Background

Gastroesophageal adenocarcinomas (GEAs), inclusive of esophageal, gastroesophageal junction (GEJ), and gastric cancers, are among the most prevalent malignancies worldwide. Although esophageal cancer and gastric cancer are the fifth and eighth most common cancers worldwide, respectively, approximately 45,000 new cases of esophagogastric cancers are diagnosed in the United States each year.¹ Efforts to molecularly classify GEA defined recurrent genomic patterns, including the subgroup of mismatch repair deficient (dMMR)/microsatellite instability high (MSI-H).^2,3 Although there is variability by age and anatomic location, dMMR/MSI-H is observed in 8% to 22% of nonmetastatic GEA and 3% to 5% of patients with advanced disease. Similar to colorectal cancer (CRC), dMMR/MSI-H status portends a more favorable prognosis compared with mismatch repair proficient (pMMR)/microsatellite stable (MSS) disease.^4–6 Clinical trials in GEA and CRC confirm the benefit for immune checkpoint inhibitor (ICI) therapy in dMMR/MSI-H tumors, and the role of immunotherapy in the locoregional and metastatic settings is rapidly evolving.^7,8 More recently, evidence supporting the use of neoadjuvant ICIs across dMMR/MSI-H cancer has emerged. Within nonmetastatic GEA, neoadjuvant immunotherapy has been associated with high rates of pathologic response and encouraging rates of event-free survival.⁹ Despite these rapid advancements, there are unresolved questions regarding testing and clinical management among this unique patient subgroup. This review summarizes progress in the diagnosis and management of patients with dMMR/MSI-H GEA.

MMR Biology

DNA polymerases accomplish an astonishing fidelity in replicating DNA, with copy errors occurring in only approximately 1 of 100,000 polymerized nucleotides.¹⁰ To address this low but constant error rate, DNA polymerases possess an intrinsic proofreading mechanism that identifies and corrects copy errors, reducing the rate by 2 orders of magnitude to approximately 1 of 10 million nucleotides. After DNA polymerase has passed through the newly synthesized DNA strand, MMR proteins act as an additional layer of protection by detecting, excising, and repairing remaining errors, such as insertions, deletions, and substitutions. This is analogous to the autocorrect function in word processing. Although there are several protein components constituting MMR machinery, a simplified model involves MSH2 binding to MSH6, forming a heterodimer called MutSalpha responsible for mismatch recognition and initiation of repair (Figure 1). MutSalpha binds double-stranded DNA at the site of mismatch and recruits additional protein-binding partners. MLH1 forms a heterodimer with PMS2, and this complex, named MutLalpha, binds MutSalpha, forming a ternary complex.¹¹ This binding interaction activates the endonuclease activity of PMS2, thus inducing single-strand DNA breaks, culminating in excision of the mismatched nucleotides from the daughter DNA strand, followed by integration of correctly matched nucleotides. MMR enzymes fail to correct errors in approximately 1 miscopied base out of 100, thereby reducing the final error rate of DNA replication to approximately 1 nucleotide in a billion.

Clinicopathologic Features of dMMR/MSI-H GEA

Defective MMR machinery can result in accumulation of excess mutations, which can facilitate tumor initiation and propagation partly via increased genomic instability.¹² In one of the largest cohorts reported comprising data from >11,000 tumor-normal pairs, including from the Cancer Genome Atlas (TCGA), the frequency of dMMR/MSI-H in gastric cancers was found to be the third most common, at 19%.^13,14 Approximately 1.63% of esophageal carcinomas were MSI-H, suggesting a high degree of variability based on anatomic site. Defective MMR can be inherited (Lynch syndrome or familial aggregation patterns) or sporadic. Sporadic cases result from loss-of-function point mutations or epigenetic silencing of MMR proteins and account for the majority of GEA hypermutated cases. Clinically, dMMR/MSI-H GEAs are enriched for advanced age, female sex, earlier stage (node negativity, fewer visceral metastases), and more distal location (more mid-to-distal stomach). Histopathologic associations include higher tumor cell pleomorphism, atypical growth patterns, mucinous features, prominent lymphoid infiltrate, and intestinal subtype^3,15–19 (Figure 2). Tumors with dMMR/MSI-H more frequently encode for non-self immunogenic neoantigens, which promote intratumor cytotoxic lymphocyte recruitment.^20,21 This is reflected by increased expression of PD-L1 and other checkpoint ligands that inhibit the immune microenvironment from tumor destruction.^21,22

MMR/MSI Testing

Several tests are available for identifying defective MMR; we recommend the 2023 ASCO overview for background.²³ Immunohistochemistry (IHC) demonstrating lack of expression of MLH1, PMS2, MSH2, and MSH6 classifies samples as dMMR. Reverse-transcriptase PCR (RT-PCR) detects abnormally high frequencies of repeat elements in segments known as microsatellites. Microsatellite lengths at 5 standardized loci (Bethesda panel) determine the microsatellite status.²⁴ Next-generation sequencing (NGS) can also detect MSI and offers the advantage of determining tumor mutation burden and other potentially relevant alterations in the same test. Several platforms are commercially available and previously reviewed.²³ Most NGS testing is ordered using tissue biopsy or resection specimens, and it remains the gold standard for molecular profiling of tumors. However, limitations include tissue availability and time needed for processing. Circulating tumor DNA assay (liquid biopsy) technology has added a new dimension to tumor profiling and is an increasingly reliable NGS test to determine MSI-H status in clinical practice. Depending on the platform used, the sensitivity of detecting mutations with lower allele frequencies may be lower in plasma compared with an adequate tissue sample; therefore, MSI status determination based off NGS testing of plasma may undercalculate mutational burden, leading to false-negative MSI-H and false-positive MSS results. However, liquid biopsy can still be useful in cases with insufficient remaining tissue for NGS testing or in patients who may need repeat NGS testing and do not wish to undergo repeat tumor biopsy.

GEAs have considerable genomic heterogeneity, which poses diagnostic and therapeutic challenges, including determination of dMMR/MSI-H status. Technical reasons accounting for MMR heterogeneity include variable sensitivity of testing modalities, antibody performance, pathologist experience, quality of tissue sampling, and quality of tissue processing.²⁵ However, discordance between testing methods is low, with one study of >3,000 cases of CRC finding a 1.6% discordance rate between IHC and molecular assays.²⁶ Kim et al²⁷ performed multiregional sampling in 79 dMMR/MSI-H tumors from patients with advanced gastric cancer and found that 8.9% demonstrated an admixture of MSS and MSI-H regions, confirming that there is a subset of dMMR/MSI-H tumors with significant heterogeneity. Interestingly, in a small phase II trial of single-agent pembrolizumab in advanced gastric cancer, the investigators observed MMR heterogeneity in a patient who did not experience response to treatment. This patient (who demonstrated rapid progression) harbored a tumor with confirmed heterogeneous staining of MLH1, underscoring that dMMR/MSI-H heterogeneity may have clinical implications for response to ICIs. Others have similarly shown heterogeneity in MMR protein staining as well as MSI-H determination by PCR/NGS.^14,16,17 Overall, dMMR/MSI-H heterogeneity is a rare phenomenon that should be considered, particularly in the clinical setting of rapid progression or primary resistance to ICIs. However, there are also many other purported mechanisms of resistance to ICIs.^28,29 Thus, there is inadequate evidence to date to support distinct management paradigms in this rare clinical scenario.

Management of Advanced dMMR/MSI-H GEA

Patients with dMMR/MSI-H GEA have an improved prognosis relative to pMMR/MSS disease, and management of advanced disease centers on shared decision-making with the goal of balancing quality and quantity of life.^2,3 The current pan-cancer dMMR/MSI-H FDA labeling for pembrolizumab is for patients whose disease has progressed following previous treatment. However, increasing evidence supports the frontline use of ICIs in dMMR/MSI-H GEA, which is our preferred approach. In examining features predictive of ICI benefit, dMMR/MSI-H status was the strongest predictor (better than PD-L1 expression) in a recent meta-analysis among modern phase III GEA trials.³⁰ In the subgroup of patients with dMMR/MSI-H tumors in the frontline phase III CheckMate 649 trial, 23 patients with MSI-H tumors received nivolumab and 11 received ipilimumab/nivolumab. When compared with 21 patients with MSI-H tumors who received chemotherapy (median overall survival [mOS], 12.3 months), nivolumab plus chemotherapy (mOS, 38.7 months) demonstrated improvement in OS (hazard ratio [HR], 0.38; 95% CI, 0.17–0.84).⁷ Ipilimumab/Nivolumab also improved survival compared with chemotherapy (mOS, not reached vs 10 months; HR, 0.28; 95% CI, 0.08–0.92)³¹ in the dMMR/MSI-H subgroup. These data, in addition to other trials for frontline dMMR CRC, highlight the ability of ICI-containing regimens to produce deep and durable responses for patients with advanced dMMR/MSI-H disease.^32–35 Similar to results from CheckMate 649, a post hoc analysis of all patients with MSI-H combined from the KEYNOTE-059, -061, and -062 trials demonstrated a survival benefit for pembrolizumab with or without chemotherapy compared with chemotherapy alone regardless of the line of therapy³⁵ (Table 1). Although KEYNOTE-062 suggested an improved progression-free survival (PFS) in patients who received chemotherapy with pembrolizumab (median PFS [mPFS], not reached; n=17) versus pembrolizumab monotherapy (mPFS, 11.2 months; n=14), the 24-month OS rate was similar regardless of whether chemotherapy was included (65% vs 71% of patients, respectively).

Table 1.

Clinical Activity of ICIs in Patients With Advanced dMMR/MSI-H GEA

VIEW TABLE

Collectively, it appears that the objective response rate (ORR) to ICIs in patients with dMMR/MSI-H GEA is 50% to 60%, with several phase III trials showing meaningful PFS and OS benefits^36–38 (Table 1). Decisions regarding use of ICI with or without chemotherapy should center on shared decision-making and a balance of toxicity and benefit. Society for Immunotherapy of Cancer (SITC) guidelines for ICI use across gastrointestinal cancers support this approach.³⁹ Given the paucity of patients with dMMR/MSI-H GEA, enrollment in clinical trials whenever feasible is preferred. Outside of a clinical trial, it is our preferred approach to incorporate ICIs in frontline management of advanced GEA, rather than preserving it for later lines of therapy. The optimal approach to patients with dMMR/MSI-H after progression on ICI-containing approaches is not currently known and remains a scientific and clinical knowledge gap.

dMMR/MSI-H Prognostic Implications in Patients With Nonmetastatic Disease

The favorable prognosis and outcomes in patients with advanced disease supported exploring dMMR/MSI-H status in patients with nonmetastatic disease. A post hoc analysis of the MAGIC trial first described the improved prognosis of dMMR/MSI-H versus pMMR/MSS gastric cancers.⁴⁰ Of 234 GEA tumors, 20 (8.5%) were MSI-H. The 5-year disease-free survival (DFS) rate was 71.8% for dMMR/MSI-H versus 52.3% for pMMR/MSS cases (P<.001). A similar difference was seen for OS, with a 5-year mOS rate of 77.5% versus 59.3% favoring dMMR/MSI-H cases (P<.001).⁵ A subsequent meta-analysis demonstrated improved 5-year OS for dMMR/MSI-H (60.5%) versus pMMR/MSS disease (49.1%), regardless of treatment received (HR, 0.71 [95% CI, 0.60–0.85]; P<.001).⁶ To address limited power and confirm prior observations, Pietrantonio et al⁵ performed an individual patient–level meta-analysis of data from the MAGIC, CLASSIC, ARTIST, and ITACA-S studies. In 1,556 trial patients, 7.8% (n=121) were found to have dMMR/MSI-H disease and the 5-year DFS and OS for this subgroup was 71.8% (HR, 1.88 [95% CI, 1.28–2.76]; P<.001) and 77.5% (HR, 1.78 [95% CI, 1.17–2.73]; P=.008), respectively.

Nonmetastatic dMMR/MSI-H GEA and Perioperative or Adjuvant Chemotherapy

There is strong suggestion that perioperative or adjuvant chemotherapy may not benefit patients with dMMR/MSI-H compared with surgery alone. In a large retrospective study (n=1,990; 8.5% dMMR/MSI-H) in patients with stage II/III GEA, adjuvant 5-FU–based chemotherapy failed to improve DFS, in contrast to a significant DFS benefit observed in those with MSS tumors.⁴¹ In a separate retrospective study comprising 1,276 patients with stage II/III GEA, the authors showed that patients with dMMR/MSI-H tumors who did not receive chemotherapy indeed demonstrated superior OS.⁴² The prognostic and predictive capacity for dMMR/MSI-H status was similarly investigated in post hoc analyses of CLASSIC and MAGIC, both randomized phase III trials studying 5-FU/platinum adjuvant and perioperative chemotherapy, respectively.^40,43 In CLASSIC, 6.8% (n=40) patients had dMMR/MSI-H tumors, and no DFS benefit with chemotherapy was seen in this subgroup (5-year DFS, 83.9% vs 85.7%; P=.931).⁴³ In MAGIC, 6.6% (n=20) of patients with dMMR/MSI-H tumors were analyzed and perioperative chemotherapy was found to be associated with worse OS (mOS, 9.6 months [95% CI, 0.1–22.5 months]) compared with patients with non-MSI tumors (mOS, 19.5 months [95% CI, 15.4–35.2 months]; HR, 2.18 [95% CI, 1.08–4.42]; P=.03).⁴⁰

In the meta-analysis by Pietrantonio et al,⁵ only patients with non-dMMR/MSI-H tumor status demonstrated a significant benefit for chemotherapy versus surgery only. Thus, the totality of data to date suggests that patients with dMMR/MSI-H tumors have favorable outcomes with surgery alone and that chemotherapy in the perioperative/adjuvant setting does not appear to improve upon surgery (and may even harm) this patient subgroup. Although the Pietrantonio analysis provides a strong rationale to test nonmetastatic GEA for MMR/MSI status, we recognize that there has not been prospective, randomized validation of the hypothesis that chemotherapy may be omitted. It is our practice to test all patients for dMMR/MSI-H status in order to have maximally informed discussions around nonmetastatic management.

ICIs in Patients With Nonmetastatic dMMR/MSI-H Tumors

Activity of ICIs in the advanced setting coupled with highly promising neoadjuvant data across dMMR/MSI-H solid tumors provides significant support for this strategy in GEA.^44–46 The phase II GERCOR NEONIPIGA study examined neoadjuvant ipilimumab and nivolumab for 3 months followed by curative-intent surgical resection and 9 months of adjuvant nivolumab in patients with MSI-H locally advanced, resectable gastric or GEJ adenocarcinomas.⁹ Of 30 patients who underwent surgery, the pathologic complete response (pCR) rate was 59%, compared with the historical standard pCR rate of approximately 15% to 20% with FLOT (fluorouracil/leucovorin/oxaliplatin/docetaxel)^9,47 (Table 2). Two patients who refused surgery had complete clinical responses confirmed with tumor-free endoscopic biopsies. Interim results from the ongoing DANTE trial of perioperative FLOT ± atezolizumab includes 7.8% (n=23) of patients with MSI-H tumors.⁴⁸ Atezolizumab was administered in conjunction with 4 cycles of preoperative and postoperative FLOT, and 8 additional cycles of maintenance atezolizumab every 2 weeks were given following completion of 4 cycles of postoperative chemoimmunotherapy. On interim analysis, patients with MSI-H tumors who received chemotherapy + atezolizumab achieved a pCR rate of 63% versus 27% in patients with MSS tumors. Reporting from cohort 1 of the phase II INFINITY trial studying neoadjuvant dual ICIs (durvalumab + tremelimumab) in MSI-H resectable gastric/GEJ tumors was recently presented. Patients received a single dose of tremelimumab and 3 doses of durvalumab (once every 4 weeks) followed by surgery. Interim analysis showed a pCR rate of 60% and encouraging early event-free survival data⁴⁹ (Table 2). Of note, this trial also contains a cohort in which nonoperative management is being evaluated in patients who experience a response to this regimen. These phase II trials demonstrate safety, no impact on surgical outcomes, high pCR rates, and encouraging event-free survival data. Additional trials are ongoing (IMAGINE, IMHOTEP, GASPAR; ClinicalTrials.gov identifiers: NCT04062656, NCT04795661, and NCT04736485, respectively). Beyond this encouraging data, we also await reporting from the phase III MATTERHORN⁵⁰ and KEYNOTE-585⁵⁰ trials studying perioperative chemoimmunotherapy in patients with resectable, locally advanced gastric/GEJ adenocarcinoma. Recently it was reported that pCR rates were improved with ICIs, but data on dMMR/MSI-H subgroups have not been reported.^50,51

Table 2.

Clinical Activity of ICIs in Patients With Nonmetastatic dMMR/MSI-H GEA

VIEW TABLE

Since the optimal strategy is not yet determined, it is best to enroll patients with locoregional dMMR/MSI-H GEA onto clinical trials where immunotherapy is incorporated into the perioperative management strategy, which is consistent with recent updates to the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Gastric Cancer.⁵² When a clinical trial is not available, it is reasonable to follow the GERCOR NEONIPIGA, DANTE, or INFINITY trial strategies; perioperative chemotherapy may be added to immunotherapy in patients with highly symptomatic disease when prompt tumor shrinkage is needed. At this time, data are insufficient to support a watch-and-wait strategy for patients with dMMR/MSI-H GEA who receive neoadjuvant immunotherapy. Future studies may evaluate neoadjuvant immunotherapy and a subsequent watch-and-wait strategy for patients with locoregional dMMR/MSI-H GEA, because esophagectomy and gastrectomy surgeries carry significant postoperative morbidity.

Conclusions and Future Directions

The treatment landscape of GEA is moving increasingly toward leveraging biologic understanding to improve outcomes for patients. A central component to this goal is biomarker testing, and we expect multiple biomarkers to ultimately impact earlier-stage management. Among defined biologic subgroups, dMMR/MSI-H GEA represents a unique population, representing 3% to 5% of patients with advanced disease and 8% to 22% of patients with nonmetastatic tumors. This patient subgroup is characterized by favorable prognosis and lesser (if any) benefit from perioperative and/or adjuvant standard chemotherapy, along with increased sensitivity to immunotherapy-based approaches. These differentiating features underscore the need for routine testing for MMR/MSI status among all patients with GEA, including expanding testing to include those with nonmetastatic disease. The promising outcomes with ICI-based approaches in nonmetastatic GEA are allowing us to explore questions that are common among the patient advocacy community, including whether there are patients who can safely forgo surgery. Evolving tools, including molecular imaging, cell-free DNA, and endoscopic improvements, may ultimately allow us to routinely bring nonoperative options to our patients with dMMR/MSI-H GEA, and we eagerly await data from ongoing studies. Although clinical activity with ICIs in the dMMR/MSI-H subgroup is highly encouraging compared with non-ICI approaches, resistant biology persists as a therapeutic challenge. Future clinical and translational work should focus on dissecting resistance mechanisms toward improved clinical outcomes. Although GEA remains a difficult cancer, present and future collaborative research and clinical trials offer new hope for patients.