Abstract
Background Paradoxically, Helicobacter pylori-positive (HP+) advanced gastric cancer patients have a better prognosis than those who are HP-negative (HP-). Immunologic and statistical analyses can be used to verify whether systematic mechanisms modulated by HP are involved in this more favorable outcome.
Methods A total of 658 advanced gastric cancer patients who underwent gastrectomy were enrolled. HP infection, mismatch repair, programmed death-ligand 1 (PD-L1), and CD4/CD8 proteins, and microsatellite instability were analyzed. Overall survival (OS) and relapse free survival (RFS) rates were analyzed after stratifying clinicopathological factors. Cox proportional hazards regression analysis was performed to identify independent prognostic factors.
Results Among 491 cases that were analyzed, 175 (36%) and 316 (64%) cases were HP+ and HP−, respectively. Analysis of RFS indicated an interaction of HP status among the subgroups for S-1 Dose (P=0.0487) and PD-L1 (P=0.016). HP+ patients in the PD-L1 group had significantly higher five-year OS and RFS than HP- patients (81% vs. 68%; P=0.0011; HR 0.477; and 76% vs. 63%; P=0.0011; HR 0.508, respectively). The five-year OS and RFS was also significantly higher for HP+ compared to HP- patients in the PD-L1-/S-1-reduced group (86% vs. 46%; p=0.0014; HR 0.205; 83% vs. 34%; p=0.001; HR 0.190, respectively). Thus, HP status was identified as one of the most potentially important independent factors to predict prolonged survival.
Conclusion Modulation of host immune system function by HP may contribute to prolonged survival in the absence of immune escape mechanisms of gastric cancer.
Introduction
Gastric cancer (GC) is the third most frequent cause of cancer death worldwide (1, 2). In Japan, the five-year overall survival (OS) for early GC exceeds 95%, but that for advanced GC is only between 40 and 70% (3-5). S-1 post-operative adjuvant chemotherapy provides a significant survival improvement for Stage II/III GCs, based on results of a series of clinical trials (6-8). In daily practice, the five-year recurrence rate after curative gastrectomy for Stage II/III GCs remains between 20 and 40% (9). Although there still is no gold standard molecular marker for advanced GC patient treatment stratification, previous studies, including ours, demonstrated that HP+ advanced GC patients have a better survival rate than HP- patients (10-14). Interestingly, similar observations have been reported from various regions around the world including Taiwan, Germany, the United States, Korea, and Japan (10-14). Advanced GC patients in these regions are likely to have received different treatments, and to be of different ethnic groups and races, in addition to being infected with different HP strains. In fact, the HP genotype cagA, is largely divided into Western and East Asian groups based on genetic polymorphisms of EPIYA motifs coded by the SHP-2 domain of cagA (15). This frequently found polymorphism in East Asian strains is suggested to be associated with a high risk of gastric carcinogenesis (16). However, most studies on HP in the context GC havefocused on carcinogenesis or early-stage cancers. The aforementioned series of previous reports suggest that HP infection may have a specific impact on treatment efficacy or anti-tumor mechanism in advanced GC patients. We reported for Japanese patients with advanced GC, five-year OS rate of 77% and 69% for HP+ than HP- patients, respectively (13). HP infection can be a risk factor of GC development from intact gastric mucosae, but the effect of HP infection after GC progression and its functional effect on advanced GC treatment efficacy is unclear (17). Importantly, the large majority (>97%) of people infected with HP never develop GC (18). If HP infection functionally affects GC progression and treatment, understanding the mechanism of the effects would be informative for the majority of the HP+ population.
HP infection in the gastric mucosa activates an innate immune response that is followed by cell-mediated and humoral immunity of the host (19, 20). At a molecular level, HP infection triggers injection of CagA protein into gastric mucosal cells, which eventually promotes cellular protein degradation (21). Although HP infection processes are thought to result in DNA damage, that may arise following HP-dependent downregulation of mismatch repair (MMR) gene expression (22, 23). MMR protein levels are reported to increase in response to HP infection, suggesting that MMR protein levels could parallel the degree of DNA damage (24). In addition, GC patients showing MSI, a hallmark genotype of DNA damage repair deficiency, have a better prognosis than those having a microsatellite stable (MSS) genotype (25, 26). Interestingly, an association between HP+/PD-L1+ and MSI in GC patients suggests a possible link between HP infection and PD-L1 expression (27-29).
In the present study, we sought to clarify whether systematic function induced by sustained HP infection may be involved in prolonged survival of patients with advanced GC. Analyses were carried out in the context of results for post-operative adjuvant chemotherapy obtained by the Northern Japan Gastric Cancer Study Consortium (13).
Methods
Patients
This study retrospectively analyzed 658 patients with primary GC who underwent R0 gastrectomy at one of 13 institutions located in Northern Japan (Hokkaido and Tohoku area) between 2009 and 2013 (9). Sample and clinical information were collected by the Northern Japan Gastric Cancer Study Consortium where the infection state of HP had previously been pathologically confirmed (13). Patient characteristics of the subset for the current study are shown in Supplementary Table 1. The multi-center study protocol was approved by the Institutional Review Board (IRB) of Iwate Medical University (H24-132, HGH24-12, and HG2020-018), where all analysis was carried out. Ethical approval for conducting this study was given by all IRB committees of all 12 participating institutions listed in the Northern Japan Gastric Cancer Study Consortium at the Note section.
Immunohistochemistry
A tissue microarray (TMA) including 658 gastric cancer specimens was made to perform immunohistochemistry (IHC) (30). Of these, the HP infection status was available for 491 cases (13). In the present study, immunostaining reactions were performed using primary antibodies against PD-L1 (22C3, Dako, Santa Clara, CA, USA, 1:50), CD4/CD8 (4B12 and SP16, Biocare Medical, Concord, Pacheco, CA, USA, 1:100), and MMR proteins including MLH1 (ES05, Agilent Technologies, Santa Clara, CA, USA, 1:80), MSH2 (FE11, Agilent Technologies, Santa Clara, CA, USA, 1:80), MSH6 (EP51, Agilent Technologies, Santa Clara, CA, USA, 1:200), and PMS2 (EP49, Agilent Technologies, Santa Clara, CA, USA, 1:50). Details of signal development protocols for each ICH are described in the Supplementary Methods.
Scoring was performed by five individuals who had no knowledge of the clinical background of the specimens. Three scientists [Y.K, S.A, S.S.N.] and two credentialed pathologists [R.S. and L.A.L.] judged the stained slides using an optical microscope. For PD-L1 scoring, cases having staining of >10% of all tumor tissues were defined as positive. CD4/CD8 and total lymphocyte count (TLC) were recorded for three randomly selected fields at 40X magnification and the counts were averaged. Samples having a count ≥1 and ≥18 for CD4 and CD8 cells, respectively, was defined positive. Samples with TLC ≥ 26 cells were defined as positive. Cases having complete loss of nuclear staining for at least one of the MMR proteins in tumor cells, but consistently preserved nuclear staining in background, non-tumor cells, were considered as having deficient MMR (dMMR). Staining for each MMR protein in < 30% of all tumor tissues was defined as negative. Samples in which nuclear staining was unaffected for all four MMR proteins tested were judged as having proficient MMR (pMMR) (31) (Supplementary Figure 2).
MSI
The MSI status could be determined for 100 cases, for which formalin fixed paraffin embedded (FFPE) sections were available for extraction of tumor and normal genomic DNA (FALCO biosystems Ltd, Waltham, MA, USA). Total DNA was extracted from FFPE tumor and paired normal tissue samples using a DNA extraction kit (Thermo Fisher Scientific, Waltham, MA, USA). MSI was evaluated with polymerase chain reaction (PCR) using primers for relapse free survival (RFS) or five mononucleotide microsatellite repeats (BAT-26, NR-21, BAT-25, MONO-27 and NR-24) according to a previously described protocol (32). Tumors showing allelic shifts in at least one of five markers were defined as MSI-positive (MSI+) and the remainder were defined as MSS normal, MSI-negative (MSI-) (33, 34) (Supplementary Figure 1).
Statistical analysis
Categorical variable analysis was performed for parameters including PD-L1, CD4, CD8, dMMR protein levels, MSI status, and clinicopathological characteristics to compare with HP status using Fisher’s exact test. OS and RFS were stratified by a combination of the parameters and estimated by the Kaplan-Meier method. The log-rank test was used to compare survival distributions, and the Cox proportional hazards model was used to determine the hazards ratio (HR). For subgroup analysis, all interaction P values were calculated with the likelihood ratio test. Survival time was defined from the date of surgery. All statistical tests were two-tailed and a P value <0.05 was considered statistically significant. Depending on the type of analysis, either JMP or GraphPad Prism was used.
Results
Immunohistochemistry
PD-L1 and CD4/CD8 IHC data were available for all 491 GC tumor samples (Table 1). There were 152 (31.0%) and 339 (69.0%) cases that were PD-L1+ and PD-L1-, respectively. The majority were CD4, (371 (75.6%) CD4- cases vs. 120 (24.4%) CD4+ cases), and for CD8, 252 (51.3%) cases and 239 (48.7%) cases were CD8+ and CD8-, respectively; The distribution of TLC status was similar (239 (48.7%) TLC+ cases vs. 252 (51.3%) TLC- cases). MMR IHC data were available for 428 GC tumor samples and most had pMMR (381 (89.0%) pMMR cases vs. 47 (11.0%) dMMR cases).
MSI
Among 100 GC cases for which FFPE sections were available for extraction of DNA, most were MSI- (88 MSI- cases (88%) vs. 12 MSI+ (12%) cases). Of the 12 MSI+ cases, 4 cases were HP+ and the remainder were HP-. Of the 88 MSI- cases, 46 and 42 cases were HP+ and HP-, respectively. Furthermore, IHC results for MMR analyses were available for 84 of the 100 samples. The concordance rate between IHC for MMR and MSI status was 83.3%. (Supplementary Table 2).
Subgroup Analysis
Survival status including OS and RFS were analyzed in terms of HP infection status (i.e., HP+ or HP-) in conjunction with previously reported parameters (13, 35) and with parameters from the current analysis (i.e., PD-L1, CD4, CD8, CD4/CD8, dMMR, and TLC). For almost all subgroups, HP+ had longer OS and RFS times. An interaction analysis by HP infection status revealed that S-1 (Full/Reduced) and PD-L1 (Positive/Negative) showed a potential interaction with P values of 0.049 and 0.016, respectively (Figure 1).
Survival Status by HP
The survival status of the related subject cohort has been reported (13, 36, 37). Briefly, outcomes were evaluated for 13.3 years (median 5.3 years) for the 491 patients. The median traceable time for survival was 5.2 years. Among 162 censored cases within the observation time, 109 and 53 cases were and were not cancer death, respectively. In the present study, OS and RFS were analyzed in two groups stratified by HP infection status in terms of PD-L1 level and S-1 treatment, which in the subgroup analysis exhibited an interaction with HP infection status (Figure 1).
OS and RFS based on HP infection status were first assessed by stratifying PD-L1 level. The five-year OS rate in the PD-L1+/HP+ group and PD-L1+/HP- was similar (69.9% (95%CI, 56.3-80.7%) vs. 72.9% (95%CI, 63.1-80.8%) (p=0.653; HR 0.879; 95%CI, 0.499-1.548; Figure 3A)), as was the five-year RFS rate (62.5% (95%CI, 48.9-74.4%) vs. 68.7% (95%CI, 58.8-77.2%) (p=0.598; HR 1.149; 95%CI, 0.683-1.935; Figure 3B). For the PD-L1- subgroup, the five-year OS was higher for the PD-L1-/HP+ group than the PD-L1-/HP- group (80.6% (95%CI, 72.5-86.8%) vs. 67.9% (95%CI, 61.3-73.8%) (p=0.0011; HR 0.477; 95%CI, 0.303-0.751; Figure 3C), as was the five-year RFS rate (75.6% (95%CI, 67.0-82.5%) vs. 62.6% (95%CI, 55.9-68.8%) (p=0.0011; HR 0.508; 95%CI, 0.335-0.771; Figure 3D).
Next, the subject patient population was narrowed down to consider those who were eligible to receive post-operative S-1 adjuvant chemotherapy (i.e., exclusion of T3N0 tumors Stage II/III according to TNM classification ver. 7). Among these patients, the five-year OS rate was similar between the PD-L1+/HP+ group (66.7% (95%CI, 51.4-79.2) and PD-L1+/HP- group (67.9% (95%CI, 56.3-77.7; p=0.652; HR 0.871; 95%CI, 0.476-1.593; Figure 4A). The five-year RFS rate was also similar between the two groups (PD-L1+/HP+ 57.5% (95%CI, 42.4-71.4) and PD-L1+/HP- 63.9% (95%CI, 52.1-74.1; p=0.535; HR 1.191; 95%CI, 0.682-2.078; Figure 4B). The five-year OS rate for the PD-L1-/HP+ group was 77.0% (95%CI, 67.5-84.4) and for the PD-L1-/HP- group was 62.7% (95%CI, 54.8-69.9) (p=0.003; HR 0.497; 95%CI, 0.310-0.797; Figure 4C); whereas the five-year RFS rate in PD-L1-/HP+ group was 70.7% (95%CI, 60.8-79.0) and for the PD-L1-/HP- group was 56.7% (95%CI, 48.8-64.1) (p=0.0022; HR 0.517; 95%CI, 0.336-0.797; Figure 4D).
Further analysis was performed to examine whether treatment effect (i.e., Surgery-only; and Surgery + S-1) could have affected the OS and RFS in the PD-L1- patients. The five-year OS rate in PD-L1-/Surgery-only/HP+ group was 71.4% (95%CI, 56.8-82.7%) and PD-L1-/Surgery-only/HP- group was 68.4% (95%CI, 58.0-77.2%) (p=0.367; HR 0.752; 95%CI, 0.404-1.40; Figure 5A); whereas the five-year RFS rate in PD-L1-/Surgery-only/HP+ group was 67.0% (95%CI, 52.2-79.1%) and PD-L1-/Surgery-only/HP- group was 65.1% (95%CI, 54.7-74.3%) (p=0.280; HR 0.721; 95%CI, 0.396-1.313; Figure 5B). The five-year OS rate in the PD-L1-/Surgery + S-1/HP+ group was 82.3% (95%CI, 69.4-90.5%) and for the PD-L1-/Surgery + S-1/HP- group was 55.3% (95%CI, 43.5-66.6%) (p=0.0008; HR 0.304; 95%CI, 0.146-0.635; Figure 5C). The five-year RFS rate in PD-L1-/Surgery + S-1/HP+ group was 74.3% (95%CI, 60.6-84.5%) and PD-L1-/Surgery + S-1/HP- group was 45.7% (95%CI, 34.3-57.5%) (p=0.0006; HR 0.350; 95%CI, 0.187-0.655; Figure 5D).
HP infection status appeared to be a strong discriminatory factor in the PD-L1- /Surgery + S-1 group. As such, the dose-related effect by S-1 was further analyzed for OS and RFS. The five-year OS rate in the PD-L1-/S-1 full dose/HP+ group was 77.3% (95%CI, 55.6-90.2%) and the PD-L1-/S-1 full dose/HP- group was 65.0% (95%CI, 46.8-79.6%) (p=0.164; HR 0.489; 95%CI, 0.176-1.361; Figure 6A). The five-year RSF rate in the PD-L1-/S-1 full dose/HP+ group was 63.6% (95%CI, 42.3-80.7%) and PD-L1-/S-1 in the full dose/HP- group was 58.6% (95%CI, 40.5-74.7%) (p=0.4171; HR 0.702; 95%CI, 0.297-1.660; Figure 6B). The five-year OS rate in PD-L1-/S-1 reduced/HP+ group was 86.2% (95%CI, 68.5-94.7%) and PD-L1-/S-1 reduced/HP- group was 46.4% (95%CI, 31.2-62.3%) (p=0.0014; HR 0.205; 95%CI, 0.07-0.602; Figure 6C); whereas the five-year RFS rate in PD-L1-/S-1 reduced/HP+ group was 82.8% (95%CI, 64.7-92.6) and PD-L1-/S-1 reduced/HP- group was 34.2% (95%CI, 21.0-50.4%) (p=0.001; HR 0.190; 95%CI, 0.072-0.498; Figure 6D).
Multivariate Analysis
To examine which parameters may be independent factors for OS or RSF, a Cox proportional hazards regression model to examine interaction among the following parameters: age, sex, pathological stage (pStage) (TNM7), S-1 Dose, histologic type, CD4, CD8, TLC, MSI, MMR, and HP*PD-L1 was carried out. There were 54 events for RSF and OS (Figure 3). To avoid overfitting, five parameters were thus selected for each analysis. Overall, the statistical model identified as independent factors pStage (TNM7), and HP*PD-L1 for advanced GC patients with >2-fold hazard ratio in the present cohort (Table2).
Discussion
The improved survival seen for advanced GC patients who are HP+ compared to those who are HP- suggests that systemic effects that are chronically induced by HP infection may play a significant role in GC treatment efficacy (13, 37). Importantly, many countries have reported similar trends, suggesting that the HP infection could have more dominant effects than a variety of treatments that are delivered in different countries and regions (10, 12, 38). For instance, the standard therapy for advanced GC varies between East Asia, Europe and North America due to heterogeneities in tumors, race, medical environment and surgical procedures (6, 39-41). However, whether HP infection actually induces a systematic effect that is directly related to the OS is unclear. Moreover, HP infection is thought to be related to gastric duodenal ulcer, idiopathic thrombocytopenic purpura, MALT lymphoma and hyperplastic polyps although most of these diseases are unlikely to directly cause death (42-45). These diseases occur only in a small percentage of people infected with HP, suggesting that HP infections can frequently persist in the host without causing disease (46). Clarification of the systematic mechanisms induced by HP infection is important since HP may have effects other than tumor induction in the majority of infected people, even though a small percentage of healthy individuals do develop GC simply due to HP infection (47).
Standardized radical gastrectomy for Stage II/III (TNM7) GC followed by S-1 adjuvant chemotherapy is a well-established treatment protocol in Japan (7). However our consortium included “surgery only” patients who underwent surgery before 2006. Therefore, although retrospective, our present data allow an important comparison to assess the effect of HP in patients who received S-1 adjuvant chemotherapy as the community standard in Japan. By profiling immune-response molecules in advanced GC patients, we demonstrated that prognosis of HP+ was significantly better than HP- patients in the PD-L1- population. This result was further pronounced by stratification of the patients in order with R0 operation, Stage II/III, S-1 adjuvant chemotherapy, and dose reduced S-1 adjuvant chemotherapy. In terms of OS and RFS, the status of other immune-related molecules, including CD4, CD8, TLC, and MMR proteins as well as the status of MSI did not show significant association with PD-L1 levels or HP infection status. In the background microenvironment, HP infection induces PD-L1 expression and T-cell activation (48). Upregulation of PD-L1 on gastric epithelial cells may serve as one negative feedback mechanism for tissue protection during T-cell inflammatory events, whereby PD-L1 delivers an inhibitory signal to T-cells (27, 49). It has been reported that the level of PD-L1 expression on gastric epithelial cells is intrinsically modulated by chronic HP infection (49, 50). An upregulation of PD-L1 expression in gastric epithelium increases the chance to trigger malignant transformation (27). In advanced GC, S-1 administration of a standard dose and regimen is associated with a favorable prognosis (51, 52), whereas our present data suggest that HP+ patients who do not have sustained PD-L1 levels on cancer cells have a potential survival benefit despite reduced adjuvant chemotherapy. In general, HP infection occurs in early age, whereas GC develops 40-50 years after exposure of HP infection (53, 54). Further investigation is warranted to clarify the way in which the degree of HP infection and subsequent host response can be individually differentiated.
All 491 cases were analyzed by the Cox proportional hazards model, which identified pStage (TMN7) and HP*PD-L1 as independent prognostic factors with a >2-fold hazards ratio. While pStage (TMN7) and age are generally associated with prognosis (7, 55), HP status and systemic immune conditions have not been investigated. Based on our current data, the identification of HP*PD-L1 as one of the prognostic factors suggests that HP infection status can provide treatment stratification for locally advanced GC patients who undergo adjuvant chemotherapy after gastrectomy.
The present study has the following limitations: (a) it was a retrospective study; (b) S-1 treatment indication was not randomized; (c) information was occasionally missing due to sample preserved conditions; (d) histological heterogeneity was not taken into account due to limited sample area; and (e) HP infection status was only assessed by histological examination. Nonetheless, our rigorous approach still suggests that PD-L1 and HP infection status have a substantial impact on the stratification of advanced GC patients.
Conclusion
For advanced GC patients who undergo postoperative chemotherapy, HP-dependent modulation of PD-L1-influenced host immune system function could contribute to prolonged survival time in the absence of functioning immune escape mechanisms.
Data Availability
None applicable
Funding
This study was supported by the Grant-in-Aid Scientific Research KAKENHI [19K09130] and [16H01578].
Notes
Satoshi S. Nishizuka received research support from Taiho Pharmaceuticals. The remaining authors declare no conflict of interest.
Consortium Members (affiliations are as of case registration)
Northern Gastric Cancer Study Consortium members: Gen Tamura, MD, PhD, Department of Laboratory Medicine, Yamagata Prefectural Central Hospital, Yamagata, Japan. Hisataka Fujiwara, MD, Takehiro Chiba, MD, Fumitaka Endo, MD, PhD, Kazushige Ishida, MD, PhD, Hirokatsu Katagiri, MD, PhD, Kei A. Sato, MD, PhD, Department of Surgery, Iwate Medical University School of Medicine, Yahaba, Japan. Chie Ito, MD, PhD, Molecular Therapeutics Laboratory, Iwate Medical University School of Medicine, Yahaba, Japan and Department of Surgery, Iwate Medical University School of Medicine, Yahaba, Japan. Kohei Kume, PhD, Molecular Therapeutics Laboratory, Iwate Medical University School of Medicine, Yahaba, Japan, Department of Surgery, Iwate Medical University School of Medicine, Yahaba, Japan, and Division of Biomedical Research & Development, Institute of Biomedical Sciences, Iwate Medical University, Yahaba, Japan. Noriyuki Uesugi, MD, and Tamotsu Sugai, MD, PhD, Department of Diagnostic Pathology, Iwate Medical University School of Medicine, Yahaba, Japan. Jun Nakajima, MD, and Akinori Takagane, MD, PhD, Department of Surgery, Hakodate Goryoukaku Hospital, Hakodate, Japan. Tatsuru Ikeda, MD, Department of Pathology, Hakodate Goryoukaku Hospital, Hakodate, Japan. Kenichi Oyama, MD, Department of Surgery, Noshiro Kousei Medical Center, Noshiro, Japan. Masayo Murakami, MD, Department of Pathology, Noshiro Kousei Medical Center, Noshiro, Japan. Chihiro Tono, MD, Department of Surgery, Iwate Prefectural Kamaishi Hospital, Kamaishi, Japan. Takashi Satoh, MD, PhD, Department of Pathology, Iwate Medical University School of Medicine, Yahaba, Japan. Shingo Mitomo, MD, PhD, and Yoshiyuki Tamasawa, MD, Department of Surgery, Japanese Red Cross Hachinohe Hospital, Hachinohe, Japan. Shun-ichi Sasou, MD, Department of Pathology, Japanese Red Cross, Hachinohe Hospital, Hachinohe, Japan. Yoshihiko Sugimura, MD, PhD, Department of Surgery, Japanese Red Cross, Morioka Hospital, Morioka, Japan. Nobuhiro Monma, MD, Department of Pathology, Japanese Red Cross, Morioka Hospital, Morioka, Japan. Masanori Takahashi, MD, and Osamu Shimooki, MD, PhD, Department of Surgery, Iwate Prefectural Kuji Hospital, Kuji, Japan. Keisa Takeda, MD, and Kazuhito
Misawa, MD, PhD, Department of Surgery, Sapporo City General Hospital, Sapporo, Japan. Yuichiro Fukasawa, MD, Department of Pathology, Sapporo City General Hospital, Sapporo, Japan. Toru Yoshida, MD, PhD, Department of Surgery, Iwate Prefectural Senmaya Hospital, Senmaya, Japan. Toru Ishikawa, MD, and Nobuo Sakashita, MD, Department of Surgery, Iwate Prefectural Miyako Hospital, Miyako, Japan. Tsutomu Sakuma, MD, Department of Pathology, Iwate Prefectural Central Hospital, Morioka, Japan. Masaya Oikawa, MD, and Takashi Tsuchiya, MD, Department of Surgery, Sendai
Open Hospital, Sendai, Japan. Takashi Sawai, MD, PhD, Department of Pathology, Sendai Open Hospital, Sendai, Japan. Ryujin Endo, MD, PhD, Division of Hepatology, Department of Internal Medicine, Iwate Medical University School of Medicine, Morioka, Japan.
Acknowledgement
The authors are grateful to all the patients, family, and medical staff who participated in this study.