Immune Activation in Mismatch Repair–Deficient Carcinogenesis: More Than Just Mutational Rate

Over the past decade, DNA mismatch repair (MMR) deficiency has emerged as a critically important biomarker with implications for the management of both early- and advanced-stage colorectal cancer (1). Approximately 10% to 15% of colorectal cancers exhibit MMR deficiency, which is characterized by a propensity for accumulating single-nucleotide mutations and insertion–deletion loops (indels) in the somatic genome, particularly within short repetitive sequences such as microsatellites (2, 3). MMR-deficient tumors often exhibit a high mutation burden and may express neoantigens generated by frameshift mutations in coding microsatellites, such as the 10-adenine mononucleotide repeat in the TGFBR2 gene (4).

As part of the standard molecular workup for colorectal cancer, MMR deficiency can be assessed on the basis of microsatellite instability (MSI) and/or loss of expression of MMR proteins in bulk tumor tissue specimens (5). MMR deficiency in the tumor is often secondary to Lynch syndrome, an autosomal-dominant hereditary cancer syndrome caused by monoallelic pathogenic germline mutations in MMR pathway genes (MLH1, MSH2, MSH6, PMS2, and EPCAM; ref. 6). More frequently, MMR deficiency occurs as a sporadic (nonhereditary) process characterized by a distinctive hyperproliferative, serrated morphology, DNA methylation abnormalities including MLH1 epigenetic silencing (CpG Island Methylator Phenotype, CIMP), and elevated frequency of activating BRAF mutations (7–9).

Altogether, dMMR colorectal cancer represents a unique molecular subtype of this disease with distinctive histopathologic features and clinical outcomes. One of the most prominent features is the enrichment of tumor stroma with infiltrating lymphocytes, and overexpression of prostaglandins and inflammatory cytokines in dMMR tumors (10–16). This inflammatory microenvironment is thought to be driven by recognition of the high burden of tumor neoantigens on MHC class I alleles by the adaptive immune system (Fig. 1, later stages). This model not only helps explain the favorable prognostic implications of MMR deficiency in colorectal cancer, but also supports the rationale for immunotherapy-based treatment strategies such as with checkpoint inhibition. In this regard, pivotal examples can be found in the setting of metastatic colorectal cancer (17, 18). In particular, the phase II study CheckMate-142 (clinicaltrials.gov ID NCT02060188) recently demonstrated the safety and durable efficacy of nivolumab [anti-programmed cell death protein 1 (PD-1)] given with or without low-dose ipilimumab [cytotoxic T lymphocyte-associated antigen 4 (CTLA-4)] as second-line therapy for patients with advanced dMMR colorectal cancer (19, 20). Similar benefit was reported by the phase II study Keynote-164 (NCT02460198) in patients treated with single-agent pembrolizumab (anti–PD-1; ref. 19). These breakthrough results have amplified interest in the potential applications of novel immunotherapy agents not only in the adjuvant therapy setting for dMMR colorectal cancer, but also in primary prevention for patients with Lynch syndrome.

A rationale for immunotherapy-based prevention (hereafter referred to as immunoprevention) strategies in Lynch syndrome is supported by multiple lines of evidence, including the identification of MMR-deficient histologically normal appearing colon crypts as the earliest definable abnormality in preneoplastic colorectal epithelium in Lynch syndrome (13). With respect to existing immunomodulatory agents, nonsteroidal anti-inflammatory drugs (NSAID) inhibit cyclo-oxygenase 2 (COX-2) and the downstream production of protumorigenic prostaglandins that promote local inflammation. Prior work has shown that NSAIDs (21), more specifically aspirin (22, 23), are associated with a modest but reliable chemopreventive benefit to reduce the risk of Lynch syndrome–related colorectal cancer (and perhaps other sites) after a continuous exposure of at least 2 years of duration. Recent preclinical work has highlighted that naproxen sodium may have greater chemopreventive efficacy than aspirin (24), although the mechanism is not yet well delineated.

Toward the goal of further improving Lynch syndrome–related cancer mortality, we propose that novel prevention strategies can be developed by elucidating the sequence of events that relate acquisition of MMR deficiency to accumulation of somatic mutations, generation of neoantigens, tumorigenesis and immune recognition, and characterizing the immune cells in the microenvironment of preneoplastic lesions (25). Such strategies would include novel immunomodulatory agents, tumor vaccines (26–29), and even low-dose immune checkpoint inhibitors. Importantly, given the unique challenges of drug development in the prevention setting, each strategy needs focused reexamination of the risks and benefits. For example, although anti–PD-1/PD-L1 antibodies may increase immune surveillance, they are also associated with significant rates of severe adverse events. These include immune-related lung, hepatic, skin, neurologic, gastrointestinal, and endocrine toxicities, some of which are fatal (30–44). Thus, although the risk:benefit ratio of PD-1/PD-L1 blockade is acceptable for patients with metastatic tumors and poor prognosis, it is almost certainly not acceptable in the setting of healthy asymptomatic Lynch syndrome patients for cancer prevention, where the tolerance for side effects is very low. PD-1/PD-L1 inhibitors also do not have clear dose response, which makes giving lower doses of these drugs for cancer prevention problematic (30–44).

Here, we will briefly review the molecular basis of neoantigen generation and immune activation as it pertains to MMR-deficient colorectal tumorigenesis. We focus particularly on the precancer state in order to shed light on possible rationales for the development of novel immunoprevention strategies.

The highly conserved MMR system facilitates repair of two important types of errors that arise during DNA replication: base pair mismatches and indels (45). Base pair mismatches occur when incorrect nucleotides are inserted into the newly synthesized strand and escape the proofreading function of DNA polymerases. Indel loops usually arise in the context of microsatellites, which are highly polymorphic short repetitive DNA sequences found throughout both prokaryotic and eukaryotic genomes (46–48). At microsatellites, the template and primer strands are prone to slippage (i.e., dissociation and reannealing) during replication. This generates a loop structure and, most importantly, a discordant number of repeated units between the template and newly synthesized strand (49). In humans, the repair process begins with binding of the MSH2/MSH6 heterodimer to the DNA defect. This is followed by recruitment of the MLH1/PMS2 heterodimer, formation of a sliding clamp structure, and then activation of exonuclease 1 (EXO1) to remove the error-laden DNA segment. The resulting gap is filled in by DNA polymerases, PCNA, and ligases (45). Failure to repair base mismatches or indels leads to propagation of single-nucleotide mutations or MSI, respectively.

The mutagenic process described above is observable not only at the population level, but also within specific individuals, particularly in the context of acquired MMR deficiency. In the case of colorectal cancer, MMR deficiency may occur as an entirely sporadic process due to aberrant hypermethylation of MLH1 in the tumor, commonly associated with the BRAF V600E mutation. However, the hereditary counterpart of this process is Lynch syndrome, which affects more than 1.1 million people in the Unites States and serves as a disease model in which to understand the relevance of MMR deficiency across many cancer types (6). Deleterious germline mutations in MLH1, MSH2, MSH6, PMS2, or EPCAM render affected individuals with only one functional allele of the respective gene. This is accompanied by somatic inactivation of the second allele (e.g., through mutation or deletion) and thus a predisposition to developing MMR-deficient neoplasms. In general, Lynch syndrome is associated with higher life-time risks of not only colorectal cancer, but also endometrial, ovarian, gastric, small bowel, and urothelial cancers (6).

To place the MMR deficiency phenotype in context, it is helpful to consider it within the mutational “signature” framework (50, 51). Recent mutational profiling efforts have revealed at least three different signatures among precancer lesions (adenomas) and carcinomas of the colorectum (52). The most common of these is “Signature 1b,” which is characterized by C>T transitions that are produced after spontaneous deamination of 5-methyl-cytosine. This signature is thought to represent a cumulative and age-related process, as it is also observed in histologically “normal” colonic mucosa that juxtaposes carcinomas with an intact MMR system (52). By contrast, sporadic and hereditary MMR deficiency are associated with “Signature 6,” in which tumors accumulate C>T transitions preferentially at NpCpG loci and indels at microsatellites. Colorectal tumors and precancers with this signature are enriched for mutations in BRAF V600E (and underenriched for alterations in APC, KRAS, and TP53; ref. 53). “Signature 6” is often referred to by the synonymous term “hypermutator,” which denotes a high tumor mutation burden (TMB) that is conventionally defined as more than 10 mutations per megabase (Mb) when measured by whole-exome sequencing (54, 55). In a related way, “Signature 10” is observed in colorectal cancer tumors that harbor error-prone DNA polymerase activity due to somatic inactivation of POLE or germline defects in POLE or POLD1 (56). Signature 10 is often termed the “ultramutator” phenotype, with a mutation rate on the order of 150 mutations per Mb (56).

We and others propose that the biological and clinical implications of MMR deficiency for immunoprevention are best understood not only with respect to overall mutation burden, but perhaps more importantly with respect to the specific functional impacts of MSI. A comprehensive analysis by Hause and colleagues demonstrated that MSI can be detected in the exomes of multiple different tumor types, albeit with varying frequency and affected loci (57). The study identified a common set of microsatellites, such as those in the coding regions of NIPBL, TCF4, and PTEN, that showed instability across many different tumor types (57). Yet, other microsatellite loci showed instability in only a specific tumor type, thus forming an instability signature. This work builds on prior efforts over the past 20 years to catalogue instable microsatellites that occur in colorectal adenomas and established carcinomas. For example, MSI has been detected in the coding regions of known oncogenes and tumor-suppressor genes such as TGFβRII, BAX, MSH3, MSH6, IGFIIR, and MRE11A (58–63), as well as others associated with immune surveillance and B2M (which is part of the neoantigen presentation machinery) in particular (64). A large set of intergenic (noncoding) microsatellite targets has also been identified (65).

MMR deficiency not only promotes the development and progression of colorectal cancer, but also contributes to the generation of tumor-specific neoantigens. Specifically, neoantigens are created when indels arise within coding microsatellites, leading to an erroneous reading frame (15). Upon translation, the frame-shifted peptide is often truncated and nonfunctional. In addition to altering downstream functions of the protein, the frameshift creates a new and foreign-appearing amino acid sequence that serves as a substrate for antigen processing and presentation via MHC class I and class II (27, 29, 66).

Taken together, these findings have contributed to the rational basis for novel primary and/or secondary immunoprevention strategies for dMMR colorectal cancer. In particular, although colorectal adenomas may lack the high TMB typically found at later stages of MMR-deficient tumorigenesis, the presence of robust and tissue-specific neoantigens indicates an opportunity to leverage the immune microenvironment to block progression into carcinomas.

Lynch syndrome serves as a disease model in which to understand how MMR deficiency and the immune microenvironment coevolve during tumorigenesis. At the earliest stage, comprehensive work by Kloor and colleagues (13) and Shia and colleagues (67) showed that MMR deficiency is present among a large proportion of nonneoplastic intestinal crypts in patients with Lynch syndrome. This observation, which is based on loss of MMR protein staining, may be explained by clonal expansion of histologically normal appearing crypts that acquired inactivating mutations in the remaining MMR gene allele. Furthermore, CD8⁺ intraepithelial lymphocytes were more abundant in these affected crypts, suggesting recognition of microsatellite-derived neoantigens in the normal crypt cells (68). Although such a hypothesis has not been definitively tested in experimental models, striking evidence comes from the observation that neoantigen-specific T cells and antibodies can be detected in the peripheral blood of Lynch syndrome without malignancy (which is more pronounced in patients with advanced MMR-deficient tumors that have higher TMB; refs. 69, 70).

Whether MMR-deficient intestinal crypts give rise to some, or all, MMR-deficient adenomas and carcinomas remains a subject of debate (15, 67, 68, 71), as some data suggest that MMR deficiency can also appear at a later step in tumorigenesis (72). Addressing this question has important implications for colorectal cancer prevention in Lynch syndrome. In particular, the prevalence of precancers (particularly adenomas) in Lynch syndrome is age- and gene mutation–dependent and ranges from 10.6% to 33% (73, 74), but only around 50% of these adenomas display MMR deficiency (75, 76). By contrast, histologically normal crypts with MMR deficiency are relatively abundant in the mucosa of healthy Lynch syndrome patients. This discrepancy raises the possibilities that either MMR-deficient adenomas develop from a different precursor lesion, or that a significant number of MMR-deficient crypts undergo “immunoediting” prior to transforming into adenomas.

Immunoediting is the process by which aberrant cell growth is halted and regressed by T-cell–mediated immunity (77–79). In cases where the lesion is not fully eradicated, immunoediting is followed by equilibrium and ultimately immune escape phases, where the remaining cells are able to evade detection by the immune system. It is therefore important to understand which intrinsic or extrinsic factors permit the formation of MMR-deficient adenomas despite early immune engagement. Notably, MMR-deficient adenomas tend to harbor significantly fewer mutations compared with carcinomas (80, 81) and yet infiltrating T cells directed against microsatellite-derived neoantigens are detectable at this stage as well (Fig. 1, precancer stage; refs. 70, 82). These observations suggested that neither having a low mutation burden nor a relatively low abundance of neoantigens fully explains immune evasion in colorectal precancers. Indeed, recent work by our group provided evidence of a robust immune activation signature in Lynch syndrome adenomas regardless of their mutation burden (81). By further characterizing the immune signature of Lynch syndrome adenomas, we also revealed global enrichment for CD4⁺ T cells and enrichment for FOXP3⁺ regulatory T cells in the subset with high mutation burdens (Fig. 1, advanced precancer stage). In addition, there was upregulation of both proinflammatory cytokines (IL12A) and checkpoint blockade (IFNG, CD274/PD1, and LAG3).

These findings correlate well with the known biology and clinical significance of immune activation in carcinomas. A high density of CD3⁺ cells in colorectal cancer is associated with longer cancer-specific survival (12, 83, 84). Similarly, the presence of CD45RO, CD8⁺, and CD4⁺ cells is associated with lower rates of metastasis, vascular, or perineural invasion, respectively (85). On the contrary, the presence of FOXP3-positive regulatory T cells in normal mucosa of patients with colorectal cancer portends a poorer prognosis (85).

Perhaps the best correlation may be found in the setting of advanced dMMR colorectal cancer, where treatment with single- or dual-checkpoint blockade now plays a pivotal role. Recent work by Turajlic and colleagues showed upregulation of multiple checkpoints in colorectal cancer tumors, including PD-1, CTLA-4, and lymphocyte activation gene 3 (LAG-3; refs. 86–89). As noted above, checkpoint blockade is effective for a plurality of patients with advanced dMMR colorectal cancer (response rates 30%–40%) and leads to durable disease control (19, 20).

Based on the evidence outlined above, at least two novel strategies for the prevention of Lynch syndrome–related colorectal cancer are currently under investigation (Table 1). First, the implication of adaptive immune resistance (PD-1, LAG3, and CTLA-4) in MMR-deficient colorectal adenomas (81) raises a key question of whether checkpoint blockade could halt the progression of such adenomas into carcinomas. The complete spectrum of factors that regulate the adaptive immune response in adenomas is yet unknown. However, given the availability of inhibitors already on the drug market and known efficacy for patients with advanced MMR-deficient colorectal cancer, the PD-1/PD-L1 axis is an especially completing target. Across multiple disease settings and cancer types, the safety profile of single- or dual-checkpoint blockade is relatively well established, as are common practice guidelines for management of immune-mediated toxicities (90). Nonetheless, clearly the safety and efficacy of such agents in the preventative setting require thorough and specific evaluation. Toward this end, a phase II single-arm study was recently opened in which adults with Lynch syndrome with MLH1 and MSH2 germline mutations (and therefore with maximum life-time risks for colorectal cancer development) and a history of partial colectomy due to advanced adenomas or colorectal cancer will receive nivolumab infusion every 3 months for up to 8 doses (clinicaltrials.gov ID NCT03631641; ref. 84). As a secondary prevention study, its primary objective is to determine the incidence of secondary adenomas and colorectal cancer among Lynch syndrome patients treated with anti–PD-1.

Second, anticancer vaccines hold significant promise not only in hereditary colorectal cancer, but in other solid tumors as well (91). For patients with Lynch syndrome, instability within coding and noncoding microsatellites yields a robust signature of tumor/tissue-specific neoantigens that may be targeted by predesigned vaccine libraries. In fact, this concept started to be explored in early 2000 triggered by meticulous efforts to catalogue the presence of instability in coding microsatellites using computational approaches coupled with labor-intensive validation via PCR-based methods (92) that led to early phase clinical trials using peptides identified as immunogenic. This approach has now recovered interest thanks to the development of improved pipelines for neoantigen identification that also incorporates immunogenicity predictions for both HLA-I and HLA-II presentation and the access to a wealth of genomic information from tumors (93–96). An example is the recent report from Scarselli and colleagues at NousCom on the identification of 209 frameshift peptide neoantigens shared across colorectal, gastric, and endometrial MSI tumors. Using a viral vector-based delivery system, the investigators observed strong immunogenicity of vaccine in mouse models (26, 97). These efforts are resulting in upcoming phase I clinical trials that are awaiting implementation and development in the following months (26, 77, 98, 99). The results of these investigations will be critical for defining the technical feasibility and safety of preventative vaccines for patients with Lynch syndrome.

The upregulation of immune checkpoints in Lynch syndrome–associated precancers despite a relatively low mutation burden suggests that neoantigen peptides are potent targets. Checkpoint blockade in the adjuvant setting may prove to be highly effective for secondary prevention in patients with Lynch syndrome or sporadic MMR-deficient colorectal cancers. However, the benefit-to-risk ratio will need to be clarified given the adverse events associated with PD-1 blockade. Neoantigen vaccination is another approach that is being used for advanced melanoma, glioblastoma, and other cancers, and repurposing this approach for primary prevention of MMR-deficient cancers in Lynch syndrome patients may be promising. We propose that there is also a rationale for combining vaccine therapy and checkpoint blockade under the hypothesis that a more specific and durable response could be generated to prevent malignant transformation of adenomas, thereby reducing risk of colorectal cancer recurrence and increasing cancer-specific survival.

E. Vilar is a consultant/advisory board member for Janssen Research and Development. No potential conflicts of interest were disclosed by the other authors.

This work was supported by grant Research Training in Academic Medical Oncology T32 Award CA009666-23 (US National Institutes of Health/National Cancer Institute) to J.A. Willis; Cancer Prevention Educational Award R25T CA057730 (US National Institutes of Health/National Cancer Institute) to K. Chang; R01 CA219463 (US National Institutes of Health/National Cancer Institute) and a gift from the Feinberg Family to E. Vilar; U01 CA233056 (US National Institutes of Health/National Cancer Institute) to S.M. Lipkin and E. Vilar; and P30 CA016672 (US National Institutes of Health/National Cancer Institute) to the University of Texas Anderson Cancer Center Core Support Grant.