MHC Hammer reveals genetic and non-genetic HLA disruption in cancer evolution

Abstract

Disruption of the class I human leukocyte antigen (HLA) molecules has important implications for immune evasion and tumor evolution. We developed major histocompatibility complex loss of heterozygosity (LOH), allele-specific mutation and measurement of expression and repression (MHC Hammer). We identified extensive variability in HLA allelic expression and pervasive HLA alternative splicing in normal lung and breast tissue. In lung TRACERx and lung and breast TCGA cohorts, 61% of lung adenocarcinoma (LUAD), 76% of lung squamous cell carcinoma (LUSC) and 35% of estrogen receptor-positive (ER+) cancers harbored class I HLA transcriptional repression, while HLA tumor-enriched alternative splicing occurred in 31%, 11% and 15% of LUAD, LUSC and ER+ cancers. Consistent with the importance of HLA dysfunction in tumor evolution, in LUADs, HLA LOH was associated with metastasis and LUAD primary tumor regions seeding a metastasis had a lower effective neoantigen burden than non-seeding regions. These data highlight the extent and importance of HLA transcriptomic disruption, including repression and alternative splicing in cancer evolution.

Fig. 1. MHC Hammer: a tool to evaluate HLA DNA and RNA disruption.

MHC Hammer assesses allelic mutations, LOH, allelic repression and allelic alternative splicing in the class I HLA genes from WES and RNA-seq data. tumor adj, tumor adjacent; WES, whole-exome sequencing; RNA-seq, RNA sequencing.

Fig. 2. HLA expression is variable in normal tissue.

a, Gene level expression in the GTEx lung and breast normal tissue samples for HLA-A, HLA-B and HLA-C (lung, n = 483 and breast, n = 392). b, The ratio of the lung to breast HLA gene expression (n = 238 patients with both a lung and breast sample). c, The fraction of tumor-adjacent normal samples with RNA AIB (top) and for the samples with AIB, the AIB ratio (bottom; lung, n = 440 and breast, n = 380). d, The allelic expression per allele type in lung and breast tissue. Only alleles with at least 30 lung and/or breast samples are included in this analysis (lung, n = 465 and breast, n = 377). e, The structure of the class I HLA molecule. f, Rates of alternative splicing in the GTEx normal lung and breast samples for the exons and introns shown along the bottom. g, The novel transcript proportion of the alternative splicing events. h, The relationship between the novel transcript proportion in the lung and breast tissues, for alternative splicing events found in both the lung and breast tissues of the same GTEx individual. The P value and correlation coefficient (r) in h are calculated using Pearson’s method. Boxplots in a–d show the median and first and third quartiles, and whiskers extend to 1.5× IQR above and below the IQR. IQR, interquartile range; AIB, allelic imbalance; AS, alternative splicing; PTC, premature termination codon; RPKM, reads per kilobase per million.

Fig. 3. Transcriptional repression of the HLA genes in lung and ER+ breast cancer.

a, Each column represents a tumor, and each box represents a region from that tumor. Each region appears as two boxes. The first top box (above line) is coloured by the regional HLA LOH status, and the second bottom box (mirrored below line), colored by whether the region has transcriptional repression of the same allele that is lost in the DNA (blue), the alternate allele (allele 2, orange) or both alleles (purple). Only tumors with a patient-matched tumor-adjacent normal sample are included in this figure. None of these tumors had a high-impact damaging HLA mutation. b, The fraction of tumors with either only HLA LOH, only repression (unexplained by genomic alterations), both HLA LOH and repression (unexplained by genomic alterations) or no HLA LOH or repression. c, The frequency of monoallelic and biallelic repression events in tumor regions without genomic HLA alterations. d, The total number of intact alleles when accounting for alleles disrupted by LOH and repression. The lighter circle indicates the number of tumor regions in total, and the superimposed darker circle indicates the number of tumor regions in the given category. e, The relationship between the tumor-to-normal ratio of NLRC5, CIITA and IFNG expression and the number of transcriptionally repressed alleles in the tumor region. The P value in e is derived from a two-sided Wilcoxon test. Boxplots in e show median and first and third quartiles, and whiskers extend up to 1.5× IQR above and below the IQR. LOH, loss of heterozygosity.

Fig. 4. HLA alternative splicing in lung and breast tumors.

a, The fraction of tumors that exhibit tumor-enriched or tumor-depleted HLA alternative splicing events in the TRACERx (LUAD and LUSC) and TCGA (ER+) cohorts. b, The three most frequent HLA alternative splicing events are shown. The fraction of tumors that exhibit the event is shown at the top and the tumor-to-normal change in the novel transcript proportion is shown at the bottom. The legend is shown in a. c, The predicted consequences of the tumor-enriched and tumor-depleted alternative splicing events. d, The purity-scaled novel transcript proportion for the tumor-enriched alternative splicing events. e, The fraction of TRACERx tumors that exhibit alternative splicing events across all protein-coding genes, split by whether the gene is classified as a lung cancer gene or not. The three HLA genes are shown in red. f, The fraction of tumor regions that do/do not have LOH or repression and do/do not have tumor-enriched alternative splicing events. g, The neoantigen count for each allele is split by whether the allele exhibits tumor-enriched alternative splicing or not. Only alleles with no genomic disruption were included. P values in e and g are derived from a two-sided Wilcoxon test. P value in f is derived from Fisher’s exact test. Boxplots in e and g show median and first and third quartiles, and whiskers extend up to 1.5× IQR above and below the IQR. AS, alternative splicing; LOH, loss of heterozygosity; TSG, tumor suppressor gene; PTC, premature termination codon; rep, repression.

Fig. 5. The role of HLA disruption in tumor evolution.

a, The heterogeneity of HLA LOH, repression and tumor-enriched alternative splicing events. b,c, Overview (b) and example (c) of convergent evolution, where the same HLA allele is disrupted via different mechanisms in different regions of the same tumor. d, The relationship between the presence of repression and the amount of CD8 T cell infiltration. e, Tumor regions with and without HLA LOH have similar levels of total HLA expression. f, LUAD tumors that have HLA LOH are more likely to metastasize. g, When accounting for LOH and repression, LUAD regions that seeded a metastasis have a lower neoantigen count than those that did not. P values in d, e and g are derived from a two-sided Wilcoxon test. P value in f is derived from Fisher’s exact test. Boxplots in d, e and g show median and first and third quartiles, and whiskers extend up to 1.5× IQR above and below the IQR. LOH, loss of heterozygosity; rep, repression; AS, alternative splicing.

Extended Data Fig. 1. The TRACERx421 cohort.

Consort diagram outlining the TRACERx421 samples used in this study. The MHC Hammer WES and RNA-seq analysis filters are outlined in the Methods. FFPE, formalin-fixed paraffin-embedded; WES, whole-exome sequencing; RNA-seq, RNA sequencing; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma.

Extended Data Fig. 2. The TCGA lung and breast cancer cohort.

Consort diagram outlining the TCGA lung and breast samples used in this study. The MHC Hammer WES and RNA-seq analysis filters are outlined in the Methods. FFPE: formalin-fixed paraffin-embedded; WES: whole-exome sequencing; RNA-seq: RNA sequencing; LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma; BRCA, breast cancer.

Extended Data Fig. 3. The GTEx cohort.

Consort diagram outlining the GTEx lung and breast samples used in this study. The MHC Hammer RNA-seq analysis filters are outlined in the Methods. WES: whole-exome sequencing; RNA-seq: RNA sequencing.

Extended Data Fig. 4. Types of alternative splicing detected by MHC Hammer.

MHC Hammer will detect 4 different types of alternative splicing: complete exon skipping, complete intron retention, partial exon skipping and partial intron retention.

Extended Data Fig. 5. The difference between the lung and breast novel transcript proportion.

For alternative splicing events that were found in both the lung and breast tissue of the same patient, the difference in the novel transcript proportion between the two tissues is shown.

Extended Data Fig. 6. HLA LOH rates in lung and breast cancer.

The rate of HLA LOH in the TRACERx421 lung cohort and the TCGA lung and breast cancer cohorts. LOH: loss of heterozygosity; LUAD: lung adenocarcinoma; LCNEC: large cell neuroendocrine carcinoma; ER+: estrogen receptor positive; ER-: estrogen receptor negative; TNBC: triple-negative breast cancer.

Extended Data Fig. 7. Neoantigen silencing due to HLA LOH and repression.

For each tumor region, the fraction of putative neoantigens predicted to bind exclusively to HLA alleles subject to LOH and/or repression. Boxplot shows median and first and third quartiles, and whiskers extend up to 1.5× IQR above and below the IQR. LOH: loss of heterozygosity; LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma.

Extended Data Fig. 8. The HLA gene regulators and HLA expression.

The relationship between total HLA expression, measured by MHC Hammer, and the expression of CIITA, IFNG, NLRC5 and TNFα (TNF) in tumor samples without HLA genomic disruption. The P value and correlation coefficient (r) are calculated using Pearson’s method. TPM, transcripts per million; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; ER+, estrogen receptor positive.

Extended Data Fig. 9. CD8 T cells and HLA alternative splicing.

The relationship between the presence of tumor-enriched alternative splicing (AS) and the amount of CD8 T cell infiltration. P values were derived from a two-sided Wilcoxon test. Boxplot shows median and first and third quartiles, and whiskers extend up to 1.5× IQR above and below the IQR. LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma; ER+: estrogen receptor positive; AS: alternative splicing.

Extended Data Fig. 10. Tumor neoantigen burden and metastasis seeding.

a, The neoantigen count of primary tumor regions, split by whether they did or did not seed a metastasis. b, The neoantigen count of primary tumor regions, restricted to reflect peptides binding only to HLA alleles without HLA LOH, split by whether the primary tumor regions did or did not seed a metastasis. P values were derived from a two-sided Wilcoxon test. Boxplot shows median and first and third quartiles, and whiskers extend up to 1.5× IQR above and below the IQR. LOH: loss of heterozygosity; LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma.