Patients with GI cancers represent a clinical population with poor outcomes after conventional treatments and represent an active avenue of investigation to meet a major medical need. Immune checkpoint inhibitors (ICIs) targeting the programmed death-1 (PD-1) protein have shown significant efficacy and are now the treatment of choice for patients with mismatch repair-deficient (dMMR) metastatic colorectal cancer (1) and offer modest benefit in patients with biliary tract cancers when combined with chemotherapy. However, dMMR accounts for only a small subset of cases and despite advancements in systemic treatments such as immunotherapy, a paucity of additional validated predictive biomarkers in these GI malignancies can limit the potential use of ICIs in these disease settings.

The advent of liquid biopsies offers an emerging platform for molecularly guided treatment stratification to identify candidates well suited for ICIs. In this instance, DNA polymerase epsilon (POLE) could be utilised which encodes the catalytic subunit of the DNA polymerase epsilon enzyme and plays a critical role in DNA replication and repair (2). Mutations in the exonuclease domain of POLE (POLE EDM) are rare but have been identified across various malignancies, with the highest frequencies observed in endometrial cancer (6%–12%) (3) and CRC (1%–2%) (4), and a few instances in gallbladder cancer (5, 6). POLE EDM in endometrial cancer is associated with a strong intratumoural T-cell response, likely due to a high neoantigen load driven by an ultra-mutated tumour phenotype, contributing to favourable clinical outcomes (7). Similarly, in stage II colorectal cancer, somatic POLE EDM are linked to elevated intratumoural immune activity, driven partly by an increased tumour infiltration of CD8+ lymphocytes and CD8A expression (8). However, given the scarcity of data regarding POLE mutant clinical population, there is a notable absence of a standardised therapeutic protocol tailored to these patients. Additionally, there is no clear guidance on the duration of ICIs where data is extrapolated from dMMR CRC patients, which often warrants an assessment by clinicians within the appropriate clinical context. There is also a paucity of data to guide whether surgery should be offered in the event of complete metabolic response (CMR).

Our study provides a foundation to address these questions via successful treatment of POLE mutant patients achieving CMR. In this study, we analysed patients with gastrointestinal cancer (GI) including colorectal or gallbladder cancer who carried a POLE EDM and were managed with ICIs after diagnosis. Clinical management, radiologic outcomes and post-operative pathological evaluations were documented and reviewed. Additionally, gene mutation profiles and tumour mutation burden (TMB) were examined in these POLE EDM cases along with molecular characteristics, including POLE non-EDM and microsatellite stability (MSS) profiles.

ICIs are reported to be ineffective among the majority of cancer patients with advanced disease (15). In the context of metastatic CRC, favourable response rates ranging from 31.1% - 65% have been limited to a small subset (5%) of MSI-H cases (16). Similarly, the Phase 3 TOPAZ-1 trial demonstrated an overall response rate limited to 26.7% with durvalumab, gemcitabine and cisplatin in biliary tract cancer patients (17). Our study highlights the potential of POLE mutation as an alternative predictive biomarker to aid patient selection for ICIs, helping to bridge a critical gap in addressing a major unmet medical need.

Our results align with previous reports in colorectal cancer that POLE mutations, which impair DNA replication proofreading, are associated with an ultramutator phenotype (18). These tumours often present with early-onset colorectal or endometrial cancer and carry a higher mutational burden than typical hypermutated cancers. POLE mutations are related to other favourable predictive factors such as high expression of PD-L1, high TMB, and infiltration of CD8⁺ cells in the tumour microenvironment (TME) (19, 20). In endometrial cancer, POLE P286R is specifically implicated in exerting its anti-tumour effects via activation of the cGAS-STING pathway (21). In our case, the high TMB driven by POLE mutations may also reflect the observation of early and sustained ICI treatment response. Although previous POLE studies have primarily emphasized the role of pathogenic EDM mutations, recent evidence suggests the increasing prognostic significance of non-EDM POLE mutations (22). Chen et al. have demonstrated that POL-non-EDM tumours can exhibit a high neo-antigen burden, and their leukocyte infiltration and CD8+ T-cell infiltration can be similar to POL-EDM tumours (22). Thus, POL-non-EDM tumours could be associated with an activated immune response even in the context of low TMB, suggesting that certain POL mutations may elicit immune activation through mechanisms independent of a hypermutator phenotype. These mechanisms may include the generation of highly immunogenic neoantigens and direct engagement of immune-related pathways, such as those involved in DNA damage response. The study also demonstrated that POL-non-EDM patients achieved significantly higher overall response rates with ICIs than POL-WT patients (58.7% vs. 23.8%, p < 0.001), even in TMB-high and TMB-low subgroups (22). Similarly, compared with POL wild type patients, significantly longer OS and PFS were also observed in POL-non-EDM patients (22). In our study, patient’s # 2 and 3 harboured both POLE EDM and POL-non-EDM, potentially contributing to the observed clinical benefit with ICIs via complementary biological pathways. Taken together, these early results suggest that POL-non-EDMs could serve as potential biomarkers to identify ICI responders in addition to POL-EDMs. Additionally, the observation that the lack of POLE detection during ICI treatment can be predictive of radiological and metabolic response in CRC further highlights the importance of incorporating ctDNA based tests for serial disease monitoring in future standard of care clinical practice.

Conversely, we observed an exceptional treatment response and prolonged survival in gallbladder cancer despite a low TMB and MSS status. To our knowledge, this is the first case of a patient with GBC harbouring a POLE R150* mutation managed with ICIs and achieving improved outcomes in the current clinical context. Previous studies have assessed nonsynonymous variants in POLE and POLD1 as potential biomarkers for response to ICI suggesting that these variants were independent predictors of prolonged OS, after adjusting for microsatellite stability and cancer type (20). Our finding alludes to the presence of TMB/MSS independent pathways conferring immuno-sensitivity in POLE R150* GBC patients, a hypothesis which warrants further exploration in future investigations. Additionally, expected survival in the absence of other actionable mutations in unresectable GBC patients is less than 1 year. Despite relapse after an iCMR, this patient successfully attained an iPR with re-introduction of chemotherapy to immunotherapy. Presence of a rare POLE variant might have influenced the lack of sustained response as compared to other patients. However, it is also possible that the underlying tumour biology of GBCs is different than compared to other GI malignancies which can become immuno-sensitive again with chemotherapy re-challenge to regenerate a metabolic response.

Data regarding management of POLE mutant patients is sparse, but emerging evidence primarily in metastatic colorectal cancer suggests a molecularly guided management approach for ICI use. Recent studies have demonstrated that such patients treated with pembrolizumab using a similar dosing regimen as our patients promotes sustained radiologic and metabolic response and in some instances, post-treatment ctDNA clearance, leading to long term survival benefit in a challenging clinical population (23–25). Decisions regarding immunotherapy duration, timing of surgical intervention following an iCMR, and treatment discontinuation were individualized and made within specialized MDT meetings. In the absence of established guidelines for POLE-mutant gastrointestinal cancers, treatment duration was guided by sustained radiologic and metabolic response, clinical tolerance, and patient preference. Surgical resection was considered only after prolonged and durable metabolic remission, with the aim of confirming pathological response and consolidating disease control rather than as a predefined treatment endpoint. Interpretation of these findings must account for the absence of standardized clinical criteria guiding ICI duration, surgical intervention following an iCMR, or treatment re-challenge in patients with POLE-mutant gastrointestinal cancers. In current practice, such decisions are extrapolated largely from mismatch repair–deficient colorectal cancer data and are further influenced by individual patient factors, disease burden, response kinetics, and treatment tolerance. In our cohort, prolonged ICI therapy, delayed surgical intervention following sustained metabolic remission, and selective discontinuation were guided by multidisciplinary consensus rather than predefined protocols. Accordingly, our observations should not be interpreted as prescriptive but rather as illustrative of a personalized, response-adaptive management approach in a rare molecular subset.

The prevalence of Lynch syndrome (LS) is another important consideration among the GI cancer clinical population. However, there are currently no guidelines recommending LS screening based on POLE mutation alone in an MMR-proficient tumour. ESMO guidelines suggest MMR IHC and/or MSI testing on tumour specimens to identify prevalence of MLH1, MSH2, MSH6 or PMS2 loss before further germline testing for LS is recommended (26). Further LS germline testing is not indicated unless there are additional high-risk factors. In our evaluated patient cohort, all patients had MSS status, thus obviating the need for further germline testing. Additionally, among the two young CRC patients below 50 years of age, none of them had a family history of cancer and did not qualify for further screening based on the NHS pathway at our institution. However, patient # 3 requested germline testing as his brother was also very young, but no evidence of any germline drivers was identified in himself or his family. Additional factors where germline evaluation would be considered include presence of multiple LS-spectrum cancers or clinical suspicion for polymerase proofreading–associated polyposis (PPAP) which were not applicable to our patients. However, this would involve germline POLE or POLD1 mutations which is distinct from LS and associated with colorectal adenomas/cancer and some endometrial cancers. In this case, the germline test would target POLE/POLD1, not MMR genes, unless clinical history suggests otherwise.

Our study is limited by the small cohort of patients (n = 3) given the rarity of this mutation and patients, limiting our ability to perform more formal statistical analysis. Additionally, the variability in commercial molecular assays and thresholds for ctDNA and TMB analysis reflects laboratory reporting without universally harmonised thresholds for all tumour types, and interpretive cut-offs for immunotherapy prediction vary across studies and tumour contexts. Moreover, different ctDNA platforms were used for longitudinal monitoring in one patient, which may introduce analytical variability when comparing ctDNA dynamics over time. The report’s retrospective nature inherently limits the ability to draw causal inferences. The presence of heterogeneous treatment regimens and additional genetic mutations in CRC patients could have also influenced outcomes in these subjects and thus the effectiveness of ICI in POLE mutant subjects warrants further investigation in larger clinical populations with standardized molecular criteria.

This study provides a unique case set highlighting the potential role of POLE genetic biomarkers in clinical decision making in the first line setting for patients with advanced GI malignancies. Evidence of significant clinical response to ICIs is noteworthy, as is the durability of response. The promising radiological improvement among all patients are substantiated with FDG PET/CT imaging results which further corroborate the functional anti-tumour effect of ICIs via complete resolution of the metabolic activity of underlying disease. The median time to achieve an iCMR for our group of patients was 8.11 months. This highlights the importance of serial FDG PET/CT imaging in such patients to accurately assess tumour functional response status which may not be adequately captured via standard CT scans. The presence of novel POLE mutations allowed us to personalize their treatment based on their unique genomic profile, which has resulted in an exceptional survival outcome and quality of life in a challenging clinical population. Identification of genomic biomarkers of ICI response for these patients with a poor prognosis is important given the long-term benefit and reduced side-effects of ICIs compared to other systemic therapies, allowing patients to continue ICIs to ensure greater duration of treatment benefit. Incorporation of minimally invasive ctDNA based liquid biopsies in routine clinical practice is vital to identify POLE mutant patients and aid patient stratification for ICI treatment to optimize outcomes in a challenging GI clinical population. While our report demonstrates a favourable response to ICIs in POLE EDMs, these findings are hypothesis-generating and require validation in larger, prospective cohorts. Given the rarity of pathogenic POLE EDMs across tumour types, future studies should adopt a basket trial design, enrolling patients with advanced solid tumours harbouring confirmed POLE EDMs, irrespective of histology. Patients could be stratified by TMB and POLE-associated mutational signatures (e.g., SBS10a/b) to explore correlations with response. The primary endpoints could include objective response rate ORR and PFS, with secondary endpoints including OS and safety. Correlative translational studies could evaluate POLE non-EDMs, TMB, mutational signatures, immune infiltrates, and other predictive biomarkers, allowing assessment of the mechanistic basis of ICI responsiveness. Future prospective studies with predefined imaging and serial ctDNA sampling schedules will be necessary to enable robust longitudinal analyses correlating molecular dynamics with radiologic response and relapse. Such a basket trial, leveraging molecular rather than histologic selection, would provide a practical framework to prospectively evaluate the hypothesis that ICIs confer clinical benefit in patients with ultra-mutated POLE EDMs, including those with gastrointestinal cancers, while addressing the low prevalence of these mutations across individual tumour types.