Extended Analysis Identifies Drug-Specific Association of 2 Distinct HLA Class II Haplotypes for Development of Immunogenicity to Adalimumab and Infliximab

Abstract

The global occurrence of inflammatory bowel diseases, including Crohn’s disease (CD), is rising, with prevalence of 0.5% to 1.0% reported in Europe and North America.1Steenholdt C. et al.Inflamm Bowel Dis. 2016; 22: 1999-2015Crossref PubMed Scopus (82) Google Scholar Biological therapies are a mainstay of patient management, with monoclonal antibodies against tumor necrosis factor (TNF), infliximab and adalimumab, widely used as first-line agents. Highly effective in some individuals, continued use of these drugs is limited by primary nonresponse (10%–40%) or secondary immune-mediated loss of response (24%–46%) to treatment in the first year.2Kennedy N.A. et al.Lancet Gastroenterol Hepatol. 2019; 4: 341-353Abstract Full Text Full Text PDF PubMed Scopus (407) Google Scholar HLA class II polymorphisms have been implicated as contributory to disease progression in numerous autoimmune diseases. A recent study based on data from the UK nationwide analysis of anti-TNF response in CD implicates HLA-DQA1∗05 in the development of antibodies against both anti-TNF biologics, with important clinical implications.3Sazonovs A. et al.Gastroenterology. 2020; 158: 189-199Abstract Full Text Full Text PDF PubMed Scopus (223) Google Scholar In this study, HLA-DQA1∗05 represents a group of allelic variants, each encoding a distinct protein: HLA-DQA1∗05:01, HLA-DQA1∗05:03 (A182S), and HLA-DQA1∗05:05 (A11T, signal peptide). The HLA class II gene locus that includes HLA-DQA1 gene is a complex locus, encoding functional and highly polymorphic molecules that present peptide antigens to T-helper cells, which provide stimulus to B-cell maturation and antibody formation. HLA class II molecules are heterodimers consisting of an alpha (A loci) and beta chain (B loci). The combination of the alpha chain with a particular beta chain, as well as amino acid differences within and outside the peptide binding groove, can affect trafficking and ligand binding (Figure 1A). It is further known that within this region a high level of linkage disequilibrium results in the frequent occurrence of extended haplotypes,4de Bakker P.I. et al.Nat Genet. 2006; 38: 1166-1172Crossref PubMed Scopus (618) Google Scholar directly linking multiple HLA allele variants across the 3 major class II genes: DR, DQ, and DP.5Price P. et al.Immunol Rev. 1999; 167: 257-274Crossref PubMed Scopus (482) Google Scholar With this biological context in mind, we hypothesized that, due to the different molecular composition of adalimumab and infliximab, their antigenic presentation and resulting potential to induce immunogenicity would differ, and be directly dependent not only on the alpha chain, but also the pairing beta chain, which mostly directs antigen-binding specificity. Using the PANTS (Personalising Anti-TNF Therapy in Crohn’s Disease) discovery cohort,3Sazonovs A. et al.Gastroenterology. 2020; 158: 189-199Abstract Full Text Full Text PDF PubMed Scopus (223) Google Scholar 1240 biologic-naïve patients with CD starting either infliximab or adalimumab, we explored the cohort at 4-digit resolution, running pairwise Pearson correlations between each HLA-DQA1∗05 variant and extended haplotypes to determine whether they should be considered as individual factors. We split the original data based on treatment, and applied Cox regression to each cohort to examine the significance of each HLA variant to immunogenicity conditioned on treatment method. Immunogenicity is defined as an anti-drug antibody concentration of >10 AU/mL, irrespective of drug level, at any time point across the initial 12-month study and 2-year follow-up. Antidrug antibodies were measured in all serum samples, at every visit, for every patient.3Sazonovs A. et al.Gastroenterology. 2020; 158: 189-199Abstract Full Text Full Text PDF PubMed Scopus (223) Google Scholar Pearson correlation analysis indicates weak correlation among the 3 HLA-DQA1∗05 protein variants (HLA-DQA1∗05:01 vs HLA-DQA1∗05:03 ρ = −0.0219; HLA-DQA1∗05:01 vs HLA-DQA1∗05:05 ρ = −0.0921; HLA-DQA1∗05:03 vs HLA-DQA1∗05:05 ρ = −0.0196), indicating these variants are biologically unrelated and should be considered separately. Analysis on HLA-DQA1∗05:03 was not progressed, as it exists in only 1 individual in the PANTS cohort. Conversely, a strong correlation is evident between alleles HLA-DQA1∗05:01, HLA-DQB1∗02:01, and HLA-DRB1∗03:01 (HLA-DQA1∗05:01 vs HLA-DQB1∗02:01 ρ = 1; HLA-DQA1∗05:01 vs HLA-DRB1∗03:01 ρ = 0.9976), indicating HLA alpha chain HLA-DQA1∗05:01 pairs exclusively with HLA beta chain HLA-DQB1∗02:01, and almost exclusively with HLA-DRB1∗03:01, confirming the known linkage disequilibrium across this region in this cohort.5Price P. et al.Immunol Rev. 1999; 167: 257-274Crossref PubMed Scopus (482) Google Scholar HLA-DQA1∗05:05 is also strongly correlated with HLA-DQB1∗03:01 and HLA-DRB1∗11:01 (HLA-DQA1∗05:05 vs HLA-DQB1∗03:01 ρ = 0.6782; HLA-DQA1∗05:05 vs HLA-DRB1∗11:01 ρ = 0.6481), noting that HLA-DQB1∗03:01 occurs in combination with alpha chains other than HLA-DQA1∗05:05 but not vice versa. Cox regression in the drug-specific cohorts demonstrates alleles HLA-DQA1∗05:01, HLA-DQB1∗02:01, and HLA-DRB1∗03:01 have no association with adalimumab immunogenicity (hazard ratio [HR] 1.071, P = .7555; HR 1.061, P = .7869; and HR 1.058, P = .7978, respectively). Conversely, alleles HLA-DQA1∗05:05, HLA-DQB1∗03:01, and HLA-DRB1∗11:01 are all significantly associated with development of adalimumab immunogenicity (HR 2.164, P = .000108; HR 1.489, P = .031163; and HR 2.912, P = 2.87262e-05, respectively). Alternatively, HLA-DQA1∗05:01, HLA-DQB1∗02:01, and HLA-DRB1∗03:01 demonstrate highly significant association to infliximab immunogenicity (HR 1.928, P = 3.071547e−09; HR 1.926, P = 2.919311e−09; and HR 2.008, P = 4.856836e−10, respectively), whereas HLA-DQA1∗05:05, HLA-DQB1∗03:01, and HLA-DRB1∗11:01 show some significance, but the association is reduced (HR 1.406, P = .004361; HR 1.278, P = .018322; and HR 1.377, P = .056476, respectively). Finally, we observed additional associations of immunogenicity for each drug in the variant of the HLA-DRB1 gene: HLA-DRB1∗03:01 (infliximab, HR 2.008, P = 4.856836e−10) and HLA-DRB1∗11:01 (adalimumab, HR 2.912, P = 2.87262e−05). Our analysis demonstrates that grouping together different HLA-DQA1∗05 alleles has important limitations that critically impact the interpretation of these data. Splitting individual alleles and extended haplotypes by treatment option (Figure 1), we show a clear difference between their effects. HLA-DQA1∗05:01 and its extended haplotype are significantly associated with immunogenicity against infliximab but NOT adalimumab (Figure 1B), whereas HLA-DQA1∗05:05 and its extended haplotype are significantly associated with immunogenicity against adalimumab AND, less strongly, infliximab (Figure 1C). This distinction is important in the context of potential clinical translation. These data demonstrate no evidence for the development of adalimumab immunogenicity for patients who possess only the HLA-DQA1∗05:01 variant and extended haplotype. Therefore, if one relies on 2D resolution of the HLA-DQA1∗05 supertype, 144 individuals in the study cohort (11.6%) currently receiving adalimumab with the HLA-DQA1∗05:01 extended haplotype would be mischaracterised as at risk of immunogenicity. Finally, our analysis shows HLA-DQA1∗05:01, HLA-DQB1∗02:01, and HLA-DRB1∗03:01 exhibit 100% linkage disequilibrium, making it impossible to determine causality based on Cox regression analysis (Figure 1B).5Price P. et al.Immunol Rev. 1999; 167: 257-274Crossref PubMed Scopus (482) Google Scholar, 6Brzyski D. et al.Genetics. 2017; 205: 61-75Crossref PubMed Scopus (72) Google Scholar, 7Lee J.C. et al.Nat Genet. 2017; 49: 262-268Crossref PubMed Scopus (184) Google Scholar Moreover, the alleles HLA-DQA1∗05:05, HLA-DQB1∗03:01, and HLA-DRB1∗11:01 are similarly indistinguishable (Figure 1C). These data highlight the complexities of deriving biological and clinical understanding of the mechanisms underlying immunogenicity to anti-TNF therapy from genetic studies of the HLA region. The effects of linkage disequilibrium and the heterodimeric nature of the antigen-presenting molecules are all important considerations. In the context of class II molecular structure and function, the contributions of both the alpha and beta chains must be considered (Figure 1A). Furthermore, the heterogeneity of response to either drug suggests potentially distinct underlying mechanisms of immunogenicity. We conclude association studies within the HLA gene locus require single HLA variant analysis to allow accurate interpretation of the data for biological understanding and potential clinical application. The authors acknowledge Aleksejs Sazonovs, Loukas Moutsianas, Carl A. Anderson, Tariq Ahmad, and the PANTS Consortium for sharing their discovery cohort data set. The authors gratefully acknowledge Paul Klenerman for helpful discussions. The figure was partially created using BioRender. Rebecca D. Powell Doherty, PhD (Conceptualization: Supporting; Formal analysis: Supporting; Investigation: Lead; Methodology: Equal; Project administration: Lead; Writing – original draft: Lead; Writing – review & editing: Equal). Hanqing Liao, PhD (Data curation: Lead; Formal analysis: Lead; Methodology: Equal; Writing – review & editing: Equal). Jack J. Satsangi, DPhil (Conceptualization: Supporting; Project administration: Supporting; Supervision: Lead; Writing – review & editing: Equal). Nicola M.N. Ternette, PhD (Conceptualization: Equal; Project administration: Supporting; Methodology: Equal; Supervision: Supporting; Writing – review & editing: Equal). Underpowered PANTS: A Response to the Conclusions of “Extended Analysis Identifies Drug-Specific Association of Two Distinct HLA Class II Haplotypes for Development of Immunogenicity to Adalimumab and Infliximab”GastroenterologyVol. 160Issue 1PreviewPowell Doherty and Liao1 present a post hoc analysis of our genome-wide association study of immunogenicity to the anti-tumor necrosis factor (TNF) drugs adalimumab and infliximab.2 In our original study, we decided a priori to maximize power to detect genetic variants associated with immunogenicity to anti-TNFs by including patients treated with either drug in the primary analysis. We reported a significant association between the allele group HLA-DQA1∗05 and time to development of antibodies to anti-TNF drugs (hazard ratio [HR], 1.90; 95% confidence interval [CI], 1.60–2.25; P = 5.88 × 10–13). Full-Text PDF

Topics & Concepts