Genomic scale analysis of racial impact on response to IFN-α

Zoltan Posa,1, Silvia Sellerib, Tara L. Spiveya, Jeanne K. Wangc, Hui Liua, Andrea Worschechd,e, Marianna Sabatinof, Alessandro Monacoa, Susan F. Leitmang, Andras Falush,i, Ena Wanga, Harvey J. Altera,1, and Francesco M. Marincolaa,1

aInfectious Disease and Immunogenetics Section, Department of Transfusion Medicine, Clinical Center, and Center for Human Immunology, National Institutes of Health, Bethesda, MD 20892; bSan Raffaele Telethon Institute for Therapy, Scientific Institute H. S. Raffaele, Milan, 20132, Italy; cDivision of Medical Imaging and Hematology Products, Office of Oncology Drug Products, Center for Drug Evaluation and Research, Food and Drug Administration, Silver Spring, MD 20993; dGenelux Corporation, San Diego Science Center, San Diego, CA 92109; eVirchow Center for Experimental Biomedicine, Institute for Biochemistry and Institute for Molecular Infection Biology, University of Würzburg, Würzburg 97074, Germany; fCell Processing Section, Department of Transfusion Medicine, Clinical Center, National Institutes of Health, Bethesda, MD 20892; gBlood Services Section, Department of Transfusion Medicine, Clinical Center, National Institutes of Health, Bethesda, MD 20892; hDepartment of Genetics, Cell and Immunobiology, Semmelweis University, Budapest 1089, Hungary; and iInflammation Biology and Immungenomics Research Group, Hungarian Academy of Sciences, Semmelweis University, Budapest 1089, Hungary

Contributed by Harvey J. Alter, November 25, 2009 (sent for review October 23, 2009) Limited responsiveness to IFN-α in hepatitis C virus (HCV)-infected HCV infection suppresses IFN-α signaling, and HCV viral par- African-Americans compared to European Americans (AAs vs. EAs) ticles are capable of suppressing STAT1 transcription and hinders the management of HCV. Here, we studied healthy non- phosphorylation, decreasing its half-life and inhibiting nuclear HCV-infected AA and EA subjects to test whether immune cell transport and DNA binding of activated STAT1 (7). Hence, in response to IFN-α is determined directly by race. We compared HCV-infected individuals, it is difficult to conclusively test baseline and IFN-α-induced signal transducer and activator of tran- whether the poor response of AAs to IFN-α/ribavarin treatment scription (STAT)-1, STAT-2, STAT-3, STAT-4, and STAT-5 protein and and their impaired response to IFN-α ex vivo is due to a genetic phosphorylation levels in purified T cells, global transcription, and trait that directly affects IFN signaling or is secondary to a dif- a genomewide single-nucleotide polymorphism (SNP) profile of ferential effect of HCV infection on the immune response of healthy AA and EA blood donors. In contrast to HCV-infected indi- AAs compared to EAs. Moreover, in HCV-infected patients it is viduals, healthy AAs displayed no evidence of reduced STAT acti- not possible to separate race-dependent markers of IFN-α vation or IFN-α-stimulated compared to EAs. responsiveness from genetic polymorphisms indirectly affecting Although >200 reacted to IFN-α treatment, race had no resistance to viral infection or resistance to viral interference impact on any of them. The only gene differentially expressed with IFN-α signaling. In this study, we provide evidence that in by the two races (NUDT3, P < 10−7) was not affected by IFN-α the absence of HCV infection, race does not affect ex vivo IFN-α and bears no known relationship to IFN-α signaling or HCV patho- responsiveness. Racial traits affecting IFN-α treatment in HCV genesis. Genomewide analysis confirmed the self-proclaimed infection may be restricted to polymorphisms in genes directly racial attribution of most donors, and numerous race-associated interacting with or affected by HCV during the natural history of SNPs were identified within loci involved in IFN-α signaling, the disease. although they clearly did not affect responsiveness in the absence of HCV. We conclude that racial differences observed in HCV- Results infected patients in the responsiveness to IFN-α are unrelated to IFN-α Signal Transduction Is Highly Conserved Between Healthy AA inherent racial differences in IFN-α signaling and more likely due and EA. STAT1 phosphorylation (STAT-P) is a key component of to polymorphisms affecting the hosts’ response to HCV, which in the Janus kinase (JAK)-STAT signaling pathway correlated to turn may lead to a distinct disease pathophysiology responsible the activation of IFN-stimulated genes (ISGs) (8). Altered for altered IFN signaling and treatment response. phosphorylation of STAT1 in response to IFN-α stimulation of PBMC in vitro was proposed as a predictor of poor respon- African-American | European American | microarray | signal transducers siveness to therapy of HCV with the same agent particularly in and activators of transcription | hepatitis C virus AAs (2). Therefore, we compared STAT-P induction between healthy EAs and AAs. Pilot studies in our lab confirmed pub- urrent treatment of chronic hepatitis C virus (HCV) infec- lished reports (9) that among all PBMC subpopulations, T cells tion relies on combinational therapy with the antiviral cyto- and monocytes express STAT1 at the highest level and phos- C fi α kine IFN-α and the antiviral nucleoside analog ribavirin. A major phorylate it most ef ciently upon IFN- stimulation (Fig. S1)

factor affecting success of therapy is the HCV-infected hosts’ and that the 200 IU/mL concentration commonly used for IMMUNOLOGY responsiveness to IFN-α. HCV-infected African-Americans PBMC stimulation is roughly equivalent to that required for the (AAs) exhibit limited responsiveness to IFN-α compared to ED50 of STAT1 phosphorylation in T cells (ED50 IFN-α = 213 IU/ other races, such as European Americans (EAs), resulting in a massively diminished AA:EA odds ratio for achievement of – Author contributions: Z.P., S.F.L., E.W., H.J.A., and F.M.M. designed research; Z.P., S.S., T.L.S., sustained virological response (1 5). In line with this clinical J.K.W., H.L., and A.W. performed research; Z.P., T.L.S., M.S., A.M., E.W., and F.M.M. con- reality, peripheral blood mononuclear cells (PBMC) of HCV- tributed new reagents/analytic tools; Z.P., E.W., and F.M.M. analyzed data; and Z.P., S.F.L., infected AA patients display limited signal transducer and acti- A.F., H.J.A., and F.M.M. wrote the paper. vator of transcription (STAT)1 phosphorylation compared with The authors declare no conflict of interest. EA patients and an altered gene expression profile upon stim- Freely available online through the PNAS open access option. ulation with IFN-α ex vivo (2). These observations led to the Data deposition: The data reported in this paper have been deposited in the Gene Ex- hypothesis that the geo-ethnic background of Americans might pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE17952). impact IFN-α signaling upstream of STAT1 and compromise 1To whom correspondence may be addressed at: Infectious Disease and Immunogenetics responsiveness of AAs to therapy. Section, Department of Transfusion Medicine, Clinical Center, National Institutes of Health, Building 10, R1C711, 9000 Rockville Pike, Bethesda, MD 20892. E-mail: [email protected]. Interestingly, polymorphisms of genes involved in type I IFN nih.gov, [email protected] or [email protected]. signaling or function have been associated with racial differences This article contains supporting information online at www.pnas.org/cgi/content/full/ in the outcome of chronic HCV infection (6). Furthermore, 0913491107/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0913491107 PNAS | January 12, 2010 | vol. 107 | no. 2 | 803–808 Downloaded by guest on September 30, 2021 mL; Fig. S2). On the basis of this observation, the critical role of levels (Fig. 1 B and C), or the ratio of STAT or STAT-P positive T cells in resolving HCV infection (10), and their relative cells (Fig. S3). Moreover, baseline and IFN-α-induced STAT abundance compared to monocytes, we used T cells purified by protein phosphorylation levels and phosphorylation fold changes negative selection. were highly variable and showed generally poor correlation In contrast to the marked differences observed in HCV- among STAT proteins (with the exception of STAT1-P and infected patients (2), STAT1-P fold change (FC) upon IFN-α STAT3-P; R2 = 0.522, P value <0.001), clearly challenging the stimulation was not significantly different between healthy AA assumption that STAT1 phosphorylation can exclusively repre- and EA donors (Fig. 1B), nor were baseline and induced STAT1- sent the complexity of the IFN-α response ex vivo (Fig. 1 A P levels or levels of STAT1 protein (Fig. 1C). In addition, race and D). did not affect responsiveness as measured by ratios of STAT1-P To test whether the lack of racial difference in IFN-α-induced or STAT1-positive cells in control and IFN-α-affected samples STAT1-P in healthy individuals was limited to IFN-α, T cells (Fig. S3). The only difference observed between the two ethnic were treated with cytokines using signaling mechanisms similar groups was intraracial response heterogeneity, which was found to IFN-α (IFN-β), highly similar intracellular signaling but dis- to be greater in AAs, compared to EAs. By arbitrarily splitting tinct receptor usage (IFN-λ), overlapping at one or more levels the response phenotype into low, medium, and high (STAT1-P of the JAK/STAT pathway (IL-2 and IFN-γ), or, as a negative FC <1.5, 1.5–3.0, and >3.0, respectively), AA values scattered in control, unrelated to IFN-α signaling (IL-4). No race-associated a uniform distribution among all categories, whereas EA values difference in induced STAT1 phosphorylation was observed prevailed in the medium response group (Fig. 1B, P = 0.028, χ2- regardless of the cytokine tested (Fig. 2A). However, we test). In smaller substudies comparing 12 donors from each of observed that induced STAT1-P FC was highly correlated among the two races, we evaluated all other STAT proteins known to be cytokines, underlining the limited specificity of STAT1-P FC as a affected by IFN-α in T cells (STAT2, STAT3, STAT4, and marker of in vitro IFN-α responsiveness (Fig. 2 B–E). Finally, in STAT5). We found no significant differences between races in line with others’ observations (11), we observed that CD3+ T baseline or induced STAT phosphorylation levels, STAT protein cells’ in vitro responsiveness to IFN-λ is very limited (Fig. 2A).

Fig. 1. No difference between healthy EAs’ and AAs’ IFN-α-induced STAT1, STAT2, STAT3, STAT4, and STAT5 activation. Activity of STAT1, STAT2, STAT3, STAT4, and STAT5 signal pathways is displayed, as affected by IFN-α and race. (A) Representative flow cytometry data for each STAT protein. (B) IFN-α-induced fold change in STAT protein phosphorylation. (C) STAT protein levels (STAT Prot), baseline (Ph C), and IFN-α-induced (Ph IFNa) STAT phosphorylation levels. Median is indicated by horizontal lines, and P values obtained from Student's t-tests or the Mann–Whitney rank sum test, as appropriate, are shown above sample groups. (D) Correlation between STAT1 and other STAT proteins’ phosphorylation in the same samples along with R and P values obtained from Pearson's correlation analyses.

804 | www.pnas.org/cgi/doi/10.1073/pnas.0913491107 Pos et al. Downloaded by guest on September 30, 2021 tion, antigen presentation, and intracellular recognition of viral pathogens, etc. (Table 2). In contrast, race was found to have virtually no impact on the gene expression profile of the analyzed samples, as there was only one transcript that was both significant and reproducibly different between races. The only transcript affected by race was NUDT3, whose expression was decreased in AAs. NUDT3 is a gene involved in nucleoside metabolism, playing a role in the removal of some toxic nucleotide metabolites, expressed almost ubiquitously in all human tissues, but with no known relationship to HCV infection, IFN-α signaling, antiviral responses, or other immune-related processes. Hence, ethnic differences in NUDT3 expression are unlikely to explain the differential response to IFN-α of HCV-infected individuals. In summary, our data sug- gest a marked dissociation between the number of race-affected genes compared to IFN-α-induced genes in healthy donors. Interestingly, this dissociation remained highly consistent regardless of statistical selection criteria of DEGs (Fig. 4A). Moreover, genes potentially affected by race did not overlap with IFN-α-affected genes at any FC or α-levels chosen for analysis (Fig. 4A); in addition, mixed-effects model ANOVA Fig. 2. Healthy EAs and AAs do not differ in activation of STAT1 by various confirmed that there were no genes affected by an interaction cytokines, using signal transduction pathways overlapping with IFN-α.(A) between race and IFN-α treatment (genes expressed at different Fold change STAT1 phosphorylation induced by various cytokines capable of α activating STAT1, as indicated, with IL-4 serving as negative control. (B–E) levels depending on race only before or after IFN- treatment) Correlations between STAT1 phosphorylation induced by various cytokines at any α-levels and FC criteria used (Fig. 4A). In contrast to along with R and P values obtained from Pearson's correlation analyses. HCV-infected patients, race did not affect average fold change of ISGs in healthy donors (Fig. 4B). As opposed to IFN-α treatment, race had virtually no effect on the ISG expression Healthy Individuals Exhibit Activation of ISGs Regardless of Race. profile as judged by unsupervised hierarchical clustering or Global gene expression profiling was used to identify genes whose multidimensional scaling (Fig. 4 C and D). expression could be affected by race, IFN-α treatment, both race and treatment, or interactions between race and treatment. To IFN-α-Independent Activation of a Subgroup of ISGs Accounts for optimize the trade-off between size and reproducibility/reliability Baseline Heterogeneity Among Healthy Donors. Whereas ISG of the list of differentially expressed genes (DEGs), several fold expression following IFN-α stimulation was consistent among change cutoff values (FC >1, >1.5, and >2.0 in >25% of all sam- individuals of either race, clear heterogeneity was observed in − − ples) and various P-value criteria (P < 10 3–10 9) were tested by the expression of a subgroup of ISGs in baseline conditions analyzing a “training” array set (Table 1), using mixed-effects (unstimulated T cells). As in chronic HCV, pretreatment acti- vation of prominent classical ISGs and canonical ISG pathways model ANOVA. The resulting gene lists were retested on a sec- α ond, nonoverlapping “test” array set (Table 1) for their reprodu- has been reported to be associated with nonresponse to IFN- therapy (12, 13). We decided to further analyze unstable ISGs in cibility, i.e., their ability to reproduce sample separation by major detail. In contrast to HCV-infected patients, this subgroup of experimental factors (race, treatment) in an independent sample transcripts (further referred to as baseline unstable ISGs, Fig. 3 set by unsupervised hierarchical clustering. A B fi > P < −7 and ) included genes whose function is not directly associated In general, gene sets de ned by FC 1.5 and 10 were with canonical IFN-regulated pathways, whereas the rest of the found to be the most reproducible at this sample size and ISGs (designated as baseline stable ISGs) were selectively composition (156 arrays total), while still containing the most regulated by IFN-α stimulation (see Table S2 for lists of baseline α comprehensive DEG list. Using these criteria, IFN- treatment stable and unstable ISGs). Comparative analysis of the training A affected 222 transcripts (205 annotated, Fig. 3 , see Table S1 for and test sample sets by hierarchical clustering confirmed that this detailed annotation and expression data). Pathway analysis of phenomenon affected a reproducible set of genes in both the genes significantly affected by IFN-α treatment confirmed func- training and the test sample sets analyzed (Fig. 3 A and B). tional reliability of these findings, disclosing that these ISGs Ingenuity pathway analysis (IPA) suggested that baseline IMMUNOLOGY constitute canonical gene networks involved in well-known IFN- unstable ISGs of healthy individuals covered different aspects of affected processes, such as IFN-α signaling, protein ubiquitina- cellular metabolism, in sharp contrast with baseline stable ISGs

Table 1. Summary of donor numbers, age, race, and gender distribution in all experimental sample sets analyzed in this study Sample Total no. Size African: No. Age, years: AA male: No. AA female: No. European: No. Age, years: EA male: No. EA female: No. set donors (n) (% total) Mean ± SEM (% total) (% total) (% total) Mean ± SEM (% total) (% total)

Master 78 156 37 (47.4) 45.9 ± 1.5 30 (38.5) 7 (8.9) 41 (52.6) 49.1 ± 2.3 32 (41.0) 9 (11.5) Training 39 78 18 (46.2) 47.9 ± 2.2 14 (35.9) 4 (10.3) 21 (53.8) 47.8 ± 3.2 16 (41.0) 5 (12.8) Test 39 78 19 (48.7) 44.0 ± 2.1 16 (41.0) 3 (7.7) 20 (51.3) 51.0 ± 3.4 16 (41.0) 4 (10.3) Repeat 15 60 9 (60.0) 45.1 ± 2.0 8 (53.3) 1 (6.7) 6 (40.0) 45.1 ± 2.0 6 (40.0) 0 (0.0)

The master set includes all donors analyzed in this study, and the training and test sets represent two nonoverlapping halves of the master set; all donors of the master, training, and test sets are represented by one single blood donation in all analyses. Blood donations of the third repeat set were derived from donors donating twice in two random independent time points over a period of 1 year; in this set, each donor is represented by two donations. Statistical group (sample) size of sample sets (n) is equal to (the number of donors) × (number of donations/donor) × (2), as each donation has been split into a control and an IFN-α–treated parallel in all analyses.

Pos et al. PNAS | January 12, 2010 | vol. 107 | no. 2 | 805 Downloaded by guest on September 30, 2021 Fig. 3. Characterization of IFN-α-stimulated genes, assessment of response heterogeneity, and stability of individual differences in healthy EAs and AAs. (A) − IFN-α-stimulated genes (ISGs) identified by microarrays and mixed-effects model ANOVA (P > 10 7,FC>1.5) in a “training” sample set consisting of 78 samples. (B) The same ISG set is tested for reproducibility by unsupervised hierarchical clustering of an independent “test” sample set of 78 samples, collected from different individuals. Untreated controls are labeled with purple and IFN-α-treated samples with pink bars. Major heterogeneity between individuals is shown in the form of ISGs displaying frequent, IFN-α-independent activation in the control groups. These genes, designated as baseline unstable ISGs, are shown marked with yellow bars. (C) Limited stability of this phenomenon over time in a third, “repeat” sample set consisting of samples derived from identical donors at two different time points. Samples derived from the same individuals are shown color coded.

clearly associated with canonical IFN-α signaling pathways such immune cellular functions exerted by IFN-α, predominantly as protein ubiquitination, antigen presentation, and JAK-STAT those related to metabolic processes. We next analyzed whether signaling, etc. (Table 2). Hence, gene expression analysis sug- preactivation of the baseline unstable ISG module was a stable gests that immunity-related core IFN-α functions are relatively phenotype for a discrete set of individuals. Using the repeat homogenously activated by IFN-α in healthy individuals and are, sample set, consisting of repeated collections derived from the in general, inactivated in baseline conditions. Most individual same donors at different time points, we observed that activation heterogeneity is present in genes associated with diverse non- of baseline unstable ISG was not idiosyncratic and could vary at different donations from the same individual over time (Fig. 3C); thus, baseline variations in unstable ISGs are also unlikely to be Table 2. Functional distribution of all baseline stable and genetically determined. unstable ISGs among canonical Ingenuity pathways We found that STAT1-P FC showed very limited correlation with ISG expression. Individuals with low, medium, and high All ISGs −Log(P) Ratio STAT1-P FC values presented some limited coclustering, but no IFN signaling 10.60 0.31 striking separation by ISG expression patterns (Fig. 4C). Con- Activation of IRF by cytosolic PRRs 5.02 0.09 sistent with this, multidimensional scaling revealed that lower Polyamine regulation in colon cancer 3.62 0.09 numbers of STAT1-P-positive cells usually lead to weaker Protein ubiquitination pathway 2.66 0.03 responses at the ISG induction (P < 0.001); however, generally Antigen presentation pathway 2.17 0.08 this marker was not able to predict the response at the ISG level Growth hormone signaling 2.12 0.05 (Fig. S4). Prolactin signaling 2.01 0.05 RIG1-like receptors in antiviral responses 1.80 0.06 Whole-Genome SNP Profiling Confirms Racial Identity and Discloses a List of Race-Associated Polymorphisms of ISGs. Whole-genome SNP Baseline stable ISGs profiling was used to analyze the distribution of SNP patterns IFN signaling 13.10 0.31 between EAs and AAs, with particular focus on IFN-α-related Activation of IRF by cytosolic PRRs 6.89 0.09 genes. It has been shown recently that Africans residing in Africa Protein ubiquitination pathway 4.35 0.03 can be heterogeneous at the SNP level, showing distinct sepa- Polyamine regulation in colon cancer 3.66 0.07 ration from other non-African ethnic groups, whereas African- Prolactin signaling 3.01 0.05 Americans have a more homogeneous West African origin and Antigen presentation pathway 2.96 0.08 are often admixed to non-African populations (14). Thus, a JAK/STAT signaling 2.19 0.05 confirmation of the self-proclaimed ethnicity of individual Gly, Ser, and Thr metabolism 2.15 0.02 donors was deemed necessary to enhance the accuracy of our study. Whole-genome SNP arrays corresponded accurately to the Baseline unstable ISGs subjective racial self-identification of donors; analysis of 848,365 CNTF signaling 1.54 0.04 SNPs that passed instrumental and methodological quality con- Pyrimidine metabolism 1.42 0.01 trol (QC) criteria (of ∼960,000 present in the array platform) Inositol metabolism 1.33 0.02 clustered donors according to their self-proclaimed race with Dopamine receptor signaling 1.22 0.02 only two exceptions (Fig. 5A, two self-proclaimed AAs clustering Ceramide signaling 1.20 0.02 with EAs). Finally, DNA samples of blood donors whose samples Fatty acid biosynthesis 1.19 0.02 were obtained on two different occasions clustered together and Phospholipid degradation 1.16 0.02 displayed a virtually identical genotype in repeat samples Purine metabolism 1.15 0.01 (Fig. 5B). − fi Race bore a huge impact on the SNP profile, resulting in Log(P value) indicates signi cance of association with a canonical Ingen- fi fi uity pathway; the eight most significant pathways are shown. Ratio values the identi cation of 26,026 SNPs speci c to the AA population, are calculated by dividing the number of genes that meet cutoff criteria by more or less evenly distributed along the genome (Fig. 5C) the total number of genes that make up the pathway. [as 848,365 SNPs passed the QC filters, the P value after the

806 | www.pnas.org/cgi/doi/10.1073/pnas.0913491107 Pos et al. Downloaded by guest on September 30, 2021 Fig. 4. Healthy EAs and AAs do not differ in IFN-α-induced ISG activation. (A) Summary of the number of genes affected by IFN-α treatment, race, and interaction between race and treatment as defined by mixed-effects model ANOVAs applying various combinations of fold change and significance criteria. (B) Analysis of the impact of race on average ISG response intensity by displaying trimmed mean (5–5% on both ends) fold change values of individual ISGs, ranked by average fold change. (C) The impact of IFN-α Fig. 5. Identification of SNPs linked to EA and AA genetic background, ISGs, treatment on ISG expression in comparison with race and STAT1 phos- and IFN-α response by whole-genome SNP analysis. (A)Efficient separation of phorylation by unsupervised hierarchical clustering. Samples are color coded EA and AA donors according to self-proclaimed race by unsupervised hier- for treatment (purple, control; pink, IFNa), race (pale blue, European; yellow, archical clustering of 848,365 SNPs; self-proclaimed race is shown by color code African-American), and STAT1-P FC (pale gray, low STAT1-P FC; medium (pale blue, European; yellow, African-American). (B) Accuracy of genomic- gray, medium STAT1-P FC; dark gray, high STAT1-P FC). (D and E) Results of a scale SNP profiling by showing virtually identical SNP patterns obtained from similar analysis after reduction of ISG data complexity to three dimensions (x, repeated donations (labeled a–e). (C) Distribution of 26,026 SNPs significantly y, and z) by multidimensional scaling (MDS). MDS reveals that axis x, rep- linked to AA background on all human (Dataset S1), including resenting the principal difference between all samples, is bona fide equiv- 158 race-linked, ISG associated SNPs (Dataset S2). (D) Significantly race-linked alent with the IFN-α treatment effect, whereas race is virtually irrelevant. SNPs on 19, which contains a suspected hotspot of IFN-α response. (E) None of the SNPs detected in the hotspot region (blue frame) was significantly linked to race. In C–E,they axis indicates uncorrected P values, IMMUNOLOGY Bonferroni correction for multiple testing (0.05/848,365) was P < whereas horizontal red lines indicate Bonferroni-corrected significance − < 5.89 × 10E 8; see Dataset S1 for a full list of SNPs linked to race]. thresholds, equivalent to a global P value of P 0.05. Concordantly, of the 5,320 SNPs linked to the 222 ISGs identified fi by gene expression pro ling, many were affected by race (see full ences between healthy EAs and AAs in the SNPs genotyped in the list of 158 SNPs associated to both race and ISGs in Dataset S2). IFN-λ 3/2 region (Fig. 5E). These findings are remarkable, considering that in spite of the relatively large number of race-linked SNPs in ISGs, none had a Discussion fi signi cant racial impact on the mRNA-level expression of its Our study shows that previously reported racial differences in associated gene. responsiveness to IFN-α under pathological conditions are not Finally, it was proposed that chromosome 19 contains a hotspot α present in healthy individuals. Thus, differences in clinical of several SNPs strongly affecting IFN- responsiveness in chronic α HCV infection and spontaneous HCV clearance (15–17), most of response to IFN- between HCV-infected AAs and EAs are not them associated to the IL28B/A (also known as IFN-λ 3/2) region; due to an inherent defect(s) in the signal transduction machinery α thus, additional analysis was performed focusing on this region. or the transcriptional regulation of ISGs downstream of IFN- Although race-linked SNPs are numerous, and evenly distributed signaling, but rather to the way AAs interact with or respond to on chromosome 19 (Fig. 5D), we did not find significant differ- HCV and how extensively or rapidly this response changes over

Pos et al. PNAS | January 12, 2010 | vol. 107 | no. 2 | 807 Downloaded by guest on September 30, 2021 time during the course of infection. The key determinant could Materials and Methods be how racial polymorphisms interact with the virus or virus– Detailed methods are available in SI Materials and Methods. drug interactions rather than with the drug itself. Among possi- ble explanations, both races may react to IFN-α similarly but (i) Blood Samples. Ninety-six consecutively collected blood donations, donated HCV suppresses IFN-α signaling in AAs more efficiently than in for research purposes with informed consent, were collected from 78 healthy EAs, (ii) the immune response to HCV among AAs is somehow EA and AA donors by the Department of Transfusion Medicine, Clinical Center, National Institutes of Health with Institutional Review Board approval (n =78, different, creating a difference in the course of the disease that Table 1). results in a decreased efficiency of IFN-α,or(iii) patients with chronic HCV infection represent a population different from Sample Processing and Treatments. Whole blood samples were processed by healthy individuals or blood donors, and hence the genetic leukapheresis and Ficoll density gradient centrifugation, and untouched T background of the one is not representative for the other. cells were isolated by magnetic-bead associated cell sorting, using Miltenyi’s Our findings are in line with the recent observation (15–17) that Human Pan T cell isolation kit for negative selection. T cell purity was + in chronic HCV patients, responsiveness to IFN-α therapy is not checked by staining CD3 cells for flow cytometry and found to be 92–95%. α fl affected by SNPs related to IFN-α or IFN-α-stimulated genes, but Cells were stimulated with IFN- 2b (200 IU/mL) for 15 min for ow cytom- etry or 6 h for gene expression profiling. strongly influenced by those associated to IFN-λ loci. Likely, λ variants of IFN- , which is highly expressed in hepatocytes, may Flow Cytometry. Cells were double stained for intracellular STAT1, -2, -3, -4, and differentially alter the biology of HCV infection in the liver and at -5 protein levels and STAT1, -2, -3, -4, and -5 phosphorylation after fixation with the systemic level, which, in turn, may influence the responsive- paraformaldehyde (PFA) and subsequent methanol permeabilization (see ness of immune cells to IFN-α stimulation. The finding that IFN-λ Table S3 for details). Statistical evaluation was done using Student's t test or SNPs also affect spontaneous clearance of the virus (18) suggests the Mann–Whitney rank sum test to compare means, the chi-square test to α that genetic differences can affect disease course from very early compare the distribution of races between IFN- response phenotypes, and Pearson's correlation for correlation studies. phases of HCV infection independent of therapy, hence indirectly α affecting the outcome of IFN- therapy. Gene Expression Arrays. Total RNA was processed by a two-cycle amplification Recently, others reported that lymphocytes and other circulat- procedure as described elsewhere (20) and hybridized to whole-genome ing mononuclear cells from patients with cancer suffer reduced human 36K oligo arrays, representing 25,100 unique human genes of the phosphorylation of STAT proteins in response to IFN-α (19). This Operon Array–Ready Oligo Set version 4.0, printed in house, phenomenon is observable at later stages of cancer progression using oligos purchased from Operon. ArrayBRB's mixed-effects model ANOVA (stage II to IV), suggesting that a sufficient bulk of tumor needs to was used to identify genes significantly affected by race, IFN-α treatment, or be present to induce systemic effects. It is, therefore, possible that interaction between race and treatment. Functional gene network analysis was performed using the Ingenuity pathway analysis system. a chronic inflammatory process, whether induced by chronic viral infection or by cancer, may be responsible for the altered innate Whole-Genome SNP Arrays. Genomic DNA was subjected to array-based SNP immune responses and that AAs may be more susceptible in the analysis, using Affymetrix’s Genome-Wide Human SNP Nsp/Sty Assay Kit 6.0 context of HCV infection. Interestingly, these authors observed per manufacturer's instructions. Data were analyzed using the SNP associa- that alterations in STAT protein phosphorylation affect most tion analysis module of the Partek GS software package. SNP association to PBMC subpopulations and they are predominantly pathway spe- experimental categories such as race, treatment response, etc., was defined χ2 cific rather than cell-type specific. Although we did not address all by performing -tests using the allele association model of Partek GS. PBMC populations and focused on the most commonly inves- fi ACKNOWLEDGMENTS. We thank Thomas R. O'Brien, Division of Cancer Epi- tigated T cell population, it is likely that these ndings could be demiology and Genetics, National Cancer Institute, National Institutes of generalized on the basis of others’ experience. Health, for his useful comments.

1. Reddy KR, et al. (1999) Racial differences in responses to therapy with interferon 10. Rehermann B, Nascimbeni M (2005) Immunology of hepatitis B virus and hepatitis C in chronic hepatitis C. Consensus Interferon. Study Group. Hepatology 30: virus infection. Nat Rev Immunol 5:215–229. 787–793. 11. Donnelly RP, Sheikh F, Kotenko SV, Dickensheets H (2004) The expanded family of class 2. He XS, et al. (2006) Global transcriptional response to interferon is a determinant of II cytokines that share the IL-10 receptor-2 (IL-10R2) chain. J Leukoc Biol 76:314–321. HCV treatment outcome and is modified by race. Hepatology 44:352–359. 12. Chen L, et al. (2009) ISG15, a ubiquitin-like interferon stimulated gene, promotes 3. Layden-Almer JE, Ribeiro RM, Wiley T, Perelson AS, Layden TJ (2003) Viral dynamics hepatitis C virus production in vitro: Implications for chronic infection and response to and response differences in HCV-infected African American and white patients treatment. J Gen Virol, in press. treated with IFN and ribavirin. Hepatology 37:1343–1350. 13. MacQuillan GC, et al. (2002) Intrahepatic MxA and PKR protein expression in chronic 4. Muir AJ, Bornstein JD, Killenberg PG, Atlantic Coast Hepatitis Treatment Group (2004) hepatitis C virus infection. J Med Virol 68:197–205. Peginterferon alfa-2b and ribavirin for the treatment of chronic hepatitis C in blacks 14. Tishkoff SA, et al. (2009) The genetic structure and history of Africans and African and non-Hispanic whites. N Engl J Med 350:2265–2271. Americans. Science 324:1035–1044. 5. Hepburn MJ, Hepburn LM, Cantu NS, Lapeer MG, Lawitz EJ (2004) Differences in 15. Ge D, et al. (2009) Genetic variation in IL28B predicts hepatitis C treatment-induced treatment outcome for hepatitis C among ethnic groups. Am J Med 117:163–168. viral clearance. Nature 461:399–401. 6. Su X, et al.; Virahep-C Study Group (2008) Association of single nucleotide 16. Suppiah V, et al. (2009) IL28B is associated with response to chronic hepatitis C polymorphisms in interferon signaling pathway genes and interferon-stimulated interferon-alpha and ribavirin therapy. Nat Genet 41: 1100–1104. genes with the response to interferon therapy for chronic hepatitis C. J Hepatol 49: 17. Tanaka Y, et al. (2009) Genome-wide association of IL28B with response to pegylated 184–191. interferon-alpha and ribavirin therapy for chronic hepatitis C. Nat Genet 41: 7. Sklan EH, Charuworn P, Pang PS, Glenn JS (2009) Mechanisms of HCV survival in the 1105–1109. host. Nat Rev Gastroenterol Hepatol 6:217–227. 18. Thomas DL, et al. (2009) Genetic variation in IL28B and spontaneous clearance of 8. Platanias LC (2005) Mechanisms of type-I- and type-II-interferon-mediated signalling. hepatitis C virus. Nature 461: 798–801. Nat Rev Immunol 5:375–386. 19. Critchley-Thorne RJ, et al. (2009) Impaired interferon signaling is a common immune 9. Lesinski GB, et al. (2004) Multiparametric flow cytometric analysis of inter-patient defect in human cancer. Proc Natl Acad Sci USA 106:9010–9015. variation in STAT1 phosphorylation following interferon Alfa immunotherapy. J Natl 20. Wang E (2005) RNA amplification for successful gene profiling analysis. J Transl Med Cancer Inst 96:1331–1342. 3:28.

808 | www.pnas.org/cgi/doi/10.1073/pnas.0913491107 Pos et al. Downloaded by guest on September 30, 2021