HLA-DQA1*05 Carriage Associated With Development of Anti-Drug Antibodies to Infliximab and Adalimumab in Patients With Crohn’s Disease

Journal Pre-proof HLA-DQA1*05 Carriage Associated With Development of Anti-Drug Antibodies to Infliximab and Adalimumab in Patients With Crohn’s Disease Aleksejs Sazonovs, BSc, Nicholas A. Kennedy, MBBS, Loukas Moutsianas, DPhil, Graham A. Heap, PhD, Daniel L. Rice, MSc, Mark Reppell, PhD, Claire M. Bewshea, MSc, Neil Chanchlani, MBChB, Gareth J. Walker, MBBS, Mandy H. Perry, PhD, Timothy J. McDonald, PhD, Charlie W. Lees, PhD, J.R. Fraser Cummings, DPhil, Miles Parkes, DM, John C. Mansfield, MD, Jeffrey C. Barrett, DPhil, Dermot McGovern, DPhil, James R. Goodhand, MBBS, Carl A. Anderson, PhD, Tariq Ahmad, DPhil, PANTS consortium PII: DOI: Reference:

S0016-5085(19)41414-5 https://doi.org/10.1053/j.gastro.2019.09.041 YGAST 62934

To appear in: Gastroenterology Accepted Date: 29 September 2019 Please cite this article as: Sazonovs A, Kennedy NA, Moutsianas L, Heap GA, Rice DL, Reppell M, Bewshea CM, Chanchlani N, Walker GJ, Perry MH, McDonald TJ, Lees CW, Cummings JRF, Parkes M, Mansfield JC, Barrett JC, McGovern D, Goodhand JR, Anderson CA, Ahmad T, PANTS consortium, HLA-DQA1*05 Carriage Associated With Development of Anti-Drug Antibodies to Infliximab and Adalimumab in Patients With Crohn’s Disease, Gastroenterology (2019), doi: https://doi.org/10.1053/ j.gastro.2019.09.041. 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 by the AGA Institute

HLA-DQA1*05 is associated with immunogenicity to anti-TNF therapy in biologic naive patients with Crohn’s disease % with no anti-drug antibodies

100%

Infliximab

80% 60% 40% 20% 0%

0

100

200

300

400

500

600

700

Days

Position on chromosome 6 (mb)

1240 patients with Crohn’s disease treated with infliximab or adalimumab ± immunomodulator

Longitudinal measurement of anti-drug antibodies

Genome-wide association study identifies DQA1*05; hazard ratio 1.90 (95%CI 1.60-2.25)

% with no anti-drug antibodies

100%

Adalimumab

80% 60% Copies of Immunomodulator at visit 1 DQA1*05

40%

0%

Yes No Yes No

0 0 ≥1 ≥1

20%

0

100

200

300

400 Days

500

600

700

HLA-DQA1*05 Carriage Associated With Development of Anti-Drug Antibodies to Infliximab and Adalimumab in Patients With Crohn’s Disease Short title: HLA-DQA1*05 carriage associated with immunogenicity to infliximab and adalimumab Authors: Aleksejs Sazonovs,1 BSc; Nicholas A. Kennedy,2,3 MBBS; Loukas Moutsianas,1 DPhil; Graham A. Heap,2,4 PhD; Daniel L. Rice,1 MSc; Mark Reppell,4 PhD; Claire M. Bewshea,3 MSc; Neil Chanchlani, 2,3 MBChB; Gareth J. Walker,2,3 MBBS; Mandy H. Perry,5 PhD; Timothy J. McDonald,5 PhD; Charlie W Lees,6 PhD; J.R. Fraser Cummings,7,8 DPhil; Miles Parkes,9 DM; John C. Mansfield,10 MD; Jeffrey C. Barrett,1 DPhil; Dermot McGovern,11 DPhil; James R Goodhand,2,3 MBBS; Carl A. Anderson*,1 PhD; Tariq Ahmad*,2,3 DPhil & PANTS consortium. 1. Wellcome Sanger Institute, Wellcome Genome Campus, Hinxton, UK 2. Department of Gastroenterology, Royal Devon and Exeter Hospital NHS Foundation Trust, Exeter, UK 3. Exeter IBD and Pharmacogenetics Research Group, University of Exeter, Exeter, UK 4. AbbVie Inc. North Chicago, Illinois, USA 5. Department of Blood Science, Royal Devon and Exeter Hospital NHS Foundation Trust, Exeter, UK 6. Department of Gastroenterology, Western General Hospital, NHS Lothian, Edinburgh, UK 7. Department of Gastroenterology, University Hospital Southampton NHS Foundation Trust, Southampton, UK 8. Faculty of Experimental Medicine, University of Southampton 9. Department of Gastroenterology, Addenbrooke's Hospital, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK 10. Department of Gastroenterology, Newcastle Upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, UK 11. F. Widjaja Foundation Inflammatory Bowel and Immunobiology Research Institute, Cedars-Sinai Medical Center, Los Angeles, USA *denotes equal contribution Grant support: PANTS is an investigator-led study funded by CORE (renamed Guts UK in 2018), the research charity of the British Society of Gastroenterology, and by unrestricted educational grants from Abbvie Inc, USA, Merck Sharp & Dohme Ltd, UK, NAPP pharmaceuticals Ltd, UK, Pfizer Ltd, USA, Celltrion Healthcare, South Korea, and Cure Crohn’s Colitis (Scottish IBD Charity). HLA typing was supported through a sponsored research agreement with AbbVie Inc. The drug and anti-drug antibody assays were provided at reduced cost by Immundiagnostik AG. Laboratory tests were undertaken by the Exeter Blood Sciences Laboratory at the Royal Devon & Exeter (RD&E) NHS Trust (https://www.exeterlaboratory.com). The sponsor of the PANTS study is the Royal Devon and Exeter NHS Foundation Trust.

A.S., L.M., D.L.R., J.C.B. and C.A.A. are funded by the Wellcome Trust, who also funded the genotyping of all samples (206194). T.M. is funded by an NIHR HEE Senior Lectureship. The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, HEE, NIHR, or the Department of Health. None of the listed funders had a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and decision to submit the manuscript for publication. Abbreviations: TNF, Tumor Necrosis Factor; HLA, Human Leucocyte Antigen; HR, Hazard ratio; PANTS, Personalising Anti-TNF Therapy in Crohn’s disease; HRC, Haplotype Reference Consortium; MHC, Major Histocompatibility Complex; UC, ulcerative colitis; IBD-U, inflammatory bowel disease type-unclassified; AIC, Akaike information criterion Corresponding authors: Dr Carl A. Anderson [email protected] Genomics of Inflammation and Immunity Research Group, Morgan Building, Wellcome Sanger Institute, Wellcome Genome Campus, Hinxton, CB10 1SA, UK. Dr Tariq Ahmad [email protected] Exeter IBD and Pharmacogenetics Research Group, RILD building, Barrack Road, Exeter, EX2 5DW, UK. Direct Dial: + 44 (0) 1392 406218 Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and declare: N.A.K has consulted for Falk and received honoraria from Falk, Allergan, Pharmacosmos and Takeda for unrelated topics and is a deputy editor of Alimentary Pharmacology & Therapeutics Journal. G.A.H. reports non-financial support from AbbVie, outside the submitted work; and that he is now an employee of AbbVie and owns stock in the company. M.R. is an employee of AbbVie and owns stock in the company. N.C. is funded by Crohn’s and Colitis UK for a clinical research fellowship. G.J.W. has consulted for AbbVie and received honoraria from Falk and AbbVie for unrelated topics and a fellowship from NIHR. C.W.L reports grants and personal fees from Abbvie, personal fees from GSK, Janssen, Takeda, Amgen, Pfizer, Samsung Bioepis, Celgene, grants and personal fees from Gilead, outside the submitted work. J.R.F.C. reports personal fees from Abbvie, grants and personal fees from Takeda, Hospira and Biogen, personal fees from Janssen, MSD, Celltrion, NAPP, and Sandoz, outside the submitted work. J.C.B. reports grants from Wellcome Trust, during the conduct of the study and is now an employee of Genomics plc, outside the submitted work. D.McG. reports grants from NIH and Helmsley Charitable Trust, during the conduct of the study; grants and personal fees from Janssen, Precision IBD Inc, Second Genome, Qu Biologics, Pfizer, Gilead, personal fees from Takeda, outside the submitted work. J.R.G received honoraria from Falk, Abbvie and Shield therapeutics for unrelated topics. T.A. has received unrestricted research grants, advisory board fees, speaker honorariums and support to attend international meetings from AbbVie, Merck, Janssen, Takeda, Ferring, Tillotts, Ferring, Pfizer, NAPP, Celltrion, Hospira for unrelated topics, no financial relationships with any organizations that might have an interest in the submitted

1

work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. A.S., L.M., D.R., C.M.B., M.H.P., T.M., M.P., J.C.M., C.A.A. have no conflicts of interest to declare. Transcript Profiling: Summary statistics will be deposited on the NHGRI-EBI GWAS Catalog upon publication. Genotype data that supports this study has been deposited in the European Genomephenome Archive (EGA) under the accession code EGAS00001000924. Writing assistance: None Author Contributions: T.A., C.A.A., J.C.B., G.A.H., C.M.B., T.M., M.H.P., N.A.K. participated in the conception and design of the work: C.M.B was the project manager and coordinated recruitment. A.S., L.M., D.L.R., J.C.B., C.A.A, M.R, G.J.W, N.A.K. were involved in the acquisition, analysis or interpretation of data. C.W.L., J.R.F.C., M.P., J.C.M., D.L.R, N.A.K. were involved in the acquisition of the replication cohort. The statistical analysis was performed by A.S., N.A.K., L.M., J.C.B., C.A.A. Drafting of the manuscript was conducted by A.S., N.A.K., L.M., G.A.H., C.M.B., N.C., D.McG., J.R.G., C.A.A, T.A. All the authors contributed to the critical review and final approval of the manuscript. T.A. and C.A.A. obtained the funding for the study. All remaining authors contributed by submitting a substantial number of samples in line with ICMJE criteria. We would also like to acknowledge the study coordinators of the Exeter IBD Research Group, United Kingdom: Marian Parkinson and Helen Gardner-Thorpe for their ongoing administrative support to the study. We would also like to thank the NIHR UK IBD BioResourse for help establishing the replication cohort.

2

Abstract Background & Aims: Anti-tumor necrosis factor (anti-TNF) therapies are the most widely used biologic drugs for treating immune-mediated diseases, but repeated administration can induce the formation of anti-drug antibodies. The ability to identify patients at increased risk for development of anti-drug antibodies would facilitate selection of therapy and use of preventative strategies. Methods: We performed a genome-wide association study to identify variants associated with time to development of anti-drug antibodies in a discovery cohort of 1240 biologicnaïve patients with Crohn’s disease starting infliximab or adalimumab therapy. Immunogenicity was defined as an anti-drug antibody titer ≥10 AU/ml in a drug-tolerant ELISA assay. Significant association signals were confirmed in a replication cohort of 178 patients with inflammatory bowel disease. Results: The HLA-DQA1*05 allele, carried by approximately 40% of Europeans, significantly increased the rate of immunogenicity (hazard ratio [HR], 1.90; 95% CI, 1.60–2.25; P=5.88×10-13). The highest rates of immunogenicity, 92% at 1 year, were observed in patients treated with infliximab monotherapy who carried HLA-DQA1*05; conversely the lowest rates of immunogenicity, 10% at 1 year, were observed in patients treated with adalimumab combination therapy who did not carry HLA-DQA1*05. We confirmed this finding in the replication cohort (HR, 2.00; 95% CI, 1.35–2.98; P=6.60×10-4). This association was consistent for patients treated with adalimumab (HR, 1.89; 95% CI, 1.32–2.70) or infliximab (HR, 1.92; 95% CI, 1.57–2.33), and for patients treated with anti-TNF therapy alone (HR, 1.75; 95% CI, 1.37–2.22) or in combination with an immunomodulator (HR, 2.01; 95% CI, 1.57–2.58). Conclusions: In an observational study, we found a genome-wide significant association between HLA-DQA1*05 and the development of antibodies against anti-TNF agents. A randomized controlled biomarker trial is required to determine whether pretreatment testing for HLA-DQA1*05 improves patient outcomes by helping physicians select anti-TNF and combination therapies. ClinicalTrials.gov no: NCT03088449 KEY WORDS: PANTS, GWAS, loss of response, drug persistence

3

Background Anti-tumor necrosis factor (anti-TNF) therapies are the most widely used biologics for treating immune-mediated diseases, and in 2018, accounted for a global expenditure in excess of $23.5 billion.1 Repeated administration often induces the formation of anti-drug antibodies (immunogenicity) leading to treatment failure.2

Immunogenicity is more common in patients treated with infliximab (a murine-human chimeric monoclonal antibody) than adalimumab (a fully human monoclonal antibody) and is a major cause of low anti-TNF drug level, infusion reactions, and non-remission in patients with Crohn’s disease.2,3 Combination immunomodulator therapy reduces the risk of immunogenicity to both adalimumab and infliximab, and for infliximab, improves treatment outcomes.2,4–6 Despite these benefits, many patients are still treated with anti-TNF monotherapy because of concerns about the increased risk of adverse drug reactions, opportunistic infections, and malignancies associated with combination therapy with immunomodulators.7–9

The ability to identify patients at increased risk of immunogenicity may influence the choice of anti-TNF treatment and the use of preventative strategies, including combination therapy with immunomodulators. However, our understanding of the cellular and molecular mechanisms underpinning immunogenicity to biologics is limited. Retrospective candidate gene studies have suggested variants in FCGR3A10 and HLA-DRB111,12 increase susceptibility to immunogenicity to anti-TNF therapy. These associations did not achieve genome-wide significance and are yet to be independently replicated. Here, we report the first genetic locus robustly associated with immunogenicity to anti-TNF therapies.

4

Methods The Personalising Anti-TNF Therapy in Crohn’s disease (PANTS) study is a UK-wide, multicenter, prospective observational cohort reporting the treatment failure rates of the anti-TNF drugs infliximab (originator, Remicade [Merck Sharp & Dohme, UK] and biosimilar, CT-P13 [Celltrion, South Korea]), and adalimumab (Humira [Abbvie, USA]) in 1610 anti-TNFnaïve patients with Crohn’s disease.2

At inclusion, subjects were aged six years or over and had active luminal Crohn’s disease involving the colon and/or small intestine. Choice of anti-TNF drug and use of concomitant immunomodulator therapy was at the discretion of the treating physician as part of usual care. Patients were initially studied for 12 months or until drug withdrawal. In the first year, study visits were scheduled at first dose, post-induction (weeks 12-14), weeks 30, 54, and at treatment failure. For infliximab-treated patients, additional visits occurred at each infusion. After 12 months, patients were invited to continue follow-up for a further two years. Drug persistence was defined as the duration of time from initiation of anti-TNF therapy to exit from the study due to treatment failure. Patients who exited the study for other reasons, declined to participate in the two-year extension, or were lost to follow-up were censored at the time of last drug dose or study visit.

At each visit, serum infliximab or adalimumab drug and anti-drug antibody levels were analyzed using total antibody enzyme linked immunosorbent assays.13 See the Supplementary Appendix for a detailed description of drug and anti-drug antibody testing. The total antibody, unlike the more commonly reported free antibody assay, includes a drug-antibody disassociation step that allows the assessment of anti-drug antibodies in the

5

presence of drug. We defined immunogenicity as an anti-drug antibody concentration of ≥10 AU/mL, irrespective of drug level, at one or more time points.

DNA was extracted from pre-treatment blood samples from 1524 individuals in the PANTS cohort and genotyping undertaken using the Illumina CoreExome microarray. See the Supplementary Appendix for a detailed description of the genetic analyses. We excluded individuals of non-European ancestry (identified using principal component analysis), one individual from each related pair (defined as a pi-hat>0.187), and those with an outlying number of missing or heterozygous genotypes. 1323 individuals remained in the study following quality control, of which 1240 had drug and anti-drug antibody level data available (Figure S2 in the Supplementary Appendix). With its case-control design, our study has more than 80% power to detect genome-wide significant evidence of association (α = 5×10-8) to variants with a minor allele frequency greater than 22% and a relative risk more than 1.4 (Figure S3 in the Supplemental Appendix).14 Baseline demographics for these patients are shown in Table S1 in the Supplementary Appendix.

Variants with a genotype call rate of <95%, or with significant evidence of deviation from Hardy-Weinberg equilibrium (P<1×10-10) were excluded. Single nucleotide polymorphisms were imputed via the Sanger Imputation Service using the Haplotype Reference Consortium (HRC) panel, and 7 578 947 variants with an information content metric (INFO) score > 0.4 were subsequently taken forward for analysis.15 Human leukocyte antigen (HLA) types were imputed at 2-, and 4-digit resolution for the following loci: HLA-A, HLA-C, HLA-B, HLA-DRB1, HLA-DQA1, HLA-DQB1, and HLA-DPB1. Long-read sequencing of these HLA alleles was undertaken to assess the accuracy of our imputation.

6

We assembled an independent cohort to replicate significant findings from the discovery cohort. This comprised 107 Crohn’s disease, 64 ulcerative colitis (UC), and 7 inflammatory bowel disease type-unclassified (IBD-U) patients with cross-sectional drug and anti-drug antibody levels measured as part of routine clinical practice. The samples were genotyped using either the Illumina CoreExome array (N=164)16 or the Affymetrix 500k array (N=14).17 Quality control and imputation methods were the same as in the discovery cohort.

Statistical analysis Rates of immunogenicity were estimated using the Kaplan-Meier method. Clinical outcomes and genetic association tests with time to anti-drug antibody development were performed using multivariable Cox proportional hazards regression: sex, drug type (infliximab or adalimumab), immunomodulator use, and the first within-sample principal component, were included as covariates (See Table S3 in the Supplementary Appendix). Patients who did not develop immunogenicity during the study were censored at the point of last observation. Post-hoc sensitivity analyses were undertaken to test our genetic findings with immunogenicity, firstly, at progressively higher antibody thresholds; secondly, to simulate a free-antibody assay and thirdly, excluding patients with a single anti-drug antibody level > 10 AU/ml and subsequent negative anti-drug antibodies <10 AU/ml.

The Akaike information criterion (AIC) was used to compare non-nested models to assess if the mode of inheritance was dominant or additive, and to determine whether HLA allele group, specific HLA alleles, or amino acid sequence best explained the association. The fixed effects Q statistic was used to perform tests of heterogeneity of effect; this test is an

7

extension of Cochran’s Q-test and examines whether the observed effect size variability is larger than expected by chance. Interaction tests of the differential effects of drug type (infliximab vs adalimumab and Remicade vs CT-P13) and combination therapy (immunomodulator vs no immunomodulator) conditional on the genotype were performed. Mann-Whitney U tests were used to compare serum levels of anti-drug antibodies at week 54 stratified by anti-TNF drug and immunomodulator use.

Ethics The South West Research Ethics committee approved the study (REC reference: 12/SW/0323) in January, 2013. Patients were included after providing informed, written consent. The protocol is available online (www.ibdresearch.co.uk).

Results Within the first 12 months, 44% of patients developed anti-drug antibodies (95% CI, 0.41 to 0.48), and 62% of patients did so within 36 months (95% CI, 0.57 to 0.67). After correcting for immunomodulator use, the rate of immunogenicity was greater in patients treated with infliximab (N=742) than adalimumab (N=498) (hazard ratio (HR), 3.21; 95% CI, 2.61 to 3.95; P=1.18×10-28). In a model including drug-type as a covariate, rates of immunogenicity were greater in patients treated with anti-TNF monotherapy (N=544) compared to combination therapy with immunomodulators (N=696), (HR, 2.30; 95% CI, 1.94 to 2.75; P<6.10×10-21).

8

A locus within the HLA region is associated with time to immunogenicity We identified a genome-wide significant association on chromosome 6 with time to development of immunogenicity, with the most associated SNP, rs2097432 (b38_pos: 6:32622994; HR, 1.70; 95% CI, 1.48 to 1.94; P=4.24×10-13), falling within the major histocompatibility complex (MHC) region (Fig. 1 and Figure S4-5 in the Supplementary Appendix). We replicated this association in our independent cohort of 178 patients with IBD (HR, 1.69; 95% CI, 1.26 to 2.28; P=8.80×10-4). A variant on chromosome 11, rs12721026 (b38_pos: 11:116835452; HR, 0.46; 95% CI, 0.33 to 0.63; P=4.76×10-8) also reached genomewide significance in our discovery analysis, though the association was not replicated in our independent cohort (HR, 0.85; 95% CI, 0.49 to 1.44; P=0.51).

Fine-mapping of the signal in the HLA region At the HLA allele group level (2-digit resolution), only HLA-DQA1*05 achieved genome-wide significance (HR, 1.90; 95% CI, 1.60 to 2.25; P=5.88×10-13) (Fig. 2). At the specific allele level (4-digit resolution), no single allele reached genome-wide significance. The two most common HLA-DQA1*05 subtype alleles, HLA-DQA1*05:01 (HR, 1.57; 95% CI, 1.33 to 1.85; P=4.24×10-7) and HLA-DQA1*05:05 (HR, 1.48; 95% CI, 1.24 to 1.78; P=5.54×10-5), had similar effects on time to immunogenicity and a model containing these two 4-digit alleles was virtually indistinguishable from a model including only HLA-DQA1*05 (AIC05=6659.07 versus AIC05:01&05:05=6659.50). We did not identify any amino acids that better fit the data than HLA-DQA1*05. We observed >99% concordance between imputed and sequenced HLA genotypes at HLA-DQA1 (Table S4 in the Supplementary Appendix).

9

To formally assess the inheritance pattern of HLA-DQA1*05 mediated immunogenicity, we compared the fit of additive and dominant models and found that the dominant model gave a better fit (AICDOM=6652.12 vs AICADD=6659.07), and stronger association signal for HLADQA1*05 (HR, 1.90; 95% CI, 1.60 to 2.25; P=5.88×10-13) (Fig. 3 and Table S2 in the Supplementary Appendix). We also looked for non-additive effects across all other HLA alleles, but the model assuming a dominant effect for HLA-DQA1*05 remained the best fit to the data. The HLA-DQA1*05 association was confirmed in our replication cohort (entire cohort (HR, 2.00; 95% CI, 1.35 to 2.98; P=6.60×10-4), Crohn’s disease-only subset (HR=2.26, 95% CI: 1.33 to 3.84, p=0.003), and UC-only subset of the replication cohort (HR=2.02, 95% CI: 1.08-3.79 p=0.03), again with a better fit for the dominant model (AICDOM=942.51 vs AICADD=944.81). After conditioning on HLA-DQA1*05 we did not identify any secondary signals of association with time to immunogenicity within the MHC region (Figure S6 in the Supplementary Appendix).

Sensitivity analyses showed that the effect size of HLA-DQA1*05 carriage on immunogenicity was similar across subgroups (Fig. 4A and 4B): firstly, when the threshold for defining immunogenicity was increased from 10AU/mL to > 150 AU/mL. Secondly, when we simulated a drug-sensitive instead of a drug-tolerant assay, where immunogenicity was defined as an anti-drug antibody titer ≥10 AU/ml without detectable drug (HR, 1.57; 95% CI, 1.23-2.01; P = 3.66 × 10-4). Thirdly, when we removed patients with a one-off transient antidrug antibody level ≥10 AU/ml (HR, 1.94; 95% CI, 1.62-2.32; P = 8.46 × 10-13).

10

The effect of HLA-DQA1*05 across drug and treatment regimes While immunogenicity rates were lower with adalimumab-treated compared to infliximabtreated patients, we did not detect a significant difference in the effect of HLA-DQA1*05 on the immunogenicity rate for these two drugs (HR, 1.89; 95% CI, 1.32-2.70 in adalimumab-, HR, 1.92; 95% CI, 1.57-2.33 in infliximab-treated patients; Phet=0.91) (Fig. 5). We also found no significant evidence for heterogeneity of effect of HLA-DQA1*05 on immunogenicity between patients treated with the infliximab originator, Remicade, and its biosimilar CT-P13 (Phet=0.23) (Figure S7 in the Supplementary Appendix). Likewise, we did not detect any significant heterogeneity of effect of HLA-DQA1*05 carriage on immunogenicity for individuals on monotherapy (HR, 1.75; 95% CI, 1.37-2.22) versus combination therapy (HR, 2.01; 95% CI, 1.57-2.58) with immunomodulators (Phet=0.14). In addition, we did not identify any significant interactions between HLA-DQA1*05 and the clinical covariates (drug type: p=0.83; mono- vs combination therapy: p=0.71; Remicade vs CT-P13: p=0.59).

The highest rates of immunogenicity, 92% at 1 year, were observed in patients treated with infliximab monotherapy who carried HLA-DQA1*05 (Fig. 6A). Conversely, the lowest rates of immunogenicity, 10% at 1 year, were observed in patients treated with adalimumab combination therapy who did not carry HLA-DQA1*05 (Fig 6B). Our final model, which includes HLA-DQA1*05 status, sex, drug, and immunomodulator usage, explained 18% of the variance in immunogenicity to anti-TNF in our cohort.

Having demonstrated that HLA-DQA1*05 was associated with time to immunogenicity, we sought associations with anti-drug antibody titers after one year of treatment and

11

subsequent non-persistence on drug. Carriage of HLA-DQA1*05 was associated with higher maximal anti-drug antibody titers (Pinfliximab = 8×10-10; Padalimumab = 0.002). We observed lower drug persistence rates to year three in patients treated with an anti-TNF drug without an immunomodulator (Fig. 7A and 7B); the optimal model here used the interaction between immunomodulator use and HLA-DQA1*05 (DQA1*05: HR, 1.40; 95% CI 1.08-1.80; P=0.011, immunomodulator use: HR, 0.74; 95% CI, 0.58-0.94, P=0.014, interaction between DQA1*05 and immunomodulator use: HR, 0.65; 95% CI, 0.45-0.95; P=0.026).

Discussion Immunogenicity to biologic therapies is a major concern for patients, regulatory authorities, and the pharmaceutical industry. We report the first genome-wide significant association with immunogenicity to anti-TNF therapy using the largest prospective cohort study of infliximab and adalimumab in Crohn’s disease. We have demonstrated that carriage of one or more HLA-DQA1*05 alleles confers an almost two-fold risk of immunogenicity to anti-TNF therapy, irrespective of concomitant immunomodulator use or drug type (infliximab [Remicade or CT-P13], or adalimumab). Fine-mapping and confirmatory sequencing of the HLA identified that the specific alleles HLA-DQA1*05:01 and HLA-DQA1*05:05 mediated most of this risk. Carriage of HLA-DQA1*05 was associated with higher anti-drug antibody levels and lower drug persistence rates, although further studies are needed to more accurately quantify the relationship between HLA-DQA1*05 and the risk of anti-TNF treatment failure. Based on our data in patients with luminal Crohn’s disease, all patients treated with an antiTNF should be prescribed an immunomodulator to lower the risk of immunogenicity. We

12

hypothesize that for patients who carry HLA-DQA1*05 in whom immunomodulators are contraindicated or not tolerated, clinicians might advise against the use of anti-TNF drugs, particularly infliximab. This is because 90% of patients who carry HLA-DQA1*05 who are treated with infliximab monotherapy have evidence of immunogenicity by week 54. In contrast, patients who do not carry HLA-DQA1*05 might be given the choice between adalimumab or infliximab combination therapy. Patients without the risk allele and a history of adverse drug reactions to thiopurines and/or methotrexate or who are at high risk of opportunistic infections might be spared the additional risks of combination therapy and treated with adalimumab monotherapy. A randomized controlled biomarker trial is required to explore these hypotheses and confirm whether HLA-DQA1*05 testing may help direct treatment choices in order to improve clinical outcomes.

The shared genetic association between HLA-DQA1*05 and immunogenicity to infliximab and adalimumab may explain the widely reported diminishing returns of switching between anti-TNF therapies at the time of loss of response.18,19 If the immunogenic effect of HLADQA1*05 extends to other therapeutic antibodies, then subjects who carry the variant may be candidates for non-antibody modality therapies, such as small molecule drugs.

Allelic variation in the HLA-DQA1 gene has been linked to aberrant adaptive immune responses. The HLA class II gene HLA-DQA1 is expressed by antigen presenting cells and encodes the alpha chain of the HLA-DQ heterodimer that forms part of the antigen binding site where epitopes are presented to T-helper cells. Relevant to immunogenicity, carriage of HLA-DQA1*05 has been associated with celiac disease, type 1 diabetes, and protection

13

against rheumatoid arthritis and pulmonary tuberculosis.20–24 Several hypotheses have been proposed, but exactly how specific HLA alleles contribute to disease pathogenesis or, in this case, increased immunogenicity, remains unknown. HLA-DQ1A*05 may serve as a useful biomarker of immunogenicity risk and may impact how the next-generation of anti-TNF drugs are designed to minimize HLA-DQA1*05 mediated immunogenicity. Previous studies have shown that it is possible to map and eliminate potential immunogenic T cell epitopes with the aim of producing safer and more durable biologic drugs.25,26 However, caution needs to be exercised to ensure protein sequence modifications designed to reduce the risk of immunogenicity to patients carrying HLADQA1*05 do not put a different group of patients at risk.

Multiple assays are available to detect anti-drug antibodies and there is no universally accepted, validated threshold to diagnose immunogenicity. We deliberately chose a total, or drug tolerant assay, that permits the measurement of anti-drug antibodies in the presence of drug, in order to minimize the number of false negative patients assigned to the control group. We then validated the manufacturer’s positivity threshold in independent experiments in 500 drug-naïve controls and confirmed that the recommend cut-off of 10 AU/mL corresponds to the 99th centile of the anti-drug antibody titer distribution. In support of this threshold we have recently demonstrated that even modestly elevated antidrug antibodies levels (10-30 Au/ml) at weeks 14 and 54 of treatment are associated with lower drug levels at these time points, and non-remission at week 54.2 In addition, sensitivity analyses confirmed that the association and effect size between HLA-DQA1*05 and immunogenicity remained at progressively higher diagnostic thresholds for

14

immunogenicity, when we simulated a free-assay, and when we removed patients with transient antibodies. Finally, HLA-DQA1*05 was associated with the quantitative trait of maximal anti-drug antibody titer.

We acknowledge several important limitations of this study. Firstly, we may have underestimated the contribution of HLA-DQA1*05 to immunogenicity because of the short duration of follow-up in patients who did not continue in the study beyond the first year. Secondly, because we designed a schedule of visits to minimize patients’ inconvenience, there were fewer assessments for those treated with adalimumab than infliximab. As a result, we might have underestimated rates of immunogenicity amongst adalimumabtreated patients.

Our genome-wide association study was limited to patients with luminal Crohn’s disease of European descent. Given that HLA-DQA1*05 is not associated with IBD risk27 the percentage of carriers among our patients (39%) was similar to that reported in an independent British population cohort (38%).28 As such, we hypothesize that HLA-DQA1*05 will make a similar contribution to anti-TNF immunogenicity in other patient populations where the allele is not associated to disease susceptibility (e.g. ankylosing spondylitis, perianal Crohn’s disease). Due to the wide variation in the frequency of HLA-DQA1*05 across ethnic groups,28 further studies are required to assess the contribution of HLA-DQA1*05 to immunogenicity across populations. Whether HLA-DQA1*05 is also associated with immunogenicity to other biologic drugs also needs to be determined.

15

Conclusion We report the first genome-wide significant association with immunogenicity to biologic drugs. Carriage of HLA-DQA1*05 almost doubles the rate of anti-TNF anti-drug antibody development, independent of immunomodulator use, for both infliximab and adalimumab. To minimize the risk of immunogenicity, pre-treatment genetic testing for HLA-DQA1*05 may help personalize the choice of anti-TNF and the need for combination therapy with an immunomodulator.

16

References 1.

IQVIA Institute. The Global Use of Medicine in 2019 and Outlook to 2023. 2019. Available at: https://www.iqvia.com/-/media/iqvia/pdfs/institute-reports/the-global-use-of-medicine-in2019-and-outlook-to-2023.pdf?_=1557505804444.

2.

Kennedy NA, Heap GA, Green HD, et al. Predictors of anti-TNF treatment failure in anti-TNFnaive patients with active luminal Crohn’s disease: a prospective, multicentre, cohort study. Lancet Gastroenterol Hepatol 2019;1253:1–13.

3.

Vermeire S, Gils A, Accossato P, et al. Immunogenicity of biologics in inflammatory bowel disease. Therap Adv Gastroenterol 2018;11:1756283X17750355. Available at: http://www.ncbi.nlm.nih.gov/pubmed/29383030.

4.

Colombel JF, Sandborn WJ, Reinisch W, et al. Infliximab, Azathioprine, or Combination Therapy for Crohn’s Disease. N Engl J Med 2010;362:1383–1395.

5.

Panaccione R, Ghosh S, Middleton S, et al. Combination Therapy With Infliximab and Azathioprine Is Superior to Monotherapy With Either Agent in Ulcerative Colitis. Gastroenterology 2014;146:392–400.e3.

6.

Matsumoto T, Motoya S, Watanabe K, et al. Adalimumab Monotherapy and a Combination with Azathioprine for Crohn’s Disease: A Prospective, Randomized Trial. J Crohns Colitis 2016;10:1259–1266. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27566367.

7.

Lichtenstein GR, Feagan BG, Cohen RD, et al. Drug therapies and the risk of malignancy in Crohn’s disease: results from the TREATTM Registry. Am J Gastroenterol 2014;109:212–23.

8.

Osterman MT, Sandborn WJ, Colombel J-F, et al. Increased risk of malignancy with adalimumab combination therapy, compared with monotherapy, for Crohn’s disease. Gastroenterology 2014;146:941–9.

9.

Lemaitre M, Kirchgesner J, Rudnichi A, et al. Association Between Use of Thiopurines or Tumor Necrosis Factor Antagonists Alone or in Combination and Risk of Lymphoma in Patients With Inflammatory Bowel Disease. JAMA 2017;318:1679–1686. Available at: http://www.ncbi.nlm.nih.gov/pubmed/29114832.

10.

Romero-Cara P, Torres-Moreno D, Pedregosa J, et al. A FCGR3A Polymorphism Predicts Antidrug Antibodies in Chronic Inflammatory Bowel Disease Patients Treated With Anti-TNF. Int J Med Sci 2018;15:10–15.

11.

Billiet T, Casteele N Vande, Stappen T Van, et al. Immunogenicity to infliximab is associated with HLA-DRB1. Gut 2015;64:1344–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25876612.

12.

Liu M, Degner J, Davis JW, et al. Identification of HLA-DRB1 association to adalimumab

17

immunogenicity. PLoS One 2018;13:e0195325. 13.

Perry M, Bewshea C, Brown R, et al. Infliximab and adalimumab are stable in whole blood clotted samples for seven days at room temperature. Ann Clin Biochem 2015;52:672–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25780249.

14.

Johnson JL, Abecasis GR. GAS Power Calculator: web-based power calculator for genetic association studies. bioRxiv 2017:164343.

15.

McCarthy S, Das S, Kretzschmar W, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat Genet 2016;48:1279–83.

16.

Lange KM de, Moutsianas L, Lee JC, et al. Genome-wide association study implicates immune activation of multiple integrin genes in inflammatory bowel disease. Nat Genet 2017;49:256– 261.

17.

Wellcome Trust Case Control Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 2007;447:661–678.

18.

Panaccione R, Loftus E V, Binion D, et al. Efficacy and safety of adalimumab in Canadian patients with moderate to severe Crohn’s disease: results of the Adalimumab in Canadian SubjeCts with ModErate to Severe Crohn’s DiseaSe (ACCESS) trial. Can J Gastroenterol 2011;25:419–25.

19.

Sandborn WJ, Rutgeerts P, Enns R, et al. Adalimumab Induction Therapy for Crohn Disease Previously Treated with Infliximab. Ann Intern Med 2007;146:829.

20.

Megiorni F, Pizzuti A. HLA-DQA1 and HLA-DQB1 in Celiac disease predisposition: practical implications of the HLA molecular typing. J Biomed Sci 2012;19:88.

21.

Oliveira-Cortez A, Melo AC, Chaves VE, et al. Do HLA class II genes protect against pulmonary tuberculosis? A systematic review and meta-analysis. Eur J Clin Microbiol Infect Dis 2016;35:1567–80.

22.

Reddy MPL, Wang H, Liu S, et al. Association between type 1 diabetes and GWAS SNPs in the southeast US Caucasian population. Genes Immun 2011;12:208–212.

23.

Eleftherohorinou H, Hoggart CJ, Wright VJ, et al. Pathway-driven gene stability selection of two rheumatoid arthritis GWAS identifies and validates new susceptibility genes in receptor mediated signalling pathways. Hum Mol Genet 2011;20:3494–506.

24.

Ursum J, Weijden MAC van der, Schaardenburg D van, et al. IL10 GGC haplotype is positively and HLA-DQA1*05-DQB1*02 is negatively associated with radiographic progression in undifferentiated arthritis. J Rheumatol 2010;37:1431–8.

25.

Groot AS De, Knopp PM, Martin W. De-immunization of therapeutic proteins by T-cell epitope modification. Dev Biol (Basel) 2005;122:171–94.

18

26.

Sathish JG, Sethu S, Bielsky M-C, et al. Challenges and approaches for the development of safer immunomodulatory biologics. Nat Rev Drug Discov 2013;12:306–24. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23535934.

27.

Goyette P, Boucher G, Mallon D, et al. High-density mapping of the MHC identifies a shared role for HLA-DRB1*01:03 in inflammatory bowel diseases and heterozygous advantage in ulcerative colitis. Nat Genet 2015;47:172–179.

28.

González-Galarza FF, Takeshita LYC, Santos EJM, et al. Allele frequency net 2015 update: new features for HLA epitopes, KIR and disease and HLA adverse drug reaction associations. Nucleic Acids Res 2015;43:D784–D788.

19

Figure Captions Figure 1: A regional plot of the association results with the MHC region on chromosome 6. Midpoint positions of the HLA alleles across the MHC region are shown in red on the x-axis. SNPs that passed the genome-wide significance threshold (P=5×10-8) are shown above the red horizontal dashed line with the most significant SNP in red. SNPs correlated with the lead SNP (r2>0.05) are color-coded from purple to yellow. Pairwise genotype correlation (r2) between SNPs was calculated using genotype data from the non-Finnish European population of the 1000 Genomes Project. Figure 2: Effect sizes of the most strongly associated SNP, HLA alleles, and amino acids of time to immunogenicity. Blue lines represent 95% CIs. Association test P-values are shown in parentheses Figure 3: Kaplan–Meier estimator showing the rate of anti-drug antibody development, stratified by the number of HLA-DQA1*05 alleles carried. Orange, blue, and red indicate 0, 1, and 2 copies of DQA1*05 allele, respectively. Carriers of one or two copies of the allele have a similar rate of immunogenicity development, and a dominant model is a better fit for the data than an additive model (AICDOM=6652.12 vs AICADD=6659.07). X-axis truncated at 700 days, due to the low number of observations for longer time periods. Figure 4: A Sensitivity analysis of the effect size of DQA1*05 association and time to immunogenicity B Sensitivity analysis of the significance of DQA1*05 association and time to immunogenicity In the primary analyses, immunogenicity was defined as an anti-drug antibody concentration ≥10 AU/mL, irrespective of drug concentration (red dot). We repeated the time to immunogenicity analysis varying this definition from ≥5 to ≥200 AU/mL. Figure 5: HLA-DQA1*05 has a consistent effect on immunogenicity in different patient subgroups. We repeated the proportional hazard association analysis, separating the full cohort into subgroups by drug and therapy type. Estimated hazard ratios and standard errors between the pairings were compared using a heterogeneity of effects test (P>0.05), suggesting that the effect of DQA1*05 on immunogenicity is not affected by these clinical covariates. Blue lines represent 95% CIs. Association test P-values are shown in parentheses Figure 6: A Anti-drug antibody development - infliximab B Anti-drug antibody development - adalimumab Kaplan–Meier estimator showing the rate of anti-drug antibody development (A and B), stratified by carriage of HLA-DQA1*05 alleles and treatment regime. Dotted lines indicate patients undergoing anti-TNF monotherapy; solid lines indicate combination therapy with

20

immunomodulators. Red indicates carriers of the HLA-DQA1*05 allele (1 or 2 copies); blue indicates non-carriers. For both drugs and treatment regimes, immunogenicity rates are higher for HLA-DQA1*05 carriers. The X-axis was truncated at 700 days due to the low number of observations. Figure 7: A Drug persistence - infliximab B Drug persistence – adalimumab Kaplan–Meier estimator showing the rate of drug persistence (A and B), stratified by carriage of HLA-DQA1*05 alleles and treatment regime. Dotted lines indicate patients undergoing anti-TNF monotherapy; solid lines indicate combination therapy with immunomodulators. Red indicates carriers of the HLA-DQA1*05 allele (1 or 2 copies); blue indicates non-carriers. For both drugs and treatment regimes, drug persistence rates are higher for HLA-DQA1*05 carriers. The X-axis was truncated at 700 days due to the low number of observations.

21

12

0.8

8

0.6

Top SNP r2

10

6

0.4

4 0.2 31

32

Position on chromosome 6 (mb)

33

DPB1

30

DRB1 DQA1 DQB1

0

C B

2 A

-log10(P)

1.0

rs2097432

0.0

Top SNP

rs2097432 (P = 4.55×10 13)

Allele groups

HLA-DQA1*05 (P = 5.88×10 13) HLA-DQA1*05:01 (P = 5.49×10 7)

Specific alleles

HLA-DQA1*05:05 (P = 3.39×10 5) AA-107-ile (P = 4.02×10 13)

Amino acids 0.8

AA-175-lys (P = 4.02×10 13) 1.0

1.2

1.4

1.6 1.8 Hazard ratio

2.0

2.2

2.4

% with no anti-drug antibodies

100% 80% 60% 40%

DQA1*05 0 copies 1 copy 2 copies

20% 0%

0

100

200

300

Number at risk 626 0 copies 752 320 1 copy 410 57 2 copies 75

478 216 38

361 137 25

Days

400

500

600

700

216 90 18

173 72 16

133 43 9

119 39 6

Figure 4: A Sensitivity analysis of the effect size of DQA1*05 association and time to immunogenicity

4.0 3.5

HR

3.0 2.5 2.0 1.5 1.0 0

25 50 75 100 125 150 175 Minimal anti-drug antibody concentration (AU/mL)

200

B Sensitivity analysis of the significance of DQA1*05 association and time to immunogenicity

12

-log10(P)

10 8 6 4 2 0

0

25 50 75 100 125 150 175 Minimal anti-drug antibody concentration (AU/mL)

200

In the primary analyses, immunogenicity was defined as an anti-drug antibody concentration ≥10 AU/mL, irrespective of drug concentration (red dot). We repeated the time to immunogenicity analysis varying this definition from ≥5 to ≥200 AU/mL.

Full cohort (N = 1240; P = 5.88×10 13)

-DQA1*05 × covariate interaction

Adalimumab (N = 498; P = 5.05×10 4)

P = 0.83

Infliximab (N = 742; P = 1.82×10 10) Monotherapy (N = 544; P = 7.38×10 6)

P = 0.71 0.75

Combination therapy (N = 696; P = 4.26×10 8) 1.00

1.25

1.50

1.75 Hazard ratio

2.00

2.25

2.50

2.75

0 copies of 0 copies of 1 copy of 1 copy of

Figure 6: A Drug persistence -infliximab Infliximab

100%

% drug persistence

DQA1*05, immunomodulators on Visit 1 DQA1*05, no immunomodulators on Visit 1 DQA1*05, immunomodulators on Visit 1 DQA1*05, no immunomodulators on Visit 1

80% 60% 40% 20% 0%

0

100

Number at risk 270 252 177 160 176 163 117 103

200

300

400

500

600

700

185 101 123 57

132 79 85 37

117 62 77 31

104 55 65 26

400

500

600

700

86 76 52 42

64 60 38 32

58 49 29 24

50 45 24 23

Days 232 137 156 82

205 119 141 72

B Drug persistence – adalimumab

Adalimumab

% drug persistence

100% 80% 60% 40% 20% 0%

0

100

Number at risk 147 137 157 142 101 95 89 79

200

300 Days

128 127 83 68

114 116 75 56

Kaplan–Meier estimator showing the rate of drug persistence (A and B), stratified by carriage of HLADQA1*05 alleles and treatment regime. Dotted lines indicate patients undergoing anti-TNF monotherapy; solid lines indicate combination therapy with immunomodulators. Red indicates carriers of the HLA-DQA1*05 allele (1 or 2 copies); blue indicates non-carriers. For both drugs and treatment regimes, immunogenicity and drug persistence rates are higher for HLA-DQA1*05 carriers. The X-axis was truncated at 700 days due to the low number of observations.

0 copies of 0 copies of 1 copy of 1 copy of

% with no anti-drug antibodies

Figure 7: A Anti-drug antibody development - infliximab

DQA1*05, immunomodulators on Visit 1 DQA1*05, no immunomodulators on Visit 1 DQA1*05, immunomodulators on Visit 1 DQA1*05, no immunomodulators on Visit 1

Infliximab

100% 80% 60% 40% 20% 0%

0

100

Number at risk 240 270 125 177 151 176 75 118

200

300

400

500

600

700

84 30 38 6

70 24 37 5

55 19 23 3

49 17 22 3

400

500

600

700

55 47 35 29

42 37 25 21

29 30 15 11

25 28 11 9

Days 185 76 107 29

145 50 63 9

% with no anti-drug antibodies

B Anti-drug antibody development - adalimumab

Adalimumab

100% 80% 60% 40% 20% 0%

0

100

Number at risk 147 129 158 132 102 87 89 64

200

300 Days

113 104 69 49

89 77 55 35

Kaplan–Meier estimator showing the rate of anti-drug antibody development (A and B), stratified by carriage of HLA-DQA1*05 alleles and treatment regime. Dotted lines indicate patients undergoing antiTNF monotherapy; solid lines indicate combination therapy with immunomodulators. Red indicates carriers of the HLA-DQA1*05 allele (1 or 2 copies); blue indicates non-carriers. For both drugs and treatment regimes, immunogenicity and drug persistence rates are higher for HLA-DQA1*05 carriers. The X-axis was truncated at 700 days due to the low number of observations.

What you need to know: BACKGROUND AND CONTEXT: Anti-tumor necrosis factor (anti-TNF) therapies are the most widely used biologic drugs for treating immune-mediated diseases, but repeated administration can induce the formation of anti-drug antibodies, increasing the risk of treatment failure. NEW FINDINGS: In a genome-wide association study, we identified a variant (HLA-DQA1*05) that increases risk of development of antibodies against infliximab and adalimumab 2-fold in patients with Crohn’s disease, regardless of concomitant immunomodulator use. LIMITATIONS: The study was limited to patients of European descent, 40% of whom carry the risk allele. IMPACT: Testing patients for HLA-DQA1*05 might help physicians decide whether they should receive anti-TNF and combination immunomodulator therapy. Lay Summary: We identified a genetic variant that increases risk of development of antibodies against anti-TNF agents, which reduces their effectiveness in treatment of Crohn’s disease.

Appendix Sazonovs et al. HLA-DQA1*05 is associated with immunogenicity to anti-TNF therapy

Table of Contents IBD Pharmacogenetics Study Group .............................................................................................. 2 Supplementary Methods ............................................................................................................. 10 Supplementary Tables.................................................................................................................. 13 Table S1. Baseline demographic and clinical characteristics of the 1240 individuals from the PANTS cohort.......................................................................................................................... 173 Table S2. Comparison between different models for the observed effect in the Major Histocompatibility Complex region ......................................................................................... 15 Table S3. Covariates used in the final model .......................................................................... 16 Table S4. Confusion matrix comparing HLA imputation vs HLA typing of HLA-DQA1*05 ..... 17 Supplementary Figures ................................................................................................................ 18 Figure S1. Principal component analysis on imputed data from 1323 individuals in the PANTS study ............................................................................................................................. 18 Figure S2. Flowchart describing the cohort used for the time to immunogenicity genetic analysis………………………………………………………………………………………………………………………….…….19 Figure S3. Power calculation assuming a case-control setting………………………………………...........20 Figure S4. Quantile-quantile plot for Cox proportional hazards model analysis of time to immunogenicity........................................................................................................................ 21 Figure S5. Manhattan plot for Cox proportional hazards model analysis of time to immunogenicity........................................................................................................................ 22 Figure S6. Residual association signal in the MHC region, after conditioning on HLADQA1*05 ................................................................................................................................. 233 Figure S7. HLA-DQA1*05 has a consistent effect on immunogenicity between patients treated with the infliximab originator, Remicade, and its biosimilar CT-P13 ........................ 24 References .................................................................................................................................... 25

1

IBD Pharmacogenetics Study Group All UK gastroenterologists were invited to participate in the PANTS study which was promoted through the UK National Institute for Health Research and the British Society of Gastroenterology. Hospital or Trust name

City

Name

Job Title

Tameside Hospital NHS Foundation Trust

Ashton U Lyne

Dr Vinod Patel

Consultant Gastroenterologist

Basildon and Thurrock University Hospitals NHS Foundation Trust

Basildon

Dr Zia Mazhar

Consultant Gastroenterologist

Hampshire Hospitals NHS Foundation Trust

Basingstoke

Dr Rebecca Saich

Consultant Gastroenterologist

Royal United Hospital

Bath

Dr Ben Colleypriest

Consultant Gastroenterologist

Ulster Hospital

Belfast

Dr Tony C Tham

Consultant Gastroenterologist

University Hospital's Birmingham NHS Foundation Trust

Birmingham

Dr Tariq H Iqbal

Consultant Gastroenterologist

East Lancashire NHS Teaching Trust

Blackburn

Dr Vishal Kaushik

Consultant Gastroenterologist

Blackpool Teaching Hospitals NHS Foundation Trust

Blackpool

Dr Senthil Murugesan

Consultant Gastroenterologist

Bolton NHS Trust

Bolton

Dr Salil Singh

Consultant Gastroenterologist

Royal Bournemouth Hospital

Bournemouth

Dr Sean Weaver

Consultant Gastroenterologist

Bradford Teaching Hospitals Foundation Trust - (St Lukes Hospital &Bradford Royal Infirmary)

Bradford

Dr Cathryn Preston

Consultant Gastroenterologist

Brighton and Sussex University Hospitals NHS Trust

Brighton

Dr Assad Butt

Paediatric Consultant Gastroenterologist

Brighton and Sussex University Hospitals NHS Trust

Brighton

Dr Melissa Smith

Consultant Gastroenterologist

University Hospitals Bristol NHS Foundation Trust

Bristol

Dr Dharamveer Basude

Consultant Paediatric Gastroenterologist

University Hospitals Bristol NHS Foundation Trust

Bristol

Dr Amanda Beale

Consultant Gastroenterologist

2

Frimley Park Hospital NHS Foundation Trust

Camberley

Dr Sarah Langlands

Consultant Gastroenterologist

Frimley Park Hospital NHS Foundation Trust

Camberley

Dr Natalie Direkze

Consultant gastroenterologist

Cambridge University Hospitals NHS Foundation Trust

Cambridge

Dr Miles Parkes

Consultant Gastroenterologist

Cambridge University Hospitals NHS Foundation Trust

Cambridge

Dr Franco Torrente

Consultant Paediatric Gastroenterologist

Cambridge University Hospitals NHS Foundation Trust

Cambridge

Dr Juan De La Revella Negro

Research fellow

North Cumbria University Hospitals NHS Trust

Carlisle

Dr Chris Ewen MacDonald

Consultant Gastroenterologist

Ashford & St Peter's Hospitals NHS Foundation Trust

Chertsey

Dr Stephen M Evans

Consultant Gastroenterologist

St Peter's Hospital

Chertsey

Dr Anton V J Gunasekera

Consultant Gastroenterologist

Ashford & St Peter's Hospitals NHS Foundation Trust

Chertsey

Dr Alka Thakur

Paediatric Consultant

Chesterfield Royal NHS Foundation Trust

Chesterfield

Dr David Elphick

Consultant Gastroenterologist

Colchester Hospital University NHS Foundation Trust

Colchester

Dr Achuth Shenoy

Consultant Gastroenterologist

University Hospitals Coventry and Warwickshire NHS Trust

Coventry

Prof Chuka U Nwokolo

Consultant Gastroenterologist

County Durham and Darlington NHS Foundation Trust

Darlington

Dr Anjan Dhar

Consultant Gastroenterologist & Hon. Clinical Lecturer

Derby Hospital NHS Foundation NHS Trust

Derby

Dr Andrew T Cole

Consultant Gastroenterologist

Doncaster and Bassetlaw Hospitals NHS Foundation Trust

Doncaster

Dr Anurag Agrawal

Consultant Gastroenterologist

Dorset County Hospital NHS Foundation Trust

Dorchester

Dr Stephen Bridger

Consultant Gastroenterologist

3

Dorset County Hospitals Foundation Trust

Dorchester

Dr Julie Doherty

Paediatric Consultant

Dudley Group NHS Foundation Trust

Dudley

Dr Sheldon C Cooper

Consultant Gastroenterologist

Russells Hall Hospital, The Dudley Group NHS Foundation Trust

Dudley

Dr Shanika de Silva

Consultant Gastroenterologist

Ninewells Hospital & Medical School

Dundee

Dr Craig Mowat

Consultant Gastroenterologist

East Sussex Healthcare Trust

Eastborne

Dr Phillip Mayhead

Consultant Gastroenterologist

NHS Lothian

Edinburgh

Dr Charlie Lees

Consultant Gastroenterologist and Honorary Senior Lecturer

NHS Lothian

Edinburgh

Dr Gareth Jones

Research fellow

Royal Devon and Exeter NHS Foundation Trust

Exeter

Dr Tariq Ahmad

Consultant Gastroenterologist

Royal Devon and Exeter NHS Foundation Trust

Exeter

Dr James W Hart

Consultant Paediatrician

Glasgow Royal Infirmary

Glasgow

Dr Daniel R Gaya

Consultant Gastroenterologist

Royal Hospital for Children

Glasgow

Prof Richard K Russell

Consultant Paediatric Gastroenterologist

Royal Hospital for Children

Glasgow

Dr Lisa Gervais

Research fellow

Gloucestershire Hospitals NHS Trust

Gloucester

Dr Paul Dunckley

Consultant Gastroneterologist

United Lincolnshire Hospitals NHS Trust

Grantham

Dr Tariq Mahmood

Consultant Gastroenterologist

James Paget University Hospitals NHS Foundation Trust

Great Yarmouth

Dr Paul J R Banim

Consultant Gastroneterologist

Calderdale and Huddersfield NHS Trust

Halifax

Dr Sunil Sonwalkar

Consultant Gastroenterologist

Princess Alexandra Hospital NHS Trust

Harlow

Dr Deb Ghosh

Consultant Gastroenterologist

Princess Alexandra Hospital NHS Trust

Harlow

Dr Rosemary H Phillips

Consultant Gastroenterologist

4

Hull and East Yorkshire NHS Trust

Hull

Dr Amer Azaz

Paediatric Consultant Gastroenterologist

Hull and East Yorkshire NHS Trust

Hull

Dr Shaji Sebastian

Consultant Gastroenterologist

Airedale NHS Foundation Trust

Keighley

Dr Richard Shenderey

Consultant Gastroenterologist

Crosshouse Hospital

Kilmarnock

Dr Lawrence Armstrong

Consultant Paediatrician

Crosshouse Hospital

Kilmarnock

Dr Claire Bell

Research fellow

The Queen Elizabeth Hospital NHS Foundation Trust

Kings Lynn

Dr Radhakrishnan Hariraj

Consultant Gastroenterologist

Kingston Hospital NHS Trust

Kingston upon Thames

Dr Helen Matthews

Consultant Gastroenterologist

NHS Fife

Kirkcaldy

Dr Hasnain Jafferbhoy

Consultant Gastroenterologist

Leeds Teaching Hospitals NHS Trust

Leeds

Dr Christian P Selinger

Consultant Gastroenterologist

Leeds Teaching Hospitals NHS Trust

Leeds

Dr Veena Zamvar

Paediatric Consultant Gastroenteorlogist

University Hospitals of Leicester NHS Trust

Leicester

Prof John S De Caestecker

Consultant Gastroenterologist

University Hospitals of Leicester NHS Trust

Leicester

Dr Anne Willmott

Paediatric Consultant Gastroenterologist

Mid Cheshire Hospitals NHS Foundation Trust

Leighton

Mr Richard Miller

Research Nurse

United Lincolnshire Hospitals NHS Trust

Lincoln

Dr Palani Sathish Babu

Consultant Gastroenterologist

Alder Hey Childrens Hospital

Liverpool

Dr Christos Tzivinikos

Consultant Paediatric Gastroenterologist

University College London Hospitals NHS Foundation Trust

London

Dr Stuart L Bloom

Consultant Gastroenterologist

Kings College Hospital NHS Foundation Trust

London

Dr Guy Chung-Faye

Consultant Gastroenterologist

Royal London Childrens Hospital, Barts Health NHS Trust

London

Prof Nicholas M Croft

Paediatric Consultant Gastroenterologist

5

Chelsea & Westminster Hospital

London

Dr John ME Fell

Consultant Paediatric Gastroenterologist

Chelsea and Westminster Hospital NHS Foundation

London

Dr Marcus Harbord

Consultant Gastroenterologist

North West London Hospitals NHS Trust

London

Dr Ailsa Hart

Consultant Gastroenterologist

Kings College Hospital NHS Foundation Trust

London

Dr Ben Hope

Consultant Paediatrician

Guys & St Thomas' NHS Foundation Trust

London

Dr Peter M Irving

Consultant Gastroenterologist

Barts and The London NHS Trust

London

Prof James O Lindsay

Consultant Gastroenterologist

Guy's and St Thomas' NHS trust

London

Dr Joel E Mawdsley

Gastroenterology Consultant

Lewisham and Greenwich Healthcare NHS Trust

London

Dr Alistair McNair

Consultant Gastroenterologist

Chelsea and Westminster Hospital NHS Foundation

London

Dr Kevin J Monahan

Consultant Gastroenterologist

Royal Free London NHS Foundation Trust

London

Dr Charles D Murray

Consultant Gastroenterologist

Imperial College Healthcare NHS Trust

London

Prof Timothy Orchard

Consultant Gastroenterologist

St George's Healthcare NHS Trust

London

Dr Thankam Paul

Paediatric Consultant Gastroenterologist

St George's Healthcare NHS Trust

London

Dr Richard Pollok

Reader and Consultant Gastroenterologist

Great Ormond Street Hospital for Children NHS Foundation Trust

London

Dr Neil Shah

Consultant Gastroenterologist

North West London Hospitals NHS Trust

London

Dr Sonia Bouri

Research fellow

The Luton & Dunstable University Hospital

Luton

Dr Matt W Johnson

Consultant Gastroenterologist

Luton and Dunstable Hospital Foundation Trust

Luton

Dr Anita Modi

Paediatric Consultant with Allergy and Gastroenterology interest

6

The Luton & Dunstable University Hospital

Luton

Dr Kasamu Dawa Kabiru

Research fellow

Maidstone and Tunbridge Wells NHS Trust

Maidstone

Dr B K Baburajan

Consultant Gastroenterologist

Maidstone and Tunbridge Wells NHS Trust

Maidstone

Prof Bim Bhaduri

Paediatric Consultant Gastroenterologist

Manchester University Hospitals NHS Foundation Trust

Manchester

Dr Andrew Adebayo Fagbemi

Consultant Gastroenterologist

Central Manchester University Hospitals NHS Foundation Trust

Manchester

Dr Scott Levison

Consultant Gastroenterologist

The Pennine Acute Hospitals NHS Trust

Manchester

Dr Jimmy K Limdi

Consultant Gastroenterologist

Manchester University NHS Foundation Trust, Wythenshawe Hospital

Manchester

Dr Gill Watts

Consultant Gastroenterologist

Sherwood Forest Hospitals NHS Foundation Trust

Mansfield

Dr Stephen Foley

Consultant Gastroenterologist

South Tees Hospital NHS Foundation Trust

Middlesbrough

Dr Arvind Ramadas

Consultant Gastroenterologist

Milton Keynes Hospital NHS Foundation Trust

Milton Keynes

Dr George MacFaul

Consultant Gastroenterologist

Newcastle Upon Tyne Hospital Trust

Newcastle

Dr John Mansfield

Consultant Gastroenterologist

Isle of Wight NHS Foundation Trust

Newport

Dr Leonie Grellier

Consultant Gastroenterologist

Norfolk & Norwich University Hospital NHS Foundation Trust

Norwich

Dr Mary-Anne Morris

Consultant Paediatric Gastroenterologist

Norfolk & Norwich University Hospital NHS Foundation Trust

Norwich

Dr Mark Tremelling

Consultant Gastroenterologist

Nottingham University Hospitals NHS Trust

Nottingham

Prof Chris Hawkey

Consultant Gastroenterologist

Nottingham University Hospitals NHS Trust

Nottingham

Dr Sian Kirkham

Consultant Paediatric Gastroenterologist

Nottingham University Hospitals NHS Trust

Nottingham

Dr Charles PJ Charlton

Consultant gastroenterologist

7

Oxford University Hospitals NHS Foundation Trust

Oxford

Dr Astor Rodrigues

Paediatric Consultant Gastroenterologist

Oxford University Hospitals NHS Trust

Oxford

Prof Alison Simmons

Consultant Gastroenterologist

Plymouth Hospitals NHS Trust

Plymouth

Dr Stephen J Lewis

Consultant Gastroenterologist

Poole Hospital NHS Foundation Trust

Poole

Dr Jonathon Snook

Consultant Gastroenterologist

Poole Hospital NHS Foundation Trust

Poole

Dr Mark Tighe

Paediatric Consultant with interest in Oncology and Gastroenterology

Portsmouth Hospitals NHS Trust

Portsmouth

Dr Patrick M Goggin

Consultant Gastroenterologist

Royal Berkshire NHS Foundation Trust

Reading

Dr Aminda N De Silva

Consultant Gastroenterologist

Salford Royal NHS Foundation Trust

Salford

Prof Simon Lal

Consultant Gastroenterologist

Shrewsbury and Telford Hospital NHS Trust

Shrewsbury

Dr Mark S Smith

Consultant Gastroenterologist

South Tyneside NHS Foundation Trust

South Shields

Dr Simon Panter

Consultant Gastroenterologist

Southampton University Hospitals NHS Trust

Southampton

Dr Fraser Cummings

Consultant Gastroenterologist

Southampton University Hospitals NHS Trust

Southampton

Dr Suranga Dharmisari

Research fellow

East and North Herts NHS Trust

Stevenage

Dr Martyn Carter

Consultant Gastroenterologist

NHS Forth Valley

Stirling

Dr David Watts

Consultant Gastroenterologist

Stockport NHS foundation Trust

Stockport

Dr Zahid Mahmood

Consultant Gastroenterologist

North Tees and Hartlepool NHS Foundation Trust

Stockton

Dr Bruce McLain

Paediatric Consultant Gastroenterologist

University Hospitals of North Staffordshire

Stoke-on Trent

Dr Sandip Sen

Consultant Gastroenterologist

University Hospitals of North Midlands NHS Trust

Stoke-on-Trent

Dr Anna J Pigott

Consultant Paediatric Gastroenterologist

8

City Hospitals Sunderland NHS Foundation Trust

Sunderland

Dr David Hobday

Consultant Gastroenterologist

Taunton and Somerset NHS Foundation Trust

Taunton

Dr Emma Wesley

Consultant Gastroenterologist

South Devon Healthcare NHS Foundation Trust

Torquay

Dr Richard Johnston

Consultant Gastroenterologist

South Devon Healthcare NHS Foundation Trust

Torquay

Dr Cathryn Edwards

Consultant gastroenterologist

Royal Cornwall Hospitals NHS Trust

Truro

Dr John Beckly

Consultant Gastroenterologist

Mid Yorkshire Hospitals NHS Trust

Wakefield

Dr Deven Vani

Consultant Physician & Gastroenterologist

Warrington& Halton NHS Foundation

Warrington

Dr Subramaniam Ramakrishnan

Consultant Gastroenterologist

West Hertfordshire Hospitals NHS Trust

Watford

Dr Rakesh Chaudhary

Consultant Gastroenterologist

Sandwell and West Birmingham Hospitals NHS Trust

West Bromwich

Dr Nigel J Trudgill

Consultant Gastroenterologist

Sandwell and West Birmingham Hospitals NHS Trust

West Bromwich

Dr Rachel Cooney

Consultant gastroenterologist

Weston Area Health NHS Trust

Weston-SuperMare

Dr Andy Bell

Consultant Gastroenterologist

Royal Albert Edward Infirmary, Wrightington, Wigan & Leigh NHS Foundation Trust

Wigan

Dr Neeraj Prasad

Consultant Gastroenterologist

Hampshire Hospitals NHS Foundation Trust

Winchester

Dr John N Gordon

Consultant Gastroenterologist

Royal Wolverhampton Hospitals NHS Trust

Wolverhampton

Prof Matthew J Brookes

Consultant Gastroenterologist

Western Sussex Hospitals NHS Trust

Worthing

Dr Andy Li

Consultant Gastroenterologist

Yeovil District Hospital NHS Foundation Trust

Yeovil

Dr Stephen Gore

Consultant Gastroenterologist

9

Supplementary Methods Measurement of drug and antidrug antibody levels 1

As previously described, serum infliximab and adalimumab drug levels were analyzed on the Dynex (Chantilly Virginia, USA) DS2 automated Enzyme-Linked ImmunoSorbent Assay (ELISA) platform, using the Immundiagnostik (Immundiagnostik AG, Bensheim, Germany) IDKmonitor® drug (K9655 infliximab drug level and K9657 adalimumab drug level) and total antibody ELISA assays (K9654 infliximab total anti-drug antibody and K9651 adalimumab total anti-drug antibody). These assays allow quantitative determination of free infliximab and adalimumab using a sandwich ELISA technique. The IDK monitor infliximab drug level assay has a measuring range of 0.8-45mg/L, with an intra-assay CV of <9.7% and an inter-assay CV of <11.0%. The IDK monitor adalimumab drug level assay has a measuring range of 0.8-45mg/L, with an intra-assay CV of <2.6% and an inter-assay CV of <13.0%.

Positive anti-drug antibody status was defined in line with the manufacturer's recommendations as a concentration ≥10 AU/mL, irrespective of drug level. Serum infliximab and adalimumab total anti-drug antibody levels were analysed on the Dynex DS2 automated ELISA platform using the IDK monitor infliximab and adalimumab total anti-drug-antibody ELISA assays. These assays allow the detection of total antibodies against infliximab; measuring free and bound antibodies against infliximab. The IDK monitor Infliximab total ADA assay has a measuring range of 10-400 AU/mL, with an intra-assay CV of <4.7% and an inter-assay CV of <12.6%. The IDK monitor adalimumab drug level assay has a measuring range of 10-200 AU/mL, with an intraassay CV of <7.2% and an inter-assay CV of <5.9%. All assays, including drug and anti-drug antibody levels, were tested for stability. In independent experiments, we confirmed that this cut-off corresponds to the 98th centile of the anti-drug antibody titre distribution in more than 500 drug-naïve controls.

Genotyping, quality control and genotype imputation Genotyping of DNA was undertaken using the Illumina CoreExome microarray, with genotype calls made using 2

optiCall. Pre-imputation quality control (QC) of samples and single nucleotide polymorphisms (SNPs) was 3

performed as previously described. We excluded individuals of non-European ancestry (identified using

10

principal component analysis), one individual from related pairs (defined as a pi-hat>0.1875), and individuals with an outlying number of missing or heterozygous genotypes. SNP genotypes were imputed via the Sanger Imputation Service using the Haplotype Reference Consortium 4

(HRC) panel. Following imputation we ran additional variant and sample QC procedures. We calculated the first 15 principal components (PC) for the 1323 samples that passed the QC criteria (Figure S1A in the Supplementary Appendix). The Tracy-Widom test showed that only the first principal component explained a significant proportion of the variance. In addition, we confirmed the European ancestry of our cohort by performing principal component analyses with 3500 patient samples from the 1000 Genomes Project (1KGP) 5

(Figure S1B and C in the Supplementary Appendix). Following standard practices, we discarded poorly imputed SNPs, defined as having an information content metric (INFO) score <0.4. We removed SNPs -10

significantly deviated from Hardy-Weinberg equilibrium (P<1×10 ), with a call-rate ≤0.95 or minor allele frequency ≤1%, leaving 7,578,947 variants. PCA and genotype filtering were performed using the Hail framework (version 0.1).

6

7

HLA imputation was carried out using the HIBAG package in R, using pre-fit classifiers trained specifically for the CoreExome genotyping microarray on individuals of European ancestry. HLA types were imputed at 2-, and 4-digit resolution for the following loci: HLA-A, HLA-C, HLA-B, HLA-DRB1, HLA-DQA1, HLA-DQB1, and HLADPB1. In addition, we obtained amino acid sequences for all the imputed HLA alleles, using the IPD-IMGT/HLA 8

database. Following the recommended best practices for HIBAG, 17 HLA-allele and amino acid calls with posterior probability <0.5 were set to missing for the given individual. To confirm our imputation and genetic association we carried out long-read sequencing of the HLA (Histogenetics, New York, USA).

Statistical Analyses All analyses were two tailed and p-values <0.05 were considered significant. Univariable analyses using Fisher’s exact and Mann-Whitney U tests were used to identify differences in baseline characteristics between infliximab- and adalimumab-treated patients. Mann-Whitney U tests were used to compare serum levels of anti-drug antibodies at week 52 stratified by anti-TNF drug and immunomodulator use. Rates of immunogenicity were estimated using the Kaplan-Meier method, and comparative analyses were performed using univariable and multivariable Cox proportional hazards regression. For immunogenicity, patients were censored after their last drug and anti-drug antibody measurement or at week 54. Optimal thresholds for drug

11

levels were determined graphically by plotting outcome against intervals of drug level and looking for the threshold beyond which further increases were not associated with improvement in outcome.

Both Akaike information criterion (AIC) and Bayesian information criterion (BIC), calculated using the ‘survival’ R package, were used for model comparison across non-nested models; these estimates the goodness of fit while penalising for the number of parameters to prevent overfitting. Lower values of AIC and/or BIC indicate a better fit (Table S2 in the Supplementary Appendix).

Cox proportional hazards regression was used to test genetic variants and for association with time to antidrug antibody development, assuming an additive genetic effect. Patients who did not develop immunogenicity during the study were censored at the point of last observation. The association analyses were performed using a likelihood ratio test with sex, drug type (infliximab or adalimumab), immunomodulator use, and the first principal component as covariates (Table S3 in Supplemental Appendix), 9

with the SurvivalGWAS_SV software. Kaplan-Meier estimates were produced using the Lifelines package for Python.

12

Supplementary Tables Variable

Level

infliximab n = 742

Sex (male)

adalimumab n = 498

p

353 (47.6%)

234 (47.0%)

31.3 (21.2 - 46.0)

37.6 (28.7 - 50.3)

Disease duration (years)

2.6 (0.7 - 9.8)

3.1 (0.8 - 11.5)

Age at diagnosis (years)

24.4 (16.4 - 35.9)

29.5 (21.8 - 41.8)

L1

207 (28.2%)

157 (32.0%)

L2

190 (25.9%)

108 (22.0%)

L3

332 (45.2%)

223 (45.4%)

L4

6 (0.8%)

3 (0.6%)

78 (10.6%)

23 (4.7%)

1.8 × 10

B1

455 (61.7%)

277 (56.3%)

1.9 × 10

B2

196 (26.6%)

184 (37.4%)

B3

86 (11.7%)

31 (6.3%)

104 (14.0%)

40 (8.0%)

339 (45.7%)

196 (39.4%)

mercaptopurine

57 (7.7%)

32 (6.4%)

methotrexate

50 (6.7%)

24 (4.8%)

2 (0.3%)

0 (0.0%)

294 (39.6%)

246 (49.4%)

Steroids

220 (29.6%)

133 (26.7%)

0.275

Previous resectional surgery

170 (22.9%)

122 (24.5%)

0.539

6.0 (3.0 - 9.0)

5.0 (3.0 - 8.0)

0.224

Age (years)

Montreal location

Montreal L4 Montreal behaviour

Perianal Immunomodulator

azathioprine

tacrolimus none

HBI sPCDAI

25.0 (10.0 - 50.0)

BMI

23.1 (20.0 - 27.2)

13

24.6 (21.6 - 28.4)

0.862 -11

1.1 × 10

0.033 -11

1.7 × 10

0.334

-4

-5

0.009

-7

6.5 × 10

Hemoglobin (g/L)

-10

126.0 (115.8 - 136.0)

132.0 (122.0 - 141.0)

8.0 (6.2 - 10.3)

7.9 (6.3 - 9.8)

343.0 (281.0 - 412.8)

312.0 (256.0 - 386.0)

39.0 (34.0 - 42.0)

39.0 (36.0 - 43.0)

CRP (mg/L)

8.0 (3.0 - 21.0)

6.0 (2.0 - 13.0)

5.1 × 10

Fecal calprotectin (μg/g)

402 (149 - 855)

292 (130 - 603)

1.4 × 10

9

White cell count (×10 /L) 9

Platelet count (×10 /L) Albumin (g/L)

2.2 × 10

0.324 -6

3.4 × 10

0.051 -5

Table S1. Baseline demographic and clinical characteristics of the 1240 individuals from the PANTS cohort

14

-4

Genetic effect included in the model

HR

95% CIs

P

Akaike information criterion (AIC)

Bayesian information criterion (BIC)

AA-107-ile (dominant)

1.90

1.60

2.26

4.00×10

-13

6649.45

6675.04

AA-175-lys (dominant)

1.90

1.60

2.26

4.00×10

-13

6649.45

6675.04

DQA1*05 (dominant)

1.90

1.60

2.25

5.88×10

-13

6652.12

6677.73

AA-107-ile (additive)

1.59

1.40

1.80

1.50×10

-11

6656.55

6682.14

AA-175-lys (additive)

1.59

1.40

1.80

1.50×10

-11

6656.55

6682.14

DQA1*05 (additive)

1.58

1.39

1.80

1.94×10

-11

6659.07

6684.67

DQA1*05:01 and DQA1*05:05 (both additive)

1.61 1.54

1.36 1.28

1.90 1.85

1.19×10 -5 1.37×10

-7

6659.50

6690.20

DRB1*03 (additive)

1.58

1.34

1.85

2.43×10

-7

6661.30

6686.89

DRB1*03 (dominant)

1.67

1.38

2.03

1.48×10

-7

6662.46

6688.06

rs2097432 (additive)

1.69

1.48

1.94

4.24×10

-13

6674.24

6699.85

rs2097432 (dominant)

1.92

1.61

2.28

4.55×10

-13

6674.34

6699.95

DQA1*05:01 (additive)

1.57

1.33

1.85

4.24×10

-7

6676.74

6702.33

DQA1*05:01 (dominant)

1.67

1.38

2.03

5.49×10

-7

6677.26

6702.85

DQA1*05:05 (dominant)

1.67

1.27

1.90

3.39×10

-5

6684.97

6710.56

DQA1*05:05 (additive)

1.48

1.24

1.78

5.54×10

-5

6685.84

6711.43

“Null” model with no genotype parameter

–

–

–

6725.43

6745.92

–

Table S2. Comparison between different models for the observed effect in the MHC region Models are sorted by Akaike information criterion (AIC) score (column 6), which is a measure of model fit to the data. All models included immunomodulator status, drug type, sex, and first principal component as covariates (see Supplemental Methods and Table S3 in Supplemental Appendix for details). Hazard ratios, Confidence intervals, and p-values are calculated using SurvivalGWAS_SV and AIC is calculated using R surv package, as described in Methods. The effect assumed for the main genetic effect (additive or dominant) is shown in parentheses in the first column. Hazard ratios with 95% Confidence Intervals for the genetic effect are shown in columns 2-4, and the respective p-values in column 5. The last column contains the BIC score for each model. The ranking of the models is identical whether AIC or BIC is used.

15

Covariate

HR

95% CIs

P -34

Drug (0 – infliximab; 1 – adalimumab)

3.27

2.67

4.02

2.82×10

Immunomodulators (0 – yes; 1 – no)

2.41

2.02

2.87

1.84×10

HLA-DQA1*05 (dominant)

1.90

1.60

2.25

5.88×10

Sex (0 – female; 1 – male)

0.93

0.78

1.11

0.41

PC1 (continuous)

0.33

0.03

3.60

0.38

Table S3. Covariates used in the final model

16

-22

-13

Typed

Imputed

Homozygote DQA1*05

Heterozygote DQA1*05

No DQA1*05 Alleles

Homozygote DQA1*05

75

0

1

Heterozygote DQA1*05

0

424

0

No DQA1*05 Alleles

0

0

772

Table S4. Confusion matrix comparing HLA imputation versus HLA typing of HLA-DQA1*05 The imputation procedure from genotype data is described in the Supplementary Methods. Comparison performed amongst 1272 of the 1323 individuals in the PANTS cohort.

17

Supplementary Figures

Figure S1. Principal component analysis (PCA) on imputed data from 1323 individuals in the PANTS study PANEL A: PCA on the PANTS cohort alone. Blue dots show individuals who develop immunogenicity and orange dots show individuals who did not. First principal component was not associated with immunogenicity status (P=0.99). PANELS B, C: PCA analysis of the 1323 individuals from the PANTS cohort together with the 3500 individuals from the 1000 Genomes Project (1KGP) cohort. Panel B: PCA confirmed that PANTS samples clustered together with the individuals of European ancestry from the 1KGP. Panel C: A detailed (zoomed-in) view of PANTS samples plotted together with European samples from the 1KGP.

18

Figure S2. Flowchart describing the cohort used for the time to immunogenicity genetic analysis Abbreviations: UC = ulcerative colitis, IBD-U: inflammatory bowel disease unclassified, anti-TNF: anti-tumor necrosis factor, CRP: C-reactive protein, INFO: information content metric

19

Figure S3. Power calculation assuming a case-control setting The following parameters were used: alpha = 5e-8, prevalence of trait = 44% (rate of immunogenicity at 12 months in our cohort). In a case-control setting, we have 100% power to detect variants of similar frequency and effect as rs2097432 (MAF=0.27 in gnomAD, OR=1.89) amongst the patients developing immunogenicity.

20

Figure S4. Quantile-quantile plot for Cox proportional hazards model analysis of time to immunogenicity Using a regression model implemented in the GenABEL package in R, we estimated the inflation factor λ to be -5 1.02 (SE=2 × 10 ), suggesting a good fit to the uniform distribution. See Table S3 in the Supplementary Appendix for the covariates used in this model.

21

Figure S5. Manhattan plot for Cox proportional hazards model analysis of time to immunogenicity

22

Figure S6. Residual association signal in the MHC region, after conditioning on HLA-DQA1*05 Midpoint positions of the HLA alleles across the MHC region are shown in red on the x-axis. The SNP most strongly associated with time to immunogenicity (rs2097432) during the initial analysis is marked with a red dot. No other SNPs passed the genome-wide significance threshold (red dashed line), suggesting that HLADQA1*05 explains the chromosome 6 signal (Figure 1).

23

Figure S7. HLA-DQA1*05 has a consistent effect on immunogenicity between patients treated with the infliximab originator, Remicade, and its biosimilar CT-P13

24

References 1.

Kennedy NA, Heap GA, Green HD, et al. Predictors of anti-TNF treatment failure in anti-TNF-naive patients with active luminal Crohn’s disease: a prospective, multicentre, cohort study. Lancet Gastroenterol Hepatol 2019;1253(19):1–13.

2.

Shah TS, Liu JZ, Floyd JAB, et al. OptiCall: A robust genotype-calling algorithm for rare, low-frequency and common variants. Bioinformatics 2012;28(12):1598–603.

3.

de Lange KM, Moutsianas L, Lee JC, et al. Genome-wide association study implicates immune activation of multiple integrin genes in inflammatory bowel disease. Nat Genet 2017;49(2):256–61.

4.

McCarthy S, Das S, Kretzschmar W, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat Genet 2016;48(10):1279–83.

5.

1000 Genomes Project Consortium, Auton A, Brooks LD, et al. A global reference for human genetic variation. Nature [Internet] 2015;526(7571):68–74. Available from: http://www.ncbi.nlm.nih.gov/pubmed/26432245

6.

Neal Lab (Broad Institute of MIT and Harvard and MGH. Hail [Internet]. [cited 2018 Oct 3];Available from: https://github.com/hail-is/hail

7.

Zheng X, Shen J, Cox C, et al. HIBAG--HLA genotype imputation with attribute bagging. Pharmacogenomics J [Internet] 2014;14(2):192–200. Available from: http://www.ncbi.nlm.nih.gov/pubmed/23712092

8.

Robinson J, Mistry K, McWilliam H, Lopez R, Marsh SGE. IPD--the Immuno Polymorphism Database. Nucleic Acids Res [Internet] 2010;38(Database issue):D863-9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/19875415

9.

Syed H, Jorgensen AL, Morris AP. SurvivalGWAS_SV: software for the analysis of genome-wide association studies of imputed genotypes with “time-to-event” outcomes. BMC Bioinformatics [Internet] 2017;18(1):265. Available from: http://www.ncbi.nlm.nih.gov/pubmed/28525968

25