Post-hoc biomarker analyses of T4a/T4b gastric cancer from patients recruited into SAMIT, a phase III randomized controlled trial

Takashi Oshima Department of Gastrointestinal Surgery, Kanagawa Cancer Center, Yokohama, 241-8515 Akira Tsuburaya (  tuburayaa@gmail.com ) Department of Surgery, Ozawa Hospital, Odawara, 250-0012 Kazuhiro Yoshida Department of Surgical Oncology, Gifu University Graduate School of Medicine, Gifu, 501-1194 Takaki Yoshikawa Department of Gastric Surgery, National Cancer Center Hospital, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045 Yohei Miyagi Kanagawa Cancer Center Research Institute, Yokohama, 241-8515 Yasushi Rino Department of Surgery, Yokohama City University, Yokohama, 236-0004 Munetaka Masuda Department of Surgery, Yokohama City University, Yokohama, 236-0004 Jia Guan Department of Clinical Biostatistics, Graduate School of Medicine Kyoto University, Kyoto, 606-8501 Patrick Tan Cancer Science Institute of Singapore, National University of Singapore, Singapore, 117599 Heike Grabsch Department of Pathology, GROW School for Oncology and Development Biology, Maastricht University Medical Center+, Maastricht Junichi Sakamoto Tokai Central Hospital, Kakamigahara, 504-8601 Shiro Tanaka Department of Clinical Biostatistics, Graduate School of Medicine Kyoto University, Kyoto, 606-8501


Introduction
In Japan, 134,650 patients were diagnosed with gastric cancer (GC) in 2019, of which 25,850 had stage II/III disease according to UICC TNM 8 th edition 1,2 . The standard treatment for patients with stage II/III GC in Japan is curative D2 gastrectomy followed by uorinated-pyrimidine-based chemotherapy 3 based on the results from the Japanese ACTS-GC and Korean CLASSIC randomized phase III trials 4,5,6,7 . However, despite improved overall survival (OS) with adjuvant chemotherapy, the ve-year OS rate of patients with stage III GC remains unsatisfactory. While RNA expression studies from the ACTS-GC trial identi ed several novel GC biomarkers, none of them were signi cantly associated with the S-1 treatment effect 8, 9,10,11 . In a post-hoc analysis of resection specimens from the CLASSIC trial, the RNA expression levels of four genes (GZMB, WARS, SFRP4, and CDX1) were able to stratify patients by risk of recurrence and to predict the bene t of adjuvant chemotherapy 12 .
In addition to uorinated-pyrimidine and platinum-based anticancer drugs, uorinated-pyrimidine combined with taxanes such as paclitaxel or docetaxel has been considered for GC treatment 13 . Taxanebased anticancer drugs have a lower incidence of nephrotoxicity or neuropathy than platinum-based compounds, such as cisplatin or oxaliplatin, and can be administered safely in an outpatient setting. Recently, a randomized phase III study in patients with curatively resected pathological stage (pStage) III GC (JACCRO GC-07 trial) reported signi cantly longer three-year recurrence-free survival in patients treated with adjuvant S-1 plus docetaxel than with adjuvant S-1 alone 14 . Based on these results, chemotherapy with S-1 plus docetaxel after D2 gastrectomy was recommended as the new standard of care for patients with pStage III GC in Japan.
The Stomach Cancer Adjuvant Multi-Institutional Group Trial (SAMIT) was a randomized phase III trial that investigated whether (1) sequential treatment (paclitaxel treatment followed by tegafur, uracil (UFT), or S-1) was superior to monotherapy (UFT or S-1) as adjuvant chemotherapy, (2) S-1 was non-inferior compared to UFT as an adjuvant chemotherapeutic agent in patients with T4a/T4b GC, and (3) sequential treatment with paclitaxel followed by uorinated-pyrimidine was superior to uorinatedpyrimidine monotherapy as adjuvant chemotherapy for T4a/T4b GC. The results showed that S-1 was non-inferior compared to UFT and that sequential treatment improved disease-free survival (DFS) only in patients with stage IIIB GC 15 .
The JACCRO GC-07 trial and SAMIT both demonstrated improved outcomes in patients with stage III GC treated with adjuvant chemotherapy using uorinated-pyrimidine and taxane-based anticancer drugs. However, the recurrence rate within 2 years after surgery was 75.3% in the JACCRO GC-07 study (pStage III GC) and 40.0% in the SAMIT (pT4a/pT4b GC). Therefore, we hypothesized that adjuvant chemotherapy using uorinated-pyrimidine plus taxane-based anticancer drugs may only be effective in a subset of patients. If these patients can be identi ed using biomarker assessment performed in the gastrectomy specimens, adjuvant treatment regimens can then be personalized to improve patient outcomes.
In this study, we performed a post-hoc analysis of the tissue samples collected from patients recruited in the SAMIT and analyzed a comprehensive panel of RNA expression-based biomarkers to identify genes that might be suitable for selecting patients who are likely to bene t more from sequential paclitaxel and uorinated-pyrimidine adjuvant chemotherapy than from adjuvant uorinated-pyrimidine monotherapy.

Results
Patients and sample collection FFPE samples were obtained from 556 patients who participated in the SAMIT. Twenty-nine patients were subsequently excluded because of insu cient RNA. Therefore, biomarker analysis was eventually performed in 527 patients (94.7%; Fig. 1). The characteristics of the patients included in the current study were representative of those of the entire SAMIT population (Table 1). Except for sex (more males in the sequential paclitaxel treatment group (p=0.04)), the clinical and pathological characteristics were well balanced between the sequential paclitaxel treatment and monotherapy subgroups (Supplementary Table  S1, Online Resource 1). The median follow-up time from randomization was 56.8 (interquartile range (IQR)=45.3-69.8 months) and 59.1 months (IQR=46.2-72.8 months) for patients in the monotherapy and sequential paclitaxel arms, respectively.

Predictive biomarkers
We conducted multivariable Cox regression analysis to assess the potential relationship between gene expression level and OS, DFS, or cumulative incidence of relapse after sequential paclitaxel therapy; the genes were ranked based on the interaction-related p-values (Table 2: top 10 genes and Supplementary  Table S2, Online Resource 1: all genes). VSNL1 and CD44v were the only genes with mRNA expression levels that were statistically signi cant as predictive biomarkers of sequential paclitaxel treatment for all three endpoints (Table 3).
Patients with low expression of VSNL1, CD44v, or both had signi cantly longer OS and DFS after sequential paclitaxel treatment than after monotherapy (Fig. 2a, b). A total of 191 (36.2%) patients showed combined low expression of both genes, which was related to the most signi cant bene t from sequential paclitaxel treatment compared to monotherapy (Table 4). No such effect was observed in the cumulative incidence of relapse (Fig. 2c).
The OS improvement in patients with low VSNL1 and CD44v expression treated with sequential paclitaxel remained signi cant in multivariable analysis after adjustment for clinical and pathological factors in the treatment group (Table 4).
Patient strati cation based on the pTNM stage showed that OS improvement in response to sequential paclitaxel treatment in patients with low VSNL1 and/or CD44v expression was greatest in patients with stage IIIB GC (Table 4, Fig. 3).

Internal validation
The overall performance of the different statistical models, including the interactions between VSNL1 expression and the treatment group, as well as the clinical and pathological factors for OS prediction with C statistics using the bootstrap 0.632+ estimator (0.7111) and apparent estimator (0.7266), was evaluated. The accuracy of OS prediction based on CD44v and VSNL1 expression levels was comparable when the apparent estimator was used (0.7252), whereas it was not su ciently accurate when the bootstrap 0.632+ estimator was used (0.7083) (Supplementary Table S3, Online Resource 1).

Discussion
This study is the third large-scale biomarker study in patients with gastric cancer treated with adjuvant chemotherapy in a randomized trial setting. Although previous studies using clinical samples from the ACTS-GC trial revealed several novel molecular GC biomarkers, signi cant interactions between S-1 treatment and the RNA expression level of a set of 63 genes could not be identi ed 8, 9,10,11 . In a study of clinical samples from the CLASSIC trial, the RNA expression levels of four genes were able to stratify GC patients based on high, medium, or low risk of recurrence, or predict bene t from adjuvant chemotherapy. An approximately 7% difference was observed in the 5-year survival rate when patients receiving adjuvant chemotherapy were retrospectively strati ed based on the four-gene classi er 12 .
Although several candidate biomarkers of resistance or sensitivity to paclitaxel have previously been suggested 16,17,18,19,20,21,22,23,24, none have been validated in a second independent series. Hence, there remains a clinical need to validate the proposed biomarkers and/or identify new biomarkers that can be used in routine clinical practice to identify patients likely to bene t from paclitaxel therapy 25 . Moreover, multiple studies have reported an association between the expression of several genes or proteins and bene ts from paclitaxel in different tumor types 26,27,28,29,30 . For example, CCND1 overexpression promotes paclitaxel-induced apoptosis in breast cancer 27 . Members of the BCL-2 and P-glycoprotein families, such as ABCB1, have been reported to be involved in paclitaxel resistance in esophageal cancer 28 . SPARC expression in tumor stromal cells has been suggested as a potential negative predictor of paclitaxel treatment in patients with lung cancer 29,30 . However, the expression levels of these genes were not signi cantly associated with patient outcomes in the current study. This may be related to the cancer type, sample size, case mix, ethnic differences, or methodological differences.
In the current study, we identi ed low expression levels of VSNL1 and/or CD44v as potential novel predictive biomarkers of bene t from paclitaxel chemotherapy after curative D2 gastrectomy. VSNL1 encodes visinin-like protein 1 (VILIP-1) 31 . It has been suggested that the absence or reduced expression of VILIP-1 results in increased cancer cell motility, suggesting a potential tumor suppressor function of this protein 32 . Additionally, VILIP-1 was shown to prevent epithelial-mesenchymal transition of cancer cells by regulating the expression of the transcription factor SNAIL1 in a cAMP-dependent manner 33 . The mechanisms underlying the relationship between VSNL1 expression and the bene ts of paclitaxel chemotherapy remain to be elucidated. CD44v is a cell-surface molecule that senses, integrates, and relays cellular microenvironmental signals to membrane-associated cytoskeletal proteins or to the nucleus and regulates the expression of numerous genes encoding cell behavior-related proteins 34,35,36 . CD44v has been identi ed as a prognostic and cancer stem cell marker in several different types of cancer, including GC, and has been reported to regulate cancer stemness-related properties, including selfrenewal, tumor initiation, aggressiveness, relapse, chemoradiotherapy resistance, and metastasis 37, 38 . However, the biological mechanisms underlying the survival bene t of sequential paclitaxel and uorinated-pyrimidine adjuvant chemotherapy in patients with GC with low CD44v and/or VSNL1 expression are yet to be clari ed.
To our knowledge, this is the rst and most comprehensive study to explore and identify potential RNA expression-based biomarkers for the prediction of survival bene t from sequential paclitaxel and uorinated-pyrimidine adjuvant chemotherapy in patients with GC. However, this study has certain limitations. Although we demonstrated that the study cohort was representative of the entire SAMIT patient cohort with respect to clinicopathological characteristics, including survival, we were only able to retrieve material from approximately one-third of the original SAMIT population. Furthermore, we only analyzed RNA samples from a single tissue block. Therefore, intratumoral heterogeneity at the gene expression level was not assessed.
In conclusion, our study, to the best of our knowledge, is the rst to identify biomarkers for selecting patients with locally advanced GC who most likely bene t from adjuvant chemotherapy with sequential paclitaxel and uorinated-pyrimidine treatment after curative D2 gastrectomy. The validation of our ndings in a second independent series followed by a prospective trial is necessary before we can consider using these biomarkers in routine clinical practice. Nevertheless, personalized adjuvant chemotherapy using these biomarkers may improve treatment outcomes in patients with locally advanced GC. Our results may further help patient selection in clinical trials for biomarker-oriented adjuvant chemotherapy.

Patients and sample collection
This study was approved by the Institutional Review Board (IRB) of Kanagawa Cancer Center, the central institute for this study (approval number: 26 -42), as well as the IRBs of all institutions that participated in this study. Representative blocks from formalin-xed, para n-embedded (FFPE) gastrectomy specimens were collected retrospectively from participating institutions according to the following inclusion criteria: (1) patients were participants in the SAMIT, (2) FFPE blocks or unstained cut sections were available, and (3) the translational study protocol was approved by the IRB of the respective institution. Samples were collected by the data center of Kanagawa Cancer Center and shipped to Yokohama City University for RNA extraction and analysis. Sections (each 10-μm thick) were cut from the FFPE blocks and stored at 4°C until microdissection.

RNA extraction and complementary DNA (cDNA) synthesis
Hematoxylin and eosin-stained slides were reviewed, and the area with the highest tumor content was outlined manually. After manual microdissection, total RNA was isolated using NucleoSpin® FFPE RNA XS (MACHEREY-NAGEL GmbH & Co. KG, Düren, Germany). For RNA quality control, the OD 260 /OD 280 ratio was measured using a NanoDrop 2000 (Thermo Fisher Scienti c Inc., MA, USA; RRID:SCR_018042). The total RNA integrity number was measured using an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Waldbronn, Germany, RRID:SCR_018043). To con rm that the total RNA samples were not contaminated with DNA, RNA18S1 expression was evaluated by quantitative real-time PCR (qRT-PCR) in each sample before cDNA preparation. cDNA was prepared from samples that passed all the quality control checks. cDNA was synthesized from 0.4 µg of total RNA using an iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Inc., CA, USA), diluted to 0.2 µg/µL with distilled water, and stored at −20°C until use.
qRT-PCR qRT-PCR was performed using the QuantiFast TM Probe Assay (QIAGEN, Venlo, Netherlands) and

Gene selection
The 105 selected genes included 63 genes analyzed in an exploratory biomarker study of ACTS-GC trial participants 10 . Of these, 57 genes have been previously reported as biomarkers of paclitaxel resistance or sensitivity. The functional annotation of each gene using DAVID 6.7, is outlined in Supplementary Table S4 (Online Resource 1).

De ning the predictive value of the biomarkers
The mRNA expression level of each gene was classi ed as low versus high using the median mRNA expression level as a cut-off point, as described previously 40 . If the mRNA expression level of a particular gene was below 1.0×10 -8 ng/μL, the expression level was set to '0.00'. The value of a biomarker in predicting the bene t of sequential paclitaxel based on the OS, DFS, and cumulative incidence of relapse was determined by examining the p-values of the interaction between the dichotomized gene expression level and the treatment group (sequential paclitaxel versus monotherapy) after adjusting for clinical and pathological factors using Cox regression or Fine-Gray models 41,42 . The genes were ranked according to treatment interaction-related p-values. Values were considered signi cant at p<0.05. Additionally, we combined the expression levels of selected genes to identify sensitive and non-sensitive patient subsets.

Internal validation
We adopted an internal validation strategy, as proposed by Wahl et al. 43 , to address the potential overestimation of the standard error owing to multiple imputations and optimism in the predictive performance. We used Harrell's C statistics to analyze the predictive performance of the survival data and addressed the optimistic bias by Harrell's C statistics using the bootstrap 0.632+ method with 20 bootstrap samples from the original dataset with replacement, followed by multiple imputations.

Statistical analysis
The pre-de ned statistical analysis plan for this study has been reported previously 40 . The primary and secondary endpoints were the OS and DFS, respectively. The OS and DFS curves were constructed using the Kaplan-Meier method, and the cumulative incidence curves of relapse were constructed using the Aalen-Johansen method 44 to compare sequential paclitaxel and monotherapy, considering the expression levels of the selected genes either individually or in combination. The adjusted hazard ratios (HRs), 95% con dence intervals (CIs), and p-values of the major treatment effects and interactions were estimated for the entire patient population and subgroups according to the UICC TNM 8 th ed stage 2 . We used multiple imputations to handle missing clinical and pathological factor data and generated 20 multiply imputed datasets for parameter estimates. The reported p-values were two-tailed, and the major effects and interactions were considered statistically signi cant at p<0.05. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC, USA).

Declarations Ethical statement
All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964 and later versions. Informed consent or substitute for it was obtained from all patients for inclusion in the study.

Acknowledgments
We are grateful to Kazuaki Tanabe Table 3. Effects of sequential paclitaxel followed by UFT or S-1 on overall survival, diseasefree survival, and cumulative incidence of relapse, based on gene expression levels    Figure 1 Flowchart of SAMIT patients available for primary analysis and subsequent biomarker analysis. Formalin-xed, para n-embedded (FFPE) samples were available from 556 SAMIT patients. Twenty-nine patients had to be excluded owing to insu cient RNA.

Figure 2
Kaplan-Meier curves by gene expression level in the sequential paclitaxel and monotherapy arms. Patients with low RNA expression levels of VSNL1, CD44v, or both had signi cantly longer overall survival (a), longer disease-free survival (b), and lower cumulative incidence of relapse (c) after sequential paclitaxel treatment than after monotherapy.

Figure 3
Forest plot of the study results. After patient strati cation based on the pTNM stage, the survival bene t from sequential paclitaxel treatment was greater among patients with stage IIIB gastric cancer with low expression of either gene or combined low gene expression. The association between low expression levels of VSNL1 and CD44v and potential bene ts from sequential paclitaxel treatment were signi cant for disease-free survival and cumulative incidence of relapse.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.