Very low mutation burden is a feature of inflamed recurrent glioblastomas responsive to cancer immun

Several immunotherapy clinical trials in recurrent glioblastoma have reported long-term survival benefits in 10–20% of patients. Here we perform genomic analysis of tumor tissue from recurrent WHO grade IV glioblastoma patients acquired prior to immunotherapy intervention. We report that very low tumor mutation burden is associated with longer survival after recombinant polio virotherapy or after immune checkpoint blockade in recurrent glioblastoma patients. A relationship between tumor mutation burden and survival is not observed in cohorts of immunotherapy naïve newly diagnosed or recurrent glioblastoma patients. Transcriptomic analyses reveal an inverse relationship between tumor mutation burden and enrichment of inflammatory gene signatures in cohorts of recurrent, but not newly diagnosed glioblastoma tumors, implying that a relationship between tumor mutation burden and tumor-intrinsic inflammation evolves upon recurrence.

Despite aggressive standard of care (SOC) therapy, recurrence of WHO grade IV malignant glioma (glioblastoma, GBM) is virtually inevitable. No effective salvage therapy exists for recurrent GBM (rGBM), for which median survival is ~12 months. A subset of rGBM patients exhibited durable long-term survival1 following intratumor infusion of a recombinant rhino:poliovirus hybrid, PVSRIPO2. PVSRIPO kills neoplastic cells and induces innate inflammation that primes anti-tumor T cells3. Similarly, subsets of rGBM patients have responded to other virotherapy approaches4 or immune checkpoint blockade (ICB)5.

Such dichotomous responses have been observed in other indications after ICB therapy6, inspiring investigation into predictive biomarkers of ICB response7. High tumor mutation burden (TMB) is associated with response to ICB in several cancer types, with gliomas among the noted exceptions8. Rather, some hypermutant gliomas with mismatch repair (MMR) deficiency are less responsive to PD1 blockade than gliomas with lower TMB9, and a non-significant inverse relationship between TMB and radiographic/histological responses to PD1 blockade was observed in an rGBM patient cohort10. Using clinical and genomic information from three immunotherapy clinical trial cohorts1,8,10 (Table 1), we sought to define common features of the subsets of immunotherapy-responsive rGBM patients.

Here we show that rGBM patients carrying low TMB survive longer after polio virotherapy and ICB. rGBM tumors with lower TMB have enriched inflammatory gene signatures relative to rGBM tumors with higher TMB levels. This correlation is not observed in primary GBM tumors, indicating that a relationship between tumor-intrinsic inflammation and TMB develops upon recurrence in GBM.

Very low TMB identifies rGBM patients with longer survival after immunotherapy

We determined that rGBM tumors from patients with longer survival (>20 months) after PVSRIPO treatment harbored very low TMB (<0.6 mutations/Mb, Supplementary Fig 1, n = 21). Stratifying patient survival after PVSRIPO treatment by median TMB (1.3 mutations/Mb) confirmed longer survival of rGBM patients carrying ≤ median TMB (Fig. 1a); of note only 2/21 patients in the > median TMB strata were hypermutated (>10 mutations/Mb). TMB was previously shown to correlate with time to recurrence from initial surgery11, and a lack of non-synonymous mutations in TP53 (ref. 12). Accordingly, shorter time to first recurrence (Fig. 1b, preceding PVSRIPO treatment, n = 45) and lack of TP53 mutation (Supplementary Fig. 2, n = 34) also identified patients with longer survival after PVSRIPO therapy. Survival differences upon stratification by TMB, time to recurrence, and TP53 mutation status were maintained after excluding patients with IDH1 mutation and/or MGMT promoter methylation (Supplementary Fig. 2); survival differences upon stratification by TMB were also observed after exclusion of patients with PTEN mutations (Supplementary Fig. 2k). Suggesting these findings are not related to steroid dosing, stratification of cumulative pretreatment dose of steroids (2 days prior to, and the day of, treatment) did not reveal survival differences (Supplementary Fig. 2k, l).

To test if very low TMB associates with longer survival after ICB in rGBM patients, we compared survival after treatment with ICB in two cohorts stratified by TMB8,10. Patients with ≤ median TMB lived longer after receiving ICB therapy than those with >median TMB in both cohorts (Fig. 1c (n = 81; 3 hypermutants total), d (n = 29; no hypermutants)). We confirmed that stratification of post-PVSRIPO/ICB survival by TMB remained significant following exclusion of patients with hypermutation, indicating these differences were not driven by unfavorable responses in hypermutated patients (Supplementary Fig. 3). Very low TMB is associated with favorable survival in WHO grade III anaplastic astrocytoma, but not in WHO grade IV GBM9,13. Indeed, we confirmed no survival differences in immunotherapy-naïve primary (n = 277) or rGBM (n = 132)11 patient cohorts upon stratification by TMB (Fig. 1e, f), or by time to first recurrence (Fig. 1g, n = 131). Together these findings imply that increased survival of immunotherapy-treated GBM patients with very low TMB is due to immunotherapy response.

Tumor immune status inversely associates with TMB in rGBM tumors

High TMB associates with inflammatory profiles in select tumor types, potentially explaining its link with favorable response to ICB therapy14. Inflammation-associated gene expression measured via single sample gene set enrichment analysis (ssGSEA)15 revealed minimal differences in tumors with TMB below vs. above median in IDH1-wildtype (wt) primary GBM (pGBM, TCGA) patients (Fig. 2a; n = 193). To test this in rGBM, we sequenced RNA from all available pre-treatment tumor biopsy tissue samples in the PVSRIPO cohort (Supplementary Fig. 4, n = 14, all IDH1-wt); we also tested other previously sequenced IDH1-wt rGBM cohorts (refs. 10,16, and TCGA). In contrast to IDH1-wt pGBM, stratification of IDH1-wt rGBM ssGSEA scores by median TMB revealed markedly higher inflammation in rGBM tumors with lower TMB (Fig. 2b, Supplementary Fig. 5, n = 58); this correlation was also evident along a continuum (Supplementary Fig. 6).

Neoantigen depletion/immunoediting may explain a relationship between low TMB and immune activity. The ratio of nonsynonymous mutations to total TMB was lower in tumors with lower overall TMB in the PVSRIPO-treated cohort (Fig. 2c; n = 18). Likewise, in a larger cohort of rGBM tumors11, the ratios of observed vs. computationally expected neoantigens (Fig. 2d; n = 115) were non-significantly lower in tumors carrying lower overall TMB; this trend was exclusive to rGBM (Supplementary Fig. 7). Thus, rGBM tumors with very low TMB are immunologically engaged and may be immunoedited; these relationships were not apparent in primary, treatment-naïve GBM.

To determine how an association between TMB and immunological activity emerges in rGBM, we analyzed TMB and transcriptome immune signatures from paired primary and recurrent IDH1wt GBM tumors (Fig. 2e, f; n = 25 pairs, all <10 mutations/Mb)16. As observed for pGBM tumors from TCGA (Fig. 2a), there was limited correlation between TMB and inflammation-associated ssGSEA scores in the primary tumors; yet a correlation was observed in the recurrent tumors from the same patients (Fig. 2e and Supplementary Fig. 8). Calculating the change in these features upon recurrence revealed enrichment in cytolytic score (CYT) and MHC-class II (MHC-II) gene sets in patients with relatively lower TMB at recurrence (rTMB; Fig. 2f). Decreases in TMB were also primarily observed in patients with very low rTMB. Thus, enrichment of inflammatory signatures upon GBM recurrence—concurrent with TMB suppression—may explain the relationship between very low TMB and immune status of rGBM. Notably, CYT and MHC-II gene enrichment scores were quantitatively similar between primary and recurrent tumors at a cohort level, despite heterogenous changes at the patient level (Supplementary Fig. 8d, e). Together these observations indicate that a relationship between TMB and immune status evolves upon recurrence.

Our findings, in conjunction with recent studies in hypermutant rGBM patients9,17, suggest utility of TMB to predict immunotherapy outcome in rGBM. Notably our study included only a handful of hypermutant patients that did not drive the differences observed upon stratification by TMB; rather relative to non-hypermutant rGBM patients with moderate TMB levels, patients with very low TMB survived longer after immunotherapy. ICB and PVSRIPO immunotherapy engage antitumor T cell immunity3. Immunologically active tumors are a feature of ICB-responsive patients in other indications14, in some cases in patients with low TMB18, and very low TMB associates with increased tumor-intrinsic inflammation in rGBM (Fig. 2). Thus, we postulate that the baseline inflammatory status of rGBM tumors determines their susceptibility to immunotherapy. Correlation between tumor immune status and TMB was robust in rGBM, but not pGBM, suggesting that this relationship is a consequence of SOC therapy.

Since neoantigen depletion was also observed in rGBM patients with very low TMB, a possible explanation for the link between TMB and tumor immune status is that mutation levels, the landscape of which evolve during/after SOC therapy19, are restricted in a subset of patients with active and/or functional immune surveillance. Nonetheless, low TMB tumors are equally lethal to their high TMB counterparts in immunotherapy naïve patients, indicating that despite such potential immune engagement/immune-surveillance, immune evasion still occurs. Thus, it is possible that tumors with low TMB, which are more likely to be inflamed in rGBM patients, may employ different strategies to evade immune surveillance relative to their higher TMB counterparts, and that the immune evasion mechanisms of inflamed rGBM tumors may be successfully targeted by immunotherapy. Possibly suggesting an immune subversive mechanism: low TMB also associated with higher MDSC/myeloid cell density within tumors (Supplementary Fig. 4c, d), cell types well-known for suppressing effector functions of antitumor T cells20—including via the PD1/PD-L1 axis21,22,23—that are also potentially reprogrammed by PVSRIPO therapy3,24. Alternatively, as a shorter time to recurrence also correlates with lower TMB and favorable response to PVSRIPO, shorter duration of SOC, or shorter time from initial surgery, may allow for superior response to immunotherapy. In support of the latter, surgery has been shown to bolster efficacy of neoadjuvant αPD1 therapy25.

Thus, TMB itself may not be a causative driver of response to immunotherapy/tumor inflammation, but rather may merely reflect the immunological status of tumors or one of several other potential co-related features. Such features may include time to recurrence (and extent of tumor evolution), TP53 mutation, as well as any differences in the clinical care between patients with high vs. low TMB (e.g. differences in long-term corticosteroid exposure). Our study is insufficient to identify a specific number or threshold of TMB that predicts immunotherapy response in rGBM patients, but reveals an unexpected correlation between TMB, tumor-intrinsic inflammation, and survival after immunotherapy in rGBM patients that may hold predictive biomarker potential upon further validation/investigation. Future studies are needed to determine how SOC therapy alters the tumor immune status of GBM, and how reliably TMB, as well as potential TMB-related features, predicts response to immunotherapy in rGBM patients.