Burden of tumor mutations, neoepitopes, and other variants are weak predictors of cancer immunotherapy response and overall survival

Abstract

Background: Tumor mutational burden (TMB; the quantity of aberrant nucleotide sequences a given tumor may harbor) has been associated with response to immune checkpoint inhibitor therapy and is gaining broad acceptance as a result. However, TMB harbors intrinsic variability across cancer types, and its assessment and interpretation are poorly standardized.

Methods: Using a standardized approach, we quantify the robustness of TMB as a metric and its potential as a predictor of immunotherapy response and survival among a diverse cohort of cancer patients. We also explore the additive predictive potential of RNA-derived variants and neoepitope burden, incorporating several novel metrics of immunogenic potential.

Results: We find that TMB is a partial predictor of immunotherapy response in melanoma and non-small cell lung cancer, but not renal cell carcinoma. We find that TMB is predictive of overall survival in melanoma patients receiving immunotherapy, but not in an immunotherapy-naive population. We also find that it is an unstable metric with potentially problematic repercussions for clinical cohort classification. We finally note minimal additional predictive benefit to assessing neoepitope burden or its bulk derivatives, including RNA-derived sources of neoepitopes.

Conclusions: We find sufficient cause to suggest that the predictive clinical value of TMB should not be overstated or oversimplified. While it is readily quantified, TMB is at best a limited surrogate biomarker of immunotherapy response. The data do not support isolated use of TMB in renal cell carcinoma.

Keywords: Immunotherapy response; Neoantigens; Neoepitope burden; Neoepitopes; Retained introns; Splice junctions; TMB; Tumor mutational burden; Tumor variant burden.

Fig. 1

Per-patient distribution of mutation and neoepitope burdens across 7 cancer types. a The number of somatic DNA variants per patient (scaled for sequence coverage) are shown along the y-axis, with each dot representing an individual cancer patient (cancer types shown along the x-axis). Note that MMR-deficient cancers here represent a cohort of three different cancer types including colon, endometrial, and thyroid with evidence of mismatch repair deficiency as determined by polymerase chain reaction or immunohistochemistry [9]. Red colored dots correspond to patients with microsatellite instability as determined by mSINGS (see “Methods”). b The number of putative neoepitopes per patient are shown along the y-axis, with each dot representing an individual cancer patient (cancer types shown along the x-axis). Abbreviations as follows: MMR = mismatch repair

Fig. 2

Per-patient distribution of overall tumor variant burden and its components. The number of total tumor variants per patient is shown along the y-axis, with the numbers of retained introns (RI), tumor-specific exon-exon junctions (Jx), insertions/deletions (Indel), and single nucleotide variants (SNV) shown in green, blue, red, and purple, respectively. The data for each individual patient is displayed as stacked bars along the x-axis, sorted from left to right by the number of single nucleotide variants (from highest to lowest)

Fig. 3

Robustness of putative neoepitope presentation. a The number of unique patient-matched HLA alleles that are predicted to present an individual neoepitope is shown along the x-axis, with the y-axis (log scale) corresponding to the overall percent of neoepitopes sharing that same robustness of HLA presentation. Red and blue curves denote the best fit line based on linear regression for all neoepitopes and those resulting from cancer driver mutations, respectively. The surrounding red and light blue shading denotes the 95% confidence interval for all and driver-derived neoepitopes, respectively. Individual data points are shown as open circles, whose diameter corresponds to the number of neoepitopes as shown by the corresponding scale at right. b The total number of unique patient-matched HLA alleles that are predicted to present one or more neoepitopes arising from a single DNA mutation is shown along the x-axis, with the y-axis corresponding to the overall percent of mutations sharing that same robustness of HLA presentation. Red and blue curves denote the best fit line based on local polynomial regression for all mutations and cancer driver mutations, respectively. The surrounding red and light blue shading denotes the 95% confidence interval for all and driver mutations, respectively. Individual data points are shown as open circles, whose diameter corresponds to the number of mutations as shown by the corresponding scale at right. c The percentage of total variants that are predicted to be presented by one or more patient-matched HLA alleles is shown along the y-axis, with the x-axis corresponding to the number of unique HLA alleles for that patient. Red and blue curves denote the best fit line based on linear regression for all mutations and cancer driver mutations, respectively. The surrounding red and light blue shading denotes the 95% confidence interval for all and driver mutations, respectively. Individual data points are shown as open circles, whose diameter corresponds to the number of mutations as shown by the corresponding scale at right. Note that a predicted HLA binding affinity threshold of ≤ 500 nM was used in all cases (see “Methods”)

Fig. 4

Receiver operating characteristic curves of predictive capacity of 11 different mutation/neoepitope burden metrics. The upper panels depict the true positive rate (sensitivity, y-axis) and false positive rate (1-specificity, x-axis) for each metric across all probability thresholds. The four panels represent models for four different cohorts based on different subsets of patients: All Cancers, which includes all patients, and Melanoma, RCC, and NSCLC, which include only melanoma, RCC, and NSCLC patients, respectively. The table in the lower panel reports the area under the curve (AUC) for each metric (columns) applied to a different cancer cohort (rows), with colors above the methods indicating the color of the corresponding curve in the upper panels. TMB is used as a predictor in both raw (TMB1) and coverage-adjusted (TMB2) forms, as well as in a multiplicative combination with patient HLA allele count (TMB1*HLA). Neoepitope burden (NB) is used as a predictor in both raw and extended formats (see “Methods”). Extended neoepitope burden metrics include number of amino acid mismatches (M), number of HLA alleles predicted to bind each epitope (A), and number of transcripts expressing each epitope in TCGA (T), along with their multiplicative combinations. Bold-faced values indicate the best value for each cancer cohort

Fig. 5

Receiver operating characteristic curves of predictive capacity of nine different variant/neoepitope burden metrics. The upper panels depict the true positive rate (sensitivity, y-axis) and false positive rate (1-specificity, x-axis) for each metric across all probability thresholds. The three panels represent models for three different cohorts based on different subsets of patients: All Cancers, which includes all patients, and Melanoma, and RCC, which include only melanoma and RCC patients, respectively. The table in the lower panel reports the area under the curve (AUC) for each metric (columns) applied to a different cancer cohort (rows), with colors above the methods indicating the color of the corresponding curve in the upper panels. TMB and TVB are used as predictors in the raw formats. Jx represents the number of tumor-specific junctions per patient, and RI represents the number of retained introns per patient, with RI epitopes representing neoepitopes derived from those retained introns. Neoepitope burden is used as predictor in its RNA-feature-extended formats (see “Methods”). Extended neoepitope burden metrics include number of expressed transcripts for each epitope (E), number of amino acid mismatches (M), number of HLA alleles predicted to bind each epitope (A), and number of transcripts expressing each epitope in TCGA (T), along with their multiplicative combinations

Fig. 6

Overall survival among cancer patients with high and low TMB. a Kaplan-Meier curves for immunotherapy-treated (+ICI) and immunotherapy-naive (−ICI) Stage III-IV melanoma patients with high TMB (> 80th percentile) are shown in red, and dark gray, respectively, while immunotherapy-treated (+ICI) and immunotherapy-naive (−ICI) patients with low TMB (≤ 80th percentile) are shown in blue and light gray, respectively. The underlying table corresponds to the number of patients at risk of death at each timepoint. Note: TCGA SKCM patient data (−ICI) is censored at 2885 days (maximal follow-up in immunotherapy-treated cohort) to emphasize comparable time-scales. b Kaplan-Meier curves for the immunotherapy-treated (+ICI) and immunotherapy-naive (−ICI) metastatic (Stage IV) renal cell carcinoma patients with high TMB (> 80th percentile) are shown in red, and dark gray, respectively, while immunotherapy-treated (+ICI) and immunotherapy-naive (−ICI) patients with low TMB (≤ 80th percentile) are shown in blue and light gray, respectively. The underlying table corresponds to the number of patients at risk of death at each timepoint. Note: TCGA KIRC patient data is censored at 1724 days (maximal follow-up in immunotherapy-treated cohort) to emphasize comparable time-scales

Fig. 7

Receiver operating characteristic curves of predictive capacity of coverage-adjusted TMB from 7 different variant calling methods: consensus calling (see Methods), MuSE [51], MuTect [52], Pindel [53], RADIA [54], SomaticSniper [22], and VarScan 2 [23]. The upper panels depict the true positive rate (sensitivity, y-axis) and false positive rate (1-specificity, x-axis) for each method across all probability thresholds. The four panels represent models for four different cohorts based on different subsets of patients: All Cancers, which includes all patients, and Melanoma, RCC, and NSCLC, which include only melanoma, RCC, and NSCLC, respectively. The table in the lower panel reports the area under the curve (AUC) for each method (columns) applied to a different cancer cohort (rows), with colors above the methods indicating the color of the corresponding curve in the upper panels. TMB as determined by consensus calling (see “Methods”) is compared to the individual variant calling tools used in consensus calling. RCC = renal cell carcinoma, NSCLC = non-small cell lung cancer