Why This Study Matters
Immune checkpoint blockade has transformed care for metastatic colorectal cancers with deficient DNA mismatch repair (dMMR), and anti-PD-1 therapy is now standard of care for this population. Yet resistance remains common — roughly half of patients with metastatic dMMR colorectal cancer do not respond to first-line checkpoint blockade. Unlike several other solid tumor types, neither overall tumor mutation burden nor PD-L1 expression reliably predicts which dMMR patients will benefit, leaving clinicians without a strong biomarker to guide expectations or sequencing decisions.
This study, led by investigators at Mayo Clinic and published in Clinical Cancer Research, asked whether a more granular, quantitative measure of microsatellite instability (MSI) burden — rather than the conventional binary dMMR/MSI-high classification — could better predict response and survival after anti-PD-1 therapy, and whether additional genomic and immune-repertoire features could refine that picture further.
Study Design
Researchers retrospectively identified consecutive patients with dMMR metastatic colorectal cancer, confirmed by immunohistochemistry for mismatch repair proteins, who were treated with anti-PD-1 monotherapy at Mayo Clinic between 2015 and 2022. Pre-treatment, surgically resected primary tumors were profiled on the ImmunoID NeXT exome and transcriptome platform, generating tumor mutation data, gene expression, neoantigen characterization, HLA typing and loss-of-heterozygosity analysis, and T-cell and B-cell receptor repertoire profiling.
MSI burden was quantified computationally with the MSIsensor-pro algorithm as the percentage of unstable microsatellite loci detected by whole-exome sequencing, rather than relying on a simple positive/negative MSI-high call. Objective response was assessed by RECIST version 1.1, and associations with progression-free and overall survival were evaluated with Cox proportional hazards models.
Patient Population
The cohort's median age at the start of immunotherapy was 73.7 years (range, 62.4–83.0). Fifteen patients (38.5%) received anti-PD-1 monotherapy as first-line treatment, while 24 (61.5%) received it as second-line or later therapy after prior chemotherapy. Nineteen of 39 tumors (48.7%) carried a BRAF V600E mutation and 4 (10.3%) had a KRAS mutation; 20 patients (51.3%) had liver metastases. By germline testing, 27 patients had sporadic dMMR tumors, 9 had Lynch syndrome, and germline status was unknown in 3.
Of the 32 patients with evaluable outcome and imaging data, 14 (43.8%) were classified as responders — 7 complete responses and 7 partial responses — and 18 (56.3%) as non-responders (3 stable disease, 15 progressive disease). Median progression-free survival across the cohort was 14.9 months (range, 2.1–93.0), with 18 progression or death events recorded for that endpoint.
Primary Endpoint Results: MSI Burden and Outcome
The study's central finding was a strong, dose-like relationship between quantitative MSI burden and both response and survival. Responders had a significantly higher median MSIsensor-pro score than non-responders (18.6 vs 9.47; P = 0.018), and patients who achieved a complete response had the highest scores of all (median 22.6% vs 9.5% in non-responders; P = 0.0061).
By contrast, two biomarkers commonly used to guide immunotherapy in other tumor types showed no predictive value here. Overall neoantigen burden was not associated with progression-free survival (HR, 1.00; 95% CI, 0.97–1.02; P = 0.80) or overall survival (HR, 1.00; 95% CI, 0.98–1.03; P = 0.70), and tumor mutation burden was similarly non-predictive of response (median 125 in responders vs 48.4 in non-responders; P = 0.65) or progression-free survival (HR, 1.0; 95% CI, 1.00–1.00; P = 0.687).
These findings suggest that distinct features of the T-cell receptor repertoire may influence different aspects of clinical outcome, with repertoire evenness potentially facilitating initial tumor regression, whereas durable disease control following checkpoint blockade may depend on expansion of specific tumor-reactive clonotypes rather than maintenance of a highly diverse T-cell repertoire.
— Discussion, Sinicrope et al., Clin Cancer Res 2026;32:3073–83
Additional Biomarker Findings: Immune Repertoire, HLA, and Tumor Microenvironment
Beyond MSI burden, the study examined adaptive immune repertoire features and HLA genetics. Higher T-cell receptor alpha-chain evenness — reflecting a more balanced clonotype distribution — was associated with improved best overall response (OR, 4.07; 95% CI, 1.08–15.36; P = 0.038) and was significantly higher in responders than non-responders (median 0.85 vs 0.63; P = 0.015). Paradoxically, higher Rao–Stirling T-cell receptor diversity was associated with inferior progression-free survival for both the alpha chain (HR, 2.01; 95% CI, 1.12–3.60; P = 0.020) and beta chain (HR, 2.04; 95% CI, 1.13–3.66; P = 0.018), suggesting that very broad repertoires may lack the dominant tumor-reactive clones needed for durable control. Greater B-cell receptor diversity, in contrast, was associated with improved progression-free survival (Shannon entropy HR, 0.47; 95% CI, 0.23–0.95; P = 0.035).
Among HLA class I alleles, HLA-B*07:02 was associated with higher odds of objective response (OR, 19.0; 95% CI, 1.80–468.2; P = 0.025) but, unexpectedly, with worse progression-free survival (HR, 5.29; 95% CI, 1.56–18.00; P = 0.008) — a discordance the authors attribute to possible immunoediting and the emergence of antigen-presentation escape after an initial strong response. HLA-B*57:01 and HLA-C*02:02 were each associated with inferior progression-free survival but not with response. Overall HLA class I gene expression and HLA loss-of-heterozygosity were not significantly associated with outcomes.
Tumor microenvironment analysis using gene-expression deconvolution found that higher intratumoral NK-cell infiltration was associated with shorter progression-free survival (HR, 1.78; 95% CI, 1.14–2.78; P = 0.01), and higher CD8+ T-cell (HR, 2.24; 95% CI, 1.13–4.45; P = 0.02) and cytotoxic lymphocyte infiltration (HR, 2.03; 95% CI, 1.13–3.63; P = 0.02) were both associated with worse overall survival — a counterintuitive pattern the authors link to markers of T-cell exhaustion that correlate with these immune-cell populations. Consistent with this, non-responders showed a significantly higher immune checkpoint gene expression signature — combining PD-L1, CTLA-4, TIM-3, LAG-3, PD-1, PD-L2, and TIGIT expression — than responders (P = 0.005).
Subgroup Analyses
| Subgroup | Reported Finding |
|---|---|
| BRAF V600E-mutant tumors | Unstable microsatellite loci percentage was significantly higher in responders than non-responders, and higher MSIsensor-pro quartiles (Q2–4 vs Q1) were associated with significantly improved progression-free survival; this association was not seen in BRAF wild-type tumors. Exact hazard ratios and confidence intervals for this comparison were not available in the accessible full text. |
| Liver metastases present vs absent | No statistically significant differences in objective response or survival were observed between subgroups; exact statistics were not reported. |
| First-line vs second-line-or-later checkpoint blockade | Higher MSIsensor-pro quartiles (Q2–4 vs Q1) were significantly associated with improved progression-free survival in both treatment-line subgroups; exact statistics for this stratified comparison were not reported. |
Per the study's own limitations, these subgroup analyses were exploratory and the overall cohort of 39 patients is small; the authors describe their findings as hypothesis-generating and call for confirmation in larger, independent cohorts.
Safety Profile
Not reported. This is a retrospective, correlative biomarker analysis of tumor and immune features in patients who received standard-of-care anti-PD-1 therapy; the publication does not report treatment-related adverse event rates or a formal safety analysis, and none should be inferred.
Interpretation and Broader Context
The authors conclude that quantitative MSI burden, measured continuously rather than as a binary MSI-high call, tracked consistently with response and survival benefit from anti-PD-1 therapy in this cohort, and that the relationship appears to be mediated mechanistically through increased neoantigen clonality rather than sheer neoantigen quantity. They note that, because all patients in the study received PD-1 blockade, the design cannot formally distinguish a predictive biomarker — one that specifically identifies immunotherapy responders — from a purely prognostic one associated with better outcomes regardless of treatment, an important caveat for interpreting the results.
The identification of specific HLA class I alleles associated with outcomes, and the counterintuitive associations between certain immune-cell infiltration measures and worse survival, add mechanistic texture to ongoing efforts to understand resistance to checkpoint blockade in dMMR colorectal cancer, particularly around T-cell exhaustion. The authors frame these results as hypothesis-generating given the single-center, retrospective design and modest sample size, and call for validation in independent cohorts before any of these biomarkers could inform clinical decision-making.
In a 39-patient Mayo Clinic cohort of dMMR metastatic colorectal cancer treated with anti-PD-1 therapy, quantitative MSI burden — the percentage of unstable microsatellite loci, not just a binary MSI-high call — tracked strongly with response and survival: each additional percentage point lowered the risk of progression by about 10% and complete responders carried roughly double the MSI burden of non-responders. Established biomarkers like overall tumor mutation burden and total neoantigen count showed no predictive value in this population. The findings are hypothesis-generating from a small, single-center, retrospective series that cannot separate a predictive biomarker from a purely prognostic one, and the authors call for validation in larger, independent cohorts before quantitative MSI burden could inform treatment decisions.