonstrated that by day 3, naphthalene treatment had caused a large reduction in CC10 pos cells in the airways, an increase in the number of CGRP pos cells per NEB, and a substantial increase in BRE-eGFP reporter activity at airway branch-points and terminal bronchioles ( Figure 8A ).This pattern of CC10 localization and eGFP reporter expression corresponds with previously determined anatomic locations that maintain chemically resistant epithelial progenitor cells [61, 62, 63, 64] . Analysis at different time points after naphthalene-treatment demonstrated a gradual increase in the number of eGFP pos bronchial epithelial cells that was evident on day 2 and maximum increase of 15-fold over control levels by day-9 post-naphthalene treatment ( Figure 8B ). Up to day-2, the increase in eGFP pos airway epithelial cell numbers involved CGRP pos and CGRP neg -CC10 lowcells ( Figures 8C and D) . Thereafter, although the total number of eGFP pos -CGRP pos cells remained stable from days 2-9, a wave of eGFP pos -CC10 low -CGRP neg cells that reached maximum around day 3 was followed by a wave of eGFP pos -CC10 pos cells that reached plateau around day-9 ( Figure 8D) . Interestingly, the distribution of eGFP and CC10 expression recapitulated the pattern observed during early lung development i.e. eGFP pos -CC10 pos cells separating eGFP pos -CC10 low from apparently newly formed eGFP neg -CC10 pos cells. The kinetics of appearance and relative localization of these cells within the tissues are consistent with the notion that they may represent sequential stages of development from the eGFP pos -CC10 low to the eGFP neg -CC10 pos phenotype. Bleomycin is an anti-cancer drug, which induces lung injury and fibrosis [60] . As shown in Figures 9A and 9B , treatment of adult BRE-eGFP animals by intra-tracheal administration of bleomycin resulted in substantial reactivation of the eGFP reporter in type-II pneumocytes (Pro-SpC pos ). During the first days of treatment the eGFP pos cells were primarily cuboidal SpC pos cells. From day-4 onwards both cuboidal and squamous eGFP pos cells were found in the alveolar regions. Interestingly, the squamous eGFP pos cells were T1a-expressing type-I pneumocytes and were always localized in the borders between affected and apparently normal tissue ( Figure 9C ). We postulate that the cuboidal eGFP pos cells observed in the early stages of the bleomycin induced response represent a reparatory progenitor population that expands and differentiates into the eGFP pos -T1a low cells observed in the later stages to repair the damaged alveolar epithelium. Further studies will be required, however, to validate this hypothesis. Collectively, our findings demonstrate that the canonical BMP pathway is re-activated during adult lung injury in a manner that bears resemblance to the activation of this pathway during early lung development, strongly supporting its importance during adult lung tissue injury repair. Utilizing a transgenic reporter mouse line harboring a BMPresponsive eGFP reporter allele we were able to construct a detailed spatiotemporal map of canonical BMP signalling during early lung development and adult lung tissue injury repair. Our studies demonstrated that during the pseudoglandular stage, when branching morphogenesis characterises lung development, canonical BMP pathway is active mainly in the vascular network and the airway smooth muscle layer. Only after the completion of branching morphogenesis and the initiation of epithelial cell differentiation canonical BMP pathway activation commences in airway and alveolar epithelial cell that are enriched in progenitors that can form epithelial colonies in Matrigel in-vitro. The BRE-eGFP reporter is reactivated in mesenchyme-free distal epithelial buds cultured in Matrigel suggesting that the low reporter activity in intact lungs is the result of a dynamic interplay between the endoderm and the mesenchyme and moreover suggest that the elusive, mesenchyme derived, epithelial growth promoting signal postulated by Bragg et al. [51] may act by regulating negatively activation of the canonical BMP pathway in the distal epithelium buds. BRE-eGFP expression, in agreement with earlier reports describing the temporal expression pattern of some BMP pathway components during late lung development [65] , peaks around birth and returns to very low levels upon completion of lung development. Remarkably, severe depletion of Clara cells in the adult lung by naphthalene treatment, leads to re-expression of the eGFP reporter around NEBs and terminal bronchioles, areas known to harbour airway progenitor cells. Likewise, injury of the alveolar epithelium in the adult lung by bleomycin treatment, leads to re-expression of eGFP initially in SpC pos cuboidal epithelial cells, which we postulated to be alveolar epithelial progenitors and subsequently in squamous T1a low cells which we postulated to be derived from the former population and on the way to differentiate to Type-I pneumonocytes. Our findings are compatible with the notion that during the late stages of lung development the role of canonical BMP signalling pathway is to maintain the undifferentiated state of the airway and alveolar epithelial progenitors preventing premature exhaustion of their pools and securing the continuous supply of the differentiated epithelial cells that are required for a fully developed lung. The reactivation of the BMP responsive reporter in adult animals undergoing repair of severe epithelial injury, where the same requirements regarding proper management of the undifferentiated epithelial progenitor pools apply, is also compatible with this conjecture. An analogous function has been previously proposed for BMP signaling regarding the maintenance of the undifferentiated state of pluripotent mouse embryonic stem (ES) cells [66, 67] and the maintenance of the undifferentiated, multipotent state of distal lung progenitor cells [45, 68] . The construction of a spatiotemporal map of canonical BMP signalling during lung development and adult lung tissue injury repair (summarised in Figure 10 and table S2) will facilitate interpretation of earlier genetic studies and guide rational targeting of BMP signaling components in future experiments. Moreover, the ease isolation of living eGFP pos i.e. BMP responding, subpopulations of lung cells by cell sorting will greatly facilitate definitive clarification of the mechanisms of action of this signaling system during early lung development and repair of lung tissue injury in the adult. Figure S3 FGF10, BMP4 and a Sonic Hedgehog agonist affect expression of the BRE-eGFP reporter. A) Whole E12 lung-explants were cultured on Nuclepore membranes for eight hours in the presence of vehicle, FGF10 (250 ng/ml), BMP4 (100 ng/ml) or Sonic Hedgehog agonist Purmorphamin (500 nM). Upper panel shows bright field images of representative explants. Lower panel shows eGFP expression in the same explants. B) Confocal images of tissue sections prepared from the explants described above stained for eGFP (green stain). Nuclei were counterstained with DAPI. Note the strong upregulation of eGFP expression in the FGF10 and BMP4 treated explants, and the Figure S4 Correlation between pSmad1/5/8 and BRE-eGFP immune-staining in P1 lung tissue sections. Adjacent tissue section of lungs from BRE-eGFP reporter animals were stained with anti-GFP or anti-pSmad1/5/8 antibodies as described in materials and methods. A) Staining of adjacent sections of a large airway with anti-GFP and anti pSmad1/5/ 8antibodies or normal rabbit IgG fraction as isotype control. B) Representative images from the indicated tissue regions demonstrating that tissue areas with intense BRE-eGFP expression coincide with regions exhibiting pSma1/5/8 immuno-staining. The scale bar corresponds to 25 mm. (TIF) Figure S5 The zone of eGFP pos CC10 pos cells coincides with the NEB-rich portion of the airway tree. A) Confocal image of a lung tissue section derived from a P1 BRE-eGFP animal stained for eGFP (green staining), CC10 (red staining) and CGRP (white staining). The images illustrate the zone in the airway-tree where co-expression of eGFP and CC10 occurs. The lower images, showing the CGRP and DAPI channels of the upper images, illustrate the coincidence of eGFP pos -CC10 pos zone with the NEB-rich regions of the airway tree. B) Confocal images of the transitional zone between the eGFP pos -CC10 neg and eGFP neg -CC10pos domains of the airways demonstrating the preferential association of eGFP pos -CC10 low cells with NEBs. The image depicting NEBs is obtained from a section sequential to the ones depicting CC10 and eGFP expression. Nuclei were counterstained with DAPI (blue staining). (TIF) Figure S6 Minimal eGFP expression in the Foxj-1 pos airway ciliated cells. Representative confocal images demonstrating minimal activation of the BRE-eGFP transgene in FoxJ1 pos ciliated airway epithelial cells. Tissue section collected from E17.5, E19.5, P1, P15 and adult BRE-eGFP transgenic lungs were stained for eGFP (green staining), FoxJ1 (red staining) and CGRP (white staining). Nuclei were counterstained with DAPI (blue staining). The images demonstrate the presence of remark-ably low number of eGFP pos -FoxJ1 pos epithelial cell (depicted with yellow arrows). (TIF) Figure S7 A small, however, detectable number of BRE-eGFP developing type-I pneumocytes (T1a pos ) express eGFP. Representative confocal images of lung sections of E14.5, E17.5, E19.5, P1, P15 and adult BRE-eGFP transgenic lungs stained for eGFP (green staining), CD31 (red staining) and T1a (white staining). Nuclei were counterstained with DAPI (blue staining). The Scale bars are 20 mm. The images demonstrate that whereas the majority of the developing endothelial cells express eGFP, only a small number of type-I pneumocytes, shown in the inserts of the P1 and P15 image, are eGFP pos . (TIF)

Dengue Virus Serotype 2 Blocks Extracellular Signal-Regulated Kinase and Nuclear Factor-κB Activation to Downregulate Cytokine Production

BACKGROUND: Dengue virus (DENV) infection is the most common mosquito-borne viral disease threatening human health around the world. Type I interferon (IFN) and cytokine production are crucial in the innate immune system. We previously reported that DENV serotype 2 (DENV-2) induced low levels of interferon regulatory factor 3 and NF-κB activation, thus leading to reduced production of IFN-β in the early phase of infection. Here, we determined whether DENV infection not only hampers type I IFN activation but also cytokine production triggered by Toll-like receptor (TLR) signaling. METHODOLOGY/PRINCIPAL FINDINGS: We used quantitative RT-PCR and found that only low levels of IFN-β and inflammatory cytokines such as interleukin 10 (IL-10), IL-12 and tumor necrosis factor α (TNFα) mRNA were detected in DENV-2–infected bone-marrow–derived dendritic cells. Furthermore, DENV-2 infection repressed cytokine production triggered by TLR signaling. To elucidate the molecular mechanisms underlying this suppression event, we measured NF-κB activation by p65 nuclear translocation and luciferase reporter assay and found that NF-κB activation triggered by TLR ligands was blocked by DENV-2 infection. As well, extracellular signal-regulated kinase (ERK) activity was suppressed by DENV-2 infection. CONCLUSIONS/SIGNIFICANCE: To downregulate the host innate immunity, DENV-2 by itself is a weak inducer of type I IFN and cytokines, furthermore DENV-2 can also block the TLR-triggered ERK–NF-κB activation and cytokine production.

Dengue virus (serotypes DENV-1, -2, -3 and -4) is a positivestrand RNA virus belonging to the family Flaviviridae, genus Flavivirus. Mosquitoes transmitting DENV in humans has been a major cause of dengue diseases in tropical and subtropical counties; approximately one-third of the world's population is at risk of the infection. People infected with DENV typically show self-limited febrile dengue fever and dengue hemorrhagic fever (DHF). Life-threatening dengue shock syndrome (DSS) is more likely to occur after a second DENV infection [1, 2] . Other than supportive treatments, no specific therapy is available for denguerelated diseases. Several tetravalent DENV vaccine candidates are under development, but an effective, safe and affordable dengue vaccine remains elusive [3] . DENV infects multiple organs and cell types in humans. Particularly, the mononuclear phagocyte lineage of macrophages, monocyte-derived dendritic cells (DCs) and skin Langerhans cells are the primary cell targets [4, 5, 6] . Similar cellular tropism of macrophages and DCs was observed in the experimental DENV mouse infection model [7] . Macrophages and DCs are the most crucial cell types in innate immunity and rapidly produce type I interferons (IFNs) and cytokines to fight against microbe invasion. Type I IFNs are potent inhibitors of virus replication. Therefore, many pathogenic viruses have developed strategies to escape the IFN-triggered anti-viral effects. More than 170 different virusencoded IFN antagonists from 93 distinct viruses have been described [8] . For example, hepatitis C virus (HCV), a member of Flaviviridae, evades innate immunity by cleaving mitochondrial antiviral signaling protein, an IFN stimulator, with its protease NS3/4A [9] . The nonstructural proteins NS4B of DENV-2, West Nile virus (WNV), and yellow fever virus (YFV) block the activation of STAT1 in cells stimulated with type I IFN [10, 11] . We previously found that two flaviviruses, Japanese encephalitis virus and DENV-2, trigger type I IFN transcription through an RIG-I-dependent signaling cascade to activate interferon regulatory factor (IRF) and NF-kB. However, JEV induced higher activation of IRF3 and NF-kB than DENV-2 in human A549 cells [12] . Furthermore, type I IFN production triggered by doublestranded RNA (dsRNA) stimulation was blocked in DENV-2infected human DCs [13, 14] . Microarray results from rhesus macaques also indicated that type I IFN, interleukin 10 (IL-10), IL-8, IL-6 and tumor necrosis factor a (TNFa) were not upregulated with DENV-1 infection [15] . Therefore, DENV-2 might modulate the induction pathway of type I IFN and other cytokines. The Toll-like receptor (TLR) family is one of the best-studied pattern-recognition receptor families and is responsible for sensing invading pathogens [16] . Different TLRs recognize the different molecular patterns of microorganisms; for example, TLR4 recognizes lipopolysaccharide (LPS) and TLR3 recognizes dsRNA. Engagement of TLRs with their ligands triggers signal cascades to activate IRFs and NF-kB, thus leading to production of cytokines and type I IFN [17] . NF-kB activation is crucial for cytokine induction, and many viruses evolve various strategies to manipulate NF-kB signaling [18, 19] . Several RNA virus-encoded proteins, such as HCV NS5B, SARS CoV M protein, measles virus V protein, and enterovirus 71 2C, inhibit NF-kB activation directly or indirectly [20, 21, 22, 23] . In this study, we investigated whether DENV-2 could block type I IFN and cytokine induction triggered by TLR signaling. We studied the influence of DENV-2 infection on activation of NF-kB and extracellular signal-regulated protein kinase (ERK). The DENV-2 PL046 strain (Genbank accession: AJ968413.1) was isolated from a Taiwanese DF patient. The DENV-2 prototype New Guinea C (NGC) strain was kindly provided by D. J. Gubler of the Centers for Disease Control and Prevention, USA. These viruses were propagated in the mosquito cell line C6/ 36 (ATCC: CRL-1660) grown in RPMI 1640 medium containing 5% fetal bovine serum (FBS) [24] . The J774A.1 mouse macrophage cell line (ATCC: TIB-67), A549 human lung epithelial carcinoma cell line (ATCC: CCL-185), and African green monkey kidney epithelial cell line Vero (ATCC: CCL-81) were cultured in DMEM medium supplemented with 10% FBS (Invitrogen). The TLR3 ligand polyinosine-polycytidylic acid (polyI:C) and TLR9 ligand CpG oligodeoxynucleotides 1826 (CpG ODN 1826; hereafter CpG) were from InvivoGen. The TLR4 ligand LPS (Sigma-Aldrich), anti-ERK antibody, anti-phospho-ERK antibody (Cell Signaling, catalog# 9102 and 9101S) and anti-NF-kB p65 antibody (sc-372, Santa Cruz Biotechnology) were used. Female C57BL/6 mice at 6-8 weeks old were used in accordance with the guidelines of Kaohsiung Veterans General Hospital animal care and use committee under the approved animal study protocol (VGHKS-99-A028). BMDCs were generated by culturing bone-marrow hematopoietic cells with FMS-like tyrosine kinase 3 ligand (Flt3L) for 8 days [25, 26] . For DENV-2 infection, 10 6 DCs were adsorbed with DENV-2 at a multiplicity of infection (MOI) of 5 for 1 h. After removing the virus inoculant, cells were incubated with complete medium. For stimulation with TLR ligands, DCs were incubated with 100 mg/ml polyI:C or 1 mg/ml CpG. TRIzol reagent (Invitrogen) was used for total RNA extraction, and cDNA was synthesized from 0.5 mg total RNA by Superscript III reverse transcriptase (Invitrogen). qPCR amplification was done with 4 ng cDNA in 10 ml SYBR Green PCR master mix (Applied Biosystems) with 3 mM of primers in the ABI Prism 7000 Sequence Detection System (Applied Biosystems). Transcript levels were normalized to that of hypoxanthine phosphoribosyltransferase ( Table S1 . Cells were fixed with 4% paraformaldehyde for 30 min, then permeabilized with 0.5% Triton X-100 for 10 min. After 2 washes with PBS, cells were blocked with 10% skim milk in PBS. NF-kB p65 subcellular location was determined by immunostaining with rabbit anti-NF-kB p65, then Alexa Fluor-568-conjugated goat antirabbit IgG antibody (Invitrogen). DENV-2 NS3 was detected by a specific monoclonal antibody against NS3 (#YH3304, 1:500 dilution, Yao-Hong Biotechnology) plus Alexa Fluor-488-conjugated goat anti-mouse IgG antibody (Invitrogen). Fluorescence signals were observed under a fluorescence microscope (Olympus BX51). TurboFect transfection reagent (Fermentas) was used for transient transfection following the manufacturer's protocol. Cells cultured in 12-well plate were transfected with NF-kB-or IFN-b-Luc reporter plasmids [12, 25] . pRL-TK (Promega), encoding Renilla luciferase under an HSV thymidine kinase promoter, was used as an internal control. After transfection for 24 h, cells were infected with DENV-2; in some experiments, cells were further stimulated with LPS or polyI:C (both 1 mg/ml). Cell lysates were collected at the indicated times for dual-luciferase assays (Promega). Relative firefly luciferase activity was normalized to Renilla luciferase activity. Cells were lysed in RIPA buffer (150 mM NaCl, 0.5% sodium deoxycholate, 1% NP40, 0.1% SDS, 50 mM Tris-HCl [pH 8.0]) containing protease inhibitor and phosphatase inhibitor cocktails (Roche). Harvested cell extracts were separated by 10% SDS-PAGE and transferred to PVDF membranes, which were reacted with primary antibody, and then horseradish peroxidase-conjugated secondary antibody (Jackson ImmunoResearch Laboratory) and visualized with an enhance chemiluminescence system (Thermo). Images were acquired by a digital image system (UVP or Fujifilm). To study the effect of DENV-2 on modulating innate immune response, we determined the levels of type I IFN and cytokines in DENV-2-infected mouse BMDCs by quantitative RT-PCR. The TLR3 ligand polyI:C and the TLR9 ligand CpG greatly stimulated the mRNA expression of IFN-b (124-fold induction at 2 h by polyI:C and 531-fold induction at 6 h by CpG), IL-10 (77-fold at 12 h by polyI:C and 74-fold at 36 h by CpG), IL-12p40 (381-fold at 24 h by polyI:C and 1551-fold at 6 h by CpG), and TNFa (64-fold at 2 h by polyI:C and 180-fold at 6 h by CpG) ( Figure 1A -D, left panels). However, levels of IFN-b and these cytokines were much lower in cells with DENV-2 infection (Figure 1 , left and right panels): especially, IL-10 was not induced by DENV-2 infection. The kinetic of DENV-2 replication in BMDCs was measured by qPCR with primers specific for DENV-2 59-UTR ( Figure 1E ) and by immunofluorescence staining with antibody specific against DENV-2 NS3 ( Figure 1F ). DENV-2 replication peaked around 12-24 h post infection, in consistence with cytokine induction peaked around 12-36 h post infection. Therefore, infection with DENV-2 inefficiently triggered type I IFN expression and that of other cytokines. Since DENV-2 infection would produce intracellular viral RNA to turn on RLR and TLR signaling cascades for cytokine production, weak induction of DENV-2 for these cytokine genes ( Figure 1 ) implies that DENV-2 may interfere with a common signaling pathway for inducing type I IFN and other inflammatory cytokines. To determine whether DENV infection could suppress the cytokine induction triggered by TLR signaling, we investigated infection with a murine macrophage cell line J774A.1. Consistent with previous report that J774A.1 is susceptible to DENV infection [27] , DENV-2 infection in J774A.1 was shown by immunofluorescence staining with antibody specific against DENV-2 NS3 (Figure S1A) , and by qPCR with DENV-2 59 UTR primers ( Figure S1B ). J774A.1 cells were mock-infected or infected with DENV-2, then stimulated with LPS or polyI:C. LPS and polyI:C readily promoted the expression of IFN-b (Figure 2A and B; left panels) and IL-10 ( Figure 2A and B right panels), but with DENV-2, the expression was diminished. Similar results were noticed in DENV-2-infected BMDCs that polyI:C-activated IFN-b and IL-10 were reduced with DENV-2 infection ( Figure 2C ). Thus, DENV-2 appears to interfere with IFN-b and cytokine production triggered by TLRs signaling cascade. NF-kB is an essential molecule in the TLR signaling pathway for inducing type I IFN-b and cytokines [16, 28, 29] . In our infection system, two strains of DENV-2, PL046 and NGC, triggered low degree of NFkB p65 nuclear translocation. Nuclear staining of p65 was noted in 7% and 13% of PL046-and NGCinfected cells, respectively ( Figure S2A panels b and c) ; in contrast, about 80% of the control virus infected cells showed nuclear staining of NF-kB p65 ( Figure S2A, panel d) . Thus, we checked whether DENV-2 could suppress NF-kB activation triggered by TLRs engagement. With use of a NF-kB-dependent luciferase reporter, NF-kB activation stimulated by the TLR ligands LPS and polyI:C was readily blocked in DENV-2-infected cells at 12 and 24 h post stimulation ( Figure 3A ). Immunoblotting result of DENV-2 NS3 indicated that the virus was replicating in A549 cells and was not affected by LPS and polyI:C posttreatment ( Figure 3B ). As well, DENV-2 infection blocked LPS-triggered NF-kB activation in Vero cells as measured by nuclear translocation of NF-kB p65 ( Figure 4A) ; similar results were also observed in J774A.1 macrophages ( Figure S2B ). Nuclear translocation of NF-kB p65 was slightly increased in A549 cells up to 60 h post DENV-2 infection, but the polyI:C-triggered p65 nuclear translocation was greatly blocked in DENV-2 infected cells ( Figure S2C ). We also tested the influence of TLR singling in DENV-2 replication, polyI:C pretreatment blocked DENV-2 replication; whereas, the antiviral effect of polyI:C was not seen in cells with established DENV-2 infection ( Figure 4B ). This data is consistent with previous report [30] and suggest that DENV-2 inhibits TLR signaling to benefit its replication. DENV-2 replication in these two cell lines, A549 and Vero, was confirmed by qPCR with viral 59 UTR primers and plaque assay for virion production ( Figure S3 ). Thus, DENV-2 downregulates TLR-activated NF-kB, which leads to reduced cytokine expression. The activity of ERK is associated with the expression of type I IFN and cytokines, particularly IL-10 [31, 32, 33] . We thus further checked whether DENV-2 targets ERK activation by examining the phosphorylated ERK (p-ERK). DENV-2 infection did not activate ERK1/2 phosphorylation and even impaired the basal level of p-ERK1/2 ( Figure 5A ). Furthermore, polyI:C-and LPS-stimulated phosphorylation of ERK1/2 was decreased with DENV-2 infection ( Figure 5B and 5C), so DENV blocks TLR-mediated ERK activation to modulate both arms of the innate immunity response to infection: type I IFN and cytokines. The pathogenesis mechanism of severe DHF/DSS with DENV infection is not well understood, but evidence suggests that the magnitude of DENV replication and its regulation of innate and adaptive immunity may both contribute [34] . The incidence of DHF/DSS is higher in people with previous exposure to different serotypes of DENV, and antibody-dependent enhancement (ADE) may be a mechanism for DHF/DSS [1] . Enhancing antibodies may increase viral entry and increase the number of infected cells. DENV infection via the ADE route has also been shown to downregulate several genes of the innate immunity system, resulting in suppression of the innate response and increase of DENV replication [35, 36, 37] . In this study, we further demonstrate that DENV per se is a weak cytokine inducer because IL-10, IL-12, and TNFa were induced to a lower extent in DENV-2infected BMDCs. Furthermore, even in the absence of enhancing antibody, DENV could block NF-kB activation and cytokine induction triggered by TLR signaling. Downregulation of the cytokine system may provide a growth advantage for DENV to propagate in host macrophages and DCs. IL-10 is a potent immunosuppressor produced by several immune cells [38] , and its expression can be triggered by TLR [31, 32] . IL-10 is critical in suppressing excessive inflammation and immunopathologic conditions caused by the host immune system responding to infections [39, 40] . DHF patients showed high serum levels of IL-10, which may be involved in the pathogenesis of severe dengue disease [41, 42] . In a WNV animal infection model, IL-10 downregulated T cell-mediated immunity and had a negative role in antiviral immunity [43] . Thus, our findings that DENV-2 failed to induce IL-10 expression in BMDCs (Figure 1 ) are unexpected but are consistent with previous reports that IL-10 is not produced from mature human CD1a+ DCs infected with DENV-2 [44, 45] and that microarray data from DENV-1-infected rhesus macaques showed no transcription of IL-10 or other cytokine genes [15] . In addition, we found that DENV-2 could inhibit TLR-triggered IL-10 expression. Therefore, the high IL-10 expression found in patients with dengue-related diseases might not be simply stimulated by the DENV itself. Instead, it may be resulted from uncontrolled DENV replication, which then triggers increased levels of immune activation and increased IL-10 production. DENV infection through the ADE route often induces IL-10 expression, which worsens the host anti-viral system and results in increased DENV production [35, 37, 45] . Thus, DENV infection via ADE or non-ADE routes might have different effects in the immune system. TLR signaling cascades are mainly controlled by the MyD88dependent and TRIF-dependent pathways, which both lead to activation of NF-kB and mitogen activated protein kinases (MAPKs) [17, 46] . NF-kB and MAPKs have been suggested to have critical roles in cytokine induction [47, 48] . A recent study showed that constitutive intestinal NF-kB activation does not lead to destructive inflammation unless accompanied by activation of MAPKs such as p38 and ERK [49] . Our results that DENV-2 blocked activation of NF-kB, as well as ERK1/2, support our findings of cytokine production hampered in DENV-2-infected cells. Because both NFkB and ERK1/2 were affected, DENV may suppress an upstream molecular event such as TLR gene expression, as that has been reported for DENV infection through an ADE route [37] . We also detected TLR genes expression in DENV-2 infected J774A.1 macrophages and found that polyI:Ctriggered TLR4, TLR5 and TLR13 expression was significantly downregulated by DENV-2 ( Figure S4 , panels a-c), but not for that of TRL6, TLR7, TLR8 and TLR2 ( Figure S4 , panels d-g). In contrast, DENV-2 enhanced the gene expression of TLR1 and TLR3 triggered by polyI:C stimulation ( Figure S4 , panels h-i), suggesting that DENV-2 may modulate certain TLR genes expression in macrophage. Other possibilities, such as whether DENV protease, found to block IFN-b promoter activation [13] , may target common molecules involved in type I IFN and cytokine production in TLR signaling, or whether ISG15 that is induced by DENV-2 and functions as an inhibitor of type I IFN production [50] may also contribute to immune evasion, remain to be further studied. Taiwanese DENV-2 strain PL046 used in this study was isolated from patient with DF, and this virus has been used in the studies of viral pathogenesis and host responses mechanism in vitro and in vivo [7, 12, 51, 52, 53] . Other groups have reported that DENV-2 strains MON601 (a laboratory strain of NGC) and 16681 downregulate the activation of NF-kB and production of type I IFN and TNFa in human DC, macrophage, and Huh7 cells [14, 54] . A recent report by Chase A. J. et al. [55] also revealed that human DCs infected with several endemic DENV-2 strains failed to polarize the naïve CD4 + T cells to effectors, suggesting a defect on T cell priming for DENV-infected DCs [55] . However, one of the strain ARA6894 did not show such kind of inhibition effect, as ARA6894-infected DCs triggered CD4 + Th1 polarization with high expression of IFNc and TNFa. Interestingly, the polyprotein sequences of strain ARA6894 contain nonsynonymous amino acids that are not present in other DENV-2 strains such as PL046, 16681 and NGC [55] , suggesting that different impacts on DC's function between these DENV-2 strains might be due to viral genome diversity. Dengue may be the most important arboviral disease potentially affecting 2 to 3 billion people living in tropical and subtropical areas. DENV mainly infects monocytes, macrophages and DCs that are also the most important innate immune cells. The balance between the protective and pathological immune responses likely contributes to the DENV infection outcomes. Our results demonstrating that DENV can modulate the signaling events triggered by several TLRs are of interest and provide an explanation for how DENV may skew the host immune system. Immunofluorescence analysis showed the subcellular localization of NF-kB p65 (red, panels a-d) and the detection of viral proteins DENV-2 NS3 or JEV NS1 (green, panels e-h). (B) J774A.1 macrophages were infected with DENV-2 PL046 for 48 h before stimulation with LPS (1 mg/ml). After 6 h of LPS treatment, the localization of NF-kB p65 was determined by immunostaining with anti-NF-kB p65 antibody (red fluorescence, panels a-d). DENV-2 infection was determined by anti-DENV-2 NS3 antibody (green fluorescence, panels e-h), and the DAPI presents the nuclear counter stain (blue fluorescence, panels i-l). Representative cells from the same field are shown for each experimental group. (C) A549 cells were infected with DENV-2 PL046 (MOI of 5) for the indicated times. For polyI:C stimulation, DENV-2-infected cells at 36 h p.i. were transfected with polyI:C (2 mg) and incubated for another 24 h, meaning DENV-2 infection for a total of 60 h. For the mock control group, polyI:C stimulation was conducted by polyI:C (2 mg) transfection for 24 h. The immunofluorescence staining of NF-kB P65 was performed as described above, and the DENV-2 infected cells with nuclear p65 were counted. Data are mean6SD from 3 determinations. Table S1 . (TIF)

High Influenza A Virus Infection Rates in Mallards Bred for Hunting in the Camargue, South of France

During the last decade, the role of wildlife in emerging pathogen transmission to domestic animals has often been pointed out. Conversely, far less attention has been paid to pathogen transmission from domestic animals to wildlife. Here, we focus on the case of game restocking, which implies the release of millions of animals worldwide each year. We conducted a 2-year study in the Camargue (Southern France) to investigate the influence of hand-reared Mallard releases on avian influenza virus dynamics in surrounding wildlife. We sampled Mallards (cloacal swabs) from several game duck facilities in 2009 and 2010 before their release. A very high (99%) infection rate caused by an H10N7 strain was detected in the game bird facility we sampled in 2009. We did not detect this strain in shot ducks we sampled, neither during the 2008/2009 nor the 2009/2010 hunting seasons. In 2010 infection rates ranged from 0 to 24% in hand-reared ducks. The 2009 H10N7 strain was fully sequenced. It results from multiple reassortment events between Eurasian low pathogenic strains. Interestingly, H10N7 strains had previously caused human infections in Egypt and Australia. The H10 and N7 segments we sequenced were clearly distinct from the Australian ones but they belonged to the same large cluster as the Egyptian ones. We did not observe any mutation linked to increased virulence, transmission to mammals, or antiviral resistance in the H10N7 strain we identified. Our results indicate that the potential role of hand-reared Mallards in influenza virus epizootics must be taken into account given the likely risk of viral exchange between game bird facilities and wild habitats, owing to duck rearing conditions. Measures implemented to limit transmission from wildlife to domestic animals as well as measures to control transmission from domestic animals to wild ones need to be equally reinforced.

During the last decade, awareness concerning the intimate links between human and animal health has rapidly increased in the context of disease emergence [1] . Indeed, approximately 80% of the infectious diseases that recently emerged were zoonotic [2] . The role of wildlife in emerging pathogen transmission to humans and domestic animals has in many cases been pointed out [3±5] . Conversely, pathogen transmission from domestic animals to wildlife has received far less attention, although the importance of this issue was often mentioned [6, 7] . Indeed, contacts between wildlife and livestock or their environment sometimes result in wildlife diseases with conservation issues [8, 9] . Besides, handreared animal releases into the wild for either conservation or exploitation purposes represent a particular case in which handreared individuals eventually share natural habitats with their wild congeners. In both cases such releases can dramatically influence disease dynamics in the surrounding wild animal populations [ 10± 14] . In the present study we focused on the case of game restocking, which implies the release of millions of individuals worldwide each year [15] . Birds are the most frequently involved, with millions of individuals being released annually in Europe only. For example more than 3 millions red-legged partridges are released annually in Spain [15] , and ca. 1.4 million Mallards are being so in France [16] . The Camargue region, a complex network of wetlands situated in the Rhone delta, is a major duck winter quarter [17] and a central place for wildfowl hunting in France [18] . Hunting is also among the most important economic activities in the area, which is one of the reasons for the massive Mallard releases in the Camargue [19] . At least 30 000 hand-reared individuals are released annually in the region [20] . Maximum Mallard numbers in the wild are reached in September after the beginning of the hunting season, with 56 500 individuals on average over the last seven years (Gauthier-Clerc, unpubl. data), these numbers certainly include a mixing of wild and released Mallards. Given its central position on the flyway of many European migratory species, the Camargue is also a potential hotspot for the introduction and transmission of bird-borne pathogens [21] . For this reason, avian influenza viruses (AIV) have been studied since 2004 in the area. These negative-sense single stranded RNA viruses belonging to the Orthomyxoviridae family are commonly characterized by the combination of their surface proteins: hemagglutinin (HA) and neuraminidase (NA) [22, 23] . AIVs are highly variable and undergo continuous genetic evolution via two mechanisms: i) accumulation of point mutations at each replication cycle, ii) reassortment involving gene segment exchanges that occur when a cell is co-infected by different viruses [24] . These mechanisms contribute to the emergence of new variants with the ability to transmit to new hosts and/or with epidemic or even pandemic potential. Aquatic birds, particularly Anseriforms (ducks, geese and swans) and Charadriiforms (gulls, terns and shorebirds) constitute their major natural reservoir [23, 25] . The AIV circulating in wild birds are usually low pathogenic ones (LPAIV). LPAIV generally have little impact on their host [26, 27] , although some studies have reported a possible influence on migration capacities [28] . Besides, when LPAIV of H5 or H7 subtypes are transmitted from wild birds to domestic ones reared in artificial environments, their virulence can evolve to high pathogenicity [29, 30] . Highly pathogenic avian influenza viruses (HPAIV), such as HP H5N1 strains currently circulating in Asia and Africa, are still of great economic concern, notably due to the cost of preventive actions including vaccination and massive birds culling [31] . Moreover, HPAIV infections represent a threat for human health since 603 HPAIV H5N1 human infections including 356 fatal cases have been reported worldwide since 2003 [32] . Relatively high prevalence of AIV was regularly detected in the wintering Mallard population of the Camargue (e.g. 5.4% prevalence during the 2006±2007 hunting season) [33] . Moreover a seasonal infection pattern was identified in Mallards during autumn and winter, with higher infection rates in early fall [33] . Mallards hence represent a focal study species in AIV research. Indeed wild Mallards are one of the main low pathogenic AIV natural reservoir host [34] , and have proven to be healthy carriers of some of the H5N1 HPAIV strains [35] . However, to our knowledge no study ever aimed at investigating the potential role of hand-reared Mallards released for hunting in the epidemiology of AIV, despite the very large number of ducks being released in the wild annually. To clear this gap we conducted a 2-year study in the Camargue to investigate the potential influence of hand-reared Mallard releases on AIV dynamics in surrounding wildlife. We first hypothesized that, owing to high density rearing conditions and to their genetic uniformity, hand-reared Mallards should be highly susceptible to AIV infections and could play an amplification role in AIV dynamics. This phenomenon has already been pointed out in red-legged partridge (Alectoris rufa) reared for hunting in Spain, where Escherichia coli prevalence was much higher in hand-reared populations before their release than in the wild ones [36] . To test this assumption we collected cloacal swabs from Mallards reared for hunting in several game bird facilities (GBF) in 2009 and 2010, and analysed these samples to measure AIV prevalence. Second, we hypothesized that AIV exchange occurs between wild and hand-reared Mallards, potentially leading either to the circulation of new strains in wild populations or to the amplification and dispersal of wild strains. Indeed, no barrier prevents AIV exchange between wild birds and hand-reared ducks in the GBF since water flows exist between pens and ponds used by wild birds. Moreover, as the GBF roofs are made of nets wild birds can deposit feces in the pens. Finally, hand-reared ducks are in direct contact with wild ones after their release. To investigate these issues, we tested shot waterfowl before and after handreared Mallards were sampled. Noteworthy, genetic analyses suggest that 76% of hunted Mallards in the Camargue have a captive origin [37] . Considering the low annual survival of released Mallards (0.8±15.9% depending on the release site) [37] , the individuals we tested from the Camargue hunting bags certainly included a large proportion of ducks released some months before being shot. These released ducks cannot be differentiated morphologically from the wild ones [38] . Here we hence analyzed shot ducks as a whole since they are a representative sample of the Mallard population wintering in the Camargue, which is composed of individuals of both wild and captive origin that share habitats and can thus be considered as a single epidemiological unit. Our third hypothesis was that any AIV strain potentially found in captive reared Mallards might present genetic characteristics linked to its circulation in domestic populations. We therefore performed a molecular study of the identified strains in order to look for such characteristics, and in particular to test for the presence of mutations known to be associated with increased virulence, since it has been highlighted that artificial environments are favorable to the appearance of such mutations in birds [30] . We also searched for mutations linked with transmission to other species including humans, since some studies proved that they can be acquired during their transmission among birds [39] . Thirdly, we tested for the presence of mutations conferring resistance to common antiviral drugs, since such resistance has recently been recorded in wild birds, notably in Sweden [40] . Lastly, full sequence analysis of some strains was performed, so as to get insight into their geographic origin through a phylogenetic study, and to determine their relatedness with strains, which have caused human infections in the past. The infection rates in the different GBF were determined by real-time RT-PCR targeting the conserved M gene of AIV (Table 1) . A very high infection rate (99%) was observed in the single GBF sampled in 2009. In 2010 the infection rates ranged between 0 and 24%, being of 8% in the farm infected in 2009. Initial subtyping, searching for H5, H7, H9, N1, and N7 by realtime RT-PCR performed on positive samples led to the identification of an H10N7 strain in the GBF in 2009 and further testing for H10 by real-time RT-PCR showed that the outbreak involved a single H10N7 virus (named H10N7 Camargue below). The viral strains involved in the infections observed in 2010 were LPAIV. H5, H7, H9, H10, and N7 subtypes were searched among them but none was detected. In shot ducks, mean AIV infection rates were 15% and 5% during the 2008/2009 and the 2009/2010 hunting seasons, respectively ( Table 2) . During these two periods the maximum monthly infection rates (respectively 20% and 16%) were observed in September (Table 2 ). All the involved strains were LPAIV but an important proportion of these were of the H5 subtype (2008/ 2009:17%; 2009/2010:30%). No H7 or H10N7 subtype was detected in AIV found in shot ducks. Phylogeny. The H10N7 Camargue strain was isolated in embryonated chicken eggs and amplified. Determination of the sequence of the whole genome was performed in order to gain insights into its phylogeny and molecular characteristics. The phylogenetic analysis of four H10N7 Camargue isolates confirmed that they were very closely related to each other and formed a cluster (Figures 1 and S1 ). Analysis of each of the sequences of the 8 viral segments highlighted some shared characteristics (Figures 1 and S1). First, the sequences of all segments were clearly distinct from those of LPAIV strains from the North-American lineage. Second, none were closely related to highly pathogenic viruses (H5N1 or H7N7) that circulated in Europe and Asia. However, the PB2 segment was more closely related to the Asian HP H5N1 cluster than to most of the European LPAIV strains we included in our phylogenetic analysis. This finding suggests an ancient Asian origin of this segment. Third, the sequences of all segments were closely related to Eurasian strains. Interestingly, the most closelyrelated Eurasian strains differed between segments. As an example the LPAIV strain A/mute swan/Hungary/5973/2007 (H7N7) was very closely related to the H10N7 Camargue virus for the N7 segment, but less so for the M segment and belonged to a distant phylogenetic group for the PB2 segment. In the same way, the LPAIV strain A/mallard/Netherlands/9/2005 (H7N7) was closely related to the H10N7 Camargue virus for the M segment, whereas its PB2 segment belonged to a distinct phylogenetic group. African strains that we included in our analysis were part of Eurasian clusters and clearly distinct from American strains. For one of them, A/Pekin duck/South Africa/AI1642/2009 (H10N7) the NS, N7 and H10 sequences were closely related to those of the H10N7 Camargue virus but available sequences of two other segments (NP and M) were clearly distinct. Finally, it is important to note that based on the analysis of partial sequences of the H10 (data not shown) the H10N7 strains that caused human infections in Australia (Genebank accession numbers ADG58106 and ADG58107; Ratnamohan,V.M. and Dwyer, D.E. unpublished) are clearly distinct from the H10N7 Camargue strain. On the contrary Egyptian H10N7 strains that circulated in birds in 2004, i.e. at the time of the detection of two human infections [41] , are relatively closely related to the H10N7 Camargue strain. However, this relatedness is difficult to interpret since the involved nodes are not strongly supported statistically (Figure 1) . Study of the extremities of the viral segments. We also determined the sequences of the extremities including the full noncoding sequences of all segments of the H10N7 Camargue strain, which to our knowledge was not performed before for any H10N7 strain. The extremities of the N7 segment revealed two characteristics that confirmed the phylogenetic distinction between the H10N7 Camargue strain and those of the American lineage. First, in the non-coding region of the 59 extremity, an A was found at position 1440, as commonly seen in Eurasian strains, while strains of the American lineage possess a T at this position. Second, at position 145±147 of the 39 extremity that belongs to the coding region the Camargue strain, like other Eurasian ones, has a supplementary amino acid compared to strains of the American lineage. Molecular characterization. The molecular analysis of the H10N7 Camargue strain is detailed in Table 3 . The focus strain is a low pathogenic one since the HA segment possesses a monobasic cleavage site. HA, NA and PB2 sequences all presented some typical avian characteristics (see Table 3 ). No known determinants of adaptation to mammals (humans or mice) were detected. In addition, we did not identify any mutation known to be linked with increased virulence in mammals. Yet, we detected the V149A mutation in PB2, which is associated with high virulence in chickens. However, 149A was also observed in low pathogenic H7 strains sampled from wild and domestic ducks in China. This suggests that this mutation alone does not increase the virulence in ducks [42] . Furthermore, no mutation was detected in the H10N7 Camargue strain at positions known to be linked with reduced susceptibility or resistance to neuraminidase inhibitors or M2blockers. This study confirms our first hypothesis that hand-reared Mallards are highly sensitive to AIV infections. Indeed, we detected high infection rates in the GBF we sampled in 2009 (99%) and in 2010 (up to 24%), whereas the prevalence usually observed in duck populations very rarely exceeds 20% in the wild [33, 43, 44] . This shows that massive outbreaks can occasionally affect GBF, potentially impacting on AIV dynamics in the wild populations with which hand-reared Mallards are in contact before and after their release into the wild. This pattern has already been highlighted in Spain where rabbit (Oryctolagus cuniculus) restocking caused sarcoptic mange dispersion in the wild rabbit population, leading to massive decrease in numbers [11] . The possible outcomes of GBF infection onto wild duck populations should here depend on the origin of the strain involved. Indeed, low pathogenic AIV strains that naturally circulate in wild ducks have little impact on their health [26] . Our data did not permit to determine the origin of the strain we identified in 2009. Indeed, we did not detect the H10N7 Camargue strain among the samples collected on shot Mallards, neither during the hunting season that preceded the release of the infected individuals, nor during the following one. Our sample size for hunted Mallards was reasonably large (n = 299 in 2008/2009 and 555 in 2009/2010). Moreover, the sites where we sampled hand-reared and shot Mallards were close to each other (see Figure 2 ; for instance one of the hunting estates we sampled was within 4 km of the GBF where we detected an AIV outbreak). Nevertheless, our sampling effort was limited in time to the hunting seasons: we do not have any data from February to August 2009. We therefore cannot exclude that the H10N7 strain originated from wild birds. Population density during the breeding season is low in wild Mallard [45] , and thus, even if the H10N7 Camargue virus originated from wild birds, hand-reared Mallards would then have played an amplifying role. In addition, commercial trade of Mallard chicks is common and most of the juveniles sampled in this study traveled 600 km from the producer when one-day old. It is also possible that the H10N7 Camargue strain originated from the birthplace of the chicks. If so, domestic ducks would have both played an amplifying role and potentially a dispersal one, since they could have spread a new strain in wild duck populations. The lack of samples during the weeks that directly followed the release of the positive Mallards (because the hunting season had not yet started) could explain why we did not detect the H10N7 Camargue virus in wild individuals. Furthermore, the lack of large-scale dispersal of that H10N7 strain could be due to different parameters. First, we sampled the hand-reared ducks of the GBF1 21 days before their release. It is possible that only few birds were still excreting viruses when released into the wild, since excretion time in birds can be as short as 2 days (although it can also last up to 30 days in some cases) [46±48]. Second, a parallel capturerecapture demographic study ran in the Camargue has shown that hand-reared Mallards exhibit low monthly survival before the hunting season (only 44% survive from release until the onset of the hunting period on average) [49] . Moreover, their dispersal capacities appear to be very low [49] . The same pattern has been observed for red-legged partridges released in Spain [13] . Although intestinal parasites were much more numerous in the individuals living in the hunting estates where domestic birds were released, this was interestingly not observed in the neighboring estates [13] . It is possible that the poor survival and low dispersal of the hand-reared Mallards after release limited the potential spread of the H10N7 Camargue virus within the Mallard population in the wild. Besides, we did not detect any H5 subtype circulating in the GBF whereas 23% of all the AIV we isolated from samples collected on shot Mallards in the Camargue belonged to this subtype. Again we could not highlight any AIV exchange between ducks living in captivity and in the wild. Yet, hand-reared ducks are exposed to wild bird feces and GBF are connected to wild bird habitats through common use of water. Thus, AIV exchanges are very likely to exist. Our results highlighted that: i) GBF represent an epidemiological compartment into which important AIV outbreaks can occur; ii) a significant proportion of Mallards wintering in the Camargue are infected by LPAIV, including H5 strains that are known to be able to evolve to HPAIV in domestic birds [30] . Knowing that rearing conditions in the GBF may favor AIV exchanges between wild and hand-reared Mallards, these findings outline the need for influenza surveillance in GBF to prevent HPAIV outbreaks in both wild and domestic birds in the region. We used the phylogenetical analysis of whole-genome sequence of the H10N7 Camargue strain as a supplementary tool to get insights into its evolutionary and geographical origins. The four isolates studied clearly belonged to a cluster that did not include other strains. The infections we detected were therefore certainly due to a single introduction event. All the viral segments of the H10N7 Camargue strain belong to the Eurasiatic lineage also including African strains. The sequences of the extremities of the N7 segment, that exhibit features characteristic of the Eurasian lineage viruses, confirmed this finding. The Eurasian lineage is clearly distinct from the American lineage and our observations further support the fact that inter-hemispheric AIV exchanges are rare. The Camargue strain was also distinct from the H10N7 viruses that previously caused human infections in Australia, but belonged to the same large cluster as the Egyptian H10N7 strains that circulated in birds when H10N7 human infections occurred in this country. These Egyptian strains did not cause severe disease. Nevertheless, this information raises concern about a possible transmission to humans of strains such as the Camargue one, in particular to the GBF owners and hunters who live in close contact with ducks. Interestingly, the strains that are genetically closest relatives differ for each segment of the H10N7 Camargue strain. The Camargue strain thus seems to result from multiple reassortments between different Eurasiatic strains, a common phenomenon which has been illustrated by many studies [50, 51] . Regrettably, we could not determine if the H10N7 Camargue strain was more closely related to strains observed in the wild or in GBF. Indeed, the data associated with viral sequences from AIV infected Mallards in GenBank are often too scarce to determine if the sample was taken on a wild or a domestic individual. This underlines the need for more detailed information on AIV sequences included in common databases. The molecular analysis of the H10N7 Camargue strain allowed us to question whether it may exhibit some characteristics related to its circulation in an artificial environment, namely some mutations linked to increased virulence, resistance to antiviral drugs or transmission to humans. Our results do not support this assumption. The sequences of the studied segments presented typical avian characteristics. No mutation known to contribute to transmission to humans was detected. Moreover, the Camargue strain is a low pathogenic one like all viruses isolated during our research program in the Camargue since 2005 [33, 52] . No mutation conferring antiviral resistance was detected. This analysis therefore failed to highlight any risk factor linked to this particular strain. Yet, one must keep in mind that we only looked for mutations previously described in the literature, while many others may remain undiscovered so far. It is also important to stress that most experimental studies focusing on mutations modifying host specificity and virulence are carried out on chicken, although it has been demonstrated that phenotypic outcomes of a given mutation can greatly differ from one host species to another [42] . In conclusion, our results point out the important role that could be played by hand-reared ducks released for hunting in AIV dynamics. As we did not detect similar AIV strains in shot and hand-reared ducks we could not prove that AIV exchanges exist between these two epidemiological compartments. Yet, due to rearing conditions in GBF, AIV exchange risks seem to be high enough to urge for sanitary control of hand-reared animals prior to their release into the wild, which appears to be highly insufficient so far. Such surveillance would also prevent HPAIV circulation that may arise from the evolution in GBF of H5 LPAIV that proved to be commonly infecting free-living ducks in the Camargue. The world organization for animal health (OIE) stresses that surveillance of AIV infection should be applied to all domesticated birds including those used``for restocking supplies of game'' [53] , but control measures generally appear to be poor in game bird rearing estates. This surveillance gap has recently been highlighted in the USA, where game bird holders reported very variable sanitary practices [54] . The problem appears similar in Europe, including France. Additionally, according to French law, game birds should be ringed to allow for the differentiation of wild and released individuals [55] . Most of GBF owners ignore this obligation. If released birds were clearly identified the specific role they may have in AIV dynamics could be addressed both before and after their release. Our study illustrates the reality of the epidemiological consequences that can result from surveillance gaps and the knowledge that could be gained if released individuals were recognizable. Sharing knowledge and strictly controlling viral exchanges between wild birds and all kinds of domestic ones represent major steps to anticipate and face HPAIV epizootics. This study has been submitted for approval and the results reviewed by the Scientific Council of the Tour du Valat Foundation. Birds were handled, ringed and sampled under the supervision of a veterinarian (Michel Gauthier-Clerc) and a registered duck ringer of the « Museum National d'Histoire Naturelle » of Paris (Matthieu Guillemain). All procedures were carried out in accordance with the permit delivered by the prefect of Paris to the « Office National de la Chasse et de la Faune Sauvage » (permit n 2009±014). (Table 1) . Samples were stored in 3 ml of universal transport medium (Biolys kit) and frozen at 280uC until molecular analysis was performed. RNA isolation from 140 ml of each sample was carried out using a Macherey-Nagel NucleoSpin 96 virus system with RNA elution into a final volume of 60 ml. Influenza A virus was detected by real-time RT-PCR targeting the conserved matrix as describe previously [56] . All real-time RT-PCR assays were performed on a LightCycler 480 (Roche Diagnostic) in a final volume of 10 ml with 2.5 ml RNA, 0.5 mM of each primer, 0.2 mM probe and 0.4 ml enzyme mix using a Superscript III Platinum One-Step quantitative RT-PCR system (Invitrogen). The reaction was carried out with the following temperature profile: 15 min at 45uC, 3 min at 95 C, 50 cycles of 10 s at 95uC, 10 s at 55uC, 20 s at 72uC, and finally 30 s at 40uC. Positive samples were further tested for H5, H7, H9, N1 and N7 subtypes using the same real-time RT-PCR technique (primers available upon request). For 10 RT-PCR positive samples, 200 ml of specimen was inoculated in the allantoic cavity of 10-days-old embryonated hen's eggs. The allantoic fluid was harvested after 72 h at 37uC and influenza A virus was detected by hemagglutination assay with hen erythrocytes. Viral RNA was purified from allantoic fluid using a QIAamp viral RNA mini kit (Qiagen) following manufacturer's instructions. HA subtype of virus isolates was determined by RT-PCR and sequencing of the HA 0 cleavage site region using a universal set of primers as described previously [57] . HA sequence was analyzed with the basic local alignment search tool available from NCBI (data not shown) and confirmed by sequencing of the whole HA gene using a set of H10-specific primers (primer sequences available upon request). The NA subtype was deduced from this analysis and confirmed by RT-PCR and sequencing of 172 nt of the NA using a set of H10specific primers (primer sequences available upon request). Amplification of viral RNA extracted from 4 virus isolates was carried out using a Superscript Platinum One-step RT-PCR system (Invitrogen) and primers specific of each segment. Sequencing was done using a Big Dye Terminator V1.1 kit and a sequencer ABI DNA Analyzer 3730XL (Applied Biosystems). Sequences of all primers are available upon request. Sequences of the whole viral genome were analyzed with CLC Main Workbench 5.6.1. We performed alignments for the 8 segments using sequences available from the Influenza Sequence Database with CLC Main Workbench 5.6.1. Phylogenetic trees were constructed using maximum parsimony (MP) methods with the dnapars program of the PHYLIP 3.68 package and the maximum likelihood (ML) with the software PhyML 2.4.4. Evolutionary model was selected using Model Generator 0.85 [58] . Nodal supports were assessed with 100 bootstrap replicates generated for each method. Determination of 39 and 59 End Non Coding Sequences of the 8 Segments of the H10N7 Virus Viral genomic RNA was extracted using the QIAamp Viral RNA Mini kit (Qiagen) from 140 ml of allantoic fluid according to the manufacturer's recommendations. The RNA was eluted in 60 ml of RNase-free water and the 39 and 59 NC regions were amplified as previously described [59] using an anchored (dT) 14 oligonucleotide and primers specific for the coding sequences of both segments for reverse transcription and amplification. After purification, the PCR products were sequenced with internal oligonucleotides using a Big Dye terminator sequencing kit and an automated sequencer (Perkin Elmer). All sequences of the primers are available from the authors upon request. One-step real-time RT-PCR assays were developed to be specific of the virus identified. All influenza A positive samples detected between August 2008 and July 2010 on wild and domestic Mallards were tested for H10 and N7 subtypes respectively. Detection was performed in the same conditions as described above except annealing temperature was decreased from 55uC to 50uC using the following primers and probes: H10-780Fw

Establishment of a Reverse Genetics System for Studying Human Bocavirus in Human Airway Epithelia

Human bocavirus 1 (HBoV1) has been identified as one of the etiological agents of wheezing in young children with acute respiratory-tract infections. In this study, we have obtained the sequence of a full-length HBoV1 genome (including both termini) using viral DNA extracted from a nasopharyngeal aspirate of an infected patient, cloned the full-length HBoV1 genome, and demonstrated DNA replication, encapsidation of the ssDNA genome, and release of the HBoV1 virions from human embryonic kidney 293 cells. The HBoV1 virions generated from this cell line-based production system exhibits a typical icosahedral structure of approximately 26 nm in diameter, and is capable of productively infecting polarized primary human airway epithelia (HAE) from the apical surface. Infected HAE showed hallmarks of lung airway-tract injury, including disruption of the tight junction barrier, loss of cilia and epithelial cell hypertrophy. Notably, polarized HAE cultured from an immortalized airway epithelial cell line, CuFi-8 (originally derived from a cystic fibrosis patient), also supported productive infection of HBoV1. Thus, we have established a reverse genetics system and generated the first cell line-based culture system for the study of HBoV1 infection, which will significantly advance the study of HBoV1 replication and pathogenesis.

Human bocavirus 1 (HBoV1) was initially identified in 2005, in nasopharyngeal aspirates of patients with acute respiratory-tract infections (ARTI) [1] . It was found to be associated with ARTI in children, at a detection rate of 2-19% [2] [3] [4] [5] . Three additional human bocaviruses, HBoV2, 3 and 4, discovered in human stool samples, have since been phylogenetically and serologically characterized [6] [7] [8] [9] . However, whether these are associated with any diseases is currently unknown. HBoV1 is commonly detected in association with other respiratory viruses, and is the fourth most common respiratory virus (after respiratory syncytial virus (RSV), adenovirus and rhinovirus) in infants less than 2 years of age who are hospitalized for the treatment of acute wheezing [2, [10] [11] [12] . Indeed, ARTI is one of the leading causes of hospitalization of young children in developed countries [13, 14] . Acute HBoV1 infection, diagnosed by a virus load of .10 4 genome copies (gc)/ ml in respiratory samples, viraemia, or by detection of HBoV1specific IgM or of an increase in the levels of IgG antibodies, results in respiratory illness [2, [15] [16] [17] [18] [19] [20] . Recent descriptions of lifethreatening HBoV1 infections in pediatric patients in association with high virus loads or diagnostic HBoV1-specific antibodies [21] [22] [23] , in addition to a recent longitudinal study of children from infants to puberty, documenting a clear association of acute primary HBoV1 infection with respiratory symptoms [24] , strongly support that HBoV1 is an etiological agent of both upper and lower ARTI. HBoV1 has been classified as a new member of the genus Bocavirus of the family Parvoviridae [25] , of which bovine parvovirus (BPV1) and minute virus of canines (MVC) are the prototypes [26, 27] . In comparison with the BPV1 and MVC genomes, the HBoV1 genome sequences obtained previously appeared to exclude the two termini, and therefore, were incomplete [28] . However, sequencing of the head-to-tail junctions of HBoV1 and HBoV3 ''episomes,'' which had been amplified in DNA samples extracted from HBoV1-infected differentiated human epithelial cells and from intestinal biopsies of HBoV3-infected patients, respectively, revealed portions of the HBoV termini [29, 30] . Notably, these sequences were conserved with the terminal sequences of BPV1 and MVC [28] . In vitro HBoV1 infection has been reported only once in welldifferentiated human airway epithelia (HAE) [31] . That study provided only minimal information on virus replication, and did not include observations of pathophysiology. Obviously, the lack of a sustainable and highly reproducible system that enables highyield virus production, as well as the ability to conduct reverse genetics is a significant barrier to further elucidation of HBoV1 replication and pathogenesis. In the current study, we have successfully sequenced the full-length HBoV1 genome and cloned it in a plasmid referred to as pIHBoV1. Furthermore, we have demonstrated that transfection of human embryonic kidney 293 (HEK293) cells with pIHBoV1 results in efficient production of HBoV1 virions at a high titer, and that these virions are able to productively infect both primary and conditionally transformed polarized HAE. The terminal hairpins of the HBoV1 genome are typical of those of the genus Bocavirus A head-to-tail junction of an HBoV1 episome identified in an HBoV1-infected HAE [28, 29] was found to possess two sequences (39-CGCGCGTA-59 and 39-GATTAG-59) identical to parts of the BPV1 left-end hairpin (LEH) [27, 32] . This finding suggested that the head sequence is part of the HBoV1 LEH (nucleotides in blue; Figure 1A ). We therefore used the head sequence as the 39 end of a reverse primer (RHBoV1_LEH). Together with a forward primer (FHBoV1_nt1), which anchors the 39 end of the HBoV1 genome predicted from the BPV1 LEH, we amplified the hairpin of the LEH from a viral DNA extract (1.2610 8 gc/ml) prepared from a nasopharyngeal aspirate taken from an HBoV1-infected patient (HBoV1 Salvador1 isolate) [17] . Only one specific DNA band was detected at approximately (,)150-bp ( Figure 1D , lane 1). Sequencing of this DNA revealed a novel sequence of the HBoV1 LEH (nucleotides in red between the two arrows; Figures 1A and S1A). Because the LEHs of the prototype bocaviruses BPV1 and MVC are asymmetric [27, 32] , we set up another PCR reaction with a forward primer located in the hairpin (FHBoV1_LEH) and a reverse primer targeting a sequence downstream of the LEH at nt 576 (RHBoV1_nt576; Figure 1B ). Sequencing of a DNA fragment ( Figure S1B ), detected as expected as a ,600-bp band ( Figure 1D, lane 3) , confirmed the presence of the novel joint sequence and the LEH ( Figure 1B) . The tail of the HBoV1 head-to-tail junction [28, 29] was found to contain a sequence (59- GCG CCT TAG TTA TAT ATA ACA T -39) identical to that of the right-end hairpin (REH) of the other prototypic bocavirus MVC [27] . We thus speculated that the entire HBoV1 REH is similar in structure to its MVC counterpart. Using a reverse primer targeted to this sequence (RHBoV1_nt5464) and a forward primer located upstream of the REH (FHBoV1_nt5201), we were able to amplify a specific ,300-bp-long DNA fragment ( Figure 1D , lane 5). Sequencing confirmed the presence of the palindromic hairpin of the predicted REH (nucleotides in red; Figures 1C and S1C), and revealed two novel nucleotides at the end of the hairpin (GC in red; Figure 1C ). These results indicate that we have identified, for the first time, both the LEH and REH of the HBoV1 genome from a clinical specimen, and confirm that the HBoV1 genome structure is typical of the genus Bocavirus. A full-length HBoV1 clone (pIHBoV1) is capable of replicating and producing progeny virus in HEK293 cells We also cloned and sequenced the non-structural (NS) and capsid (VP) protein-coding (NSVP) genes of the HBoV1 Salvador1 isolate from the patient-extracted viral DNA. We then ligated the LEH, NSVP genes and REH into pBBSmaI using strategies diagramed in Figure S2 , and refer to this full-length clone as pIHBoV1. We have deposited the sequence of the full-length genome of the isolate in GenBank (JQ923422). As we previously showed that HEK293 cells support replication of the DNA of an autonomous human parvovirus (B19V) in the presence of adenovirus helper genes or adenovirus [33] , we first investigated whether the adenovirus helper function is necessary for pIHBoV1 replication in HEK293 cells. Specifically, we transfected pIHBoV1 into HEK293 cells (untreated or infected with adenovirus), alone or with pHelper. Interestingly, we found that pIHBoV1 replicated well in the absence of helper virus. Indeed, all the three representative forms of replicated bocavirus DNA [27, 34] (DpnI digestion-resistant dRF DNA, mRF DNA and ssDNA) were detected in each test case, and at similar levels ( Figure 2A ). DpnI digestion-resistant DNA bands are newly replicated DNA in cells as DpnI digestion only cleaves plasmid DNA prepared from prokaryotic cells, which is methylated at the dam site [35] . In contrast, these DNA forms of the viral genome were absent in pIHBoV1-transfected primary airway epithelial cells (NHBE; Figure 2B , lanes 7&8) and present at very low levels (over 20 times lower than in pIHBoV1-transfected HEK293 cells) in pIHBoV1-transfected human airway epithelial cell lines BEAS-2B ( Figure 2B , lanes 5&6), A549 and 16HBE14o-( Figure 2C ), even in the presence of adenovirus. Thus, replication in these cells appears to be non-existent or poor in these contexts. To confirm the specificity of DNA replication and the identity of the DpnI-resistant DNA bands, we disrupted the ORFs encoding viral proteins NS1, NP1, VP1 and VP2 in pIHBoV1; knockout of expression of the corresponding viral protein was confirmed by Western blot analysis. When the NS1 ORF was disrupted, no DpnI digestion-resistant DNA was detected ( Figure 2D , lane 4), confirming that replication of this DNA requires NS1. Notably, when the NP1 ORF was disrupted, an RF DNA band was detected but it was very weak ( Figure 2D , lane 6), suggesting that NP1 is also involved. When the VP2 ORF was knocked out, the ssDNA band disappeared, but this was not the case when VP1 was disrupted (VP2 was still expressed; Figure 2D , compare lanes 7 to 9), these findings are consistent with a role for Human bocavirus 1 (HBoV1) has been identified as one of the etiological agents of wheezing in young children with acute respiratory-tract infections. HBoV1 productively infects polarized primary human airway epithelia. However, no cell lines permissive to HBoV1 infection have yet been established. More importantly, the sequences at both ends of the HBoV1 genome have remained unknown. We have resolved both of these issues in this study. We have sequenced a full-length HBoV1 genome and cloned it into a plasmid. We further demonstrated that this HBoV1 plasmid replicated and produced viruses in human embryonic kidney 293 cells. Infection of these HBoV1 progeny virions produced obvious cytopathogenic effects in polarized human airway epithelia, which were represented by disruption of the epithelial barrier. Moreover, we identified an airway epithelial cell line supporting HBoV1 infection, when it was polarized. This is the first study to obtain the full-length HBoV1 genome, to demonstrate pathogenesis of HBoV1 infection in human airway epithelia, and to identify the first cell line to support productive HBoV1 infection. the capsid formation in packaging of the parvoviral ssDNA genome [36] [37] [38] . The presence of the ssDNA band in pIHBoV1-transfected HEK293 cells suggested that progeny virions were produced. To prove this, we carried out large-scale pIHBoV1 transfection and CsCl equilibrium centrifugation to purify the virus that was produced. We fractionated the CsCl gradient, and found the highest HBoV1 gc (1-5610 8 gc/ml) at a density of 1.40 mg/ml, which is typical of the parvovirus virion. Electron microscopy analysis revealed that purified virus displayed a typical icosahedral structure, with a diameter of ,26 nm ( Figure 2E ). Collectively, these findings confirm that we have generated a full-length clone of HBoV1 capable of replicating and producing progeny virus in transfected HEK293 cells. The infectivity of the HBoV1 virions purified from pIHBoV1transfected HEK293 cells was examined in polarized primary HAE, the in vitro culture model known to be permissive to HBoV1 infection [31] . Three sets (different donors, culture lots #B29-11, B31-11 and B33-11) of B-HAE were generated, and these were Figure 1 . Sequencing the terminal hairpins of the HBoV1 Salvador1 isolate. Sequence and predicted structure of the left-end, LEH (A&B), and right-end, REH (C), hairpins are shown and diagramed, with PCR primers indicated by arrowed lines. PCR products were analyzed by electrophoresis on 2% agarose gels; the expected DNA bands are indicated by arrowheads (D). In both the LEH and REH, nucleotides in red represent new sequences identified in this study, nucleotides in blue represent sequences identified from the head-tail junction of an HBoV1 episome [28, 29] , and nucleotides in black are the 59end and 39end sequences of the incomplete HBoV1 genome (GenBank: JQ411251). doi:10.1371/journal.ppat.1002899.g001 infected with HBoV1 from the apical side. Initially the B-HAE cultures were infected with various amounts of virus, and when a multiplicity of infection (MOI) of ,750 gc/cell was used, most of the cells (,80%) were positive for anti-NS1 staining (indicating that the viral genome had replicated and that genes encoded by it had been expressed) at 5 days post-infection (p.i.). This MOI was subsequently used for apical infection. Notably, B29-11, B31-11 and B33-11 HAE each supported productive HBoV1 infection ( Figures 3 and S3 ). Immunofluorescence (IF) analysis of infected B31-11 HAE at 12 days p.i. showed that virtually all the cells expressed NS1 and NP1 ( Figures 3A and 3B ), and that a good portion of the infected cells expressed capsid proteins (VP1/2; Figure 3C ). The production of progeny virus following HBoV1 infection was monitored daily by collecting samples from both the apical and basolateral chambers of the HAE culture and carrying out (2) been infected with Ad as indicated. Lanes 1-8 in panel B were analyzed on the same gel, and the gels shown in panels B&C were transferred and blotted together. Ten ng of the HBoV1 dsDNA genome (,5.6-kb), excised from pIHBoV1 using the SalI and XhoI sites, was used as a control (Ctrl) for DpnI digestion in panels A-C. (D) HEK293 cells were transfected with pIHBoV1 and its various mutants as indicated. At 48 h post-transfection, Hirt DNA was extracted and digested with (+) or without (2) DpnI, followed by Southern blotting using the HBoV1 dsDNA genome as a probe. dRF DNA, double replicative form DNA; mRF DNA, monomer replicative form DNA. (E) Negative staining electron micrograph. Purified HBoV1 particles were negatively stained and examined by a transmission electron microscopy. Bar indicates 100-nm. doi:10.1371/journal.ppat.1002899.g002 HBoV1-specific quantitative PCR (qPCR; Figures 4A and S3B ). In the case of B33-11 B-HAE, apical release was obviously initiated at 3 days p.i., then continued to increase to a peak of ,10 8 gc/ml at 5-7 days p.i., then decreased slightly through day 10 p.i. and was maintained at a level of ,10 7 gc/ml through day 22 p.i. ( Figure 4A ). The total virus yield from one Millicell insert of 0.6 cm 2 over a 24-h interval was greater than 2610 10 gc. This result suggested that productive HBoV1 infection of primary B-HAE is persistent. Notably, in the B-HAE cultures from both donors, virus was also continuously released from the basolateral side, keeping pace with apical secretion throughout, though at levels about one log lower than the release from the apical surface ( Figures 4A and S3B ). The genomes of the progeny virions released from infected B-HAE were amplified and sequenced using the primers listed in Figure 1 and primers spanning the NSVP genes between the termini. The result showed an identical sequence with that of the HBoV1 Salvador isolate (Genbank JQ923422). Additionally, no virus was detected in mock-infected B-HAE (data not shown). Taken together, these results demonstrate that the HBoV1 virions produced by pIHBoV1 transfection is capable of infecting polarized primary HAE cultures from cells derived from various donors and releasing identical progeny virions from infected primary HAE. More importantly, we found that productive HBoV1 infection was persistent. Although no gross cytopathic effects were observed in HBoV1infected B-HAE, histology analysis of mock-vs. HBoV1-infected epithelia (B33-11) revealed morphological differences: infected B-HAE did not feature obvious cilia at 7 days p.i., and was significantly thinner than the mock-infected one on average at 22 days p.i. ( Figure 4B ). We further monitored the transepithelial electrical resistance (TEER) during infection of B-HAE, and found that at 6 days p.i., it was reduced from a value of ,1,200 to ,400 V.cm 2 , while the mock-infected B-HAE maintained the initial TEER ( Figure 4C) . Notably, the decrease in TEER in the infected B-HAE was accompanied by an increase in HBoV1 secretion ( Figure 4A ). To confirm a role for HBoV1 infection in disruption of the barrier function of the epithelium, we examined the distribution of the tight junction protein Zona occludens-1 (ZO-1) [39] . Infected B-HAE showed dissociation of ZO-1 from the periphery of cells started from 7 days p.i., compared with mock-infected B-HAE ( Figure 5A ), which likely plays a role in reducing TEER. Cumulatively, these results demonstrate that HBoV1 infection disrupts the integrity of HAE and that this may involve breakdown of polarity and redistribution of the tight junction protein ZO-1. To confirm a role for HBoV1 infection in the loss of cilia, we examined expression of the b-tubulin IV, which is a marker of cilia [40, 41] . In HBoV1-infected B-HAE, expression of b-tubulin IV was drastically decreased at 7 days p.i., and was not detected at 22 days p.i., in contrast to that in mock-infected B-HAE ( Figure 5B ). These results confirmed that HBoV1 infection caused the loss of cilia in infected B-HAE. Notably, infected B-HAE showed changes of nuclear enlargement, which became obvious at 22 days p.i. (Figure 5 , DAPI), indicating airway epithelial cell hypertrophy. Collectively, we found that productive HBoV1 infection disrupted the tight junction barrier, lead to the loss of cilia and airway epithelial cell hypertrophy. These are hallmarks of respiratory tract injury when a loss of epithelial cell polarity occurs. Although primary HAE cultures support HBoV1 infection, their usefulness is limited by the variability between donors, tissue availability and high cost. We thus explored alternative cell culture models for their abilities to support HBoV1 infection. Using the purified HBoV1, we examined HEK293 cells, other common epithelial cell lines permissive to common respiratory viruses [42] , including HeLa, MDCK, MRC-5, LLC-MK2 and Vero-E6, and several transformed or immortalized human airway epithelial cell lines (A549, BEAS-2B, 16HBE14o- [43] , NuLi-1 and CuFi-8 [44] ), as well as primary NHBE cells for the ability to support infection in conventional monolayer culture. All were negative for HBoV1 infection as determined by IF analysis (data not shown). We next speculated that since some respiratory viruses infect polarized HAE but not undifferentiated cells [45] , some characteristics of the polarized epithelia may be critical for HBoV1 infection. We thus polarized immortalized cells (NuLi-1 and CuFi- 8) at an air-liquid interface (ALI) for one month. Once polarization was confirmed by detection of a TEER of .500 V.cm 2 , the cultures were infected with HBoV1, under the same conditions as used for primary B-HAE cultures. Notably, IF analysis revealed that at 10 days p.i., HBoV1-infected CuFi-HAE (differentiated from CuFi-8 cells) was uniformly positive for NS1 ( Figure 6A ), whereas the HBoV1-infected NuLi-HAE (differentiated from NuLi-1 cells) was not ( Figure S4) . Moreover, the CuFi-HAE did express HBoV1 NS1, NP1 and VP1/VP2 proteins (Figures 6B and 6C ). The kinetics of virus release from the apical surface was similar to that of a primary B-HAE infected with virus at a similar titer (maximally 2610 7 gc/ml), although virus release from the basolateral surface was undetectable ( Figure 6D ). HBoV1 infection also resulted in a decrease in the thickness of the epithelium ( Figure 6E ), and dissociation of the tight junction protein ZO-1 from the epithelial cell peripheries ( Figure 6F) . Collectively, these findings demonstrate that the immortalized cell line CuFi-8 [44] , when cultured and polarized at an ALI, supports HBoV1 infection, and recapitulates the infection phenotypes observed in primary HAE, including destruction of the airway epithelial structure. In this study, we have cloned the full-length HBoV1 genome and identified its terminal hairpins. Virions produced from transfection of this clone into HEK293 cells are capable of infecting polarized HAE cultures. Thus, we have established a reverse genetics system that overcomes the critical barriers to studying the molecular biology and pathogenesis of HBoV1, using an in vitro culture model system of HAE. It is notable that the HBoV1 terminal hairpins appear to be hybrid relicts of the prototype bocavirus BPV1 at the LEH, but of MVC at the REH [28] . Replication of HBoV1 DNA in HEK293 cells revealed typical replicative intermediates of parvoviral DNA. Although the head-tail junctions are unexpected in the replication of autonomous parvoviruses, they were likely generated during the cycle of rolling hairpin-dependent DNA replication [46] . Therefore, we believe that the replication of HBoV1 DNA basically follows the model of rolling hairpin-dependent DNA replication of autonomous parvoviruses, with terminal and junction resolutions at the REH and LEH, respectively [46] . The replication of parvoviral DNA depends on entry into S phase of the cell cycle or the presence of helper viruses [46, 47] . In this regard, it is puzzling that mature, uninjured airway epithelia are mitotically quiescent (,1% of cells dividing) [48] [49] [50] , as are the majority of the cells in polarized HAE (in the G0 phase of the cell cycle). However, recombinant adeno-associated virus (AAV; in genus Dependovirus of the family of Parvoviridae) infects HAE apically and expresses reporter genes [51] [52] [53] . Gene expression by recombinant AAV requires a conversion of the ssDNA viral genome to a doublestranded DNA form that is capable to be transcribed [54] . This conversion involves DNA synthesis. Hence, we hypothesize that HBoV1 employs a similar approach to synthesize its replicative form DNA. Notably, wild type AAV infected primary HAE apically and replicated when adenovirus was co-infected [55] . The exact mechanism of how HBoV1 replicates in normal HAE will be an interesting topic for further investigation. The airway epithelium, a ciliated pseudo-stratified columnar epithelium, represents the first barrier against inhaled microbes and actively prevents the entry of respiratory pathogens. It consists of ciliated cells, basal cells and secretory goblet cells that together with the mucosal immune system, provide local defense mechanisms for the mucociliary clearance of inhaled microorganisms [56] . The polarized ciliated primary HAE, which is generated by growing isolated tracheobronchial epithelial cells at an ALI for on average one month, forms a pseudo-stratified, mucociliary epithelium and displays morphologic and phenotypic characteristics resembling those of the in vivo human cartilaginous airway epithelium of the lung [57, 58] . Recent studies have revealed that this model system recapitulates important characteristics of interactions between respiratory viruses and their host cells [41, 45, [59] [60] [61] [62] . In the current study, we have examined primary B-HAE cultures obtained from three different donors. HBoV1 infection of primary B-HAE was persistent and caused morphological changes of the epithelia, i.e. disruption of the tight barrier junctions, loss of cilia and epithelial cell hypertrophy. The loss of the former, plasma membrane structures that seal the perimeters of the polarized epithelial cells of the monolayer, is known to damage the cell barrier necessary to maintain vectorial secretion, absorption and transport. ZO-1, which we monitored here, is specifically associated with the tight junctions and remains the standard marker for these structures. Similarly, cilia play important roles in airway epithelia, in that they drive inhaled particles that adhere to mucus secreted by goblet cells outward [63] . HBoV1 infection compromises barrier function, and thus potentially increases permeability of the airway epithelia to allergens and susceptibility to secondary infections by microbes. The observed shedding of virus from the basolateral surface of infected primary HAE, albeit at a lower level (,1 log lower than that from the apical surface), is consistent with the facts that HBoV1 infection disrupted the polarity of the pseudo-stratified epithelial barrier and resulted in the leakage to the basolateral chamber. This explanation is also supported by HBoV1 infection of CuFi-HAE, where disruption of the tight junction structure was less severe and virus was released only from the apical membrane. The induction of leakage by HBoV1 also suggests a mechanism that accounts for the viraemia observed in HBoV1-infected patients [5] . Further disease pathology could be accounted for by infection-induced loss of cilia of the airway epithelia; a lack of cilia is often responsible for bronchiolitis [64] [65] [66] . Therefore, our study provides direct evidence that HBoV1 is pathogenic to polarized HAE, which serves as in vitro model of the lung [57, 58] . Since HBoV1 is frequently detected with other respiratory viruses in infants hospitalized for acute wheezing [2, [10] [11] [12] , the apparent pathological changes observed in HBoV1-infected HAE suggest that prior-infection of HBoV1 likely facilitates the progression of co-infection-driven pathogenesis in the patient. The kinetics of virus release from the apical chamber of HAE infected with the progeny virus of pIHBoV1 (cloned from the clinical Salvador1 isolate) was similar to that following infection with the HBoV1 Bonn1 isolate, a clinical specimen [31] . We believe that our study of HBoV1 infection of primary HAE reproduces infection of the virus from clinical specimens. In addition, we generated virus from a pIHBoV1-b clone, which contains the NSVP genes from the prototype HBoV1 st2 isolate [1] . Infection of primary B-HAE with this st2 virus resulted in a level of virus production similar to that observed here using the Salvador1 isolate (data not shown). We believe that our study with the laboratory-produced HBoV1 Salvador1 represents infection of HBoV1 of clinical specimens in HAE. The MOI used for infection in the current study was high. However, it should be noted that this titer is based on the physical numbers of virion particles as there are no practical methods for determining the infectious titer of HBoV1 preparations. It should also be taken into consideration that extensive parvovirus inactivation occurs during the purification process, i.e. during CsCl equilibrium ultracentrifugation [67] . Virus infection of HAE most likely reflects HBoV1 infection of the lung airways in patients with a high virus load in respiratory secretions [5] . The fact that pIHBoV1 did not replicate well in undifferentiated human airway epithelial cells (Figures 2B and 2C) indicates that polarization and differentiation of the HAE is critical for HBoV1 DNA replication. Nevertheless, polarized NuLi-HAE, which is derived from normal human airway epithelial cells, did not support HBoV1 infection, but the CuFi-HAE derived from airway epithelial cells isolated from a cystic fibrosis patient did. The CuFi-HAE is unique relative to the others in that it retains the capacity to develop epithelia that actively transport in Na + but not Cl 2 because of the mutation in the cystic fibrosis gene [44] . Given the high complexity of the airway epithelium, we speculate that the permissiveness of HBoV1 infection is dependent on various steps of virus infection, e.g. attachment, entry, intracellular trafficking, and DNA replication of the virus. Nevertheless, a polarized CuFi-HAE model derived from the CuFi-8 cell line represents a novel stable cell culture model that is providing unexpected insights into the infection characteristics of HBoV1. Although HBoV1 infection of CuFi-HAE reproduced disruption of the barrier tight junctions like that seen also in primary B-HAE, the absence of virus on the basolateral side implies that in HAE the secretion of HBoV1 is apically polarized. We speculate that the milder damage of tight junctions in these cells might prevent virus release from the basolateral side of infected CuFi-HAE. Further studies will focus on understanding the permissiveness of CuFi-HAE to HBoV1 infection and on the reason for the ease of infection of an HAE with a cystic fibrosis phenotype. It has been shown that HBoV1 remains detectable in the upper airways of patients for weeks and months, even up to half a year [68] [69] [70] [71] . However, the mechanism behind this persistence, i.e. whether it is due to persistent replication and shedding, passive persistence after primary infection, or recurrent mucosal surface contamination, has remained unknown. Our results in in vitro HAE cultures showed that HBoV1 is able to replicate and shed from both the apical and basolateral surfaces at least for three weeks, supporting the notion that shedding of the virus from the airways is a long-lasting process. This may further explain why a high rate of co-infection, or co-detection, between HBoV1 and other respiratory viruses has been reported [5] . Since recombinant AAV persists as an episome in transduced tissues, which prolongs gene expression [72, 73] , it is possible that also the HBoV1 genome can be presented as an episome [29, 30] for long term expression and replication. Apparently, the mechanism underlying this feature of HBoV1 infection warrants further investigation. However, in contrast to the other human-pathogenic B19V, HBoV1 does not seem to persist in human tissues for many years [74] . In conclusion, our findings indicate that the innovative reverse genetics system for studying HBoV1 infection that we describe here will enable us to elucidate the mechanism of HBoV1 replication and pathogenesis in a polarized HAE. Our system mimics natural HBoV1 infection of the in vivo human cartilaginous airway epithelia. The pathogenesis of HBoV1 in co-infection with other respiratory viruses and in conditions of lung diseases is a focus of future study. [43] ), as well as NuLi-1 and CuFi-8 (Tissue and Cell Culture Core, Center for Gene Therapy, University of Iowa). NuLi-1 and CuFi-8 were immortalized from normal and cystic fibrosis human primary airway cells, respectively, by expressing hTERT and HPV E6/E7 genes [44] . Primary Clonetics normal human bronchial/tracheal epithelial cells (NHBE) were purchased from Lonza (Walkersville, MD). Cells were cultured in media following instructions provided by the supplier. Human airway epithelium cultures. Polarized primary HAE, termed as primary B-HAE, was generated by growing isolated human airway (tracheobronchial) epithelial cells (three HAE cultures were generated from different donors) on collagencoated, semipermeable membrane inserts (0.6 cm 2 , Millicell-PCF; Millipore, Billerica, MA), and then allowing them to differentiate at an air-liquid interface (ALI); this was carried out at the Tissue and Cell Culture Core of the Center for Gene Therapy, University of Iowa [44, 58, 75, 76] . After 3-4 weeks of culture at an ALI, the polarity of the HAE was determined based on the transepithelial electrical resistance (TEER) using an epithelial Volt-Ohm Meter (Millipore) and the relationship to infectability was assessed; a value of over 1,000 V.cm 2 was required for HBoV1 infection. CuFi-and NuLi-HAE were generated following the same method as above, but using the immortalized airway epithelial cell lines, CuFi-8 and NuLi-1, respectively. The primary B-, CuFi-, and NuLi-HAE were cultured, differentiated and maintained in (50%:50%) DMEM:F12 medium containing 2% Ultroser G (Pall BioSepra, Cergy-Staint-Christophe, France). A nasopharyngeal aspirate was obtained from a child with community-acquired pneumonia in Salvador, Brazil, who had an acute HBoV1 infection (seroconversion, viraemia, and over 10 4 gc of HBoV1 per ml of aspirate) [17] . Viral DNA was extracted according to a method described previously [77] . The sequence of the head-to-tail junction of the HBoV1 episome suggests that HBoV LEH and REH share similarities both in structure and sequence with that of the BPV LEH and MVC REH, respectively [27, 29] . Based on this information [28] , we designed primers to amplify the HBoV1 termini, which are shown in Table 1 and Figure 1 . The Phusion high fidelity PCR kit (NEB, Ipswich, MA) was used following the manufactures' instructions, to amplify the left-end hairpin (LEH) and the rightend hairpin (REH) of HBoV1. Briefly, the DNA denaturation at 98uC for 30 s was followed by 35 cycles of: denaturing at 98uC for 10 s; annealing at 55uC for 15 s; and extension at 72uC for 30 s. Following the final cycle, extension was continued at 72uC for 10 min. The PCR products were analyzed by electrophoresis in a 2% agarose gel. DNA bands were extracted using the QIAquick gel extraction kit (Qiagen, Valencia, CA), and the extracted DNA was directly sequenced at MCLAB (South San Francisco, CA), using primers complementary to the extended sequences on the forward and reverse amplification primers. PCR-generated DNA was cloned in pGEM-T vector (Promega, Madison, WI), and DNAs isolated from cultures of individual clones were subsequently sequenced. Construction of the pBB vector. We first constructed a pBBSmaI vector by inserting a linker of 59-SalI-SacII-KpnI-SmaI-ApaI-SphI-KpnI-HindIII-XhoI-39 into a vector backbone (pProEX HTb vector; Invitrogen) generated from the B19V infectious clone pM20 [78] by removing all of the B19V sequence (SalI-digestion). All cloning work was carried out in the Escherichia coli strain of Sure 2 (Agilent, La Jolla, CA). All the nucleotide numbers of HBoV1 refer to the HBoV1 full-length genome (GenBank accession no.:JQ923422). Cloning of the left-end hairpin ( Figure S2B ). The DNA fragment SalI-BglII-nt93-518(BspEI)-576-XhoI-HindIII (containing the HBoV1 sequence nt 93-576), was amplified from the viral DNA and inserted into SaII/HindIII-digested pBBSmaI, to produce pBB2.1. Another DNA, SalI-nt1-86-BclI (containing HBoV1 nt 1-86 sequence), was synthesized according to the LEH sequence obtained in Figures 1A and 1B , and placed between the SalI and BglII sites in pBB2.1, with ligation of the BclI and BglII sites reproducing the HBoV1 sequence nt 87-92. The resultant plasmid harboring the 59 HBoV1 nt 1-576 sequence with an intact LEH is designated pBB-LEH. Cloning of the right-end hairpin ( Figure S2C ). The DNA fragment SalI-nt4097-4139(BglII)-5427(KasI)-ApaI (containing the HBoV1 nt 4097-5427 sequence) was amplified from viral DNA and inserted into SaII/ApaI-digested pBBSmaI, resulting in pBB2.2. Another DNA fragment, ApaI-nt5460(KasI)-5543-XhoI (containing HBoV1 nt 5460-5543 sequence) was synthesized based on the REH sequence ( Figure 1C ) and placed between the ApaI and HindIII sequences in pBB2.2, resulting in pBB-REH(D5428-5459). The missing short fragment between the two KasI sites encompassing nt 5428-5459 was recovered by a synthesized KasI linker based on the REH sequence ( Figure 1C ) and inserted into KasI-digested pBB-REH(D5428-5459). The resultant plasmid harboring the 39 HBoV1 nt 4097-5543 sequence with an intact REH is designated pBB-REH. Cloning of the pIHBoV1 ( Figure S2D ). The HBoV1 DNA fragment SalI-nt1-518(BspEI)-576-XhoI, which was obtained from SalI/XhoI-digestion of pBB-LEH, was ligated into SalI-digested pBB-REH, resulting in pBB-LEH(BspEI/BglII)REH. The larger fragment produced by digestion of this plasmid with BspEI/BglII was ligated to the HBoV1 DNA fragment nt 518(BspEI)-4139(BglII), which was amplified from the viral DNA. The final construct containing the full-length HBoV1 (nt 1-5543) was designated pIHBoV1. Construction of pIHBoV1 mutants. pIHBoV1NS1(2) and pIHBoV1NP1 (2) were constructed by mutating HBoV1 nt 542 from T to A, and nt 2588 from G to A, resulting in stop codons that lead to early termination of the NS1 and NP1 ORFs, respectively. Similarly, pIHBoV1VP1(2) and pIHBoV1VP2 (2) were generated by mutating HBoV1 nt 3205 from T to A, and nt 3540 from T to G, disrupting VP1 and VP2 ORFs, respectively. Cells grown in 60-mm dishes were transfected with 2 mg of plasmid as indicated in Figure 2 ; the Lipofectamine and Plus reagents (Invitrogen/Life Technologies, Carlsbad, CA) were used as previously described [79] . For some of the transfection experiments, HEK293 cells were cotransfected with 2 mg of pHelper plasmid (Agilent), which contains the adenovirus 5 (Ad5) E2a, E4orf6, and VA genes, or infected with adenovirus type 5 (Ad) at an MOI of 5 as previously described [79] . Low molecular weight (Hirt) DNA was extracted from transfected cells, digested with DpnI (or left undigested) and analyzed by Southern blotting as previously described [80] . Cells were lysed, separated by SDS-8% polyacrylamide gel electrophoresis (PAGE), and blotted with antibodies as indicated as previously described [81] . HEK293 cells were cultured on fifteen 150-mm plates in DMEM-10%FCS, and transfected with 15 mg of pIHBoV1 per dish using LipoD293 (SignaGen, Gaithersburg, MD). After being maintained for 48 h at 5% CO 2 and 37uC, the cells were collected, resuspended in 10 ml of phosphate buffered saline, pH7.4 (PBS), and lysed by subjecting them to four freezing (2196uC) and thawing (37uC) cycles. The cell lysate was then spun at 10,000 rpm for 30 min. The supernatant was collected and assessed on a continuous CsCl gradient. In brief, the density was adjusted to 1.40 g/ml by adding CsCl, and the sample was loaded into an 11ml centrifuge tube and spun in a Sorvall TH641 rotor at 36,000 rpm, for 36 h at 20uC. Fractions of 550 ml (20 fractions) were collected with a Piston Gradient Fractionator (BioComp, Fredericton, NB, Canada), and the density of each was determined by an Abbe's Refractometer. Viral DNA was extracted from each fraction and quantified with respect to the number of HBoV1 gc, using HBoV1-specific qPCR as described below. Those fractions containing the highest numbers of HBoV1 gc were dialyzed against PBS, and then viewed by electron microscope and used to infect HAE cultures. Observation by electron microscopy (EM) The final purified virus preparation was concentrated by ,5fold, and adsorbed for 1 min on a 300-mesh copper EM grid coated with a carbon film, followed by washing with deionized water for 5 s and staining with 1% uranyl acetate for 1 min. The grid was air dried, and was inspected on a 200 kV Tecnai F20 G2 transmission electron microscope equipped with a field emission gun. Fully differentiated primary B-(each of the three distinct subtypes), CuFi-and NuLi-HAE were cultured in Millicell inserts (0.6 cm 2 ; Millipore) and inoculated with 150 ml of purified HBoV1 (1610 7 gc/ml in phosphate buffered saline, pH7.4; PBS) from the apical surface (at a multiplicity of infection, MOI, of ,750 gc/cell; an average of 2610 6 cells per insert). For each of the HAE, a 2-h incubation was followed by aspiration of the virus from the apical chamber and by three washes of the cells with 200 ml of PBS to remove unbound virus. The HAEs were then further cultured at an ALI. For conventional monolayer cells, cells cultured in chamber slides (Lab-Tek II; Nalge Nunc) were infected with purified HBoV1 at an MOI of 1,000 gc/cell. After HBoV1 infection, ALI membranes were fixed with 3.7% paraformaldehyde in PBS at room temperature for 15 min. The fixed membranes were cut into several small pieces, washed in PBS three times for 5 min, and permeabilized with 0.2% Triton X-100 for 15 min at room temperature. The membranes were then incubated with primary antibody at a dilution of 1:100 in PBS with 2% FCS for 1 h at 37uC. This was followed by incubation with a fluorescein isothiocyanate-or rhodamine-conjugated secondary antibody. Confocal images were taken with an Eclipse C1 Plus confocal microscope (Nikon, Melville, NY) controlled by Nikon EZ-C1 software. Primary antibodies used were anti-(HBoV1) NS1, NP1 and VP1/2 antibodies, as reported previously [82] . For infected cells cultured in chamber slides, IF analysis was carried out as previously described [83] . Virus samples were collected from both the apical and basolateral surfaces at multiple time points. Apical washing and harvesting was performed by adding 200 ml of PBS to the apical chamber, incubating the samples for 10 min at 37uC and 5% CO 2 , and removing and storing the 200 ml of PBS from the apical chamber. Thereafter, 50 ml of medium were collected from each basolateral chamber. Aliquots (100 ml apical or 50 ml basolateral) of the samples were incubated with 25 units of Benzonase (Sigma, St Louis, MO) for 2 h at 37uC, and then digested with 20 ml of proteinase K (15 mg/ ml) at 56uC for 10 min. Viral DNA was extracted using QIAamp blood mini kit (Qiagen), and eluted in 100 ml or 50 ml of deionized H 2 O. The extracted DNA was quantified with respect to the number of HBoV1 gc, by a qPCR method that has been used previously [84] . Briefly, the pskHBoV1 plasmid [82] , which contains the HBoV1 sequence (nt 1-5299), was used as a control (1 gc = 5.4610 212 mg) to establish a standard curve for absolute quantification with an Applied Biosystems 7500 Fast system (Foster City, CA). The amplicon primers and the PrimeTime duallabeled probe were designed by Primer Express (version 2.0.0; Applied Biosystems/Life Technologies) and synthesized at IDT Inc. (Coralville, Iowa). Their sequences are as follows (GenBank: JQ411251): forward primers, 59-GCA CAG CCA CGT GAC GAA-39 (nt 2391 to 2408); reverse primer, 59-TGG ACT CCC TTT TCT TTT GTA GGA-39 (nt 2466 to 2443); and PrimeTime probe, 59 6FAM-TGA GCT CAG GGA ATA TGA AAG ACA AGC ATC G-39 Iowa Black FQ (nt 2411 to 2441). Premix Ex Taq (Takara Bio USA, Madison, WI) was used for qPCR following a standard protocol. 2.5 ml of extracted DNA was used in a reaction volume of 25 ml. On the last day of infection, the HAE on the Millicell inserts were washed with PBS and fixed in 4% paraformaldehyde for ,30 min. The fixed membranes were cut into several small pieces, and washed with PBS three times. Each membrane fragment was transferred to 20% sucrose in a 15-ml conical tube and allowed to drop to the bottom; it was then embedded vertically in cryoprotectant OCT in an orientation that enabled sectioning of the membrane perpendicular to the blade. Cryostat sections were cut at a thickness of 10 mm, placed onto slides, and stained with hematoxylin and eosin (H&E). Images were taken with a Nikon 80i fluorescence microscope at a magnification of 660. Figure S1 Sequencing of PCR DNA fragments. PCR DNA fragments indicated by arrowheads in Figure 1D were extracted and sequenced. A representative result of sequencing is shown in each chromatogram. The sequences between the arrows in the chromatograms (A-C) show the sequences which are complementary to those sequences between the arrows in the hairpin drawings in Figure 1A -C, respectively. (TIF) Figure S2 Construction of a full-length pIHBoV1 clone. (A) HBoV1 genome. The full-length genome of HBoV1 is diagramed with structures of the left-end hairpin (LEH) and rightend hairpin (REH) in a form of negative ssDNA from 39end to 59end. Restriction enzyme sites of BspEI and BglII in the replicative form (RF) DNA are shown. (B) Cloning of the LEH. PCR-amplified DNA fragments from the LEH, shown in red, were ligated into pBBSmaI or its derivative. (C) Cloning of the REH. PCR-amplified or synthesized HBoV1 DNA fragment from the REH, shown in blue, were ligated into pBBSmaI or its derivatives. (D) Cloning of the pIHBoV1. The pIHBoV1 was constructed by ligating HBoV1 DNA nt 1-517 digested from pBB-LEH and nt 518-4139 amplified from viral DNA extract (HBoV1 Salvador isolate) into the pBB-REH that contains HBoV1 REH (nt 4140-5543). All the numbers are nucleotide numbers of the HBoV1 genome (Genbank JQ923422). (TIF) Figure S3 Kinetics of virus release from HBoV1 infection of primary B31-11 and B29-11 HAE. Primary B-HAE (donor B31-11 or B29-11) was infected with purified HBoV1 at an MOI of ,750 genome copy numbers (gc)/cell. Virus was collected from the apical chamber (A), or from both the apical and basolateral chambers (B) for detection of nuclease-resistant viral gc. Averages and standard deviations are shown. ND, not determined. (TIF) Figure S4 Immunofluorescence analysis of HBoV1infected HAE polarized from NuLi-1 cells (NuLi-HAE). NuLi-1 cells were polarized by growth at an ALI for 4 weeks on Millicell inserts of 0.6 cm 2 , until a transepithelial electrical resistance (TEER) of .500 V.cm 2 was detected. Polarized HAE was infected with purified HBoV1 at an MOI of ,750 gc/cell. At 10 days p.i., infected NuLi-HAE was fixed and stained with an anti-(HBoV1)NS1 antibody. Nuclei were stained with DAPI and cells were visualized by confocal microscopy at a magnification of 640. (TIF)

Personal, Occupational, and Public Health Perspectives on Dealing with the First Case of Influenza A (H1N1) in the United Arab Emirates

New epidemics of infectious diseases often involve health care workers. In this short communication we present a case report of a health care professional who became the first case of influenza H1N1 virus to be notified in the United Arab Emirates. There are several issues related to workplace considerations and general public health, including preventive measures, the need for isolation of the patient, dealing with contacts, return to work, and communication with the workforce.

In recent years influenza viruses have circulated in seasonal (H3N2, H1N1) and avian (including H5N1) forms. There has been concern that Influenza A (H5N1), a worldwide cause of large poultry outbreaks, which by December 2009 had affected 467 persons (282 deaths), would drift or shift to become the next pandemic strain [1] . However in April 2009 'Swine flu' caused by a new strain of influenza A, Pandemic (H1N1) 2009 emerged. This has now become the dominant strain producing an illness that is transmitted in the same way as seasonal influen- This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. za, which in most cases is mild, which can be effectively treated with antiviral drugs and for which a vaccine is now available. By the end of 2009 many countries were still reporting disease activity and an impact on health-care services [2] . In the early days of the H1N1 pandemic, when there was uncertainty about the infectivity and virulence of the new virus, a more precautionary approach to management was advocated. This included laboratory testing of suspected cases, contact tracing, isolation of cases and contacts, anti-viral medication for treatment and prophylaxis, and clinical surveillance and follow-up. In this short case report we describe the personal experience and management of the first case of H1N1 reported in the United Arab Emirates (UAE). The patient was a 48 year-old male academic public health physician who had just returned to the Middle East after www.e-shaw.org spending a week with his family in Saskatoon in Canada. His journey to the UAE was via Calgary and Heathrow airport in London, UK. He started feeling lethargic, and developed a sore throat, with cough and high fever for around 10 hours since the night of his arrival in Dubai, UAE. This led him to consult the on-duty infectious disease consultant at the Emergency Department of a local hospital at around 8:00 am the following day. The consultation included a discussion of any possible exposure to H1N1 since Canada was recognized then as experiencing a large number of cases of the infection. A combined influenza A&B antigen screen on a nasopharyngeal swab was positive, and an additional swab and a blood sample were then sent for further confirmatory testing. He was prescribed Oseltamivir 75 mg orally twice daily for five days, azithromycin 500 mg daily and paracetamol 500 mg three times daily for three days, and advised to remain at home until the confirmatory test results were available. By the next morning, the patient's fever and sore throat had subsided and he was feeling better. Despite the very low, but nevertheless real risk of having 'swine flu', the patient had to make some important difficult decisions regarding his state of health and his work deadlines. His work place was a university campus and as he had no lectures that day, he had no need to be in contact with any students. All scheduled appointments on his calendar for the day were cancelled, but he decided to proceed with a ten-minute scheduled presentation to six of his peers regarding a large research grant proposal. A mask was not worn during the presentation, and he returned home immediately after the event. The patient was alone at home but one of his relatives came to visit him unannounced, accompanied by his wife and a ten year old child from a neighboring town. They stayed at his home for that night, as the distance for return travel was considerable. On the next day the patient received a call from the Health Authority confirming Influenza A (H1N1) infection and he was therefore in the unenviable although historical position of being the first reported case of H1N1 infection in the UAE. The patient was admitted to hospital with airborne and contact isolation, where he completed the rest of the maximum recommended 10 days quarantine period. The visiting couple and child also had to stay at the patient's home for 10 days of quarantine and all also received prophylactic medicine (Oseltamivir). No lab tests were advised. As a public health physician, the index case had considered the H1N1 situation before commencing his travels to Canada. At that time (May 8 2009), the World Health Organization (WHO) did not recommend restricting travel, although some individual national authorities were advising against non-essential travel. The advice on the various websites seemed very pragmatic: observe basic hygiene, hand-washing and cough etiquette; do not travel when ill and seek medical advice if you become ill after your return. The patient's route to Canada took him through London (34 cases reported in the UK at that time) and Toronto (15 cases in Ontario) to Saskatoon (2 cases). By the time he was due to return to the UAE from Saskatoon via Calgary, the number of cases in Canada had increased from 242 to 496 with 19 in Saskatchewan and 67 in Alberta. During his stay in Saskatoon he did not recall meeting anyone with respiratory symptoms and he was quite well on his journey back to Dubai. He was therefore not certain where and from whom he caught the infection. This case raised several issues related to workplace and general public health. Measures taken by the UAE government to prevent an influenza epidemic include the installation of thermal scanners at Dubai, Sharjah and Abu Dhabi airports (three major international airports in the United Arab Emirates). The individual was afebrile and symptom-free on arrival at the airport, and so was not detained for further enquiry. The thermal scanners will detect individuals with fever from whatever cause, but will not necessarily detect those with early H1N1 infection, especially if they are afebrile [3] . Effective and timely communication is essential to allay unwarranted concerns from the public and at the workplace. Queries from the media were channeled to a senior member of the administration from the office of the Dean -to ensure consistency in the information provided. He was briefed by public health physicians, occupational health physicians and hospital clinicians dealing directly with the case. A central news release was provided to staff and students on H1N1 reiterating the importance of hygiene in regards to limitation of transmission. The workplace was a university campus. This case did not have any lectures or meetings with students. Contact with a few coworkers was transient (not more than 15 minutes in the same area). These contacts were counselled on the low likelihood of acquiring the infection. They were informed about seeking medical advice if they had any other reasons for concern or if they developed H1N1 symptoms. Doctors, nurses and ancillary healthcare workers looking after the case while in hospital were briefed on hygiene and infection control procedures. N-95 masks, gloves and gowns were provided to health-care staff. The Health department took prompt action. Family members with close contact were quarantined at home. They were given a prophylactic course of Oseltamivir. Adequate supplies of food and provisions and maintenance of phone communiwww.e-shaw.org cation was confirmed. The public health department dealt with general queries from the public. Official release of information and contact with the WHO was through the federal Ministry of Health. The airline that transported the case from Canada to the UAE sought to contact passengers in the rows adjacent to the passenger's allocated seat. None of those who were traced developed any flu-like illness within the incubation period following the timing of the flight. Where new epidemics of infectious diseases appear, history has shown that the cases have often included healthcare workers, and their family members [4] . The index case for Ebola infection was a hospital laboratory worker, and secondary cases occurred in other healthcare workers and within the family. Two-thirds of the deaths from the early outbreaks of Ebola infection occurred in healthcare staff. The early cases of SARS and H5N1 infection included doctors and nurses [5] [6] [7] . The likelihood of healthcare staff being affected in such infections is high, especially in the absence of adequate preventive measures, or if there is poor compliance with recommended precautions. In this particular first reported case of H1N1 infection in the UAE, prompt and appropriate action resulted in the individual being treated, the risk of transmission being reduced, and the provision of information being timely and adequate. None of the known contacts developed signs and symptoms of the disease. It was not possible to contact the taxi driver who shared the same vehicle with the case during the hour long journey from the airport home, but there were no reports of infection in any Dubai taxi driver in the 2 weeks following the journey. We now believe that even if they are infectious, clinicians who practice good respiratory and hand hygiene will limit the risk of transmission to others. Standard and droplet precautions should be in place [8] . Standard Precautions minimize exposure to potentially infected blood and body fluids and include hand hygiene and the use of appropriate personal protective equipment. Droplet precautions require that a medical mask is worn when working within one meter of the patient and that when performing aerosol-generating procedures, further measures are taken including the use of eye protection, N-95 masks or other equivalent or more effective respirators and other personal protective equipment. In addition, respiratory or cough etiquette should be observed so that all persons cover their mouth and nose with a disposable tissue when coughing or sneezing, and then disposing the used tissue promptly. Within the healthcare setting, administrative, environmental and engineering controls such as frequent cleaning of work areas should also be in place. Generally it will not be appropriate to conduct contact tracing of patients or to provide anti-viral prophylaxis. However if there has been a particular type of contact between a healthcare worker and a patient (for example intubation) or a patient is at high risk of severe or complicated infection, then further risk assessment is indicated with a view to offering prophylaxis. An alternative approach, if practical, is to monitor exposed persons and administer antiviral treatment when symptoms develop. When a vaccine becomes available the first priority should be to immunize healthcare staff. When pandemic influenza is widespread in a community it will inevitably have consequences for the workplace not least because that is a setting where transmission can occur. In these circumstances occupational health practitioners should be prepared to lead a consistent and proportionate response. Staff with influenza will be diagnosed on the basis of symptoms. The clinical diagnostic criteria are fever (≥ 38 o C) or a history of fever and two or more symptoms of an influenza-like illness i.e. cough, sore throat, headache etc. Those who satisfy this case definition should be sent home and advised not to work until fully recovered. A risk assessment should be carried out and the risk of transmission to other staff members should be considered in terms of the excess risk compared to acquiring the infection from other community sources. Stories about the new H1N1 case in town appeared daily and reflected public anxiety. The media can play an important role in allaying the fears of the community by providing adequate and accurate information. The installation of thermal scanners at points of entry has their limitations, and is not recommended by the WHO. Studies indicate many of its drawbacks, including a low positive predictive value of 3.5% [2] . An unpublished population study carried out in the UAE during October 2009 by medical students investigated the impact of the recent H1N1 pandemic on the parents of primary school children. They found that while the majority of parents had good knowledge of H1N1 and its mode of transmission, many had mistaken beliefs about the origin of the virus, for example thinking that it had been genetically engineered. Parents reported changing their behaviour because of H1N1, taking measures such as cancelling travel plans and restricting socializing. Also, while most had confidence in the way in which the authorities had managed the pandemic, they continued to worry that their families were at risk of infection and were not persuaded of the safety of available vaccines. In conclusions, as for many epidemics, dealing with initial cases is often a key to successful subsequent management of further outbreaks. This case documents the experience of a public health physician as a patient in an infectious disease epidemic, with lessons for occupational and public health management. The lack of further transmission from this first case in the UAE may be a combination of good and effective public health intervention, or serendipity. Even though H1N1 has high infectivity with low case fatality rates, the number of cases globally declined, and WHO declared the end of the pandemic on 10 th August 2010.

Human Anti-CCR4 Minibody Gene Transfer for the Treatment of Cutaneous T-Cell Lymphoma

BACKGROUND: Although several therapeutic options have become available for patients with Cutaneous T-cell Lymphoma (CTCL), no therapy has been curative. Recent studies have demonstrated that CTCL cells overexpress the CC chemokine receptor 4 (CCR4). METHODOLOGY/PRINCIPAL FINDINGS: In this study, a xenograft model of CTCL was established and a recombinant adeno-associated viral serotype 8 (AAV8) vector expressing a humanized single-chain variable fragment (scFv)-Fc fusion (scFvFc or “minibody”) of anti-CCR4 monoclonal antibody (mAb) h1567 was evaluated for curative treatment. Human CCR4(+) tumor-bearing mice treated once with intravenous infusion of AAV8 virions encoding the h1567 (AAV8-h1567) minibody showed anti-tumor activity in vivo and increased survival. The AAV8-h1567 minibody notably increased the number of tumor-infiltrating Ly-6G(+) FcγRIIIa(CD16A)(+) murine neutrophils in the tumor xenografts over that of AAV8-control minibody treated mice. Furthermore, in CCR4(+) tumor-bearing mice co-treated with AAV8-h1567 minibody and infused with human peripheral blood mononuclear cells (PBMCs), marked tumor infiltration of human CD16A(+) CD56(+) NK cells was observed. The h1567 minibody also induced in vitro ADCC activity through both mouse neutrophils and human NK cells. CONCLUSIONS/SIGNIFICANCE: Overall, our data demonstrate that the in vivo anti-tumor activity of h1567 minibody is mediated, at least in part, through CD16A(+) immune effector cell ADCC mechanisms. These data further demonstrate the utility of the AAV-minibody gene transfer system in the rapid evaluation of candidate anti-tumor mAbs and the potency of h1567 as a potential novel therapy for CTCL.

Cutaneous T cell lymphomas (CTCLs) are a clinically heterogeneous group of lymphoproliferative malignancies characterized by the clonal accumulation of mature and skin-homing memory T cells. Mycosis fungoides (MF), which is the most common and indolent form of CTCL, accounts for 50%-60% of CTCL cases [1] ; primary cutaneous CD30 + lymphoproliferative disorders, more specifically primary cutaneous anaplastic large cell lymphoma (PC-ALCL) -the second most common CTCL, accounts for circa 30%; and Sézary syndrome, which is an aggressive leukemic variant of CTCL, affects approximately 5% of patients. These patients exhibit significant immune dysfunction [2, 3] because of the global dysregulation of T cells, which is due to an unknown etiology [4, 5] . Bacterial sepsis is the terminal event in most patients with advanced cancer. Current therapies for patients with advanced CTCL, including its leukemic variant, are only palliative, and extensive long-term remissions are rare. The poor 5-year survival rate of these patients receiving existing therapies clearly emphasizes the need for the development of new targeted therapies in this fatal disease [6] . Over the past few years, several studies have described the expression of chemokine receptors in the skin and blood of CTCL patients, including the uniformly high expression of CC chemokine receptor 4 (CCR4) [7, 8, 9, 10] . CCR4 is highly expressed in both leukemic CTCL including Sézary syndrome and in MF, both in the very early stages (patch and plaque stages) of the disease and in large cell transformations [7, 8, 10, 11, 12] . It is also expressed on circa 60% of PC-ATCL cells [1] . In a recently published consensus article regarding the classification of CTCL, it is clear that CCR4 is expressed in the vast majority of CTCL cells, regardless of their histological subtype [1] . On the other hand, expression of CCR4 is limited amongst non-malignant cells [13] . It is not present on neutrophils, monocytes, or B cells [14] . It is absent on naïve T cells, and present on fewer than half of all memory T cells [15] . expression of CCR4 by tumor cells is associated with their skin involvement, CCR4 also has an important role in normal and tumor immunity [13, 14] . CCR4 is expressed at high levels on T regulatory cells (Tregs) that can migrate to tumor cells that secrete the CCR4 chemokines CCL17 and CCL21 to facilitate evasion from immune surveillance [16, 17] . High expression of the these CCR4 ligands has been detected in CTCL lesions [11] , breast cancer [16] , ovarian cancer [17] and oral squamous cell carcinoma [18] . Thus, targeted therapy against CCR4 may be an attractive treatment option for these malignancies, not only to directly kill the CCR4 + tumor cells, but also to overcome the suppressive effect of CCR4 + Tregs on the host antitumor immune response. Monoclonal antibody (mAb)-based immunotherapies have become the standard therapy in an increasing number of human cancers [19, 20] . Tumor targeting with a human mAb directed against tumor-associated markers, such as CCR4, might provide a powerful therapeutic strategy against CTCL. In this study, we used recombinant adeno-associated viral (AAV) vector-mediated antibody gene transfer into SCID-BEIGE mice to evaluate the effectiveness of h1567, a novel humanized anti-CCR4 mAb to inhibit CCR4 + tumor cell growth and increase survival. The CCR4-specific antibody gene was packaged into an AAV vector and then delivered by a single direct intravenous (i.v.) injection which leads to the endogenous synthesis and durable expression of therapeutic antibody levels for months. Intravenous delivery of this h1567 minibody-encoding AAV vector allowed for rapid and accurate assessment of its therapeutic potential, thereby avoiding ex vivo manipulations involved in the production and purification of therapeutic mAbs. In vivo studies using therapeutic mAb gene transfer after CCR4+ tumor cell implantation demonstrated the potent antitumor activity of the mAb h1567. In addition, the in vivo effector cells that mediate tumor cell killing through h1567 Fc binding to Fcc receptors, namely FccRIIIa (CD16A), were delineated. These studies suggest that mAb 1567 can serve as an effective antibodydirected therapy for immunodepleting malignant CTCL cells and may minimize collateral damage to the already compromised immune system. Furthermore, in the context of anti-cancer mAb therapies that require frequent and repeated administration, this AAV-based therapeutic antibody gene transfer strategy might serve as an alternative platform for their delivery. In vitro and in vivo Expression of AAV8-encoding Anti-CCR4 h1567 A modified scFvFc minibody format was used as the antibody moiety in the AAV8 vector, in which the V domains of heavy (VH) and light (VL) chains of the humanized scFv h1567 were fused to the coding region of the hinge and constant domains 2 and 3 (CH2 and CH3) of the human IgG1 heavy chain, to yield bivalent binding to the target molecule hCCR4 (Figure 1a ) (DK. Chang et al., in press). The resulting recombinant AAV8 vector was used for both in vitro protein synthesis and virus production for in vivo antibody gene delivery. In a pilot dosing study, nude mice received a single injection of two different concentrations of AAV8-h1567 via intravenous tail vein injection. Serum h1567 minibody levels were followed for 15 weeks. H1567 minibody levels rose for the first 2-3 weeks, reaching levels of circa 65 and 96 ug/ml for the low (0.8610 11 vg/mouse) and high (2.0610 11 vg/mouse) vector doses, respectively and then through the remaining weeks of the study leveled off at near peak levels for the high dosed vector and circa 1/3rd that level (,35 ug/ml) for the low dosed vector ( Figure S1 ). Because 2610 11 vg per mouse gave higher serum levels of h1567, this vector concentration was used in the subsequent in vivo studies. CCR4 + Mac-1 tumor cells grow well in SCID-BEIGE mice and therefore we established a SCID-BEIGE/Mac-1 xenograft tumor model to evaluate the efficacy of AAV8-h1567 therapeutic minibody gene transfer. In SCID-BEIGE mice treated with a single intravenous tail vein injection of the AAV8 vectors, a timedependent increase in serum concentrations of the control 11A and h1567 minibodies, reaching steady state levels of circa 50 ug/ ml after 7-14 days and remaining at those peak levels through day 28, the last day of the study ( Figure 1b) . The control 11A is a irrelevant minibody that is directed against SARS Spike protein [21] . To determine whether the AAV8-minibody transduction in vivo could result in production of properly folded scFvFc, protein A-purified minibodies recovered from serum of SCID-BEIGE mice three weeks following intravenous delivery of AAV8 vectors were analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting. As shown in Figure 1c , when examined under reducing conditions, the 11A and h1567 minibodies recovered from both in vitro and in vivo sources showed bands at the expected size for scFvFc, circa 60 kD. Analysis under non-reducing conditions showed dimer formation (mol wt circa 120 kD), thereby confirming that the minibodies were divalent in vitro and in vivo ( Figure 1c ). In addition, the ease of recovery of the AAV8-derived minibodies from serum using affinity purification on protein A, their reactivity on Western blot with the anti-human Fc antibody, and their stable dimer formation confirms the proper folding and structural integrity of their CH2-CH3 domains (Figure 1c and 1d ). Binding Activity of h1567 Minibody in Serum Following AAV8-mediated Gene Transfer To determine the functional integrity of the AAV8-derived scFvFc minibodies, sera obtained from mice 14 days after in vivo AAV8 transduction were examined for the level of binding to CCR4 by flow cytometry. As shown in Figure 1e , the secreted h1567 minibody in the mouse serum could specifically bind to the CCR4 + Mac-1 cells and CCR4 + 293T cells but not to parental 293T cells, indicating that the scFv domain was correctly folded and that it retained full antigen-binding activity. Irrelevant 11A minibody, which served as a negative control, did not bind to CCR4-expressing cells. The therapeutic effects of AAV8-h1567 gene transfer were next evaluated in vivo in SCID-BEIGE mice that carried subcutaneously implanted Mac-1 tumor xenografts. Groups of 4 mice were given a single intravenous injection of AAV8-h1567 or control AAV8-11A vector on day 7 after tumor inoculation and tumor volume was assessed twice weekly. As shown in Figure 2a , a single injection of AAV8-h1567 resulted in significantly reduced tumor growth compared with AAV8-11A treated mice or PBS control treated mice (P,0.01 at day 18, P,0.0005 at day 21). Mouse survival was monitored for up to 2 months. Tumor-bearing mice treated with AAV8-h1567 significantly outlived (P,0.005) mice treated with AAV8-11A or untreated mice ( Figure 2b ). Since SCID-BEIGE mice lack T and B lymphocytes as well as functional natural killer (NK) cells, it is possible that the CCR4 + Mac-1 tumor cells were eliminated by h1567 through neutrophildependent ADCC as neutrophils are intact in SCID-BEIGE mice and they express FccRIIIA receptors which have been shown to mediated ADCC [22, 23] . Tumor sections were excised 21 days after AAV8 gene transfer and analyzed histologically for expression of Ly6G, a member of the Ly-6 family of glycosylphosphatidylinositol (GPI)-anchored proteins expressed on murine neutrophils [24, 25] . Immunostaining of tumors sections with neutrophil-specific Ly-6G mAb confirmed infiltration of neutrophils into tumors treated with AAV8-h1567 (Figure 2c , upper-left and middle panels) but not with AAV8-11A (Figure 2c , lower-left and right panels). Quantification of the neutrophil infiltration demonstrated a marked accumulation of Ly-6G+ staining cells only in the h1567 treated mice (Figure 2d) . To further assess the h1567-mediated, mouse neutrophildependent tumor cell killing, in vitro ADCC assay was carried out using purified SCID-BEIGE mouse neutrophils and h1567 minibody. Coculturing Mac-1 cells with mouse neutrophils in the presence of h1567 at the effector to target ratio of 80:1 resulted in significant neutrophil-mediated ADCC as measured by lactate dehydrogenase (LDH) release from Mac-1 cell (Figure 2e ). Control 11A minibody was not able to induce neutrophilmediated cytotoxicity. These in vitro results correlate with the observed anti-tumor activity in vivo and suggest that the antitumor activity of the h1567 minibody in this CTCL murine model is mediated, at least in part, through Fcc receptor IIIA (CD16A) engagement on mouse neutrophils to induce ADCC effector functions. The therapeutic CTCL model was further extended to evaluate the role of human effector cells in tumor cell killing using bioluminescence imaging (BLI) of luciferase expressing CCR4 + Mac-1 cells established by retroviral transduction. Ten SCID-BEIGE mice that were grafted with 1610 6 CCR4 + Mac-1 cells and developed equivalent sized tumors as detected on day 7 by BLI were divided into two groups. Eleven days after initial tumor cell inoculation, the AAV8-minibody vectors were administered intravenously. Next, human PBMCs (hPBMCs) were given by intraperitoneal injection 7 days after AAV vector administration. As shown in Figure 3a , treatment with AAV8-h1567 and hPBMCs resulted in substantial tumor growth inhibition compared to AAV8-11A plus hPBMC treated mice. Quantitative monitoring of tumor growth by in vivo BLI correlated with visible tumor growth, further confirming the tumor growth inhibition effect of AAV8-h1567 compared with control group (Figure 3b) . A significant difference was observed between the control AAV8-11A and therapeutic AAV8-h1567 groups on days 40, 42, and 45 after tumor inoculation by caliper measurement and by days 25 and 38 by BLI (Figure 3a and b) . Real-time whole-body BLI of a representative mouse showed that tumor growth was considerably inhibited in mice treated with AAV8-h1567 compared with control mice over the treatment period ( Figure 3c) . Analysis of micro-computed tomography/positron emission tomography (mCT/PET) images also revealed tumor growth inhibition with AAV8-h1567 treatment compared with the control group. While both AAV8-h1567 and AAV8-11A showed primary tumor growth 28 days after tumor inoculation, the tumor cells became much more locally invasive in the AAV8-11A treated group and showed increased metabolic activity as indicated by the accumulation of the PET tracer 18 F-fluorodeoxyglucose (FDG) in whole-body images of mice (Figure 3d) . To further assess the in vivo mechanisms of tumor cell killing in the AAV8-h1567 plus human PBMC treated group, the role of human NK cells, which also express FccRIIIA receptors, was evaluated. In the AAV8-h1567 treatment group, a substantial increase in tumor-infiltrating human NK cells was observed, as shown by the intense CD56 immunostaining compared with control 11A treated mice (Figure 4a) . Quantitative color deconvolution analysis showed a significantly increased staining in the mouse group treated with AAV8-h1567 compared with the control group treated with AAV8-11A (P,0.01; Figure 4b ). Human NK cell-mediated ADCC activity was also evaluated in vitro using purified human NK cells as effector cells. As shown in Figure 4c , human NK cells were able to kill Mac-1 target cells in the presence of h1567 in a dose dependent fashion. Control 11A minibody showed only very low levels of killing. As both mouse neutrophils and human NK cells express FccRIIIA receptors (CD16A) on their surface that can bind h1567, these in vitro and in vivo data strongly support that h1567 mediated killing occurs, at least in part, through FccRIIIA engagement and activation of immune cell effector functions. In this study, an AAV8-based therapeutic antibody gene transfer model was developed to evaluate a novel humanized anti-CCR4 monoclonal antibody h1567 as a therapeutic drug candidate against CTCL. The SCID-BEIGE mice that were used to establish this CTCL model lack T and B cells and functional NK cells [26] . High level, durable expression of the h1567 minibody was achieved after a single intravenous injection and significant anti-tumor activity against CCR4 + Mac-1 cells was seen in two animal treatment studies. These results provide the first in vivo evidence that mAb h1567 may be clinically active against CTCL cells and suggest that further studies should be undertaken to investigate its clinical efficacy. Remarkable among the findings of this study is that a single intravenous tail vein treatment with AAV8-h1567 minibody resulted in a dose dependent increase in serum minibody levels that steadily increased over a two week period and remained at near peak levels through the end of this 15 week study ( Figure S1 , Figure 1b ). The integrity of the minibodies was demonstrated biochemically in vitro and in vivo by several parameters including their CCR4 + binding activity, stable dimer formation and ease of recovery by protein A chromatography ( Figure 1 c and 1d) . This scFvFc minibody format may be ideal for experimental AAV8 delivery since conventional mAb expression is derived from heavy and light chain genes, and it can be difficult to achieve the balanced expression of two genes within a single AAV vector that has a small packaging capacity (less than 5 kb), although a 2A selfprocessing peptide and furin cleavage have been successfully used to drive the expression of full-length rat IgG [27, 28] . For cancer immunotherapy, scFvFc minibodies appear to be promising because they have been shown to be functionally comparable with full-length IgG and have been successfully used to treat various tumors in preclinical studies [29, 30, 31] . As an added benefit, along with their bivalent antigen-binding avidity and intact antibody effector functions, they comprise a single polypeptide chain that does not require balanced heavy and light antibody chain heterocomplex associations and have a smaller molecular weight for better tissue penetration compared with whole IgG molecules [32] . The functional integrity of the h1567 minibody was also shown by its potent in vivo anti-tumor and in vitro killing activities. In the first treatment study (model 1), marked inhibition of Mac-1 tumor cell growth and increased survival was seen (Figure 2a and b ) even though these SCID-BEIGE mice have profound immune cell defects including lack of T and B cells as well as impaired macrophage and NK cell effector functions [26] . Further IHC staining of the paraffin-embedded tumor tissues revealed a predominant infiltration of Ly-6G + CD16A + neutrophils only in the h1567 minibody but not control 11A minibody treated mice (Figure 2c) . Furthermore, in vitro ADCC assay using purified mouse neutrophils as an effector cells demonstrated that h1567 induced significant lysis of CCR4 + Mac-1 cells, while no lysis was seen with 11A ( Figure 2e) . These results support the view that neutrophil-mediated ADCC is involved in anti-tumor activities following AAV8-h1567 gene delivery in the SCID-BEIGE CTCL mouse model. The therapeutic SCID-BEIGE CTCL model was further extended to evaluate the role of human effector cells in tumor cell killing. In the second treatment study (model 2), AAV8-h1567 gene delivery together with human PBMCs was evaluated and a significant inhibition of CCR4 + Mac-1 tumor cell growth was again seen (Figure 3) . The therapeutic effect of the h1567 was monitored using tumor size measurements and BLI. In comparison with the measurement of tumor volume, BLI analysis enabled earlier tumor detection and revealed extensive cell death in response to h1567 treatment (Figure 3c ). The addition of microPET and CT images provided three-dimensional analysis of the primary tumor and further evaluation of the effectiveness of the AAV8-h1567 treatment in vivo. PET imaging indicated invasive tumor cell infiltration into surrounding tissues which was not seen in the h1567 treated mice (Figure 3d) . The FccRIIIA receptor (CD16A) is the dominant FccR involved in human NK cell-mediated ADCC. Treatment of CCR4 + Mac-1 tumor bearing mice with AAV8-h1567 and human PBMCs resulted in a marked increase in the number of tumorinfiltrating human CD56 + NK cells, suggesting that CD16A which is expressed on human NK cells is involved in this tumor cell killing through it's interaction with the Fc portion of h1567, a finding that has been experimentally confirmed through Fc mutagenesis studies (data not shown). Moreover, in vitro ADCC studies with purified human NK cells demonstrated a concentration dependent killing by h1567 (Figure 4c) . Thus the unifying observations from both treatment studies strongly suggest that the in vivo anti-tumor activity of h1567 is mediated, at least in large part, by ADCC through engagement of FccRIIIA on mouse neutrophils (model 1) and human NK cells (model 2). Since Fc gamma receptors (FccRs) of different types are present on a variety of effector cell populations, including NK cells, dendritic cells, macrophages, monocytes and neutrophils [33, 34] , it is possible that FccR engagement on other immune effector cells, not investigated in this study could also be involved. MAb therapy for advanced CTCL has been proposed [3] and numerous trials with alemtuzumab (anti-CD52) have shown modest to moderate clinical effects [35, 36, 37] . A recent trial with low dose alemtuzumab has shown complete remission in 50% of patients with refractory leukemic forms of the disease and without infectious disease complications although it was found completely ineffective in the treatment of MF [6] . A mAb to CD4 (GenMab) has been designated an orphan drug for the treatment of MF by the FDA [38] . While both approaches are designed to eliminate CTCL cells, there can be significant adverse effects from either treatment. CD52 is expressed by virtually all T and B cells, and the elimination of all CD4-positive cells has well-known negative consequences [39] . Clonal malignant T cells in these CTCL patients express uniformly high levels of CCR4, but variable to low levels of other skin homing addressins, including CLA, CCR10 and CCR6. CCR7, which is also uniformly highly expressed on leukemic variants of CTCL with T CM phenotype, is not expressed on the phenotypic T EM cells that are found in MF skin lesions [7] . Thus, only CCR4 is uniformly expressed on all forms of CTCL and has a restricted expression pattern on normal T cells, including Tregs [40] . Indeed, a subset of malignant T cells in some CTCL have been shown to act as CCR4 + Tregs to suppress antitumor responses and may fuel disease progression [41] . A therapeutic mAb that could preferentially target all forms of the disease and reverse Treg mediated immune suppression would be a major advance in the effective therapy of CTCL. The activity of mAb1567 in abrogating Treg mediated suppression of T effector cell function is described elsewhere (DK. Chang et al., in press). MAb KM0761, is another humanized anti-CCR4 mAb that has shown promising results in CTCL animal studies [42] and in clinical trials for refractory Adult T-cell leukemia (ATLL) and peripheral T cell lymphoma (PLCL) where good clinical activity without severe adverse side effects was seen [43, 44] . Our data support further exploration of the clinical potential of therapeutic mAbs that target CCR4 in CTCL. In summary, the results of the present study have validated the utility of an AAV8-based therapeutic minibody gene transfer platform for the rapid experimental evaluation of mAbs for the treatment of human cancer. Furthermore, this study showed that the AAV8-h1567 minibody inhibited the primary CCR4 + tumor burden, suppressed local metastasis and prolonged the survival time in tumor-bearing SCID-BEIGE mice. We remain hopeful that additional studies will support this humanized mAb1567 moving from bench to bedside. The human skin-tropic Anaplastic large-cell lymphoma (ALCL) cell line Mac-1, which was originally isolated in the laboratory of Marshall E. Kadin at Harvard Medical School [45] , was cultured in RPMI medium supplemented with 10% fetal bovine serum (FBS), 0.06 mM 2-mercaptoethanol, and 500 mg/ml G418. Immunophenotyping of the Mac-1 cell line showed the expression of all known tumor-specific chemokine receptors, including high levels of CCR4, CCR7, and CXCR4. This MAC-1 cell line was stably transduced with a luciferase encoding retrovirus. HEK 293 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% FBS and 1% penicillin/streptomycin (Invitrogen). All cells and cultures were maintained at 37uC in a 5% CO 2 humidified incubator. Human PBMCs obtained from the Dana-Farber Blood Center were purified by a Ficoll-Hypaque density gradient centrifugation as described in the general protocol of Miltenyi Biotec Inc. (Auburn, CA). Mouse neutrophils were isolated from SCID-BEIGE mouse blood by Percoll density gradient centrifugation, as described [46] . Human NK cells were isolated from human PBMC using the NK cell isolation kit, according to the manufacturer's protocol (Miltenyi Biotec, CA). To construct the scFvFc h1567 minibody expression cassette, the scFvFc h1567 gene was PCR-amplified from a plasmid coding for the humanized anti-human CCR4 antibody that is derived from heavy and light antibody chains of mAb 1567 (R&D Systems, Inc) previously cloned in our laboratory (DK. Chang et al., in press) and inserted into the AAV-cloning vector pTRUF (obtained from the University of Iowa Viral Vector Core) at the restriction sites of Sfi1 and Not1. Consequently, to efficiently direct the expression and secretion of the single chain mAb, the pTRUF vector was modified by inserting the human IgG VH4 leader sequence and the Fc sequence (hinge, CH2 and CH3 domains) of the human IgG1 flanked by 145-bp and AAV2-inverted terminal repeats (ITRs) (Figure 1a ). Recombinant AAV8 viral vectors were produced using a helper virus-free system with some modifications [47] . Low-passage human HEK 293 cells were cotransfected by linear polyethylenimine (Polysciences) with three plasmids: the AAV cis-plasmid pTRUF encoding the human mAb gene expression cassette flanked with ITRs; the AAV-packaging plasmid p5e18 (2/8) containing AAV2 rep and AV8 cap genes; and the Ad helper plasmid pXX6-80 containing the VA RNA, E2, and E4 genes required for AAV propagation (obtained from Dr. Jim Wilson, University of Pennsylvania) [48] . At 48 h post-transfection, the cells were harvested, and the AAV virus extracted by freezing and thawing the cells. Subsequently, AAV was purified by two sequential iodixanol density gradients, concentrated, then desalted by centrifugation through Biomax 100-K filters (Millipore) according to the manufacturer's instructions. Viral titers were determined as genome copy titers (vg), by quantitative real-time PCR using primers and probe speicific for AAV vector pTRUF [49] . Forward primer (59-TCTGAGTAGGTGTCATTC-TATTCTGGG-39) is located at the end of the 39-poly(A), and reverse primer (59-CACTAGGGGTTCCTAGATCTCTCCC-39) is at the beginning of the 39 inverted terminal repeat (ITR). The probe (59-TCTTCCCAATCCTCCCCCTTGCTGTC-3; FAM/TAMRA) is located in between. Larger quantity of the AAV serotype 8 vectors encoding scFvFc 11A, control minibody specific for SARS [21] , and scFvFc h1567 were produced at Harvard Gene Therapy Initiative (Harvard Institute of Medicine, Boston, MA) and used in the animal studies. Mice were monitored for tumor development and progression by both caliber measurement and Xenogen BLI. The latter was initiated for the monitoring of tumor growth 7 days after tumor implantation, which was repeated once a week. Mice were anesthetized with 3.5% isoflurane in an induction chamber, which was followed by the intraperitoneal administration of 50 mg/ml D-luciferin. For imaging, mice were maintained under 1.5% isoflurane anesthesia that was delivered through a nose cone. Whole body images were repeatedly acquired until the maximum peak of photon number was confirmed during various exposure times (10 s-1 min). Data were quantified using the time point that gave the highest photon number during the scanning time and analyzed using the Living Imaging software (Caliper Life Sciences, Hopkinton, MA). PET/CT scans were performed at the Harvard Medical School Imaging Core Facility. Mice were fasted for 12 h before the 18 F-FDG injections, but provided water ad libitum. For 18 F-FDG injection and imaging, mice were anesthetized using 2% isoflurane. The animals were then intraperitoneally injected with 7.4 MBq (200 mCi) of 18 F-FDG, allowed to regain consciousness, and then kept at 37uC until imaging. Imaging was started 30 min after the intraperitoneal injection. Mice were imaged in a chamber that minimized positioning errors between PET and CT to less than 1 mm. Image acquisition time was 10 min. Images were analyzed using AMIDE software [50] . All regions of interest were defined on fused PET/CT images to ensure reproducible positioning. HEK 293T cells (ATCC, Manassas, VA) were transfected with the AAV-coding plasmid containing the minibody-expressing constructs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). Three days after transfection, the minibodies were purified from the supernatants with protein A sepharose affinity chromatography. The in vivo production of AAV8-minibodies was generated by i.v. injections into SCID-BEIGE mice as described above. Levels of minibodies in the serum were measured in duplicate using a human IgG ELISA quantitation kit according to the manufacturer's protocol (Bethyl Laboratories, Inc., Montgomery, TX). Western immunoblotting was performed on protein A column purified samples containing in vitro synthesized minibodies and in vivo AAV8-derived minibodies. The proteins were separated by SDS-PAGE under reducing or nonreducing conditions and electrophoretically transferred onto a nitrocellulose membrane using the iBLot dry blotting system (Invitrogen). After blocking with 5% skim milk overnight, the blot was probed with an APconjugated human IgG-Fc antibody that was diluted 1:30,000 in blocking buffer for 1 h at room temperature. Excess conjugate was removed by five washes with Phosphate buffered saline containing 0.1% Tween 20 (PBS-T). The detection of protein was performed by incubating the membrane with BCIP/NBT alkaline phosphatase substrate (KPL). The biological activity of the in vivo AAV8-derived h1567 minbodies was analyzed by fluorescence-activated cell sorting (FACS) for binding activity. Mac-1 cells or 293T-CCR4 cells were washed with PBS supplemented with 0.5% bovine serum albumin (PBS-B) and then incubated with in vivo produced h1567 for 1 h at room temperature, which was followed by incubation with antihuman IgG-Fc conjugated to fluorescein isothiocyanate (FITC). Flow cytometric analysis was performed using BD FacsCalibur (BD Biosciences, San Jose, CA) and FlowJo data analysis software (Tree Star, Inc., Ashland, OR). Immunohistochemical staining was performed at DFCI/Harvard Cancer Center Research Pathology Core. For qualitative and quantitative immunohistochemical analysis, formalin-fixed and paraffin-embedded tissue sections were stained with antibodies directed against Ly-6G on the surface of mouse neutrophils and human CD56 antigen on human NK cells. The stained slides were then scanned using the Aperio ImageScope (Aperio Technologies, Inc., Vista, CA), and full tumor sections were selected for quantitative analyses. The percentage of positively stained cells in the entire tumor sections was calculated using a color deconvolution algorithm. In vitro Antibody-dependent Cell Cytotoxicity Assay ADCC was performed using the lactate dehydrogenase (LDH) release assay method, according to the CytoTox96 non-radioactive cytotoxicity assay procedure specified by the manufacturer (Promega, Madison, WI). Mouse neutrophils purified from SCID-BEIGE mouse or purified human NK cells from PBMC was used as effector cells and CCR4+ Mac1 tumor cells were used as target cells. Briefly, purified SCID-BEIGE mouse neutrophils or NK cells were plated at a density of 1610 4 cells per well in a roundbottom 96-well plate in the presence of h1567 or 11A minibodies. After 1 h of incubation, freshly prepared effector cells were added at an effector-target cell ratio (E:T) of 80:1 (mouse neutrophils) or 2:1 (human NK cells). After 2 h incubation at 37uC, supernatants of each well were recovered by centrifugation at 3006g for 5 min. LDH activity in the supernatant was determined by measuring absorbance at a wavelength of 490 nm. The cytotoxicity (%) was calculated according to the following formula: where E is the LDH release by effector-target coculture, SE the spontaneous release of the LDH from the effector cells, ST the spontaneous release of the LDH from the target cells and M the maximum release of the LDH from the target cells incubated with lysis solution (10% Triton-X). All measurements were done in triplicate. Statistical analyses were performed using 2-way ANOVA with Bonferroni post hoc tests and unpaired 2-tailed t-tests using GraphPad Prism 5 (GraphPad Software, Inc., La Jolla, CA). P values less than 0.05 were considered statistically significant. Figure S1 Dose dependent expression of h1567 anti-CCR4 minibody. Nude mice (4 mice per group) were treated one time by tail vein injection with AAV8-h1567 viral vectors at the two concentrations shows. PBS buffer treated mice served as controls. Mice were bled at the indicated time points over 15 weeks and their h1567 scFv-Fc levels were determined by ELISA on anti-human Ig capture and detection. (TIF)

The VNTR Polymorphism of the DC-SIGNR Gene and Susceptibility to HIV-1 Infection: A Meta-Analysis

BACKGROUND: Dendritic cell-specific intercellular adhesion molecule-3-grabbing nonintegrin related (DC-SIGNR) can bind to the human immunodeficiency virus-1 (HIV-1) gp120 envelope glycoprotein and is thus important for the host-pathogen interaction in HIV-1 infection. Studies of the association between the variable number tandem repeat (VNTR) polymorphism of the DC-SIGNR gene and HIV-1 susceptibility have produced controversial results. METHODS AND FINDINGS: We conducted a meta-analysis of the data contained in the literature to clarify these findings. In total, 10 studies consisting of 2683 HIV-1 patients and 3263 controls (2130 healthy controls and 1133 HIV-1 exposed but seronegative (HESN) controls) were included. Odds ratios (ORs) with 95% confidence intervals (95% CIs) were assessed in the main analyses. Further stratified analyses by ethnicity and sample size were performed. By dividing the controls into two groups, healthy controls and HIV-1 exposed but seronegative (HESN) controls, we explored different genetic models to detect any association between the VNTR polymorphism and predisposition to HIV-1 infection. The results showed that the 5-repeat allele carriers (OR = 0.84, 95% CI = 0.73–0.96) and the 5/5 homozygous (OR = 0.68, 95% CI = 0.50–0.93) had significantly reduced risk when using the HIV-1 exposed but seronegative (HESN) as controls. The stratified analyses by ethnicity and sample size confirmed these findings. However, a low to moderate degree of heterogeneity was also found across studies. CONCLUSIONS: Our findings demonstrate that the VNTR polymorphism of the DC-SIGNR gene is associated with a moderate effect on host susceptibility to HIV-1 infection. Similar to the 32-bp deletion in the chemokine receptor-5 gene (CCR5Δ32), the DC-SIGNR VNTR 5-repeat allele might have a role in resistance to HIV infection, particularly in Asian populations.

The incidence of acquired immunodeficiency syndrome (AIDS) has increased over the past few decades. Up to now, nearly 34 million people suffered from human immunodeficiency virus-1 (HIV-1) infection, and an estimated 2.7 million people were newly infected with the virus in 2010 (http://www.who.int/features/ factfiles/hiv/facts/en/index3.html). However, the natural course of HIV-1 infection and the susceptibility to infection after exposure are highly heterogeneous among individuals [1, 2] . Despite of high-risk behavior and/or multiple exposures to HIV-1, some individuals remained seronegative, or uninfected. These individuals may have a different course of progression to AIDS and may have different clinical outcomes. It is a common observation that some infected individuals became symptomatic within 2-3 years while others remained asymptomatic for more than 10-15 years [3] . Last year, NIH published a conference report about definition of HIV-exposed seronegative (HESN) individuals. It is now a consensus to define HESN from several group of individuals who are at high risk of exposure which include : (1) the commercial sex workers, (2) people with hemophilia, (3) discordant couples, (4) intravenous drug users, and (5) mother-to-child transmission [4] . In the past few years, many studies revealed that host immunogenetics, including genetic polymorphisms, play important roles in host resistance to HIV-1 infection and predict different progressions to AIDS after infection [5] [6] [7] . Polymorphisms of certain chemokines and chemokine receptors have been reported to play important roles in individual variability in response to HIV-1/AIDS. The most investigated variation is the 32-bp deletion in the chemokine receptor-5 (CCR5 D32) gene, which was shown to confer resistance to HIV-1 infection in homozygous carriers, and its role has been investigated in a clinical context [8, 9] . Similar to CCR5, DC-SIGNR is potentially an important gene affecting host susceptibility to HIV-1 infection and disease progression. DC-SIGN (dendritic cell-specific intercellular adhesion molecule-3-grabbing nonintegrin) is able to bind the HIV-1 gp120 surface glycoprotein with high affinity and is able to enhance trans-infection of T cells by HIV-1 [10] . DC-SIGNR is a homolog of DC-SIGN with 77% amino acid identity and preferential expression in liver and lymph node epithelial cells. DC-SIGNR functions similarly to DC-SIGN in capturing HIV-1 and enhancing HIV-1 infection of T cells [11] . Both DC-SIGN and DC-SIGNR are organized into 3 domains: (1) an N-terminal cytoplasmic region followed by a transmembrane domain, (2) a neck-region containing a variable number tandem repeat (VNTR) of a conserved 23 amino acid sequence, and (3) a C-terminal extracellular domain with a C-type carbohydrate recognition domain (CRD) involved in pathogen binding [12] . While the CRD forms complicated carbohydrates with high mannose-containing ligands, the neck-region is essential for lectin tetramerization and influences the capability of CRD to interact with pathogens such as HIV-1. Because the VNTR of the neck-region in DC-SIGNR is highly polymorphic, ranging from 3 to 9 repeats in worldwide populations, the polymorphism has been widely studied with regard to host genetic predisposition to HIV-1 infection [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] . However, the results are controversial and inconclusive, most likely because of the different ethnic populations used in the different studies or lacking the statistical power in any individual study to produce a reliable conclusion. Thus, a comprehensive analysis is critical. In this study, a meta-analysis was performed to investigate the association between the VNTR polymorphism of the DC-SIGNR gene and host susceptibility to HIV-1 infection. We also performed a stratified analysis by ethnicity and sample size to explore the variation in the relationship between VNTR polymorphism and HIV-1 infection risk among different ethnic populations. We performed a literature search in PubMed, MEDLINE, Embase, Wanfang Database and Weipu Database for studies published through April 2011 to use in this meta-analysis. The key terms were ''DC-SIGNR or CLEC4M or CD209L or L-SIGN or CD299 or DC-SIGN2'', ''HIV-1'', and ''polymorphism'' in various combinations. The search was limited to articles published in English and Chinese. We also manually searched the references matching the above criteria to identify additional studies. The eligible investigations met the following criteria: (1) the studies were case-control designed to explore the association between VNTR polymorphism in the DC-SIGNR gene and HIV-1 susceptibility; (2) the studies provided data on the distribution of VNTR polymorphism in the case-control population; (3) the studies were published in the English or Chinese languages. Studies were excluded if they did not contain enough data for meta-analysis, if they were abstracts or reviews, or if they were duplicated within other included studies. Two authors perused the papers and extracted the data according to the above inclusion and exclusion criteria. The following characteristics were collected from the eligible studies: first author, year of publication, country of studied population, ethnicity, sample size of the cohorts, number of HIV-1 patients, number of normal healthy controls, number of HIV-1 exposed but seronegative (HESN) controls, and the distribution of the DC-SIGNR VNTR genotype among case and control groups. ORs with 95% CI were calculated to assess the association between VNTR polymorphism in the DC-SIGNR gene and HIV-1 susceptibility. In the primary analysis, an allelic association test was performed for the prevalent alleles, which were (1) the 7repeat allele, (2) the 5-repeat allele, (3) the 6-repeat allele and (4) the 9-repeat allele. These alleles represented the four most prevalent alleles in human population. To correct for multiple testing for these 4 alleles, a Bonferroni correction was also applied to determine the corrected p-value (p-value corrected). After that it was found that 5-repeat allele had a protective effect. An exhaustive round of post-hoc analysis was carried out to identify if any particular genetic model was in operation. In this post-hoc analysis, the pooled ORs were performed between HIV-1 patients and control groups in the following genetic models: (1) the homozygote proportion [24, 25] , (2) the 7/7 genotype vs. the other genotypes, (3) the 5/5 genotype vs. the other genotypes, (4) the 5/7 genotype vs. the other genotypes, (5) the 6/7 genotype vs. the other genotypes, (6) the 7/9 genotype vs. the other genotypes. As it is an explorative attempt to delineate the mode of operation account for the protective effect in the genetic test, no Bonferroni correction was carried out in this stage. In all the genetic models, the types of controls were used: (1) healthy normal or population controls and (2) HIV-1 exposed but seronegative (HESN) controls (see the Table S1 for detailed breakdown). Because the remaining rare genotypes and alleles of the VNTR locus have negligible frequencies in the cohorts, their comparisons were not conducted in this analysis. Stratified analyses were performed by ethnicity and sample size. Heterogeneity across the publications was assessed with the Cochran's ,chi. 2 test (Q-test) [26] , and p,0.10 was considered statistically significant. The I 2 test was also conducted to evaluate heterogeneity between studies. A high heterogeneity is considered present when I 2 .50% and much higher when I 2 .75% [27] . The pooled OR was calculated by a fixed effect model (using the Mantel-Haenszel method) or a random effect model (using the DerSimonian-Laird method) according to the heterogeneity among studies [28] . The statistical significance of OR was analyzed by the Z test, and p,0.05 was considered statistically significant. Publication bias was evaluated by Egger's and Begg's tests and was considered significant if p,0.05 [29, 30] . Sensitivity analysis was performed by sequentially excluding individual studies to assess the stability of the results. Statistical analyses were performed using the Revman 5.1 software and STATA 10.0 software. The p value was two-tailed and was considered to be statistically significant if its value was ,0.05. As shown in Figure 1 , a total of 115 results were identified after an initial search from the selected electronic databases. After screening the titles and abstracts, 40 articles were selected for further review. Among them, five reviews were excluded; 22 studies were excluded for not referring to the association between the DC-SIGNR VNTR polymorphism and HIV-1 susceptibility. Finally, one study was excluded because it lacked extractable data; and two studies were excluded because the data or sample sets overlapped with another one. Thus, a total of 10 studies were suitable for meta-analysis [13] [14] [15] [16] [17] [18] [20] [21] [22] [23] . Among them, seven studies were conducted with Asians [13, [16] [17] [18] 20, 21, 23] , two were conducted with European-Americans [14, 15] and one was conducted with Africans [22] . Two studies only included subjects of HIV-1 patients and HIV-1 exposed but seronegative (HESN) controls [21, 22] . Three studies included subjects of HIV-1 patients and healthy controls [15, 18, 20] , while the remaining five studies included subjects of HIV-1 patients, HIV-1 exposed but seronegative (HESN) controls, and healthy controls [13, 14, 16, 17, 23] . The total number of samples involved in the 10 eligible studies was 5946, which included 2683 HIV-1 patients and 3263 controls (2130 healthy controls and 1133 HIV-1 exposed but seronegative (HESN) controls). The characteristics of each study are listed in Table 1 . After pooling the data from the 10 studies for meta-analysis, the results were calculated according to different genetic models. For each genetic model, two comparisons were carried out separately using either the healthy samples or the HIV-1 exposed but seronegative samples (HESN) as controls. In the primary analysis for allelic risk model of the 5-repeat allele vs. the other alleles, using the healthy individuals as controls, the value of X 2 was 9.52 with 7 degrees of freedom, and p was 0.22 in a fixed effect model (Figure 2A ). The I 2 was 27%, suggesting a low to moderate heterogeneity. The fixed effect model was applied to synthesize the data. Overall, the OR was 1.03 (95% CI = 0.93-1.15) and was not significant (Figure 2A ). On the other hand, using the HIV-1 exposed but seronegative (HESN) samples as controls, the value of X 2 was 8.84 with 6 degrees of freedom, and p was 0.18 ( Figure 2B ). The I 2 was 32%, suggesting a moderate heterogeneity. Then the fixed effect model was applied to synthesize the data and the pooled OR was calculated. The pooled OR value was significant at 0.84 (95% CI = 0.73-0.96, p = 0.01, Figure 2B ). The Bonferroni corrected P-value was 0.05. These results suggested that the 5-repeat allele carriers tended to be associated with resistance to HIV-1 infection in HIV-1 exposed seronegative (HESN) individuals. In the genetic models, we analyzed the heterogeneity of 5/5 vs. the other genotypes by using the healthy subjects as controls. The results indicated a low heterogeneity (X 2 = 9.30, I 2 = 25%, p = 0.23) ( Figure 3A ). Next, we pooled the 8 studies under the fixed effect model. The pooled OR was 0.90 (95% CI = 0.71-1.44, p = 0.38) ( Figure 3A) . Then, we analyzed the heterogeneity of 5/5 vs. the other genotypes by using the HIV-1 exposed but seronegative samples subjects (HESN) as controls ( Figure 3B) . The results indicated a low to moderate heterogeneity (X 2 = 7.09, I 2 = 29%, p = 0.21). The fixed effect model was then used to synthesize the data. Overall, the pooled OR was significant at 0.68 (95% CI = 0.50-0.93, p = 0.01) ( Figure 3B ). These results suggested that the 5/5 homozygous genotype showed a significantly reduced risk of HIV-1 infection in HIV-1 exposed seronegative individuals (HESN) but not in healthy individuals. A summary of the results of all the comparisons with different genetic models was listed in Table S1 . We also performed stratified analyses by ethnicity and sample size to explore potential sources of heterogeneity and examine the relationship between the DC-SIGNR VNTR polymorphism and susceptibility to HIV-1 infection. The results are also summarized in Table S2 . Most of the results of the stratified analyses were consistent with the main analysis. For ethnicity, the studies were divided into two subgroups, one subgroup of Asian descendants (7 studies) and the other subgroup of European-American descendants (2 studies). The resistance to HIV-1 infection found among 5/5 homozygous subjects was predominantly due to the HIV-1 exposed seronegative (HESN) subjects of the Asian population (OR = 0.58, 95% CI = 0.39-0.85, p = 0.006, Pheterogeneity = 0.34 and I 2 = 12%) ( Figure 4A ). The 5-repeat allele presented the trend of having the protective effect among Asian (OR = 0.84, 95% CI = 0.71-1.00, p = 0.05, Pheterogeneity = 0.20 and I 2 = 33%), though the p value was marginal ( Figure 4B ). This result also showed that the sample size of the Asian population in present study was not large enough to achieve enough statistical power to obtain significant observations. As there was only one study performed with African and European-Americans involving HIV-1 patients and HIV-1 exposed seronegative (HESN) controls, the stratified analyses in these populations were not performed. For sample size, the studies were stratified into two subgroups, one comprising studies with more than 200 subjects and one with fewer than 200 subjects. In these subgroups, significantly reduced HIV-1 infection of 5/5 homozygous (OR = 0.69, 95% CI = 0.50-0.94, and p = 0.02) and 5-repeat allele carriers (OR = 0.82, 95% CI = 0.68-0.98, and p = 0.03) was found in the subgroup with a sample size of more than 200 subjects, but not in the subgroup with fewer than 200 subjects (Table S2 ). Publication bias was assessed by Begg's test and Egger's test (data not shown) in the total population and all the subgroups. For the genotypic risk model, 5/5 vs. other genotypes, in which the HIV-1 exposed but seronegative samples (HESN) subjects were used as controls, both the Begg's test (p = 0.462) and the Egger's test (p = 0.215) did not present any significantly statistical evidence of publication bias. For the allelic risk model of 5-repeat allele vs. other alleles, in which the HIV-1 exposed but seronegative Sensitivity analysis was performed by deleting one study at one time to assess the stability of the pooled ORs. None of the corresponding pooled ORs was statistically changed, implying the stability of the results. By using the healthy samples as controls, the comparison was carried out between the HIV-1 patients and healthy samples; (B) By using the HIV-1 exposed but seronegative (HESN) samples as controls, the comparison was carried out between the HIV-1 patients and the HIV-1 exposed but seronegative (HESN) samples. Note: A Bonferroni correction for multiple testing of 4 was applied to get a corrected P value in this primary analysis. The association between 5-repeat allele had a corrected P value of 0.05. doi:10.1371/journal.pone.0042972.g002 A potential role of host genetic factors in the predisposition to HIV-1 infection has been suggested by different reports [8, 31, 32] . DC-SIGNR serves as an HIV-1 ligand to facilitate HIV-1 virion infection into adjacent CD4+T cells in trans and has been the subject of many recent studies. The VNTR polymorphism in its neck-region was found to be associated with host susceptibility to HIV-1 infection, but the conclusions were controversial. In the present study, we performed a meta-analysis of 10 eligible studies with 2683 HIV-1 patients and 3263 controls to elucidate the relationship between the VNTR polymorphism and HIV-1 infection risk. The strength of the present analysis is based on the accumulation of published data that provides enough information to generate a more precise conclusion. By using different genetic models, we could preliminarily estimate the effect of the allele frequency, genotype frequency, and homozygous proportion. The results indicated that the 5/5 homozygous genotype was associated with resistance to HIV-1 infection. The protective effect was most predominant in HIV-1 exposed seronegative (HESN) individuals. The stratified analyses by ethnicity and sample size confirmed these findings. Thus, it is important to emphasize that in studies concerning HIV-1 susceptibility, taking samples from HIV-1 exposed seronegative (HESN) individuals as controls may be more powerful than taking random healthy individuals as controls. In future studies, collecting both HIV-1 exposed seronegative individuals (HESN) and random healthy individuals as controls would be helpful for clarifying the relationship between candidate genes and host susceptibility to HIV-1 infection. But under the definition and criteria of HIV-1 exposed seronegative individuals (HESN), it is still a heterogeneous group. A workshop sponsored by NIH was conducted in 2011 and a consistent definition and criteria were drawn in the term of HIV-exposed seronegative (HESN) individuals [4] . In the following studies, the collection of the HESN as controls should strictly obey these definitions, which would be helpful in explaining natural HIV-1 protection in individuals exposed to HIV-1 who remain seronegative or demonstrate resistance to infection. Pooled data analysis by ethnicity was only performed in Asians since only one study involving HIV-1 patients and HIV-1 exposed seronegative controls (HESN) was separately performed in European-Americans and Africans. More studies on these populations are needed in future works, which would help us better understand the relationship between the VNTR polymorphism and HIV-1 infection risk in different ethnic populations. Our findings are consistent with the study by Rathore et al. [16] but are different from other reports [14, 15, 17] . The potential explanation for the discrepancy may be the limited sample number included in the single study and the relative impact of this polymorphism in the different populations. As shown, the associations were only confirmed in stratified analyses by combining studies with a sample size greater than 200 subjects, which emphasized the importance of having sufficient power with a large enough sample size and the requirement of further largescale analysis. Although the function of the association between the DC-SIGNR VNTR polymorphism and host susceptibility to HIV-1 infection has not been fully explored, outcomes of this metaanalysis suggest that the DC-SIGNR VNTR polymorphism had an impact on host susceptibility to HIV-1 infection. A study by Xu By using the healthy samples as controls, the comparison was carried out between the HIV-1 patients and healthy samples; (B) By using the HIV-1 exposed but seronegative (HESN) samples as controls, the comparison was carried out between the HIV-1 patients and the HIV-1 exposed but seronegative (HESN) samples. doi:10.1371/journal.pone.0042972.g003 et al. [18] showed that patients carrying the 5-repeat DC-SIGNR allele had significantly lower HIV-RNA levels compared with those observed in patients with the 7-and 9-repeat DC-SIGNR alleles, though they did not observe that the different genotypes/ alleles were associated with CD4+ T cell numbers. In vitro studies also demonstrated that DC-SIGNR with equal to or less than 5repeat alleles displayed unstable homozygous or heterozygous DC-SIGNR aggregates [33] accompanied by changes in affinity for HIV-1gp120 glycoprotein [34] . Data from these in vivo/vitro functional studies gave some support to the hypothesis that the 5/5 homozygous genotype might confer more resistance to HIV-1 infection. However, these associations were far from conclusive. More functional in vivo studies are needed. Although meta-analysis is a powerful statistical method, inherent limitations of this study should be addressed. First, we only included the studies written in English and Chinese, and the related reports in other languages were not included, which might bias our conclusion in this study. Second, publication bias could not be excluded though the test showed negative results. The studies reporting significant associations between certain genotypes and reduced susceptibility to HIV infection would be more readily published while the studies with nonsignificant associations would be more difficult to publish. Third, most of the studies were conducted using Asian population groups. In the stratified analysis by ethnicity, there was only one study performed with Africans and two studies with European Americans, each of which had a sample size too small to achieve enough statistical power to obtain significant observations. In fact, many association studies showed different results for different populations. Thus, further studies are warranted in other ethnic populations to evaluate the possible ethnic differences of the VNTR polymorphism and HIV-1 susceptibility. Fourth, gene-gene and gene-environment interactions may influence host susceptibility to HIV-1 infection. In fact, many genes have been proven to influence HIV-1 infection risk, but we did not have enough data to eliminate these interfering factors. Fifth, as most studies did not mentioned about potential population stratification of patients and controls samples, we cannot rule out a role of population structure in the observed association. Finally, further stratified analyses of patients and HESN individuals by infection exposure routes (sexual contact, intravenous drug use, etc.) could not be performed because the data detailing the infection route for the HIV-1 patients were lacking. Because DC-SIGNR is expressed at relatively lower levels in tissues in vivo [35] , the association between the VNTR polymorphism and HIV-1 infection risk by different infection routes might be different. Humans show remarkable variation in vulnerability to infection by HIV-1 and especially in the clinical outcomes after infection. Understanding why some people establish and maintain effective control of HIV-1 and others do not is a priority in the effort to (A) By using the HIV-1 exposed but seronegative (HESN) samples as controls, the comparison was carried out between the HIV-1 patients and the HIV-1 exposed but seronegative (HESN) samples under the genotypic risk model of 5/5 vs. other genotypes. (B) By using the HIV-1 exposed but seronegative (HESN) samples as controls, the comparison was carried out between the HIV-1 patients and the HIV-1 exposed but seronegative (HESN) samples under the allelic risk model of 5-repeat allele vs. other alleles. doi:10.1371/journal.pone.0042972.g004 develop new treatments for HIV/AIDS. To our knowledge, this is the first meta-analysis to assess the relationship between the DC-SIGNR VNTR polymorphism and HIV-1 infection risk. Our results showed that the 5/5 homozygous genotype was associated with resistance to HIV-1 in HIV-1 exposed seronegative subjects (HESN), especially in the Asian population. Future studies in different ethnic populations and with clear infection routes should be performed to evaluate these associations. Table S1 The breakdown of HESN controls in the included studies of this meta-analysis according the routine of exposure. (DOC)

Toll-like receptors are critical for clearance of Brucella and play different roles in development of adaptive immunity following aerosol challenge in mice

Brucella spp. cause undulant fever in humans and brucellosis in variety of other animals. Both innate and adaptive immunity have been shown to be important in controlling Brucella infection. Toll-like receptors (TLRs) represent a group of pattern recognition receptors (PRRs) that play critical roles in the host innate immune response, as well as development of adaptive immunity. In the current report, we investigated the role of TLR signaling in the clearance of Brucella and development of adaptive immunity in TLR2(−/−), TLR4(−/−), or MyD88(−/−) mice following aerosol exposure to B. melitensis 16 M. Consistent with previous reports, MyD88 is required for efficient clearance of Brucella from all three organs (lung, spleen, and liver). The results reveal Th2-skewed immune responses in TLR2(−/−) mice late in infection and support a TLR2 requirement for efficient clearance of Brucella from the lungs, but not from the spleen or liver. Similarly, TLR4 is required for efficient clearance of Brucella from the lung, but exhibits a minor contribution to clearance from the spleen and no demonstrable contribution to clearance from the liver. Lymphocyte proliferation assays suggest that the TLRs are not involved in the development of cell-mediated memory response to Brucella antigen. Antibody detection reveals that TLR2 and 4 are required to generate early antigen-specific IgG, but not during the late stages of infection. TLR2 and 4 are only transiently required for IgM production and not at all for IgA production. In contrast, MyD88 is essential for antigen specific IgG production late in infection, but is not required for IgM generation over the course of infection. Surprisingly, despite the prominent role for MyD88 in clearance from all tissues, MyD88-knockout mice express significantly higher levels of serum IgA. These results confirm the important role of MyD88 in controlling infection in the spleen while providing evidence of a prominent contribution to protection in other tissues. In addition, although TLR4 and TLR2 contribute little to control of spleen infection, a significant contribution to clearance of lung infection is described.

The genus Brucella is a group of Gram-negative, facultative intracellular bacteria that cause brucellosis, a reproductive disease in ruminants, and undulating fever in humans. Brucellosis is one of the most important worldwide zoonotic diseases. Ten species have been identified to date, three of which, including B. melitensis, B. abortus, and B. suis are virulent in humans and represent a significant threat to public health (Atluri et al., 2011) . Humans often become infected following inhalation of particles carrying the bacteria or consumption of dairy products contaminated with the organism. Although vaccination is used to successfully reduce the spread of disease, the risk remains high in underdeveloped nations. There are currently no vaccines available that are safe for use in humans, and although generally effective, antibiotic treatments do not always prevent disease recrudescence. As a result of these factors and concern over their potential weaponization, NIH and the CDC/USDA have classified these three species as category B agents. Both innate and adaptive immunity have been described as contributing to the control of Brucella infection (Baldwin and Parent, 2002; Dornand et al., 2002; Baldwin and Goenka, 2006) . The role of innate immunity against infection by this pathogen has drawn recent attention as a result of awareness of the role of innate immunity in the establishment of infection and the development of adaptive immunity (Weiss et al., 2005; Oliveira et al., 2011) . In contrast, adaptive immunity, including cell-mediated and humoral responses, has been the prominent focus of Brucella research over the past few decades. The innate immune system is composed of a variety of cellular and humoral components, which are the first line of the host defense against invading pathogens. Recognition relies on pattern recognition receptors (PRRs) expressed on/in the cellular components of the innate immune system. Toll-like receptors (TLRs) are the best characterized PRRs. Receptorligand interaction via TLRs induces the production of antimicrobial peptides and proinflammatory cytokines through NF-κB, mitogen-activated protein kinase (MAPK) and other signaling pathways (Kawai and Akira, 2006) . As a result, TLR signaling is critical to development of the host innate immune response, including recruitment of dendritic cells (DCs) and T effector cells, and upregulation of MHC I and II on antigen presenting cells (APCs) and by extension adaptive immunity against infection. 10 TLRs in human and 13 in the mouse have been identified to date (Kawai and Akira, 2006) . TLR2, TLR4, TLR5, and TLR9 recognizing lipopeptide, lipopolysaccharides, flagellin and CpG DNA, respectively, are known to be important in controlling bacterial infections. With the exception of TLR3, the TLRs require the adapter molecule myeloid differentiation factor 88 (MyD88) for signal transduction (Kawai and Akira, 2007) . As expected, MyD88 have been shown to be essential for clearance of Brucella infection from mice (Weiss et al., 2005; Copin et al., 2007; Macedo et al., 2008) . Several groups have investigated the contribution of TLR signaling to innate immunity against Brucella infection in the mouse model. The consensus opinion is that TLR2 is not required to control Brucella infection in the mouse (Campos et al., 2004; Copin et al., 2007) . However, TLR2 has been shown to be important for cytokine production (Huang et al., 2003; Giambartolomei et al., 2004; Weiss et al., 2005; Macedo et al., 2008; Zwerdling et al., 2008) , MHC-II expression and down regulation of the type I receptor for the Fc portion of IgG (FcγRI, CD64) (Barrionuevo et al., 2011) in tissue culture. The role of TLR4 in Brucella infection remains controversial. Some studies suggest that TLR4 is required to control Brucella replication in the mouse (Campos et al., 2004; Copin et al., 2007; Macedo et al., 2008) ; others indicate that TLR4 is not involved (Weiss et al., 2005; Barquero-Calvo et al., 2007) . The use of different Brucella strains/species in these experiments may account for the observed differences. The influence of the route of infection on the role of TLRs in studies with other pathogens suggests a need to do so with brucellosis, as is the role of TLR-mediated innate immunity in the development of adaptive immune response. These studies employed intraperitoneal (i.p.) inoculation, which is not a typical route for Brucella infection. As a result of these reports, the consensus of scientific opinion is that MyD88 contributes significantly to the control of Brucella infection while TLR-based signaling plays a lesser role at best. The absence of any role for TLR signaling is consistent with results indicating that the Brucella protein TcpB interrupts TLR-based signaling by promoting degradation of MyD88 adaptor like protein, MAL (Chaudhary et al., 2012) , and modification of Brucella LPS reduces agonist activity (Duenas et al., 2004) . However, these studies have been restricted to an atypical route of exposure, and the potential for differential TLR expression associated with different tissues has not been considered (Juarez et al., 2010) . Since inhalation represents a major concern to public health, experiments were undertaken to determine the role of TLR signaling in the control of Brucella infection following aerosol exposure. The current study, investigates the roles of TLR2, TLR4, and MyD88 in clearance of Brucella following respiratory exposure and development of adaptive immune response against Brucella. Since the mucosal/respiratory system is the primary portal of entry for human infection documented in many laboratory incidents and relevant to biothreats, it is important to understand the pathogenesis and immune responses resulting from aerosol exposure. Since TLR signaling is an important component bridging innate and adaptive immunity (Iwasaki and Medzhitov, 2004) , failure to clear Brucella from organs by TLR signaling deficient mice could be due to defects in the development of adaptive immunity. To test this hypothesis, we determined cellular-and humoral-mediated adaptive immunity in these TLR knockout mice following Brucella infection. The results reveal that TLR signaling exhibits significant differences in control and clearance of Brucella from selected tissues and in the associated development of an adaptive immune response. The bacterial strain used in these experiments, B. melitensis 16 M was obtained from ATCC (#23456) and recovered from an aborted goat fetus. Bacterial cultures were prepared as previously described (Pei and Ficht, 2004) . Briefly, B. melitensis 16 M was cultured in TSB for 24 h, pelleted by centrifugation at 20,000 ×g, washed with and resuspended in PBS (pH 7.4) at a density of approximately 5 × 10 9 CFU/ml (Kahl-McDonagh et al., 2007) . Breeding pairs of TLR2 −/− , TLR4 −/− , and MyD88 −/− mice were obtained from Dr. S. Akira (Osaka University, Osaka, Japan) via Dr. Michael Berton (University of Texas at San Antonio, San Antonio, TX) and colonies maintained by Comparative Medicine Program (CMP) personnel at TAMU. C57BL/6 control mice were purchased from Jackson Laboratory. All mice were housed in BSL-3 suite in the CMP at Texas A&M University. Mice of both sexes between 8 and 12 weeks old were used in the experiments. Euthanasia was performed using carbon dioxide inhalation. All personnel working with animals received training in rodent handling and euthanasia via the CMP. All animal work was performed in compliance with the Public Health Service (PHS) Policy on Humane Care and Use of Laboratory Animals as described by the National Research Council's Institute for Laboratory Animal Research (ILAR) Guide for the Care and Use of Laboratory Animals. Aerosols were generated via nebulization of a Brucella suspension (∼ 5 × 10 9 CFU/ml in PBS) into a Madison Chamber (Madison, Wisconsin) according to the manufacturer's instructions (Kahl-McDonagh et al., 2007) . The aerosol chamber was located in a biosafety level 3 (BSL3) facility in the Laboratory Animal Research and Resources (LARR) building staffed and managed by the CMP at Texas A&M University. To assure personal safety, the principal investigator implemented a strict safety protocol, and all procedures were approved by the IBC and IACUC. Following aerosol exposure, one group of mice (n = 5) were immediately euthanized via carbon dioxide inhalation and the lungs collected to determine the infecting dose (Kahl-McDonagh et al., 2007) . At 1, 2, 4, 6, 8 , and 10 weeks after exposure, mice were euthanized and lung, spleen and liver were collected and homogenized in 1 ml each of sterile water containing 0.5% (v/v) Tween-20. One hundred microliters of appropriate dilution of the homogenized tissue were plated on Farrell's selective medium plates followed by incubation at 37 • C for at least 3 days. The experimental limit of detection was determined to be 10 CFU per lung, spleen or gram of live, and the results described represent the accumulated data from 4 independent aerosol exposures. At 8-week post infection (p.i.), one-half of the spleen from each euthanized mouse was used to produce single cell suspensions as previously described. The cell density was adjusted to 2 × 10 6 per ml using complete RPMI-1640 containing 10% (v/v) heatinactivated FBS, 1 mM non-essential amino acid, 100 μg/ml of penicillin and 100 U/ml streptomycin. The cell suspension was dispensed to 24-well plates with 1 ml/well and stimulated with heat killed Brucella (HKB) or ConA (2 μg/ml) for three days. Following stimulation, supernatants were collected and cytokine synthesis characterized. Cells collected from 100 μl of the suspension were lysed using 1% (v/v) Triton X-100, and the LDH released from live cells was determined using CytoTox 96™ Nonradioactive Cytotoxicity Assay kit (Promega) following the manufacturer's instructions. IFN-γ levels in the culture supernatants were determined 72 h post stimulation using sandwich ELISA kits (PeproTech Inc., Rocky Hill, NJ) according to the manufacturer's instructions (Pei et al., 2008) . Blood samples were collected just prior to euthanization from mice at 1, 2, 4, 6, 8, and 10 weeks post exposure. Sera were separated and anti-Brucella antibodies, including IgG, IgG 1 , IgG 2a , IgM, and IgA were evaluated using ELISA. ELISA plates (Nunc) were coated with B. melitensis 16 M cell lysate. Briefly, 100 μl of 1 μg/ml lysate in bicarbonate/carbonate buffer (2.93 g NaHCO 3 , 1.59 g Na 2 CO 3 , 0.203 g MgCl 2 in 1 L of distilled water, pH 9.6) was added to microtiter wells (96 wells/plate) and incubated overnight at 4 • C (Pei and Collisson, 2005) . Non-specific binding was blocked via incubation with 5% (w/v) non-fat milk in PBS for 1 h at room temperature. The antigen-coated wells were incubated 2 h at room temperature (25 • C) with mouse sera diluted in PBS (pH 7.4) containing 0.05% (v/v) Tween-20 (PBS-T) 200 times for IgG1, IgG 2a , IgM, IgA detection or 1000 times for IgG detection. The plates were then incubated with HRP-conjugated goat anti-mouse immunoglobulins (IgG 1 , IgG 2a , IgM, IgA, and IgG from Kirkegard-Perry Labs (KPL) for 1 h at room temperature. The wells were washed with PBS-T between steps to remove any unbound materials. Color development followed the addition of 100 μl of TMB substrate KPL. The reaction was terminated by the addition of 50 μl of stop buffer (1 M H 2 SO 4 ) and OD 450 was determined using an ELISA reader (Bio-Rad). Statistical significance was determined using one-way ANOVA or Student's t-test. P-values <0.05 * and <0.01 * * were considered to be significant. In these experiments, the infecting dose determined using lungs collected from control C57BL/6 mice (n = 5) immediately following challenge is 4.26 ± 0.28 log CFU/mouse, which is consistent with a previous report (Kahl-McDonagh et al., 2007) . Bacterial load recovered from the lungs of TLR2 −/− , TLR4 −/− , and MyD88 −/− mice during the first two weeks following exposure is not significantly different from the C57BL/6 control mice (Figure 1 , panels W1-2), indicating that TLR2, TLR4, and MyD88 are not required to clear lung infection early. A significant difference in bacterial burden is detectable by week 4 and extends through week 10, indicating that the contribution of TLR2, TLR4, and MyD88 to the control of Brucella infection in the lungs occurs via adaptive immunity (Figure 1 ). By week 4 p.i., there was a significant difference in the clearance of the organism from each of the knockouts. This pattern persisted through week 10, although significance is lost due to the reduction in the number of animals available at this time point. The absence of MyD88 appears to have a greater impact on the clearance of infection from the lungs over this period, as the bacterial burden in the lungs of MyD88 −/− mice is significantly higher than that in TLR2 −/− and TLR4 −/− mice by week 8 p.i. (Figure 1, panel W8 ). The kinetics of Brucella clearance from the lungs is summarized in Figure 1K , and taken together, these results indicate that TLR2, TLR4, and MyD88 contribute to Brucella clearance from the lungs following aerosol challenge. Infection and the kinetics of clearance of Brucella from the spleens of mice is notably different from that observed in the lung. Initially, the organism exhibits a delay in systemic distribution to the spleen that requires up to 4 weeks to achieve maximum burden. The delay in systemic spread of infection is evident in the spleen and less dramatically in the liver (see below), with bacterial burden in 9 of 13 spleens from TLR4 −/− and 10 of 13 from MyD88 −/− mice below the detection limit (10 cfu/spleen). This compares with a delay in only 1 of 17 TLR2 −/− mice and 5 of 20 C57BL/6 mice. The MyD88 −/− knockout appears to have a reduced ability to control the infection as evidenced by the significant difference in spleen colonization by week 4 (Figure 2 , panel W4). Once established infection proceeds similarly, although the MyD88 −/− and TLR4 −/− mice exhibit a delay in clearance relative to the TLR2 −/− and C57BL6 mice ( week, which is accompanied by a delay in clearance at later times. Again, the exaggerated decline observed between weeks 8 and 10 in TLR4 −/− mice may be an artifact of the reduced animals numbers. Similar to the results described for the spleen, systemic spread of the organism to the liver is delayed in TLR4 −/− and MyD88 −/− mice, and is associated with a corresponding delay in the clearance of infection. Here again, TLR2 −/− mice are indistinguishable from control mice in the ability to restrict systemic infection (Figures 2 and 3) . As observed in the spleen, the capacity to control infection is most dramatically affected in the MyD88 −/− mice (Figure 3 , panels W1-10). From week 2 and onward, TLR2 −/− and TLR4 −/− mice controlled infection and cleared the organisms from the livers indistinguishably from the C57BL/6 mice (Figure 3, panel W10 ). Taken together, these results reveal that, in the mouse liver; (1) TLR2 and TLR4 are not required to control infection or clearance; (2) the absence of TLR4 reduces accumulation of Brucella in the liver, but has little effect on late stage clearance; (3) MyD88 is critical in early control and late clearance of the organism. To determine the development of cell-mediated immunity, the splenocytes collected at 8 weeks p.i. from knockout and control mice are stimulated with Brucella antigen ex vivo. Using an LDH release assay to measure lymphocyte proliferation, no significant differences in proliferation was detected for TLR2 −/− , TLR4 −/− , MyD88 −/− and control C57BL/6 mice ( Figure 4A ). This result indicates that cell-mediated adaptive immunity against Brucella is unimpaired in these knockout mice. Consistent with the results of the lymphocyte proliferation assay, IFN-γ levels detected in splenocyte supernatants collected from the knockout mice are not significantly different from C57BL/6 mice ( Figure 4B ). To determine the role of TLR signaling in antibody production and isotype switching, anti-Brucella IgM, IgG, IgG 1 , IgG 2a , and IgA responses were determined following aerosol infection. The results presented reveal detectable levels of IgM by 2 weeks p.i. that peaks by 4 weeks p.i. (Figure 5 , panel A). IgM levels begin to decline in the MyD88 −/− and C57BL/6 mice at 4 weeks p.i. while, IgM levels in the TLR2 −/− and TLR4 −/− mice plateau at 4 weeks with a gradual decline by 10 weeks p.i. Any requirement for TLR2 and TLR4 in IgM production is transient (only at 4 weeks p.i.). and IgM levels were never significantly different between MyD88 −/− and C57BL/6 mice throughout the experiment. Total Brucella-specific IgG in wild-type mice is detectable starting 4 weeks post-infection (Figure 5, panel B) , and increases continuously over the next 6 weeks. In contrast, elevation of the Brucella specific IgG levels is delayed in all of the knockout mice. Comparison among the groups revealed that, between weeks 4 and 8 p.i., IgG levels in TLR2 −/− and TLR4 −/− mice are significantly lower than that observed in the C57BL/6 mice. IgG levels increase in TLR4 −/− and TLR2 −/− mice by 10 weeks p.i. to levels that are not significantly different from the control C57BL/6 mice. In contrast, IgG levels in MyD88 −/− mice exhibit a significant reduction in titer over the course of infection. Serum IgA is not detected in the TLR2 −/− and TLR4 −/− mice, nor in C57BL/6 mice out to week 10 ( Figure 5 , panel C). Interestingly, serum IgA is prominent in the sera of MyD88 −/− mice and continuously increases over the course of the experiment. Recent studies have shown that TLRs play a critical role in determining the fate of naive T cells, directing them toward either Th1 or Th2 responses (Dabbagh and Lewis, 2003; Xu et al., 2004) . To characterize the role of TLR signaling in determining a Th1 or Th2 response, IgG 1 (Th2) and IgG 2a (Th1) levels and the IgG 1 /IgG 2a ratio are calculated following Brucella aerosol exposure. IgG 1 and IgG 2a production are above background levels only 8 weeks p.i. (Figure 5 , panels D and E). By week 10, TLR2 and C57BL/6 mice produce significantly elevated levels of IgG 1 (Figure 5 , panel D) indicating TLR2 is not required for IgG 1 production. In contrast, IgG 1 levels in TLR4 and MyD88 knockout mice are significantly reduced, indicating TLR4 and MyD88 are essential for IgG 1 production (Figure 5, panel D) . Relatively low levels of IgG 2a are present at 10 weeks p.i. in all groups, but IgG 2a levels in TLR2 and MyD88 knockout mice are significantly reduced relative to control mice. IgG 1 /IgG 2a The kinetics of Brucella spp. clearance from infected mice differs depending on the genetic background of the strain. In C57BL/6 mice, infection is at or below the limit of detection in the liver as early as 6 weeks p.i. or by 8 weeks in the lung (Figures 1 and 3) . In contrast, these same organs in BALB/c mice contain persistent levels of organism 8 weeks p.i. and beyond (Kahl-McDonagh et al., 2007) . These results are consistent with previous reports describing enhanced sensitivity of BALB/c mice to Brucella infection following i.p. inoculation (Copin et al., 2007) , confirming the expected immunological outcome despite the use of different routes of inoculation. One explanation for the improved clearance from C57BL6 mice is generation of a TH1-skewed immune response as opposed to a TH2-skewed response in BALB/c mice (Watanabe et al., 2004) . TLR activation induces the maturation of APCs, enhances antigen presentation, up-regulation of co-stimulatory molecules and cytokine production. Cytokine profiles produced by APCs control CD4 + T cell differentiation into either TH1 or TH2 cells. Engagement of different TLRs is expected to affect cytokine production, with significant effect on CD4 + T cell differentiation (Dabbagh and Lewis, 2003) . Previous studies have shown that Brucella infection induces mainly a TH1 response during acute disease (Agranovich et al., 1999; Pasquali et al., 2001; Giambartolomei et al., 2002; Rafiei et al., 2006; Khatun et al., 2009) . It is clear that TLR signaling plays a critical role in the activation of the host innate immune response, including cytokine and chemokine secretion and up-regulation of co-stimulatory molecules in APCs (Kawai and Akira, 2006) . Recent studies have shown that these activations indirectly affect the development of adaptive immunity (Iwasaki and Medzhitov, 2010) . In addition, it has been demonstrated that functional TLRs are expressed on various T and B cell subsets (Bekeredjian-Ding and Jego, 2009; Booth et al., 2011; Kulkarni et al., 2011) . For example, functional expression of TLRs has been extensively investigated on γδ-T cells (Wesch et al., 2011) that have been shown to be important in controlling Brucella infection (Bertotto et al., 1993; Ottones et al., 2000) . Interaction of TLRs with their respective microbial ligands provides a third signal for B cell activation, which is essential for optimal antigen-specific antibody production and class switch recombination (Bekeredjian-Ding and Jego, 2009; Booth et al., 2011) . Brucella are adept at inhibiting the host immune response. Brucella virulence derives from expression of an LPS with reduced agonist activity that limits activation of innate immunity and development of an effective adaptive response that promotes invasion and establishment of a replicative niche. In addition, Brucella express a protein TcpB (or Btp1) that restricts the proinflammatory response by interfering with MyD88 function. The result is enhanced degradation of MAL (MyD88-adapter-like protein) and restricted expression of NFκB. However, our current results indicate infection varies in different tissues, and that this variation is attributable in part to various PRRs, including TLR2, TLR4, and downstream signaling partners, like MyD88. Interference with TLR signaling by the organism has been described in numerous studies, as has the prominent contribution of MyD88 in limiting infection. The results reported here confirm the importance of MyD88 in the control of infection and extend those findings to include three tissues; spleen, liver and lung. In addition, the experiments performed confirm previous findings concerning the lack of significant contribution of TLR4 and TLR2 with regard to spleen (and liver) infection, but reveal their significant contribution preventing lung infection. In an effort to explain the control exerted by each of these host factors, an evaluation was made of the corresponding immune response. Since cellular immunity is critical in controlling Brucella infection, splenocytes were used to characterize the role of TLR2, TLR4 and MyD88 in the development of cellmediated adaptive immunity against Brucella. Finding that the loss of TLR2, TLR4 and MyD88 had no demonstrable effect on the development of memory T cell response against Brucella infection was initially a surprise (Figure 4) . However, recent experiments have revealed a disruption of TLR signaling resulting from enhanced degradation of MyD88-adaptor like protein (MAL) that is consistent with previous reports documenting a minimal role for TLR signaling in memory T cell development against other viral and bacterial infections (Way et al., 2003; Fremond et al., 2004; Heer et al., 2007; Seibert et al., 2010; McBride et al., 2011) . These results are also in agreement with a recent study showing that IRAK4 (interleukin-1 receptor associated kinase 4), one of the kinases recruited by MyD88 upon stimulation, is not required for generation of CD4+ and CD8+ T cells producing IFN-γ in the late stage of Brucella infection (Oliveira et al., 2011) . These results suggest that TLR-independent signaling is likely involved in the development of cellular adaptive immunity against Brucella. The contribution of humoral immunity to the outcome of Brucella infection is controversial. IgM may enhance opsonic uptake or the activity of complement. However, the differences in IgM production between knockouts are only transitory and have no correlation with differences in immune clearance. In general, IgG levels are significantly reduced in each of the knockout mice. However, the effect in the TLR4 and TLR2 knockouts is transient, while reduced IgG levels are consistently observed in the absence of MyD88 function. A second marker associated with the loss of MyD88 is the significant level of circulating IgA observed in the MyD88 knockout mice. Surprisingly, serum IgA levels were elevated in MyD88 −/− mice, but not in TLR2 −/− and TLR4 −/− mice following Brucella infection, suggesting that the IgA class switch recombination or secretion is enhanced in MyD88 knockout mice. Alternatively, higher Brucella burden in the lung constantly stimulates IgA secretion. However, this hypothesis is not supported by the fact that serum IgA is not detected in the TLR2 and TLR4 knockout mice despite similar levels of bacterial burden in the lung. In fact, this phenomenon has been previously reported in MyD88 −/− mice orally vaccinated with attenuated Salmonella expressing a Streptococcus pneumoniae surface antigen (Park et al., 2008) . It could prove useful to understand the biological mechanism involved and its potential application to vaccine design. The contribution of either elevated IgA or suppressed IgG to the inability of MyD88 knockouts to control infection is under investigation. Since a reduction in the IgG1/IgG2a ratio is associated with the development of a T H 1 response and protection against Brucella infection, IgG, IgG1 and IgG2a levels were characterized in TLR2, TLR4 and MyD88 following aerosol exposure (Figure 5 , panel F). The results reveal a significant contribution of TLR4 and MyD88 to IgG1 production and the importance of TLR2 and MyD88 to IgG2a production. Examination of the IgG1/IgG2a ratio suggests the dependence of a protective T H 1 response on expression of MyD88 and TLR2, but not TLR4. This is not a totally unexpected result since the Brucella LPS, and/or TcpB primarily exert their influence on TLR4-based signaling. The contribution of MyD88 and TLR2 to protective immune response as determined by the reduced bacterial burden in selected tissues is borne out in previous discussion. To reiterate, MyD88 −/− exhibits significant delays in clearance from all three tissues examined; TLR2 and TLR4 contribute to clearance from the lung alone. The importance of TLR signaling in the induction of antibody production remains controversial. In one such study, it was shown that TLR is not required for mice to generate robust antibody response (Gavin et al., 2006) . Other researchers report that TLR signaling is critical for antibody production and isotype switching (Pasare and Medzhitov, 2005; Heer et al., 2007) . A recent study by Weiss et al. show that TLR2, TLR4, TLR2/TLR4, and MyD88 are not required for anti-Brucella IgG production (Weiss et al., 2005) . In contrast, our results reveal that production of Brucella specific IgG is delayed in TLR2 and TLR4 knockout mice and impaired in MyD88 knockout mice. Possible reasons for the discrepancies include the use of different Brucella species, B. melitensis 16 M vs B. abortus S19, differences in virulence and different routes of inoculation (aerosol exposure vs. intraperitoneal injection). Since IgM production is not affected by MyD88 deficiency and only transiently affected by TLR2 and TLR4 absence, we conclude that TLR2, TLR4 and MyD88 are critical for IgG class switch recombination, which might be critical to the control of Brucella infection, and should be considered when Brucella designing vaccines. In our current report, we have clearly demonstrated that TLR2 and TLR4 are critical to clearance of Brucella infection from the lung, although less prominently than MyD88. This contrasts significantly with the clearance profile for the spleen and liver and may arise as a result of differences in the TLR expression pattern or levels reported for different tissues (Nishimura and Naito, 2005) . It should be noted that based on results of previous studies Frontiers in Cellular and Infection Microbiology www.frontiersin.org September 2012 | Volume 2 | Article 115 | 8 following i.p. inoculation in which only the spleens were examined, it was concluded that Brucella infection was unaffected by the loss of TLR2, however, bacterial load in the lungs and livers were not determined in these studies. (Campos et al., 2004; Weiss et al., 2005; Copin et al., 2007; Macedo et al., 2008) . In agreement with previous studies, the results confirm MyD88 as a critical factor in the control of Brucella infection in the spleen (Campos et al., 2004; Weiss et al., 2005; Barquero-Calvo et al., 2007; Copin et al., 2007; Macedo et al., 2008) . However, caution is warranted in interpretation of the significance of host factors based on the analysis of a single tissue. Recent results in our laboratory indicate that Brucella may be recovered from the lung following i.p. inoculation, indicating a need to consider persistence in this tissue irrespective of the route of infection. Another unexpected observation in this study was the delay in infection of the spleens and livers of TLR4 −/− and MyD88 −/− mice 1-2 weeks post-inoculation. These results suggest that TLR4 and MyD88 play significant roles in mediating Brucella dissemination from the lungs to other tissues, which is consistent with a previous observation indicating a role for TLR4 in the uptake of smooth Brucella by macrophage (Pei et al., 2008) , and that translocation of bacteria across mucosal or intestinal barriers is mediated by TLR4 (Neal et al., 2006) . Evaluation of Brucella infection dynamics derived from different routes of entry during the first week post aerosol exposure may be expected to confirm these results and potentially suggest post-exposure treatments. Over millions of year's co-evolution between host and pathogen have resulted in strategies to survive in the host and cause disease presumably without eliminating the host entirely. Recent studies reveal that bacteria exploit the host TLR pathway, resulting either in enhanced virulence or immune suppression (Arpaia et al., 2011; Round et al., 2011) . Our previous study has shown that Brucella can use TLR4 to gain entry into the host cells (Pei et al., 2008) . It is possible that Brucella inhibit IgA class switch recombination or secretion via MyD88 by an unknown mechanism in order to initiate infection through mucosal surfaces. Overall, our data demonstrate for the first time that TLR2, TLR4, MyD88 are essential for clearance of Brucella from the lung following aerosol exposure. Although TLR2, TLR4 and MyD88 are not required for the development of cell-mediated adaptive immunity, they play diverse roles in Brucella antigen specific antibody production and antibody class switching. The information obtained from this study will greatly facilitate efforts to understand immunity to Brucella and the rational design of novel vaccines against Brucella infection.

Diagnostic Devices for Isothermal Nucleic Acid Amplification

Since the development of the polymerase chain reaction (PCR) technique, genomic information has been retrievable from lesser amounts of DNA than previously possible. PCR-based amplifications require high-precision instruments to perform temperature cycling reactions; further, they are cumbersome for routine clinical use. However, the use of isothermal approaches can eliminate many complications associated with thermocycling. The application of diagnostic devices for isothermal DNA amplification has recently been studied extensively. In this paper, we describe the basic concepts of several isothermal amplification approaches and review recent progress in diagnostic device development.

Nucleic acid amplification is one of the most valuable tools in nucleic acid detection because it can amplify fewer than 10 target copies, significantly improving assay sensitivity. The polymerase chain reaction (PCR) was introduced by Mullis [1] and has since become an indispensable tool in numerous OPEN ACCESS molecular research and diagnostic applications. Related advanced technologies, such as multiplex PCR, nested PCR, real-time PCR, and reverse transcription PCR (RT-PCR), have been used for bimolecular analysis. However, there are numerous features confining the applicability of PCR. The approach requires thermal cycling instrumentation, considerable expertise, and a substantial amount of space in routine diagnostic laboratories, thus limiting its use to highly sophisticated facilities. These limitations in current PCR-based techniques have spurred the development of a new molecular-biological technique known as isothermal nucleic acid amplification. The major difference between PCR and isothermal amplification are the temperature reaction condition requirements. Stringent reaction conditions, including thermal cycling steps at specific temperatures, are employed in PCR, whereas only a single optimal reaction temperature is required for the entire isothermal amplification reaction, thus providing simpler and more effective reaction conditions without expensive equipment. Additionally, isothermal DNA amplification produces longer DNA fragments than the conventional PCR technique. Overall, isothermal nucleic acid amplifications have greater amplification efficiency and produce higher DNA yields than PCR owing to their undisrupted and sustained enzyme activity. With the advent of microfabrication technology, one of the directions taken to address the future needs of bioanalysis and clinical diagnosis is the development of micro total analysis systems (µTAS) or labs-on-a-chip (LOC). This scaling down capability supports an exceptional ability to miniaturize various functional units such as pumps and reactors, making it possible to integrate and automate processes into a microsystem. Additionally, it offers important advantages over bulk or large-scale analysis including rapid assay results, high-throughput screening, and low consumption of reagents. Further, the energy required for microfabrication and operation is remarkably reduced. Most importantly, these benefits make microchip systems amenable to near-patient and point-of-care testing. The development of DNA amplification microinstruments began in the 1990s, when the concepts of integrated microfluidic devices were introduced to take advantage of microfabrication technology for biological and chemical analyses [2] . To establish such a system, it was desirable to create a totally integrated device performing a series of specific molecular functions such as nucleic acid extraction and purification, nucleic acid amplification and detection, and other supporting analysis techniques, with minimal dead volumes. Owing to the overwhelming quantity of literature available on isothermal DNA amplification devices, we will describe the strategies of five major isothermal techniques. Because several reviews have previously focused on isothermal methods in bioanalysis applications [3] [4] [5] , we focus mainly on recent advances in the rational design and fabrication of integrated DNA microchips. The measurements of amplified DNA using different approaches will also be reviewed. Finally, future challenges and perspectives on diagnostic device construction are described. Isothermal approaches can facilitate rapid target amplification through single-temperature incubation, reducing system complexity compared to PCR-based methods. Established isothermal amplification methods differ in terms of complexity (multiple enzymes or primers), attainable sensitivity, and specificity. In this section, we introduce the main isothermal methods used in diagnostic systems, including nucleic acid sequence-based amplification, strand displacement from the molecular beacon, and the fluorescence provides a real-time monitoring of NASBA progress [5, 14, 15] . Recent effort has shown that an automated NASBA system, NucliSENS EasyQ, can perform simultaneous amplification and detection using fluorescence quantification. The detection of amplification products takes place in a single closed tube to significantly reduce contamination risks. This platform also helps decrease the hands-on time and provides rapid results (within 4 h), thus becoming a potentially suitable device for diagnostic applications [16] [17] [18] [19] . Although the NucliSENS EasyQ platform can obtain measurements simply and rapidly at central laboratories, the system has had limited application outside of this context. With the goal of bedside monitoring, many researchers have reported on integrated analysis systems that make it possible to shift NASBA applications from high-cost, tabletop systems to low-cost, portable devices. Esch et al. developed a NASBA assay in conjunction with fluorescence detection on a microfluidic device [20] . This device consisted of a polydimethylsiloxane (PDMS) block with a single channel, placed on a gold-coated glass slide at the device's center to immobilize the probe. Detection was accomplished using a sandwich hybridization of the NASBA products between capture probes and reporter probes tagged with carboxyfluorescein-filled liposomes. This technique had a detection limit of 5 fmol/L for a sample size of 12.5 μL. A later publication by Dimov et al. reported a microfluidic diagnostic device that integrated solid-phase extraction, real-time fluorescence detection, and a NASBA assay [21] . The integrated microfluidic NASBA chip consisted of two reaction chambers: a silica bead-bed chamber for sample purification and concentration, and a NASBA chamber for RNA amplification. To improve the efficiency of the NASBA reaction, all chambers were incubated with bovine serum albumin overnight before the reaction was started. Adequate amounts of the NASBA product were obtained after a reaction time of 30 min. Earlier this year, Zhao et al. introduced the concept of an integrated microfluidic chip-based system to monitor pathogens in a water environment with femtomolar sensitivity. The system, called immuno-NASBA, combined the versatility of enzyme-linked immunosorbent assay (ELISA) with the amplification power of NASBA [22] . The device was modeled on a 96-well ELISA microplate with 43 reaction chambers so that it would be fully compatible with a conventional reader. Moreover, the chip contained six parallel reaction channels to perform the simultaneous detection of six targets. Immuno-NASBA diagnostic devices have powerful potential to be applied for the diagnosis of various infectious diseases. Strand displacement amplification (SDA) was described in 1992 [23] and was improved in the same year [24] . There are four sequence-specific primers used in this isothermal amplification. The first set of primers (S 1 and S 2 ) is designed to have single-stranded restriction enzyme recognition site overhangs, and the second set of the primers (B 1 and B 2 ) represent the bumper primers. The DNA target is first denatured by heat and each strand is allowed to hybridize with two primers (S 1 and B 1 ), which are annealed to the DNA template. The B 1 extended product displaces the extension from the S 1 primer, which can hybridize to the opposite strand primers (B 2 and S 2 ). Thus, newly synthesized DNA that has been extended from the primers is cleaved by a corresponding restriction endonuclease, and the amplification is repeated by the polymerase, thus generating the newly synthesized strands rolling-circle and circle-to-circle amplification and the subsequent microchip electrophoretic analysis of bacterial genes (Figure 3 (B)) [60, 61] . A clinical sample was detectable in less than 65 min after the reaction was initiated. In addition to single-target detection, RCA is also desirable for multiple-analyte sensing assays because amplified products are considered to be localized at the array spot [62] . An array of real-time RCA in combination with the parallelism of arrays was developed by Yang et al. for protein quantitation down to the low nanomolar range [53] . Konry et al. constructed a two-layer sandwich assay on microbead surfaces for the combined detection of DNA and protein molecules in a single approach [63] . This array chip achieved detection limits of 1 pM and 10 fM for target DNA and proteins, respectively. Loop-mediated isothermal amplification (LAMP) is one of the DNA amplification technologies that employ a constant temperature [64] . The Bst polymerase plays a key role in the LAMP reaction process. The Bst polymerase, which is derived from Bacillus stearothermophilus living in hot springs with temperature around 70 °C, has polymerize activity, 5'-3' exonuclease activity, and strand displacement ability. At a suitable temperature, Bst polymerase with strand displacement activity can separate the non-template strand from the template DNA without the thermal cycles of the PCR process, which uses Taq polymerase to synthesize new DNA strands. Subtle primer design is also necessary for a successful LAMP reaction. In the first stage of the reaction, the so-called outer and inner primer pairs can make dumbbell-like loop DNA strands from the target DNA templates, and the dumbbell-like DNA strands become the new template DNA for the next step ( Figure 4) . The dumbbell-like DNA strands then continue replicating to become a flower-like long-chain DNA product [65] . In addition to these two primer pairs, a third pair known as loop primers has been designed and proven to be beneficial in accelerating the amplification process. A good primer design not only ensures successful execution of LAMP, but also increases the sensitivity and specificity of the reaction result [66] . Thus, the LAMP reaction is carried out by three pairs of primers in an isothermal condition. Compared to the PCR, the reaction time of LAMP is shorter while the sensitivity and specificity are almost the same or even better. For fixed temperature heating, the heater component of the device can be simpler relative to traditional DNA amplification instruments. These features afford LAMP strong potential as a disease screening method based on the economic benefits of clinical point-of-care devices with simpler designs. Because of convenience, high efficiency, and the specificity of LAMP, it has been applied to many DNA screening tests, especially virus detection. Microfluidic chips have been applied to the detection of LAMP reactions in recent years. Some chips are used only for guiding the reaction buffer and DNA solution to the reaction chamber, whereas others are combined with additional technologies such as nanostructures for sample concentrating, electrophoresis, magnetics beads, etc. A microfluidics chip made of PMMA has been used for the turbidity detection of the hepatitis B virus (HBV) LAMP reaction by our group [67, 68] . With a disposable LAMP microreactor and optical fiber-based turbidimetry device, as shown in Figure 4 , the lowest limitation for detection of the HBV DNA template was 50 copies/25 μL with the critical detecting time set at 30 min. Helicase-dependent amplification (HDA) is based on natural DNA replication mechanisms. Initially, the coordinated action of helicases unwinds and separates the template DNA duplex. The primer can hybridize with the free single-stranded templates, and the subsequent extension by a DNA polymerase will result in DNA amplification (Figure 6(A) ). The original reaction reported in the literature is performed at 37 °C for the entire process, and more than a million-fold amplification of DNA fragments can be achieved from nanogram quantities of genomic DNA [79] . Unlike the PCR, HDA uses helicases instead of heat, thus eliminating the need for any denaturation steps. Nevertheless, two additional accessory proteins are required in this approach: MutL to stimulate helicase unwinding activity and a single-strand binding (SSB) protein to prevent premature re-association of the separated ssDNA. A thermostable helicase may be also advantageous for HDA. Recently, a new helicase was developed from Thermoanaerobacter tengcongensis, which can be operated at temperatures from 45 °C to 65 °C [80] , so HDA reactions are now generally performed at the higher temperature of 65 °C. The use of thermostable helicase led researchers to abandon both the MutL and SSB proteins, while simultaneously improving the DNA yield of the reaction [81] . This simple thermal management option makes HDA very attractive for the development of simple portable DNA diagnostic devices and point-of-care testing. Recently, electrochemical methods for the detection of DNA in combination with HDA have been developed. A DNA-based sensor for the detection of M. tuberculosis using the electrochemical detection of gold nanoparticles was developed [82] . The dextrin-coated gold nanoparticles (AuNPs) used as a reporter can be electrochemically detected on a screen-printed carbon electrode chip ( Figure 6(B) ). Kivlehan et al. developed a real-time electrochemical method for HDA using the monitoring of intercalating redox probes [83] . The binding of redox probes to the HDA products (amplified double-stranded DNA) led to less electrochemically detectability, compared with the probes' free counterpart. This method of electrochemical HDA detection does not require the immobilization of the probe on the electrode; real-time isothermal HDA reactions with 48-electrochemical microwells can be performed in 1 h. Therefore, it has the potential to be a reliable method for sequence-specific DNA detection. Lateral flow test strips provide a promising tool for the development of point-of-care nucleic acid biosensors. Consequently, HDA has been employed with an embedded lateral-flow DNA detection strip for end-point assay to detect HIV-1 in human plasma [84] . The principle of this approach is based on a sandwich immunoassay using two probes: a fluorescein isothiocyanate (FITC)-labeled capture probe and a biotin-labeled detection probe. The HDA products hybridize with the capture probes and detection probes to form the complex. The hybrids are bound to streptavidin-conjugated color particles and are captured on the test zone by the interaction between the target DNA-FITC capture probe and an anti-FITC antibody. The accumulation of color beads in the test zone of the fiberglass paper is visualized as a characteristic red band. This assay provides the satisfactory detection of HIV-1 RNA at 50 copies/assay. This disposable amplicon detection device based on HDA has also been applied to the herpes simplex virus [85] and Mycobacterium tuberculosis diagnosis [86] and shows a performance comparable with conventional detection assay. Nevertheless, sample preparation, target amplification, and nucleic acid testing are conducted as distinct steps. Trau's group has proposed a beacon-assisted detection amplification (BAD-AMP) by DNA polymerization in conjunction with the nicking event [90, 91] . Two enzymes are used in BAD-AMP: the DNA polymerase that replicates the DNA target on the beacon and the nicking endonuclease that cuts the replicated single strand at the recognition position. Initially, the reaction can be activated by the addition of target DNA to switch the conformation of the beacon. When a new DNA is synthesized, the target is displaced by the polymerase with strand-displacement activity. This polymerization eventually leads to the newly synthesized DNA strand with a recognition sequence for the DNA endonuclease. This allows an enzyme to nick the DNA strand, such that the polymerase can also displace the nicked strand. BAD-AMP leads to exponential amplification by repeating cycles of polymerase and endonuclease activity. Because this strategy is a relatively simple technique, BAD-AMP has also been applied for the construction of molecular logic gates [92] . Hybridization chain reaction (HCR) is a short DNA amplification technique that is based on hybridization and strand-exchange reactions for selective and specific extension [93] . Two complementary, kinetically trapped DNA hairpins coexist in solution until the introduction of target strands initiates a cascade of hybridization events. Because there is no requirement for enzyme amplification of the signal, HCR can be performed at room temperature. The major drawback of HCR is that it provides linear amplification only, compared to the PCR, which produces exponential amplification. Various approaches with labeled hairpin probes have been reported to improve the sensitivity of targets [94] [95] [96] [97] . Although HCR is the simplest method among the isothermal nucleic acid amplifications, there are no reports on the development of an integrated HCR chip. The aim of this review was to briefly describe the current state of the art of diagnostic devices for isothermal nucleic acid amplification. The isothermal strategy has been a versatile and powerful technique applied in the detection of microbial and viral pathogens, among many other uses in the diagnostic laboratory. The combination of the properties derived from isothermal amplification and biosensing platforms proved a valuable strategy for simplifying the analytical science of nucleic acid detection. In reviewing the various detection configurations, we observed that integrated microchip systems are particularly desirable because these systems provide significant advantages in convenience and cost-effectiveness, simultaneously simplifying operational procedures and shortening analysis times. To date, the development of chip-based isothermal assay systems has received great attention, whereas achieving a higher degree of portability remains a challenge. No device reported thus far is clearly superior, resulting in the possibility that sensing platforms based on different isothermal amplifications may find their way to market. Commercialization requires further improvement in on-chip sample pretreatment, miniaturization of detectors, decrease in power consumption, and the establishment of quality control. We can expect the full integration of all components on disposable credit-card-sized systems for isothermal nucleic acid amplification and detection in the near future. Given the great effort being invested in isothermal DNA microchip systems, there is no doubt that they will provide significant contributions to point-of-care diagnostics and decentralized testing.

Clinical characteristics of pediatric hospitalizations associated with 2009 pandemic influenza A (H1N1) in Northern Bavaria, Germany

BACKGROUND: The 2009 pandemic influenza A (H1N1) (PIA) virus infected large parts of the pediatric population with a wide clinical spectrum and an initially unknown complication rate. The aims of our study were to define clinical characteristics and outcome of pandemic influenza A (H1N1) 2009-associated hospitalizations (PIAH) in children <18 years of age. All hospitalized cases of children <18 years of age with laboratory-confirmed pandemic influenza A (H1N1) 2009 in the region of Wuerzburg (Northern Bavaria, Germany) between July 2009 and March 2010 were identified. For these children a medical chart review was performed to determine their clinical characteristics and complications. RESULTS: Between July 2009 and March 2010, 94 PIAH (62% males) occurred in children <18 years of age, with a median age of 7 years (IQR: 3–12 years). Underlying diseases and predisposing factors were documented in 40 (43%) children; obesity (n = 12, 30%), asthma (n = 10, 25%) and neurologic disorders (n = 8, 20%) were most frequently reported. Sixteen (17%) children received oxygen supplementation; three (3%) children required mechanical ventilation. Six (6%) children were admitted to an intensive care unit, four of them with underlying chronic diseases. CONCLUSIONS: Most PIAH demonstrated a benign course of disease. However, six children (6%) needed treatment at an intensive care unit for severe complications.

Influenza is a common cause of illness in children, predominantly treated as outpatients. For seasonal influenza, annual incidences for influenza-associated hospitalizations were estimated as 90 (CI 95%: 80-110) /100,000 in children <5 years old in the USA [1] , and as 123/100,000 in children <6 years of age in a German region [2] . In 2009, an influenza pandemic was caused by a new influenza A/H1N1 virus (PIA). In more than 214 countries laboratory-confirmed PIA cases were reported, including over 18,097 deaths [3] . During this pandemic especially children appeared to be affected. In the USA, about 87,000 PIA hospitalizations (PIAH) and 1,280 fatalities occurred in children <18 years of age from April 2009 to April 2010, representing 32% of all PIAH and 10% of all PIA-associated fatalities in the US population [4] . In Germany, first cases of PIA were reported in April 2009; the first wave of the pandemic lasted from calendar week 42/2009 to 02/2010 [5] . The sentinel surveillance system for laboratory-confirmed PIA reported 226,137 cases and 253 fatalities for the total population until April 2010 [6] . A nationwide surveillance of critically ill children admitted to pediatric intensive care units (PICU) and fatalities associated with PIA estimated an incidence rate of 2.8 cases/100,000 children in infants * Correspondence: Liese_J@kinderklinik.uni-wuerzburg.de † Equal contributors 1 <1 year of age and 0.8 cases/100,000 in children <15 years of age [7] . Thus far, there is only limited data on clinical characteristics and outcome of PIAH in Germany [8] . In the current study, we therefore investigated clinical characteristics of all pediatric PIAH in a defined geographical region. All children and adolescents <18 years of age with laboratory-confirmed PIA admitted to one of the three hospitals covering pediatric in-patients in the area of Wuerzburg (Northern Bavaria / Germany) from July 2009 to March 2010, covering the influenza season 2009/2010, were included. The study population size in our defined catchment area corresponded to approximately 60,300 children younger than 18 years of age (Bavarian State Office for Statistics and Data Processing 2009, https://www.statistik.bayern.de). Respiratory samples (nasopharyngeal aspirate or nasopharyngeal swab or tracheal secretion) from all three hospitals were routinely sent to the Institute of Virology and Immunobiology of the University of Wuerzburg and analyzed using reverse transcriptase-polymerase-chain -reaction (RT-PCR) or direct immunofluorescence (DIF) testing for influenza A and B. All specimens tested positive for influenza by DIF were determined by PCR for subtype determination. Laboratory-confirmed PIA-infections were reported by the Institute of Virology and Immunobiology. For all identified PIAH patients <18 years of age a medical chart review was conducted by a representative of the study coordination centre (University Children's Hospital, Wuerzburg). Date of admission, number of days in hospital, data on treatment, demographic, clinical and epidemiological data were obtained using a standardized and anonymous data collection form. All data were collected for the duration of hospital stay. For analysis of clinical parameters, all identified PIAH patients with symptoms starting before hospitalization or less than 72 h after hospitalization were included. Nosocomial infections with symptoms starting >72 h after hospitalization (n = 10 cases, including one fatality) were excluded from the present analysis and will be described elsewhere. Data were analyzed descriptively using SPSS (version 18.0 and 19.0, Chicago, IL). Continuous data were reported as median with interquartile range (IQR) and categorical variables as percentage of patients. The study was approved by the Institutional Ethics Review Board of the University of Wuerzburg. Figure 2 ). Of all 94 patients, 88 (94%; 60% males, median 7 (2-12) years of age) were admitted to a general ward and stayed for a median of four (3-6) days in hospital. Six children (6%; 83% males, age range 0.1-16 years) were admitted to an intensive care unit (ICU), with a median stay at ICU of three (2-6) days and of seven (3-18) days in hospital. For 48 patients (51%), a contact with a person with suspected (n = 35) or confirmed (n = 29) PIA was reported as potential source of infection; most frequently, contacts to siblings (n = 30) or the child's mother (n = 15) had been documented. The onset of symptoms, reported for 92 (98%) patients, occurred at a median of two (1-3) days before admission to hospital. The most frequently reported symptoms at admission were cough in 75 (80% of the 94 children), fever in 73 (78%), rhinorrhea in 48 (51%), and refusal of food or drink in 36 (38%) cases ( Table 1 ). The most frequent diagnoses were upper respiratory tract infection in 84 (89%) children, bronchitis in 20 (21%; including nine (10%) cases of obstructive bronchitis), and pneumonia in 16 (17%) children, including one (1%) case of secondary bacterial pneumonia. Other reported diagnoses were fever convulsion in seven (7%), laryngitis/croup in six (6%), and otitis media in three (3%) children (multiple diseases per patient possible). At least one underlying medical condition was documented for 40 (43%) of the 94 children; the most frequent condition was obesity defined as Body Mass Index >90th percentile (n = 12, 30% of all 40 children with predisposing factors), asthma (n = 10, 25%), neurologic disorders (n = 8, 20%), preterm birth, allergic diseases and other chronic diseases (each n = 6, 15%). Oseltamivir was administered in 23 (25%) children for a median of three (2-5) days. A total of 28 (30%) children were treated with antibiotics, administered orally for a median of four (1-6) days and intravenously for a median of five (3-6) days. A total of 16 (17%) children received oxygen supplementation; three (3%) children required mechanical ventilation for a median of four (range 2-5) days. Patients admitted to an intensive care unit ICU treatment was reported for six (6%) of the 94 patients. Four of them had known underlying chronic diseases or predisposing factors: i) A two-month-old former preterm (male) with congenital heart defect was admitted due to desaturation and pneumonia. After treatment with oxygen and antibiotics, he was discharged after three days on PICU and 31 days in hospital. ii) A ten-year-old boy with chronic neurologic and lung diseases was treated at PICU for one day due to fever convulsion. iii) A ten-year-old boy with tuberous sclerosis as chronic disease was admitted to PICU due to fever convulsion and was intubated and mechanically ventilated at the PICU for five days because of hypopnoea. iiii) A sixteen-year-old girl with obesity (BMI >30) and nicotine abuse as risk factors was treated at PICU due to pneumonia and acute respiratory distress syndrome. She was mechanically ventilated at the PICU for four days due to respiratory failure. Three of 94 hospitalized children (3%), two with underlying diseases, had been vaccinated against PIA five days (n = 2) or about two months before hospitalization (n = 1). One child without underlying diseases had been vaccinated against seasonal influenza before hospitalization (no information about date of vaccination). Only two (5%) of 40 children with underlying chronic conditions for whom influenza vaccination is generally recommended in Germany had received PIA vaccination, compared to one child (2%) of the 54 children without underlying chronic conditions. None of the six children with severe complications had received any influenza vaccination. During the influenza A/H1N1 epidemic a total of 94 children with PCR-confirmed PIA were hospitalized in the three hospitals in Wuerzburg, Northern Bavaria, which cover the pediatric population of the city of Wuerzburg and its surroundings. The clinical course was mostly benign with cough (80%), fever (78%), and rhinorrhea (51%) as predominant symptoms. Only six percent of PIAH were admitted to an ICU; which is about three times lower than reported from Argentina, Canada and USA (17-19%) [9] [10] [11] but similar to 8% found in a UK hospital [12] and identical to 6% of ICU admissions reported from a large hospital in Hamburg (Northern Germany) [8] . In Germany, both in our study and in Hamburg, a clearly lower percentage of patients with PIA received treatment with oseltamivir (25% and 28%, respectively) and antibiotics (30% and 25%, respectively) [8] , when compared to antiviral (46-99%) [9] [10] [11] and antibiotic (74-86%) treatment in other PIAH studies [9] [10] [11] . In Argentina, oxygen was supplemented five times more often (82%) than was to be observed in our study (17%) [11] . Mechanical ventilation was required six times more often (17%) than documented in our study (3%), whereas data from USA and Canada revealed a similar frequency (6%) for hospitalized patients [9, 10] . The low rates of severe influenza cases in the present study correspond with the results of an earlier study on severe seasonal influenza in Germany [13] . Hence, on the one hand, the higher primary and secondary complication rates among hospitalized patients in other countries may reflect a real increase in the complication rate, due to delayed treatment with a limited access to primary health care. On the other hand, observed heterogeneity in the severity of hospitalized patients may result from differences in hospitalization access, with a higher threshold for hospitalization in countries with lower socio-economic status or limited health insurance (as in the USA) compared to Germany [11] . Underlying diseases were documented in 43% of PIAH, with asthma (25% of all children with predisposing factors) reported as one of the most frequent conditions. These results are comparable to the 32-40% of cases with underlying diseases (predominantly led by asthma) reported by most other surveys on pediatric PIAH [8, 11, 12, 14] . Of 40 patients with underlying diseases, 10% received ICU treatment, in contrast to only 4% of 54 previously healthy patients, indicating a more severe course of disease in risk group children. However, the majority of children with underlying disease had an uncomplicated course of disease. It may be assumed that at least in part they were hospitalized for pre-emptive treatment and monitoring of possible complications. In contrast, a recent study from seven Austrian hospitals on PIA patients seeking emergency medical care reported underlying chronic conditions only in 13% of PIA patients <18 years of age [15] . In October 2009, the German Advisory Board on Immunization (STIKO) recommended vaccination against PIA for selected risk groups. For children, it was recommended that primarily children above six months of age with underlying diseases, such as chronic diseases of the air ways, cardiovascular system, liver or kidneys, should be vaccinated. Secondarily, healthy children should be vaccinated as well [16] . The first PIA vaccine was available in Germany at the end of October 2009. In our study, 40 (43%) children suffered from underlying diseases and, hence, ideally should have been vaccinated against influenza. Of these 40 children, two children were younger than six months and 16 children became ill before the PIA vaccine was available. Of the remaining 22 children with underlying diseases only two (9%) had received a PIA vaccination. Only for one child (2%) out of 51 children without predisposing factors aged above six months a PIA vaccination was reported. The low PIA vaccination coverage found in our study is confirmed by results from cross-sectional surveys in children <14 years of age in Germany (8% coverage) [17] , and from a German surveillance study on severe PIA cases <15 years of age admitted to intensive care units (9% coverage) [7] . Potential limitations may result from the differences in criteria for inpatient treatment and use of diagnostic methods depending on the individual decision by the admitting physician. The number of hospitalizations corresponded to a conservative incidence estimate of at least 118 PIAH per 100,000 children <18 years of age. However, this may considerably underestimate the true pediatric PIAH incidence as only patients with laboratory-confirmed PIA were included; children hospitalized with respiratory symptoms or influenza-like illness without being tested for influenza were not captured in this study. The course of PIAH was predominantly benign; 43% occurred in children with chronic underlying diseases. Severe complications implying treatment at ICU occurred in six (6%) of the children, including four children with chronic underlying diseases. Better acceptance and higher vaccination coverage of risk group children with the recommended and available PIA vaccine could have prevented a considerable number of pediatric PIAH in Germany.

Pip6-PMO, A New Generation of Peptide-oligonucleotide Conjugates With Improved Cardiac Exon Skipping Activity for DMD Treatment

Antisense oligonucleotides (AOs) are currently the most promising therapeutic intervention for Duchenne muscular dystrophy (DMD). AOs modulate dystrophin pre-mRNA splicing, thereby specifically restoring the dystrophin reading frame and generating a truncated but semifunctional dystrophin protein. Challenges in the development of this approach are the relatively poor systemic AO delivery and inefficient dystrophin correction in affected non-skeletal muscle tissues, including the heart. We have previously reported impressive heart activity including high-splicing efficiency and dystrophin restoration following a single administration of an arginine-rich cell-penetrating peptide (CPPs) conjugated to a phosphorodiamidate morpholino oligonucleotide (PMO): Pip5e-PMO. However, the mechanisms underlying this activity are poorly understood. Here, we report studies involving single dose administration (12.5 mg/kg) of derivatives of Pip5e-PMO, consecutively assigned as Pip6-PMOs. These peptide-PMOs comprise alterations to the central hydrophobic core of the Pip5e peptide and illustrate that certain changes to the peptide sequence improves its activity; however, partial deletions within the hydrophobic core abolish its efficiency. Our data indicate that the hydrophobic core of the Pip sequences is critical for PMO delivery to the heart and that specific modifications to this region can enhance activity further. The results have implications for therapeutic PMO development for DMD.

Duchenne muscular dystrophy (DMD) is a severe muscle wasting disorder caused by a disruption of the dystrophin mRNA reading frame resulting in an out-of-frame transcript and a non-functional dystrophin protein. 1 In the past decade, a number of new treatments for DMD have been investigated, of which antisense oligonucleotide (AO)-mediated splice correction is one of the most promising approaches. [2] [3] [4] AOs modulate dystrophin pre-mRNA splicing, by specifically restoring the reading frame of the dystrophin gene via exon skipping, and therefore generate truncated but semi-functional dystrophin protein isoforms. In vivo studies in the DMD mouse model, mdx, have shown that systemic delivery of naked phosphorodiamidate morpholino oligomers (PMO) 5 and 2′-O-methyl (2′OMe) 6 AOs are not capable of restoring significant dystrophin protein in cardiac muscle. Notably, even direct intra-cardiac injections of naked AOs resulted in very low exon skipping. 7 And although clinical trials with 2′OMe 8, 9 and PMO 10,11 chemistries have shown great promise, there is still a need for further optimization to improve AO delivery to all skeletal muscles and to the heart. This is critical as respiratory complications 12 and cardiac dysfunction 13 are the major causes of premature death in DMD patients. In particular the cardiomyopathy becomes clinically apparent at ~10 years of age, and is exhibited in all DMD patients by the age of 18. 13, 14 While some studies have shown that the restoration of dystrophin in respiratory muscles can improve cardiac function in the absence of restored dystrophin protein in the heart, 15, 16 there is concern that any improvement in skeletal muscle function in the absence of cardiac correction may worsen the cardiac disease progression due to increases in cardiac work load. [17] [18] [19] It is therefore highly desirable that future therapies endeavor to restore dystrophin in cardiac as well as in skeletal muscles. Cell-penetrating peptides (CPPs), which may be readily conjugated to charge neutral AOs such as PMO, have shown potential for improved systemic delivery. CPPs that contain cationic amino acids, particularly multiple arginines, are highly effective in enhancing AO delivery due to their unique ability to deliver associated cargoes across the plasma and endosomal membranes. 20 Various arginine-rich peptides have been found to be particularly effective for delivery of such charge neutral AOs, [21] [22] [23] leading to systemic dystrophin production. However, dystrophin protein restoration in heart has typically required repeated 24 and/or very high dose administrations. 25 Corinne Betts 1 , Amer F Saleh 2 , Andrey A Arzumanov 2 , Suzan M Hammond 1 , Caroline Godfrey 1 , Thibault Coursindel 2 , Michael J Gait 2 and Matthew JA Wood 1 Antisense oligonucleotides (AOs) are currently the most promising therapeutic intervention for Duchenne muscular dystrophy (DMD). AOs modulate dystrophin pre-mRNA splicing, thereby specifically restoring the dystrophin reading frame and generating a truncated but semifunctional dystrophin protein. Challenges in the development of this approach are the relatively poor systemic AO delivery and inefficient dystrophin correction in affected non-skeletal muscle tissues, including the heart. We have previously reported impressive heart activity including high-splicing efficiency and dystrophin restoration following a single administration of an arginine-rich cell-penetrating peptide (CPPs) conjugated to a phosphorodiamidate morpholino oligonucleotide (PMO): Pip5e-PMO. However, the mechanisms underlying this activity are poorly understood. Here, we report studies involving single dose administration (12.5 mg/kg) of derivatives of Pip5e-PMO, consecutively assigned as Pip6-PMOs. These peptide-PMOs comprise alterations to the central hydrophobic core of the Pip5e peptide and illustrate that certain changes to the peptide sequence improves its activity; however, partial deletions within the hydrophobic core abolish its efficiency. Our data indicate that the hydrophobic core of the Pip sequences is critical for PMO delivery to the heart and that specific modifications to this region can enhance activity further. 26 has also been investigated, namely R6-Penetratin which contains six additional arginines. 27 Employing R6-Penetratin as the lead peptide, a series of peptide nucleic acids/PMO internalization peptides (Pips) were derived that were found to be much more stable to serum proteolysis. 28 Two such Pip peptides, Pip2a and Pip2b, conjugated to a dystrophin exon 23-specific (peptide nucleic acids) AO, were shown to be capable of inducing strong exon skipping and dystrophin positive fibres following intramuscular injection into the tibialis anterior (TA) muscle of the mdx mouse. Further optimization of this peptide series was carried out as conjugates to PMO, and Pip5e-PMO was identified as the most efficient peptide-PMO conjugate capable of inducing high levels of exon skipping and dystrophin restoration body wide, including in the heart, following a single dose intravenous administration. 29 The Pip5e structure comprises a hydrophobic core region flanked on each side by arginine-rich domains containing aminohexanoyl (X) and β-alanine (B) spacers. By analogy with the previous arginine-rich B peptide, 22 it was thought that the high arginine content of Pip5e contributed to overall delivery efficiency into all muscle tissues, whereas the hydrophobic region might be important for heart muscle delivery. We now report the results of a series of mutations to the hydrophobic core region of the Pip5e peptide, where this central core region amino acid sequence is reversed, scrambled, or partially deleted. These changes affect the levels of exon skipping and dystrophin restoration in multiple muscle groups, including the heart, following a single, low dose intravenous injection of the corresponding Pip6-PMO conjugates. The results show that a core length of 5 amino acids (5-aa) appears to be essential for heart dystrophin production, since reductions in core length reduced cardiac activity. Unexpectedly, an arginine residue was tolerated in one position of the hydrophobic core, but two arginine residues were not tolerated, nor an arginine in a different position. Surprisingly, skeletal dystrophin production was also reduced in these two latter cases. Our previous lead Pip series CPP, Pip5e, 29 contains two arginine-rich flanking regions and a central hydrophobic core. To further probe the composition requirements of the hydrophobic core for maintenance of good heart dystrophin production, we synthesized a range of Pip5e derivative peptides (Pip6 a-f) (Figure 1a) where mutations were made only to the hydrophobic core region, for example scrambled and partially deleted core region peptides. All peptides contained the same number of arginine residues (10) in the flanking sequences as in Pip5e, with the exception of Pip6e. These peptides were conjugated to a 25-mer PMO complementary to dystrophin exon 23, 30,31 previously validated for exon skipping in mdx mice. In contrast to the method of conjugation to the 5′ end of PMO that we utilised previously, 29 Pip6-PMO conjugates were prepared by conjugation of the 3′ end of the PMO to the C-terminal carboxylic acid moiety of the Pip peptide (Figure 1b) . We reported that there was no significant difference between the in vivo dystrophin production or exon skipping activity for Pip5e-PMO conjugated to the 3′ end of the PMO or to the 5′ end and therefore chose to utilise 3′ end conjugation for these experiments. 32 The exon skipping potential of Pip6-PMO conjugates was evaluated in differentiated mouse H2K mdx myotubes in the absence of any transfection agent (Figure 2 ) at concentrations ranging from 0.125 to 1 µmol/l. This showed that exon skipping activity in cultured muscle cells was very similar for all these constructs, including Pip5e-PMO. These results differ from the previous Pip5 series, 29 where the flanking argininerich sequences mostly contained a fixed number of arginine residues (10) but where spacings were varied through alternative placement of aminohexanoyl and β-alanine units. This resulted in small variations in exon skipping activity that correlated well with in vivo activity. In the case of Pip6 sequences, the flanking arginine-rich sequences are identical (with the exception of Pip6e, which is identical except for one arginine immediately preceding the core which is displaced into the second position of the core). The results demonstrate that cellular exon skipping activity does not depend on the sequence or length of the hydrophobic core. Note that we have previously shown that major changes in in vitro exon skipping activities are correlated instead with the total numbers of arginine residues. 33 Alterations to the Pip5e hydrophobic sequence improve splicing activity in vivo Given the potency of Pip5e-PMO in heart tissue, the aim of altering the sequence of the hydrophobic core (whilst maintaining the 5-aa length) was to identify peptides that might be more efficient at lower doses. These modifications included inversion of the hydrophobic region (Pip6a), substitution of tyrosine by isoleucine (Pip6b), substitution of glutamine in the Pip6a sequence by displacement of the arginine immediately flanking the core in the first arginine-rich flanking region (Pip6e), and a scrambled hydrophobic core sequence (Pip6f). All Pip6 peptide-PMO conjugates were administered to mdx mice as single 12.5 mg/kg intravenous injections via the tail vein and tissues were harvested 2 weeks later and assessed for activity at both RNA and protein levels. Immunohistochemical staining of dystrophin expression for all 5-aa core Pip6-PMOs revealed high levels of dystrophin production in skeletal muscles including the TA, diaphragm, and the heart (Figure 3 ). Immunohistochemical staining quantification (Figure 4a , b) was performed as previously described 16, 34 and was achieved by taking four representative frames of the dystrophin staining and correlating this with laminin staining for each section (n = 3) of the quadriceps, diaphragm and heart for each peptide-PMO treatment. Untreated mdx and treated mdx mice were normalized to C57BL10 mice. This method allows comparison of the staining intensity of dystrophin at the sarcolemma relative to laminin for each treatment group. Intensity ratios are normalized to C57BL10 samples and each region of interest at the sarcolemma (120 regions for each treatment group) is plotted on a scatter graph. The relative intensity values obtained for all four of the 5-aa core Pip6-PMO conjugates were significantly different to those of untreated mdx mice for the quadriceps and diaphragm ( Figure 4b and Table 1 ). There were very similar dystrophin restoration levels in the quadriceps (percent recovery score-%RS-range between 21.10 and 33.44%; Figure 4b ) and in the diaphragm for all treatments, with the exception of Pip6b which had a higher recovery score in the diaphragm (%RS range between 38.87 and 48.43%, Pip6b 56.72%; Figure 4b ). All 5-aa core Pip6-PMO-treated mice exhibited high dystrophin intensity values in the heart with the exception of Pip6e (other Pip6-PMOs were statistically significant compared to mdx = P < 0.0001; Table 1 ). Pip6a-and Pip6b-PMO conjugates displayed the highest recovery scores, as observed in Figure 4b (%RS 37.66% and 34.22%, respectively) closely followed by Pip6f-PMO (%RS 26.24%) and then Pip5e-PMO (%RS 17.32%). When directly compared to Pip5e-PMO treatment, only Pip6a-PMO was significantly better in the heart ( Table 2 ). These 5-aa core Pip6-PMOs were also shown to restore other dystrophin complex proteins, namely neuronal nitric oxide synthase, α-sarcoglycan, and β-sarcoglycan as illustrated by immunohistochemical staining in the TA muscle (Supplementary Figure S4) . The PCR and western blot analyses exhibited similar results to the immunostaining. The reverse transcription-PCR (RT-PCR) representative images (Figure 4d ) illustrate high exon skipping efficiency in all tissues analyzed. This is better shown by quantitative real-time PCR (qRT-PCR) results for the quadriceps, diaphragm and heart (Figure 4c) . The delta 23 transcript is normalized against "total dystrophin" for each muscle group (n = 3). Quantification of this data revealed similar levels of ∆23 skipping in the quadriceps of all 5-aa Pip6-PMO-treated mice. The data trends suggest that Pip6f-and Pip5e-PMO show the highest exon skipping in the diaphragm, and Pip6f-PMO the highest in heart (for splicing mean values, see Supplementary Figure S2a ). Western blots (Figure 4e) were performed on the tissues of each mouse and were quantified against a 50 and 10% C57BL10 control. These results were averaged and are presented in Supplementary Figure S2b. Pip6a-, Pip6b-, Pip6e-, and Pip6f-PMO conjugates exhibited the highest dystrophin protein restoration in the TA and quadriceps muscles. The levels of dystrophin restoration in the diaphragm were uniform across all of these treatments, whereas in the case of the heart, Pip6b-and Pip6f-PMO conjugates showed the highest dystrophin restoration. Protein restoration as measured by immunohistochemical staining is consistently higher than protein restoration 4 calculated by western blot analysis. These differences may be attributed to the differing "housing proteins" used i.e., dystrophin restoration quantified by immunohistochemical staining is normalized against laminin, whereas western blot analysis uses α-actinin for normalisation. Quantification of western blots has only recently been reported for dystrophin and currently uses chemiluminescence methods. It may therefore be judicious to give greater weight to the trends in dystrophin protein levels revealed by western blot rather than to the absolute values. Therefore, considering the results overall, mdx mice treated with each of the four 5-aa core Pip6-PMOs (Pip6a-, Pip6b-, Pip6e-, and Pip6f-PMO) appear to demonstrate improved dystrophin production and exon skipping in TA, quadriceps, and heart muscles compared with the previous lead candidate, Pip5e-PMO. In addition, these 5-aa core Pip6-PMOs do not exhibit evidence of toxicity, as assessed by plasma levels of relevant toxicity biomarkers, alanine aminotransferase, aspartate aminotransferase, and creatine kinase (see Supplementary Figure S5a ). Blood urea nitrogen and creatinine levels were similar to untreated mdx levels (see Supplementary Figure S5b ). All Pip6-PMO treatment groups exhibit similar biomarker levels to untreated C57BL10 controls. In addition to the need to identify Pip-PMOs with high efficiency and cardiac delivery, it was a further aim to better define those elements of the hydrophobic core of Pip peptides that are important for heart delivery. To this end, Pip peptides containing partial deletions of the hydrophobic core by 1 aa (removal of tyrosine; Pip6c) and by 2 aa (removal of isoleucine and tyrosine; Pip6d) were synthesised as PMO conjugates (Figure 1a) . Following treatment of mdx mice, immunohistochemical staining was performed and this revealed some dystrophin expression in skeletal muscles such as the TA and diaphragm for both these deletion Pip6-PMOs ( Figure 5) . Quantification of the immunohistochemical staining revealed the lowest dystrophin restoration in the quadriceps with Pip6d-treatment, closely followed by Pip6c-PMO when compared to the other Pip6-PMOs (Figure 6a,b and Table 1 ). Similarly, Pip6d-PMO displayed the lowest dystrophin restoration in the diaphragm. The recovery scores for Pip6c-and Pip6d-PMO conjugates in the heart were very low, indicating their poor efficiency (Pip6c %RS -1.06% and Pip6d 2.50%; Figure 6b) . These results are corroborated by the PCR and western blot analyses. The RT-PCR representative images ( Figure 6d ) and the qRT-PCR exon skipping results (Figure 6c ) both indicate reduced exon skipping in mdx mice treated with Pip6c-and Pip6d-PMO conjugates in quadriceps and diaphragm and negligible exon skipping in the heart. Western blot analysis revealed inefficient dystrophin protein production in the TA and quadriceps muscles and negligible dystrophin restoration in the heart (Figure 6e and Supplementary Figure S2b) . These results show that the length of the hydrophobic core is crucial not only for good heart dystrophin production but also for activity in some other muscle groups. Therefore, the arginine content of the CPP alone is not the sole predictor of dystrophin production and exon skipping efficiency for this class of peptides. Altering the position of arginine in the hydrophobic core or adding a second arginine is detrimental to dystrophin production The repositioning of an arginine from a flanking region into the core was unexpectedly well-tolerated (Pip6e-PMO) Two further Pip6-PMO conjugates were thus synthesized (Figure 1a) as derivatives of Pip6e-PMO. Pip6g-PMO contained a second arginine residue, which was moved from the second flanking region into the central hydrophobic core, and Pip6h-PMO contains an inversion of the Pip6e hydrophobic region, such that the single arginine location is altered within the core. (e) Representative western blot images for each treatment. Ten micrograms of total protein was loaded (TA, quadriceps, gastrocnemius, diaphragm, heart, and abdomen muscles) relative to 50% (5 µg protein) and 10% (1 µg) C57BL10 controls, and normalized to α-actinin loading control (for quantification see Supplementary Figure S2a) . PMO, phosphorodiamidate morpholino oligonucleotide. Surprisingly, these changes to the hydrophobic core resulted in further reductions in dystrophin expression not only in heart but also in all other tissues, as observed in the immunohistochemical staining (Figure 7a ) and in the quantifications thereof (Figure 7b,c) . The immunohistochemical staining representative images revealed very few dystrophin positive fibers in all tissues with the exception of the TA and quadriceps (Figure 7a) . With reference to the quantifications, Pip6g-PMO was not significantly different to untreated mdx in the quadriceps or diaphragm (Supplementary Figure S3a) . Statistical significance tables for immunohistochemical staining of quadriceps, diaphragm, and heart muscles for Pip6a-f-treated mdx mice relative to untreated mdx mice. Statistical significance was determined using a repeated measures, multilevel statistical model (****P < 0.0001, ***P < 0.001, **P < 0.01*P <0.05, N/S, not significant). Statistical significance tables for immunohistochemical staining of quadriceps, diaphragm, and heart muscles for Pip6a-f-treated mdx mice relative to Pip5e-treated mice. Statistical significance was determined using a repeated measures, multilevel statistical model (****P < 0.0001, ***P < 0.001, **P < 0.01*P <0.05, N/S, not significant). Both Pip6e-PMO derivatives were not significantly different to untreated mdx in heart muscles, illustrating the general inefficiency of these two peptides. Similarly, these Pip6e-PMO derivatives showed reduced efficiency in exon skipping, as illustrated in representative RT-PCR images (Figure 7e ) and in qRT-PCR analyses (Figure 7d and Supplementary Figure S3c ) in all tissues. Western blots revealed negligible dystrophin protein restoration (Figure 7f and Supplementary Figure S3d ) in all tissues with the exception of the TA muscle. These data show that an increase in the number of arginines or alteration in the location of the single arginine in the hydrophobic region of Pip6e are detrimental to both heart as well as skeletal muscle dystrophin production. The most promising therapy to date for the severely debilitating neuromuscular disorder DMD is treatment with AOs, which restores the reading frame of the dystrophin pre-mRNA by exon skipping. Two AOs, a PMO 10,11 and a 2′OMe oligonucleotide, 8, 9 are currently in clinical trials and the early Ten micrograms of total protein was loaded (TA, quadriceps, gastrocnemius, diaphragm, heart, and abdomen muscles) relative to 50% (5 µg protein) and 10% (1 µg) C57BL10 controls, and normalized to α-actinin loading control (for quantification see Supplementary Figure S2a) . PMO, phosphorodiamidate morpholino oligonucleotide. promising results have increased hope for DMD patients. However, studies involving the administration of very high doses of naked PMO into mdx mice have shown only partial restoration of dystrophin in body-wide skeletal muscles and negligible correction in heart. 16, 35 The necessity to correct dystrophin in the heart is ever more apparent following studies whereby the correction of the skeletal phenotype resulted in an increase in the cardiac workload and thus further progression of the cardiomyopathy. 17, 18 The discovery that CPP-conjugated PMOs can achieve much more effective dystrophin correction in mdx mice than naked PMOs has brought renewed promise for enhanced AO efficacy by improving cellular and in vivo delivery. We previously reported a promising peptide-PMO candidate, Pip5e-PMO, capable of restoring dystrophin protein to high levels in all muscle types, including heart, following a single 25 mg/kg administration. 29 In addition to arginine-rich sequences, Pip peptides contain a 5-aa hydrophobic section not present in the previous B-peptide lead, 26 which seemed likely to be responsible for the improved heart activity. The Pip6 series was developed as derivatives of Pip5e-PMO in an attempt to cast light on aspects of the hydrophobic core required for heart dystrophin production and also to identify even more active Pip-PMO conjugates. Our study using a moderate, single dose administration regimen has produced some interesting and sometimes surprising results. A key finding is that maintenance of the 5-aa length of the hydrophobic core region is imperative for good heart dystrophin production. One might imagine that diminished efficiency of dystrophin restoration in the heart for Pip6c-PMO and Pip6d-PMO, with sequential amino acid deletions in the core, might be correlated with the resultant lower hydrophobicity and hence a reduced capacity to enter the cell. 36 However, the in vitro results would suggest that all of these constructs are capable of entering the cells as they are all fully capable of exon skipping in muscle cells (Figure 2) . Thus, the length of the 5-aa hydrophobic core must affect a different parameter essential for in vivo heart delivery. Enhanced uptake into whole heart slices of fluorescently labeled Pip5e-PMO, compared to B-PMO, suggested instead that crossing of another barrier (for example, the endothelial lining to the heart) is improved. 30 Further heart studies are continuing with a Pip6-PMO that may help to address this issue. More surprising perhaps is that for Pip6c-PMO and Pip6d-PMO there was also some loss of dystrophin production in other muscle types. This suggests that the hydrophobic/cationic balance and/or the precise spacings of hydrophobic and cationic residues in the CPP impose more subtle effects on in vivo delivery parameters. Another clear conclusion arising from the Pip6-PMO analogues is that a specific order of hydrophobic residues within the hydrophobic core is less important at maintaining the heart dystrophin production, since an inverted sequence (Pip6a), a single substitution of an equally hydrophobic residue (Pip6b), and a scrambled sequence (Pip6f) were at least as active as Pip5e-PMO, and more efficient in heart and some muscle groups (Figure 4 and Supplementary Figure S2) . These results provide evidence that the hydrophobic core of Pip peptides is unlikely to contain a particular amino acid sequence that recognizes a specific receptor in a membrane barrier required to penetrate heart tissue, but instead the core acts as a hydrophobic spacer of some kind. Most surprising however is that Pip6e-PMO did induce some dystrophin splicing and protein restoration in heart muscle as indicated by the western and qRT-PCR results (note: not significantly different in immunohistochemical staining quantification). In the Pip-6e peptide, one arginine residue is moved into the hydrophobic core, which also results in alignment of a hydrophobic X residue adjacent to the core (X-YRFLI). One might have expected heart dystrophin production to have been completely lost in this conjugate, since a cationic amino acid (arginine) is now included in the core. By contrast, such heart activity was lost for Pip6h-PMO (X-ILFRY core) and the double arginine core conjugate Pip6g-PMO (X-YRFRLI-X core). Unexpectedly, dystrophin production was also lost in quadriceps and diaphragm for both Pip6g-and Pip6h-PMO. The unanticipated inconsistencies within the activity results for Pip6e, Pip6g, and Pip6h, and the losses of activities for Pip6c-and Pip6d-PMO, are perhaps best explained by the realization that precise spacings of the arginine residue within the Pip peptides with respect to both the outer hydrophobic amino acid spacers (X and B) and the inner hydrophobic core residues may drastically alter the pharmacological properties of each conjugate. This might occur not only through alteration in cationic/hydrophobic balance but alternatively due to secondary or tertiary structure changes of the Pip-PMOs, which could in turn affect serum protein binding or another parameter that alters the circulatory half-life, or which could affect the ability to traverse barriers required to penetrate muscle tissues. Such more subtle effects will require a more wide-ranging pharmacological and biophysical study, upon which we are currently embarking. For a complete understanding of the role of the hydrophobic core within Pip peptides, one would ideally wish to study the activities, pharmacology and biophysical parameters of a much larger range of sequence-variant Pip-PMO conjugates. The need for multi-mg synthesis of each conjugate as well as the availability of a large number of mdx 37 ). This is greatly promising for the Pip6-PMOs, as it would not be considered the optimal peptide yet still demonstrated the significant therapeutic effect of this group of peptides. These new leads provide a good basis for identification of a Pip-PMO candidate suitable for detailed physiological studies of muscle and heart function, as well as thorough toxicity profiling including dose escalation studies, in anticipation that one such Pip-PMO will proceed to clinical trial. Synthesis of peptide-PMO conjugates. Peptides were synthesized by standard Fmoc chemistry and purified by highperformance liquid chromatography. The PMO sequence (5 ′-GGCCAAACCTCGGCTTACCTGAAAT-3′) was purchased from Gene Tools LLC (Corvallis, OR). Peptides were conjugated to PMO through an amide linkage at the 3′ end of the PMO, followed by purification by high-performance liquid chromatography and analyzed by MALDI-TOF MS as previously described in preliminary communication. 32 Full details of synthesis including improvements to the experimental procedures are described in detail in the Supplementary Materials and Methods. Peptide-PMO conjugates were dissolved in sterile water and filtered through a 0.22-µm cellulose acetate membrane before use. Conjugates of PMO of Pip6a, Pip6b, Pip6e, and Pip6f were found to be predominantly stable and of similar stability to Pip5e-PMO in 100% serum for 2 hours at 37 °C, as seen by high-performance liquid chromatography and MALDI-TOF mass spectral analysis. The conjugates all showed similar degradation patterns, and intact conjugates were still observed up to 4 hours (data not shown). In vitro assays: exon skipping in mdx mouse myotubes. H2K mdx myotubes were prepared and incubated with peptide-PMO conjugates in the absence of any transfection agent at concentrations of 0.125, 0.25, 0.5, and 1.0 µmol/l by the method described previously. 29 The products of nested RT-PCR from total isolated RNA were examined by electrophoresis on a 2% agarose gel. Quantification of ∆23 transcript levels was calculated using densitometry. The MTS cell viability test (Promega, Madison, WI) showed 100% survival at the highest concentrations of peptide-PMO conjugates used in the study (data not shown). Animals and intravenous injections. Four and a half month old to 5½-month-old mdx mice were used in these experiments (n = 3). The experiments were carried out in the Biomedical Sciences Unit, University of Oxford according to procedures authorized by the UK Home Office. Pip6-PMO conjugates were prepared in 0.9% saline solution at a final dose of 12.5 mg/kg. The 160 µl total volume was administered via the tail vein of anaesthetized mice. Two weeks later mice were sacrificed by CO 2 inhalation, and muscles and other tissues harvested and snap-frozen in cooled isopentane before storage at -80 °C. Immunohistochemistry and quantification of dystrophin expression. Transverse sections of tissue samples were cut (8-µm thick) for the examination of dystrophin expression. For dystrophin visualisation and quantification, sections were costained with rabbit-anti-dystrophin (Abcam, Cambridge, MA) and rat anti-laminin (Sigma, St Louis, MO), and detected by goat-anti-rabbit immunoglobulin G Alexa 594 and goat-antirat immunoglobulin G 488 secondary antibodies, respectively (Invitrogen, Carlsbad, CA). Images were captured using a Leica DM IRB microscope and Axiovision software (Carl Zeiss, Cambridge, UK). Quantitative immunohistochemistry was performed as previously described. 16, 34 A representative image for each treatment was taken. For quantification, four representative frames of the dystrophin and correlating laminin were taken for each section (n = 3) of the quadriceps, diaphragm and heart for each treatment. Using ImagePro software, 10 regions of interest were randomly placed on the laminin image which was overlaid on the corresponding dystrophin image. The minimum and maximum fluorescence intensity for 120 regions were recorded for each treatment. The intensity difference was calculated for each region to correct for background fluorescence and untreated mdx and treated mdx were normalized to C57BL10. These values were plotted on a scatter graph. The "relative intensity means" were calculated using a multilevel statistics model. Using these values, the percentage recovery score was calculated by implementing the following equation, as described on the TREAT-NMD website (http://www.treat-nmd.eu/downloads/file/sops/dmd/MDX/ DMD_M.1.1_001.pdf): (dystrophin recovery of treated mdx mice-dystrophin recovery of untreated mdx mice)/(dystrophin recovery of C57BL10 mice-dystrophin recovery of untreated mdx mice). Staining of dystrophin associated proteins was performed as previously described 29 using a MOM blocking kit (Vector Labs, Burlingame, CA) and α-sarcoglycan and α-dystroglycan (Novocastra, Newcastle-Upon-Tyne, UK) antibodies (1:100 dilution). Neuronal nitric oxide synthase staining was performed using a goat anti-rabbit antibody (Abcam). Exon skipping in mdx mouse tissues. Total RNA was extracted from control and treated mouse tissues using TRIzol reagent (Invitrogen) following manufacturer's instructions. RT-PCR: Four hundred nanograms of RNA template was used in a 50 µl reverse transcription reaction using One Step RT-PCR Kit (Qiagen, Hilden, Germany) and gene-specific primers (Ex 20-26, Fwd: 5′-CAG AAT TCT GCC AAT TGC TGA G-3′, Rev: 5′-TTC TTC AGC TTG TGT CAT CC-3′). Cycle conditions: 50 °C for 30 minutes, followed by 30 cycles of 30 seconds at 94 °C, 1 minute at 58 °C, and 2 minutes at 72 °C. Two microliters of cDNA was further amplified in a 50 µl nested PCR (QIAGEN PCR kit) using the following cycle control, mdx untreated and Pip6g-and Pip6h-PMO-treated mice. Dystrophin immunostaining in TA, quadriceps, gastrocnemius, diaphragm, heart, and abdomen muscle groups for C57BL10, mdx-untreated and mdx-treated mice are shown. (b) Quantification of dystrophin immunohistochemical staining relative to laminin counter-stain in quadriceps, diaphragm, and heart muscles of C57BL10, mdx-untreated and mdx-treated mice. Relative intensity values for each region of interest (120 regions) are plotted and the model estimate averages calculated (presented in c) from the repeated measures, multilevel statistical model. For statistical significance tables see Supplementary Figure S3a ,b. Percentage recovery score is represented below. (d) Percentage ∆23 exon skipping as determined by quantitative real time (q-RT)-PCR in quadriceps, diaphragm and heart muscles. (e) Representative real-time (RT)-PCR images demonstrating exon skipping (skipped) in TA, quadriceps, gastrocnemius, diaphragm, heart, and abdomen muscles. The top band indicates full-length (FL) or unskipped transcript. (f) Representative western blot images for each treatment. Ten micrograms of total protein was loaded (TA, quadriceps, gastrocnemius, diaphragm, heart, and abdomen muscles) relative to 50% (5 µg protein) and 10% (1 µg) C57BL10 controls, and normalized to α-actinin loading protein (for quantification, see Supplementary Figure S3c) . PMO, phosphorodiamidate morpholino oligonucleotide. conditions: 94 °C for 30 seconds, 58 °C for 1 minute, and 72 °C for 1 minute for 24 cycles (Ex 20-26: Fwd: CCC AGT CTA CCA CCC TAT CAG AGC, Rev: CCT GCC TTT AAG GCT TCC TT). PCR products were examined by electrophoresis on a 2% agarose gel. Quantitative real time PCR: Two micrograms of RNA was reverse transcribed using a High Capacity cDNA Synthesis kit (Applied Biosystems, Branchburg, NJ). Exon skipping qPCR was performed using Syber green Kits (Applied Biosystems), primer sets (IDT) and the StepOne Plus Real-Time PCR system (Applied Biosystems). Primer sets used were as follows: total dystrophin transcripts, ex19-20: Fwd: GCCATAG-CACGAGAAAAAGC, Rev: GCATTAACACCCTCATTTGC; Delta23 dmd transcript, Fwd: GCG CTA TCA GGA GAC AAT GAG, Rev: GTT TTT ATG TGA TTC TGT AAT TTC CC. Plasmids (total dystrophin and delta 23 skipped) were used for the standard curve. Protein extraction and western blot. Control and treated muscle samples were homogenised in lysis buffer comprising 75 mmol/l Tris-HCl (pH 6.5) and 10% sodium dodecyl sulphate complemented with 5% 2-mercaptoethanol. Samples were heated at 100 °C for 3 minutes before centrifugation and removal of supernatant. Protein levels were measured by Bradford assay (Sigma) and quantified using BSA standards. Ten to fifteen micrograms of protein of untreated and treated mdx sample, and 50% and 10% of these concentrations of C57BL10 protein (positive control) were loaded onto 3-8% Tris-Acetate gels. Proteins were blotted onto polyvinylidene fluoride membrane and probed for dystrophin using DYS1 (Novocastra) and loading control, α-actinin (Sigma), antibodies. Primary antibody was detected by binding of horseradish peroxidase-conjugated anti-mouse immunoglobulin G with lumigen. Western blots were imaged (LiCOR Biosciences, Lincoln, NE) and analyzed using the Odyssey imaging system. Clinical biochemistry. Plasma samples were taken from the jugular vein of mdx mice immediately following sacrifice by CO 2 inhalation. Analysis of toxicity biomarkers was performed by a clinical pathology lab, Mary Lyon Centre, MRC, Harwell, UK. Statistical analysis. All data reported mean values ± SEM. A multilevel, repeated measures model was implemented for this study. The multilevel statistical approach builds upon traditional statistical methods and is being increasingly implemented in the social, medical and biological sciences. [38] [39] [40] [41] The model used for this study takes into account the multiple "relative intensity units" (level 1) for each mouse (level 2) for each treatment (level 3) as performed in the immunohistochemical staining quantification. In this example mdx untreated mice and Pip5e-PMO-treated mice were applied as the constant/ fixed parameter, to which the other treatments and wild-type control were compared. This was following a Box-Cox power transformation which was performed to ensure a normal distribution. Statistical analysis was performed using MLwIN version 2.25. Figure S1 . HPLC chromatogram and MALDI-TOF data for Pip6e-PMO. Figure S2 . qRT-PCR mean values table and quantification of western blots for C57BL10 control, mdx-untreated and Pip6-PMO-treated mdx mice, following a single 12.5 mg/kg, i.v. injection. Figure S3 . Statistical tables for quantitative immunohistochemical staining, qRT-PCR mean values table and quantification of western blots for C57BL10 control, mdx untreated and Pip6e-PMO derivative (Pip6g and Pip6h) treated mdx mice, following a single 12.5 mg/kg, i.v. injection. Figure S4 . Immunohistochemical staining of dystrophin complex proteins in C57BL10 control, mdx untreated, the 5-aa hydrophobic core Pip6-PMO-treated mice. Figure S5 . Toxicity assays assessed in blood samples of C57BL10 control, mdx-untreated, Pip6-PMO-and Pip5e-PMO-treated mdx mice, following a single 12.5 mg/kg, i.v. injection.

Clinical characteristics and outcome of Penicillium marneffei infection among HIV-infected patients in northern Vietnam

OBJECTIVE: This study reports the clinical characteristics and outcome of HIV-associated Penicilliummarneffei infection in northern Vietnam. METHODS: We conducted a retrospective chart review of all patients with laboratory confirmed Penicilliummarneffei infection admitted to the National Hospital for Tropical Diseases in Hanoi, Vietnam, between July 2006 and September 2009. RESULTS: 127 patients with P. marneffei infection were identified. All were HIV-infected; median CD4+ T-cell count was 24 cells/μl (IQR:12-48); 76% were men. Common clinical features were fever (92.9%), skin lesions (82.6%), hepatomegaly (61.4%), lymphadenopathy (40.2%), weight loss (59.1%) and cough (49.6%). Concurrent opportunistic infections were present in 22.0%; half of those had tuberculosis. Initial treatment regimens were: itraconazole or ketoconazole capsule (77.2%), amphotericin B (20.5%), and fluconazole (1.6%). In-hospital mortality was 12.6% and showed no significant difference in patients treated with itraconazole (or ketoconazole) and amphotericin B (p = 0.43). Dyspnea, ascites, and increased LDH level were independent predictors of mortality. No seasonality was observed. CONCLUSION: The clinical features, treatments and outcomes of HIV-associated P. marneffei infection in northern Vietnam are similar to those reported in other endemic regions. Dyspnea was an important predictor of mortality. More patients were treated with itraconazole than amphotericin B and no significant difference in treatment outcome was observed. It would be of clinical value to compare the efficacy of oral itraconazole and amphotericin B in a clinical trial.

Penicillium marneffei can cause a fatal systemic mycosis in immunosuppressed patients and is one of theleading causesof mortality in people living with Human Immunodeficiency Virus (HIV) in South-East Asia [1] [2] [3] . Penicilliosis presents primarily as a disseminated disease in HIVinfected patients with CD4+ T-cell count <100 cells/μL, involving the blood stream, skin, liver, spleen, lymph nodes, bone marrow, lung and gastrointestinal tract [2, 4] . Typical umbilicated skin lesions are present in 70% of patients, facilitating early empirical antifungal therapy and resulting in better outcomes [4] . Laboratory diagnosis is made by microscopy and culture of skin lesions, blood, lymph node, or other body fluids [3] .The majority of patients respond well to either amphotericin B or itraconazole treatment [2, [4] [5] [6] ; however no randomized controlled trials have been conducted to evaluate treatment choices for penicilliosis. We conducted a retrospective patient chart review of all patients with a compatible clinical syndrome and culture confirmed P. marneffei infection admitted to the National Hospital for Tropical Diseases (NHTD) in Hanoi between July 2006 and September 2009. Data collected included: demographics, admission date, clinical characteristics, HIV status, CD4+ T-cell count, laboratory investigations, concurrent opportunistic infections (OIs), treatment, and hospital outcome. P. marneffei was cultured from clinical specimens according to standard culture techniques [7] . For seasonality analysis, we assessed the number of penicilliosis admissions during dry versus rainy months in relation to all HIV-related admissions identified from hospital records. For outcome analysis three variables were used: 1. Survival including clinical improvement, defined as regression of symptoms such as fever, skin lesions, lymphadenopathy, hepatomegaly and splenomegaly, or resolution; 2. death at hospital discharge; 3. Lost to follow up due to that the patient were taken home or referred to another health facility and could not be contacted. We performed univariate and multivariate logistic regressions using forward stepwise selection of predictor variables. Statistically significant results (p < 0.05) from the univariate analysis were entered into a multivariate logistic regression model where continuous variables were dichotomized based on the mean values and after exclusion of correlated variables. The analysis was performed using Statistics Package of Social Sciences (SPSS version 19, USA). P. marneffei infection was diagnosed in 127 patients, 42.5% by both blood and skin lesion culture, 29.1% by blood culture alone, and 28.3% by skin lesion culture alone. The average age was 32 years (range 21-50); the majority was male (75.6%), and 81.9% came from provinces outside of Hanoi in northern Vietnam. The reported route of HIV infection was intravenous drug use (IDU) (37.0%), commercial sex (29.9%), husband to wife (15.7%), combination of IDU and commercial sex (7.9%) and unknown (9.0%). 29.9% had been on antiretroviral therapy (ART) prior to the diagnosis of penicilliosis; the mean duration of ART was 10.8 weeks (SD 18.4). The clinical and laboratory features are shown in Table 1 and Table 2 . The median CD4+ T-cell count on admission was 24 cells/μl (IQR: 12-48). Other OIs were diagnosed (n = 28, 22%) patients; including tuberculosis (n = 14, 11.0%), Pneumocystis jirovecii pneumonia (PCP) (n = 6, 4.7%), varicella-zoster virus (n = 4, 3.1%) as well as toxoplasmosis, cytomegalovirus retinitis and herpes simplex virus type 1 infection all (n = 1, 0.8%). Five patients also had bacteremia with the following pathogens: Escherichia coli, Staphylococcus aureus, Streptococcus pyogenes and Salmonella spp. Six patients were previously diagnosed and received treatment for penicilliosis; however, none had started ART. In addition to blood and skin lesion cultures, there were 9 CSF cultures performed of which one was positive for P. marneffei and 11 lymph node aspirate cultures were all negative. The initial choices of antifungal treatments are listed in Table 3 . Outcome at discharge was clinical improvement or resolution in 107 (84.2%) patients and death in 16 (12.6%). Nine patients were discharged early due to request from the family members, or referred to other hospital. Five of them were assessed within a month after discharge and four patients were lost to follow-up. The mean duration of treatment in the hospital was 15.5 days (SD 11). For diseased patients deaths occurred early in the course of treatment after an average of 6.6 days (SD 10), compared to survivors, who were treated in average 16.7 days (SD 10)(p < 0.001). Among 98 patients receiving itraconazole or ketoconazole, 85 (86.7%) had an improvement; 13 (13.3%) died. By comparison, among 26 patients receiving amphotericin B, 24 (92.3%) had an improvement; and two (7.7%) patients died. The mortality differences between the two treatment groups did not reach statistical significance (p = 0.43, χ 2test). Results of the logistic regression are shown in Table 1 for clinical variables and Table 2 for laboratory variables. Univariate analysis showed significantly higher risk of death among patients with dyspnea, defined as a combination of subjective sensation of difficulty breathing and observed tachypnea, ascites, jaundice, splenomegaly (Table 1) , increased levels of alanine transaminase (AST), bilirubin, lactate dehydrogenase (LDH), white blood cell count, blood urea, thrombocytopenia and prolonged prothromb in time (Table 2 ). However, in the multivariate analysis, only dyspnea, ascites and increased LDH levels remained independent predictors of mortality. Of the 11 patients that died 6 had dyspnea, none of them had tuberculosis or Pneumocystis jiroveci. However, two were diagnosed with sepsis, one Escherichia coli and one Streptococcus pyogenes. In total 23 patients had X-ray confirmed lesions of the lungs, 5 of these also had dyspnea (p = 0.015), 4 had pulmonary tuberculosis and 2 had Pneumocystis jiroveci, 4 died, of those 2 were diagnosed with tuberculosis. Of the 5 patients with dyspnea and lung lesions one was diagnosed with Pneumocystis jiroveci and none with tuberculosis. Assessment of seasonality was performed for the year 2007 and 2008. During these two years P.marneffei accounted for 87 of 793 (11.0%) of all HIV related admissions. The number of penicilliosis admissions was 43 during the hot rainy months (May to October) and 44 during the cooler dry months (November to April). The number of all HIV-related admissions was 463 during the rainy months and 330 during the dry months. The proportion of penicilliosis admissions in relation to all HIV-related admissions comparing dry versus rainy season did not show statistically significant difference (p = 0.07, χ 2test). Penicilliosis accounted for 11% of all HIV-related admissions at NHTD in Hanoi during 2007 and 2008 which is higher than the 4.4% reported from the major referral hospital for infectious diseases in Ho Chi Minh City, southern Vietnam [4] . However as NHTD is a specialized tertiary hospital, and most (82%) of the cases were referred, and because other epidemiological data were lacking, it cannot be concluded that there is a difference in disease prevalence between northern and southern Vietnam. The clinical features of disseminated penicillosis are consistent with other studies including profound immunosuppression (median CD4+ T-cell count: 24 cells/μl) and high rate of co-infections with other opportunistic pathogens [2, 4, 6] . One third of the patients were already on ART for in average 10 days, this indicates that many patients had an ongoing P. marneffei infection that was not revealed before initiation of ART, but was probably unmasked by immune reconstitution inflammatory syndrome (IRIS) after initiation of ART, this has earlier been reported in a few case studies [8, 9] . In our study the presence of dyspnea, ascites, and high LDH levels on admission independently predicted hospital mortality. Of the six patients that had dyspnea and died no one had tuberculosis or Pneumocystis jiroveci diagnosis, however two had septicemia. This may indicate that pulmonary involvement (i.e. P. marneffei pneumonia) drives disease severity or that the disease severity result in pulmonary lesions. This is consistent with a prior study showing that high respiratory rates and dyspnea among penicilliosis patients predicted poor hospital outcome [4] . Of the 11 patients with dyspnea 5 also had lung lesions, of these one was diagnosed with Pneumocystis jiroveci and non with tuberculosis. Lung involvement of P. marneffei has been shown in Taiwan where it was found to be the most common cause of cavities in the lungs of immunosuppressed HIV infected patients [10] . The majority of patients with dyspnea did not have lung lesions, hence the dyspnea might be due to the severe condition with multi-organ involvement and acute respiratory distress syndrome (ARDS). It should be noted that there could be some under diagnosis of tuberculosis due to the poor sensitivity for sputum microcopy and culture in immunosuppressed individuals as well as for Pneumocystis jiroveci as microscopy of sputum, obtained by nebulizer or bronchoscopy, is not routinely performed. Typical skin lesions were present in 80% of patients in this study. The pathogenesis and prognosis of skin involvement is poorly understood. Although a previous study showed that presence of typical skin lesions facilitates early initiation of empirical antifungal treatment and results in better outcome [4] , it is unclear whether skin involvement itself is a prognostic marker. In the absence of skin lesions, the differential diagnoses for an AIDS-associated febrile illness with reticuloendothelial system involvement are broad and include: Mycobacterium tuberculosis, Mycobacterium Avium Complex, histoplasmosis and cryptococcosis among others [11] . This poses a major challenge in diagnosis and treatment, especially in areas with poor access to blood culture and other diagnostic laboratory. Penicilliosis should be considered in all severely immunosuppressed HIV patients with one or more of the common presentations, skin lesions, lymphadenopathy, hepatomegaly, splenomegaly, ascites, jaundice and dyspnoea, who have been in P. marneffei endemic areas, and empirical antifungal treatment in very ill patients may be indicated. In this study 22% of the patients had a concurrent OI; of whom half had tuberculosis. Hence, multiple OIs, particularly pulmonary tuberculosis, need to be considered. Tuberculosis and penicilliosis co-infection poses a major therapeutic dilemma in resource-poor settings as rifampicin is a potent P450 inducer and markedly reduces itraconazole concentrations [12] Amphotericin B is recommended for patients with concurrent tuberculosis treatment. An alternative tuberculosis drug rifabutin is not available in most penicilliosis endemic areas. The WHO recommended treatment for severe penicilliosis, amphotericin B, was only given to 20.5% of patients. The majority (77.2%) was treated with either itraconazole or ketoconazole. Although itraconazole is a recommended alternative treatment for mild to moderate disease or when amphotericin B is unavailable [13] , in Vietnam both mild and severe cases are commonly treated with oral itraconazole because amphotericin B is often not available or is too expensive. In our study there was no significant difference between itraconazole and amphotericin B in treatment outcome. So far no randomized, prospective treatment trials have been conducted to compare the efficacy of different antifungal treatment regimens for penicilliosis. Compared to northern Thailand and southern Vietnam where cases peak in the rainy season [4] , seasonality was not observed in our cohort. It should be noted that our small sample size and short time frame does not enable any conclusive results about seasonality. However, the cool and dry season in northern Vietnam is often damp with high humidity which might have an impact on the reservoirs of P. marneffei. This study shows that penicilliosis in northern Vietnam presents with similar clinical characteristics as in other endemic areas, and that dyspnea is an important predictor of mortality. It is common practice to treat patients with oral itraconazole rather than amphotericin B, and no significant difference in treatment outcome was observed. It would be of clinical value to compare the efficacy of oral itraconazole and amphotericin B in a clinical trial to develop evidenced based guidelines. Abbreviations HIV: Human Immunodeficiency Virus; NHTD: National Hospital for Tropical Diseases; IDU: Intravenous drug use. The author declares that they have no competing interests Authors' contributions NTLH -conception and design, acquisition of data, analysis and interpretation of data, has been involved in drafting the manuscript and have given final approval of the version to be published. ML -Analysis and interpretation of data, has been involved in drafting the manuscript and have given final approval of the version to be published. HFLW -conception and design, acquisition of data, interpretation of data, revising it critically for important intellectual content and have given final approval of the version to be published. DTT -Laboratory analysis and have given final approval of the version to be published. WT -conception and design, acquisition of data, revising it critically for important intellectual content and have given final approval of the version to be published. PH -conception and design, acquisition of data, interpretation of data, revising it critically for important intellectual content and have given final approval of the version to be published. NVT-Laboratory analysis and have given final approval of the version to be published. NTMH -Acquisition of data and have given final approval of the version to be published. TL -Revising it critically for important intellectual content and have given final approval of the version to be published. NVK -conception and design, acquisition of data and have given final approval of the version to be published. All authors read and approved the final manuscript.

Prospective application of clinician-performed lung ultrasonography during the 2009 H1N1 influenza A pandemic: distinguishing viral from bacterial pneumonia

BACKGROUND: Emergency department visits quadrupled with the initial onset and surge during the 2009 H1N1 influenza pandemic in New York City from April to June 2009. This time period was unique in that >90% of the circulating virus was surveyed to be the novel 2009 H1N1 influenza A according to the New York City Department of Health. We describe our experience using lung ultrasound in a case series of patients with respiratory symptoms requiring chest X-ray during the initial onset and surge of the 2009 H1N1 influenza pandemic. METHODS: We describe a case series of patients from a prospective observational cohort study of lung ultrasound, enrolling patients requiring chest X-ray for suspected pneumonia that coincided with the onset and surge of the 2009 H1N1 influenza pandemic. RESULTS: Twenty pandemic 2009 H1N1 influenza patients requiring chest X-ray were enrolled during this time period. Median age was 6.7 years. Lung ultrasound via modified Bedside Lung Ultrasound in Emergency protocol assisted in the identification of viral pneumonia (n = 15; 75%), viral pneumonia with superimposed bacterial pneumonia (n = 7; 35%), isolated bacterial pneumonia only (n = 1; 5%), and no findings of viral or bacterial pneumonia (n = 4; 20%) in this cohort of patients. Based on 54 observations, interobserver agreement for distinguishing viral from bacterial pneumonia using lung ultrasound was ĸ = 0.82 (0.63 to 0.99). CONCLUSIONS: Lung ultrasound may be used to distinguish viral from bacterial pneumonia. Lung ultrasound may be useful during epidemics or pandemics of acute respiratory illnesses for rapid point-of-care triage and management of patients.

Emergency department visits quadrupled with the initial onset and surge during the 2009 H1N1 influenza pandemic in New York City (NYC) from April to June 2009 (Figures 1 and 2) [1, 2] . This time period was unique in that >90% of the circulating virus was surveyed to be the novel 2009 H1N1 influenza A according to the New York City Department of Health. Five-hundred sixtyseven patients requiring hospitalization were confirmed with the 2009 H1N1 influenza A in NYC [1] . In NYC, there were 16 deaths, 46% of admitted patients were <18 years old and 20% were <5 years old [2] . Eighty percent of confirmed cases had a known underlying risk condition, most commonly asthma (40% of confirmed cases) [1] . This fourfold increase in patient volume presented logistical challenges for emergency departments [1] . In response to mass casualty incident-type conditions and overcrowding, emergency departments in New York City added staffing and created alternate sites of care to accommodate the increased patient volume. Increased demand for chest radiography for those patients with more severe disease led to increased delays and length of stay for those patients with suspected, but non-severe pneumonia. Clinicians are challenged by the diagnostic dilemma that influenza cannot reliably be distinguished from other acute respiratory illnesses on the basis of clinical presentation alone [3] . Rapid viral antigen testing for diagnosis, which under ideal situations can yield results within 30 min, is not practical nor cost-effective in pandemic conditions [3] . Point-of-care ultrasound has been demonstrated to identify, in real-time, various pathologies of the lung, such as pneumonia, viral pneumonia, and acute respiratory distress syndrome (ARDS) [4] [5] [6] [7] [8] [9] [10] An algorithm for differentiating between various respiratory pathologies has been described ( Figure 3 ) [4] , and evidence-based recommendations regarding the use of point-of-care lung ultrasound have recently been published [11] . The use of lung ultrasound during the 2009 H1N1 influenza pandemic in adults has also been recently described [12] . We describe a prospective case series of children in whom clinicianperformed lung ultrasonography was used to differentiate between different respiratory pathologies and assessed interobserver agreement of these ultrasound findings during the initial onset and surge of the 2009 H1N1 pandemic (April to June 2009). We describe a subcohort of patients who required chest X-ray for suspected pneumonia and were enrolled into a prospective study of lung ultrasound for diagnosing pneumonia that coincided with the onset and surge of the 2009 H1N1 influenza pandemic from April to June 2009 [1, 2, 13] . We also describe the application of a modified Bedside Lung Ultrasound in Emergency (BLUE) protocol [4] with posterior thorax scanning ( Figure 3 ) during the This study was approved by our institutional review board. The study population consisted of a convenience sample of patients who met predetermined inclusion criteria and in whom informed consent had been obtained and documented from the patient or guardian for enrollment into the study. Inclusion criteria consisted of patients < 21 years of age presenting to the emergency department with clinical suspicion of pneumonia requiring chest X-ray for eva-luationWe excluded those patients who presented the following: (1) arrival in the emergency department with a chest X-ray, (2) a confirmed diagnosis of pneumonia by diagnostic imaging, or (3) hemodynamic instability. Enrolled patients had a screening history and physical examination performed at the time of triage to determine eligibility into the study. After informed consent was obtained, enrolled patients had clinical exam findings documented on a standardized form and underwent point-of-care lung ultrasound examination. An ultrasound machine with a linear array transducer at 7.5 to 10 MHz (Sonosite Micromaxx, Bothell, WA, USA) was used to image the lungs in perpendicular planes (transverse, parasagittal, and coronal) in the midclavicular line anteriorly and posteriorly on the chest and the midaxillary line from the axillae to diaphragm (Figure 4 ). Using a six-zone lung ultrasound scanning protocol similar to that described by Copetti et al. [7] , we defined and classified patients as positive or negative for viral pneumonia based on the presence of small subpleural consolidations usually <0.5 cm ( Figure 5 and Additional file 1) and/or individual B-lines or confluent B-lines (echogenic vertical lines arising from the pleural line to the bottom of the ultrasound screen; Figure 6 and Additional file 2) [7] . These ultrasound findings are similar to those described in interstitial syndrome which is defined as three or more B-lines in a given lung region [10, 14, 15] . A-lines (horizontal, reverberation artifacts of the pleural line; Figure 7 left) which indicate areas of the normal lung were also noted when present [10, 14] . Patients were classified as positive or negative for bacterial pneumonia based on the presence or absence of lung consolidation with air bronchograms [6, 7, 16] seen on ultrasound (Figures 7 right, 8, and Additional file 3). A clinical course with follow-up after 2 weeks (via electronic medical record and telephone interview) was used to determine disposition and outcomes of enrolled patients. Clinicians performing and interpreting ultrasound were blinded to chest X-ray results, and when performed per hospital protocol for possible admission, viral antigen testing results. Bacterial pneumonia on chest X-ray (posterior-anterior and lateral views) was classified based on the attending pediatric radiologist reading for 'consolidation, 'infiltrate, or 'pneumonia. For analysis purposes only, viral pneumonia on chest X-ray was defined as 'peri-bronchial cuffing, 'peri-bronchial thickening, or 'increased interstitial markings identified by the pediatric radiologist. Pediatric radiologists were blinded to the lung ultrasound results. Ultrasound images and videos were reviewed between two blinded investigator sonologists (enrolling sonologist and reviewing sonologist) to determine interobserver agreement by unweighted Cohens Kappa for viral pneumonia (small subpleural consolidation and/or B-lines), normal lung ultrasound pattern (A-lines), and bacterial pneumonia (lung consolidation with sonographic air bronchograms). Patient demographic and study characteristics are presented in Table 1 . Twenty pandemic 2009 H1N1 influenza patients requiring chest X-ray (CXR) were enrolled during this time period. Distribution of diagnoses based on lung ultrasound findings, chest X-ray findings, and clinical outcomes using a modified BLUE protocol [4] is presented in Table 2 . Interobserver agreement for ultrasound findings of lung consolidation with air bronchograms, B-lines or small subpleural consolidations, and A-lines by Cohen's Kappa was 0.82 (95% confidence interval (CI), 0.63 to 0.99) ( Table 3) . Ultrasound findings of lung consolidation with sonographic air bronchograms [6, 7, 16] correlated 100% with chest X-ray findings of bacterial pneumonia (reported as consolidation or infiltrate) in eight patients. All of these patients were confirmed to have pneumonia based on the clinical course at 2-week follow-up. This represented a doubling (40% vs. 20%) in the prevalence rate of bacterial pneumonia in our study during the H1N1influenza A onset and surge time period compared to the time period prior to the onset of H1N1 influenza A. The prevalence of viral lung ultrasound findings increased from approximately 50% for the overall study [13] to 75% during the surge of H1N1 influenza. Chest X-ray findings for viral pneumonia (most commonly described as peri-bronchial thickening or peri-bronchial cuffing) were present in 8 of 15 (53%) patients identified as having viral pneumonia on ultrasound. Seven of these 15 patients with viral pneumonia based on ultrasound had superimposed bacterial pneumonia also identified by ultrasound (Figure 7 and Additional file 4). All four patients in our series that required hospitalization had viral and bacterial pneumonia based on ultrasound. All patients in our series were recovering or recovered from their influenza illness on follow-up after 2 weeks. All admitted patients were subsequently confirmed with the 2009 H1N1 influenza A by the New York City Department of Health. Per hospital protocol for possible hospital admission, four of nine patients tested positive for influenza A by viral antigen testing, despite the New York City Department of Health reporting >90% of the circulating virus during this pandemic time period was the novel influenza A H1N1 [1] . One infant in the cohort was co-infected with respiratory syncytial virus based on viral antigen testing. Three patients, all <5 years of age requiring hospital admission had evidence of both bacterial and viral pneumonia on ultrasound. The only patient requiring ICU admission, a 20-year-old female, was intubated after deteriorating during her ED stay with persistent hypotension and septic shock from a left lower lobe bacterial pneumonia. This patient initially presented with an influenza-like illness and acute abdominal pain. To our knowledge, this is the first prospective series describing the use of lung ultrasound in children as a potential real-time diagnostic triage tool during a mass casualty-type incident due to an acute respiratory illness pandemic surge [17, 18] . Testa et al. have reported on similar lung ultrasound findings in adults during the 2009 H1N1 influenza A pandemic [12] . Single case reports of clinician-performed lung ultrasound to monitor the progression of H1N1 influenza-associated ARDS [19] and point-of-care echocardiography to diagnose H1N1 influenza myocarditis [20] have been described. Retrospective reports of the role of ultrasound in mass casualty incidents during disasters such as earthquakes have also been described [21, 22] . Lichtenstein et al. described an algorithm using lung ultrasonography to distinguish between various respiratory pathologies of the lung [4] . We modified Lichtenstein's BLUE protocol [4] to recognize basic lung ultrasound patterns to distinguish between the normal unaffected lung, viral pneumonia pattern, and bacterial pneumonia (Figure 3 ). Scanning the posterior thorax was added to increase the sensitivity of the protocol [23] . Point-of-care lung ultrasound was able to identify, in real-time, four groups of pandemic patients: viral pneumonia only (subpleural consolidations and/or B-lines or confluent B-lines), bacterial pneumonia only (lung consolidation with sonographic air bronchograms), both viral and bacterial pneumonia (Figure 7) , and normal lungs (A-lines only). Our calculated Kappa was 0.82, which means that the interobserver agreement in distinguishing between these ultrasound findings was excellent. These ultrasound findings facilitated triage and immediate decision making regarding the need for respiratory isolation in a negative pressure room without waiting for chest X-ray. Our median time to chest X-ray tripled (Table 1 ) during the pandemic compared to a time period prior to the pandemic. Our time to chest X-ray interpretation during the pandemic was longer than the median of 98 min reported by Zanobetti et al. in the study of emergency department lung ultrasound in nonpandemic conditions [5] . When lung consolidation with sonographic air bronchograms was visualized, point-of-care ultrasound facilitated the immediate decision to treat with antibiotics, without waiting for chest X-ray. Visualization of viral pneumonia on ultrasound may be useful to assist in the decision to initiate immediate empiric treatment with antiviral medication for future pandemic or epidemic influenza patients. In a large cohort of hospitalized H1N1 influenza A pandemic patients, only 73% of patients with radiographic evidence of pneumonia received antiviral drugs, whereas 97% received antibiotics [24] . Better recognition of viral pneumonia by ultrasound may impact outcomes, as available data have shown treatment with antiviral medication reduces mortality in hospitalized patients with influenza, even when therapy is initiated after 48 h of illness onset [24] . Our sample size was limited by the inability to enroll during the surge of pandemic patients due to time and resource constraints. Selection bias from convenience sampling may have occurred because patients were more likely to have been enrolled at less busier or better staffed times. In general, the patients in this series had illnesses severe enough to warrant investigation with chest X-ray. Thus, information about less ill or asymptomatic pandemic patients is lacking. Although our calculated interobserver agreement for lung ultrasound to distinguish between viral and bacterial pneumonia is high, the number of total observations was limited, and this is reflected in our wide 95% confidence intervals. However, it is notable that our point estimate Kappa for ultrasound is higher than the reported interobserver agreement for chest X-ray for pneumonia by pediatric radiologists, 0.51 (0.39 to 0.64) [25] . Due to the large numbers of patients presenting to our emergency department during the pandemic, only hospitalized patients (four patients in our series) were confirmed with 2009 H1N1 influenza A [1] . Finding small subpleural consolidations and/or B-lines on ultrasound allows the recognition of viral pneumonia from bacterial pneumonia (lung consolidation with sonographic air bronchograms), but it is unknown if different viruses have unique lung ultrasound patterns (e.g., influenza A from RSV). We could not report test performance characteristics, such as sensitivity and specificity, as there was no practical reference gold standard for viral pneumonia at the time our study was conducted. Additionally, chest X-ray cannot be used as a gold standard for viral pneumonia. However, according to the New York City Department of Health, >90% of the circulating virus during this pandemic time period was the novel influenza A H1N1 [1] .

Serum Levels of Gelatinase Associated Lipocalin as Indicator of the Inflammatory Status in Coronary Artery Disease

Background. Atherosclerosis is a chronic inflammatory disease and the acute clinical manifestations represent acute on chronic inflammation. Neutrophil gelatinase-associated lipocalin (NGAL) is found in the granules of human neutrophils, with many diverse functions. The aim of this study was to evaluate the hypothesis that levels NGAL in blood may reflect the inflammatory process in various stages of coronary artery disease. Methods. We studied 140 patients, with SA 40, UA 35, NSTEMI 40, and STEMI 25, and 20 healthy controls. Serum NGAL was measured upon admission and before coronary angiography. Results. Significant differences were observed in median serum-NGAL(ng/mL) between patients with SA (79.23 (IQR, 37.50–100.32)), when compared with UA (108.00 (68.34–177.59)), NSTEMI (166.49 (109.24–247.20)), and STEMI (178.63 (111.18–305.92)) patients and controls (50.31 (44.30–69.78)) with significant incremental value from SA to STEMI. We observed a positive and significant correlation between serum-NGAL and hs-CRP (spearman coefficient rho = 0.685, P < 0.0001) as well as with neutrophil counts (r = 0.511, P < 0.0001). Conclusions. In patients with coronary artery disease serum levels of NGAL increase and reflect the degree of inflammatory process. In patients with acute coronary syndromes, serum levels of NGAL have high negative predictive value and reflecting the inflammatory status could show the severity of coronary clinical syndrome.

Systemic inflammation participates in atherosclerosis evolution from the early development of endothelial dysfunction, to formation of mature atheromatic plaques, to the ultimate endpoint, rupture, and thrombotic complications [1] . Plaque rupture with the formation of an occlusive thrombus is the cause of acute coronary syndromes (ACS) [2] . Inflammatory cells, involving activated neutrophils, are more frequently found in plaques vulnerable to rupture [3] . Neutrophil activation has been reported in unstable angina (UA) and acute myocardial infarction (AMI) but not in patients with stable angina (SA) [4] [5] [6] [7] [8] [9] [10] . This activation seems to precede myocardial injury in patients with AMI [11] . Therefore biomarkers of neutrophil activation could be of prognostic and even diagnostic importance. Recent studies have shown that gelatinase B also known as matrix metalloproteinase-9 (MMP-9), an endopeptidase capable of degrading the extracellular matrix, is thought to be associated with atherosclerosis, and plaque rupture [12, 13] . Therefore, MMP-9 is considered to be an important mediator of vascular remodeling and plaque instability. The MMP-9 action is enhanced b neutrophil gelatinase-associated lipocalin (NGAL), also known as lipocalin-2, a 25 kDa glycoprotein, that is, found in the granules of human neutrophils, with many diverse functions, such as scavenger of bacterial products, modulator of inflammation, iron trafficking, and apoptosis [14] . The formation of a complex with NGAL and MMP-9 is crucial for atherosclerotic plaque erosion and thrombus formation [15] . NGAL is also produced by kidney tubular cells in response to various ischemic or toxic insults and has 2 International Journal of Inflammation been proposed as an early biomarker for the diagnosis of acute kidney injury [16, 17] . In this study, we hypothesized that levels NGAL in blood may reflect the extent of neutrophil activation in various stages of ACS and could discriminate various types of ACS (UA, NSTEMI, and STEMI) and stable from unstable coronary syndromes. One hundred and seventy consecutive patients programmed for coronary angiography to the Invasive Cardiology Department of the KAT General Hospital Athens, Greece, were recruited for this study, from June 2010 to October 2010. The study was performed according to the principles of the Declaration of Helsinki and was approved by the hospital's ethics committee. Written informed consent was obtained from all participating patients. Thirty patients were excluded from the study. Exclusion criteria included a negative coronary angiography in patients with a typical chest pain which was considered as angina or had a false positive single photon emission computed tomography (SPECT), any surgery in the previous six months, liver disease, end stage renal disease, renal cardiac or liver transplantation, neoplasia, and infection since all these can affect serum-NGAL levels. The 140 patients who fulfilled the study criteria after the clinical assessment and final diagnosis were divided into the following 4 groups: SA (n = 40), UA (n = 35), NSTEMI (n = 40), and STEMI (n = 25). Twenty (20) healthy amateur athletes without risk factors served as control group (Figure 1 ). The demographics and clinical characteristics of patients and controls are shown on Table 1 . All patients, upon presentation in emergency room, underwent an initial clinical assessment that included clinical history, physical examination, 12-lead ECG, continuous ECG monitoring, and standard blood tests (including white blood cell, polymorphonuclear neutrophil counts, and troponin-I). These tests were repeated at 6 at 12 and 24 hours as long as clinically indicated. To determine the final diagnosis for each patient 2 cardiologists blinded to NGAL results reviewed all patients available records (including patient history, laboratory results, radiologic testing, ECG, echocardiography, and coronary angiography) at the completion of their hospital stay. The SA group consisted of patients with angiographically documented organic coronary stenosis >70% by quantitative coronary angiography in major arteries who had chronic symptoms of angina or a positive SPECT test. UA was diagnosed in patients with typical angina at rest, or a sudden increase in episodes of a previously stable angina. AMI was diagnosed when there was evidence of myocardial necrosis in a clinical setting consistent with myocardial ischemia. Necrosis was diagnosed by a rising and/or falling pattern of troponin-I with at least one value above the cutoff value (defined as the 99th percentile of a normal population where the assay shows an imprecision <10%). Our troponin-I assay fulfills the imprecision criteria for concentration >0.2 ng/mL. Serum and K 2 EDTA-plasma samples were collected from all SA patients in the morning before the coronary angiography. In all ACS patients PCI was performed within 24 hours from admission and the blood samples were collected on admission. From all healthy subjects, samples were collected in the morning and before training. Serum samples were kept frozen at −80 • C until tested. Total white blood cell count (WBC), and peripheral polymorphonuclear neutrophil count (PMN) were assessed using the Cell-Dyn Sapphire haematology analyzer (Abbott, Chicago, Il, USA). Serum creatinine was measured with a modified jaffe method on Architect ci16200 analyzer (Abbott, Chicago, Il, USA). High-sensitivity CRP (hs-CRP) was measured with a turbidimetric assay on the same analyzer. Troponin-I was measured with a chemiluminescent immunoassay on the same analyzer. Serum-NGAL was measured with an ELISA (Bioporto, Gentofte, Denmark). This was performed with NCSS statistical program. Normality of distributions for quantitative data was tested with the Shapiro-Wilk test. For normally The mean age and the mean BMI of the patients did not differ significantly among the four groups whereas the controls were significantly younger and their BMI was significantly lower (ANOVA-test). The proportion of diabetic patients did not differ significantly among four patient groups (chisquare = 1.69, P = 0.639) as well as the proportion of patients with hypertension and dyslipidemia (chi-square = 1.63, P = 0.652). Finally smoking habits did not differ significantly in the first three patient groups while it was significantly higher in group 4. Those risk factors were absent from our controls. (50.31 ng/mL) ( Table 2 and Figure 2 ). Also significant were the differences observed between healthy controls and SA patients, between UA patients and patients with AMI (NSTEMI or STEMI). Patients with STEMI had higher levels of NGAL than patients with NSTEMI but the difference was nonsignificant ( Table 2) . Markers. The median plasma levels of hs-CRP were similar in patients with SA (0.40 mg/dL) and those with UA (0.69 mg/dL) and were significantly higher than the levels in the control group (0.12 mg/dL). Hs-CRP levels were significantly increased in patients with NSTEMI (1.20 mg/dL) and STEMI (6.76 mg/dL). In order to further investigate the relationship between serum-NGAL and hs-CRP, we performed regression analysis between serum-NGAL and hs-CRP (Figure 3 ). This analysis revealed that there is a linear and positive correlation between hs-CRP and serum NGAL (spearman rank correlation coefficient rho = 0.685, P < 0.0001). The differences that were observed among the four patient groups in WBC and PMN counts were statistically significant (P < 0.001, ANOVA-test). There was a positive and significant correlation between serum-NGAL and WBC (r = 0.510, P < 0.0001) and PMN (r = 0.511, P < 0.0001) counts (Figure 4) . In a multivariate regression analysis model entering as independent parameters age, serum creatinine, hs-CRP, and PMN count, we identified only hs-CRP (P < 0.005) and PMN count (P < 0.0001) as independent predictors of serum-NGAL levels. curve for serum-NGAL hs-CRP and PMN counts ( Figures 5 and 6 ). The diagnostic value for serum-NGAL in discriminating patients with UA, from those with SA is high (AUC = 0.852) and better than of hs-CRP (AUC = 0.735) or PMN count (AUC = 0.761). If we use as cutoff for serum-NGAL 83.74 ng/mL, we can predict an UA event with sensitivity and specificity, 82.8% and 75%, respectively. The negative predictive value of this cutoff is high (97.28%). The diagnostic value for serum-NGAL in discriminating ACS patients, from patients with SA is high (AUC = 0.929) and better than of hs-CRP (AUC = 0.794) and PMN count (AUC = 0.830). If we use as cut-off for serum-NGAL 89.29 ng/mL, we can discriminate an ACS patient from a stable patient with sensitivity and specificity, 89.3% and 81.6%, respectively. The negative predictive value of this cutoff is high (98.65%). In this study, we demonstrated that serum levels of NGAL are higher in patients with CAD than in healthy controls patients. Among ACS patients, these levels are gradually elevated according to the severity of the coronary clinical syndrome (UA, NSTEMI, and STEMI). Also serum levels of NGAL are higher in patients with ACS than in patients with SA and could be used, with high negative value, to discriminate patients with stable or unstable coronary syndromes. The relevance of NGAL to cardiovascular disease (CVD) remains primarily unknown. Elevated plasma NGAL levels were associated with atherosclerosis and were implicated as a predictor for cardiovascular mortality after cerebrovascular ischemia, possibly because of activation of blood leukocytes [18] [19] [20] . Although in recent reports has been shown that NGAL is present in atherosclerotic plaques and in human abdominal aortic aneurisms, raising the possibility that expression of NGAL can be induced in vascular cells during atherogenesis, the underlying mechanism for the induction of NGAL in vascular cells remains unknown [15, 21] . In further analysis the main source of NGAL was found to be neutrophils, probably recruited in the vascular wall by platelet activation [21] . NGAL is considered to have a protective effect on MMP-9 and enhancing its proteolytic activity, could be considered as an important factor indirectly contributing to the progression of aneurism as well as involved in the physiologic and pathologic remodeling of vessel walls. This view is further supported by the observation that similar neutrophil NGAL/MMP-9 overexpression can be found in atherosclerotic plaques, particularly those with intramural haemorrhagic debris and central necrosis [15, 22] . The above evidence supports the clinical observations that highcirculating leucocyte (particularly neutrophil) counts are independent predictors of recurrent ischaemic attacks. This may be explained by their presence in the necrotic core of unstable plaques and by their proteolytic activity towards atherosclerotic tissue and secondary mobilization of thromboembolic fragments [23] . The evidence derived from these experimental studies, showing the close link between neutrophils, their products and the natural history of atherosclerosis, and its complications, generated clinical studies that investigated the clinical utility of serum-NGAL measurements. In two recent studies it was found that serum levels of NGAL were significantly elevated in patients with angiographically confirmed CAD compared to those with normal arteries or controls [24, 25] . Our data agree with these reports since we found that levels of serum-NGAL are significantly higher in patients with all clinical syndromes of CAD than in healthy controls, reinforcing the utility of NGAL as biomarker of detection and the extent of CAD. The expression of NGAL from vascular cells during atherogenesis can also explain the differences between patients with SA and control subjects with no risk factors observed in our study. In addition to its induction in the vessels after mechanistic injury, previous studies suggest that NGAL is strongly upregulated in atherosclerotic lesions and also in the heart after ischemic injury [15] . It is possible that NGAL produced by vascular cells could also be secreted into the systemic circulation. Inflammation plays a critical role not only in development and progression of atherosclerosis but also in pathogenesis of the destabilization of atherosclerotic plaque that leads to ACS [1, 26] . Activation and degranulation of polymorphonuclear neutrophils and probably an underestimated critical components of an acute coronary inflammation event. Infiltrating macrophages and neutrophils participate in the transformation of stable coronary artery plaques to unstable lesions with a thin fibrous cap [27] . It has been repeatedly reported that thrombosed plaques were densely infiltrated by neutrophils and macrophages [28, 29] . Macrophages and neutrophils and some other types of leukocytes produce various proteolytic enzymes which facilitate the rupture of plaques by thinning and weakening their normally thick and firm cap [30, 31] . NGAL is one protein, that is, produced not only by the distressed kidney but also by activated neutrophils and by the vascular wall cells. Recent studies have shown that neutrophils are the main source of NGAL in blood [32, 33] . Increase in serum NGAL resulting from activation of neutrophils may reflect an acute systemic inflammatory response to events such as stroke, renal failure, or infection [18, [34] [35] [36] but are also linked with the presence of chronic inflammatory diseases such as atherosclerosis [18] whose acute clinical manifestations represent acute on chronic inflammation. Besides neutrophils, NGAL is also expressed by epithelial cells, renal tubular cells, and hepatocytes during inflammation or injury [37] [38] [39] . Our data agree with the above studies since we found a positive correlation between levels of serum-NGAL and systemic inflammation (expressed by the serum hs-CRP levels and neutrophil count), and also serum levels of NGAL were higher in patients with ACS than with SA. The higher levels of serum-NGAL observed in patients with ACS compared to SA could be explained by International Journal of Inflammation 7 the fact that neutrophil activation is present only in patients with acute coronary events (10, 11) . Also, our results, as far as patients with SA and AMI, are similar with the findings of a recent published study which showed that the plasma level of NGAL is higher in patients with AMI compared with the patients with stable CAD [40] . In clinical practice, levels of serum-NGAL have a high negative predictive value, 97.28% and 98.65% for patients with UA and ACS, respectively. So, serum-NGAL could be used in discriminating of patients with ACS or especially UA from whom with SA or without CAD, giving the possibility to exclude patients with symptoms similar to angina but not having true ACS. As far as the gradual increase of serum-NGAL, according to the seriousness of unstable coronary clinical syndrome, this could reflect the intensity of the inflammatory reaction, as it is expressed by the incremental increase of hs-CRP and neutrophil count and their combination with serum NGAL. Especially between serum-NGAL and hs-CRP, the correlation is linear and positive. In conclusion, our study shows that serum levels of NGAL increase in patients with CAD with every coronary clinical syndrome and reflect the inflammatory status in the same population. Having high negative predictive value could be used as a marker for the discrimination of SA or chest pain without CAD from those with ACS. Also in patients with ACS, serum levels of NGAL reflecting the inflammatory status could show the severity of coronary clinical syndrome (UA, NSTEMI, and STEMI). Acute coronary syndrome CAD: Coronary artery disease SA: Stable angina UA: Unstable angina AMI: Acute myocardial infarction NGAL: Neutrophil gelatinase associated lipocalin NSTEMI: Non-ST-elevation myocardial infarction STEMI: ST-elevation myocardial infarction PCI: Percutaneous coronary intervention.

Activity based protein profiling to detect serine hydrolase alterations in virus infected cells

Activity-based protein profiling (ABPP) is a newly emerging technique that uses active site-directed probes to monitor the functional status of enzymes. Serine hydrolases are one of the largest families of enzymes in mammals. More than 200 serine hydrolases have been identified, but little is known about their specific roles. Serine hydrolases are involved in a variety of physiological functions, including digestion, immune response, blood coagulation, and reproduction. ABPP has been used recently to investigate host–virus interactions and to understand the molecular pathogenesis of virus infections. Monitoring the altered serine hydrolases during viral infection gives insight into the catalytic activity of these enzymes that will help to identify novel targets for diagnostic and therapeutic application. This review presents the usefulness of ABPP in detecting and analyzing functional annotation of host cell serine hydrolases as a result of host–virus interaction.

Most enzymes are tightly regulated post-translationally. Many enzymes are synthesized as zymogens, which are functionally inactive. Moreover, enzyme functions can be changed by alterations in pH and binding to inhibitors. Thus, methods that allow direct quantification of protein activities rather than simply protein abundance are required to delineate distinct protein functions in physiological and pathological events. Activity-based protein profiling (ABPP) is a chemoproteomic platform for monitoring active proteins or enzymes. ABPP utilizes chemical probes to interrogate the functional state of large numbers of enzymes in complex proteomes in vitro or in vivo biological systems. ABPP probes consist of two key elements: (1) a reactive group/warhead (e.g., small molecule inhibitors, substrate-based scaffolds, or protein-reactive molecules) for binding and covalently labeling the active sites of many members of a given enzyme class (or classes), and (2) a reporter tag for the detection, enrichment, and/or identification of labeled enzymes from proteomes. A variety of reporter tags are used in ABPP, such as fluorophores (e.g., rhodamine) for visualization, biotin for enrichment as well as "clickable" handles, such as azides and acetylenes for in vivo or in situ labeling of proteins. The linker region is a flexible chain of varying length and hydrophobicity that connects and acts as a spacer between the warhead and the reporter tag. Serine hydrolases represent one of the largest and most diverse classes of enzymes in higher eukaryotes, collectively composing about 3% of the predicted Drosophila proteome (Rubin et al., 2000) and about 1% of all predicted expressed human genes (Lander et al., 2001) . Serine hydrolases are involved in a variety of physiological and pathological processes including blood coagulation (Kalafatis et al., 1997) , T cell cytotoxicity (Smyth et al., 1996) , inflammation (Bonventre et al., 1997) , neural plasticity (Yoshida and Shiosaka, 1999) , neurotransmitter catabolism (Taylor, 1991; Cravatt et al., 1996) , peptide/protein processing (Steiner, 1998) , protein/lipid digestion (Lowe, 1997) , angiogenesis (Mignatti and Rifkin, 1996) , emphysema (Kato, 1999) , and cancer (DeClerck et al., 1997) . Serine hydrolases also perform crucial functions in bacteria and viruses, where they contribute to pathogen life cycle (Steuber and Hilgenfeld, 2010) , virulence (White et al., 2011) , and drug resistance (Damblon et al., 1996) . Most enzymes hydrolyze metabolites, peptides or post-translational ester and thioester modifications on proteins. Because of the biological importance of serine hydrolases, clinically approved drugs target members of this enzyme class to treat diseases such as obesity (Henness and Perry, 2006) , diabetes (Thornberry and Weber, 2007) , microbial infections (Kluge and Petter, 2010) , and Alzheimer's disease (Racchi et al., 2004) . Proteolytic cleavages of viral proteins by cellular or viral proteases are necessary for host cell attachment, invasion, and reproduction of viral progeny. Host serine proteases are essential for the influenza virus life cycle because the viral hemagglutinin is synthesized as a precursor which requires proteolytic maturation (Garten and Klenk, 2008) . Recently, the non-structural 3 protease (NS3 -is a chymotrypsin-like serine protease which requires a polypeptide cofactor NS2B for activation) has been shown to be responsible for cleavage of the viral polyprotein precursor and to play a pivotal role in the replication of flaviviruses (Falgout et al., 1991; Mukhopadhyay et al., 2005; Chappell et al., 2006) including Hepatitis C (HCV), West Nile virus, and Dengue virus. NS3 also facilitates viral pathogenicity by cleaving host proteins and down-regulating the innate immune response of the cell (Failla et al., 1994; Meylan et al., 2005) . In fact, site-directed mutagenesis www.frontiersin.org that focused on the NS3 cleavage sites in the polyprotein precursor abolishes viral infectivity (Chappell et al., 2006) . Cell culture models provided important clues about potential inhibition of several protease inhibitors against NS3 for Dengue virus and West Nile virus (Cregar-Hernandez et al., 2011; Steuer et al., 2011) . Clinical trials of NS3 serine protease inhibitors showed good success rates (Lee et al., 2012) as anti-HCV. Therefore, NS3 is one of the most promising targets for drug development against Flaviviridae infections (Kolykhalov et al., 2000; Chappell et al., 2008) . Other serine proteases involved in the pathogenesis and virus life cycles are being considered as targets for chemotherapy. The catalytic activity of the herpes simplex virus type 1 serine protease is essential for viral nucleocapsid formation and for viral replication (Gao et al., 1994) . A trypsin-like serine protease is involved in pseudorabies viral penetration of the basement membrane during mucosal invasion (Glorieux et al., 2011) . Serine protease inhibitors inhibit pseudorabies virus invasion in basal membranes. A vaccinia virus serine protease inhibitor prevents virus induced cell fusion (Law and Smith, 1992) . Serine protease inhibitor AEBSF and pAB significantly reduce influenza A virus replication in mouse models (Bahgat et al., 2011) . However, identification of active SHS and their functional characterization are necessary for better understanding the molecular pathogenesis and development of antiviral strategies. All serine hydrolases possess a common catalytic mechanism that involves activation of a conserved serine nucleophile for attack on a substrate ester/thioester/amide bond to form an acylenzyme intermediate, followed by water-catalyzed hydrolysis of this intermediate to liberate the product. The greatly enhanced nucleophilicity of the catalytic serine renders it susceptible to covalent modification by many types of electrophiles, including fluorophosphonates (FPs) and aryl phosphonates, sulfonyl fluorides, and carbamates (Alexander and Cravatt, 2005; Jessani et al., 2005; Okerberg et al., 2005) . FPs are highly reactive and provide broad coverage, with the capacity to react with nearly all essential serine hydrolases (Bachovchin et al., 2010) . Therefore, they are ideal reagents to use for ABPP of serine hydrolases (Liu et al., 1999; Patricelli et al., 2001) . However, certain serine proteases displayed restricted substrate selectivities that reduce their labeling with FPs. To address this limitation of FPs, selective inhibitors (e.g., carbamates, triazole ureas) have been introduced to probe the function of individual serine hydrolase in biological systems (Bachovchin et al., 2010; Adibekian et al., 2011) . Microarray technologies in the field of genomics (transcriptomics), and mass spectrometry and bioinformatics technologies in proteomics, have facilitated the specific and global analyses of genes and their expression, and this has accelerated understanding the molecular basis of disease. These technologies, coupled with two-dimensional gel electrophoresis, mass spectrometry enhanced with chromatographic separations such as MudPIT (Shaw et al., 2008) , or isotope coding-ICAT (Yan et al., 2004) , iTRAQ (Lu et al., 2012) , and SILAC (Coombs et al., 2010) , have provided valuable insight into the quantitative differences in protein abundance during virus infections. However, these methods lack the inherent ability to profile and distinguish proteins according to their actual biological activities or functional state, which has more important bearings on understanding the implications of these macromolecules in vivo (Barglow and Cravatt, 2007) . The lack of functional assessment of these other omic methods has prompted the development of alternative strategies such as ABPP, for the discovery and characterization of enzyme activities within highly complex biological samples. A typical target discovery experiment would comparatively analyze two or more proteomes by ABPP to identify enzymes with differing levels of activity (Figure 1) . The differentially expressed serine hydrolases in healthy and diseased samples can be hypothesized to regulate the host-virus interaction. The testing of such hypotheses, of course, requires further experimentation for validation (e.g., functional interference of the target enzyme). ABPP has been used to profile a number of enzyme classes including proteases, hydrolases, oxidoreductases, and isomerases in the process of host-virus interaction Schlieker et al., 2005; Wang et al., 2006; Gredmark et al., 2007; Jarosinski et al., 2007; Shah et al., 2010) . Profiling of hydrolases in Huh7 cells replicating HCV identified CES1 (carboxylesterase 1) as a differentially active enzyme which has an important role in HCV propagation (Blais et al., 2010) . We have examined the activity of serine hydrolases during reovirus, Influenza A, and Sindbis virus replication in cell culture in different cell lines. Differential serine hydrolase activities were induced by different viruses and alterations of serine hydrolases were dependent on the time course of viral infection. Several of these differentially active serine hydrolases represent possible virus-host interactions that could be targeted for development of antivirals. ABPP can also be used as a competitive screen to identify both reversible and irreversible enzyme inhibitors and also to confirm target inhibition because inhibitors have the ability to block probe labeling of enzymes (Kidd et al., 2001; Greenbaum et al., 2002; Leung et al., 2003; Adibekian et al., 2011) . Competitive ABPP has already led to the discovery of selective serine inhibitors (e.g., carbamates, trizole ureas) for several enzymes (e.g., peptidase, lipases), which have in turn been used to test the function of these proteins in living systems (Bachovchin et al., 2010; Adibekian et al., 2011 ). An alternative omic strategy would be to examine libraries of commercially available protease inhibitors for their ability to inhibit a virus' pathological process; this would potentially lead to development of novel therapeutic options. Quantification of differentially expressed active proteins after virus infection is essential for better analysis of results, particularly when examining enzymes. It is difficult to compare the altered serine hydrolases between healthy and infected samples by simply visualizing gel images or merely by mass spectrometry. To address this problem an advanced quantitative mass spectrometry-based method called ABPP-SILAC (stable isotope labeling with amino acid in cell culture) has been used to identify alterations in the levels of active enzyme targets (Everley et al., 2007) and in small molecule-binding proteins in cell lysates (Ong et al., 2009) . Comparative ABPP -SILAC can be used to quantify more accurately the intricate changes in host proteins caused by viral infection. Similarly, competitive ABPP-SILAC is valuable to identify inhibited enzymes during global screening of inhibitors. Many enzymes and metabolites display difficult physicochemical properties that complicate their analysis in biological samples, and many metabolic pathways that enzymes regulate in a disease-specific context are not understood. These challenges can be addressed by applying innovative metabolomics and ABPP approaches to mapping biochemical pathways that support disease. Using selective inhibitors developed through competitive ABPP efforts or RNA interference technology, the function of an enzyme of interest can be specifically blocked, and then the metabolites that the enzyme regulates can be profiled. In this manner, not only can the substrates and products of an enzyme in specific (patho)physiological contexts be examined, but also the metabolic networks that the enzyme regulates can be identified and annotated. Collectively, this platform will allow identification of novel biochemical roles of already characterized enzymes, or may allow the identification of metabolic roles of completely uncharacterized enzymes. Understanding the mechanisms by which viruses develop resistance is a vital component of the fight against viral diseases, and can lengthen the lifespan of existing antivirals. Potentially any antivirus molecule could be transformed into an activity-based or affinity-based probe, allowing isolation and characterization of enzymes that detoxify the antiviral drug. ABPP with live cell imaging may provide additional insight into understanding the pathogenesis due to viral infection (Furman et al., 2009) . Identification and functional characterization of serine hydrolases involved in pathogenesis and virulence of viruses would be a novel approach to uncover molecular processes at the basis of viral diseases. Natural products represent an important treasure box of biologically active molecules, from which many drug candidates have been developed (Newman and Cragg, 2007) . Since a large number of the proteome remains functionally uncharacterized and is therefore difficult to assemble into larger biochemical networks, competitive ABPP will inevitably accelerate the development of novel inhibitors from natural products. This mini review describes briefly a limited number of approaches involved in profiling serine hydrolases during viral infection and assigning catalytic functions to previously uncharacterized serine hydrolases. Visualization of the altered active serine hydrolase in situ during viral disease progression, trying to fully understand mechanisms of resistance and developing new antiviral therapeutics and viral diagnostics will make the ABPP application more worthwhile for the field of virology. www.frontiersin.org

Murphy's law—if anything can go wrong, it will: Problems in phage electron microscopy

The quality of bacteriophage electron microscopy appears to be on a downward course since the 1980s. This coincides with the introduction of digital electron microscopes and a general lowering of standards, possibly due to the disappearance of several world-class electron microscopists The most important problem seems to be poor contrast. Positive staining is frequently not recognized as an undesirable artifact. Phage parts, bacterial debris, and aberrant or damaged phage particles may be misdiagnosed as bacterial viruses. Digital electron microscopes often seem to be operated without magnification control because this is difficult and inconvenient. In summary, most phage electron microscopy problems may be attributed to human failure. Journals are a last-ditch defense and have a heavy responsibility in selecting competent reviewers and rejecting, or not, unsatisfactory articles.

Legend has it that Murphy's law was formulated in 1949 by a Captain Edward A. Murphy, then at Edwards Air Force Base in California. 15 There are several variants, all meant to express a perverse outcome. Murphy's law certainly applies to bacteriophage electron microscopy. This type of investigation is a multistep procedure that depends on expensive and complicated instruments, refined techniques, bacteriophages, and, not least, the investigator. Problems and errors beset every step. Artifacts in particular have received little attention and will be the focus of this article. Negative staining of viruses, arguably the technically simplest and most important single method in virology, was introduced in 1956. Hall and later Brenner and Horne used phosphotungstic acid for contrasting plant viruses. 8, 12 This technique was extended in 1959 to coliphage T2. 9 Viruses and their components stood out as white on a dark background with unprecedented clarity. Negative staining by phosphotungstates rapidly superseeded shadowing for virus visualization. Uranyl salts and ammonium molybdate were introduced later. The standards of viral electron microscopy were set in the 1970s. 11, 20, 21 Today, negative staining is done almost exclusively with uranyl acetate (UA) and phosphotungstate (PT) salts (Figs. 1 and 2) and has been applied to thousands of viruses. At present, at least 6,300 bacterial and archaeal viruses have been examined in the electron microscope after negative staining. 6 Transmission electron microsocpes (TEMs) fall into three categories. (1) Conventional or "manual," also called "analogue" microscopes using mechanical devices for stage drive and aperture alignment and analog potentiometers or variable resistors for electronic controls. Images are recorded on photographic film or plates. (2) "Digital" microscopes using far fewer controls due to a computer-based system with digital potentiometers, electronic stage drives, and a digital alignment via a centralized computer interface. Objects are generally visualized on a monitor screen and images are acquired via a "chargedcouple device" or CCD sensor, replacing film or plates as the recording media. (3) Hybrids or conventional microscopes equipped with a digital camera. Since about 1985, "manual" electron microscopes were gradually replaced by "digital" instruments. These instruments were marketed almost simultanously by three major competing companies that did so with meager or no instructions relating to contrast and image quality. An additional modification by microscope manufacturers was to change the focal length of the objective pole piece (zeta or Z angle) to increase fidelity at the cost of reduced contrast. All these factors could explain a wave of poor electron microscopical images 1 which, seemingly, has not abated. Unfortunately, this wave coincided with the death of several famous and highly skilled electron microscopists, such as E. Kellenberger in Switzerland, D.E. Bradley in Canada, and A.S. Tikhonenko in Russia, who could have stemmed the tide. To assess this problem, one of us (HWA) analyzed 155 publications with phage micrographs originating in 28 countries, published between 2007 and January 2012 in over 50 journals. Twothirds (109) were of poor quality, namely contrastless, unsharp, astigmatic, of small size and low magnification (below 150,000¾). Only 46 could be considered as good or acceptable. The adjectives "poor and good" are somewhat subjective and we apologize for this. The micrographs had been obtained with a wide selection of about 50 types or models of electron microscopes, most of which were "digital." The vast majority of articles presented several pictures. "Manual" microscopes were apparently disappearing fast. Clearly, poor electron microscopy is a global problem. Electron microscopes. The most frequent types of transmission electron microscopes (TEMs) used in phage research are produced by JEOL (Japanese Electron Optical Laboratories), Hitachi and FEI, a successor of Philips (Eindhoven, The Netherlands), followed distantly by Carl Zeiss, Germany. One also finds a few old AEM (England) and Tesla (Czechia) instruments. The prevalence of JEOL and Hitachi instruments may be attributed to their relatively low cost. The most popular types are the JEOL 1200 EX, the Hitachi H-1700, and three instruments of the Philips family (FEI CM 100, Technai Spirit or Morgagni). A few "manual" microscopes, for example the Philips EM 300, persist and seem to be, because of their high quality, the objects of a cult among inconditionals of electron microscopy. FEI-Tecnai electron microscopes appear to be the most highly evolved TEMs and have indeed produced excellent micrographs, whereas, for example, the JEOL 1200 is often associated with poor pictures. This impression is superficial. Indeed, good and poor pictures have been produced with JEOL, Hitachi and Philips-FEI instruments. In conclusion, electron microscopes are above all "operator-sensitive." Poor pictures must generally be attributed to poor maintenance and untrained users. There is no cure but know-how. The dirty picture. Quite often, specimens are not purified in any way and investigators examine crude lysates. This is almost certain to go wrong. Crude lysates contain huge amounts of impurities, notably proteins (Figs. S4 and S19). This results in contrastless, flat images and the presence of bacterial debris and even complete bacteria. The debris may mimic certain types of bacteriophages ( Table 2) . Images of nonpurified specimens should not be accepted for publication. Purification can be achieved by centrifugation in sucrose or CsCl density gradients; unfortunately, this technique is limited to a small number of specialized laboratories. Instead, we recommend to sediment phages in Eppendorf tubes followed by two washes in buffer or even tap water. Using fixed-angle rotors, phages can be sedimented in as little as 1 h at 25,000 g. It is not necessary to stain phages for long minutes or to fix them. Staining is almost instantaneous and fixation is merely a complication. While purification is mandatory for any real investigation, crude lysates can be examined for quick checks of phage presence and identity. It is also possible to examine phages extracted from a lysed agar surface. 23 For this, a drop of staining solution is deposited on the agar and the Petri dish is agitated gently. The drop is then touched with a grid and excess liquid is drawn off. The technically deficient picture. The root causes of unsharp, "fuzzy" images (Figs. S5 and S6) are poor microscope maintenance and misalignment of column components. Both focusing and astigmatism correction can go wrong and produce unsharp images. Much has been said in manuals of electron microscopy (e.g., in ref. 7) about focusing and astigmatism and this does not need to be repeated. Digital microscopes allow the investigator to use the FFT (Fast fourier transform) function as a measure of astigmatism (Fig. 3) . This function is available in FEI, Hitachi, and JEOL instruments and, for example, the widespread AMT, Gatan, and OlympusSis cameras. If fuzziness and astigmatism persist, the electron microscope should be checked by the installer. Test specimens for resolution checks and conventional astigmatism correction ("holey" grids) are commercially available and a must for the serious microscopist. Unsharp images are generated by: (1) Misalignment of the whole EM column and/or the objective aperture. (2) Object drift due to a specimen support film that breaks, does not adhere to the grid, or has not been stabilized with carbon, thus causing the the substrate to charge and float. (3) Discharges in the electron gun altering the focus. (4) Over-or under-focusing by the user (Fig. S7) . (5) Astigmatism, generally caused by a dirty object holder or objective lens diaphragm, resulting in fuzziness and fine parallel lines instead the normal grain of micrographs (Fig. S8) . Contrast. Poor contrast (Figs. S5 and S6) seems to be a pervasive, if not the main problem of "digital" electron microscopy (EM). Indeed, many "digital" micrographs are lamentably dark and poorly contrasted. This is not an intrinsic limitation of "digital" microscopes or cameras, but rather due to inappropriate parameters or complete misunderstanding of the dynamic signal range during image acquisition. In "manual" instruments, poor contrast can be overcome by apertures, highcontrast films, or elongating the focal distance of the objective lens, for example in the old Siemens Elmiskop I. Finally, contrast can be improved in the darkroom by means of fast developers, polycontrast papers and optical filters. Unfortunately, this is very difficult with "digital" electron microscopes where contrast must be adjusted before or after image acquisition. CCD digital technology offers a wide range of gray scale values. They can be selected to favor the mid-tonal range, thus reducing the extreme dark and white values. For example, a 12-bit image comprises 4,096 levels from black to white. If the software is set to adjust the predominant intensity to a middle tonality, all values will be adjusted accordingly. The mid-tones typically yield lowcontrast images. Contrast levels should therefore be set to exclude a given percentage of black and white outliers. After image acquisition, the histogram camera software allows for adjustments of gamma (a function of luminance), contrast, and brightness. 19 The observer is left to find his or her best combination by trial and error. Contrast may also be improved by third-party software (e.g., Adobe Photoshop or Image J; http// rsbweb.nih.gov/ij/). This seems often to be misunderstood or ignored. Poorly contrasted digital images should be a thing of the past. Farming out. The high cost of electron microscopes and the specialized knowledge needed to operate them generated a questionable, even damnable development: the "farming out" of investigations to central laboratories or institutions carrying out examinations for a fee. Anything might go wrong during this procedure. It might be acceptable if phage researchers themselves have access to the electron microscope used or the technician that carries our investigations, is properly qualified or backed by an experienced phage researcher. Unfortunately, this seems to be rarely the case and has resulted in situations when examinations are performed by incompetents without instructions or personal interest in the subject, or simply unused to magnifications above 200.000¾. The investigating laboratories sometimes charge outlandish fees for essentially worthless data. This should be resisted. We propose fees of $50 US for access to an electron microscope and $75 US for an investigation aided by a technician or a scientist. The journals. Murphy's law also rules the final step of publication. Indeed, a good micrograph can be ruined by a journal that reduces it to postage stamp size or darkens it beyond recognition. Part of this may be attributed to the now frequent procedure of outsourcing manuscript assembly to distant offices, e.g., in South East Asia. Only protests will help here and sometimes do. Positive staining. This is the most frequent artifact in virus electron microscopy. Uranyl salts cause both negative and positive staining, while phosphotungstates and molybdates cause negative staining only ( Table 1) . The principal incriminated stain is uranyl acetate (UA). In positive staining (Figs. S9 and S13), virus particles themselves are stained and then may appear black on a white background. This is due to the strong affinity of uranyl ions to dsDNA 13 and seen in any viruses with compact masses of DNA, e.g., phage heads and adeno-or herpes-viruses. It never occurs in filamentous or ssRNA phages. Positively stained phage heads are invariably shrunken by about 10-15% 5, 16 and show neither capsomers nor transverse edges. Phage tails are not stained and appear as shadows. Consequently, these viruses can rarely be identified by EM. Their dimensions are perfectly useless and should not be published. If positive staining is generally an undesirable artifact, it has however two important applications. It allows (A) the visualization of phage heads in sectioned bacteria and (B) facilitates the quantification of aquatic viruses. 4 The reasons for positive staining are unknown. Both negative and positive staining may occur in adjacent areas of the same EM grid (Figs. S10, S12 and S13). The addition of protein to UA solutions seems to alleviate the incidence of positive staining. Halo formation. UA positive staining is often accompanied a gray halo around the virus capsid that increases with the time of irradiation (Figs. S12 and S13). The halo has unsharp margins and resembles an envelope. Its origin and significance are unknown. UA crystals and precipitates. UA tends to crystallize on the grid, often starting with viruses as crystallization origin. Crystals come in many varieties. Some are small (Fig. S14) and feathery or appear as long black needles. Others are flat, large sheets (Figs. S15 and S16). One also observes, albeit rarely, membranaceous precipitates around phage particles. Viruses within crystals often appear as brilliant white structures of abnormally large size (Fig. S17) . Swollen proteins. As an acidic stain (pH 4 to 4.5), UA causes proteinic structures to swell. 5 Compared with PT, the walls of UA-stained empty phage capsids and tails appear as relatively thick and less sharply defined. Typically, contractile tails are 2 nm larger in UA than in PT. Tail length is not affected. On the other hand, UA acts as a fixative and preservative, so that UA-stained specimens can be kept over years. Phage heads are better preserved in UA than in PT. Artifacts caused by phosphotungstate. Surprisingly, these are few. Compared with UA, and depending on the phage, heads tend to be rounded and are sometimes broken and empty. From the medium. Unwashed preparations contain proteins, sugars, and cell debris (below). They yield dirty, poorly contrasted, even opaque images with little structural details (Fig. S4) . PEG (polyethylene glycol), which is frequently used to concentrate phages, produces similar effects. PEG will collect anything: DNA, proteins, cell debris, phages and phage debris. Fortunately, PEG is is easily removed by washing. In one recent case, four washes were needed to obtain a clean preparation. CsCl, used for phage purification, may persist after incomplete dialysis. CsCl forms flat crystals, but does not interfere with staining. Arborescent or flat NaCl crystals are frequent in lysates from halophilic bacteria which require NaCl for growth, e.g., Vibrio and Halobacterium. Again, bacteriophages may act as nuclei of crystallization (Fig. S18) . NaCl crystals may obscure phage particles, but do not alter the quality of staining. In lysates from bacteria prepared with meatbased media, e.g., of clostridia, one may observe occasional fibers of striated muscle and bundles of double-walled rings which resemble phage tails (Fig. S41) and may be be attributed to self-assembly under the action of phospholipase C. 3 From bacteria (Table 2 ). Bacteria contribute proteins, DNA, ribosomes, cytoplasm, capsular material, pili, flagella, and fragments of plasma membrane and cell wall. Proteins and cytoplasm may interfere with the staining process, but are not a source of error. However, this is the case with the other cellular constituents that one may find in a lysate. Their presence depends very much on the phage host. For example, slime and capsular material are present in huge amounts in lysates of Acinetobacter and Xanthomonas and cell wall debris abound in those of Pseudomonas. Slime and capsular material normally appear as rounded elements, but can be distorted by centrifugation and then superficially resemble filamentous viruses (Inoviridae) (Figs. S20 and S21) or even tailed phages (Fig. S27) . Similarly, pili and debris of flagella may be mistaken as filamentous phages and short debris of flagella, of 100 to 200 nm length, may be confused with contractile phage tails. Occasionally, a fragment of a pilus still attached to a piece of plasma membrane may suggest the presence of a tailed phage. These elements should not be much of a problem as they are easily identified at magnifications above 150,000¾. Chloroform, sometimes used for sterilization of lysates, is dangerous as it may produce a thick smear of particles thought to be lipopolysaccharide (LPS). This goo interferes with staining and can make observations impossible. The amount of smear probably depends on the bacterium and seems to be particularly high in lysates of Brucella spp and Cronobacter sakazakii (Fig. S22) . Cell wall and plasma membrane fragments (Figs. S23 and S26) may be misdiagnosed as enveloped pleomorphic viruses, novel viruses, or tailless phages. This error is easily explained. Upon bacteriophage lysis, bacteria fall into pieces, many of which are round and have indeed the size of plasmaviruses or cystoviruses (70-80 nm). The latter are characterized by an envelope surrounding an isometric capsid. When superposed over each other, some cellular elements exhibit what seems to be external and internal (Fig. S25) . Cell wall fragments carrying an adsorbed phage tail may be taken at low magnification for tailed phages (Fig. S26) . These membrane fragments are much more than an inconsequential nuisance because phage counts in seawater and freshwater are increasingly often based on fluorescent microscopy. 19 As it happens, bacterial DNA can associate with membrane fragments during lysis 18 and will stain with fluorescent dies. The fluorescent membrane fragments are approximately of the size of phage heads and appear as tiny green dots simulating phages. So far, investigations of phage prevalence in the environment seem to have neglected this potentially serious source of errors. 18 From Bacteriophages (Tables 2 and 3) Bacterial viruses produce a wide variety of abnormal particles ( Table 3) . Nearly all of them have been observed in tailed phages, although the filamentous inoviruses sometimes give rise to doublelength particles. Tailed phages are classified into three families according to tail structure: Myoviridae with contractile tails, Siphoviridae with long, noncontractile tails and Podoviridae with short tails. Nature and frequency of "freaks" or "monsters" vary with the complexity of phages. Some aberrations have a genetic basis, others reflect errors of assembly, and still others result from propagating phages and their hosts on the wrong substrate, e.g., a medium containing amino acid analogs. 10 The subject has been reviewed elsewhere in some detail. 2 Phage T4 and its relatives have a particular propensity to produce aberrant particles. Some sources of error (and Mr Murphy's and the electron microscopist's delight) are: (1) Abnormally long tails (Fig. S28 ). They are found in very numerous siphoviruses, are extremely rare in myoviruses, and have never been reported in podoviruses. Normal tails may appear short when the head partly covers the tail (Figs. S29 and S30). (2) Isolated contracted tails, which have been interpreted by inexperienced observers as complete tailed phages and even novel species of viruses. (4) Proheads, recognizable by their small size and wavy outline (Fig. S33) , may be mistaken as complete isometric viruses or phage debris. (5) Mottled heads and polyheads (Figs. S34 and S35); they seem products of faulty phage synthesis. (6) Freak particles with two tails, two tail sheaths (Figs. S36 and S37) or even two heads. The latter were observed in Lactococcus lactis phages of the c2 species (not shown). (7) Myoviruses mimicking as siphoviruses after loss of their tail sheath (Fig. S38) . (8) Head-size variants (Fig. S40) , notably in phages with prolate heads www.landesbioscience.com Bacteriophage producing particles with short or isometric capsids. (9) Broken tails, suggesting the presence of podoviruses or isometric phages instead of sipho-or myo-viruses. 22 (10) Virus-like particles (Figs. S41 and S43) and true, but deformed phages (Fig. S44) . The presence of abnormal or damaged particles indicates a contamination by other phages or the presence of lysogenic phages produced by the phage host. The observer has to decide whether there are different phage populations or aberrant particles, e.g., a spectrum of phage tails of different length. The latter indicates the presence of a malformation. The matter requires prolonged observation at magnifications above 150,000¾. Exact magnification depends first on the adjustments made on installing an electron microscope. It must be controlled later by means of test specimens (e.g., beef liver catalase crystals 13 or T4 phage tails) because magnification may change over time and at every repair. Latex crystals and diffraction grating replicas are for low magnification only and to be rejected. In "manual" TEMs, magnification is easily and rapidly adjusted in the darkroom by means of a photographical enlarger. In "digital" TEMs, magnification can be controlled, if necessary, by calculation of correction factors. This potentially tedious procedure seems to be very unpopular with today's phage electron microscopists. Indeed, magnification control is seldom mentioned in recent phage descriptions and one suspects that it is rarely done. As a result, particle dimensions from "digital" TEMs sometimes appear as products of fantasy. Lastly, electron microscopy depends heavily on the quality of observations and interpretations. 19 For example, a common error is to call every phage head with a hexagonal outline an icosahedron although, geometrically speaking, it could also be an octahedron or a dodecahedron ( Table 2 ). The past five years have generated scores of strange publications. Some investigators of soil phages saw novel viruses in every round or oval particle or isolated tail sheath, others interpreted particles with contracted and extended tails of the same Bacillus myovirus as members of different species, and still others confused negative and positive staining or myo-, sipho-and podo-viruses. This denotes a decline in the general knowledge of viruses and of bacteriophages in particular. Misdiagnosis is particularly frequent in the interpretation of natural samples (water, soil and feces) because these may contain almost anything: algal, plant and vertebrate viruses, muscle fibers, abiogenic material, and of course the omnipresent bacterial debris of any kind. Ultimately, the human factor is the root cause of most problems: failure to maintain, repair and align an electron microscope; failure to focus properly and to correct astigmatism; improper or no specimen preparation; failure to recognize artifacts and fake viruses; absence of magnification control; finally failure to consult the now very abundant literature on electron microscopy in general and that on phages in particular. This translates as "everything that can go wrong, will." Fortunately, there are reasons to be optimistic as any man-made problems can be corrected by humans. Supplemental materials may be found here: www.landesbioscience.com/journals/ bacteriophage/article/20693

Herbal Products: Benefits, Limits, and Applications in Chronic Liver Disease

Complementary and alternative medicine soughts and encompasses a wide range of approaches; its use begun in ancient China at the time of Xia dynasty and in India during the Vedic period, but thanks to its long-lasting curative effect, easy availability, natural way of healing, and poor side-effects it is gaining importance throughout the world in clinical practice. We conducted a review describing the effects and the limits of using herbal products in chronic liver disease, focusing our attention on those most known, such as quercetin or curcumin. We tried to describe their pharmacokinetics, biological properties, and their beneficial effects (as antioxidant role) in metabolic, alcoholic, and viral hepatitis (considering that oxidative stress is the common pathway of chronic liver diseases of different etiology). The main limit of applicability of CAM comes from the lacking of randomized, placebo-controlled clinical trials giving a real proof of efficacy of those products, so that anecdotal success and personal experience are frequently the driving force for acceptance of CAM in the population.

Complementary and alternative medicine (CAM) therapies sought and encompass a wide range of approaches, including two broad categories: exogenous chemicals such as herbal supplements, vitamins, or plant extract, and natural or selftherapies (NST) techniques including relaxation, meditation, prayer, hypnosis, biofeedback, or physical strengthening [1] . The use of herbal medicine began in ancient China at the time of Xia dynasty and in India during the Vedic period [2] . With the revolution of the natural sciences and evidence-based medicine, the divide between Western and Eastern medicines appeared to widen, with CAM reaching an increasing popularity in western countries through years (from 34% of the population in 1990 to 48% in 2004) [3] . The age-old system of herbal medicine is being revived by day-to-day practice for its long-lasting curative effect, easy availability, natural way of healing, and less side-effects, so that today herbal medicines are gaining importance and expanding throughout the world [4] . The widespread use of CAM is emphasized among people with chronic disease, since it promotes greater personal control over health decision, empowers people to manage their chronic condition, and helps to avoid dissatisfaction often associated with conventional health care [5] . CAM is believed to be safer and better than standard medical practice because they are "natural" or are based on a religious, philosophical or a strongly felt concept of "wellness" and health. Treatments with herbal medicine concentrate on reestablishing or reinforcing natural healing processes and wellness [6] . Despite increasing popularity, communication about the use of CAM between physicians and patients is limited: most physicians know little about CAM and patients avoid discussing CAM because they fear being received with indifference [7] . Moreover, physicians used to focus attention on potential toxicities, even though identification of toxicity from herbal preparations is often difficult, because patients generally selfmedicate with these and may withhold this information. Toxic hepatitis is the most common adverse reaction resulting from the use of CAM [8] , often associated with the concomitant consumption of hepatotoxic ingredients such as acetaminophen and nonsteroidal anti-inflammatory agents or with hepatotoxicity of herbal ingredients themselves [9] . Physicians and health care providers need to become familiar with these products and to recognize potential interaction between conventional drugs and herbals, considering their actual diffusion [10] . Botanical medicines have been used traditionally by herbalists and indigenous healers worldwide for the prevention and treatment of liver disease. Clinical research in this century has confirmed the efficacy of several plants in the treatment of liver disease, so the fact that the patients with chronic liver disease seek primary or adjunctive herbal treatment is not surprising. Particularly, silymarin (an extract of milk thistle) is the most popular product taken by subjects with liver disease and especially by those with hepatitis C virus infection [11] . Seeff et al. [12] found that 41% of outpatients with diagnosis of liver disease had used some form of CAM. Herbal products are often used to improve well-being and quality of life [13] and to ameliorate side effects in patients on antiviral treatment, as fatigue, irritability, and depression: lessening of these symptoms might permit a higher compliance and avoid the need to limit the dose and finally withdraw inter-feron. A systematic review about the use of CAM in chronic hepatitis C has been conducted by Coon and Ernst [14] , The authors cited fourteen randomized clinical trials considering the combined use of herbal products and interferon-alfa during antiviral treatment. Although difficulty in extrapolation and interpretation of results because of different methodological limits of the considered studies, the authors found that several herbal products and supplements (vitamin E, thymic extract, zinc, traditional Chinese medicine, Glycyrrhiza glabra, and oxymatrine) could exert potential virological and biochemical effects in the treatment of chronic hepatitis C infection, as a greater clearance of HCV-RNA and normalization of liver enzymes. As shown in various studies [15, 16] , the use of CAM could be predicted by social, cultural, and geographic factors: sex, age, higher education level, or marriage status of patients are associated with a different use of herbal products. The aim of this study is to describe the potential role, benefits, and limits of some of known widespread herbal products in chronic liver disease. We conducted an updated research on Pubmed and Medline in order to refer to more recent articles about this issue. Quercetin is one of the major flavonoids, which represent a class of naturally occurring polyphenolic compounds, ubiquitously present in photosynthesising cells. The intake of flavones and flavonols is determined as 23-24 mg/day and quercetin, the main flavonol present in our diet, represents 70% of this intake. Quercetin is found in fruits (apple) and vegetables, especially onions [17] . Various ways of supplementing quercetin are possible, including a pure supplement or a diet intervention using a food component with a high quercetin content. Supplement usually contains only the aglycon form of quercetin, whereas a food component normally comprises high amounts of various quercetin derivatives that might have a better biological availability than the aglycon itself. Another advantage of a dietary supplementation versus a "conventional" supplement might be a better compliance, especially in long-term use [18] . The absorption of quercetin is considerably enhanced by its conjugation with a sugar group. After their facilitated uptake by means of carrier-mediated transport, quercetin glycosides often become hydrolysed by intracellular βglucosidases. After absorption, quercetin becomes metabolized in various organs including the small intestine, colon, liver, and kidney. Metabolites formed in the small intestine and liver are mainly the result of phase II metabolism by biotransformation enzymes and, therefore, include the methylated, sulphated, and glucuronidated forms. Moreover, bacterial ring fission of the aglycon occurs in both small intestine and colon, resulting in the breakdown of the backbone structure of quercetin and the subsequent formation of smaller phenolics [19] . Quercetin appears to have many beneficial effects on human health, including cardiovascular protection, anticancer activity, antiulcer and antiallergy activity, cataract prevention, and anti-inflammatory effects. Quercetin has been shown to be an excellent in vitro antioxidant. Within the flavonoid family, it is the most potent scavenger of ROS (reactive oxygen species), including O 2 [20] . The pathogenesis and progression of ALD are associated with free radical injury and oxidative stress, which could be partially attenuated by antioxidants and free radical scavengers. Lipid metabolism disorder and oxidative stress play an important role on the development and progression of ALD, and mitochondria compartment is presumed to be the main source and susceptible target of intracellular ROS. The hypothesis that quercetin could prevent ethanolinduced oxidative damage in hepatocytes has been investigated. In animal studies, prophylaxis with quercetinameliorated ethanol-stimulated mitochondrial dysfunction manifested by decreased membrane potential and by induced permeability transition though suppressing glutathione depletion, enzymatic inactivation of manganese superoxide dismutase, and glutathione peroxidase, ROS overgeneration, and lipid peroxidation in mitochondria. Quercetin, thus, may protect rat, especially hepatic mitochondria, from chronic ethanol toxicity through its hypolipidemic effect and antioxidative role, highlighting a promising preventive strategy for ALD by naturally occurring phytochemicals [21, 22] . Evidence-Based Complementary and Alternative Medicine 3 Quercetin tends to downregulate the ethanol-induced expression of glutathionine peroxidase 4 (GPX4). Furthermore, it tends to reduce the expression of SOD2 induced by ethanol, to downregulate the expression of Gadd45b at the presence of ethanol, which could permit to explain DNA demethylation associated with the upregulation of gene expression in experimental ALD [23] . Another study evaluated the effect of quercetin on the parameters classically associated with alcohol liver injury, as lactate dehydrogenase (LDH), aspartate transaminase (AST), malondialdehyde (MDA), glutathione (GSH), superoxide dismutase (SOD), and catalase (CAT) in order to address the alterations of cell damage and antioxidant state after quercetin intervention; the ethanol-intoxicated (100 mM for 8 h) rat primary hepatocytes were simultaneously treated, pretreated (2 h) and posttreated (2 h) with quercetin. The toxic insult of ethanol on hepatocytes was challenged by quercetin and biochemical parameters almost returned to the level of control group when hepatocytes were treated with quercetin at the dose of 50muM for 2-4 h before ethanol exposure [24] . A recent study elucidates also a neuroprotective effect of quercetin in alcohol-induced neuropathy through modulation of membrane-bound inorganic phosphate enzyme and inhibition of release of oxidoinflammatory mediators, such as malondialdehyde (MDA), myeloperoxidase (MPO) MPO, and nitric oxide (NO) [25] . In conclusion, pretreatment with quercetin provided protection against ethanol-induced oxidative stress in hepatocytes and may be used as a new natural drug for the prevention and/or treatment of ALD. Antioxidants significantly reduced the oxidative stress induced by ethanol intoxication, increased membrane integrity, and also increased organ regeneration [26] . Nonalcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver disease in the United States. It represents the hepatic manifestation of metabolic derangements, known as metabolic syndrome, with insulin resistance playing the major role. Nonalcoholic fatty liver disease includes a variety of histological conditions (ranging from liver steatosis and steatohepatitis, to fibrosis and hepatocarcinoma), all characterized by an increased accumulation/deposition of fat within the liver and associated with alterations in hepatic and systemic inflammatory state. Hepatocytes of a primary cell culture that are exposed to high glucose, insulin, and linoleic acid concentration respond with lipid accumulation, oxidative stress up to cell death. Regarding the role of quercetin in NAFLD, it has been show that in mice fed with a Western diet chronic dietary intake of quercetin reduces liver fat accumulation and improves systemic parameters related to metabolic syndrome, probably mainly through decreasing oxidative stress and reducing expression of genes related to steatosis (as PPARα) [27] . Another study was aimed to examine the hypoglycemic and insulin-sensitizing capacity of onion peel extract (OPE, containing a high content of quercetin) in high fat diet/streptozotocin-induced diabetic rats and to elucidate the mechanism of its insulin-sensitizing effect. OPE might improve glucose response and insulin resistance associated with type-2 diabetes by alleviating metabolic dysregulation of free fatty acids, by suppressing oxidative stress, upregulating glucose uptake at peripheral tissues, and/or downregulating inflammatory gene expression in liver. Moreover, in most cases, OPE showed greater potency than pure quercetin equivalent. These findings provide a basis for the use of onion peel to improve insulin insensitivity in type-2 diabetes [28] . In hepatocytes from normal rats, the decrease in de novo fatty acid and TAG synthesis induced by quercetin represent a potential mechanism contributing to the reported hypotriacylglycerolemic effect of this agent [29] . The hepatic response to chronic noxious stimuli may lead to liver fibrosis and to preneoplastic cirrhotic liver. Fibrogenic cells activate in response to a variety of cytokines, growth factors, and inflammatory mediators. The involvement of members of the epidermal growth factor family in this process has been suggested. Amphiregulin is an epidermal growth factor receptor (EGFR) ligand, specifically induced upon liver injury. Recent study investigated the effects of quercetin on the amphiregulin/EGFR signal and on the activation of downstream pathways leading to cell growth: quercetin-ameliorated activation of survival pathways and downregulated the expression of genes related to inflammation and precancerous conditions. Suppression of amphiregulin/EGFR signals may contribute to this effect [30] . Phytochemicals exert antiviral activity and may play a potential therapeutic role in hepatitis C virus (HCV) infection: these aspects were investigated by several studies. Park et al. [31] investigated about the antiviral activity in HCV-infected patients of derivates of 7-O-arylmethylquercetins. Only five quercetin derivatives showed selective antiviral activity in HCV replicon cell-based assay. Recent study studied the quercetin as a potential nontoxic anti-HCV agent in reducing viral production by inhibiting both NS3 and heat shock proteins (that are essential for HCV replication). It was found that quercetin inhibit NS3 activity in a specific dose-dependent manner in vitro catalysis assay. Moreover, as analyzed in the subgenomic HCV RNA replicon system, quercetin seemed to exert adjunctive effect: to inhibit HCV RNA replication and production in the HCV infectious cell culture system (HCVcc), as analyzed by the focusforming unit reduction assay and HCV RNA real-time PCR. The inhibitory effect of quercetin was also obtained when using a model system in which NS3-engineered substrates were introduced in NS3-expressing cells, providing evidence that inhibition in vivo could be directed to NS3 and does not involve other HCV proteins. This work demonstrates that quercetin has a direct inhibitory effect on the HCV NS3 protease [32] . (9) showed anti-HBV activity in vitro. Anti-HBV activity was closely related to the parent structure of these compounds (agigenin > luteolin > quercetin) as well as to the number of glucoside (flavone monoglucoside > flavone diglucoside). The structure of these agent also influences their cytotoxicity (flavone > flavone monoglucoside > flavone diglucoside). In addition, the substitution of acyl group on glucoside may be important to keep their anti-HBV activities (galloyl > feruloyl > coumaroyl) [33] . Quercetin did not show activities against HBeAg secretion: this limit might be due to the absence of the saccharide group in their structures [34] . Curcumin is a low-molecular-weight polyphenol derived from the rhizomes of turmeric (curcuma longa). It represents a yellow pigment widely used as a coloring agent and spice in many foods. It has various beneficial pharmacological effects including antioxidant, anti-inflammatory, anticarcinogenic, hypocholesterolemic, antibacterial, antispasmodic, anticoagulant, and hepatoprotective activities [35] . Phase I clinical trials have shown that curcumin is safe even at high doses (12 g/day) in humans but exhibit poor bioavailability. Despite the promising biological effects of curcumin, low plasma and tissue levels of curcumin due to its poor oral bioavailability and absorption, rapid metabolism and systemic elimination in both rodents and humans [36] may be responsible for the unfavorable pharmacokinetic of this molecule. To improve the bioavailability of curcumin, numerous approaches have been undertaken. These approaches involve, first, the use of adjuvant like piperine that interferes with glucuronidation; second, the use of liposomal curcumin; third, curcumin nanoparticles; fourth, the use of curcumin phospholipid complex; fifth, the use of structural analogues of curcumin (e.g., EF-24) [35, 36] . Liver fi-brosis can be explained with an increased deposition of extracellular matrix (ECM). Chronic alcohol abuse is one of the main causes of liver fibrosis. Ingestion of polyunsaturated fatty acids (PUFAs) together with alcohol can aggravate the toxicity of alcohol. The degree of abnormal ECM degradation depends on the ratio of active matrix metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs). A recent work studied the influence of bis-desmethoxy curcumin analog (BDMC-A) on the expression of MMPs and TIMPs during alcohol and DeltaPUFA-induced liver toxicity. Administration of BDMC-A significantly decreased the levels of collagen and TIMPs and positively modulated the expression of MMPs. From this study, we can conclude that BDMC-A influences MMPs, TIMPs expression and that it acts as an efficient anti-fibrotic agent [37] . It has been demonstrated the potential protective effect of curcumin pretreatment against ethanol-induced hepatocytes oxidative damage, with emphasis on heme oxygenase-1 (HO-1) induction. Ethanol exposure resulted in a sustained malondialdehyde (MDA) elevation, glutathione (GSH) depletion, and evident release of cellular lactate dehydrogenase (LDH) and aspartate aminotransferase (AST), which was significantly ameliorated by curcumin pretreatment. In addition, dose-and time-dependent induction of HO-1 was involved in such hepatoprotective effects by curcumin. Curcumin exerts hepatoprotective properties against ethanol involving HO-1 induction, which provide new insights into the pharmacological targets of curcumin in the prevention of alcoholic liver disease [38] . To study the mechanism of curcumin-attenuated inflammation and liver pathology in early stage of alcoholic liver disease, female Sprague-Dawley rats were divided into four groups and treated with ethanol or curcumin via an intragastric tube for 4 weeks. A control group was treated with distilled water and an ethanol group was treated with ethanol (7.5 g/kg bw). Treatment groups were fed with ethanol supplemented with curcumin (400 or 1 200 mg/kg bw). The liver histopathology in ethanol group revealed mild-to-moderate steatosis and mild necroinflammation. Hepatic MDA, hepatocyte apoptosis, and NF-kappaB activation increased significantly in ethanol-treated group when compared with control. Curcumin treatments resulted in improving of liver pathology, decreasing the elevation of hepatic MDA, and inhibition of NF-kappaB activation. The 400 mg/kg bw of curcumin treatment also revealed a trend of decreased hepatocyte apoptosis. However, the results of SOD activity, PPARgamma protein expression showed no difference among the groups. In conclusion, curcumin improved liver histopathology in early stage of ethanol-induced liver injury by reduction of oxidative stress and inhibition of NF-kappaB activation [39] . In animal study, it has been demonstrated the capacity of curcumin to improve liver histology of NAFLD. The NASH model was induced by highfat diet combined with carbon tetrachloride. These rats were successively treated with curcumin and curcumin derivative. The results showed a remarkable reduction in serum ALT (U/L), AST (U/L) in rats treated with curcumin derivatives. The degrees of fibrosis were also significantly alleviated. The reduction of the gene transcriptions of TNF-alpha, NF-kappaB, and HMG-CoA was the mechanism proposed by wich curcumin exercise its beneficial effects in NASH. The results of this study indicate, moreover, that the watersoluble curcumin derivative displays superior bioavailability to the parent curcumin, which can effectively improve the lipid metabolism and delay the progression of hepatic fibrosis in rats with steatohepatitis [40] . Interestingly, it also demonstrates that the activating effect of LDL can be reversed by curcumin. The hypocholesterolemic action of curcumin was reported by numerous studies: curcumin seems to reduce serum cholesterol concentrations by increasing hepatic expression of LDL receptors, by blocking LDL oxidation, increasing bile acid secretion, and faecal excretion of cholesterol, repressing the expression of genes involved in cholesterol biosynthesis and protecting against liver injury and fibrogenesis [41, 42] . Curcumin induces apoptosis and blocks proliferation of hepatic stellate cells (HSCs), and, via activation of PPARγ, inhibits extracellular matrix formation, downregulates the expression of LDL receptors, induces SREBP-1c,and increases the fat-storing capacity of HSC, and it may thereby restore this "protective" functionality, proving of therapeutic usage in preventing liver steatosis and fibrosis [43] . An Italian study tested whether the administration of curcumin limits fibrogenic evolution in a murine model of nonalcoholic steatohepatitis. They demonstrated that curcumin decreased the intrahepatic gene expression of monocyte chemoattractant protein-1, CD11b, procollagen type I, and tissue inhibitor of metalloprotease (TIMP)-1, together with protein levels of alpha-smooth muscle actin, a marker of fibrogenic cells. In addition, curcumin reduced the generation of reactive oxygen species in cultured HSCs and inhibited the secretion of TIMP-1 both in basal conditions and after the induction of oxidative stress. This study so proposed that curcumin administration effectively limits the development and progression of fibrosis in mice with experimental steatohepatitis and reduces TIMP-1 secretion and oxidative stress in cultured stellate cells [44] . High consumption of dietary fructose is an important contributory factor in the development of hepatic steatosis in insulin or leptin resistance. The effects of curcumin on fructose-induced hypertriglyceridemia and liver steatosis and its preventive mechanisms in rats have been investigated. Curcumin reduced serum insulin and leptin levels in fructose-fed rats and it protects against fructose-induced hypertriglyceridemia and hepatic steatosis by inhibition of PTP1B (hepatic protein tyrosine phosphatase 1B) and subsequent improvement of insulin and leptin sensitivity in the liver of rats. This PTP1B inhibitory property may represent a promising role for curcumin to treat fructoseinduced hepatic steatosis induced by hepatic insulin and leptin resistance [45, 46] . Curcumin is known to exert antiviral activity against influenza virus, adenovirus, coxsackievirus, and the human immunodeficiency virus [47, 48] . However, it remains to be determined whether curcumin can inhibit the replication of hepatitis C virus (HCV). A study showed that curcumin decreases HCV gene expression via suppression of the Akt-SREBP-1 activation, not by NF-kappaB pathway. The combination of curcumin and IFN-alpha exerted profound inhibitory effects on HCV replication. These results indicate that curcumin can suppress HCV replication in vitro and may be potentially useful as novel anti-HCV reagents [49] . Hepatitis B virus (HBV) infects the liver and uses its cell host for gene expression and propagation. Therefore, since targeting host factors is essential for HBV gene expression, it could represent a potential anti-viral strategy. Curcumin treatment could complement the antiviral activity of the nucleotide/nucleoside analogues, which are considered as the gold standards for anti-HBV therapy. The combination of Lamivudine and curcumin treatments resulted in an enhanced suppression of HBV expression by up to 75%, as compared to nontreated cells. These results suggest that curcumin may work synergistically with the current anti-HBV nucleotide/nucleoside analogous, and that this combination may result in a better suppression of HBV. Moreover, curcumin inhibits HBV gene expression and replication, by downregulating PGC-1alpha, a starvationinduced protein that initiates the gluconeogenesis cascade and that has been shown to robustly coactivate HBV transcription [50, 51] . Silybum marianum, also known as milk thistle, is a member of Asteraceae family and is well recognized as a hepatoprotective herbal medicine. Silymarin is a lipophilic extract of the milk thistle seeds. It is composed of three isomers of flavonolignans (silybin, silydianin, and silychristin), and two flavonoids (taxifolin and quercetin). Silymarin has revealed poor absorption, rapid metabolism, and ultimately poor oral bioavailability. For optimum silymarin bioavailability, issues of solubility, permeability, metabolism, and excretion must be addressed. An array of methods have been described in recent years that can improve its bioavailability, including complexation with βcyclodextrins, solid dispersion method, formation of microparticles and nanoparticles, self-microemulsifying drug delivery systems, micelles, liposomes, and phytosomes [52] . Silymarin possesses various pharmacological activities, including hepatoprotective, antioxidant, anti-inflammatory, anticancer, and cardioprotective effects. Silybum marianum is the most well-researched plant in the treatment of liver diseases. Silymarin has been shown to have a variety of anti-inflammatory effects on liver, including mast cell stabilization, inhibition of neutrophil migration, and Kupffer cell inhibition [53] . Silymarin is commonly prescribed in cases of cirrhosis or viral hepatitis. Hepatocyte models were proposed as a platform for screening of herbal 6 Evidence-Based Complementary and Alternative Medicine component against ethanol hepatotoxicity. Nanosilibinin, for the first time, found to perform significant protection against ethanol-induced hepatotoxicity while silibinin in normal particles could not inhibit such toxicity. This protection of nanosilibinin might be related to its high bioavailability compared to normal insoluble silibinin and could act as an antioxidative and antisteatosis agent against ethanol-induced hepatotoxicity [54] . The affect of silymarin on the levels of serum ALT and GGT in ethanol-induced hepatotoxicity in albino rats has also been tested. Eighteen male albino rats aged 6-8 weeks, weighing 150-200 gm, were divided into 3 groups of 6 rats each: group A as control, group B as rats taking ethanol at a dose of 0.6 mL (0.5 gm)/100 gm/day, and group C taking ethanol and silymarin at a dose of 0.5 gm/100 gm/day and 20 mg/100 gm/day, respectively, for 8 weeks. Silymarin tends to normalize liver function test in alcoholic liver disease [55] . Acute ethanol administration caused prominent hepatic microvesicular steatosis with mild necrosis and an elevation of serum ALT activity, induced a significant decrease in hepatic glutathione in conjunction with enhanced lipid peroxidation (oxidative stress) and increased hepatic TNF (necrosis factor-alpha) production. Supplementation with a standardized silymarin with its both antioxidant and antiinflammation properties decreases TNF production and attenuated these adverse changes induced by acute ethanol administration. In view of its nontoxic nature, it may be developed as an effective therapeutic agent for alcoholinduced liver disease by its antioxidative stress and antiinflammatory features [56] . Silymarin showed a significant hypocholesterolemic effect compared to the diet model with high fat-diet (HFD). Moreover, silymarin significantly reduces TG levels compared to HFD group. The elevation of transaminases usually reflects necrosis of hepatocytes: with silymarin, ALT levels (a specific index of hepatic necrosis) were particularly reduced [57] . Assuming that oxidative stress leads to chronic liver damage, Loguercio et al. conducted a study about the antioxidant activity of silybin conjugated with vitamin E and phospholipids. Eighty-five patients were divided into 2 groups: those affected by nonalcoholic fatty liver disease (group A) and those with HCV-related chronic hepatitis associated with nonalcoholic fatty liver disease (group B), nonresponders to antiviral treatment. The treatment consisted of silybin/vitamin E/phospholipids. After treatment, group A showed a significant reduction in ultrasonographic scores for liver steatosis. Liver enzyme levels, hyperinsulinemia, and indexes of liver fibrosis showed an improvement in treated individuals. A significant correlation among indexes of fibrosis, body mass index, insulinemia, plasma levels of transforming growth factorbeta, tumor necrosis factor-alpha, degree of steatosis, and gamma-glutamyl transpeptidase was observed. Our data suggest that silybin conjugated with vitamin E and phospholipids could be used as a complementary approach in the treatment of patients with chronic liver damage [58] . Silymarin showed antiviral effects against hepatitis C virus cell culture (HCVcc) infection that included inhibition of virus entry, RNA and protein expression, and infectious virus production. Silymarin did not block HCVcc binding to cells but inhibited the entry of several viral pseudoparticles (pp), and fusion of HCVpp with liposomes. Silymarin also blocked cell-to-cell spread of virus [59] . Pegylated interferon (PEG-IFN) plus Ribavirin therapy is the current treatment for the patient with chronic hepatitis C. The main goal of therapy is to achieve a sustained virological response (SVR is defined as undetectable HCV-RNA in peripheral blood determined with the most sensitive polymerase chain reaction technique 24 weeks after the end of treatment). This goal is practically equivalent with eradication of HCV infection and cure of the underlying HCV-induced liver disease. This therapy is effective only in half of patients, because of important side-effects, resistance, and high cost related to therapy. Silymarin inhibits both HCV RNA (in a dose-dependent manner) and HCV core expression thanks to its direct effect against HCV 3a core or activation of JAK/STAT pathways, resulting in inhibition of HCV core gene, by phosphorylation of Stat1 on tyrosine and serine [60] . Silymarin but not silibinin inhibited genotype 2a NS5B RNA-dependent RNA polymerase (RdRp) activity at concentrations 5 to 10 times higher than required for anti-HCVcc effects. Furthermore, silymarin had inefficient activity on the genotype 1b. Although inhibition of in vitro NS5B polymerase activity is demonstrable, the mechanisms of silymarin's antiviral action appear to include blocking of virus entry and transmission by targeting the host cell [60, 61] . In another study, patients with chronic hepatitis C performing silymarin (650 mg/day) for 6 months improved serum HCV-RNA titer, serum aminotransferases (ALT, AST), hepatic fibrosis, and patient's quality of life. After the treatment, nine patients were found with negative HCV-RNA and statistically significant improvement in results of liver fibrosis markers was found only in fibrosis group. So, for its antioxidant and anti-inflammatory actions, sylimarin could result useful in reducing hepatic inflammation in chronic liver disease, including HCV-related damage. It has been hypothesized that decreased hepatic inflammation-due to both direct and indirect effects of silymarin in decreasing viral replication has the potential to induce long-term benefit to the infected liver [62] . Since oxidative stress may play a pathogenetic role in chronic hepatitis C, and sustained virological response to antiviral therapy is limited in HCV1 genotype infection, a double-blind study was performed in HCV1 patients treated with pegylated interferon + ribavirin in order to assess the Evidence-Based Complementary and Alternative Medicine 7 efficacy of supplementation with the antioxidant flavonoid silymarin. In the silymarin group, a more rapid decrease in the malondialdehyde level as well as a marked decrease in superoxide dismutase and an increase in myeloperoxidase activity after twelve months of treatment were found. In particular, alanine aminotransferase normalized in 6/16 (versus control 9/16) cases and sustained virological response occurred in 3/16 (versus 7/16) patients [63] . Few recent data discuss about the role of silymarin in the hepatitis B. We only reported silymarin beneficial effects in early stages of liver pathogenesis, in preventing and delaying liver carcinogenesis. This drug should be considered as a potential chemopreventive agent for HBV-related hepatocarcinogenesis [64] . Betaine is a naturally occurring dietary compound that is also synthesized in vivo from choline. In vivo, betaine acts as a methyl donor for the conversion of homocysteine to methionine and its also functions as an osmolyte. Chronic ethanol exposure has been shown both to decrease hepatic concentrations of Sadenosylmethionine (SAM) and plasma concentrations of folate in animal and human studies and to increase plasma concentrations of homocysteine and hepatic levels of Sadenosylhomocysteine (SAH) [65] . In the liver, betaine can transfer its methyl group to homocysteine in order to form methionine. This can result in decreased concentrations of homocysteine and increased concentrations of methionine in the liver, resulting in decreased hepatic concentrations of SAH, whereas the latter can increase hepatic SAM concentrations, which leads to an increased SAM : SAH. An elevated SAM : SAH can trigger a cascade of events leading to formation of proper VLDL, to export of triacylglycerol and attenuation of fatty liver. Increased hepatic concentrations of SAM can activate cystathionine-synthase and lead to upregulation of the transsulfuration pathway, to increase synthesis of glutathione and attenuate oxidative stress. Thus, betaine can ameliorate ALD by attenuating fatty liver, inflammation, and fibrosis [66] . The role of mitochondrial dysfunction in the pathogenesis of alcoholic liver disease has been long documented by multiple laboratories. The dietary supplementation with betaine protects against ethanol-induced loss in oxidative phosphorylation system proteins. Even if the exact mechanism for this protection at the organelle level is not known, betaine shows to preserve the function of the electron transport chain, to maintain the integrity of the liver, and to protect against the development of alcoholic liver injury by preventing NOS 2 induction and NO generation [67] . Moreover, specific changes are associated with the normalization of hepatic SAM : SAH ratio and maintenance of methylation potential in response to betaine supplementation during chronic ethanol ingestion. Chronic alcohol administration increases gut-derived endotoxin in the portal circulation, activating Kupffer cells to produce several proinflammatory cytokines such as tumor necrosis factor-α (TNF-α) and interleukin (IL)-1. Ethanol administration can also lead to the synthesis of Toll-like receptor 4 (TLR4) protein and its gene expression in Kupffer cells, indicating that TLR4 may play a major role in the development of alcohol-induced liver injury. The intragastric ethanol-fed rat model, which reproduces the pathological features of early alcohol-induced liver injury, was used to observe the changes of TLR4 expression and the effect of betaine in alcohol-induced liver injury animal models. It has been suggested that betaine can prevent alcoholinduced liver injury effectively and improve liver function. The hepatoprotective mechanism of betaine is probably related to the inhibition of endotoxin/TLR4 signaling pathways. In rats with alcohol-induced liver injury, betaine feeding can decrease the levels of serum ALT, AST, endotoxin, TNF-α, IFN-γ, and IL-18, reduce the expressions of TLR4, and improve the degree of hepatic steatosis and inflammation in liver tissues. In summary, the results of this study show that the expression of TLR4 increased significantly in ethanol-fed rats. Betaine administration can inhibit TLR4 expression, which may be the mechanism of protection from alcoholic liver injury exerted by betaine [68] . The role of betaine in the treatment of NASH has been evaluated in human studies. Oral administration of betaine glucuronate in NASH patients for 8 weeks reduced both hepatic steatosis by 25% and hepatomegaly by 8%, and it significantly attenuated serum concentrations of AST, ALT, and glutamyl transferase. Similarly, a marked improvement in the degree of steatosis, necroinflammatory grade, and stage of fibrosis was obtained after treatment with betaine [69] . Nonalcoholic fatty liver disease (NAFLD) is a common liver disease, associated with insulin resistance. Betaine treatment would prevent or treat NAFLD in mice. Betaine reduces fasting glucose, insulin, triglycerides, and hepatic fat in mice submitted to a moderate high-fat diet (MHF). Betaine significantly improved insulin resistance and hepatic steatosis. Betaine treatment reversed the inhibition of hepatic insulin signaling in mHF and in insulin-resistant HepG2 cells, including normalization of insulin receptor substrate 1 (IRS1) phosphorylation and of signaling pathways for gluconeogenesis and glycogen synthesis. We can conclude that betaine treatment prevents and treats fatty liver in a moderate high-dietary-fat model of NAFL D in mice [70] . Moreover, betaine supplementation alleviated hepatic pathological changes, which were concomitant with attenuated insulin resistance (as shown by improved homeostasis model assessment of basal insulin resistance values and glucose tolerance test) and corrected abnormal adipokine productions (adiponectin, resistin, and leptin). Specifically, 8 Evidence-Based Complementary and Alternative Medicine betaine supplementation enhanced insulin sensitivity in adipose tissue as shown by improved extracellular signalregulated kinases 1/2 and protein kinase B activations. In adipocytes, freshly isolated from mice fed a high-fat diet, pretreatment of betaine enhanced the insulin signaling pathway and improved adipokine productions. Further investigation using whole liver tissues revealed that betaine supplementation alleviated high-fat diet-induced endoplasmic reticulum stress response in adipose tissue as shown by attenuated glucose-regulated protein 78/C/EBP homologous protein (CHOP) protein abundance and c-Jun NH2-terminal kinase activation [71] . Song et al. showed that betaine significantly attenuated hepatic steatosis induced by high-sucrose diet (animal model), and this change was associated with increased activation of hepatic AMP-activated protein kinase (AMPK) and attenuated lipogenic capability (enzyme activities and gene expression) in the liver [72] . The cause of failure of antiviral treatment with standard therapy, that is, pegylated interferon alpha (pegIFNalpha) combined with ribavirin, in half patients, is unknown, but viral interference with IFNalpha signal transduction through the Jak-STAT pathway could be considered. The expression of HCV proteins leads to an impairment of Jak-STAT signaling because of the inhibition of STAT1 methylation. Unmethylated STAT1 is less active since it can be bound and inactivated by its inhibitor, protein inhibitor of activated STAT1 (PIAS1). The treatment of cells with S-adenosyl-L-methionine (AdoMet) and betaine could restore STAT1 methylation and improve IFN alpha signaling. Furthermore, the antiviral effect of IFNalpha in cell culture could be significantly enhanced by the addition of AdoMet and betaine. S-adenosyl-L-methionine (SAMe) and betaine potentiate IFNα signaling in cultured cells expressing hepatitis C virus (HCV) proteins and enhance the inhibitory effect of IFNα on HCV replicons. SAMe and betaine were found to be safe when used with pegIFNα/ribavirin [73] . In conclusion, the addition of these drugs to antiviral standard therapy for patients with chronic hepatitis C could overcome the problem of drug resistance [74] . Homocysteine, a sulfuric amino acid involved in methionine metabolism, belongs to the group of intracellular thiols. Hyperhomocysteinemia is frequent in the Caucasian and its role in vascular pathology has been clearly established. In hepatology, experimental data in transgenic mice deficient in homocysteine metabolism enzymes have shown the presence of severe liver steatosis with occasional steatohepatitis. In chronic hepatitis C, preliminary data have shown that hyperhomocysteinemia is an independent risk factor for steatosis or even fibrosis. The physiopathological mechanism has now begun to be better understood. On one hand, there is a strong correlation between homocysteine and insulin resistance whatever its etiology. On the other hand, homocysteine has a direct effect on the liver, resulting in overexpression of SREBP-1 and favoring steatosis. It stimulates proinflammatory cytokine secretion such as NF kappa B increasing the risk of NASH. Finally, homocysteine could increase the risk of fibrosis by stimulating TIMP 1. Moreover, hepatitis C virus induces hypomethylation of STAT 1 and could decrease the antiviral activity of interferon. Results from in vitro studies have shown that the normalization of STAT 1 methylation by bringing betaine and S Adenosyl Methionine (which belongs to homocysteine cycle) restores the antiviral activity of interferon. Finally, treatment of hyperhomocysteinemia could have favorable consequences in steatohepatites and HCV infection [75, 76] . Only few data exist about the role of betaine in histological or clinical effects of hepatitis-B-virusinfected patients. Glycyrrhiza glabra (licorice root), a perennial herb cultivated in temperate and subtropical regions of the world and native to Mediterranean region as well as to central and South-Western Asia, belongs to the Leguminosae family, genus Glycyrrhiza [77] . The aqueous extract of this plant contains Glycyrrizin (GL), a conjugate of two molecules of glucuronic acid and one of 18β-glycyrrhetic acid (GA) [78] and other substances as flavonoids, hydroxycoumarins and beta-sitosterols [79] . Stronger neominophagen C (SNMC) is a product used in Japan for the treatment of acute and chronic hepatitis. This solution is administered intravenously, 80-200 mg/day, for variable periods of time, and contains 0.2% glycyrrhizin, 0.1% cysteine, and 2% glycine in physiological solution. In the United States, glycyrrhizin is available in a multiplicity of nonstandardized oral formulations found over the country [80, 81] . GL injected i.v., is partially metabolized to 3-monoglucuronyl-glycyrrhetinic acid (3MGA) in the liver by lysosomal β-D-glucuronidase, and GL and 3MGA could be excreted with bile. The biliary-excreted GL and 3MGA are hydrolyzed by intestinal bacteria into GA, which is reabsorbed into the bloodstream. Orally administered GL is enzymatically hydrolyzed to GA by intestinal bacterial flora before absorption into the bloodstream. The circulated GA is further metabolized to 3MGA in the liver by UDPglucuronyl transferase and then excreted with bile into the intestine [82] . The pharmacokinetic characteristics of i.v. administration of GL in patients with liver disease was studied in a Japanese and European report: SNMC has linear pharmacokinetics up to 200 mg, and steady state in achieved after two weeks of 200 mg doses administrated six times per week [83] . Potential interactions may occur with drugs metabolized by CYP450 3A4, although those have not been reported to date [80, 81] . Decreases of potassium, sodium retention, worsening of ascites, and hypertension are possible adverse effects due to the 11-hydroxy-steroid dehydrogenase inhibitory activity of GL and GA [84] . However, published data show no increased rate of these side-effects during treatment although documentation of toxicity is poor in most reports [77] . The use of GL in acute and chronic hepatitis is due to its hepatoprotective, immunomodulatory, and antiinflammatory effect. It reduced ischemia/reperfusion (I/R-) induced liver injury [85] and inhibited hight-mobility group box 1 (HMGB1), an inflammatory cytokine that acted in inflammation and organ damage in hepatic I/R-injury [85, 86] . Many studies shown that GL attenuated inflammatory responses due to decreased activation of nuclear factor kB (NF-kB) and mitogen-activated protein kinase (MAPK) pathways [87] , moreover, inhibited the production of LPSinduced nitric oxide (NO) and tumor necrosis factor-α (TNF-α), prostaglandin E2, intracellular reactive oxygen species (ROS), proinflammatory interleukins as IL-4, IL-5, IL-6, IL-18, IL-1β [88] [89] [90] [91] [92] , and increased the production of anti-infiammatory interleukins as IL-12 and IL-10 [92] . In animal studies, GL inhibits CD4+ T-cell and tumor necrosis factor-(TNF-) mediated cytotoxicity [93] , activated NK cells and extrathymic T lymphocyte differentiation [94, 95] , and promoted maturation of dendritic cells [92] . GL inhibited serum AST and ALT levels and histologically inhibited the infiltration of inflammatory cells and the spreading of degenerative areas of hepatocytes in an animal model of concanavalin A-induced liver injury [96] . Many studies showed GL have antiviral activity (reviewed in [97] ). A mechanism proposed for explain this propriety is the membrane stabilizing effect, as demonstrated in rat hepatocytes incubated with antibody raised against rat liver cell membranes: rat hepatocytes released AST after incubation with antiliver cell antibody in the presence of complement, and the endogenous phospholipase A2 activity was increased, but glycyrrhizin suppressed phospholipase A2 activity and reduced transaminase level [98] . A more recent study confirms this propriety in HIV and Influenza A virus [99] . Despite a precedent review [100] evidenced that SNMC acts as an antiinflammatory or cytoprotective drug but does not have antiviral properties, a recent in vitro study found that GL inhibit HCV full-length viral particles and HCV core gene expression or function in a dose-dependent manner and had synergistic effect with interferon [101] and a European randomized trial showed biochemical effects of 26-week treatment with SNMC in patients with chronic hepatitis C [102] . Glycyrrhizin-modified glycosylation and blocked sialylation of hepatitis B surface antigen (HBsAg) [103] . An in vitro study, measuring the release of surface protein (HBsAg) and HBV-DNA in transfected HepG2 2.2.15 cells, showed that this compound had a moderate ability in reducing viral production [104] . Long-term clinical trials in Japan and The Netherlands demonstrate that interferon nonresponder patients with chronic hepatitis C and fibrosis stage 3 or 4 have a reduced incidence rate of HCC after glycyrrhizin therapy normalizes ALT levels [105, 106] . Other well-diagnosed studies are needed to better define the role of GL in HBV and HCVrelated liver disease. Steatosis. 18 beta-glycyrrhetinic acid (18α-GL) can suppress the activation of hepatic stellate cells (HCSs) and induce their apoptosis by blocking the translocation of NF-kappaB into the nucleus, furthermore, it promoted the proliferation of hepatocytes in rats with CCl4-induced liver fibrosis [107] . GA inhibited type I collagen synthesis and progression of liver fibrosis probably through the suppression of collagen gene (COL1A2) promoter [108] . In transgenic mice expressing the HCV polyprotein fed an excess iron diet, SNMC prevented hepatic steatosis: this product attenuated ultrastructural alterations of mitochondria of the liver, activated mitochondrial β-oxidation with increased expression of carnitine palmitoyl transferase I and decreased the production of reactive oxygen species. Wu et al. found that 18α-GL, the biologically active metabolite of GL, prevented FFA-induced lipid accumulation and cell apoptosis in in vitro HepG2 NAFLD models and also prevented high-fat-diet-induced hepatic lipotoxicity and liver injury in vivo rat NAFLD models. GA stabilized lysosomal membranes, inhibited cathepsin B expression and enzyme activity, inhibited mitochondrial cytochrome c release, and reduced FFA-induced oxidative stress [109] . A recent review revealed that triterpenoids, which are also found in Glycyzirra glabra extract, had antitumor activities. Triterpenoids could induce apoptosis in various cancer cells by activating various proapoptotic signaling cascades. The molecular mechanisms involved include inhibition of various oncogenic and antiapoptotic signaling pathways and suppression or nuclear translocation of transcription factors including NF-κB [110] . In a human hepatoma cell line, the expression of junB mRNA, a tumor suppressor gene, and JUNG protein is highly increased by GL treatment [111] . Zhao et al. studied the β-Cyclodextrin/glycyrrhizic acid functionalized quantum dots (β-CD/GA-functionalized QDs [112] , and found that this drug has proapoptotic effects in hepatocarcinoma cells. β-CD/GA-functionalized QDs triggered G0/G1 phase arrest and induced apoptosis through an reactive oxygen species mediated mitochondrial dysfunction pathway. 7.1. Definition, Pharmacokinetics, and Biological Aspects. The genus Phyllantus (Euphorbiaceae) consist of about 6500 species in 300 genera, of which 200 are American, 100 African, 70 from Madagascar, and the remaining are Asian and Australian [113] . Many species are used in traditional medicine mainly in India and China to treat several diseases [114] and a morphological analysis of samples of Phyllanthus used in raw drug trade in southern India shown that 76% of the market samples contained P. amarus as the predominant species (>95%) and other five different species, namely, P. debilis, P. fraternus, P. urinaria, P. maderaspatensis, and P. kozhikodianus, were found in the remaining 24% of the shops [115] . P. amarus has been widely studied because it is the most commonly used in the Indian Ayurvedic medicine in the treatment of gastrointestinal and genitourinary diseases. Its main active components are ligans, as phyllanthin and hypophyllanthin, and flavonoids, alkaloids, hydrolysable tannins, polyphenols, triterperens, sterols, and volatile oil [113] . Many animal studies evidenced the hepatoprotective activity of P. amarus. The extract enhanced liver and serum alanine transaminase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), acid phosphatase (ACP), glutathione-S transferase (GST). Furthermore lipid peroxidation level was significantly reduced in ethanol and CCl4-induced liver disease animal models [116] [117] [118] [119] . The administration of P. amarus extract significantly decreased the levels of collagen and tissue inhibitors of matrix metalloproteinases (TIMPs) and positively modulated the expression of matrix metalloproteinases (MMPs) in rats with alcohol and thermally oxidized polyunsaturated fatty acid-(PUFA-) induced hepatic fibrosis [120] . In mice with aflatoxin B1-induced liver damage, P. amarus extract lowered down the content of thiobarbituric acid reactive substances (TBARSs) and enhanced the reduced glutathione level and the activities of antioxidant enzymes, glutathione peroxidase (GPx), glutathione-S-transferase (GST), superoxide dismutase (SOD), and catalase (CAT) [121] . In a recent animal study, it was found a synergistic effect between silymarin and P. amarus, especially with ethanoic extract of P. amarus, due to its higher concentration of phyllanthin in comparison to aqueous extract against CClinduced nepatotoxicity [122] . The hepatoprotective action was studied also in other Phyllanthus species, as P. simplex [123] , P. atropurpureus [124] , P. acidus [125, 126] , P. fraternus [127] , P. emblica [128] , P. urinaria, and P. maderaspatensis [129, 130] , and significant antioxidant and anti-inflammatory activities were found especially in P. simplex extracts because of its high phenolic content [123] . Finally, the antioxidant activity was compared between P. virgatus and P. amarus and was found that the first had higher cytotoxicity, higher free radical scavenging activity, and more inhibition of peroxidation capacity [131] . Preclinical studies have shown that Emblica officinalis (P. emblica) protect against ethanol-induced hepatotoxicity (reviewed in [132] ). Animal studies showed that the fruit extract improved plasma enzymes level, reduced lipid peroxidation, and restored the enzymatic and nonenzymatic antioxidants level in alcohol-induced liver disease, this action is probably due to tannoid, flavonoid and NO scavenging compounds present in the extract [133] [134] [135] . Phyllanthin restored the antioxidant capability of rat hepatocytes including level of total glutathione, and activities of superoxide dismutase (SOD) and glutathione reductase (GR) which were reduced by ethanol [136] and P. amarus-modified alcohol and thermally oxidized PUFA-induced fibrosis in rats because decreased the levels of collagen and TIMPs and positively modulated the expression of MMPs [120] . 7.3. Phyllanthus in HCV and HBV Infection. P. amarus extract was used, in 1988, in a preliminary study involving 37 patients with chronic hepatitis B. 22 of 37 treated patients had cleared the virus 2-3 weeks after the end of the treatment period and only one of 23 placebo treated became HBsAg negative [137] . The mechanism of action appears to be related to the suppressive effect of Phyllanthus extract on HBsAg secretion and HBsAg mRNA expression [138] and the inhibition of hepatitis B virus polymerase activity [139] . A recent study isolated a polyphenolic compound, 1,2,4,6tetra-O-galloyl-β-D-glucose (1246TGG), from P. emblica, and found that treatment with 1246TGG (6.25 μg/mL, 3.13 μg/mL), reduced both HBsAg and HBeAg levels in culture supernatant of HepG2.2.15 cells [140] . The role of Phyllanthus spp. in the treatment of chronic hepatitis B was studied in several reports that were evaluated in a recent Cochrane review. The authors included a total of 16 randomized trials but only one compared Phyllanthus with placebo and found no significant difference in HBeAg seroconversion after the end of treatment or followup. Fifteen trials compared Phyllanthus plus an antiviral drug like interferon alpha, lamivudine, adefovir dipivoxil, thymosin, vidarabine, or conventional treatment with the same antiviral drug alone and found that the combined treatment affect serum HBV DNA, serum HBeAg, and HBeAg seroconversion. The authors conclude that Phyllanthus in combination with an antiviral drug may be better than the same antiviral drug alone but clinical trials with large sample size and low risk of bias are needed to confirm these findings [141] . The metanoic extracts of P. amarus root and leaf are also recently studied for the treatment of chronic hepatitis C and the root extract showed significant inhibition of HCV-NS3 protease enzyme; whereas the leaf extract showed considerable inhibition of NS5B in the in vitro assays. Both extracts significantly inhibited replication of HCV monocistronic replicon RNA and HCV H77S viral RNA in HCV cell culture system. Furthermore, addition of root extract together with IFN-α showed additive effect in the inhibition of HCV RNA replication [142] . P. emblica and P. urinaria are Phyllanthus species most studied for cancer treatment. Water extract of P. urinaria induces apoptosis by DNA fragmentation and increased caspase-3 activity, reduces the viability of numerous cancer cells lines probably by telomerase suppression activity, and reduces the angiogenesis as suppressing MMP-2 secretion and inhibiting MMP-2 activity through zinc chelation [143] . P. emblica and some of its phytochemicals such as gallic acid, ellagic acid, pyrogallol, some norsesquiterpenoids, corilagin, geraniin, elaeocarpusin, and prodelphinidins B1 and B2 possess antineoplastic effects. It also possess other properties that are efficacious in the treatment and prevention of cancer as radiomodulatory, chemomodulatory, chemopreventive effects, free radical scavenging, antioxidant, anti-inflammatory, antimutagenic, and immunomodulatory activities [144] . In liver, P. emblica and P. urinaria inhibited HepG2 cell growth and five other cancer cell lines [145, 146] . Progallin A isolated from the acetic ether part of the leaves inhibited the proliferation of BEL-7404 cells, upregulated Bax, and downregulated Bcl-2 expression [147] . Defatted methanolic fruit extract of P. emblica suppressed carcinogen-induced response in rat liver with diethylnitrosamine-induced hepatocarcinoma [148] . Liv.52 is an Ayurvedic medicine that was used for 50 years in in the prevention and treatment of viral hepatitis, alcoholic liver disease, early cirrhosis, and a variety of conditions as protein energy malnutrition, loss of appetite, and others. It is composed by Capparis spinosa, Cichorium intybus, Mandur bhamsa, Solarium nigrum, Terminalia arjuna, Cassia occidentalis, Achillea millefolium, and Tamarix gallica [149] . The potential cytoprotective effect of Liv.52 was studied in vitro studies: it improved copper [150] and tert-butyl hydroperoxide (t-BHP) [151] toxicity in HepG2 cells by inhibition of lipid peroxidation, and increase of GSH content and antioxidant enzyme activity. Another recent study found that Liv.52 abrogated the ethanol-induced PPARγ suppression and ethanol-induced TNFα gene expression, it also upregulated PPARγ mRNA [152] . Pretreatment with low (2.6 mL/kg/day) and higher doses (5.2 mL/kg/day) of Liv.52 reversed paracetamol-induced liver toxicity in mice [153, 154] . Few randomized controlled clinical trials were made and results were conflicting [155] , recently, a double-blind, placebo-controlled study reported that cirrhotic patients treated for 6 months with Liv.52 had significantly better Child-Pugh score, decreased ascites, and decreased serum ALT and AST levels compared with placebo group [156] . Hepatocellular carcinoma (HCC) is one of the most frequent cancers in the world and its incidence has been increasing recently in countries including the United States of America, western Europe, and eastern Asia [157] . Systemic chemotherapy plays a palliative role while yields unsatisfactory response rates, which is partly due to the poor selectivity and low uptake efficiency of chemotherapeutic drugs in tumor [158] . Since the prognosis of cirrhotic patients seems to be largely influenced by the development of HCC, every attempt should be performed to prevent HCC in such a highrisk group. Oka et al. reported in a randomized controlled trial that a kind of medicinal herb, "Sho-saiko-to" could significantly decrease hepatic carcinogenesis rate in patients wiith cirrhosis [159] . Moreover, a number of clinical and laboratory studies have been done in the past decades in order to provide the scientific basis for the effectiveness of traditional Chinese medicine against cancer. However, actually, there are a number of contradictory reports due to various factors, as inconsistency in treatment schemes, limited sampling sizes and lack of quality assurance of the herbal products well-designed randomized controlled trials (RCT). In general, most of the published clinical studies are trials without rigorous randomization or they involved single group pre-post, cohort, time series, or matched case-control studies [160] . Herbs are generally used in combination as "formulas," in the belief that in this way their benefits were enhanced and side effects reduced. Moreover, practitioners can adjust or customize the formulas to suit individual cancer patients. Through synergistic interactions between different effective ingredients, the herbal preparation can exert its effects in several ways: (i) they can protect the noncancerous cells and tissues in the body from the possible damage caused by chemo/radiotherapy; (ii) they can enhance the potency of chemo/radiotherapy; (iii) they can reduce inflammatory and infectious complications in the tissues surrounding the carcinoma; (iv) they can enhance immunity and body resistance; (v) they can improve general condition and quality of life; (vi) they can prolong the life span of the patients in the late stages of cancer [161] . The anticancer herbal drugs can be divided into three categories based on their target: (i) drugs that uniquely target topoisomerases (Topos) and perturb DNA replication; (ii) drugs that kill tumor cells through apoptotic pathways; (iii) drugs that alter signaling pathway(s) required for the maintenance of transforming phenotypes of the tumor cells. The cellular and animal studies have provided strong molecular evidences for the anticancer activities of the herbal medicine; however, several important questions remain to be answered. Specifically, three specific issues that will require focused attention: (i) more well-designed clinical trials to support the effectiveness and the safety of TCM in the management of cancers; (ii) new parameters based on the unique properties and theory of TCM to assess the clinical efficacy of TCM in clinical trials; (iii) new approaches to research, given the nature of TCM herbs as being fundamentally different from drugs. Undoubtedly, a clinical study of TCM treatment is more difficult and complicated than the study of single compound drugs. In addition, the effects, as well as the toxicity, of individual herbs or single compounds derived from the herb cannot completely reflect the benefits and toxicity of the herbal combination [162] . As a goal, to develop TCM into rational cancer therapy, more well-designed intensive clinical evaluations and translational laboratory studies are absolutely needed. Also, close collaboration between TCM and conventional Western medicine professions and a combination of TCM with modern multidisciplinary cutting-edge technologies, such as omic methodology on systems biology [163] , would provide us with an attractive and effective strategy to achieve this goal. Although there are many therapeutic strategies including chemotherapy to treat cancer, high systemic toxicity and drug resistance limit the success full outcomes in most cases. Accordingly, several new strategies are being developed to control and treat cancer. One approach could be a combination of and effective phytochemicals with chemotherapeutic agent, which, when combined, would enhance efficacy and reduce toxicity to tissues [164] . Several herbs and plants with different pharmacological properties are known to be rich of sources of chemical constituents that may have a potential for the prevention and the treatment of several human cancers. 9.1. Curcumin. Curcumin has been shown with chemopreventive and chemotherapeutic properties against tumors in animal models and clinical trials [165] [166] [167] . The anticancer effects of curcumin have been documented in many cancers; it induces apoptosis through the death receptor mediated pathway and mitochondrial dysfunction and also induces DNA damage response by cleaving caspase-3. In addition, curcumin induces cell cycle arrest by downregulating the protein expression on cdc2 and inhibits the proliferation of human hepatocellular carcinoma J5 cells in a time-and dosedependent manner. Glycyrrhizin has been shown to successfully prevent the occurrence of primary HCC in patients with HCV-related chronic liver disease by unknown mechanisms [168] . One of the principal roles of long-term administration of glycyrrhizin in decreasing the carcinogenesis rate seemed to be anti-inflammatory ones, which would retrieve an active carcinogenic process. It has been shown that quercetin inhibits the growth of hepatoma cells in dose-and time-dependent manners. Particularly, in a recent study, quercetin treatment of hepatoma cells resulted in changes of cell cycle, reducing HCC progression [169] . Most of studies involving quercetin and HCC analyzed cotreatment with different chemotherapeutics. A study showed that BB-102 (a recombinant adenovirus vector expressing the human p53, GM-CSF and B7-1 genes) and quercetin synergetically suppress HCC cell proliferation and induce HCC cell apoptosis, suggesting a possible use as a combined anticancer agent [170] . In a different study, the authors explored the effect of combination treatment of quercetin in combination with roscovitine in hepatoma cells. Results showed that roscovitine in combination with quercetin can be considered as a potential therapeutic target for treatment of HCC [171] . Furthermore, it has been demonstrate that reactive oxygen species production is involved in quercetin-induced apoptosis in human HCC cell lines so quercetin induces favorable changes in the antioxidant defense system of hepatoma cells that prevent or delay conditions which favor cellular oxidative stress [172] . Otherwise, quercetin, by inducing oxidative stress, potentiates the apoptotic action of 2-methoxyestradiol in human hepatoma cells [173] . The chemopreventive effect of silymarin on HCC has been established in several studies using in vitro and in vivo methods; it can exert a beneficial effect on the balance of cell survival and apoptosis by interfering cytokines. In addition, anti-inflammatory activity and inhibitory effect of silymarin on the development of metastases have also been detected. In some neoplastic diseases, silymarin can similarly be administered as adjuvant therapy. 9.5. Phyllanthus. Phyllantus Emblica exhibits a variety of pharmacological effects including antiinflammatory, antipyretic, antioxidant and anti-mutagenic effects [174] . The active principles of extracts of P. emblica have demonstrated anti-proliferative effects in several cancer cell lines both in vivo and in vivo, thanks to their ability to interfere with cell cycle regulation via the inhibition of cdc 25 phosphatase and partial inhibition of cdc 2 kinase activity [175] . A study examined the growth inhibitory effect of P. emblica on human hepatocellular carcinoma (HepG2) and its synergistic effect with doxorubicin and cisplatin: the effect of chemotherapeutic agents may be modified by combination of P. emblica and be synergistically enhanced in some cases [176] . Depending on the combination ratio, the doses for each drug for a given degree of effect in the combination may be reduced. The mechanism involved in this interaction between chemotherapeutic drugs and plant extracts remains unclear and should be further evaluated. Although research on complementary and alternative medicine (CAM) therapies is still limited, this systematic review has revealed sufficient evidence to conclude that CAM, particularly the herbal products examined can be effective for certain conditions. There are reliable evidences of potential therapeutic benefit. At the same time, the more limited state of knowledge regarding the side effects of this herbal products are studied in this issue. These "natural products" have multiple pharmacological actions on various human physiological systems that would support the treatment of chronic disease like cancer. Moreover, the use of herbal medicines is safe compared with synthetic drugs. Further studies are required to determine the molecular mechanisms of their active ingredients. The limitations of available clinical trials with regard to establishing safety are the same as those for establishing efficacy. Several studies remark the importance of their protective effects for their principle antioxidants effects useful because it may help to prevent carcinogenicity-associated proliferative processes, but there are not recent publication about their toxicity or their side effects derived by their cronic or acute use. Anyway, if it presents, the side effects are poor (i.e., Glycyrrizin can induce hypokalemia, sodium retention, increase in body weight, and elevated blood pressure) [177] . Finally, hepatic damage from conventional drugs is widely acknowledged and most physicians are well aware of them. It is important to remember that acute and/or chronic liver damage occurred after ingestion of some Chinese herbs, herbals that contain pyrrolizidine alkaloids, germander, greater celandine, kava, atractylis gummifera, callilepsis laureola, senna alkaloids, chaparral, and many others. Several herbals have been identified as a cause of acute and chronic hepatitis, cholestasis, drug-induced autoimmunity, vascular lesions, and evenhepatic failure [79] ( Table 1 ). Oxidative stress is the common pathway of chronic liver diseases of different etiology (both viral and alcoholic). CAM seems to exert an antioxidant and antifibrotic effect on liver (even if histological proof of these actions is not provided in all studies), so its use alone or in association with etiologic and causal standard therapies is actually common. For the majority of herbal products, proof of efficacy by randomized, placebo-controlled clinical trials is often lacking. Anecdotal success and personal experience are frequently the driving force for acceptance of CAM in the population [178] . In contrast to pharmaceuticals, CAM are usually distributed as "food supplements" and not evaluated formally for safety and efficacy; variations in methods of harvesting, preparing, and extracting the herb, which can result in dramatically different levels of certain alkaloids. The biologically active substances have been structurally defined and standardized for only a few of the herb: in most countries, their use is neither regulated nor controlled [179] . It has been clearly shown that herbal products can protect the liver from oxidative injury, promote virus elimination, block fibrogenesis, or inhibit tumor growth, but the active molecules must be isolated and tested in suitable culture and animal experiments and finally in randomized, placebocontrolled studies to enable rational clinical use of the agents [180] .

Identification of pneumonia and influenza deaths using the death certificate pipeline

BACKGROUND: Death records are a rich source of data, which can be used to assist with public surveillance and/or decision support. However, to use this type of data for such purposes it has to be transformed into a coded format to make it computable. Because the cause of death in the certificates is reported as free text, encoding the data is currently the single largest barrier of using death certificates for surveillance. Therefore, the purpose of this study was to demonstrate the feasibility of using a pipeline, composed of a detection rule and a natural language processor, for the real time encoding of death certificates using the identification of pneumonia and influenza cases as an example and demonstrating that its accuracy is comparable to existing methods. RESULTS: A Death Certificates Pipeline (DCP) was developed to automatically code death certificates and identify pneumonia and influenza cases. The pipeline used MetaMap to code death certificates from the Utah Department of Health for the year 2008. The output of MetaMap was then accessed by detection rules which flagged pneumonia and influenza cases based on the Centers of Disease and Control and Prevention (CDC) case definition. The output from the DCP was compared with the current method used by the CDC and with a keyword search. Recall, precision, positive predictive value and F-measure with respect to the CDC method were calculated for the two other methods considered here. The two different techniques compared here with the CDC method showed the following recall/ precision results: DCP: 0.998/0.98 and keyword searching: 0.96/0.96. The F-measure were 0.99 and 0.96 respectively (DCP and keyword searching). Both the keyword and the DCP can run in interactive form with modest computer resources, but DCP showed superior performance. CONCLUSION: The pipeline proposed here for coding death certificates and the detection of cases is feasible and can be extended to other conditions. This method provides an alternative that allows for coding free-text death certificates in real time that may increase its utilization not only in the public health domain but also for biomedical researchers and developers. TRIAL REGISTRATION: This study did not involved any clinical trials.

The ongoing monitoring of mortality is crucial to detect and estimate the magnitude of deaths during epidemics, emergence of new diseases (for example, seasonal or pandemic influenza, AIDS, SARS), and the impact of extreme environmental conditions on a population such as heat waves or other relevant public health events or threats [1, 2] . The surveillance of vital statistics is not a novel idea; mortality surveillance has played an integral part in public health since the London Bills of Mortality were devised in the seventeenth century [3] . The Bills served as an early warning tool against bubonic plague by monitoring deaths from the 1635 to the 1830s. Today, mortality surveillance continues to be a critical activity for public health agencies throughout the world [4] [5] [6] [7] . Pneumonia and influenza are serious public health threats and are a cause of substantial morbidity and mortality worldwide; for instance, the World Health Organization (WHO) estimates seasonal influenza causes between 250,000 to 500,000 deaths worldwide each year [8] while pneumonia kills more than 4 million people worldwide every year [9] . Worldwide, the morbidity and mortality of influenza and pneumonia have a considerable economic impact in the form of hospital and other health care costs. Each year in the United States approximately 3 million persons acquire pneumonia and, depending on the severity of the influenza season, 15 to 61 million people in the US contract influenza [9] . These numbers contribute to approximately 1.3 million hospitalizations, of which 1.1 million are pneumonia cases [10] and the remainder for influenza [11] . Moreover, pneumonia cases and influenza together cost the American economy 40.2 billion dollars in 2005 [12] . In The Netherlands it has been estimated that influenza accounts for 3713 and 744 days of hospitalization per 100,000 highrisk and low-risk elderly, respectively [13] . Due to the public health burden and the unpredictability of an influenza season, strong pneumonia and influenza surveillance systems are a priority for health authorities. Mortality monitoring is an important tool for the surveillance of pneumonia and influenza which can aid in the rapid detection and estimates of excess deaths and inform and evaluate the effect of vaccination and control programs. Traditionally, influenza mortality surveillance often uses the category of "pneumonia and influenza" (P-I) on death certificates as an indicator of the severity of an influenza season or to identify trends within a season; however, only a small proportion of these deaths are influenza related. It has been reported that only 8. [5] [6] [7] [8] [9] .8% of all pneumonia and influenza deaths are influenza related [14, 15] . The non-influenza-related pneumonia deaths tend to be stable from year to year and fluctuations in this category are largely driven by the prevalence and severity of seasonal influenza. As a result, the P-I category is an important sentinel indicator. In the US, death certificates are the primary data source for mortality surveillance whose findings are widely used to exemplify epidemics and measure the severity of influenza seasons [16] . Currently, there are three systems to monitor influenza-related mortality; one system in particular, the 122 Cities Mortality Reporting System, provides a rapid assessment of pneumonia and influenza mortality [6] . Each week, this system summarizes the total number of death certificates filed in 122 US cities, as well as the number of deaths due to pneumonia and influenza. However, even these data can be delayed by approximately 2-3 weeks from the times of death. This delay can be attributed to one of the following reasons: 1) timeliness of death registration and 2) reviewing of the death certificates to identify pneumonia and influenza deaths [6, 16, 17] . The registration and reviewing of death certificates varies by states and, as a result, there is variability in length of time to report a death to CDC. For instance, states with paper-based death registration system typically perform manual reviews of the death certificates which can take up to 3 weeks; however states with electronic death registration systems (EDRS) may perform automatic reviews which can decrease this time significantly. The current 122 Cities Mortality Reporting System surveillance system also lacks flexibility for expanding the number of conditions and/or the geographic distribution. Moreover, the unavailability of coded death records due to the complexity of the National Center of Health Statistics (NCHS) coding process results in multiple strategies to identify common outbreaks such as pneumonia and influenza deaths, which greatly vary by jurisdiction. To bypass the lengthy NCHS process, a variety of approaches have been attempted that are close to 'realtime' but less than optimal. For instance, in Utah keyword searching is used to identify pneumonia and influenza deaths; although this method is fast and easy to implement, it can easily result in the over or under estimation of cases. This can occur by missing cases due to misspelled terms, synonyms, variations, or the selection of strings containing the search term. Other research groups [18, 19] have demonstrated the feasibility of using mortality data for real time surveillance but all used "free text" search for the string "pneumonia", "flu" or "influenza." As noted earlier, although this method can provide the semi quantitative measurements for disease surveillance purposes, keyword searches can also result in an array of problems that result from complexities of human language such as causal relationships and synonyms [20] . Therefore, the lack of coded death data that may not be available for months [21] seriously limits the use of death records in automated systems. At this time, there is little published on the automatic assignment of codes to death certificates for automatic case detection. Currently the coding of death certificates is a complex process which involves many entities. In the US, where we are focusing this study, the codes on death certificates that are generated by the National Center for Health Statistics (NCHS) depend on information reported on the death certificate by the medical examiner, coroner, or another certifier, and there is substantial variation in how certifiers interpret and adhere to causeof-death definitions [22] . The cause of death literals are coded into International Classification of Diseases Tenth Revision (ICD-10) [23] and the underlying and multiplecause-of-death codes are selected based on the World Health Organization coding rules. These coding rules have been automated by CDC with the development the Mortality Medical Data System (MMDS) which consists of four programs: Super Mortality Medical Indexing Classification and Retrieval (SuperMICAR) Data Entry; Mortality Medical Indexing Classification and Retrieval (MICAR); Automated Classification of Medical Entities (ACME) and TRANSAX (Translation Axes). SuperMI-CAR was designed to facilitate the entry of literal text of causes of death in death certificates and convert them into standardized expressions acceptable by MICAR. It contains a dictionary which assigns an entity reference number (ERN) to statements on the death certificate. These ERNs are fed into MICAR200 which transforms the ERNS into ICD-10 codes by using specific mortality coding rules; the rules require look-up files and a dictionary. ACME and TRANSAX then selects the underlying and multiple causes of death respectively. ICD-10 codes from MICAR200 are fed into ACME which assigns the underlying cause of death using decision tables. The decision table contains all possible pairings of diseases for which the first disease can cause the second. In the latest version of the system, ACME is comprised of eight decision tables including three tables of valid and invalid codes, causal relationships (General Principle and Rule 1), and direct sequel (Rule 3), and three other tables needed by modification rules. Figure 1 provides the workflow for the MMDS system. Of the 2.3 million deaths that occur each year 80-85 percent are automatically coded through Super-MICAR, and the remaining records are then manually coded by nosologists, a medical classification specialist [24] ; this is a tedious and lengthy process lasting up to 3 months. Although the automation process has decreased the time required for coding death data to 1-2 weeks, the national vital statistics data is not available for at least two years. Therefore, local health department still manually code records or perform basic process techniques to quickly characterize disease patterns [25] . Records that were processed through Super-MICAR or were manually coded are then processed through the remaining components (MICAR200, ACME and TRANSAX) of MMDS. In 1999, MICAR200 had a throughput rate of 95-97%, while ACME rate was 98 percent. Moreover, based on a reliability study, ACME error rate for selecting the underlying cause is at onehalf percent, while TRANSAX, the multiple cause codes had a one-half percent error rate [26] . Due to the high processing rates and low error rates, MMDS is considered by practitioners as the gold standard for the processing and coding of death certificates in the US and other countries (such as Canada, the United Kingdom (UK) and Australia). Therefore, we used the codes produced by this system as the "gold standard" when comparing with the methods developed here. In 1997, the US Steering Committee to Reengineer the Death Registration Process (a task force representing federal agencies, the National Center for Health Statistics and the Social Security Administration, and professional organizations representing funeral directors, physicians, medical examiners, coroners, hospitals, medical records professionals, and vital records and statistics officials (NAPHSIS) published the report "Toward an Electronic Death Registration System in the United States: Report of the Steering Committee to Reengineer the Death Registration Process." This report explained the feasibility of developing electronic death registration in the United States [27] and argued that these electronic death records have the potential to be an effective source of information for nation-wide tracking and detecting of disease outbreaks. However, little actions have been taken to implement such recommendations in a comprehensive manner. As of July 2011, electronic death registration systems were operating in 36 states, the District of Colombia, and in development or planning stage in a dozen others [28] . Information representing the 'cause of death' field on the death certificates is free text. One major goal of natural language processing (NLP) is to extract and encode data from free-texts. There have been many research groups developing NLP systems to aid in clinical research, decision support, quality assurance, the automation of encoding free text data and disease surveillance [29] [30] [31] . Although, there have been a few NLP applications to the public health domain [32, 33] , little is known about its capability to automatically code death certificates for outbreak and disease surveillance. Recently, Medical Match Master (MMM) [25] , developed by Riedl et al at the University of California Davis, was used to match unstructured cause of death phrases to concepts and semantic types within the Unified Medical Language System (UMLS). The system annotates each death phrase input with two types of information, the Concept Unique Identifier, CUI, and a semantic type both assigned by the UMLS. MMM was able to identify an exact concept identifier (CUI) from the UMLS for over 50% of 'cause of death' phrases. Although, the focus of this study was to use NLP techniques to process death certificates, the description of this system reported in the literature did not show how well coded data from an NLP tool along with predefined rules can detect countable cases for a specific disease or condition. The purpose of our project is to create a pipeline which automatically encodes death certificates using a NLP tool and identify deaths related to pneumonia and influenza which provides daily and/or weekly counts. We compared the new technique developed here with keyword searching and MMDS as exemplars of the easiest possible approach and the current "gold standard", respectively. The comparison of the techniques was done by calculating recall, precision, F-measure, positive predictive value and agreement (Cohen's Kappa). We obtained 14,440 de-identified electronic death records all with multiple-cause-of-death from the Utah Department of Health (UDOH) for the period 1 January 2008 to 31 December 2008. The records included a section describing the disease or condition directly leading to death, and any antecedent causes, co-morbid conditions and other significant contributing conditions. An example of a paper and electronic death certificate are shown in Figures 2 and 3 respectively. All death certificates used in this study have been processed using the Mortality Medical Data System (MMDS) and the record axis codes were received from UDOH. For our study we randomly selected 6,450 (45%) records. All death records included in the study were previously also coded by NCHS into ICD-10, but this information was not used for our coding, it was only used as posteriori to assess to quality of the automatic coding. We chose to apply the Centers of Disease Control and Prevention case definition of pneumonia and influenza deaths defined by CDC's epidemiologist staff through personal communication. Therefore, the operational definition for deaths from influenza includes deaths from all types of influenza with the exception of deaths from HAEMOPHILUS INFLUENZAE infection and deaths from PARAINFLUENZAE VIRUS infection. Pneumonia deaths include deaths from all types of pneumonia including pneumonia due to H. influenza and pneumonia due to parainfluenzae virus. The exceptions include aspiration pneumonia (O74.0, O29, O89.0, J69.-and P24.-)1, pneumonitis (J84.1, J67-J70), and pneumonia due to pneumococcal meningitis (J13, G00.1) 1. Pneumonia and influenza related deaths were defined as one of the diagnoses listed in Table 1 which were reported in any cause of death field. These codes were selected through manual review of the ICD-10 version 2007 manual [23] . The Death Certificates Pipeline, DCP, was developed to identify pneumonia and influenza cases. The pipeline consisted of two components. The first component of the system was the natural language processor, for which we used MetaMap [34] , and the second component was the definitional rules that were applied to the output generated by MetaMap. The study procedures for this pipeline included: preprocessing, NLP, extraction of coded data and the detection of pneumonia and influenza cases (Figure 4 ). Spelling errors are common on death certificates; therefore, the death records were first processed through a spell checker to identify misspellings. Although the UMLS SL has a spell suggestion tool called GSPELL [35] [36] [37] , we decided not to use it and chose to utilize ASPELL [38] . Our motivation for this decision was based upon an evaluation which showed ASPELL outperforming GSPELL; ASPELL performed better on three areas of performance which were (2) whether the correct word was ranked in the top ten; and (3) whether the correct word was found at all [35] . PERL (www.perl.org), a high-level computer programming language that aids in the manipulation and processing of large volume of text data was then used to prepare the cause of death free text for NLP. The preprocessing also involved the removal of non-ASCII characters; this was a required technical step for MetaMap processing. Step 2: Natural language processing MetaMap was used to convert the electronic death records to coded descriptions appropriate for the rule based system. MetaMap [34] , developed by the National Library of Medicine (NLM), is useful in identifying biomedical concepts from free-form textual input and maps them into concepts from the Unified Medical Language System (UMLS) Metathesaurus [34, 39] . MetaMap works by breaking the inputted text into words or phrases, map them to standard terms, and then match the terms to concepts in the Unified Medical Language System (UMLS) [40] . For each matched phrase, MetaMap classifies it into a semantic type then returns the concept unique identifier (CUI) and the mapping options which are ranked according to the strength of the mapping. output from MetaMap. Text bolded in the output from NLP represent the code and its corresponding phrase. Step 3: Extraction of coded data The data produced by MetaMap (XML format) was processed through a PERL script to extract the inputted text and its corresponding meta-mapped CUIs. This extracted data was outputted to a text document. Step 4: Identification of P-I deaths The identification of pneumonia and influenza cases involved two steps: 1) identifying CUIs relating to pneumonia and influenza and 2) use of the CUIs to create a rules based algorithm to identify cases. Details of each step are explained in the following paragraphs. To determine which CUI codes were relevant for identifying pneumonia and influenza deaths it was necessary to create a "CUI code list" that represents all the ICD-10 codes of interest (see Table 1 ). To create this list, we generated a subset of the UMLS 2010 AB database [41] using the Metamorphosys [40] tool provided by the National Library of Medicine, NLM. The UMLS database includes many vocabularies, therefore, to determine which vocabularies are relevant to our aims we used the procedure used by Riedl Three queries were performed on the subset described above to map pneumonia and influenza ICD-10 codes to CUIs and identify related pneumonia and influenza concepts. Each query was then placed in a separate database, all duplicates were removed and a sub-query was run to ensure that only the ICD-10 codes in Table 1 were included in this list. This produced 241 distinct concept identifiers (CUIs) relating to pneumonia or influenza. These codes were used to develop the rules to identify the cases of interest. The coded data produced by MetaMap was accessed by rules, aimed at identifying the presence of pneumonia and influenza based on the coded data. The rules for identifying these deaths used the CUI code list described above. The rule looks at each cause of death field (Underlying Cause, Additional Causes, etc.) to flag records with relevant codes. These rules used boolean operators (And, Or, Not) and if-then statements to create a chain of rules ( Figure 5 ). The list of cases identified by our automated detection system was compared with those identified by two other methods: a) keyword searching and b) the reference standard: the ICD-10 codes given by the CDC MMDS method. For key-word searching we followed the process To evaluate the performance of both techniques against the reference standard, we needed to specify what constituted a match. Each death record is associated to a unique number; therefore, we considered a match if the unique identifier was identified by the comparator and also found by the reference standard. Three standard measures were used to evaluate the performance of one method in relation to the reference standard used in this study: precision (equivalent to positive predictive value; recall (equivalent to sensitivity or true positive rate), and F-measure. Kappa statistics were used to assess agreement and McNemar's test was used to analyze the significance between the two methods. All calculations were performed in R [42] . To calculate these values, pneumonia and influenza related deaths were examined by comparing the reference standard output vs. the two comparators: DCP and keyword search. For both comparators, the deaths were counted and categorized as TRUE POSI-TIVES (cases found by the comparator-pneumonia deaths being correctly classified); FALSE POSITIVES (incorrect cases found by the comparator-the number of pneumonia and influenza deaths incorrectly identified by the comparator); FALSE NEGATIVES (correct cases not found by the comparator-the number of pneumonia deaths not identified by the comparator). Precision, recall and F-score were calculated as follows: Precision = True Positives/(True Positives + False Positives) (1) Recall = True Positives/(True Positives + False Negatives) (2) F-measure = 2 *(P R/ P + R) (3) McNemar's test was also calculated to evaluate the significance of the difference between the two comparators. To calculate this value a confusion matrix was created where A is the number of times both methods have correct predictions; B is the number of times method 1 has a correct prediction and method 2 has a wrong prediction; C is the number of times method 2 has a correct prediction and method 1 has a wrong prediction; D is the number of times both methods have incorrect predictions. Ethics approval was not required for this study. Identifying variables that could be used for re-identifying individuals were excluded from the study data. The records were processed and analyzed on a server with two Opteron Dual-Core 2.8 GHz processors and 16 GB RAM at the Center of High Performance Computing at the University of Utah. Using keyword searching the CPU processing time to identify pneumonia and influenza cases was 0.21 seconds and the wall time was 0.37 seconds. For the DCP, the total CPU processing time was 881.83 seconds. The NLP portion of the pipeline attributed to 99.4 percent of the processing time (NLP-877 seconds). While the DCP execution time is much longer, still it is well within the "in real time" realm. For instance, it would take 6,364.3 seconds CPU time seconds for DCP to code and flag all the weekly death records of the US ( 46,523). Recall and precision were calculated at a 0.95 confidence intervals; the F-measure was also calculated. The performance of each method is described below. Of the 6,450 records analyzed keyword search identified 473 records as pneumonia and influenza deaths, 21 being identified as false positives. Precision for keyword searching was calculated at 96%. Of the 21 false positives, 6 records correctly mentioned pneumonia in the cause of death text but their corresponding ICD-10 codes failed to provide any code related to pneumonia, while 2 records were flagged because it included the sub-string "pneumonia" in the additional cause of death field. The death literal for these two records were "bacteremia due to Streptococcus pneumonia" and "Streptococcal Pneumoniae Septicemia", The remaining 13 errors were due to the entry of the death literals; in all cases the negation of 'aspiration pneumonia' either due to: 1) 'pneumonia' being in a separate cause of death field to 'aspiration' or 2) 'pneumonia' not being directly followed by 'aspiration' in the death text (example "pneumonia due to secondary aspiration"). A total of 20 false negatives were recorded, yielding a recall of 96%. The false negatives could be generalized into two categories: 1) misspellings of pneumonia on the death certificated (n = 8) and 2) appropriate pneumonia or influenza ICD-10 code was coded but the death literals did not mention an appropriate scanned phrase (n = 12). F-measure was also calculated at 96%. A high level of agreement was seen among keyword searching and the reference standard (kappa 0.95). Utilizing the Death Certificates Pipeline (DCP), we identified 481 records as pneumonia and influenza deaths, 9 of which were false positives. The precision for this method was calculated at 98%. Like the keyword searching method, of the 9 false positives, 6 records mentioned pneumonia in the cause of death field but their corresponding ICD-10 codes failed to provide any code related to pneumonia and the remaining errors were due to the reporting of aspiration pneumonia on the death certificate. This method had only 1 false negative for the death literal stating "recurrent aspiration with pneumonia", thus yielding a recall at 99.8%, being less than keyword searching. F-measure was calculated at 99%. The level of agreement between the pipeline and the gold standard was almost perfect with a Cohen's kappa of 0.988. The precision and recall scores that are reported above suggest that the DCP is a better method for identifying pneumonia and influenza deaths than keywordsearching. Therefore, we investigated if this observation is supported by statistical analysis. Performing a Fisher's exact test at α = 0.05, significant difference was seen for both recall (p = 1.742e-05) and precision (p = 0.026). The McNemar's test result also showed DCP to be a better method with a p-value = 2.152e-05. For the 472 pneumonia and influenza cases found by the reference standard, DCP correctly identified 471 cases, missed one case and incorrectly flagged nine cases. Most failures were due to discrepancies between the death literal and its respective ICD-10 code. For the only case which the pipeline did not match, the phrase 'recurrent aspiration with pneumonia' was present in the death literal. MetaMap coded this literal as aspiration pneumonia which was excluded from the CUI code list, but its respective ICD-10 included J189. For the 9 additional cases which were not present in the reference standard, we noticed two categories of errors: 1)cases where the string 'pneumonia' is present in the death literal but not coded into ICD-10 and 2) the reporting of aspiration pneumonia on the death certificate. The first category of errors was not due to MetaMap or the rule algorithm, but perhaps due to the coding process. As described earlier, MMDS produces entity axis and record axis codes. The entity axis codes would be a more appropriate reference standard for they provide the ICD 10 codes for the conditions or events reported as listed by the death certifier and maintains the order as written on the death certificate [43] ; but as noted earlier only the record axis codes were made available for this study. The algorithm used to produce record axis codes from the entity axis data removes duplicate codes and contradictory diagnoses within the entity axis data to produce the more standardized record axis [44] . For example, if a medical examiner reports pneumonia with chronic obstructive pulmonary disease both conditions will be shown in entity axis code data. However, in record axis code data, they will be replaced with a single condition: Chronic obstructive pulmonary disease with acute lower respiratory infection (J44.0). We were unable to verify that codes related to pneumonia were present in the entity axis codes for the six cases; therefore, we can only speculate the reason for this failure. The second category of errors was due to the reporting of aspiration pneumonia on the death certificate. In cases where the string "aspiration" and pneumonia" were not reported in the same text field MetaMap processed the string separately thus yielding two codes: one for aspiration and the other pneumonia, instead of one code for "aspiration pneumonia" [C0032290]. In an initial review of MetaMap we found MetaMap had difficulties processing the phrase "pneumonia secondary to acute aspiration", therefore, our rule detection algorithm excluded cases where the code for pneumonia and aspiration were present in the same text field. To our knowledge, this is the first published report on using a natural language processing tool and the UMLS to identify pneumonia and influenza deaths from death certificates. We found that automated coding and identification of pneumonia and influenza deaths is possible and computationally efficient. The Death Certificates Pipeline developed here was statistically different to keyword searching and has higher recall and precision when compared to the current semi-automatic methods in use by the CDC. A good recall is required to help capture the 'true' P-I deaths and a good precision is needed to avoidoverestimating the number of P-I deaths. This study also indicated that keyword searching underestimated pneumonia and influenza deaths in Utah. The simple keyword search method not only decreased recall and precision but also reduced the level of agreement. When reporting counts for surveillance purposes it's best to be as accurate as possible; however, there's a trade-off between recall and precision. For disease surveillance, increased precision enables public health officials to more accurately focus resources for control and prevention, therefore, although both methods had good precision the pipeline developed would be more advantageous to utilize. MetaMap did an excellent job at extracting cause of deaths from free-form text which is consistent with the results of Reid et Al [25] . Most of the concepts were present in the UMLS which attributed good recall. Both recall and precision depended on the comprehensiveness of the CUI code list. The performance of this system is determined largely by the coverage of terms and sources in the UMLS. Both keyword searching and the system's weakest point is its lack of precision. Most of the concepts the system did not identify had either the aspiration text in another field or pneumonia was mentioned in the cause of death text but not coded (9 cases fit these criteria). The sample size was sufficient to show difference between the two methods. It is important to note that utilizing trained nosologists, who would manually code the death certificates, would have developed an absolute gold standard which may or may not be a better reference standard than ICD-10 codes. However, our motivation for utilizing ICD codes was influenced due to the fact that the use of ICD codes to identify all-cause pneumonia has been examined and has showed to be a valid tool for the identification of these cases [45, 46] . In terms of timing, while keyword searching is faster than the DCP, our method is also sub 1/10 second range, which implies that it is possible to process the daily Utah deaths (~40) in approximately 5.47 seconds and all deaths in the US (~6646) in approximately 909.17 seconds using current hardware. This timing would be much faster than the minimum of two weeks to receive the coded data from the current CDC process. Moreover, these timings make it apparent that this system can be integrated in a real time surveillance system without introducing any additional bottlenecks. There are several potential limitations with this analysis. First, the generalizability of the findings is limited because the death records were only from one institution. Although death certificates have a standardized format, the death registration process and the reviewing of death records differ by institutions. UDOH utilizes keyword searching to identify pneumonia and influenza cases, other institutions may use more accurate (manual review) or less accurate methods for finding cases. Second, a separate evaluation of the NLP component of the DCP was not performed. Further research is needed to examine the use of NLP on electronic death records across institutions and countries which may have different documentation procedures. This study shows that it is feasible to achieve high levels of accuracy when using NLP tools to identify cases of pneumonia and influenza cases from electronic death records while still providing a system that can be used for real time coding of death certificates. Identification of concept identifiers related to the CDC's case definition of pneumonia and influenza was very important in producing a highly accurate rule for the identification of these cases. Future work will aim to improve the preprocessing phase of the pipeline by providing the inclusion of the spellchecker used by the CDC's Mortality Medical Data System. Future work will also involve evaluating the flexibility (e.g. identification of different diseases) of the system to deploy the pipeline tool, along with other public health related analytical tools, as a grid service to provide to real time public health surveillance tool that uses data and services under the control of different administrative domains. We have shown that it is feasible to automate the coding of electronic death records for real-time surveillance of deaths of public health concern. The performance of the Pipeline outperformed the performance of current methods, keyword searching, in the identification of pneumonia and influenza related deaths from death certificates. Therefore, the Pipeline has the potential to aid in the encoding of death certificates and is flexible to identify deaths due to other conditions of interest as the need arises.

Functional Analysis of Rift Valley Fever Virus NSs Encoding a Partial Truncation

Rift Valley fever virus (RVFV), belongs to genus Phlebovirus of the family Bunyaviridae, causes high rates of abortion and fetal malformation in infected ruminants as well as causing neurological disorders, blindness, or lethal hemorrhagic fever in humans. RVFV is classified as a category A priority pathogen and a select agent in the U.S., and currently there are no therapeutics available for RVF patients. NSs protein, a major virulence factor of RVFV, inhibits host transcription including interferon (IFN)-β mRNA synthesis and promotes degradation of dsRNA-dependent protein kinase (PKR). NSs self-associates at the C-terminus 17 aa., while NSs at aa.210–230 binds to Sin3A-associated protein (SAP30) to inhibit the activation of IFN-β promoter. Thus, we hypothesize that NSs function(s) can be abolished by truncation of specific domains, and co-expression of nonfunctional NSs with intact NSs will result in the attenuation of NSs function by dominant-negative effect. Unexpectedly, we found that RVFV NSs truncated at aa. 6–30, 31–55, 56–80, 81–105, 106–130, 131–155, 156–180, 181–205, 206–230, 231–248 or 249–265 lack functions of IFN–β mRNA synthesis inhibition and degradation of PKR. Truncated NSs were less stable in infected cells, while nuclear localization was inhibited in NSs lacking either of aa.81–105, 106–130, 131–155, 156–180, 181–205, 206–230 or 231–248. Furthermore, none of truncated NSs had exhibited significant dominant-negative functions for NSs-mediated IFN-β suppression or PKR degradation upon co-expression in cells infected with RVFV. We also found that any of truncated NSs except for intact NSs does not interact with RVFV NSs even in the presence of intact C-terminus self-association domain. Our results suggest that conformational integrity of NSs is important for the stability, cellular localization and biological functions of RVFV NSs, and the co-expression of truncated NSs does not exhibit dominant-negative phenotype.

Rift Valley fever virus (RVFV) belongs to genus Phlebovirus of the family Bunyaviridae, and is a mosquito-borne zoonotic pathogen which causes Rift Valley fever (RVF). RVF is characterized by an acute febrile illness, hemorrhagic fever, neurological disorder or blindness in humans [1, 2, 3, 4] . In ruminants, RVFV induces a high rate of abortion or fetal malformation as well as lethal hepatitis in newborn lambs [5] . The first recognized outbreak of RVF occurred in Kenya in 1930 [6] , and RVFV has spread from endemic region in sub-Saharan Africa into Egypt [7] , Madagascar and the Arabian Peninsula [8, 9, 10, 11, 12] . The potential threat of RVFV introduction into non-endemic countries raises concern of agriculture and public health [13, 14, 15] . RVFV is a risk group 3 pathogen, Category A pathogen and an overlap select agent by the CDC/USDA [16] . The handling of wild-type (wt) RVFV within the U.S. requires BSL3+ or BSL4 facilities. Live-attenuated MP-12 vaccine strain is excluded from select agent rule, and handled at BSL2 level. MP-12 encodes for functional NSs protein, which is useful for the analyses of various NSs functions at BSL2 level [17, 18, 19] . RVFV has a tripartite negative-stranded RNA genome, referred to as Small (S)-, Medium (M)-and Large (L)-segment. The Ssegment encodes for N and NSs genes in an ambi-sense manner, M-segments encodes for NSm, 78-kD protein, NSm-Gn, Gn, and Gc proteins, and L-segment encodes for RNA-dependent RNA polymerase [20, 21, 22, 23] . NSs is a major virulence factor of RVFV and inhibits host general transcription through sequestration of TFIIH p44 [24] or promotion of TFIIH p62 subunits degradation [19] . NSs also inhibits host antiviral response by inhibiting the activation of interferon (IFN)-b promoter through interaction with Sin3A-associated protein (SAP30) at aa.210-230 [25, 26] , and promotion of dsRNA-dependent protein kinase (PKR) degradation [17, 27, 28] . Developing countermeasures against RVFV is important for the prevention of RVF outbreaks or decreasing impact of RVFV introduction. A number of candidate vaccines are under development including live-attenuated vaccine [29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40] , formalin-inactivated vaccine [41, 42] , subunit vaccine [43] , virus-like particle [44, 45, 46, 47] , nonspreading RVFV replicon [48, 49] , viral vectors [50, 51, 52, 53, 54, 55, 56] , and DNA vaccines [57, 58, 59] . For treatment of ongoing outbreaks, several antivirals have been tested for RVFV infection. Liposome-encapsulated ribavirin is effective to treat RVFV infection in mice [60] , while a potent IFN inducer, polyriboinosinic-polyribocytidylic acid stabilized with poly-L-Lysine and carboxymethyl cellulose [Poly(LCIC)] are effective in combination with ribavirin [61] . Therapeutic administration of IFN-a into rhesus monkeys infected with RVFV also limits RVFV replication [62] . The 6-fluoro-3-hydroxy-2-pyrazinecarboxamide (T-705) is shown to be more effective to inhibit RVFV replication than ribavirin [63] , while antiviral small molecule, LJ001 was shown to be effective to numerous enveloped viruses including RVFV [64] . These studies suggest that increased innate immune responses could inhibit RVFV replication in addition to antivirals specific to viral proteins. Since the major virulence factor, NSs protein, is an antagonist of IFN responses, the direct attenuation of NSs function may increase host innate immune responses in cells infected with RVFV potentially limiting RVFV replication. Thus, we aimed to attenuate RVFV NSs function(s) by co-expressing nonfunctional NSs. It has been shown that recombinant MP-12 virus encoding truncated NSs at aa.210 to 230 (SAP30-binding domain) does not inhibit IFN-b mRNA synthesis [26] . Thus, we employed a similar strategy to abolish a part of NSs functions by truncating each 17 to 25 aa. Co-expression of such truncated NSs may exhibit dominant-negative phenotype by self-association through the Cterminus at the self-association domain (aa.249 to 265) [65] . In this study, we generated NSs encoding deletions of 17 to 25 amino acids, characterized the functions of truncated NSs, and analyzed the dominant-negative effects of co-expressed truncated NSs in cells infected with RVFV. Generation of recombinant MP-12 encoding NSs encoding a 17 to 25 aa. truncation Using reverse genetics for the RVFV MP-12 strain, we recovered 11 recombinant MP-12 viruses which encode NSs protein with a 17 to 25 aa. truncation (Fig. 1 ). These NSs mutants exhibited different plaque phenotypes in plaque assay (Fig. S1 ) suggesting possible variation of attenuation by each NSs mutant. The plaques of NSD6-30 and NSD56-80 were clear in neutral red stain, while other mutants made turbid plaques. To test the functions of each NSs protein, VeroE6 cells (type-I IFNincompetent) were mock-infected or infected with MP-12, rMP12-C13type (a control lacking NSs functions) [18] , and NSs mutants using an moi of 3. After 16 hours, cells were collected and the abundance of PKR was measured ( Fig. 2A) as described previously [17] . As expected, PKR was not detectable in cells infected with MP-12 by post-translational downregulation [17] , but was detected in cells mock-infected or infected with rMP12-C13type. However, cells infected with the MP-12 encoding partially truncated NSs did not decrease PKR abundance. Next, we tested if partial deletions within the NSs gene would affect the inhibition of IFN-b mRNA synthesis. Type-I IFNcompetent A549 cells were mock-infected or infected with MP-12, rMP12-C13type or NSs mutants at an moi of 3, and then total RNA was extracted at 7 hpi. Northern blot was performed using RNA probe specific to human IFN-b, ISG56 or RVFV anti-viralsense S RNA/N mRNA as described previously [66, 67] . We tested ISG56 gene, one of the genes controlled under IFN-stimulated response element (ISRE), to confirm the inhibition of host transcription suppression including IFN-b mRNA by NSs. As expected, cells infected with MP-12 inhibited the synthesis of IFNb and ISG56 mRNA, while those infected with rMP12-C13type induced both IFN-b and ISG56 mRNA (Fig. 2B) . Interestingly, none of NSs mutants, including NSD210-230 lacking SAP30binding domain, had inhibited IFN-b mRNA synthesis. Viral replication of those mutants was significantly decreased in type-I IFN-competent MRC-5 cells (Fig. S2) . Therefore, it was concluded that a series of MP-12 encoding partially truncated NSs gene does not degrade PKR and inhibit IFN-b mRNA synthesis. In Western blot as shown in Fig. 2A , all NSs mutants, except for NSD249-265, could be detectable by using anti-RVFV mouse polyclonal antibody. It was possible the anti-RVFV polyclonal antibody does not sufficiently contain antibodies reactive to linear epitopes except for the C-terminus. Thus, we next tested the accumulation of NSD249-265 by using indirect immunofluorescent assay to know if the same anti-RVFV polyclonal antibody can recognize conformational epitopes on NSD249-265. 293 cells were transfected with in vitro synthesized RNA encoding NSs of MP-12, NSD249-265 or chloramphenicol acetyltransferase (CAT) (control), and the cells were fixed with methanol at 16 hours post transfection. Nuclear filamentous inclusion was observed in cells expressing NSs of MP-12 or NSD249-265, while the specific signals of NSs accumulation were weaker in cells expressing NSD249-265 than those expressing MP-12 NSs (Fig. 3) . NSD249-265 was also accumulated in cytoplasm. Next, we tested the cellular localization of truncated NSs other than NSD249-265 by Western blot (Fig. 4) . We did not include NSD249-265 for the experiment as no antibodies were available to detect this NSs in Western blot. 293 cells were mock-infected or infected with MP-12 or NSs truncation mutants at an moi of 3. Cells were collected at 16 hpi, and nuclear and cellular fractions were analyzed for the presence of NSs proteins. MP-12 NSs were accumulated both at cytoplasm and nucleus, while N proteins were exclusively localized at cytoplasm, which is consistent with previous study [65] . Abundant accumulation of NSs in nucleus was only observed in cells infected with NSsD6-30, NSsD 31-55 and NSsD56-80, while other mutants, NSsD 81-105, NSsD106-130, NSsD131-155, NSsD156-180, NSsD181-205 and NSsD206-230, NSsD231-248 poorly accumulated NSs in nucleus. To find if co-expression of nonfunctional NSs is able to attenuate PKR degradation function of MP-12 NSs, VeroE6 cells were infected with MP-12 and then co-infected with rMP12-C13type or one of the NSs truncation mutants using an moi of 3. Cells were collected at 16 hpi and Western blot was used to measure abundance of PKR and RVFV NSs. However, it was found that levels of MP-12 NSs accumulation were not identical to those expressing truncated NSs (Fig. S3) . Only cells co-infected with NSD6-30 and NSD56-80 allowed an efficient accumulation of MP-12 NSs. As a result, PKR was abundantly detected in cells infected with MP-12 and NSs mutants. We attempted to allow accumulation of MP-12 NSs by using 293 cells. Cells were mock-infected or infected with rMP12-NSs-Flag, which encode Flag-tagged NSs in place of intact NSs, at an moi of 3, and subsequently mock-transfected (a control) or immediately transfected with in vitro synthesized RNA encoding CAT (a control), or NSs mutants, as described previously [17] . Cells were collected at 16 hpi, and the abundance of PKR and RVFV proteins were analyzed by Western blot. All cells infected with rMP12-NSs-Flag accumulated abundant levels of NSs (Fig. 5A) . PKR was degraded in infected cells mock-transfected or transfected with CAT RNA, while PKR was also degraded in infected cells transfected with RNA encoding NSD6-30, NSD31-55, NSD56-80, NSD81-105, NSD106-130, NSD131-155, NSD156-180, NSD181-205, NSD206-230 or NSD231-248. On the other hand, the expression of NSD249-265 lacking the Cterminus self-association domain slightly increased the abundance of PKR in cells infected with rMP12-NSs-Flag. We next tested the effect of co-expression of NSs mutants in the inhibition of IFN-b mRNA synthesis by MP-12 NSs. A549 cells were mock-infected or infected with rMP12-NSs-Flag using moi of 3 and then either mock-transfected or immediately transfected with in vitro synthesized RNA encoding CAT or NSs mutants. Total RNA was extracted at 7 hpi, and Northern blot was performed as described above. None of cells transfected with RNA encoding NSs mutants increased the synthesis of IFN-b mRNA (Fig. 5B) . On the other hand, cells transfected with RNA encoding NSD249-265 slightly increased ISG56 mRNA abundance (Fig. 5B) . These results suggest that nonfunctional NSs encoding the C-terminus self-association domain do not have dominant-negative function, while those lacking the C-terminus domain slightly inhibit PKR degradation as well as ISG56 mRNA synthesis. Co-affinity precipitation studies were conducted with use of Strep-tagged protein purification to know if over-expressed truncated NSs can interact with MP-12 NSs in infected cells. 293 cells were infected with moi 3 of rMP-12-NSs-SF (recombinant virus tagged with tandem Strep-Tag and Flag) [19] and were then transfected using the in vitro synthesized capped RNA encoding each of the truncated NSs mutants. After 6 hours, newly synthesized host and viral proteins were labeled with [ 35 S] methionine/cysteine for 4 hours. Whole cell lysates were mixed with Strep-Tactin beads, and SF-tagged MP-12 NSs and bound host and viral proteins were precipitated. Presence of NSs bands were visualized with autoradiography ( Fig. 6 ). As expected, MP-12 NSs was co-precipitated with NSs-SF, and MP-12 NSs was migrating slightly faster than that of NSs-SF. The expression of truncated NSs was lower than that of MP-12 NSs (see input), while the expression of NSsD206-230 was not detectable. The same phenomenon was observed in a repeated experiment, suggesting the instability of NSsD206-230 expression in this experiment. On the other hand, we could not detect co-precipitation of any truncated NSs with NSs-SF. Collectively, our results suggest that those truncated NSs accumulates in cells at low level, mislocalizes, and do not interrupt the MP-12 NSs functions by co-expression. As NSs lacking C-terminus exhibited slight dominant-negative effect on PKR degradation, we hypothesized that an intact sequence at aa.1 to 248 is required for the dominant-negative effect. We next tested the effect of the C-terminus on the dominant-negative effect. We substituted two sequential acidic amino acids triplets located at the C-terminus with alanines as shown in Fig. 7A; i.e., NSs-E253-255A/D257-259A, NSsD257-259A or NSs-E253-255A. We found that NSs-E253-255A could form filamentous inclusion bodies (data not shown), and accumulation was equivalent to that of MP-12 NSs (Fig. 7B) . On the other hand, NSs-E253-255A/D257-259A or NSsD257-259A did not accumulate in cells efficiently, and the NSs could not be detected with IFA (data not shown). We then tested PKR degradation function and IFN-b mRNA suppression function of those mutants using VeroE6 cells and A549 cells using the same method as described above. Cells infected with NSs-E253-255A degraded PKR (Fig. 7B) , and inhibited the synthesis of IFN-b mRNA (Fig. 7C) , while those infected with NSs-E253-255A/ D257-259A or NSsD257-259A did not degrade PKR and did not inhibit IFN-b mRNA synthesis. The results suggest that the glutamic acid at aa.253 to 255 can be replaced without affecting [26] and C-terminus selfassociation domain (aa.249-265) [65] . The rMP12-C13type (C13type) encodes an in-frame truncation of aa. 16-198, NSs functions, while aspartic acid at aa.257 to 259 is important for NSs stability. We next tested the effect of co-expression of NSs-E253-255A/ D257-259A, NSsD257-259A or NSs-E253-255A in infected cells, as described above. 293 cells were mock-infected or infected with rMP12-NSs-Flag using a moi of 3, then cells were mocktransfected or transfected with in vitro synthesized RNA encoding CAT, NSD249-265, NSs-E253-255A/D257-259A, NSsD257-259A or NSs-E253-255A. Cells were collected at 16 hpi for Western blot analysis. As shown in Fig. 8A , co-expression of NSs-E253-255A/D257-259A or NSs-D257-259A or NSs-E253-255A did not inhibit PKR degradation by rMP12-NSs-Flag, while that of NSD249-265 very slightly increased the PKR abundance, which is consistent with the result in Fig. 5A . To test the co-expression effect of those NSs mutants in IFN-b mRNA synthesis, A549 cells were mock-infected or infected with rMP12-NSs-Flag at a moi of 3, and immediately transfected with in vitro synthesized RNA encoding NSs-E253-255A/D257-259A, NSsD257-259A or NSs-E253-255A. Total RNA was extracted at 7 hpi, and Northern blot was performed to detect IFN-b, ISG56 or RVFV S-RNA/N mRNA. None of those mutants attenuated NSsmediated IFN-b mRNA synthesis suppression (Fig. 8B) . These results suggest that attenuation of PKR degradation function might occur due to the accumulation of nonfunctional truncated NSs with some stability (Fig. 3) by the lack of C-terminus 17 amino acids residues. Next, we tested whether the co-expression of NSD249-265 can inhibit MP-12 replication. A549 cells were infected with MP-12 using an moi of 0.01, and then were either mock-transfected or immediately transfected with in vitro synthesized RNA encoding CAT (RNA transfection control), NSD249-265 or NSs-E253-255A (a control with functional NSs). Culture supernatants were harvested at 72 hpi for viral titration using plaque assay. RVFV titer was significantly decreased by the CAT RNA transfection when compared to mock-transfection control. Transfection with RNA encoding NSD249-265 did not further decrease RVFV titer compared to CAT RNA transfected control, while RNA encoding NSs-E253-255A increased RVFV titer significantly. These results suggest that co-expression of NSD249-265 NSs does not significantly decrease viral replication, while that of NSs-E253-255A facilitates RVFV replication by inhibiting IFN-b and PKR in transfected cells. Overall, co-expression of truncated NSs inhibited neither NSs functions nor RVFV replication efficiently. Even though NSs encode self-association domain at the Cterminus domain, the expressed protein mislocalizes in cells, and does not maintain the stability of intact NSs, which minimizes the dominant-negative effect on MP-12 NSs. Discussion Dominant-negative suppression of viral replication has been characterized in a number of different viral proteins [68, 69, 70, 71, 72, 73] . For RVFV, L proteins form oligomer, and exhibit dominant-negative function [74] . In this study, we used NSs protein as a target protein of dominant-negative suppression, because a lack of NSs dramatically attenuates RVFV [75, 76] . Since NSs can self-associate and form filamentous inclusion bodies in infected cells [65, 77] , we hypothesized that co-expression of nonfunctional NSs with the C-terminal self-association domain in cells infected with RVFV allows incorporation of such nonfunctional NSs into the NSs filament, and attenuates a part of NSs functions. However, our results were not as expected; 1) most of truncated NSs localized at cytoplasm, 2) all of truncated NSs did Our results suggest truncation of NSs causes mis-folding and/or mis-localization of protein, which might abolish the ability to interact with authentic MP-12 NSs through the C-terminus selfassociation domain. Unexpectedly, only the co-expression of NSs lacking C-terminus self-association domain (NSD249-265) slightly inhibited PKR degradation by MP-12 NSs. The NSD249-265 could accumulate both in cytoplasm and nucleus, which is consistent with previous study [65] . We speculate that NSD249-265 could compete with host factors required for PKR degradation with intact NSs. In the meantime, co-expression of NSD249-265 NSs did not result in a significant decrease of RVFV replication when compared to CAT RNA transfection control. Therefore, the co-expression of NSs fragments in infected cells might not be an effective strategy to inhibit RVFV replication in vivo. Another novel finding is that all of truncation mutants; i.e., NSD6-30, NSD31-55, NSD56-80, NSD81-105, NSD106-130, NSD131-155, NSD156-180, NSD181-205, NSD206-230, NSD231-248 and NSD249-265, had lacked both PKR degradation and IFN-b suppression functions. This suggests that conformation structure might be important for those NSs functions rather than the presence of linear domain. Our results suggest that those truncated NSs decrease accumulation level, and change the localization pattern in cells. The stability and cellular localization of NSs, which are probably controlled by conformational domain, might be important for biological functions of NSs. Although the C-terminus 17 amino acids were determined as a self-association domain important for filament formation [65] , our result suggests that filament formation does not occur by NSs encoding an in-frame truncation with 25 amino acids. It is possible that the C-terminus 17 amino acids are the prerequisite of NSs self-association, and other structural domains play a role in the filament formation through the C-terminus domain. Our result showed that NSD206-230 dominantly localized at cytoplasm. Previous study showed that rec-ZHD210-230 (recombinant ZH548 encoding an in-frame truncation in NSs at 210-230) could induce IFN-b mRNA due to a lack of SAP30-binding domain in infected cells [26] . It was also shown that NSs binding to SAP30 is required for NSs filaments to target pericentromeric DNA and induce nuclear anomalies [78] . They described that rec-ZHD210-230 expresses a stable NSs protein located in the nucleus in the discussion [78] . Thus, it might be possible that NSs encoding 20 aa. does not change the nuclear localization. On the other hand, NSD6-30, NSD31-55, NSD56-80 and NSD249-265, all of which encode SAP30-binding domain at aa.210-130, could be accumulated in nucleus, whereas they did not inhibit IFN-b gene, suggesting that NSs has another structural requirement to inhibit IFN-b gene activation in addition to SAP30 binding. The requirement of NSs localization and IFN-b gene suppression should be further studied to understand the detailed mechanism of IFN-b gene regulation by RVFV NSs. We also characterized the role of C-terminus acidic residues in NSs functions. We found that both NSs-E253-255A/D257-259A and NSsD257-259A were not abundantly accumulated in infected cells. On the other hand, the NSs-E253-255A accumulated efficiently in cells, and showed a similar phenotype with authentic NSs. Thus, the aspartic acids at 257-259 but not glutamic acid at 253-255 must be important for the stability of NSs. Since the accumulation of NSs-E253-255A/D257-259A and NSsD257-259A were very low, we could not determine whether those mutants are lacking the functions to degrade PKR or inhibit IFNb mRNA synthesis. We noted that the co-infection of recombinant MP-12 encoding truncated NSs with MP-12, except for NSD6-30 and NSD56-80, had resulted in dominant accumulation of truncated NSs. This effect may possibly occur at the transcription or translation level rather than post-translation level, since MP-12 NSs can also accumulate with truncated NSs when RNA transfection was used for truncated NSs expression. If a selective viral transcription of truncated NSs mRNA, or a selective translation of truncated NSs proteins could occur in co-infected cells, then a virus exhibiting these traits may be useful for post-exposure vaccination. However, new studies will be required to detail this mechanism. In summary, short in-frame truncations of RVFV NSs affect the expression level and cellular localization, which lessen or abolish biological functions of NSs most probably due to the lack of functional conformation domains. Thus, co-expression of truncated nonfunctional NSs in RVFV-infected cells does not attenuate NSs functions of RVFV. VeroE6 cells (ATCC CRL-1586), 293 cells (ATCC CRL-1573) and A549 cells (ATCC CCL-185) were maintained in Dulbecco's modified minimum essential medium (DMEM) containing 10% fetal calf serum (FCS). BHK/T7-9 cells that stably express T7 RNA polymerase [79] were maintained in MEM-alpha containing 10% FCS. Penicilin (100 U/ml) and streptomycin (100 g/ml) were added to the culture media. The plasmid encoding anti-viral-sense of MP-12 S-segment at the downstream of the T7 promoter, pProT7-S(+), was described previously [18] . Serial deletion of 75 bp (25 aa.) was introduced into the NSs open reading frame (ORF) of pProT7-S(+) by sitedirected mutagenesis with Pfu Turbo DNA polymerase (Stratagene), designated as pProT7-S(+)-NSD6-30, NSD31-55, NSD56-80, NSD81-105, NSD106-130, NSD131-155, NSD156-180, NSD181-205, NSD206-230 or NSD231-248, respectively. For C-terminus mutant, the PCR fragment encoding NSs ORF with C-terminus 51 bp (17 aa.) deletion was amplified, and cloned between HpaI and SpeI of pProT7-S(+) [18] , designated as NSD249-265. The alanine substitutions of NSs-E253-255A/ D257-259A, NSs-E253-255A or NSsD257-259A were made onto pProT7-S(+) plasmid by site-directed mutagenesis with Pfu Ultra (Agilent Technologies), and designated as pProT7-NSs-E253-255A/D257-259A, NSs-E253-255A or NSsD257-259A, respectively. NSs ORF of those NSs mutants were amplified by PCR with Phusion High Fidelity DNA polymerase (New England Biolab), and cloned into pcDNA3.1mycHisA (Invitrogen) between KpnI and XhoI, and designated as pcDNA3.1mycHisA-NSD6- were transfected with those plasmids as described previously [18] . Total RNA was extracted from infected or mock-infected cells using TRIzol reagent. Denatured RNA was separated on 1% denaturing agarose-formaldehyde gels and transferred onto a nylon membrane (Roche Applied Science). Northern blot analysis was performed as described previously with strand-specific RNA probes to detect RVFV anti-sense S-segment/N mRNA, human IFN-b mRNA, or human ISG56 mRNA [66, 67] . Western blot analysis was performed as described previously [17] . The membranes were incubated with anti-PKR monoclonal A549 cells were infected with rMP12-NSs-Flag at an moi of 0.01, and mock-transfected or immediately transfected with in vitro synthesized RNA encoding CAT, NSD249-265 or NSs-E253-255A. At 72 hpi, culture supernatants were collected, and plaque assay was performed as described previously [18, 80] . The pcDNA3.1mycHisA plasmids encoding CAT [17] , NSD6-30, NSD31-55, NSD56-80, NSD81-105, NSD106-130, NSD131-155, NSD156-180, NSD181-205, NSD206-230, NSD231-248, NSD249-265, NSs-E253-255A/D257-259A, NSs-E253-255A or NSsD257-259A were linearized, and in vitro transcribed by using mMESSAGE mMACHINE T7 Ultra kit (Ambion) according to manufacturer's instruction. The linearized CAT DNA contained myc-His tag at the 39end. Transfection of in vitro synthesized RNA was performed by using TransIT-mRNA Transfection Kit (Mirus) according to manufacturer's instruction. Figure 8 . Co-expression of truncated NSs in RVFV-infected cells. 293 cells were mock-infected or infected with rMP12-NSs-Flag at an moi of 3, and mock-transfected or immediately transfected with in vitro transcribed RNA encoding CAT (control) or NSs with indicated NSs mutants. Cells were collected at 16 hpi, and NSs-Flag/NSs (a-RVFV antibody), NSs-Flag (a-Flag antibody), PKR (anti-PKR antibody), CAT-myc (anti-myc antibody) and b-actin (anti-actin antibody) were detected by Western blot. (B) A549 cells were mock-infected or infected with rMP12-NSs-Flag at an moi of 3, and mock-transfected or immediately transfected with in vitro transcribed RNA encoding CAT or NSs with indicated NSs mutants. As a control to induce IFN-b and ISG56 mRNA synthesis, A549 cells were infected with rMP12-C13type (C13type) at an moi of 3. Total RNA was extracted at 7 hpi, and IFN-b mRNA, ISG56 mRNA and RVFV anti-viral-sense S-RNA/N mRNA were detected by Northern blot with specific RNA probe. Representative data from at least 3 independent experiments are shown. (C) A549 cells were infected with MP-12 at moi of 0.01, and mock-transfected or immediately transfected with in vitro transcribed RNA encoding CAT or NSs of NSD249-265 or NSs-E253-255A. Culture supernatants were harvested at 72 hpi, and virus titers were measured by plaque assay. P-values of unpaired Student's t-test are shown (*; p,0.05, ns; not significant). doi:10.1371/journal.pone.0045730.g008 VeroE6 cells were mock-infected or infected with MP-12 or recombinant MP-12 encoding partially truncated NSs at an moi of 4 in 6-well plate. At 16 hpi, cells were collected, and washed once in PBS. Then, cytoplasmic fraction was lyzed with PBS containing 1% TritonX-100 on ice for 5 min. After centrifugation at 10,000 xg at 4 u C for 5 min, nuclear fraction was washed once with cold PBS, and resuspended in PBS containing 1% TritonX-100. Both cytoplasmic and nuclear fractions were mixed with 26 SDS sample buffer, and subjected to SDS-PAGE and Western blot analysis. Co-affinity precipitation SF (Strep-Flag)-tagged protein was precipitated with Strep-Tactin magnetic beads (Qiagen). 293 cells were first infected with rMP-12 tagged with SF-tag (moi 3) and were then transfected with in vitro synthesized capped RNA (encoding NSD6-30, NSD31-55, NSD56-80, NSD81-105, NSD106-130, NSD131-155, NSD156-180, NSD181-205, NSD206-230, NSD231-248, or NSD249-265). After incubation for 6 hours, newly synthesized proteins were then labeled with [ 35 S] methionine/cysteine (MP Biomedicals). Using cell lysates, SF-tagged proteins were precipitated with Strep-Tactin beads according to manufacturer's instructions. Then, co-precipitated proteins were analyzed by separating on 10% SDS-PAGE gel and followed by autoradiography. Unpaired Student's t-test was performed by using the Graphpad Prism 5.03 program (Graphpad Software Inc.) for the comparison of two groups, All the recombinant DNA and RVFV were created upon the approval of the Notification of Use by the Institutional Biosafety Committee at UTMB. Figure S1 Plaque phenotypes of MP-12 encoding NSs mutants. VeroE6 cells were infected with indicated virus as 10-fold dilution, and overlaid with 0.6% noble agar containing 5% FBS and 5% Tryptose phosphate broth in MEM as described previously [80] . Second overlay of agar containing 0.011% of neutral red was performed at 3 dpi. Plaques at 4 dpi are shown. (TIF) Figure S2 Titer of MP-12 NSs mutants in MRC-5 cells. Human lung diploid MRC-5 cells were infected with MP-12, rMP12-C13type (C13type) or NSs mutants encoding indicated truncations at an moi of 0.01. At 72 hpi, culture supernatants were collected, and virus titers were measured by plaque assay using VeroE6 cells. Means and standard deviations of 3 independent experiments are shown. **p,0.01, Student's unpaired t-test compared to MP-12. (TIF) Figure S3 Co-infection of MP-12 and MP-12 encoding truncated NSs. VeroE6 cells were mock-infected or infected with a mixture of MP-12 (an moi of 3) and either of rMP12-C13type (C13type) or indicated NSs truncation mutants (an moi of 3). Cells were collected at 16 hpi, and PKR (anti-PKR antibody), NSs and N (anti-RVFV antibody) and b-actin (anti-actin antibody) were detected by Western blot. (TIF)

Predictive and Reactive Distribution of Vaccines and Antivirals during Cross-Regional Pandemic Outbreaks

As recently pointed out by the Institute of Medicine, the existing pandemic mitigation models lack the dynamic decision support capability. We develop a large-scale simulation-driven optimization model for generating dynamic predictive distribution of vaccines and antivirals over a network of regional pandemic outbreaks. The model incorporates measures of morbidity, mortality, and social distancing, translated into the cost of lost productivity and medical expenses. The performance of the strategy is compared to that of the reactive myopic policy, using a sample outbreak in Fla, USA, with an affected population of over four millions. The comparison is implemented at different levels of vaccine and antiviral availability and administration capacity. Sensitivity analysis is performed to assess the impact of variability of some critical factors on policy performance. The model is intended to support public health policy making for effective distribution of limited mitigation resources.

As of July 2010, WHO has reported 501 confirmed human cases of avian influenza A/(H5N1) which resulted in 287 deaths worldwide [1] . At the same time, the statistics for the H1N1 2009 outbreak has so far included 214 countries with a total reported number of infections and deaths of 419,289 and 18,239, respectively [2] . Today, an ominous expectation exists that the next pandemic will be triggered by a highly pathogenic virus, to which there is little or no pre-existing immunity in humans [3] . The nation's ability to mitigate a pandemic influenza depends on the available emergency response resources and infrastructure, and, at present, challenges abound. Predicting the exact virus subtype remains a difficult task, and even when identified, reaching an adequate vaccine supply can currently take up to nine months [4, 5] . Even if the existing vaccines prove to be potent, their availability will be limited by high production and inventory costs [6, 7] and also will be constrained by the supply of antiviral drugs, healthcare providers, hospital beds, medical supplies, and logistics. Hence, pandemic mitigation will have to be done amidst limited availability of resources and supporting infrastructure. This challenge has been acknowledged by WHO [7] and echoed by the HHS and CDC [8, 9] . The existing models on pandemic influenza (PI) containment and mitigation aims to address various complex aspects of the pandemic evolution process including: (i) the mechanism of disease progression, from the initial contact and infection transmission to the asymptomatic phase, manifestation of symptoms, and the final health outcome [10] [11] [12] , (ii) the population dynamics, including individual susceptibility [13, 14] and transmissibility [10, [15] [16] [17] , and behavioral factors affecting infection generation and effectiveness of interventions [18] [19] [20] , (iii) the impact of pharmaceutical and nonpharmaceutical measures, including vaccination [21] [22] [23] , antiviral therapy [24] [25] [26] , social distancing [27] [28] [29] [30] [31] and travel restrictions, and the use of low-cost measures, such as face masks and hand washing [26, [32] [33] [34] . Recently, the modeling efforts have focused on combining pharmaceutical and nonpharmaceutical interventions in search for synergistic strategies, aimed at better resource utilization. Most of such approaches attempt implementing a form of social distancing followed by application of pharmaceutical measures. For significant contributions in this area see [33, [35] [36] [37] [38] [39] [40] [41] . One of the most notable among these efforts is a 2006-07 initiative by MIDAS [42] , which cross-examined independent simulation models of PI spread in rural areas of Asia [43, 44] , USA and UK [45, 46] , and the city of Chicago [47] , respectively. MIDAS crossvalidated the models by simulating the city of Chicago, with 8.6M inhabitants and implementing a targeted layered containment [48, 49] . The research findings of MIDAS and some other groups [12, 33] were used in a recent "Modeling Community Containment for Pandemic Influenza" report by IOM, to formulate a set of recommendations for PI mitigation [50] . These findings were also used in a pandemic preparedness guidance developed by CDC [51] . At the same time, The IOM report [50] points out several limitations of the MIDAS models, observing that "because of the significant constraints placed on the models . . . the scope of models should be expanded." The IOM recommends "to adapt or develop decision-aid models that can . . . provide real-time feedback . . . and include the costs and benefits of intervention strategies." Our literature review yields a similar observation that most existing approaches focus on assessment of a priori defined strategies, and virtually none of the models are capable of "learning," that is, adapting to changes in the pandemic progress, or even predicting them, to generate dynamic strategies. Such a strategy has the advantage of being developed dynamically, as the pandemic spreads, by selecting a mix of available mitigation options at each decision epoch, based on both the present state of the pandemic and its predicted evolution. In an attempt to address the IOM recommendations, we present a simulation optimization model for developing predictive resource distribution over a network of regional outbreaks. The underlying simulation model mimics the disease and population dynamics of each of the affected regions (Sections 2.1 and 2.2). As the pandemic spreads from region to region, the optimization model distributes mitigation resources, including stockpiles of vaccines and antiviral and administration capacities (Section 2.3). The model seeks to minimize the impact of ongoing outbreaks and the expected impact of potential outbreaks, using measures of morbidity, mortality, and social distancing, translated into the cost of lost productivity and medical expenses. The methodology is calibrated and implemented on a sample outbreak in Fla, USA with over 4M inhabitants (Section 3). The strategy is compared to the reactive myopic policy, which allocates resources from one actual outbreak region to the next, each time trying to cover the entire regional population at risk, regardless of the resource availability. The comparison is done at different levels of vaccine and antiviral availability and administration capacity. We also present a sensitivity analysis for assessing the impact of variability of some critical factors, including: (i) antiviral efficacy, (ii) social distancing conformance, and (iii) CDC response delay. The objective of our methodology is to generate a progressive allocation of the total resource availability over a network of regional outbreaks. The methodology incorporates (i) a cross-regional simulation model, (ii) a set of single-region simulation models, and (iii) an embedded optimization model. We consider a network of regions with each of which classified as either unaffected, ongoing outbreak, or contained outbreak ( Figure 1) . The cross-regional simulation model connects the regions by air and land travel. The single-region simulation models mimic the population and disease dynamics of each ongoing region, impacted by intervention measures. The pandemic can spread from ongoing to unaffected regions by infectious travelers who pass through regional border control. At every new regional outbreak epoch, the optimization model allocates available resources to the new outbreak region (actual distribution) and unaffected regions (virtual distribution). Daily statistics is collected for each ongoing region, including the number of infected, deceased, and quarantined cases, for different age groups. As a regional outbreak is contained, its societal and economic costs are calculated. In Sections 2.1-2.3, we present the details of the simulation and optimization models. A testbed illustration and a comparison of our strategy to the myopic policy is given in Section 3. A schematic of the cross-regional simulation model is shown in Figure 2 . The model is initialized by creating population entities and mixing groups, for each region. A pandemic is started by an infectious case injected into a randomly chosen region. The details of the resulting regional contact dynamics and infection transmission are given in Section 2.2. As the infected cases start seeking medical help, a new regional outbreak is detected. A resource distribution is then determined and returned to the single-region model. The outbreak can Begin cross-regional simulation The single-region model subsumes the following components (see Figure 3 ): (i) population dynamics (mixing groups and schedules), (ii) contact and infection process, (iii) disease natural history, and (iv) mitigation strategies, including social distancing, vaccination, and antiviral application. The model collects detailed statistics, including number of infected, recovered, deceased, and quarantined cases, for different age groups. For a contained outbreak, its societal and economic costs are calculated. The societal cost includes the cost of lost lifetime productivity of the deceased; the economic cost includes the cost of medical expenses of the recovered and deceased and the cost of lost productivity of the quarantined [52] . Each region is modeled as a set of population centers formed by mixing groups or places where individuals come into contact with each other during the course of their social interaction. Examples of mixing groups include households, offices, schools, universities, shopping centers, entertainment centers, and so forth, [53] . Each individual is assigned a set of attributes such as age, gender, parenthood, workplace, infection susceptibility, and probability of travel, among others. Each person is also assigned Δt time-discrete (e.g., Δt = 1 hour) weekday and weekend schedules, which depend on: (i) person's age, parenthood, and employment status, (ii) disease status, (iii) travel status, and (iv) person's compliance to social distancing decrees [54] . As their schedules advance, the individuals circulate throughout the mixing groups and come into contact with each other (see Section 2.2.2). It is assumed that at any point of time, an individual belongs to one of the following compartments (see Figure 4 ): susceptible, contacted (by an infectious individual), infected (asymptomatic or symptomatic), and recovered/deceased. In what follows, we present the infection transmission and disease natural history model, which delineates the transitions between the above compartments. Process. Infection transmission occurs during contact events between susceptible and infectious cases, which take place in the mixing groups. At the beginning of every Δt period (e.g., one hour), for each mixing group g, the simulation tracks the total number of infectious cases, n g , present in the group. It is assumed that each infectious case generates r g per Δt unit of time new contacts [46] , chosen randomly (uniformly) from the pool of susceptibles present in the group. We also assume the following: (i) during Δt period, a susceptible may come into contact with at most one infectious case and (ii) each contact exposure lasts Δt units of time. Once a susceptible has started her contact exposure at time t, she will develop infection at time t + Δt with a certain probability that is calculated as shown below. Let L i (t) be a nonnegative continuous random variable that represents the duration of contact exposure, starting at time t, required for susceptible i to become infected. We assume that L i (t) is distributed exponentially with mean 1/λ i (t), where λ i (t) represents the instantaneous force of infection applied to susceptible i at time t [55] [56] [57] . The probability that susceptible i, whose contact exposure has started at time t, will develop infection at time t + Δt is then given as A schematic of the disease natural history is shown in Figure 5 . During the incubation phase, the infected case stays asymptomatic. At the end of the latency phase, she enters the infectious phase [44, 46, 48] . She becomes symptomatic at the end of the incubation period. At the end of the infectious phase, she enters the period leading to a health outcome, which culminates in her recovery or death. Mortality for influenza-like diseases is a complex process affected by many factors and variables, most of which have limited accurate data support available from past pandemics. Furthermore, the time of death can sometimes be weeks following the disease episode (which is often attributable to pneumonia-related complications [58] ). Because of the uncertainty underlying the mortality process, we adopted an age-based form of the mortality probability of infected i, as follows: where μ i is the age-dependent base mortality probability of infected i, ρ i is her status of antiviral therapy (0 or 1), and τ is the antiviral efficacy measured as the relative decrease in the base probability [44] . We assume that a recovered case develops full immunity but continues circulating in the region. Mitigation is initiated upon detection of a critical number of confirmed infected cases [59] , which triggers resource distribution and deployment. The model incorporates a certain delay for deploying field responders. Pharmaceutical intervention (PHI) includes vaccination and antiviral application. Vaccination is targeted at individuals at risk [60] to reduce their infection susceptibility. The vaccine takes a certain period to become effective [61] . Vaccination is constrained by the allocated stockpile and administration capacity, measured in terms of the immunizer-hours. We assume that as some symptomatic cases seek medical help [62, 63] , those at risk of them will receive an antiviral. The process is constrained by the allocated stockpile and administration capacity, measured in terms of the number of certified providers. Both vaccination and antiviral application are affected by a number of sociobehavioral factors, including conformance of the target population, degree of risk perception, and compliance of healthcare personnel [64] [65] [66] . The conformance level of the population at risk can be affected, among other factors, by the demographics and income level [67] [68] [69] [70] [71] as well as by the quality of public information available [54] . The degree of risk perception can be influenced by the negative experience developed during previous pharmaceutical campaigns [72, 73] , as well as by public fear and rumors [74, 75] . Nonpharmaceutical intervention (NPI) includes social distancing and travel restrictions. We adopted a CDC guidance [51] , which establishes five categories of pandemic severity and recommends quarantine and closure options according to the category. The categories are determined based on the value of the case fatality ratio (CFR), the proportion of fatalities in the total infected population. For CFR values lower than 0.1% (Category 1), voluntary at-home isolation of infected cases is implemented. For CFR values Influenza Research and Treatment 5 between 0.1% and 1.0% (Categories 2 and 3), in addition to at-home isolation, the following measures are recommended: (i) voluntary quarantine of household members of infected cases and (ii) child and adult social distancing. For CFR values exceeding 1.0% (Categories 4 and 5), all the above measures are implemented. As the effectiveness of social distancing is affected by some of the behavioral factors listed above [54] , we assume a certain social distancing conformance level. Travel restrictions considered in the model included regional air and land border control for infected travelers. Figure 2 , the optimization model is invoked at the beginning of every nth new regional outbreak epoch (n = 1, 2, . . .), starting from the initial outbreak region (n = 1). The objective of the model is to allocate some of the available mitigation resources to the new outbreak region (actual distribution) while reserving the rest of the quantities for potential outbreak regions (virtual distribution). By doing so, the model seeks to progressively minimize the impact of ongoing outbreaks and the expected impact of potential outbreaks, spreading from the ongoing locations. Mitigation resources can include stockpiles of vaccines and antivirals, administration capacity, hospital beds, medical supplies, and social distancing enforcement resources, among others. The predictive mechanism of the optimization model is based on a set of regression equations obtained using single-region simulation models. In what follows, we present the construction of the optimization model and explain the solution algorithm for the overall simulation-based optimization methodology. We introduce the following general terminology and notation: The optimization criterion (objective function) of the model incorporates measures of expected societal and economic costs of the pandemic: the societal cost includes the cost of lost lifetime productivity of the deceased; the economic cost includes the cost of medical expenses of the recovered and deceased and the cost of lost productivity of the quarantined. To compute these costs, the following impact measures of morbidity, mortality, and quarantine are used, for each region k: To estimate these measures, we use the following regression models obtained using a single-region simulation of each region k: where δ i ·· denotes the regression coefficient associated with resource i and δ im ·· is the regression coefficient for the interaction between resources i and m. Similar models are used for Y hk , D hk , and V hk . The above relationships between the impact measures and the resource distributions ought to be determined a priori of implementing a cross-regional scenario (see Section 3). Here, we consider each region k as the initial outbreak region. We assume, however, that as the pandemic evolves, the disease infectivity will naturally subside. Hence, the regression equations need to be re-estimated at every new outbreak epoch, for each region k ∈ C n , using the singleregion simulation models, where each simulation must be initialized to the current outbreak status in region k in the cross-regional simulation. As an alternative to using a computationally burdensome approach of re-estimating the regression equations, a modeler may choose to use a certain decay factor α n [76] to adjust the estimates of the regional impact measures at every nth outbreak epoch, in the following way: In addition, we use the following regression model to estimate the probability of pandemic spread from affected region l to unaffected region k, as a function of resources allocated to region l, which, in turn, impact the number of outgoing infectious travelers from the region: where γ i ·· denotes the regression coefficient associated with resource i, γ im ·· is the regression coefficient associated with interaction between resources i and m, and γ 0 ·· represents the intercept. Consequently, the total outbreak probability for unaffected region k can be found as p k = l∈B n p lk . As in the case of the impact measures, the estimates of the regional outbreak probabilities need to be progressively re-estimated or adjusted using a scheme similar to (4), as follows: 6 Finally, we calculate the total cost of an outbreak in region k at the nth decision epoch as follows: where m h is total medical cost of an infected case in age group h over his/her disease period, w h is total cost of lost wages of an infected case in age group h over his/her disease period, w h is cost of lost lifetime wages of a deceased case in age group h, and w h is daily cost of lost wages of a non-infected case in age group h who complies with quarantine. The model. The optimization model has the following form. Minimize TC n j q 1 j , q 2 j , . . . , q r j + s∈C n TC n s q 1s , q 2s , . . . , q rs · p n s subject to The first term of the objective function represents the total cost of the new outbreak j, estimated at the nth outbreak epoch, based on the actual resource distribution {q 1 j , q 2 j , . . . , q r j } (see (7)). The second term represents the total expected cost of outbreaks in currently unaffected regions, based on the virtual distributions {q 1s , q 2s , . . . , q rs } (7) and the regional outbreak probabilities p n s (6) . The set of constraints assures that for each resource i, the total quantity allocated (current and virtual, both nonnegative) does not exceed the total resource availability at the nth decision epoch. Note that both the objective function and the availability constraints are nonlinear in the decision variables. (1) Estimate regression equations for each region using the single-region simulation model. (2) Begin the cross-regional simulation model. (4) Select randomly the initial outbreak region j. Set n = 1. (c) Re-estimate regression equations for each region k ∈ B n ∪ C n using the single-region simulations, where each simulation is initialized to the current outbreak status in the region (alternatively, use (4) and (6)). (d) Solve the resource distribution model for region j. (e) Update the total resource availabilities. (10) Calculate the total cost for each contained region and update the overall pandemic cost. To illustrate the use of our methodology, we present a sample H5N1 outbreak scenario including four counties in Fla, USA: Hillsborough, Miami Dade, Duval, and Leon, with populations of 1.0, 2.2, 0.8, and 0.25 million people, respectively. A basic unit of time for population and disease dynamics models was taken to be Δt = 1 hour. Regional simulations were run for a period (up to 180 days) until the daily infection rate approached near zero (see Section 3.3). Below, we present the details on selecting model parameter values. Most of the testbed data can be found in the supplement [77] . Models. Demographic and social dynamics data for each region [77] were extracted from the U.S. Census [78] and the National Household Travel Survey [79] . Daily (hourly) schedules [77] were adopted from [53] . Each infected person was assigned a daily travel probability of 0.24% [79] , of which 7% was by air and 93% by land. The probabilities of travel among the four regions were calculated using traffic volume data [80] [81] [82] [83] , see Table 1 . Infection detection probabilities for border control for symptomatic cases were assumed to be 95% and 90%, for air and land, respectively [84] . The instantaneous force of infection applied to contact i at time t ((1), [57] ) was modeled as Influenza Research and Treatment 7 where α i is the age-dependent base instantaneous infection probability of contact i, θ i (t) is her status of vaccination at time t (0 or 1), and δ is the vaccine efficacy, measured as the reduction in the base instantaneous infection probability (achieved after 10 days [61] ). The values of age-dependent base instantaneous infection probabilities were adopted from [46] (see Table 2 ). The disease natural history included a latent period of 29 hours (1.21 days), an incubation period of 46 hours (1.92 days), an infectiousness period from 29 to 127 hours (1.21 to 5.29 days), and a period leading to health outcome from 127 to 240 hours (5.29 to 10 days) [85] . Base mortality probabilities (μ i in (2)) were found using the statistics recommended by the Working Group on Pandemic Preparedness and Influenza Response [52] . This data shows the percentage of mortality for age-based high-risk cases (HRC) ( Table 3 , columns 1-3). Mortality probabilities (column 4) were estimated under the assumption that highrisk cases are expected to account for 85% of the total number of fatalities, for each age group [52] . Single-region simulation models were calibrated using two common measures of pandemic severity [35, 45, 46] : the basic reproduction number (R 0 ) and the infection attack rate (IAR). R 0 is defined as the average number of secondary infections produced by a typical infected case in a totally susceptible population. IAR is defined as the ratio of the total number of infections over the pandemic period to the size of the initial susceptible population. To determine R 0 , all infected cases inside the simulation were classified by generation of infection, as in [33, 43] . The value of R 0 was calculated as the average reproduction number of a typical generation in the early stage of the pandemic, with no interventions implemented (the baseline scenario) [33] . Historically, R 0 values for PI ranged between 1.4 and 3.9 [37, 43] . To attain similar values, we calibrated the hourly contact rates of mixing groups [77] (original rates were adopted from [46] ). For the four regions, the average baseline value of R 0 was 2.54, which represented a high transmissibility scenario. The values of regional baseline IAR averaged 0.538. Mitigation resources included stockpiles of vaccines and antiviral and administration capacities (Section 3.4). A 24-hour delay was assumed for deployment of resources and filed responders [59] . PHI. The vaccination risk group included healthcare providers [66] , and individuals younger than 5 years (excluding younger than 12 months old) and older than 65 years [60] . The risk group for antiviral included symptomatic individuals below 15 years and above 55 years [60, 86] . The efficacy levels for the vaccine (δ in (9)) and antiviral (τ in (2)) were assumed to be 40% [44, 87] and 70%, respectively. We did not consider the use of antiviral for a mass prophylactic reduction of infection susceptibility due to the limited antiviral availability [9] and the risk of emergence of antiviral resistant transmissible virus strains [26] . We assumed a 90% target population conformance for both vaccination and antiviral treatment [64] . The immunity development period for the vaccine was taken as 10 days [61] . A version of the CDC guidance for quarantine and isolation for Category 5 was implemented (Section 2.2.4, [51] ). Once the reported CFR value had reached 1.0%, the following policy was declared and remained in effect for 14 days [51] : (i) individuals below a certain age ξ (22 years) stayed at home during the entire policy duration, (ii) of the remaining population, a certain proportion φ [88] stayed at home and was allowed a one-hour leave, every three days, to buy essential supplies, and (iii) the remaining (1 − φ) noncompliant proportion followed a regular schedule. All testbed scenarios considered the quarantine conformance level φ equal to 80% [54] . An outbreak was considered contained, if the daily infection rate did not exceed five cases, for seven consecutive days. Once contained, a region was simulated for an additional 10 days for accurate estimation of the pandemic statistics. A 2 5 statistical design of experiment [89] was used to estimate the regression coefficient values of the significant decision factors and their interactions (see Section 2.3; the values of adjusted R 2 ranged from 96.36% to 99.97%). The simulation code was developed using C++. The running time for a cross-regional simulation replicate involving over four million inhabitants was between 17 and 26 minutes (depending on the initial outbreak region, with a total of 150 replicates) on a Pentium 3.40 GHz with 4.0 GB of RAM. The performance of the DPO and myopic policies is compared at different levels of resource availability. Table 4 summarizes the total vaccine and antiviral requirements for each region, based on the composition of 8 Influenza Research and Treatment Average daily cost of lost productivity of a non-infected quarantined case (20-99) $432.54 theregional risk groups (see Section 3.3). Table 5 shows the per capita costs of lost productivity and medical expenses, which were adopted from [52] and adjusted for inflation for the year of 2010 [90] . Comparison of the two strategies is done at the levels of 20%, 50%, and 80% of the total resource requirement shown in Table 4 . Figures 6(a) and 6(b) show the policy comparison in the form of the 95% confidence intervals (CI) for the average number of infected and deceased, respectively. Figure 7 also shows the policy comparison using the 95% CI for the average total pandemic cost, calculated using the pandemic statistics, and the per capita costs from Table 5 . For illustrative purposes, we also show the average number of regional outbreaks, for each policy, at different levels of resource availability, in the testbed scenario involving four regions, with the Hillsborough as the initial outbreak region ( Table 6) . It can be observed that the values of all impact measures exhibit a downward trend, for both DPO and myopic policies, as the total resource availability increases from 20% to 80%. An increased total resource availability not only helps alleviating the pandemic impact inside the ongoing regions but also reduces the probability of spread to the unaffected regions. For both policies, as the total resource availability approaches the total resource requirement (starting from approximately 60%), the impact numbers show a converging behavior, whereby the marginal utility of additional resource availability diminishes. This behavior can be explained by noting that the total resource requirements were determined assuming the worst case scenario when all (four) regions would be affected and ought to provided with enough resources to cover their respective regional populations at risk. It can also be seen that on average, the DPO policy outperforms the myopic approach at all levels, which can attest to a more efficient resource utilization achieved by the DPO policy (see also Table 6 ). The difference in the policy performance is particularly noticeable at the lower levels of resource availability, and it gradually diminishes, as the resource availability increases and becomes closer to be sufficient to cover the entire populations at risk in all regions. It can also be noted that the variability in the performance of the DPO strategy is generally smaller than that of the myopic policy. In general, for both strategies, the performance variability decreases with higher availability of resources. In this section, we assess the marginal impact of variability of some of the critical factors. The impact was measured separately by the change in the total pandemic cost and the number of deaths (averaged over multiple replicates), resulting from a unit change in a decision factor value, one factor at a time. Factors under consideration included: (i) antiviral efficacy, (ii) social distancing conformance, and (iii) CDC response delay. We have used all four regions, separately, as initial outbreak regions for each type of sensitivity analysis. The results (patterns) were rather similar. Due to limited space, we have opted to show the results for only one initial region, chosen arbitrarily, for each of the three types of sensitivity studies. While Duval County was selected as the initial outbreak region to show the sensitivity results on antiviral efficacy, Hillsborough and Miami Dade were used as the initial regions to show the results on, respectively, social distancing conformance and CDC response delay. Figure 8 depicts the sensitivity of the average total cost and average total deaths to antiviral efficacy values between 0% and 80%. As expected, for both policies, the curves for the average number of deaths exhibit a decreasing trend which is almost linear for the values of τ between 0% and 40%. As the value of τ approaches 70%, the curves start exhibit a converging behavior. The curves for the average total pandemic cost exhibit a similar pattern for both policies. It can be noted that the performance of both policies is somewhat identical for low antiviral efficacy (between 0% and 30%). However, the performance of the DPO policy improves consistently as τ increases which can be attributed to a more discretionary allocation of the antiviral stockpile by the DPO policy. Reduction of the contact intensity through quarantine and social distancing has proven to be one of the most effective containment measures, especially in the early stages of the pandemic [27, 30, 31, 41] . Figure 9 shows the sensitivity of the average total cost and average total deaths to the social distancing conformance ranging between 60% and 80%. We observed that for both impact measures, the DPO policy demonstrated a better performance with the difference ranging from $3B to $26B in the total cost and from 1,400 to 20,000 in the number of fatalities. The biggest difference in performance was achieved at the lower-to-medium levels of conformance (between 65% and 72%). As the conformance level approached 80%, the dominating impact of social distancing masked the effect of better utilization of vaccines and antivirals achieved by the DPO strategy. The CDC response delay corresponds to the interval of time from the moment an outbreak is detected to a complete deployment of mitigation resources. Depending on the disease infectivity, CDC response delay may represent one of the most critical factors in the mitigation process. Figure 10 shows how the performance of both policies was significantly impacted by this factor. The DPO policy showed a uniformly better performance with the difference ranging between $3B to $4B in the average total cost, and between 800 to 1,800 in the average number of mortalities, over the range (24- Figure 10 : Sensitivity analysis for CDC response delay. As recently pointed by the IOM, the existing models for PI mitigation fall short of providing dynamic decision support which would incorporate "the costs and benefits of intervention" [50] . In this paper, we present a large-scale simulation optimization model which is attempted at filling this gap. The model supports dynamic predictive resource distribution over a network of regions exposed to the pandemic. The model aims to balance both the ongoing and potential outbreak impact, which is measured in terms of morbidity, mortality, and social distancing, translated into the cost of lost productivity and medical expenses. The model was calibrated using historic pandemic data and compared to the myopic strategy, using a sample outbreak in Fla, USA, with over 4 million inhabitants. Summary of the main results. In the testbed scenario, for both strategies, the marginal utility of additional resource availability was found to be diminishing, as the total resource availability approached the total requirement. In the testbed scenario, the DPO strategy on average outperformed the myopic policy. As opposed to the DPO strategy, the myopic policy is reactive, rather than predictive, as it allocates resources regardless of the remaining availability and the overall cross-regional pandemic status. In contrast, the DPO model distributes resources trying to balance the impact of actual outbreaks and the expected impact of potential outbreaks. It does so by exploiting regionspecific effectiveness of mitigation resources and dynamic reassessment of pandemic spread probabilities, using a set of regression submodels. Hence, we believe that in scenarios involving regions with a more heterogeneous demographics, the DPO policy will likely to perform even better and with less variability than the myopic strategy. We also note that the difference in the model performance was particularly noticeable at lower levels of resource availability, which is in accordance with a higher marginal utility of additional availability at that levels. We thus believe that the DPO model can be particularly useful in scenarios with very limited resources. Contributions of the paper. The simulation optimization methodology presented in this paper is one of the first attempts to offer dynamic predictive decision support for pandemic mitigation, which incorporates measures of societal and economic costs. Our comparison study of the DPO versus myopic cross-regional resource distribution is also novel. Additionally, our simulation model represents one of the first of its kind in considering a broader range of social behavioral aspects, including vaccination and antiviral treatment conformance. The simulation features a flexible design which can be particularized to a broader range of PHI and NPI and even more granular mixing groups. We also developed a decision-aid simulator which is made available to the general public through our web site at http://imse.eng.usf.edu/pandemics.aspx. The tool is intended to assist public health decision makers in implementing what-if analysis for assessment of mitigation options and development of policy guidelines. Examples of such guidelines include vaccine and antiviral risk groups, social distancing policies (e.g., thresholds for declaration/lifting and closure options), and travel restrictions. Limitations of the model. Lack of reliable data prevented us from considering geo-spatial aspects of mixing group formation. We also did not consider the impact of public education and the use of personal protective measures (e.g., face masks) on transmission, again due to a lack of effectiveness data [91] . We did not study the marginal effectiveness of individual resources due to a considerable uncertainty about the transmissibility of an emerging pandemic virus and efficacy of vaccine and antiviral. For the same reason, the vaccine and antiviral risk groups considered in the testbed can be adjusted, as different prioritization schemes have been suggested. The form of social distancing implemented in the testbed can also be modified as a variety of schemes can be found in the literature, including those based on geographical and social targeting. Effectiveness of these approaches is substantially influenced by the compliance factor, for which limited accurate data support exists. It will thus be vital to gather the most detailed data on the epidemiology of a new virus and the population dynamics early in the evolution of a pandemic, and expeditiously analyze the data to adjust the interventions accordingly.

A Novel Vaccine Using Nanoparticle Platform to Present Immunogenic M2e against Avian Influenza Infection

Using peptide nanoparticle technology, we have designed two novel vaccine constructs representing M2e in monomeric (Mono-M2e) and tetrameric (Tetra-M2e) forms. Groups of specific pathogen free (SPF) chickens were immunized intramuscularly with Mono-M2e or Tetra-M2e with and without an adjuvant. Two weeks after the second boost, chickens were challenged with 107.2 EID50 of H5N2 low pathogenicity avian influenza (LPAI) virus. M2e-specific antibody responses to each of the vaccine constructs were tested by ELISA. Vaccinated chickens exhibited increased M2e-specific IgG responses for each of the constructs as compared to a non-vaccinated group. However, the vaccine construct Tetra-M2e elicited a significantly higher antibody response when it was used with an adjuvant. On the other hand, virus neutralization assays indicated that immune protection is not by way of neutralizing antibodies. The level of protection was evaluated using quantitative real time PCR at 4, 6, and 8 days post-challenge with H5N2 LPAI by measuring virus shedding from trachea and cloaca. The Tetra-M2e with adjuvant offered statistically significant (P < 0.05) protection against subtype H5N2 LPAI by reduction of the AI virus shedding. The results suggest that the self-assembling polypeptide nanoparticle shows promise as a potential platform for a development of a vaccine against AI.

Avian influenza (AI) is a devastating poultry disease with serious economic consequences to the commercial poultry industry. AI is also a significant public health concern because of recent highly pathogenic H5N1 avian influenza outbreaks causing also human deaths in Asia, Europe, and North Africa. According to the world health organization (WHO) update, 2011, since 2003, 520 confirmed cases of human infection with H5N1 have been reported, of which 307 died due to disease complications. However, other avian influenza viruses including low-pathogenic avian influenza (LPAI) can also be a risk to public health. For instance, the LPAI subtype H9N2 infection in chickens is mild to asymptomatic and easily overlooked. However, it shares similar receptor binding epitopes with human influenza viruses and can infect humans [1] . There is a risk for LPAI subtypes H5 and H7 to become high-pathogenic avian influenza (HPAI) viruses in chickens due to constant virus shedding and transmission to new birds within the flock or neighboring flocks [2, 3] . Vaccination is an effective way for prevention of viral diseases in poultry. However, routine vaccination against AI has not been widely practiced throughout the world mainly for surveillance reasons [1, 2] . When there is the desire for routine vaccination, constant 2 Influenza Research and Treatment reformulation of AI vaccines is required according to the circulating field virus, which can be cumbersome in the case of an immediate outbreak. Current vaccines against AI viruses can reduce mortality, clinical signs, shedding, and transmission of the virus in poultry, but they are not capable of preventing infection and virus replication [4] . The design of a universal influenza vaccine has been the major focus of researchers in the influenza vaccinology field. The external domain of matrix protein 2 (M2e) has been one of the main interests for the generation of a universal AI vaccine. The M2e is encoded by a separate open reading frame of segment 7 of the influenza virus genome, is located in the viral envelope, and projects from the surface of the virus as tetramers [5, 6] . The M2 is composed of 97 amino acids which forms 3 domains: the external domain, the transmembrane domain, and the internal domain. The external domain of M2 (M2e) is recognized by the host's immune system [7] [8] [9] . Initially, vaccination of ferrets with whole M-or M2-expressing recombinant vaccinia virus showed no evidence of protection [10] . However, later vaccine constructs using plasmid and recombinant salmonella expressing M or M2 induced significant protection in terms of reduction in virus growth and mortality in mice and chickens, respectively [11] [12] [13] . A multiple antigenic peptide construct containing M2e (M2e-MAP) induced strong M2especific antibody titers in the serum of mice and resulted in significant protection against influenza virus challenge [13] . Liang et al., 1994 [14] showed the importance of CD4 + T cells for nasal resistance and protection against the virus. It is assumed that M2e-specific memory T h cells also may have an important role in protection against the virus in the nose and trachea of mice [13] . De Filette et al., 2005 [15] used the hepatitis B virus core particle (HBc) as a carrier and fused M2e (conserved region of human influenza A virus) to either the C-terminus of HBc or inserted it in the immune-dominant loop of HBc. Immunization of mice with this M2e-HBc vaccine was 100% protective against lethal challenge [15] [16] [17] . Antigenic epitopes of pathogens are peptides that are capable of inducing an immune response. However, their small size limits their immunogenicity. Therefore, usually a larger carrier protein, such as bovine serum albumin (BSA), keyhole limpet hemocyanin (KLH), or a virus-like particle (VLP), is required for optimal immunogenicity [18] . Structural organization of the epitope on the carrier is critical for inducing stronger immune responses. Denis et al., 2007 [19] demonstrated that a monomeric form of M2e peptide was not immunogenic and Huleatt et al., 2008 [20] tried to solve that problem by adding 4 copies of the M2e peptide in their platform. Here, we used peptide nanoparticles as a platform to display the M2e peptide to the host's immune system. These nanoparticles represent a novel type of repetitive antigen display system which allows presenting the M2e peptide in high density in both, either in its monomeric or its tetrameric form. This idea was first presented in Raman et al., 2006 [21] ; the monomeric peptide is composed of two coiled coils connected by a short linker region. The association between the coiled coils induces self-assembly of the monomers into Figure 1 : Computer model (a): the pentameric-trimeric architecture of Mono-M2e and the fully assembled icosahedral nanoparticle. (b) Tetra-M2e, with tetrameric-trimeric architecture, and the resulting octahedral nanoparticle. In both images, green: pentameric coiled coil, turquoise: tetrameric coiled coil, and blue: trimeric coiled coil. Red represents M2e in either its monomeric or tetrameric state. spherical nanoparticles with either icosahedral or octahedral symmetry ( Figure 1 ) according to our computer models. The potential for these nanoparticles to serve as platforms for vaccines is apparent. As opposed to live attenuated vaccines, there is no risk of infection within the vaccinated population [21] . Furthermore, the ease and speed of protein expression, purification, and self-assembly into nanoparticles reduce the cost and time of large-scale production. The concept has been successfully used for the design of malaria [22] and SARS [23] vaccines prototypes. Here, we present the biophysical characterization of the nanoparticles and an immunological profiling using chickens as test animals. The results suggest that the selfassembling polypeptide nanoparticle shows promise as a potential vaccine against AI. 2.1. Nanoparticle Synthesis. The DNA coding for the nanoparticle constructs was prepared using standard molecular biology procedures. Shortly, plasmids containing the peptide monomers (Table 1) were constructed by cloning complementary oligonucleotides (CCCGGGGGGGCAGCGGC AGCCTGCTGACCGAAGTGGAAACCCCGACCCGCAAC-GGCTGGGAATAATGAATTC) encoding the avian M2e epitope with flanking residues (ARGGSGSLLTEVETPTRNGW-E * * E) into the XmaI/EcoRI restriction sites of the basic SAPN expression construct to yield Mono-M2e. To Table 1 : Summary of self-assembling nanoparticle peptide sequences. Peptide sequence The peptide Mono-M2e is composed of a pentameric coiled coil (green) and a trimeric coiled coil (blue). Tetra-M2e uses the same trimer but has a tetrameric coiled coil (turquoise). In both sequences, the M2e epitope is shown in red. Other amino acid residues, such as linkers and his-tags, are shown in black. M2eN-GCN4 consists of M2e attached to the tetrameric GCN4 coiled coil, shown in brown. Monomeric M2e, used for ELISA, coating is shown in red. generate Tetra-M2e, we first cloned the tetrameric oligomerization domain of tetrabrachion into the BamHI/BssHII restriction sites of pPEP-T ( Figure 5 ), before cloning complementary oligonucleotides (ATGCATCCCTGGTTCC GCGTGGAAGCCTGCTGACCGAAGTGGAAACCCCGAC-CCGCAACGGCTGGGAATGCAAATGCAGCGATAGCAGC GGATCC) coding for the slightly longer avian M2e sequence (HASLLTEVETPTRNGWECKCSDSSGS) including flanking residues into the N-terminal NsiI/BamHI restriction sites. The M2e-GCN4 construct was made by replacing the nanoparticle fragment of Tetra-M2e with the GCN4 sequence. The plasmids were transformed into Escherichia coli BL21 (DE3) cells, which were grown in Luria broth with ampicillin at 37 • C. Expression was induced with isopropyl β-D-thiogalactopyranoside. Four hours after induction, cells were removed from 37 • C and harvested by centrifugation at 4,000 ×g for 15 min. The cell pellet was stored at −20 • C. The pellet was thawed on ice and suspended in a lysis buffer consisting of 9 M urea, 100 mM NaH 2 PO 4 , 10 mM Tris pH 8, 20 mM imidazole, and 0.2 mM Tris-2-carboxyethl phosphine (TCEP). Cells were lysed by sonication and the lysate was cleared by centrifuging at 30.500 ×g for 45 min. The cleared lysate was incubated with Ni-NTA Agarose Beads (Qiagen, Valencia, CA, USA) for at least 1 hour. The column was washed with lysis buffer and then a buffer containing 9 M urea, 500 mM NaH 2 PO 4 , 10 mM tris pH 8, 20 mM imidazole, and 0.2 mM TCEP. Protein was eluted with a pH gradient: 9 M urea, 100 mM NaH 2 PO 4 , 20 mM citrate, 20 mM imidazole, and 0.2 mM TCEP. Subsequent washes were done at pH 6.3, 5.9, and 4.3. Following the pH gradient, a gradient of lysis buffer with increasing imidazole strength was used to further elute the protein. Purity was assessed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) as shown in Figure 6 . The protein solution was filtered with a 0.1 μm polyvinylidene fluoride membrane filter (Millipore Billerica, MA, USA). Nanoparticle self-assembly was performed by dialysis into buffer containing 8 M urea, 20 mM Tris pH 7.5, 150 mM NaCl, and 5% glycerol, at a protein concentration of 0.1 mg/mL. This was followed by dialysis into the same buffer containing decreasing concentrations of urea: 6 M, 4 M, 2 M, 1 M, and two changes of the same buffer without urea. Following self-assembly, the nanoparticle solution was again filtered with a 0.1 μm filter. Scattering. Dynamic light scattering experiments were carried out on a Zetasizer Nano S Instrument (Malvern, Worcestershire, UK), with a 633 nm He-Ne laser. All measurements were carried out at 25 • C in a buffer containing 20 mM Tris pH 7.5, 150 mM NaCl, and 5% glycerol. Samples were negatively stained with 1% uranyl acetate (SPI Supplies, Westchester, PA, USA) and observed with a FEI Tecnai T12 S/TEM at an accelerating voltage of 80 kV (FEI, Hillsboro, Oregon). The peptide concentration of the constructs was about 0.05 mg/mL. Samples were dialyzed into 20 mM sodium phosphate pH 7.5, 150 mM NaCl, and 5% glycerol and concentrated or diluted to a peptide concentration of about 0.13 mg/mL for Mono-M2e and about 0.05 mg/mL for Tetra-M2e. Circular dichroism measurements were performed at room temperature using an Applied Photophysics (Surrey, UK) Pi Star 180 spectropolarimeter, taking measurements from 200 to 250 nm. The influenza virus used in the direct challenge AI study was A/Turkey/CA/D0208651-C/02 H5N2 low pathogenic. Influenza A/Turkey/Wisconsin/1/1966 H9N2 low pathogenic was used for hyperimmune serum production provided by Charles River Avian Vaccine Services (Storrs, CT). Viruses were grown and titered in 9-to 11-dayold embryonated specific pathogen-free (SPF) chicken eggs as previously described [24] . Groups. SPF P2a line (B19/ B19) white Leghorn chickens eggs were obtained from Cornell University, Ithaca, NY. The eggs were hatched in the University of Connecticut Poultry Farm and after the hatch, the chickens were moved to the Office of Animal Research Services (OARS) facilities. After 2 weeks in the brooders with free access to water and a standard starter diet, the chickens were divided into groups, bled for baseline serology, transferred to isolators equipped with high-efficiency particulate air (HEPA) filters, and were provided commercial diets and water ad libitum. A previously described plaque reduction assay was modified and used to evaluate the virus neutralization activity of collected sera after vaccination [25] . Briefly, serum samples from each treatment group were pooled. An equal volume of a 1 : 10 dilution of pooled serum and LPAI subtype H5N2 was mixed and incubated for 30 min at 37 • C. A commercially available anti-M2 antibody (ProSci-Inc, Poway, CA) was used in a 1 : 1000 dilution as a control for antibody activity. Chicken embryo kidney cell (CEKC) monolayers in 6-well plates were washed twice with prewarmed PBS and 400 μL of the above mixture was added to the CEKC monolayer. The plates were incubated for 60 min at 37 • C. Then, the inoculums were removed and after 2 washes with prewarmed PBS, they were overlaid with 0.8% agar (University of Connecticut Cell Culture Facility) in Minimum Essential Medium Eagle (MEM). After 72 h, the plates were checked for plaque formation and for further evaluation were fixed with 99% methanol and stained with crystal violet for plaque counting. ELISA. The M2e epitopes, including the nanoparticle platforms with M2e epitopes (Tetra-M2e and Mono-M2e) and M2e linked to GCN4, (M2eN-GCN4), were used for coating of the ELISA plates. Briefly, individual wells of the flat-bottom 96-well Immulon 1B plates (NUNC/Thermo Fisher Scientific, Rochester, NY) were coated with 5 μg/mL of tetrameric M2e (M2eN-GCN4) or the nanoparticle of interest. Antigen adhesion was allowed to proceed at 4 • C overnight. Plates were rinsed with 2% Tween 20 in phosphate-buffered saline (PBS) (PBS/Tween 20 Ther-moFisher) and blocked with a 3% BSA in PBS solution. Plates were incubated at 37 • C for 3-4 h or 4 • C overnight (preliminary studies showed that there was no difference in the result). After incubation, plates were rinsed 4 times with PBS/Tween 20 and incubated for 1 h at room temperature with the previously collected sera. Briefly, 2-fold serial dilutions of each serum sample were prepared in a PBS solution containing 0.2 to 0.5% BSA. Hyperimmune serum from previously infected birds with the LPAI subtype H9N2 or commercial anti-M2e antibody were used as positive controls; and sera from healthy, unvaccinated birds were used as a negative control. After appropriate washes, peroxidaseconjugated goat anti-chicken IgY (Sigma Aldrich,) was prepared in a 1 : 10,000 dilution in PBS and was added to each well and plates were incubated for an additional hour at room temperature. After subsequent rinsing, the plates were developed using 3,3 ,5,5 tetramethylbenzidine (TMB) peroxidase substrate (Thermo Fisher Scientific Inc., Rockford, IL) followed by a room temperature incubation period of 15 to 30 min. The absorbance was read in a SpectraMax 250 microplate reader (Molecular Devices, Sunnyvale, CA) at 450 nm. In order to generate a standard curve for realtime PCR, we transcribed standard RNA in vitro using T7 RiboMAX Express Large-Scale RNA Production System (Promega, Madison, WI). Briefly, RNA extraction was done by using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer. The coding region of the M gene from the LPAI subtype H5N2 was amplified using previously described universal primers [26] . RT-PCR was performed using a Qiagen One-Step RT-PCR kit (Qiagen, Valencia, CA) according to the standard manufacturer's protocol. PCR products were visualized by electrophoresis through ethidium-bromide-stained (0.5 μg/mL) 1.5% (40 mM Tris-Acetate pH 7.8, 0.1 mM EDTA) agarose gels under UV light. The amplified fragment was excised from the gel and cDNA was recovered from agarose gel using a QIAquick Gel Extraction Kit (Qiagen, Valencia, CA) according to the manufacturer's protocol and the purified DNA product was ligated with a pCR 2.1 vector (Invitrogen, Carlsland, CA) according to the manufacturer's protocol to generate the pCR-M5 plasmid. For further confirmation, PCR positive plasmids were sequenced in the DNA Biotechnology Facility of the University of Connecticut. Six micrograms of plasmid DNA was linearized using 6 units of the restriction enzyme Bam HI for 4 h at 37 • C. Then, linearized DNA was used as a template in an in vitro transcription reaction with the T7 RiboMAX Express Large-Scale RNA Production System (Promega, Madison, WI) according to the manufacturer's recommendation. After the in vitro transcription reaction at 37 • C for 1 h, the possible remaining plasmid DNA was digested by DNase I and purified RNA was quantified with a spectrophotometer NanoDrop ND-1000 (Thermo Fisher Scientific, Wilmington, DE). The copy numbers of purified RNA were determined using a previously described method [27] and was used for the generation of a real time standard curve. In this study, real-time RT-PCR was performed using the previously published primers M+25: AGA TGA GTC TTC TAA CCG AGG TCG and M-124: TGC AAA AAC ATC TTC AAG TCT CTG for quantification of viral load [28] . RNA extraction was done on each swab sample followed by PCR in duplicate or triplicate using 5 μL of RNA per each PCR reaction. Briefly, the Power SYBR Green RNA-to-CT 1-Step Kit (Applied Biosystem, Foster City, CA) was used with a 20 μL reaction mixture. For each PCR run, standards were designated for the plate and viral loads were calculated using fluorescence data acquired at the end of each annealing step. The amount of unknown sample was extrapolated based on the standard curve and was reported as viral copy number. Prior to vaccination and challenge study, a pilot study was performed to evaluate pathogenicity Influenza Research and Treatment 5 Upon determination of the peak of virus shedding and appropriate infectious dose in the pilot study, the vaccination and challenge trial was initiated. Briefly, 42 SPF chickens were divided into six groups of seven and received their first inoculation at 2 weeks of age followed by two boosters, two weeks apart, at 10 weeks and 12 weeks after hatch as described in Table 2 . Preceding injection, nanoparticles were concentrated using Amicon Centrifugal Filter units with a 100 kDa MWCO (Millipore, Billerica, MA). Concentration was determined by absorbance at 280 nm and nanoparticle quality was assured by DLS. The nanoparticle vaccine constructs were emulsified with either Freund's complete adjuvant (prime) or Freund's incomplete adjuvant (boosters) and injected into the pectoral muscle of each chicken. Two weeks after the second booster, the birds, except for those in the negative control group, were challenged with 10 7.2 EID 50 LPAI subtype H5N2. Briefly, each bird received 1 mL allantoic fluid containing 10 7.2 EID 50 LPAI subtype H5N2 divided among the eyes, nasal cavity, and oropharynx, while temporarily blocking the fresh air delivery to the isolator. Fresh air was resumed after 5-10 min the following challenge of the last bird in the isolator. Although the clinical signs associated with LPAI viruses are rare, they were observed for possible clinical symptoms daily; and the presence of the symptoms and their severity was recorded. Tracheal and cloacal swabs were taken from each bird at days 2, 4, 6, and 8 after challenge and they were placed in a 3.0 mL UVT tube (Becton, Dickinson, NJ). Blood samples from each bird were collected before each booster as well as two weeks after the second booster prior to challenge. Each blood sample was collected in a separating blood tube and serum was separated by placing the tubes at 37 • C for 1 h then at room temperature overnight followed by a 5 to 10 min centrifugation at 1000 rpm at 4 • C. Then, the collected serum samples were stored at −20 • C until analysis. Design. An obvious model for a selfassembling protein particle is a viral capsid. The capsids of spherical viruses often have icosahedral symmetry, due to their need to build a large encapsulating structure from many copies of the same, or only few, capsid proteins. An icosahedron is the most efficient way to accomplish this. By utilizing pentameric and trimeric coiled coils, we have built a self-assembling nanoparticle which uses the threefold and fivefold symmetry of an icosahedrons [21] . The pentameric coiledcoil motif of the monomer is taken from Cartilage Oligomeric Matrix Protein (COMP) and the trimer is a de novo designed coiled coil. Self-assembly occurs when the coiled-coil domains of different monomers associate, forming the icosahedral nanoparticle ( Figure 1 ). A nanoparticle with this sort of architecture can then be used as a vaccine platform by extending the ends of the monomer with an epitope sequence. The Mono-M2e species of nanoparticle follows this plan ( Table 1) . As a result, it repetitively displays a monomeric form of M2e on the surface of the nanoparticles. The M2e peptide on the icosahedral nanoparticles lacks its C-terminal five residues to avoid problems with disulfide crosslinking that presumably require the native tetrameric conformation for proper formation. Although the simplest icosahedral particle with T1 icosahedral symmetry is made from 60 polypeptide chains, it may also be possible that Mono-M2e particles possess higher triangulation numbers, resulting in particles with an even greater molecular mass. On the other hand, the native conformation of M2 is a tetramer. Hence, to elicit conformationally specific antibodies, the M2e antigen displayed by a vaccine particle should ideally be tetrameric. With that in mind, we designed the Tetra-M2e peptide ( Table 1) . Instead of a pentameric coiled coil, this polypeptide uses the tetrameric coiledcoil motif from the protein tetrabrachion [29] . Self-assembly using this peptide would result in a nanoparticle with threefold and fourfold symmetry axes or octahedral symmetry. As opposed to the larger icosahedral Mono-M2e, this octahedral particle would only have 24 polypeptide chains. In addition, the epitope is now constrained to its native tetrameric conformation. The full-length M2e contains two cysteine residues. The formation of disulfide bridges between the cysteines of adjacent chains under oxidizing conditions is thought to stabilize their tetrameric conformation. The speed and ease of protein expression and purification, as well as of the self-assembly process, contribute to the overall viability of this technology as a vaccine platform. To facilitate purification, we have included polyhistidine tags at the N-terminal ends of the peptides. To enable detection of antibodies against tetrameric M2e, the peptide M2eN-GCN4 was designed (Table 1 ). M2e is linked to a GCN4, a coiled coil whose oligomerization state can be determined by the identity of amino acid residues in key a and d positions of the coiled coil. In this case, the tetrameric version of GCN4 was used [30] . By affixing M2e to a tetrameric protein, we can constrain it in its tetrameric conformation. The effect is similar to that experienced by the ends of the tetrameric coiled coil from tetrabrachion of the Tetra-M2e nanoparticle. However, the coiled coil sequence is different. This guarantees that any antibodies bound to M2e-GCN4 are specific for the tetrameric version of M2e and not against the coiled coil or other parts of the nanoparticle. Mono-M2e formed particles whose hydrodynamic diameters have a distribution which peaks at 34.5 nm, while the distribution of Tetra-M2e peaks at 22.9 nm (Figure 2 ). It is also noteworthy that the size distribution peak of Mono-M2e is broader than that of Tetra-M2e, suggesting that the former has a higher degree of polydispersity. The results were confirmed by transmission electron microscopy ( Figure 3) . We can see that nanoparticles were formed and that their diameters are comparable with those measured by dynamic light scattering. It can be seen from the micrographs that neither Mono-M2e nor Tetra-M2e form nanoparticles with perfectly spherical morphology. This may in some way explain the polydispersity observed by dynamic light scattering. Structure. The double minima found by circular dichroism confirm the alpha helical structure of the nanoparticles (Figure 4 ). It appears that Tetra-M2e exhibits this behavior much less than Mono-M2e. This may be partly due to the larger M2e epitope sequence in the Tetra-M2e peptide as compared to that used in Mono-M2e. The plaque reduction assay performed by using pooled serum from chickens inoculated with Tetra-M2e did not show a significant (P > 0.05) difference compared to control nonvaccinated chicken serum and commercial anti-M2e antibody. The anti-M2e immune response was monitored by determining the titer of the M2e-specific IgY at three different time points (2 weeks after each inoculation). Chickens after each inoculation developed high levels of antibody against the injected construct and anamnestic response clearly was seen when the plates were coated with Mono-M2e and Tetra-M2e nanoparticles and M2e-GCN4 (tetrameric M2e), respectively (Table 1, Figures 7 and 8) . For further investigation of the antibodies, the level of M2e specific antibody was measured using plates coated with tetrameric M2e-GNC4 peptide to evaluate the specific antibody against tetrameric M2e rather than the whole particle. The result of this study indicated that in chickens, after the second booster, the antibody levels are not at the same level as our previous results in mice with the same backbone but a different (malaria) epitope had been shown [23] . The dose level was also higher than what was shown to be required in mice. This could be because of the lower haplotype-specific immunogenicity of the particles in chickens, the route of administration in mice (intraperitoneal and intranasally), the body weight of the mice compared with chickens, and different immune system repertoires of mammalian and avian species. In future studies, changing the administration route can be another approach to reducing the dose of vaccine construct. We also coated the plate with inactivated purified virus to observe seroconversion of the chickens after challenge with the virus at 2 weeks after the last boost. Results indicated that whole virus response was higher as expected with hyperimmune serum (Figure 9 (a)), however, ELISA response from chicken vaccinated with tetra-M2e and with whole virus reacted similarly on GCN-M2e coated plate (Figure 9(b) ). The protective efficacy of the anti-M2e antibody responses induced by different constructs was assessed by evaluation of viral shedding post challenge. To determine the peak of shedding, viral copy number was Figure 10 . We determined viral loads in tracheal and cloacal swabs samples on day 8 following challenge with 10 7.2 EID 50 LPAI subtype H5N2. Reduction of cloacal and oropharyngeal shedding in vaccinated birds was significant in chickens vaccinated with Tetra-M2e with Freund's adjuvant. Virus shedding was evaluated at day 4 and day 6 after challenge; the swabs were tested for virus load ( Figure 11 ). It is seen that virus shedding reduction starts at day 4 post infection with a significant decrease at day 8 post infection. Currently available vaccines induce antibodies against specific field strains or closely related avian influenza strains. Most of these vaccines are killed virus vaccines that induce short-lived immunity and are lacking a broad cross-reactive humoral immune response. Recently, the generation of a universal influenza vaccine using conserved peptide regions among several influenza virus strains has been an area of interest in the human influenza vaccine field. M2e is a highly conserved region among influenza viruses and it has been studied as a possible universal vaccine candidate against human influenza virus infection [16, 17] . In the present study, protection efficiency of two different nanoparticle constructs harboring M2e was studied as possible vaccine candidates for low-pathogenicity avian influenza infection. Biophysical analysis confirms that they are of relatively regular shape and size, but there is some degree of heterogeneity. Though molecular weight measurements still remain to be carried out and we have no high resolution Influenza Research and Treatment structural data, we assumed that nanoparticles assembled in a state close to what was expected, that is, icosahedral and octahedral nanoparticles, respectively. This will repetitively display M2e in both, either in its monomeric or its tetrameric form. There is speculation as to how the polyhistidine tag at the N-terminal end of the monomers may affect the selfassembly process, the final nanoparticle structure, or the immunogenicity of the vaccine, but attempts at producing his-tag free versions of the nanoparticle constructs either did not reliably provide pure protein or never adequately self-assembled. Similarly, we attempted to include CD4 T cell epitopes to increase the immune response, but this also interfered with nanoparticle formation. Since many variables can affect the host's virus shedding and the course of disease [31, 32] , prior to evaluation of vaccine constructs, the pathogenicity after challenge with the LPAI subtype H5N2 virus was evaluated. A biphasic virus shedding was observed in this study. For the LPAI subtype H5N2, the peaks for tracheal and cloacal shedding were at days 4 and 8 postinfection. The Tetra-M2e vaccine construct provided a significant viral load reduction at the peak of viral shedding in immunized chickens. Chickens immunized with Tetra-M2e that harbors the tetrameric M2e with Freund's adjuvant showed a clear reduction in cloacal and tracheal excretion of LPAI compared to challenge control groups. The results of immunization with Mono-M2e with adjuvant and Tetra-M2e without adjuvant were also promising and by improving both B-and T-cell epitopes of those constructs, desirable results may be obtained. In vaccine design, repetitive Bcell epitope display is considered a strategy for improving the humoral immune response [33, 34] . In addition to repetitive antigen display on the nanoparticle, we were able to present M2e in its native tetrameric conformation. The correlation between high protection and antibody response specific for tetrameric M2e elicited by Tetra-M2e supports our assumption of tetrameric M2e presentation. The results of our studies show that tetrameric M2e stimulates a more specific immune response compared to the monomeric presentation and induces a significant protection against homologous virus challenge. The fact that a large portion of the antibody response is directed against the carrier and not only against the epitope(s) (Figures 7 and 8) can be explained by the fact that significant portions of the core of the nanoparticles are also exposed to the immune system (compare Figure 1) and hence, these portions will also induce a significant immune response. In this study, we showed that anti-M2e antibodies are not neutralizing antibodies; however they are capable of binding to the M2 proteins that are abundantly presented on the surface of the infected cells (data not shown). These can describe an efficient delayed clearance of the virus in M2e vaccinated chickens based on the previously described NK cell involvement in ADCC [35] . Significant improvement of virus clearance in vaccinated chickens with tetrameric M2e may be considered in a new vaccination strategy by vaccinating chickens with both a killed vaccine and a nanoparticle vaccine in order to provide robust protection, cross-reactive immunity, and clearance in case of emerging new strains of the virus. However, there remains the risk that such vaccination may cause a long-term persistence of HPAI in poultry flocks, because the vaccine could not prevent the viral infection but rather suppresses the symptoms of HPAI virus-infected chickens by reducing the virus shedding in chicken. Thereby, especially in the case of HPAI infection, the vaccination may make the infection less visible and the eradication of virus more difficult, and consequently it may provide a good opportunity for HPAI virus to survive and persist in poultry flocks for a long time. For this reason, we plan to design new nanoparticle constructs that also contain fragments of hemagglutinin in addition to the M2e domain. Immunization would then result in the generation of neutralizing hemagglutinin-specific antibodies in addition to the disease modulating M2e-specific antibodies. In this study, we evaluated a new approach to immunizing chickens against AI that uses a nanoparticle platform to carry an antigenic epitope. Further designing and testing of new nanoparticle vaccines should demonstrate that they are effective tools for stimulation of an immune response against M2e and other B-or T-cell epitopes. Therefore, application of the nanoparticle platform facilitates the development of a new generation of vaccines that harbor conserved epitopes of avian influenza viruses and would not be rendered ineffective by viral mutations such as antigenic shifts and drifts. The nanotechnology described here offers the opportunity to rapidly produce new vaccines according to the emergence of new strains of influenza virus without going through the time-consuming steps of production currently used in manufacturing commercial influenza vaccines. For future studies, the chicken's LPAI infection model needs to be improved to evaluate clinical signs and higher virus shedding. This may help to better evaluate virus shedding, specifically cloacal virus shedding. Also, vaccination and HPAI challenge may be used to evaluate the vaccine efficiency in protection against high-pathogenicity AI viruses. Avian influenza M2e: Ectodomain of matrix protein 2 LPAI: Low pathogenicity avian influenza HPAI: High pathogenicity avian influenza SPF: Specific pathogen free.

Factors Associated with Increased Risk Perception of Pandemic Influenza in Australia

The aim of this study was to assess factors associated with increased risk perception of pandemic influenza in Australia. The sample consisted of 2081 Australian adults aged 16 years and older who completed a short three item pandemic influenza question module which was incorporated into the NSW Health Adult Population Health Survey during the first quarter of 2007. After adjusting for covariates, multivariate analysis indicated that those living in rural regions were significantly more likely to perceive a high risk that a pandemic influenza would occur, while those with poor self-rated health perceived both a high likelihood of pandemic and high concern that self/family would be directly affected were such an event to occur. Those who spoke a language other than English at home and those on low incomes and younger people (16–24 years) were significantly more likely to have changed the way they lived their lives due to the possibility of pandemic influenza, compared to those who spoke only English at home, middle-high income earners, and older age groups, respectively. This data provides an Australian population baseline against which the risk perceptions of demographic subgroups regarding the current, and potential future pandemics, can be compared and monitored.

The pattern of recurrence of pandemics since the mideighteenth century indicates that pandemics occur about every 30 years [1] . Prior to 2009, expert consensus was that another pandemic influenza was almost inevitable [2] [3] [4] [5] [6] [7] , and although the H5N1 avian viruses were the most likely candidate for an influenza outbreak, the unexpected H1N1 swine influenza reached pandemic in June, 2009. With previous influenza pandemics and the current H1N1 influenza pandemic arriving with little to no warning, we are afforded a unique opportunity to prepare for the next pandemic threat, which has the potential to be more severe than the current pandemic. Important for preparation is knowledge about the public's response to such a threat, and a key component to the public's response is their perception of risk. Knowing how a risk is perceived is essential for preparing an effective plan for risk communication, and may be predictive of the public's response. In a study of the NSW population, Barr et al. [8] found that respondents with higher levels of risk perception reported more willingness to comply with public health behaviours in the event of an outbreak of influenza. Similar results were found in Hong Kong [9] and Italy [10] , where respondents in both studies with an increased perception of risk were more likely to be engaged in risk-reducing behaviours. In 2007, 14.9% of the NSW population reported that they thought pandemic influenza was very or extremely likely to occur and 45.5% were very or extremely concerned that they or their family would be affected by an influenza pandemic should it occur [8] . What may be of particular importance however, is how risk perception varies within the population. Risk perception may be affected by factors such as awareness of a hazard, cultural and social factors or the experience or memory of a prior similar hazard, all of which may result in variation in risk perception among individuals. Lau et al. [11] found in a Hong Kong sample that the odds of females reporting worry about themselves or their families contracting an outbreak of avian influenza if it is to occurr were 1.6 times higher than the odds of males reporting such worry. De Zwart et al. [12] similarly found that women and older respondents scored significantly higher on a composite measure of risk perception (combining perceived seriousness of threat and vulnerability to threat) than men and younger respondents, respectively. In an Italian population, Di Giuseppe et al. [10] found that risk perception was higher for respondents with lower socioeconomic status and lower education. In preparation for a pandemic influenza outbreak, the Australian Government recommends a number of measures the general public could take, such as having enough food, water, and essential items to enable a household to be confined at home for up to 14 days [13] , ensuring such food is rotated and use by dates are checked regularly [13] ; practicing good personal hygiene, and teaching children about hand washing and cough etiquette [14] .The World Health Organisation has also recommended seasonal influenza vaccinations for health care workers to reduce the risk of genetic shifts in the influenza virus [15] .The preparation of the general public for an outbreak of influenza may be a key strategy in preventing the spread of the disease in the event of a pandemic. Thus it is important to identify subpopulations in Australia who are more and less likely to have changed their life in response to the possibility of pandemic influenza. The aim of the current study was to obtain baseline Australian data on factors associated with perceptions of the likelihood of pandemic influenza, concern for self and family in the event of an influenza pandemic and broad changes in living as a result of the threat of pandemic influenza. A short three item pandemic influenza question module was developed as the first part of a larger module of questions on potential threats. These questions were field tested and inserted into the New South Wales Population Health Survey, administered between 22 January and 31 March, 2007 [8] . The New South Wales Population Health Survey is a continuous telephone survey including questions on health behaviours, health status, and access to health services of the state population using the in-house CATI facility of the New South Wales Department of Health [16] . Households were contacted using random digit dialing. Up to 7 calls were made to establish initial contact with a household, and 5 calls were made in order to contact a selected respondent. Only residential phone numbers were used in the sample, as residential phone coverage in Australia still remains high [17] and results from persons who only have mobile phones has been shown to be comparable in the United States [18, 19] . Interviews were conducted in English, Arabic, Chinese, Greek, Italian, or Vietnamese, depending on respondent preference. More details of the sampling approach can be found in the 2007 NSW Health survey report [20] . A three item pandemic influenza question module was developed which addressed pandemic influenza threat perceptions. The wording of the questions was as follows: (1) How likely do you think it is that pandemic influenza will occur in Australia? (2) If a pandemic influenza were to occur in Australia, how concerned would you be that you or your family would be affected by it? (3) How much have you changed the way you live your life because of the possibility of an influenza pandemic? All responses were coded on a five-point Likert scale. Response options for all questions were "not at all", "a little", "moderately", "very", and "extremely". In addition, "do not know" and "refused" responses were coded as missing. Analysis. Data analysis was performed using the "SVY" commands of Stata version 9.2 (Stata Corp, College Station, TX, USA), which allowed for adjustments for sampling weights. The five-point Likert-scale responses were dichotomised. The definitions of the variables used are as follows: (1) Pandemic influenza likely to occur: the proportion of households aged 16 years and older who rated pandemic influenza as very or extremely likely to occur. (2) Concern for self/family: the proportion of households aged 16 years and older who were very or extremely concerned that self/family would be directly affected if pandemic influenza were to occur. To determine factors associated with risk perception, the dichotomized risk question indicators and the "composite" indicators were used as outcome measures and these were investigated using the following set of independent variables: age, gender, marital status, children in household, location (urban/rural) as defined by respondents' area health region, born in Australia, speaking a language other than English at home, highest level of formal education, household income, living alone, self-rated health status, and psychological distress. Self-rated health status was assessed with the question "Overall, how would you rate your health during the past 4 weeks?" with possible responses being "excellent", "very good", "good", "fair", "poor", and "very poor". Responses of "very good" and "good" were combined and reported as "good" self-rated health, and responses of "poor" and Influenza Research and Treatment 3 "very poor" were combined and reported as "poor" selfrated health. Psychological distress was assessed using the Kessler 10 measure (K10). The K10 provides a measure of nonspecific psychological distress. Questions in the K10 include "In the past 4 weeks about how often did you feel . . ." "tired out for no good reason", "nervous", "so nervous that nothing could calm you down", "hopeless", "restless of fidgety", "so restless you could not sit still", "depressed", "that everything was an effort", "so sad that nothing could cheer you up", and "worthless". Possible responses were "all of the time", "most of the time", "some of the time", "a little of the time", and "none of the time". The K10 provides a score ranging from 10-50. For the current analysis a score below 22 was considered as low-psychological distress, and a score of 22 or above was considered as high-psychological distress. Multiple survey logistic regression using stepwise backwards model was used in order to identify the factors significantly associated with risk perception. All variables with statistical significance of P ≤ .05 were retained in the final model. In total, 2081 state residents aged 16 and over completed the module on pandemic influenza. The overall response rate was 65%. The key demographics of the weighted survey were comparable to Australian Bureau of Statistics (ABS) 2006 Australian population census data [21] . Multiple survey logistic regression analyses using a backward stepwise method were performed for the nine outcome variables. Table 1 shows that Australian households who lived in rural areas were significantly more likely to think that pandemic influenza was very or extremely likely to occur than those in urban region (AOR = 1.59 (95% CI: 1.02-2.49, P = .041)). Respondents with poor self-rated health were also significantly more likely to think that pandemic influenza was very or extremely likely to occur, compared to those with good self-rated health (AOR = 1.92 (95% CI: 1.12-3.31, P = .018)) and were also more likely to report being very or extremely concerned that self or family would be directly affected if a pandemic was to occur (AOR = 1.64 (95% CI: 1.09-2.47, P = .017). Those from low income households, those who spoke a language other than English, and young people (16-24 years) were more likely to have changed the way they lived their lives because of the possibility of pandemic influenza, compared to their respective reference groups. Table 1 also shows that respondents who lived in rural areas and respondents who reported poor self-rated health were significantly more likely to report combined indicator (1) than those who lived in urban areas and those with good self-rated health. The odds of respondents with highpsychological distress reporting combined indicator (2) were 3.03 (AOR = 3.03) times higher than the odds of respondents with low-psychological distress levels reporting combined indicator (2). The aim of this study was to assess factors associated with increased risk perception of pandemic influenza in Australia, increased concern for self and family if a pandemic influenza were to occur in Australia and associated changes in living due to the threat of such an event. Particular strengths of this study are the population-based sampling method and appropriate adjustment for sampling weight to reflect the population of interest. Generally, pandemic influenza was not regarded as a high threat by NSW residents, with only 14.9% reporting that they felt pandemic influenza was very or extremely likely to occur. Those living in rural areas and those with poor self-rated health were more likely to report pandemic influenza very or extremely likely to occur compared with those living in urban areas and those with good self-rated health, respectively. Although not regarded as a high threat by Australians, 45.5% of respondents said they would be very or extremely concerned for self and family in the event of a pandemic influenza. Respondents with poor self-rated health were more likely to report more concern for self and family if an influenza pandemic occurred as compared to respondents with good self-rated health. These results are dissimilar to those in prior studies. Although the studies of Lau et al. [11] and De Zwart et al. [12] found that females scored higher on risk perception than males, gender was not a significant risk factor for high-perceived pandemic likelihood or concern for self and family in this study. Similarly, although in prior studies older respondents [12] , those with lower socioeconomic status [10] and lower education [10] reported significantly higher risk perception, in the current study none of these were risk factors for high-perceived pandemic likelihood or concern for self and family. What is common to all these groups is that they represent the groups typically most vulnerable to concern due to a focal threat. The prior studies were conducted on populations responding to a tangible threat as they were conducted around either the time of the avian influenza outbreaks or on populations which were most affected by the avian influenza or the SARS outbreaks. As this study was conducted on an Australian population which was not directly affected by avian influenza or SARS and where influenza was not a media focus, the null findings in this study may indicate that the threat of pandemic influenza was so general and distal that it did not have the capacity to concern even the portion of the population normally most sensitive to threat. It is not surprising that individuals with poor self-rated health reported greater risk perception and concern for self and others than individuals with good self-rated health. The health concerns of these individuals may lead them to believe they are more susceptible to infection or complications which may occur with an outbreak of pandemic influenza. Also, since poor self-rated health has been associated with increased levels of anxiety [22, 23] and distress [24] , there may be a heightened focus on, concern about, and belief in the likelihood of major external threats such as a pandemic influenza would represent. Such anxiety and distress may 4 Influenza Research and Treatment also lead these individuals to be more concerned about others as well as themselves in the event of a pandemic influenza. The potential link between self-rated health and heightened risk perception and concern for self and other in the event of a risky external event such as pandemic influenza warrants further examination. We can only speculate as to why individuals living in rural areas believed pandemic influenza was more likely to occur than those living in urban areas. Perhaps individuals living in rural areas are more broadly aware of disease transmission and its health and economic consequences, including the possibility of influenza transmission from animals to human. However, though individuals living in rural areas believed that pandemic influenza was more likely to occur than individuals living in urban areas, they did not display more concern for self and family should a pandemic influenza occur. This is not unexpected given that the influenza virus is more easily transmitted from person to person in crowded environments, and that rural environments are typically not densely populated. It might be expected however that individuals living in urban environments may be more concerned for self and family in the case of a pandemic influenza, as urban environments are typically crowded. This, however, was not the case in this study. Similarly, it is of particular note that in this study concern for self and family did not increase when there were children or elderly in the household, despite individuals in these age groups being particularly vulnerable to influenza morbidity and mortality. As suggested above, these null results might reflect that pandemic influenza is too distal a threat for concern for the whole Australian population. Perhaps higher levels of perceived likelihood and concern for self and family in the context of a specific imminent threat (e.g., swine flu) are required for significant group differentiation to emerge. Generally, a minority of people had changed the way they live their life because of the possibility of pandemic influenza, with only 23.8% reporting they had changed their life at all. This is not surprising as the current data also indicate that few Australians believed pandemic influenza was likely to occur. Households which had a lower income, households which spoke a language other than English and those respondents aged between 16 and 24 were more likely to have changed the way they lived their life because of the possibility of a pandemic influenza than those households with middle-high income, those who only spoke English and those older than 24 years, respectively. Further investigation into specific actions people take to change their lives in response to the threat of a pandemic influenza may provide useful information. Interestingly, all of the factors associated with living changes in the case of a pandemic influenza are independent of the factors associated with perceived likelihood and concern for self and family in the case of pandemic influenza. That is, these groups are reporting living changes in the absence of heightened perceived threat or concern relative to the remainder of the population. This suggests that these groups may not be changing their way of life because they feel pandemic influenza is more likely than the remainder of the population, or because they feel themselves or their families are particularly vulnerable should pandemic influenza occur, but for some other reason. It is possible that these results may be due to methodological issues. These groups (lower income, language other than English, and younger respondents) are somewhat marginalized groups, that we would expect to have a higher threat perception and concern. The results therefore may be due to the broad nature of the question, which may have tapped into a more general and pervasive sense of threat vulnerability within the community, such as upswings in terrorism, war, and climate change, which may have been felt more strongly in more exposed or vulnerable groups. Similarly, the response set for this question was extremely broad compared to the remaining questions. A respondent was considered to have changed his way of life if he reported to have done so a "little", "moderately", "very", or "extremely". This is in contrast to the perception and concern questions where only responses of "very" and "extremely" were included in analyses. Changed life a "little" could be interpreted by some respondents as an increased feeling of threat which represents a change in effect rather than in behaviour. Even though individuals with poor self-rated health believed that pandemic influenza was more likely to occur and felt more concern for self and family in the case of a pandemic influenza than those with good self-rated health, they were not more likely to report changing their life as a result of the possibility of pandemic influenza. Again, these populations may be responding to a more distal than proximal threat. Though they report some concern, it may not have been of the extent to prompt actual changes in way of living. More practically, neither the government nor the media were concerned at this time with promoting that the population makes life changes as a result of the threat of pandemic influenza, nor what these actions should be. In a previous study, the role of concern for self and family was a key factor associated with likely compliance with protective health behaviours [25] . This suggests the benefit of risk communication messages that strategically heighten and then utilise public concern when a pandemic has or is likely to occur to increase compliance behaviours. For example, risk communication strategies could selectively target sub-population for whom risk beliefs are particularly low; in the current study these groups are urban populations and populations with good self-rated health. However, some authors have cautioned that increasing the risk perception of the population through such strategies risks societal estrangement and may frighten health care workers, first responders, and those who would have contact with the public in the event of a pandemic [26] . When consensus is reached regarding the optimal level of risk perception required for specific populations to elicit appropriate protective responses, the results of this study may be useful to guide which population groups these artificially inflating or deflating risk communication messages should be targeted. It is likely that due to the recent H1N1 swine influenza pandemic that the current risk perceptions of the population are significantly different to those reported in this paper. Further research could examine changes in risk perceptions following this current pandemic for the whole population as well as the subpopulations examined in this paper. As such, the results of this paper provide a baseline measure for which future studies on risk perceptions can be compared. The population may have also recently made changes in daily living as a result of the H1N1 pandemic, as information on preventative measures such as personal hygiene have featured prominently in social marketing messages since its outbreak. This represents a response to a pandemic rather than a preventative measure for a potential pandemic. However, it would be important to determine which subpopulations maintain key behaviours (e.g., sneeze etiquette) following the end of the current pandemic, as this information can assist in the prevention of future pandemic threats. With the outbreak of the current H1N1 swine influenza, research has emerged which has reexamined pandemic influenza attitudes and reactions in the Australian population. In research conducted during the World Health Organization (WHO) Phase 5 of the swine flu pandemic (between 2 May and 29 May, 2009) [27] , 21% of a Sydney-based sample ranked their risk of catching pandemic influenza as high. The same authors conducted a similar survey also on a Sydney-based population during the WHO Phase 6 of the swine flu pandemic (between September and October, 2009) and found that 17.4% believed they had a high to very high risk of acquiring H1N1 influenza [28] . In a CATI survey conducted between August and September, 2009, in an Australian nationally representative sample, Eastwood et al. [29] found that of the respondents, 5% were extremely concerned and 17% were quite concerned that they or a member of their family would contract swine influenza. Although in the latter study consideration was given to the reasons for concern (e.g., close family member/friend in high-risk group, having an underlying illness, being employed in a position with high public contact), none of these studies examined factors associated with increased perceived risk for acquiring swine H1N1 influenza. These results provide interesting comparison to the results of the current study. When the likelihood of pandemic influenza occurring was generally considered to be low, 45.5% of the population reported they would be concerned for themselves and their family should it occur. However, in the midst of the current pandemic, individuals perceived less risk to the self which has decreased as the influenza pandemic has progressed [27, 28] , and with as few as 22% [29] of the population reporting concern that they or a family member would contract the virus. Lastly, consideration should be given to the limitations of the current study. The main limitation is that the study was conducted using telephone interviews which may have introduced selection bias. However, residential phone coverage in Australia remains high [17] , and a large number of studies on SARS and avian influenza have utilized this method. Also, risk perception and protective behaviours are likely to be mediated by a number of factors in addition to those identified in this study. Factors such as anxiety, risk perception of others, media, and recent events such as the current swine flu pandemic are all factors likely to affect risk perception. However, despite these limitations the results of this study suggest that it may be appropriate to direct risk communication strategies to individuals living in urban populations and individuals with good self-rated health, which may result in an increased likelihood of appropriate protective responses if an influenza pandemic was to occur. Data in this study further suggest that in contexts where pandemic influenza is generally not regarded as a high threat by the population, messages highlighting actions individuals can take to prepare for a pandemic influenza should be directed to households with higher incomes, households which do not speak a language other than English, and individuals above the age of 24 years. Finally, data from this study provide an Australian population baseline against which factors associated with risk perception related to outbreaks of pandemic influenza, both current and future, can be compared.

Blow Flies Were One of the Possible Candidates for Transmission of Highly Pathogenic H5N1 Avian Influenza Virus during the 2004 Outbreaks in Japan

The 2003-2004 H5N1 highly pathogenic avian influenza (HPAI) outbreaks in Japan were the first such outbreaks in 79 years in Japan. Epidemic outbreaks have been occurring in Southeast Asia, with the most recent in 2010. Knowledge of the transmission route responsible for the HPAI outbreaks in these countries remains elusive. Our studies strongly suggested that field and laboratory studies focusing on mechanical transmission by blow flies should be considered to control H5N1 avian influenza outbreaks, in particular in epidemic areas, where there are high densities of different fly species throughout the year. In this paper, we review these field and laboratory entomological studies and discuss the possibility of blow flies transmitting H5N1 viruses.

The H5N1 subtype of highly pathogenic avian influenza (HPAI) A virus has frequently infected wild and domestic ducks in Asia, causing huge economic damage to both poultry farms and governments in the affected countries. Most avian influenza viruses do not infect humans, but the 1997 outbreak of the H5N1 virus in Hong Kong [1, 2] alerted the medical community that some subtypes of avian influenza viruses include highly pathogenic strains that can affect humans. In this influenza virus outbreak, there were 6 deaths in the 18 human cases caused by the H5N1 subtype [3] . As of August 2, 2010, WHO has identified 502 human cases of H5N1 influenza around the world, and 298 of these were fatal [4] . In particular, H5N1 outbreaks have occurred recently in Egypt, Indonesia, and Vietnam. Therefore, H5N1 influenza virus can cause serious public health problems in birds and humans and is one of the most infectious avian diseases transmissible to humans. From January 2004 to March 2004, there were outbreaks of acute, highly transmissible, lethal diseases in chickens at four poultry farms in Japan: one in Oita, one in Yamaguchi, and two in Kyoto Prefecture ( Figure 1 ). Virus isolates from infected chickens were all identified as influenza A virus of the H5N1 subtype [5] . Such highly pathogenic avian influenza (HPAI) epidemics had not been reported in Japan for 79 years. Two avian influenza outbreaks at poultry farms in Tamba Town, Kyoto Prefecture, were the last two outbreaks of the 2004 avian epidemics in Japan. Since then, there were outbreaks of H5N1 avian influenza in Okayama and Miyazaki Prefectures in 2007. The H5N1 virus was also isolated from dead Whooper swans, Cygnus cygnus, in 2008 in Towada Lake, Akita Prefecture [6] . In addition, outbreaks of other subtypes of avian influenza virus have frequently occurred in Japan. The H5N2 avian influenza was reported in Ibaraki and Saitama Prefectures in 2005 and 2006, the H3 subtype was reported in Saitama Prefecture in 2009, and the H7 subtype was in Aichi Prefecture in 2009. We know of no report suggesting that H5N1 virus could be transmitted efficaciously from person to person, but the possibility remains that such transmission could evolve [7] . Tamba Town (35 • 9 42 N and 135 • 26 31 E) is located in a hilly area 150-300 m above sea level, 50 km northwest of Kyoto City, Japan ( Figure 2 Knowledge of the transmission route responsible for the HPAI outbreaks in Southeast and East Asian countries still remains elusive [5, 9] . Four hypotheses have been suggested for transmission of H5N1 in the HPAI outbreaks in Japan [6] : (1) H5N1-virus-infected chickens may have been imported from other countries, (2) materials (e.g., vehicles and egg containers) from infected area may have been used, (3) viruses may have been carried on clothes, boots, hands, and so forth, and (4) infected wild birds may have carried H5N1 virus into poultry farms to infect chickens. In particular, it has been suggested that migratory birds carried the viruses and subsequently infected domestic and/or wild ducks [9] . HPAI outbreaks in Japan. Therefore, it is very likely that the 2004 epidemics H5N1 virus was transmitted to Japan from East Asian countries. The important question is how could the H5N1 virus be transmitted from virus-positive migratory wild birds to domestic poultry in Japan? We have noted that it is unlikely that wild birds directly transmitted influenza viruses to poultry in Japan, because all of the Japanese poultry farms, where H5N1 virus outbreaks occurred, had fowling nets in place to prevent entry of wild birds. However, flying insects (e.g., flies) can easily get through the nets and invade a poultry farm. We have shown that a chicken can eat all 31 blow flies put inside its cage in just 7 min [10] . A chicken can catch and break down the body of a fly and even catch and swallow a fly in flight. Therefore, we have been interested in whether, if a chicken eats blow flies carrying H5N1 virus, the chickens might become infected and develop symptoms of H5N1 influenza. Although the spring season in Japan is cold, some fly species are present. However, no studies have been reported on the possible role of flies in transmission of H5N1 influenza virus. Therefore, an entomological survey was conducted in March 2004 to investigate the possibility of blow flies transmitting H5N1 virus, using flies collected from around the infected poultry farm in Tamba Town for virus detection and isolation. Blow fly collection was carried out on 10-11 March 2004, just after the H5N1 outbreak at poultry farm B [11] . A sunny place protected from strong wind was selected, and rotten fish bait was placed on the ground. A total of 926 flies were collected within a 2.3 km radius of poultry farm A in Tamba Town ( Figure 2 ), representing eight fly species with >80% of the collected flies identified as either Calliphora nigribarbis Vollenhoven or Aldrichina grahami (Aldrich) (Figure 3 ). Influenza A virus matrix protein (M) and hemagglutinin (HA) genes were detected in the intestinal organs, crop, and gut of C. nigribarbis and A. grahami by reverse transcriptionpolymerase chain reaction (RT-PCR) [11] . The prevalence of H5 subtype virus (20-30%) was higher in flies of both species collected 600-700 m from poultry farm A and lower (10%) in flies collected >2 km from poultry farm A. We found that nearly 5% of C. nigribarbis collected around the affected areas contained infectious H5N1 viruses. viral M, HA, and NA genes were amplified by PCR with universal primers, full-length sequences were analyzed, and sizes were found to be 991, 1,707, and 1,362 bp, respectively. These sequences showed high similarity to those of strains from chickens (A/chicken/Kyoto/3/2004) and crows (A/crows/Kyoto/53/2004) isolated during the 2004 outbreaks in Kyoto, with >99.9% identity for all three genes. The virus from C. nigribarbis (A/blow fly/Kyoto/93/2004) was characterized as an H5N1 subtype influenza A virus based on neuraminidase gene (NA) sequences. In addition, the HA1-HA2 connecting peptide sequence in the HA gene segment was RERRRKKR↓G. Finally, virus isolated from C. nigribarbis was characterized as a highly pathogenic H5N1 subtype influenza A virus. To investigate whether H5N1 virus could survive in the blow fly C. nigribarbis, we monitored the titer of infectious virus in flies after they were exposed to the H5N1 avian influenza A virus (A/duck/Hyogo/35/2001) [10] . Fifty female blow flies (Kyoto strain C. nigribarbis), approximately 14 days old, were put into a 20 cm 3 fly cage for 3 h at 20 • C with a piece of cotton impregnated with 10 8 EID 50 /mL allantoic fluid from an H5N1 virus (A/duck/Hyogo/35/2001 [10] )-infected egg diluted with MEM diluents [11] . Following the 3 h virus exposure, the blow flies were individually reared at 20 • C or 10 • C until tested. Incubation at 10 • C was chosen because the average temperature around Tamba Town was 3.6 • C in February and 6.4 • C in March, with average daytime highs of 11.1 • C in February and 13.1 • C in March [12] . Crops and intestines dissected from flies' bodies at various times after virus exposure were used for virus isolation and titration. Virus was isolated from fly crops and intestines up to 24 h after exposure and from feces and vomit matter of 1 of 3 blow flies at 48 h after exposure (Table 1) Two species of the blow fly, C. nigribarbis and A. graham, are categorized as larger-sized fly species in particular in comparison to the house fly Musca domestica (L.); its body size is 5-8 mm (Figure 2 ). The body length of female C. nigribarbis is 11-15 mm and approximately 1.5 times larger than that of female A. grahami (8-13 mm) . The capacity of the crop of female C. nigribarbis (average = 23 mL) is approximately five times greater than that of female A. grahami (average = 4.4 mL) [11] . The consumption rate of both C. nigribarbis and A. grahami might have been high because of their large body size. In fact, virus genes were found more often in C. nigribarbis than in A. grahami [11] . Stable flies, Muscina stabulans (Falle'n) and M. angustifrons (Loew), collected at the same collection sites and the same time as the fly surveillance in Kyoto, showed much smaller body size than C. nigribarbis and A. grahami, and no virus was detected in these smaller-sized flies [11] . Blow flies prefer to lick animal carcasses and droppings. If food for blow flies is contaminated by pathogens, the flies might ingest significant numbers of pathogens. One possible mechanism for mechanical transmission of pathogens by blow flies is regurgitation and the feces on the food source [13, 14] . The effectiveness of mechanical transmission through regurgitation may depend on the viability and titer of pathogens in the fly's body. The accumulated droppings at a poultry farm should be a good breeding site for blow flies. If the flies reproduced at a poultry farm, they should have many opportunities for contact with viruses in the feces of infected chickens and/or their dead bodies. Calliphora nigribarbis has a characteristic temperate-zone life cycle. For example, in Japan, they become more active between winter and spring for migration and reproduction [15, 16] . It is well known that C. nigribarbis has excellent flight capacity and high dispersal ability. They have been identified by weather ships at stations located on the Pacific Ocean and East China Sea, 300-450 km from Kyushu Island, Japan [17] . It was also suggested that the number of flies found in autumn in the Kyushu District appears to increase due to their transoceanic migration [18] . Female blow flies can survive for about one year in Japan [17] , in comparison to the house fly Musca domestica which has a mean longevity of 34.2 days [19] . The longevity and high dispersal ability of blow flies may also result in wide dispersion of viruses that they carry. Mark-release-recapture experiments conducted at Tamba Town in 2005 suggested that C. nigribarbis generally could migrate up to 2-3 km in 24 h [20] . The distance between the two poultry farms in Kyoto prefecture, where the two H5N1 virus outbreaks took place in 2004, was approximately 4 km. In fact, 10% of all C. nigribarbis flies collected at a site intermediate between the affected farms expressed H5N1 virus genes [11] . Viable titers of H5N1 influenza virus, but not virus replication, were detected for up to 24 h in the crop and intestine of virus-exposed C. nigribarbis [10] . The presence of infectious virus in blow flies for 24 h could have a strong implication for virus dispersion since blow flies, with their excellent flight capacity, could transport the H5N1 virus over significant distances. In addition, H5N1 virus has been isolated from feces and vomit matter of blow flies at the 48 h postexposure, but virus titers in flies at 48 h were lower than that of the virus-containing cotton used in these experiments. This suggested that the viability of influenza virus decreases steadily in the blow fly crop and intestine, although some infectious virus remains for longer than 24 h. Therefore, C. nigribarbis could transport H5N1 virus to poultry farms 2-3 km apart. How often do H5 influenza viruses migrate to Japan? Which subtypes of H5 influenza virus migrate to Japan? To following previous studies [10, 11] . No H5 influenza virus gene was detected from a total number of 96 fly pools examined ( It is well known that the domestic house fly, Musca domestica spp., and some other fly species can transmit many kinds of pathogens mechanically [21] [22] [23] [24] [25] . In particular, M. domestica spp. are the most important fly species at poultry farms [26] with regard to mechanical transmission of >30 different pathogens [13] , for example, bacteria, protozoa, viruses, and parasite oocysts and eggs. Some viruses can be transported to animals by contact with contaminated body surfaces of flies. In the case of the house fly, it has been shown that rotavirus can be mechanically transported by contaminated fly surfaces [21] . House flies frequently defecate while feeding and resting on food surfaces [27] . However, in studies of C. nigribarbis, neither defecation nor vomiting was observed within 24 h after feeding (data not shown). The body surface of the house fly could be contaminated by viruses easier than that of blow flies. This would suggest that the mechanisms of virus transmission by blow flies could be different from those of house flies. Therefore, to evaluate virus transmission mechanisms that are more complex than contact with a contaminated fly surface, blow fly intestinal organs, crop, and gut were analyzed for their possible role in transmission of avian influenza virus. A seasonal consideration is that M. domestica vicina populations are generally highest in the summer in Japan. In fact, no house fly was found around any poultry farm or pigpen in Tamba Town during our survey in March. Therefore, it seems reasonable that winter blow flies may be involved in transmission of winter pathogens, like influenza virus, by maintaining minimum infectious titers. We have suggested here that blow flies are likely candidates for mechanical transmission of HPAI because of their ecological and physiological characteristics as reviewed here. In fact, blow flies have already been recognized as important vectors for mechanical transmission of several serious infectious diseases, that is, poxvirus [28] , rabbit hemorrhagic disease [29] , and paratuberculosis [30] . Recently, it has been reported that the H5N1 viral gene was detected in house flies [31] and engorged mosquitoes [32] . We suggest that mechanical transmission by flies may also be involved in the outbreak and pandemic of infectious diseases other than HPAI. However, although there are high densities of a variety of fly species during all seasons in Southeast Asia, their ability to transmit viruses has not been evaluated. The prevalence of H5N1 avian influenza is still a public health problem for birds and humans. Therefore, field and laboratory studies on mechanical transmission of pathogens by flies would be very important for controlling H5N1 avian influenza outbreaks, at least in epidemic Southeast Asian countries. Recently, the H5N1 virus surveillance conducted in Indonesia suggested that pigs are at risk of infection during outbreaks and pigs can serve as intermediate hosts in which this avian virus can adapt to mammals [33] . They also found the evidence of pig-to-pig transmission of this virus without any significant influenza-like signs. The transmission mechanism of this virus became more complicated and serious. As we introduced in the previous section, blow flies prefer to lick carcasses and droppings of not only chickens but also pigs. Furthermore, we can assume that the flies can access the pigpen easier than the poultry farm. This finding from Indonesia [33] strongly suggest that it is important to pay attention to pigpens as well as poultry farms within 2-3 km, where viable H5N1 viruses are transmitted by blow flies.

Examining the knowledge, attitudes and practices of domestic and international university students towards seasonal and pandemic influenza

BACKGROUND: Prior to the availability of the specific pandemic vaccine, strategies to mitigate the impact of the disease typically involved antiviral treatment and “non-pharmaceutical” community interventions. However, compliance with these strategies is linked to risk perceptions, perceived severity and perceived effectiveness of the strategies. In 2010, we undertook a study to examine the knowledge, attitudes, risk perceptions, practices and barriers towards influenza and infection control strategies amongst domestic and international university students. METHODS: A study using qualitative methods that incorporated 20 semi-structured interviews was undertaken with domestic and international undergraduate and postgraduate university students based at one university in Sydney, Australia. Participants were invited to discuss their perceptions of influenza (seasonal vs. pandemic) in terms of perceived severity and impact, and attitudes towards infection control measures including hand-washing and the use of social distancing, isolation or cough etiquette. RESULTS: While participants were generally knowledgeable about influenza transmission, they were unable to accurately define what ‘pandemic influenza’ meant. While avian flu or SARS were mistaken as examples of past pandemics, almost all participants were able to associate the recent “swine flu” situation as an example of a pandemic event. Not surprisingly, it was uncommon for participants to identify university students as being at risk of catching pandemic influenza. Amongst those interviewed, it was felt that ‘students’ were capable of fighting off any illness. The participant’s nominated hand washing as the most feasible and acceptable compared with social distancing and mask use. CONCLUSIONS: Given the high levels of interaction that occurs in a university setting, it is really important that students are informed about disease transmission and about risk of infection. It may be necessary to emphasize that pandemic influenza could pose a real threat to them, that it is important to protect oneself from infection and that infection control measures can be effective.

Public cooperation in complying with infection control measures is required to minimize the spread of infectious diseases. Previous studies have demonstrated the positive correlation between willingness to adhere to the recommendations around infection control practices and perceived infectiousness and severity of the disease [1] [2] [3] [4] , perceptions about the effectiveness of control measures [5] and trust in the information being provided by national and international public health authorities [1] . From the literature published to date on the general public's risk perceptions and behaviour changes during the 2009 influenza A/H1N1 pandemic [1, 6, 7] , higher risk perception scores were reported from Asian countries than from Western countries. For example, participants from studies conducted in India [8] , Saudi Arabia [6] and Hong Kong [9] expressed higher concern and perceived susceptibility levels than the respondents from studies conducted in the UK [1] and Australia [3] . While these variations may be correlated with methodological issues or the time period during the 2009 pandemic in which the study was conducted (i.e. May 2009 versus November/December 2009), if the trends are accurate it has the potential to affect the speed and extent to which infection control measures are accepted. During the height of the pandemic, we undertook a study which aimed to measure the perceptions and responses of staff and students at our University [10] . While a large proportion of the sample reported either "no anxiety" or "disinterest", Asian respondents were significantly (p < 0.001) more likely to believe that the pandemic was serious compared to their counterparts from other regions. Although, most participants reported not adopting any specific behaviour changes, those who did were significantly more likely to be of Asian origin. In order to further explore these trends amongst our domestic and international university students, we used qualitative methods to explore their attitudes, risk perceptions and adoption of health behaviour interventions against seasonal and pandemic influenza. This study was carried out from May to August 2010. Qualitative semi-structured interviews were undertaken at the University of New South Wales (UNSW) in Sydney, Australia. The relevant Human Research Ethics Committee located at the university approved this study. Students attending the main campus of the university were approached to participate in the study. Two methods were used to identify potential participants. Firstly, the interviewer (JP) directly approached a convenience sample of students who were located in the food halls and recreation areas of the university campus and invited them to participate. In the latter half of the study, a snowball approach was used. The snowball approach was adopted due to problems with identifying and recruiting postgraduate students. They constitute a considerably smaller percentage of the total student body, often are enrolled externally and attend classes in the late afternoon/evening. Students were classified on the basis of their enrolment status: undergraduate vs. postgraduate, and domestic vs. international. We aimed to recruit a sample of students from each classification and hence we firstly screened the student to identify their enrolment status. Students enrolled in the bachelor of medicine were excluded as it was assumed they would not be representative of the general student body, and would have had more exposure to issues surrounding disease spread and control. Participation was voluntary and written consent was obtained. During the study period, pandemic influenza H1N1 activity remained low and sporadic cases of pandemic influenza continued to be reported without evidence of sustained community transmission [11] . The study researchers collaboratively developed an interview guide. Questions were shaped to cover the key areas of interest that included: knowledge, perceived severity, risk perceptions and concerns towards seasonal and pandemic influenza and personal health seeking behaviours and practices. Small variations in the questions were used to provide relevance for the overseas students. For example, we explored whether the international students believed their personal risk varied when located at home versus while residing in Australia and whether they had adopted or discontinued any health related behaviours whilst studying in Australia. We were not prescriptive around the term 'pandemic influenza'; instead we left it up to the student to interpret what they felt it meant to them. Pre-designed prompts were employed throughout the interview to trigger interviewees' thought. An interview face sheet was used to collect demographic information (sex, age, enrolment status etc.) from the participants. All interviews were conducted by JM and lasted up to one hour in length. The interviews were recorded and transcribed verbatim. Two investigators (HS and JM) developed a list of themes after the analysis of one-quarter of the transcripts. An agreed framework was then applied to another subsample of transcripts and further modified. Using this final framework, all of the transcripts were analysed and coded. Text was organized within the identified themes of the developed framework. No software was used in the process. A total of 20 university students' aged ≥18 years completed the interview (RR: 35 %). The participants ranged in age between 21 to 30 years and 70 % were born overseas (14/20). International students were over-represented in our sample (50 %, 10/20) compared to the actual proportion enrolled at UNSW (25 %). There was a reasonably level of knowledge amongst the participants about the transmission modes and common symptoms of influenza and the common cold. A number of international students associated the occurrence of seasonal influenza with the temperature drop in winter; however they did not elaborated on the mechanisms of the connection. "....people tend to get sick, the flu during winter because of the change in temperature, it gets colder. . .." (International postgraduate student) Many of the participants were unable to accurately define what pandemic influenza was. While avian flu or SARS were mistaken as examples of past pandemics, almost all participants were able to associate the recent "swine flu" situation as an example of a pandemic event. In comparison to seasonal influenza and colds, participants generally perceived pandemic influenza as being more serious. Pandemic influenza was associated with increased numbers of medical consultations/hospitalizations and a higher mortality rate. However, there were a few sceptical participants who were doubtful about the actual disease impact and felt that it was only "promoted" as being serious by the government and the media. Young children and the elderly were nominated as being the most vulnerable groups during a pandemic outbreak due to their 'sub-optimal immune systems'. Participants believed that children were less conscious about hygiene and were therefore more likely to be exposed to other infected children or contaminated objects in a school environment. On the other hand, teenagers and young adults (20s-30s), the 'physically and socially healthy' and the 'well educated' , were considered to be at lowest risk of contracting pandemic influenza. In regards to differences in risk between racially or culturally diverse groups, one participant commented that people or cultures that have frequent proximate interaction with each other were at risk of contracting the disease during pandemic. While another suggested that in countries where there are higher levels of respect for traditional medicines over western medicines, people may also be at risk. Amongst the international students it was suggested that the risk of contracting pandemic influenza is higher for people in their 'home towns' because of differences in their health care system, population density, personal hygiene practices and environmental quality. ". . .culture. . ...where they interact with a lot of people. . . such as Italians, they're very outgoing. . ., lots of interaction, proximate to each other, perhaps they're more prone," (Domestic postgraduate student) ". . ..some cultures where they have let say more respect for traditional medicine than modern medicine. . .. . ..are also going to be a problem. . .that's why lots of pandemic in say Asia and Africa, and not so much in Europe or America" (International postgraduate student) Not surprisingly, only a few postgraduate participants stated that university students were at risk of catching pandemic influenza. When participants were required to rate their self-perceived risk of contracting pandemic influenza in Australia during an outbreak, almost all of them rated themselves at the low end of the scale. Being young and leading a healthy lifestyle were the major reasons provided to justify the low self-perceived risks levels. Only two overseas participants considered themselves at the 'relatively high risk' end. However, they presented very different justifications for this ranking, as one thought that their 'adventurous lifestyle' put them at risk and the other because of the dynamic nature of the university environment. ". . .may be I travel a lot more than the other people, and I go to polluted environments, institutions. . .. and I meet people I don't know. "(International undergraduate student) Five of the international students perceived themselves to be at a higher risk of catching pandemic influenza in their 'home town' than in Australia. Differences in population density, quality of transport, connectivity to other countries, hygiene levels, accessibility to and quality of health care were the main reasons given for the differences. Regular hand washing, cough etiquette (covering mouth and nose when coughing or sneezing), and avoiding the sick were suggested as good strategies to prevent becoming infected with pandemic influenza. The use of social distancing/isolation or masks/respirators was not very popular. Social avoidance was considered to be the most difficult intervention to comply with and impractical due to the vast amount of human interaction existing in the society. One international student also felt that it was impolite to maintain a distance to a sick acquaintance or relative. The use of masks was dismissed, as they were considered uncomfortable, inconvenient and unnecessary. Moreover, participants believed that wearing mask would cause embarrassment and social stigma. Outbreaks of seasonal influenza amongst student and university populations have been previously reported [12, 13] . These outbreaks have resulted in increased absenteeism, impaired school performance, and increased health care utilization [14] . The first reported university outbreak of 2009 pandemic H1N1 occurred at the University of Delaware (UD), affecting an estimated 10 % of the student population. It spread rapidly through the University of Delaware community with a surge in illness over a 2-week period. Although severe illness was rare in this instance, the authors documented that the outbreak caused a substantial burden and challenge to the university health care system [15] . In Japan, Uchida et al. reported that the infection rate among university students they surveyed ranged from 4.3 % to 15.5 % during the 2009 pandemic [16] . The authors suggested that continued exposure to sick individuals and disease transmission occurred during the pandemic, mainly through university club activities. While our participants were knowledgeable about the modes of transmission of influenza, very few were able to accurately describe what 'pandemic influenza' actually meant. The participants had heard of 'swine flu' , however only a few demonstrated a high level of knowledge around how it originated. There was a lot of confusion around the role that animals play in regard to 'pandemics'. Unconfirmed beliefs and misconceptions regarding pandemic influenza H1N1 2009 have been previously documented [17, 18] . In accordance with most of the pre-and post pandemic general public studies conducted worldwide [1, 2, 8, 19, 20] , our participants held a common belief that they were not at risk of acquiring the disease. Amongst our participants, it was felt that they were protected against the infection because they were 'fit and young'. This sense of non-vulnerability has also been previously documented in our previous university study [10] and amongst dormitory housed university students (aged18-23 years) in the USA [21] . However, given the low level of comprehension about pandemic influenza, the general public may be over or underestimating their level of risk towards acquiring the disease and the health consequences if infected (serious illness, need for hospitalization, mortality risk). As highlighted through the interviews, our students believed in the classic picture of morbidity attributable to the flu, such that only the very young, the elderly, those with co-morbidities and those with weakened immunity are at risk. This result is consistent with the risk groups identified by participants in previous studies [4] . Given this low level of anxiety towards the pandemic, it is perhaps not surprising that the students did not undertake any behavioural changes in response to the H1N1 pandemic, as highlighted here and in our previous study [10] . During the 2009 H1N1 pandemic, posters developed by the Commonwealth Department of Health and Ageing and UNSW were placed in high traffic areas. They focused on: (1) encouraging faculty, staff and students to stay at home if symptomatic (i.e. with a fever, cough, and runny nose) and to protect each other; (2) cough/sneeze etiquette (i.e. "cover your mouth and nose when you cough and sneeze" and "dispose of used tissues in the bin) (3) hand hygiene (i.e. "Wash your hands properly and regularly"). Our participants considered regular hand washing, cough etiquette (covering mouth and nose when coughing or sneezing), and avoiding the sick as good strategies to prevent infection. During the early and peak pandemic periods, hand washing was found to be the most accepted intervention among university students in Hong Kong [22] , Korea [23] , United States [24] and Australia [10] . Young people such as our university students may be more amenable to hand hygiene as a strategy because of a number of reasons. Firstly, these practices are community learnt and represent actions that the person has been encouraged to carry out from a young age. Secondly, hygiene-based measures pose minimal disruptions to daily routine. However, this is just a hypothesis and was not explored in depth in the study. Amongst our participants, mask use, as an infection control strategy was extremely unpopular. In many western settings, where medical mask/respirator use is generally restricted to the hospital setting, it is not unanticipated that people would associate embarrassment and social stigma with the use of these products. At the University, it is extremely rare to see a student wearing a medical mask. This maybe because students believe that masks are uncomfortable, inconvenient and unnecessary. Habit is an important influence on routine behaviour [25] , including hygiene behaviour [26] , such that despite their best intentions people may find it difficult to implement new hygiene measures during a pandemic if they have not previously made these a habit. The implementation of infection control behaviours appears to depend on a number of environmental (e.g. time, energy, availability of facilities, social norms), and motivational (e.g. social responsibility, social relationships, selfishness) factors. In the future however, the level of adoption of measures such as masks will fluctuate with changes in perceptions of risks and the perceived infectiousness and severity of the disease. The use of voluntary home quarantine, social distancing, and school dismissal to prevent the transmission of pandemic influenza is a standard inclusion in most countries pandemic plans [27] [28] [29] . However, lower acceptance of isolation and social distancing, which can disrupt routine and enjoyable activities, has been observed in prior studies [30, 31] . When participants were asked to comment on how they felt about the use of these interventions they stated that they were not in favour of adopting these actions and would find them extremely difficult to comply with. During an outbreak of pandemic H1N1 virus infection at a large public university in April 2009, Mitchell et al. undertook an online survey of students, faculty, and staff to assess knowledge of and adherence to university recommended non-pharmaceutical interventions [24] . They found that amongst the students with an acute respiratory infection (ARI), 44 % reported leaving campus for >1 day while sick, 35 % had visitors and only 34 % reported missing days of class. Most students attended class or work, went out in public, and participated in purely social activities (including having visitors) while having an ARI. Aside from not wanting to miss these important events, it could be suggested that low risk perceptions and mixed messages about the severity of the 2009 influenza H1N1 pandemic and about the actual need for isolation and social distancing would probably have contributed to a low acceptance rate. There are a number of logistical issues that universities and other institutions need to contend with instigating measures such as isolation. For example, universities may have large numbers of students living on or around the campus. While some of these facilities are self-contained, others have large common dining, entertainment and study rooms. The difficulty of introducing home quarantine in this setting is that many of these students (especially international and interstate students) may be unable to leave the campus facilities and would end up having to care and cater for themselves. Given the inevitability of future disease outbreaks or pandemic, universities must undertake efforts ensure that the needs of the students are catered for in these situations. Students, their parents, and other members of the university community must be involved with planning for these events so that feasible action plans are developed. These plans must ensure that there is continuity for the student. A strength of this study was using interviews that allowed to uncover in greater depth the attitudes and perceptions of the students. However, there are several limitations in this study. These include: (1) over reporting: as the study was conducted through face-to-face interviews with our interviewer, it may have resulted in an overreporting of infection control behaviours to avoid embarrassment or judgement; (2) recall bias: some questions in the interview guide required participants to recall their past experiences during the 2009 H1N1 pandemic, therefore recalling errors may have occurred; (3) representation: as the study was undertaken one year after the pandemic, the responses received may not represent the attitudes that participants held during the pandemic; (4) participation rate was low. Communicating to students effectively about the spread of influenza and the need to adopt preventative measures on a large campus presents a challenge. University officers need to find a balance between promoting and educating, while trying not to incite unnecessary fear. In the event of prolonged public health threats, such as infectious disease disasters, online messaging and regularly updated web sites have been shown to be timely and effective in providing risk communication and health messages [32] . However, it has also been demonstrated that pandemic influenza-specific web sites are among the least accessible and most difficult to understand compared with web sites addressing other types of disasters [33] . Poor accessibility can significantly undermine the effectiveness of university pandemic preparedness efforts and limit the ability of individuals to make well informed decisions during pandemics [34] . Other mediums favoured by young adults such as popular internet sites (Facebook or twitter) should be considered as a possible means of information provision to this susceptible cohort and in increasing uptake of preventative health advice. Education campaigns targeting young adults could also utilise the university networks and information gateways, or distribute information through universitywide emails and newsletters. Given the high levels of interaction that occurs in a university setting, it is really important that students are informed about disease transmission and about risk of infection. It may be necessary to emphasize that pandemic influenza could pose a real threat to them, that it is important to protect oneself from infection and that infection control measures can be effective. Our participants believed that it would be extremely difficult to comply with infection control measures such as social distancing. In this university setting, practical measures may also be needed to support implementation, such as education, reminders and provision of hand gel. Raina MacIntyre receives funding from influenza vaccine manufacturers GSK and CSL Biotherapies for investigator-driven research. These payments were not associated with this study. The remaining authors have no competing interests.

Use of functional gene arrays for elucidating in situ biodegradation

Microarrays have revolutionized the study of microbiology by providing a high-throughput method for examining thousands of genes with a single test and overcome the limitations of many culture-independent approaches. Functional gene arrays (FGA) probe a wide range of genes involved in a variety of functions of interest to microbial ecology (e.g., carbon degradation, N fixation, metal resistance) from many different microorganisms, cultured and uncultured. The most comprehensive FGA to date is the GeoChip array, which targets tens of thousands of genes involved in the geochemical cycling of carbon, nitrogen, phosphorus, and sulfur, metal resistance and reduction, energy processing, antibiotic resistance and contaminant degradation as well as phylogenetic information (gyrB). Since the development of GeoChips, many studies have been performed using this FGA and have shown it to be a powerful tool for rapid, sensitive, and specific examination of microbial communities in a high-throughput manner. As such, the GeoChip is well-suited for linking geochemical processes with microbial community function and structure. This technology has been used successfully to examine microbial communities before, during, and after in situ bioremediation at a variety of contaminated sites. These studies have expanded our understanding of biodegradation and bioremediation processes and the associated microorganisms and environmental conditions responsible. This review provides an overview of FGA development with a focus on the GeoChip and highlights specific GeoChip studies involving in situ bioremediation.

As the most phylogenetically and functionally diverse group of organisms on the planet (estimated 2000-50,000 microbial species per gram of soil; Torsvik et al., 1990; Hong et al., 2006; Schloss and Handelsman, 2006; Roesch et al., 2007) , microorganisms are critical to ecosystem functioning and are involved in the biogeochemical cycling of carbon, nitrogen, sulfur, phosphorus, and metals, as well as degradation or stabilization of contaminants in the environment. However, because a vast majority (>99%) of microorganisms remain uncultured (Amann et al., 1995; Fuhrman and Campbell, 1998; Whitman et al., 1998) , culture-independent approaches must be used to gain a comprehensive picture of microbial communities. However, many of the culture-independent methods, such as 16S rRNA genebased cloning or quantitative PCR, require a PCR amplification step, which introduces well-known biases (Suzuki and Giovannoni, 1996; Warnecke et al., 1997; Lueders and Friedrich, 2003) . In addition, since many functional genes have too much variance or too few sequences available, conserved PCR primers cannot be designed for many functional genes. Even if primers could be designed for many functional genes, performing PCR with many different primer sets would be cost-and time-prohibitive. Microarrays allow the examination of thousands of genes at one time without the need for PCR amplification of each gene. Since microarrays were first shown to be valuable for the study of microbial communities (Guschin et al., 1997) , several types have been designed to examine microbial communities. These include (i) phylogenetic oligonucleotide arrays (POA), designed to examine phylogenetic relatedness or community composition using 16S rRNA or other conserved phylogenetic genes (Small et al., 2001; Loy et al., 2002; Wilson et al., 2002; Brodie et al., 2006) ; (ii) community genome arrays (CGA), designed to examine the relatedness of microbial species or strains or to identify community members using whole-genomic DNA probes Zhang et al., 2004) ; (iii) metagenomic arrays (MGA), designed as a highthroughput screening method using environmental clone library inserts as probes (Sebat et al., 2003; Mockler and Ecker, 2005; Gresham et al., 2008) ; (iv) whole-genome ORF arrays (WGA), designed to examine gene expression of individual microorganisms using probes for all ORFs in one or more genomes (Wilson et al., 1999) , but can also be used for comparative genomics (Murray et al., 2001) ; and (v) functional gene arrays (FGAs), designed to examine multiple functional genes at one time using probes for key genes involved in microbial functional processes of interest He et al., 2007 He et al., , 2010a . This review will focus on FGAs. for examining and monitoring microbial communities and is even being used for metatranscriptome analysis (van Vliet, 2010) . However, while many of the technical challenges of microarrays and high-throughput sequencing have been overcome, each still has some distinct advantages and disadvantages, which make them ideal as complementary approaches: (i) Random sampling errors. In most sequencing studies, only a small proportion of the microbial community is actually sampled (McKenna et al., 2008) and while theoretically with true random sampling the probability of sampling the same fraction of the community multiple times is low , one would expect that dominant populations would have a greater chance of being sampled multiple times. These sampling errors can result in low reproducibility between technical replicates (17.2 ± 2.3% for two replicates; 8.2 ± 2.3% for three; Zhou et al., 2011) . Microarrays, in contrast, interrogate all samples against the same set of sequences (probes), so that the same population is sampled each time. (ii) Relative abundance. Abundance of individual species will vary greatly within microbial communities. With sequencing-based approaches there will be a bias toward the most abundant sequences in the environment so that many of the obtained sequences will represent the most abundant species/sequences while possibly missing lesser abundant species/sequences. Microarrays are not affected in the same way since lesser abundance sequences will still hybridize to their corresponding probe and as long as it is above the detection limit, it will be detected. (iii) New sequence detection. One of the greatest advantages of sequencing is that new sequences are easily detected since any sequences in the sample can be sequenced (open system). Microarrays, in contrast, can detect only the limited number of sequences covered by the probe set on the array (closed system), as such, it is not able to detect new sequences. Some new microarray techniques have been developed to allow the discovery of new sequences. Capture microarrays have been developed, which use lower stringency conditions to hybridize or "capture" sequence variants Okou et al., 2007) . These captured sequences are then washed off and sequenced. An array with probes specific for viral families has been developed that uses the hybridization pattern to classify novel viruses (Wang et al., 2002; Ksiazek et al., 2003) . As such, microarray and sequencing approaches could be used to maximize the benefits and minimize the deficiencies of each. The first FGA developed used PCR-amplicon probes and targeted four N-cycling genes (nirS, nirK, amoA, and pmoA; Wu et al., 2001) . However, since PCR-amplicons were used, only a limited number of genes could be included because conserved primers can only be designed for a few functional genes. In addition, it would be cost-and time-prohibitive to amplify genes from hundreds of microorganisms or clones in order to achieve a truly diverse probe set. Most microarrays now use oligonucleotide probes, which are more specific (Zhou, 2003) , can be easily customized (Denef et al., 2003; Zhou, 2003; Gentry et al., 2006) , and are relatively inexpensive. Since then several different FGAs have been developed, although most of these cover only a limited number of genes and focus on specific functional groups or locations. For example, FGAs have been designed to examine methanotrophs (Bodrossy et al., 2003 (Bodrossy et al., , 2006 Stralis-Pavese et al., 2004) , N-cycling genes (Taroncher-Oldenburg et al., 2003; Jenkins et al., 2004; Steward et al., 2004; Zhang et al., 2007a) , pathogens and virulence factors (Call et al., 2003; Kostić et al., 2005; Cleven et al., 2006; Miller et al., 2008; Palka-Santini et al., 2009) , rhizobial isolates (Bontemps et al., 2005) , and acid mine drainage (AMD) and bioleaching systems (Yin et al., 2007) . The most comprehensive FGAs reported to date are the GeoChip arrays. The GeoChip 1.0 had 2006 oligonucleotide probes (50-mers) for genes involved in nitrification, denitrification, nitrogen fixation, methane oxidation, sulfate reduction (Tiquia et al., 2004) , organic contaminant degradation, and metal resistance (Rhee et al., 2004) . This array was used in several studies examining microbial communities at uranium (U)-contaminated sites (Wu et al., 2006a; Waldron et al., 2009) , in the Gulf of Mexico (Wu et al., 2008) , and under different land use strategies (Zhang et al., 2007b) and showed FGAs to be useful for microbial community studies. GeoChip 2.0 was developed to provide a truly comprehensive probe set for multiple functional gene categories and to provide increased specificity for highly homologous gene variants . GeoChip 2.0 contains 24,243 (50-mer) oligonucleotide probes targeting ∼10,000 functional genes from 150 gene families involved in the geochemical cycling of C, N, and P cycling, sulfate reduction, metal reduction and resistance, and organic contaminant degradation. This array has been used in numerous studies to examine microbial communities at metals contaminated sites (Gao et al., 2007; Van Nostrand et al., 2009 , oil or diesel-contaminated sites (Rodríguez-Martínez et al., 2006; Liang et al., 2009a,b) , coral mucus (Kimes et al., 2009) , lake or river samples (Taş et al., 2009; Parnell et al., 2010) , deep sea samples (Mason et al., 2009; Wang et al., 2009) , Antarctic soils (Yergeau et al., 2007) and to examine the taxa-area relationship . GeoChip 3.0 covers 56,990 sequences from 292 gene families, greatly increasing the number of genes and categories covered compared to GeoChip 2.0 and added new control features (He et al., 2010a) . New gene categories include antibiotic resistance, energy processing, and phylogenetic markers (i.e., gyrB). A set of 16S rRNA gene probes were added as positive controls, human, plant, or hyperthermophile gene probes were added as negative controls, and a common oligo reference standard (CORS) was added for data normalization and comparison. The CORS is composed of an artificial sequence probe that is co-spotted with each gene probe and the complementary CORS target, labeled with a contrasting fluorescent dye to the sample, which is then spiked into each sample prior to hybridization (Liang et al., 2010) . The signal intensity of the CORS probe can then used to normalize the signal intensity of the sample and allows comparison of samples hybridized at different times. In addition, a computational pipeline has been developed for GeoChip probe design and data analysis. The GeoChip 3.0 has been used to examine microbial communities associated with elevated CO 2 (He et al., 2010b) , to examine communities within coal formation production waters (Wawrik et al., 2012b) or rhizosphere communities in As-contaminated sites (Xiong et al., 2010) . GeoChip 4.0, the newest version, is synthesized by Nimblegen (Madison, WI, USA) in their 12-plex format and contains 83,992 probes targeting 152,414 genes in 410 gene categories (Lu et al., 2012a) . In addition to added genes in most categories, new categories added include stress response, antibiotic resistance, and bacteriophage genes. It has been used to examine microbial communities during the 2010 Gulf oil spill (Lu et al., 2012a) , GeoChip covers a wide range of functional genes and currently includes sequences from bacteria, Achaea, fungi, and viruses. The first step in designing new probes for the array is deciding which processes should be included. Then genes for enzymes or proteins that are key to the process of interest are selected. These could be catalytic subunits or proteins with recognition sites or that provide functional specificity. Next, keywords are selected to search public sequence databases (e.g., GenBank). The keywords should be as broad as possible since proteins from different microorganisms may be annotated differently or have more general or specific annotations. Once the sequences are downloaded, they are confirmed by HMMER alignment 1 with preselected seed sequences. The seed sequences are those sequences for which the protein identity and function have been experimentally confirmed. This is a critical step in the design process and these sequences should be selected with care. The HMMER confirmed sequences are then used to design gene-or group-specific 50-mer oligonucleotide probes using new versions of the CommOligo software (Li et al., 2005) and experimentally determined criteria based on sequence homology (≤90% identity for gene-specific probes, and ≥96% for group-specific probes), continuous stretch length (≤20 bases for gene-specific probes, and ≥35 for group-specific probes), and free energy (≥35 kJ mol −1 for gene-specific probes, and ≤60 kJ mol −1 for group-specific probes; He et al., 2005b; Liebich et al., 2006) . The probes are then BLASTed against the GenBank database to confirm specificity. Keywords, downloaded sequences, seed sequences, HMMER confirmed sequences, and designed probes are stored in corresponding databases for use in future array updates. The newly designed probe sets can then be commercially synthesized. Several options are available for producing arrays. Synthesized oligonucleotide probes can be spotted onto nylon membranes or glass slides (Taroncher-Oldenburg et al., 2003; Rhee et al., 2004; Tiquia et al., 2004) . Glass slides are more frequently used since they have less background fluorescence (Schena et al., 1995 (Schena et al., , 1996 and allow higher probe density (Ehrenreich, 2006) . Probes can also be added to slides using bubble Jet printing (Okamoto et al., 2000) , laser-induced forward transfer (Serra et al., 2004) , or photolithography (Chen et al., 2009 ). In addition, a few companies, such as Agilent or Affymetrix, synthesize custom microarrays using a customer's probe set. GeoChip can be hybridized with either DNA or RNA. Most DNA samples used for GeoChip analysis are extracted using a 1 http://hmmer.wustl.edu/ well-established freeze-grind method with detergent lysis (Zhou et al., 1996; Hurt et al., 2001) since it provides high molecular weight DNA, important for later amplification steps. The use of RNA presents some challenges as mRNA is unstable and has a low abundance in environmental samples. Several papers have described methods for extracting environmental RNA, including a protocol for the dual extraction of both DNA and RNA Burgmann et al., 2003) or RNA alone (McGrath et al., 2008; Poretsky et al., 2009a) . Methods for mRNA enrichment include size separation by gel electrophoresis (McGrath et al., 2008) or use of commercial kits [MICROBExpress (Ambion) and/or mRNA-ONLY (Epicentre Biotechnologies); Poretsky et al., 2009b; Mettel et al., 2010] . Size separation obtained 115-155 ng mRNA from 4.6-5.3 μg total RNA (McGrath et al., 2008) . Using commercial kits, Mettel et al. (2010) were able to obtain 140-530 ng of mRNA from 0.4-2.0 μg total RNA per 0.5 g soil. Nucleic acid quality is of great importance for microarray analysis. DNA and RNA should have an A 260 to A 280 ratio ∼1.8 and >1.9, respectively and an A 260 to A 320 ≥ 1.7. The A 260 to A 320 ratio is most important in determining microarray success (Ning et al., 2009) . Some environmental samples, especially those with high humics, can be difficult to purify up to the necessary level. A gel purification strategy followed by a phenol-chloroform-butanol extraction (Xie et al., 2007; Liang et al., 2011) has been successful with a wide range of soil and sediment samples. Large amounts of DNA (e.g., 1 μg) or RNA (e.g., 5 μg) are needed for GeoChip hybridization. However, it can be difficult to get sufficient quantities of nucleic acid from some types of samples (e.g., water) or the sample is too difficult or impossible to replace to use such large quantities of nucleic acid. In this case, amplification of DNA or RNA can be done using either whole community genome amplification (WCGA; Wu et al., 2006a) or whole community RNA amplification (WCRA; Gao et al., 2007) . WCGA uses the Templiphi 500 amplification kit (phi 29 DNA polymerase, GE Healthcare, Piscataway, NJ, USA) with a modified amplification buffer and using 1-100 ng DNA provides a sensitive (10 fg detection limit) and representative amplification (<0.5% of amplified genes showed >2-fold difference from unamplified; Wu et al., 2006a) . WCRA provides a representative amplification with 50-100 ng of starting material. There are commercial kits available for microbial RNA amplification such as the MessageAmp TM II-Bacteria RNA Amplification Kit (Life Technologies, Grand Island, NY, USA). There are also other commercially available methods for WCGA. Wang et al. (2011) compared two of these (Bacillus stearothermophilus DNA polymerase (Bst) and REPLI-g; Qiagen, Valencia, CA, USA) with the modified Templiphi kit (Wu et al., 2006a) . The amplification bias for all methods was relatively low (<3-fold). Less bias was observed with REPLI-g and Templiphi for pure culture DNA and with REPLI-g for community DNA while Bst showed the least inhibition by lesser quality DNA. The amplified (or unamplified) nucleic acids are directly labeled with a fluorescent dye (Cy3 or Cy5) using random priming with the Klenow fragment of DNA polymerase for DNA (Wu et al., 2006a) or Superscript TM II/III RNase H-reverse transcriptase for RNA (He et al., 2005b) . The labeled DNA/RNA is then purified and dried for hybridization. The labeled nucleic acids are then hybridized to the microarray at 42-50 • C with 40-50% formamide (He et al., , 2010a Lu et al., 2012a) . Hybridization specificity can be adjusted by varying the temperature or the formamide concentration (the effective hybridization temperature increases by 0.6 • C for every 1% of formamide). Hybridized slides are then scanned and analyzed by quantifying the pixel density (intensity) of each spot using image analysis software. Commercial manufacturers often have their own analysis software or other microarray analysis software can be used, such as GenePix Pro (Molecular Devices, Sunnyvale, CA, USA), GeneSpotter (MicroDiscovery, San Diego, CA, USA), or ImaGene (BioDiscovery, El Segundo, CA, USA). For GeoChip data, there is a data analysis pipeline 2 for rapid preprocessing and data analysis. Poor and low quality spots and outliers, based on Grubbs' test of outliers (Grubbs, 1969) , are removed and then the signal intensities of all spots are normalized. Positive spots can be determined using signal-to-noise ratio [SNR = (signal mean − background mean)/background standard deviation], signal-to-both-standard-deviations ratio [SSDR = (signal mean − background mean)/(signal standard deviation − background standard deviation)] (He and Zhou, 2008) , or signalto-background ratio (SBR = signal mean/background mean) (Loy et al., 2002) . Due to the large volume of data obtained from GeoChip, data analysis can be very challenging. The data has a multivariate structure and the number of variables is much larger than the number of observations (p n). To assist users with data analysis steps, a pipeline is available which performs many of the common analyses 3 . Some common descriptive statistics used include relative abundance of gene categories or subcategories, richness and diversity (α and β) indices, and percentages of gene overlap between samples. To compare the overall community structure, unconstrained ordination [principal component analysis (PCA) and correspondence analysis (CA)] to reduce the dimensionality of variables in order to maximize the visible variability of the data or hierarchical cluster analysis (HCA), which groups communities based on the similarity of their gene profiles, can be used. To compare communities, response ratios, which compare the signal intensity of genes between conditions (Luo et al., 2006; Liang et al., 2009a) , t-tests, ANOVA, and dissimilarity tests can be used. Several methods can be used to examine the relationship between communities and environmental parameters. These include constrained ordination, such as canonical correspondence analysis (CCA; ter Braak, 1986), distance-based redundancy analysis (db-RDA; Legendre and Anderson, 1999), variation partitioning analysis (VPA; Økland and Eilertsen, 1994; Ramette and Tiedje, 2007) , and Mantel test. A relatively new analysis method is the random matrix theorybased (Mehta, 1990) neural network analysis (NNA) used to examine gene relationships within microbial ecological networks . Having high-quality nucleic acids (non-degraded, large fragments to improve amplification yields, absence of inhibitors or contaminants which may impede subsequent amplification and labeling steps) is the most important criterion for successful microarray experiments. Nucleic acids can be purified using commercial kits although the presence of humic acids and other contaminants can still be a problem. If large amounts of DNA are present, an agarose gel purification followed by phenol-chloroform-butanol extraction (Xie et al., 2007; Liang et al., 2009b) can be used, but large amounts of DNA are lost with this method so it is not practical for low abundance samples. So, better purification methods with high recovery yields are needed. One of the main objectives in developing FGAs was to provide a truly comprehensive probe set . Each new GeoChip version has expanded the coverage of gene variants and expands the number of genes covered (He et al., , 2010a Lu et al., 2012a) . This continued expansion is challenging as the number of gene sequences available is constantly increasing as new sequences are being submitted to public databases. While the GeoChip design pipeline 2 has an automated update feature which uses the previously selected key words and seed sequences to search the NCBI database, downloading new sequences and designing probes is still time consuming due to the sheer volume of sequences available. As such, better and faster computation systems are needed. In addition, available microarray probe density limits are rapidly being approached as the number of GeoChip probes increases. So, new methods of array construction to increase probe density are needed. Two key issues for microarray hybridization of microbial communities are specificity and sensitivity since environmental communities can have such vast diversities. Both of these conditions can be improved at various stages of microarray design, construction, target preparation, or hybridization. During probe design, determining the best criteria for probe design, such as using experimentally determined design criteria (He et al., 2005b; Liebich et al., 2006) can improve specificity, thus decreasing the number of false positives . Probe length also affects specificity and sensitivity; longer probes are more sensitive, but less specific (Denef et al., 2003; He et al., 2005a) . The method of array synthesis can also affect sensitivity and specificity. Increasing the probe concentration per spot can increase sensitivity (Cho and Tiedje, 2002; Relógio et al., 2002; Zhou and Thompson, 2002) . However, this may also decrease specificity by decreasing the overall probe signal intensity (Denef et al., 2003) . The choice of array surface can also be important as use of unmodified array slides can decrease background fluorescence thus requiring a lower signal fluorescence for detection (Kumar et al., 2000; Gudnason et al., 2008) . Target preparation strategies can also affect these parameters. Amplification of community DNA can increase sensitivity. WCGA was able to representatively amplify 1-250 ng of community DNA (Wu et al., 2006a) , increasing the detection limit from 25 ng to 10 pg (2 bacterial cells); however, using such small quantities of starting material greatly increases the amplification bias compared to the bias observed with 1 ng of DNA. Labeling methods can also affect sensitivity. For example, cyanine dye-doped nanoparticles or tyramide signal amplification labeling are able to increase sensitivity 10-fold (Denef et al., 2003; Zhou and Zhou, 2004) . Hybridization conditions can also be used to increase specificity and sensitivity. Temperature and formamide concentration can be modified to adjust stringency thus altering specificity . A lower hybridization solution volume and mixing during hybridization (Adey et al., 2002) have both been shown to increase sensitivity. Decreasing ozone levels, which can degrade cy-dye signal (Branham et al., 2007) , can also improve sensitivity. Most GeoChip analysis has involved the use of DNA, so that only gene abundance can be determined. These changes can be used to infer microbial activity, but cannot provide direct proof of activity. mRNA can be used for FGA analysis to monitor activity (Dennis et al., 2003; Bodrossy et al., 2006; Gao et al., 2007; Wawrik et al., 2012a) , although as mentioned above, working with environmental RNA can be challenging. Stable isotope probing (SIP) has also been used with GeoChip to monitor microbial activity (Leigh et al., 2007) . Gao et al. (2007) used amplified community mRNA from a denitrifying fluidized bed reactor to examine microbial activity. Genes for nitrate and nitrite reduction, organic contaminant degradation, sulfite reduction, and polyphosphate kinase were detected, consistent with reactor operation (Gao et al., 2007) . Another study used amplified community mRNA to examine nitrate utilization in marine bacterial communities (Wawrik et al., 2012a) . Hybridization results indicated activity by ureC, nirS, nirK, narG, nosZ, napA, nrfA, amoA, and nifH genes, indicating that urea cycling, denitrification, dissimilatory nitrate, nitrite reduction, and N fixation were occurring (Wawrik et al., 2012a) . Another method of monitoring microbial activity with GeoChip is to combine it with SIP (Leigh et al., 2007) . Microcosms were set up from soil samples collected from the root zone of a tree growing in a PCB-contaminated site and fed 13 C-labeled or unlabeled biphenyl. Genes involved in biphenyl degradation were detected as were other organic contaminant degradation genes including those for degradation of benzoate, catechol, naphthalene, and phenol. Several GeoChip-related studies have examined microbial communities from U-contaminated groundwater at the U.S. Department of Energy (DOE) Oak Ridge Integrated Field Research Challenge (OR-IFRC) site. Groundwater samples covering a range of contamination levels and an uncontaminated background sample were compared using GeoChip 1.0 (Wu et al., 2006a) . Samples from the uncontaminated site and those with lower levels of contaminants had higher functional gene diversity and gene numbers. In addition, as expected based on the contaminants present at this site, genes for denitrification, organic contaminant degradation, metal resistance, and sulfite reduction (dsr) were detected. A similar sample set using the same array was examined in greater detail in a later study (Waldron et al., 2009) . In this study, sulfate, pH, U, and Tc were found to be the most important drivers in determining the microbial community structure, with pH and the combination of U and Tc explaining ∼21% of the variance observed or 29-40% when all four variables were included. Another study at this site examined a pilot-scale field bioremediation system which used ethanol as an electron donor to stimulate microbial communities and immobilize U(VI) by reduction to U(IV) (Luo et al., 2006; Wu et al., 2006b,c) . GeoChip 2.0 was used to examine the microbial communities during different phases of operation. A period of active U(VI) reduction occurred after initial start-up (days 137-304). During this period U(VI) reduction was relatively rapid and genes associated with denitrification, sulfate reduction, and Fe(III) reduction increased in abundance, suggesting that these populations were involved in U(VI) reduction . This active reduction was followed by a maintenance period during which the low level of U(VI) was maintained, and the denitrifying, sulfateand Fe(III)-reducing communities remained in higher abundance. Next, the stability of the bioreduced U(IV) was examined by allowing the system to enter periods of starvation (ethanol injections were halted) and reoxidation (dissolved O 2 entered the system). The functional communities showed distinct clustering patterns based on whether the system received ethanol or not, indicating a shift in community structure with the return of ethanol injections . While total gene numbers increased once ethanol injection was restarted, the relative abundance of each gene group changed little during and after starvation, indicating a functionally diverse community which could be stimulated after adverse conditions. Chemical oxygen demand (COD, i.e., ethanol) was the most important driver in determining community structure, but temperature, sulfate, and U(VI) were also important. In this same remediation system, the sediment microbial community was examined with GeoChip 2.0 after 2 years of operation (Xu et al., 2010) . Sediment samples were collected from 11 wells, 5 from the outer loop and 6 from the inner loop. Results revealed significant differences between the microbial communities in the inner and outer loops. The inner loop communities had higher gene numbers and greater diversity than those in the outer loop and inner and outer loop samples were grouped separately based on hierarchical clustering and principle component analysis, indicating that the ethanol injections stimulated the microbial communities in the inner loop. In addition, genes important for U(VI) reduction such as cytochrome c, dsr, and denitrification as well as genes involved in metal resistance and organic contaminant degradation were enriched in the inner loop where electron donor was added. This study demonstrated the importance of U(VI)reducing populations for the maintenance of reducing U(IV) in this bioremediation system. Another GeoChip 2.0 study examined groundwater microbial communities at a field site examining the use of acetate to stimulate U(VI)-reducing microorganisms in the subsurface at the Old Rifle site, a former U ore processing facility in Rifle CO (Liang www.frontiersin.org et al., 2012) . The study compared communities taken during a shift from sulfate to Fe(III)-reducing conditions. The overall community structure changed with the switch from Fe(III)to sulfate-reducing conditions and were reflective of the redox conditions at the site. Sulfate-reducing and methane-generating microorganisms increased in abundance under sulfate-reducing conditions. Acetate, U(VI) and redox potential were important environmental variables in determining the microbial community structure. Xie et al. (2011) examined five AMD sites in China using GeoChip 2.0 to determine the functional diversity and metabolic potential of microbial communities in these sites and to determine how the communities responded to environmental conditions. The sites showed a great deal of variability in regards to the microbial communities with ∼150-1000 functional genes detected in each sample. Most of the genes represented on the GeoChip that were involved in C, N, S cycling and metal resistance were detected in all of the AMD sites. Results indicated that the immediate environmental conditions were important in forming the variations in the functional structure of microbial communities as opposed to spatial distance. There was a positive correlation between Zn resistance gene abundance and Zn concentration but not for other metals. However, the concentrations of B, Co, Cu, La, Mg, and S were significantly correlated with the community structure in these communities. Overall, results suggested that AMD microbial communities may not be as simple as previously thought. GeoChip 2.0 has also been used to probe pure culture isolates for the presence of specific genes. Four Ni-resistant Gram-positive actinomycetes were hybridized to GeoChip to get a better idea of what metal resistance genes were present (Van Nostrand et al., 2007) . Genes associated with resistance to Al, As, Cd, Cr, Cu, Hg, Ni, Te, and Zn were detected. Microbial communities from the rhizosphere of the arsenichyperaccumulating plant Pteris vittata and non-rhizosphere samples were examined using GeoChip 3.0 (Xiong et al., 2010) . The functional gene diversity was significantly correlated with As concentration. Interestingly, As contaminated rhizosphere samples had higher functional gene diversity than non-rhizosphere samples even though the non-rhizosphere samples had a lower level of As. In addition, greater numbers of As resistance genes, with higher signal intensities, were detected in rhizosphere samples compared to non-rhizosphere samples and very few genes were detected in both environments, suggesting that the rhizosphere and non-rhizosphere microbial communities were distinct. Results suggested that the P. vittata rhizosphere may protect the microbial communities from As contamination. Another study used GeoChip 2.0 to examine microbial communities in Zn-and Cd-contaminated soil microcosms with or without Thlaspi caerulescens, a Cd and Zn hyperaccumulator plant (Epelde et al., 2010) . Higher numbers of functional genes were detected in the contaminated samples than in uncontaminated samples and in planted samples compared to unplanted. Thirty-five to forty-seven percent of the variation in community structure observed was explained by metal concentrations. All of the Cd and/or Zn resistance genes (12) were detected in the contaminated, planted samples while only 7 were detected in the contaminated/unplanted samples. Substrate-induced respiration, K concentration, and nitrate concentration were the most important environmental variables in determining the functional community structure. The microbial community associated with a bioremediation system comprised of a fluidized bed reactor to clean dieselcontaminated groundwater in Vega Baja, Puerto Rico was examined with the GeoChip 1.0 (Rodríguez-Martínez et al., 2006) . Genes involved in the degradation of diesel fuel and other organic contaminants (acetylene, aniline, benzoate, biphenyl, cyclohexanol, methyl tert-butyl ether, naphthalene, phthalate, protocatechuate, and toluene) were detected. Increased signal intensities for genes involved in anaerobic benzoate degradation indicated a shift toward anaerobiosis over time, a conclusion supported by other experimental evidence. Liang et al. (2009b) examined the effect of different bioremediation treatments on microbial communities using laboratory scale bioremediation systems with sediment from contaminated oil fields and inoculated with oil degrading enrichment cultures. The systems were incubated 242 days, treated with ozone, and incubated an additional 125 days. Many oil degradation genes (benzene, benzoate, catechol, polyaromatic hydrocarbon aromatics, protocatechuate, phthalate) were detected with GeoChip 2.0. Ozonation treatment resulted in an almost 50% reduction in the number of functional genes detected. Gene numbers increased again after a recovery period and the community retained the ability to degrade oil. Another study used GeoChip 2.0 to characterize microbial communities along an oil contaminant gradient and found a decreased number of functional genes as the contaminant levels increased although genes involved in the degradation of biphenyl, catechol, and protocatechuate increased in the more contaminated samples (Liang et al., 2009a) . The most important environmental factors in determining the microbial community structure were oil concentration and soil available nitrogen. Liang et al. (2011) collected contaminated and uncontaminated soils from five oil fields across China in order to determine whether oil contamination or geographic location played a larger role in determining the microbial community structure. Results from GeoChip 2.0 indicated that communities from uncontaminated sites had higher functional gene diversity than those from contaminated sites in the same geographical area. Overall, the microbial communities clustered based on geographic location; however, when only organic contaminant degradation genes were examined, the contaminated samples clustered together. Geographic location was able to explain ∼33% of the microbial community variation observed, oil explained ∼10% of the variation, and soil geochemistry explained another 12%, while the remainder (∼41%) was unexplained. GeoChip 4.0 was used to compare microbial communities in oil-contaminated water to those from uncontaminated water in order to understand the effects of the 2010 Gulf of Mexico oil spill Lu et al., 2012a) . Results indicated that after only 40 days the presence of the hydrocarbon plume Frontiers in Microbiology | Microbiotechnology, Ecotoxicology and Bioremediation (1100 m depth) caused a significant shift in the microbial community functional structure and composition and that indigenous microorganisms, similar to known petroleum degraders, were stimulated by the hydrocarbon plume. Many genes associated with hydrocarbon degradation were significantly enriched in plume samples Lu et al., 2012a) . Genes that were enriched in plume samples included those for naphthalene 1,2-dioxygenase, β-oxidation of benzylsuccinate, cyclohexanone 1,2-monooxygenase, and alkene monooxygenase (Lu et al., 2012a) . These findings suggest that the microbial communities in the Gulf of Mexico were capable of intrinsic bioremediation and that the presence of the oil stimulated the oil-degrading community members and were important in determining the fate of the deep-sea oil spill. In a study using GeoChip 2.0 to examine three atrazinecontaminated aquifers and a background site, Liebich et al. (2009) detected more genes in the background site compared to the contaminated sites. The aquifer with the highest level of contamination had the highest number of genes, most involved in contaminant degradation, compared to the other contaminated samples. Atrazine-degradation genes were detected in all contaminated samples and verified by PCR. These results indicated that even small amounts of contaminant were enough to select for specific degrading populations. River sediments from industrial pollutant and pesticidecontaminated sites were examined with GeoChip 2.0 and the results indicated that contaminant level was not a major driver in these systems (Taş et al., 2009) . Instead, C/N ratio, depth, total Kjeldahl N, and location were the strongest drivers in determining the community structure. Most of the reductive dehalogenation genes detected were from Dehalococcoides spp., suggesting that this microorganism may play an important role in contaminant degradation in this system. GeoChip 2.0 was used to examine phenanthrene-spiked soil microcosms to examine the effect of phenanthrene on microbial communities (Ding et al., 2012) . Communities were examined after a 21-day incubation and compared with communities from day 0. A larger number of genes were detected in spiked soils compared to the control soils. Genes showing an increase in the spiked soils included dioxygenases involved in aromatic compound degradation, genes involved in the degradation of PAHs (nahA, rhda, nahQ, narR) , and genes involved in the degradation of one-ring aromatic compounds. In addition, an overall shift in community composition and structure was noted in spiked soils as determined by non-metric multidimensional scaling. Another study examined microbial communities associated with a leachate-contaminated landfill using GeoChip 3.0 (Lu et al., 2012b) . Groundwater samples were collected from wells along a flowpath of the landfill. Communities directly under the landfill and in the closest well had significantly lower functional gene diversity and richness. Genes involved in the anaerobic degradation of organic contaminates such as aromatic acids (bclA, bbs, tutFDG), phenoxyacetic acid herbicides (ftdA) atrazine (atzABC, trzN, trzA, trzE) were detected in all wells. Based on canonical correspondence analysis, the environmental variables (pH, sulfate, ammonia, and dissolved organic carbon) had significant effects on the community structure. The GeoChip arrays have been shown to be powerful tools in linking microbial function to ecosystem processes and are able to provide sensitive, specific, and potentially quantitative information. Use of this array in bioremediation studies have expanded our understanding of the microbial processes and communities at work in these sites and provide information necessary for the successful improvement and application of bioremediation strategies. Over the past decade, great improvements have been made in regards to microarray technology, design, and application. However, there are still technical hurdles that need to be overcome to further improve sensitivity and specificity in addition to better methods of nucleic acid extraction and purification. Improved bioinformatics tools are also needed to assist with data processing and analysis. The effort for preparing this review was supported by the Office of Science, Office of Biological and Environmental Research, of the U. S. Department of Energy under Contract No. DE-AC02-05CH11231 through ENIGMA -Ecosystems and Networks Integrated with Genes and Molecular Assemblies (http://enigma.lbl.gov), a Scientific Focus Area Program at Lawrence Berkeley National Laboratory and the Oklahoma Applied Research Support (OARS), Oklahoma Center for the Advancement of Science and Technology (OCAST), the State of Oklahoma through the Project AR062-034.

Functional Studies of ssDNA Binding Ability of MarR Family Protein TcaR from Staphylococcus epidermidis

The negative transcription regulator of the ica locus, TcaR, regulates proteins involved in the biosynthesis of poly-N-acetylglucosamine (PNAG). Absence of TcaR increases PNAG production and promotes biofilm formation in Staphylococci. Previously, the 3D structure of TcaR in its apo form and its complex structure with several antibiotics have been analyzed. However, the detailed mechanism of multiple antibiotic resistance regulator (MarR) family proteins such as TcaR is unclear and only restricted on the binding ability of double-strand DNA (dsDNA). Here we show by electrophoretic mobility shift assay (EMSA), electron microscopy (EM), circular dichroism (CD), and Biacore analysis that TcaR can interact strongly with single-stranded DNA (ssDNA), thereby identifying a new role in MarR family proteins. Moreover, we show that TcaR preferentially binds 33-mer ssDNA over double-stranded DNA and inhibits viral ssDNA replication. In contrast, such ssDNA binding properties were not observed for other MarR family protein and TetR family protein, suggesting that the results from our studies are not an artifact due to simple charge interactions between TcaR and ssDNA. Overall, these results suggest a novel role for TcaR in regulation of DNA replication. We anticipate that the results of this work will extend our understanding of MarR family protein and broaden the development of new therapeutic strategies for Staphylococci.

Staphylococci are among the most common causes of bacterial infection in the community and pose a major danger to human health. S. aureus is the most well-known of the species which produce hospital-and community-acquired infections, with methicillin-resistant S. aureus presenting a serious public health threat [1] . S. epidermidis is the sister species of S. aureus which often causes infection in immunocompromised individuals or those following damage to the epithelium. Both of them produce biofilm to protect themselves from host immune system and enhance their resistance to antibiotic chemotherapy [2] . The key component of the biofilm extracellular matrix in Staphylococci is polysaccharide intercellular adhesin (PIA) [3] , an essential factor in biofilm formation composed of homopolymer of b-1,6-linked N-acetylglucosamine (GlcNAc). The production of PIA depends on the expression of the icaADBC operon, and TcaR and IcaR are a weak and a strong negative regulator of transcription of the ica locus, respectively [4] . The transcription regulator TcaR is a member of the MarR family, and is involved in teicoplanin and methicillin resistance in Staphylococci [5] . The MarR family proteins function as regulators of protein expression and these regulated proteins confer resistance to multiple antibiotics, household disinfectants, organic solvent virulence factors, and oxidative stress agents [6, 7, 8, 9, 10, 11] . The crystal structures of a number of MarR family proteins have been reported, including MarR from Escherichia coli [12] , OhrR from Bacillis subtilis [13] , MexR from Pseudomonas aeruginosa [14] , MarR from Xanthomonas campestris [15] , SlyA from Salmonella typhimurium [16] , AdcR from Streptococcus pneumonia [10] and TcaR, which is studied in this work, from S. epidermidis [17] . These structures revealed that MarR family proteins are all homodimers. The overall structure of each monomer could be divided into two functional domains, one is the dimerization domain and the other is the winged helix-turn-helix (wHTH) DNA binding domain. The N and the C-terminal a-helices (a1, 5, 6) of one monomer interdigitate with those of the other monomer to produce dimerization interaction. In addition, the wHTH DNA binding domain is composed of a2-a3-a4-bA-W1-bB which adopts the winged-helix-fold, and the amino acid sequences of this domain are highly conserved. As the MarR-type proteins can act as positive, negative, or bifunctional regulators, TcaR also acts as a multi-functional regulator. It is not only as a regulatory factor to affect the transcription of icaADBC [4] , the first regulator reported for cell wall-anchored proteins (SpA and sasF), but also as the regulator of sarS [18, 19] . We previously described the 3D structures of TcaR in its apo form and in complex with salicylate as well as several aminoglycoside and b-lactam antibiotics [17] . In this research, comparison of the native TcaR structures and those in complexes indicated that the regulation mechanism involves a large conformational shift in the DNA binding lobe. Several antimicrobial compounds inhibited TcaR-DNA interaction and further induced biofilm formation in S. epidermidis. In the present study, we found that TcaR could interact with ssDNA and the result demonstrated that TcaR shows a stronger preference toward GCrich ssDNA than to dsDNA by using EMSA, CD, and Biacore experiments. However, the detailed mechanism of the interaction between TcaR and ssDNA still remains to be elucidated. In order to investigate the regulation mechanism of the ssDNA binding ability of TcaR, we applied electron microscopy (EM) technique to reveal TcaR-ssDNA complex. Furthermore, we clarified the role of TcaR-ssDNA interaction by in vitro replication assay and in vivo plaque assay. Taken together, these results provide an in-depth investigation on the multiple functions of TcaR in S. epidermidis. Strong TcaR Binding to ssDNA Oligomers Revealed by EMSA TcaR is known to bind and regulate the ica promoter [4] . We previously identified that TcaR most strongly interacts with IcaR DNA1 (a 33-mer pseudo-palindromic sequence containing consensus sequence TTNNAA) compared with other designed IcaR DNA fragments [17] . However, when using the sense strand of IcaR DNA1 (IcaR DNA1S) and the antisense strand of IcaR DNA1 (IcaR DNA1A) ( Figure 1A ) as controls in electrophoretic mobility shift assays (EMSA), the result demonstrated that TcaR shows a stronger preferences toward ssDNA fragments (IcaR DNA1S and DNA1A) ( Figure 1B) . To determine the type and length of the TcaR-binding site on ssDNA, a series of GC-rich and AT-rich ssDNA segments were designed ( Figure 1A ) [20, 21] . Their TcaR binding ability was tested using EMSA. As shown in Figure 1C , TcaR does not significantly interact with 17-mer GCrich (GC17) and AT-rich (AT17) ssDNA oligomers, but shows strong interaction with 33-mer GC-rich (GC33) and AT-rich (AT33) ssDNA sequences with a preference toward the 33-mer ssDNA sequence with a molar ratio of 1:1. Thus, we suggest that TcaR prefers binding to the 33-mer ssDNA. In order to evaluate the minimal DNA binding length of TcaR, GC-rich fragments of different lengths were designed. As seen in Figure 1D , GC-rich fragments with 33, 29, and 25 bases showed similar binding strength to TcaR; with TcaR forming a large, apparently multimeric complex with GC33, a small complex with GC25, and both small and large complexes with GC29 in EMSA. These results indicated that the minimal observed ssDNA fragment size to allow TcaR binding ranges between 17 to 25mer; providing useful information for the design of a DNA fragment with precise length suitable for crystal packing. Up to now, only three MarR family protein complex structures have been reported, and the first one is complexed with dsDNA [13, 16, 22] . The second one is complexed with salicylate [12, 17, 22, 23] and we discovered the third case which is complexed with antibiotics [17] . We have already obtained TcaR-ssDNA crystals and collected X-ray diffraction data to 3.6 Å resolution at SPring-8 (Hyogo, Japan), beamline BL12B2. However, the phase problem is still the main challenge and the works are currently under progress. Moreover, to investigate whether TcaR preferentially binds to ssDNA or dsDNA, the ability of ssDNA to compete with the TcaR-dsDNA complex was evaluated. For the competition assay, the IcaR DNA1 probe was preincubated with TcaR (dimer) protein to allow formation of the dsDNA-TcaR complex prior to mixing with increasing amounts of single-stranded GC33 DNA. It has been known that ssDNA products have lower migration velocity compared to its dsDNA counterparts in polyacrylamide electrophoresis [24, 25] . As shown in Figure 1E , ssDNA, as a competitor, interfered the binding of TcaR to the dsDNA, suggesting a binding preference for ssDNA. To further confirm this result, IcaR DNA1 and GC33 ssDNA oligomers were mixed, and their interaction strengths with TcaR were compared using EMSA ( Figure 1F ). Findings indicated that increasing the concentration of TcaR produces a ssDNA band shift greater than that for dsDNA, confirming a stronger interaction between TcaR and ssDNA. Moreover, to investigate possible pH effect of ssDNA binding activity of TcaR, a series of buffers with increasing pH were tested for their potential interfere in TcaR-ssDNA binding. As shown in Figure 1G , the EMSA results showed that TcaR had a strongest affinity for GC33 at pH 8.0 and the affinity was reduced by decreasing pH. Consequently, the result indicates that the ssDNA binding activity of TcaR is pH-dependent. To clarify whether the ssDNA binding site of TcaR is close, or identical, to the dsDNA binding site, a TcaR quadruple mutant (4 positively charged amino acids responsible for DNA binding mutated to alanines to produce R71A/K73A/R93A/K95A) [17] was designed and its ssDNA and dsDNA binding ability tested. As seen in Figure 1H , the mutant failed to interact with either dsDNA or ssDNA. This indicated that these amino acids are essential for binding in both ssDNA and dsDNA. The MarR protein of E. coli is a multidrug binding transcription regulator. A wide range of compounds, including 2,4-dinitrophenol, plumbagin, menadione, and salicylate, attenuate and inhibit its association with cognate DNA [26] . In our previous study, salicylate and multiple antibiotics interfered with the transcriptional repressor activity of TcaR [17] . These findings prompted the current investigation into the possible effects of antibiotics on the ssDNA binding ability of TcaR. Here, to investigate the possible effect of some drugs on TcaR, four compounds were tested for their potential inhibition on TcaR-ssDNA interaction. These include one beta-lactam antibiotics (ampicillin) that contain a b-lactam nucleus in their molecular structure and act by inhibiting the synthesis of the peptidoglycan layer of bacterial cell walls, two aminoglycoside antibiotics (kanamycin and gentamicin) that composed of several sugar groups and amino groups, and bacteriostatic antimicrobial (chloramphenicol) which is considered as a prototypical broad-spectrum antibiotic. As shown in Figure 1I , kanamycin, chloramphenicol and gentamycin interfered with the ssDNA (GC33) binding activity of TcaR at a concentration of 2 mM and this effect was more pronounced at a higher concentration, suggesting that antibiotics inhibit formation of the TcaR-ssDNA complex. Results indicated that ampicillin also antagonized the ssDNA binding activity of TcaR at a concentration of 20 mM. This result is consistent with the observation seen in SPR, as discussed below. Taken together, we believe that diverse kinds of antibiotics may interact with TcaR to regulate its ssDNA binding ability. The binding affinity of GC33 and TcaR was determined quantitatively using surface plasmon resonance. Increasing concentrations of TcaR were passed across a flow cell coated with ssDNA GC33 and the binding response was recorded as changes in response units (RU) after subtraction of the binding response for the reference flow cell. Figure 2A shows a representative sensorgram. Analysis of the sensorgram data indicates k a for the interaction of TcaR with the ssDNA GC33 is 8.8610 5 M 21 s 21 ; k d for the interaction of TcaR with the ssDNA GC33 is 9.5610 23 s 21 (Table 1) . To investigate whether TcaR binds to different types and different lengths of DNA molecules, Biacore experiment was used to test the binding of TcaR to DNA fragments of biotin-labeled ssDNA and hairpin DNA duplex (IcaR DNA1). Consistent with previous observations in Fig. 1B , TcaR binds to ssDNA GC33 and ssDNA AT33 with the higher affinity and the association rates of TcaR to IcaR DNA1, and ssDNA GC17, ranging from 4.3610 5 M 21 s 21 to 1.2610 5 M 21 s 21 , are much lower (shown in Figure 2B ). However, when look at a wider range of DNA sizes, TcaR showed no detectable binding ability to ssDNA 8-mer GC8 but the highest binding ability to ssDNA GC99. The association rate with ssDNA GC99 is 36-fold higher than with ssDNA GC33, along with a 20-fold higher off-rate, suggested that cooperative binding of TcaR may contribute significantly in its ssDNA binding activity (Table 1) . Interestingly, the association rate of GC33 ssDNA is two-times higher than IcaR DNA1, but the dissociation rate of IcaR DNA1 is the lowest compared to the DNA fragments tested in this study, suggesting different modes of interaction occurs in ssDNA-TcaR and dsDNA-TcaR complex ( Figure 2B ). Furthermore, in order to determine whether other MarR family and TetR family proteins such as SAR2349 from S. aureus and . IcaR DNA1 probe duplex of 1 mM was pre-incubated with 2 mM TcaR (dimer) at room temperature for 15 min before mixing with increasing concentration of GC33 ssDNA, followed by the same procedure as described in the legend to Figure 1B . (F) Competition experiment was carried out to compare the binding strength of TcaR to ssDNA and to dsDNA. In the EMSA analysis, 1 mM IcaR DNA1 probe duplex was pre-incubated with 1 mM GC33 ssDNA fragment for 15 min at room temperature before mixing with TcaR protein of increasing concentration. (G) The effects of pH to the TcaR binding to ssDNA GC33 in EMSA experiment. Molar ratio of GC33:TcaR was 1:2. (H) EMSA analysis of mutant TcaR protein binding to ssDNA (GC33) and dsDNA (IcaR DNA1) fragments with different protein-DNA ratio. (I) EMSA analysis of the binding of TcaR protein to ssDNA(GC33) in the present of different types of antibiotics. GC33 ssDNA probe of 1 mM was preincubated with 2 mM TcaR (dimer) at room temperature for 15 min before mixing with 2 mM or 20 mM antibiotics, followed by the same procedure as described in the legend to Figure 1B IcaR from S. epidermidis have the ssDNA binding ability, we conducted a series of SPR experiments to analyze the binding ability of SAR2349 and IcaR proteins to GC33 ssDNA ( Figure 2C ). The result shows that not all MarR family proteins have this ssDNA binding ability, thus pointing to the specific ssDNA-binding feature of TcaR. In addition, we have previously demonstrated that antibiotics appear to antagonize the ssDNA binding activity of TcaR ( Figure 1I ). Therefore, a measurement for the effect of kanamycin and ampicillin to the GC33 ssDNA binding affinity of TcaR is conducted using surface plasmon resonance to confirm the result. As seen in Figure 2D , the affinity between TcaR and GC33 ssDNA is shown by a decrease in RU values in the presence of antibiotics. This was especially apparent with kanamycin, which yield the lower binding capacity. Taken together, we demonstrate that TcaR shows a higher binding affinity to ssDNA than to dsDNA, and several antibiotics could regulate the ssDNA binding activity of TcaR. Conformational changes of TcaR in response to ssDNA were monitored using CD spectroscopy [27] . As shown in Figure 3 , the CD spectra of TcaR protein were scanned from 200 to 250 nm in the presence of GC33. With increasing concentrations of GC33 ssDNA, the CD spectrum shows a concomitant decrease in the intensity of the negative peak at 222 and 210 nm, revealing the conformational change of TcaR. Furthermore, in order to examine whether TcaR shows a binding ability towards much longer ssDNA fragments such as viral wx174-ssDNA, the interaction between them was also examined with increasing concentration (0, 2.5, 10 mM) of viral In order to further confirm our finding that TcaR forms complex with viral ssDNA, the M13 and wX174 phage ssDNA circles were used as probes in EMSA to evaluate TcaR binding. As shown in Figure 4A , TcaR reduced the mobility of the M13 and wX174 ssDNA, but S. epidermidis IcaR and S. aureus MarR family protein SAR2349 had no specific interactions with ssDNA. This indicated that TcaR has strong viral ssDNA-binding ability. It is also worth noting that other MarR family protein and TetR family protein do not have such ssDNA binding properties with high affinity, suggesting that the results from TcaR studies are not an artifact due to simple charge interactions between TcaR and ssDNA. Furthermore, since the attempt to obtain the TcaR-ssDNA complex structure was not successful, we resorted to EM analysis to image TcaR-wx174 complex and to pursue its 3-D reconstruction. After staining for 4 min with 2.5% uranyl acetate, EM analysis was performed with a Tecnai tm G 2 Spirit Bio TWIN (FEI CO., The Netherlands) using 120kV acceleration voltage. As seen in Figure 4B -D, EM imaging revealed that no complex was found in the negative control sections, whereas TcaR form a nucleoprotein filament with a circular viral wX174-ssDNA fully covered with proteins, suggesting strong and cooperatively interaction between viral ssDNA and TcaR. This is consistent with the EMSA results we observed. We are now testing another EM method as described by Lurz R et al. [28] to confirm the cooperative binding between the TcaR and viral wX174-ssDNA. A distinct group of DNA-binding proteins called the ssDNAbinding proteins (SBP) could specifically bind ssDNA and be used in processes where the double helix is separated, including DNA replication, transcription, and recombination. Because TcaR is known as a MarR family transcription regulator that binds to specific dsDNA sequence with the winged helix-turn-helix (wHTH) DNA binding domain, the ssDNA binding ability of TcaR may not be involved in transcription. In order to clarify the role of TcaR-viral ssDNA interaction, our approach is to examine it with in vitro replication. We used single-primed M13 replication assay to measure the ability of purified TcaR protein to convert a primed single-strand M13 template to the duplex form in a manner that requires processive DNA synthesis. As seen in Figure 5A , M13-based in vitro DNA replication assay showed that the addition of TcaR protein to the reaction mixture, and incubation for up to 30 min, resulted in almost no DNA replication activity compared to controls. This indicated that TcaR markedly inhibits DNA replication and that the mechanism of inhibition, at least in part, involves interaction with viral ssDNA. These results suggest a possible role for TcaR in bacteriophage resistance. Since 1980, investigators have developed an increasing number of bacteriophage therapies for the treatment, or prophylaxis, of bacterial infectious diseases [29, 30, 31] . Reports have described that appropriately administered phages can treat lethal infectious diseases caused by gram-negative and gram-positive bacteria, such as Pseudomonas aeruginae, Klebsiella pneumoniae, Enterococcus faecium, and S. aureus [32, 33, 34, 35] . Antibiotic resistance has become a global public health concern; thus investigators are extensively reevalu-ating phage therapies to fully exploit their antimicrobial potential [36, 37] . However, phages encounter a variety of different antiviral mechanisms during their infection of bacterial cells, such as prevention of phage adsorption and DNA entry, cutting of phage nucleic acids, and abortive infection systems [38] . Most reported antiphage systems have been shown to be relevant to the dsDNA phage, but not ssDNA, ssRNA, or dsRNA phages. To further confirm and clarify the first relationship between the TcaR protein and ssDNA phage resistance, the standard plaque assay was performed in E. coli as a model system since little is known about the ssDNA phage infecting Staphylococci. As seen in Figure 5B , induction of the TcaR protein in E. coli conferred increased host resistance to ssDNA phage (M13 and wX174) infection. However, a TcaR-expressing strain did not demonstrate reduced sensitivity to dsDNA phage Lambda (l) infection, suggesting that the phage resistance was caused by TcaR-viral ssDNA complex. The observed biological differences point to a remarkable plasticity of TcaR. These findings, thus, may support a hypothesis that TcaR might interfere with viral ssDNA replication and establish a link between TcaR and ssDNA phage resistance. The MarR family transcriptional regulators serve as sensors of changing environments, allowing pathogenic bacteria to survive and persist in a dynamic environment [39] . However, up to now, the knowledge of MarR family protein-nucleic acid interaction has been focused on dsDNA and the MarR family protein-ssDNA interaction ability as well as their contribution to the multiple functions of TcaR is yet to be discovered. Better understanding of these interactions not only will benefit the understanding of many biological mechanisms but also is expected to provide a concept for designing a new therapy for Streptococci. In this work, we present the first attempt to investigate the TcaR-ssDNA interaction. The information of TcaR-ssDNA binding mode and the minimal binding length that we obtained from EMSA analysis and Biacore will be helpful for us to obtain TcaR-ssDNA complexed structure successfully. Moreover, we used in vitro replication assay and plaque assay to elucidate the specific biological role of the ssDNA binding ability of TcaR. Such observations may help us understand the mechanism of antibiotic resistance in the MarR family regulators. The IcaR and TcaR proteins were expressed in E. coli BL21 (DE3) and purified as already described [17, 40] . The MarR homologous gene, SAR2349, was amplified directly from the S. aureus MRSA252 genome by polymerase chain reaction (PCR) and subsequently cloned into expression vector pET-32. This construct was transferred into E. coli of Arctic Express TM (DE3) RIL strain. The His 6 -tagged wild-type protein was over-expressed in Difco Luria-Bertani (LB) broth containing 50 mg/l ampicillin to an optical density at 600 nm of 0.5-0.6 and then induced with 0.5 mM IPTG (isopropyl-b-D-thiogalactopyranoside). Cells were grown for 2 days at 16uC. The cells were then harvested by centrifugation at 12,000 g for 30 min and disrupted by Constant Cell Disruption System (CONSTANT SYSTEM Ltd, UK) with lysis buffer containing 20 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 20 mM imidazole. The homogenate was centrifuged at 27,000g for 30 min and the cell-free extract was loaded onto a Ni 2+ -NTA column, which had been previously equilibrated with lysis buffer. The column was washed with lysis buffer, and the His 6 -tagged SAR2349 was subsequently eluted by a linear gradient of imidazole from 10 mM to 500 mM. His-tagged SAR2349 eluted was dialyzed twice against 5 liters of buffer (20 mM Tris-HCl, pH 8.0, and 500 mM NaCl) and then subjected to Thrombin digestion to remove the tag. The mixture was then passed through another Ni 2+ -NTA column, and subsequently untagged SAR2349 protein was dialyzed twice against 3 liters of buffer (20 mM Tris-HCl, pH 8.0) and then passed through a Q-Sepharose anion-exchange column for further purification, and subsequently SAR2349 was eluted by a linear gradient of 10 mM to 500 mM NaCl-containing buffer and then dialyzed twice against 5 liters of buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 2 mM DTT) for storage. The purified SAR2349 protein was finally concentrated by 3 kDa cut-off size membrane of Amicon ultra-15 centrifugal filter units (Millipore, MA, USA) for storage at -80uC. The six oligonucleotide probes used in EMSA experiments were purchased from MDBio Inc. (Taiwan) ( Figure 1A ). The viron wX174 and M13 ssDNA were purchased from New England Biolabd (USA). For the preparation of double-stranded IcaR DNA1, equimolar amounts (100 mM each) of complementary oligonucleotides were mixed, fully denatured by heating at 95uC for 5 min in 10 mM Tris-HCl pH 8.0, 20 mM NaCl and allowed to cool gradually to room temperature. Gel shift assays were performed by incubating 1 mM of ssDNA or dsDNA with 1-4 mM purified recombinant proteins under binding conditions (20 mM Tris-HCl, pH 8.0, 150 mM KCl, 0.1 mM MgCl 2 , 0.05 mM EDTA, 12.5% Glycerol, 0.1 mM DTT and 1 mg/ml BSA) for 15 min at room temperature with gentle vortex. After incubation, 15 ml of the reaction solution was mixed with 3 ml of the sample loading dye and subsequently electrophoresed on 6% preequilibrated polyacrylamide gels in 1/2 Tris/acetate/EDTA (TAE) at 100 V for 30 min and visualized using SYBR Green I nucleic acid gel stain (Invitrogen). For competition assay, IcaR DNA1 probe duplex of 1 mM was pre-incubated with 2 mM TcaR (dimer) at room temperature for 15 min, then mixed with increasing concentration of GC33 ssDNA. In the assay for analyzing the effect of different antibiotics on the interaction between TcaR and ssDNA, 1 mM GC33 probe was pre-incubated with 2 mM TcaR (dimer) at room temperature for 15 min before mixing with 2 mM or 20 mM antibiotics. The affinity, association and dissociation between the drug and the DNA duplexes were measured using a BIAcore 3000A surface plasmon resonance (SPR) instrument (Pharmacia, Uppsala, Sweden) with a SensorChip SA5 from Pharmacia by monitoring the refractive index change of the sensor chip surface [41, 42, 43] . These changes are proportional to the amount of bound analyte. The SPR angle change is reported as resonance units (RU). Five fragments of 59 biotinylated oligonucleotides probes purified by PAGE were purchased from MDBio Inc. (Taiwan). Activation buffer (100 mM NaCl, 50 mM NaOH) was injected for 1 min (20 ml) to remove any unbound streptavidin from the sensor chip. To control the amount of the DNA bound to the SA chip surface, 200 nM of the biotinated oligonucleotides were immobilized manually onto the surface of a streptavidin chip until 120 RU was reached in the first cell. The chip surface was then washed with 10 ml of 10 mM HCl to eliminate non-specific binding. The second flow cell was unmodified and served as a control. Different concentration of TcaR, IcaR, and SAR2349 proteins were injected at a flow rate of 30 ml/min in 50 mM Tris, 150 mM NaCl, pH 7.5 for 170 s to reach equilibrium. Blank buffer solution was then passed over the chip to initiate the dissociation reaction. At the end of each cycle, the surface was recovered with two 30 s injections of 0.025% SDS. SPR-binding constant is analyzed as described previously [44] . Sensorgrams for the interactions between DNA and TcaR were analyzed using BIA evaluation software to determine the association and dissociation rate constant (k a /k d ). In the assay analyzing the effect of antibiotics on the interaction between TcaR and ssDNA, TcaR protein with 640 nM kanamycin or ampicillin in 50 mM Tris, 150 mM NaCl, pH 7.5 was injected on to the sensor chip. Circular dichroism (CD) spectra were obtained using a JASCO-815 CD spectropolarimeter. Temperature was controlled by circulating water at the desired temperature in the cell jacket. TcaR and DNA samples were prepared under the conditions identical to those prepared for EMSA assay. The CD spectra were collected between 250 and 200 nm with 1 nm bandwidth at 1 nm intervals. All spectra were obtained from an average of five scans. The photomultiplier absorbance did not exceed 600 V during the analysis. CD spectra were normalized by subtraction of the background scan with buffer alone. The mean residue ellipticity, [h], was calculated based on the equation, [h] = MRW6hl/ 106l6c, where MRW is the mean residue weight, hl is the measured ellipticity in milidegrees at wavelength l, l is the cuvette pathlength (0.1 cm), and c is the protein concentration in g/mL. The TcaR proteins (0.3 mM) were first incubated at 30uC for 15 min in reaction buffer [20 mM Tris-HCl, pH 8.0, 150 mM KCl, 0.1 mM MgCl 2 , 0.05 mM EDTA, 12.5% glycerol, 10 mM DTT, 12 mM circular viron wX174 ssDNA (5386 nucleotides in length), 0.2 M ammonium acetate], and then chilled on ice to stop the reaction. The reaction product was diluted 100-fold with EM sample dilution buffer (2 mM MgCl 2 , 0.5 mM DTT, 10 mM HEPES pH 7.0). A droplet (4 ml) was placed for 1 min at room temperature on a copper grid (300 mesh, Pelco, USA) coated with fresh carbon. The excess buffer was then carefully blotted away from the edge of the grid with Whatman #1 filter paper (Whatman Inc., USA). After staining for 1 min with 2% uranyl acetate, excess liquid was removed and samples were dried at room temperature. Bio-transmission EM was performed with a Tecnai F20 Bio TWIN (FEI Co., Netherlands) using an acceleration voltage of 200 kV. Images were recorded with a slow scan CCD camera (Gatan MultiScan TM 600, USA) at a resolution of at least 4k64k pixels. For M13 replication assay, 250 mM of single primed M13mp18 ssDNA was incubated with/without 2 mM TcaR protein in the reaction mixture (20 ml) containing 25 unit Klenow fragment, 25 mM NaCl, 7 mM MgCl 2 , 1 mM EDTA, and 0.5 mM DTT for 3 min at 30uC to allow replication complexes to assemble at the primer template junction [45, 46] . Replication was allowed to proceed by addition of 60 mM dNTP. After incubation at 30uC for 30 min, the reactions were terminated by addition of 10 mM Tris-HCl, 5 mM EDTA, 0.5% SDS, and 50 mg of proteinase K (with total volume of 20 ml) and incubated at 50uC for 1 h. Conventional electrophoresis was then performed to verify the result of DNA replication (20-cm 0.8% agarose gel, 15 mA, 0.5X TBE buffer). The bands are visualized using SYBR Green I nucleic acid gel stain (Invitrogen) and quantified by Quantify One (BIO-RAD, USA). Host sensitivity to phages was tested using a virulent variant of phage (M13, wX174 and c) and E. coli BL21 (DE3) RIL transformed with engineered pET-16b-TcaR plasmid containing lacI gene and lac operator as host [47, 48, 49] . Cells were grown in LB media until the optical density (OD 600 ) reached 0.6. TcaR protein was then induced by adding a final concentration of 0.1 mM IPTG and used in plaque assays as previously described [50, 51] . Plaque assays were performed in triplicate. Plates and topagar contained LB and above mentioned concentrations of inducers. The sensitivity of the host to phage infection was calculated as the efficiency of plaquing, which is the plaque count ratio of a non-IPTG set to the IPTG set [52] . Error-bars were calculated as one standard deviation. The atomic coordinates and structure factors for the TcaR-RNA complex have been deposited in the wwPDB with accession numbers of 4EJT.

Marine Organism Cell Biology and Regulatory Sequence Discoveryin Comparative Functional Genomics

The use of bioinformatics to integrate phenotypic and genomic data from mammalian models is well established as a means of understanding human biology and disease. Beyond direct biomedical applications of these approaches in predicting structure–function relationships between coding sequences and protein activities, comparative studies also promote understanding of molecular evolution and the relationship between genomic sequence and morphological and physiological specialization. Recently recognized is the potential of comparative studies to identify functionally significant regulatory regions and to generate experimentally testable hypotheses that contribute to understanding mechanisms that regulate gene expression, including transcriptional activity, alternative splicing and transcript stability. Functional tests of hypotheses generated by computational approaches require experimentally tractable in vitro systems, including cell cultures. Comparative sequence analysis strategies that use genomic sequences from a variety of evolutionarily diverse organisms are critical for identifying conserved regulatory motifs in the 5′-upstream, 3′-downstream and introns of genes. Genomic sequences and gene orthologues in the first aquatic vertebrate and protovertebrate organisms to be fully sequenced (Fugu rubripes, Ciona intestinalis, Tetraodon nigroviridis, Danio rerio) as well as in the elasmobranchs, spiny dogfish shark (Squalus acanthias) and little skate (Raja erinacea), and marine invertebrate models such as the sea urchin (Strongylocentrotus purpuratus) are valuable in the prediction of putative genomic regulatory regions. Cell cultures have been derived for these and other model species. Data and tools resulting from these kinds of studies will contribute to understanding transcriptional regulation of biomedically important genes and provide new avenues for medical therapeutics and disease prevention.

the species and provided additional insights about genomic evolution. Another area of interest that stands to benefit greatly from genomic data is that of gene regulation. Understanding in this field is fragmented and, as with any scientific discipline, the concepts and questions that can be conceived and addressed experimentally are dependent on available technological approaches. Mechanisms by which expression of a single gene is regulated can be extremely complicated. Multiple phosphorylation-or ligand-dependent nuclear receptors that homo-or heterodimerize may be required to achieve activity. Each of these receptors may have different activation specificity or duration, even when acting via the same regulatory DNA sequence such as classical proximal promoter elements. These receptors may also work in combination with other transcription factors that function at sites more distal from the proximal promoter or in introns. Alternatively spliced transcripts represent another complex aspect of gene expression regulation that is influenced by extracellular and intracellular signaling but is not well understood (Stamm et al. 2005) . Furthermore, individual genes often are part of a broader, coordinately regulated network of genes that function to elicit a set of cellular responses (Wagner 1999) . Through such mechanisms, ligands may, for instance, induce their own metabolism or export, a process that further complicates understanding of gene regulation and that also has critical implications for models of pharmacokinetics and drug efficacy. Experimental identification of functional genomic sequences depends heavily on cell culture and other techniques of in vitro cell biology. Traditionally, identification of gene regulatory regions has been limited by the labor-intensiveness of the requisite strategies. Generally only regions close to the transcriptional start sites have been experimentally tractable for detailed examination, despite evidence that important regulatory regions exist more than 10 kb upstream or downstream from the coding region or in introns of genes (Rowntree et al. 2001) . Identification of specific functional sequences through the generation of deletion constructs is limited in the sequence size that can be analyzed and is restricted to examination of single genes. Transfection of cells with reporter constructs containing putative proximal promoters may elicit strong activation when treated with receptor-specific ligands in culture, while in vivo studies may yield inconsistent results, suggesting that these receptors and transcription factors are a subset of a larger complex of regulators of gene expression. Techniques such as DNASE I hypersensitivity studies and gel shifts are excellent methods for testing functionality of putative regulatory elements; however, they are not efficient for screening candidate sequences. By contrast, computational analyses allow the examination of a significantly larger region when predicting conserved regulatory regions and signals. Computational techniques lend themselves to the identification of patterns or clusters of regulatory motifs and prediction of coregulated genes, and generate targeted predictions of candidate regulatory sequences and signaling molecules that can then be directly tested in functional and mutagenesis studies (Hughes et al. 2000; Loots et al. 2000; Pennacchio and Rubin 2003; Ovcharenko et al. 2005) . Understanding causative relationships between specific regulatory elements and expression patterns would greatly enhance the ability to predict disease-associated genes (Pennacchio and Rubin 2003) . However, even in computational approaches the ability to identify and predict the regulatory functions of non-coding sequences has been limited (Pennacchio and Rubin 2003) . Pairwise comparisons have helped to predict functionally conserved regions, however the statistical accuracy of these predictions is increased when more than two sequences are used (Dubchak et al. 2000) . Several studies suggest that comparative analysis of multiple evolutionarily diverse organisms facilitate the prediction of functionally important non-coding regions (Dubchak et al. 2000; Matys et al. 2003; Thomas et al. 2003) . Comparisons of genome sequences from evolutionarily diverse organisms also elucidate regulatory regions that are specific to a particular species or group of species. Recent additions of the African tree frog (Xenopus tropicalis), chicken (Gallus gallus) and dog (Canis familiaris) to the sequencing pipeline will provide important complementary perspectives for tetrapods, but data from more divergent vertebrate genomes is still needed. Different rates of molecular evolution, including gene duplication, underscore the importance of the examination of genomic sequences from an evolutionary range of organisms for comparative sequence analyses. When such an approach is taken, the contrast of divergent sequences differentiates between functionally conserved regions and generally conserved regions that simply reflect a lack of divergence time (Dubchak et al. 2000) . Using sequences from evolutionarily diverse organisms may provide the necessary divergence to identify functionally conserved regions in genes that have evolved slowly. The increasing availability of genomic data for non-mammalian organisms and similarities to the human genome underscore the value of these organisms as models of a variety of diseases in humans (Aparicio et al. 2002; Dehal et al. 2002; Ballatori et al. 2003) . The identification of conserved coding and regulatory regions is enhanced by including divergent sequences in comparative studies (Thomas et al. 2003 ) because these sequences provide more stringent filters for detecting conserved, and presumably functionally important elements (Dubchak et al. 2000; Thomas et al. 2003; Ahituv et al. 2004) . Consequently, sequences from evolutionarily distant marine vertebrates and protovertebrates are being used in comparative studies with increasing frequency. The pufferfish (Fugu rubripes) genome sequencing project supported this approach as it led to the discovery of nearly 1000 human genes not previously described (Aparicio et al. 2002) . Sequence comparisons of the Hoxb-4 gene in mouse and Fugu identified novel regulatory elements that directed subsets of the full Hoxb-4 expression pattern in transgenic mice (Aparicio et al. 1995; Amores et al. 2004) . In a recent comparative analysis of the HoxA cluster in human, horned shark and zebrafish (Chiu et al. 2002) , extensive conservation of non-coding sequence motifs was found between the human and shark sequences, whereas zebrafish sequences exhibited significant loss of conservation. The majority of newly identified regulatory elements for this cluster of genes were identical to known binding sites for regulatory proteins as defined in the transcription factor database, TRANSFAC (http://www.biobase.de; Matys et al. 2003) , demonstrating the accuracy of this approach (Matys et al. 2003; Santini et al. 2003) . Special physiological attributes exhibited by many evolutionarily diverse aquatic organisms have led to the increased appreciation of their importance as models of human disease. These characteristics have been exploited experimentally to further understanding of human immunology, genomics, stem cell and cancer biology, pharmacology, toxicology and neurobiology. Historically, marine vertebrates provided critical insights into fundamental mechanisms of physiological processes, and the value of these organisms has not been diminished by the advent of molecular approaches. Membrane transporters that are the sites of action of diuretic drugs were first cloned from specialized organs in marine species (Gamba et al. 1993; Xu et al. 1994) . Mutagenesis studies in bony fish generated a spectrum of biologically relevant and distinct phenotypes (Naruse et al. 2004; Walter et al. 2004 ). Large-scale genetic screens produced more than 500 zebrafish mutants, many with phenotypes similar to human disorders (Dooley and Zon 2000) . Immunogenetic studies in carp, salmonids and other species are contributing to the growing database of marine vertebrate genomics Fujiki et al. 2001 Fujiki et al. , 2003 Tomana et al. 2002; Ishikawa et al. 2004) . Although increasing numbers of vertebrate genomes are being sequenced, there are still stretches of evolutionary history without representation (Thomas and Touchman 2002) . Among aquatic organisms, teleosts are represented by significant EST and genomic sequencing, such as those in zebrafish and pufferfish. Until recently, chondrichthyes, which include elasmobranchs (sharks, rays and skates) were the only major line of gnathostomes, or jawed vertebrates, for which there are no major initiatives to generate genomic sequences. Sequence data from elasmobranchs have provided unique insights into conserved functional domains of genes associated with human liver function, multidrug resistance, cystic fibrosis, G protein coupled receptors, natriuretic peptide receptors, and other biomedically relevant genes (Valentich and Forrest 1991; Henson et al. 1997; Aller et al. 1999; Silva et al. 1999; Greger et al. 1999; Waldegger et al. 1999; Ke et al. 2002; Wang et al. 2002; Yang et al. 2002; Cai et al. 2001 Cai et al. , 2003 Mattingly et al. 2004b) . In March, 2005 the National Human Genome Research Institute of the National Institutes of Health (NIH), USA, announced that it will fund the whole genome sequencing of Raja erinacea. In the announcement of the Skate Genome Sequencing Initiative (http://www.genome.gov/ 13014443), NIH states: 'The skate (related to many species of shark and cartilaginous fish) was chosen because it belongs to the first group of primitive vertebrates that developed jaws, an important step in vertebrate evolution. Other innovations in this group of animals include an adaptive immune system similar to that of humans, a closed and pressurized circulatory system, and myelination of the nervous system. Understanding these systems of the skate at a genetic level will help scientists identify the minimum set of genes that create a nervous system or develop a jaw, possibly illustrating how these systems have evolved in humans, and how they sometimes go wrong.' Chondrichthyes (cartilaginous) fish appeared approximately 450 million years ago. Elasmobranchs comprise most chondrichthyan organisms. They exhibit fundamental vertebrate characteristics, including a recombinatorial immune system (Hinds and Litman 1986; Adelman et al. 2004) , and are also the oldest existing vertebrates with circulatory system-related signaling molecules and receptors such as platelet-derived growth factor and adenosine receptors. The specialized rectal gland of Squalus acanthias, the spiny dogfish shark, has greatly facilitated the study of cystic fibrosis, sodium and chloride secretion (Devor et al. 1995; Lehrich et al. 1995; Forrest 1996; Henson et al. 1997; Lehrich et al. 1998; Silva et al. 1999; Aller et al. 1999; Greger et al. 1999; Waldegger et al. 1999; Ke et al. 2002; Yang et al. 2002) . Unlike many primary cell cultures that dedifferentiate and lose transport polarity immediately after isolation, primary cultures of shark rectal gland tubular cells maintain fully differentiated function and expression of all known receptors, transporters, ion channels and signal transduction pathways in vitro (Valentich and Forrest 1991; Devor et al. 1995; Lehrich et al. 1995; Aller et al. 1999; Greger et al. 1999; Waldegger et al. 1999; Ke et al. 2002; Yang et al. 2002; Mattingly et al. 2004b) . A comparison of the properties of the cloned shark and human cystic fibrosis transmembrane regulator (CFTR) has provided insights into structural domains related to functional differences in the normal and mutant proteins (Marshall et al. 1991) . The spiny dogfish shark CFTR protein is 72% identical to the human ortholog and comparison of the human and shark CFTR sequences revealed conservation of five cyclic AMP-dependent kinase phosphorylation sites and three residues that, when mutated in the human protein, are associated with cystic fibrosis. The coding sequences and functions of a number of medically relevant genes are conserved in the spiny dogfish shark and little skate (Cai et al. 2001 (Cai et al. , 2003 Wang et al. 2002; Yang et al. 2002; Mattingly et al. 2004b ). Primary hepatocytes from little skate retain hepatobiliary polarity for at least 8 h and possibly up to several days in culture, offering particular advantages for studies of liverspecific functions . Genomic information applied to existing physiological data in these systems, along with the further development of in vitro cell culture systems, will allow the testing of molecular hypotheses and understanding of regulatory mechanisms that are directly applicable to human biology. Targeted sequencing of well-defined genomic regions generated from BAC clones has been extremely useful in providing supporting genomic information in species for which complete genome data are not available . In addition to constructing four-fold coverage bacterial artificial chromosome libraries from sperm DNA of dogfish shark and little skate (http://www.mdibl.org/research/ skategenome.shtml), an expressed sequence tag (EST) sequencing project is underway to substantially increase the availability of sequence data for these model organisms (Mattingly et al. 2004b ). Over 10,000 sequences are publicly available through the EST database at the National Center for Biotechnology Information (dbEST; Boguski et al. 1993 ) and http://www.mdibl.org/decypher. These data sets are updated as new sequences become available. One of the remarkable findings of the human genome sequencing project was the discovery that coding regions account for only 5% of the genome (Venter et al. 2001) . The remaining sequence consists of repetitive DNA ($40-45%) and extensive non-coding regions for which there is very little functional information. It is within these regions that significant regulatory information presumably is concealed. The major challenge in the postgenomic age is uncovering important functional regions within this non-coding DNA. The rapid development of sequence analysis software tools, increasing availability of genomic data, and recognition of the importance of comparing data from diverse organisms are allowing scientists to make fruitful inroads to understanding genomic structure and gene expression regulation. A brief summary of software tools that are valuable for identifying regulatory information from crossspecies comparative analyses follows. The University of California Santa Cruz Genomic Browser (http://genome.ucsc.edu/; Karolchik et al. 2003 ) is among the most popular databases for querying and retrieving genomic sequences of interest. It currently provides access to genomic assemblies from 23 organisms, including 10 vertebrates, 8 insects, 2 nematodes, a sea squirt, baker's yeast, and the SARS virus. Whether genomic regions are retrieved from an existing database or sequenced locally, there are an increasing number of options for analysis and annotation. Several computational tools allow identification of genes and exon boundaries in genomic sequences. GENSCAN (Burge and Karlin 1997) , TWINSCAN (Korf et al. 2001) , MZEF (Zhang 1997) , and Gene Recognition and Analysis Internet Link (GRAIL; Uberbacher and Mural 1991) are among the most popular. Most gene finding programs were optimized to predict genes in sequences from mammalian models; as a result accuracy is sometimes reduced when using sequences from more divergent organisms or nonvertebrates. These programs use different strategies and are based on current, but incomplete understanding of gene structure. Therefore, predictions from multiple programs should be combined computationally to enhance accuracy and confidence (Rogic et al. 2002) . Quality control strategies can be used to increase accuracy of and confidence in results from gene finding tools. First, the abundance of repetitive elements in genomic sequences can distort predictions of genes and exon locations. To counter this problem, masking genomic sequences is recommended, before gene analysis, using a program like RepeatMasker (A.F. A. Smit et al. 1996, unpublished) . The effectiveness of Repeat-Masker with marine and other evolutionarily divergent organisms is not yet clear because interspersed repeats are specific to a species or group of species. Sequences from such organisms should be evaluated under masked and unmasked conditions. Second, aligning ESTs or cDNAs with genomic sequences using programs like Spidey (http://www.ncbi.nlm.nih.gov/spidey/; Wheelan et al. 2001 ) refines predictions of exon and intron boundaries, promoter regions, and splice sites. This approach presents challenges for transcripts that are expressed at very low levels, have significant tissue-or age-specific requirements, or are from species for which minimal sequencing has been done (Schwartz et al. 2000) . A new, publicly available resource, the Comparative Toxicogenomics Database (CTD; http://ctd.mdibl.org; Mattingly et al. 2003 Mattingly et al. , 2004a , provides multiple alignment and phylogenetic analysis results with sequences from diverse organisms for biomedically significant genes and proteins. CTD provides access to data valuable for identifying homologous genomic sequences and confirming gene and gene feature predictions. Identification of gene features greatly improves interpretation of subsequent multiple sequence analysis results. Aligning multiple genomic sequences is becoming widely accepted as a powerful mechanism for identifying important functional regions such as regulatory elements. MultiPipMaker (http:// pipmaker.bx.psu.edu/pipmaker/; Schwartz et al. 2003 ) and mVISTA (http://gsd.lbl.gov/vista/index.shtml; Bray et al. 2003; Frazer et al. 2004 ) are two of the most popular web servers that conveniently combine alignment engines with visualization capabilities. MultiPipMaker, and a new server zPicture (http://zpicture.dcode.org/; Ovcharenko et al. 2004) , use the local alignment program BLASTZ as their alignment engine (Schwartz et al. 2000) ; VISTA uses AVID, a global alignment program (Bray et al. 2003; Frazer et al. 2004 ). Local alignment tools find high-scoring, short matching segments and extend these regions based on a scoring threshold. Local alignments may permit greater diversity between sequences by finding regional similarities. Segments of similarity need not be conserved in order or orientation. This feature may be advantageous for finding conserved transcription factor binding sites, which are very short and prone to reordering (Bray et al. 2003) . High similarity of short regions, however, does not necessarily imply homology (a gene derived from a common ancestral gene) and can lead to false implications of relatedness among sequences that are not homologs. By contrast, global alignments do require that the order and orientation of similar regions is conserved, because similar architecture is often observed in homologous sequences (Bray et al. 2003) . MultiPipMaker and VISTA provide visualization options for alignments that include percent similarity and usersubmitted annotations (e.g., exon locations, repetitive elements). It is important to note that local and global alignment tools are being refined so rapidly that it is becoming difficult to distinguish between them (Frazer et al. 2004) . Readers are referred to two recent reviews of alignment programs for detailed comparisons (Frazer et al. 2003; Pollard et al. 2004) . A major challenge to identifying transcription factor binding sites (TFBSs) is that they tend to be short, degenerate, and occur frequently throughout the genome. Analysis of a single sequence usually leads to an abundance of false positive predictions of TFBSs. Several programs have been developed to respond to this challenge based on two principles. First, functional regulatory elements are often conserved evolutionarily; therefore, identifying TFBSs that are conserved or aligned in multiple sequences may effectively filter false positive predictions (Ovcharenko et al. 2005) . Second, gene expression often results from coordinate activation of multiple, proximal regulatory elements; therefore identifying TFBSs in clusters, rather than isolation, may enhance confidence in the functionality of predicted TFBSs. These principles have been leveraged, albeit differently, by rVISTA (http:// gsd.lbl.gov/vista/index.shtml; Loots et al. 2002; Loots and Ovcharenko 2004) , which is a member of the VISTA suite of tools and is also integrated with zPicture, and the newly launched Mulan (http://mulan.dcode.org/; Ovcharenko et al. 2005) . Both servers use profiles of transcription factor binding sites from the TRANSFAC database (http://www.biobase.de; Matys et al. 2003) . Because TRANSFAC and other similar resources only contain information for known transcription factor binding sites, they are inherently incomplete. Furthermore, existing tools do not address other important regulatory features such as properties related to protein-protein interactions and chromatin structure, and clusters of binding sites that may have been reshuffled between organisms over evolutionary time (Loots et al. 2002) . Cell culture of marine genomic model species and experimental verification of predictions from comparative analysis of genomic sequences Using in vitro cell culture systems, the functional significance of conserved, putative regulatory sequences predicted through comparative computational analysis in, for instance, the 5¢upstream region of select genes can be tested experimentally. The availability of sufficient genomic information facilitates targeted studies to evaluate such potential functional regulatory regions. Comparative experimental studies can be designed employing cell lines derived from any species, though mammalian cell lines have thus far been favored (Mather and Barnes 1997; Barnes and Sato 2000) . Reporter constructs containing regulatory regions of select genes can be generated using up to 5 kb upstream of the transcriptional start site of the relevant genes; these are inserted upstream of a reporter gene such as that for an enhanced fluorescent protein. In the last decade, significant progress has been made in development of marine and freshwater organism cell lines with utility for genomic studies. A variety of zebrafish cell lines have been developed, some of which maintain a normal karyotype for extended periods in vitro (Barnes and Collodi 2005) . Zebrafish cells in culture can be transfected with plasmid DNA using adaptations of approaches common for mammalian cells in vitro, and transient transfection methods have identified expression of genes under control of a number of mammalian and fish promoters. In addition, cell cultures from pufferfish (genera Fugu and Tetraodon) provide a biological complement to the genomic libraries derived to study the molecular biology of these animals, allowing the extension of this model to experiments in functional genomics. Multipassage cell cultures have been established from embryo and adult tissues of species of both of these pufferfish genera (Barnes and Collodi 2005) . One of the Fugu rubripes cultures has been maintained for more than 200 population doublings, and flow cytometry showed that the relative amount of DNA present in cultured cells was approximately 15% of that in human cells, as predicted by biochemical analysis. Telomerase, an enzyme associated with indefinite proliferation in mammalian cell cultures, was easily detectable in these cells, suggesting that the cultures are capable of indefinite growth. Until recently, the lack of in vitro culture systems for elasmobranch models has been a major limitation for the use of these species in comparative functional genomics. Cold-water marine animals such as the little skate and dogfish shark are useful models for physiology and genomics, but homologous in vitro systems with which investigators can test hypotheses or confirm predictions at a molecular level are essential for widespread use of these models. Expression of elasmobranch genes in heterologous systems such as mammalian cell cultures or Xenopus oocytes may be compromised by differences in membrane lipid composition and missing or interfering accessory and messenger proteins. Heterologous systems are also not adaptable to mechanistic studies using genetically altered, dominant negative molecules. Another advantage derived from cell cultures from species such as shark and skate (that may be only seasonably available) is that they make biological material available yearround in any laboratory. Studies on regulation of gene expression in dogfish shark and little skate are beginning to benefit from the expanding genomic databases for these animals. Skate embryo-derived cell cultures also provide new avenues for elasmobranch research in embryology and organogenesis, toxicology, neurobiology, genome regulation and comparative stem cell biology. Cells have been cultured from the rectal gland, eye, brain, kidney and early embryo of Squalus acanthias. Medium is either LDF, developed for zebrafish cells (Barnes and Collodi 2005) or VCM (Valentich and Forrest 1991) , a urea-timethylamine oxide (TMAO)-containing medium that supports the short-term culture of shark rectal gland cells. In some cases cells are plated onto dishes pretreated with collagen. Medium is further supplemented with a variety of cell typespecific peptide growth factors, nutrients and purified proteins at a range of concentrations. These include insulin, transferrin, epidermal growth factor (EGF), basic fibroblast growth factor (FGF), transforming growth factor-beta, vitamins A and E, selenium, mercaptoethanol, dexamethasone, fetal calf serum, shark serum and shark yolk extract. For example shark brain cells can be grown in primary culture in VCM containing EGF, FGF, insulin, transferrin, selenium, a chemically-defined lipid supplement and vitamin E on a fibronectin substratum (Figure 1 ). Expression of genes from plasmids has been achieved in primary dogfish shark rectal gland cell cultures by lipofection using the CMV promoter to direct expression of the gene of interest ( Figure 2 ). Differentiated function also is maintained, as evidenced by expression of CFTR and vasoactive intestinal peptide receptor (VIPR) mRNA detected by reverse-transcription polymerase chain reaction (RT-PCR) assay (Figures 3 and 4) . Assay for telomerase on cultures of a variety of dogfish shark and little skate cell types showed that all cultures tested were positive, although specific activity varied almost 100-fold among different cell types. Assay for cell proliferation by in situ bromo-deoxyuridine (BrdU) incorporation also was carried out on cultures from shark brain, kidney and rectal gland. The assay involves an overnight incubation with BrdU, followed by an immunoassay identifying cells synthesizing DNA during the time of incubation and incorporating the nucleoside analogue. The results showed that cells in the cultures were synthesizing DNA. The most active synthesis was seen in cultures from early shark embryos. Medium was supplemented with insulin, tranferrin, selenous acid, EGF, FGF, Lglutamine, chemically defined lipids, non-essential amino acids, heat-inactivated fetal bovine serum (heat treated for 30 min at 56°C) and shark yolk extract. Conditioned medium from the cells was stimulatory for other shark cell cultures, including the shark rectal gland cells ( Figure 5) . A normalized c-DNA library has been made from these cells and attempts will be made to identify the cell type by extensive EST analysis. In addition to the scarcity of cell lines from marine vertebrates, a persistent absence of cell lines from non-arthropod invertebrates has stymied research on many valuable species useful in comparative genomics and a variety of other disciplines (Bayne 1998; Rinkevich 1999) . Selection of the echinoderm Strongylocentrotus purpuratus (purple sea urchin) as a model species for genomic sequencing has enhanced the need for cell lines from this species in particular. We recently have explored cell culture using both this species and the closely related Strongylocentrotus droebachi- ensis (green sea urchin). Cells from Polian vesicles and axial organ (Figures 6 and 7) appear to have proliferated in vitro for more than 8 months (Parton and Bayne 2004, 2005) . Other cultures yielded thraustochytrid protists that are common parasites of marine invertebrates worldwide (Rinkevich 1999) . Basal nutrient culture medium was LDF diluted with 4 volumes of filtered sea water (Kawamura and Fujiwara 1995) with antibiotics (penicillin 200 U/ml, streptomycin 200 lg/ ml, ampicillin 25 lg/ml), sodium bicarbonate at 0.18 g/l and HEPES at 15 mM. Culture vessels were held at 20°C in ambient air. Additives included fetal calf serum (heat inactivated) at 1-3%; insulin at 10 lg/ml; transferrin at 10 lg/ml; FGF at 50 ng/ml; EGF at 50 ng/ml; b-mercapto-etha-nol at 55 lM; chemically defined lipid supplement at 1 ll/ml; selenium at 10 nM, and L-glutamine at 200 lM. Transcriptional regulation involves complex interactions of diverse proteins, or transcription factors, with specific DNA sequences in the noncoding regions of target genes. Furthermore, cells respond to environmental stimuli and to developmental signals by altering expression of gene networks. The limited number of transcription factors suggests that few, if any of these proteins exert their activity exclusively on a single gene; rather, they bind to conserved sites in several genes to coordinate their expression (Wagner 1999; Pennacchio and Rubin 2003) . In situations in which there are no clinically acceptable inhibitors or other modulators of clinically relevant proteins, elucidating mechanisms by which these genes are regulated and identifying other coordinately regulated genes may reveal novel strategies by which disease processes may be disrupted or controlled. Comparative studies provide insight into the possible common ancestor of a gene, trace the accumulation of mutations over time and suggest selective pressures that influence the expression and functions of genes (Pennacchio and Rubin 2003) . Comparisons have provided useful predictions about which sequences are minimally essential for function as well as those that may be important on a species-specific level (Aparicio et al. 1995) . Comparative genomic computational approaches continue to identify conserved regions in non-coding sequences. The challenges will be to determine which sequences are functionally significant and to identify coordinately regulated genes and common regulatory pathways. Understanding such networks, the transcription factor binding sites and genes involved in disease states may reveal alternative points of intervention and contribute to a more predictive approach to molecular medicine. Significant progress has been made in the development of powerful sequence analysis tools. Although often optimized for sequences from more traditional model organisms, they offer great value for comparative studies that include evolutionarily divergent sequences. They should be used cautiously, however, with awareness of their limitations, which are largely a reflection of current scientific knowledge and understanding of genome structure. As the diversity of available sequence data continues to increase, it will drive refinements and development of new tools and provide important insights into the evolution and essential mechanisms of gene expression regulation.

Regulated multicistronic expression technology for mammalian metabolic engineering

Contemporary basic research is rapidly revealing increasingly complex molecular regulatory networks which are often interconnected via key signal integrators. These connections among regulatory and catalytic networks often frustrate bioengineers as promising metabolic engineering strategies are bypassed by compensatory metabolic responses or cause unexpected, undesired outcomes such as apoptosis, product protein degradation or inappropriate post- translational modification. Therefore, for metabolic engineering to achieve greater success in mammalian cell culture processes and to become important for future applications such as gene therapy and tissue engineering, this technology must be enhanced to allow simultaneous, in cases conditional, reshaping of metabolic pathways to access difficult-to-attain cell states. Recent advances in this new territory of multigene metabolic engineering are intimately linked to the development of multicistronic expression technology which allows the simultaneous, and in some cases, regulated expression of several genes in mammalian cells. Here we review recent achievements in multicistronic expression technology in view of multigene metabolic engineering.

There are two general levels of genetic engineering in which a suitable production cell line is generated; (i) stable introduction of the genetic information for the product protein and (ii) an optional metabolic engineering step to improve cellular activities by the manipulation of enzymatic, transport, and regulatory functions of the cell (Bailey, 1991) . Metabolic engineering of animal cells has already been proven useful for improving diverse key characteristics of cultured cells including cell viability (apoptosis engineering: Cotter and Al-Rubeai, 1995; Mastrangelo and Betenbaugh, 1998) , product quality (glycosylation engineering: Bailey et al., 1998; Jenkins et al., 1996) , product yield (controlled proliferation technology: Fussenegger et al., 1997a; Fussenegger et al., 1998a; Fussenegger et al., 1998b; Papoutsakis, 1998) and growth in protein-free medium (cell-cycle engi- * Author for all correspondence. neering: Renner et al., 1995; Lee et al., 1996; Rivard et al., 1996; Greulich and Erikson, 1998) . Most of these successes have been realised by the addition of a single gene to the host cell's genome. However, just as single-gene interpretations of human disease have limited scope (Lander and Schork, 1994) , one-gene metabolic engineering cultured cells cannot access anything approaching the full potential set of useful engineered phenotypes (Papoutsakis, 1998) . Owing to the genetic complexity of higher eukaryotic cells and the absence of sophisticated genetic tools (compared to those for several microbial hosts), introduction of heterologous genetic information into mammalian hosts is usually achieved by cotransfection of a selection marker and the gene of interest with subsequent selection for clones containing the marker, and as empirical experience has shown, often also include the cotransfected gene (Kaufman and Sharp, 1982) . Many undesired phenomena accompany this haphazard genetic engineering of mammalian cells because of the undefined, mechanistically obscure selection of random integration sites in different stable clones, giving rise to variability in product expression levels, genetic stability, and second order effects on growth, viability, and productivity resulting from disruption of host genes (or regulatory loci) at the integration site. Recently, chromosomal locations of some industrially relevant mammalian cells lines have been found which show high transcription and stability for integration of transgenes (Karreman et al., 1996) . Screening for such sites is a time-consuming process that involves establishment of a genetic platform for subsequent targeted integration. However, unlike the situation in mouse ES stem cells (Hicks et al., 1997) gene targeting is difficult to achieve in most industrially relevant cell lines because they seem to lack necessary basic recombination machinery, and therefore they require installation of complex heterologous site-specific recombination systems (Fukushige and Sauer, 1992; Karreman et al., 1996) . Regardless of the method of integration and the chosen combination of product and metabolic engineering genes, it is desirable to manipulate the cell in a minimal number of steps. This goal is addressed by technology for simultaneous cloning and subsequent expression of multiple genes in a desired host. Besides providing a platform for future metabolic engineering breakthroughs, multicistronic expression technology should speed basic functional genomic research and new applications in tissue engineering and gene therapy. Here we review recent developments in multicistronic expression technology and their use to enable one-step multigene metabolic engineering, positive feedback regulation circuits and auto-selective expression systems. Bacteria have evolved expression units called operons which unite functionally related genes under the control of a single promoter, thus enabling coordinated, simultaneous and rapid expression of metabolically coordinated genes in response to specific environmental signals or physiological constraints (the classic example is the lactose operon; Dickson et al., 1975) . Individual genes in a bacterial operon are preceded by characteristic sequences, socalled ribosomal binding sites (RBS), for translation-initiation at appropriate points within a single mRNA molecule. In contrast to bacterial multigene transcripts, most eukaryotic mRNAs are monocistronic, and optimal translation of the encoded gene relies on a post-transcriptional 5 modification (capping) for ribosome binding and subsequent AUG-scanning (Shatkin, 1985; Kozak, 1989) . However, other capindependent modes of translation-initiation such as leaky scanning, termination-reinitiation, and internal initiation are used in rare cases (Jackson et al., 1995; Table 1; Figure 1 ). As part of their pathogenic life cycle, picornaviruses have evolved specific genetic elements (internal ribosomal entry sites; IRES) in their 5 nontranslated leader regions (ntr) which adopt a particular secondary structure able to attract eukaryotic ribosomes and to allow internal translation-initiation (Belsham and Sonenberg, 1996 ; Table 1 ). The pivotal role of IRES in picornaviral pathogenesis is based on the expression of a viral protease which cleaves the host cap-binding translation-initiation factor eIF4G and allows redirection of host translation machinery for exclusive translation-initiation of IRES-containing viral mRNAs (Etchison et al., 1982; Pelletier and Sonenberg, 1988; Jackson et al., 1990; Belsham and Sonenberg, 1996; Rueckert, 1996) . IRES-like elements are present in other viral systems and were recently discovered in eukaryotic cells which give certain mRNA molecules cap-independent translation ability in response to viral infection or stress conditions, as was shown for immunoglobulin heavy chain binding protein (Bip) and the cap-binding protein eIF4G (Macejak and Sarnow, 1991; Gan and Rhoads, 1996) (Table 1) . Cap-independent translation can also enforce an alternative translation start site, resulting in translation of different proteins from the same mRNA, such as that mediated by the human fibroblast growth factor 2 (FGF2) (Vagner et al., 1995) . Recently, IRES elements were also identified in the translation regulation of developmentally regulated genes such as the homeotic gene Antennapedia or Ultrabithorax of Drosophila (Oh et al., 1992; Ye et al., 1997) , the genes for human insulinlike growth factor (IGF-II) (Teerink et al., 1995) , and the platelet-derived growth factor B (developmental IRES or D-IRES; Bernstein et al., 1997) . The potential for cap-independent translation-initiation has also been found in yeast and Xenopus oocytes (Iizuka et al., 1994; Kneiper and Rhoads, 1997) . Furthermore, the finding of an internal ribosomal entry site in the 5 untranslated region of c-myc suggests that IRES- Figure 1 . Strategies for simultaneous and in some cases regulated expression of more than one gene in mammalian cells. Key genetic elements for expression in mammalian cells such as the promoter (P), internal ribosomal entry sites of polioviralorigin (IRES) or derived from the encephalomyocarditis virus (CITE), the splice donor (SD) and acceptor (SA) and the polyadenylation site (PA) as well as for regulated gene expression including the tetracycline-responsive transactivator (tTA) and the tet-responsive promoter (P hCMV * −1 ) are indicated. In some cases translation is shown below the genetic configuration (mRNA, ribosome, proteins). Iizuka et al., 1994 mediated translational control may be vital for higher organisms as aberrant translational regulation of cmyc is likely to play a role in tumorigenesis (Stoneley et al., 1998) . Despite the potential of IRES as key elements of operon-like multicistronic expression units in mammalian genomes, such genetic configurations seem to have rarely evolved in a natural context, perhaps because most complex and fine-tuned regulatory cir-cuits in mammalian cells are best configured with independent regulation of individual genes. Since the transcription and translation of separate cotransfected genes is not strictly correlated, the reliability of product expression based on selection of the cotransfected marker gene can be very low. Further- more, even under high selection pressure, the genetic stability of the expressed product can not be assured in long term cultivations. The combination of product and marker genes on the same vector does not completely alleviate these complications. For these reasons, dicistronic genetic configurations were developed. The first dicistronic constructs used IRES elements of picornaviral origin or from encephalomyocarditis virus (EMCV) for cap-independent translation of the second cistron while the first cistron relied on classical cap-dependent translation-initiation (Pelletier and Sonnenberg, 1988; Kaufman et al., 1991) . Although the two IRES elements differ completely at the sequence level, their secondary structure is very similar and typical for such internal translation initiators. A large number of dicistronic product-marker configurations have since been developed for many different applications. Table 2 gives an overview of recent dicistronic expression vectors. Although genetic combinations used for dicistronic expression vary among different applications, EMCV and picornaviral IRES remained the most popular cap-independent translation-initiators for dicistronic configurations because these elements function in a wide variety of cell lines including the industrially relevant CHO and BHK cell lines (Borman et al., 1997) . However, recent reports of varying translation-initiation capabilities of IRES in different host cell environments and discovery of new IRES elements is stimulating new development to apply these IRES elements for multicistronic expression (Bernstein et al., 1997; Schumacher and Wirth, 1997) . Dicistronic genetic configurations which contain the marker gene in the second cistron enable autoselective expression in addition to simultaneous and coordinated gene expression. Resistance to the marker gene or expression of the reporter gene is only possible if all 5 -encoded genetic elements are intact. This intrinsic self-selective program was found to be very reliable, with nearly all of the resistant cells also expressing the desired product gene (Gurtu et al., 1996; Rees et al., 1996) . Furthermore, product-marker configurations can also be used for efficient generation and screening of high producing cell clones: IRES-based translation-initiation of the second cistron is usually less efficient compared to cap-dependent translation, and can be decreased further by loss-infunction mutations of the IRES or the marker gene itself. The overall lower translation efficiency or activity is then compensated under high selective pressure by integration of the dicistronic expression unit into chromosomal sites with high transcriptional activity (Kaufman et al., 1991; Gurtu et al., 1996; Rees et al., 1996) . Certainly, simultaneous expression of two genetic traits can also be achieved by gene fusions (Krömer et al., 1997 ; pTracer plasmids of Clontech) or recently developed splicing expression technology (Lucas et al., 1996; Figure 1 ), but gene fusion strategies are limited in functional applications or may lead to fusion products with altered physiologic specificities, and splicing-based two-gene expression leads to unequal expression levels of both proteins. Only IRESbased dicistronic expression guarantees simultaneous and coordinated expression of both transgenes at comparable levels for multi-subunit proteins (for example antibodies) which enables genetic configurations for a wide variety of contemporary research and development applications which are also listed in Table 2 (Dirks et al., 1993; Dirks et al., 1994; Fussenegger et al., 1997a) . Furthermore, IRES-mediated expression systems can be extended beyond the dicistronic level to tri-or even quattrocistronic artificial eukaryotic operons (Fussenegger et al., 1997b; Fussenegger et al., 1998c) . Despite the numerous expression vectors available containing dicistronic expression units ( Table 2) , most of these expression systems express a marker or reporter gene in a fixed configuration, leaving only one cistron free for heterologous gene expression. However, for one-step transfection of a product protein, metabolic engineering, and a selection marker in a single expression unit, multicistronic artificial mammalian operons with 3 or even 4 cistrons are desirable. We recently reported the construction of a novel vector family, pTRIDENT, for tricistronic gene expression in mammalian cells (Fussenegger et al., 1998c; Fig-ure 2) . A single promoter allows high level expression and, in some vectors, adjustable transcription of all three genes. Whereas the first cistron is translated in a classical cap-dependent manner, translation-initiation of the subsequent two cistrons rely on IRES elements of picornaviral (IRES; pTRIDENT1) and EMCV origin (denoted here CITE, cap-independent translation enhancer; third cistron; pTRIDENT3; Fussenegger et al., 1998c) . Tricistronic pTRIDENT1-and pTRIDENT3-derived test vectors encoding the model product gene SEAP (secreted alkaline phosphatase; first cistron), a metabolic engineering determinant (the cyclindependent kinase inhibitor p21 (CDI) second cistron), and the reporter gene GFP (green fluorescent protein; third cistron) were transfected into a CHO cell derivative which allows tetracycline-responsive gene expression (Fussenegger et al., 1998c) . Both tricistronic configurations were stable in CHO cells and showed strict simultaneous, coordinated as well as regulated expression of all three cistrons. The expression levels of individual cistrons were assessed by comparison to respective values of isogenic monocistronic expression vectors. Although expression levels of genes encoded on different cistrons are largely dependent on the overall stability of the polycistronic mRNA and therefore a direct function of the genetic configuration of encoded genes, our test vectors showed similar expression levels to those provided by the monocistronic vector on the first two cistrons and approximately 35% (CITE) to 50% (IRES II) lower expression levels on the third cistron. For enhanced translation-initiation of the third cistron, CITE was specially mutated (CITE * ) to avoid erroneous translation-initiation at upstream ATG start codons (Jackson et al., 1990; Kaufman et al., 1991; Davies and Kaufman, 1992; Rees et al., 1996; Fussenegger et al., 1998c) . Initially, the use of pTRIDENT3 derivatives (IRES-CITE) was preferred over double IREScontaining counterparts because pTRIDENT3-based vectors show a slightly higher translation efficiency of the third cistron, and they contain no duplicated sequence elements (IRESI-IRESII; pTRIDENT1) which bear the risk of recombination-mediated deletion of the second cistron. However, genetic rearrangements or deletions in double IRES-containing pTRIDENT1 derivatives were never observed during cloning steps in recA − E.coli nor in mammalian cells (Fussenegger et al., 1998a and 1998c) . pTRIDENT vector backbones encode a bacterial ampicillin resistance and origin of replication (ori) Figure 2 . Examples for multicistronic expression in mammalian cells: pTRIDENT1 and pQuattro-tTA. Both vectors allow tetracycline-regulated expression of all transgenes. Since pQuattro-tTA encodes all genetic elements for regulated gene expression in its multicistronic expression unit, pQuattro-tTA allows autoregulation and can be used to achieve one-step regulated expression the genes of interest in any cell line where the internal ribosomal entry sites of polioviral origin function (IRES, CITE). for high copy number amplification of these plasmids (Figure 2 ). High copy number amplifications in bacterial hosts is a prerequisite for large-scale transient transfection protocols which are becoming increasingly popular for industrial R&D applications (Fussenegger et al., 1997a) . The tricistronic expression unit contains three multiple cloning sites (MCS) with up to 18 unique restriction sites, many for 8 bptargeting, rare-cutting enzymes to allow sequential, complication-free cloning of all three transgenes into pTRIDENT. The general modular set-up of the key genetic elements in the pTRIDENT series, including, promoter, IRES elements, polyadenylation site, and vector backbone with their well selected flanking (or sometimes internal) restriction sites or MCS allows straightforward elimination or exchange of cistrons between existing conventional monocistronic or pTRI-DENT expression vectors. Also, the modular set-up enables rapid adaptation of the pTRIDENT vector concept for special applications and stimulates future developments in expression vector design. Based on the compatibility of pTRIDENT to existing vector families, for example the one presented by Dirks et al. (1993 and , recent developments of the growing pTRIDENT family resulted in tricistronic vector derivatives with various constitutive (P SV 40 , P MP SV ), tetracycline-and ecdysone-regulated promoters (P hCMV * −1 ; P EC ) and in construction of auto-regulated, self-selective, one-step transfection systems described below. Pioneering reports by Suzuki and Ollis (1990) and Al-Rubeai et al. (1992) showing increased specific productivity of growth-inhibited hybridoma cells stimulated research on chemical culture additives to arrest cell growth and initiated efforts to control cell growth by controlled overexpression or inhibition of selected genes. Three successful one-gene metabolic engineering strategies have been developed to reversibly control mammalian cell growth: (i) estrogen-regulated overexpression of the interferon-responsive factor (IRF-1), a transcription factor which is upregulated by interferons as response to viral cell invasion, in BHK cells (Koester et al., 1995) ; (ii) dexamethasoneinducible suppression of the key transcription factor c-jun by antisense technology in Friend murine erythroleukemia cells (F-MEL) (Kim et al., 1998) ; and (iii) tetracycline-regulated overexpression of negative key regulators of the cell-cycle including the tumor suppressor p53 and the CDIs p21 and p27 in CHO cells (Fussenegger et al., 1997a; Fussenegger et al., 1998a and 1998b) . Overexpression of IRF-1 resulted in cell-cycle-independent growth arrest, but heterologous gene expression was not enhanced unless the exogenous genes were placed under control of IRF1-responsive promoters. Furthermore, IRF-1overexpressing BHK cells rapidly die, probably by an apoptosis-independent pathway (Koester et al., 1995; Müller et al., 1998) . On the contrary, c-jun suppression leads to sustained G0-phase arrest of F-MEL cells for over two weeks and protects these cells against apoptosis (Kim et al., 1998) . Unfortunately, this promising antisense technology remains to be assessed in an industrially relevant cell line and in connection with cloned protein production. However, G0-arrested cells have previously shown to produce exogenous protein at a lower rate (Kim et al., 1998) . In another strategy, transient tetracycline-responsive overexpression of p53, p21 or p27 in a dicistronic configuration (SEAP-p53; SEAP-p21; SEAP-p27) led to G1-specific cell-cycle arrest, and in each case was accompanied by an up to 4-fold increase in SEAP production compared to proliferation-competent control cells (Fussenegger et al., 1997a) . These results compare favourably with those from G1-arrested, temperature-sensitive CHO cells generated by random mutagenesis, which also showed a 3-4-fold higher heterologous protein production upon growth arrest but retained low cell viability at elevated permissive temperatures (Jenkins and Hovey, 1993) . However, in a stable genetic configuration in CHO cells, only SEAP-p27 overexpression lead to a significant increase in productivity, with specific SEAP productivity increasing by 15-fold compared to control cells (Fussenegger et al., 1998b) . Intracellular p21 levels were probably insufficiently high to cause significant growth inhibition, and p53-based cell-cycle arrest led to rapid decrease in cell viability accompanied by cell morphologies indicative of apoptosis, even when achieved by overexpression of the apoptosis-deficient mutant p53175P (Rowan et al., 1996) , a phenomenon which could not be observed with p27-induced G1-arrest (Fussenegger et al., 1998b) . The failure to produce stable growth-controllable CHO cells by p21-mediated overexpression exemplifies current limitations of one-gene metabolic engineering strategies. Although global regulatory proteins certainly exist, such key metabolic effectors are rare, difficult to find and their overexpression may imbalance fine-tuned interconnected cellular circuits, as seems the case with overexpression of p53. Using tricistronic expression technology we extended the SEAP-p21-encoding dicistronic configu-ration by an additional cistron harbouring the differentiation factor, CCAAT/enhancer binding protein α (C/EBPα) (pSS5; Figure 3 ). C/EBPα has been shown to stabilise p21 at the protein level and also to induce endogenous p21 alleles (Timchenko et al., 1996) . Using this tricistronic set-up for metabolic engineering, the induction of conditional growth arrest of CHO cells was successful, and the sustained cell-cycle arrest achieved was accompanied by an up to 15-fold higher specific SEAP productivity compared to proliferationcompetent control cells, similar to that achieved by p27-based one-gene metabolic engineering (Fussenegger et al., 1998a and 1998c) . In a further preventive measure against possible apoptosis, which was strongly suggested by morphologies of p53 overexpressing cells, we linked SEAP-p27 expression with the expression of the survival gene bcl-x L in a tricistronic configuration (pDD6; Figure 3 ). bcl-x L belongs to the family of bcl-2 anti-apoptosis genes which have been successfully used to suppress apoptosis in production cell lines (Cotter and Al-Rubeai, 1995; Mastrangelo and Betenbaugh, 1998) . Although overexpression of SEAP-p27-bcl-x L induced sustained growth arrest in CHO cells like its dicistronic counterpart, the specific SEAP productivity of arrested cells was increased by an additional factor of three, which corresponds to 30-times higher specific SEAP productivity than respective proliferation-competent control cell lines (Fussenegger et al., 1998a) . This unexpected effect of bcl-x L expression cannot be explained based on the current knowledge of cell-cycle and apoptosis regulatory pathways, and further investigations are needed to reveal the mechanism of this new, apparently apoptosis-unrelated effect of bcl-x L . Thus, using controlled proliferation technology as an example, multigene metabolic engineering has proven to be useful for achieving difficult-to-attain cell culture states, and the combinatorial expression of an intuitively unrelated gene revealed previously unknown functions and potential molecular links of complex cellular pathways. There is much current interest in the development of regulatable expression systems in basic functional genomic research, since externally regulated transcription enables the effects of a particular gene product Figure 3 . Tricistronic expression vectors enabling multigene metabolic engineering. Both vectors express the model product gene, the secreted alkaline phosphatase (SEAP) and one of the cell-cycle inhibitors p21 (pSS5) and p27 (pDD6). While the expression of p27 is sufficient to cause cell-cycle arrest and result in enhanced specific SEAP productivity which is additionally increased by coexpression of the anti-apoptosis gene bcl-x L , cell-cycle arrest using p21 is only effective when the differentiation factor c/ebpα is coexpressed and stabilizes p21. to be assessed in an identical genetic background. Regulated gene expression is also gaining increasing importance for biotechnological applications since it allows conditional metabolic engineering and achievement of specific cell culture states in a timely manner (Fussenegger et al., 1997a; Fussenegger et al., 1998a) . For example, regulated metabolic engineering in a multicistronic configuration allows differentiation of a cell culture process into two stages: a nonproductive growth phase in which the cells are rapidly expanded to the desired cell density, and a subsequent non-proliferating production phase where the cells can devote all of their metabolic capabilities to the production of protein instead of biomass. Several in vivo regulated eukaryotic promoters have been described (Schweinfest et al., 1988; Israel and Kaufman, 1989; Ko et al., 1989; Hu and Davidson; Mattioni et al., 1994) and used for regulated gene expression. However, as most of these regulated promoters are derived from regulatory circuits which mediate metabolic responses, the corresponding regulating external stimuli may lead to undesired pleiotropic effects. More successful transcriptional regulation circuits rely on basic regulatory machineries of heterologous origin which are genetically adapted for use in mammalian cells. Besides the lac switch (Fieck et al., 1992) and the ecdysoneresponsive system , the tetracyclineregulatable expression system (tet system; Gossen and Bujard, 1992) is by far the most popular. The tet system consists of two separate genetic entities, the tet-responsive transactivator (tTA) and the tTA responsive promoter, P hCMV * −1 each of which represents a chimeric genetic configuration composed, respectively, of a protein fusion between the bacterial tet repressor and the VP16 domain of the herpes simplex virus (tTA) and a genetic fusion which places a tet operator adjacent to a minimal cmV promoter (P hCMV * −1 ). While the bacterial parts are responsible for promoter recognition and integrate responsiveness to the external stimulus tetracycline, the viral parts initiate transcription (Gossen and Bujard, 1992) . Despite its success story, the tet system has two major limitations. First, prior to introduction of the regulated transgene, each host cell line must be engineered to express tTA in a fashion which affords efficient tetracycline-mediated control of P hCMV * −1initiated transcription. Screening for this phenotype is tedious and time-consuming. Cotransfection of the tTA-expression plasmid and the vector encoding the gene of interest or the transfection of a single vector with a combination of both genes, are not recommended since, in either case, the close proximity of the two genes in the host chromosome may cause the enhancer of tTA-driving promoter expression to interfere with P hCMV * −1 thus leading to hardly regulatable configurations (Gossen and Bujard, 1992) . Second, a transcriptional 'squelching' effect by the VP16 transactivator domain may be lethal for the host cell, even at moderate expression levels (Gill and Ptashne, 1988) . Consequently, since the activity of P hCMV * −1 is proportional to the intracellular tTA levels, moderate tTA expression may lead to apparently low expression levels of regulated transgenes (Furth et al., 1994) . Several improvements have been made to alleviate these complications including (i) fusion of the tTA to the ligand-binding domain of the estrogen receptor to control the transfer of tTA into the nucleus (Iida et al., 1996) , (ii) fusion of tTA to a nuclear localisation signal enabling tight regulation and high-level induction (Yoshida and Hamada, 1997) , (iii) construction of a regulatory cascade by controlling tTA expression by another higher-order control system, for example, by the lac switch (Aubrecht et al., 1996) , (iv) construction of P hCMV * −1 derivatives harbouring minimal promoters of various sources which show altered regulatory features, promoter strength and tTA responsiveness (Hoffmann et al., 1997) , and (v) placing tTA expression itself under control of P hCMV * −1 to prevent accumulation of toxic tTA levels prior to induction (Shocket et al., 1995) . However, all of these improvements still require two rounds of transfection for their implementation. We recently reported one-step, auto-regulated and auto-selective multicistronic mammalian expression systems which included the tTA in a multicistronic, pTRIDENT-based or quattrocistronic configuration (pQuattro-tTA; Fussenegger et al., 1997b ; Figure 2 ). Since the tTA gene is encoded on the multicistronic expression unit itself, little or no tTA is expressed under repressive conditions. This genetic configuration alleviates intracellular accumulation of toxic tTA levels. However, when the auto-regulated system is induced, the few tTA molecules originating from the leakiness of P hCMV * −1 activate this tet-responsive promoter ( Figure 1 ). Since tTA is itself encoded on the artificial operon, every round of transcription generates also an additional tTA message resulting in a positive feedback regulation system with high tTA levels and consequently high level expression of all cocistronically expressed transgenes. Since all genetic elements essential for regulated gene expression are united in a single vector, these autoregulated pTRIDENT derivatives mediate one-step regulated gene expression in various cell lines including CHO, BHK and HeLa cells (Fussenegger et al., 1997b) . Previously, HeLa cells have been reported to be very sensitive to squelching, and prolonged screening procedures are usually necessary to select HeLa clones with moderate tTA expression to avoid this problem (Gossen and Bujard, 1992) . To our knowledge, no convincing tet-regulation gene expression has previously been established in BHK cells apart from a recent report by Sekigushi and Hunter (1998) which shows very high background under repressed conditions and only 10-fold induction. On the contrary, our positive feedback regulation system showed both tight regulation as well as high levels of tet-responsive gene expression in all these cell lines with no signs of squelching. The lack of squelching is rather surprising, considering that the positive feedback circuit is expected to produce high intracellular levels of tTA. However, recent experiments with a monocistronic positive feedback configuration in transgenic animals also showed no detrimental effects (Shocket et al., 1995) . Positive feedback configurations with tTA in the last cistron consist of both essential and interdependent elements for regulated expression, with P hCMV * −1 and tTA at the perimeters of the multicistronic expression unit. This set-up harbours an intrinsic, auto-selective program which guarantees full length transcripts and maintains the functional integrity of all genetic elements encoded on this autoregulatory operon. Recently, a similar autoregulated dicistronic expression system was reported (Shocket et al., 1995; Hofmann et al., 1996; Zhang et al., 1997) , but only pTRIDENT-based or pQuattro-tTA systems allow one-step, autoregulated and auto-selective multigene metabolic engineering in industrially-relevant cell lines (Fussenegger et al., 1997b) . Since one-gene metabolic engineering will necessarily reach its limits when coping with today's increasingly complex challenges, the recent development of artificial eukaryotic operons enables effective multigene metabolic engineering of mammalian cells. This greatly expands possibilities to reprogram interconnected cellular networks in desired ways to improve key characteristics of mammalian cells. Besides use in next-generation multigene metabolic engineering, multicistronic expression units are expected to have great impact on very specific applications including (i) straightforward combinatorial evaluation of gene functions and metabolic networks, (ii) one-step transfection, selection and maintenance of difficult-toexpress (multi-subunit) proteins, (iii) selection of high producer cell lines, and (iv) genetic immunisation and gene therapy in combination using sense, antisense or ribozyme technology.

Production of high titre disabled infectious single cycle (DISC) HSV from a microcarrier culture

Disabled Infectious Single Cycle (DISC) HSV-2 has been cultured in the complimentary cell line CR2 to provide high titre bulk material suitable for the purification of the virus as a live viral vaccine. CR2 cells are cultured on the microcarrier Cytodex-1 at 5 g l-1 in small scale (1 l) and larger scale (15 l) reactors. The cells are infected at an MOI of 0.01 pfu cell-1 and the culture harvested 60–72 h later. The infected cells are removed from the microcarriers by the addition of a hypotonic saline and the virus released by low-pressure disruption techniques. Virus titres achieved are compared to the standard roller bottle process. The resulting material is the starting point for the purification of the DISC-HSV virus.

The incidence of genital herpes is high and increasing world-wide (Corey, 1993 ). An important aim in combating the disease is achieving an effective vaccine which can act against both the primary and recurrent disease caused by herpes simplex virus. The company strategy has therefore been to adapt live viruses by genetic manipulation to introduce an acceptable margin of safety through the development of DISC (Disabled Infectious Single Cycle) virus vaccines. These are viruses which lack an essential gene, and are therefore unable to undergo multi-cycle replication in a vaccinated host. They can, however be prepared in a way that allows them to go through a single cycle of replication in cells of the vaccinee, leading to an effective immune response. It has been shown that herpes simplex viruses lacking the essential glycoprotein H (gH) gene can be used as effective vaccines (Farrell et al., 1994; McLean et al., 1994) . The virus is therefore genetically inactivated so that it is unable to spread within the host. * Part of this work was presented as a poster at the ESACT meeting in Tours, France. September 1997. Currently the DISC-HSV is propagated in a complementing cell line derived from a Vero (African Green monkey kidney) cell line approved by the World Health Organisation (WHO) for use in vaccine manufacture (WHO, 1989) . The Vero cells were modified to contain the HSV-2 gH gene under the control of the HSV-1 glycoprotein D (gD) gene promoter (Boursnell et al., 1997) . The complementing cell line was designated CR2. Because the gD promoter requires additional HSV proteins for its induction, gH is regulated so that it is only produced in the cell following an infection with virus. Virus produced from CR2 cells can infect normal cells, but can only perform one cycle of replication. The progeny virus from this replication cycle lack the gH protein and are therefore non-infectious ( Figure 1 ). Previously it has been shown that a DISC gHdeleted HSV-1 can protect against HSV-1 challenge in the mouse ear model (Farrell et al., 1994) . The DISC virus, by virtue of its capacity for a single round of replication in normal cells, is more potent than a non-replicating, inactivated virus preparation. A gHdeleted HSV-2 virus has also been tested as a vaccine in a guinea pig model. Animals vaccinated with DISC HSV-2 showed complete protection against primary HSV-2 induced disease, even when challenged six months after vaccination (Boursnell et al., 1997) . The animals were also almost completely protected against recurrent disease. For the vaccine to become a viable option as a commercial product, it must be manufactured on a large scale with appropriate processing to meet the demands of the regulatory requirements of safety and efficacy. Currently the DISC-HSV is propagated in the CR2 cell line. The CR2 cell line is an adherent cell and therefore routine culture was performed in roller bottle cultures. Whilst this method of cell growth and virus production is suitable for development work it is not a desirable system for the manufacture of the virus on a larger scale. The use of microcarriers as a support for anchorage-dependent cells has been reported previously (Hu et al., 1985; Van Wezel, 1973) . Anchorage-dependant cell lines have been cultured to produce a variety of viruses on a large scale (Talbot et al., 1989; Lesko et al., 1993; Baijot et al., 1987; Meignier, 1978) . Some cell lines such as Vero and MRC5 have been propagated on microcarriers to produce human viral vaccines (Fabry et al., 1989; Griffiths et al., 1980; Montagnon et al., 1981) . Griffiths et al. (1982) have demonstrated HSV-2 production from cells cultured on low (2 g l −1 ) concentrations of Cytodex microcarriers. The yields obtained were ten-fold lower when compared to those achieved from a roller bottle system. It was hoped that our current microcarrier system would produce equivalent productivity when compared to the standard roller bottle process. The aim of this work was to evaluate the production of DISC-HSV in a microcarrier based culture system in comparison to a conventional roller bottle culture process. The work set out to determine the conditions for the optimum growth of cells and ultimately, production of high titre virus. Production of the virus was initially determined in small-scale (1 l) cultures. Scale-up of the production system was then demonstrated at the 15 l scale. This paper details the way in which we have approached this issue of manufacturing DISC-HSV only as far as the initial upstream bulk harvest product and does not discuss the subsequent virus purification. Viruses: The DISC-HSV virus, grown in gHexpressing CR2 cells was constructed at Cantab Pharmaceuticals, Cambridge, U.K. Cells: Two cell lines are used in our study. One has been designated CR1 and the other CR2. The CR1 cell line is a line used for the assay of the virus only. It is a Vero derived cell line that has been modified to express the glycoprotein H (gH) gene derived from HSV-1 this gene is under the control of the HSV-1 glycoprotein D (gD) promoter. CR2 cells are a modified Vero cell from the WHO Vero accredited bank (No. 88020401; European Collection of Animal Cell Cultures ECACC, Porton Down, U.K.) that has been modified to express the glycoprotein H (gH) gene derived from HSV-2. This gene is under the control of the HSV-1 glycoprotein D (gD) promoter. The CR1 cells are solely used to assay the virus in the TCID 50 assay whereas the CR2 cells are the production cells for manufacture of the DISC-HSV. Cells were cultivated in Dulbecco's Modified Eagles Medium (DMEM with high glucose, Life Technologies, U.K.) supplemented with 5% Foetal Bovine Serum (PABCO, New Zealand origin). The bottles used had a total surface area of 850 cm 2 and were obtained from Corning. Cytodex 1 microcarriers: (Pharmacia Biotech, U.K.): The microcarriers were used at a density of 5 g l −1 and were prepared according to the manufacturers instructions. Before use they were pre-conditioned in DMEM (5%FBS). Hypotonic saline: The solution comprised of Na 2 HPO 4 2.29 g l −1 , NaH 2 PO 4 .2H 2 O 0.599 g l −1 NaCl 0.58 g l −1 . The reagents were sourced from BDH. Roller bottle cultures were seeded with a total of 2 × 10 7 CR2 cells per roller bottle. The cultures contained 100 ml of DMEM (5% FBS) per bottle culture. Cultures had a five day cell growth period at 37 • C prior to infection with DISC-HSV. At infection the cell monolayers were washed with Dulbecco's PBS to decrease the level of any contaminating BSA from the FBS. Serum-free DMEM was added to the cells as a maintenance medium throughout the virus production phase. Cultures were infected with a working seed preparation of DISC-HSV at a multiplicity of infection (MOI) of 0.01 pfu cell −1 . The temperature of the cultures during the virus production phase was decreased and maintained at 34 • C. Approximately 64-68 h post infection cell monolayers have between 90-100% CPE. This observed level of CPE is ideal for harvesting the DISC-HSV. The harvest method involved pouring off the medium and adding 10 ml of hypotonic saline. The cultures were then incubated at 34 • C for 5-10 min. The cells and virus were collected by scraping the surface with a plastic scraper. Production of DISC-HSV in microcarrier culture 1 l and 15 l scale The microcarriers were prepared as according to instructions from Pharmacia. They were prepared in a siliconised Schott bottle of suitable size. All microcarriers were conditioned by washing with DMEM (5% FBS) prior to addition to the vessel. The microcarriers for the 1 l cultures were washed twice with 250 ml of complete medium whilst the microcarriers for the 15 l cultures were washed twice with 1500 ml of medium. All vessels were obtained from FT Applikon Ltd. For the 1 l cultures a 2 l total volume jacketed glass vessel was used. A BioBench 20 l total volume, jacketed stainless steel vessel was used to culture the 15 l cultures. Both vessels had spin filters (76 µm mesh size) and used reverse marine impellers. The 2 l vessels had an H:D ratio of 1.5 and the 20 l vessel had an H:D ratio of 2.2. The growth of our CR2 cells was examined on several commercially available microcarriers. The range used included the following: -Cytodex 1 (Pharmacia), Cytodex 2 (Pharmacia), Cytodex 3 (Pharmacia), Cultisphere (Cellon Sarl), Cytopore 1 (Pharmacia), Cytopore 2 (Pharmacia), Cytoline 2 (Pharmacia) and FACT (Solohill). The microcarrier Cytodex 1 was chosen as the production carrier because the cell density obtained using this carrier was the greatest from the least amount of microcarriers used. Cytodex 3 microcarriers gave a similar cell number as expected but this microcarrier has a pig skin collagen layer on the bead. Cytodex 1 was selected in preference to the other microcarriers because of good cell growth and to remove the issue of animal products in the manufacturing process. We currently use the Cytodex 1 at a level of 5 g l −1 . The growth medium used was Dulbecco's Modified Eagle Medium (DMEM) supplemented with 5% Foetal Approximately 75% of the final volume of cultivation medium and 5 g l −1 of microcarriers were pre-incubated in the culture vessel overnight. The culture parameter set points were controlled by using the Applikon 1030 controlling system for both vessels. The culture pH was maintained by controlling at a set point of 7.2±0.2 by sparging air/CO 2 (90:10%) into the spin filter. Dissolved oxygen concentrations were kept above a minimum set point of 30% saturation by sparging when necessary into a spin filter with air for the 1 l cultures and with pure oxygen in the 15 l cultures. The culture temperature was maintained at 37 • C throughout the cell growth period by use of a recirculating thermocirculator attached to the jacket of the 2 l vessels and in the case of the 20 l vessels the Applikon controlling system used a similar integral recirculating thermocirculator. The cell inoculum for cultivation on microcarriers was prepared from late exponential CR2 cultures in 850 cm 2 roller bottles. Cultures were inoculated at a density of approximately ten CR2 cells per microcarrier and 5 g of Cytodex 1 l −1 (Figure 2 ). Cell growth was monitored by estimating nuclei released from a sample of microcarrier culture taken from the reactor after incubation in 0.1 M citric acid containing 0.1% (w/v) crystal violet (Sanford et al., 1951) . Glucose and lactate concentrations were determined using an off-line YSI 2700 glucose/lactate analyser (YSI, U.K. Ltd). Glucose and lactate concentrations were controlled by partial media changes on days 2 and 3 for the 1 l culture and days 2, 3, 4 and 5 for the 15 l culture. The 15 l culture was maintained for an additional 48 h period compared to the small scale 1 l cultures in Figure 5 . The bionebuliser is a low pressure shear disruption method which uses nitrogen as a carrier gas. The disrupter is depicted here schematically indicating how the process disrupts cells. Using a carrier gas pressure of 50 psi cells are drawn up and nebulised. This cell mixture is directed to a ceramic ball bearing where they impact. The cells are disrupted to release the virus and the whole disrupted cell/virus solution is collected. order to provide a maximal cell density at the point of infection. Confluent microcarriers (Figure 3) , approximately 100 h post inoculation in the 1 l cultures and approximately 140-160 h in the 15 l cultures were infected with DISC-HSV at an MOI of 0.01 pfu cell −1 . The agitation was stopped and the microcarriers were allowed to settle out. The medium was then removed to waste. The 1 l cultures were washed by addition of 1 l of Dulbecco's PBS with agitation. The PBS wash was removed and this step repeated a further two times. The 15 l cultures were washed by addition of 10 l of Dulbecco's PBS with agitation. The PBS wash was removed to waste and this step was repeated a further two times. After washing, the original culture volume was replenished with serum free DMEM. The temperature during the infection period was decreased and maintained at 34 • C. Infected cells (Figure 4) were harvested from the microcarriers 60-72 h post infection when the cells on the majority of the microcarriers showed 100% CPE. Harvesting was accomplished by removal of the media from the culture, addition of hypotonic saline at a volume equal to 10 ml hypotonic saline per roller bottle equivalent of surface area of microcarriers used in the culture. The detached cells were separated from Cytodex 1 by filtration through a sterilisable 70µm mesh in the bottom of the vessel. The resulting cell/virus supernatant was subjected to low pressure disruption (Degouys et al., 1997) , (Figure 5) , to release the DISC-HSV virus from the cells and cellular debris. Virus titres were calculated using an in-house TCID 50 assay with subsequent conversion to pfu ml −1 . The cell and virus harvests from either pooled roller bottle cultures or microcarrier cultures were passed through a bionebuliser (Glas-Col) at a pressure of 50 psi using sterile filtered nitrogen as the carrier gas. The carrier gas causes a vacuum over the hole leading to the cell suspension. This vacuum draws the cell suspension up the tubing and allows the cell suspension to become mixed with the carrier gas. The mixture then flows through the orifice to impact on the target a ceramic ball to disrupt the cells in the suspension. The flow rate of the gas determines the density of cells in the mixture and the target ceramic ball can be adjusted to maintain a good bionebulisation of the cells. The process is depicted in Figure 5 . This low-pressure shear system has been used previously to disrupt cells (Degouys et al., 1997) . The DISC virus infectious titre (TCID 50 ml −1 ) was estimated in CR1 cells on 96-well plates infected with serial tenfold dilution's of virus material. The CPE was read 3-4 days post infection using a TCID 50 ml −1 method (Reed and Muench, 1938) . The TCID 50 ml −1 assay is used with a subsequent conversion to pfu ml −1 as described by Dougherty (1964) . The TCID 50 dose can be measured by a number of means and the method of Reed and Muench (1938) was employed in this case. This TCID 50 is the usual end point for many virus assays. The conversion to pfu ml −1 takes into account the distribution of virus in the suspension according to Poisson's Law. According to Poisson distribution the TCID 50 would contain 0.69 infectious particles per unit volume. Therefore this correction factor can only be applied when chance distribution of virus in suspension occurs (Dougherty, 1964) . Roller bottle process CR2 cells were cultured in roller bottles for 5 days until a cell density of 1 × 10 8 cells was typically achieved. At this density 100% cell confluency was observed. Roller bottle cultures infected with DISC-HSV at an MOI of 0.01, exhibited 100% CPE after approximately 65 h. Harvesting the roller bottles with hypotonic saline it was possible to achieve 10 9 pfu per bottle. Figure 6 . Typical growth of CR2 cells and DISC-HSV production at 1 l scale. The total DISC-HSV virus titre produced in a 1 l vessel is approximately 1.5 × 10 10 pfu. Figure 7 . Typical growth of CR2 cells and DISC-HSV production at 15 l scale. The amount of DISC-HSV virus titre able to be produced from a 15 l culture is 7 × 10 11 pfu. This demonstrates a considerable scale up potential for a microcarrier production system. Based on available surface area for cell attachment, a 1 l microcarrier culture is equivalent to approximately 25 × 850 cm 2 roller bottles in terms of potential virus production capability (Table 1) . A typical growth curve for CR2 cells on Cytodex 1 microcarriers is shown in Figure 6 . Cells multiplied exponentially during the first 4 days before reaching stationary phase. A cell density of 1.5-2.0 × 10 9 total cells was typically obtained after approximately 100 h growth. At this density a confluent monolayer of CR2 cells covered the microcarriers (Figure 3 ). At 100% confluency the cells appeared elongated, resembling the surface of a golf ball. Cells ready for harvest exhibited complete CPE (Figure 4 ) 60-72 h after infection. The cultures were harvested before they became detached from the microcarriers. The morphological appearance of the cells at both critical stages was identical to the appearance observed in roller bottle cultures. Infectious DISC-HSV released from the CR2 cells was quantified using a TCID 50 assay and compared to the amount obtained from the roller bottle culture harvests. The results are shown in Table 1 . Typically 1.5-2.0 × 10 10 total pfu from a 1 l culture was regularly achieved, with approximately a 20 hour period during which the culture could be harvested without significant decrease of the titre. The total surface area available in the culture is equivalent to 25 roller bottles. This enables a comparison in terms of productivity to be made. We routinely achieve 2.0-2.5 × 10 10 total pfu from 25 roller bottle cultures. Therefore, the two production systems yield similar total amounts of DISC-HSV. Based on available surface area for cell attachment, a 15 l culture is equivalent to approximately 388 × 850 cm 2 roller bottles in terms of potential virus production (Table 1) . A typical growth curve for CR2 cells on Cytodex 1 microcarriers grown at the 15 l scale is shown in Figure 7 . Cells multiplied exponentially during the first 6 days before reaching stationary phase. Typically a cell density of 4.0-5.0 × 10 10 total cells was obtained after approximately 150 h. After this growth period a confluent monolayer of CR2 cells (Figure 3 ) covered the microcarriers. When the cells exhibited complete CPE (Figure 4 ) 60-72 h after infection, the cultures were harvested. Infectious DISC-HSV released from the CR2 cells was quantified using a TCID 50 assay and was again compared to the titre achieved from the equivalent roller bottle culture harvests. The results are shown in Table 1 . Typically 4-7.0 × 10 11 total pfu from a 15 l culture was achieved, with a period of approximately 9 h during which this high titre was sustained. From an equivalent number of 388 roller bottles we would expect to achieve approximately 3.88 × 10 11 total pfu. Our results, therefore, compare very favourably with the expectations from roller bottle cultures and the results achieved from the 1 l cultures. Cytodex 1 microcarriers can be used as a Vero cell culture support for the production of DISC-HSV virus. It could be envisaged that the microcarrier washing and conditioning procedure at a large manufacturing scale could be extremely time consuming. It is expected that the microcarriers will be sterilised in-situ inside the production vessels. Preliminary work at 35 l scale employs the use of an internal sieve to aid the washing and conditioning process. At present at the 35 l scale this step takes approximately 2 h to complete. This washing and conditioning step may be an extensive time constraint at a manufacturing scale. Tackling the engineering problem in terms of pipe work and increasing flow rates could reduce this time. This is an issue that is under review and will require suitable development time to reduce it satisfactorily. Yields achieved from the microcarrier cultures were comparable to those obtained in the standard roller bottle culture systems (Table 1) . It has been demonstrated that a 15 l microcarrier culture can produce an equivalent amount of virus as that achieved from 700 × 850 cm 2 roller bottles despite having the surface area of only 400 roller bottles. This maintenance of the cell productivity has not always been observed when virus production systems have been scaled up. Griffiths et al. (1982) had a tenfold decrease in cell productivity of HSV per cell when switching from roller bottle cultures to a microcarrier system using Vero cells. This may have been an artefact of poor virus release from the cells. In comparison, when culturing the HSV-2 on MRC-5 cells, a significant but smaller decrease in production levels was observed when changing from roller bottle cultures to microcarrier cultures. This observation reinforces the issues made in some early virus production work reported by Giard et al., 1977. They proposed that there are specific virus/cell line requirements coupled with a specific optimisation of the conditions not only for the growth of the cells but also for the virus production phase. The production of hepatitis A in microcarrier culture was reported to be 8-fold lower than when cultured in conventional flask cultures (Widell et al., 1984) . The maintenance of productivity seen in this study is very encouraging in moving the culture system forward to a larger scale which may be suitable for the production of Phase III material and for the supply of commercial requirements. Due to the productivity's shown here, production at a 15 l scale would be large enough to supply material for Phase I and Phase II studies. Additional benefits of this system arise in the need for decreased culture medium. The very large culture surface area to volume ratio offered by the microcarrier system provides high cell yields in a minimal volume. When compared with the roller bottle system approximately half the volume of medium is required to produce an equivalent cell density (Van Wezel, 1972) . This leads to savings on the cost for culture medium and in particular for costly serum additions. The number of direct manipulations is also reduced when switching to a microcarrier based production system. This decreases the labour intensity of the process, minimising costs of materials and the overall process time. The virus is produced from a single batch culture with one set of aseptic manipulations throughout the process rather than multiple manipulations that are required when using a roller bottle system. If the system can be operated at a larger scale (e.g. >50 l) and still maintain DISC-HSV productivity then it would become a viable production method. Passaging Vero cells to a large production vessel (500 l) has been suggested (Baijot et al., 1987) , but it likely that the trypsinisation method will be all important in such a process. The high productivity cultures using perfusion techniques, such as the production system used for the poliovirus production at 20 g l −1 of microcarriers (Fabry et al., 1989) , may also be employed for high titre virus production. However, further development work would be needed to adopt a similar production system for our DISC-HSV vaccine. In our laboratory we are considering serial passaging and perfusion techniques to improve our production capabilities further. From the results reported in this present study we are confident that we have a productive scaleable microcarrier system for our DISC-HSV vaccine that will allow us to produce adequate material for both Phase III studies and commercial product.

Diagnosis of influenza viruses with special reference to novel H1N1 2009 influenza virus

On 15 April and 17 April 2009, novel swineorigin influenza A (H1N1) virus was identifi ed in specimens obtained from two epidemiologically unlinked patients in the United States. The ongoing outbreak of novel H1N1 2009 influenza (swine influenza) has caused more than 3,99,232 laboratory confi rmed cases of pandemic influenza H1N1 and over 4735 deaths globally. This novel 2009 influenza virus designated as H1N1 A/swine/California/04/2009 virus is not zoonotic swine flu and is transmitted from person to person and has higher transmissibility then that of seasonal influenza viruses. In India the novel H1N1 virus infection has been reported from all over the country. A total of 68,919 samples from clinically suspected persons have been tested for influenza A H1N1 across the country and 13,330 (18.9%) of them have been found positive with 427 deaths. At the All India Institute of Medical Sciences, New Delhi India, we tested 1096 clinical samples for the presence of novel H1N1 influenza virus and seasonal influenza viruses. Of these 1096 samples, 194 samples (17.7%) were positive for novel H1N1 influenza virus and 197 samples (18%) were positive for seasonal influenza viruses. During outbreaks of emerging infectious diseases accurate and rapid diagnosis is critical for minimizing further spread through timely implementation of appropriate vaccines and antiviral treatment. Since the symptoms of novel H1N1 influenza infection are not specifi c, laboratory confi rmation of suspected cases is of prime importance.

The current outbreak of swine infl uenza that originated in Mexico in March 2009 has spread to more than 80 countries causing more than 3,99,232 laboratory confi rmed cases of pandemic infl uenza H1N1 globally and over 4735 deaths reported to World Health Organization (WHO)as of 11 October 2009 [1] . The WHO declared pandemic alert stage 6 on 11 June 2009, indicating an ongoing infl uenza pandemic [2] . The 2009 swine fl u virus designated H1N1 A/swine/California/04/2009 is not zoonotic swine fl u and is not transmitted from pigs to humans, but rather from person to person and has higher transmissibility than seasonal infl uenza viruses [3] . In humans, H1N1 swine fl u presents as an infl uenza-like illness (ILI) with symptoms similar to seasonal infl uenza, i.e. fever, cough, sore throat, runny nose, muscle pains, severe headache, however, a considerable proportion of patients reported vomiting or diarrhea which is unusual in seasonal infl uenza [4, 5] . Since these symptoms are not specifi c to swine fl u, early in the pandemic physicians were advised to consider swine infl uenza in the differential diagnosis of patients with acute febrile respiratory illness who had returned from Mexico or been in contact with persons with confi rmed swine fl u [6] . This new strain of H1N1 swine infl uenza has a unique combination of genes from both North American and Eurasian swine lineages that has not been identifi ed previously in either swine or human populations [7] . The virus appears to be a result of reassortment of two swine infl uenza viruses, one from North America and one from Europe with the North American virus itself the product of previous re-assortments, carrying an avian PB2 gene for at least 10 years and a human PB1 gene since 1993. The virus also has genome segments of avian origin. Hence scientists call this novel strain as a "quadruple reassortant" virus. The hemagglutinin (HA) gene is similar to that of swine fl u viruses present in pigs in United States since 1999, where as neuraminidase (NA) and matrix (M) genes resemble viruses present in European pigs. Viruses with this genetic makeup have not previously been found in humans or pigs. In India the novel H1N1 virus infection has been reported from all over the country. The most affected states are Maharshtra, Delhi, Tamil Nadu, Karnataka, Andhra Pradesh, Haryana, Kerala, Uttar Pradesh and Gujarat. As on 21 October 2009, a total of 68,919 samples from clinically suspected persons have been tested for infl uenza A H1N1 in government laboratories and a few private laboratories across the country and 13,330 (18.9%) of them have been found positive with 427 deaths [8] . Genomic analysis of the 2009 infl uenza A (H1N1) virus in humans indicates that it is closely related to reassortant swine infl uenza A viruses isolated in North America, Europe and Asia [ Fig. 1 ] [9] [10] [11] . The segments coding for the polymerase complex, hemagglutinin, nuclear protein, and non-structural proteins show high similarity with the swine H1N2 infl uenza A viruses isolated in North America in the late 1990s. The segments coding for the neuraminidase and the matrix proteins of the new human H1N1 virus are, however, distantly related to swine viruses isolated in Europe in the early 1990s. In particular, the closest isolated relatives of the neuraminidase segment have 94.4% similarity at the nucleotide level with European swine infl uenza A virus strains from 1992 [11] . The incubation period for novel H1N1 2009 infection appears to range from 2 to 7 days; however, additional information is needed. On the basis of data regarding viral shedding from studies of seasonal infl uenza, most patients with novel H1N1 2009 infection might shed virus from 1 day before the onset of symptoms through 5 to 7 days after the onset of symptoms or until symptoms resolve; in young children and in immunocompromised or severely ill patients, the infectious period might be longer [12] . Patients who are at highest risk for severe complications of novel H1N1 2009 infection are likely to include but may not be limited to groups at highest risk for severe seasonal infl uenza: children under the age of 5 years, adults 65 years of age or older, children and adults of any age with underlying chronic medical conditions and pregnant women [13] . Two classes of antiviral medication are available for the treatment of seasonal human infl uenza: neuraminidase inhibitors (oseltamivir and zanamivir) and adamantanes (rimantadine and amantadine). During the 2008-2009 infl uenza season, almost all circulating human infl uenza A (H1N1) viruses in the United States were resistant to oseltamivir [14] . However, genetic and phenotypic analyses indicate that novel H1N1 2009 is susceptible to oseltamivir and zanamivir but resistant to the adamantanes [15] . The Centers for Disease Control and Prevention (CDC), Atlanta, GA, USA has recommended that given the severity of illness observed among some patients with novel H1N1 2009 infection, therapy with neuraminidase inhibitors should be prioritized for hospitalized patients with suspected or confi rmed novel H1N1 2009 infection and for patients who are at high risk for complications from seasonal infl uenza. A number of different laboratory diagnostic tests can be used for detecting the presence of novel H1N1 infl uenza virus in respiratory specimens, including direct antigen detection tests, virus isolation in cell culture, or detection of infl uenza-specifi c RNA by real-time reverse transcriptasepolymerase chain reaction (Real-time RT-PCR). During outbreaks of emerging infectious diseases accurate and rapid diagnosis is critical for minimizing further spread through timely implementation of appropriate vaccines, antiviral treatment and prophylaxis where available and other public health-based nonpharmaceutical measures. Appropriate treatment of patients with respiratory illness depends on accurate and timely diagnosis and early diagnosis of infl uenza can reduce the inappropriate use of antibiotics and provide the option of using antiviral therapy. Preferred respiratory samples for infl uenza testing include nasopharyngeal or nasal swab, throat swab and nasal wash or aspirate, depending on which type of test is used. Samples should be collected within the fi rst 4 days of illness. Routine serological testing for infl uenza requires paired acute and convalescent sera, does not provide results to help with clinical decision-making. Serological testing results for human infl uenza on a single serum specimen is not interpretable and is not recommended. All respiratory specimens should be kept at 4°C for no longer than 72 hours before testing and ideally should be tested within 24 hours of collection. If storage longer than 72 hours is necessary, clinical specimens should be stored at -70°C [16]. Antigen detection tests; also known as rapid infl uenza diagnostic tests (RIDTs) detect infl uenza viral antigens in clinical specimens. These rapid infl uenza diagnostic tests can provide results within 30 min or less. Hence the results are available in a clinically relevant time period. Diagnostic tests for detection of novel H1N1 infl uenza virus antigen may be of two main types: direct fl uorescent antibody (DFA) tests and rapid enzyme/optical immunoassays or assay for NA enzymatic activity. Direct fl uorescent antibody (DFA) staining of clinical specimens using specifi c monoclonal antibodies against novel H1N1 infl uenza virus antigen can be a reliable and relatively rapid technique for the pandemic novel H1N1 infl uenza virus detection. Studies of DFA detection of infl uenza viruses have shown highly variable results with sensitivities ranging from 40% to 100%. Recent analytical studies indicate that commercially available RIDTs can detect novel infl uenza A (H1N1) virus [17] . In a study Chan et al. showed that the rapid antigen tests they evaluated in their study have comparable sensitivity for detection of novel H1N1 infl uenza and seasonal infl uenza viruses [18] . Data on analytical sensitivity for detection of different viruses does not directly refl ect clinical sensitivity on patient specimens. However, only limited data have been published on the performance of RIDTs compared with RT-PCR for detecting the presence of novel infl uenza A (H1N1) virus in clinical specimens [19] . Compared to RT-PCR, the sensitivity of RIDTs for detecting novel infl uenza A (H1N1) virus infections ranged from 10% to 70%. The sensitivity of RIDTs to detect novel infl uenza A (H1N1) virus is equal to or lower than the sensitivity to detect seasonal infl uenza viruses [17] . Although these rapid tests do not differentiate between novel H1N1 2009 infl uenza virus and seasonal infl uenza A or even between subtypes H1 and H3 but they may provide useful information that might impact patient care. Understanding the limitations of rapid tests is very important to appropriately interpret results for clinical management of the disease [20] . Novel H1N1 infl uenza virus detection can also be achieved by inoculating the clinical specimen on MDCK cells for virus isolation with subsequent characterization by hemagglutination inhibition (HI) and neuraminidase inhibition tests using monospecifi c antiserum. Although the cell culture method is sensitive, it requires viable virus, needs expertise and at least 6-8 days to grow the virus to a level where cells are examined for cytopathic effect (CPE). Virus isolation is not only labor-intensive it is timeconsuming also and requires a week for declaring a sample positive or negative hence not appropriate for an epidemic situation. Although the extreme genetic variability of infl uenza viruses is a challenge for design of molecular-based diagnostic tests. Reverse transcriptase-polymerase chain reaction (RT-PCR) is a widely used molecular tool that has been applied to both infl uenza virus detection and subtype characterization of virus isolates. Most infl uenza A PCR assays in use target conserved regions of the M gene and therefore should detect infl uenza A from all established subtypes, including the newly emergent novel H1N1 infl uenza. However, such methods need to be complemented with a rapid subtyping test to distinguish seasonal infl uenza A from novel H1N1 2009 infl uenza virus. Multiplex PCR testing for the detection of respiratory viruses has seen major advances over the past decade resulting in the development of several commercially available tests. These tests can amplify one or more genes from a number of respiratory viruses and detect amplifi ed products using microgene arrays. One such assay the xTAGTM RVP test was developed in 2005 immediately following SARS and H5N1 infl uenza and was designed to detect and type the three infl uenza A subtypes circulating at that time viz. H1, H3 and H5 [21, 22] . A limitation of PCR methods is that false-negative results may occur due to sequence variation in primer and probe targets and is particularly relevant for the detection of emerging viruses. However the use of multiple targets can reduce such limitations, and may serve as a means of confi rming positive results. Mahony The effi ciency and performance of nucleic acid amplifi cationbased assays depends on the amount and quality of sample template. Nucleic acid amplifi cation assays, including reverse transcriptase RT-PCR (rRT-PCR), and real-time RT-PCR are the most sensitive and specifi c infl uenza virus diagnostic assays. Real-time RT-PCR remains the method of choice for clinical diagnosis of novel H1N1 2009 virus in respiratory specimens and for differentiating it from seasonal infl uenza viruses [25] . Laboratory tests, such as real-time RT-PCR should be prioritized for hospitalized patients to diagnose 2009 H1N1 infl uenza and immunocompromised persons with suspected infl uenza where RIDT or DFA testing is negative or to determine infl uenza A virus subtype in patients who have died from suspected or confi rmed infl uenza A virus infection. The CDC has developed and recommended a realtime RT-PCR asaay for detection and characterization of novel H1N1 infl uenza. The assay includes a panel of oligonucleotide primers and dual-labeled hydrolysis (Taqman®) probes. The assay can be used to detect and characterize the novel H1N1 virus (swine infl uenza) in respiratory specimens and viral cultures. The assay has InfA primer and probe set designed for universal detection of type A infl uenza viruses and swInfA primer and probe set to specifi cally detect all swine infl uenza A viruses. The assay also includes a set of specifi c primer and probes for HA gene to specifi cally detect swine H1 infl uenza virus in specimens positive with SwInfA primers and probes. The assay can be applied on a wide range of specimens such as broncheoalveolar lavage, tracheal aspirates, sputum, nasopharyngeal or oropharyngeal aspirates or washes, and nasopharyngeal or oropharyngeal swabs taken from suspect swine infl uenza A infected patients. Recently Carr et al. developed an M gene-based real-time reverse transcriptase polymerase chain reaction (rtRT-PCR) assay for the detection of novel H1N1 2009 infl uenza virus that does not cross-react with human seasonal infl uenza A viruses (subtypes H1N1 and H3N2) [26] . An internal control should be included for each and every clinical sample tested for novel H1N1 2009 virus. The inclusion of internal control ensures proper specimen collection, processing and RNA extraction. The CDC realtime RT-PCR protocol uses Human RNaseP gene (RNP) as internal control for human nucleic acids. No template controls and positive template controls should also be included in each run. A human specimen control provides a secondary negative control that further validates the nucleic extraction procedure and reagent integrity. The no template control reactions should not exhibit fl uorescence growth curves that cross the threshold line. Although the real-time RT-PCR is highly sensitive and specifi c assay for novel H1N1 virus detection, the limitations include need of trained personnel for assay set up and result interpretation, false negative results which may occur if inadequate numbers of organisms are present in the specimen due to improper collection, transport, handling or excess of DNA/RNA template in the reaction and initial cost of machine. (Table 1) (Fig. 2) . There is no perfect test for the diagnosis of infl uenza. Virus culture, the present 'gold-standard test' is not 100% sensitive and does not provide results in a time-frame that allows optimal use of potentially effective antiviral treatment. Although rapid diagnostic tests provide results in less than 30 minutes, they are signifi cantly less sensitive and do not differentiate between different subtypes of infl uenza A virus. Rapid testing is only offered after the fi rst culture-confi rmed cases of infl uenza are reported from the community. Molecular assays; reverse transcriptase polymerase chain reaction (RT-PCR) and real-time RT-PCR targeting conserved regions of infl uenza virus genome have advantages over other methods and provide sensitive, highly specifi c and rapid diagnosis. The realtime RT-PCR should be the method of choice and both in-house developed and CDC-developed real-time PCR assays can be used for the specifi c detection of novel H1N1 2009 infl uenza virus.

Influenza A: From highly pathogenic H5N1 to pandemic 2009 H1N1. Epidemiology and clinical features

The last decade has seen the emergence of two new influenza A subtypes and they have become a cause of concern for the global community. These are the highly pathogenic H5N1 influenza A virus (H5N1) and the Pandemic 2009 influenza H1N1 virus. Since 2003 the H5N1 virus has caused widespread disease and death in poultry, mainly in south East Asia and Africa. In humans the number of cases infected with this virus is few but the mortality has been about 60%. Most patients have presented with severe pneumonia and acute respiratory distress syndrome. The second influenza virus, the pandemic H1N1 2009, emerged in Mexico in March this year. This virus acquired the ability for sustained human to human spread and within a few months spread throughout the world and infected over 4 lakh individuals. The symptoms of infection with this virus are similar to seasonal influenza but it currently affecting younger individuals more often. Fortunately the mortality has been low. Both these new influenza viruses are currently circulating and have different clinical and epidemiological characteristics.

Infl uenza is an acute respiratory disease which affects the upper and/or lower respiratory tracts. Infl uenza outbreaks occur every year and globally account for about 3-5 million severe cases with 250,000 to 500,000 deaths annually. It is caused by infl uenza viruses which are of three types: A, B and C and they can all affect humans. Infl uenza A viruses have 2 main surface glycoproteins -hemagglutinin (HA) and neuraminidase (NA) -which have 16 and 9 subtypes respectively. The infl uenza A virus subtypes are classifi ed on the basis of the different HA and NA glycoprotein subtype combinations. All the subtypes can affect birds which are the natural hosts. Only a few subtypes are capable of infecting humans. The past few years have seen the emergence of two new infl uenza A subtypes which have become a cause of concern for the global community. These are the highly pathogenic H5N1 infl uenza A virus (H5N1) and the pandemic 2009 infl uenza H1N1 virus. The H5N1 virus emerged initially in 1997 and then in 2003. Since 2003 this virus has caused widespread disease and death in poultry. In humans the number of cases infected with this virus is few but the mortality is very high. The second infl uenza virus, the pandemic H1N1 or the swine origin infl uenza A (H1N1) virus emerged this year and within a few months quickly spread throughout the world. The virus acquired the ability for effi cient human-tohuman spread which resulted in a large number of infected individuals. Fortunately the mortality has been low. Infl uenza A viruses are dynamic and can evolve by two processes, antigenic drift and antigenic shift. This capability of antigenic variation is responsible for the severe outbreaks of disease. Antigenic drift occurs by point mutations in the two genes coding for HA and NA and this causes minor changes in surface proteins. This leads to a new strain which is not recognized by antibodies to previous infl uenza strains. Antigenic shift is a major change through genetic reassortment which produces a novel infl uenza A subtype in humans. This occurs through mixing of human and animal infl uenza A virus genes or by animal to human transmission. A pandemic occurs when a new type of infl uenza A virus is introduced in humans that can cause a serious illness and is capable of sustained human to human transmission. H5N1 is a highly pathogenic avian infl uenza virus which has caused a widespread epizootic illness among birds. Although the virus is widely present in birds in various parts of the world, human disease from H5N1 has been uncommon as bird to human transmission is ineffi cient and human to human transmission is rare. The virus could cause an infl uenza pandemic if it attains the ability for effi cient and sustained human to human transmission. This is of great concern as there is little natural immunity to H5N1 virus in humans and the disease in humans is severe with a high mortality. The emergence of H5N1 infl uenza in humans for the fi rst time occurred in 1997 in Hong Kong. Eighteen people admitted with a respiratory illness were found to be H5N1 positive. Six of them (33%) died. This outbreak was epidemiologically linked to H5N1 infection in a live bird market in Hong Kong. At this time 1.4 million poultry was culled in Hong Kong, the market was disinfected and import of poultry from mainland China was halted. This outbreak was successfully controlled with these measures [1]. In 2003, 2 members of a 5 member family from Hong Kong were infected with H5N1 virus after travelling to China. The source of the infection could not be confi rmed. One of these two persons died while the other, a young boy recovered. Subsequently an outbreak of H5N1 infection in poultry and humans occurred in south East Asia. In 2003-2004, fatal and severe respiratory infection, mostly pneumonia and respiratory failure, were seen in China, Thailand and Vietnam. All of these were associated with an outbreak of H5N1 in poultry. A total of 50 cases with 36 deaths were reported with the mode of disease transmission being from sick or dead poultry to humans. Only one case of human-to-human transmission from a sick child to mother was reported in Thailand [1]. In 2005, 98 human cases with 43 mortalities were reported from Cambodia, China, Indonesia, Thailand and Vietnam. Again all these outbreaks were associated with an ongoing H5N1 infection in poultry. The next year the cases and geographical area increased and 115 cases with 79 deaths were reported from China, Cambodia, Azerbaijan, Djibouti, Egypt, Indonesia, Iraq, Thailand and Turkey. Most of these were again from contact with dead or live infected poultry. The only evidence of human to human transmission was in a family cluster of 8 in Indonesia in which 7 individuals died [1]. In 2007, 86 human cases with 59 mortalities were reported from nine countries: Cambodia, China, Egypt, Indonesia, Laos, Myanmar, Nigeria, Pakistan and Vietnam [1]. A total of 442 confi rmed cases of avian fl u (H5N1) have been reported till 24 September 2009 from the above-mentioned 15 countries with 262 deaths [2].The mortality rate is around 60% which is very high. Infection with H5N1 in humans has so far remained confi ned to individuals with close contact with infected birds or surfaces or objects heavily contaminated with their droppings. This virus has not acquired the ability for sustained human to human spread and therefore has not been able to infect a large number of individuals and cause a pandemic. 90% of patients infected with H5N1 are less than 40 years old with a median age of 18 years [3] . The mortality is highest among the 10-19 years age group and lower in people more than 50 years old. The reason for lower infection rate and mortality in older people has not been ascertained. The route of transmission is usually from birds to humans. There is usually a history of exposure to dead or sick poultry/wild birds, their secretions or excretions during the week prior to illness. Activities involving close contact such as defeathering, preparing poultry for cooking, holding or playing with sick poultry, handling fi ghting cocks and eating raw or undercooked poultry products have been implicated [4] [5] [6] [7] . Human-to-human transmission has only been documented in 1 case of transmission from sick child to mother and possibly 1 case of a cluster of 8 patients in a family in Indonesia in 2006 [7, 8] . The incubation period of H5N1 infection in humans has usually ranged from 2-5 days, though clinical features have appeared even up to 8-17 days after exposure . Almost all patients present with high grade fever (>38°C). Cough and dyspnea are seen commonly (about 90%). Sore throat is seen in around half the patients with rhinitis and upper respiratory symptoms being less common. Headache, myalgia and weakness have also been reported. Gastrointestinal symptoms such as diarrhea, vomiting and abdominal pain may also be present. Watery diarrhea may precede respiratory symptoms by a week [9] [10] [11] . One case with fever, diarrhea, seizure and coma has been reported from Vietnam leading to a clinical diagnosis of encephalitis. H5N1 was detected from CSF, serum, throat and fecal samples [12] . Conjunctivitis is also described occasionally [13, 14] . Most patients develop lower respiratory features early during the illness. Dyspnea usually develops after a median of 5 days of initial symptoms [15] . Respiratory distress, tachypnea and inspiratory crackles are commonly seen. Most patients have clinical and radiological features of pneumonia which is seen to rapidly progress to respiratory failure with manifestations of adult respiratory distress syndrome (ARDS). In a report from Thailand, the median time for progression to ARDS was 6 days (range 4 to 13 days) [15] . Complications such as multiorgan failure, cardiac dilatation, arrhythmias, ventilator-associated pneumonia, pulmonary hemorrhage, pneumothorax, sepsis syndrome, Reye's syndrome and pancytopenia have been described. A very high mortality of more than 60% has been reported though the risk factors for severe disease are not clear. In 1997, old age, delayed hospitalization and lower respiratory infection were found to be associated with severe disease with children less than 6 years having milder disease. However, recent H5N1 infections have caused high mortality rates in infants and young children. Knowledge related to epidemiology and clinical features remains incomplete and coordination between affected countries is needed to fully understand the profi le of this new viral infection. Novel strains of infl uenza virus arise due to antigenic shifts and drifts. These strains have very different surface glycoproteins which did not exist in human strains before. A pandemic occurs when such a virus emerges in humans with effi cient human to human transmission. As there is very little or no immunity against it, the virus quickly infects a large number of individuals in all age groups. A pandemic has been expected for long and it was feared that H5N1 avian infl uenza virus which caused severe disease in clusters of humans was the most likely candidate virus to cause a pandemic. For any pandemic to start three conditions need to be met: A infl uenza virus subtype not seen in humans for at least a generation should emerge; 2. The new virus should have the ability to infect and replicate effi ciently in humans; and 3. The new virus should have developed the ability for easy and sustained human-to-human spread. H5N1 virus was unable to cause a pandemic due to its ineffective human to human transmission. In March 2009, a novel strain of H1N1 infl uenza virus evolved from a reassortant between triple re-assortant swine infl uenza viruses in North American pigs and infl uenza A virus circulating in Eurasian pigs [16, 17] . This combination had not been seen previously. By the end of April 2009, WHO declared the emergence of human cases of H1N1 swine infl uenza virus. On 11 June 2009, the WHO raised the pandemic alert from phase 5 to phase 6 and announced that the world was at the beginning of an infl uenza pandemic [18] . The Centers for Disease Control, Atlanta, USA compiled and analyzed data collected from the beginning of the outbreak till 24 July 2009 and this gives a fair idea of the demographic characteristics of H1N1 infl uenza. During this period, 43,771 cases were reported in the USA with the majority being in people under 24 years of age [21] . Few cases were reported in people older than 60 years. The age distribution of the number of cases per 100,000 according to this analysis is given in table 1 [22] . In India, all ages have been affected, with the age group 15-34 years being the worst affected. The incidence has been low in children below 3 and people above 60 years of age [23] . The hospitalization rate was highest in children less than 4 years followed by the 5-24 year age group. Hospitalization was less common in 25-65 years age group and increased thereafter. This is unlike seasonal fl u infection where the elderly and young children are found to be at higher risk for fl u related complications [24] . Complications were also higher in people with underlying diseases such as asthma, cardiac diseases, renal diseases and in pregnancy. Obesity has also found to predispose to severe disease [25] . Human-to-human transmission is through inhalation of respiratory droplets which are expelled when an infected person sneezes or coughs or by contact with surfaces that have been contaminated by respiratory secretions and then by touching the mouth or nose. The rate of secondary transmission has been found to be 22-33% [26] . In a study in Kenya the secondary household transmission was found to be 26% [27] . The transmissibility in schools has been found to be around 20% [28] . The incubation period is between 1 and 7 days and an infected person can transmit the infection from a day prior to onset of symptoms to a day after symptoms have subsided. The clinical picture of H1N1 infl uenza encompasses a wide spectrum ranging from the mild self-limiting upper respiratory illness to lower respiratory infection including ARDS, cardiac involvement, neurological involvement, multiorgan failure, septicemia and death. H1N1 most commonly causes a mild respiratory illness with fever, cough, sore throat, dyspnea, rhinorrhea, myalgias, chills, headache and fatigue. Diarrhea and vomiting are more commonly seen than with seasonal fl u. In a study of patients from April to June 2009 in the USA, gastrointestinal symptoms were seen in 39% patients [29] . Fever and cough are the most common features seen in 93% and 83% respectively [30]. It is usually a self limiting, mild illness but may occasionally present as a serious illness needing admission. The hospitalization rate in the USA between 15 April and 24 June 2009 was 2.21% with the highest rate of hospitalization being seen in people below 24 years. Most patients who required admission did so within 1-7 days, with a median of 4 days, from the onset of illness [31] . 73% of these patients had some underlying disease which predisposed them to a severe disease and complications. The most common underlying disorders were asthma (28%), neurological disorders (21%), diabetes (15%), immunosuppression (15%) and cardiovascular disorders (13%), with the others being chronic renal disorder, chronic obstructive pulmonary disorder and pregnancy. Obesity was found in 29% of the hospitalized patients with or without other risk factors. 25% of the hospitalized patients were critically ill [29] . Patients who required ICU admission, vasopressors, inhaled oxygen at FiO2 more than 60% or required mechanical ventilation were said to be critically ill [31] . Younger people are at higher risk of being critically ill with infants and people between 26 and 64 years being the worst affected, the mean age being between 30 and 40 years of age. Around one-third of these patients were young or middle aged adults and were not pregnant and had no underlying disorder [32] . Most patients who were critically ill presented with fever, cough, dyspnea, myalgias, malaise, weakness, tachypnoea, tachycardia, hypotension, cyanosis and low oxygen saturation. The most common presentation was with adult respiratory distress syndrome or an acute lung injury picture. 31% of these patients had superadded bacterial pneumonia. About 60-80% of these patients required mechanical ventilation [31] . Other complications that have been reported include myocarditis, pericarditis, encephalitis, seizures, myositis, multiorgan failure and toxic shock syndrome [33] . The mortality in this group has been found to be around 18% with older age, requirement of mechanical ventilation and co-morbid conditions being major risk factors. A large number of deaths are seen in young to middle aged adults due to the higher incidence in this group [31, 32] . Till date, the pandemic 2009 H1N1 infl uenza virus has spread rapidly and caused a massive burden of disease around the world. It affects the younger population more frequently and can cause severe illness in a small proportion of people. The mortality rate is low and similar to seasonal infl uenza. How the virus will behave in subsequent months in terms of virulence and morbidity is unclear. A better understanding of the disease will help us prepare for the future. The last decade or so has seen the emergence of two new infl uenza A viruses which have a different epidemiolgical and behavior pattern. The H5N1 virus has remained mainly an avian virus with human spread being limited and occuring only in persons coming in close contact with infected or dead poultry. The virus is however lethal causing a very high mortality. The pandemic 2009 infl uenza A (H1N1) virus on the other hand has emerged as a new human virus with the ability for effi cient human to human spread. Within a span of less than 6 months this virus has spread to more than 191 countries and emerged as the fi rst pandemic virus of this century. Fortunately, this virus causes a low mortality and hopefully this pandemic will be as mild as the last two infl uenza pandemic that occurred in 1957 and 1968. The fear that this virus may become more virulent and lead to a more severe pandemic as occurred in 1918 still exists. We therefore need to be vigilant and prepared if this happens in subsequent waves.

Lessons learned from the 1918–1919 influenza pandemic

The 1918 influenza pandemic was one of the most virulent strains of influenza in history. Phylogenic evidence of the novel H1N1 strain of influenza discovered in Mexico last spring (2009) links it to the 1918 influenza strain. With information gained from analyzing viral genetics, public health records and advances in medical science we can confront the 2009 H1N1 influenza on a global scale. The paper analyses the causes and characteristics of a pandemic, and major issues in controlling the spread of the disease. Wide public vaccination and open communication between government and health sciences professionals will be an essential and vital component in managing the 2009 H1N1 pandemic and any future pandemics.

The infl uenza pandemic of 1918 is generally ranked second only to the 14th century "Black Death" plaque in terms of relativity and absolute mortality. [1] In fact, the death toll of the First World War failed to infl ict the human casualty rate that the 1918-1919 Infl uenza Pandemic did. Little progress has been made toward understanding the condition responsible for the extreme virulence of the "1918 type," and/or the conditions necessary to prevent the reappearance of this infl uenza. Unlike the typical "fl u" that strikes the very young, chronically ill and the elderly, this fl u would attack and kill healthy young adults. Taubenberger [2] reported that deaths resulting from the infl uenza and pneumonia for the 15-34-year-old cohort were 20 times higher in 1918 than any previous time, and 99% of excess deaths among people under 65 years of age [3] This strain of infl uenza killed so many people that it reduced the life expectancy of the United States with ten years during its course. The present H1N1 strain discovered in the spring of 2009, almost 90 years from the onset of the 1918 pandemic, is resurrecting this specter from the past. This novel H1N1 infl uenza strain emerged from a quiet village in Mexico. The Mexican government responded on 24 April 2009, closing schools, canceling public gatherings in Mexico City and surrounding states until 6 May 2009 [4] . This drastic step may have slowed down the regional spread of H1N1 in Mexico, but it had already left the country through international trade and travel. As in past quarantines, this quarantine would fail as well [5] . The early focus on slowing this new strain by the health community and governmental agencies would be depended on non-pharmaceutical interventions focusing on measures to: Limit international spread of the virus (e.g. travel screening and restrictions); 2. Reduce the spread within national and local populations; 3. Reduce an individual person's risk for infection; and 4. Communicate risk to the public. Infl uenza pandemics occur as a result of two different mechanisms: novel emergence of an avian descendent virus (as in the 1918 virus) or a modifi cation of a human adapted virus by genetic mixing as in the reassortment of novel hemagglutinin (HA) with or without non-committent neuraminidase, (NA) as in the 1968 H3N2 strain [19] . The pandemics of 1948, 1957 and 1968 were caused by variations of the infl uenza virus resulting from this shuffl ing of the eight gene pairs within the virus. Infl uenza pandemics occur in three waves. The typical fi rst wave or initial outbreak, as in the spring of 1918, was relatively mild, starting from the Midwest and spreading along the rail lines with soldiers from Ft. Funston, Kansas, modern day Ft. Riley [2, 3, 14, 16] . The novel H1N1 discovered in Mexico also had such a humble origin. Both patient zeros reported little problems and made complete recoveries. Unlike the Mexican outbreak, the 1918 spring outbreak was not even noted in the index in the 1918 volumes of the Journals of the American Medical Association. Infl uenza was not a reportable disease until 1925 in the United States: the only evidence of the early occurrence was the registration of deaths reported as uncomplicated cases of pneumonia by physicians to various public health departments [3, 8, 9] . Those who had suffered from the earlier spring infl uenza generally suffered less discomfort in the second wave which would occur in the early summer of 1918 in Europe, affecting the outcome of the war. The third and most deadly wave of the infl uenza would occur later that year in the late fall. Despite the obvious differences between the strains in each wave, it is suggested that the more virulent form of infl uenza was genetically derived from the spring infl uenza [3, 10, 12] . The antigenic composition of the 1918 virus is related to the H1N1 viral group. Phylogenetic studies indicate that the virus responsible for the 1918 infl uenza and viruses that provided gene segments for the Asian/1957 and Hong Kong/1968 pandemics are still circulating in wild birds, with few or no mutations [10,11,16,] . The extreme virulence of fall 1918 infl uenza strain has been blamed on severe pathology with acute pulmonary edema, as well as hemorrhage with acute bronchiolitis, aveolitis and bronchopneumonia [20] it is believed that the severe infl ammation of the lungs initiated high levels of cytokines resulting in a depletion of neutrophils and alveolar macrophages, causing death. The stronger the host's immune system, the stronger response to the infl uenza infection, and greater release of cytokines to counter the virus [19, 20] . This is what resulted in the much higher than average mortality among the 15-40-year-old cohort. Patterson and Pyle [12] , Crosby [8] and many other researchers believe that a strain of pneumonia bacteria accompanied the virus [3, 12] . Noyes [3, 14, 17] noted that the nation's people were stricken and died from the illness at differing rates, just as the cities were hit at differing rates. There was no correlation between populations, or even geographical demographics. Sex and age both played a major factor in determining the susceptibility to the disease of the individual. Females were stricken in rates greater than males, and young adults were sickened in greater numbers that other age cohorts [3, 13 and 14] . The current international strain of novel H1N1 virus discovered in Mexico is derived from two unrelated swine viruses, one associated with a fourth generation of the 1918 human infl uenza virus with which acts to recombine the viruses and its progeny [4, 18] . Essentially the virus continues by shuffl ing its eight genes in the avian reservoir to eventually be passed to swine and other mammals before the encounter with humans. Seldom are there transfer of infl uenza between humans and avian. Pigs act as the "transformers or converters" for the various infl uenza viruses and let loose the world new "strains" of the infl uenza virus. The intervening passage continues to be through the domesticated pig, as they have the mechanism to convert the arrangement of the sialic acids to become receptive to human cells [21, 22] . The arrangement of the genes, determine the protein sheath structure and the interaction between the antibody defenses of the host. A mere change of one amino acid can change the impact of the infection, from a mild discomfort to a killer virus [14] . One of the major concerns of the 2009 H1N1 is this very issue; will it change between the waves of outbreaks? In the 1918 pandemic control was sought after, as each wave of the infl uenza outbreak proved deadlier than the previous. Communication between the medical community, the government, media and the public was non-existent. Because of the war, information concerning the infl uenza was blocked, except for neutral Spain, which won the honor of being the name sake for this deadly virus: "The Spanish Infl uenza". In Britain, the battle between preventive and curative medicine raged. Previously these medical approaches were championed by the Medical Offi cers of Health in their efforts to prevent illness, and a therapeutic practice by private physicians [1] . The curative physicians, primarily general practitioners, were overwhelmed by the cases of infl uenza and relied on traditional methods to cope with the illness ranging from aspirin, quinine, opium, ammonia, alcohol, camphor, eucalyptus and iodine to musk, wet packs, blood serum, creosote, turpentine, cinnamon and turtle soup [1, 2] . At best, these prescriptions offered symptomatic relief, and some actually harmed the patients. In the United States, the 1918 pandemic was met with local quarantines, and large public gatherings were discouraged, and like Great Britain folk remedies were widely used [2, 8, 22] . There was no coordinated information from medical authorities to the communities on how to cope with this outbreak. The best advice that was offered then was common sense: bed rest and careful nursing to avoid complications, which is still issued today. The issue of vaccination was not available until 1931, when a viral growth in embryonated hens' eggs was discovered, and in the 1940s, the US military developed the fi rst approved inactivated vaccines for infl uenza, which were used in the Second World War with limited success. Vaccination is the best defense against the infl uenza if the right strain is predicted, and if there are no mutations after the administration of the vaccine. In the 1957 and 1968, infl uenza vaccination programs were credited with the reduction of the severity of both pandemics. The 1976 swine fl u scare provided the general public a forum to exam the role of vaccination, when the "predicted" swine fl u pandemic failed to materialize after hundreds of thousands of Americans were vaccinated [15] . As shared by many authorities, the directions of pandemics are diffi cult to predict. Successive pandemics and pandemiclike episodes have been decreasing in severity over time. This is due to advances in medicine, public health and understanding the genetics of the disease, but this may also refl ect the evolutionary course of the virus, to that favors transmissibility with minimal pathogenicity. A virus that kills its host, or causes its host to remain at home, will not be transmitted [10] . The swine fl u scare did not result in a pandemic: was it because it lost its pathogenicity or because of the massive immunization clinics which would result in a "herd immunization"? Harvey Fineberg from the Institute of Medicine, reviewed the1976 swine fl u scare, and shared these fi ve principles facts for preparing the public for a pandemic: (i) Build a base for decision-making; (ii) Think thoroughly each decision point; (iii) Consider and maintain good ties to the media; (iv) Maintain long-term credibility; and (v) Think twice about medical knowledge [15] . These fi ve lessons have been applied to several situations since 1976, i.e. the foot and mouth disease outbreak in UK, and the recent SARS outbreak and the bird fl u threat of only a few years ago. The application of these fi ve procedures has prevented panic, allowed professionals to do their jobs, and provided the media sound and correct information. Current health and government offi cials have been implementing these fi ve principles, and the 2009 H1N1 pandemic preparedness has been successful thus far. Steps have been implemented at the various levels of government in the United States, and most institutions where people gather have procedures in place to cope with a pandemic (see Table 1 ). Infl uenza pandemics will not be prevented, but with lessons learned from previous pandemics, the effect of infl uenza pandemics can be reduced. Governments and health professionals must continue to maintain surveillance against all diseases, share information and work together in developing vaccines. As learned in 1918-1919, infl uenza respects no politics, nor borders.

Differential Effects of IL-12 on Tregs and Non-Treg T Cells: Roles of IFN-γ, IL-2 and IL-2R

Complex interactions between effector T cells and Foxp3(+) regulatory T cells (Treg) contribute to clinical outcomes in cancer, and autoimmune and infectious diseases. Previous work showed that IL-12 reversed Treg-mediated suppression of CD4(+)Foxp3(−) T cell (Tconv) proliferation. We and others have also shown that Tregs express T-bet and IFN-γ at sites of Th1 inflammation and that IL-12 induces IFN-γ production by Tregs in vitro. To investigate whether loss of immunosuppression occurs when IFN-γ is expressed by Tregs we treated mouse lymphocyte cultures with IL-12. IFN-γ expression did not decrease the ability of Tregs to suppress Tconv proliferation. Rather, IL-12 treatment decreased Treg frequency and Foxp3 levels in Tregs. We further showed that IL-12 increased IL-2R expression on Tconv and CD8 T cells, diminished its expression on Tregs and decreased IL-2 production by Tconv and CD8 T cells. Together, these IL-12 mediated changes favored the outgrowth of non-Tregs. Additionally, we showed that treatment with a second cytokine, IL-27, decreased IL-2 expression without augmenting Tconv and CD8 T cell proliferation. Notably, IL-27 only slightly modified levels of IL-2R on non-Treg T cells. Together, these results show that IL-12 has multiple effects that modify the balance between Tregs and non-Tregs and support an important role for relative levels of IL-2R but not for IFN-γ expression in IL-12-mediated reversal of Treg immunosuppression.

Natural regulatory T cells (Tregs), identified by the expression of Foxp3, play an important role in down-regulating immune responses and an imbalance between numbers of Tregs and conventional CD4 (Tconv) and CD8 T cells contributes to outcomes in cancer and autoimmune and infectious diseases [1] [2] [3] [4] . Interactions between Tregs and non-Tregs result in modifications of the function of both cell types, with the goal of minimizing detrimental side effects of pro-inflammatory immune responses while maximizing the efficacy of processes such as pathogen clearance. Further, the function of both cell types and the balance between the pro-and anti-inflammatory immune responses may be affected by cytokines and other molecules produced by cells such as dendritic cells, in the inflammatory milieu. While Tregs are immunosuppressive at sites of inflammation, Tregs may express transcription factors and cytokines that parallel those of Tconv cells. Thus, Tregs express T-bet and IRF4 at sites of Th1 and Th2 inflammation, respectively [5] [6] [7] . Additionally, we and others have shown that Foxp3 + Tregs express IFN-c during infections [5, 8] . Unfractionated populations of Tregs harvested from inflammatory sites, which include cells expressing IFN-c, remain immunosuppressive. Suppression can be demonstrated in T cell proliferation assays after stimulation with either anti-CD3 mAb or virus-specific peptides [8] . In contrast, other studies suggest that a pro-inflammatory milieu may result in diminished Treg suppressive function [9] [10] [11] . In a model of experimental autoimmune encephalomyelitis (EAE), CNS-derived epitope MOG -specific Tregs showed reduced ability to inhibit proliferation of Tconv of the same specificity isolated from the inflamed CNS [9] . This reduction in inhibitory function was IL-6 and TNF-dependent. IL-12 has also been shown to enhance activation and proliferation of Tconvs even if Tregs are present, possibly reflecting an IL-12-mediated reduction in Treg suppressive function [12] . In other studies, IL-12 was shown to induce IFNc production by Tregs in vitro and in vivo [5, 13] ; IFN-c expression by Tregs may indicate that these cells are transiting to Th1 effector cells [13] . Thus, in apparent conflict with the results using suppression assays, these data suggest that IFN-c expression is a marker for reduced immunosuppressive ability. However whether IFN-c expression actually affects Treg suppressive function has not been assessed experimentally. Collectively these results suggest that IL-12 has independent effects on Tconv and CD8 T cells as opposed to Tregs. Consequently, the effect of IL-12 (or other cytokines) on the T cell response in a total lymphocyte culture will reflect the summation of effects on the different cell types. In order to delineate the roles of these cytokines, particularly IL-12, on Treg function, we treated lymphocytes harvested from spleens and lymph nodes of naïve mice with a panel of cytokines. Treatment with IL-12 induced IFN-c expression by both Tconv and Tregs and reduced Treg frequency and Foxp3 expression. We also demonstrated that IFN-c + Tregs are as immunosuppressive as IFN-c -Tregs. Our results showed that IL-12 functioned, in part, by diminishing IL-2 production by Tconv and CD8 T cells in the mixed cultures and by down-regulating IL-2 receptor expression on Tregs but up-regulating it on non-Treg T cells. 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. Mice were housed in the Animal Care Facility at the University of Iowa. The protocol was approved by the University of Iowa Animal Care and Use Committee (Protocol Number: 1007161). All efforts were made to minimize animal suffering. Specific pathogen-free C57BL/6 (B6) and B6/Thy1.1 mice were purchased from the National Cancer Institute, Bethesda, MD. Foxp3 gfp mice on a B6 background were kindly provided by Dr. A. Rudensky (Sloan-Kettering Institute) and were bred to a Thy1.1 background. IL-12Rb2 2/2 mice on a B6 background were obtained from Dr. J. Harty (University of Iowa). Lymphocytes prepared from lymph nodes and/or spleens of mice were stimulated with soluble anti-CD3 mAb (0.5 mg/ml) in the presence of various cytokines. Unless indicated, cells were cultured at 1610 6 cells/ml in 24-well tissue culture plates (2 ml/well) and IL-12 was used at 1 ng/ml. To analyze cell proliferation, cells were labeled with CFSE (2 mM, Invitrogen) before culture. To detect intracellular IFN-c or IL-2 production by T cell subpopulations, PMA (50 ng/ml, Sigma), ionomycin (1 mg/ml, Sigma) and Golgi plug (1 ml/ml, BD Biosciences) were added for the last 4 hr of culture. Treatment with TLR Agonists 5x10 6 B16-Flt3L cells [14] (obtained from Dr. J. Harty) were inoculated subcutaneously into 12-wk old B6 mice. Twelve to fourteen days post inoculation, spleens were harvested, digested with collagenase and DCs were isolated using anti-CD11c microbeads (Miltenyi Biotec, Auburn, CA). T cells were enriched from spleens of B6 or IL-12Rb2 2/2 mice using a Pan T Cell Isolation Kit II (Miltenyi Biotec). 1610 5 T cells were cultured with 1610 4 DCs in the presence of anti-CD3 mAb and medium or IL-12 or LPS (1 mg/ ml) or CpG (1 mM) for 72 hr in a 96-well round-bottom plate. In parallel wells, 2610 5 unfractionated lymphocytes were treated under the same conditions. Samples were analyzed in triplicate. To evaluate the function of IFN-c producing Tregs, lymphocytes were prepared from Foxp3 gfp /Thy1.1 mice and stimulated with anti-CD3 mAb in the presence of IL-12 or medium as control. After 66 hr, cells were harvested, washed and enriched for CD4 + T cells using a CD4 T Cell Isolation Kit II (Miltenyi Biotech). GFP + Tregs were then sorted. Approximately .90% and ,3% of Tregs expressed IFN-c in the IL-12 and medium groups, respectively. Responder Tconvs were isolated from naïve Foxp3 gfp (Thy1.2) mice. For suppression assays, Tregs were co-cultured with CFSE-labeled (2.5 mM) responder Tconvs at the indicated ratios (Tregs plus responders = 5610 4 cells/well) in 96-well round bottom plates in the absence of IL-12. Wells also contained 2610 5 T-cell depleted splenocytes (irradiated at 2500 rad) and anti-CD3 mAb. After 66 hr, Thy1.1 2 Thy1.2 + Tconv cells were analyzed for CFSE dilution by flow cytometry. To evaluate the function of Tregs in the presence of IL-12, responder T cells were enriched from naïve B6/Thy1.1 mice using a Pan T cell Isolation Kit II, and Tregs were isolated from B6 mice using a CD4 + CD25 + Regulatory T cell Isolation Kit (Miltenyi Biotec). For suppression assays, Tregs were mixed with CFSElabeled (2.5 mM) responder T cells at the indicated ratios (Tregs plus responders = 5610 4 cells/well). Cells were cultured in the presence of 2610 5 irradiated splenocytes and anti-CD3 mAb with or without IL-12 in 96-well round bottom plates. After 66 hr, Thy1.1 + Tconvs and Thy1.1 + CD8 T cells were analyzed for CFSE dilution by flow cytometry. The Division Index (DI) was obtained using FlowJo software (Tree Star, Inc., Ashland, OR). A normalized DI was calculated as follows: % normalized DI = 100% 6(DI of responders plus Tregs/DI of responders only). A Foxp3 Staining Buffer Set (eBioscience) was used for Foxp3 or T-bet staining or when cells were analyzed for Foxp3 and cytokine expression simultaneously; otherwise, BD Cytofix/Cytoperm and Perm/Wash buffers (BD Biosciences) were used in intracellular cytokine staining assays. Cell sorting was performed with a FACSDiva or FACSAria and cell analysis with a FACSCalibur or LSRII (BD Biosciences). Lymphocytes from lymph nodes of B6 mice were stimulated with anti-CD3 mAb in the presence of IL-12 (1 ng/ml) or IFN-c (100 ng/ml) in 24-well plates as described above. Supernatants were collected at 48 hr after culture. IL-2 ELISAs were performed using reagents and protocols provided by the manufacturer (eBioscience, Mouse IL-2 ELISA Ready-SET-Go kit). Samples were plated in duplicate. Data are presented as means 6 standard errors of the means (SEM). Differences between two groups were determined by Student's two-tailed unpaired t tests or paired t tests (when normalized DI were compared), using GraphPad Prism 5.03 software. Differences with values of P,0.05 were considered significant. *, P,0.05; **, P,0.01; ***, P,0.001. We and others showed that Tregs express IFN-c at sites of inflammation and can be induced to express IFN-c in vitro [5, 8, 13] . In specific, IFN-c is expressed by virus-specific Tregs in the CNS of mice infected with the rJ2.2 strain of mouse hepatitis virus (MHV) [8] ; we used information obtained from this experimental system as the basis for the approach described here. To determine whether cytokines had a role in IFN-c expression by Tregs, we initially focused on cytokines that are up-regulated in the CNS of MHV-infected mice and remain elevated as the infection resolves [15] . Among these cytokines, IL-12, IL-6 and IFN-c have roles in T cell polarization and Treg development [16] [17] [18] [19] [20] , so we first evaluated their participation in Treg-mediated IFN-c expression. Lymphocytes were prepared from the lymph nodes of naïve B6 mice, labeled with CFSE and treated with each cytokine and anti-CD3 mAb for 66 hours. Cells were then stimulated for 4 hours with PMA and ionomycin and evaluated for T cell proliferation and IFN-c expression. As shown in Fig. 1B , IL-12 but not IL-6 or IFN-c induced IFNc production by Tregs and Tconvs although IL-12 induced lower levels of IFN-c in Tregs than in Tconvs. After 42 hours, the majority of Tconv and Tregs expressed IFN-c; by 66 hrs nearly all of the cells were IFN-c + (Fig. 1C) . Further, IL-12 treatment increased proliferation of Tconvs (Fig. 1B, D) and CD8 T cells (data not shown), but, in contrast, diminished frequency and proliferation of Tregs (Fig. 1A, D, E) . IL-12 also decreased levels of Foxp3 in Tregs (Fig. 1F ). The net result was a decreased proportion of Tregs expressing reduced levels of Foxp3 in IL-12treated cell cultures (Fig. 1A, E) . IL-6 reduced both Tconv and Treg proliferation although the effects on Treg proliferation were greater (Fig. 1B, D) . IL-12 treatment increases expression of T-bet, a transcription factor associated with Th1 development [19, 20] ; we and others previously showed that T-bet was up-regulated in Tregs localized to sites of Th1-type inflammation [5, 7, 8] . Consistent with these results, T-bet was also increased in Foxp3 + Tregs as well as in Tconvs after IL-12 treatment (Fig. 2) . To examine the role of IL-12 in IFN-c induction in Tregs when inflammation is induced by a more physiological mechanism, we treated samples with TLR agonists, LPS or CpG. T cells (B6 or IL-12Rb2 2/2 ) and splenic DCs (B6) were isolated and incubated together in the presence of anti-CD3 mAb and LPS or CpG. The fraction of Tregs expressing IFN-c increased after LPS or CpG exposure and was similar to levels obtained in cultures treated with IL-12 (Fig. 3 ). LPS and CpG likely functioned via effects on DCs because no IFN-c was induced when unfractionated lymph nodederived lymphocytes were treated with LPS or CpG in the absence of added DCs (data not shown). CpG or LPS-mediated augmentation in the proportion of cells expressing IFN-c was largely dependent on IL-12, since the fraction of IFN-c + Tregs increased only slightly in cultures containing IL-12Rb2 2/2 T cells (Fig. 3) . Of note, IL-12Rb2 2/2 mice are deficient in IL-12 but not IL-23 signaling [21] . Mixed populations of IFN-c + and IFN-c -Tregs are able to suppress proliferation of both anti-CD3 mAb and MHV peptidestimulated T cells [8] . However, it was not possible from these experiments to determine whether IFN-c expression attenuated suppressive function. To address this question, we treated spleen and lymph node-derived lymphocytes isolated from Foxp3 gfp mice with anti-CD3 mAb in the presence or absence of IL-12. GFP + Tregs were then sorted from the mixed cultures. In the presence of IL-12, .90% of all GFP + cells expressed IFN-c while 90-98% IFN-c -Tregs were detected in cultures treated with anti-CD3 mAb alone. Both sets of Tregs (IFN-c + or IFN-c -) were then mixed with CFSE-labeled Tconvs. Cells were then stimulated with anti-CD3 antibody in the absence of IL-12. As shown in Fig. 4 , IFN-c + and IFN-c -Tregs inhibited the proliferation of CFSE-labeled Tconv to the same extent. In these assays, we compared equal numbers of IFN-c + and IFN-c -Tregs for suppressive function. However, results in Fig. 1 showed that the numbers of Tregs and their relative expression of Foxp3 were diminished in IL-12-treated cultures. Further, a previous report showed that Treg suppressive function was alleviated by IL-12 [12] . To reconcile these disparate results, we assessed Treg suppression in the presence of IL-12 using Tregs and responder T cells enriched from lymph nodes and spleens. Responder T cells were labeled with CFSE and co-cultured with unlabeled Tregs at several cell ratios. We observed diminished suppression of CFSE-labeled CD4 T cells in the presence of IL-12, in agreement with published results (Fig. 5) . Suppression of CD8 T cell proliferation was almost completely abrogated if IL-12 was present in the culture. Treg proliferation was diminished in the presence of IL-12, resulting in an effective change in ratio of the two cell types within a single culture (Fig. 1C, D) . However, it is also possible that IL-12 augmented Treg cell death or induced conversion of Tregs to Tconvs, which would have the same effect. To address the role of apoptosis, we measured AnnexinV expression on Tconvs and Tregs in lymph nodes cultures. IL-12 treatment did not increase the percentage of either AnnexinV + Tregs or Tconvs in the cultures, demonstrating that there was no preferential Treg death (Fig. 6) . Loss of Foxp3 expression by Tregs has been demonstrated in vivo, although whether this represents Treg instability is controversial [22] [23] [24] . IL-12 induces IFN-c expression by Tregs (Fig. 1B) ; if IL-12 also mediated Treg conversion, this would contribute to a decrease in Treg, and an increase in Tconv numbers. To examine this possibility, Foxp3 gfp /Thy1.1 Tregs were purified from naïve spleens and incubated with unfractionated lymphocytes from Foxp3 gfp /Thy1.2 mice in the presence of anti-CD3 mAb and IL-12 or media. As shown in Fig. 7 , IL-12 did not induce an appreciable level of Treg (Foxp3 + ) to Tconv (Foxp3 2 ) conversion when compared to samples exposed to medium alone. Treg development, homeostasis and suppressive function are dependent upon IL-2 [25] [26] [27] [28] [29] . Further, IL-12 diminishes IL-2 expression by CD4 T cells [30] . Initially, we showed that IL-2 expression by Tconvs was modestly reduced after 48 hr with more substantial effects noted after 72 hr of IL-12 treatment (Fig. 8A ). More strikingly, IL-2 production by CD8 T cells was greatly reduced after both 48 and 72 hr (Fig. 8A ). In agreement with this diminished expression, levels of IL-2, measured by ELISA, were reduced in IL-12-treated cultures at 48 hr (Fig. 8B) . T cell proliferation is augmented by IL-2, especially in tissues at sites of inflammation [31] [32] [33] , yet the results in Fig. 1 show that Tconv proliferation is increased by IL-12 treatment. One possible explanation for these results is that IL-12 preferentially induces Effects of IL-12 on IL-2 and IL-2R Expression PLOS ONE | www.plosone.org IL-2R expression on Tconvs; IL-12 was previously shown to enhance IL-2R expression on CD8 T cells [34] . As shown in Fig. 9 (A-B, D-E), this is indeed the case. Treatment with IL-12 in the presence of anti-CD3 mAb induced up-regulation of both the a (CD25) and b (CD122) chains of the IL-2R on Tconv and CD8 T cells in a dose-dependent manner. In contrast, IL-12 treatment diminished CD25 expression on Tregs while not affecting CD122 expression, reducing the ability of Tregs to compete for IL-2 in the culture. Of note, even though IL-12 induced increased IFN-c expression by Tconv and CD8 T cells, IFN-c, by itself, was unable to induce diminished expression of IL-2 (Fig. 8B) or IL-2R upregulation on non-Treg T cells (Fig. 9C) . As shown in Fig. 1 , one of the effects of IL-12 treatment was decreased Foxp3 expression by Tregs. Previous studies showed that IL-2 is required for maximal Foxp3 expression [35] but whether IL-12-mediated reduction in Foxp3 expression occurred through effects on IL-2 signaling was not examined. To address the role for IL-2 signaling in IL-12-mediated effects, we treated cells with exogenous IL-2. IL-2 treatment partially reversed the effect of IL-12, resulting in higher levels of Foxp3 in Tregs (Fig. 9F ). [30] . To determine whether IL-27 has similar effects as IL-12, we treated lymphocyte cultures with anti-CD3 mAb and IL-27 or medium. Similarly to IL-12, IL-27 diminished IL-2 production by Tconvs and CD8 T cells (Fig. 10A ) and down-regulated Foxp3 expression by Tregs (Fig. 10B) . However, unlike IL-12, IL-27 decreased proliferation of Tconvs and Tregs and had no effect on CD8 T cell expansion (Fig. 10C , D) resulting in a relatively slight decrease in Treg frequency (Fig. 10E) . The effects of IL-12 were mediated both by diminished IL-2 expression and changes in the relative amounts of IL-2R on Tregs and Tconvs and CD8 T cells. IL-27 reduced CD25 and CD122 expression on Tregs but only marginally affected expression on non-Tregs (Fig. 10F) . Collectively, these results suggest that Tregs are better able to compete for IL-2 in IL-27 compared to IL-12-treated cultures. A previous report, using co-cultures of purified Tconvs and Tregs, showed that IL-12 could relieve Treg-mediated suppression of Tconv proliferation and activation [12] . Here, we extend these observations to analyze the effects of IL-12 on T cells in unfractionated lymphoid cultures and identify a basis for at least some of the effects of IL-12. We show that IL-12 decreases IL-2 production by Tconv and CD8 T cells and differentially modulates IL-2R expression on Tregs and non-Tregs. IL-2R is up-regulated on Tconv and CD8 T cells and down-regulated on Tregs. When coupled with decreased IL-2 production, this results in a milieu that favors non-Treg proliferation. The effect of IL-12 on CD25 expression by CD8 T cells is especially notable. In the absence of IL-12, levels of CD25 are much greater on Tregs than CD8 T cells, but treatment with IL-12 at 1 ng/ml results in levels of CD25 on CD8 T cells that exceed those on Tregs in the same culture (Fig. 8B, C) . Levels on CD8 T cells are substantially higher than those on Tconv cells, explaining why IL-12 more readily relieves Treg-mediated suppression of CD8 T cell proliferation. These results are consistent with previous reports showing that IL-2R is down-regulated on Tregs in inflamed tissues in the setting of infectious or autoimmune diseases and that IL-2 levels are reduced at these sites, resulting in the outgrowth of non-Treg T cells [5, 36, 37] . One probable consequence of these changes in IL-2R expression, in conjunction with IL-12-mediated decreases in IL-2 production, is that Tregs are less able to compete for the smaller amount of IL-2 present in the culture. These changes result in diminished Treg proliferation and perhaps decreased Treg Effects of IL-12 on IL-2 and IL-2R Expression PLOS ONE | www.plosone.org suppressive function on a per cell basis (as measured by decreased Foxp3 expression (Fig. 1E) ), resulting in decreased total Treg suppressive capacity in the culture. The importance of IL-2 for Treg function is well established [29, 35, 36] and has been confirmed in clinical studies. In patients with either hepatitis C virus-induced vasculitis or graft versus host disease, treatment with low doses of IL-2 therapy increased Treg numbers and decreased clinical disease [38, 39] . These results are also consistent with others showing that Treg numbers were increased when cancer patients were treated with low dose IL-2 [40, 41] . Exogenously administered IL-2 also corrects IL-12mediated decreases in Foxp3 levels in Tregs (Fig. 8B) ; whether low level IL-2 treatment also increases Foxp3 levels in Tregs in patients, in addition to increasing Treg numbers, has not yet been addressed. Further support for an important role for IL-2R levels on Tconv and CD8 T cells comes from studies of IL-27. IL-27 is a cytokine with pro-and anti-inflammatory properties [42] [43] [44] [45] . Mice transgenic for the expression of IL-27 exhibit diminished IL-2 production, which results in decreased numbers of Tregs and subsequently, a systemic inflammatory disease [46] . We showed that IL-27 diminished IL-2 expression, consistent with these results, but IL-27 only modestly decreased Treg frequency (Fig. 9D ). Consistent with a role for IL-2R expression on non-Tregs, there were minimal changes in IL-2R levels on these cells after IL-27 treatment (Fig. 9E) . Thus, under these conditions, Tregs were presumably able to compete for IL-2 even in the presence of reduced amounts of IL-2. Another consequence of IL-12 treatment is the induction of IFN-c expression by Tregs. Tregs express IFN-c at sites of inflammation under conditions of Th1-type lethal and non-lethal infections [5, 8] . It has been difficult to determine whether IFN-c affects Treg suppressive function, since Tregs in these settings include both IFN-c + and IFN-ccells. Here we showed that IFN-c expression did not diminish Treg suppressive function. These results are in agreement with others showing that in vitro-generated IFN-c + and IFN-c -Tregs were equally suppressive in a mouse model of colitis [13] . In other experimental settings, loss of Foxp3 and gain of inflammatory cytokine expression suggested that Tregs converted into effector T cells [22, 47] . If Tregs at sites of inflammation can generally express IFN-c and still remain suppressive, these results raise the possibility that IFN-c + Tregs, instead of entering the effector T cell pool, could lose IFN-c and revert to a classic Treg phenotype. Whether these cells would preferentially re-express IFN-c on repeat exposure to antigen remains to be determined. Several studies including this one demonstrate an important role for IL-12 in IFN-c expression by Tconv and CD8 T cells and Tregs in vitro [5, 13] . Further, IL-12 is critical for IFN-c production by Tregs in mice with autoimmune colitis [13] . However, the requirement for IL-12 in IFN-c expression by Tconv and CD8 T cells or Tregs in virus-infected animals is less clear. To investigate whether IL-12 signaling is critical for IFN-c production by Tregs in virus-infected mice, we infected IL-12Rb2 2/2 mice with the mildly neurovirulent rJ2.2 strain of MHV [48] . We detected slightly lower frequencies of virus-specific IFN-c expressing Tregs and Tconv in IL-12Rb2 2/2 mice, consistent with a previous report showing that IL-12 modestly decreased the frequencies of virus-specific T cells in MHV-infected IL-12p35 2/2 mice [49] . The absence of IL-12 signalling may not have had a substantial effect because we detected no differences in CD25 expression on Tregs and virus-specific non-Treg T cells when cells harvested from the brains of IL-12Rb2 2/2 mice and B6 mice were compared (data not shown). Our results are also consistent with others showing IFN-c production was relatively normal in IL-12 2/2 or IL-12R 2/2 mice infected with viruses such as lymphocytic choriomeningitis virus, influenza A virus, adenovirus or hepatotropic strains of MHV. In contrast, IL-12 was essential for IFN-c expression in mice infected with some intracellular nonviral pathogens, such as Leishmania species and Toxoplasma gondii [50] . In summary, our results show that Treg-mediated suppression of T cell proliferation, especially that of CD8 T cells is nearly ablated in the presence of IL-12. IL-12 induces IFN-c expression by Tregs, but this does not affect Treg immunosuppressive function. Rather, the ability of IL-12 to reverse Treg immunosuppression results, at least in part, from effects on IL-2 and IL-2R expression and by extension, differential IL-2 signaling in mixed cultures of lymphocytes.

Genetic Variation in the TNF Gene Is Associated with Susceptibility to Severe Sepsis, but Not with Mortality

BACKGROUND: Tumor necrosis factor (TNF) and TNF receptor superfamily (TNFR)-mediated immune response play an essential role in the pathogenesis of severe sepsis. Studies examining associations of TNF and lymphotoxin-α (LTA) single nucleotide polymorphisms (SNPs) with severe sepsis have produced conflicting results. The objective of this study was to investigate whether genetic variation in TNF, LTA, TNFRSF1A and TNFRSF1B was associated with susceptibility to or death from severe sepsis in Chinese Han population. METHODOLOGY/PRINCIPAL FINDINGS: Ten SNPs in TNF, LTA, TNFRSF1A and TNFRSF1B were genotyped in samples of patients with severe sepsis (n = 432), sepsis (n = 384) and healthy controls (n = 624). Our results showed that rs1800629, a SNP in the promoter region of TNF, was significantly associated with risk for severe sepsis. The minor allele frequency of rs1800629 was significantly higher in severe sepsis patients than that in both healthy controls (P(adj) = 0.00046, odds ratio (OR)(adj) = 1.92) and sepsis patients (P(adj) = 0.002, OR(adj) = 1.56). Further, we investigated the correlation between rs1800629 genotypes and TNF-α concentrations in peripheral blood mononuclear cells (PBMCs) of healthy volunteers exposed to lipopolysaccharides (LPS) ex vivo, and the association between rs1800629 and TNF-α serum levels in severe sepsis patients. After exposure to LPS, the TNF-α concentration in culture supernatants of PBMCs was significantly higher in the subjects with AA+AG genotypes than that with GG genotype (P = 0.007). Moreover, in patients with severe sepsis, individuals with AA+AG genotypes had significantly higher TNF-α serum concentrations than those with GG genotype (P(adj) = 0.02). However, there were no significant associations between SNPs in the four candidate genes and 30 day mortality for patients with severe sepsis. CONCLUSIONS/SIGNIFICANCE: Our findings suggested that the functional TNF gene SNP rs1800629 was strongly associated with susceptibility to severe sepsis, but not with lethality in Chinese Han population.

Sepsis is an infection-initiated and inflammation-induced syndrome. Despite progress in the development of antibiotics and other supportive care therapies, severe sepsis remains an unconquered challenge for the clinicians with an unacceptable high mortality rate of 30%-50% [1] . The response to infection is diverse among different individuals. Given the same therapies, most sepsis patients will recover and do well, while a small, but significant portion, will develop severe sepsis and multiple organ system failure, refractory hypotension and death [2, 3] . Currently, more and more evidence showed that genetic factors played an important role in the development and severity of sepsis [4, 5, 6, 7, 8, 9, 10, 11] . Common sequence variants within genes involved in pro-inflammatory response have received particular attention [12, 13] . Although the pathogenesis of sepsis remains incompletely understood, an excessive pro-inflammatory response has been established as a fundamental component of severe sepsis [14] . The proinflammatory cytokine TNF-a is an essential component in the host immune response to infection and has been widely reported to be an important mediator in severe sepsis and septic shock. High circulating levels of TNF-a were correlated with poor outcomes in sepsis patients [15] . TNF-a and lymphotoxin-a (LT-a) share the same receptors as well as many biological activities, and they are central mediators of immune responses [16] . TNF-a and LT-a are encoded by adjacent gene loci in the central or class III region of the human major histocompatibility complex (MHC), between the HLA class I and II genes on the short arm of chromosome 6 [17] . Several SNPs within the promoter region of TNF (2238, 2308, 2857, 2863, 21031) and the first intron of LTA (+252) were thought to influence TNF-a and LT-a production, and have therefore been identified as candidate variants that might influence susceptibility to and/or outcomes from severe sepsis and infectious diseases [18, 19, 20, 21, 22, 23, 24, 25] . In particular, rs1800629 (TNF 2308) and rs909253 (LTA +252) have been the focus of many investigations on sepsis. Although several studies have identified associations for rs1800629 and rs909253 with sepsis risk or outcomes [18, 19, 20, 23, 24] , other studies have not replicated the associations [21, 26, 27, 28] . This inconsistency may be due to small samples size studied and ethnic differences [29] . TNF-a and LT-a exert their pleiotropic functions by activating intracellular signaling cascades via binding to two types of receptors, TNFR-1 (encoded by the TNFRSF1A gene) and TNFR-2 (encoded by the TNFRSF1B gene) [30] . TNFR1deficient mice are resistant to endotoxic shock and have prolonged survival with less hypothermia [31] . TNFR2 influences the biological activity of TNF-a and LT-a both in a membranebound and a soluble form. Membrane-bound TNFR2 facilitates activation of nuclear factor (NF)-kB and mitogen-activated protein kinase signaling cascades upon binding with TNF-a and LT-a, whereas soluble TNFR2 is capable of binding and inactivating circulating TNF-a and LT-a [16, 30] . Moreover, animal studies showed that TNFR2 mediated protective effects in the development of severe sepsis [31] . Recent studies proposed that genetic variation in TNFRSF1A and TNFRSF1B was associated with susceptibility to inflammatory and autoimmune diseases, such as tuberculosis, systemic lupus erythematosus, rheumatoid arthritis and Crohn's disease [32, 33, 34, 35] . However, to date, only one study investigated the role of TNFRSF1A and TNFRSF1B polymorphisms in sepsis susceptibility and mortality [36] . Considering the important role of TNF-a, LT-a, TNFR1 and TNFR2 in the pathogenesis of severe sepsis, we hypothesized that genetic variation in TNF, LTA, TNFRSF1A and TNFRSF1B might be associated with susceptibility to and outcomes from severe sepsis in Chinese Han population. To test this hypothesis, we conducted a relatively large-scale case-control study enrolling 432 severe sepsis patients, 384 sepsis patients and 624 healthy individuals to investigate the association of genetic variants in TNF, LTA, TNFRSF1A and TNFRSF1B with severe sepsis susceptibility and prognosis in Chinese Han population. Furthermore, we investigated the association between the genotypes of the TNF gene SNP rs1800629 and TNF-a concentration in culture supernatants of LPS simulated PBMCs obtained from healthy donors and in serum from severe sepsis patients. A total of 432 severe sepsis patients, 384 sepsis patients and 624 health volunteers were enrolled in this case-control study. According to the mortality within 30 days, severe sepsis patients were divided into survivor and non-survivor groups. The baseline characteristics and clinical data of all subjects are shown in Table 1 . The average age and proportion of male among the severe sepsis, sepsis and healthy control groups did not show significant difference. The primary source of infection in severe sepsis patients was the lungs (69.9%), followed by abdomen (21.8%), blood stream (3.5%), urinary tract (2.5%) and others (2.3%). The overall 30-day mortality rate of severe sepsis patients was 36.1%. The mean APACHE II and SOFA score in nonsurvivor group was higher than that in survivor group (P,0.05). Association Analyses of TNF, LTA, TNFRSF1A and TNFRSF1B SNPs with Susceptibility to Severe Sepsis The genotyping success rates of all tested SNPs ranged from 95% to 99% and none of the ten SNPs diverged significantly from Hardy-Weinberg equilibrium (P.0.05) ( Table 2 ). The allele and genotype distributions of all tested SNPs in severe sepsis patients, sepsis patients and healthy controls are listed in Table 3 . Our results showed that rs1800629 (2308G/A), located in the promoter region of TNF, was associated with significantly increased risk for severe sepsis. The frequency of rs1800629A in severe sepsis patients was significantly higher than that in both the healthy control subjects (P = 0.00028, OR = 2.08) and the sepsis patients (P = 0.00035, OR = 2.39), and the difference remained significant after Bonferroni correction. Moreover, in multivariate analyses after adjustment for covariates, rs1800629A was still significantly associated with the development of severe sepsis when compared with healthy control group (P adj = 0.00046, ORadj = 1.92) and sepsis group (P adj = 0.002, OR adj = 1.56). The genotype distribution of rs1800629 in the severe sepsis group was also significantly different from that in the healthy control group (P adj = 0.003) and the sepsis group (P adj = 0.004), and the significance remained after Bonferroni correction. However, the difference of the allele and genotype frequencies of rs1800629 between subjects with sepsis and healthy controls were not statistically significant (P.0.05). When we analyzed the allele and genotype distributions of the other nine SNPs (rs361525, rs1799724, rs1799964, rs767455, rs4149570, rs1061622, rs3397, rs1800630 and rs909253), no significant difference was found between the severe sepsis, sepsis and healthy control groups (Table 3) . Association Analyses of TNF, LTA, TNFRSF1A and TNFRSF1B SNPs with Severe Sepsis Outcomes We next investigated the association between all tested SNPs and 30-day mortality. The overall 30-day mortality rate among severe sepsis patients was 36.1%. We compared the allele and genotype distributions of all tested SNPs between survivors and non survivors of severe sepsis patients. No association was observed between TNF, LTA, TNFRSF1A and TNFRSF1B variants and 30-day mortality in the severe sepsis cohort in either the unadjusted or adjusted models (Table 4 ). To determine whether rs1800629 genotypes influenced TNF-a production, we investigated TNF-a levels in culture supernatants of PBMCs obtained from 24 healthy volunteers. We observed a significant association between TNF-a levels and rs1800629 genotypes under the LPS-stimulated condition. AA+AG genotypes were associated with higher levels of TNF-a compared with GG genotype after LPS stimulation (P = 0.007) ( Figure 1 ). However, no significant association was observed under the unstimulated condition. Furthermore, we measured TNF-a serum concentrations in 120 severe sepsis patients, including 104 patients with rs1800629GG genotype, 14 patients with GA genotype and 2 patients with AA genotype. Our results showed that rs1800629A allele was associated with higher TNF-a serum concentrations on the first day of severe sepsis. As shown in Figure 2 , the serum concentration of TNF-a in severe sepsis patients with AA+AG genotypes was significantly higher than that of patients with GG genotype (550.4673.6 pg/mL vs. 488.0668.5 pg/mL, P = 0.001). To control confounding variables, we used the possible confound- ing factors (age, gender and APACHE II scores) as covariates in a linear regression model and found that the rs1800629 genotypes remained associated with TNF serum concentration (P adj = 0.02). Several genetic variants within genes involved in pro-inflammatory response have been associated with morbidity and mortality in patients with severe sepsis or septic shock, which is a complex and multifactorial syndrome [2] . TNF-a is an important pro-inflammatory cytokine involved in sepsis; several functional SNPs in TNF and LTA have been extensively studied in sepsis [29, 37] . However, results from previous studies were inconsistent. Discrepancies among previous studies may have resulted from differences in the populations studied, sepsis phenotype or imprecise definition of phenotype and limited sample size [38] . Considering these factors that might affect the results, we designed the present study with large samples size to achieve greater statistical power. Furthermore, all the samples were recruited from central Chinese Han population, thus the ethnic heterogeneity could be eliminated. To our knowledge, this was the first relatively large-scale study investigating associations of genetic variants within TNF, LTA, TNFRSF1A and TNFRSF1B with severe sepsis in Chinese Han population. Our results provided evidence that rs1800629, a functional SNP in the promoter region of TNF, was significantly associated with susceptibility to severe sepsis in Chinese Han population. The association between rs1800629 and severe sepsis risk may be explained by its influence on the expression of TNF-a. In this study, we found that the risk allele (rs1800629A) is associated with increased TNF-a production in PBMCs from healthy subjects after stimulation with LPS. Moreover, we found that TNF-a serum levels in severe sepsis patients with AA+AG genotypes for rs1800629 were significantly higher than those in the individuals with GG genotype. Previous studies also showed that rs1800629A allele was associated with a six fold higher expression of both basal and induced TNF mRNA [39, 40] . Menges et al. found that the plasma TNF-a concentrations in patients with sepsis secondary to severe traumatic injury were significantly elevated in rs1800629A carriers on the first day after admission and for the following 14 days [41] . As TNF-a plays a pivotal role in the pathogenesis of severe sepsis in response to infection, it is reasonable to assume that patients with rs1800629A allele might produce a higher amount of TNF-a, and therefore become more susceptible to severe sepsis. In contrast to the TNF 2308G/A polymorphism, the LTA +252A/G was not associated with the development of severe sepsis in our study. Our data showed that 2308G/A of TNF and +252A/G of LTA were in weak linkage disequilibrium (LD) (D9 = 0.118) in Chinese Han population. The LD pattern is quite dissimilar to Caucasian population, which might result from the racial difference [24] . Rs1800629 was not associated with mortality among subjects with severe sepsis in our study. This was consistent with the study by Stuber et al., which demonstrated that the rs1800629 genotypes were not associated with poorer prognosis in severe sepsis. However, they did not find an association between rs1800629 genotypes and plasma TNF-a levels [26] . Recently, Teuffel et al. conducted a systematic review and meta-analysis, which also concluded that rs1800629 (TNF 2308 AA/AG, TNF2) was associated with susceptibility to sepsis, but not with sepsis mortality [29] . Several studies have proposed that genetic variation in TNFRSF1A and TNFRSF1B was associated with susceptibility to inflammatory and autoimmune diseases, such as tuberculosis, systemic lupus erythematosus, rheumatoid arthritis and Crohn's disease [32, 33, 34, 35] . However, up to now, only one case control study investigated associations between TNFRSF1A and TNFRSF1B polymorphisms and sepsis susceptibility [36] . Four potentially functional SNPs in TNFRSF1A and TNFRSF1B were genotyped in our study. However, none showed association with susceptibility to or death from severe sepsis in Chinese Han population. Our findings were consistent with the results of Gordon et al. that five functional SNPs in TNFRSF1A and TNFRSF1B were not associated with susceptibility to or outcomes from sepsis in Caucasian population [36] . Potential limitations of this study should be addressed. First, although we knew that different pathogens had different impact on severity and outcomes of sepsis, we did not perform stratification analysis by different pathogens due to small number of cases with a definite microbiologic diagnosis. Second, we did not resequence these genes or select tag SNPs for genotyping. Instead, only ten potentially functional SNPs in TNF, LTA, TNFRSF1A and TNFRSF1B were included in our study, which was far from comprehensive. Indeed, these four genes are highly polymorphic. Therefore, it was possible that some important SNPs might be missed or the observed association might be due to other polymorphisms in LD with the studied ones. Additionally, assuming the prevalence of 0.01 for severe sepsis and using a significance level of 0.05, our study with 432 severe sepsis patients and 624 healthy controls had about 80% power to detect a 5% risk allele with an odds ratio of 1.63. Variant with an effect size smaller than this cannot convincingly be excluded based on these results. Therefore, our results cannot exclude variant associations with weaker effects between severe sepsis and the other three candidate genes (LTA, TNFRSF1A, TNFRSF1B). A more highly powered study involving thousands of subjects may yet exclude the role of these variants in severe sepsis susceptibility and outcomes. In conclusion, our relatively large scale association study demonstrated that individuals with a functional variant in the promoter region of TNF may confer susceptibility to severe sepsis. However, common functional genetic variants in TNF, LTA, TNFRSF1A and TNFRSF1B were not associated with severe sepsis mortality in Chinese Han population. This study was approved by the Ethics Study Board of Zhongshan Hospital, Fudan University, Shanghai, China (Record no: 2006-23). Written informed consent was obtained from patients or the next of kin, carers or guardians on the behalf of the participants before enrollment. From May 2005 to March 2011, a total of 432 severe sepsis patients, 384 sepsis patients and 624 ethnic-matched healthy controls were enrolled in this study ( Table 1) . The severe sepsis patients were those admitted to the Emergency, Surgical and Respiratory ICU at Zhongshan Hospital. The sepsis patients were those admitted to Zhongshan Hospital, but did not develop severe sepsis during hospital stay. The sepsis patients were considered as at risk controls for severe sepsis. Of 384 sepsis patients, 174 patients overlapped with that from our previous study [42] . Another 210 sepsis patients were collected between May 2008 and March 2011, and these patients were not included in our previous study. Sepsis patients recruited in the current study included multitrauma subjects and patients with a history of chronic heart, renal, liver or pulmonary failure, thus they spent a long time (more than 8 days on average) on ICU (Table 1) . Sex-and age-matched controls were selected from healthy blood donors. To reduce the potential confounding from ethnic backgrounds, only central Han Chinese individuals were recruited in this study. The diagnosis of sepsis was based on the criteria presented at the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference in 1992 [43] . Severe sepsis was defined as sepsis in combination with infection-induced acute organ dysfunction in at least one organ. Acute organ dysfunction was defined as Sequential Organ Failure Assessment (SOFA) scores more than 2 for the organ in question. Baseline characteristics (age, gender and previous health status), as well as clinical data including Acute Physiology and Chronic Health Evaluation II (APACHE II) and SOFA scores, source of infection, microbiology and ICU mortality were obtained after the patient met sepsis criteria. The APACHE II and SOFA scores were calculated in the first 24 hours after the diagnosis of sepsis and severe sepsis. All patients included in the protocol were followed up for 30 days or hospital discharge. When cultures were absent or negative, the source of infection was determined by two senior physicians. Exclusion criteria included age below 18 years, pregnancy, severe chronic respiratory disease, severe chronic liver disease (defined as a Child-Pugh score of .10), malignancy, using of high-dose immunosuppressive therapy and AIDS diagnosis. Questionnaires were obtained from all control subjects to document smoking status, and history of chronic illness or severe sepsis. Healthy controls were defined as individuals without any recent acute illness, chronic illness or history of sepsis and severe sepsis. Previous studies found that several functional SNPs in TNF, LTA, TNFRSF1A and TNFRSF1B were associated with inflammatory and autoimmune diseases. In our study, SNPs in TNF, LTA, and TNFRSF1A and TNFRSF1B were selected based on the following criteria: (1) location within the gene region (promoter, intron, exon, 39UTR and 59UTR); (2) association with inflammatory and autoimmune diseases such as sepsis, asthma, tuberculosis, systemic lupus erythematosus, rheumatoid arthritis and Crohn's disease in more than two studies. A total of ten SNPs were selected and genotyped in our study. Location and characterization of all selected SNPs were listed in Table 2 . Genomic DNA was extracted from whole blood with a FlexiGene DNA Kit (Qiagen, Hilden, Germany) in accordance with the protocol of the manufacturer. Six SNPs (rs1800629, rs1799724, rs361525, rs1800630, rs1799964 and rs909253) in TNF and LTA were selected and genotyped by direct sequencing. The sequencing reactions were performed using Applied Biosystems BigDye (version 3.1) chemistry (Applied Biosystem, Foster City, CA, USA), and the sequences were resolved using an ABI 3730 Genetic Analyzer. Analyses of the sequence traces were performed using the Staden package and were double scored by a second operator. The primers and PCR protocols used were shown in Table S1 . Four SNPs in TNFRSF1A (rs767455, rs4149570) and TNFRSF1B (rs1061622, rs3397) were selected and genotyped on the GenomeLab SNPstream high-throughput 12-plex genotyping platform (Beckman Coulter, Fullerton, CA) following the manufacturer's instructions. The primers for PCR and single base extension were performed with Beckman Coulter Autoprimer software and shown in Table S2 . To determine the associations between rs1800629 genotypes and TNF-a levels in PBMCs, we investigated 15 subjects with rs1800629GG genotype, 8 subjects with GA genotype and 1 subject with AA genotype. PBMCs were derived by using Ficoll gradient density centrifugation method. Isolated PBMCs were plated at a density of 1610 6 cells/ml in 24-well plates and cultured in RPMI 1640 medium with 10% FBS at 37 uC with 5% CO 2 . The cells were then incubated for 6 hours in presence or absence of 100 ng/ml Escherichia coli 0111:B4 LPS (Sigma, USA). After incubation, supernatants were harvested and stored at 280uC until use. Blood samples (5 mL) were collected within 24 hours of meeting criteria for severe sepsis. Samples were centrifuged at 4uC for 10 min at 3200 rpm within 60 min after collection. Then the serum was stored at 280uC until use. TNF-a level was determined by human ELISA kit (R&D Systems, USA) according to the manufacturer's protocol. The genotype data of cases and controls was analyzed for deviations from Hardy-Weinberg equilibrium by the Haploview v4.1 software [44] . The differences in allele and genotype distributions between severe sepsis and control groups were compared using x 2 -test or Fisher's exact test when appropriate. The test for association with genotypes used the global genotype test in the 362 contingency table. Allele frequencies of cases and controls were used to calculate the OR and the 95% CI. Multivariate logistic regression was used to adjust for potential confounding factors. When comparing severe sepsis group to sepsis group, we entered the genotypes or alleles in the multivariate models controlling for the confounding variables including age, gender, history of diseases, source of infection, APACHE II and SOFA scores. When comparing severe sepsis patients to healthy controls, age and gender were included in the multivariate models. The Bonferroni method was used to correct for multiple comparisons where applicable. The power analysis was performed using the Genetic Power Calculator web tool [45] . A two tailed Pvalue of ,0.05 was considered statistically significant, whereas a value of corrected P,(0.05/number of tests) was considered significant after Bonferroni correction. Continuous variables were described as either a mean 6 standard deviation, or as a median with interquartile range. TNF-a serum levels between individuals with different rs1800629 genotypes (AA+GA vs. GG) were compared by Student's t-test. To determine whether an association with rs1800629 genotypes might depend on other potential confounding factors for TNF-a serum levels, we investigated the association of rs1800629 genotypes by adding the polymorphisms to a linear regression model controlling for age, gender and APACHE II scores. The software used for statistical calculations was SPSS 15.0 (SPSS Inc., Chicago, IL, USA) unless specified. Table S1 Primers and PCR protocols for six SNPs in TNF and LTA. (DOC)

Identification, Characterization and Application of a G-Quadruplex Structured DNA Aptamer against Cancer Biomarker Protein Anterior Gradient Homolog 2

BACKGROUND: Anterior gradient homolog 2 (AGR2) is a functional protein with critical roles in a diverse range of biological systems, including vertebrate tissue development, inflammatory tissue injury responses, and cancer progression. Clinical studies have shown that the AGR2 protein is overexpressed in a wide range of human cancers, including carcinomas of the esophagus, pancreas, breast, prostate, and lung, making the protein as a potential cancer biomarker. However, the general biochemical functions of AGR2 in human cells remain undefined, and the signaling mechanisms that drive AGR2 to inhibit p53 are still not clearly illustrated. Therefore, it is of great interest to develop molecular probes specifically recognizing AGR2 for its detection and for the elucidation of AGR2-associated molecular mechanism. METHODOLOGY/PRINCIPAL FINDINGS: Through a bead-based and flow cytometry monitored SELEX technology, we have identified a group of DNA aptamers that can specifically bind to AGR2 with K(d) values in the nanomolar range after 14 rounds of selections. Aptamer C14B was chosen to further study, due to its high binding affinity and specificity. The optimized and shortened C14B1 has special G-rich characteristics, and the G-rich region of this binding motif was further characterized to reveal an intramolecular parallel G-quadruplex by CD spectroscopy and UV spectroscopy. Our experiments confirmed that the stability of the G-quadruplex structure was strongly dependent on the nature of the monovalent ions and the formation of G-quadruplex structure was also important for the binding capacity of C14B1 to the target. Furthermore, we have designed a kind of allosteric molecule beacon (aMB) probe for selective and sensitive detection of AGR2. CONCLUSION/SIGNIFICANCE: In this work, we have developed new aptamer probes for specific recognition of the AGR2. Structural study have identified that the binding motif of aptamer is an intramolecular parallel G-quadruplex structure and its structure and binding affinity are strongly dependent on the nature of the monovalent ion. Furthermore, with our design of AGR2-aMB, AGR2 could be sensitively and selectively detected. This aptamer probe has great potential to serve as a useful tool for early diagnosis and prognosis of cancer and for fundamental research to elucidate the biochemical functions of AGR2.

Anterior gradient homolog 2 (AGR2) was identified initially as a secretory factor expressed in the anterior region of the dorsal ectoderm in Xenopuslaevis embryos, where it was postulated to mediate the specification of dorsoanterior ectodermal fate, particularly in the formation of the cement gland [1] . Clinical studies have further shown that the AGR2 protein is overexpressed in a wide range of human cancers, including carcinomas of the esophagus, pancreas, breast, prostate, and lung [2] [3] [4] [5] [6] . More biological studies in these cancer cell lines have indicated a significant role for AGR2 in tumor-associated pathways, including tumor growth, cellular transformation, cell migration, limb regeneration, and metastasis [5, [7] [8] [9] . However, the general biochemical functions of AGR2 in human cells remain undefined, and the signaling mechanisms that drive AGR2 to inhibit p53 are still not clearly illustrated [10] . Therefore, the development of molecular ligands specifically recognizing AGR2 is of great significance to early diagnosis and prognosis of cancer and to fundamental research for the elucidation of the biochemical functions of AGR2. Various ligands have been developed for specific molecular recognition, such as small molecules, antibodies, and peptides [11] [12] [13] . More recently, another type of molecular ligand, named aptamer, has drawn significant attention. Aptamers, singlestranded modified or unmodified oligonucleotides (RNA or DNA), are generated through in vitro selection process or SELEX (Systematic Evolution of Ligands by EXponential enrichment) with high binding affinity and specificity towards defined targets [14, 15] . The selected aptamers can recognize a wide variety of targets, including small molecules, proteins, cells and tissues relying on their diverse tertiary structures. Compared to antibodies, aptamers have low molecular weight, fast tissue penetration rate, high stability and low immunogenesis [16] . They can be chemically synthesized with low cost and modified easily with various reporters [17] . Furthermore, they can be ligated and/or amplified by enzymes in vitro [18] . These advantages make aptamers promising ligands for medical and pharmaceutical research, such as drug development, disease diagnosis, and targeted therapy [19] . The possibilities provided by aptamers are enormous, and some aptamers have already shown many important applications in bioanalysisand biomedicine [20] [21] [22] [23] . Particularly, several aptamers have been generated against cancer-related proteins, such as PDGF, VEGF, HER3, NFkB, tenascin-C, or PMSA [24] [25] [26] . Many aptameric sensors, probes and assays have been developed to allow sensitive and selective detection of these cancer biomarker proteins [27] . For instance, Yang et al has reported a lightswitching excimer aptamer probes for sensitive quantitative detection of PDGF in cell media [28] . Kwon et al have developed a functionalized polypyrrole nanotube with aptamer to build a VEGF biosensor [29] . Aptamers have also been applied for molecular imaging to in vivo characterize the complex pathogenic activities that accompany tumor growth for disease early diagnosis and pathogenesis measurement [30] [31] [32] [33] . Since the targets for aptamers could be intracellular, extracellular or cell-surface biomolecules, various therapeutic methods have been developed using the aptamers as targeting reagents [34] [35] [36] [37] , which greatly broaden the range of targeted therapy. In addition, some therapeutically useful aptamers have been found to inhibit protein-protein interactions, such as receptor-ligand interactions, and thereby function as antagonists [38] . In this study, using the bead-based and flow cytometry monitored SELEX technology, we aimed to obtain specific aptamers to AGR2 and study theirs structure and potential function. Beads-based SELEX allowed the use of simple, yet effective, flow cytometry analysis to monitor the progress of the selection, avoiding the tedious, time consuming and radioactive EMSA process [39] [40] [41] [42] [43] . After 14 rounds of selection, we have identified a group of DNA aptamers that specifically bound to AGR2 with high affinities. Structural studies on one of the aptamer sequences, C14B, revealed an intramolecular parallel Gquadruplex, and its structure and binding affinity to AGR2 depend on K + ion intensively. Furthermore, we designed an allosteric molecule beacon AGR2-aMB based on the identified aptamer, which enables simple, sensitive and selective detection AGR2. The aptamer sequences and AGR2-aMB reported in this study are potentially useful tools for early diagnosis and prognosis of cancer and for fundamental research to elucidate the biochemical functions of AGR2. To identify aptamers against AGR2, recombinant AGR2 was fused with glutathione-S-transferase (GST) to facilitate the attachment of the protein to solid supports (Sepharose GSHbeads). The resulting AGR2-GST-beads were used as the positive target in SELEX while the GST-beads as negative control to remove non-specific surface binding sequences. The process of in vitro sepharose-bead-based SELEX is schematically illustrated in Figure 1 . An 87-nucleotide (87-nt) single-stranded DNA (ssDNA) library with 45 random bases flanked by two primer sequences (22-nt and 20-nt) was subjected to the SELEX procedure. The library was first allowed to interact with excess negative control beads, and only the DNA sequences that did not bound to the GST-beads were collected. The collected sequences were then incubated with AGR2-GST-beads. After rigorous washing, those sequences that either did not bind, or only weakly bound to the target were discarded. Only the sequences that bound strongly enough were retained on beads, and the bead-ssDNA complexes were collected and amplified by PCR for the next round of selection. After multiple rounds of selection, the subtraction process efficiently reduced the DNA sequences that bound to the GST beads, while those AGR2-specific aptamer candidates were gradually enriched. The progress of the selection process was monitored by flow cytometry. The stronger binding of DNA library to AGR2, the more FAM labeled sequences bound to the beads, thus the higher fluorescence intensity the beads would emit. With the increasing number of selection cycles, steady increases in fluorescence intensity on the target beads were observed (Figure 2a) . The binding affinity of the enriched library after 14 rounds of selection was determined to be in the nanomolar range (K d = 64.165.4 nM), while there was no observable binding of the library to control beads ( Figure 2b ). These results suggested that the DNA aptamers specifically recognizing AGR2 were enriched during the selection process. After 14 rounds of selection, the enriched DNA pool was cloned and sequenced. The sequencing data for clones were analyzed by using sequencing analysis software Clustal W 6.0 [44] . The sequences were grouped based on the homology similarity of the DNA sequences from individual clone. Among the sequences from 62 clones, there were two subfamilies each containing multiple sequences with common features. One subfamily is guanosine-rich sequences (22 clones) , and the other is thymine-rich sequences (40 clones) ( Figure S1 ). Four sequences were chosen and synthesized for further characterization (Table S1 ): three sequences from Grich subfamily, C14A (appeared 2 times), C14B (appeared 3 times) and C14C (appeared 5 times), and one sequence C14D (appeared 2 times) from T-rich subfamily. Only one T-rich sequence was chosen because of T-rich sequences tend to from non-rigid tertiary structures and the chance of being aptamer was thought to be very low. As shown in Figure 2c , C14B can bind AGR2 with the highest affinity (K d = 13.167.2 nM). Titration of C14B to GSTbeads revealed no observable binding, establishing that the binding target of C14B was indeed AGR2. Other three sequences have similar binding constants to AGR2 but a little weaker than C14B, in which the K d of C14A is 20.965.2 nM, C14C is 44.667.0 nM and C14D is 48.4615.6 nM. ( Figure S2 ). Since C14B is the best aptamer we obtained from the four sequences tested, it was choses for future optimization and characterization. The length of C14B is 87mer, which is disadvantageous for future applications because it is inconvenient and expensive to synthesize such a long sequence. To identify the binding region of the aptamer, the marginal sequences of C14Bwas gradually truncated. Two truncated sequencesC14B0 and C14B1 from C14B were shown in Table 1 . Subsequent binding affinity experiments revealed that both C14B0 (8.563.6 nM) and C14B1 (19.165.1 nM) have similar K d to AGR2 as that of original C14B (13.167.2 nM). The majority of eliminated sequences were primer sequences, implying that the binding region of the aptamer was the middle of random sequences and the primer sequences do not or contribute little to the binding affinity of aptamer. C14B1 has five portions of poly-G and we designed five truncated sequences by removing one of poly-G portion each (Table S2) . Removal of any poly-G portion would destroy its binding affinity to certain extend ( Figure S3) , suggesting the whole G-motif is required for aptamer binding. Taking together, these results indicate that C14B1, which was only 33 mer, was the essential binding region to AGR2. Thus, C14B1 was applied for further characterization and probe design. In order to demonstrate the specific interaction between AGR2 and C14B1, three control proteins BSA, thrombin and trypsin were coupled with the NHS-sepharose-beads and then were incubated with C14B1. As shown in Figure 3 , C14B1 could bind to target AGR2 strongly, while there were no or little binding affinity towards thrombin, BSA and trypsin, demonstrating the high selectivity of the aptamer. Further examination revealed that C14B1 has multiple stretches of guanines (CGGGTGGGAGTTGTGGGGGGGGGTGG-GAGGGT). It is well-known that guanine-rich sequences can fold into four-stranded secondary structures called quadruplexes. According to G-Quadruplex prediction formula d(G 3+ N 1-7 G 3+ N 1- Figure 1 . Schematics of systematic evolution of DNA aptamers against AGR2. In sepharose-beads-based SELEX process for protein AGR2, the GST-beads were incubated with the ssDNA library for counter-selection to remove nonspecific sequences. The unbound DNAs were then incubated with AGR2-GST-beads for target-selection. After harshly washing, the AGR2-specific DNA sequences were subsequently amplified by PCR for the next round of selection, or for cloning and sequencing to identify individual aptamers after flow cytometry analysis. doi:10.1371/journal.pone.0046393.g001 single stranded G-quadruplex structure, the dependence of melting temperature on the concentration of C14B1 was studied [48] . The melting temperature of C14B1 was found to be 59uC, which was independent of oligonucleotide concentration ( Figure S4 ), indicating that the aptamer forms an intramolecular Gquadruplex. Binding Affinity and Structure Stability of C14B1 is K + Dependent Many DNA aptamers have been found to form G-quadruplex structures [49] [50] [51] . It is well established that the existence of a monovalent cation (especially potassium) in the center of these tetrads can significantly stabilize G-quadruplexes [52] [53] [54] . Thus, we investigated how a cation would affect the structural stability of C14B1 and its binding activity towards to AGR2. As shown in Figure 4a , the stability of the C14B1 G-quadruplex structure is strongly dependent on the presence of the monovalent ion. With 60 mM of KCl, strong CD peaks were observed, suggesting the formation of stable G-quadruplex structure. Replacing KCl with the same concentration of LiCl, NaCl, and MgCl 2 led to a dramatic intensity decrease in CD. We further studied the effect of K + concentration on the stability of G-quadruplex. In Figure 4b , addition of 0.1 mM K + in phosphate buffer dramatically increased the CD intensity at 240 and 260 nm. CD absorption intensity enhanced with the increase of K + concentration and reached a plateau when K + concentration was higher than 20 mM. The binding affinity of C14B1 at different concentrations of K + was also investigated. As the flow cytometry results shown in Figure 4c , very weak binding of C14B1 towards AGR2-beads was observed when there was no K + in the buffer, and the binding affinity kept increasing with the addition of K + . The binding constants of C14B1 were measured and compared in the presence or absence of K + (Figure 4d ). The K d value in the presence of K + was determined to be 6.261.9 nM. The results demonstrated that formation of G-quadruplex is important for the binding capability of aptamer C14B1 to its target, which is highly K + dependent. Clinical studies have already shown that the AGR2 protein is overexpressed in a wide range of human cancers. The sensitive and selective detection of AGR2 is thus of great importance to early cancer diagnostics. Herein, by applying the selected and optimized aptamer C14B1, we designed and developed an allosteric molecular beacon against AGR2, named AGR2-aMB, which converts the molecule recognition property of aptamer to fluorescence flow cytometry signal for AGR2 sensing [55] . Figure 5 illustrates the working principle of AGR2-aMB. An AGR2-aMB is a ssDNA consisting of an streptavidin (SA) aptamer sequence [56] , a C14B1 sequence, a short sequence complimentary to a small part of the SA aptamer sequence, and a fluorophore. A stable hairpin structure is formed by intramolecular hybridization between the SA aptamer sequence and the complementary sequence, temporarily disabling the probe's ability to bind with SA beads. Consequently, when incubated with SA beads, no probe can bind to SA beads, and the beads display very low fluorescence. In the presence of AGR2, however, the C14B1 sequence in the loop of aMB binds to the target sequence, which in turn disrupts the hairpin structure to free the SA binding sequence, thereby activating the probe's binding affinity to SA beads. Thus, the AGR2-bound probe will bind to SA beads, which will fluoresce strongly because the probe is FAM labeled. Target molecules can be detected and quantified by reading the fluorescence intensity of SA beads through flow cytometer. We designed AGR2-aMB, with the following sequence: CGACGCACCGATCGCAGGTTCGGGATTTTCGGGTGG-GAGTTGTGGGGGGGGGTGGGAGGGTT-FAM, where the AGR2 binding region is underlined and the SA aptamer sequence is in bold. Stable hairpin structure is formed by intramolecular hybridization ( Figure S5 ). In our experiment, when only AGR2-aMB incubated with SA beads, no probe can bind to SA beads, and the beads fluoresced very weakly. In the presence of AGR2, however, the fluorescence intensity of SA beads increase significantly, suggesting the binding of AGR2-bound probe to SA beads. With the increase of AGR2 concentration, steady increases in fluorescence intensity on the target beads were observed. AGR2 at 100 nM concentration can be easily detected (Figure 6a) . A series of proteins including BSA, trypsin, thrombin, and IgG were employed as controls (500 nM), and no distinct fluorescent signals were observed (Figure 6b) , indicating an excellent specificity of the AGR2-aMB towards target molecule. The results demonstrated the sensitivity and specificity of the allosteric molecular beacon probe, implying its potential for application in real sample analysis, such as protein function study and disease diagnosis. In conclusion, we have developed new aptamer probes for specific recognition of the AGR2. In the SELEX process, AGR2-GST fused protein was used as the target protein, and linked to sepharose beads through the mild, yet specific, noncovalent GST-Glutathione interaction. Beads-based SELEX allowed the use of simple, yet effective, flow cytometry analysis to monitor the progress of the selection, avoiding the tedious, time consuming and radioactive EMSA process. Through multiple rounds of selection with GST as a control, we have identified aptamers that selectively recognize AGR2 with nanomolar K d values. CD measurements and melting-temperature assays demonstrated that the optimal aptamer C14B1 forms an intramolecular parallel G-quadruplex structure and its structure and binding affinity are strongly dependent on the nature of the monovalent ion. Furthermore, with our design of AGR2-aMB, AGR2 could be sensitively and selectively detected The aptamer sequences and sensors reported here has great potential to serve as a useful tool for early diagnosis and prognosis of cancer and for fundamental research to elucidate the biochemical functions of AGR2. The HPLC-purified library containing a central randomized sequence of 45 nucleotides flanked by two 20-nt and 22-nt primer hybridization sites. Initial library was: 59-TCT CGG ACG CGT GTG GTC GG-N45-C TCG CTG CCT GGC CCT AGA GTG-39, forward primer: 59-FAM-TCT CGG ACG CGT GTG The plasmid pGEX-GST-AGR2 was transformed into the engineering strain BL-21 and the GST-tagged AGR2 protein was expressed. After purification with glutathione sepharose beads (GE healthcare) by affinity chromatography, the GST-AGR2 fused protein was linked to sepharose beads for positive selection. After being washed several times with W1 buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.07% b-mercaptoethanol, 1% Triton X-100), the final purified AGR2-GST-beads were stored in sterilized PBS buffer. Flow cytometry analysis with TAMRA-labeled anti-AGR2 monoclonal antibody (Santa Cruz) and SDS-PAGE indicated that the GST-AGR2 fused protein was successfully linked to the sepharose beads. The procedures of selection were as follows. The ssDNA pool (200 pmol) dissolved in binding buffer (200 mL, 16PBS buffer containing 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 and 2.0 mM KH 2 PO 4 , pH 7.4) was denatured by heating at 95uC for 5 min, then quickly cooled on ice for 10 min, and subsequently incubated for another 10 min at room temperature before binding. The ssDNA pool was then incubated with negative GST beads (1.0610 5 beads) for counter selection to remove sequences non-specifically binding to GST beads. After filtering with a homemade filter column, the filtrate was incubated with positive AGR2-GST-beads (1.0610 5 beads) at 37uC for 45 min. The unbound or nonspecifically bound oligoes were removed by filtration. The sequences bound to the target-coated beads were then amplified by PCR with FAM and biotin-labeled primers (5-15 cycles of 0.5 min at 94uC, 0.5 min at 53uC, and 0.5 min at 72uC, followed by 5 min at 72uC; the Easytaq plus polymerase and dNTPs were obtained from Transgen Beijing). After denaturation in NaOH (0.1 M), the selected sense ssDNA was separated from the biotinylated antisense ssDNA strand on streptavidin-coated sepharose beads (GE healthcare) and used for next round selection. For the first-round selection, the amount of initial ssDNA pool was 5 nmol, dissolved in 500 mL binding buffer, and the counter-selection step was eliminated. To acquire aptamers with high affinity and specificity, the selection strength was enhanced gradually by increasing the number of washes (from three to ten times with 200 mL 16PBS buffer each) and decreasing the amount of the ssDNA library per round (from 200 to 150 pmol). The resulting pool from the 14 th round was PCR amplified, cloned and sequenced (Shanghai Sangon sequencing facility). The resulting 62 sequences were subjected to multiple sequence alignment analysis by using Clustal W 6.0 software to discover highly conserved motifs in groups of selected DNA sequences. The discovered consensus sequences with high repeats among selected pools were then chemically synthesized for further testing. To monitor the enrichment of aptamers after selection, the FAM-labeled ssDNA pool was incubated with 5610 4 AGR2-GST-beads or GST-beads in binding buffer (200 mL) at 37uC for 45 min. Beads were washed three times with 200 mL binding buffer by means of filtration, and suspended in binding buffer (250 mL). The fluorescence intensity of the resulting beads was monitored with a FACSAria cytometer (Becton Dickinson Immuno cytometry systems) by counting 10000 events. The binding affinities of aptamers were determined by incubating AGR2-GST-beads (5610 4 ) with various concentrations of FAMlabeled aptamers (pre-heat-treated) in binding buffer (200 mL) at 37uC for 45 min in the dark. Beads were then washed three times with the binding buffer, then resuspended in binding buffer (250 mL) and subjected to flow cytometry analysis. The FAMlabeled unselected ssDNA library was used as negative control for the nonspecific binding evaluation. All binding experiments were repeated two to four times. The mean fluorescence intensity of target protein labeled by aptamers was used to evaluate binding affinity by subtracting the mean fluorescence intensity of nonspecific binding produced by unselected ssDNA library. The dissociation constants (K d ) of the fluorescent ligands were obtained by fitting the dependence of fluorescence intensity of specific binding on the concentration of the ligands to the equation (1): Y = B max X/(K d +X) using SigmaPlot software. CD measurements were carried out on a Jasco J-810 spectropolarimeter equipped with a programmable temperature-control unit (Julabo HP-4). The concentration of DNA samples were 2 mM. Before the CD spectrum measurement, the DNA samples were annealed by heating to 95uC for 5 min, then rapidly cooled on ice for 10 min, and incubated for another 10 min at room temperature. The spectra from 400 to 200 nm were obtained by using 1 nm slit width and 0.1 nm scanning resolution. Each CD spectrum was an average of 8 scans with the buffer background subtracted. UV absorbance and melting studies were carried out on an Agilent 8453 spectrophotometer equipped with a programmable temperature-control unit (Agilent 89090A). Melting temperatures (T m ) were taken as the temperature of half-dissociation of the quadruplex and were obtained from the maximum of the first derivative dA/dT plots at 295 nm. The heat-treated DNA solutions at several concentrations were introduced into a quartz cuvette and overlaid with a thin layer of silicone oil to prevent evaporation. The optical path length was 1 cm. Absorbance and temperature were recorded every 2uC. Evaluation of AGR2-aMB AGR2-aMB (40 nM) was annealed by heating to 95uC for 5 minutes in 200 mL Tris-HCl buffer (25 mM Tris-HCl, 120 mM NaCl, 0.5 mM KCl, 2 mM MgCl 2 at pH 7.4), and then rapidly cooled on ice for 10 min, and subsequently waiting for another 10 min at room temperature before use. To the solution of AGR2-aMB, various concentrations of AGR2 were added and the resulting solutions were allowed to incubate at 37uC for 30 min. To the mixture, 1 mL of streptavidin beads (about 5610 4 beads) were added and the mixture was allowed to set for 45 min in dark. After washing twice with Tris-HCl buffer, SA beads were filtered to eliminate unbound AGR2-aMB and resuspended in Tris-HCl buffer solution 250 mL before flow cytometry analysis. In selectivity tests, BSA (Tagene Biotechnology Xiamen), trypsin (Dingguo Beijing), thrombin (alfa) and IgG (Dingguo Beijing) were used. Figure S1 The sequences of 62 clones. One subfamily is guanosine-rich sequences (22 clones) , and the other is thymine-rich sequences (40 clones). (TIF) Figure S2 The dissociation constant measurement of C14A, C14B, C14C and C14D against AGR2 GST and GSH. (TIF) Figure S3 Flow cytometry assay to monitor the binding of C14B1 and its five truncated sequences with a) AGR2 (target protein) and b) GST (control protein). (TIF) Figure S4 UV thermal-denaturation experiment of C14B1. Denaturation profiles obtained at 295 nm for the aptamer at three different concentrations (2 mM, 4 mM, 8 mM). The T m (59uC) at 295 nm is independent of oligonucleotide concentration, indicating that the aptamer forms an intramolecular G-quartet. (TIF) Figure S5 The secondary structure of AGR-aMB. Stable hairpin structure is formed by intramolecular hybridization between the SA aptamer sequence and the complementary sequence of C14B 1 . (TIF)

Flt3L Combined with Rapamycin Promotes Cardiac Allograft Tolerance by Inducing Regulatory Dendritic Cells and Allograft Autophagy in Mice

The induction of immune tolerance is still a formidable challenge in organ transplantation. Dendritic cells (DCs) play an important role in orchestrating immune responses by either mediating protective immune responses or inducing antigen specific tolerance. Previous studies demonstrated that the fms-like tyrosine kinase 3 receptor (Flt3) and its ligand (Flt3L) play an essential role in the regulation of DC commitment and development. Here, we report a synergic effect between Flt3L and low-dose rapamycin (Rapa) in the protection of allograft rejction. It was found that Flt3L combined with Rapa significantly prolonged murine cardiac allograft survival time as compared with that of untreated recipients or recipients treated with Rapa or Flt3L alone. Mechanistic studies revealed that Flt3L combined with low-dose of Rapa induced the generation of tolerogenic DCs along with the production of CD25(+) Foxp3(+) regulatory T cells and IL-10 secretion. We also observed enhanced autophagy in the cardiac allograft, which could be another asset contributing to the enhanced allograft survival. All together, these data suggest that Flt3L combined with low-dose of Rapa could be an effective therapeutic approach to induce tolerance in clinical setting of transplantation.

Organ transplantation has become the primary treatment for patients with end-stage organ failure [1] . Although the application of immunosuppressive drugs has contributed significantly to the success of allograft survival, side effects resulted from immunosuppression or drug toxicity also markedly impact the quality of life of recipients [2, 3] . Therefore, establishment of strategies aimed at inducing allograft tolerance is wanting. In the setting of transplantation, DCs actively mediate graft rejection by presenting donor-derived alloantigens to naïve T cells. However, there is emerging evidence indicating that other than mediation of allograft rejection, DCs also possess the capability to induce allograft tolerance [4] . Therefore, DCs are capable of inducing either immune response or tolerance depending on their activation and maturation status [5, 6] . Flt3, a member of the tyrosine-kinase receptor family, was initially cloned from fetal liver of cells with hematopoietic stem cell activity [7] . Flt3L is the ligand for Flt3, which is a key regulator for DC commitment and development. Mice administered with Flt3L displayed the expansion of certain subtypes of DCs such as the plasmacytoid DCs (pDCs) and conventional DCs (cDCs) in the spleen [8] . Studies have shown that pDCs play an important role in the induction of Tregs in vivo manifested by that their precursors are able to prolong graft survival [9] . Studies in hematopoietic cell transplantation also revealed that Flt3L is a potent mobilizer to induce murine hematopoietic chimerism, while long-term persistence of donor hematopoietic cells in peripheral lymphoid and non-lymphoid tissues of organ allograft recipients is postulated to be an essential prerequisite for the induction of donor-specific tolerance [10, 11] . Autophagy in eukaryotic cells is a cellular quality control process to deliver cytoplasmic constituents for lysosomal degradation, which enables cells to recycle nutrients for survival during nutrient starvation [12] . Rapamycin, an inhibitor of the mammalian target of rapamycin (mTOR) has been used to stimulate autophagy. In the steady state, autophagy-mediated antigen processing in thymic epithelial cells could have a crucial role in the induction and maintenance of CD4 + T-cell tolerance [12, 13] . In this study, we examined the effects of Flt3L in combination with Rapamycin in a BALB/c-to-C57BL/6 cardiac allograft model. Our results demonstrate that Flt3L combined with lowdose of Rapamycin prevented acute allograft rejection and prolonged allograft survival time. The enhanced allograft survival is associated with the induction of tolerogenic DCs along with the production of CD25 + Foxp3 + regulatory T cells and increased allograft autophagy. Our data suggest that Flt3L combined with low-dose of Rapa could be a promising therapeutic strategy in clinical transplantation. We first sought to investigate the regulatory effect of Flt3L combined with a short-term of low-dose Rapa treatment on allograft acute rejection. A murine cardiac allograft transplantation model was employed to address this question. As shown in Figure 1A , untreated recipients rapidly rejected their grafts (7.660.516 days) along with typical pathological features of acute rejection (Fig. 1B) . Similar pathological features were observed in recipients treated with either PBS or rGST (controls, Figure 1B ). As expected, administration of Rapa significantly increased graft mean survival time (MST) (16.962 .767 days, Fig. 1A) . Similarly, recipients treated with Flt3L doubled allograft MST (13.161.287 days, Fig. 1A ). In sharp contrast, a significant enhanced graft survival was observed in mice treated with Flt3L and low-dose of Rapa (36.4618.704 days, Fig. 1A) . Remarkably, about 20% of recipients showed long-term graft survival (MST.100 days, Fig. 1A ). In consistent with these results, a significant reduction for inflammatory infiltration was observed in the allograft sections (Fig. 1B ). We next examined DC phenotypes in the recipient mice. It was noted that a significant increase in the number of pDCs in mice treated with Flt3L or Flt3L with low-dose of Rapa as compared with that of control mice ( Fig. 2A) . Similarly, we also detected a significantly higher proportion of splenic CD8a + DCs in Flt3L or Flt3L/Rapa treated mice ( Fig. 2A) . Since both CD8a + DCs and pDCs have regulatory functions [14] , Our data suggest that Flt3Lcombined with Rapa could induce the expansion of regulatory DCs in the recipient mice. Rapamycin has been shown to inhibit the up-regulation of maturation markers expressed by DCs [15] , while immature DCs have been demonstrated with capability to prolong allograft survival [16, 17] . In line with this notion, Flt3L combined with Rapa significantly blocked CD11c + DC maturation as manifested by the decreased CD40, CD80 expression (Fig. 2B ). Since CCR9 + pDCs had previously been reported to be potent inducers for Tregs [18] , we then investigated whether CCR9 + pDCs are implicated in the induction of allograft tolerance. Indeed, recipients treated with Flt3L and low-dose of Rapa displayed a significant increase for CCR9 + pDCs as compared with those recipients treated with Flt3L or Rapa alone or other groups (Fig. 2C ). Flt3L Combined with Rapa Promotes the Production of CD4 + CD25 + Foxp3 + and CD8 + CD25 + Foxp3 + T Cells The above results prompted us to examine the proportion of CD4 + CD25 + Foxp3 + and CD8 + CD25 + Foxp3 + Tregs. To this end, splenic T cells isolated from each group of recipient mice were stained for intracellular Foxp3 after co-staining with anti-CD4/ CD25 antibodies or anti-CD3/CD8/CD25 antibodies, and the cells were next analyzed by flow cytometry. A significant increase of CD4 + CD25 + Foxp3 + and CD8 + CD25 + Foxp3 + Tregs was observed in recipients treated with Flt3L and Rapa as compared with recipients in other groups (Fig. 3 ). We then tested the reactivity of bulk T cells isolated from the spleens of the recipients against donor-derived DCs. As shown in Figure 4A , T cells derived from Flt3L/Rapa treated mice exhibited a significantly lower proliferative response against donor BALB/c DCs in MLR when compared with that of T cells derived from untreated, rGST-treated and PBS-treated recipients. Unexpectedly, a similar proliferation rate was noted in recipients treated with either Flt3L or Rapa alone as compared with that of recipients treated with Flt3L/Rapa. This result promoted us to further examine cytokine production in the culture supernatants by ELISA. Of Interestingly note, the production of IL-10 was significantly higher in the Flt3L/Rapa treated mice. More importantly, although T cells derived from Flt3L or Rapa treated mice showed similar potency for proliferation (Fig. 4A ), but their capacity for secretion of IL-10 was similar to those T cells from control mice (Fig. 4B ). Collectively, our data suggest that Flt3L/ Rapa treatment enhances IL-10 secretion which could favor the induction of T anergy. Given that Flt3L/Rapa therapy resulted in reproducible longterm allograft survival in about 20% of recipients without immunosuppression ( Fig. 1A) , we thus employed adoptive transfer studies to characterize the cell subpopulations for sustained allograft survival. For this purpose, CD4 + T cells, CD8 + T cells, pDCs and total splenocytes were first isolated from long-term survival recipients, and then adoptively transferred into naive recipients. The mice were next transplanted with cardiac allografts as described. As shown in Figure 5 , transferred CD8 + T cells and pDCs exhibited the ability to prolong allograft survival. Unexpectedly, compared with CD8 + T cells and pDCs, adoptive transfer of CD4 + T cells failed to provide protection against allograft rejection. To determine whether the prolonged allograft survival was associated with enhanced autophagy, we examined the levels of autophagy-related protein beclin-1, microtubule associated pro-tein1 light chain 3 (LC3) I and II by Western blot analysis. As shown in Fig. 6A , only low levels of Beclin-1 and LC3 II were detected in control mice (rGST, untreated and PBS groups), but higher levels of Beclin-1 and LC3 II were observed in mice treated with Flt3L or Rapa alone. Importantly, the highest Beclin-1 and LC3 II expressions were noted in Flt3L/Rapa treated mice. To further confirm these results, we did immunohistochemical analysis, and similar results were obtained (Fig. 6A ). To examine the functional relevance for the enhanced autophagy in allograft survival, we performed similar transplantation along with the administration of 3-MA, an autophagy inhibitor [19, 20] . As expected, administration of 3-MA significantly repressed autophagy as determined by Western blot and immunohitochemical analysis (Fig. 6B) . Remarkably, repression of autophagy significantly attenuated Flt3L/Rapa therapy mediated protection against allograft rejection (Fig. 6C ). In consistent with this result, we also observed a decrease for the number of regulatory T cells and regulatory DCs (Fig. 6D) . Taken together, these data indicate that autophagy plays an essential role in Flt3L/Rapa therapy mediated protection against allograft rejection. Despite the benefit of Flt3L and Rapa for anti-rejection in setting of transplantation, the effect for Flt3L or Rapa based mono-therapy on allograft survival remains poor [21, 22, 23, 24] . In the present study, we have demonstrated that Flt3Lcombined with low-dose of Rapa could markedly increase graft survival time as compared with that of Flt3L or Rapa alone. The enhanced protective effect was achieved by promoting the generation of tolerogenic DCs and Tregs along with increased autophagy in the allograft. It has been reported that DCs expanded by Flt3L only express low levels of costimulatory molecules, and therefore, they are able to delay allograft rejection in transplantation settings [25, 26] . In contrast to Flt3L or Rapa alone, our data show that Flt3L combined with low-dose of Rapa are more potent to expand DCs with tolerogenic characteristics. Moreover, short-term of low-dose Rapa exerts the effect to expand Tregs while to avoid side effect due to high-dose of Rapa. Several DC subsets have been reported to promote tolerance induction [4, 27] . The CD11c + CD8a + DC subset could promote peripheral self-tolerance and prolong the survival of cardiac allografts in rodents [28, 29] . More recently, pDC precursors have been identified and are implicated in the induction of allogeneic Tcell hyporesponsiveness [9, 30] . In line with these findings, our phenotypic analysis of splenic DCs in the recipient mice revealed that Flt3L/Rapa treatment increased the proportion of DC subsets with regulatory effect such as CD11c + CD8a + DCs and pDCs. It is known that immature DCs not only fail to prime T cells effectively, but also serve to promote tolerance induction [31] . Our studies show that Flt3L combined with Rapa could block DC maturation and increase the ratio of CCR9 + pDCs, while CCR9 + pDCs are immature pDCs with potency to suppress antigen-specific immune responses both in vitro and in vivo [18] . Regulatory T cells are a subpopulation of T cells that have the ability to suppress immune reactions and maintain tolerance. They were believed to play a key role in transplant tolerance induction. The most important representative of regulatory T cells is the naturally occurring regulatory T cells (Tregs), with a CD4 + CD25 + Foxp3 + phenotype. They were found to mediate donor-specific Figure 2 . DCs from recipients treated with Flt3L/Rapa display a tolerogenic phenotype. Spleen cells isolated from mice treated with Flt3L/ Rapa and other control mice at the time of rejection or at study endpoint (POD 100) were stained with PE-cy7-labeled anti-mouse CD11c mAb, Per-cy5.5-labeled anti-mouse B220 mAb, APC-cy7-labeled anti-mouse CD8a mAb, FITC-labeled anti-mouse CCR9 mAb, APC-labeled anti-mouse CD80 mAb and APC-labeled anti-mouse CD40 mAb. CD11c + and CD11c + B220 + gated DCs were analyzed by flow cytometry for expression of the various DC markers. (A) The bar graph was a summary of percentages of pDC and CD8a + DC in the recipients. (B) CD11c + -gated DCs were analyzed for expression of costimulatory markers CD80 and CD40 to assess their maturation state. (C) The percentage of CCR9 + cell of gated pDC subsets was assessed by FACS analysis. The results are representative of three independently performed experiments (* P,0.05). doi:10.1371/journal.pone.0046230.g002 Figure 3 . Flt3L combined with Rapa promotes the production of CD4 + CD25 + Foxp3 + and CD8 + CD25 + Foxp3 + T cells. Spleen cells isolated from mice treated with Flt3L/Rapa and other control mice at the time of rejection or at study endpoint (POD 100) were analyzed to determine the proportion of CD4 + CD25 + Foxp3 + and CD8 + CD25 + Foxp3 + T cells. (A, B) The expression of Foxp3 by CD4 + CD25 + and CD8 + CD25 + T cells was analyzed by flow cytometry after intracellular staining. The bar graph was a summary of percentages of CD4 + CD25 + Foxp3 + T cells and CD8 + CD25 + Foxp3 + T cells in the recipients. The data shown are representative of three independent experiments that yielded comparable results (* P,0.05). doi:10.1371/journal.pone.0046230.g003 unresponsiveness in some stable renal [32, 33, 34, 35, 36] and liver [37] transplant patients. In addition, adoptive transfers of Tregs could prevent allograft rejection and prolong graft survival in the mouse models [38] . Previously, addition of rapamycin to human and murine T cell cultures preferentially expanded regulatory T cells [39, 40] . Our data indicated that infusion of Flt3L with Rapa and Rapa alone enhances the production of CD4 + CD25 + Foxp3 + Tregs. In addition, we found a significant increase of splenic CD8 + CD25 + Foxp3 + T cell populations in the recipient mice treated with Flt3L/Rapa, while CD8 + CD25 + Foxp3 + T cell have been found being able to inhibit antigen-specific CD4 + and CD8 + T cell responses by cell contact inhibition, secretion of cytokines and other ways [41] . Our adoptive transfer studies showed that CD8 + T cells are responsible for induction of long-lasting tolerance toward donor alloantigen. We therefore proposed that Flt3L-mobilized recipient DCs might process indirect antigen presentation of allopeptides to induce tolerance for host CD8 + alloreactive T cells, which are the main effector mechanism of cellular rejection across an MHC class I barrier. This notion is well supported by a recent study, in which Rapa-conditioned, alloAg-pulsed DC can present acquired MHCpeptide complexes from donor cells and modulate directly-reactive CD8 + T cell function, and through which, these DCs enhance Agspecific CD8 + T cell undergoing apoptosis [42] . Furthermore, CD8 + CD28 -T cells may also play an important role in tolerance induction. These T cells represent a transient state of effector CD8 + T cells, which promotes the production of the immunosuppressive cytokine IL-10 [43, 44] . Emerging data suggest that the appearance of these T cells in transplant recipients is associated with better graft acceptance along with stable function [45, 46, 47] . Finally, our data indicate that autophagy plays an important role in the protective effect of Flt3L in combination with Rapa on cardiac allograft survival. The possible mechanisms is that autophagy delivers self-proteins for MHC class II loading, and their peptidic fragments are essential for the deletion of selfreactive T cells in the thymus. Recent studies have also suggested that the possible link between autophagy and tolerance is that autophagy plays a role in the removal of apoptotic cell corpses [48] . Regarding to the effect of autophagy on allograft survival seems controversial. In a rat liver transplantation study, it has been shown that autophagy-associated hepatocyte death triggers liver graft dysfunction, and suppression of autophagy prevent cold ischemia-warm reperfusion injury associated with liver transplantation [49] . Thus, further relevant investigations are needed. In summary, Flt3L combined with Rapamycin is able to promote the prolongation of allograft survival. This protective effect is associated with the generation of tolerogenic DC subsets, CD25 + Foxp3 + regulatory T cells (Tregs), and enhanced allograft autophagy. Six to eight week-old female C57BL/6 (H-2 b ), BALB/c (H-2 d ) mice were obtained from the animal facilities at Tongji Medical College. All mice were maintained under specific pathogen-free conditions and the studies were carried out in compliance with the institutional animal care and use guidelines. DNA sequence encoding Flt3L (696 bp) was amplified from the C57BL/6 mouse spleen cDNA using the following oligonucleotides: 59-CGG GAT CCA TGA CAG TGC TGG CGC C-39 and 59-CGG AAT TCA TCC TAG GGA TGG GAG-39. The resulting products were subsequently cloned into a pGEX-4T3 vector with a glutathione S-transferase tag. The plasmid was transformed into Escherichia coli strain BL21. Recombinant Flt3L expression was induced by addition of 1 mM IPTG and then purified using the glutathione sepharose 4B resin columns as instructed by the manufacturer. A GST vector protein was also expressed and purified to homogeneity. The proteins were passed over polymyxin B columns (PIERCE) to remove any contaminated LPS and further concentrated by 3KD Micropore Filters. The purified recombinant Flt3L and rGST were stored in aliquots at 280uC until use. Heterotopic cardiac transplantation was performed as previously described [50] . In this study, transplant surgery involved the transfer of fully MHC-mismatched hearts from BALB/c (H-2 d ) donors to C57BL/6 (H-2 b ) recipients. Heart beat of the grafts was monitored and evaluated daily by direct abdominal palpation to detect the state of cardiac health/rejection. Heart transplant recipients were randomly assigned to six groups (n = 10): Group 1, rGST alone; Group 2, untreated control; Group 3, PBS alone; Group 4, Flt3L alone (10 mg/day, i.v) for 10 consecutive days before heart transplantation; Group 5, Rapa alone (2 mg/kg/day; p.o, POD0-15); Group 6, combination treatment of Flt3L and Rapa. Allograft samples were fixed in 4% paraformaldehyde. Serial sections (5 mm in thickness) were prepared using a microtome and stained with hematoxylin/eosin for the analysis of pathological changes. Spleen cells isolated from mice treated with Flt3L/Rapa and other control mice were incubated with anti-CD16/CD32 monoclonal antibody to block FcR, and then stained with PE-cy7-labeled anti-mouse CD11c mAb, Per-cy5.5-labeled antimouse B220 mAb, APC-cy7-labeled anti-CD8a mAb, FITClabeled anti-mouse CCR9 mAb, APC-labeled anti-mouse CD80 mAb and APC-labeled anti-mouse CD40 mAb. CD11c + and CD11c + B220 + gated DCs were analyzed by flow cytometry for expression of the various DC markers. For analysis of regulatory T cells, splenocytes were isolated from the recipient mice and subsequently co-stained with CD4, CD25 or CD3, CD8a CD25 antibodies. The cells were further stained with Foxp3 using established techniques [51] . Cells were then analyzed using FACS flow cytometer as previously reported [52] . Splenic T cells (purified using nylon wool columns) isolated from naive or grafted C57BL/6 mice serving as responders (2610 6 / well) were incubated for 7 days in the presence of Mitomycin Ctreated (50 mg/ml, Sigma-Aldrich) DCs used as stimulators derived from naive BALB/c mice (4610 5 /well, DC: T ratio of 1:5) in 12-well flat bottom plates (Nunc, Roskilde, Denmark). Cultures were kept at 37uC in 5% CO 2 atmosphere. Supernatants were recovered after 7 days for determination of IL-10 production. Cytokine level was quantified using Mouse IL-10 ELISA MAX TM Deluxe. T-cell proliferation was assessed by serial dilution of the intracellular CFSE dye as described previously [53] . Proteins (50 mg) extracted from cardiac allograft were subjected to electrophoresis on a 12% SDS-PAGE and then transferred onto PVDF membranes. The membranes were then incubated with primary antibodies for Beclin-1, LC3B and GAPDH at 37uC for 2 h, respectively. The blots were visualized using enhanced chemiluminescence (Amersham). After deparaffin and rehydration, the paraffin-embedded heart sections (5 mm) were treated with 3% H 2 O 2 for 5 min. The nonspecific proteins were blocked with 5-10% goat serum for 10 min. For Beclin-1 staining, specimens were incubated with a rabbit antimouse Bcelin-1 Monoclonal antibody (1:200) at 4uC overnight, followed by incubation with a HRP conjugated goat anti-rabbit secondary antibody. The sections were finally incubated with DAB chromogenic substrate and counterstained with hematoxylin. Total T cells were first enriched with T cells using nylon wool columns, and then subjected to isolation of mouse CD4 + T cells using the (L3T4) MicroBeads, and CD8 + T cells using the mouse CD8a (Ly-2) MicroBeads as instructed (Miltenyi Biotec, CA)., pDCs were isolated using a mouse Plasmacytoid Dendritic Cell Isolation Kit II as instructed (Miltenyi Biotec, CA). The purified CD4 + T cells (5610 6 ), CD8 + T cells (5610 6 ), pDCs (5610 5 ) and total splenocytes (5610 6 ) originated from recipients with long-term survival allograft were then adoptively transferred into naive recipients, respectively. One day after the adoptive transfer, the mice were transplanted with cardiac allografts and monitored for allograft survival as described above. The data are presented as means 6 SD. Statistical differences were determined by one-way ANOVA. Allograft survival differences between groups were determined using the log-rank test. P values less than 0.05 were considered with statistical significant.

Renalase's Expression and Distribution in Renal Tissue and Cells

To study renalase's expression and distribution in renal tissues and cells, renalase coded DNA vaccine was constructed, and anti-renalase monoclonal antibodies were produced using DNA immunization and hybridoma technique, followed by further investigation with immunological testing and western blotting to detect the expression and distribution of renalase among the renal tissue and cells. Anti-renalase monoclonal antibodies were successfully prepared by using DNA immunization technique. Further studies with anti-renalase monoclonal antibody showed that renalase expressed in glomeruli, tubule, mesangial cells, podocytes, renal tubule epithelial cells and its cells supernatant. Renalase is wildly expressed in kidney, including glomeruli, tubule, mesangial cells, podocytes and tubule epithelial cells, and may be secreted by tubule epithelial cells primarily.

Renalase is a newly discovered monoamine oxidase enzyme originating from renal tissues [1] . It degrades circulating catecholamines, regulates blood pressure and cardiac function, and is closely associated with cardiovascular diseases and chronic kidney disease (CKD) [2] [3] [4] . Renalase is a protein made of 342 amino acids with a molecular weight of 37.8 KDa approximately. The N-terminate of renalase contains one signal peptide, one flavin adenine dinucleotide binding site and one monoamine oxidase domain, and 13.2% of its amino acid sequence is similar to the monoamine oxidase A. Obtaining recombinant renalase protein and preparation of monoclonal antibodies are the essential steps for the study of renalase's function, expression and distribution in renal tissues and cells. Recently we have been using recombinant renalase protein produced by prokaryotic expression system to develop monoclonal antibodies, and we also have tried to use recombinant protein produced by eukaryotic expression systems such as baculovirus etc, to develop monoclonal antibodies. However, obtaining large volume of eukaryotic expressed recombinant renalase protein has not been such an easy task [5, 6] . DNA immunization technique is a vaccination method, which is a fast and effective way of stimulating body to generate immune reaction against target protein [7] [8] . After DNA vaccine's uptake and processing by muscle cells, the natural structure of target protein remains well preserved. In recent years, DNA vaccine techniques have been used to obtain monoclonal antibodies, especially when it is difficult to get protein antigens. Recently we have successfully utilized such techniques to prepare monoclonal antibodies for retinol binding protein (RBP4) [9] , and established immunological testing methods as well as explored renalase related DNA vaccine techniques [10] . On this basis, we utilized DNA immunization technique to prepare anti-renalase monoclonal antibodies and those antibodies were used to analyze renalase expression in renal tissues and cells. Primer for renalase gene was designed as Sp 59-ATAA-GAATGCGGCCGCATGGCGCAGGT GCTGATC -39, As 59-GGAAGATCTCTAAATATAATTCTTTAAAGCTT -39, and then renalase gene was replicated using RT-PCR technique. Renalase gene was then inserted in pBudCE4.1 (Invitrogen, Carlsbad, CA, USA) vector plasmid, after which the immunization plasmid was completed. Renalase protein was obtained using prokaryotic expression system. After being purified and measured for concentration, it then was stored at 220uC [5] . HEK293T cells purchased from ATCC (ATCC No. CRL-11268 TM , USA) were cultured in Dulbecco's Modified Eagle Medium (Invitrogen, USA) containing 10% fetal bovine serum. Plasmids were transfected into HEK 293T cells using Lipofectamine 2000 Reagent (Invitrogen, USA), and this process was carried out according to the manual guidelines. After 48 hours, culture media was removed and rinsed once with PBS (pH 7.4), and then dissolved (0.2 ml/hole). After the collection, the sample was heated for 5 min at 100uC, and then stored at 220uC. Six weeks old BALB/c mice (Shanghai Experimental Animal Center of Chinese Academy of Sciences) were kept in SPF grade animal room. Animals were used according to the feeding and utilization guidelines prescribed by Chinese Academy of Science. Animal immunization was done as per the recommended methods in the literature [11] : 100 mg plasmids (100 ml sterile PBS, pH 7.4) was given intramuscularly in the quadriceps. Immediately after the injection, electric impulse stimulation was given with ECM830 (BTX, Holliston, USA). Stimulation parameters were square wave, 100 V/50 ms, positive and negative impulses given three times each, with an interval of 1 second. A total of three injections were given with an interval of 3 weeks. Three weeks after the last injection, cell fusion was carried out. A shot of intra-abdominal booster of recombinant renalase protein was given three days before the cell fusion. 10 days after the third injection, blood was withdrawn from the orbit, kept for 1 hour at the temperature of 37uC, followed by 24 hours at 4uC, and then centrifuged. Upper layer of colorless or pale yellow serum was collected and stored at 220uC for further use. The classic method of monoclonal antibody preparation was used [12, 13] . In brief, selected one mouse that had high sensitive activation to renalase protein, harvested the spleen, then extracted spleen cells and myeloma cell line SP2/0, and then performed cell fusion with 50%(w/v)polyethylene glycol in a 50 ml centrifuge tube (Corning, New York, USA). Following the cell fusion, cells were subjected to centrifugation and transferred to 96-well plate, and then with HAT added, cell selection was performed. Ten days after the cell-fusion, cells selection was performed with selecting ELISA positive holes. Anti-renalase monoclonal antibodies were used as primary antibodies and SABC testing kit (Boster, Wuhan, China) was used to detect renalase's expression in renal tissues. Each testing step was performed according to the testing manual guidelines. The renal tissues subjected to testing were from a post-traumatic kidney. With self-prepared anti-renalase antibodies used as primary antibodies and HRP labeled goat-anti-mouse IgG (Shanghai Immune Biotech, China) used as secondary antibodies, western blotting was carried out to detect renalase's expression in After the renal tissues were obtained, 3 mm thickness frozen sections or renal cell climbing slices were made and fixed with acetone, and self-prepared anti-renalase monoclonal antibodies were added as primary antibodies. Then, FITC labeled goat-antimouse IgG (Santa Cruz, USA) were added as secondary antibody, slide sealed with glycerol and observed under fluorescence microscope. Renalase coded gene was inserted into the expression vector pBudCE4.1 (contains one His Tag gene after the sequence of multiple cloning sites), and the correct sequencing was verified. After transfection in vitro expression of pBudCE4.1-Renalase was verified by western blotting demonstrated in Fig. 1 , renalase was detected by anti-His Tag monoclonal antibody. No response was observed in empty vector, indicating that renalase can be expressed in mammalian cells. The titer of antibody from mice serum was tested before the cell fusion. One mouse with high titer was selected as spleen cells donor and cell fusion was done with cell line SP2/0. The serum titer of anti-renalase was 1:32000 and the titer of normal mouse was below 1:1000. After 3 cycles of cloning, two hybridoma cell lines were obtained. Culture supernatant was collected and ascites was prepared. Analysis of culture supernatant showed an antibody titer of 1:128000, and an antibody sub type IgG1/k. Titration ELISA of purified anti-renalase monoclonal antibody was over 1:512000, higher than that of poly-serum titer 1:128000. Western blotting was performed by using prokaryotic expressed renalase protein as the sample and the prepared anti-renalase monoclonal antibodies as primary antibodies. The results showed the recombinant renalase protein can be recognized by the monoclonal antibody (Fig. 2) . Renalase's expression in renal tissues can be detected by immunofluorescence Indirect immunofluorescence testing was conducted by using anti-renalase monoclonal antibody. The results showed that renalase expressed in cytoplasm of glomerular mesangial cells (Fig. 3A) . And renalase can also be detected for its expression in cytoplasm of tubule epithelial cells (Fig. 3C ). Immunohistochemical testing was carried out with renalase monoclonal antibodies as the primary antibodies. The results showed that renalase expressed in renal glomeruli (Fig. 4A) and tubule (Fig. 4C ). Indirect immunofluorescence testing was conducted with antirenalase monoclonal antibodies as the primary antibodies. The results showed that renalase expressed in cytoplasm of mesangial cells (Fig. 5A), podocytes (Fig. 5C ) and renal tubular epithelial cells (Fig. 5E ). Western blotting was done with anti-renalase monoclonal antibodies as the primary antibodies. The results showed that mesangial cells (Fig. 6A) , podocytes (Fig. 6C ) and renal tubular epithelial cells (Fig. 6E ) cultured in vitro also could express renalase. But only tubular epithelial cells' supernatant could be detected the expression of renalase, as shown in Fig. 6F . Recent studies showed that elevated levels of catecholamine in renal failure patients compared to normal individuals may be associated with hypertension and cardiovascular complications of CKD. But the exact mechanism of it is still unclear [14] . Renalase was discovered in 2005, it is an enzyme expressed in kidneys that can degrade circulating catecholamines. This finding has changed the previous understanding about renal physiology and neurophysiology [15] [16] [17] . In patients with CKD, renalase production in serum and tissues decreases. Current studies indicate that the decreased renalase production is associated with hypertension, as well as increased levels of circulating catecholamines [18] . In 2007, renalase-like substance was successfully cloned from mice by Wang [19] . Renalase is a highly conserved protein, of which 95% of amino acids are similar to that of primates. This is probably because renalase coded genes have evolved from a same ancestor gene [4] . Studies have shown that kidneys are the main source of renalase production. In vitro study has shown that renalase can degrade catecholamines and has strongest enzymatic hydrolysis action against dopamine, followed by adrenaline and noradrenaline. In vivo studies indicate that catecholamine activates renalase precursor which promotes deactivation of catecholamines and regulation of cardiovascular function [20, 21] . As of today, renalase is the only enzyme known to be secreted into blood that can degrade circulating catecholamine and may have a great value in prevention of kidney diseases and cardiovascular diseases. Discovery of renalase has drawn the attention of the scientific world [22] [23] [24] . Even though there have been some debates (Boomsma F), much importance is given to its significance as further studies are carried out [25] . It may have a great value in further understanding increased sympathetic activity and mechanism of cardiovascular complications in CKD patients as well as a prevention of CKD and cardiovascular diseases [26] [27] [28] . Both in vitro and animal studies have demonstrated the possible effects of renalase in chronic renal failure and cardiovascular diseases [29] . It can be speculated that replacement or supplementation of renalase may bring up a promising treatment. At present, there are many ongoing studies on genetic background of renalase. Zhao et al have found that renalase gene polymorphysim is associated with primary hypertension, indicat-ing this gene may become the novel marker of genetic susceptibility in essential hypertension [30] . Farzaneh-Far discovered that renalase gene polymorphism also is associated with ventricular hypertrophy [31]. Buraczynska found that an association between renalase gene polymorphism and hypertension in type 2 diabetes [32] . Stec et al investigated patients with end stage renal disease and found that renalase gene polymorphism is associated with hypertension among these patients [33] . Despite the role of renalase in regulating serum catecholamines and blood pressure, recent studies demonstrated that extra-renal renalase is also important. In renalase gene knocked-out mice, serum urea nitrogen, creatinine and aldosterone levels were unaffected, and cardiac systolic function was intact. However, mild ventricular hypertrophy was presented with decreased tolerance to ischemia, and increased risk of myocardial infarction, as much as 3 times higher than that of wild mice. Recombinant renalase replacement therapy could abort such abnormal changes [34] . Gu et al recently reported that renalase's expression is influenced by renal perfusion and abnormality of the renal perfusion is probably implicated in elevated serum catecholamine levels in cardiac failure [35] . Both renalase gene and protein are also observed in central and peripheral nerve systems [26] . Renalase is found in hypothalamus, pons, medulla oblongata and spinal cord, where the presympathetic and preganglionic neurons are located. It is highly plausible that renalase is also involved in regulating the central sympathetic output. Other extra-renal sources of renalase (i.e. skeletal muscles, small intestine and vasculature) are yet to be determined. The regulation of renalase's expression is another enigma that requires further studies. As of today, there have been some animal experiments that showed decreased renalase production associated with high intake of sodium and phosphate [36] . Renal denervation, a potential anti-hypertension treatment [37] , was reported to increase plasma renalase content and renalase expression in the kidneys in spontaneously hypertensive rats [38] . Data from clinical studies has shown that renalase level significantly decrease in patients with stroke and chronic kidney diseases, especially those on renal replacement therapy [39] [40] [41] , and is normalized in kidney and heart transplant recipients [42, 43] . However what parts of kidney and which renal cell renalase is expressed in are not clear yet. In order to obtain high sensitive anti-renalase antibody, this study used DNA immunization which is an established laboratory technique that has been done successfully. Study was conducted using self-made anti-renalase monoclonal antibody, and findings were consistent with the literature, that is, renalase was expressed in both glomeruli and renal tubule [1] . Further in vitro testing showed renalase expressed in mesangial cells, podocytes and tubular epithelial cells, especially only tubular epithelial cells can secrete renalase to the supernatant. Therefore, tubular epithelial cells may be the major cells that can secrete renalase. Such observation has not been reported. These study findings indicate that anti-renalase monoclonal antobodies used in this study could identify natural renalase protein and it is a good experiment tool. Renalase is widely expressed in glomeruli, tubules, mesangial cells, podocytes and tubular epithelial cells, but its physiological function is not clear yet. Further studies about this novel protein's molecular structure as well as function, carry a great value. At present, a lot more works need to be done on renalase's expression, regulation, function and replacement therapy etc.

Virus contaminations of cell cultures – A biotechnological view

In contrast to contamination by microbes and mycoplasma, which can be relatively easily detected, viral contamination present a serious threat because of the difficulty in detecting some viruses and the lack of effective methods of treating infected cell cultures. While some viruses are capable of causing morphological changes to infected cells (e.g. cytopathic effect)which are detectable by microscopy some viral contaminations result in the integration of the viral genome as provirus, this causes no visual evidence, by means of modification of the cellular morphology. Virus production from such cell lines, are potentially dangerous for other cell cultures (in research labs)by cross contaminations, or for operators and patients (in the case of the production of injectable biologicals) because of potential infection. The only way to keep cell cultures for research, development, and the biotech industry virus-free is the prevention of such contaminations. Cell cultures can become contaminated by the following means: firstly, they may already be contaminated as primary cultures (because the source of the cells was already infected), secondly, they were contaminated due to the use of contaminated raw materials, or thirdly, they were contaminated via an animal passage. This overview describes the problems and risks associated with viral contaminations in animal cell culture, describes the origins of these contaminations as well as the most important virsuses associated with viral contaminations in cell culture. In addition, ways to prevent viral contaminations as well as measures undertaken to avoid and assess risks for viral contaminations as performed in the biotech industry are briefly described.

Since the development of viral vaccines, animal cell technology has been used for the production of biologicals for prophylaxis and therapy of humans and animals. As many of these products are injectables, the microbiological safety is of particular importance and is a permanent concern. Whereas microbial contaminations (fungi, yeasts, and bacteria) can be rather easily detected via cultivation methods, the detection of mycoplasma and viruses is more difficult because they are not observable by routine light microscopy. However, fluorescence microscopy, mycoplasma amplification and culture techniques, ELISA and PCR are well developed for determining the extent of mycoplasma contamination. Viral contamination, on the other hand, represents a greater concern because viruses require more complex and frequently need highly sophisticated detection methods (see later). In addition the potential for viruses to cause silent infection of cell culture needs to be addressed, negative results do not always signify that there is no virus contamination. Viral infection can originate from contaminated cell lines, contaminated raw materials, or from a GMP breakdown in the production and purification process (Minor, 1996) leading to a virus-contaminated final product. In addition, although downstream processing is able to eliminate or inactivate certain viruses, not all viruses can be eliminated in such a way because they can be resistant to elimination and/or inactivation steps. Viral infection can be highly pathogenic, and in contrast to microbial infection, there are frequently no effective treatments available, this requires serious consideration to be given to the prevention of contamination. As a result of this everything possible has to be done in order to maintain the entire manufacturing process, and thus the final product, virus-free. In this review, the problem of viral contaminations in animal cell culture will be presented with special emphasis on animal cell technology used for the production of biologicals for prophylaxis and therapy. In addition, this article will suggest actions which can be taken, in order to assure the absence of viral contaminations. These may include the use of production media devoid of animal derived substances, validation of viral clearance in downstream processing or analytics for detecting adventitious viruses in cell culture and final biological product. At the end of this article, the implications for the more basic research laboratory will be discussed. Viral contamination of cultured cells is associated with several problems: -In contrast to bacterial and fungal contamination, viral contamination cannnot be easily detected, because they cannot be observed by normal light microscopy. It is only when a viral contamination leads to a morphological modification of the cultured cells, such as a cytopathic effect, that a contamination by a virus can be suspected. Silent infection by viruses with no observable morphological modification of the infected cell are clearly of greater concern. -A restricted number of viruses can infect cells and can integrate as a provirus, as in the case of Adeno-Associated virus (AAV), for instance. In this case, the provirus is present, but cannot replicate without the assistance of a helper virus, when one is present both virus species will be produced, together (Mayor and Melnick, 1966; Mayor et al., 1967; Hoggan et al., 1972; Berns et al., 1975) . -Virus contaminated cell cultures represent a risk for the operators, collaborators, patients, as well as for non-contaminated cell cultures. -Cells contaminated with some viruses can show a change to their susceptibility to infection by other viruses. For example some safety testing proto-cols require indicator cells to be used to show the presence of virus, if these are chronically infected by viruses this reduces their susceptibility to other virus species, this in turn, can lead to false negative results, because the virus to be detected can no longer infect the indicator cells. -As a general statement, cell lines contaminated by viruses cannot be treated to become virus free. The result of this is that potentially valuable cell lines will have to be discarded and replaced by new, non-contaminated cells. One of the few exceptions to this rule is the case of Lactate Dehydrogenase virus (LDV). Cultures of cells recovered from a passage in infected animals contain this virus; however, as this virus cannot infect the cells, it will be diluted during subsequent in vitro passaging and thus will be lost (Nicklas et al., 1993; Nakai et al., 2000) . In order to avoid viral contamination or reduce the incidence it is helpful to know the source of the contamination. Viruses can be introduced by a limited number of different routes and knowing this provides the possibility to avoid infection. The identified routes of infection are: (i) the cells used to produce the production cells are already contaminated by exogenous virus because the cells were already contaminated at source, e.g., the donor animal from which the cells were explanted (see 'Contamination via the cell source'), the pre-culture (in vitro), or the in vivo passage in an animal led to the virus contamination (see 'Passages via virus infected animals'). (ii) Endogenous viruses, such as retroviruses are a particular concern. Several cell lines of biotechnological importance, such as murine hybridomas or Chinese Hamster Ovary (CHO) cells, contain endogenous retroviruses and can produce retroviral particles during production. In the case of murine retroviruses these can be capable of replication, as observed with hybridoma cells, or which are incapable of replication, as in the case of CHO-cells (see 'Cell lines of biotechnological interest-endogenous retroviruses and other cell associated/latent viruses'). (iii) The cell cultures can be contaminated by viruses which were present in the animal derived materials used in the manipulation or for the growth of the cells. These types of materials include serum or trypsin (see 'Use of contaminated raw materials'). Animal derived raw materials are of particular concern as many different animal viruses can potentially be present originating from the use of infected source animals. Non-animal derived raw materials can also be contaminated by viruses due to contact with virus shed from animals or man from production until its eventual use in the medium of which they are a component. (iv) Finally, errors made by the operator can also result in viral contamination of animal cells (see 'Handling errors of the operator'). In the following paragraph, these issues will be described in more detail. Primary cells derived from explants or continuous cell lines immortalised/transformed by viruses can be contaminated by adventitious viruses. Primary cell cultures derived from animal tissues are seldom used for the production of vaccines for humans but are more frequently applied for veterinary use. When the donor animals are already latently infected with viruses the subsequent in vitro cultures of cells derived from these animals may be infected. If such primary cells are then used for the production of viral vaccines, these vaccines are likely to be contaminated with the adventitious virus. In the period between 1954 and 1961, when primary kidney cells from macaque or rhesus monkeys were used for the production of poliovirus vaccine, this vaccine was frequently contaminated by SV40. The source of this virus was from the kidney cells of infected monkeys (Sweet and Hilleman, 1960; Shah and Nathanson, 1976 ) (for details, see 'Passages via viruses infected animals'). Young immunocompetent rhesus or macaque monkeys can readily be infected with SV40 by the oral, intranasal, and subcutaneous routes, and viremia and viruria occur in these animals (Shah et al., 1969) . The use of such animals as source for kidney cells may lead automatically to a SV40 contaminated primary kidney cell culture because the virus may persist in the kidneys in a latent form (Shah et al., 1969) . Similar observations were made with secondary lamb kidney cells, which are widely used for the production of veterinary vaccines. In the case described, an attenuated Aujeczky's Disease viral vaccine was produced on lamb kidney cells. The master virus stock used for the production had become contaminated with the Border Disease virus due to the contamination of the cell culture used for its production. The vaccination of the sows with this vaccine during the first third of their pregnancy led to the infection of the fetuses which led to a disease similar to classical swine fever (Vannier et al., 1988) . Leiter et al. (1978) reported on the establishment of a mouse epithelial pancreatic cell line which was persistently infected by a polyoma virus. The origin of this infection was not completely clear, but it seemed probable that the mouse which was the source of these cells was also the source of the virus. Finally, all cell lines which have been established by using a virus transformation (e.g. EBV-transformed B-lymphocytes, such as the Namalva cell line (Klein et al., 1972; Butler, 1991) ) are potentially able to produce the virus used for the transformation and therefore also represent a potential infection risk for the operators, the cell culture lab, and the patients receiving a biological produced with such cell lines. The in vitro production of viral vaccines began with the demonstration that explanted embryonic tissue could be used for the production of poliovirus (Enders et al., 1949) . Subsequently primary cells, in particular primary monkey kidney cells, were used for the production of this virus. Although the use of such cells was very convenient and the first accepted way to produce a viral vaccine, the source of these cells (the primary monkey kidney cells) was associated with a number of problems particularly with the introduction of a number of viral contaminats. The advantage of the use of primary cells for the vaccine production system was that it was the first in vitro system for the production of viral vaccines, that high titers of viruses could be obtained, and that the system could be scaled up to a reactor stage by using microcarriers . However, it has been recognised that, this system could be prone to contamination originating from the donor animals. The frequency of contaminants may have been due initially because, the animals were caught in the wild with no control of the disease risks with these animals, leading to a high incidence of contaminations by adventitious viruses. Stones (1977) reported that 40-80% of the cultures of kidney cells from Vervet monkeys were positive for adventitious agents. As mentioned, many of the viruses which infect primates are pathogenic for humans, the species barrier is being able to be crossed (Eloit, 1997) . The notion of the species barrier is sometimes seen as the ultimate rampart that will protect humans against animal viruses, and this can be true for a number of animal viruses which are not able to intitate infection in human cells. However, the species barrier relies on a number of inate features of the immune system which are bypassed in the case of medicinal products (injectables). In addition, the barrier may only be a quantitative issue (quantity of active virus, e.g. the LD 50 of rabies virus in mice is 10 6 times higher by the oral than by the intracerebral route) and is a question of the route of administration (e.g. the mucous membrane is a rather efficient barrier, however, when a virus is inoculated parenterally, the species barrier is no longer valid). Finally, even in the case of an abortive replication cycle, cell transformation via the expression of early viral genes is still possible, eventually leading to a transformed phenotype of non-permissive cells (e.g., mouse cells are transformed by SV40, which in monkey cells leads only to a lytic cycle, or non-permissive cells transformed by bovine polyomavirus) (Eloit, 1997 (Eloit, , 1999 . Zoonotic infection where humans are infected by animal pathogens is frequently observed in nature. Only two examples are described here: Minor (1996) reported on 28 cases of infections occuring between 1932 and 1987 in which individuals who had close contacts with macaque monkeys infected with a highly pathogenic virus (Herpes simiae or B virus) fell ill. This virus is latent in these monkeys, but causes fatal disease in man. Twenty cases were fatal and seven had severe sequelae. The second example concerns the contamination by the Marburg virus. Smith et al. (1967) reported that the contamination of monkeys with this filovirus led to an infection of 31 operators/monkey handlers of which 7 died. These monkeys had been imported into Germany for vaccine production by using their kidney cells. More details can be found in Peters et al. (1992) . A further relevant example from the production of poliovirus vaccine is provided by SV40, a polyomavirus which is an extremely common infection of macaques and rhesus monkeys (Shah and Nathanson, 1976) where it persists in the kidneys in a latent form without causing a cytopathic effect (Sweet and Hilleman, 1960) . This virus grows much more slowly than poliovirus and thus an infection might not be observed during the vaccine production. It is estimated that almost everybody who was vaccinated against poliovirus between 1954 and 1961 also recieved viable SV40 together with the poliovirus vaccine. This is true for the live attenuated as well as for the inactivated poliovaccine, because, in the first case, no inactivation step was applied and, in the second case, the formalin inactivation step was insufficient to inactivate SV40 (Sweet and Hilleman, 1960) . Whereas in the case of the live vaccine, the route by mouth is a poor route for infection with SV40 (Sweet and Hilleman, 1960) , others were injected with inactivated poliovaccine together with infectious SV40. A long-term follow-up study with a small number of individuals, as well as the observation that, in spite of the large number of vaccinees (10-30 millions of 98 millions who were vaccinated, or almost 90% of individuals under 20 yr in 1961 (Shah and Nathanson, 1976) ) which are believed to have received infectious SV40, showed no corresponding increase in related cancers (Shah and Nathanson, 1976) . It should be mentioned, however, that (i) DNA-sequences of SV40 have been detected in association with different human tumors and at an higher incidence in mesotheliomas (Horaud, 1997) , (ii) that SV40 isolated from primary monkey kidney cells by Sweet and Hilleman (1960) induced sarcomas in newborn hamsters (Eddy et al., 1961) , and (iii) that SV40 is oncogenic for laboratory rodents (Magrath, 1991) . In 1970, Hoggan reported the detection of latent infections of AAV in human and monkey kidney cells. He and his coworkers screened cell lines intended for vaccine production and found that approximately 1% of human embryonic kidney cells and 20% of African green monkey kidney cells produced AAV when infected with helper adenovirus. These results suggested that AAV stayed in an integrated form in the absence of helper virus and that this inapparent infection was a rather frequent natural occurrence. Hoggan et al. (1972) could prove by using Detroit 6 cells that a deliberate infection with AAV in the absence of helper virus led to a latent infection. A superinfection with helper adenovirus led to the induction of AAV replication. It is evident that steps can be taken to reduce the risk of contamination risk in primary cell cultures. In practice, different actions are possible: use of virus free animals, use of kidney cells from animals which are less susceptible to virus infections, establishment of veterinary examination and quarantine of animals intended for use, and/or use of diploid or continuous cell lines (use of the cell bank concept/seed stock concept); all reduce the frequency of contamination, Use of captive-bred monkeys versus use of imported wild-caught animals. Besides the advantage of independence from the diminishing wild populations and the sparing of these, it is evident that breeding under controlled conditions led to the production of better quality animals especially in respect of virus infections (Van Steenis et al., 1980) . Van Wezel et al. (1978) could show, by performing serological tests on 18 captive-bred cynomolgous monkeys and 40 imported wild caught parent animals that most of the wild caught animals were positive for antibodies against herpes simplex B, parainfluenza 3, or measles virus, whereas two thirds of the captive bred animals were only positive for antibodies against rotaviruses. Twenty out of 36 imported animals were positive for foamy virus 1 antibodies whereas these antibodies were not observed in the animals bred in captivity (Table I) . During production, control cultures are generally established in parallel to the cultures used for the production of polio vaccine. Van Steenis et al. (1980) compared the frequency of contaminated primary, secondary, tertiary, and quaternary cultures (flasks as well as microcarrier cultures) and stated that all cultures derived from captive-bred animals were free of virusinduced cytopathic effects or hemadsorption, whereas 30 out of 45 kidney cultures from single imported wild caught animals and all of 71 cultures derived from multiple animals showed cytopathology (Table II) . Reduction of the frequency of contaminations. Statistically, the use of fewer animals (kidneys) will increase the probability to establish virus free cultures. As primary kidney cells from monkeys can be amplified to reactor scale cultures by using microcarriers, the number of animals (kidneys) per batch could be considerably reduced (Van Wezel et al., 1978 , 1980 . The calculations done by Van Wezel et al. (1978) in- dicated very clearly that this approach could reduce the number of animals necessary for production purposes of Polio virus vaccine by 5-7 fold. Shah and Nathanson (1976) calculated the probability to obtain kidney cultures free from SV40 with respect to the number of animals used per production batch. By using one animal the frequency of SV40 contamination was about 20%. However, the frequency increased to 70% when the kidneys of two to three animals were pooled and was 100% when the kidneys of more than 10 animals were pooled, indicating very clearly that the increase of the number of animals per batch increased considerably the probability of the presence of virus contamination. Use of kidney cells from animals, which are less susceptible for virus infections. One way to reduce the contamination by human pathogenic virus is to change the species of the animal as donor of the primary cells. Shah and Nathanson (1976) proposed that new world spider monkeys should be used instead of the rhesus monkeys or macaques because SV40 does not readily multiply in cells from spider monkeys. On the other hand, macaque monkeys can be infected with Herpes simiae or B viruses, which are highly pathogenic for humans. The replacement of the macaques as donors by African Green Monkeys, which are not susceptible to infection by herpes simiae virus, would be the best precaution in this case (Minor, 1996) . Other means. In addition to the above mentioned measures, there are some other measures which can be performed in order to increase biological safety. The animals intended for use should be examined for their health status and must pass through a quarantine regime. For safety reasons, there has to be routine use of in vitro and in vivo culture systems for detection of viruses in any case. However, the best means to increase the biological safety of the produced viral vaccines is the use of diploid or continuous cell lines, because it can be determined that such cells are free of animal derived viruses: This can be achieved by establishing master (seed stock) and working (distribution and user stocks) cell banks which have been rigorously tested and validated for the absence of microbial as well as viral contaminants (see chapter by Freshney and the section on 'Testing-virus screening in cell banks' of this article). By this means producers of viral vaccines and all other biotech products can make use of a homogeneous pool of characterized cells from which each production run will be started, in the knowledge that they are free of any contaminant (because they have been tested) (Berthold et al., 1996) . In addition, by using the seed stock/working stock concept for the viral inoculum the manufacturer of viral vaccines can use a tested and validated stock of virus inocula of which one aliquot is used for the infection of each production run. Many contaminating murine viruses, such as Minute virus of mice (MVM), K virus, Mouse Encephalomyelitis virus, and Mouse Adenovirus have been isolated from contaminated virus pools. Viruses, such as A similar study was published by Nicklas et al. (1993) , however, revealing a lower rate of contamina- (Nicklas et al., 1993) Origin tions probably due to the improvement of the microbiological quality of the laboratory rodents. Of 297 tumors examinated, 75 (25.3%) were contaminated. Considerable differences were observed for in vivo (36.6% positive of 186 tumors) and in vitro (6.3% positive of 111 tumors) passaged transplantable tumors. Mouse transplantable tumors showed the highest frequency of contamination, whereas tumors of other species showed much lower frequencies (Table IV) . Contamination with reovirus 3 and MVM was found in 4 (3.7%) of 109 cell lines tested, and in 2 of 60 monoclonal antibody bulk preparations. With respect to LCMV, Bhatt et al. (1986) reported its isolation from transplantable tumor cell lines. The testing of tumor cell lines revealed that 16 out of 55 in vivo tumor samples and one out of eight in vitro samples were positive. A similar situation was found in a New Jersey research institute, where human cell lines and tumor cell lines were passaged via nude mice for the development of monoclonal immunodiagnostics and immunotherapeutic agents. This LCMV contamination led to the outbreak of laboratory-acquired human LCMV infection (Mahy et al., 1991) . LCMV contaminated hamster tumor cell lines have also been responsible for an outbreak of infections occurred in medical center personal at the University of Rochester (Hay, 1991) . Mouse hybridomas are of particular concern because, first, these cells have been created by fusion of mouse spleenocytes with mouse myeloma cells, second, many hybridoma cells have been cultivated in animals, and third they are used for the produc-tion of injectables. This signifies that mouse viruses are potential contaminants of these cell lines and their products. These viral contaminants can be divided into two groups; group 1 contains viruses which are also known to cause human diseases or to be able to infect human cells, while group 2 contains other mouse viruses (Table V) (Minor, 1996) . Although Ectromelia virus is listed in group 2, cultures infected with this virus are only processed in the special P4 unit available at the NIH (Hay, 1991) . Ectromelia, a member of the orthopoxviurs group, is a natural pathogen in mice, and is able to replicate in all mouse lymphoma lines, in some hybridoma cell lines, and in BS-C-1 cells (Buller et al., 1987) . Consequently, the ATCC has screened its collection of murine cell lines but no characteristic cytopathic effects have been observed (Hay, 1991) . Moore (1992) listed the mouse viruses which had been detected in production cell banks of hybridomas: LMCV, MVM, Sendai, LDH, and epizootic diarrhea virus of infant mice. The present view is that hybridoma cell lines should be tested for the viruses indicated in Table V , and only those should be used for biotechnological applications which are free of these viruses. The only acceptable viral particles in bulk supernatants from hybridoma cell lines are those of endogenous origin. The passage of human tumor cells in nude mice can also lead to the infection of these cells by murine endogenous retroviruses. Crawford et al. (1979) reported the contamination of a human nasopharyngeal carcinoma with murine endogenous xenotropic retroviruses after a passage in a nude mouse. As these Table V . Viruses potentially infecting rodent cells (Minor, 1996) Group contaminations are of animal origin, it is necessary to verify the contamination status of laboratory animals. Minor (1996) indicated that all mouse strains, which were received at NIBSC from breeders of laboratory animals, were tested positive for MVM and Sendai virus. Finally working with rat cell lines, which have been passaged via rats, is also a concern because they can be contaminated by different rat viruses, in general, and by Hantaan virus, in particular. Hantaan virus has been isolated in cell culture from rat immunocytomas. Transplantation into LOU/M/Wsl rats and storage of passaged immunocytomas at -70 • C over a period of 8-10 yr did not eliminate the virus. Lloyd and Jones (1986) also showed that the passage of rat immunocytomas in infected LOU/Wsl rats led to a contamination of these cell lines. Hantaan virus is a silent pathogen in rats and mice, but causes disease in humans (hemorrhagic fever with renal syndrome) after infection (Leduc et al., 1985) . Laboratory animal care workers working with infected animals as well as persons handling contaminated rat immunocytoma cell lines have been infected with this virus (Leduc et al., 1985; Lloyd and Jones, 1986; Mahy et al., 1991) . For this reason, all ATCC certified cell lines of rat origin have been screened for the presence of hantavirus. The following rat cell lines and rat hybridomas appeared free of hantavirus infection: CCL 38, 43, 45, 47, 82, 82.1, 97, 107, 144, 149, 165, 192, 216; CRL 1213 CRL , 1278 CRL , 1439 CRL , 1442 CRL , 1444 CRL , 1446 CRL , 1458 CRL , 1468 CRL , 1476 CRL , 1492 CRL , 1548 CRL , 1569 CRL , 1570 CRL , 1571 CRL , 1578 CRL , 1589 CRL , 1592 CRL , 1600 CRL , 1601 CRL , 1602 CRL , 1603 CRL , 1604 CRL , 1607 CRL , 1631 CRL , 1655 CRL , 1662 CRL , 1674 HB 58, 88, 90, 92, 100, 132; and TIB 104, 105, 106, 107, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 145, 146, 164, 166, 168, 175, 183, 184, 207, 210, 211, 213 (Leduc et al., 1985) . It should also be mentioned that cell lines of other species can also be contaminated by hantavirus. This virus had been detected in a human lung carcinoma cell line (A549) (French et al., 1981) and in Vero E6 cells (McCormick et al., 1982) . Precautions: The precautions which should be taken to reduce the risk of contamination of cells during passage in animals: (i) Cells which have been once passaged in animals have to be screened for the absence of microbial and viral contaminants normally found in the animal species; (ii) Animal passages should be avoided as much as possible, the contamination risk by mouse pathogens is reduced 6-fold when these cells are cultivated in vitro (Nicklas et al., 1993) ; and (iii) only tested laboratory-bred animals (virus-defined, specific pathogen free), and no wild caught animals, should be used for animal passages of mammalian cells. The most important cell lines of biotechnological interest, mouse hybridomas and CHO cells, are known to contain endogenous retroviruses (ERVs) and are known to produce retroviral particles. ERVs exist in 2 forms, a viral form and a proviral form. Retroviral proviruses are transmitted through germ cells and are present in the genomes of almost all vertebrates thus far studied. Humans possess many ERV genomes related to mammalian C type retroviruses and to A, B and D types of this family of viruses. The presence of ERV-like sequences in a cell line to be used in the production of a biological is a potential cause for concern because of the possibility that the endogenous retrovirus may be activated and result in infectious virus being present. particles (IAP) as well as budding C-type particles (Table VI) . With respect to IAPs, Anderson et al. (1990) demonstrated, that the CHO cells' genome contained approximately 300 copies of viral sequence per haploid genome. No intact open reading frame for gag, pol, or env could be detected in clones of either family (Anderson et al., 1990) . In addition, no infectivity has been associated with A-type retroviruses from CHO cells (Kuff and Lueders, 1988; Adamson, 1998) . CHO cells also produce C-type particles at concentrations of <10 3 -10 6 ml −1 . The presence of virus particles was correlated with detectable reversed transcriptase activity (a retrovirus specific enzyme) (Dinowitz et al., 1992) (Table VII) . As for the IAPs, each CHO cell contains between 100 and 300 copies/genome (Dinowitz et al., 1992) . No evidence of infectivity could be detected (Dinowitz et al., 1992; Adamson, 1998) , this may be due to the lack of functional open reading frames, rendering the retroviruses incapable of encoding an intact endonuclease (Dinowitz et al., 1992) . Hybridoma and murine plasmacytoma cell lines. As for CHO cells, plasmacytoma and hybridoma cells produce IAPs and C-type particles (Table VI) (Spriggs and Krueger, 1979; Weiss, 1982) . As for CHO IAPs, hybridoma IAPs stay inside the cells and are noninfectious because they are devoid of intact open reading frames. However, unlike CHO C-type retroviral particles, hybridoma C-type particles have the ability to replicate in several different cell lines including a small number of human cell lines (lung fibroblasts, RD cells (Weiss, 1982; Levy, 1983; Adamson, 1998) ). By using electron microscopy, the retroviruses present in cell culture supernatant have been as high as 10 9 particles per ml (Moore, 1992) . About one in 10 4 -10 6 particles is infectious. In this context, Froud et al. (1997) presented data from Lonza Biologics (former Celltech Biologics), indicating that all hybridoma cell lines processed by the company produced retroviral particles. Five to six percent of the cells produced infectious mouse ecotropic retrovirus, whereas almost all (about 85%) mouse cell lines tested produced low levels of infectious mouse xenotropic retrovirus (X-MLV) when the cell banks were tested. All mouse cell lines commonly used for antibody or recombinant protein production are derived from the MOPC21 tumor of a female BALB/c mouse. This indicates that all cell lines, clones, and subclones which are derived from the MOPC21 tumor (the plasmacytomas P3X63.Ag8.653, NS1, NS0, and the hybridoma SP2/0.Ag14) produce X-MLV (Froud et al., 1997) . Infectious retrovirus has also been found in mouse/human hybridomas. In co-cultivation studies it could be established that these retroviruses were of the X-MLV type, and no human retroviruses have been detected in any mouse/human hybridoma or genetically engineered human cell line (Moore, 1992) . Although not detected so far, the possibility of molecular recombinations leading to pseudotyped particles is a concern. Other cell associated viruses and safety considerations Mouse/human hybridomas can be established by fusing human EBV transformed lymphoblastoid cells with mouse plasmacytomas. As these cells were EBV transformed, the hybridoma cells are potential EBV producers. Cells transformed by EBV are potential EBV producers (e.g. Namalva), and the downstream processing protocol as well as the safety testing have to take this fact into account (Cartwright, 1994; Robertson, 1996) (see Section on 'Process validation -downstream -processing -viral clearance'). BHK cells are also transformed rodent cells and it was possible to induce production of R-type particles in these cells (Moore, 1992) . Table VIII presents a short résumé on cell lines of biotechnological interest which contain endogenous viruses or latent proviruses and are therefore producers or potential producers of these viruses. Whereas plasmacytomas, hybridomas, CHO and BHK cells contain endogenous viruses which integrated into the genome Table VII . Characteristics of C-type particles isolated from CHO cells (Dinowitz et al., 1992) Reversed transcriptase activity Detected in highly concentrated (4000-7000-fold) culture fluids from some cell lines; Mn 2+ preferring. Similar to other C-type particles in sucrose density gradients (about 1.13-1.16 g ml −1 ). Nucleotide homology to other C-type Endnuclease region contains significant homology particles with mammalian C-type retroviruses. No intact open reading frames detected in cloned cDNA sequences. About 100-300 copies per CHO genome. Proteins P30 core protein is related to murine and other C-type retroviruses. No infectivity detected by direct inoculation of reverse transcriptase-containing concentrates or cocultivation of CHO cells with a battery of cell lines. of the respective species after an infection millions of years ago, the presence of EBV (due to EBV transformation of human B-lymphocytes) or sequences of parvoviruses (e.g. AAV (latent infection leading to integration) due to natural contamination of human embryonic kidney cells and African Green Monkey kidney cells (Hoggan, 1970) , or Procine Parvovirus (latent infection) (Fikrig and Tattersall, 1992 ) (see Section on 'Trypsin')) are due to recent events and can eventually be avoided by using tested cell lines (absence of the respective sequences). It is evident that all biological products derived from cell lines containing endogenous retroviruses or other latent viruses have to be characterized for the presence of virus. In addition, in order to increase their biological safety, first, all biotech products derived from such cell cultures have to be rigorously tested for the absence of retroviral activities/viruses or latent viruses, and second, the purification protocols for biotech products derived from these cells have to be validated for their capacity to eliminate or inactivate retroviruses/latent viruses. An important potential source of viral contaminations are raw materials used for the preparation of culture media in animal cell technology. Although any component of the culture medium can theoretically by contaminated by viruses, the materials with the highest probability of viral contamination are those derived from animal origin. The classical animal cell culture technology currently makes use of several raw materials of animal or human origin. This is true for the production of viral vaccines for human or veterinary use or many other biological therapeutic. Animal sera as medium additive is the most widespread animal derived material used today. Fetal, new born or adult bovine sera and in some cases also horse sera are used. Trypsin, mainly from pig pancreas, is a very important detachment agent for all adherent cells. Other animal or human derived substances are often used in association with the replacement of serum by serumfree media, but which can eventually also be found as excipients: human albumin, protein hydrolysates (casein, gelatin, etc.), and human transferrin. Other substances of animal origin are some amino acids, which are derived from complete hydrolysis of proteins. However, because of the chemical conditions used in their production, there is less risk than that of materials prepared without this process. Although the use of entirely chemically defined media devoid of any animal derived substance reduces the risk of viral contaminations, it is important to mention that this risk will never be zero. Several zoonotic viruses are known and can be transmitted from animal sources (Eloit, 1999) . Because of the recognised risks from these agents careful sourcing and screening can easily prevent the risk of their transmission. However, it should be noted that other viruses not known to be harmful for humans might be infectious and might lead to severe disease. Contamination problems associated with the use of serum Bovine serum might be contaminated by many different bovine viruses. Although theoretically qualitycontrolled serum should have been tested for all possible viruses, this is not possible for economic reasons and may not even be necessary. It is evident that each serum batch has to be tested for those viruses which are ubiquitous and known risks, such as the Bovine Viral Diarrhoea Virus (BVDV). However, there is also the question of the geographical origin of the serum which may indicate the need for additional virus tests. It is only necessary to test for those viruses, which are present in the geographical region from which the serum is coming. Table IX presents bovine viruses for which tests have to be performed depending on the geographical origin of the serum. The testing Table IX . Specific tests applicable to the screening of calf serum and porcine trypsin used for the production of medicinal products for human use (Eloit, 1999) Calf serum Trypsin (Table X) . In the following, the most important bovine viruses are presented in more detail. BVDV. Bovine viral diarrhoea/mucosal disease is one of the most important viral diseases of cattle. The natural prevalence is very high with approximately 80% of cattle being seropositive and 1-2% of these animals being persistently viremic animals due to immune tolerance which occurs after infection of the fetus (Kniazeff, 1973) . The infection rate has been increased by the uncontrolled use of live vaccines and by heterologous vaccines fortituously contaminated with BVDV virus (Kreeft et al., 1990) . Together with Hog Cholera and Border Disease virus of sheep, BVDV constitutes the pestivirus group. Bolin et al. (1991) have studied the frequency of contamination of fetal calf serum with BVDV and reported that 332 of 1608 raw fetal serum samples (20.6%) derived from the abbatoirs were positive for this virus, 224 of these samples (13.9%) contained antibodies against BVDV and 3.1% of the samples (50/1608) were positive for both, BVDV and antibodies against BVDV (Table XI ). They have also tested commercial fetal calf serum for cell culture and detec-ted BVDV in 47% of the samples (90/190): 88 contained non-cytolytic and only two contained cytolytic BVDV isolates. Two percent of these samples (3/190) were positive for Infectious Bovine Rhinothracheitis virus isolates. Wessman and Levings (1999) have reported similar results, indicating that 32 to 68% of fetal bovine serum samples (pooled one liter lots from two bovine fetuses) were rejected for presence of BVDV or antibodies against BVDV, in the period of 1990-1997. These studies indicate the importance of the problem of BVDV contamination in fetal calf serum and several conclusions can be made: first, veterinary diagnostic laboratories should avoid the use of fetal calf serum in diagnostic procedures for pestivirus infections, second, there is a significant risk that adventitious BVDV from fetal calf serum may lead to contaminations in the veterinary biologicals industry (see Section on 'Other substances of animal/human origin and non-animal derived substances'), and third, the results indicate a very high rate of fetal infection with BVDV, possibly reflecting a failure in hygiene issues or in control measures (Bolin et al., 1991) . Finally the most important question concerns the contamination status of cell lines in culture collections, because many cell lines have existed for many years in these collections and might have been contaminated in periods when no or fewer tests were performed for proving the absence of BVDV. The question most relevant to the use of fetal calf serum for animal cell culture and animal cell culture technology is: which cells are/were contaminated by BVDV and are the cells from different species as easily contaminated as bovine cells or is there any species barrier. In this context, Bolin et al. (1994) performed a survey of cell lines from the American Type Culture Collection and observed the following contamination status: Using immunocytochemical procedures and PCR amplification, 13 of 41 ATCC cell lines were tested BVDV positive: these cell lines were derived from cattle, sheep, goat, deer, bison, rabbit, and domestic cat. Attemps to experimentally infect 14 different cell lines from animals, which were not found positive in the survey of the ATCC cell lines, led to the result that all swine cell lines (3/3: MPK, ESK-4, and one other) and most rabbit (3 out of 4: Sf 1 Ep, R9AB, RAB-9) and cat cell lines (3 out of 4: CRFK, AK-D, NCE-F161) became infected with BVDV, whereas hamster (BHK-21), human (IMR-90), dog (MDCK), rabbit (SIRC) and cat (Fc3Tg) cells were refractory to BVDV infections. The results concerning monkey cells (LLC MK2) were variable -no clear answer was obtained. Wessman and Levings (1999) reported that the following cell lines could be infected with BVDV: bovine cells (EBK, MDBK, BoTur, primary and continuous kidney cell lines, lung, trachea, and aortic endothe-lium), sheep choroid plexus and lamb kidney cells, monkey kidney cells (Vero and others), mosquito cells, porcine cells (PK-15 and others, testis, minipig kidney cells), goat cells (kidney and oesophagus), cat cells (lung, CRFK, tongue, feline embryo), rabbit kidney cells (RK-13), and others. Harasawa and Mizusawa (1995) published a study on the pestivirus contamination of cell stocks of the Japanese Cancer Research Resources Bank. Fifteen out of 20 cell lines (75%) were positive using RT-PCR. Whereas bovine cell lines (HH, MDBK, CPA, CPAE, EBTr, Ch1Es) were contaminated with genotypes I, II, and III, cell lines of dog, cat, and primate origin were contaminated with genotype II of BVDV (HeLa, MOLT-4, U937, WI-38, WiDr, CV-1, Vero, MDCK, CRFK). Roehe and Edwards (1994) assessed the ability of 11 pestiviruses from pig, eight from cattle, and five from sheep to replicate in cells of porcine (PK-15), bovine (BT) and ovine (SCP) origin. The pattern of replication in different cell types varied between different isolates of the same virus species. These results indicate that the virus suceptibilities of a species are not completely predictible and that many cells derived from other species than cattle can be infected by BVDV. of importance. Viral infections lead to respiratory syndrome in cattle (shipping fever) and abortion in bovines. In 68% of calves significant antibody levels against PI-3 have been detected. The virus can be easily replicated in primary (bovine kidney cells) and established bovine cells (EBTr, MDBK) (Kniazeff, 1973) . tracheitis -a very common infection of cattle -, abortion, pustular vulvovaginitis, meningoencephalitis of calves, and conjunctivitis. The incidence of infection is high with viremia a common feature. It replicates in leukocytes and can stay there latently. It is never found free in the bloodstream. It replicates in bovine cells, but also in cell cultures from: elk, mule deer, sheep, L cells, chick embryo cells, pig, human (amnion (probably also a clon of HeLa cells (see chapter by Masters)), HeLa), and primary monkey kidney cells (Kniazeff, 1973) . BPyV. Bovine polyomavirus belongs to the family of the polyomaviruses. These viruses have been isolated by several laboratories (e.g., from monkey kidney and other cells cultured in the presence of bovine serum. By infecting permissive cells, this virus leads to a cytopathic effect, whereas non-permissive cells are transformed and they acquire certain properties of a malignant cell. As for the other bovine derived viruses, BPyV is an ubiquitous virus and about 40% of the calves are seropositive in the first month after birth. In the subsequent months this seropositivity decreases to about 11% at an age of one year, however, the older the animals become seropositive again and the final percentage of seropositivity is beyond 80%. It was also shown that bovine fetuses were infected in utero, leading to the presence of antibodies against this virus in fetal bovine serum batches (in about 12% of the tested batches). Despite this rather high incidence of infection in fetuses, no known disease is associated with this virus, neither for cattle nor for humans. Using PCR, 14/20 European serum batches (70%) contained BPyV DNA sequences . A similar frequency of contamination was observed in North American serum batches (Kappler et al., 1996; Van der Noordaa et al., 1999) . There was no correlation between the PCR results and the presence/absence of antibodies against BPyV, however, could show that all PCR-positive sera contained infectious BPyV. IBR or BHV-1. BPyV can infect calf kidney cells, and monkey kidney cells (Vero, BSC-1, CV-1, RITA) but also human embryonic kidney cell cultures (Waldeck and Sauer, 1977; Wognum et al., 1984) . The virus does not seem to replicate in mouse 3T3 cells, nor in the human embryonic lung cell WI-38 (Waldeck and Sauer, 1977) . The BPy virus is known to lead to cell transformation and tumorigenesis, which is induced by the expression of the large-T antigen . All adherently growing cells have to be detached for passaging from time to time. To facilitate this the enzyme trypsin is frequently used. As for serum, trypsin is an animal derived product, generally from porcine pancreas. Therefore, similar safety criteria as for serum have to be applied for trypsin. A special concern is Porcine Parvovirus (PPV). Latent parvovirus contamination has been found in many permanent human cell lines. The first contamination was observed by Hallauer and Kronauer (1960) when they observed that some control cultures (non-infected with Yellow Fever virus) yielded a different hemagglutinating agent, unrelated to Yellow Fever virus, when subjected to their isotonic, high pH glycine extraction buffer (= physiological stress). Further studies identified this infectious agent as a member of the parvovirus group. Following this, Hallauer et al. (1971) isolated 38 parvoviruses in 43 permanent human cell lines obtained from 19 laboratories. Some cell lines showed signs of degenerescence when arriving into the laboratory of Hallauer, others appeared completely normal. Three different serotypes of parvovirus could be identified, the origin of them is not really known. However, the recovery of the same serotypes from different laboratories suggests a common source, such as a reagent used in cell cultivation. One of the serotypes had been identified as PPV, indicating that the use of contaminated trypsin lots was the probable source of contamination (Fikrig and Tattersall, 1992) . The definite proof that porcine trypsin was the source for cell culture contaminations by PPV was apported by Croghan et al. (1973) , because they detected the same serotype in commercial trypsin lots. It could be shown that various cell lines from different species can be infected by PPV, such as human continuous cell lines (Lu 106, HeLa, and the following HeLa clones (see chapter by Masters): KB, Amnion, and Hep-2) or swine kidney cells (PK 15) (Hallauer et al., 1971) . Recently, a new group of viruses, the circoviridae, was described. The circoviridae are very small viruses (non-enveloped, circular single stranded DNA, diameter of 17 nm) and are very resistant against most of the inactivation methods currently used. This group of viruses was found in Japanese patients suffering from non A to G hepatitis, described as TT virus (Nishizawa et al. 1997) , and also found in chickens, where it is described as Chicken Anemia virus (Yuasa et al., 1979) . This virus seems to be ubiquitous in humans because DNA of the TT virus was identified in plasma of 76% of French blood donors (Biagini et al., 2000) . It is evident that such a virus might be a problem when human derived proteins are used, because this it is very resistant against most inactivation methods. Tischer et al. (1982) described a circovirus in pigs, and it has been reported that the swine cell line PK-15 was chronically contaminated by this type of virus. A serological study showed that 20 out of 22 randomly collected pig sera contained specific antibodies against the virus, whereas no specific antibodies could be detected in sera from rabbits, mice, calves, and man, including the laboratory staff working with this virus (Tischer et al., 1982) . The virus exists as 2 subtypes, type 2 porcine circovirus replicates actively in porcine fetuses (Sanchez et al., 2001) and is associated with abortions, reproductive failure and postweaning multisystemic wasting disease in swine (O'Connor et al., 2001; Ellis et al., 2001) , however, it seems that this virus is very porcine specific . Although there is no record of Porcine Circovirus being able to infect man, precautions should be taken when porcine derived substances, such as trypsin, are used in animal cell technology. As this type of virus is very resistant it is preferable to avoid contaminations of the cell culture, and thus of the biotech product, than to try to separate the product from the virus (due to stability reasons autoclaving the final product is not possible). For the moment, these viruses are most frequently detected by PCR, presently there is no good culture method available. The highest risk is associated with the use of human or animal derived substances. With respect to material of human origin, it is evident that there exists an important risk of viral transmission, because of the absence of any species barrier to infection. Human sourced raw materials should be checked for the absence of viruses, like Hepatitis B virus (HBV), Hepatitis C virus (HCV), Human Immunodeficiency virus (HIV), and EBV, or CMV (Committee for Proprietary Medicinal Products: Ad Hoc Working Party on Biotechnology/Pharmacy and Working Party on Safety Medicines, 1992). The testing for these viruses is most frequently accomplished by PCR assay. The extent of the use of human derived materials is limited with only human transferrin and human serum albumin still in use. The development of serum-and protein-free media leads more and more to media devoid of any animal/human derived substance. Other substances of animal origin, such as certain amino acids, lipids, or other protein additives (peptones) are also potentially contaminated by viruses and have to be rigorously tested for virus absence or should be replaced by non-animal derived compounds (Merten, 1999; Jayme, 1999) . Finally it should be mentioned that adventitious viruses can also be introduced via contaminated nonanimal derived substances (medium components), as observed by the contamination of a manufactured material by viruses of extraneous origin. The most widely reported case was that occuring with the manufacturer Genentech (Garnick, 1996) (see next section). The scientific literature gives few descriptions of viral contamination in biotechnological productions of recombinant proteins or viral vaccines (Table X) . Nine cases have been published indicating that viral contamination can be acquired via the serum source (in 7 out of 9 cases) or the culture medium (in 1 out of 9 cases). Biotechnological products were contaminated by different viruses, by those leading to a cytopathic effect (epizootic haemorrhagic disease virus) but in most cases by those which have no effect on the morphology of the contaminated cells (Minute Virus of Mice (MVM), BVDV, BPyV, bluetongue virus, reovirus). In cases where the viral infection leads to observable morphological effects, the contamination is easily detectable. However, those viruses which do not lead to a modification of the morphology and/or growth characteristics of the cells require methods such as RT-PCR, PCR, bulk in vitro testing. Or alternatively the onset of diseases in animals administered with the test material. These virus positive products were not delivered in the case of the products destined for human use, because the virus detection was performed before product release (Garnick, 1996; Rabenau et al., 1993; Harasawa and Sasaki, 1995) , however, with respect to the live attenuated veterinary vaccines, several incidents of disease in vaccinated animals were reported (Kreeft et al., 1990; Wilbur et al., 1994; Falcone et al., 2000) . It should be mentioned here, that the case of MVM contamination of CHO cultures for the production of TPA, did not lead to a cytopathic effect and could not be detected without specific virus tests (Garnick, 1996) . In contrast, Nettleton and Rweyemamu (1980) and Hughes (1996) reported on a MVM contamination of BHK-21 cells for veterinary vaccine production, which was detected via persistent cell deaths of these cells. It could be shown that the serum batch was the origin of this contamination. It is evident that rigourous testing of raw materials is necessary because of the following: -In the case of the production of recombinant proteins for human use, virus contaminations can only be eliminated with difficulty from the bulk product. Should a virus present in the bulk product be able to be eliminated during downstream processing, the FDA will generally not accept the final product after purification (Burstyn, 1996) . -The problem associated with live attenuated virus vaccines is that these products cannot be treated for virus inactivation because the active ingredient would be inactivated at the same time. Such products require a more extended quality control testing for the raw materials. For instance, new serum batches should be tested for a more extensive range of bovine viruses and in particular for those viruses, which are of relevance for the final application of the product. For instance, in the case of the contamination of a canine vaccine by bluetongue virus which lead to the death of some bitches (Table X) (Wilbur et al., 1994) , the application of a specific test would have avoided this incidence. -In conclusion, the best solution for reducing the risks of viral contaminations is the use of raw materials which are not of animal or human origin, but of plant or microbial origin or produced by chemical synthesis. It is evident that such an approach will not eliminate the risk of viral contaminations, but represents an important step towards risk minimization. Sourcing. This approach is clearly limited to agents for which there is a well-documented specific geographical distribution. Such examples are quite rare and only the case of TSE agents will be mentioned here, where sourcing of bovine serum from diseasefree countries (a geographical choice) is possible. This approach, however, can also be used for viruses such as bluetongue virus. Screening. In principle, all raw materials, of animal origine or produced by chemical synthesis, have to be rigourously tested and should fulfill certain quality attributes, when used for the production of biological injectables (GMP-guidelines). The characteristics which are frequently required to be described in raw materials are identity (testing, tracability, labels), purity (testing, inspection, vendor certificate of analysis), suitability for intended use (process validation, vendor audit programme, performance testing if needed), tracability (vendor audit programme, vendor certification, certificates of analysis, contractual obligations under change control, labelling, control). For more details, consult Lubiniecki and Shadle (1997) . Although it would be desirable that all raw materials should be tested for the absence of adventitious agents, in order to be sure that they are safe, this is often impracticable. Therefore there are two approaches: first, tests are employed which are based on the detection of general characteristics of viruses (cytopathic effects, haemadsorption) and, second, specific tests using imunological and/or PCR methods are employed for detecting virus antigens or specific viral sequences after amplification in permissive cells (see Table IX for testing of bovine serum and porcine trypsin, see Section on 'Testing -virus screening in cell banks'). However, such a screening gives only a limited guarantee of safety because of the following: -Complete testing can be impracticable on a batch to batch basis. In most cases, screening will only be done for certain viruses, e.g. for BVDV, IBR, and PI-3 in the case of bovine serum, for porcine parvovirus in the case of trypsin, because these are the most probable viral contaminants. However, depending on the geographical origin of the serum or the trypsin, additional tests for viruses which are present in that geographical area from where the raw material is coming may have to be performed (Table IX) . If a raw material is of animal origin, screening tests should also include the use of cells of the species of origin. With respect to the use of serum for the GMP production of biologicals for human use, the EMEA proposes in a draft that more viral screening tests should be performed for proving the absence of BVDV, IBR, PI-3, Bovine adenovirus, Bovine Parvovirus, Bovine Respiratory Syncytial virus, Bovine Re- virus tests, otherwise they pass undetected. -Due to sampling size, low titers of some adventitious viruses can remain undetected but may be amplified during the manufactureing process. In this context it should be remarked that screening methods are not always sufficient because contaminated serum batches which had passed as uncontaminated have been detected by and Yanagi et al., (1996) ; contamination of serum samples with BPyV and BVDV, respectively. Recommendations for fetal bovine serum quality (Hansen and Foster, 1997) . Although the best choice would be a serum-free cell culture process which is devoid of any animal or human derived substances, this is not always possible. Where serum supplementation is necessary, the serum should be of high quality. In addition to the absence of viruses the hemoglobin level should be <10 mg, the endotoxin level should be below 10 Eµ ml −1 , and there should be a reliable tracability to countries without BSE nor foot and mouth disease. A serial filtration using 40 nm poresize filters should be used and the serum should be γ irradiated with >25 kGy using validated procedures. For veterinary use, the radiation dose should be 35 kGy. More on the quality control of bovine serum used for the production of viral vaccines for human use can be found in Mareschal (1999) . Other precautions: The screening of animal derived raw materials for the presence of adventitious viruses is of utmost importance, however, as already mentioned, screening has its limits, because it is impractical to screen for all theoretical viruses, and other new viruses might emerge for which no tests are available. Because of this supplementary precautions have to be undertaken for reducing the risk of viral contamination, by inactivating or eliminating at least viruses of families, which are susceptible for inactivation and/or elimination. With respect to the treatment of fetal bovine serum for animal cell technology, γ irradition, and UVC irradiation are used. Heat treatment as well as treatment with peracetic acid are possible, however, they are not really used. Some treatments are presented in the following: (a) Nanofiltration (Troccoli et al., 1998; Aranha-Creado et al., 1997; Graf et al., 1999) . If the size difference (molecular weight) between the raw material and the virus is large enough, viruses can be removed by nanofiltration, which makes use of pore cut offs of 50, 35 nm, and even 15 nm. Filters with a pore cut offs of 50 nm can be used to eliminate viruses which have a diameter larger than 50 nm, such as retroviruses or influenza A virus (80-120 nm) (typical log titer reduction in a validation study: >6.3). However, such filters only partly reduce the quantity of poliovirus (28-30 nm), and viruses of a size of 25 nm (model particle: Bacteriophage PP7) pass without any significant retention (log titer reduction: <1-8.5, depending on the buffer system used) (Graf et al., 1999) . It should be mentioned that the composition of the medium/buffer system in which the virus is placed, has an effect on the log titer reduction of viruses which are below the molecular weight cut-off of the membrane used. An example of typical virus retention data for a commercial hydrophilic PVDF membrane filter is shown in Table XII . Improved virus retention can be obtained by using pore cut offs of 35 nm. Using a 35 nm membrane in line with two prefilters (one 75 nm filter followed by a first 35 nm filter) led to log titer reductions of >4.3 for Hepatitis A Virus (HAV) and Encephalomyocarditis Virus, although both viruses are smaller (28-30 nm) than the cut-off of the filters (Troccoli et al., 1998) . All viruses larger than 35 nm were completely removed. Finally the use of a 35 nm membrane filter followed by a 15 nm pore size membrane filter assures a log reduction factor of >6.7 and >5.8 for HAV and BVDV, respectively, signifying that in principle all potential adventitious viruses (also the small ones) can be removed from the product (Johnston et al., 2000) . Nanofiltration is mainly used as a final step in the production of biologicals purified from human plasma and of recombinant DNA-derived products. Fetal calf serum is often filtered three times using a cut-off of 100 nm (removal of, for instance, IBR and PI-3, but not of BVDV). Only some companies provide fetal bovine serum which is serially filtered through 40 nm pore size filters (Hanson and Foster, 1997) , because this cut-off allows also the elimination of, for instance, BVDV, which has a size of 45-55 nm (see Table XIII ). (b) γ Irradition (Plavsic et al., 1999) . γ irradition (using a 60 Co source) is a very efficient and straightforward means for inactivating many different virus types. As animal derived substances such as serum can be contaminated by adventitious viruses, γ irradition is, after routine quality control for virus detection, the best method to increase the safety of using serum in the production of animal cell culture derived biologicals. Using PETG (polyethylene terephthalate G copolymer) bottles with 500 ml of frozen serum (-40 • C) inoculated with model virus (Hanson and Foster, 1997) , validation experiments have been performed to determine the optimal radiation dose to inactivate relevant bovine and porcine viruses, and in parallel to assure that the irradiated serum has still a sufficient growth supporting ability. Plavsic et al. (1999) could show that at a radiation dose of 25 kGy, all tested viruses (Bovine Reovirus, Porcine Parvovirus, Canine Adenovirus, IBR, and BVDV) showed a significant decline in titer. An exposure of 35 kGy led to titers of all viruses tested falling below the detection level (≤0.5 TCID 50 ml −1 ). Even very resistant viruses, such as the Porcine Parvovirus, could be reduced to below the detection level. For all viruses tested the log reduction factor was at least 6.78 (Table XIV) . Willkommen et al. (1999) reported an overview on virus inactivation and removal from serum and serum substitutes. With respect to the efficiency of PPV inactivation by γ irradition they indicated that even after application of a radiation dose of 40 kGy, a TCID 50 of 5.3 per ml was observed, indicating that the log reduction was only about 2. This difference with data published by Plavsic et al. (1999) might be due to differences in the design of the respective studies. However, with respect to the other viruses tested (BVDV, IBR, PI-3, reovirus 3), no differences in the inactivation doses were observed. It should be mentioned here, that sera are normally irradiated using a dose of 20 to 25 kGy. For veterinary use, the radiation dose has to be 35 kGy. A very important consideration is the capacity of the irradiated serum to support cell growth. By performing long-term standard cultures (three passages in a medium supplemented with 5% of the irradiated sera), Plavsic et al. (1999) were able to show that in principle all tested cell lines could be cultivated, but also that different cells reacted relatively differently on the radiation doses used. Whereas low passage BHK cells, Vero (only slightly), and CHO cells displayed an inverse relationship between growth and radiation dose, high passage BHK cells and the human diploid fibroblasts, WI-38 and MRC-5 -the latter are of special interest for vaccine production and virology -did not display growth decline as a function of radiation dose (Table XV ). None of the tested cell lines showed a modifed morphology. The advantages of irradiation is that it is easy and safe and does not leave residual molecules in the final product as when chemical inactivation methods are used. Today γ irradition is mandatory in Europe for fetal bovine serum. Validation experiments of the γ irradiation (dose: 25-35 kGy) of porcine pancreatic trypsin powder indicated a 6.7 log 10 reduction of the median tissue culture infective dose (TCID 50 ) (Erickson et al., 1989) . (c) UVC irradiation (Kurth et al., 1999) : A second rather easy and safe method is UVC irradiation for inactivating adventitious agents. As distinct from γ irradition, the UVC irradiation has to be performed by using a continuous flow through irradiator. An irradiation time of 8±1 s and a fluence of 0.1 J cm −2 are used normally. The principle of this type of irradiation is a DNA excitation leading to electron transfer (→ 8-hydroxoy-guanine), photohydration (→ cytosin hydrates), photoaddition and dimerization (→ Pyr -Pyr, Thy -Ade, Pyrimidine (6-4) Pyrimidone). UVC irradiation is very effective for selected virus groups, especially for those with single-stranded nucleic acids. The data shown in Table XIII indicate that all tested single-stranded viruses were rather efficiently inactivated with a log clearance ranging from >5.5 to 8. Only the reovirus 3 which has a double stranded RNA shows a lower log reduction (4). Long term growth assays did not reveal any reduction in the ability of the UVC irradiated sera (used at 1%) to support cell growth. (d) Other treatments: Substances of animal origin, such as serum or trypsin, or final biotech products can also be treated by other methods for reducing the eventual viral burden. These treatments can be of chemical nature, such as treatment with peracetic acid (Hughes, 1996) , solvents (e.g. 1% Tween 80 and 0.3% tri-n-butyl-phosphate at 25 • C for 8.5 h for the treatment of plasma derived Factor IX, works only for lipid enveloped viruses) (Johnston et al., 2000) , or imines (Brown et al., 1999) , or physical methods, such as heating (Hughes, 1996) or the reduction of the pH to 4.5. With respect to heat treatment, it is less effective than irradiation methods (Willkommen et al., 1999) and the serum composition is too much altered (Hanson and Foster, 1997) , leading to a rather weak growth promotion. The addition of a chemical substance, such as peracetic acid, is not ideal because a chemically reactive substance is added which might also lead to an inactivation of some medium compounds. In spite of this, virus inactivation based on the treatment with peracetic acid is rather effective for inactivating resistant virus, such as poliovirus Table XV . The effect of increasing doses of gamma radiation on the ability of fetal bovine serum to support the long term growth of selected cell lines in culture (expressed as percent of growth of control cultures) (Plavsic et al., 1999) (Sprössig and Mücke, 1967) , as well as for maintaining the growth supporting ability of the serum. Other adventitous agents like mycoplasmas and bacteria are also efficiently inactivated (Schweizer et al., 1972) . (e) Replacement of animal derived substances by non-animal/non-human derived substances: Although viral screening tests are efficient for detecting adventitious viruses, the best remedy for avoiding the presence of these viruses is the use of non-animal/nonhuman derived raw materials. This is not a complete assurance for the absence of virus, but a considerable risk reduction, since adventitious viruses might also be introduced via non animal derived raw materials, such as medium components as shown by Garnick (1996) . In this context it should be mentioned that most of the recent biologicals based on the use of animal cell technology are produced in serum-free or protein-free media (Froud, 1999; Merten, 1999) . With respect to the production of viral vaccines, the first serum-free viral vaccines were developed and tested in clinical studies (Brands et al., 1999; Kistner et al., 1999) , and are going to be put on the market. Recently, a study concerning a veterinary live virus vaccine, which was produced under protein-free conditions (devoid of any animal or human derived substances), showed that such a vaccine was as efficacious and safe as a classically produced vaccine (under serum-conditions) (Makoschey et al., 2002) . This indicates very clearly that the use of serum for the production of viral vaccines, in particular, and of biologicals, in general, is an anachronism and that the efficient replacement of non-animal derived serum-supplements is feasible. Operator induced biological contaminations in cell culture is a multifaceted problem involving the unex-pected introduction of other animal cells (see chapter by Masters), microbial (see chapter by Drexler and Uphoff), and viral contaminants. There are few reports on operator induced viral contaminations. The potential exists, however, as reports have appeared documenting the considerable stability of Rhinoviruses, Respiratory Syncytial virus, Rotaviruses, and others, in aerosols on worker's hand and safety hood surface (for more details, see Hay, 1991) . In general, viral contaminations of cell lines cannot be treated and contaminated cultures should be discarded, with the exception of LDV. This virus causes a life long viremia in infected mice without any clinical signs, and each sample of these animals is virus contaminated. Because LDV requires primary mouse macrophages for replication it cannot survive repeated in vitro subcultivations, leading to a loss of this virus in infected in vitro cultures. Another elimination method is the passage of the contaminated cell line/tumor in another species, for example nude rats (Nicklas et al., 1993; Nakai et al., 2000) . The absence of virus can only be assured by performing a rigorous testing programme, which includes all steps in a bioprocess: master cell bank, working cell bank, the raw materials, the unprocessed bulk harvest, late expanded cells, and the final product. A summary on the tests to perform is presented in Table XVI . Whereas research cell banks are mostly tested for sterility and absence of mycoplasmas, GMP cell banks a Used for preparation of WCB or in production runs starting from MCB or WCB. b Cells at the end of a typical production run are tested to determine the virus load in case of retrovirus like particle bearing production cells. Only few harvests need to be tested (validation). c Old cells may be from production runs (as post production cells) or from a separate culture kept in continuous culture for a long period and prepared for this analysis of 'limit of cell age' only (as late expanded cells). Extensive testing performed as part of the qualification of the MCB regarding absence of latent virus, inducible by cultivation on production conditions. d A very large sample volume for testing would be required for statistics of a sufficiently sensitive detection of low virus titers. for the production of biologicals for parenteral applications have to be tested much more rigorously. ICH Topic Q5A (1997) suggests the following virus tests for different cell banks: 'A master cell bank has to be extensively screened for both endogeneous and non-endogeneous viral contaminants. For heterohybridoma cell lines in which one or more partners are human or non-human primate in origin, tests should be performed in order to detect viruses of human or nonhuman primate origin as viral contaminants arising from these cells may pose a particular hazard. Testing for non-endogeneous viruses should include in vitro and in vivo inoculation tests and any other specific tests, including species-specific tests such as mouse antibody (MAP) test, that are appropriate, based on the passage history of the cell line, to detect possible contaminating viruses.' 'The working cell bank as a starting cell substrate for drug production should be tested for adventitious viruses either by direct testing or by analysis of cells at the limit of in vitro cell age, initiated from the WCB. When appropriate non-endogenous virus tests have been performed on the MCB and cells cultured up to or beyond the limit of in vitro age have been derived from the WCB and used for testing for the presence of adventitious viruses, similar tests need not to be performed on the initial WCB. Antibody production tests (MAP, RAP, or HAP) are usually not necessary for the WCB. An alternative approach in which full tests are carried out on the WCB rather than on the MCB would also be acceptable. ' 'The limit of in vitro cell age used for production should be based on data derived from production cells expanded under pilot-plant scale or commercial-scale conditions to the proposed in vitro cell age or beyond. Generally, the production cells are obtained by expansion of the WCB; the MCB could also be used to prepare the production cells. Cells at the limit of in vitro cell age should be evaluated once for those endogenous viruses that may have been undetected in the MCB and WCB. The performance of suitable tests (e.g. in vitro and in vivo) at least once on cells at the limit of in vitro cell age used for production would provide further assurance that the production process is not prone to contamination by adventitious virus. If any adventitious viruses are detected at this level, the process should be carefully checked in order to determine the cause of the contamination, and completely redesigned if necessary.' The detection of adventitious viruses in cell banks has to follow two principles -use of detection methods for specific viruses such as MAP, HAP, RAP (mouse, hamster, rat antibody production tests -examination of serum antibody levels against specific viruses or enzyme activity after a specified period), and different specific PCRs, as well as the use of general tests which may indicate the presence of one or more of a large variety of different viruses. General tests include the in vitro test for adventitous virus. This test involves the inoculation of different cell lines -a human, a primate and a bovine (if bovine material was used for the production, otherwise a cell line from the species of origin of the cell substrate), and the production cell line (co-cultivation test). The RTase assay is a general test which will detect the presence of all viruses which contain reverse transcriptase enzyme. The last general test which is normally applied is the, in vivo tests where animals are used to identify the presence of virus. The animals are treated using different inoculation routes (the health of these animals should be monitored and any abnormality should be investigated to establish the cause of the illness). Finally, electron microscopic examination is also a general test which can be used for detecting adventitious viruses in the case of rather high virus loads. More details can be found in the articles by Poiley (1990) , by Werz et al. (1997) , and in the ICH Topic Q5A (1997). When a producer cell line of murine origin is used, the consensus opinion among regulators is that all known murine viruses should be tested for. If a cell of human origin is involved in production, then there should be tests for human viruses, such as HIV, HTLV, EBV, CMV, HHV6 and HHV7. Human-mouse heterohybridoma cell lines have to be tested for both human and murine viruses (Robertson, 1996) . In general, the unprocessed bulk material (pool of harvests of cells and culture media) should be tested for viruses after the end of production and before any downstream processing. The scope, extent, and frequency of virus testing on the unprocessed bulk should be determined by taking several points into consideration including the nature of the cell lines used to produce the desired product, the results and extent of virus tests performed during the qualification of the cell line, the cultivation method, raw material sources and results of viral clearance studies. In vitro screening tests, using several cell lines, are generally employed for testing. If appropriate, a PCR test or other suitable methods may be used. The presence of an adventitious virus will block the further use and processing of the harvest material. For screening of raw materials, mainly serum, see Section on 'Sourcing, screening and other precautions'. In summary general virus tests are vital because past incidents of viral contaminations have derived from viruses not known to be present within the production systems. Therefore the approach involving a variety of both general and specific tests applied at more than one stage of manufacture in combination with viral elimination steps during subsequent processing to assure the safety of the product is of utmost importance. The use of modern biotechnology for the production of biopharmaceuticals allows the treatment of diseases, which could not be treated previously. However, virus infection and replication is an inherent risk during cultivation of mammalian cells. Raw material testing, rigorous characterization of the master and the working cell bank as well as testing of the final bulk product (before downstream processing and virus inactivation steps have been performed) lead to a considerable increase in the viral safety. Although viral contamination might happen during production only preventive measures can be taken in fermentation and product biosynthesis. Thus the downstream processing is an integral part of the manufacturing process and has to be validated in order to assess its potential to eliminate, clear, or inactivate viruses. The downstream processing has two, sometimes contradictory aims: (i) purify the product to the required purity at high recovery using the lowest number of steps possible, and (ii) eventually increase the number of purification steps in order to assure the elimination of potential viral contaminants. The general difficulty resides in the physico-chemical properties of the product. The properties that maintain the beneficial effects of a product are often very similar to those carried by all viruses, in particular when the bioproduct is a live virus (for vaccination or for gene therapy purposes). Therefore only a limited spectrum of techniques can be used for virus inactivation/elimination. To assess the capability of individual process steps to remove viruses these steps have to be tested with live model viruses for which clearing factors can then be calculated. These, so called, viral clearance studies have the objective to demonstrate the capacity of different steps of the purification process to eliminate or inactivate adventitious agents acquired during the production process (contaminated cells, raw materials, process failure, etc.). They are performed via spiking experiments. As viruses vary greatly in properties such as size, resistance to inactivation, type of envelope, type and structure of their genome, the model viruses used for these spiking experiments have to be selected in order to cover the whole spectrum of potentially present viruses. However, in order to assure absence of viral contaminations derived from the producer cell line in the product, the downstream protocol has also to be validated for its capacity to inactivate or eliminte these specific viruses (e.g. retroviruses derived from hybridoma and CHO cells, or EBV derived from human lymphoblastoid cell lines). In general, it is difficult to show more than a five log removal on any given step due to the titers of the model virus used. More details can be found in the following references: Fritsch (1992) , Cartwright (1994) , Werz et al. (1997) , ICH Topic Q5A (1997), and Larzul (1999) . It is evident that a research laboratory cannot afford all tests, which have to be performed by the biotech industry in the case of GMP production. In addition, there is no need for such exhaustive testing unless the materials are to be used in the treatment of human patients. However, most of the issues described in the chapters on 'Problems associated with viral contaminations' and on 'Origin of viral contaminations' are valid for everyone working in the field of animal cell culture. The use of validated cell lines, shown to be 'virus-free', is the best choice because cell collections, such as the ATCC (www.atcc.org), perform entrance tests for new cell lines for assuring the absence of mycoplasma, bacteria, fungi, protozoa, and cytopathic viruses and can guarantee a certain microbial quality for the delivered cell lines. The German Cell Line Bank (www.dsmz.de) provides cells, most of them have been tested for the absence of HIV-I, HTLV-I and II, EBV, HBV, HCV, and HHV-8 by using PCR or RT-PCR. The absence of retroviruses is proven by performing a reverse transcriptase assay. However, cell lines obtained from such collections, can still be contaminated by viruses because viruses, which do not lead to a cytopathic effect, are not detected by the tests commonly used, on one side (ATCC), or only tests for detecting human pathogenviruses have been performed on the other side (DSMZ). If a very important cell line of a research laboratory has to be more rigorously screened, commercial screening services are needed and their use is recommended. The use of controlled animals, free of microbial contaminants, for animal passages of cells and of controlled raw materials derived from accredited dealers, who are performing viral screens (e.g. for virus absence in serum and trypsin preparations), for the preparation of culture media are steps in the right direction for reducing the risk for viral contaminations. The rules concerning the general cell culture operation procedures for avoiding viral contaminations are largely the same as for preventing mycoplasmal contaminations (see chapter by Drexler and Uphoff). Viral contaminations are a serious threat for animal cell cultures and may lead to false results in research, development, and virus screening, to viral contaminations in the biologicals derived from the contaminated cultures and finally to an infection of the treated patient. Fortunately due to rigorous testing of the animals used as the source of explants, the production of ascites or the passage of cells, of raw materials, of the cell strains and cell lines in use, and finally of bulk and the final product can prevent potentially dangerous viral contaminations. Existing data demonstrate that contamination of cells and harvests by viruses can occur for products of biotechnology, and while the frequency may be low it is not zero. For instance, routine testing of cell lines of biotech interest revealed a contamination frequency with adventitious viruses of less than 1% (Moore, 1992) . However, the possibility of new emerging viruses and the permanently existing risk of contaminations by adventitious agents and viruses leads to the conclusion that the user of animal cells as well as the producer of biotech products by using animal cells have to be attentive to this possible threat and that they have to assure the absence of adventitious agents/viruses by any mean. Only then, animal cell technology biotech products can be used for the benefit of everyone.

DENV Inhibits Type I IFN Production in Infected Cells by Cleaving Human STING

Dengue virus (DENV) is a pathogen with a high impact on human health. It replicates in a wide range of cells involved in the immune response. To efficiently infect humans, DENV must evade or inhibit fundamental elements of the innate immune system, namely the type I interferon response. DENV circumvents the host immune response by expressing proteins that antagonize the cellular innate immunity. We have recently documented the inhibition of type I IFN production by the proteolytic activity of DENV NS2B3 protease complex in human monocyte derived dendritic cells (MDDCs). In the present report we identify the human adaptor molecule STING as a target of the NS2B3 protease complex. We characterize the mechanism of inhibition of type I IFN production in primary human MDDCs by this viral factor. Using different human and mouse primary cells lacking STING, we show enhanced DENV replication. Conversely, mutated versions of STING that cannot be cleaved by the DENV NS2B3 protease induced higher levels of type I IFN after infection with DENV. Additionally, we show that DENV NS2B3 is not able to degrade the mouse version of STING, a phenomenon that severely restricts the replication of DENV in mouse cells, suggesting that STING plays a key role in the inhibition of DENV infection and spread in mice.

Viral infections have a vast impact on human health, resulting in hundreds of thousands of deaths yearly. To replicate and spread, these intracellular pathogens subvert the host cellular defense machinery. Dengue virus (DENV) is the most prevalent arbovirus in humans, and productively infects cells that are involved in the immune response, such as monocytes, B cells, macrophages and dendritic cells (DCs) among others [1, 2, 3, 4, 5] . Like most viruses, DENV has evolved in order to inhibit or evade different aspects of the innate immune system, the first line of human defense against microbes. DCs are antigen presenting cells (APCs) and some of the first cells that interact with the virus after the bite of an infected mosquito. Infection of these cells induces their activation, which results in their migration to the lymph nodes where the virus can infect other susceptible cells. The kinetics of infection of different cells in the immune system is not well documented, due to the lack of immune-competent mouse models for dengue disease. Nevertheless, in mice defective for type I IFN signaling, one of the most accepted current models for dengue disease, it has been shown that DCs and macrophages are productively infected by DENV [3, 4, 5, 6] reviewed in [7] . DENV is a single stranded RNA virus of positive polarity that, after entering the cytoplasm of the host cell, releases its genome and synthesizes a polyprotein using the cellular machinery, as a first event of the viral cycle. The DENV polyprotein is cleaved by the viral protease complex (NS2B3) and cellular proteases, including furin [8] . After this processing, some of the viral proteins have the ability to rearrange the ER membrane and create the micro-environment necessary for the production of de novo synthesized viral genomic RNA. During this event, DENV accumulates products with conserved molecular structures, like RNA with 59-triphosphate moiety or double stranded RNA, also referred to as pathogen associated molecular patterns (PAMPs). These foreign molecules are ligands of different cellular receptors engaged in their recognition, known as pattern recognition receptors (PRRs). PRRs are mainly divided into two different classes depending on their localization, associated with either the membrane or the cytoplasm. The Toll-like receptor (TLR) family is composed of membrane proteins with domains that are designed to detect extracellular PAMPs. On the other hand, the cytosolic DExD/Hbox RNA helicase proteins that contain caspase-recruiting domains (CARDs), referred to as RIG-I and MDA-5, can detect specific PAMPs present in the cytoplasm. These last two cytoplasmic sensors together with the TLR family members (TLR3/TLR7/TLR8) have been described so far as the most relevant DENV sensors [9, 10, 11] . After recognition of the mentioned PAMPs by the C-terminal helicase domain of RIG-I and MDA-5, these undergo a conformational change that exposes their CARD domains and promote the interaction with different down-stream molecules. One of the most well studied downstream molecules, referred as IPS-1 (also known as, MAVS, CARDIF or VISA), is located in the outer membrane of the mitochondria and transmits the signal via different molecules, including the tumor necrosis factor receptor associated factors 6 and 3 (TRAF6 and TRAF3) and the IkB kinase (IKK) family members (TBK1, IKKa, IKKb and IKKe) among other cellular factors [reviewed in [12] ]. Recently three different groups, using cDNA library screening of genes that induced the IFNb promoter, described an adaptor protein that localizes in the endoplasmic reticulum (ER). This protein was named as stimulator of the interferon gene (STING) [13] , mediator of IRF3 activation (MITA) [14] and endoplasmic reticulum IFN stimulator (ERIS) [15] . Also the same protein, referred to as MYPS, was previously identified as a mediator of anti-major histocompatibility complex II monoclonal antibody-induced apoptosis in B-lymphoma cells [16] . STING is highly expressed in several immune cells, including macrophages and DCs, as well as endothelial and epithelial cells [13] . This protein can interact with RIG-I and IPS-1, but not with MDA-5, and the signaling mediated by this adaptor is independent of the sensing by the TLR family members [17] . In two recent reports, it was documented that STING is involved in the pathway that mediates the detection of pathogens with DNA genomes [18] and has a role as a direct sensor of cyclic dinucleotides, a signaling molecule produced exclusively by bacteria and archea [19] . Activation of STING by some of these stimuli leads to its relocalization with TBK1 from the ER to perinuclear vesicles containing the subunit of the exocyst complex 5 (Sec5) followed by the phosphorylation of TBK-1 and the subsequent activation of the transcription factors IRF3/7 and NFkB, which translocate to the nucleus and complex with ATF2/c-Jun to induce the expression of type I IFN and pro-inflammatory cytokines [17] . A remarkable hallmark of highly virulent human pathogens is the ability, acquired through evolution, to inhibit this innate immune response by the expression of viral factors that affect one or several steps of the above described signaling cascade. Some of the most notorious examples are the influenza virus NS1 protein, that targets RIG-I for degradation, minimizing the sensing of influenza virus PAMPs by this PRR [20] or the Hepatitis C virus NS34A protease complex that cleaves the adaptor IPS-1 to interrupt the signaling that ends with the activation of IRF3, NFkB and the subsequent production of type I IFN in human hepatocytes [21] . Our group has documented that DENV is a weak inducer of type I interferon in human DCs, in particular when compared with other viruses that competently produce these cytokines in large amounts, such as Newcastle disease virus (NDV) [22] and Semliki Forest virus (SFV) [23] . This lack of type I IFN production by DCs infected with DENV results in an impaired ability of those DCs to prime T cells toward Th1 immunity, an effect that can be reversed by the addition of IFNb [24] . Nevertheless, DENV is able to induce the expression of some pro-inflammatory cytokines at early times post infection, which we hypothesize is a strategy used by this virus to attract more cells to the site of infection by allowing the expression of some chemo-attractants by infected cells. Our group described that the infection by DENV does not induce the phosphorylation of IRF3 in human primary cells, resulting in an inhibition of type I IFN production [24] . In a subsequent report, we examined the ability of DENV-infected DCs to respond to a variety of type I IFN-triggering signals using potent stimulators such as NDV, SeV, SFV, or TLR-3 ligand poly(I:C) [25] . This effect is viral dose dependent and takes place as early as 2 hours after DENV infection. We also showed that the inhibition of IFNa/b production after NDV infection in DENV-infected DCs is not a bystander effect, implying an active role of the DENVinfected DC population in the inhibition of IFNa/b. By using an NDV vector strategy to express the individual DENV nonstructural proteins (NS2A, NS2B3, NS4A and NS4B), we showed that only the recombinant NDV expressing the protease complex NS2B3 inhibited IFNa expression in infected MDDCs, as compared to NDV alone. Similar results were obtained using an IFNb promoter activity assay in 293T cells. Catalytically inactive NS2B3 mutants showed a diminished inhibition of this phenotype, which highlighted the important role for the protease activity of the NS2B3 protein as inhibitor of the type I IFN production. Interestingly, the proteolytic core of NS2B3, consisting of the last 40 amino acids of NS2B and the first 180 amino acids of NS3, was enough to reduce the activation of the IFNb promoter by a strong stimulus, such as Sendai virus (SeV) infection. DENV has also been shown to express inhibitors of the type I IFN signaling cascade [26] and has been shown to encode for at least four non-structural proteins NS2A, NS4A, NS4B and NS5 that target different components of this pathway. The most remarkable example is the proteasomal degradation of human STAT2 by the NS5, a phenomenon that does not occur in mouse cells, which makes mouse STAT2 a restriction factor for DENV replication in these animals [27, 28, 29, 30, 31] . In summary, DENV can successfully inhibit two fundamental steps of the innate immune system, both the inhibition of the type I IFN production and the signaling. In this way, DENV reduces the expression of hundreds of interferon inducible genes that would otherwise establish the antiviral state and control the spread of the infection in the host. In the present report, we describe the mechanism of inhibition of type I IFN production by DENV in primary human and mouse cells and identify the human adaptor molecule STING as a target Dengue virus (DENV) is a pathogen with a high impact in human health that replicates in a wide range of cells of the immune system. To efficiently infect humans, DENV must evade or inhibit fundamental elements of the innate immune system, namely the type I interferon response (IFN). Thus, DENV can inhibit type I IFN signaling (described by several groups), and type I IFN production (described by our group). We documented the inhibition of type I IFN production in human monocyte derived dendritic cells (MDDCs) with an otherwise strong cytokine and chemokine profile in those cells and that the NS2B3 protease complex of DENV functions as an antagonist of type I IFN production, and its proteolytic activity is necessary for this event. Here we identify the human adaptor molecule STING as a target of the NS2B3 protease complex and characterize the mechanism of inhibition of the type I IFN production in primary human MDDCs mediated by this viral factor. We also describe that DENV NS2B3 cannot degrade the mouse version of STING, a phenomenon that strictly restricts the replication of DENV in mouse cells, suggesting that STING plays a key role in the inhibition of DENV infection and spread in mice. of the DENV NS2B3 protease complex. We demonstrate that the proteolytic activity of this viral factor is crucial for the cleavage and degradation of STING and this phenomenon impairs the production of type I IFN in DENV infected cells. Furthermore, we show that DENV NS2B3 is not able to cleave the mouse version of STING. Using STING double knockout mouse embryonic fibroblast (MEFs) and human dendritic cells, we demonstrate the relevant role of this host factor in the restriction of DENV replication in mouse cells. This is the first report showing STING as a target for cleavage and degradation by a viral protein to inhibit innate immune responses and as a host restriction factor for virus infection in primary cells. Previous results from our laboratory showed that dengue virus inhibits type I IFN production in human primary dendritic cells and that this inhibition requires a proteolytically active NS2B3 protease complex [24, 25] . In order to identify potential NS2B3 targets we performed a bioinformatic search for potential DENV protease cleavage sites contained within members of the type I IFN pathway [32] . We identified putative cleavage sites in several known members of the type I IFN pathway (see table 1 ). After testing the factors shown in table 1 for their susceptibility to be cleaved by the DENV protease NS2B3, we observed that only STING was cleaved in our experimental set up (data not shown and figure 1). We have generated a wild type DENV-NS2B3, and a proteolytically inactive version (NS2B3-S135A), ( Figure 1A ) by direct mutagenesis [25] that were used to analyze the potential cleavage of STING by the DENV protease complex. When we compared the human amino acid sequence of human STING to its mouse counterpart we noticed that the putative NS2B3 cleavage site in hSTING which is situated at the beginning of transmembrane domain 3 ( Figure 1B ), is absent in mouse STING. In order to test the susceptibility of human and mouse STING proteins to proteolytic cleavage or degradation, we co-expressed a Cterminally HA-tagged STING alongside a wild type or catalytically active DENV-NS2B3, and a proteolytically inactive version (NS2B3-S135A) in 293T cells, ( Figure 1A ) and analyzed them by western blot (Figures 1C, 1D ). In the presence of WT NS2B3 we observed the full-length 42 kDa human STING and an additional band of about 32 KD, which is consistent with a C-terminal region product of a cleavage occurring within the first 96 aa of STING ( Figure 1C ). The additional band was not visible when the mouse version of STING or a catalytically inactive NS2B3 was used ( Figure 1C and 1D ). This putative cleavage site for the DENV NS2B3 lies very close to the conserved cysteine motif C88xxC91, or redox motif, recently described to be required for dimerization of STING and subsequent signaling in the type I IFN production pathway [33] . Regardless of their susceptibility to cleavage by the DENV NS2B3 complex, both the human and mouse versions of STING co-immunoprecipitated with the WT DENV NS2B3 complex and the proteolytically inactive mutant NS2B3 S135A ( Figure 1E , lanes 2, 3, 6 and 7). To map the putative cleavage site of human STING for the DENV NS2B3 complex, we mutated the sequence corresponding to the first three amino acids of the human site, RRG (shown in figure 1B as hSTING in red) with the sequence corresponding to the amino acids HCM found in the mouse version of STING (shown in figure 1B as mSTING). These recombinant versions of STING were co-transfected into 293T cells with the WT and mutant version of the NS2B3 protease and the ability of the DENV protease to cleave STING ( Figure 1F , lane 5) was drastically reduced when the mouse sequence was present in hSTING ( Figure 1F , lane 2). These data confirm the requirement for amino acids RRG for efficient cleavage of STING by the DENV NS2B3. However, replacement of the corresponding amino acid sequence of mouse STING (IHCM) by the human putative cleavage sequence (LRRG) does not render mouse STING susceptible to cleavage by the DENV protease ( Figure 1G , lane 2), suggesting that additional flanking amino acids are required for this cleavage. Altogether, these results strongly suggest that STING is a target for NS2B3 in human cells and possibly a restriction factor for DENV infection in the mouse. To test whether endogenous STING undergoes the same NS2B3-dependent processing as in overexpression experiments in 293T cells, we infected human MDDCs with DENV-2 (16681 strain) and analyzed the cell lysates by western blot at different time points ( Figure 1H ). Infection of MDDCs by DENV resulted in the degradation of STING that could be detected at 24 and 48 hours post infection (hpi) ( Figure 1H , lanes 9 and 10) which correlate with peak expression levels of the NS2B3 (as detected with NS3 specific antibodies). As expected, this degradation of STING was not observed in MDDCs treated with UV-inactivated DENV or mock treated cells ( Figure 1H , lanes1-5 and 11-15). These data demonstrate that DENV NS2B3-dependent cleavage of endogenous human STING occurs in cells relevant to DENV infection (MDDCs), and therefore has the potential to play a crucial role in inhibition of type I IFN production. We next investigated whether STING cleavage by DENV NS2B3 had an impact on its ability to mediate the signaling necessary for type I IFN production. We transfected 293T cells with either hSTING (Figures 2A and 2B ) or mSTING (Figures 2C and 2D) and the three different versions of the DENV protease: wild type, the proteolytically inactive version (NS2B3-S135A) and the proteolytic core (NS2Bh-NS3pro) alongside luciferase reporter constructs driven by either an IFNb promoter (IFNb-Luc) or by three IRF3/7 binding sites (p55-C1B-Luc) (kindly provided by Dr. Megan Shaw and shown in figure 2E schematically) [34] . As shown in figures 2A and 2B, cleavage of hSTING by the DENV NS2B3-WT and NS2Bh-NS3pro greatly inhibited its ability to activate both reporter constructs, while transfection of the mutant version of the protease did not. Conversely, the DENV NS2B3 had minimal or no impact on the ability of mSTING to induce activation of either of the reporter constructs used. We did not observe any DENV NS2B3-dependent inhibition when the human adaptor TBK1 was used as a positive control to induce the IFNb promoter, demonstrating that the inhibition during dengue infection occurs upstream of this adaptor (data not shown). Taking together, these data demonstrate that cleavage of hSTING by the DENV NS2B3 precludes the induction of type I IFN responses. Moreover, to validate the observed results in human primary cells, we used E. coli DNA, a known inducer of STING signaling [35] to treat human MDDCs previously infected with DENV, UV-inactivated DENV or mock treated (as shown schematically in figure 2F ). Figures 2G, 2H and 2I show that only live DENV but not UV inactivated DENV was able to inhibit the induction of IFNb, IFNa or ISG15 by this ligand in human MDDCs. To validate the results described using p55-C1B and IFNb promoter assays in a primary cell model, we measured IFNa/b production upon infection of MDDCs either with DENV or a Semliki forest virus (SFV) expressing the DENV NS2B3 protease complex or the mutant version of the protease (NS2B3-S135A) as a control [23, 25] . Consistent with our earlier report [24] , human MDDCs infected with DENV-2 (16681 strain) were unable to produce IFNa/b. Furthermore, SFV-NS2B3 induced significantly lower levels of IFNa/b mRNA than the SFV-NS2B3-S135A ( Figures 3A and 3B ). Interestingly, SFV-NS2B3-WT induced significantly higher expression of TNFa at early times postinfection compared to SFV-NS2B3-S135A control ( Figure 3C ). As shown previously, this would suggest an involvement of the NS2B3 protease complex in the expression of this pro-inflammatory cytokine [24] . As expected, the infection of MDDCs by DENV upregulated the expression of STING in these cells ( Figure 3D ). Figure 3E shows the kinetics of infection by DENV in MDDCs, with the peak of viral RNA at 48 hpi. In contrast, the SFV vectors used in these studies show low levels of viral RNA at late times after treatment ( Figure 3F ), since these vectors are replication deficient [36] . Then we infected mouse bone marrow-DCs (BM-DCs) using the same viruses, and analyzed the gene induction profile in those cells. As expected, infection of BM-DCs by DENV was rapidly controlled, consistent with the inability of DENV to infect mouse cells, and showed an opposite kinetic of viral RNA synthesis compared to the observed pattern in human DCs, with a modest induction of cytokines ( Figure 3G to 3K). SFV is an alphavirus that can replicate in mouse cells, and DCs in particular. In this context, the SFV-NS2B3 exhibited a higher induction of both IFNa/b genes compared to the SFV-NS2B3-S135A control, showing an opposite profile than that observed in human DCs, in which SFV-NS2B3-S135A induced higher levels of IFNa/b genes ( Figure 3G and 3H) . Again, the kinetics of infection of the SFV vectors in mouse DCs ( Figure 3L) show very low levels of viral RNA, consistent with their lack of productive infection in these cells [36] . The observed phenomenon agrees with the inability of recombinant DENV-NS2B3 to cleave mouse STING and decrease the activity of the IFNb and p55-C1B promoters induced by this adaptor protein ( Figure 2C and 2D). To explore STING's impact on the DENV replication in mouse cells, we used WT (Sting +/+ ) and STING double knockout (Sting 2/2 ) mouse embryonic fibroblasts (MEFs). First, we infected WT and Sting 2/2 MEFs with two different DENV-2 strains, 16681 and NGC (a strain that was obtained after several passages in mouse brain) [37] , with an MOI of 5. Then, we measured the ability of the two DENV-2 strains to induce IFNb production, to replicate in these cells and to release infectious particles from those cells ( Figure 4A-4F ). Both DENV-2 strains induced significantly higher levels of IFNb in WT MEFs as compared to the Sting 2/2 MEFs ( Figure 4A and 4D) , underlining the relevance of STING in the signaling of type I IFN upon the infection with DENV. Consistent with the observed low induction of IFNb, Sting 2/2 MEFs were permissive to DENV replication while, despite the high MOI used, replication of DENV 16681 and NGC was rapidly controlled in WT MEFs ( Figure 4B and 4E) . The production of infectious particles by the two DENV-2 strains in WT and Sting 2/2 MEFs was measured by plaque assay and shows that the KO MEFs were more permissive to DENV infection than the WT MEFs and have very different peaks of infection ( Figure 2C and 2F). To test whether our observation was independent from the high MOI and viral strains used, we repeated the infection using different DENV serotypes (DENV-2 16681 strain, DENV-3 PR-6 strain and DENV-4 H-241 strain) and an MOI of 1, with similar results (data not shown). These results are likely due to the inability of DENV to inhibit the type I IFN signaling in mouse cells and the establishment of the antiviral state [31] Mutation of cleavage site for DENV-NS2B3 restores the ability of STING to induce type I IFN production upon DENV infection To confirm the relevance of STING cleavage by the DENV-NS2B3 on the inhibition of type I IFN production upon DENV infection, we transduced Sting 2/2 MEFs with lentiviruses expressing either WT human STING (STING-WT) or a mutant (uncleavable) version, that harbors the mouse STING sequence at the NS2B3-cleavage site (STING-MUT) ( Figure 1B and 1F) . Twenty-four hours after transduction MEFs were infected with DENV-2 at an MOI of 1 (strains 16681 and NGC) or mock treated and the levels of IFNb, IFNa and viral RNA were measured by qRT-PCR after total RNA extraction from the cells at different times post infection ( Figure 5A-5F) . A schematic representation of the lentiviruses used is shown in figure 5G . For these experiments, as shown in figure 5A and 5D, MEFs expressing the uncleavable version of STING (STING-MUT) expressed significantly higher levels of IFNb mRNA when compared to MEFs expressing WT STING (STING-WT) or to the control MEFs (GFP) confirming that the cleavage of STING by DENV-NS2B3 is necessary for the inhibition of IFNb production in infected cells. The induction of IFNb was detected as early as 2 hpi. Infection with the two DENV-2 strains also induced higher levels of IFNa mRNA in MEFs expressing STING-MUT than in the MEFs expressing STING-WT and the GFP control ( Figure 5B and 5E ). Under these experimental conditions, the replication of the mouse adapted DENV-2 (NGC) was increased in Sting 2/2 MEFs expressing wild type STING as compared with STING-MUT ( Figure 5F ). In the case of 16681 strain, a significant increase of replication was observed only with Sting 2/2 MEFs (GFP) at 48 hpi., and no significant difference was observed in MEFs expressing the two versions of STING, presumably due to a high level of lentiviral-expressed STING that could overwhelm the ability of this non mouse-adapted DENV strain to replicate in this system ( Figure 5C ). As shown in figure 5 , the lack of STING cleavage was sufficient to increase the expression of type I IFN in MEFs infected with DENV. To validate these results in primary human cells, we transduced MDDCs from three different donors with the STINGexpressing lentiviruses and the GFP-only control ( Figure 5G ). STING transduced DCs were then infected with DENV-2 at an MOI of 5 and we assessed the production of IFNb in those cells. As shown in Figure 6A (showing one representative donor out of three), there were no significant differences between the levels of STING-WT and STING-MUT mRNA. This demonstrates that any difference in antiviral effect observed with the two different versions of STING is independent of the expression levels for these proteins. Upon DENV-2 infection, DCs expressing STING-MUT produced higher levels of IFNb when compared with STING-WT or GFP controls (showing statistical significance at 2 and 12 hpi). The induction of IFNb messenger RNA was detected as early as 2 hpi, which is in agreement with the results described with MEFs ( Figures 4, 5 and 6B ). These data confirm that detection of DENV infection by DCs takes place at early times post infection and that STING cleavage by DENV NS2B3 is fundamental to inhibit the signaling mediated by this adaptor in human cells. We next measured DENV replication kinetics and we found that viral RNA levels were significantly lower in DCs over-expressing STING-MUT when compared with STING-WT and GFP control ( Figure 6C ). Suggesting that the inability of DENV to cleave mutant STING and inhibit the induction of type I IFN has a direct impact on its replication kinetic and the accumulation of viral RNA in those cells. To determine STING's impact on DENV replication in primary human MDDCs, we used RNA interference (RNAi) to silence its endogenous expression. A decrease of STING mRNA level was observed when specific siRNAs were used compared to two scrambled control siRNAs ( Figure 7A ). As expected, the previously observed upregulation of STING after DENV infection was controlled by the STING siRNAs ( Figure 7A ). As a consequence, the reduction in STING expression resulted in an increase of DENV replication, illustrated in Figure 7B . When the viral progeny released by those infected MDDCs was quantified by plaque assay, the six donors treated with STING specific siRNA, showed viral production under this experimental conditions, however when scrambled siRNA was used, only three out of the six donors released detectable viral progeny in the supernatant ( Figure 7C ). Data shown in figures 7A and 7B correspond to donor 4 in figure 7C . Taken together, these data confirm that STING is a crucial restriction factor of DENV replication in human dendritic cells, since its silencing increases the levels of DENV replication in those cells. Different populations of cells were isolated from human blood and subsequently infected with DENV-2 at MOI of 1 and 12 h after infection supernatants were collected and RNA was extracted from cells. DENV-2 RNA was detected in all cells tested including plasmacytoid DCs (pDCs), B cells, blood circulating DCs (cDCs), monocytes as well as in monocyte-derived DCs (MDDCs) ( Figure 8A ). We also analyzed the cytokine and chemokine expression profile in all those cells by qRT-PCR (data not shown) and by multiplex ELISA (Figure 8B ) in the supernatants at 12 hpi. We observed a marked chemokine response (IL-8 and MIP1b) in monocytes, MDDCs, B cells and cDCs at this early time point, ( Figure 8B ). However pDCs did not show any significant chemokine profile after DENV-2 infection at this time point ( Figure 8B ). More interestingly, there was no significant type I IFN production observed in any of the cells tested by qRT-PCR (data not shown) or ELISA ( Figure 8B ). These data suggest that there is a coordinated and distinct kinetic of infection of DENV-2 in different cell populations in blood and there is a lack of type I IFN production in those cells after infection with this virus, at least at this early time point. The early time point of 12 hpi was chosen to obtain sufficient numbers of pDCs, since these cells have short half-lives and downregulate their specific cell surface markers in a rapid fashion. Nevertheless, we have previously reported that as early as 8 hpi, pDCs are able to produce type I IFN after infection with other viruses, such as NDV [24] . To rule out that the lack of type I IFN production resulted from lack of cell to cell interactions, we infected whole PBMCs with DENV-2 (MOI of 1) and 18 h after infection cell supernatants were collected. Figure 8C shows multiplex ELISA data of cell supernatants from those cultures. While there is a clear IL-8 response to DENV-2 infection in PBMCs, consistent with the strong IL-8 signature observed in sera from patients [38] , there was no detectable IFNa secretion from infected PBMCs ( Figure 8C ). This suggests that DENV-2 may inhibit type I IFN production in susceptible cells within those cultures. Macrophages have been shown to support DENV infection in animal models, and have been proposed to play an important role during early phases of dengue virus infection [39, 40] . We tested if monocyte-derived macrophages (MDMs) when infected with DENV were able to produce type I IFN. Figure 8D shows that macrophages are efficiently infected with DENV, with an early peak of replication and produce TNFa and IL-6 after DENV infection and after SFV expressing the WT and mutant versions of the NS2B3 DENV protease complex (Figures 8E, 8F ). Under these experimental conditions we were unable to detect IFNa released by macrophages after DENV infection. Interestingly, when compared to MDDCs infected with the same viruses ( Figure 8H) , macrophages produce at least 10-fold lower levels of IFNa after SFV infection, and the inhibitory effect of the DENV protease in this system was less apparent (8G and 8H). Activation of innate immunity due to the detection of viral replication products in the cell leads to the expression of hundreds of antiviral genes that controls the spread of the infection [12] . The inhibition of different steps implicated in these molecular pathways by viruses has been a matter of extensive study for several years. It has been demonstrated by others and by our group that DENV can inhibit both the production and signaling of type I IFN by the expression of viral proteins. In this way, DENV can mitigate the immune response induced by the host upon infection [24, 25, 29] . Here we have identified the human adaptor molecule STING as a protein with a predominant role in the recognition of DENV by the innate immune system. This adaptor protein was described to reside in the ER, a cellular organelle intimately related to the DENV replication process. Also, STING has been described as part of the TRAP (translocon associated protein) complex that can associate with RIG-I and IPS-1, two proteins with relevant roles in viral detection [41] . Ishikawa et al. also described an inhibition of the STING mediated IFNb production by the yellow fever virus (YFV) NS4B [42] . However, when we tried to replicate these results using the DENV NS4B, this viral protein was unable to decrease the induction of luciferase mediated by STING in an IFNb promoter assay (data not shown). By co-expression experiments of human STING with the DENV NS2B3 protease complex we observed a specific cleavage at (94-RRGA-99), a site described as a putative target for DENV NS2B3 [32] that generated a cleaved band of approximately 32 KDa (Figure 1C and 1F) . Interestingly, by analysis of the sequence alignment between human STING and its mouse version we, observed a drastic difference in the amino acid sequence in this region (94-HCMA-99) (shown in figure 1B ) and the inability to cleave the mouse STING by the DENV NS2B3 was confirmed by co-expression experiments ( Figure 1D and 1G) . Furthermore, the impact that the STING cleavage by NS2B3 had on the signaling of IFNb production pathway was subsequently demonstrated using IFNb and p55-C1B promoter systems ( Figure 2 ). In these experiments, a reduction in luciferase induction was only observed for human STING, suggesting that the cleavage confirmed by WB (showed in Figures 1C and 1F ) impaired the ability of this adaptor to induce IFNb. Recently, Jin et al. described a series of mutations in hSTING that were implicated in the activation/dimerization and subsequent induction of interferon. Interestingly, two cysteines located at C88XXC91 were fundamental for the proper induction of type I IFN after stimulation [33] . It could be interesting to investigate the presence of mutations at the cleavage site of STING for DENV NS2B3 in the human population, to identify a natural resistance to DENV infection. While this manuscript was under review, Yu and colleagues reported by overexpression experiments that the DENV protease can cleave the adaptor molecule MITA, [43] . In the present report we provide important data on the role of this adaptor molecule in primary human and mouse cells and during the context of DENV infection. We also confirmed the cleavage and degradation of STING by the DENV NS2B3 protease in human MDDCs in the context of DENV infection, since it is important to validate these findings in a relevant primary cell system and during virus infection ( Figure 1H ). In primary human cells as well as in mouse cells, such as MEFs and DCs, we also found that the presence of human STING allowed for greater DENV replication and the presence of mouse STING seemed to restrict DENV replication (Figure 3, figure 4 and figure 7) . We also show that the NS2B3 protease of DENV has specificity for the human STING and not for the mouse homologue of this protein ( Figure 5 and figure 6 ), suggesting that STING may be an important restriction factor in mice. Several viral proteases have been described as proteins that modulate cellular pathways, allowing many viruses to modify the intra and extracellular environment to promote optimal conditions for replication and spread. Some of the most remarkable characteristics observed at early times after infection by DENV are the lack of IFNa/b induction and a robust induction of proinflammatory cytokines like TNFa [24, 25] . As it was described in our previous work, DENV infection can abrogate IRF3 phosphorylation, but has no impact on NF-kB activity [25] . Using recombinant viruses expressing DENV-NS2B3 we observed a clear effect on the induction of TNFa in human DCs, similar to that observed with DENV infection (Figure 3 and figure 8) . Also, the inhibition of luciferase activity driven by p55-C1B promoter was considerably more efficient when compared with IFNbpromoter, since p55-C1B only harbors sites for IRF3/7 transcription factors, and IFNb-Luc has also has response elements for NFkb and AP-1 transcription factors (Figures 2B and 2D ). Furthermore, Ishikawa et al. overexpressed STING in 293T cells in the presence of different promoters driving the luciferase gene. Interestingly, STING stimulated IFNb promoter up to 400-fold, IRF3 response element (PRDIII-I-Luc) up to 1,000-fold, and NFkb responsive promoter (NF-kb-Luc) only up to 12-fold [41] . This observation suggested that STING is fundamentally involved in phosphorylation of IRF3, and under these experimental conditions showed a 100 fold less influence on NF-kb induction. Taken together, these observations suggest that DENV NS2B3 protease inhibits IFNb production by cleavage of the adaptor STING without modifying the observed NF-kb activity induced after infection by DENV. Further work exploring the impact that DENV-NS2B3 has on the induction of NF-kb activity in infected cells would confirm a putative role of this viral factor in the modulation of innate immune response by the induction of proinflammatory cytokines, a hallmark phenomenon observed during infection by DENV [44] . It is becoming increasingly clear that STING is a crucial adaptor in immune cells after infection with different viruses, such as HIV and DENV, among others [35, 45, 46] . These viruses require activation of their target cells in order to establish infection, or in the case of DENV to induce viremia in the host. Nevertheless, all viruses need to limit or inhibit the production of type I IFN in infected cells to avoid the establishment of an antiviral state in those cells. STING could be instrumental in these types of virus infections, since it can discriminate between the induction of type I IFN and the activation of the NFkB pathway. Along those lines, this report shows a novel mechanism of inhibition of IFN production by an RNA virus, namely DENV targeting STING. By inhibiting only type I IFN but not the NF-kB pathway, DENV induces a specific profile in infected human MDDCs and other susceptible primary cells that allow the virus to efficiently reach the lymph nodes and spread in the infected host, culminating in the production of viremia. All our experiments were performed in the context of primary infections with DENV, since we believe that the early events in primary infections dictate the quality of adaptive immune responses and the outcome of the infection. By targeting DCs and inhibiting the production of type I IFN in those cells, DENV may be able to efficiently modulate the generation of adaptive immune responses and establish infection in the host [24, 25] . It has been proposed that during DENV infection IPS-1 may be responsible for controlling early viral replication and type I IFN production [47] , while the IFN signaling pathway (JAK/STAT) may control late viral replication and type I IFN production in DENV infected cells [48] . It is possible that interactions between STING and IPS-1 [46] may be disrupted by the DENV NS2B3 targeting of STING. This mayinhibit type I IFN production early during DENV infection in susceptible cells, although the NS2B3 has not been shown to directly interact with IPS-1. Further experiments are required to understand these complex interplays between different signaling molecules in human primary cells infected with DENV. Our experiments demonstrate that primary human cells implicated in dengue virus infection, such as dendritic cells, macrophages, monocytes and B cells can support DENV replication, although at different levels. Interestingly, DENV infection did not induce type I IFN production in any of those human primary cells tested (Figure 8) . These different blood cells may play different roles during DENV infection in humans, such as being involved in the initial infection or in the final stage of viremia. Also, since the mouse models that support DENV replication and recapitulate dengue symptoms are deficient in type I IFN responses or are reconstituted with human immune cells [6, 49] , the inhibition of type I IFN in infected cells seems to be crucial for the establishment of infection by DENV. Our data on different cells from blood also show that the inhibition of type I IFN production by DENV is not a DC specific phenomenon ( Figure 8) . The inability of DENV to replicate in wild type mouse cells is well documented, and many attempts have been made to develop a competent animal model to study DENV infection [50] . The data presented in this manuscript, showing the ability of DENV to replicate in Sting 2/2 MEF (Figure 4) , open new approaches to develop a mouse model to study DENV infection and also highlights the requirement that type I IFN production has on the innate immune system and for the control of invading pathogens. Ashour et al. described the adaptor STAT2 as a restriction factor for DENV replication in mouse cells [31] . Based on our combined data, an interesting approach would be the development of a transgenic mouse model with humanized STING and STAT2. This approach could provide an immune competent mouse model for DENV that eliminates two of the potential bottlenecks that exist for DENV replication in mice. The animal protocol used in this study was reviewed and approved by the University of Miami Institutional Animal Care and Use Committee (IACUC) under IACUC protocol 11-181 ''Host Defense and the Regulation of Interferon Production: STING.'' The University of Miami has an Animal Welfare Assurance on file with the Office of Laboratory Animal Welfare (OLAW), National Institutes of Health. The assurance number is #A-3224-01, effective July 11, 2007 . Additionally, as of July 20, 2010, the Council on Accreditation of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC International) has continued the University of Miami's full accreditation. Vero, 293T and mouse embryonic fibroblast (MEFs), were cultured in Dulbecco's modified essential medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Baby hamster kidney cells (BHK) were grown in Glasgow minimal essential medium (MEM) supplemented with 10% FBS, and 20 mM HEPES. Mosquito cells derived from Aedes albopictus, clone C6/36, were expanded at 33uC in RPMI medium with 10% FBS. All media were supplemented with 100 U/ml of L-glutamine and 100 mg/ml of penicillin-streptomycin. All tissue culture reagents were purchased from Invitrogen. Dengue virus serotype 2 (DENV-2) strains 16681 and New Guinea C were used in this study. DENV was grown in C6/36 insect cells for 6 days as described elsewhere [51] . Briefly, C6/36 cells were infected at a multiplicity of infection (MOI) of 0.01, and 6 days after infection, cell supernatants were collected, clarified, and stored at 80uC. The titers of DENV stocks were determined by limiting-dilution plaque assay on BHK cells [52] . Semliki Forest virus (SFV) expressing GFP and DENV-NS2B3 were generated as described previously [36] and titrated in BHK cells by immunofluorescence [53] . Lentiviral vector constructs were built using conventional molecular biology techniques. Briefly, human STING cDNA was PCR amplified from pcDNA3.1 hSTING [13] and cloned into a lentiviral vector derived from pHR SIN CSGW [54] ( Figure 5G ). Mutations in the NS2B3 cleavage site at positions 94-96 of hSTING were obtained by overlap PCR. Residues RRG were changed to the corresponding murine sequence HCM. Lentiviral vector derived viruses were obtained by transfection of HEK 293T with 3 plasmids encoding STING, HIV-1 Gag-Pol, and VSV-G respectively [55] . Viral supernatants were harvested 48 and 72 hours post-transfection, 0.45 mm filtered, concentrated at 14,000 g for 6 hours over a 20% sucrose cushion and frozen at 280uC until used. Monocytes, pDCs, B cells and circulating CD11c + DCs (cDCs) were isolated from blood of healthy donors (New York Blood Center) using Miltenyi isolation kits. CD14 + clinimacs, for monocytes, CD123/BDCA4 kit for pDCs and BDCA1 kit for cDCs. B cells were isolated as part of the BDCA1 kit for isolation of cDCs according to manufacturers' instructions. The purity of each cell population was tested by flow cytometry as described below and was routinely 85-95% for CD14 + cells, 87-90% for pDCs, and 95-99% for both MDDCs and cDCs. Samples of 5610 5 isolated cell populations were infected with DENV-2, 16681 at the indicated MOI in a total volume of 100 ml of DC media for 1 hour at 37 C. Then, DC media supplemented with 4% HS was added up to a final concentration of 10 6 cells/ml and cells were incubated for the remainder of the infections at 37 C. At the indicated times, cell supernatants were collected and cell pellets were used for RNA extractions. Whole PBMCs were used after ficoll centrifugation and samples of 60610 6 PBMCs were infected with DENV-2 at MOI of 1 or left uninfected. After 1 hour DC media 4% HS was added. Eighteen hpi supernatants were collected and isolation of the different cell populations after DENV-2 infection was carried out as described above. Human MDDCs were obtained from healthy human blood donors (New York Blood Center), following a standard protocol as previously described [24] and described above. Briefly, after Ficoll-Hypaque gradient centrifugation, CD14+ cells were isolated from the mononuclear fraction using a MACS CD14 isolation kit (Milteny Biotec) according to the manufacturer's directions. CD14+ cells were then differentiated to naïve DCs by incubation during 5 to 6 days in DC medium (RPMI supplemented with 100 U/ml L-glutamine, 100 g/ml penicillin-streptomycin, and 1 mM sodium pyruvate) with the presence of 500 U/ml human granulocyte-macrophage colony-stimulated factor (GM-CSF) (Pe-proTech), 1,000 U/ml human interleukin 4 (IL-4) (PeproTech), and 10% FBS (Hyclone). To generate MDMs, monocytes were cultured in the presence of 2000 U/ml human granulocytemacrophage colony-stimulated factor (GM-CSF) for 10 days, and media was replenished (with same concentration of GMCSF) at days 2, 5 and 8. The purity of each cell population was confirmed by flow cytometry analysis and was at least 99% for MDDCs and 95% for MDMs. Femurs and tibia of wild-type C57BL/6 mice (Jackson) were soaked in 70% ethanol, washed with RPMI (Invitrogen), and epiphyses were cut to expose the bone marrow. The bones were flushed with RPMI supplemented with 10% fetal bovine serum (FBS) to extract the bone marrow. Cells were pelleted by centrifugation, washed once with RPMI and resuspended in ammonium chloride red blood cell lysis buffer. RBC lysis was performed for 1 minute at room temperature, stopped with RPMI-FBS and cells were collected by centrifugation. Bone marrow cells were seeded in 6-well dishes in RPMI containing 10% FBS, 50 U/ml Penicillin (Invitrogen), 50 mg/ml Streptomycin (Invitrogen), 20 ng/ml GM-CSF (Peprotech), 10 ng/ml IL-4 (eBioscience), and 40 mM beta-mercaptoethanol (BIO-RAD). Cells were cultured for 5 days at 37 degrees C, 5% CO2 and fresh media was added every 2-3 days. Human and mouse DCs were obtained as described above, and at day 5 of culture, samples of 1610 6 cells were resuspended in 100 ml of DC-medium and were infected for 45 min at 37uC with the indicated MOI of virus (diluted in DC media) or with DC medium (mock group) in a total volume of 200 ml. After the adsorption period, DC medium supplemented with 10% FBS was added up to a final volume of 1 ml, and cells were incubated for the appropriate time at 37uC. siRNA transfection and Dengue infection of human MDDCs 2.5610 4 MDDCs were seeded per well in 96 well plates and transfected with the corresponding siRNA using the StemFect RNA transfection kit (Stemgent), according to the manufacturer instructions. Chemically synthesized 27mer siRNA duplexes were obtained from OriGene Technologies, Inc. The sequences of the STING siRNA oligonucleotides used in this study are as follows: siSTING -1:59-rGrGrCrArUrGrGrUrCrArUrArUrUrArCrAr-UrCrGrGrArUAT-39. siSTING-2:59-rArCrCrUrGrUrGrArArArUrGrGrGrArUrCr-ArUrArArUrCAC-39. siSTING-3:59-rGrGrArUrUrCrGrArArCrUrUrArCrArArUr-CrArGrCrArUTA-39. Two random non-coding control siRNA were used: Sc-1 (59-rUrArCrGrUrArCrUrArUrCrGrCrGrCrGrGAT-3) from Qiagen, and Sc-2 (universal scrambled negative control siRNA duplex SR30004) from OriGene Technologies, INC. 48 h after transfection, cells were infected with Dengue virus at an MOI of 1. Briefly, cells were centrifuged (4006g, 10 min), the media was removed and 25 ml of RPMI containing the appropriate amount of virus was added and the plates were incubated for 45 min at 37uC. Then, 75 ml of RPMI with 10% FBS were added and cells were incubated at 37uC for the indicated hours. Subsequently, cells were recovered by centrifugation for 10 min at 4006g, and the cell pellets were lysed for RNA isolation. Plated monocytes were transduced as previously described [56] . In brief, freshly isolated monocytes were transduced with VSV-G pseudo-typed SIV VLPs and each lentiviral vector construct for 3 h by spinoculation, in the presence of 2 mg/mL polybrene (Sigma). Subsequently, cells were washed, resuspended in regular growth medium described before for the generation of monocytes derived human dendritic cells and incubated for 5 days at 37uC until stimulation. At day 5 post-transduction, MDDCs were infected with DENV as described before, and cell pellets were collected at the indicated time points, and lysed for RNA isolation. 293T cells were transfected by using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. A type I IFN production antagonist assay was performed as described previously [25] using IFNb-Luc and p55-C1B-Luc [34, 57] . 293T cells seeded on 24-well plates were transiently transfected with 50 ng of the luciferase reporter plasmid together with a total of 400 ng of various expression plasmids or empty control plasmids. As an internal control, 50 ng pRL-TK was transfected simultaneously. Then, 24 or 48 h later, the luciferase activity in the total cell lysate was measured by using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's directions. Transfection of 293T cells and infection of human DCs was performed as described above. Cell lysates were obtained after incubation of cells with RIPA lysis buffer (Sigma Aldrich) supplemented with complete protease inhibitor (Roche) and resuspended in a total of 50 ml of Laemmli sample buffer (Bio-Rad). Crude lysates were either boiled for 10 min or incubated at 42uC for 20 min and then kept on ice. Each sample was loaded in a polyacrylamide-SDS gel, and the proteins were electrophoretically separated by conventional methods. Proteins were transferred to nitrocellulose, and blots were incubated with anti-HA, anti-FLAG, anti-Actin anti-GAPDH (Sigma Aldrich) and rabbit polyclonal antibodies anti-hSTING [41] and anti-DENV NS3 (kind gift of Dr. Andrea Gamarnik), and developed using SNAP ID detection system (Millipore), following the manufacturer's instructions. Antibody-protein complexes were detected using a Western Lighting chemiluminescence system (Perkin Elmer). RNA from different cells was extracted using Trizol (Invitrogen), followed by a treatment with DNase using DNA-free Ambion. The concentration was evaluated in a spectrophotometer at 260 nm, and 500 ng of RNA were reverse transcribed using the iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. Evaluation of the expression of human and mouse cytokines from different cell types and viral RNA was carried out using iQ SYBR green Supermix (Bio-Rad) according to the manufacturer's instructions. The PCR temperature profile was 95uC for 10 min, followed by 40 cycles of 95uC for 10 s and 60uC for 60 s. Expression levels for individual mRNAs were calculated based on their CT values using two different housekeeping genes (human: rps11 and a-tubulin genes) and (mouse: S18 and b-Actin) to normalize the data. One paired two tailed Student's t-test was used to analyze data. Data considered significant demonstrated p values less than 0.05.

Acute Reactogenicity after Intramuscular Immunization with Recombinant Vesicular Stomatitis Virus Is Linked to Production of IL-1β

Vaccines based on live viruses are attractive because they are immunogenic, cost-effective, and can be delivered by multiple routes. However, live virus vaccines also cause reactogenic side effects such as fever, myalgia, and injection site pain that have reduced their acceptance in the clinic. Several recent studies have linked vaccine-induced reactogenic side effects to production of the pro-inflammatory cytokine interleukin-1β (IL-1β) in humans. Our objective was therefore to determine whether IL-1β contributed to pathology after immunization with recombinant vesicular stomatitis virus (rVSV) vaccine vectors, and if so, to identify strategies by which IL-1β mediated pathology might be reduced without compromising immunogenicity. We found that an rVSV vaccine induced local and systemic production of IL-1β in vivo, and that accumulation of IL-1β correlated with acute pathology after rVSV immunization. rVSV-induced pathology was reduced in mice deficient in the IL-1 receptor Type I, but the IL-1R−/− mice were fully protected from lethal rechallenge with a high dose of VSV. This result demonstrated that IL-1 contributed to reactogenicity of the rVSV, but was dispensable for induction of protective immunity. The amount of IL-1β detected in mice deficient in either caspase-1 or the inflammasome adaptor molecule ASC after rVSV immunization was not significantly different than that produced by wild type animals, and caspase-1−/− and ASC−/− mice were only partially protected from rVSV-induced pathology. Those data support the idea that some of the IL-1β expressed in vivo in response to VSV may be activated by a caspase-1 and ASC-independent mechanism. Together these results suggest that rVSV vectors engineered to suppress the induction of IL-1β, or signaling through the IL-1R would be less reactogenic in vivo, but would retain their immunogenicity and protective capacity. Such rVSV would be highly desirable as either vaccine vectors or oncolytic therapies, and would likely be better tolerated in human vaccinees.

Vaccines based on attenuated viral vectors (e.g. poxviruses, adenovirus) are highly immunogenic, relatively inexpensive to produce, and can be delivered by multiple routes. These advantages have accelerated the development of virus-vectored vaccines, especially as concerns about emerging infectious diseases such as avian influenza and multi-drug resistant tuberculosis increase. However, pre-existing anti-vector immunity can significantly impair the ability to raise immune responses to vaccination in the pre-immune host. That suggests that it will be advantageous to prioritize development of viral vaccine vectors to which preexisting immunity is rare in the human population. Vaccine vectors based on recombinant vesicular stomatitis virus (rVSV) are highly immunogenic and can protect small animals and nonhuman primates against a range of infectious diseases [1, 2, 3, 4, 5] . VSV is also a potent oncolytic agent, and has been shown to be superior to nine other viruses tested against multifocal glioma [6, 7] and other cancers [8, 9] . Natural infection of humans with VSV is extremely uncommon, and therefore pre-existing immunity to rVSV vectors is almost non-existent [10, 11, 12] . Despite these advantages, development of rVSV for human clinical use continues to be delayed because of concerns about vectorassociated pathology. Small laboratory animals immunized with rVSV vaccines display clinical signs of acute illness and lose up to 20% of their pre-immunization body weight in the first four days after immunization [13, 14] . Although most animals recover, undesirable side effects such as these would be unacceptable in human vaccinees. Weight loss is less pronounced in non-human primates immunized with rVSVs, but the one human to date therapeutically immunized with a single cycle non-replicating experimental rVSV vaccine against Ebola virus developed a high fever and transient viremia in the first 24 hours after vector administration [15] . That result demonstrated that even single cycle vectors can induce significant adverse side effects, and suggested that alternate strategies to reduce residual viral vector reactogenicity are required. For rVSV-based vaccines to move forward to clinical use it is important to identify ways to eliminate rVSV vector-associated pathology that do not at the same time compromise vector immunogenicity or oncolytic properties. It has been shown previously that TNF-a produced in response to intranasal immunization with rVSV vaccine vectors contributes to, but is not the only factor responsible for, rapid weight loss after rVSV immunization [14] . In those studies, TNF-a deficient mice were partially protected from acute pathology after VSV challenge, and TNF-a was not required for the generation of humoral immune responses to rVSV [14] . We undertook the present study to expand on those observations, focusing on the importance of IL-1. We addressed intramuscular immunization, because that is the most likely route of administration for rVSV vectors in humans. Mice deficient in the interleukin-1 receptor (IL-1R2/2) are protected from weight loss after immunization with VSV vaccine vectors Mice immunized with recombinant VSV (rVSV) vaccine vectors lose weight rapidly after immunization. Weight loss is usually maximal by the second day after immunization but most animals recover, with weight returning to normal by five to seven days after immunization. Interleukin-1b (IL-1b) is a pro-inflammatory cytokine that causes acute weight loss and fever in mice and humans [16, 17, 18] , and which stimulates the production of other pro-inflammatory mediators such as IL-6 [17, 19] and prostaglandin E2 (PGE 2 ) [20, 21] . To determine whether IL-1 can contribute to the induction of acute pathology after administration of rVSV vaccine vectors, we immunized groups of C57BL/6 wild type (WT, n = 4) or IL-1 receptor deficient (IL-1R2/2, n = 3) mice with a single intramuscular injection of 5610 8 PFU live replication-competent rVSV. All animals survived the immunization, but wild type mice lost approximately 10% of their preimmunization body weight by the first day after immunization ( Figure 1 ) and did not fully regain their pre-immunization weight until eight days after immunization. In contrast, IL-1R2/2 mice had minimal pathology, losing only a small amount (,5%) of weight on the first day after immunization and returning to their starting weight by the second day after immunization. IL-1R2/2 mice lost significantly less weight than wild type control animals after challenge (P = 0.03 from day 1-day 4 after challenge, Mann-Whitney test). Because mice deficient in the IL-1R cannot respond to IL-1, this result demonstrated that either IL-1a or IL-1b, both of which bind to the IL-1R, contributed to acute pathology after rVSV immunization. Mice immunized with VSV vaccine vectors produce IL-1b systemically and at the injection site Both IL-1a and IL-1b are regarded as pro-inflammatory cytokines, but IL-1b rather than IL-1a has been more commonly associated with symptoms such as weight loss and fever that are induced by infection with a live virus or other immune stimulus [16, 17] . To determine whether intramuscular immunization with VSV vaccine vectors induced production of IL-1b in vivo we challenged wild type C57BL/6 mice with rVSV and used an ELISA to quantitate the amount of IL-1b produced locally (at the injection site and in the draining lymph node) and systemically (in the blood). As shown in Figure 2A -C, mice immunized intramuscularly with rVSV had accumulations of IL-1b in the blood, in the quadriceps muscle, and in the popliteal lymph node that drains the quadriceps at 12 and 24 hours after immunization with VSV rwt. These results confirmed that IL-1b was produced in vivo both locally and systemically after intramuscular rVSV challenge, and was consistent with the results of Poeck et al. who found that wild-type mice injected intravenously with rVSV had detectable levels of IL-1b in the serum at six hours after challenge [22] . Our results were also consistent with the hypothesis that IL-1b caused the acute pathology that occurs after immunization with rVSV vaccine vectors, and suggested that reducing the production of IL-1b after rVSV immunization might correspondingly reduce reactogenicity and other undesirable side effects of vaccination. Mice deficient in the IL-1R control rVSV replication, have normal cellular and humoral immune responses, and are immune to high dose re-challenge IL-1b causes some deleterious effects such as fever and anorexia, but has also been shown to contribute positively to the generation of adaptive immune responses induced by live virus infection [23] . The role of IL-1 in control of VSV replication has not been examined previously, and reports describing the role of IL-1 in control of other viruses in vivo are not in agreement. Because our ultimate goal is to be able to generate less reactogenic rVSV vectors by attenuating the host inflammatory response, it was important to determine whether IL-1 was required for the control of vaccine vector replication, generation of humoral and cellular immune responses, or to generate protection against rechallenge. To investigate these questions we intramuscularly immunized mice deficient in the IL-1 receptor (IL-1R2/2) with 5610 8 PFU of rVSV and measured viral loads, induction of antibody and T cell responses, and protection from rechallenge. As shown in Figure 3A , viral loads in the quadriceps muscle (injection site) were not significantly different between wild type and IL-1R2/2 mice (n = 3 per group) at 24 hours after infection, which is the peak of VSV replication in vivo. Also, virus was not detected 2) are protected from acute weight loss after intramuscular immunization with rVSV. Eight to ten week old female C57BL/6 mice were immunized with a single intramuscular injection to the left rear quadriceps of 5610 8 PFU of live replicating rVSV. Graph shows average percent initial weight by day for each group of mice beginning on the day of immunization (Day 0). IL-1R2/2 mice (n = 3) lost less weight and recovered more quickly than wild type (WT, n = 4) mice after immunization. The difference in weight loss was significant from day 1-day 4 after challenge (P = 0.03, Mann-Whitney test). This experiment has been performed two times with consistent results. doi:10.1371/journal.pone.0046516.g001 in the blood of any infected animal. Those results demonstrated that IL-1 is not required for the control of VSV replication in vivo. As shown in Figure 3B , serum neutralizing antibody titers against VSV were not significantly different in IL-1R2/2 (n = 4 per group per timepoint) and wild type mice (n = 5 per group per timepoint) at any time after immunization. When we used an MHC Class I tetramer to measure CD8 T cell responses to an immunodominant H-2 K b -restricted epitope (N-RGYVYQGL-C) in the VSV N protein [24] , IL-1R2/2 mice had slightly fewer anti-VSV N specific CD8 T cells in the blood than did wild type animals ( Figure 3C ) at 14 and 28 days after immunization, although the difference was only statistically significant at 14 days after immunization (P = 0.03, Two-tailed T test). Finally, to determine whether these immune responses were sufficient to protect wild type and IL-1R2/2 mice from re-challenge with rVSV, we challenged all animals intranasally with 1610 8 PFU of rVSV at eight weeks after the primary infection. As shown in Figure 3D , pre-immune wild type and pre-immune IL-1R2/2 mice were fully protected from rechallenge, while naïve wild type animals (n = 7, open triangles, Figure 3D ), lost up to 20% of their pre-infection body weight. Two out of the seven naïve animals succumbed to infection. Together, these results indicated that IL-1 was not required either for the control of rVSV replication in vivo, or for the generation of protective anti-VSV immune responses. The results also supported the idea that suppressing the production of or response to IL-1b in response to rVSV vector administration would not render rVSV vectors unsafe or non-immunogenic. Mice deficient in inflammasome adaptor molecule ASC are partially protected from acute weight loss after immunization with rVSV vaccine vectors Because the ultimate goal of these studies was to devise strategies by which we could suppress IL-1b production and thereby reduce the pathology of rVSV vectors in vivo, we sought to determine the mechanism by which IL-1b was being produced in response to VSV. IL-1b is synthesized as an inactive precursor molecule (pro-IL-1b), which must be cleaved either intracellularly by endogenous protease caspase-1 [25, 26, 27] , or extracellularly by matrix metalloprotease 9 [28] or other neutrophil [29] and mast cell-associated proteases [30] to become biologically active. It was shown recently that murine bone marrow derived dendritic cells (BMDC) infected with VSV in vitro produce IL-1b via formation of an inflammasome composed of RNA helicase RIG-I, adaptor molecule ASC, and caspase-1 [22] . Because the authors did not test whether RIG-I, ASC, and caspase-1 were required to produce IL-1b in response to VSV in vivo we used caspase-1-deficient and ASC-deficient mice to determine the effects of the absence of these molecules on acute pathology after rVSV immunization. It was not practical to test RIG-I deficient mice for IL-1b induction, because RIG-I deficient mice do not produce IFN in response to VSV and therefore rapidly succumb to infection [31] . We challenged ASC-deficient mice (ASC2/2) and wild type C57BL/6 mice intramuscularly with 5610 8 PFU rVSV as in Figures 1, 2 , and 3 and measured production of IL-1b and acute pathology after immunization. As shown in Figure 4A , the systemic and local production of IL-1b was not significantly reduced in ASC2/2 mice relative wild type control mice immunized in parallel. Consistent with that result, and with our prediction that IL-1b induces acute weight loss after rVSV infection, ASC2/2 (n = 5) mice lost slightly less weight after rVSV challenge than did wild type C57BL/6 (n = 6) mice infected in parallel, with the difference only reaching statistical significance only on the second day after challenge (P = 0.004, Mann Whitney Test, Figure 4B ). A second experiment replicated these findings almost exactly, with ASC2/2 mice (n = 6) showing slightly enhanced protection relative wild type animals (n = 10, Figure S1 ). The difference in weight loss between wild type and ASC2/2 mice was small but highly reproducible, and ASC2/2 mice were never protected to the same extent as IL-1R2/2 mice. To confirm that observation, we infected the three groups in parallel (5610 8 PFU rVSV intramuscular). As shown in Figure 4C , IL-1R2/2 mice (n = 4) lost significantly less weight than wild type (n = 5) mice (days 1-4 after challenge P,0.05 via one way ANOVA with Bonferroni multiple comparison test) and recovered their pre-immunization body weight more quickly than either wild type or ASC2/2 mice. The difference in weight loss between IL-1R2/2 and ASC2/2 mice was significant on the first and second day after challenge (P,0.05 via one way ANOVA with Bonferroni multiple comparison test). Similar to the results obtained in IL-1R2/2 mice, ASC2/2 mice made equivalent humoral responses and slightly reduced cellular responses to VSV, and were fully protected from high dose rechallenge ( Figure 4D -F). Taken together, these results demonstrated that production of IL-1b in vivo after intramuscular rVSV immunization occurred independent of inflammasome adaptor molecule ASC. Because ASC deficient mice were not protected fully from acute pathology after rVSV immunization, strategies that suppress the function of ASC would be predicted to partially but not completely abrogate acute pathology after rVSV immunization. Mice deficient in caspase-1 are partially protected from acute weight loss after immunization with rVSV vaccine vectors Similarly, when we immunized mice deficient in caspase-1 (caspase-12/2) or wild type mice with 1610 9 PFU of rVSV, caspase-1 deficient mice did not have significantly reduced levels of IL-1b relative wild type mice ( Figure 5A ) when we measured IL-1b production at the injection site and in the draining lymph node by ELISA. Consistent with those data, caspase-1 deficient mice immunized intramuscularly with 5610 8 PFU of rVSV as in Figure 4B -C (n = 5 per group), were partially but not completely protected from acute weight loss ( Figure 5B ) relative wild type mice (n = 5), with the difference in weight loss between wild type and caspase-12/2 mice being significant only on days 2 and 3 after challenge (P,0.05, Mann Whitney test). Caspase-1 deficient mice controlled viral replication as well as wild type mice ( Figure 5C , n = 6 per group), with no significant difference in viral loads in the quadriceps muscle of infected animals at 24 hours after infection. As with IL-1R2/2 mice, no virus was recovered from the blood of infected mice. Caspase-1 deficient mice made robust humoral ( Figure 5D ) and cellular ( Figure 5E ) responses to VSV, which were not significantly different than those of wild type mice. We undertook this study with the goal of determining which cytokines induced by rVSV contribute to acute pathology after intramuscular immunization. This is important because while VSV is a highly promising vaccine vector and oncolytic agent, its clinical development has lagged behind that of other live viruses because of concerns about vector-associated pathology. Fever, myalgia, and the ''sickness response'' are induced by many live viral or bacterial vaccines [32, 33, 34, 35, 36] , and these vaccineinduced side effects are among the leading reasons why some individuals elect not to receive protective vaccines [37, 38, 39] . Consistent with our results obtained in the mouse model and presented here, several recent studies have correlated increased levels of IL-1 or other pro-inflammatory cytokines with the induction of high fevers or other adverse events in response to live There was no significant difference in viral loads between the two groups. In Panels B-D, adult female wild type (n = 5) or IL-1R2/2 (n = 4) mice were immunized intramuscularly with a single injection of 5610 8 PFU rVSV in the rear quadriceps. At the indicated timepoints after infection mice were bled and humoral and cellular immune responses were assayed. Panel B shows average anti-VSV neutralizing antibody responses by group as measured by microneutralization assay. Error bars represent the upper and lower limits of the 95% confidence interval. Panel C shows average percent CD8 T cells specific for the VSV N1 epitope as measured by MHC Class I tetramer. At 14 days after immunization, WT mice had significantly more (P = 0.03, Twotailed T test) VSV N specific CD8 T cells than IL-1R2/2 mice, but by day 28 the difference was no longer significant (P = 0.13). At eight weeks after the primary infection all mice were challenged intranasally with a semi-lethal dose of rVSV (1610 8 PFU). A cohort of naïve wild type mice (n = 7) was challenged at the same time. All pre-immune mice had robust immunity to rechallenge (Panel D) and did not lose weight or exhibit other signs of pathology. Two of the naïve mice succumbed to infection. Days on which naïve animals succumbed are indicated with an asterisk on the graph. doi:10.1371/journal.pone.0046516.g003 Figure 4 . Mice deficient in the inflammasome adaptor ASC (ASC2/2) are partially protected from acute weight loss after intramuscular immunization with rVSV. Groups of adult female C57BL/6 wild type or ASC2/2 mice were immunized with two injections (5610 8 PFU per injection) of rVSV in each rear quadriceps, or were sham inoculated with sterile PBS. At 12 and 24 hours after infection mice (n = at least 4 per timepoint), were sacrificed and IL-1b in the blood (Panel A left), draining popliteal LN (Panel A middle), and quadriceps muscle (Panel A right) was quantitated via ELISA. The amount of IL-1b produced by wild type and ASC2/2 mice was not significantly different at any time or in any organ. The comparison of IL-1b production by wild type and ASC2/2 mice has been performed twice with consistent results. Panel B shows average percent initial weight for wild type (n = 6) and ASC2/2 (n = 5) mice after intramuscular challenge with 5610 8 PFU of rVSV. The comparison of WT and ASC2/2 mice has been performed four times with consistent results. Panel C shows average percent initial weight for wild type (n = 5), ASC2/2 (n = 5), and IL-1R2/2 (n = 4) mice infected with rVSV. IL-1R2/2 mice lost significantly less weight than wild type (days 2-4) or ASC2/2 mice (days 1-2) (P,0.05 via one way ANOVA with Bonferroni test). The comparison of WT, IL-1R2/2, and ASC2/2 mice has been performed twice with consistent results. Panel D shows average serum neutralizing titers for WT (n = ) and ASC2/2 (n = ) mice after primary immunization with VSV. There were no significant differences in neutralizing titer between the two groups. Panel E shows average percent CD8 T cells specific for the VSV N1 epitope as measured by MHC Class I tetramer. At 14 days after immunization, WT mice (n = 9) had significantly more (P = 0.001, Two-tailed T test) VSV N specific CD8 T cells than ASC2/2 mice (n = 9), but by day 28 the difference was no longer significant (P = 0.07). At eight weeks after the primary infection preimmune WT (n = 10) and ASC2/2 (n = 6) mice were challenged intranasally with a semi-lethal dose of rVSV (1610 8 PFU). A cohort of naïve wild type mice (n = 5) was challenged at the same time. All pre-immune mice had robust immunity to rechallenge (Panel F) and did not lose weight or exhibit other signs of pathology. One of the naïve mice succumbed to infection. doi:10.1371/journal.pone.0046516.g004 Figure 5 . Mice deficient in caspase-1 are partially protected from acute weight loss after intramuscular immunization. Groups of adult C57BL/6 wild type or caspase 12/2 mice were immunized with two injections (5610 8 PFU per injection) of rVSV in each rear quadriceps, or were sham inoculated with sterile PBS. At 24 hours after infection mice (n = 2-3 mice per timepoint), were sacrificed and IL-1b in the draining popliteal LN (Panel A left) and quadriceps muscle (Panel A right) was quantitated via ELISA. The amount of IL-1b produced by wild type and caspase 12/2 mice was not significantly different in either organ. The comparison of IL-1b production by wild type and caspase 12/2 mice has been performed twice with consistent results. Panel B shows average percent initial weight for wild type and caspase 12/2 mice (n = 5 per group) after intramuscular challenge with 5610 8 PFU of rVSV. The comparison of WT and caspase 12/2 mice has been performed twice with consistent results. Caspase 12/2 mice lost significantly less weight than wild type controls on days 2 and 3 after challenge (P,0.05 via Mann Whitney test). Panel C shows average viral loads in the quadriceps muscle of infected mice (n = 6 per group). Data is compiled from two identically performed experiments. Viral loads in WT and caspase 12/2 mice were not significantly different. Panel D shows average serum neutralizing antibody titers for wild type and caspase 12/2 mice immunized with rVSV (n = 5 per group per timepoint). Error bars represent the upper and lower limits of the 95% confidence interval. Panel E shows average percent 6 SEM of CD8 T cells in the blood binding to an MHC Class I tetramer recognizing an immunodominant epitope within VSV N. There were no significant differences in the antibody or CD8 T cell responses between the two groups at any time. The comparison of humoral and cellular immune responses in WT and caspase 12/2 mice has been performed twice with consistent results. doi:10.1371/journal.pone.0046516.g005 virus vaccination in humans [36, 40] , or have identified genetic polymorphisms in the IL-1 gene which predispose individuals to severe adverse events after receiving live virus vaccines [41] . For these reasons it is important to determine the ways in which proinflammatory cytokines contribute to reactogenicity of VSV vectors in particular, not only because those findings could advance development of VSV-based therapeutics, but also because our findings might help to inform the development of other live virus vaccines and oncotherapies. Previous studies in which VSV vectors were delivered to mice intranasally showed that intranasal immunization with VSV induces the pro-inflammatory cytokine TNF-a, and that TNF-a production directly correlated with weight loss and acute pathology [14] . Although intranasal immunization has many advantages (needle free delivery, induction of mucosal immunity, etc), a significant drawback to intranasal immunization with rVSV is the risk of neurotropic spread of the virus. VSV instilled in the nose rapidly colonizes the olfactory neurons, and migrates into the brain [42] . The neurovirulence of VSV vectors has been significantly reduced by attenuating the capacity of the VSV vector to replicate [43] , but even a remote chance of neurotropic spread will likely prevent use of the intranasal route for human inoculations. Therefore, we decided to determine which cytokines were responsible for acute pathology after intramuscular immunization, which is regarded as a safer route by which to administer potentially neurotropic agents. We show here that mice deficient in the interleukin-1 receptor (IL-1R) are significantly protected from weight loss after intramuscular challenge with VSV. Although these results do not preclude the contribution of other inflammatory processes to pathology, they do positively identify IL-1 as an important target for intervention. The IL-1 receptor binds IL-1a and IL-1b. Therefore because IL-1R2/2 mice were protected from pathology, it was possible that IL-1a, IL-1b, or both cytokines contributed to VSVassociated pathology in vivo. IL-1b has been well characterized as a mediator of acute inflammatory responses in mice and humans, namely the induction of fever and cachexia, the acute phase response, and upregulation of other inflammatory mediators such as IL-6 in response to infectious or non-infectious stimuli [44, 45] , reviewed in [46] . IL-1b is more commonly associated with these pathologies than is IL-1a. For example, IL-1b deficient mice, but not IL-1a deficient mice, are protected from fever after injection of turpentine [16, 17] . Similarly, IL-1b, but not IL-1a, is the etiologic agent of hereditary periodic fever syndromes and other ''autoinflammatory'' diseases in humans [47, 48] . IL-1a is more commonly associated with ''sterile'' inflammation that accompanies apoptotic cell death [49] , and with the exception of adenovirus mediated inflammation [50] has not been associated with pathology arising from viral infection. For these reasons, we predict that IL-1b is the primary mediator of pathology in our system, but further experiments will be required to formally exclude a role for IL-1a. Once we had established that IL-1 contributed to acute pathology after intramuscular VSV inoculation, the most important question to pursue was whether IL-1 would be required for control of VSV replication in vivo, and for the induction of immune responses to the VSV vector. If IL-1 were not required for control of virus replication, or for generation of protective immune responses, then that would suggest that engineering rVSV vectors to suppress IL-1 production and/or signaling would be a rational approach to reducing rVSV associated pathology. Although three separate reports have now described the induction of IL-1 in response to VSV challenge in vivo [22, 51, 52] , none of these had examined whether or not the induced IL-1 contributed positively to or was required for generation of immune responses or protection. In other viral infection models where this has been investigated, published reports are not in agreement. For example, Schmitz et al found that IL-1R2/2 mice infected with influenza had a significantly higher rate of mortality than did wild type control animals, but the increase in mortality did not correlate with a higher virus burden in the IL-1R2/2 mice [53] . Three additional reports examined the role of IL-1b, as well as components of the Nlrp3 inflammasome, in control of influenza infection [23, 54, 55] . In these, two out of three found transiently elevated viral loads in mice in which IL-1b production was decreased [23, 54] , the other report found no difference in virus burden [55] . Similarly, only one of the three reports [23] found that a reduction in IL-1b correlated with a reduction in cellular immune responses-the others either did not examine immune responses or did not find a correlation. In this study, we found that IL-1R2/2 mice controlled viral loads as well as wild type control animals. Viral loads at the injection site were not significantly different between groups, and neither group had a detectable viremia after challenge. This result highlights the relative safety of the intramuscular immunization route. Equally important was that IL-1R2/2 mice challenged with VSV made robust humoral and cellular responses to VSV antigens, and were fully protected from a high dose intranasal rechallenge with VSV. Interestingly, we did observe levels of anti-VSV CD8 T cells in the IL-1R2/2 mice and in ASC2/2 mice were transiently but significantly decreased, which is consistent with the report of Iwasaki et al, who found that anti-influenza CD8 T cell responses were slightly decreased in mice with reduced levels of IL-1b production [23] . In our studies, the number of anti-VSV specific CD8 T cells was significantly lower in IL-1R2/2 and ASC2/2 mice at 14 days after infection, but was not significantly different at either 7 or 28 days after infection. This finding warrants further investigation, and follow-up studies will focus on determining whether CD8 T cells primed in the absence of IL-1 signaling are functionally different (e.g. in cytotoxic capacity, or in acquisition of central memory phenotype) from those primed in IL-1 intact animals. In summary, these results support the idea that IL-1 production is not required either for control of VSV replication in vivo, or for the induction of protective immune responses to VSV antigens after intramuscular immunization. Because the relative magnitude of immune response to foreign antigens expressed by VSV is generally similar to the relative magnitude of the immune response to VSVs own antigens [56] , we predict that IL-1 would also not be required for the induction of immune responses to vaccine antigens expressed by VSV. Nonetheless, determining whether IL-1 is important for the response to foreign (vaccine) antigens expressed by an rVSV will be an important direction for further study. The second question arising from our results is how the VSVinduced production of IL-1 might be suppressed. One means of suppressing the biological response to IL-1 in humans is via injection of recombinant IL-1 receptor antagonists. Currently marketed as the drugs Anakinra or Kineret, these agents effect a systemic suppression of IL-1a and IL-1b. Although systemic suppression of IL-1 is an effective treatment for some conditions (severe rheumatoid arthritis, for example) [57] , systemic suppression of IL-1 also carries the risk of enhanced susceptibility to infection [57, 58] . Therefore we decided to determine whether there might be a way of suppressing IL-1 only in the cells directly infected with VSV, or within the focus of VSV-infected cells. To determine the appropriate target molecule(s) to effect local suppression of IL-1, it was necessary to determine the mechanism by which IL-1 was being produced in our model. Several years ago, Muruve et al reported that adenovirus activated human monocytes (THP-1 cells) to produce IL-1b in vitro, and that IL-1b production was dependent upon NLRP3 and caspase-1 [51] . In the supplement to that manuscript, the authors reported that infection of THP-1 cells with VSV did not induce IL-1b production, although detailed methods for that study were not provided. In contrast to that, two more recent studies reported that VSV-infected THP-1 cells do produce IL-1b in vitro. In the first study, Poeck et al found that VSV induced production of IL-1b was dependent upon caspase-1, but not upon NLRP3 [22] . In contrast to that, Rajan et al found VSV mediated induction of IL-1b from THP-1 cells to be dependent on NLRP3, and dependent on caspase-1 [52] . It is likely that the discrepancies in these reports were due to subtle differences in experimental procedures, but the different outcomes reported by each of these groups highlight the complexity of elucidating the precise mechanism(s) by which VSV may induce IL-1. In our studies we found that VSV induced robust production of IL-1b in vivo and that IL-1b was still produced in mice lacking caspase-1 or ASC. One caveat to those findings is that ELISA assays such as the one used here do not rigorously discriminate between detection of pro-IL-1b and detection of the cleaved active IL-1b. Therefore it is possible that the some of the IL-1b detected in caspase2/2 mice was not biologically active. Precise determination of the amount of active IL-1b present in caspase2/2 mice will require the development of better detection reagents. Despite this limitation, our results support a model in which there are multiple pathways by which mature IL-1b is produced in vivo in response to VSV. The data are also consistent with the idea that the manner in which mature IL-1b is produced (via caspase-1 or not) may vary with the cell in which the IL-1b is induced. We further observed that mice lacking caspase-1 or ASC were not protected from acute pathology to the same extent that IL-1R2/ 2 mice were. That meant that VSV vectors engineered to suppress components of the RIG-I/ASC/caspase-1 inflammasome, or the Nlrp3/ASC/caspase-1 inflammasome, would be unlikely to be significantly less reactogenic than the parent vector. On the other hand, rVSV designed to suppress the biological response to IL-1b (for example via inclusion of a soluble IL-1b trap), or to IL-1a and IL-1b (via inclusion of a soluble IL-1 receptor antagonist) might reduce the biological response to IL-1 in vivo and therefore be less reactogenic. In summary, we have shown here that IL-1 contributes to acute pathology after intramuscular immunization with VSV. IL-1 was not required for control of viral replication, for the induction of cellular or humoral immune responses, or for development of protective immunity to rechallenge. These results add to our understanding of the role for IL-1 in promoting immunity to viral challenge, and support the notion that the requirement for IL-1b in promoting adaptive immunity may vary according to the type and dose of pathogen encountered as well as the route of exposure. Finally, by identifying IL-1b as a major source of reactogenicity for rVSV vaccines, we are able to propose a novel strategy to ameliorate side effects without compromising immunogenicity. All animal studies were reviewed and approved by the Duke University Institutional Animal Care and Use Committee. Vesicular stomatitis virus (Indiana strain) was originally obtained from Dr. John Rose (Yale University). Virus was propagated on BHK-21 cells (ATCC CCL-10) and titered using a standard plaque assay. Eight to ten-week-old C57BL/6 wildtype and IL-1 receptor type 1 deficient (IL-1R2/2, strain name B6.129S7-Il1r1 tm1Imx/J ) mice were obtained from Jackson Laboratories. Caspase12/2 on the C57BL/6 background were generously provided by Dr. Fayyaz Sutterwala and Dr. Richard Flavell. ASC2/2 and Nlrp32/2 mice on the C57BL/6 background were generously provided from Genentech Inc. (San Francisco, CA). Mice obtained commercially were housed for at least 1 week before experiments were initiated. Mice were housed in microisolator cages in a biosafety level 2-equipped animal facility. Viral stocks were diluted to appropriate titers in serum-free DMEM. For intramuscular immunization (i.m.), mice were injected with the indicated amount of virus(es) in 50 ml total volume. For intranasal (i.n.) vaccination, mice were lightly anesthetized with isoflurane using a vaporizer and administered the indicated amount of virus in 30 ml total volume. The Institutional Animal Care and Use Committee of Duke University approved all animal experiments. Mice were sacrificed via anesthetic overdose and organs removed aseptically. After dissection organs were weighed, and homogenized in sterile buffer (100 ml buffer per 0.1 g organ weight). Homogenates were titered by standard plaque assay on BHK-21 cells (ATCC CCL-10) using a semi-solid overlay to detect infectious VSV. After 48 hours the overlay was removed and the cell layer stained with crystal violet to visualize plaques. Mice were bled and then sacrificed via anesthetic overdose and organs removed aseptically. Organs were homogenized in 500 mL (100 mL for lymph nodes) of buffer containing 137 mM NaCl, 20 mM Tris-Cl pH 8.0, 5 mM EDTA, 0.05% Triton-X 100, and protease inhibitor cocktail (Roche). Serum and organ homogenate supernatants were assayed for IL-1b by ELISA (R&D Systems). Organ IL-1b amounts were normalized to the amount of protein in the samples, as determined using a bicinchoninic acid (BCA) protein assay (Thermo Scientific). Blood was obtained from mice on days 7, 14, and 28 after vaccination via cheek bleed. Heat inactivated serum was diluted in serum-free DMEM such that the final dilution in the first well of a 96-well plate was 1:10 for day 7 samples and 1:100 for day 14 and 28 samples, with subsequent two-fold dilutions. Samples were assayed in duplicate. 100 PFU of rVSV diluted in serum-free DMEM was added to each well and incubated for one hour at 37uC-5%CO 2 , after which 4,000 BHK-21 cells (ATCC CCL-10) diluted in 5%FBS-DMEM were added to each well. Plates were incubated at 37uC-5%CO 2 for three days, and cytopathic effect was observed. The neutralizing titer was defined as the highest dilution of serum that gave 100% neutralization of rVSV. To obtain peripheral blood lymphocytes blood was collected into serum free medium (DMEM) containing heparin. Blood was layered onto a Ficoll gradient and spun, after which lymphocytes were collected from the interface. Cells were washed and resuspended in DMEM containing 5% FCS. Staining was performed on freshly isolated lymphocytes as previously described [4] . Briefly, approximately 5610 6 cells were added to the wells of a 96-well V-bottom plate and were blocked with unconjugated streptavidin (Molecular Probes) and F c block (Pharmingen) for 15 min at room temperature (RT). Following a 5-min centrifugation at 5006g, lymphocytes were labeled with a FITC-conjugated anti-CD62L antibody, (Pharmingen), an allophycocyanin-conjugated anti-CD8 antibody (Pharmingen), and tetramer for 30 min at RT. The tetramer was a PE-conjugated major histocompatibility complex (MHC) class I K b tetramer (NIH Tetramer Facility) containing the H-2K b restricted peptide VSV N 53-59 (N-RGYVYQGL-C). Sham-inoculated control animals were used to determine background levels of tetramer binding. Background was routinely less than 0.1% and was subtracted from all reported percentages. Statistical comparisons were made using GraphPad Prism software. Results were considered significant when P,0.05. Figure S1 Mice deficient in the inflammasome adaptor ASC (ASC2/2) are partially protected from acute weight loss after intramuscular immunization with rVSV. Average percent initial weight for wild type (n = 10) and ASC2/2 (n = 6) mice after intramuscular challenge with 5610 8 PFU of rVSV. The difference in weight loss between wild type and ASC2/2 mice was significant on the first and second day after challenge (P,0.05, Mann Whitney test). Author Contributions

Competition between Influenza A Virus Genome Segments

Influenza A virus (IAV) contains a segmented negative-strand RNA genome. How IAV balances the replication and transcription of its multiple genome segments is not understood. We developed a dual competition assay based on the co-transfection of firefly or Gaussia luciferase-encoding genome segments together with plasmids encoding IAV polymerase subunits and nucleoprotein. At limiting amounts of polymerase subunits, expression of the firefly luciferase segment was negatively affected by the presence of its Gaussia luciferase counterpart, indicative of competition between reporter genome segments. This competition could be relieved by increasing or decreasing the relative amounts of firefly or Gaussia reporter segment, respectively. The balance between the luciferase expression levels was also affected by the identity of the untranslated regions (UTRs) as well as segment length. In general it appeared that genome segments displaying inherent higher expression levels were more efficient competitors of another segment. When natural genome segments were tested for their ability to suppress reporter gene expression, shorter genome segments generally reduced firefly luciferase expression to a larger extent, with the M and NS segments having the largest effect. The balance between different reporter segments was most dramatically affected by the introduction of UTR panhandle-stabilizing mutations. Furthermore, only reporter genome segments carrying these mutations were able to efficiently compete with the natural genome segments in infected cells. Our data indicate that IAV genome segments compete for available polymerases. Competition is affected by segment length, coding region, and UTRs. This competition is probably most apparent early during infection, when limiting amounts of polymerases are present, and may contribute to the regulation of segment-specific replication and transcription.

The mechanism of replication and transcription varies greatly among viruses depending on the nature and structure of their viral genomes. Negative-strand RNA viruses replicate their viral genome via the synthesis of full length positive-strand complementary RNA (cRNA) molecules that in turn serve as templates for the synthesis of negative-strand virion RNA (vRNA) genomes. The negative-strand genomes also function as templates for the production of mRNAs [1, 2] . In non-segmented negative-strand RNA viruses, sequential transcription of successive genes results in a gradient of transcript abundance that steadily decreases towards the end of the template. Thus, the expression level of each gene is governed by the gene order [3] . This does, however, not apply to all negative-strand viruses as some of them acquired segmented genomes during their evolution. Each genome segment of these viruses is individually replicated and transcribed, necessitating careful regulation of these distinctive processes to generate sufficient vRNAs and proteins for the production of progeny virions [2] . Influenza A virus (IAV) of the family Orthomyxoviridae is an enveloped, negative-strand RNA virus. The IAV genome is composed of eight different vRNA segments that altogether encode up to 13 proteins [4] [5] [6] [7] . Each vRNA and cRNA possesses untranslated regions (UTRs) of varying length at the 39 and 59 ends. The first 12 and 13 nucleotides at the 39 and 59 UTRs of the vRNAs and cRNAs are highly conserved among different RNA segments. These highly conserved partly complementary UTRs, which form a ''panhandle'' or ''corkscrew'' conformation by alternative modes of base-pairing, constitute the promoter structure for RNA synthesis [8, 9] . The panhandle conformation results from base-pairing between 59 and 39 terminal ends of the viral RNA segment with a small internal loop [10, 11] , while the corkscrew structure consists of a six base-pair RNA rod in the distal element in conjunction with two stem-loop structures of two short-range base-pairs [12] . The IAV vRNA and cRNA segments form ribonucleoprotein (RNP) complexes by association to the polymerase and to multiple copies of the nucleoprotein (NP). These RNPs may be regarded as independent molecular machines responsible for transcription and replication of each segment. The viral RNA polymerase, which consists of the PA, PB1 and PB2 subunits, recognizes the RNA promoter, and stabilizes a supercoiled conformation of the RNPs. Different models have been proposed for the regulation of transcription and replication. One model suggests that the RNA polymerase switches from a transcriptase, used for mRNA synthesis, to a replicase, used for cRNA and vRNA synthesis, which is triggered by newly synthesized NP protein [13] . Another model suggests that cRNAs can be directly synthesized from incoming vRNAs, but require newly synthesized polymerase and NP to be stabilized in RNPs [14] . More recently, Jorba and colleagues proposed a model, in which a template RNP is replicated in trans by a soluble polymerase complex, whereas transcription of the vRNA occurs in cis and the resident polymerase complex is responsible for mRNA synthesis [15] . Early studies, in which semi-quantitative hybridization techniques were used, described differential expression rates and levels of the different vRNAs. In general it appeared that the mRNAs for NS1 and NP are synthesized preferentially at the early times post infection, while the synthesis of matrix (M1) mRNA is delayed [16] [17] [18] . More recently, Vester and coworkers showed, by using quantitative RT-PCR that the vRNAs are synthesized in equimolar amounts and with similar kinetics, whereas early in infection preferential synthesis of NS1 mRNA and a delay in that of M1 mRNA was found [19] . However, how IAV temporally regulates the replication and transcription levels of its multiple genome segments is not known. Several reporter assays have been described to study and quantify IAV transcription/replication in vivo. These reporter systems usually consist of a reporter protein-encoding cDNA, flanked by 39 and 59 UTRs, inserted in an antisense orientation between a PolI promoter and a transcription terminator or ribozyme sequence. After introduction of the reporter construct into a cell, reporter gene expression is induced by co-transfection of plasmids encoding NP, PA, PB1 and PB2 (transfection assay) or by subsequent infection with a helper IAV (infection assay). Such reporter assays are very helpful to quantify virus replication or virus production, and to assess the antiviral activity of compounds including antibodies [20] [21] [22] . These assays have also been used for the mutational analysis of IAV promoter elements in vivo [12, 23, 24] . To get more insight in the mechanism by which IAV regulates and balances the replication and transcription of its genome segments, we converted the IAV single reporter assay into a dual reporter assay, by which the expression of two different luciferase genes can be monitored simultaneously. This assay more closely resembles the multiple segment transcription and replication conditions that occur in IAV infected cells than the single reporter assay. Our results indicate that different vRNA segments compete with each other, as transcription/replication of one vRNA Figure 1 . Schematic representations of the dual luciferase reporter constructs, and of transfection and infection assays. A) Schematic outline of the firefly and Gaussia luciferase reporter constructs. The firefly and Gaussia luciferase genes, flanked by 39 and 59 UTR of the NP segment, were inserted in antisense orientation between a PolI promoter and a ribozyme sequence, resulting in FNP and GNP, respectively. The extended Gaussia luciferase reporter construct (GFsNP) additionally contains the 39 terminal half of the firefly luciferase gene (indicated as Fs) behind the stop codon of the Gaussia gene. B) HEK 293T cells were transfected with one or both reporter constructs (single or co-transfection). Luciferase expression is induced by expression of viral RNA polymerases (PB1, PB2, PA) and NP either by simultaneous co-transfection of expression plasmids (transfection assay) or by infection with IAV at an MOI 1 at 24 h post-transfection (infection assay). The firefly and Gaussia luciferase expression levels can be measured consecutively using a dual luciferase assay system (Promega) 24 h post-transfection or post-infection. doi:10.1371/journal.pone.0047529.g001 segment can affect that of another. Using this multiple segment reporter assay we subsequently assessed the contribution of vRNA segment length, UTRs sequence and coding sequence to the competition between the different segments. A schematic overview of the dual luciferase reporter constructs and the assays is shown in Figure 1 . The firefly and Gaussia luciferase genes, in this example flanked by 39 and 59 UTRs of the NP segment (referred to as FNP and GNP, respectively), are inserted in an antisense orientation between a PolI promoter and a ribozyme sequence. Cells are transfected with either one or both reporter constructs (single or co-transfection). Luciferase expression is induced by expression of viral RNA polymerases and NP either by co-transfection of expression plasmids (transfection assay) or by virus infection (infection assay). The expression levels of the firefly and Gaussia luciferase reporter constructs are determined consecutively using a single tube, dual luciferase assay system. First we determined the luciferase expression levels of the firefly (FNP) and Gaussia (GNP) luciferase reporter constructs when transfected alone or in combination by using the transfection assay. As shown in Figure 2 , single transfection of each reporter gene resulted in high expression levels of both the firefly ( Fig. 2A) and Gaussia (Fig. 2B ) luciferase genes. However, when the reporter constructs were co-transfected, the firefly luciferase expression level was dramatically reduced ( Fig. 2A) , while the Gaussia luciferase expression level in the same cells was not affected when compared to the single-transfected cells (Fig. 2B) . The differential expression of firefly and Gaussia reporter plasmids when transfected alone or together can also be illustrated by plotting the normalized ratio of firefly to Gaussia luciferase activity (normFluc/ Gluc; the normalized ratio's were calculated as indicated in the Materials and Methods section). As shown in Figure 2C , this ratio is significantly decreased upon co-transfection of the two reporter constructs, when compared to the ratio of the luciferase expression levels in the single-transfected cells. Similar results were obtained at earlier and later time points post transfection (data not shown). This indicates that the balance of firefly and Gaussia luciferase expression is strongly in favor of Gaussia luciferase, when both reporter constructs are present within the same cell. Thus, the results indicate that expression of the firefly luciferase gene is Figure 2 . Competition between firefly and Gaussia luciferase reporter genome segments. Plasmids encoding firefly (FNP) or Gaussia (GNP) luciferase reporter constructs were transfected alone (Single) or in combination (Co). Luciferase expression was induced by simultaneous cotransfection of polymerase and NP expression plasmids (transfection assay). A) Firefly luciferase activity after transfection of FNP or FNP together with GNP. B) Gaussia luciferase activity after transfection of GNP or GNP together with FNP. C) Normalized ratio of firefly to Gaussia luciferase activity (Fluc/ Gluc) when FNP and GNP were transfected singly or in combination. D) Quantitative RT-PCR analysis of mRNA levels derived from FNP and GNP after single or co-transfection of these constructs. RNAs were extracted 24 h post-transfection and subjected to quantitative RT-PCR. The comparative Ct method was used to determine the relative mRNA levels using the housekeeping gene GAPDH as a reference. The mRNA levels were normalized relative to the samples in which a single reporter construct was transfected. E) Normalized ratio of firefly to Gaussia luciferase activity (Fluc/Gluc) when reporter gene constructs FNP and GNP were transfected singly or in combination. The amounts of reporter gene constructs transfected are indicated. Significant differences in A-D are indicated (**; P,0,01). doi:10.1371/journal.pone.0047529.g002 negatively affected by co-transfection of the Gaussia luciferase reporter plasmid. Very similar results were obtained when an empty plasmid (pUC18) was included in the transfection mixture when only one reporter construct was transfected ( Fig. S1A-C) . Thus, the observed differences in firefly luciferase expression do not result from a lower transfection efficiency of the firefly luciferase, but not of the Gaussia luciferase reporter construct, when an additional plasmid was included in the transfection mixture. Next, to analyze whether the observed difference in luciferase protein levels results from differences at the RNA level, we performed a quantitative RT-PCR analysis of the mRNA levels [19] . The results are shown in Figure 2D . The mRNA levels of the Gaussia reporter gene were not affected by co-transfection of the other reporter construct. However, co-transfection of the Gaussia luciferase construct significantly affected mRNA levels of the firefly luciferase gene. From these results we conclude that the observed inhibitory effect of co-transfection of the Gaussia luciferase construct on the firefly luciferase activity is a reflection of lower firefly luciferase mRNA levels. The results indicate that replication and transcription of the Gaussia and firefly luciferase genome segments are in competition with each other. If so the observed inhibitory effect of cotransfection of the Gaussia luciferase construct on the firefly luciferase expression level is expected to depend on the ratio of the transfected reporter constructs. As shown in Figure 2E , this is indeed the case. Lowering the amount of co-transfected Gaussia as well as increasing the amount of co-transfected firefly luciferase reporter plasmid shifted the balance in the competition between the firefly and Gaussia luciferase genome segments as judged from the increased normFluc/Gluc ratio (Fig. 2E ). This increased ratio resulted from altered firefly rather than Gaussia luciferase expression levels ( Fig. S2A and B ). From these results we conclude that the Gaussia luciferase genome segment is much more efficiently replicated and transcribed than its firefly luciferase counterpart, the latter of which is outcompeted by the presence of the former. The firefly and Gaussia luciferase genome segments are most likely competing for host and/or viral factors that are necessary for transcription and/or replication. To analyze whether a limiting availability of viral proteins is an important factor in the competition between firefly and Gaussia luciferase genome segments, we increased the amount of plasmids encoding the RNA polymerase subunits and NP in the transfection assay. Empty plasmid (pUC18) was included in the transfection mixture when needed to achieve the same total amount plasmid DNA for each transfection condition. Upon increasing the amounts of transfected plasmids encoding PB1/PB2/PA/NP, the normFluc/ Gluc ratio increased ,10-fold (Fig. 3A ) as a result of increased firefly luciferase expression levels ( Fig. S3A and B ). This result indicates that increased amounts of polymerase subunits and NP can alleviate the competition between the firefly and Gaussia luciferase genome segments. Increasing the amount of transfected NP-encoding plasmid alone did not affect the normFluc/Gluc ratio ( Fig. 3B ) or the absolute Fluc and Gluc levels ( Fig. S3C and D). Increasing the amount of polymerase subunit-encoding plasmids, but not of the NP-encoding plasmid, appeared to alleviate the competition between the two segments ( Fig. 3C ). However, it also negatively affected the reporter gene expression levels per se, with most dramatic effects being observed for the firefly luciferase reporter segment ( Fig. S3E and F). We speculate that this negative effect correlates with the requirement for NP for replication, which appears less stringent for short RNA templates [25, 26] . Although we did not analyze the NP and polymerase protein levels directly, our results indicate that the luciferase genome segments compete for RNA polymerase subunits and/or NP, with a limiting amount of polymerase subunits being the most likely explanation for the observed competition. However, we cannot exclude that the observed competition between reporter segments is partly caused by limiting amounts of host factors. Next we analyzed to what extent the competition between the luciferase genome segments is affected by characteristics of the genome segments themselves. An obvious difference between the firefly and Gaussia genome segments is their gene length as the firefly and Gaussia luciferase genes consist of 1653 and 558 nucleotides, respectively. Small genome segments are likely to be replicated faster than long ones. To test this hypothesis, we generated an extended Gaussia luciferase gene construct, in which part of the firefly luciferase gene (39-terminal half) was inserted immediately behind the stop codon of the Gaussia luciferase gene in the GNP plasmid (referred to as GFsNP, Fig. 1A) to produce a genome segment with exactly the same length as the firefly luciferase genome segment. The extended and the normal Gaussia luciferase reporter segments were compared for their ability to compete with the firefly luciferase reporter segment. The absolute Gaussia luciferase expression level from the GFsNP segment was lower than that from the GNP segment (Fig. S4B ), probably resulting from its extended length, for which is corrected by plotting normalized Fluc/Gluc ratios in Figure 4 . These norm-Fluc/Gluc ratios (Fig. 4A) indicate that the extended GFsNP segment is still a strong competitor of the FNP segment, although much less efficient than the smaller GNP segment. We conclude that segment size is an important factor in the competition between vRNA segments. However, our results also suggest that coding regions are important as FNP and GFsNP segments do not differ in size, while they contain identical 59 and 39UTRs. Next, we analyzed to what extent the competition between different reporter constructs is affected by the identity of the 39 and 59 UTRs. The genome segment UTRs provide signals for viral RNA transcription and replication, as well as for packaging of vRNP into virus particles [27] . The first 12 and 13 nucleotides at the 39 and 59 UTRs, which are highly conserved among the eight viral RNA segments, constitute the promoter structure for RNA synthesis [8, 9] . Also the non-conserved regions of the different segments have been implicated in viral RNA replication [28] . We inserted the 39 and 59 UTRs of the eight IAV-WSN segments (Table 1) into the extended Gaussia reporter plasmid and tested them in the competition assay with the firefly luciferase reporter construct FNP. The extended Gaussia construct was chosen instead of the short version as the former construct affects expression of the firefly luciferase construct less than the latter, resulting in a more balanced system in which it will be easier to detect UTRdependent up and down effects. Introduction of the UTRs of different segments into the extended Gaussia reporter construct affected the balance between the Gaussia and firefly luciferase expression to different extents (Fig. 4B) . Introduction of PB1, NA or NS segment UTRs resulted in balanced firefly and Gaussia luciferase expression levels as similar ratio's were observed when expressed alone or in combination (normFluc/Gluc ,1; Fig. 4B ). When the extended Gaussia construct was provided with the PB2, PA, HA and M segment UTRs, the normalized ratio's observed after co-transfection with FNP were similar to those observed when the construct containing the NP segment UTRs was used (normFluc/Gluc ,1), indicating that the balance had shifted in favor of Gaussia luciferase expression. Subsequently, we analyzed whether the absolute expression levels of the eight different Gaussia luciferase segments (Fig. S5A ) correlated with their ability to inhibit expression of firefly luciferase (Fig. S5B) . Correlation between inhibition of firefly luciferase expression and the expression of Gaussia luciferase with different UTRs resulted in an R 2 value of 0.8 (Fig. 4C) , which indicates that 80% of the variance in firefly luciferase inhibition correlates with variability in Gaussia luciferase expression levels. A similar R 2 value was obtained when the results obtained with the short GNP reporter segment were also taken into account (Fig. S6) . The results indicate that not only the expression of reporter genome segments, but also the balance between different reporter genome segments is affected by 39 and 59 UTRs, and that these two phenomena are largely correlated. Thus, Gaussia luciferase genome segments that are expressed to a higher extent are more efficient inhibitors of firefly luciferase expression driven by another genome segment. Subsequently, we analyzed to what extent reporter gene expression was affected by the presence of the natural viral genome segments. To this end, the reporter constructs were cotransfected with plasmids encoding each of the IAV genome segments under the same control of human RNA polymerase I promoter. As a control, empty plasmid (pUC18) was cotransfected. The different IAV segments significantly affected the firefly luciferase levels (Fig. 5A) . In general, the shorter segments gave stronger competition on the firefly expression compared to longer segments, with the M and NS segments having the largest effect. Correlation between inhibition on firefly luciferase expression and the gene length resulted in an R 2 value of 0.7 (Fig. 5C) , which indicates that 70% of the variance in firefly luciferase inhibition correlates with variability in genome segment length. In contrast, expression of Gaussia luciferase was hardly affected by the presence of the viral RNA segments (Fig. 5B) , while no correlation was observed between the modest decrease/increase of Gaussia luciferase expression and the genome segment length (Fig. S7) . We speculate that the lack of inhibition of Gaussia luciferase expression correlates with the very efficient replication/transcription of this reporter segment. Mutations in the 39UTR of the NP segment that increase gene expression have been described [23] . The nucleotide changes were predicted to improve base pairing of the 39 and 59 UTRs and thus to stabilize the panhandle structure. Considering the results described above, we expected that these mutations would affect competition between different reporter genome segments. The results show that reporter constructs containing these panhandlestabilizing UTRs (referred to as NPph; Fig. 6A ) indeed displayed 3-5 fold higher luciferase expression levels in the transfection assay than their wild-type UTR-containing counterparts (Fig. S8) . Remarkably, however, the normalized ratio between firefly and Gaussia luciferase was much more affected by the presence of the NPph UTRs. Introducing the NPph UTR in the background of the extended Gaussia construct (GFsNPph) resulted in a much decreased normFluc/Gluc when compared to its counterpart with the wild type NP UTRs (GFsNP; Fig. 6B) . A similar level of competition was not observed when the mutations that increase the number of base-pairs were introduced in the 59 UTR of the NP segment (referred to as NPphR; Fig. 6A ) instead of in the 39UTR. In this case, the Gaussia, rather than the firefly luciferase expression was affected by the co-transfection of both reporter plasmids (normFluc/Gluc .1; Fig. 6B and Fig. S8B ). Similar results were obtained when NPph UTR was introduced in the firefly luciferase genome segment (referred to as FNPph; Fig. 6C ). In general, normFluc/Gluc was increased when FNPph rather than FNP was used, except for the combination with GNPph ( Fig. 6C & Fig. S8C and D) . Thus, the balance between different segments is dramatically affected by the introduction of panhandle structure-stabilizing mutations in the 39UTR. Infection assay Neumann and Hobom (1995) previously reported increased reporter gene expression upon the introduction of panhandlestabilizing mutations in the 39 UTR. In their experimental system, however, the differences in reporter gene expression appeared much larger than ours. We hypothesized that this difference might be explained by Neuman and Hobom using virus infection to drive reporter gene expression, while we used transfection of polymerase subunit-and NP-encoding plasmids. Infection with IAV will not only provide viral RNA polymerase and NP, but will also introduce natural vRNPs that may compete with reporter genome segments for replication and/or transcription. Thus, in virusinfected cells, the natural virus genome segments might be preferentially replicated and transcribed over the reporter genome segments, unless the panhandle-stabilizing mutations in the 39 UTR are present. To test this hypothesis we compared reporter gene expression driven by co-transfection of expression plasmids (transfection assay) with expression driven by virus infection (infection assay). Reporter genes flanked either by natural NP UTRs or by the mutant NPph UTRs were used. Both firefly and Gaussia luciferase genes were expressed at high levels in the transfection assay, with the reporter constructs containing the NPph UTRs again displaying somewhat higher luciferase levels than their counterparts with the natural NP UTRs (Fig. 7A) . However, when using the infection assay, dramatic differences in reporter gene expression levels were observed (Fig. 7B) . Thus, while the reporter genes flanked by the NPph UTRs reached 1 to 2 fold higher expression levels than those flanked by the natural NP UTR in the transfection assay, this fold difference was much increased (130 to 160 fold) in the infection assay (Fig. 7C) . Quantitative RT PCR confirmed that mRNAs levels of the Gaussia luciferase RNAs were very similar, regardless of the presence of the natural NP UTRs or the mutant NPph UTRs in the transfection assay, but not in the infection assay (Fig. 7D) . These results are in agreement with our model, in which IAV genome segments compete for available resources, likely the viral proteins, to maximize their replication and/or transcription. Only reporter genome segments carrying panhandle-stabilizing mutations in their 39 UTR are able to efficiently compete with the natural genome segments in the infection assay. The molecular mechanisms by which IAV replicates and transcribes its genome segments have generally been well studied. However, the way by which IAV regulates and balances the replication/transcription of its 8 genome segments is much less understood. In order to study and manipulate these processes, we developed a dual reporter genome segment assay that enabled us to analyze whether the replication/transcription of one genome segment is affected by that of another. Our results indicate that this is indeed the case as luciferase expression driven from a reporter genome segment was shown to be affected by the presence of other genome segments, both in the context of virus infection and in the presence of polymerase and NP proteins provided by transfection of the expression plasmids. Furthermore, our results indicate that genome segments are likely to compete with each other for the available viral proteins and that the balance between different genome segments is affected by reporter genome segment length, by the identity of 39 and 59 UTRs, and probably also by their coding regions. Our results indicate that replication/transcription of a genome segment can be negatively affected by the presence of another genome segment. This interference became less pronounced when the length of the smaller segment was extended, indicating that genome segment length plays a role in the competition between different segments. This ''length effect'' was also observed when natural genome segments were present in addition to the reporter construct, with the shortest segments, M and NS, giving the strongest inhibition of the reporter gene expression. In agreement herewith, IAV defective interfering (DI) RNAs, which are formed by internal deletion of progenitor RNA segments, interfere with vRNA synthesis, probably because of the competitive advantage of the smaller DI RNA molecules (reviewed by Nayak [29] ). In addition, our data indicate that the coding region of the vRNA segment may also be of importance, as the extended version of the Gaussia reporter segment was still able to outcompete its firefly luciferase counterpart, albeit less efficiently than its shorter version. The segment UTRs are known to contain signals for transcription, replication and packaging of vRNP [24, 27, 28, 30] . We now show that the identity of the 39 and 59 UTRs also influences the competition between different segments. Relatively minor differences were observed when reporter genome segments with different natural UTRs were compared in the competition assay. This result is in agreement with the observation that non- conserved regions of the UTRs contribute to some but limited extent to viral RNA replication [28, 31] . However, introducing three nucleotide changes in the 39 UTR (G3A/U5C/C8U) of the NP segment, which is predicted to stabilize the UTR panhandle structure and is known to lead to increased reporter gene expression in infected cells [23] , dramatically increased the competitive ability of the reporter segment, both when replication/transcription was driven by transfection of polymerase-and NP-encoding segments and when mediated by IAV infection. Thus, while reporter segments carrying the natural NP UTRs or the mutant NPph UTRs were both efficiently expressed in the absence of competitor segments, large differences in luciferase expression were observed in favor of the luciferase segment carrying the panhandle-stabilizing mutations when other reporter segments were co-transfected or in IAV infected cells. In agreement herewith, recombinant viruses carrying two nucleotide changes (G3A/C8U) in the UTR of either the PB1 or PA segment displayed enhanced replication/transcription of the mutated segments in detriment of the wild-type UTR-bearing segments [32] . The most likely scenario suggested by our observations is that replication/transcription of one reporter segment interferes with that of another by sequestering UTR-binding proteins, probably polymerases, required for RNA synthesis. Several observations by us and others support this hypothesis: 1) increasing the amount of polymerase and NP proteins, but not of NP protein alone, alleviated the competition between different segments, 2) the polymerase proteins have been shown to bind to 59 and 39 UTRs of vRNAs, with most strong binding observed to the 59 UTR [33] , 3) introduction of mutations in the 39 UTR (NPph) that stabilize the panhandle structure and are predicted to result in increased polymerase binding [23] result in increased ability of the reporter segment to be replicated/transcribed in the presence of competitor segments ( [23] and this study), 4) introduction of similar mutations in the 59 UTR (NPphR) that are likely to interfere with polymerase binding [33] , had a negative effect on the competitive ability of the reporter construct, and 5) panhandle-stabilizing mutations in the 39 UTR (NPph), that increased the competitive ability of the reporter construct, partly compensated for replication-debilitating mutations in PB2 (R142A or E361A) [34, 35] , but not in NP (M331K or F488G) [36] (Fig. S9) , suggesting a link between the interaction of the UTR with polymerase and the ability to compete with other segments. Although our experimental system (i.e. the transfection assay) does not approach the complexity of the IAV infected cells with respect to number of vRNA segments and viral proteins present, our data suggest that IAV RNA segments compete with each other for available polymerases. This competition is expected to be most apparent early during infection, when only low amounts of polymerase are present. It is conceivable that at this stage of the infection the low level of RNA polymerase plays a critical role in the regulation of segment-specific replication and/or transcription. At later times during infection competition between vRNA segments is expected to be alleviated by the increased levels of the polymerase subunits, thereby ensuring the efficient replication/transcription of all genome segments. HEK 293T and MDCK cells were maintained in complete Dulbecco's Modified Eagle's Medium (DMEM; Gibco) containing 10% (v/v) Fetal Calf Serum (FCS; Bodinco B. V.), 100 U/ml Penicillin and 100 mg/ml Streptomycin. Influenza A/WSN/33 (H1N1) (IAV-WSN) was grown on MDCK cells. Briefly ,70% confluent MDCK cells were infected with IAV-WSN at a multiplicity of infection (MOI) of 0.02 50% tissue culture infectious dose (TCID50) per cell. Supernatant was harvested after 48 h of incubation at 37uC and cell debris was removed by centrifugation at 2,000 rpm for 10 min. Virus was stored at 280uC and TCID50 on MDCK cells was determined. Precursor firefly and Gaussia luciferase reporter gene constructs were generated by GenScript. These constructs have the following schematic make up: hepatitis delta virus (HDV) ribozyme -AloI restriction site -firefly or Gaussia luciferase gene -BaeI restriction site -human PolI promoter. These constructs did not yet contain IAV 59 and 39 UTR sequences. 39 UTR sequences were introduced by ligation of primer dimers into the AloI-digested constructs. Subsequently, primer dimers corresponding to 59 UTR sequences were ligated into the resulting BaeI-digested plasmid. The AloI and BaeI restriction recognition sites are completely removed by this procedure and the corresponding PolI transcripts are similar to genuine vRNA segments with the exception of the coding region. In addition, we generated an extended version of the Gaussia luciferase reporter gene construct by introduction of a firefly luciferase gene fragment. As a result, the length of the extended Gaussia luciferase vRNA is identical to that of the corresponding firefly luciferase-encoding vRNA. To generate this construct, a NotI-digested PCR fragment, corresponding to part of the firefly luciferase gene (nucleotide 559 -stop codon) and containing flanking NotI restriction sites, was ligated into the NotIrestricted precursor Gaussia luciferase plasmid, immediately downstream of the Gaussia luciferase gene stop codon. The IAV segment UTRs were inserted in the extended Gaussia luciferase plasmid in the same manner as described above. For a schematic overview of the reporter constructs, see Figure 1A . The protein expression plasmids encoding PB2, PB1, PA and NP (pcDNA-PB2, pcDNA-PB1, pcDNA-PA and pcDNA-NP) and transcription plasmids encoding eight IAV-WSN vRNA segments (pPOLI-PB2, pPOLI-PB1, pPOLI-PA, pPOLI-HA, pPOLI-NP, pPOLI-NA, pPOLI-M, and pPOLI-NS) were a kind gift of Dr. Ervin Fodor [37] . HEK 293T cells were seeded in 96-wells plates at a density of 10,000 cells per well and incubated overnight. For the transfection assay, cells were transfected with reporter plasmids encoding firefly or Gaussia luciferase along with expression plasmids encoding PB2, PB1, PA, or NP using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocols. Fifty nanogram of each plasmid was used in the transfection unless mentioned otherwise. For the infection assay, cells were transfected with reporter plasmids encoding firefly or Gaussia luciferase. The next day, cells were infected with IAV-WSN at a multiplicity of infection (MOI) of 1 TCID50 units per cells. Twenty-four h post-transfection or -infection, cells were lysed by incubation with Passive Lysis Buffer (Promega) for 15 min at room temperature. Cell lysates were assayed for luciferase activity using the Dual-Luciferase assay system (Promega) according to the manufacturer's protocols, and the relative light units (RLU) were determined using a Centro LB 960 Luminometer (Berthold Technologies). The ratio of firefly luciferase/Gaussia luciferase activity after co-transfection of both reporter constructs (norm-Fluc/Gluc) was normalized to the ratio of firefly luciferase/Gaussia luciferase activity after single transfection of reporter constructs, which was set at 1. Quantitative RT-PCR to determine the amount of mRNA synthesized for the reporter genes during the transfection and infection assays was performed according to Vester et al. [19] . Briefly, following the removal of the cell culture medium, cells were washed with PBS and lysed by incubation with TriZol reagent (Invitrogen) for 3 min at room temperature. The lysates were mixed with chloroform and centrifuged at 14,000 rpm for 20 min at 4uC. The water phase was collected and mixed with 70% (v/v) ethanol. Subsequent RNA purification was performed using the RNAeasy Kit (Qiagen) according to the manufacturer's protocols. The concentration of total RNA was determined using NanoDrop 1000 Spectrophotometer (Thermo Scientific). The total RNA was treated with amplification grade DNase (Invitrogen) according to the manufacturer's protocols to digest the plasmid DNA. Reverse transcription from total RNA was performed using mRNA-specific primer ( Table 2) . Reverse transcription was carried out using Superscript II reverse transcriptase (Invitrogen). Briefly, 100 ng of DNase-treated total RNA was mixed with 2 pmol of primer and 1 ml of 25 mM dNTP in the total volume of 12 ml. The mixture was incubated at 65uC for 5 min. After the cooling step to 4uC, 4 ml of 56 first strand buffer, 2 ml of 0.1 M DTT, 1 ml of RNase Inhibitor (40 U/ml) were added and the mixture was incubated at 42uC for 2 min. Reverse transcription was carried out at 42uC for 50 min after addition of 1 ml superscript II reverse transcriptase (50 U/ml) and was terminated by heating at 70uC for 15 min. Real time quantitative PCR was performed using qPCR MasterMix Plus for SYBR Green (Eurogentech) on a LightCycler 480 II (Roche). qPCR forward and reverse primers ( Table 2 ) that primed at the coding sequence of corresponding reporter gene were used to amplify cDNA. Quantitative PCR reactions were set up in triplicates according to the manufacturer's instruction by mixing 20 pmol of forward and reverse primers and 1 ml of cDNA products. The PCR mixture was incubated at 95uC for 10 min, followed by 40 cycles of 15 sec and 1 min incubations at 95uC and 60uC, respectively. To check the specificity of PCR product, melting curve analysis was performed at the end of the PCR. The comparative Ct method was used to determine the relative mRNA levels using the housekeeping gene GAPDH as a reference [38, 39] . The mRNA levels were normalized relative to the samples in which a single reporter construct was transfected. The means of multiple experiments are shown. All experiments were performed 2-4 times, with each experiment containing 4 replicates. Differences between means were determined using Student's t-test. Differences were considered significant if P,0.05. Significant differences are indicated by symbols in the figures where appropriate. Figure S1 Competition between firefly and Gaussia luciferase reporter genome segments. Plasmids encoding firefly (FNP) or Gaussia (GNP) luciferase reporter constructs were transfected alone (Single) or in combination (Co). Luciferase expression was induced by simultaneous co-transfection of polymerase and NP expression plasmids (transfection assay). A) Firefly luciferase activity after transfection of FNP with empty plasmid (pUC18) or FNP together with GNP. B) Gaussia luciferase activity after transfection of GNP with empty plasmid (pUC18) or GNP together with FNP. C) Normalized ratio of firefly to Gaussia luciferase activity (Fluc/Gluc) when FNP and GNP were transfected singly or in combination. (TIF)

Oral health in China – trends and challenges

For a long time, oral disease is one of the major problems of the public health for its high prevalence and incidence throughout the world, which is especially true for low-income populations. Since China's economic reform in 1978, great changes have taken place in China. These changes have significant impact on and have been reflected in oral disease trends in China. This paper provides an overview and assessment of the oral health status in China. It focuses on changes in the nation's demographic profile, in the marketplace, the oral disease status and trends. The paper also suggests some possible measures and strategies for bettering oral health in future China.

In recent years, chronic noncontagious diseases have become a major health problem in the world. Besides social and environmental factors, the changes in lifestyle worldwide have great effects on the changes in the disease status and trends. Oral disease is one of the major problems of the public health for its high prevalence and incidence, especially in low-income populations. Oral diseases have become one of the major maladies. This paper provides an overview of the oral health status in China, focusing on changes in the nation's demographic profile, the marketplace, and trends of Chinese dentistry. The paper also suggests some related measures and strategies for improving oral health in future China. China's population is expected to grow from 1. cing aging of its population and the senescence process develops rapidly. Since 1990, the aged population (people over 60 years) has increased at an average annual rate of 3.32%. In the 21st century, China has entered the aging society. In 2005, 11% of Chinese population, i.e. 144 million people, are older than 60 years of age. The number of Chinese senior citizens over 65 years old has amounted to 100 millions, accounting for 7.7% of the total population. By 2040, the number will have increased to about 374 million, accounting for about 25%, thus making China the largest aging society in history (Tables 1, 2 ). The change of population will have major effects on oral health [1] . In the same time, many rural people have rushed into the city, altering the rural to urban population proportion a great deal (Table 3) . The rural population has decreased from 73.6% of the total population in 1990 to 57.0% in 2005. A few years later, this proportion will further decrease to about a half. Meanwhile, the great changes in lifestyle will affect their general and oral health. These demographic changes will alter disease patterns, cultural attitudes, health behaviors and the health care delivery systems and services. [2] . Economic development has instigated the development of general medical health care and oral health care. However, despite the high growth rate of GDP, the spending on public health care, especially the oral health care, remains low, as seen by the average spending on health care per person [3] [4] [5] (Tables 4, 5). The data are calculated at current prices. Recently, China's gini coefficient (a measurement of wealth distribution of a society, the greater the value, the deeper the gap between rich and poor) keeps growing. According to the World Bank assessment, China's gini coefficient was 0.447 in 2001. In 2002, the Chinese Academy of Social Sciences calculated the gini coefficient at 0.454. In both years, China has exceeded the alerting threshold of 0.4, indicating that the income gap in China has been too wide [6] . For the difference in oral health status between rich and poor, it is the same story. Low-income populations and people in need of special care (aged people, children, disabled people) experience illness more often and need more treatment. Besides lack of money and ignorance of oral health knowledge, they do not have easy access to oral health care. According to the statistics released by the Chinese Ministry of Civil Affairs, there are more than 60 million victims of natural disasters each year, over 22.82 million low-income people who receive subsistence allowances in city, 40 million rural people who have very low incomes and live in absolute poverty, 60 million disabled people and 100 million elder people over 65 years old who need all kinds of assistance [7] . Meanwhile, the government budgetary funding only accounts for 0.3% of the national GDP. In the past two decades, both the general medical and oral health conditions of the Chinese people have been improving. However, oral diseases remain prevalent. Many factors, such as area, race, age and gender, affect the severity of oral diseases. Oral health conditions differ greatly under different influences. Caries of the primary dentition still calls for attention, as the number of children, although accounting for a decreased percentage of the total population, remains large. In some areas, caries of the permanent dentition has increased in number. The elderly with their relative and absolute number increasing have complicated medical conditions and an increased need for treatment. According to the third national epidemiological investigation on oral diseases conducted in 2005, the caries prevalence rate of children aged between 5-6 years old remains high, people of 35-44 and 65-74 years experience high caries prevalence rate and low filling rate [8] [9] (Table 6 ). With more teeth remaining in the oral cavity, the elderly are apt to develop caries and other oral disorders. Investigations conducted in Bejing and Shanghai have revealed that people aged between 65-70 years exhibit a root caries prevalence rate of 57%, a root caries index of 4.0-5.8, DFS index of 5.9-6.3 while 60%-70% of the fillings need replacement or have secondary caries underneath [8] [9] [10] [11] . The caries status in China exhibits characteristics typically found in developing countries. About 97% carious teeth of children aged 5 are left untreated, while for children aged 12, this percentage is 89%. About 78.9%-91.7% of carious teeth (including third molars) of the middle-aged and elderly people have been left untreated. This situation has not been much improved with China's economic development, challenging China's oral health care resources [8] [9] . The third national epidemiological investigation on oral diseases conducted in 2005 revealed that, gingival bleeding and calculus occur in over a half of 12-year-olds (57.7% and 59.1% respectively), most middle-aged people (77.3%, 97.3%) and many elderly people (68.0%, 88.7%) ( Table 7) . Periodontal conditions of all age groups were found to be better in cities, females and eastern areas than in rural areas, males, middle and western areas. The findings of this investigation did not differ much from the previous one [8] [9] . In all factors contributing to periodontal diseases, smoking and diabetes should not be ignored. Tobacco consumption related to oral disorders is one of the major sources of the global oral disease load. Over a half of adult periodontitis cases are triggered by it. China has become the biggest tobacco consuming country in the world. Its tobacco consumption increases at an annual rate of 5.3%. China's tobacco yield is equal to the sum production of the other 7 biggest tobacco producing countries. About 160 thousand billion cigarettes are sold in China each year, accounting for over 30% of the world total tobacco consumption [12] [13] . On the other hand, in recent years, the incidence rate of diabetes has been growing in China. According to a survey of the Chinese Academy of Endocrinology, the prevalence of diabetes is reaching 11% in urban and suburban areas. The sufferers of diabetes in China are ranked second in the world. According to the WHO assessment, there will be about 50 million diabetes patients in China by the year of 2025. According to summary of the Chinese Academy of Orthodontics, the prevalence rate of malocclusion has increased by 20% from 1950-60s [14] (Table 8 ). According to the third national epidemiological survey on oral diseases conducted in 2005, the prevalence of edentulism in 65-74 age group was 7%, lower than the finding of 1995 investigation (11%), indicating that the elderly have more remaining teeth than before. However, considering the criteria of this survey (any remaining tooth was excluded regardless of the treatment need for extraction), the actual rate should be higher [8] [9] . The prevalence of dental erosion is relatively high and appears to be growing. Statistics show that the prevalence for the 65-74 age-group is 30.2%-87.6%. With the increased consumption of sugar-containing soft drinks, the incidence for children aged 3-5 has even reached 9.3% [15] [16] . Only regional surveys contain information about oral cancer [17] [18] [19] (Table 9 ). Calculated at a prevalence rate of 81-41.2 per 100 thousand, there are 100 thousand oral cancer victims in China, with 10 thousand patients dying each year. Besides, oral soft tissue diseases, oral and facial injuries and trauma, tooth abrasion, mucosal infections and diseases, developmental disorders, craniofacial anomalies are very common. Suffering from multiple chronic diseases and more other health problems, having complicated medical and dental problems, adults especially elderly people present complicated medical problems. More than 70% elderly people take medicine, which will affect doctor's diagnosis and treatment. The treatment for old people is particularly complex. Dental therapies are frequently intricate, and the elderly also have chronic medical conditions, which complicate patient management. Dentists will face more complicated treatment and medically complicated or compromised patients. With the economical development, the dental service need will increase for the people and patient. As people retain more teeth into old age, the need for all kinds of oral health services will continue to increase. Dentists will be faced with more difficult cases in the future. At the same time, patients' expectations keep growing, their need for oral care, oral health products and medicals expand quickly. These special needs and trends will affect China's oral product market for a considerable time. The oral health care system is an arrangement that combines the patients with the service provider. So far, health care services in China are mainly paid through the ways listed in Table 10 [3 -5] . Although there are no statistics about oral health care expenses, it's believed that over 85% of the total expense is paid by the patients themselves. In recent years, the number of dentists and that of oral health education institutions have grown rapidly (Table 11) , however, the utilization efficiency of oral health resources remains low. Difficulty in visiting the doctor and paying for medication remains to be serious problems that concern people's livelihood. According to Qiang Gao, the former minister of Ministry of Health of P. R. China, at present, the development of China's health care system is faced with important opportunities as well as problems. It is an important historical mission for us to strengthen health care system reform, increase the governmental input for more preventive services, products and programs, to establish examples of successful prevention programs. Comparing the findings of the national epidemiological investigations in 1995 and in 2005, it is indicated that the oral health status of the whole population will not be improved by only increasing the number of dentists and oral health institutions. Prevention as a public policy measure should be underlined in the oral health care system. New policies must have the necessary resources to translate need, currently not met, into effective demand. Policy should be developed to emphasize dental prevention and insurance reimbursement for preventive services. It should be emphasized that the reform start with innovating the structure of oral health care system, improve public health, rural health, and enhance the construction of the medical care system in urban communities. Meanwhile, it is important to pro-mote dental care professionals to pinpoint prevention service and to gear people to participate in preventive services. Insurance should cover the field of oral preventive services. Successful experiences of preventive model should be explored. In recent years, Beijing and Shanghai municipal governments have increased the inputs in oral health care for students, making a good start. Nowadays, the effectiveness of present oral preventive methods has been proved by the evidence-based medicine. Therefore, the use of fluorides, pit and fissure sealing, oral health products and medicines should be advocated. After early diagnosis is made, the process of oral diseases can be monitored and specific advice can be given to patients. The significance of technology in decision making should not be overlooked in continuing education. Furthermore, the real challenge lies in the use of innovations rather than the conventional surgeries. Recently, the new concept of minimal invasive dentistry (MID) has become the theme in the clinical practice of dentistry. The former U.S. Secretary of Health and Human Services Dr. Louis Sullivan said: getting patients to practice better health habits is the greatest challenge in medicine. The emphasis is on promoting health, rather than preventing diseases. Oral health education and promotion through community and clinical practice will play a leading role in the future public health initiatives. Changed demographics, disease trends and scientific advances are stimulating the need of pharmaceuticals and oral health products. More patients require dental medicine rather than traditional dental treatment. Oral pharmaceuticals and dental health products are increasingly used for diagnosing, treating and preventing oral disease. Therefore, pharmaceutical solutions will play a greater role in oral care because they are passive, economic and preventive rather than surgical care. Medical agents can reach people of high caries risk who fall outside the dental delivery system and using these products require less contact with dental care providers. Medical interventions have a potential market in both urban and rural areas. Dentists will be able to save time and patient care costs. Toffler et al. [20] argued in their book, Revolutionary Wealth, that novel technologies are fortifying the fusion of producers and consumers, creating a new class, the prosumers. They pointed out, consumers can take advantage of new treatments brought about by nanotechnology and other new technology to accomplish the task that used to be executed by doctors, namely diagnosis and treatment. Such revolution will change the functioning manner of the entire business in the arena of oral health area in the future. Just as the SARS virus and the bird flu virus can be transported around the world in hours, health care information can be transmitted from one corner of the globe to another in seconds. Health care is a global concern that breaks down national boundaries. New scientific findings and technologies can arise anywhere in the world. The globalization of the health system will surely affect the area of oral health. Success in preventing and controlling oral disease in China is increasingly dependent on the ability to share knowledge and expertise with others around the world in a free and open manner. Dentistry in China must be fully involved in international organizations and activities for research, education, clinical practice, product development and distribution, and health promotion. In the same time, China will benefit from international cooperation and collaboration.

National intensive care unit bed capacity and ICU patient characteristics in a low income country

BACKGROUND: Primary health care delivery in the developing world faces many challenges. Priority setting favours HIV, TB and malaria interventions. Little is known about the challenges faced in this setting with regard to critical care medicine. The aim of this study was to analyse and categorise the diagnosis and outcomes of 1,774 patients admitted to a hospital intensive care unit (ICU) in a low-income country over a 7-year period. We also assessed the country’s ICU bed capacity and described the challenges faced in dealing with critically ill patients in this setting. FINDINGS: A retrospective audit was conducted in a general ICU in a university hospital in Uganda. Demographic data, admission diagnosis, and ICU length of stay were recorded for the 1,774 patients who presented to the ICU in the period January 2003 to December 2009. Their mean age was 35.5 years. Males accounted for 56.5% of the study population; 92.8% were indigenous, and 42.9% were referrals from upcountry units. The average mortality rate over the study period was 40.1% (n = 715). The highest mortality rate (44%) was recorded in 2004 and the lowest (33.2%) in 2005. Children accounted for 11.6% of admissions (40.1% mortality). Sepsis, ARDS, traumatic brain injuries and HIV related conditions were the most frequent admission diagnoses. A telephonic survey determined that there are 33 adult ICU beds in the whole country. CONCLUSIONS: Mortality was 40.1%, with sepsis, head injury, acute lung injury and HIV/AIDS the most common admission diagnoses. The country has a very low ICU bed capacity. Prioritising infectious diseases poses a challenge to ensuring that critical care is an essential part of the health care package in Uganda.

The prevalence of critical illness in developing countries is disproportionately high in view of the disproportionate burden of diseases such as HIV/AIDS, malaria, tuberculosis and trauma. Sub-Saharan Africa bears 25% of the global burden of disease [1] . Management of critically ill patients requires significant human, infrastructural, and financial resources. These resources are typically limited in low-income countries. Major intensive care units (ICUs) are mostly found in large hospitals in urban or metropolitan areas [2] . The most common admission criteria to these units are post-operative treatment, infectious diseases, trauma and obstetric complications [2, 3] . A recent review highlighted the paucity of knowledge regarding critical care in the developing world [4] . Knowledge of the characteristics and outcomes of critically ill patients admitted to ICUs in low-income countries may help with the identification of priorities and the resources required for improvement of the care of critically ill patients in resource-limited regions of the world. The aim of this study was to determine the admission diagnoses and outcome of patients admitted to the Mulago Hospital ICU from 2003 until 2009 and to highlight the country's ICU bed capacity. It is hoped that the findings will be a useful addition to the increasing body of evidence highlighting the disparities between critical care in high-and low-income countries. This study was a retrospective audit. The study protocol was approved by the hospital Research and Ethics Committee. Medical charts were reviewed and anonymity was preserved for each case record. The study intensive care unit is a 12-bed unit with the ability to ventilate only six patients at any one time. It provides level II ICU services to all kinds of critically ill patients. Level II care includes mechanical ventilation for longer than 24 h, and specific organ support like dialysis and inotropic infusions. The ICU can provide mechanical ventilation, post-operative care, intermittent haemodialysis, peritoneal dialysis, and basic neurocritical care. The ICU serves Mulago Hospital, which is a 1,500 bed national referral hospital, and Makerere University teaching hospital. The unit was started in the late 90s with a foreign donation and was initially run by a UK trained anaesthesiologist who has since retired. The ICU is currently staffed by three full-time ICU doctors (two internists and an anaesthesiologist) who have undergone further training in higher income countries, and 20 nurses. It receives technical support from the Department of Anaesthesia. Apart from the study ICU, Mulago hospital also has a four-bed cardiac ICU, a four-bed coronary care unit (the heart institute is a semi-autonomous unit within the hospital that caters for paying patients and open-heart surgery patients), a six-bed paediatric high dependency unit, a new five-bed obstetric high dependency unit and a neonatal special care unit that can only provide nasal CPAP ventilation. No unit in the hospital can ventilate infants or neonates. Currently, the study ICU uses early warning score criteria to admit patients, together with a first come first served basis system, due to the limited number of beds. It is estimated that about ten critically ill patients are deprived of ICU admission daily, and as a result succumb to their illnesses. An ongoing study is being conducted in the hospital to identify missed opportunities for saving such patients. The audit included all patients admitted to the study ICU from January 1, 2003, until December 31, 2009. No patient was excluded from the study. The following information was recorded for each study patient: basic demographic data (including age and gender), admission criterion, duration of stay in the ICU, and survival to ICU discharge. We also conducted a telephonic survey to establish the ICU bed capacity in the whole country. Basic descriptive statistics were used to analyse demographics data and other study variables. Logistic regression analysis was used to determine the association between different durations of ICU stay and survival to discharge. P-values <0.05 were considered statistically significance. Data are presented as mean values, with standard deviations, unless otherwise indicated. For the purposes of the telephonic survey, an ICU bed was defined as comprising a bed, a pulse oximeter, a mechanical ventilator, a suction machine and an anaesthesia provider in the vicinity. We determined that, based on our definition, there were 33 ICU beds in the whole country for a population of 33 million people (Table 1) . During the study period, 1,774 patients were admitted to the study ICU ( Table 2 ). The mean age of the study patients was 35.5 ears. The majority of the patients (56.5%) were male. Indigenous Ugandans accounted for the majority (92.8%) of the patients. Upcountry referrals constituted 42.9%, and the remaining patients were from within and around Kampala, the capital city of Uganda. The mean mortality rate over the 7-year period was 40.1% (n = 715) ( Table 3 ). The highest mortality rate (44%) was observed in 2004; the lowest mortality rate (33.2%) was observed in 2005. Children (age <18 years) accounted for 11.6% of admissions, and their mortality rate was 40.1%, with paediatric post-operative admissions being higher than paediatric medical admissions. Sepsis, ARDS, traumatic head injury, and HIV/AIDS were the most frequent admission diagnoses during the study period (Table 2) . Neurosurgical conditions accounted for the ICU admission diagnosis with the highest mortality. Patients who stayed in the hospital for 6 to 10 days were three times more likely to survive compared with patients who stayed for 1 to 5 days. Patients who stayed for 11-20 days were at twice as likely to surviveas likely to die compared with patients who stayed for 1 to 5 days ( Figure 1 ) (Table 4 ). Patients who stayed for >20 days were almost twice as likely to survive compared with patients staying for 1 to 5 days. In this retrospective audit, we aimed to determine admission patterns in our ICU during a 7-year observation period. We found that the two most common admission diagnoses were identical to those reported by ICUs located in other parts of the world [5] [6] [7] . The overall mortality rate of 40.1% is comparable to reports from other African country ICUs [5] , but much higher than that reported by ICUs in high-income regions of the world (at between 10-20.9%) [6] [7] [8] . Head injuries were a common reason for ICU admission and associated with the highest mortality rates in this audit. This is not surprising, considering that the study ICU is a general ICU and does not have specialised neurocritical care resources (e.g. facilities to measure intracranial pressure or arterial blood gases). This is despite the ICU being served by four neurosurgeons; therefore the limitations are related to infrastructure rather than skills or personnel. It was difficult to determine what proportion of deaths was preventable because reliable data for this was not available. A paper by Mock et al. estimated that improved trauma systems can avert between one and two million deaths a year in severely injured patients in low-and middle-income countries [9] . The lack of neurocritical monitoring equipment is coupled with the fact that Uganda does not have a functional emergency medical response system. This leads to inadequate transportation of trauma victims to health care facilities and delays in definitive care. There are a limited number of ICU beds in Uganda as a whole -only one ICU bed for every one million Ugandans or 0.1 ICU beds/100,000 (Table 1) . This compares poorly with South Africa (8.9/100,000), Sri Lanka (1.6/ 100,000), and the United States of America (20/100,000) [1] . This also explains the high number of referrals to Mulago hospital from upcountry centres. This limitation is further compounded by a well-documented dearth of anaesthesiologists-a critical human resource for intensive care units [10, 11] . Adequate emergency care at a crash scene (e.g. airway management, positioning, oxygen and fluid resuscitation) is known to improve trauma outcome [12] . The high number of non-helmet wearing motorcycle riders in Uganda, and in Kampala in particular, also contributes to the high injury severity and mortality rate of neurotrauma observed in this study [13] . Sepsis was also a common cause of mortality, with mortality rates higher than those reported from industrialised countries [6] [7] [8] . Although our study data cannot explain the high mortality rates associated with sepsis, it is likely that insufficient early sepsis care may have contributed. Delayed presentation of sepsis patients to the hospital, and subsequently to the ICU, is common [14] . The paucity of resources to manage patients with sepsis (e.g. insufficient amounts of fluids, unavailability of intravenous broad-spectrum antibiotics and unavailability/unreliability of microbiological diagnostics) may have prevented adequate sepsis management at the study ICU. Recently, an expert group published guidelines to help resource poor settings manage critically ill patients with sepsis [15] . The recommendations have been well received in a number of resource limited countries. The patient population included in this study is younger compared with patients admitted to ICUs in industrialised countries [16] . However, our findings are similar to those reported from surveys of critically ill patients treated in other African countries, where life expectancy is comparably low to that of Uganda [5] . Similarly, the mean length of ICU stay in this study resembled that in other parts of Africa. The finding that patients who stayed 6 to 15 days in the ICU experienced better survival to discharge than those treated for less than 5 days or longer than 2 weeks indicates that patients in the study ICU typically die early (within a few days) or relatively late (after 2 weeks). Early deaths can most likely be explained by the lack of trained staff and resources to provide adequate care for critically ill patients with a high disease severity (e.g. those with brain trauma, shock or sepsis). Children accounted for 11% of all ICU admissions with a mortality rate of 40%. This is similar to other African country ICUs [17, 18] , but considerably higher than in industrialised countries [19] . The lack of ventilators and accompanying resources in the paediatric high dependency unit at the Mulago hospital is one of the main reasons why children are admitted to the study ICU. Although our study cannot prove a causal relationship, it is likely that delayed initiation of mechanical ventilation and aggressive resuscitation could explain the high death rate in the paediatric patients in the study population There was a higher mortality in the paediatric medical group than in the surgical group, and we hypothesise that this is because a lot of the post-operative patients were elective surgical patients who were admitted for observation. Most paediatric referrals were, and continue to be, children with acute respiratory failure who are transferred from the paediatric high dependency unit because they are in need of mechanical ventilation. The relatively younger population in LICs and the fact that respiratory illness is the leading cause of deaths in under-5-year olds in such countries [20] , implies that more emphasis should be placed on strengthening paediatric critical care resources in LICs. Previous studies have suggested the need to estimate the cost effectiveness of critical care in this setting, given the relatively younger and economically active population. [1, 2] The fact that HIV is endemic in Uganda explains why HIV/AIDS was one of the most common reasons for admission in the study population. Due to the advent of easily accessible highly active anti-retroviral therapy, together with septrin prophylaxis, the incidence of HIVrelated diseases (such as pulmonary infection with Pneumocystis jiroveci, which usually presents as acute respiratory failure) has markedly decreased [21, 22] . In this survey, it was difficult to retrospectively determine from the medical records whether acute respiratory failure was due to infection or other causes. We could, however, determine that chronic obstructive pulmonary disease was a very rare cause of acute respiratory failure in our setting. Other rare HIV-related causes of ICU admission were viral encephalitis and liver failure. Obstetric admissions in our study were largely due to perioperative cardiac arrest occurring as a consequence of peripartum haemorrhage, eclampsia and/or sepsis. Following the introduction of protocolised care for peripartum emergencies and the establishment of the obstetric high dependency unit (patient monitors and more intense nursing and protocols without mechanical ventilation) at the Mulago hospital, the number of obstetric critically ill patients admitted to the study ICU dropped substantially. Limitations of this study include its retrospective nature with the consequence that it could not provide the same level of evidence as a prospective survey. Furthermore, due to the concise format of medical records, only limited data could be retrieved for this audit. For example, information on whether patients received mechanical ventilation; the volume of fluids; and drugs was not available. According to anecdotal evidence, 99% of all admissions are mechanically ventilated; however, the lack of data to support this precludes us stating this as a fact. Other ICU-relevant data would have allowed better description of the study population. More detailed data may have allowed for examination of other variables associated with mortality. It is hoped that advances in health information technology in low-income countries will result in improved reporting ability in the future. This is the largest study to date of critically ill patients in a low-income setting in sub-Saharan Africa. Our ICU study population is a young one and, even though we have limited data for comparison, high-income countries may have an older ICU population. We had a mortality rate of 40.1%, with sepsis, head injury, acute lung injury and HIV/AIDS the most common admission diagnoses. The mortality rate stayed the same over time, possibly because the admitting doctors stuck to their prognoses, and there are limitations in resources and a paucity of use of evidence-based practice. The fact that half the patients came from outside of the capital city is explained by the dearth of ICU beds in the country as a whole. Critical care remains a neglected area of health service delivery in this setting, with large numbers of patients with potentially treatable conditions not having access to such services. Further research needs to be carried out in ICUs in other resource limited settings, including a prospective study to estimate the resources required to design resource appropriate units in such settings and the impact on morbidity and mortality, especially for the most common conditions.

Influenza Virus Infection in Nonhuman Primates

To determine whether nonhuman primates are infected with influenza viruses in nature, we conducted serologic and swab studies among macaques from several parts of the world. Our detection of influenza virus and antibodies to influenza virus raises questions about the role of nonhuman primates in the ecology of influenza.

W orldwide, infections with infl uenza A viruses are associated with substantial illness and death among mammals and birds. Public health and research have placed major focus on understanding the pathogenicity of different infl uenza virus strains and characterizing new infl uenza vaccines. Nonhuman primates (NHPs), including macaques, have become popular experimental models for studying the pathogenesis and immunology of seasonal and emerging infl uenza viruses. NHPs readily seroconvert after experimental inoculation with seasonal infl uenza virus and have been used to test candidate vaccines for strains of human and avian origin. Like humans, macaques infected with infl uenza virus exhibit fever, malaise, nasal discharge, and nonproductive cough; virus replication can be detected in the nasal passages and respiratory tract (1, 2) . However, whether NHPs are infected with infl uenza viruses in nature remains unknown. Over the past decade, we have focused on the role of pet and performing monkeys in disease transmission throughout Asia. Commonly trapped in the wild, these monkeys might be sold at wet markets, the putative source of several zoonotic outbreaks (3), where they might be caged next to any number of animal species (Figure 1 , panel A) (4). Pet and performing monkeys are likely conduits for cross-species transmission of respiratory pathogens like infl uenza viruses because of their close and long-term contact with their owners, audiences, domestic animals, wild animals, and birds ( Figure 1, panel B) (5) . However, the breadth and diversity of this interface presents a challenge for monitoring the emergence of infectious diseases. We have approached this challenge by conducting longitudinal studies at several sites and collecting biological samples and behavioral data representing various contexts of human-NHP contact (4-7). We used these historical and newly acquired samples, representing various countries and contexts of human-macaque contact, to determine whether NHPs are infected with infl uenza viruses in nature. As part of our decade-long longitudinal studies, ≈200 serum samples were collected from macaques. These included pet macaques (Macaca nigra, M. nigrescens, M. hecki) from Sulawesi, Indonesia; performing macaques from Java, Indonesia (M. fascicularis) and from Bangladesh (M. mulatta); M. fascicularis macaques from the Bukit Timah and Central Catchment Nature Reserves in Singapore, where they freely interact with wild avian fauna and visitors (occasionally entering residential areas) (7); M. sylvanus macaques from the Upper Rock Nature Reserve in Gibraltar, where international tourists frequently use food to entice the macaques to climb about their heads and shoulders (6) ; and free-ranging macaques (M. fascicularis and M. nemestrina) at temple shrines or M. fascicularis macaques that range throughout a wildlife rescue center and nearby villages in Cambodia ( Figure 2 ). Serum was collected and stored as described (8) . All samples were stored on cold packs in the fi eld and transferred to dry ice for shipment to the United States, where they were then stored at −80°C. For initial screening for antibodies against infl uenza virus, serum samples were treated with receptor-destroying enzyme as described (9) and tested by using a multispecies Infl uenza A Virus NP Antibody Inhibition Test (Virusys Corporation, Taneytown, MD, USA) according to manufacturer's instructions. ELISA results indicated nucleocapsid protein antibodies against infl uenza in samples from macaques from Cambodia (29.2%), Singapore (16.7%), Sulawesi (16.1%), Bangladesh (13.3%), and Java (6.0%) ( Table 1) . Antibodies were detected in animals 1-10 years of age at the time of sampling. No infl uenza virusspecifi c antibodies were detected from the 73 total samples from Gibraltar, perhaps because persons with infl uenza virus infection infrequently travel to the Upper Rock Reserve (healthy-visitor effect) (10) or perhaps because monkeys from Gibraltar are less susceptible to infection. Seroprevalence of antibodies against infl uenza A virus, by site and collection year, human and NHP population densities, and prevalence of avian infl uenza viruses are shown in Figure 2 . Serum samples that were positive by ELISA were also screened by hemagglutination-inhibition assay as described (9) . Based on the year and location of NHP sample collection, the estimated ages of the NHPs at the time of sample collection, and the presence of avian H5 and H9 infl uenza viruses in many of these countries during the sampling period (11) (12) (13) , a panel of human vaccine strains and avian infl uenza virus strains was used in the hemagglutination-inhibition assay. Although not all ELISA-positive serum samples could be subtyped, antibodies against seasonal subtype H1N1 and H3N2 infl uenza A strains were detected from macaques in Bangladesh, Singapore, Java, and Sulawesi (Table 2) . Of the performing macaques in Bangladesh, 2 had antibodies against A/chicken/Bangladesh/5473/2010, a strain of G1 clade subtype H9N2 avian infl uenza virus. Subtype H9N2 infl uenza viruses are prevalent in poultry in Bangladesh (14) and have been detected in humans (12) . We did not detect antibodies against highly pathogenic avian infl uenza subtype H5 viruses, which might not be surprising given our relatively small sample size (Table 2) . Also given the small sample size, we were unable to perform microneutralization studies, which would be useful to perform with future samples. In 2011, to determine whether any macaques were actively infected with infl uenza virus, we collected oral swabs from 48 monkeys in Cambodia to test for infl uenza virus by real-time reverse transcription PCR as described (8) . In brief, the inside of the anesthetized and immobilized monkeys' mouths (cheeks, tongue, and gums) were swabbed. Swabs were immediately placed into viral transport media, stored, and shipped as previously described. RNA was isolated by using an Ambion MagMAX-96 AI/ND Viral RNA Isolation Kit (Life Technologies Corporation, Grand Island, NY, USA) on a Kingfi sher Flex system (Thermo Fisher Scientifi c, Waltham, MA, USA). Viral RNA was analyzed in a Bio-Rad CFX96 Real-Time PCR Detection System and a C1000 Thermocycler (Bio-Rad, Hercules, CA, USA) with TaqMan Fast Virus 1-Step Master Mix (Applied Biosystems, Foster City, CA, USA) and the InfA primer/ probe sets as described (15) . Of the 48 respiratory samples, 1 (2.1%) was positive for infl uenza virus; cycle threshold value was 38 (limit of detection is 40). Attempts to amplify longer PCR fragments of matrix, hemagglutinin, or neuraminidase genes or to isolate the virus by blind passage in embryonated chicken eggs or MDCK cells were unsuccessful. Our results indicate that NHPs that have contact with humans can be naturally infected with seasonal endemic human infl uenza viruses and with emerging pandemicrisk avian infl uenza viruses. We found serologic evidence of infection in several countries, contexts, and macaque species. Preliminary real-time reverse transcription PCR results also pointed to the presence of virus in a buccal swab from an adult macaque from Cambodia, indicating active infection at the time of sampling. On the basis of results from this study, it seems that pet, performing, and free-ranging macaques are susceptible to infl uenza virus infection. Given the close relationship between humans and NHPs in areas of the world where avian and human infl uenza viruses cocirculate, further surveillance of these populations is warranted. The ability to detect and eventually isolate strains of infl uenza virus currently infecting NHPs and humans at the animal-human interface is paramount to understanding how NHP-human interactions can affect the genetics, transmission, and public health risk for infection with infl uenza A viruses.

Label-Free Electrochemical Diagnosis of Viral Antigens with Genetically Engineered Fusion Protein

We have developed a simple electrochemical biosensing strategy for the label-free diagnosis of hepatitis B virus (HBV) on a gold electrode surface. Gold-binding polypeptide (GBP) fused with single-chain antibody (ScFv) against HBV surface antigen (HBsAg), in forms of genetically engineered protein, was utilized. This GBP-ScFv fusion protein can directly bind onto the gold substrate with the strong binding affinity between the GBP and the gold surface, while the recognition site orients toward the sample for target binding at the same time. Furthermore, this one-step immobilization strategy greatly simplifies a fabrication process without any chemical modification as well as maintaining activity of biological recognition elements. This system allows specific immobilization of proteins and sensitive detection of targets, which were verified by surface plasmon resonance analysis and successfully applied to electrochemical cyclic voltammetry and impedance spectroscopy upto 0.14 ng/mL HBsAg.

In recent years, various types of biosensors have been increasingly becoming practical and useful tools in a wide variety of analytical devices [1, 2] . The immobilization of biological elements to realize the biosensor is an essential step for the successful construction of a diagnostic system. In order to allow the detection of a small amount of target sample and improve detection performance, bioreceptor proteins must be immobilized onto biosensor chip surfaces with high density and nonspecific adsorption avoided or at least minimized. Moreover, orientation control with retention of protein conformation and activity is a required task to be established [3] . One method involves the physical adsorption via van der Waals forces, ionic binding or hydrophobic and polar forces on an insoluble matrix. This is a simple process which causes little disruption of the proteins, while it is unstable during the binding procedure due to the highly dependency against environmental conditions in maintaining its functional characteristics. Thus, the resulting receptor layer seems to form heterogeneous and random orientation. Another method can also be constructed by crosslinking functional reagents by a certain number of functional groups due to its simple procedure and strong chemical bond of proteins. This is widely used for stabilization of proteins that are covalently bound onto the support platform generated by chemical treatment. However, this method has also disadvantages as follows: the difficulty in controlling the crosslinking reaction, the gelatinous nature of the proteins and the relatively low activity of the proteins due to the specific structural features [3, 4] . In each of these methods used for the favorable performance of biosensors, a number of factors deserve careful consideration for strong binding forces between the solid surface and recognition elements, exposure of active sites on the target sample, conservation of biological activity after immobilization process, and simple protocol [3, 4] . Many researchers have studied in vivo combinatorial biotechnology, e.g., either phage or cell-surface display techniques, and developed polypeptide sequences, which can specifically bind to metals [5] [6] [7] , oxides [8, 9] , and semiconductors [10] [11] [12] . Among them, gold-binding polypeptide (GBP) is one of the genetically engineered proteins for a strong binding onto the gold surface [7, 13, 14] . Whereas many proteins and chemicals bind to the gold surface with thiol linkage, GBP does not contain a cysteine residue having thiol group. Although the definite mechanism is not clear yet, it is estimated that these polar groups in GBPs seem to coordinately interact with the gold surface within a monolayer [15, 16] . In addition, the kinetics and thermodynamics of biomimetic interactions between the GBP, and the gold surface were investigated by surface plasmon resonance (SPR) [14, 17, 18] . Compared with other thiol-based systems, the GBP binds tightly to the gold surface due to the lower standard Gibbs free energy for the bond, and the binding process is fast under aqueous conditions compatible with biological environments [14, 17, 19] . These characteristics suggest its potential applications in nano-and bio-technologies as novel agents for surface functionalization [13] . Moreover, immobilization with correct orientation of biological material is a problem of prime importance in biosensors. We employed GBP-fusion proteins in the construction of biosensor for the detection of hepatitis B viral surface antigen (HBsAg) as a model (Figure 1 ), which is a biomarker for diagnosing the hepatitis B virus (HBV). The strong affinity between the GBP and the gold surface guarantees the stability of this sensor system and orients the sensing parts outward from the solid surface, exposing them directly to the target sample [13, 16] . Furthermore, electrochemical detection has attracted considerable interest recently for miniaturized analytical systems [20, 21] , including remarkable sensitivity (approaching that of fluorescence), inherent miniaturization of both the detector and control instrumentation, independence of optical path length or sample turbidity, low cost, low-power requirements and high compatibility [22, 23] . Besides, one of the most attractive points of this method is its potential for portable assays in a variety of point-of-care testing (POCT) environments. We here developed a simple platform biosensor technology by mediating the recognition parts and the solid surface on the gold substrate. SPR analyses were used for optimization of sample concentrations and verification of target sensing. Electrical signal-based detection methods for HBV such as electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) were developed on the gold electrode surface, which has been a very versatile material in the field of biosensors. Restriction enzymes were purchased from New England Biolabs (Beverly, MA, USA). Agarose was from Cambrex BioScience Rockland (Rockland, ME, USA). 30% (w/v) acrylamide/bis solution and protein assay were purchased from Bio-Rad (Hercules, CA, USA). HBsAg PreS2 peptide (H 2 N-NSTTFHQALLDPRVRGLYFPAGG-COOH) was synthesized at Peptron (Daejeon, Korea). Ni-NTA affinity kit was from Qiagen (Hilden, Germany). Other chemicals and reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA), unless otherwise stated. All oligonucleotides were synthesized at Bioneer (Daejeon, Korea). Polymerase chain reaction (PCR) experiments were performed with a PCR Thermal Cycler (Bio-Rad) using High-Fidelity PCR System (Boehringer Mannheim, Mannheim, Germany). DNA sequences were confirmed by automatic DNA sequencer (ABI Prism model 377, Perkin Elmer, Grove, IL, USA). Cell growth was monitored by measuring the absorbance at 600 nm (OD 600 ; DU ® 650 spectrophotometer, Beckman, Fullerton, CA, USA). Cells were disrupted by using ultra-sonicator (Braun Ultrasonics, Danbury, CT, USA). SPR experiments were performed by using SPRLAB™ system (K-MAC, Daejeon, Korea) and BIAcore3000 (GE Healthcare, Uppsala, Sweden). Electrochemical detection analysis was carried out using CHI 660C Electrochemical Analyzer/Workstation (CH Instruments, Austin, TX, USA). The flow-chart for the construction of the fusion protein between the GBP and its single-chain antibody (ScFv) is shown in Figure S1 (Supplementary material). The 6HGBP-ScFv fusion protein was prepared by genetically fusing the GBP and ScFv, allowing two specific interactions between GBP and gold substrates, and the capture of HBsAg and its ScFv, respectively. For easy purification of the fusion protein by metal affinity chromatography, the coding sequence of a six-histidine (6H) was introduced at the N-terminus of the GBP. For the cloning of the fusion gene, the DNA fragments encoding 6histidine-fused GBP (6HGBP) were obtained by PCR amplification using plasmid pTacFadLGBP-1 [16] as a template, and P1 (5'-AAAATACCATATGGGCCACCATCACCATCACCACGG-3') and P2 (5'-TTCCCCATGGAGACGAATGGTACCGCTCGT-3') as primers. The PCR product was digested with NdeI and NcoI, and ligated into the same sites of pET-22b(+) (Novagen, San Diego, CA, USA) to make pET-6HGBP. For the cloning and expression of the 6HGBP-ScFv fusion gene, the DNA sequence encoding ScFv fragment was amplified by PCR using plasmid pET-ScFv-SBD [18] as a template, and P3 (5'-CAAGACCATGGGTGTCGACTGAGGAGTCTGGA-3') and P4 (5'-TCCGCTCGAGACGTTTTATTTCCAGGTAGGT-3') as primers. This PCR product was digested with NcoI and XhoI, and ligated into the same sites of pET-6HGBP to make pET-6HGBP-ScFv. ) gal dcm (DE3)] was used as a host strain for the expression of GBP-fused ScFv fragment (GBP-ScFv). Recombinant E. coli BL21(DE3) strain harboring pET-6HGBP-ScFv was cultivated in 250 mL flasks containing 100 mL Luria-Bertani medium supplemented with 2% (w/v) glucose and 100 μg/mL of ampicillin at 37 °C in a rotary shaking incubator. Cell growth was monitored by measuring the absorbance at 600 nm. At an OD 600 of 0.4, isopropyl-β-D-thiogalactopyranoside (IPTG) was added to a final concentration of 1 mM to induce the gene expression. Then, cells were further cultivated for 6 h and harvested by centrifugation. Cells were disrupted by sonication for 1 min at 40% output in cell lysis solution (Tris-NaCl buffer containing 6 M GuHCl and 5 mM imidazole, pH 8.0) and centrifuged at 16,000× g for 10 min at 4 °C. The pellet containing insoluble proteins was denatured, purified and underwent dialysis for further experiments. Since the 6HGBP-ScFv fusion protein contains 6H tag at N-terminal, they could be simply purified using Ni-chelating resin (Qiagen, Valencia, CA, USA) without further purification step. Protein concentration was determined by Bradford's method using bovine serum albumin (BSA) as a standard. The SPR experiments were carried out at 25 °C by using a repeated angle mode and a fixed angle mode. The repeated mode is the method for measuring the changes of the minimum resonance angle in specially fixed angle range by repetitive angle scanning and fitting, which is a plot of change in the reflectance intensity as a function of time (Resonance vs. Time). SPRLAB™ (K-MAC, Korea) system detects a refractive index and thickness change containing semiconductor laser source with 635 nm wavelength and dielectric silicon photodiode detector, which has incident angle range of 30-80 degree. For the HBsAg detection, the concentration of 6HGBP-ScFv fusion protein was roughly optimized at first. Sequential injection of solutions was as follows: For equilibration, phosphate-buffered saline (PBS) solution was flowed over the bare gold surface to wash away any potential contaminants. Then, several different concentrations of 6HGBP-ScFv fusion protein were injected. The optimal concentration of 6HGBP-ScFv fusion protein, determined in this process, was used for detecting various concentrations of target HBsAg in the following experiments. In the process of HBsAg detection, BSA instead of the target was injected as a negative control. Prior to the target binding, 0.5 mg/mL of BSA was injected with a flow rate at 20 μL/min for an effective blocking of nonspecific binding sites. In order to remove nonspecifically bound molecules and unbound samples, washing step was applied intermittently with PBS for about 5 min. For the binding of targets, a fixed flow rate at 5 μL/min and binding time of 5 min was applied, consuming a total sample volume of 100 μL. Electrochemical measurements including EIS and CV were performed on a conventional electrochemical cell equipped with Ag/AgCl with 3 M KCl as a reference electrode, platinum wire as a counter electrode and bare gold electrode as a working electrode. All potentials were referred to the Ag/AgCl reference electrode. Immediately prior to use, working electrodes were cleaned by five cycles of CV in a potential window of −0.5 to 0.5 V to remove any potential contaminants. Then, the clean gold surface was serially modified by sequential immersion in 50 μg/mL 6HGBP-ScFv and 0.5 mg/mL BSA as a negative control for 1 h at room temperature, respectively. In the last step, the modified electrodes were incubated with various concentrations of HBsAg solution for about 2 h. The washing step with DI water and nitrogen gas was performed after each binding event. The electrode fabrication process, namely, confirmation of binding of recognition elements onto the gold electrode was characterized by EIS and CV, while EIS was used for target detection. Electrochemical measurements including EIS and CV were performed in 1 mM ferricyanide in 0.1 M KCl (Nyquist plot). CV experiments were carried out in unstirred solutions at a scan rate of 0.1 V/s and at a fixed potential window of −0.5 to 0.5 V vs. Ag/AgCl. Impedance measurements were carried out in the frequency range from 10 4 down to 0.1 Hz with AC amplitude of 5 mV and bias potential of 0.22 V. To verify the expression level of the 6HGBP-ScFv fusion proteins, the sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of 6HGBP-ScFv fusion protein was carried out as shown in Figure S2 (Supplementary material). Compared with the wild-type E. coli BL21(DE3) strain as a negative control, E. coli BL21(DE3) harboring pET-6HGBP-ScFv expresses a thick band between marker bands of size 27.5 kDa and 37 kDa. This band size is in agreement with the size of 6HGBP-ScFv fusion protein, which was calculated as approximately 31.9 kDa. Since this fusion protein is expressed as an insoluble fraction, it was purified in denaturing condition and sequentially subjected to dialysis. According to the result, the most right band shows the 6HGBP-ScFv fusion protein purified using a Ni-chelating column resulting with good purity and quantity for further use in several sensing platforms. SPR is a label-free and sensitive spectroscopic technique used to study bioaffinity interactions on gold thin films by measuring changes of the refractive index under conditions of total internal reflection [24, 25] . To prove that GBP-fusion proteins could be functionally immobilized on the gold surface, the 6HGBP-ScFv fusion protein was flown over the gold sensor chip to monitor its binding affinity by combining the GBP-fusion approach with SPR. The dynamic and specific binding of fusion proteins onto the gold sensor chip could be directly monitored in real time by Biacore ® SPR spectroscopy ( Figure S3 in Supplementary material). A sharp increase in the SPR signal was observed upon introducing the 6HGBP-ScFv fusion protein solution onto the chip surface. About 93% of the injected total 6HGBP-ScFv fusion proteins were bound onto the gold sensor chip surface. Most 6HGBP-ScFv proteins remained bound to the gold surface after washing with PBS solution. Whereas a little decline in SPR signal during washing is due to the removal of unbound 6HGBP-ScFv proteins from the gold surface, the SPR signal sharply decreased after washing with the PBS solution when ScFv instead of 6HGBP-ScFv was flown over the gold sensor chip. About 23% of ScFv protein was nonspecifically bound to the gold surface compared with 6HGBP-ScFv fusion protein. The binding affinities were calculated by assuming a single binding site model as shown in Table S1 . Strong binding of GBP-fusion protein onto the gold chip surface suggests that various protein-protein interactions can be performed by using this system. To confirm the successful binding of a series of molecules, the detection of HBV was first performed using commercial SPR system. At first, 6HGBP-ScFv concentration was optimized as follows: 100, 50 and 10 µg/mL of 6HGBP-ScFv fusion protein was flowed over the gold surface, respectively. The fusion protein 6HGBP-ScFv was immobilized onto the gold surface via its intrinsic affinity with the gold as shown in Figure 2 . Interestingly, coverage of 6HGBP-ScFv on the gold surface didn't always increase with increasing protein concentrations, a maximal RU change appearing at concentrations of 50 µg/mL. From this result, 6HGBP-ScFv concentration of 50 µg/mL was chosen for subsequent experiments. Next, we checked the specificity of this platform with BSA as a negative control. First, 50 µg/mL of 6HGBP-ScFv was injected into the fluid phase, followed by 0.5 mg/mL BSA blocking of the binding sites on the fusion protein. Then, 10 µg/mL of control peptide (HOOC-CGPTGPTGPTGPTGPT-NH 2 ) as a negative control, 50, 10, 5, 2.5 and 1 µg/mL of target HBsAg were injected into respective channels, respectively. Below concentrations of 50 µg/mL for HBsAg, RU levels dropped rapidly while target concentrations as low as 2.5 µg/mL can be detected compared with the control sample. For 50, 10, 5, 2.5 and 1 µg/mL of target HBsAg, and 0.5 mg/mL of BSA, RU increase is 542.2 RU, 55.7 RU, 48.5 RU, 17.4 RU, 4.9 RU and 5.6 RU, respectively, showing an increasing trend of SPR signals with increasing concentration of target within this scope. To detect HBsAg, electrochemical assay was also carried out, which have traditionally received the major share of the attention in biosensor development for their various advantages such as high sensitivity, specificity and simplicity, and inherent miniaturization of modern electrical bioassays permits them to rival the most advanced optical protocols [26] . Such miniaturization allows packing of numerous microscopic electrode transducers onto a small footprint of the biosensor device, and hence the design of high-density arrays was required. In this paper, EIS and CV were performed in order to characterize the electrode fabrication process, and EIS was applied to detect HBsAg. For the determination of optimal 6HGBP-ScFv coverage, its concentration and volume were carefully calculated in regard to that used in SPR analysis. For the characterization of the electrode fabrication process, EIS and CV were performed each time after binding of each reagent. As shown in Figure 3(a) , the dotted line is the impedance spectrum obtained on the bare gold electrode. Frame circle line is an impedance spectrum of 50 µg/mL 6HGBP-ScFv fusion protein modified electrode, and solid circle line is that of after BSA blocking. Finally, solid line represents an impedance spectrum obtained after HBsAg binding. The result reveals the resistance of the electron transfer at the electrode surface increasing step by step because of the insulating effect of the binding proteins. This electron transfer resistance at the interface on the electrode surface, and solution can be determined by the diameter of the semi-circle of the curves in EIS spectra. After the gold electrodes were immobilized with GBP-ScFv, the peak current decreased dramatically with an increase of the peak-to-peak potential separation (ΔEp) for the bare gold electrode. When HBsAg was bound on the gold chip surface, the peak current more decreased and ΔEp (approximately 110 mV) increased comparing with those of the gold electrode and the GBP-ScFv immobilized chip, resulting from the electron transfer resistance of bound HBsAg molecules. BSA was used as a negative control. It can also be observed that the current responses in CV spectra (Figure 3 (b)) are decreasing in the process of electrode fabrication, which coincides with a conclusion drawn from impedance assay. These results were due to the insulating characteristics of the protein. As shown in Figure 4 , 50 µg/mL, 10 µg/mL, 100 ng/mL, 10 ng/mL and 1 ng/mL of HBsAg were detected by using this method, respectively, and they presented a rough trend of increasing electron transfer resistance with increasing target concentrations. As a negative control, BSA of 10 µg/mL was detected at the same time. Numerical data were drawn by fitting with a circuit model as shown in an inset of Figure 4 . Correspondingly, binding of 50 µg/mL, 10 µg/mL, 100 ng/mL, 10 ng/mL and 1 ng/mL of HBsAg and 10 µg/mL BSA caused electron transfer resistance increase of 8,493 Ω, 6,473 Ω, 4,047 Ω, 3,097 Ω, 2,143 Ω, 1,513 Ω and 777 Ω, respectively. Though linearity is not very strict as a function of HBsAg concentration, it demonstrated a rough linear trend into log scale. Target concentrations as low as 0.14 ng/mL of HBsAg calculated via 3-sigma rule was successfully detected compared with the negative control of 10 µg/mL of BSA ( Figure 4 ). Furthermore, a lower limit of detection (LOD) can be further expected if experimental conditions, such as 6HGBP-ScFv concentration, binding time, temperature and washing condition, were optimized. In order to check the nonspecific binding of real blood sample, we tested a fetal bovine serum of 10 µg/mL instead of HBsAg for clinical trials [27, 28] and confirmed no effect. Several procedures for HBV diagnosis with GBP-fusion protein were developed in this study. First, 6HGBP-ScFv fusion protein was prepared from simple cultivation of recombinant E. coli for HBsAg detection. This GBP-fusion protein allowed for the direct and easy immobilization onto the gold surface. Successful binding of the proteins was demonstrated by SPR optical analysis due to versatile use of gold in the sensing area. Second, this one-step immobilization process onto the solid surface is very simple. Strong affinity between the GBP and the gold surface guarantees the stability preventing possible leakage of sensing proteins. Binding of GBP onto the gold surface also might help to expose sensing molecules outward to react with their targets by arrangement for correct orientation of bioreceptor. Additionally, GBP prevents direct contact between proteins and the gold surface, which is advantageous for protein activity conservation. Furthermore, it is easy to conclude that a device capable of detecting multiple targets can be designed since the bio-recognition elements against targets can be easily fused with GBP via recombinant DNA technology or linker chemistry. Coupling with microfluidics, it can serve to minimize sample volume required as well as to decrease diagnosis time. In summary, we have developed a label-free electrochemical method for HBV detection based on a gold-specific immobilization strategy. The results showed that optimal concentration of bioreceptor was 50 μg/mL in SPR analysis, and the LOD for HBsAg was about 0.14 ng/mL in the electrochemical analysis. This EIS analysis successfully presents the effectiveness of this sensing platform. Regarding intrinsic advantageous property of the electrochemical assay, such as high sensitivity, simplicity and especially its inherent miniaturization characteristics, and a popular trend of development of electricity-based sensors in everyday life, the proposed method presented in this study has a huge potential in commercialization of a POCT device for viral diagnosis. We have to be conducted with clinical trials in the near future to determine its performance.

Development of a Plastic-Based Microfluidic Immunosensor Chip for Detection of H1N1 Influenza

Lab-on-a-chip can provide convenient and accurate diagnosis tools. In this paper, a plastic-based microfluidic immunosensor chip for the diagnosis of swine flu (H1N1) was developed by immobilizing hemagglutinin antigen on a gold surface using a genetically engineered polypeptide. A fluorescent dye-labeled antibody (Ab) was used for quantifying the concentration of Ab in the immunosensor chip using a fluorescent technique. For increasing the detection efficiency and reducing the errors, three chambers and three microchannels were designed in one microfluidic chip. This protocol could be applied to the diagnosis of other infectious diseases in a microfluidic device.

Swine-origin influenza virus, a high-risk human influenza A virus (H1N1), is a serious health threat and potential leading cause of death all around the World [1] . The World Health Organization (WHO) has reported that more than 16,000 cumulative deaths were reported from 213 countries due to H1N1 in February 2010 [2] . Several laboratory diagnostic methods have been developed to monitor the outbreaks of the virus as follows: (1) specific real-time polymerase chain reaction (PCR)-based detection method, (2) isolation of H1N1 influenza virus, (3) detection of 4-fold rise of neutralization antibodies to the virus [3, 4] . However, these methods require highly skilled-personnel and expensive laboratory instruments. In addition, they are not suitable for undeveloped countries because of the limited access to central laboratories and expensive costs. To overcome these issues, microfluidic immunoassay systems have been introduced because of their various advantages, including high throughput, high-efficiency, low-cost and minimized consumption of samples and reagents [5] . After the development of soft lithography techniques using poly(dimethylsiloxane) (PDMS), PDMS has become the most popular microfluidic device materials and offers several advantages such as easy handling, good sealing properties and high optical transparency [6] . However, the poor chemical stability in different types of organic solvents, difficulty in surface modification and mass production have limited the use of PDMS in the various applications [7] . Recently, because of the material issues, some researchers have been attempted to use plastic materials as an alternative solution. Among the various types of polymers, cyclic olefin copolymer (COC) is one of the most popular polymeric materials in the fabrication of microfluidic chips. COC is a well-known polymeric material with various advantages, including high clarity and light transmission, excellent mechanical properties and great biocompatibility [8] . Furthermore, effective immobilization of proteins is essential and important in microfluidic chips to be used as immunosensors. Several methods to immobilize antibodies on the sensor chip surface have been developed, including physical adsorption, covalent binding, and specific interaction between avidin and biotin [9, 10] . However, these previous methods have limitations in terms of denaturization, extra chemical modification and random orientation. In order to overcome these issues, Brown et al. and Park et al. developed specific gold-binding polypeptide (GBP) that endows the orientation of proteins in their functional state [11, 12] . GBP shows a strong binding affinity to the gold surface without any surface modifications [13] [14] [15] . Therefore, GBP-fusion proteins could be selectively and functionally immobilized onto the gold surface. In this study, we carefully designed microfluidic devices, and the surface of a detection chamber was coated with gold for the direct assembly of proteins. A microfluidic-based immunosensor to detect human H1N1 influenza was developed into a low-cost immunosensor based on the exploration of fluorescence signals. The detection of a specific antibody among serological assays in blood samples was performed in the microfluidic biosensor chip by immunoreactions between the GBP-recombinant influenza hemagglutinin antigen (GBP-H1a) fusion protein and its specific antibody (Ab). The GBP-H1a fusion protein as a bioreceptor and the fluorescence-labeled Ab as a marker were used to provide an excellent detection signal. In addition, the chip fabrication and sensing characteristics are reported in detail. COC was purchased from TOPAS Advanced Polymers (Frankfurt-Höchst, Germany). Unless otherwise stated, all chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used without further purification steps. Rabbit anti-H1 polyclonal Ab was obtained from AbFrontier (Seoul, Korea). The microfluidic immunosensor device is fabricated by the injection molding method. The master stamp was fabricated using a micromilling process. The size and thickness of the stamp were 95 × 95 mm and 1.2 mm, respectively. We used a plastic microinjection mold machine (A270C 400-100, ARBURG, Lossburg, Germany) to produce the microfluidic device in order to achieve cost-effectiveness and facilitate mass production. Once the chip was fabricated, chromium (50 Å) and gold (100 Å) were coated on the detection area for further immobilization of GBP-H1a. In this work, the ultrasonic bonding method with an ultrasonic bonder (2000X, Branson, Danbury, CT, USA) was employed to bond the COC chips. For this purpose a melting line with a height of 20 μm was made around microchannels. Once the ultrasonic energy was applied to the COC plates, the sonic energy was intensively localized on the top of the melting line. After the lines were immediately melted and then cooled, both COC plates were tightly bonded with each other to fabricate the plastic-based microfluidic device. Bifunctional fusion protein was created by genetically fusing GBP and H1a, allowing specific interactions between GBP and the gold substrates as well as the capture of H1a and its antibodies. As described in the previous report [16] , the DNA fragments encoding the H1N1 viral surface antigen (H1a) were amplified by PCR with forward primer (5′-CCATGGCATATGGG CCACCATCACCATCACCACGGCAA-3′) and reverse primer (5′-CCGCTCGAGCTGGCTACG CACTTTTTCATACAGGTTTTTAACGTTGCTATCGTGATAGCCGCAAGCTTGTCGACA-3′) for the construction of 6His-GBP-H1a fusion gene. Then, the PCR product was cloned into the NdeI-XhoI fragment of pET-6HGBP to make pET-6HGBP-H1a. Recombinant E. coli BL21 (DE3) strain harboring pET-6HGBP-H1a was cultivated in Luria-Bertani (LB) medium (10 g/L bacto-tryptone, 5 g/L yeast extract, and 5 g/L NaCl) supplemented with 100 µg/mL of ampicillin at 37 °C and 250 rpm. At OD 600 (DU600 ® Spectrophotometer, Beckman Coulter, Brea, CA, USA) of 0.4, cells were induced with 1 mM of isopropyl-β-D-thiogalactopyranoside (IPTG, Sigma) for the production of the fusion protein. After induction, cells were further cultured for 4 h. The cells were then harvested and disrupted by sonication (Braun Ultrasonics, Orlando, FL, USA) for 1 min at 20% output power. After centrifugation at 16,000× g for 10 min at 4 °C, the pellet containing the soluble protein fraction with high-level expressed target fusion-protein was collected for purification of the fusion protein. Because of the 6His tags, a HisTrap TM column (GE Healthcare, Chalfont St. Giles, UK) was used to purify the fusion protein without further purification steps. The protein concentration was determined by the Bradford assay with bovine serum albumin (BSA) as a standard. The binding of the GBP-H1a fusion protein onto the surface of a SPR bare gold chip was characterized by SPR measurement using a BIAcore3000 TM with an automatic flow injection system (Biacore AB, Uppsala, Sweden). All experiments were performed in phosphate-buffered saline (PBS, pH 7.4) at room temperature. A fresh SPR sensor chip was attached to a separate chip carrier for easy assembly in the SPR system. After docking and priming of the SPR chip, PBS was used to flush the activated surface and thereby minimize non-specific binding and any unbound sites by removing loosely bound material and dust. All samples were injected onto the gold chip surface at a flow rate of 5 μL/min for 10 min at room temperature using a liquid-handling micropipette in the SPR system. The surface was then washed and equilibrated using PBS. The GBP-H1a fusion protein (50 μL of a 25 μg/mL solution) was injected onto the SPR chip surface. Before binding anti-H1 Ab, 50 μL BSA of 100 μg/mL was injected to block non-specific binding for 10 min. All SPR sensorgrams were fitted globally using BIA evaluation software (Biacore). The plastic-based microfluidic devices were fabricated by an injection molding technique. The microchannels-embedded master mold was obtained by a milling technique, and it was used for casing the microfluidic devices. In this case, COC was chosen as a substrate material to produce the device due to its great advantages over the other thermoplastic polymers in terms of optical, physical and chemical properties and biocompatibility. The total cycle time of the microinjection molding process took less than 1 min to replicate the COC substrate. After location of the master mold in the injection mold machine, COC was injected through the nozzle, and the microfluidic device was released from the mold. The production process and conditions are similar to the previous research which demonstrated various microinjection conditions to form microstructures [17] . Holes of 1 mm in diameter for an inlet and outlet ports were punched to load the sample and buffer solutions into the microchannels. As shown in Figure 1 , the channel depths are 200 and 500 μm with 450 μm in width, respectively. The diameter of the detection chamber is 1 mm, and it was coated with chromium and gold by sputtering for the immobilization of GBP-H1a fusion protein. The design and fabrication process of microfluidic device are described in Figure 1(A) . The key features of microstructures including welding lines, gold deposition on detection chamber, and backflow prevention structures are schematically demonstrated in Figure 1B . Previously, thermal bonding method has been popular to bond plastic-based microfluidic device. However, this method requires high temperature and takes too much time to bond it. Recently, an ultrasonic bonding method has been developed as an alternative solution to the thermal bonding. However, welding line is essential for bonding between plastic chips in this technique. In this experiment, we carefully designed the welding lines as shown in red color in Figure 1 (B), and these lines were melted during the ultrasonic bonding process for the sealing of the chips. After coating of gold, two COC plates were placed on the ultrasonic bonder (2,000X, Branson) and set the bonder in time mode with weld frequency of 20 kHz. After the setting of the mode, the COC was bonded as following conditions: (1) 800 Pa weld pressure, (2) 0.2 s of weld time, (3) 75% of amplitude, (4) 10 s of hold time, (5) 1.5 kPa of hold pressure. Occasionally, backflow of the buffer or target solutions occurs in microfluidic channels during the injection to cause a contamination problem. Therefore, there would be required a unique microstructure in a microchannel to prevent backflow of solution and contamination of microchannel. Among various methods, two different channel depths were applied into this device as shown in Figure 1(B) . This method was worked properly due to the pressure difference in the microchannel to prevent the backflow. The integrated microfluidic immunosensor chips can reduce the analytical time compared to the conventional methods. In addition, usage of plastic facilitates low-cost mass production of disposable and easy-to-use microfluidic chips. Previously reported plastic-based immunosensors required a surface modification step for the immobilization of biomolecules. Among these methods, silanization using 3-aminopropyltriethyoxysilane (APTES) is a widely used method for coupling of biomolecules onto the inorganic substrate [18] . However, silane is difficult to react with organic materials without any hydroxyl groups present on the surface by using UV/ozone treatment [5] . In order to overcome these issues, a specially engineered peptide was developed and used to immobilize onto the specific targeted surface, especially gold. There is no requirement of complicated steps for coupling of antibody or antigen. The GBP-H1a fusion protein can be simply and selectively immobilized in the microchannel on the gold surface as described in Figure 1(C) . In this device, Y shaped inlet channels with backflow-prevented microstructure and detection chambers were designed for an efficient immunoassay as shown in Figure 2 . The key features of microstructures in microfluidic device were successfully replicated using the microinjection molding, and they were confirmed through top and tilted scanning electron microscopy (SEM, Hitachi S4800, Ibaraki, Japan) images as shown in Figure 2(C-F) . Due to the height differences between inlet channels and main channels (Figure 2(C,D) ), this could prevent the backflow of the solutions. This device is composed of three detection chambers with 1 mm in diameter for the further immobilization of GBP-H1a fusion protein as shown in Figure 2 , and these chambers may reduce the errors during analysis by averaging signals. All protuberant microstructures near to the microchannels were specially designed as welding lines. During the ultrasonic bonding process to bond the top and bottom of the microfluidic chip, these lines would be melted by concentrating the ultrasonic energies on the top of welding lines. To investigate the sensing window of the fusion protein, different concentrations (6.25, 12.5, 25, 50, 100 and 200 μg/mL, respectively) of GBP-H1a fusion protein were immobilized onto the gold chip surface by surface plasmon resonance (SPR) microfluidics. A greater shift in resonance unit (RU) was observed by SPR analysis with increasing concentration of immobilized GBP-H1a fusion protein bound on the planar surface at various concentrations as shown in Figure 3(A) . These results suggest that the SPR sensor with GBP-fusion protein implemented on the gold surface can be an effective system for biomolecular immobilization. Furthermore, the concentration of GBP-H1a fusion protein was fixed to 100 μg/mL due to its best immobilization concentration. For the subsequent binding of anti-H1 Ab, different concentrations (1.5 to 400 μg/mL) of specific antibody were bound to the GBP-H1a fusion protein on the gold sensor chip. The saturated 4,000 RU value obtained with SRP experiments implies that about 4 ng of anti-H1 Ab was immobilized onto the gold surface area of 1 mm 2 . One RU is determined as 0.0001° of resonance angle shift and equivalent to a mass change of the 1 pg/mm 2 on the SPR sensor chip surface [19, 20] . Specific anti-H1 Ab against the H1 influenza surface antigen, which is a hemagglutinin subunit having a high immunogenicity and surface probability, was applied to the GBP-H1a fusion protein-layered surface to monitor specific binding between GBP-H1a fusion protein and anti-H1 Ab by SPR biosensor as shown in Figure 3(B) . To further investigate whether this microfluidic device can be used in immunoassay, we employed GBP-H1a fusion protein, BSA as a blocking agent and Cy3-labeled anti-H1 Ab. Cy3-labeled anti-H1 Ab is a strongly fluorescent molecule, and the fluorescence-based immunoassay is more sensitive compared to the most colorimetric assays in most of the cases [5] . The whole immunosensing process was carried out by using COC microfluidic chips at room temperature. The three detection chambers are included in one microchannel to verify the sensing results which may reduce the error of the signal. In addition, one chip is composed of three different detection zones to test different concentrations of the target Abs. First of all, 100 μg/mL of GBP-H1a fusion protein was injected through the microchannel for the immobilization on the surface of gold surface. After immobilization for 1 h, all the chips were washed with PBS solution and BSA solution (1 mg/mL) was injected to the channels to prevent the non-specific binding then washed with PBS solution. After the blocking and washing process, five different concentrations of Cy3-labeled-Ab (100, 50, 10, 5, and 1 μg/mL, respectively) were injected through the microchannel and left them for 1 h. After incubation for the further reaction, all immunosensors were rinsed with PBS solution three times, and the microchannels were blown off by air. All microfluidic immunosensor chips were examined under same conditions of confocal microscopy (Carl Zeiss LSM510 Meta NLO, Göttingen, Germany) as shown in Figure 4 . At the low concentration of Cy3-labeled Ab applied, small fluorescent signals were observed in the detection chambers. As increasing the concentration of Ab, the whole detection chamber was covered with red fluorescence, and the signal intensities were also increased. In order to compare the signal intensity, the intensity profiles were also recorded because the fluorescence intensity is directly proportional to the amount of Cy3-labeled Ab attached to the surface of detection chamber. The fluorescence intensity changes at the center of chamber were measured, and their fluorescent images with same scale in Y-axis were showed. The fluorescent intensity graphs which correspond to each inserted white line also showed similar signal changing patterns compared with the fluorescent image. From these results, the specific binding of GBP-H1a was successfully immobilized on the gold surface, and the fluorescent images and emission profiles were subsequently increased due to the effective binding of Cy3-labeled anti-H1 Ab, which could be applicable in immunoassay onto the microfluidic chip surface. As a result, the fluorescent intensity depends on the number of immobilized Cy3-labeled-Ab in the detection chamber under the same incubation period. In addition, all detection chambers were analyzed using the line profiles. As increasing the amount of Ab, the intensity of line profiles also increases over the time, and the intensity line is getting increased, which is properly matching with fluorescence images. From the analyzed fluorescence results, the calibration curve, which represents the relationship between the fluorescence intensity and Cy3-labeled-Ab, is shown in Figure 4 (G). The regression equation could be expressed as: where I is the fluorescence intensity and C is the concentration (μg/mL) of Cy3-labeled-Ab with a correlation coefficient of 0.983. In this study, we report a development of a microfluidic device for the detection of human influenza by antigen-antibody interaction based on a highly transparent and inexpensive polymer. The significant fluorescence intensity changes over the different concentrations to the serological antibodies and three different chambers in one microchannel provide more accurate information to detect the H1N1 flu virus. In addition, the immunosensor chips were successfully applied for the detection of the H1N1 without any surface modification of microfluidic chip. The proposed integrated plastic-based microfluidic chip could provide a significant improvement in the miniaturization and a cost-effective way for bio-analysis systems. Therefore, this platform offers perspective for point-of-care testing diagnosis in various infectious disease areas.

Hepatitis C VLPs Delivered to Dendritic Cells by a TLR2 Targeting Lipopeptide Results in Enhanced Antibody and Cell-Mediated Responses

Although many studies provide strong evidence supporting the development of HCV virus-like particle (VLP)-based vaccines, the fact that heterologous viral vectors and/or multiple dosing regimes are required to induce protective immunity indicates that it is necessary to improve their immunogenicity. In this study, we have evaluated the use of an anionic self-adjuvanting lipopeptide containing the TLR2 agonist Pam(2)Cys (E(8)Pam(2)Cys) to enhance the immunogenicity of VLPs containing the HCV structural proteins (core, E1 and E2) of genotype 1a. While co-formulation of this lipopeptide with VLPs only resulted in marginal improvements in dendritic cell (DC) uptake, its ability to concomitantly induce DC maturation at very small doses is a feature not observed using VLPs alone or in the presence of an aluminium hydroxide-based adjuvant (Alum). Dramatically improved VLP and E2-specific antibody responses were observed in VLP+E(8)Pam(2)Cys vaccinated mice where up to 3 doses of non-adjuvanted or traditionally alum-adjuvanted VLPs was required to match the antibody titres obtained with a single dose of VLPs formulated with this lipopeptide. This result also correlated with significantly higher numbers of specific antibody secreting cells that was detected in the spleens of VLP+E(8)Pam(2)Cys vaccinated mice and greater ability of sera from these mice to neutralise the binding and uptake of VLPs by Huh7 cells. Moreover, vaccination of HLA-A2 transgenic mice with this formulation also induced better VLP-specific IFN-γ-mediated responses compared to non-adjuvanted VLPs but comparable levels to that achieved when coadministered with complete freund’s adjuvant. These results suggest overall that the immunogenicity of HCV VLPs can be significantly improved by the addition of this novel adjuvant by targeting their delivery to DCs and could therefore constitute a viable vaccine strategy for the treatment of HCV.

Hepatitis C virus (HCV) infection affects an estimated 200 million individuals worldwide and contributes to significant morbidity and mortality rates associated with liver cirrhosis and hepatocellular carcinoma. Approximately 80% of infected individuals do not clear the virus following acute infection and will develop chronic infection that can lead to end-stage liver disease and complications. Although treatment options using a combination of pegylated interferon-a and ribavirin are available, sustained clearance of the virus is only achieved in approximately 40% of individuals infected with HCV genotype 1 and 60-70% of those who are infected with genotypes 2 or 3 [1] . Recent advances in the treatment of HCV using directly acting antiviral agents (DAAs) such as boceprevir and telaprevir have improved SVR rates in both treatment naïve and experienced patients (reviewed in [2] ). However, treatment can be prolonged, expensive and also associated with substantial side effects. The development of an effective vaccine that can significantly reduce the number of new infections and improve sustained virological response rates could therefore be a useful adjunct to current therapeutic approaches and reduce the impact of infection on global health care systems. Whilst the immune correlates mediating the clearance of virus are still not entirely clear or defined, there is substantial evidence demonstrating that the development of a broad multifunctional T cell response against an array of key viral proteins such as core, E1, NS3, NS4 and NS5 during acute HCV infection is associated with disease resolution [3, 4] and may also provide a level of protection against reinfection [5] . It is also becoming increasingly apparent that such responses alone are not enough [6] and that neutralising antibodies also play an integral role in conferring protection [7, 8] and facilitating viral clearance by mechanisms including antibodydependent cellular cytotoxic mechanisms [9] . An effective HCV vaccine will need to induce antibody and cell-mediated responses and also provide cross protection against different viral genotypes and quasispecies. Neutralising antibodies induced against conserved, conformational epitopes in the viral envelope E1 and E2 glycoproteins [10] [11] [12] , notably antigenic region 3 (AR3) [13] of E2, including the critical neutralisation contact residues contained within domain I of E2 [14] and amino acids 313-327 of E1 [15] , can be broadly cross-neutralising. The fact that these antibodies neutralise different HCV genotypes highlights the importance of including epitopes from both envelope proteins for a vaccine strategy to be effective. Virus-like particles (VLPs) possess features which make them ideal vehicles for the delivery of viral antigens to the immune system; (i) antibody epitopes are presented in the native conformation for induction of potentially neutralising antibodies (ii) multiple T cell, CD4 + and CD8 + , epitopes are packaged in VLPs (iii) VLPs lack regulatory proteins as well as genetic material that could pose a risk of reversion or mutation (iv) encouraging results have been obtained using insect cell-derived recombinant VLPs expressing HCV antigens which induce virus-specific humoral and cellular responses [16] [17] [18] (v) HCV VLPs appear to possess properties favourable for dendritic cell uptake [19] and (vi) they exhibit superior immunogenicity and antigenicity over recombinant protein and DNA-based vaccine approaches [16, 17] . An important consideration in the manufacture of HCV-based VLPs is the cell-type used for their manufacture. For example, it has been shown that vaccination with recombinant HCV envelope proteins expressed in mammalian cells, but not in yeast or insect cells, protect chimpanzees from primary infection by an homologous HCV isolate [20] . Similarly, Rosa et al have demonstrated that mammalian cell-derived recombinant envelope proteins bind to human cells with higher affinity than those produced in yeast or insect cells and appear to be antigenically and functionally similar to the viral proteins produced in an infected host cell [21] . More recently, vaccination of macaques using VLPs in a prime-boost regime has been reported to induce broadly neutralising antibody responses against different HCV genotypes [22] . Although all of these studies provide encouraging results supporting the development of HCV VLP-based vaccines, the fact that heterologous viral vectors and/or unrealistic dosing regimes are required to induce protective immunity indicates that it is necessary to improve their immunogenicity. In this study we evaluate the immunogenicity of mammalian cell-derived VLPs containing structural proteins (core, E1 and E2) of HCV genotype 1a when delivered directly to dendritic cells (DCs) using a Toll-like receptor 2 (TLR2) targeting lipopeptide. This lipopeptide contains the TLR2 agonist dipalmitoyl-Sglyceryl-cysteine (Pam 2 Cys) and associates electrostatically with protein antigens significantly improving their ability to induce both humoral and cell-mediated responses [23] . The ability of lipopeptide-VLP complexes to facilitate DC uptake, induce antibody capable of inhibiting VLP entry into target cells and to elicit cell-mediated antigen-specific responses were each determined. HCV virus-like particles were constructed using a recombinant adenovirus containing encoding the HCV structural proteins (core, E1 and E2) of HCV 77H, genotype 1a. Briefly, the core/E1 genes were amplified from pBRTM_HCV 1-3011 plasmid containing the genome of HCV H77 genotype 1a (a gift from Prof C Rice). The core/E1 genes were amplified from pBRTM_HCV 1-3011 plasmid containing the genome of HCV H77 genotype 1a (a gift from Prof C Rice). The forward Core/E1 primer (59gCCTCTAgAgCCACCATgCATCACCATCACCAT-CACACAAgCACgAATCCTAAACTCAAAgAAAAACC 39) was designed to introduce an XbaI enzyme restriction site followed by a Kozak sequence, a start codon and a His(6) tag at amino terminal end of the core protein. The reverse primer of core/E1 (59 ggCTTAAgCCCggTgACgTgggTTTCC gCgTCgAC 39) was designed to amplify from the sequence downstream of the region corresponding to E1/E2 cleavage site (amino acids 383/384) and to introduce an Afl II restriction site at the 39 end. Next, the E2 genome was amplified using a forward primer (59CgACTAgT-gAAACCCACgTCACCgggggAAgTgCCggCCgC 39) that also introduced a SpeI site at the 59 end. The reverse E2 primer (59CgGATATCTCATCAC gCCTCCgCTTgggATATgAgTAA-CATCATCC 39) was designed to introduce a double stop codon and an EcoRV restriction site at the 39 end of the amplicon. The core/E1 and E2 amplicons were cloned into pGEMEasy (Promega) and subsequently subcloned and ligated to produce a construct which was verified by DNA sequencing. This construct was subsequently subcloned into pAdTrack-CMV (provided by B Vogelstein, Howard Hughes Medical Centre, Baltimore), digested with PmeI and transformed into AdEasier-1 cells by electroporation (Bio-Rad Gene Pulser) as previously described [24] . High titres of recombinant adenovirions encoding the HCV proteins (rAdHCV-CE1E2) were produced in 293T cells by serial passaging and the equivalent multiplicity of infection (MOI) was determined as described previously [24] . To produce HCV VLPs, Huh7 cells were infected with rAdHCV-CE1E2 at a MOI of 1. At 72 hours post-infection, cells were collected and disrupted using a dounce homogeniser and centrifuged at 17,000 g for 5 min. The supernatant was further centrifuged through a 30% sucrose cushion (containing 20 mM Tris pH 7.4 and 150 mM NaCl) at 178,000 g for 4 hours at 4uC. The resulting pellet was resuspended in 50 mM Tris pH 7.and 100 mM NaCl and purified through a 33% caesium chloride gradient by ultracentrifugation at 14uC at 143,000 g) for 72 hours. Twelve 1 ml gradients were recovered and dialysed against sterile PBS at 4uC overnight. Fluorescent labelling of VLPs was achieved by adding VLPs (6 mg/ml) to 2 mg/ml of fluorescein isothiocyanate (FITC; Sigma Aldrich) in 50 ml of DMSO. The suspension was vortexed vigorously and incubated overnight at 4uC before dialysis against PBS the next day. All VLP preparations were stored in aliquots at 270uC until use. Huh7 and 293T cells were grown in Dulbecco's modified Eagle's medium (DMEM; Invitrogen USA) supplemented with 10% fetal calf serum (FCS) and streptomycin 50 mg/ml at 37 C in 5% CO 2 . The syntheses of the branched anionic peptide construct containing eight N-terminal glutamic acid residues (E 8 ) using traditional Fmoc chemistry has been described previously [23] . Briefly, synthesis was carried out manually using PEG-S RAM solid support (Rapp Polymere, Tübingen, Germany; substitution factor 0.27 mmol/g). Fmoc-lysine(Mtt)-OH (Novabiochem, Läufelfingen, Switzerland) was first coupled to the support and the Fmoc protecting group present on the a-amino group then removed and Fmoc-lysine(Fmoc)-OH was then coupled to the exposed N-terminal amino group. Subsequent de-protection and acylation of another two rounds of Fmoc-lysine(Fmoc)-OH yielded eight branch points to which glutamic acid residues were coupled. The primary amino groups of the glutamic acid residues were then acetylated using a 20-fold excess of acetic anhydride and a 40-fold excess of diisopropylethylamine (DIPEA; Sigma, Australia) to generate E 8 which has an overall charge of 28. Lipidation of E 8 was then carried out by removing the Mtt protective group present on the e-amino group of the C-terminal lysine followed by acylation of the exposed e-amino group with two serially added serine residues. The Pam 2 Cys lipid moiety was then coupled according to Zeng et al [25] to generate E 8 (Pam 2 Cys) ( Figure 1A ). Following assembly, lipopeptides were cleaved from the solid phase support and all side-chain protecting groups removed with 88% TFA, 5% phenol, 2% TIPS, 5% water for 3 hours at room temperature. Lipopeptides were analysed by reversed phase highpressure liquid chromatography (RP-HPLC) using a Vydac C4 column (4.66300 mm) installed in a Waters HPLC system. The chromatogram was developed at a flow rate of 1 ml/min using 0.1% TFA in H 2 O and 0.1% TFA in acetonitrile as the limit solvent. Lipopeptides were purified if necessary. All products presented as a single major peak on analytical RP-HPLC and had the expected mass when analysed using an Agilent series 1100 ion trap mass spectrometer. A line of murine BALB/c-derived DCs (D1 cells) was prepared and propagated according to the method described by Chua et al [26] . After a minimum of 21 days in culture, cells were stained for Class II MHC using FITC conjugated anti-IA/IE antibody (Clone M5/114.15.2; Becton Dickinson, USA) and PE-conjugated CD11c (Clone 2G9; Becton Dickinson, USA) prior to use. Cells were verified to be CD11c + MHC Class II + by flow cytometry using a FACSCaliber (Becton Dickinson, USA). D1 cells (2610 5 ) were seeded onto a petri dish in 1 ml of fresh D1 media [26] and incubated at 37uC and 5% CO 2 in the presence or absence of 5 mg FITC-labelled VLPs alone or with VLPs mixed with E 8 Pam 2 Cys (0.2 pmole/ml). Cells were harvested 24 hours later and washed with FACs wash (1% FCS/5 mM EDTA in PBS) before fixation in 1% paraformaldehyde in PBS. To examine cellular association of VLPs, cell fluorescence was analysed by flow cytometry (FACSCaliber, Becton Dickinson, USA). For examination of intracellular uptake of VLPs, extracellular fluorescence was quenched by addition of an equal volume of 0.1 M citrate buffer (pH 4.0) containing 250 mg/ml trypan blue (Merck, Damstadt, Germany) prior to analysis [27] . Data were analysed using FlowJo software (Tree Star, San Carlos, CA). To assess the degree of DC maturation resulting from exposure to different adjuvants, cells were exposed to varying concentrations of aluminium hydroxide gel, Alhydrogel (Sigma Aldrich, Missouri, USA), E 8 Pam 2 Cys or lipopolysaccharide (5 mg/ml) as a positive control (LPS; Sigma Aldrich, Milwaukee, USA). In some experiments, cells were incubated with VLPs alone (5 mg/ml) or in the presence of E 8 Pam 2 Cys (0.032 mg/ml). After 16 hours, cells were harvested, washed and analysed for expression of surface Class II MHC antigen. All experimental procedures involving animals were approved by the University of Melbourne's animal ethics committee under the AEC numbers 0707207 and 061061. HLA-A2k b transgenic mice (HHD mice) were obtained from the Queensland Institute for Medical Research and bred in the Animal House Facility of the Department of Microbiology and Immunology at The University of Melbourne under specific pathogen free conditions. These mice do not express H-2D b but instead express the chimeric monochain of the a1 & a2 domains of HLA-A2.1 and the a3 cytoplasmic and transmembrane domains of H-2D b linked at its N-terminus to the C terminus of human b2 microglobulin [28] . Female HHD mice (3 per group) were inoculated subcutaneously on each side of the base of tail (50 ml per dose) with VLPs (30 mg) either alone, emulsified in an equal volume of complete Freund's adjuvant (CFA) or with an equal amount of E 8 Pam 2 Cys on days 0 and 14. Spleens were removed 28 days after the second dose and splenocytes restimulated in vitro at a concentration of 3610 6 cells/ml in RF-10 medium consisting of RPMI 1640 medium (Gibco, USA) supplemented with 10% fetal calf serum (CSL, Parkville, Australia) 7.5 mM HEPES, 2 mM L-glutamine, 76 mM 2-mercaptoethanol, 150 U/ml penicillin, 150 mg/ml streptomycin, 150 mM non-essential amino acids and 10 U/ml of recombinant IL-2 (Roche, Indianapolis, USA) at 37uC in an atmosphere of 5% CO 2 . Restimulation was carried out in the presence of 10 mg of VLPs or 10 mM of an irrelevant HCVderived HLA-A2-restricted epitope (NS5B2594-2602) that is not contained in the VLP construct. Cells were harvested 5 days later, washed and serial dilutions commencing at 5610 5 cells/ml performed in polyvinylidene fluoride (PVDF) membrane-lined 96-well plates (Millipore, Ireland) previously coated with 5 mg/ml anti-IFN-c capture antibody (clone R4-6A2-BD Pharmingen, San Diego, USA). Cells were then cultured in the presence of 3.75610 5 irradiated autologous VLP-pulsed (10 mg) splenocytes for 40 hours at 37uC and 5% CO 2 . After washing with PBST (PBS containing 0.05% Tween 20), biotinylated anti-IFN-c detection antibody (clone XMG1.2; Becton Dickinson, USA) was added and incubated for 2 hours at room temperature in a humidified atmosphere. Plates were then washed and streptavidin-alkaline phosphatase (Becton Dickinson, USA) added and incubated for a further 2 hours. Spots representative of IFN-c-producing cells were developed by the addition of 100 ml of 1 mg/ml 5-bromo-4-chloro-3-indolyl phosphate in 2-amino-2-methyl-1-propanol buffer (Sigma-Aldrich, USA) for 30 minutes. Individual spots were counted using an AID iSpot EliSpot Reader (GmbH, Strassberg, Germany). Analysis of variance in all experiments and all p values in this study were conducted and obtained using one-way ANOVA nonparametric statistical analysis and Tukey's post-hoc range tests performed with Prism 5 (GraphPad Software, La Jolla, California USA). Flat bottom 96-well polyvinyl plates were coated with either purified HCV VLPs (10 mg/ml) or recombinant E2 protein (5 mg/ ml) in PBSN 3 overnight at 4uC. Prior to coating plates with HCV VLPs, wells were pre-incubated with Galanthus nivalis lectin (2 mg/ml; Sigma Aldrich Australia) in carbonate buffer (15 mM NaCO 3 , 35 mM NaHCO 3 , 0.21 mM NaCl) for 30 minutes at room temperature Following removal of antigen, 100 ml of BSA (10 mg/ml) in PBS was added and plates incubated for 1 hour at room temperature before washing four times with PBST (PBS containing v/v 0.05% Tween-20 [Sigma Aldrich, Milwaukee, USA]). Serial dilutions of sera obtained from immunised mice were added to wells and incubated in a humidified atmosphere overnight. After washing, bound antibody was detected using horseradish peroxidase-conjugated rabbit antimouse IgG antibodies (Dako, Glostrup, Denmark) in conjunction with enzyme substrate (0.2 mM 2,2_-azino-bis 3-ethylbenzthiazoline-sulfonic acid in 50 mM citric acid containing 0.004% hydrogen peroxide). The reaction was stopped by addition of 50 ml of 0.05 M NaF. Titers of antibody are expressed as the reciprocal of the highest dilution of serum required to achieve an optical density of 0.2. For the detection of specific antibody secreting cells by ELISPOT, PVDF membrane-lined 96-well plates (Mabtech, Nacka Strand, Sweden) were coated with 100 ml of PBS containing VLPs (10 mg/ml), recombinant E2 (10 mg/ml) or anti IgG antibody (10 mg/ml) overnight at 4uC. Plates were washed 5 times with PBS and blocked for 2 hours using RPMI 1640 medium (Gibco, USA) supplemented with 20% BSA (Sigma, Australia). Wells were emptied before 5610 5 splenocytes in 200 ml of RF-10 medium was added and incubated for 36 hours at 37uC in an atmosphere of 5% CO 2 . Spot forming units representative of specific antibody-producing cells were developed as previously described for the detection of IFN-c-secreting cells except that biotinylated anti-IgG antibody and streptavidin-conjugated horseradish peroxidase (both from Mabtech, Nacka Strand, Sweden) were used as detecting reagents. In order to determine any inhibition of cell entry by VLPs using antibodies present in sera of vaccinated animals, Huh7 cells were Enhancement of HCV-VLP Immunogenicity PLOS ONE | www.plosone.org first incubated with PBS (10% FCS) for 15 min at 4uC to reduce non-specific binding of antibodies subsequently added. Cells were washed twice, resuspended in PBS (0.1% FCS) and incubated with FITC-labelled VLPs at 4uC for 1 hr. Serial dilutions of sera from vaccinated or non-vaccinated mice were then added and incubated for a further 1 hour at 37uC. For each reaction, 5610 5 Huh7 cells and 200 ng of FITC-labelled VLPs were used in a total volume of 500 ml. At the end of this incubation period, cells were washed with PBS (0.1% FCS) and then fixed in BD Cytofix (Becton Dickinson, USA). Inhibition of VLP entry was determined by flow cytometry and analysed using WEASEL 2.0 software (Walter and Eliza Hall Institute, Melbourne, Australia). Sera from vaccinated mice that demonstrated a decrease in specific cellular binding of 50% or more compared to sera from naïve mice were considered to contain neutralising antibodies [18] . HCV genotype 1a VLPs were produced by transducing a human hepatocyte-derived cell line with recombinant adenovirus containing encoding the HCV structural proteins (core, E1 and E2) of genotype 1a to produce particles that harbour antigenic resemblance to virions produced in an infected host cell. Because of the essential role that DCs play in the induction of both humoral and cell-mediated responses, we first examined the ability of a spleen-derived DC line (D1 cells) to take up fluorescein isothiocyanate-labelled HCV VLPs (FITC-VLPs). Flow cytometric analysis revealed that DCs incubated with FITC-VLPs exhibited higher whole cell fluorescence intensities compared to untreated DCs indicating the presence of cell-associated VLPs ( Figure 1B) . Exposure of DCs to VLPs pre-mixed with the lipopeptide E 8 Pam 2 Cys also resulted in higher levels of cell fluorescence compared to untreated DCs. The percentage of fluorescenated cells in these cultures was similar to cultures that contained FITC-VLPs alone ( Figure 1C) . To determine the magnitude of VLP cell uptake, intracellular fluorescence was measured by quenching extracellular fluorescence after exposure of cells to trypan blue prior to flow cytometric analysis [27] . Although the resulting fluorescence intensities of DCs incubated with FITC-VLPs were now lower following this treatment, the levels were still notably higher than untreated cells confirming the presence of intracellular FITC-VLPs ( Figure 1D ). Equivalent fluorescence cell intensities were also observed in DCs that were incubated with FITC-VLPs pre-mixed with E 8 Pam 2 Cys. However, a higher percentage of fluorescenated cells were detected in those cultures compared to those exposed to FITC-VLPs alone ( Figure 1E ) indicating that an increase in uptake of these constructs is facilitated using the lipopeptide. To investigate the ability of HCV VLPs and E 8 Pam 2 Cys to cause activation of DCs, we measured the expression of surface MHC class II molecules following incubation with the various antigens. The results (Figure 2A) indicate that untreated DCs contained two populations of cells which were MHC class II low and MHC class II high , the latter comprising ,24% of the population analysed. While the distribution of these populations was not affected by exposure to HCV VLPs alone, incubation with HCV VLPs mixed with E 8 Pam 2 Cys caused a dramatic shift in the distribution of MHC class II expressing cells such that 83% of cells were MHC class II high . The upregulation of MHC class II expression on these cells were comparable to those cultured in the presence of LPS which is a potent DC maturation stimulus. Further dosing experiments showed that increasing the concentrations of HCV VLPs to 20 mg/ml, did not induce DC activation because the percentage of MHC Class II high DCs in cultures containing escalating doses of VLPs remained similar to those containing untreated DCs ( Figure 2B ). In contrast, exposure to as little as 0.1 nmoles of E 8 Pam 2 Cys was sufficient at inducing a greater than two-fold increase in activation of DC compared to untreated cells ( Figure 2C) and was similar to the levels of activation observed with LPS. No DC activation was observed in the presence of Alhydrogel ( Figure 2D ). The presence of E 8 Pam 2 Cys in HCV VLP-containing formulations however, not only promotes uptake of HCV VLPs by DCs but also considerably increases the level of DC activation. To determine if the DC activating properties of VLP formulations containing E 8 Pam 2 Cys translate to an improvement in HCV VLP immunogenicity, BALB/c mice were inoculated with VLPs alone or with VLPs mixed with E 8 Pam 2 Cys. HCV VLP-specific antibody titres in sera obtained after one, two or three doses of each formulation were then determined by ELISA. Administration of VLPs alone in saline was able to elicit detectable titres of specific antibody that were marginally increased after each dose of antigen ( Figure 3A ). In animals that received HCV VLPs mixed with E 8 Pam 2 Cys, however, antibody levels were significantly higher, in some cases by up to ten-fold more than those from animals that received the same dose of HCV VLPs alone. In fact the titre of specific antibody induced following administration of 3 doses of HCV VLPs alone was achieved using a single dose only of HCV VLP mixed with E 8 Pam 2 Cys. When compared to animals that were inoculated with HCV VLPs formulated with Alhydrogel, an adjuvant widely used to induce antibody responses to both human and veterinary vaccines [29] , lower antibody titres were observed in these animals than in those that received the VLP-lipopeptide formulation. In examining levels of E2 specific antibodies elicited by vaccination, higher titres were once again demonstrated in animals that received 3 doses of VLPs mixed with E 8 Pam 2 Cys compared to those that were inoculated with VLPs alone or with Alhydrogel ( Figure 3B ). The hierarchical pattern of antibody responses induced by E 8 Pam 2 Cys and Alhydrogel was also confirmed by the numbers of specific antibody secreting cells that were detected in the spleens of vaccinated mice. Once again significantly higher numbers of cells secreting both HCV VLP ( Figure 4A ) or E2-specific antibodies ( Figure 4B ) were detected in animals that received HCV VLPs mixed with E 8 Pam 2 Cys than those that were inoculated with HCV VLPs alone or VLPs formulated with Alhydrogel. To assess the neutralising activity of antibodies induced by vaccination, we first set out to investigate if VLP entry into human hepatocyte cell line Huh7 could be inhibited. Pre-incubation of FITC-labelled VLPs (VLP-FITC) with PBS or naïve serum resulted in minimal inhibition of VLP entry ( Figure 5A ). However, the presence of an antibody against CD81, a cell surface molecule implicated in HCV entry into hepatocytes [30] , was able to prevent VLP entry into these cells by .90% confirming that these VLPs also utilise this molecule to facilitate cell entry. We next analysed the ability of sera obtained from mice inoculated with VLPs to inhibit the binding and entry of VLPs into Huh 7 cells ( Figure 5B ). Neutralisation of binding of VLPs to Huh7 cells was significantly greater in sera obtained from mice inoculated with VLPs in E 8 Pam 2 Cys (,50%) compared to sera obtained from mice inoculated with VLPs administered in Alhydrogel (,30%) or in saline (,20%). The ability of VLP formulations containing E 8 Pam 2 Cys to induce a cell-meditated immune response was examined by inoculating transgenic mice expressing the MHC class I (HLA-A2) allele but not endogenous H-2D b molecules [28] . Control transgenic animals were inoculated with VLPs alone or VLPs emulsified with an equal amount of complete Freund's adjuvant (CFA). Splenocytes from vaccinated animals were obtained 28 days post-inoculation and re-stimulated with antigen in vitro. The results ( Figure 6 ) of an ELISPOT assay carried out revealed significantly higher numbers of HCV VLP-specific IFN-c producing cells in the spleens of mice inoculated with VLPs in the presence of E 8 Pam 2 Cys or VLPs emulsified in CFA than in those that received VLPs alone. The development of novel, effective anti-viral vaccine strategies in recent times has seen a notable shift away from the use of traditional formulations which utilize whole inactivated or live attenuated viruses towards approaches based on recombinant subunit protein antigens which are more easily characterised and defined. VLPs offer features that make them a useful platform for delivering viral antigens in a single vaccine construct which not only minimises the risks that may be associated with preparations containing or requiring the use of a replicating pathogen but will also closely resemble native viral antigens from which they are derived. The most convincing demonstration of VLPs efficacy is the quadrivalent VLP-based vaccine Gardasil which prevents persistent infection and associated disease caused by human papillomavirus [31] . Other studies of VLP-based vaccination strategies have also shown promising results and led to Phase I testing against a number of disease indications including seasonal and pandemic influenza, Hepatitis B, malaria and HIV (reviewed in [32] ). Depending on the type of virus used to manufacture a VLP construct, studies have shown that protective responses induced by VLPs can be elicited without co-administration of adjuvant [33] [34] [35] . In many cases, however, the induction of useful immune responses may require multiple doses [36, 37] , a regime that may be impractical to implement in the field or involve a viral vector to provide an initial priming dose followed by a boost using VLPs [22, 38] . Of relevance to the present study, the use of adjuvants to enhance VLP immunogenicity has been shown to induce strong antibody responses using dose-sparing amounts of HIV [39] or Norwalk virus-derived VLPs [40] and also elicits cell-mediated were incubated with VLPs (5 mg) alone or formulated with E 8 Pam 2 Cys (0.01 nmoles/ml) or Alhydrogel (5 mg) in a total volume of 1 ml. For comparative purposes within all experiments, cells were also either left untreated, exposed to LPS (5 mg/ml) or to similar amounts of each adjuvant alone. Cell surface MHC class II expression was determined after 16 hours using a PEconjugated anti-IA/IE antibody. Cells expressing low levels of MHC Class II molecules were deemed to be immature whilst those expressing high levels were considered to be mature. Shown are representative histograms depicting cell surface MHC class II expression from one of three experiments conducted separately. MHC Class II high expressing cells are shaded in grey. For dosing experiments, cells were also incubated with increasing amounts of (B) VLPs, (C) E 8 Pam 2 Cys, (D) Alhydrogel or VLPs (5 mg) formulated with increasing amounts of (E) E 8 Pam 2 Cys, or (F) Alhydrogel. doi:10.1371/journal.pone.0047492.g002 responses that culminate in improved protection against lethal influenza viral [41, 42] and tumorigenic challenge [43] . Our previous studies have shown peptide epitope and proteinbased antigens can be made far more immunogenic when covalently attached to Pam 2 Cys in order to target their delivery via TLR2 to dendritic cells (DCs) [44] . This results in the induction of robust antibody and CD8 + T cell-mediated immune responses and has been shown for multiple indications [44] [45] [46] [47] . Each of these vaccine candidates demonstrated the ability of this simple lipid structure to dramatically enhance the immunogenicity of antigens that are otherwise immunologically inert. Nevertheless, the approach introduces complexities into the vaccine manufacturing process due to the requirement for covalent attachment of Pam 2 Cys to an antigen. The use of the anionic lipopeptide E 8 Pam 2 Cys overcomes many of the technical complexities related to this process, especially the use of covalent chemistries, by making use of electrostatic association with antigen [23] . The ability of Pam2Cys to dramatically enhance the immunogenicity of HCV VLPs was demonstrated in the improved overall antibody responses that we observed. Not only are greater antibody titres induced following vaccination with VLPs formulated with E 8 Pam 2 Cys compared to the use of VLPs alone or when co-administered with Alhydrogel but up to 3 doses of nonadjuvanted or traditionally adjuvanted antigen were required to match the titres obtained with a single dose using lipopeptide. Most importantly in the context of HCV the trend translates to improved E2-specific antibody responses and the use of lower doses of VLPs to achieve this while maintaining efficacy has major advantages by providing cost benefits to vaccine manufacturers. Our studies examining the interaction of VLPs with DCs indicate that while improvements in VLP uptake mediated by this lipopeptide is minimal, its ability to concomitantly induce the maturation of DCs at very small doses is a feature not observed using VLPs alone or VLPs administered in the presence of alum. Our previous work also demonstrated that association of antigen with charged lipopeptide facilitates trafficking of antigen to lymph Cys. Supernatants were clarified by centrifugation, incubated with Huh7 cells (5610 5 ) in a total volume of (500 ml) for 1 hour. Cells were then harvested and cellular fluorescence levels analysed by flow cytometry. All bar graphs represent the percentage reduction in VLP entry relative to baseline levels obtained using serum from naïve mice. doi:10.1371/journal.pone.0047492.g005 Figure 6 . Cell-mediated responses elicited by vaccination. HLA-A2k b transgenic mice (n = 3 per group) were inoculated (100 ml) subcutaneously at the base of the tail on days 0 and 14 with 30 mg of VLPs alone, emulsified with an equal amount of complete freund's adjuvant (CFA) or pre-mixed with 30 mg of E 8 Pam 2 Cys in saline. Splenocytes were obtained 28 days later and restimulated for 7 days in the presence of 10 mg VLPs or an irrelevant HCV-derived HLA-A2restricted epitope not part of the VLP construct. The frequency of peptide-specific T cells producing IFN-c was determined in an ELISPOT assay. Each bar represents the average number of IFN-c producing T cells and standard deviation in each group after subtracting nonspecific responses from corresponding samples stimulated with the irrelevant peptide. doi:10.1371/journal.pone.0047492.g006 Enhancement of HCV-VLP Immunogenicity PLOS ONE | www.plosone.org nodes draining from the vaccination site [23] and together with the results presented in this study provide an explanation for the dose-sparing neutralising antibody responses that we observe. Coadministration of VLPs using charged lipopeptide has the added benefit of eliciting VLP-specific cell-mediated responses in HLA-A2 transgenic mice, a fact that may also be attributed to DC targeting and activation. Given the results of the work described in this study, we conclude that the use of this branched anionic lipopeptide together with VLPs containing HCV antigens in order to provide a broad spectrum of conformational epitopes can provide benefits in terms of inducing improved levels of neutralising antibody titres and eliciting cell-mediated responses. This strategy could therefore constitute a valuable addition to the armamentarium of current VLP-based vaccine developments against HCV.

Viral metagenomic analysis of bushpigs (Potamochoerus larvatus) in Uganda identifies novel variants of Porcine parvovirus 4 and Torque teno sus virus 1 and 2

BACKGROUND: As a result of rapidly growing human populations, intensification of livestock production and increasing exploitation of wildlife habitats for animal agriculture, the interface between wildlife, livestock and humans is expanding, with potential impacts on both domestic animal and human health. Wild animals serve as reservoirs for many viruses, which may occasionally result in novel infections of domestic animals and/or the human population. Given this background, we used metagenomics to investigate the presence of viral pathogens in sera collected from bushpigs (Potamochoerus larvatus), a nocturnal species of wild Suid known to move between national parks and farmland, in Uganda. RESULTS: Application of 454 pyrosequencing demonstrated the presence of Torque teno sus virus (TTSuV), porcine parvovirus 4 (PPV4), porcine endogenous retrovirus (PERV), a GB Hepatitis C–like virus, and a Sclerotinia hypovirulence-associated-like virus in sera from the bushpigs. PCR assays for each specific virus combined with Sanger sequencing revealed two TTSuV-1 variants, one TTSuV-2 variant as well as PPV4 in the serum samples and thereby confirming the findings from the 454 sequencing. CONCLUSIONS: Using a viral metagenomic approach we have made an initial analysis of viruses present in bushpig sera and demonstrated for the first time the presence of PPV4 in a wild African Suid. In addition we identified novel variants of TTSuV-1 and 2 in bushpigs.

Conclusions: Using a viral metagenomic approach we have made an initial analysis of viruses present in bushpig sera and demonstrated for the first time the presence of PPV4 in a wild African Suid. In addition we identified novel variants of TTSuV-1 and 2 in bushpigs. Wild animals are carriers of a number of pathogens that have the potential to infect the human population and/ or domestic animals. The intensity of contact between wildlife, livestock and human population is increasing due to a number of factors, primarily human and livestock population growth leading to encroachment onto wildlife habitat [1, 2] . During recent decades a number of pathogen crossovers from wildlife to humans and livestock have occurred resulting in emerging diseases, such as SARS, Hantavirus Pulmonary syndrome 1, Nipah virus disease 1, and Hendra virus-induced diseases among others. However, it is also evident that transmission can occur both ways i.e. pathogens may spill-over to wildlife from humans and/or from livestock. One example of this is the spill-over of Canine distemper virus from domestic dogs (Canis familiaris) to African wild dogs (Lycaon pictus) in Serengeti in 1991 leading to local extinction of wild dogs in the area [2] . Apart from being carriers of novel viruses with the potential to cause disease in naïve domestic hosts, wildlife may also act as reservoirs for known viral pathogensfor example there are a number of wildlife reservoirs for foot and mouth disease virus such as African buffalo (Syncerus caffer), reindeer (Rangifer tarandus) and wild boar (Sus scrofa) [3] . However, information on the viral flora in wildlife is typically scarce or non-existent. Traditional viral detection methods, such as virus isolation, are often hindered by the inability to grow virus in cell culture. The divergence of many viruses and absence of a common viral marker gene makes detection difficult using standard molecular techniques including PCR and microarray as they are frequently targetspecific through the use of specific primers, probes and/ or antibodies. Viral metagenomics is a sequence, and culture-independent approach that has proven to be a valuable tool for the investigation not only of diseases of unknown etiology but also of the complete viral flora of different reservoirs and vectors. By providing insights into a wide range of unknown potential pathogens and revealing novel aspects of biodiversity, metagenomics is able to detect and characterise novel pathogens [4] [5] [6] . In many rural parts of the developing world, domestic livestock are kept in free-range systems, potentially allowing contact with wild animals. In some parts of Uganda, free-range scavenging by pigs is frequent. At the same time wild species of suidae such as bushpigs (Potamochoerus larvatus), with a wide distribution in Eastern and Southern Africa, live and move at the interface between the national parks and the farmland where there is an opportunity for interaction and sharing of pathogens with domestic relatives. Bushpigs are considered to be possible natural reservoirs for African swine fever virus [7] , but less is known about what other viruses bushpigs might carry. Therefore, to investigate whether bushpigs are carriers of known and or unknown porcine viruses we have in this study investigated the viral flora of bushpig sera. A total of 171,466 reads were obtained from the 454sequencing run, and after assembly 4,441 contigs were created while 32,863 reads remained unmatched (singletons). Although blastn and blastx analysis showed that a majority of the sequences were non-viral, 35 contigs and singletons sequences were identified as viral sequences ( Table 1 ). The high percentage of host genetic sequences found despite of the nuclease treatment prior to sequencing demonstrates the difficulties of completely reducing the background of the host as discussed in the review by Blomström A-L 2011 [4] . Most viral hits showed closest similarity to known pig viruses such as, Torque teno virus (TTSuV) and porcine parvovirus 4 (PPV4). By design of primers based on the sequences obtained from the 454-sequencing run, PCR assays were established to verify the presence of these viruses. These PCR assays confirm the presence of TTSuV1, TTSuV2, PPV4, and the porcine endogenous retrovirus (PERV). The GB Hepatitis C-like, and the Sclerotinia hypovirulence associated like virus, on the other hand, were not detected by the PCR approach, possibly due to the low concentration of these viruses in the samples as only one sequencing read was found for each among the total 171,466 reads obtained from the 454-sequencing run. Also it is possible that in some cases sequencing errors led to primer mismatches. Parvoviruses are small single-stranded linear DNA viruses with a genome of approximately 5000 nucleotides, which have been found in a number of species such as human, swine, cattle and gorilla [8] . In swine, porcine parvovirus (PPV) is a known agent that causes reproductive failure [9] . However, in recent years a number of new parvovirusesporcine hokovirus (PHoV) [10] , porcine bocavirus (PBoV) [11] and porcine parvovirus 4 (PPV4) [12] -have been discovered in pigs, with their potential involvement in disease currently unknown. The parvovirus sequences discovered in the investigated bushpigs showed closest similarity to PPV4. Porcine parvovirus 4 was originally discovered in 2010 in samples collected from pigs in North Carolina in 2005 after an outbreak of acute-onset of disease with high mortality [12] . Subsequently, PPV4 was reported in pigs in China where 1.84% of the investigated pigs were PPV4 positive [13] . The genome of parvoviruses consists of two major open reading frames (ORF) encoding the non-structural and the capsid proteins. However, the genomes of PPV4 as well as of PBoV contain an additional third ORF [12] . The parvovirus sequences obtained from this study could be found in the capsid and in the non-structural ORF. All reads classified as parvovirus gave significant similarity to PPV4 through Blastx searches. A PCR with primers designed to amplify PPV4-like sequences from the original extracted genetic material showed the presence of this virus in one of the three bushpig sera. Sequencing of the PCR product (GenBank accession number: JQ277337) confirmed correct amplification and sequence analysis showed that at protein level (84 amino acids), the product displayed a 75.9-77.1% sequence similarity to the PPV4 sequences available in GenBank. In addition, the phylogenetic tree generated from these data ( Figure 1 ) grouped the sequence with PPV4 when analysed together with sequences from all the different parvovirus genera (Parvovirus, Erythrovirus, Dependovirus, Amdovirus and Bocavirus) and PPV4. However, the tree also confirms the divergence of the bushpig PPV4 from the other PPV4. PPV4 has previously been reported in domestic pigs only in USA and China [12, 13] . However with this study we show the presence of a PPV4 variant in a wild suid in Uganda. Figure 1 Neighbour-joining phylogenetic tree of Parvovirus. The Neighbour-joining tree shows the phylogenetic relationship between the bushpig PPV4 sequence (84 amino acid long) and 49 sequences available from GenBank. The bushpig PPV4 is indicated with ♦. Torque teno virus was discovered 1997 in a serum sample from a patient with posttransfusion hepatitis of unknown etiology using representational difference analysis [14] . Since then the virus has been detected and characterised in a number of species such as primates, cats, dogs and pigs [15] , but the role of these viruses in disease development is still controversial. These viruses have small (approximately 2.8 kb) circular DNA genomes. Torque teno virus in pigs is divided into two different species, Torque teno sus virus 1 and 2, and prevalence studies have shown that TTSuV is widely spread in pig populations across the world [16, 17] . In a previous study [18] , we have found that 51.6% of a sample population of domestic pigs in Uganda were carrying one or both these TTSuV variants. Although, most studies have targeted domestic pigs, TTSuV have also been found in wild boar in Europe [19] . As shown in Table 1 , our data indicated the presence of both TTSuV-1 and 2 in the investigated serum samples. Both the TTSuV-1 and 2 sequence reads were located in the major open reading frame (ORF1) and all reads showed significant similarity to the respective virus in both blastn and blastx analyses. Two of the TTSuV-1 sequence reads partially overlapped and the sequence analysis indicated a significant variation between the two and therefore two different TTSuV-1 PCR assays were designed. The results from the specific PCR assays showed that one of the TTSuV-1 variants (here named TTSuV-1a) could be found in two of the three bushpig sera while the other one (here named TTSuV-1b) was found in all three. The sequence analysis of the PCR amplified TTSuV-1 products (Gen-Bank accession number JQ277338 -42) showed a sequence similarity between TTSuV-1a and b on protein level at 53-54.5% in the analysed 66 amino acid region. Compared to sequences from other studies, TTSuV-1a showed a 60.6-84.8% protein sequence similarity while TTSuV-1b was more divergent (50-56% similarity). These protein sequence identity values were similar to those seen when comparing the sequences retrieved from the GenBank with each other (59-100% similarity). The phylogenetic tree (Figure 2 ) also indicated that TTSuV-1b was more divergent than the other sequences. Sequences from different parts of the world such as China, Germany, Spain, US etc. was used in the phylogenetic study however no clear geographical clustering was seen. TTSuV-2 was confirmed in one of the bushpig sera and the amplified product GenBank accession number JQ277343) showed a protein sequence identity of 66.6-79.4% to the other TTSuV-2 sequences used in the phylogenetic study using a 313 amino acid region. This sequence identity was in the range of the similarity seen between the different TTSuV-2 sequences from other studies used for comparison (64.2-100%). In the phylogenetic study ( Figure 3 ) the sequenced TTSuV-2 grouped with the other TTSuVs but in its own clade. TTSuV-1 and 2 have previously, as mentioned, been detected both in domestic pigs across the world [16] [17] [18] and wild suidae in Europe [19] and now for the first time in bushpigs on the African continent. Studies on the genetic variability of TTSuV-1 and 2 have shown a higher genetic diversity in the coding regions compared to the untranslated region [20] . The sequence analysis of both the bushpig-derived TTSuV and those from GenBank shows a high genetic variability among the different TTSuV-1 and TTSuV-2 and also shows the co-infection of two different TTSuV-1 variants and one TTSuV-2. Endogenous retroviruses are integrated in the host genome and all vertebrates investigated have been shown to carry retroviral sequences. It is estimated that up to 10% of animal genomes are retroviral elements [21] . Also the bushpig genome has been investigated and confirmed to contain PERV [22] [23] [24] . By running the specific PERV PCR on both DNA and on DNase treated RNA we could as expected see that all the bushpigs had both integrated proviral DNA and expressed PERV RNA, indicating active viral transcription and replication. The sequenced PERV products (GenBank accession number JX566717-719) showed a high similarity (85 -89%) to those available in GenBank. Through investigating sera collected from bushpigs in Uganda by viral metagenomics, we have for the first time showed the presence of PPV4 in a wild Suid on the African continent. The region of PPV4 investigated indicates a sequence divergence relative to previously described PPV4. In addition, novel TTSuV-1 and 2 variants were found. Further sequence analysis and prevalence studies can be used to define the genetic relationships of these viruses and their distribution in both domestic pigs and in wildlife. The sera from three bushpigs collected from Lake Mburo National Park, Uganda in March 2010, as part of a research project on African swine fever epidemiology were used in this study. The animals were captured using game capture nets (50x3m, 150 mm square mesh, 3.5 mm nylon braid khaki, ALNET Ltd, South Africa) with assistance from Uganda Wildlife Authority (UWA) staff, and sedated with zolazepam and tiletamine (Zoletil forte vet 50 mg/ml + 50 mg/ml, Virbac Laboratories, France) before blood sampling from the saphenous vein. The Ministry of Agriculture Animal Industry and Fisheries and Uganda Wildlife Authority, together with Makerere University are mandated to carry out animal disease investigations in livestock and wildlife in the country. This is done by veterinarians who handle the animals under internationally recognized guidelines. Fifty microliters of serum was aliquoted for the RNA and DNA extraction respectively. One hundred and fifty microliter of 1x DNase buffer (Roche, Mannheim, Germany) was added to each aliquot of serum after which the sample was treated with nucleases -100 U DNase (Roche, Mannheim, Germany) and 2 μg RNase (Invitrogen, Carlsbad, CA, USA) for two hours at 37°C in order to degrade the host nucleic acid. Trizol was added to one of the two aliquots and RNA was extracted using a combination of Trizol and Qiagen RNAeasy kit. DNA was extracted using the Qiagen DNAeasy mini extraction kit according to the manufacturer's instructions and eluted in 50 μl elution buffer (EB). The DNA and RNA were amplified by random PCR as described earlier [25] . Before sequencing, the primers were cleaved using EcoRV (NEB, Ipswich, MA, USA) and the cleaved product was purified using the Qiagen PCR purification kit (Qiagen, Hilden, Germany) and eluted in 30 μl EB. The purified product was sequenced on 1/8 th of a pico titre plate using the 454 technology by Roche at Inqaba Biotech (South Africa). The sequences were analysed through quality check and removal of very short sequences before being assembled using CLC genomic workbench v4.6 (http://www.clcbio.com/index.php). Blastn and blastx searches were performed through the Camera 2.0 portal [26, 27] and searches through NCBI (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi). The viral blast hits with an e-value of 10 -4 or lower were further analysed. Confirmation PCRs, sequencing and phylogenetic studies PCR primers were designed based on the reads from the 454-sequencing run ( Table 2 ) and PCR assays were set up with the aim to look for each individual virus in each bushpig sera. For the RNA viruses cDNA synthesis was performed prior to PCR using Superscript III (Invitrogen, Carlsbad, CA, USA) and random primers according to the manufacturer's instructions. The PERV RNA was treated with DNase prior to the cDNA synthesis and both a "+" and a "-" RT cDNA synthesis reaction was performed. The PCR using each specific primer pair (Table 1) was performed according to the following procedure: 1x PCR buffer, 2.5 mM MgCl 2 , 1.0 mM dNTP, 0.4 μM forward primer and reverse primer each, and 1.25 U AmpliTaq Gold DNA polymerase (Applied Biosystems, Foster City, CA, USA). For each reaction, two μl DNA or cDNA from each respective bushpig was used. The amplification was performed with the following reaction conditions: a 12 minute enzyme activation step at 95°C followed by 39 cycles of 95°C for 30 seconds, 58°C for 30 seconds and 72°C for 90 seconds, finishing with one cycle for 10 minutes at 72°C. The PCR products were visualized on a 1.5% agarose gel. The PCR-positive products were purified using the QIAquick PCR purification kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions and eluted in 25 μl EB. The purified products were sequenced with standard Sanger sequencing using Big Dye Termination kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. The obtained chromatograms were edited using SeqMan (Lasergene 9, DNASTAR Inc., Madison, USA). Sequence identity plots were performed using the Bioedit software [28] . ClustalW as well as the phylogenetic analyses were carried out using Mega 5 [29] . The phylogenetic trees were constructed using the Neighbour-joining algorithm with p-distances and with a bootstrap value of 1000.

A Three-Dimensional Comparison of Tick-Borne Flavivirus Infection in Mammalian and Tick Cell Lines

Tick-borne flaviviruses (TBFV) are sustained in nature through cycling between mammalian and tick hosts. In this study, we used African green monkey kidney cells (Vero) and Ixodes scapularis tick cells (ISE6) to compare virus-induced changes in mammalian and arthropod cells. Using confocal microscopy, transmission electron microscopy (TEM), and electron tomography (ET), we examined viral protein distribution and the ultrastructural changes that occur during TBFV infection. Within host cells, flaviviruses cause complex rearrangement of cellular membranes for the purpose of virus replication. Virus infection was accompanied by a marked expansion in endoplasmic reticulum (ER) staining and markers for TBFV replication were localized mainly to the ER in both cell lines. TEM of Vero cells showed membrane-bound vesicles enclosed in a network of dilated, anastomosing ER cisternae. Virions were seen within the ER and were sometimes in paracrystalline arrays. Tubular structures or elongated vesicles were occasionally noted. In acutely and persistently infected ISE6 cells, membrane proliferation and vesicles were also noted; however, the extent of membrane expansion and the abundance of vesicles were lower and no viral particles were observed. Tubular profiles were far more prevalent in persistently infected ISE6 cells than in acutely infected cells. By ET, tubular profiles, in persistently infected tick cells, had a cross-sectional diameter of 60–100 nm, reached up to 800 nm in length, were closed at the ends, and were often arranged in fascicle-like bundles, shrouded with ER membrane. Our experiments provide analysis of viral protein localization within the context of both mammalian and arthropod cell lines as well as both acute and persistent arthropod cell infection. Additionally, we show for the first time 3D flavivirus infection in a vector cell line and the first ET of persistent flavivirus infection.

Vector-borne flaviviruses, such as Dengue (DENV), Yellow Fever, Japanese Encephalitis virus (JEV), and tick-borne encephalitis (TBEV) viruses are recognized as significant human pathogens and cause considerable mortality and morbidity worldwide. TBEV, a tick-borne flavivirus (TBFV), is responsible for 14,000 infections per year [1] and has a fatality rate of up to 40% [2] . Symptoms of TBEV infection can include fever, malaise, meningitis, and encephalitis. TBEV and other TBFV, such as Omsk Hemorrhagic Fever virus, are classified as NIAID Category C pathogens and are treated as biosafety level 4 agents in the United States. One TBFV, Langat virus (LGTV), is naturally attenuated [3, 4] , making it suitable for biosafety level 2 work and ideal for use in laboratory studies as a model for higher pathogenicity viruses. In nature, LGTV and other TBFV maintain a complex cycle between ticks and vertebrate hosts. Historically, Ixodidae ticks (hard ticks) have been considered to be the arthropod vector, but, some findings with Alkhurma virus [5] and Kyasanur Forest virus [6] suggest that the soft-bodied Ornithodoros ticks can transmit TBFV. Thus, the arthropod host-range for TBFV may be greater than assumed. The TBFV present a unique situation because the viruses persistently infect ticks and are maintained by vertical transmission across the developmental instars (larval, nymph, and adult). Horizontal transmission (from tick to vertebrate host) then allows amplification of the frequency of the virus within the tick population, as uninfected ticks feeding upon infected vertebrates can acquire the virus [7, 8] . The primary vertebrate hosts are generally small rodents; however, infection of larger mammals also occurs in endemic areas. Humans are an inadvertent, dead-end host, contracting TBFV via tick bite or less frequently by consumption of milk from infected animals [1] . The impact of TBFV infection on vertebrates can vary considerably; some reports describe persistent infection of vertebrates and vertebrate cell lines [9] [10] [11] while other laboratory studies show severe disease development in infected animals [12] [13] [14] . Like other members of the family Flaviviridae, LGTV is a singlestranded, positive-sense RNA virus. Upon infecting a cell, the virions are thought to traffic to the endosome, where they undergo structural transformation and fuse with host cell membranes, releasing the 11-kb viral RNA genome into the cytoplasm [15] . The genomic RNA, which can function as mRNA, is translated into an approximately 400 kDa polyprotein [16] that is subsequently cleaved into at least ten proteins by both viral and cellular host proteases [17] . The currently defined complement consists of three structural proteins (Capsid [C] , membrane [M] , and envelope [E]) and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) [17, 18] . Precise functions for all of the nonstructural proteins are not fully elucidated, but some roles have been determined. NS1 can be found on the cell surface but is known to be secreted [17] and is also detected in perinuclear areas where it is believed to play a part in viral replication [19] ; however, its role has not yet been clearly defined. NS3 is a RNA helicase and also acts as a viral protease using NS2B as a co-factor [20, 21] . NS4A, a small hydrophobic protein, has been implicated in the cellular changes that accompany virus reproduction [22] and has been shown to localize to foci within the perinuclear region of infected cells [23] . NS5 is the RNA-dependent RNA polymerase and methyltransferase [17] ; however, recent implications of LGTV NS5 as an innate immune response antagonist confirms a broader role for this protein beyond genome replication [24] . This broader role is also seen in NS59s wider distribution, being located in cell nuclei, perinuclear regions, and diffusely throughout the cytoplasm [25] . Following synthesis of the viral proteins, the single stranded genome is converted into a doublestranded RNA (dsRNA), which serves as a transcriptional template as well as the viral replicative form. Together with the replication complex, made up of NS1, NS2A, NS3, NS4A, and NS5, dsRNA is found in cytoplasmic, perinuclear foci throughout LGTV infected cells. This dsRNA species is an additional obligatory replicative intermediate. Dramatic cellular changes accompany flavivirus replication and have been the subject of considerable study, although most of the studies have been done in mammalian cells using the mosquitoborne flaviviruses. Replication has been linked to areas of the cell containing large amounts of virus-induced structures, including convoluted membranes, dilated cisternae, vesicles, tubules, and paracrystalline arrays [19, [26] [27] [28] . These structures are localized in a perinuclear distribution of vastly proliferated membranes which, in the case of cells infected with mosquito-borne flaviviruses, are derived from endoplasmic reticulum (ER) [29] [30] [31] . The vesicles are typically 70-150 nm in diameter [28] with a pore-like connection either between vesicles and the cytoplasm or between individual vesicles [30, 31] . Vector-borne flaviviruses, like DENV, JEV, and TBEV, must replicate in both mammalian and arthropod cells, and a limited number of comparative studies have been published [32] [33] [34] [35] [36] . The findings, generally, show agreement between mammalian and arthropod cells in virus induced structures, such as cytoplasmic membrane proliferation and vesicle formation. Some changes were noted via TEM by Senigl et al. [35] four to seven days post TBEV infection; viral particles were observed inside rough ER in mammalian cells but were visualized within vacuoles or within the cytosol in tick cells. Consequently, we have undertaken a direct comparison of LGTV infection in both mammalian and arthropod cells, and extended the results to include 3D reconstructions. In addition, we have shown for the first time the 3D structure of acute and persistent flavivirus infection in an arthropod cell line. African green monkey kidney cells (Vero, ATCC) were maintained in Dulbecco's minimal essential media (DMEM, Invitrogen) supplemented with 10% fetal calf serum (FCS) and 50 ug/ml gentamicin (complete DMEM) at 37uC in 5% CO 2 . ISE6 cells derived from I. scapularis embryonated eggs (kindly provided by Dr. Timothy Kurtti, University of Minnesota) were cultured at 32uC as previously described [37] with the addition of 10 ug/ml gentamicin to the L-15B media. A virus stock of Langat virus strain TP21 (LGTV) (kindly provided by Dr. Alexander Pletnev, NIH/NIAID) was prepared by infection of Vero cell cultures at a multiplicity of infection (MOI) of 0.005 [38, 39] . For virus titration by immunofocus assay, 1610 5 Vero cells were seeded in complete DMEM into wells of twelve-well plates. Ten-fold dilutions of LGTV were adsorbed for 60 minutes at 37uC, 5% CO 2 with constant rocking. Infected cells were then washed twice with Dulbecco's phosphate buffered saline (DPBS, Invitrogen) and fresh complete DMEM added. After three days incubation, cells were washed twice with DPBS and fixed with 100% methanol for 30 minutes. Cells were rinsed with DPBS, blocked with OptiMEM supplemented with 2% FCS for 10 minutes, and incubated with Russian Spring-Summer Encephalitis immune ascites fluid (ATCC) diluted 1:1000 in OptiMEM/2% FCS. Following two DPBS washes, cells were incubated with 1:1000 dilution of anti-mouse horseradish peroxidase antibody (Dako) in OptiMEM/2% FCS. Foci were visualized using peroxidase substrate containing 0.4 mg/ml 3, 39 diaminobenzidine (Sigma Aldrich) and 0.0135% H 2 O 2 (JT Baker) in PBS. For all experiments, Vero and ISE6 cells were infected at an MOI of 10. To establish a persistently infected culture, 1610 7 ISE6 cells were plated in a 150 cm 2 flask and incubated for 24 hours at 32uC. Cells were then infected at an MOI of 10 with LGTV TP21. Cells were passaged every two to three weeks as needed and supernatants collected for viral titer via immunofocus assay to confirm productive viral infection. Samples for immunofluorescence and electron microscopy were prepared at multiple time points post-infection. No differences were noted between samples collected at various time points. Representative images for figures depicted were acquired between 90 and 260 days post infection. The primary antibodies against viral proteins used were: mouse monoclonal anti-prM 13A10, IgG2a; mouse monoclonal anti-E 11H12, IgG2a; mouse monoclonal anti-NS1 6E11, IgG2a (all kind gifts from Dr. Connie Schmaljohn, USAMRID, Fort Detrick, Frederick, MD), chicken poly-clonal anti-NS3 (sequence: CZRDIREFVSYASGRR) and chicken polyclonal anti-NS5 (sequence: CZDRHDLHWELKLESSIF) (custom prepared by Aves Labs), mouse anti-dsRNA clone J2 (purchased from English & Scientific Consulting, Szirak, Hungary). Markers against cellular organelles used were: Alexa Fluor 594-conjugated wheat germ agglutinin (WGA, Invitrogen), and mouse monoclonal Dylight 488-conjugated protein disulfide isomerase 1D3 (PDI, Enzo Life Sciences). Secondary antibodies used were: Alexa Fluor 488-and 594-conjugated anti-mouse-and anti-rabbitspecific IgG and Alexa Fluor 647-conjugated anti-chickenspecific IgG (Invitrogen). Vero and ISE6 cells were seeded at 3610 4 cells/well in 8 well Labtek dishes (Nunc), infected with LGTV (MOI 10), and incubated for 48 hours at 37uC and 32uC, respectively. The cells were washed twice with PBS, fixed for 10 minutes in 4% paraformaldehyde (PFA)/5% sucrose in PBS, permeabilized with 0.1% Triton X-100/4% PFA for 10 minutes, washed with 50 mM glycine for 5 minutes, and blocked for 60 minutes with 2% bovine serum albumin (BSA)/PBS. Primary and secondary antibodies were incubated at a 1:1000 dilution in 2% BSA/PBS for 60 minutes each, with three five-minute washes with PBS between incubation steps. After washes, slides were dried and coverslips mounted using Prolong Gold Antifade with DAPI (49, 6-diamidino-2-phenylindole, Invitrogen). Confocal images were acquired using a Carl Zeiss LSM 710 confocal laser scanning microscope. TIFF files of individual channels and merge images were made using Bitplane Imaris 7.2 and images assembled for publication using Adobe Photoshop software. For transmission electron microscopy (TEM), Vero or ISE6 cells were seeded on 13-mm Thermanox coverslips (Nunc) in 12 well plates at 1610 5 cells/well. After 24 hours, cells were infected with LGTV TP21 at an MOI of 10 and incubated for 48 hours at 37uC and 32uC, respectively. Cells were then washed twice with DPBS and fixed with 2.5% glutaraldehyde in cacodylate buffer (100 mM sodium cacodylate, 50 mM KCl, 2.5 mM CaCl 2 ) for 30 minutes at room temperature prior to overnight incubation at 4uC. Samples were processed further in a Biowave model laboratory microwave oven, equipped with a Coldspot water circulator (Ted Pella, Inc.) as follows: washed twice in 0.1 M sodium phosphate buffer, pH 7.2, for 1 minute each; post-fixed in 1% osmium tetroxide in phosphate buffer for two cycles of 2 minutes on, 2 minutes off, and 2 minutes on; washed once in phosphate buffer for 1 minute and twice in water for 1 minute each; contrasted with 1% uranyl acetate in water for two cycles of 2 minutes on, 2 minutes off, and 2 minutes on; dehydrated in three changes of ethanol for 1 minutes each; and embedded in Spurr's resin using steps of 50%, 75%, and two changes of 100% resin in ethanol for two periods of 5 minutes each. The power output of the oven was set at 250 watts for dehydration and embedding. All other steps were performed at a setting of 167 watts. The cover slips were placed cell-side down onto resin block molds, polymerized overnight at 65uC, and removed from hardened blocks after a 5 second immersion in liquid nitrogen. Thin and semi-thick sections of approximately 70 and 250 nm, respectively, were cut using a diamond knife and a model EM UC6 microtome (Leica Microsystems). Thin sections were collected on uncoated 200 mesh copper grids. Once thoroughly dried, the sections were stained with fresh 1% lead citrate using 1 minute of microwave irradiation at 167 watts, followed by a 1 minute water wash at 167 watts and standard drop-wise washing. For electron tomography (ET), semi-thick sections were collected on carbon coated 200 mesh copper grids (Ted Pella, Inc.), dried, and stained and washed similarly, except that the microwave irradiation steps were extended to 2 minutes each. Tomography grids were then immersed briefly in goat anti-mouse IgG 10 nm gold conjugate (Ted Pella, Inc.) to add fiducial markers on both sides of the sections and then thoroughly washed with water. Thin sections were examined at 80 kV using a model H7500 transmission electron microscope (Hitachi High Technologies, Inc.). Digital images were captured using a model HR-100 camera system (Advanced Microscopy Techniques, Inc.) and processed with PhotoShop (Adobe Systems, Inc.). Semi-thick sections were examined at 120 kV using a G2 Spirit BioTwin model TEM (FEI, Inc.) equipped with a tilt stage, Xplore3D acquisition system, and model Ultrascan 1000 CCD system (Gatan, Inc.). Tilt series images were collected at 0 and +/268 degrees with 1 degree intervals. Images were aligned and reconstructed into tomogram volumes using Inspect3D software (FEI, Inc.) and rendered using Amira (Visage Imaging, Inc.). Vector-borne flaviviruses replicate in both mammalian and arthropod cells. However, the three-dimensional (3D) fine structure of TBFV replication in mammalian cells has not yet been compared to that in acutely or persistently infected tick cells. In order to undertake this comparison, baseline characterizations of acutely infected mammalian (African green monkey kidney, Vero) and tick (Ixodes scapularis embryo derived, ISE6) cells were performed by light microscopy and virus titer was assayed from culture supernatants. Uninfected Vero cell cultures formed a monolayer of uniform elongated, fibroblast-like cells that were firmly adherent to the culture vessel. Infected Vero cells first showed cytopathic effect (CPE) by 24 hpi, consisting of rounded cells and syncytia (Fig. 1A) . The CPE was progressive and by 96 hpi, Vero cells showed extensive evidence of cell death (shape irregularity, cytoplasmic condensation, blebbing, and an inability to exclude trypan blue). After 96 hours, very few cells remained attached to the culture vessel. ISE6 cell cultures, on the other hand, are a mixed population of cells consisting of clumps of loosely adherent, round cells as well as more firmly adherent, stellate cells with branching pseudopodia. In marked contrast to infected Vero cell cultures, there was no obvious evidence of CPE following acute infection of the ISE6 cells (Fig. 1A) . LGTV replicated well in both cell lines, but the titer was 2-3 logs lower in acutely infected ISE6 cells than in Vero cells (Fig. 1B) . Thus, there was a prominent difference in the cellular response to acute LGTV infection between mammalian and tick cells. For subsequent experiments, we selected a time point of 48 hpi as a compromise between virus production and integrity of cellular morphology. In nature, TBFV establish and maintain a persistent infection across the various life-stages of the arthropod host [8] . In an effort to model and study this aspect of TBFV infection in vitro, ISE6 cells were infected with LGTV, serially passaged at confluence, and examined at intervals for the induction and maintenance of a persistent infection. Interestingly, cultures of infected ISE6 cells appeared normal for a period of at least one year following initial infection. LGTV infection was maintained in the ISE6 cells, and the viral titer oscillated from a low of 2610 3 ffu/ml to a high of 2610 4 ffu/ml (Fig. 1C) . After roughly one year, approximately 50% of the cells were positive for viral proteins, and the culture supernatant at this point had a viral titer of 2610 4 ffu/ml (Fig. 1C) . Thus, a persistent infection had been established in the ISE6 cells and that infection was still productive a year later. The findings described in the previous section provided us with a model system in which to examine the distribution of several structural and non-structural proteins in infected mammalian and tick cells (Fig. 2) . The analysis is complicated by the fact that the basic cellular morphology differs substantially between the mammalian and tick cell lines. Nevertheless, in acute infections in Vero and ISE6 cells, as well as in the persistently infected ISE6 cells, M, E, and NS3 proteins were localized to the perinuclear region, although some staining was seen throughout the cytoplasm (Fig. 2B) . NS1 exhibited a more punctate cytoplasmic staining pattern and NS5 showed a diffuse cytoplasmic staining pattern with occasional evidence of nuclear staining. Similar patterns were seen for all the viral proteins in both cells lines. To more precisely evaluate the relative distribution of the proteins during infection, we co-stained specimens for multiple viral proteins. In each case, the bulk of NS3 staining was colocalized with both the viral membrane (prM) (data not shown) and envelope (E) proteins (Fig. 3A) ; although the overlap appeared less complete in the ISE6 cells. NS5 was diffusely distributed in the cytoplasm, but there was clear evidence that the staining for E overlapped (Fig. 3B) . The punctate staining pattern for NS1 also exhibited coincidence with prM, E, and NS3 but with fewer areas of colocalization with NS5 (data not shown) in both Vero and ISE6 cells. There were no appreciable differences between the acute and persistently infected ISE6 cells (Fig. 3) . Flavivirus replication and assembly takes place on membranes derived from ER and that maturation of virus particles occurs within the trans-Golgi [40] [41] [42] [43] . Thus, we next looked at LGTV protein localization in relationship to markers for ER (Dylight 488conjugated protein disulfide isomerase, PDI) or Golgi membranes (Alexa Fluor 594-conjugated wheat germ agglutinin, WGA). In both infected mammalian and tick cells (Fig. 4C) , we observed a notable increase in the amount of ER staining in comparison to mock infected cells (Fig. 4A & B) , while the amount of Golgi staining remained relatively constant (date not shown). In infected Vero cells, the bulk of structural protein E and the nonstructural protein NS3 colocalized in large concentrated areas of ER (Fig. 4C ) but a small amount of signal from E overlapped with the Golgi marker (WGA) (data not shown). Similar results were observed for both the acutely and persistently infected ISE6 cell cultures (Fig. 4C) . Thus, LGTV infection of mammalian and tick cells was accompanied by an increase in ER related structures, and furthermore, both structural and nonstructural LGTV proteins are concentrated in those areas. An obligate marker for flavivirus replication is the presence of the replicative form, double-stranded RNA (dsRNA). Therefore, to further demonstrate that LGTV replication was occurring in an ER derived compartment [44] , we co-stained infected Vero and ISE6 cells for dsRNA and either viral proteins or cellular markers. Uninfected cells showed no dsRNA staining (Fig. 5A) . In infected Vero cells, labeling for dsRNA was largely confined to areas staining for NS3 (Fig. 5B) and NS5 (Fig. 5C) . The same general pattern was seen in acutely ( Fig. 5B and 5C ) and persistently (data not shown) infected ISE6 cells, but the dsRNA signal was not as prominent as in the Vero cells. The majority of dsRNA labeling was coincident with ER staining (Fig. 6 ), but not with Golgi staining (data not shown) in acutely infected Vero or in ISE6 cells, both acutely and persistently infected. Taken together, these findings indicated that in both mammalian and In the previous sections, we showed that LGTV replication was occurring in part of the cytoplasm associated with expanded ER and that there was no obvious difference between mammalian and tick cells. However, immunofluorescence does not provide sufficient resolution to clearly define the cellular compartments associated with replication. Therefore, to compare TBFV infection in mammalian and tick cells at higher resolution, we used TEM of fixed, resin-embedded specimens to examine mock-infected Vero and ISE6 cells (Fig. 7A) , LGTV infected Vero, and either acutely or persistently infected ISE6 cells. In Vero cells, extensive areas of altered membrane proliferation (Fig. 7B-D) could be found in the majority of cells, consistent with the expansion of the ER system evident by fluorescence microscopy [27, 29, 30] . In some sections, these regions of proliferation encompassed nearly half of the cytoplasmic area, thus explaining our immunofluorescence results. Single-membrane bound vesicles, ranging in diameter from 60-100 nm, were frequently found within these proliferated ER areas, often occurring in large groups contained within dilated ER cisternae. Single, small groups, and large paracrystalline arrays of virions were readily observed (Fig. 7A) . Infrequently, tubular structures or elongated vesicles were seen (Fig. 7D) . Tubules have been seen in other reports on ultrastructure of mosquito-borne flavivirus mammalian cell infection [30, 45] and mosquito cell infection [26, 46, 47] ; however, no function has been assigned to these tubules. In acutely infected ISE6 cells, membrane proliferation and vesicles were also observed ( Fig. 7B & C) , but the extent was notably less widespread than in mammalian cells, consistent with the findings of previously published work [35] . The vesicles were also found in dilated ER cisternae and were the same diameter as those seen in Vero cells (Fig. 7C) . Virions were not detected until 96 hpi (data not shown), likely a consequence of the lower viral titer and slower replication observed in the tick cells (Fig. 1B) . The membranous tubules or vesicles were present at a slightly higher frequency than in acutely infected Vero cells. In persistently infected ISE6 cells, membrane proliferation was more extensive than in acutely infected tick cells; but, again, this was not as pronounced as in infected Vero cells (Fig. 7B) . Vesicles were again seen within proliferated, dilated ER and remained within the same size range as in acute infection. Virions were not noted in the samples we examined. One difference in the persistently infected ISE6 cells was particularly striking. Tubules were seen in almost all persistently infected ISE6 cells and often occurred in fascicle-like bundles of multiple tubules cloaked by a single membranous sheath (Fig. 7D ). The recent application of electron tomography (ET) to the study of virus infection has allowed 3D evaluation of virus replication [48] [49] [50] [51] [52] [53] including in mosquito-borne flavivirus infection [30, 31] . We employed ET of 250 nm thick, fixed, and resin-embedded specimens to enhance the findings described in the earlier sections. Tilt series were acquired, aligned using gold particles, and tomograms assembled to reveal 3D structures. In Vero cells acutely infected with LGTV, we found the vesicles and virions were contained within an anastomosing network of dilated membranes. The 60-100 nm diameter vesicles were round in shape ( Fig. 8A and Movie S1) and there were pore-like openings connecting vesicles to the cytoplasm (Fig. 9A ) and to other vesicles (Fig. 9D) . In tick cells, acute LGTV infection induced 3D rearrangements similar to those seen in Vero cells (Fig. 8B and Movie S2). The membrane proliferation, while not as marked as in Vero cells, still resulted in a continuous network that surrounded the vesicles. The majority of vesicles observed in the tick cells were round (Fig. 7C & D) ; however, slightly elongated vesicles were seen more frequently than in Vero cells. Communicating pores or openings between the vesicles and the cytoplasm (Fig. 9B ) and between individual vesicles (Fig. 9E) were also seen in the tick cells. The persistently infected tick cells also contained round vesicles enclosed within dilated membrane structures (Fig. 8C and Movie S3). The 3D analysis confirmed that the elongated profiles were in fact tubular, and that they were clustered in fascicle-like groups surrounded by a single membrane. These structures were cylindrical in shape with closed ends, varying in width from 60 to 100 nm and in length from 100 to 800 nm. Pores were seen between tubules and vesicles ( Fig. 9C & F) ; however, in our examinations of numerous tubular structures, we were unable to identify pores linking tubules with other tubules or with the cytoplasm. structure. Similar to other positive strand viruses, such as Semliki Forest virus [54] , rubella virus [52] , and poliovirus [51] , the flaviviruses induce remarkable alterations in the cytoplasmic membrane system. In our study, Vero cells acutely infected with LGTV show the ER proliferation, vesicle accumulation, and membrane network (Fig. 7B-D) described in earlier studies with mosquito-borne flaviviruses [30, 31] . A number of functions have been ascribed to the alterations, including concentration of viral replication machinery, provision of a solid state platform for viral protein synthesis and replication, and sequestration of viral dsRNA replicative form from innate immune sensors [27, 55, 56] . In general, viral protein distribution in acutely infected tick cells was similar; however, some differences were apparent. Vero cells had higher levels of ER proliferation and rearrangement as well as significantly more viral particles. This may have been the result of the greater level of replication seen in Vero cells but, it is also possible that the differences are the result of a fundamental difference between the two host cell lines. In our study, an absence of observed viral particles in LGTVinfected ISE6 cells did not allow for confirmation of previously reported differences seen with TBEV [35] between virus particle location in mammalian versus arthropod cells. However, we have augmented this earlier published work by reporting the 3D structure of acutely infected tick cells. By ET, the proliferating membranes are revealed to be a complex anastomosing system of membranes almost certainly derived from ER. In the Semliki Forest virus replicon system containing Kunjin virus proteins, expression of a single flavivirus protein, NS4A, can induce the membrane rearrangements [22] , and similar results have been reported for Dengue virus NS4A [57] , but it is uncertain at this time if this role can be assigned to NS4A in the TBFV. Future investigation is in progress to determine which TBFV protein or combination of proteins cause membrane rearrangement in mammalian and tick cells. The circular profiles seen in TEM were clearly demonstrated by ET to be spherical vesicles bound by a single membrane and to have pore-like connections to the cytoplasm or to other vesicles in Vero and ISE6 cells. The function of the vesicles is thought to be to minimize exposure of the replicative form dsRNA to innate immune sensors, such as RIG I [27, 58, 59] , while the pores are thought to provide a conduit for nucleotides, amino acids, and other components required for replication and gene expression [30] . In brief, our findings suggest marked similarities between acute replication in ISE6 and Vero cells, and further demonstrate the similarity between cellular features of acute tick-borne and mosquito-borne [30, 31] flavivirus replication. Our work also allowed a detailed examination of cellular features of persistent LGTV infection in ISE6 cells and an opportunity to compare acute and persistent infection in tick cells. Over a yearlong period, the persistently infected ISE6 cells maintained a grossly normal morphology. Furthermore, the distribution of viral proteins and their co-localization with cellular markers mirrored the acute infection. However, study by EM and 3D reconstruction revealed differences between the persistently infected and the acutely infected ISE6 cells. The persistently infected tick cells had greater changes in ER structure and ER abundance (Fig. 7B-D) than the acutely infected cells, although viral titers (Fig. 1B & C) were similar in both settings. However, the most remarkable difference between acute and persistent infection was the number of tubular structures. These structures have been noted infrequently in infected cells [26, 30, 34] , and their relevance in flavivirus replication is uncertain. In our acute mammalian and tick cell experiments, we confirmed these observations and saw tubules only occasionally (Fig. 7D, 8 , and Movies S1, S2, and S3). However, in the persistently infected tick cells, the number of tubules increased dramatically (Fig. 7D, 8C , & Movie S3). While the diameter was similar to that seen in round vesicles (60-100 nm), the length was up to eight times as long. The tubules were often arranged in membrane-bound, fascicle-like bundles ( Fig. 8C and Movie S3). The 3D reconstructions demonstrated that the tubules were closed on each end and, although closely juxtaposed, were not connected by pores to other tubules or to the cytoplasm, unlike the round vesicles. The function of the tubules is obscure at this time, and it is unclear if they represent bona fide features of replication, aberrant structures, or the result of a cellular process to restrict infection. It is possible the tubules may play a role in initiation or maintenance of the persistent infections or if the apparent lack of pores in the tubules is consequential. It Figure 6 . Localization of LGTV TP21 replication complex to endoplasmic reticulum. Cells were infected at a MOI of 10, fixed, and costained for double-stranded RNA (dsRNA, red) and endoplasmic reticulum (protein disulfide isomerase (PDI), green). Nuclei counterstaining (DAPI, blue) is only shown in the merge panel. Scale bars, 10 um. doi:10.1371/journal.pone.0047912.g006 3D LGTV Infection in Mammalian and Tick Cells PLOS ONE | www.plosone.org may be that the increase in the number of tubules is the result of the higher number of defective virus particles, which are known to exist in persistently infected cells [60] [61] [62] . Tubules may be formed as a result of a failure to correctly gather the membrane to form the round vesicles and pores that are associated with the flavivirus replication complex. Perhaps the lack of pores prevents proper replication or packaging of the viral genome. Additional cell biology or biochemistry studies may shed light on the role of the tubules. In summary, our experiments have provided the first analysis of the 3D structure of tick-borne flavivirus infection in both mammalian and arthropod host cell systems. We observed vesicles with pores connecting to other vesicles or opening to the cytosol in tick-borne flavivirus infection, similar to those seen in mosquito-borne flavivirus infection. We have shown for the first time the 3D ultrastructure of acutely and persistently, flavivirusinfected arthropod cells, facilitating the observation of the shift that occurs from round vesicles during acute TBFV-infection to the elongated tubules that dominate persistent infection. Future experiments are needed to better understand the increasing presence of tubules at the site of membrane rearrangement during persistent infection. Movie S1 LGTV TP21-induced structures in acutely infected Vero cells. Animation through a z-series and 3D surface rendering of a semi-thick section of an acutely infected Vero cell. ER is depicted in green, vesicles in blue, and virions in red. Both virions and vesicles are contained within a network of proliferated ER. The images were aligned using Inspect3D software (FEI, Inc.) and rendered using Amira (Visage Imaging, Inc., San Diego, CA). (MP4) LGTV TP21-induced structures in acutely infected ISE6 cells. Animation through a z-series and 3D surface rendering of a semi-thick section of an acutely infected ISE6 cell. ER is depicted in green and vesicles & tubules in blue. Vesicles and tubules are contained within proliferated ER. Tubules are short in length, only reaching approximately twice the diameter length. The images were aligned using Inspect3D software (FEI, Inc.) and rendered using Amira (Visage Imaging, Inc., San Diego, CA). Movie S3 LGTV TP21-induced structures in persistently infected ISE6 cells. Animation through a z-series and 3D surface rendering of a semi-thick section of a persistently infected ISE6 cell. ER is depicted in green and vesicles & tubules in blue. Numerous long tubules are seen in a large bundle; however, smaller tubules and round vesicles are also seen. The images were aligned using Inspect3D software (FEI, Inc.) and rendered using Amira (Visage Imaging, Inc., San Diego, CA). (MP4)