Inducing Hypermutability to Promote Anti–PD-1 Therapy Response

In this issue of Cancer Discovery, Crisafulli and colleagues present interim clinical and translational analyses of the ARETHUSA trial (NCT03519412), a phase II study of pembrolizumab in patients with chemorefractory mismatch repair–proficient (MMRp), RAS-mutated, and O6-methyl-guanine-DNA-methyltransferase (MGMT) promoter methylated metastatic colorectal cancer (mCRC) following induction (“priming”) therapy with temozolomide (1). Leveraging strong preclinical and clinical evidence, this study provides proof of concept for ongoing efforts to modulate the immunogenicity of “cold” MMRp mCRC and thereby enhance the efficacy of immune-checkpoint blockade.

Colorectal cancers with MMR deficiency (MMRd) or somatic hypermutation (as in the context of inactivating POLE mutations) represent molecularly distinct subtypes in which the efficacy of anti–PD-1/L1 has been established. For MMRd colorectal cancer, such efficacy is based on the presence of novel immunogenic expressed peptides (tumor neoantigens) resulting from the accumulation of somatic frameshift mutations [insertion–deletions (indel)] and to a lesser extent single-nucleotide variations [SNV (2)]. High levels of microsatellite instability (MSI-H) and elevated tumor mutation burden (TMB) are the resultant molecular phenotypes of MMRd. In addition, it is also possible to classify MMRd at the level of genome-wide, context-dependent mutational signatures, specifically 6, 15, and 26 (3). Similarly, POLE inactivation is associated with markedly elevated TMB (predominantly SNVs) and signature 10.

In contrast, MMRp colorectal cancers, the predominant molecular subtype of colorectal cancer, are nonhypermutated tumors, in which multiple prior studies have demonstrated extremely limited efficacy of immune-checkpoint inhibition. Hence, intense effort has focused on enhancing the immunosensitivity of MMRp colorectal cancers through either modulation of the tumor immune microenvironment or improvements in T-cell identification/targeting of MMRp colorectal cancer cells. In the context of this latter effort, prior evidence from studies of temozolomide-refractory human gliomas has inspired a compelling novel approach in the context of mCRC. Namely, Hunter and colleagues first reported on the emergence of a hypermutation phenotype among a subset of recurrent malignant gliomas that were treated with temozolomide, an alkylating agent commonly used in combination with radiotherapy for treatment-naïve disease (4). As shown in this and subsequent studies (5, 6), temozolomide exposure/resistance in malignant gliomas is frequently associated with the loss of MSH6 and/or other MMR gene expression, leading to the hypothesis that subclonal functional MMRd accelerates the molecular evolution and outgrowth of temozolomide-resistant tumor cell populations. Temozolomide resistance in the context of MMRd appears to be uniquely associated with mutation signature 11 (6).

Elegant in vitro and in vivo work by Bardelli and colleagues has recently demonstrated a similar phenomenon in colorectal cancer models (7), where extended treatment with temozolomide was associated with the acquisition of MMR gene mutations and increased (posttreatment) TMB. Multiple phase II studies have demonstrated favorable toxicity profiles but limited clinical efficacy of various temozolomide-based therapies among heavily pretreated patients with mCRC harboring MGMT methylation. Altogether, these findings support the approach of using DNA-damaging agents to induce mutations and therefore potentially “prime” MMRp tumors for response to immune-checkpoint inhibition.

Hence, the work by Crisafulli and colleagues offers critical insights into the feasibility and potential pitfalls of this novel approach. In prospective fashion, the investigators selected a cohort of 69 patients with MMRp, RAS-mutated, MGMT-methylated mCRC that had progressed after at least two prior lines of standard systemic therapy. Subsequently, 27 of 69 patients were treated per protocol with temozolomide until disease progression (median 3.0 months; range, 0.6–7.5). The investigators also identified three additional patients who progressed on temozolomide-based therapy given either off-protocol or in the context of a different clinical trial. Posttreatment tissue biopsies were obtained in 21 of these total 30 patients with progressive disease on temozolomide. Within this analysis cohort, the investigators undertook comprehensive somatic mutation profiling of posttreatment and (if available) pretreatment tumor biopsies using a whole-exome DNA sequencing platform and peripheral blood (germline) controls. Overall, the analysis validated three recurrent features among posttreatment samples. First, increases in TMB to ≥20 mutations per megabase (mut/Mb) were achieved in 2 of 21 (9.5%) patients when considering only clonal tumor tissue mutations [i.e., mutations with variant allele frequency (VAF) ≥10%]. This proportion was increased to 19 of 21 (90.5%) patients when considering subclonal mutations (VAF <10%). Interestingly, most of the variation in TMB was explained by SNVs only, as there was no significant variation in the number of indels present among posttreatment tumor samples. Patterns of TMB modulation were largely recapitulated using peripheral blood circulating tumor DNA (ctDNA)–based mutation profiling. Second, investigators observed the emergence of mutation signature 11 in 17 out of 21 (81%) patients. However, only 2 of these 17 patients had signature 11 detectable at the level of clonal mutations, corresponding to the same two patients with clonal TMB of ≥20 mut/Mb noted above. The remainder of patients had signature 11 detectable at the level of subclonal mutations only. Third, the investigators identified recurrent de novo MSH6 mutations in 17 of 21 (81%) posttreatment tumor samples, a finding that was exclusive to the same 17 patients in whom signature 11 was also detectable. A deleterious missense mutation (MSH6 p.T1219I) was by far the most prevalent, occurring in 16 of the 17 (94%) patients.

With respect to clinical outcomes, six patients with posttemozolomide clonal or subclonal TMB ≥20 mut/Mb were treated with pembrolizumab. Among these six patients, there were no objective responses. Three patients achieved stable disease for at least 5 months, though in two of these patients, ctDNA increased during disease stability. One patient demonstrated robust clinical benefit with durable disease control on pembrolizumab for >2 years. Of note, this patient had received temozolomide and capecitabine outside of the ARETHUSA trial but was included in the aforementioned analyses. Furthermore, this patient corresponds to one of the two patients with posttreatment biopsies that harbored TMB ≥20 mut/Mb and mutation signature 11 at the clonal mutation level. Despite the research challenges posed by the COVID-19 pandemic, Crisafulli and colleagues have identified critically important candidate biomarkers to select patients who may benefit from anti–PD-1 therapy after progression on temozolomide priming. This commendable effort also demonstrates the relative safety and feasibility of this approach, particularly one that incorporates real-time assessment of tumor molecular evolution into clinical decision-making.

However, their findings also illuminate several unanswered questions and potential pitfalls. First, although the duration of temozolomide exposure was linearly correlated with the emergence of hypermutation and signature 11, a TMB of ≥20 mut/Mb was achieved in only 9.5% of patients when evaluated at the level of clonal mutations. This finding is particularly relevant considering recent evidence suggesting that immunoreactivity and sensitivity are largely driven by clonal, not subclonal, antigens (8). Given the small sample size, the impact of clonal versus subclonal TMB on the clinical activity of anti–PD-1 therapy in MMRp mCRC requires further study. However, current and future applications of temozolomide priming may need to optimize the emergence of clonal mutations or incorporate in silico methods to identify tumors that acquire highly immunogenic alterations. Second, it is notable that TMB increases were predominantly driven by a rise in SNV burden and not indels. Although disease control >5 months with anti–PD-1 therapy was achieved in three of six patients with clonal or subclonal TMB ≥20 mut/Mb, there appeared to be no correlation with indel burden. This observation again raises the question of whether global mutation quantity is as necessary as neoantigen quality or immunogenicity. Furthermore, despite the acquisition of de novo MSH6 alterations in temozolomide-resistant tumors, the lack of corresponding increases in indel mutations highlights the need to better understand the functional and clinical relevance of MMR gene inactivation in temozolomide-resistant tumors. In essence, MSH6 mutation alone was not sufficient to predict response or disease control with pembrolizumab, but it strongly correlated with the emergence of mutation signature 11 after prolonged temozolomide treatment. Lastly, it is notable that a relatively small proportion (15.6%) of patients screened in the ARETHUSA study had tumors that harbor MGMT methylation. Importantly, the use of MGMT methylation for patient selection in this study is supported by a strong preclinical and clinical rationale. However, it also reflects a need to develop alternative priming strategies for patients with non–MGMT-methylated mCRC tumors.

Moving forward, the comprehensive work by the ARETHUSA study team [as well as parallel efforts by the MAYA study team (9)] argues that the overall approach of temozolomide-based priming in mCRC warrants further investigation and focus. There are likely several areas of opportunity to improve the efficacy of this approach. As an example, preliminary in vivo data presented by Luis Diaz and colleagues at the American Association for Cancer Research (AACR) 2022 Annual Meeting suggest that the combination of cisplatin and temozolomide may be a more potent mechanism to induce high TMB and immunogenic indel mutations compared with temozolomide alone. In addition, spatiotemporal profiling of the “primed” tumor microenvironment may also yield important insights and/or novel targets for immune modulation in this therapeutic setting. Though this approach is not ready for clinical practice, the data reported by the authors demonstrate a critical direction for future exploration in mCRC.