A Genetic Variant in the BCL2 Gene Associates with Adalimumab Response in Hidradenitis Suppurativa Clinical Trials and Regulates Expression of BCL2

Journal Pre-proof A GENETIC VARIANT IN THE BCL2 GENE ASSOCIATES WITH ADALIMUMAB RESPONSE IN HIDRADENITIS SUPPURATIVA CLINICAL TRIALS AND REGULATES EXPRESSION OF BCL2 Mohan Liu, Jacob Degner, Robert W. Georgantas, Ahmed Nader, Nael M. Mostafa, Henrique D. Teixeira, David A. Williams, Robert S. Kirsner, Anna J. Nichols, Justin Wade Davis, Jeffrey F. Waring PII: S0022-202X(19)33147-1 DOI: https://doi.org/10.1016/j.jid.2019.06.152 Reference: JID 2118 To appear in: The Journal of Investigative Dermatology Received Date: 15 February 2019 Revised Date: 14 June 2019 Accepted Date: 22 June 2019 Please cite this article as: Liu M, Degner J, Georgantas RW, Nader A, Mostafa NM, Teixeira HD, Williams DA, Kirsner RS, Nichols AJ, Davis JW, Waring JF, A GENETIC VARIANT IN THE BCL2 GENE ASSOCIATES WITH ADALIMUMAB RESPONSE IN HIDRADENITIS SUPPURATIVA CLINICAL TRIALS AND REGULATES EXPRESSION OF BCL2, The Journal of Investigative Dermatology (2019), doi: https://doi.org/10.1016/j.jid.2019.06.152. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 The Authors. Published by Elsevier, Inc. on behalf of the Society for Investigative Dermatology. A GENETIC VARIANT IN THE BCL2 GENE ASSOCIATES WITH ADALIMUMAB RESPONSE IN HIDRADENITIS SUPPURATIVA CLINICAL TRIALS AND REGULATES EXPRESSION OF BCL2 Mohan Liu1,5,*, Jacob Degner1,5, Robert W. Georgantas1, Ahmed Nader2, Nael M. Mostafa2, Henrique D. Teixeira3, David A. Williams3, Robert S. Kirsner4, Anna J. Nichols4, Justin Wade Davis1, and Jeffrey F. Waring1 1 Pharmacogenetics, Human Genetics, and Computational Genomics, Genomics Research Center, AbbVie Inc., North Chicago, IL, United States of America 2 Clinical Pharmacology and Pharmacometrics, AbbVie Inc., North Chicago, IL, United States of America 3 Global Pharmaceutical Research and Development, AbbVie Inc., North Chicago, IL, United States of America 4 Department of Dermatology & Cutaneous Surgery, University of Miami Miller School of Medicine, FL, United States of America 5 These authors contributed equally to this work * Corresponding Author: mohan.liu@abbvie.com 1 Abstract Hidradenitis Suppurativa (HS) is a chronic skin disease with strong genetic component and prevalance from 0.5% to 4%. Adalimumab is the only treatment approved by either the European Medicines Agency (EMA) or the US Food or Drug Administration (FDA) for the management of moderate-to-severe HS. To identify genetic variants associated with adalimumab response, we performed a genome-wide association study (GWAS) from the largest two Phase 3 HS clinical trials (PIONEER I and II) to date. Through direct genotyping and imputation, we tested almost 7 million genetic variants with MAF > 5% and identified one single linkage disequilibrium (LD) block, located in the intron of the BCL2 gene, which reached genome-wide significance (lead single-nucleotide polymorphism [SNP] rs 59532114; p=2.35E-08). Bioinformatic analysis and functional genomics experiments suggested correlation of the most strongly associated SNP minor allele with increased BCL2 gene and protein expression in hair follicle tissues. In reciprocal knockdown experiments, we found that BCL2 is down-regulated by TNF inhibition. These results highlight to our knowledge previously unreported pathway that involves BCL2 in response to adalimumab. Further work is required to determine how this pathway influences adalimumab effectiveness in patients suffering with HS. Introduction Hidradenitis suppurativa (HS, acne inversa) is a chronic skin disease characterized by painful, deep seated, and recurrent inflammed subcutaneous nodules. Follicular occlusion is the primary event resulting in comedones, painful nodules, abscesses and recurrent draining sinuses in the apocrine gland-bearing skin ultimately culminating in disfiguring scars (Saunte DML and Jemec 2 GBE 2017; De Vita and McGonagle 2018). The prevalence for HS is 0.5 to 4% but may be underestimated (Garg et al. 2017; Revuz et al. 2018). Acquired and constitutive risk factors associated with HS include smoking, obesity, infection and family history. Familial aggregation of HS suggests genetic factors play an important role in disease pathogenesis. For example, up to 42% of patients report a family history of HS and in some families HS can follow an autosomal dominant inheritance mode (Fitzsimmons and Guilbert 1985; Von der Werth et al. 2000). Two large multi-generation pedigrees found 100% penetrance as an autosomal dominant trait (Saleh Al-Ali et al. 2010). For the majority of patients, HS appears to be a complex trait with limited knowledge about the pathogenesis. Adalimumab is the only approved treatment for the management of moderate-to-severe HS by either the European Medicines Agency (EMA) or the US Food or Drug Administration (FDA), with an overall response rate approaching 60% (42% and 59%, PIONEER I and PIONEER II respectively (Kimball et al. 2016a)). In addition to adalimumab, common therapies for HS include surgical excision of affected sites, topical antibacterial agents, antiseptic cleansers, intralesional corticosteroids, systemic antibiotics, systemic corticosteroids or hormonal therapies, which result in varied and often only temporary improvement (Saunte DML and Jemec GBE 2017). Given the limited knowledge of the pathways involved in HS etiology or the differential HS patient response to anti-TNF treatment, we performed an unbiased genome-wide association study (GWAS) to identify variants associated with adalimumab response. This analysis resulted in a single genome-wide significant hit, which was followed up with functional experiments to determine which gene(s) in the local LD block, if any, is directly affected by the genetic variant. These functional experiments were carried out in HS relevant cell types to further illuminate which tissues/cell types are potentially relevant to the mechanism through which this variant influences adalimumab response in HS patients. 3 Results Adalimumab response genome-wide association study of HS To increase the statistical power, all participants of both studies who received adalimumab either in the initial treatment Period 1 or in Period 2 or 3 were combined (Figure 1). A GWAS against log2 fold change of abscess and inflammatory nodule count was performed, yielding a singlepeak with a significant association. Five SNPs on Chromosome 18 within this genome-wide analysis significantly associated with adalimumab response and had an effect size estimate of ~1.07 (±0.37) units of log2 AN.COUNTFC per minor allele (Table 1). All five SNPs were located in the second intron of the BCL2 gene. These variant SNP genotypes have MAFs of 0.04 in the Caucasian population in the 1000 genomes database. However, they are relatively common in African American (MAF=.27) and Asian (MAF= .12) populations. Replicating the association in other sample sources Because the two clinical trials (PIONEER I and II) are the only HS trials with adalimumab treatment for which we had access, we pursued replication in two other sources of data, although they are not all completely independent. In the first data source (Data Source 1), we tested additional time points (week 4 and week 8) that were not used in the original GWAS (GWAS identified association with response at week 12 which was the primary outcome time-point (Figure 2C). As in the week 12 data, in both week 4 and week 8, there were significant associations between the SNP allele status and the fold change of AN.COUNTFC (P = .0003 and P=.037, respectively) (Figure 2C left panel). In a second data source (Data Source 2), we compared the response of the same individuals, but in the open-label extension trial that followed the initial trials after a 12 week washout period 4 (Figure 1). All the participants enrolled in period 3 experienced a 12 week washout period, following which they were retreated with adalimumab 40 mg EW for up to 48 weeks. Within period 3, we tested the top GWAS variant for association with response in each of the five time points after returning to adalimumab dosing. We found a significant association between rs59532114 genotype and the response (p-values from .00017 at week 12 to P=.038 at week 72). However, in this post-washout time course, there was no significant association in weeks 4 and 8 (P=.14 and P=.32 respectively) (Figure 2C, right panel). All measured time-points in the postwashout time course had effect size estimates in the same direction as the original GWAS. rs59532114 association with change in AN count is drug dependent To ask if the effect of this SNP was drug dependent, we compared individuals treated with adalimumab to those who never participated in a treatment arm. As shown in the box plots in Figure 3, within the adalimumab treated group, the participants who carried the minor allele had inadequate response compared to other participants. In contrast to the general pattern of decreasing AN count after 12 weeks of adalimumab treatment, the average fold change of AN count in the minor allele group was greater than 1, indictating participants in this subgroup have increased AN count compared to their baseline (Figure 3, left box). In contrast, within the placebo group, there was no difference between reference allele carriers and minor allele carriers (P= 0.64) (Figure 3, right box), suggesting the association is dependent on adalimumab treatment. Analysis between the SNP rs59532114 genotype with Hurley stages didn’t show any association (Supplementary Figure S3). Expression Quantitative Trait Loci (eQTL) Analysis Since the significantly-associated variants are all within the second intron of BCL2, we sought to test whether they impacted this gene or other neighborhood genes using analysis of expression 5 quantitative trait loci (eQTL). We evaluated the top two equally linked SNPs, rs59532114 and rs67645778 (LD R2=1) for their association with expression levels of genes (BCL2, PHLPP1, KDSR, VPS4B, and SERPINB5) within a 1 Mb region using published eQTL data (Figure 2B), such as the Genotype-Tissue Expression (GTEx) project13. However, with allele frequency of 0.04 in the Caucasian population, there were few minor allele genotype samples in the GTEx database, so we expanded our analysis to another public dataset (GSE24277), which includes 300 LCLs (Lymphoblastoid Cell Lines) from 3 ethnic groups. We found that rs59532114 was an eQTL for BCL2 in African-Americans and Han Chinese-Americans participants (Supplementary Figure 1). We also identified a recently published whole-genome sequencing study of LCLs from UK10K (Rare genetic variants in health and disease project) participants which covered rare variants and more LCLs compared to GTEX (N=506). In this study, the SNP rs67645778, which is in equal LD with rs59532114, was in eQTL with BCL2 with p-value = 9.45E-07(Brown et al. 2017). Finally, to confirm that rs59532114 is an eQTL for BCL2, we designed a more balanced study with respect to genotype. We selected 72 LCLs from Caucasians in the 1000 Genomes Project human variation repository panel with known genotypes at rs59532114 that were either homozygous for major allele participants (n=18), heterozygous (n=34), or homozygous for the minor allele (n=18). We cultured these LCLs for 72 hours for mRNA extraction to test the SNP correlation to expression of BCL2 measured by quantitative PCR. We found higher BCL2 mRNA expression in LCLs homozygous for the minor allele genotype as compared with LCLs homozygous for the major genotype (p=5.75E-04; Figure 4A). Other genes in the region, including PHLPP1, KDSR, VPS4B, and SERPINB5, were not significantly associated with allele status (data not shown). SNP rs59532114 minor allele associated with increased baseline BCL2 protein level in LCLs 6 In the LCLs, the SNP rs59532114 minor allele was associated with increased BCL2 mRNA expression, but only a fraction of eQTL translate to functional protein differences(Battle et al. 2015). We tested if the minor allele also associated with BCL2 protein levels. We randomly selected four LCLs homozygous for the major allele and four LCLs homozygous for the minor allele from the same LCLs that were used for previous eQTL analysis. Cytoplasmic proteins were extracted for western blot analysis. As shown in the gel image, all the homozygous major allele LCLs have less BCL2 protein compared to the LCLs homozygous for the minor allele (28 KDa band; Figure 4B and C). As in the eQTL analysis, we didn’t find expression correlation with the protein levels of the neighborhood proteins (Figure 2B), such as PHLPP1, VPS4B, KDSR and SERPINB5 (Figure 4D). rs59532114 minor allele associated with increased BCL2 transcriptional activity in reporter assays To evaluate the effect of SNP rs59532114 on BCL2 transcriptional activity in HS diseaserelevant tissues, we created luciferase reporter constructs containing 250-bp of DNA surrounding this top SNP. We cloned the full-length of the 2.8 kb BCL2 promoter downstream of the SNP region (Figure 5 A and B) and performed the reporter assay in three different potentially relevent tissues for HS. To compare expression between the major and minor allele, we generated two reporter constructs with only one single nucleotide difference with the SNP genotype in the middle of the 250 bp region. One construct contained the major allele of rs59532114 (C) while the other contained the minor allele (A). Compared to the major allele, the minor allele produced a 5-fold increase in luciferase expression (p < 0.001) in the human primary outer root sheath cells, suggesting that the rs59532114 minor allele results in increased transcriptional activity of the BCL2 promoter in these cells. In contrast, transfection into 293T and HEK001 cells resulted 7 in a marginal, non-significant increase in transcriptional activity by t-test, indicating that the increase in transcriptional activity may be cell-type specific (Figure 5C). TNF-α inhibition affects BCL2 expression through TNF signaling pathway We next examined possible mechanisms by which the adalimumab target, TNF, might influence BCL2 expression in HS disease-relevant tissues. Single target knockdown of both TNF and BCL2 followed by genome-wide expression microarray analysis were performed. The same experiments were performed in both human primary outer root sheath cells and human primary keratinocytes. In the primary outer root sheath cells, we first confirmed that knockdown of TNFα resulted in significantly decreased BCL2 expression (p-value = 1.71E-04), at the mRNA level (Figure 7 left panel). In contrast, knockdown of BCL2 did not have a strong effect on TNFα expression ((p-value = 0.13) (Figure 5D, right panel), which suggests that adalimumab may down-regulate BCL2 expression partially through its effect on TNFα, and suggests that BCL2 is downstream of TNF in a regulatory pathway in human primary outer root sheath cells. Discussion We have completed the largest pharmacogenomic GWAS of participants (N=445 consented) with hidradenitis suppurativa (HS), investigating genetic variants associated with variation in response to adalimumab. We identified a single locus in BCL2 which suggested a potential role of apoptosis homeostasis in the pathophysiology of adalimumab response, in the context of HS disease. As the mechanistic link between associated SNPs and the response phenotype can often be obscure and require experimental validation, we performed several functional genomics experiments to identify a mechanism implied by our GWAS association. 8 The statistically most strongly associated variant, (rs5932114) was shown to be associated with increased expression and protein levels of BCL2 in LCLs. In addition, we demonstrated, using reporter assays, that the minor allele is also associated with increased BCL2 transcription in primary human hair outer root sheath cells, but not in other cell types. The knockdown of TNF and BCL2 separately showed BCL2 belongs to the same pathway and is down-regulated by TNF. BCL2 acts as an anti-apoptotic regulatory protein that blocks cell death and plays an important role in regulating skin homeostasis in the outer root sheath cells of the hair follicles(Abe and Tanaka 2017). It has been reported in psoriatic patients based on lesional skin immunohistochemical staining, that another TNFα inhibitor (infliximab) also regulated BCL2 protein through TNF, and that TNF was upstream of BCL2 (Kokolakis et al. 2012). Moreover, BCL2 has recently been shown to be associated with skin-related excessive hairiness (hirsutism) among Japanese(Endo et al. 2018). Interestingly, genetic mapping identified BCL2 as one of the regions controlling eccrine gland and hair follicle traits response for sweat gland density in mice(Kamberov et al. 2015). As in many other GWAS with pharmacogenomics drug response, our study has several limitations(Park et al. 2014). Our sample size is modest; however, our samples were selected from the largest 2 HS trials that have been performed (Figure 3). Critically, we deployed functional experiments to evaluate molecular mechanisms that could account for how these genetic variations impact BCL2 gene. Our results provide multiple lines of supportive evidence for a plausible mechanism to support this statistical association. Together with these findings, our results support the hypothesis that apoptotic balance may be an important component of treatment response in HS and possibly other immune mediated skin diseases. More specifically, this study provides evidence of a genetic link to adalimumab response and provides an indication that the key anti-apoptotic factor, BCL2, may be repressed in HS relevant tissues by anti-TNF 9 agents (Supplementary Figure S4). These results broaden our understanding of underlying disease mechanisms in HS, and may lead to the pursuit of additional mechanisms relevant for the treatment of HS either alone or in combination with anti-TNF agents. Material and Methods Study design This study is based on two placebo-controlled randomized clinical trials, the Pioneer I (M11-313) and Pioneer II (M11-810) Phase 3 clinical trials (NCT01468207 and NCT01468233, respectively) (Kimball et al. 2016b). Following written informed consent and Ethics Committee approvals, 445 trial participants consented DNA samples (199 from PIONEER I and 246 from PIONEER II) and were used for genotyping. In both trials, the primary efficacy endpoint was the percentage of participants achieving Hidradenitis Suppurative Clinical Response (HiSCR) at week 12. HiSCR is a binary outcome that is achieved when there is at least a 50% reduction in abscess and inflammatory nodule count (AN.COUNT) with no increase in abscess count and no increase in draining fistula compared to baseline (Kimball et al. 2016c). PIONEER I, achived a significant difference in response rates at week 12 with 41.8% of participants on adalimumab achieving HiSCR compared to 29.8% of placebo. Similarly in PIONEER II, achived a significant difference in response rates was observed (58.9% for treated vs 36.8% for placebo.) Combining participants from four separate study arms Both PIONEER I and II consisted of three periods where some participants that initially received placebo were switched to adalimumab at later periods (Figure 1A). In period 1 (24 weeks) of both trials, participants were randomized in a 1:1 ratio to receive either 40mg adalimumab EW 10 or placebo. In period 2 of PIONEER I, participants who had been on placebo during period 1 were switched to adalimumab 40mg EW. In PIONEER II, participants on placebo in Period 1 continued on placebo until week 36 at which time (period 3) they were switched to adalimumab 40mg EW in an open-label extension study. This study design resulted in four different groups that effectively received an initial 12-week adalimumab 40 mg EW dosing and efficacy determination. We combined these four treatment-arms into a single GWAS analysis to maximize statistical power. Response definition HiSCR was used as a measure of response in the clinical trials, which as a classifier has reduced statistical power of the GWAS analysis. To increase statistical power, the main quantitative component of HiSCR - Fold Change (FC) of total number of Abscesses and inflammatoryNodule (AN) Count (AN. COUNTFC) was defined and calculated as AN.COUNTWeek12 / AN.COUNTBaseline, where baseline was the first week of dosing. The data was normalized using a Log 2 transformation and this transformed metric was truncated at 2 to reduce the impact of several extreme outliers (with Log2FC > 2). However, we found that untransformed fold changes did not meet the assumptions of normal residuals in our statistical tests. Linear regression analysis was used to test a null hypothesis of no association between AN. COUNTFC and variant with minor allele frequency (MAF) ≥ 5% in our sample (implemented in PLINK linear regression feature). Dosage-based genotypes were used as opposed to allelic and took values ranging from 0-2 (PLINK dosage feature). Observed minimum serum adalimumab concentration (Ctrough) was used as a covariate. Only individuals who had nonmissing phenotype, covariate values, and who actually received adalimumab treatment in one of the periods were used in the analysis. 11 GWAS genotyping, quality control and imputation Genotyping was performed using Illumina HumanOmniExpressExome v1.2b Bead-Chip containing 964,193 SNPs. Initial genotype calling was performed and analyzed on all batches and samples using Illumina‘s Genome Studio software and were exported as PLINK format (Oros et al. 2013) for bioinformatic analysis. PyGenClean (Lemieux Perreault et al. 2013) was used for performing initial quality control (QC). At the sample QC level, of the original 445 individuals samples, 57 individuals were excluded as genetic ancestry outliers. At the SNP (marker) QC level, of 964,193 initial directly genotyped SNPs, 53,161 SNPs were removed due to quality control issues such as high missing genotype rate or deviations from Hardy-Weinberg Equilibrium (HWE) (P < 1 × 10-4 ). With the QC filtered genotypes, Beagle v. 4.1 (Browning and Browning 2016) was used to imput genotypes for 31.6 million variants, or all variants called in 1K Genomes Project Phase 3. These were further filtered for MAF > 0.05 to 6,949,858 variants for testing. Beagle was run on each chromosome separately to impute both missing genotypes at variants on the SNP chip and variants not typed on the chip. Finally, estimated minor allele dosages output from Beagle were converted to PLINK dosage format for GWAS analysis. Validation in different sample sources Since no conventional replication cohort exists to retest significant findings, we turned to several different data sources for confirmatory evidence of our association results. First, we compared AN.COUNTFC at two other time points (Week 4 and Week 8 measurements) in the same participants that were used in the original GWAS analysis. Second, we used data from a subset of participants who, after 12 weeks of initial treatment, went through a 24 week washout period, followed by a return to adalimumab treatment and measurements of AN.COUNT at 4, 8, 12, 18, 12 24, 36, and 48 weeks after return to adalimumab treatment – i.e. so called switchover placebo (PBO) participants. In this second treatment period, we measured the decay of AN.COUNTFC and tested for significant GWAS associations. Expression quantitative trait loci (eQTL) analysis eQTL analysis focused on top variant rs59532114 and genes within ±500 kb (each 1-Mb wide). LCLs of defined genotype at rs59532114 were bought from Coriell Institute NIGMS Human Genetic Cell Repository (Coriell Institute, Camden, NJ), then plated on non-adherent 12-well plates at a density of 1 × 105 cells per well, without antibiotics. RNA was extracted after 48 hrs of culture using RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Reverse transcription was performed using High-Capacity RNA-to-cDNA kit (ThermoFisher Scientific) according to the manufacturer’s instructions. Quantitative PCR (qPCR) was performed using Taqman Advance Master Mix with pre-designed Taqman probes (Thermo Fisher Scientific) for genes, include BCL2, PHLPP1, KDSR, VPS4B and SERPINB5, and compared against the control gene INST4 (cycle mean =22.97, SD=1.01). Relative quantification over the control gene was calculated for the value of 2- CT to get the expression levles. The statistical significance between means was tested using a two-tailed Student’s t test, with equal variances assumed. Each experiment was conducted in triplicates. A p-value of less than 0.05 was considered significant. Western blot analysis Four LCLs of each homozygous genotype at rs59532114 were grown in 25 mL of RPMI with 10% FBS for 72 hours. Protein was extracted and concentration was measured by BCA assay (ThermoScientific), followed by transfer to PVDF according to the manufacturer's instructions (LI-COR Biosciences). Primary antibodies against BCL2 (Novus Biologicals), VPS4B (Novus Biologicals), PHLPP1 (Abcam), KDSR (Abcam), SERPINB5 (Abcam), and ACTB 13 (Abcam) were tested since they are all located within 1 Mb of the response associated LD block. The protein bands were detected using the Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE, USA). IRDye 800CW and IRDye 680 (LI-COR Biosciences, Lincoln, NE, USA) served as secondary antibodies. All western blot band quantification procedures were performed using the Image Studio Lite 3.1 software (LI-COR Biosciences). Plasmid construction and reporter assays Promoter-driven luciferase reporter constructs were generated by insertion of gBlocks (Integrated DNA Technologies), containing the BCL2 2.84 Kb promoter, into the KpnI/MluI sites of pGL3-Basic (Promega). For the SNP rs59532114 locus, a 250 bp DNA with the SNP major allele C and the minor allele A in the middle were synthesized and cloned into the BamHI/SalI sites of BCL2-promoter construct. Human primary outer root sheath cells, HEK-293T or human primary keratinocyte cells were transfected with 500 ng luciferase reporter plasmids and 50 ng of pRL-TK transfection control plasmid with Lipofectamine 2000 (Thermo Fisher Scientific). The total amount of transfected DNA was kept constant at 500 ng for each construct. Luciferase activity was measured 48 h after transfection assayed with a dual-luciferase system and normalized to renilla activity (Promega, Madison, WI). Data Availability Statement Datasets related to this article can be found at http://shinycg.abbvienet.com/degner/private/HUMIRA_HS_GWAS_M11-313_M11810/results/IndividualGWASReports_Dosage_G+Trough/tier3_quantitative.phenotypes.txt.WEE K.12/summarize.plink.singleGWAS.html, hosted at [AbbVie, Inc.] Conflict of Interests 14 The studies were sponsored by AbbVie. AbbVie contributed to the study design, research, and interpretation of data, writing, reviewing, and approving the publication. Several authors are AbbVie employees and may hold AbbVie stocks or options. ML, JD, RG, AN, NM, HT, DW, JD and JW are employees of AbbVie. The design, study conduct, and financial support for this research were provided by AbbVie. AbbVie participated in the interpretation of data, review, and approval of the publication. Drs. Anna J. Nichols and Robert S. Kirsner have served as consultant for AbbVie. They were not funded by AbbVie. Acknowledgments We thank the participants for their participation and desire to provide their DNA for genetic research in the M11-313 and M11-810 studies. The authors wish to thank Jozef Lazar, M. D. Ph.D. of AbbVie for reviewing the manuscript. CRediT statement Conceptualization: Mohan Liu, Jacob Degner, Nael M. Mostafa, Henrique D. Teixeira, David A. Williams, Robert S. Kirsner, Anna J. Nichols, Jeffrey F. Waring; Data curation: Mohan Liu, Jacob Degner, Robert W. Georgantas, Ahmed Nader, Nael M. Mostafa, Henrique D. Teixeira, David A. Williams, Robert S. Kirsner, Anna J. Nichols, Justin Wade Davis, Jeffrey F. Waring; Formal Analysis: Jacob Degner, Justin Wade Davis; Funding Acquisition: Mohan Liu, Jeffrey F. Waring; Ivestigation: Mohan Liu, Jacob Degner, Robert W. Georgantas, Ahmed Nader, Nael M. Mostafa, Henrique D. Teixeira, David A. Williams, Jeffrey F. Waring; 15 References Abe Y, Tanaka N. Roles of the Hedgehog Signaling Pathway in Epidermal and Hair Follicle Development, Homeostasis, and Cancer. J. Dev. Biol. 2017;5(4):12 Available from: http://www.mdpi.com/2221-3759/5/4/12 Battle A, Khan Z, Wang SH, Mitrano A, Ford MJ, Pritchard JK, et al. Genomic variation. Impact of regulatory variation from RNA to protein. Science (80-. ). 2015;347(6222):664–7 Available from: http://www.sciencemag.org/content/347/6222/664.full%5Cnpapers3://publication/doi/10.1126/sci ence.1260793 Brown AA, Viñuela A, Delaneau O, Spector TD, Small KS, Dermitzakis ET. Predicting causal variants affecting expression by using whole-genome sequencing and RNA-seq from multiple human tissues. Nat. Genet. 2017;49(12):1747–51 Browning BL, Browning SR. Genotype Imputation with Millions of Reference Samples. Am. J. Hum. Genet. 2016;98(1):116–26 Endo C, Johnson TA, Morino R, Nakazono K, Kamitsuji S, Akita M, et al. Genome-wide association study in Japanese females identifies fifteen novel skin-related trait associations. Sci. Rep. 2018. Fitzsimmons JS, Guilbert PR. A family study of hidradenitis suppurativa. J. Med. Genet. 1985;22(5):367–73 Available from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1049480&tool=pmcentrez&renderty pe=abstract Garg A, Lavian J, Lin G, Strunk A, Alloo A. Incidence of hidradenitis suppurativa in the United 16 States: A sex- and age-adjusted population analysis. J. Am. Acad. Dermatol. 2017;77(1):118–22 Kamberov YG, Karlsson EK, Kamberova GL, Lieberman DE, Sabeti PC, Morgan BA, et al. A genetic basis of variation in eccrine sweat gland and hair follicle density. Proc. Natl. Acad. Sci. 2015;112(32):9932–7 Available from: http://www.pnas.org/lookup/doi/10.1073/pnas.1511680112 Kimball AB, Okun MM, Williams DA, Gottlieb AB, Papp KA, Zouboulis CC, et al. Two phase 3 trials of adalimumab for hidradenitis suppurativa. N. Engl. J. Med. 2016a;375(5):422–34 Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0- 84982161469&doi=10.1056%2FNEJMoa1504370&partnerID=40&md5=2eb973444117f095172 3613492ad603c Kimball AB, Okun MM, Williams DA, Gottlieb AB, Papp KA, Zouboulis CC, et al. Two Phase 3 Trials of Adalimumab for Hidradenitis Suppurativa. N. Engl. J. Med. 2016b;375(5):422–34 Available from: http://www.nejm.org/doi/10.1056/NEJMoa1504370 Kimball AB, Sobell JM, Zouboulis CC, Gu Y, Williams DA, Sundaram M, et al. HiSCR (Hidradenitis Suppurativa Clinical Response): a novel clinical endpoint to evaluate therapeutic outcomes in patients with hidradenitis suppurativa from the placebo-controlled portion of a phase 2 adalimumab study. J. Eur. Acad. Dermatology Venereol. 2016c;30(6):989–94 Kokolakis G, Giannikaki E, Stathopoulos E, Avramidis G, Tosca AD, Krüger-Krasagakis S. Infliximab restores the balance between pro- and anti-apoptotic proteins in regressing psoriatic lesions. Br. J. Dermatol. 2012;166(3):491–7 Lemieux Perreault LP, Provost S, Legault MA, Barhdadi A, Dubé MP. pyGenClean: Efficient tool for genetic data clean up before association testing. Bioinformatics. 2013. p. 1704–5 17 Oros KK, Arcand SL, Bayani J, Squire JA, Mes-Masson AM, Tonin PN, et al. Analysis of genomic abnormalities in tumors: A review of available methods for Illumina two-color SNP genotyping and evaluation of performance. Cancer Genet. 2013;206(4):103–15 Park HW, Dahlin A, Tse S, Duan QL, Schuemann B, Martinez FD, et al. Genetic predictors associated with improvement of asthma symptoms in response to inhaled corticosteroids. J. Allergy Clin. Immunol. 2014;133(3) Revuz JE, Canoui-Poitrine F, Wolkenstein P, Viallette C, Gabison G, Pouget F, et al. Prevalence and factors associated with hidradenitis suppurativa: Results from two case-control studies. J. Am. Acad. Dermatol. 2018;59(4):596–601 Available from: http://dx.doi.org/10.1016/j.jaad.2008.06.020 Saleh Al-Ali FM, Ratnamala U, Mehta TY, Naveed M, Al-Ali MT, Al-Khaja N, et al. Hidradenitis suppurativa (or Acne inversa) with autosomal dominant inheritance is not linked to chromosome 1p21.1-1q25.3 region. Exp. Dermatol. 2010. p. 851–3 Saunte DML and Jemec GBE. Hidradenitis suppurativa: Advances in diagnosis and treatment. JAMA - J. Am. Med. Assoc. 2017;318(20):2019–32 Available from: https://jamanetwork.com/journals/jama/articlepdf/2664466/jama_saunte_2017_rv_170008.pdf%0 Ahttp://ovidsp.ovid.com/ovidweb.cgi?T=JS&PAGE=reference&D=emexb&NEWS=N&AN=619 564432 De Vita V, McGonagle D. Hidradenitis suppurativa as an autoinflammatory keratinization disease. J. Allergy Clin. Immunol. 2018; Von der Werth JM, Williams HC, Raeburn JA. The clinical genetics of hidradenitis suppurativa revisited. Br. J. Dermatol. 2000;142(5):947–53 18 Table 1. SNPs in the chr18 loci associated with HS to adalimumab response in Europeanancestry Chr rsID a b Genes Allele chr18 rs59532114 60830032 BCL2 C/A 0.04 2.35E-08 chr18 rs67645778 60836331 BCL2 C/T 0.04 3.65E-08 chr18 rs11877475 60831189 BCL2 T/C 0.04 4.26E-08 chr18 rs11877911 60841744 BCL2 C/G 0.04 4.3E-08 chr18 rs11152370 60835509 BCL2 G/C 0.04 4.62E-08 chr18 rs17070798 60840128 BCL2 A/G 0.04 5.43E-08 chr18 rs60405107 60405107 BCL2 C/G 0.04 1.51E-07 chr18 rs141057516 60836654 BCL2 C/T 0.01 2.38E-07 chr18 rs145099651 60836658 BCL2 C/T 0.02 2.38E-07 Position EAF P value EAF, effect allele frequency in Caucasian; OR, odds (log-additive) ratio; 95% CI, 95% confidence interval. a b SNP position, build 37. Major/Minor allele 19 Figure Legends: Figure 1. GWAS study design for HS response to adalimumab. a. To maximize the statistical power, subjects in red boxes were included. AN.COUNT: abscess and inflammatory nodule count, PBO: placebo, EW: every week. Ancestry outliers: subjects from European cluster with > 1.9 times the estimated standard deviation of Utah residents with Northern and Western European ancestry from the CEPH (CEU) collection reference cluster in the 1000 Genomes Reference Panel. b. Overall decay of Log2 AN.COUNTFC. Red line represents Log2 AN.COUNT of fold change after treatment (4, 8, and 12 weeks) compared to baseline (0 week). Change above the red line, represents symptoms worsening, while below the redline, represents symptoms improved. Each circle represents one patient. c. Distribution of AN.COUNTFC at week 12. On average with each treatment time point, the percentage of AN.COUNT decreases to 50% of the original level. All the y-xiles of AN.COUNT fold changes was Log2 transformed. Figure 2. Overview of adalimumab response GWAS results in HS. a. Manhattan plot for a null hypothesis of no additive association to genotype was tested for each variant and –log10 p-value is given. b. Zoomed plot surrounding the genome-wide significant SNPs on chromosome 18 in the BCL2 gene. Top SNP rs59532114 in purple, and LD SNPs with r2 > 0.8 – red; r2 > 0.6 – orange; r2 > 0.4 – green. c. Additional data sources support the association analysis. Data source 1 with are subjects at different time points (week 4, 8 and 12) (P= .0003, .037, and <5E-08 respectively). Data source 2 20 are subjects but in 24 weeks wash-out and return to adalimumab treatment and significantly difference in week 12 (P= .00017). Light Blue box: patients homozygous for the major allele (CC); Dark Blue box: patients heterozygous of minor allele (CA). Figure 3. Box plot shows SNP the minor allele at rs59532114 associated with increased AN.COUNTFC and less response at week 12 in HS trials in adalimumab treated participants (P< 0.001). Each dot represents the AN.COUNTFC of a participant. In the Placebo group, no significant difference between the major and minor allele of rs59532114 (P=0.64). The horizontal lines within each box represents the lower and upper borders of each box with 25th and 75th percentiles, respectively, and the whiskers mark the 95% confidence intervals. WT: patients homozygous for the major allele (genotype CC); W/V: patients heterozygous for the miner allele (genotype CA). Figure 4. The minor allele of rs59532114 genotype-carried LCLs associates with increased BCL2 mRNA and protein levels. a. QRT-PCR of BCL2 expression based on different genotype classes of LCLs(CC n=18; CA n=36; AA n=18). The homozygote genotype of minor allele (AA) genotype was associated with increased mRNA expression of BCL2. b. Western Blot and c. quantification of BCL2 protein (28 kDa) was more highly expressed in the minor allele homozygote LCLs compared to major allele homozygote LCLs. ACTB (42 kDa) was used as a control (upper). d. No differences between minor allele homozygote LCLs to major allele homozygote LCLs for other proteins (BCL2, ACTB, PHLPP1, KDSR, VPS4B and SERPINB5)locate with 1 Mb around the SNP. 21 Figure 5: SNP rs59532114 allele affects BCL2 transcriptional activity in primary human outer root sheath cells. a. A 2.84 kb DNA region of the BCL2 promoter was cloned in dual luciferase constructs. b. rs59532114 major or minor allele resides in the middle of the 250 bp DNA sequence were tested. Firefly luciferase values were normalized to the co-transfected Renilla luciferase to correct transfection efficiency. c. As compared to the “WT" major allele (C), the minor allele (A) gave a significant increase in BCL2 transcriptional activity in human primary outer root sheath, but not in 293T and HEK001 cells. The error bar represents standard deviation among the replicates. d. TNF-alpha up-regulates BCL2 expression in human primary outer root sheath cells. Microarray analysis was conducted after transi-transfection. 22 Supplemental Figure Titles and Legends: Figure S1. Principal component analysis (PCA) plots and sample familial relatedness plots for GWAS sample quality control. PCA plots for GWAS samples together with Hapmap reference samples. Ancestry outliers are colored black and highlighted with red circles were excluded (b) IBS sample relatedness plots of the indicating familial relationships discovered among study participants. For first degree relatives, one family member was chosen at random to exclude from the GWAS. For twins/sample duplicates, clinical data such as age and sex were used to determine these were sample duplicates and both samples were excluded from analysis. Figure S2. The minor allele of rs59532114 genotype-carried LCLs associates with increased BCL2 mRNA expression in African American and Han Chinese American LCLs data in published GEO dataset (GSE24277). Affymetrix microarray U133 results of BCL2 expression based on different genotype classes of LCLs were extracted from 100 African American LCLs (left) and 100 Han Chinese American LCLs (right). The heterozygous genotype of minor allele (Het WV) genotype was associated with increased mRNA expression of BCL2 from published GEO data (GSE24277). Figure S3. Plot of association between rs59532114 genotype carriers with baseline patients’ Hurley stage. There was no significant association to baseline disease severity as measured by AN count (P=0.69). We additionally checked for an association between genotype and Hurley stage but found that there was an equal split between Hurley stage II and Hurley stage III subjects in the two genotype classes (CC and CA) of rs59532114 (55% Hurley stage II in homozygote major 1 allele genotypes vs 57% Hurley Stage II in heterozygotes; P=1). Figure S4. In Hidradenitis Suppurativa patients with BCL-2 in the TNF pathways of pathogenesis (left panel). Under adalimumab treatement, the homozygouse of major allele (CC) rs59532114 carriers associate with decreased BCL2 expression and decreased anti-apoptosis responses to adalimumab (middle panel). However, the minor allele (A) of rs59532114 carriers associated with increased BCL-2 expression and with strong anti-apoptosis with lack of response to adalimumab treatment (right panel). 2