Time to Think about HLA-Based Diagnostics in Lung Cancer?

In this issue of Clinical Cancer Research, Jiang and colleagues examine HLA evolutionary divergence (HED) as a putative biomarker to select patients who might benefit from camrelizumab, an anti-programmed cell death 1 (anti-PD1) agent when added to platinum-based chemotherapy (1). The authors used data from two randomized clinical trials in the People's Republic of China, CameL and CameL-sq, for treatment-naïve patients with locally advanced or metastatic non-squamous (EGFR and ALK mutated excluded) and squamous non–small cell lung cancer (NSCLC), respectively. HED was calculated as the mean distance between HLA alleles in each of the HLA class I loci (HLA-A, HLA-B, and HLA-C) and patients with HED values in the top quartile were classified as HED-high. In both cohorts, patients in the HED-high group had higher objective response rate, progression-free and overall survival when camrelizumab was added to chemotherapy but this was not true for the HED-low group. It is noteworthy that HED was not prognostic of outcomes in The Cancer Genome Atlas lung adenocarcinoma and squamous cell carcinoma cohorts, where patients generally did not receive treatment with immune checkpoint inhibitors (ICI). Analysis of the immune infiltrates in the tumor microenvironment at the single-cell level identified dominance of cell populations promoting, versus suppressing antitumor immunity in the HED-high tumors, whereas the opposite was observed in HED-low cases. Overall, the data support the use of HED in addition to established biomarkers like programmed cell death ligand 1 tumor proportion score (PDL1 TPS) and tumor mutation burden (TMB) to predict benefit from anti-PD(L)1 inhibitors for patients with NSCLC.

Anti-PD1 and anti-PDL1 ICIs provide benefit to a fraction of patients with NSCLC, either as single agents or in combination with chemotherapy. Adverse events from those agents along with financial toxicity underline the importance of biomarkers to guide the benefit and risk discussion with individual patients. In the current state of the art, membranous expression of PDL1 in tumor cells assessed by PDL1 TPS, and TMB expressed as number of mutations per megabase, have emerged as predictors of response to anti-PD(L)1 inhibitory antibodies at the clinical level (2, 3). Both those biomarkers however have important limitations, and the majority of patients with advanced NSCLC receives treatment with ICIs regardless of PDL1 TPS or TMB. On the one hand, tumors with no or little PDL1 expression might still benefit from anti-PD(L)1 agents, especially when added to chemotherapy. On the other hand, TMB calculation methods are analytically heterogeneous. Importantly, neither biomarker captures the ability of HLA to present tumor neoantigens, identified as non-self, to T cells.

A number of studies have tested the ability of HLA-based biomarkers to predict outcomes for patients with NSCLC and other cancers who receive treatment with ICI. Patients with HLA A*03 had less benefit from immunotherapy in a pan-cancer study that included 3,335 patients who received ICI and 10,917 patients who received alternative therapies (4). Although the mechanism for this effect is unclear, it is possible that HLA A*03 interacts with inhibitory innate immune system receptors and also might be associated with enhanced T regulatory activity. Although HLA A*03 is less common in Asian patients, it is possible that individual HLA alleles with variable prevalence in different populations shape the benefit from ICI.

Less than complete HLA class I heterozygosity has an impact on the diversity of the immunopeptidome and the calculated HED. Either germline homozygosity in at least one HLA class I locus, or somatic HLA class I LOH can result in less than six distinct HLA class I alleles in the tumor. Both have been proposed as biomarkers for ICI for patients with lung cancer (5, 6). Germline HLA class I homozygosity is encountered in 20% to 25% of patients with NSCLC (5, 6). Somatic HLA LOH for specific HLA alleles might be clonal or subclonal, and marks an evolutionary milestone for the tumor, as it is associated with an increase in the number of mutations (6, 7). It is interesting that HLA LOH failed to predict benefit from immunotherapy in the study by Jiang and colleagues. It is notable however, that the predictive value of HED increases when only patients with complete HLA heterozygosity in all class I loci are analyzed. These data reflect that HED might recapitulate the diversity of presented peptides more accurately than HLA heterozygosity status (less than six vs. six distinct HLA class I alleles), and yet the somatic loss of an HLA allele marks an HLA-independent phase of tumor evolution that alters the predictive value of HED. It is important to emphasize that an inflection point of tumor evolution is expected with HLA LOH, but not with germline HLA homozygosity.

Estimation of the neopeptidome diversity is highly relevant, as tumors are less proficient to escape immunity that is directed to diverse targets. The quality of individual peptide targets is another aspect of antitumor immunity that could influence immune editing of tumors. The majority of HLA-restricted and T cell–based immune reactions to tumors is triggered by tumor-associated antigens generated by passenger mutations (8, 9), but reactions to peptides containing driver mutations are occasionally possible. Immunity to driver mutations can be harnessed therapeutically, as it was emphatically shown for EGFR mutations in lung cancer and KRAS G12D in colorectal cancer (10, 11). It will be useful to develop biomarkers for ICIs that capture the quality, in addition to the diversity of T-cell targets in the future.

There are a number of cautionary points. First, the data used in the study by Jiang and colleagues are derived from the CameL and CameL-sq studies which were conducted in China. HLA genotype is interrelated with ethnicity, and therefore the predictive value of HED, as well as the optimal cut-off point to define HED-high patients might be discordant for patients of different ethnic groups. Validation of the predictive role of HED for anti-PD(L)1 treatment in studies like the Keynote 189, the Keynote-407, the EMPOWER-Lung 3, the IMpower150, and the IMpower130 studies will be very useful to this direction. Second, whether HED can predict benefit from CTLA4 inhibitors when given in combination with anti-PD(L)1 is currently unknown. Third, standardization of the cut-off point to separate patients with high versus low HED is required before possible clinical application: the top quartile of a population is widely used; however, an absolute number is more practical for the prospective patient. Finally, it is currently unknown whether HED or any HLA-based biomarker can have predictive value in the perioperative setting.

In conclusion, the authors are to be congratulated for their rigor. The study introduces HED as an attractive biomarker for ICI in patients with NSCLC receiving treatment with chemotherapy and anti-PD(L)1. HED is reflective of immunopeptidome diversity (Fig. 1) and from this perspective augments the predictive value of established biomarkers. Pending prospective validation, the field is getting closer to incorporate HLA typing in the standard diagnostic algorithm of NSCLC.

A. Dimou reports personal fees from Roche/Genentech, Intellisphere, TP Therapeutics, Guardant Health, AnHeart Therapeutics, and ChromaCode; and grants from Syntrix Pharmaceuticals, AstraZeneca, Novartis, AnHeart Therapeutics, Merck, Sorrento Therapeutics, Guardant Health, and Philogen outside the submitted work. No other disclosures were reported.