Several biomarkers like tumour mutational burden, PDL-1 expression, microsatellite instability (MSI) status, neoantigen load and HLA status have been recently evaluated to determine their utility to predict response to ICI therapy[6, 14, 22]. We have previously published on the utility of circulating tumour cell PDL1 expression  and serum vascular endothelial growth factor (VEGF) levels  as potential blood-based predictive biomarkers of response to anti-PD1 therapy in advanced melanoma. Here we report our findings of HLA genotyping as a predictive biomarker from a cohort of advanced melanoma patients treated with pembrolizumab. Overall, we found HLA-B27 to be a strong independent predictor of poor outcome to therapy.
Our results failed to demonstrate a predictive association between HLA Class I or II genotypes and survival in advanced melanoma cohort treated with pembrolizumab. Our findings are in contrast to the conclusions made by Chowell et al. , who found better treatment outcomes for patients with HLA Class I heterozygosity. This cohort was markedly heterogeneous, including patients with advanced melanoma or lung cancer in addition to some other tumour types. The sample size of their cohort was substantially larger (n = 1535, with 269 melanoma patients in cohort 1) compared to our cohort (n = 113). Notably, the Chowell et al. study included different ICI therapies with patients receiving anti-PD1 and/or anti-CTLA4 therapy primarily in the 2nd or 3rd line. In contrast, we focused on melanoma patients treated with anti-PD1 (pembrolizumab), primarily in the 1st line setting (73%). This distinction is of clinical relevance as single agent anti-PD1 is administrated commonly as first line therapy for advanced melanoma as well as in the adjuvant settings[26, 27].
Studies evaluating the utility of HLA status as a predictive biomarker have yielded mixed results. In addition to the study by Chowell et al, two other studies have recently been published in advanced non-small cell lung cancer. The findings from Abed et al. suggest that homozygosity at one or more HLA-I loci was associated with shorter OS and PFS in patients with advanced non-small cell lung cancer with PDL1 score ≥ 50%. The study also found that carriers of HLA-A02 supertype had better survival outcomes. However, in contrast to their findings, Negrao et al found that HLA class I genotype did not correlate with survival in advanced non-small cell lung cancer patients treated with ICI therapy. They concluded that the impact of HLA class I diversity may be disease specific and that tumour genomic and immune markers might carry stronger predictive significance.
An important finding of our study was the association of the HLA-B27 supertype with pembrolizumab treatment outcomes. Albeit not statistically significant, HLA-B62 and B44 supertypes were found to be associated with shorter and longer PFS, respectively, similar to the report by Chowell et al.. However, HLA-B27 positivity demonstrated a strong negative association with both PFS and OS, even after controlling for multiple cofounder variables. Notably, this novel finding was validated in an independent cohort, again focusing on anti-PD1 treated melanoma patients.
HLA-B27 is associated with the presence of seronegative spondyloarthropathies , such as ankylosing spondylitis (AS), psoriatic arthritis, and Reiter's syndrome (reactive arthritis). HLA B27 is also strongly associated with acute anterior uveitis. These associations are amongst the strongest of any HLA antigen to a human disease, although the pathogenetic mechanism(s) involved remain unknown.
Another line of evidence suggests that HLA-B27 affects the composition of the gut microbiome, which in turn can modify the immune system and thereby affect health and disease . This is particularly relevant, given the strong evidence that gut microbiota not only serves a biomarker of response to ICI , but its modification through faecal microbiota transplant can serve as a companion intervention to improve anti-PD1 therapy efficacy. Thus, further studies should seek to evaluate patients’ microbiomes and ICI outcomes in relation to HLA-27 supertype.
There are several limitations of our study. Our sample size is small meaning we might not have been able to able to appreciate small differences and determine the true magnitude of impact of HLA status on response to ICI therapy. Our multivariate analysis did not include other confounding factors reported to regulate anti-tumour immune response including tumour immune signature/score, tumour mutational burden or gut microbiome[6, 34]. We did not have access to tumour tissue and hence were unable to calculate the impact of somatic loss of HLA heterozygosity on survival outcomes. A quarter of our patients were ECOG 2 + in contrast to the clinical trial population where the vast majority of the patients are ECOG ≤ 1. However, our cohort is more reflective of the real-world population.
In conclusion, the results of our study underscore that prognostic biomarkers may be only relevant in very specific clinical context. In particular, we report on the poor outcomes observed in melanoma patients with the HLA-B27 supertype when treated with pembrolizumab monotherapy. This observation requires validation in an independent study, as it may inform the need to treat patients with certain HLA supertypes associated with poorer OS with combination therapy to improve their long-term outcomes. Additionally, HLA-B27 carriers might be better suited to treatment with single agent ICI to avoid the exaggerated risk of immune side effects with combination ICI given the association of this supertype with other autoimmune syndromes.