The Impact of Molecular Subtype on Efficacy of Chemotherapy and Checkpoint Inhibition in Advanced Gastric Cancer

Gastric cancer is the fifth most common type of cancer and the third leading cause of death from cancer globally (1). A combination of fluoropyrimidines and a platinum agent (with trastuzumab for HER2-positive cases) as first-line therapy is the standard treatment in patients with advanced gastric cancer (AGC; refs. 2–6). Second-line therapy includes taxane agents with or without ramucirumab or irinotecan (7–12). In third-line or later therapy, two anti–PD-1 inhibitors have been approved for AGC: pembrolizumab by the U.S. Food and Drug Administration (FDA) for PD-L1–positive tumors, and nivolumab in Asian countries, regardless of PD-L1 status (13, 14). Despite the recent increase in treatment options, the median overall survival time of patients with AGC treated with a combination and sequential regimen involving these agents was less than 15 months.

The molecular characterization of gastric cancer has recently been rapidly evolving. As shown in The Cancer Genome Atlas and The Asian Cancer Research Group (15, 16), the microsatellite instability–high (MSI-H) subtype exhibits a greater number of mutations in multiple genes and hypermethylation compared with other subtypes, which contributes to the enhanced expression of neoantigens and subsequent high infiltration of CD8⁺ T cells. Indeed, the FDA-approved pembrolizumab for patients with MSI-H/mismatch repair-deficient (MMR-D) solid tumors including AGC based on the durable response shown in several trials (14, 17–19). Also, Epstein–Barr virus- (EBV)-positive status has been reported to be associated with a high amplification rate of the CD274 gene (which encodes PD-L1) and the PDCD1LG2 gene (which encodes PD-L2; refs. 15, 16), as well as high expression of these immune checkpoints, which might lead to favorable outcomes for AGC following the use of immune checkpoint inhibitors (20). However, little is known about the association between molecular subtypes and the efficacy of standard chemotherapy or the relative efficacy of immune checkpoint inhibitors for AGC. Therefore, we investigated the efficacy of standard first- and second-line chemotherapy for various clinical molecular subtypes: MMR-D, EBV⁺, HER2⁺, and others (all negative). The impact of these subtypes on the efficacy of subsequent anti–PD-1 therapy was also analyzed in this study.

We performed a single-institute study to evaluate the efficacy of standard first- (fluoropyrimidine + platinum ± trastuzumab) and second-line (taxanes ± ramucirumab) chemotherapy, and subsequent anti–PD-1 therapy in patients with AGC of four clinical molecular subtypes: MMR-D, EBV⁺, HER2⁺, and others (all negative). The eligibility criteria were as follows: (i) an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–2; (ii) histologically proven, unresectable, locally advanced, or metastatic gastric adenocarcinoma; (iii) adequate bone marrow, hepatic, and renal function; (iv) received systemic chemotherapy from October 2015 to July 2018; and (v) with available molecular features (MMR/EBV/HER2 tests were all required, while PD-L1 expression or gene alterations were not). Patients with recurrence within 6 months of adjuvant chemotherapy were excluded from first-line analysis. All patients provided written informed consent for the biomarker analysis. The study protocol was approved by the Institutional Review Board at the National Cancer Center Japan. This study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki.

Molecular characteristics, such as the status of HER2, PD-L1, mismatch repair (MMR), and EBV, were analyzed using formalin-fixed paraffin-embedded tissue specimens from archival tissue samples. As described previously (21), IHC using a monoclonal anti-HER2 antibody [PATHWAY HER2 (4B5), Ventana] and FISH using the PathVysion HER-2 Probe Kit (Abbott Laboratories) were performed to assess HER2 status. HER2 positivity was defined as an IHC of 3+ or an IHC of 2+ and a positive FISH result according to College of American Pathologists/American Society of Clinical Oncology criteria (22). PD-L1 expression was assessed by IHC using an anti-PD-L1 rabbit mAb (Clone SP142 or SP263, Ventana) and measured using the combined positive score (CPS), defined as the number of PD-L1–positive cells (tumor cells, lymphocytes, and macrophages) as a proportion of the total number of tumor cells multiplied by 100. MMR status was assessed by IHC using mAbs for anti-mutL homolog 1 (MLH1, ES05), anti-mutS homolog 2 (MSH2, FE11), anti-postmeiotic segregation increased 2 (PMS2, EP51), and anti-mutS homolog 6 (MSH6, EP49; Agilent Technologies); tumors that lacked MLH1, MSH2, PMS2, or MSH6 expression were considered MMR-deficient (MMR-D), whereas those that maintained expression of MLH1, MSH2, PMS2, and MSH6 were considered MMR-proficient (MMR-P). Chromogenic in situ hybridization for EBV-encoded RNA (EBER) using fluorescein-labeled oligonucleotide probes (INFORM EBER Probe, Ventana) was performed to evaluate EBV status (23). All specimens in this study were reviewed by T. Kuwata.

Genomic alterations were analyzed using Oncomine Comprehensive Assay version 3 or Oncomine Cancer Research Panel (Thermo Fisher Scientific). TMB was defined as the number of nonsynonymous mutations, including indels, per megabase (mt/Mb) of genome examined in tumor tissue.

We evaluated the objective response rate (ORR), the disease control rate (DCR), and progression-free survival (PFS). Tumor response was assessed in patients with measurable lesions using the Response Evaluation Criteria in Solid Tumors version 1.1. The ORR was defined as the proportion of patients with the best overall response of complete response (CR) or partial response (PR). The DCR was defined as the proportion of patients with the best overall response of CR, PR, or stable disease. The PFS was defined as the interval from the start of treatment until disease progression or death from any cause or the last follow-up visit and estimated by the Kaplan–Meier method. The χ² test or Fisher exact test was used to compare baseline characteristics and response rates. PFS was also compared between each molecular subtype using the Cox proportional hazards model and presented as an HR with 95% confidence intervals (CIs). Confounders in the multivariate analyzes of PFS were prespecified according to previously reported prognostic factors in AGC receiving standard chemotherapy (24–28); ECOG PS (1–2 vs 0), liver metastasis (yes vs no), peritoneal metastasis (yes vs no), number of metastatic sites (2 or more vs 1), prior gastrectomy (no vs yes), ALP [≥ upper limit normal (ULN) vs < ULN], histologic type (diffuse vs intestinal), and a measurable lesion (yes vs no). Statistical analyzes were performed using SPSS Statistics software V22 (IBM). All tests were two-sided; P < 0.05 was considered to indicate statistical significance.

A total of 410 patients who received systemic chemotherapy from October 2015 to July 2018 were enrolled in this study. The proportions of patients who went on to second-line and third line treatment were 94% and 56%, respectively. All 410 specimens were collected from primary tumor samples before chemotherapy; 338 were biopsy specimens, and 72 were surgical specimens. All 410 patients were tested for MMR/EBV/HER2. Furthermore, results of PD-L1 expression and genomic alterations were available in 375 and 259 patients, respectively. PD-L1 expression and genomic alterations were analyzed using same tumor blocks. PD-L1 expression was assessed by IHC using mainly SP263 (82%) and partially SP142 (18%). Of the 410 patients in the overall population, MMR-D, EBV⁺, HER2⁺, and all-negative subtypes were identified in 24 (5.9%), 17 (4.1%), 56 (13.7%), and 313 (76.3%) patients, respectively, with no overlapping among molecular subtypes. Patient characteristics are shown in Table 1. In patients with MMR-D, the frequency of lymph node metastasis, CPS ≥10, TMB ≥10, PIK3CA mutation, and KRAS mutation was significantly higher compared with their frequency in all-negative subtype patients. One patient with MMR-D with appendix cancer had lack of MSH2 and MSH6 expressions and underwent germline testing for MMR genes, resulting in a diagnosis as lynch syndrome with MMR gene mutation. All other 23 patients with MMR-D with lack of MLH1 and PMS2 expression did not undergo germline testing. In EBV⁺ patients, the frequency of PIK3CA mutation, CPS ≥1, and CPS ≥10 was higher compared with their frequency in all-negative subtype patients, although there were no statistically significant differences.

Among 410 patients, 285 patients received standard first-line chemotherapy and 252 patients received standard second-line chemotherapy (Fig. 1): Among 24 patients with MMR-D, 16 patients received standard first-line chemotherapy (the other eight patients received nonstandard chemotherapy) and 11 patients (including two patients who received nonstandard first-line chemotherapy) received standard second-line chemotherapy; Among 17 EBV⁺ patients, nine patients received standard first-line chemotherapy (the other eight patients received nonstandard chemotherapy) and 11 patients (including five patients who received nonstandard first-line chemotherapy) received standard second-line chemotherapy. In 243 patients with measurable disease who received standard first-line chemotherapy, the ORRs were 31%, 62%, 60%, and 49% in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (Table 2A). The median PFS times with first-line chemotherapy were 4.2, 6.0, 7.5, and 7.6 months in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (Fig. 2A). PFS in patients with MMR-D tended to be shorter compared with that in all-negative patients, with an HR of 1.64 (95% CI, 0.95–2.84; P = 0.075).

This difference was found to be statistically significant by multivariate analysis after adjustment for confounding factors (HR, 1.97; 95% CIs, 1.09–3.53; P = 0.022), suggesting shorter PFS following first-line chemotherapy in patients with MMR-D (Table 3). No other differences between the subgroups were statistically significant.

In 238 patients with measurable disease who received standard second-line chemotherapy, the ORRs were 30%, 40%, 22%, and 27% in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (Table 2B). The median PFS times following second-line chemotherapy were 3.4, 6.6, 3.7, and 3.9 months in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (Fig. 2B). There were no significant differences in efficacy seen among the four subtypes following second-line chemotherapy.

Among 410 patients, 110 patients received subsequent anti–PD-1 antibodies (Fig. 1): Among 24 patients with MMR-D, 12 patients received subsequent anti–PD-1 therapy (four patients in the second line and eight patients in the third or later line); Among 17 EBV⁺ patients, six patients received subsequent anti–PD-1 therapy (two patients in the second line and four patients in the third or later line). In 110 patients with measurable disease receiving subsequent anti–PD-1 antibodies, the ORRs were 58%, 33%, 7%, and 13% in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (P = 0.001, MMR-D vs all negative; Table 2C). The median PFS times were 13.0, 3.7, 1.6, and 1.9 months in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (Fig. 2C). Patients with MMR-D had a significantly longer PFS following anti–PD-1 therapy compared with the PFS in all-negative patients (HR, 0.14; 95% CI: 0.06–0.36; P < 0.001). The patient diagnosed as Lynch syndrome showed durable objective response to anti–PD-1 therapy (shown in Fig. 3 as patients no. 6). Patients with EBV⁺ subtypes also tended to show longer PFS compared with PFS in all-negative patients (HR, 0.48; 95% CI, 0.22–1.05; P = 0.064). Among patients receiving subsequent anti–PD-1 therapy, the percentage of patients who showed longer PFS following anti–PD-1 therapy compared with earlier-line chemotherapy were 83%, 33%, 14%, and 26% in MMR-D, EBV⁺, HER2⁺, and all-negative subtypes, respectively (Swimmer plot of PFS times with each line chemotherapy in 12 patients with MMR-D who received subsequent ant-PD-1 therapy is shown in Fig. 3).

Comparisons of clinical and molecular features between responders (CR or PR) and nonresponders (SD or PD) in each subtype with anti–PD-1 therapy were also available in Supplementary Table S1. In all-negative subtype, higher ORRs were observed in patients with PS 0 than in those with PS ≥1 (25% vs 6%) and in patients with TMB ≥10 than in those with TMB <10 (17% vs 9%), although these were not statistically significant (Supplementary Table S1D).

We investigated the efficacy of systemic chemotherapy in patients with AGC according to their different clinical molecular subtypes. To our knowledge, this is the first report to provide detailed information on the impact of these subtypes on responses to first- or second-line standard chemotherapy and the relative efficacy of subsequent anti–PD-1 therapy for AGC.

In our patient cohort, the MMR-D subtype was associated with significantly shorter PFS and a lower ORR following first-line chemotherapy compared with other subtypes. This observation is consistent with a small earlier study that demonstrated shorter PFS following first-line chemotherapy for AGC patients with MSI-H compared with those with non-MSI-H tumors (29). Also, a recent meta-analysis of four randomized trials (MAGIC, CLASSIC, ARTIST, and ITACA-S) showed that patients with MSI-H locally advanced gastroesophageal cancer did not exhibit any survival benefit from perioperative cytotoxic chemotherapy (30). Several preclinical studies using cancer cell line models demonstrated that a loss of MMR results in an inability to detect DNA damage and activate apoptosis of tumor cells, leading to resistance to DNA-damaging cytotoxic agents, such as fluoropyrimidine and platinum compounds (31–34). However, a phase III study, KEYNOTE-062, which compared the efficacy of cytotoxic agents with that of pembrolizumab monotherapy in patients with untreated AGC showed that the ORRs to first-line standard chemotherapy in MSI-H AGC were almost equal to those of the whole population (35). Thus, further analysis regarding the impact of MSI/MMR status on the efficacy of combinations of fluoropyrimidines and a platinum agent is needed. In a second-line setting, we found that there were no significant differences in efficacy among the clinical molecular subtypes. Taxanes exert antitumor effects by impairing microtubule function and inhibiting cell division, and its process is not associated with MMR function, which might be one of the reasons for the almost equal efficacy of second-line chemotherapy for MMR-D AGC and the other subtypes.

Consistent with previous reports (19, 20, 29), subsequent anti–PD-1 therapy was more effective in patients with AGC with MMR-D. Importantly, most patients with MMR-D showed an objective response and longer PFS compared with earlier-line chemotherapy. These results support the earlier use of immune checkpoint inhibitors for AGC patients with MSI-H/MMR-D.

In a recent phase II study of pembrolizumab, extremely high ORRs (100%) were reported in 6 patients with EBV⁺ AGC (20), while one of four patients (25%) with EBV⁺ AGC achieved a PR in a phase Ib/II study of toripalimab (a PD-1 antibody; ref. 36). All responders with EBV⁺ in these studies showed PD-L1–positive status. In our study, two (one with CPS ≥10 and one with unknown CPS status) of six patients (33%) with EBV⁺ achieved a PR. Among four patients with CPS evaluation, one of four patients (25%) with CPS ≥ 1 achieved a PR, while one of two patients (50%) with CPS ≥ 10 did. The impact of EBV status as well as PD-L1 expression on the efficacy of immune checkpoint inhibitors should be evaluated in a larger cohort.

Although an exploratory subgroup analysis of a phase III trial, ATTRACTION-2, showed that nivolumab improved clinical outcomes compared with placebo regardless of prior trastuzumab use in patients with AGC (37), HER2 alterations in gastric cancer were associated with decreased immunogenicity in terms of immune-related gene mRNA expression, immune infiltrates, and neoantigen level (38). Indeed, patients with HER2⁺ AGC in our study had shorter PFS and a lower ORR with anti–PD-1 therapy, which warrants further evaluation in a larger cohort.

There were some limitations inherent in this study. First, this was a single-institution study with a limited sample size. Owing to the small number of patients with MMR-D- and EBV-positive status, we could not evaluate the exact impact of these molecular factors on the efficacy of standard chemotherapy, which warrants further evaluation with a larger cohort. Second, timing of radiographic evaluation was not specified prospectively due to retrospective nature of the study. Third, PD-L1 expression or gene alterations were not analyzed in all patients who received systemic chemotherapy. Fourth, we used SP263 (82%) or SP142 (18%) assay for PD-L1 assessment rather than 22C3 assay. A previous study showed that a higher proportion of PD-L1+ status (≥1% in tumor cells or immune cells) was observed in patients with lung cancer with SP263 than in those with 22C3 (39). This difference in staining might be associated with the relatively higher PD-L1 rate in this study. Finally, overall survival was not evaluated in this study as it was affected by sequential therapies.

In conclusion, MMR-D might result in poor clinical outcomes with first-line chemotherapy for patients with AGC compared with other subtypes. However, subsequent anti–PD-1 therapy achieved favorable outcomes compared with outcomes following prior chemotherapy in most patients with MMR-D, supporting the earlier use of immune checkpoint inhibitors.

D. Kotani reports receiving speakers bureau honoraria from Takeda, MerckBiopharma, Chugai, Lilly, Taiho, and Sysmex. Y. Kuboki reports receiving speakers bureau honoraria from MSD and Ono. H. Taniguchi reports receiving speakers bureau honoraria from Ono Pharma, MSD, and Taiho. T. Kojima is an employee/paid consultant for Bristol-Myers Squibb, and reports receiving commercial research grants from Chugai Pharma, MSD, Astellas Amgen BioPharama, Shionogi, and Ono Pharmaceutical. T. Yoshino reports receiving other commercial research support from Novartis Pharma K.K., MSD.K.K., Sumitomo Dainippon Pharma Co., Ltd., Chugai Pharmaceutical CO., LTD., Sanofi K.K, Daiichi Sankyo Company, Limited, PAREXEL International Inc., Ono Pharmaceutical CO., LTD., GlaxoSmithKline K.K., and Boehringer Ingelheim Japan, Inc. T. Kuwata reports receiving commercial research grants from Ono and Daiichi-Sankyo, and speakers bureau honoraria from Chugai, Taiho, and MSD. K. Shitara is an employee/paid consultant for Astellas Pharma, Lilly, Bristol-Myers Squibb, Takeda, Pfizer, Ono Pharmaceutical, Taiho, MSD, Novartis, Abbvie, and GlaxoSmithKline, reports receiving commercial research grants from Lilly, Ono Pharmaceutical, Dainippon Sumitomo Pharma, MSD, Daiichi Sankyo, Taiho Pharmaceutical, Chugai Pharma, Astellas Pharma, and Medi Science, and holds ownership interest (including patents) in Novartis, Abbvie, and Yakult. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Kubota, A. Kawazoe, T. Yoshino, K. Shitara

Development of methodology: Y. Kubota, A. Kawazoe, T. Yoshino, K. Shitara

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Kubota, S. Mishima, Y. Kuboki, H. Taniguchi, T. Kojima, T. Yoshino, T. Kuwata, K. Shitara

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Kubota, A. Kawazoe, A. Sasaki, Y. Nakamura, H. Taniguchi, T. Doi, T. Yoshino, K. Shitara

Writing, review, and/or revision of the manuscript: Y. Kubota, A. Kawazoe, A. Sasaki, K. Sawada, Y. Nakamura, D. Kotani, Y. Kuboki, H. Taniguchi, T. Kojima, T. Yoshino, G. Ishii, K. Shitara

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Kubota, A. Sasaki, S. Mishima, K. Shitara

Study supervision: T. Doi, K. Shitara

This study was supported by a research funding from National Cancer Center Hospital East.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.