Genetic variation in antigen presentation and cancer immunotherapy

Abstract

Human leukocyte antigen (HLA) molecules are critical mediators of anti-tumor immune responses. In this issue of Immunity, Chhibber et al. (2021) challenge previous associations between germline HLA zygosity and immunotherapy outcomes, demonstrating that germline HLA genotypes and diversity alone are not independent biomarkers of anti-PD1 clinical efficacy.

Tumor-associated antigen presentation is an essential component of anti-tumor immune surveillance and can be affected by both germline and somatic alterations at the HLA class I and II loci. The highly polymorphic HLA class I genes (HLA-A, HLA-B, and HLA-C) encode for major histocompatibility complex class I (MHC-I) molecules that harbor significant inter-individual heterogeneity in the spectrum of peptides that can be presented. MHC-I molecules display cytosolic (8–11 amino acid) peptide fragments from endogenously degraded self or non-self proteins as antigens on the surface of nucleated cells to T cell receptors of CD8⁺ cytotoxic T lymphocytes. In the tumor microenvironment, recognition of tumor-derived peptides presented by MHC-I molecules triggers the activation of CD8⁺ T cells, triggering an antitumor immune response. MHC-II molecules encoded by HLA class II genes are involved in antigen presentation on professional antigen presenting cells, such as dendritic cells, promoting T cell activation by displaying peptides to CD4⁺ T cell subsets. Antigen presentation by MHC-II molecules on human cells involves three prominent loci on chromosome 6 (HLA-DR, HLA-DQ, and HLA-DP), which are in linkage disequilibrium and often inherited as a haplotype.

Conceptually, a high burden of mutation-associated neoantigens presented on MHC-I molecules would flag the tumor for destruction by the adaptive immune system. Therefore, the diversity of the HLA repertoire may directly impact the magnitude of the anti-tumor immune response especially in the context of immune checkpoint blockade (ICB). Recent studies have suggested that HLA class I germline zygosity and evolutionary divergence between HLA alleles of an individual may provide a primary measure of diversity at the MHC-antigenic peptide complex, which can in turn impact clinical outcomes with ICB (referred to as the divergent allele advantage hypothesis) (Chowell et al., 2018, 2019). However, several groups have reported discrepant findings across studies performed in independent patient cohorts and in tumor types underrepresented in initial analyses, calling into question the significance of HLA class I germline variation as a primary determinant of treatment outcomes (Anagnostou et al., 2020; Litchfield et al., 2021). In this issue of Immunity, Chhibber et al. (2021) performed a comprehensive analysis of HLA germline diversity in over 3,500 patients enrolled in 17 clinical trials of pembrolizumab monotherapy across eight tumor types and one basket trial, showing that germline HLA class I or II diversity alone is an insufficient indicator of therapeutic response.

Given the significance of genetic ancestry for germline haplotype analyses, the authors first stratified patients into more homogeneous populations and focused on the European ancestry subpopulation (Chhibber et al., 2021). Use of ancestry controls is an important aspect of genetic association studies to ensure that population-based differences in response or survival do not confound key study associations. The impact of genetic ancestry was exemplified in a recent study by Carrot-Zhang et al., 2021, showing that genetic ancestry may contribute to somatic tumor mutational burden and specific driver alterations (e.g., EGFR), highlighting the importance of appropriately controlling for genetic ancestry in germline HLA analyses. Chhibber et al. then investigated the landscape of germline HLA variation across pembrolizumab-treated patients. These analyses revealed that HLA diversity as measured by genotype, heterozygosity (defined by the presence of two distinct alleles) at each class I or Il locus, and evolutionary divergence (assessed using Grantham distance to measure the sequence divergence between peptide binding domains for any two given HLA alleles for a gene) across class I or Il loci was not associated with therapy response across all patients or within individual tumor types studied. As germline HLA heterozygosity at the allele level may not fully capture broader peptide binding potential in patients, heterozygosity at HLA-A and HLA-B was further assessed at the supertype (sets of HLA alleles with overlapping peptide-anchoring motifs) level, with results confirming a similar lack of association with patient response. To further improve the reliability of comparisons with previously published datasets (Chowell et al., 2018, 2019), the authors also performed a tumor-type-focused analysis of HLA class I heterozygosity across melanoma and non-small cell lung cancer (NSCLC) trial cohorts but identified no significant link to either survival outcomes or best overall response. Comparable trends were observed during an independent exploratory analysis of patients from the KEYNOTE-158 trial, who were stratified according to PD-L1 status and tumor mutational burden. The impact of individual HLA class I and II alleles on pembrolizumab efficacy was additionally assessed within each trial cohort, as well as across tumor types and collectively across all patients, again demonstrating no significant association with clinical outcomes (Chhibber et al., 2021).

These findings are in line with previous smaller studies of immunotherapy-treated patients with NSCLC (Anagnostou et al., 2020; Montesion et al., 2021) and argue against the notion that HLA germline genotypes and diversity alone are linked to patient responses to ICB (Chowell et al., 2018, 2019). Chowell et al. analyzed HLA class I genotypes across two ICB-treated cohorts of melanoma and NSCLC patients, demonstrating that homozygosity in ≥ 1 HLA class I loci was significantly associated with reduced overall survival (OS), independent of tumor stage, tumor mutation burden, or patient-related factors. In follow-up, the group also reported that the effect of germline HLA class I heterozygosity on tumor immunosurveillance was modulated by sequence divergence between the peptide binding domains, which reflect the peptide binding potential of MHC-I molecules and the variety of antigens presented for CD8⁺ T cell recognition (Chowell et al., 2019). A high HLA evolutionary divergence was reported to be significantly associated with longer OS, independent of other clinical variables (Chowell et al., 2019). Furthermore, certain HLA class I supertypes have been linked with therapeutic responses to ICB; in patients with advanced melanoma receiving anti-CTLA4 treatment, the B44 supertype (in particular the HLA-B*18:01, HLA-B*44:02, HLA-B*44:03, HLA-B*44:05, and HLA-B*50:01 alleles) was associated with longer OS (Chowell et al., 2018). Notably, these findings were not confirmed by Chhibber et al. (2021), who found no associations between the B44 supertype and clinical outcomes.

In thinking about tumor immunoediting and rejection in the context of immunotherapy, one should consider a constellation of tumor- and immune-related processes, which can lead to the collective modulation of functional antigen presentation capacity within the tumor microenvironment (Figure 1). Chief among these considerations is the need to understand HLA genetic variation within the context of the tumor evolutionary trajectory. A growing body of evidence indicates that germline HLA types can shape the evolutionary landscape of cancer by exerting a negative selective pressure on mutations leading to highly presented peptides (Marty et al., 2017; McGranahan et al., 2017). This inherent complexity of the temporal cascade of tumor-immune interactions, as cancer cells go through evolutionary bottlenecks during their natural clonal evolution, can be further amplified under the selective pressure of therapy. To this end, further investigation into potential correlations between germline HLA diversity and the promiscuity (variation in the absolute number of peptides a given HLA allele can bind) of individual alleles, as well as somatic changes in zygosity and somatic HLA loss-of-heterozygosity, which can be influenced by tumor neoantigen profiles (Montesion et al., 2021), is warranted to fully capture the overall antigen presentation capacity and ultimately the “visibility” of tumors. The number of unique HLA alleles in the tumor cells, which accounts for somatic loss-of-heterozygosity events, especially when combined with the tumor mutation burden, may better capture the presented antigen repertoire and thus reflect clinical outcomes with ICB. Germline and somatic HLA variation may be linked with anti-tumor immune responses and clinical outcomes in a context-dependent manner determined by the genomic background of the tumor and the cancer lineage. It is also plausible that germline HLA variation is most informative when the extremes are considered and maximal heterozygosity is compared to maximal homozygosity. Furthermore, non-genetic mechanisms of allele-specific repression such as epigenetic silencing may affect both HLA genes and neoantigen-encoding genes; these mechanisms can be harnessed by tumors to evade immune surveillance (Rosenthal et al., 2019).

Figure 1. Integrated approach to evaluate tumor and host-related factors that affect tumor visibility and can moderate clinical outcomes with ICB.

Germline HLA genotypes and diversity should be considered as a component of a multi-modal model rather than an independent feature of response to immunotherapy. An integrated approach may be better suited to interpret the complexity of the tumor-immune system interactions and determine tumor “visibility.” To this end, additional key features include somatic HLA status, which can modulate the impact of germline HLA diversity, allele-specific repression of HLA and neoantigen-containing genes by epigenetic silencing, controlled for genetic ancestry and considered in the context of an evolving tumor, and tumor microenvironment under the selective pressure of therapy.

The findings of Chhibber et al. do no support an association between HLA germline zygosity and response to ICB. Given the multi-faceted crosstalk between cancer and immune cells reflected in the multi-factorial nature of responses to ICB, germline HLA allelic diversity represents a component of a set of multi-modal features that orchestrate the tumor-immune system interactions. Future studies incorporating germline and somatic HLA variation alongside cancer and tumor microenvironment features will be key toward the integration of antigen presentation variation in predictive models for cancer immunotherapy.