Genetic alterations in gastric cancer patients according to sex

To date, few reports have investigated the genetic alterations and clinicopathological features in gastric cancer (GC) according to sex. In total, 2673 GC patients receiving curative surgery were enrolled. Among the 2673 GC patients, 1979 (74.0%) patients were male. After propensity-score matching, 846 patients were enrolled for the analysis, including 423 males and 423 females. There was no significant difference in the clinicopathological features between the sexes. Regarding the initial recurrence pattern, the males were more likely to develop tumor recurrence and liver metastasis than the females, especially in stage III GC. Regarding the molecular analysis, the males had higher PD-L1 expression than the females, especially in stage III GC. In addition, the patients aged ≥ 65 years had higher PD-L1 expression than the patients younger than 65 years. The multivariate analysis demonstrated that sex was among the independent prognostic factors affecting overall survival (OS) and disease-free survival (DFS). Among the patients with liver metastases, PD-L1 expression was more common among the aged male patients. The males were associated with more tumor recurrence and higher PD-L1 expression than the females, especially in stage III GC. For GC patients with liver metastases, PD-L1 testing is recommended, especially among aged male patients.


INTRODUCTION
Gastric cancer (GC) is the sixth most common cancer and the second most common cause of cancer-related deaths worldwide [1]. Surgical resection is the only curative treatment for GC. The most frequent age of GC onset is between 50 and 70 years. The factors associated with gastric carcinogenesis include Helicobacter pylori infection, Epstein-Barr virus infection, genetic mutations, and microsatellite instability (MSI). [2,3].
A male predominance in upper gastrointestinal tract cancers, including esophageal cancer and GC, is observed in almost all populations, and the male gender is a well-established risk factor [4,5]. Some possible reasons contributing to the different incidences between males and females include the protective effects of hormones in females, differences in body iron storage, and different dietary habits and lifestyles [6]. However, strong evidence explaining this difference in incidence is lacking.
It has been reported that in young GC patients, MSI is more common in females than males [7]. However, there is no significant difference in genetic mutations in hMLH3, BRAF, and KRAS between the sexes in young GC patients [7]. Programmed death-ligand 1 (PD-L1) expression is reportedly more common in males GC As shown in Table 1, after propensity-score matching at a 1: 1 ratio, 846 patients were enrolled for the subsequent analysis, including 423 males and 423 females. There is no significant difference in the clinicopathological features between the males and females (Table 1), and no difference was observed between the sexes in each TNM stage (Table 2).

Initial recurrence patterns
The mean follow-up time was 91.9 months, which was similar between males and females (91.5 vs. 92.4 months, P=0.875). The longest follow-up time was 344.6 months. Among the 846 patients, 207 patients (24.5%) had tumor recurrence during follow-up; 445 patients died during follow-up, and 49 patients (5.8%) had follow-up time shorter than 3 years; among them, there were 22 stage I diseases, 13 stage II diseases, and 14 stage III diseases. Among the 49 patients, four patients had tumor recurrence, including three patients with stage III diseases and one patient with stage II disease.
As shown in Table 3, the males were more likely to develop tumor recurrence, distant metastases, and liver metastases than the females. In stage I GC, the males were more likely to develop tumor recurrence than the females. In stage II GC, there was no difference in the initial recurrence patterns between the sexes. In stage III GC, the males were more likely to develop tumor recurrence and liver metastases than the females.

Survival analysis
As shown in Figure 1, the males had significantly lower 5-year overall survival (OS) (60.9% vs. 64.0%, P=0.030, Figure 1A) and disease-free survival (DFS) (56.9% vs. 62.0%, P=0.015, Figure 1B) rates than the females. As shown in Figure 2, in stage I GC, the males had significantly lower 5-year OS (85.9% vs. 90.5%, P=0.047, Figure 2A) and DFS (83.6% vs. 88.8%, P=0.032, Figure  2B) rates than the females. In stage II GC, there was no significant difference in the OS (69.9% vs. 71.2%, P=0.526, Figure 2C) and DFS (65.0% vs. 66.7%, P=0.557, Figure 2D) rates between the sexes. In stage III GC, there was no significant difference in the OS (25.1% vs. 25.6%, P=0.436, Figure 2E) and DFS (19.8% vs. 24.2%, P=0.268, Figure 2F) rates between the sexes. Tables 4 and 5, the univariate analysis demonstrated that age, gender, tumor location, gross appearance, adjuvant chemotherapy, lymphovascular invasion, and pathological T and N categories were associated with OS and DFS. The aforementioned eight variables were included in a multivariate Cox proportional hazards model to adjust for the effects of covariates. The multivariate analysis demonstrated that age, gender, gross appearance, and pathological T and N categories were independent prognostic factors affecting OS and DFS.

Analysis of genetic alterations
Genetic alterations were analyzed in 438 GC patients with available GC tissues. As shown in Table 6, the males had higher PD-L1 expression than the females (31.1% vs. 20.2%, P=0.020). Patients aged 65 years or older had higher PD-L1 expression than those younger than 65 years (33.1% vs. 20.8%, P=0.005).
In stage I GC, the females were more likely to have MSI-H tumors than the males (17.9% vs. 4.1%, P=0.033). In stage III GC, the males were more likely to exhibit PD-L1 expression than the females (31.2% vs. 16.4%, P=0.033). Among the 225 stage III GC, the patients aged 65 years or older were more likely to exhibit PD-L1 expression than the patients younger than 65 years (29% vs. 14.9%, P=0.013).

DISCUSSION
To minimize selection bias, we performed propensityscore matching and analyzed the clinicopathological feature differences between the sexes in our GC patients. Our results demonstrate that the males are associated with more tumor recurrence, more liver metastasis, and worse prognoses than the females with AGING GC. In addition, the molecular analysis demonstrated that PD-L1 expression was more common among the aged male patients, especially those with stage III GC. Among the patients with liver metastases, PD-L1 expression was more common among the aged male patients.
In the present study, sex was one of the independent prognostic factors of OS and DFS. Although gross appearance and pathological T and N categories were significantly related to each other, all of the five independent prognostic factors were still significantly associated with OS and DFS by multivariate analysis. It seems that all the five factors play an important role in GC patient prognosis.
A meta-analysis [11] showed that MSI-H GC was associated with females, an older age, and intestinaltype GC. In the present study, a significantly higher tumor recurrence rate and a significantly higher frequency of MSI-H tumors was observed in the females than the males (17.9% vs. 4.1%, P=0.003) in stage I GC only. MSI-H tumors were associated with a better prognosis than MSI-L/S tumors in GC [12]. The    PD-L1 expression has been reported to be higher in metastatic liver tumors than primary gastric tumor cells and tumor infiltrating lymphocytes [13]. Consequently, immunotherapy may be beneficial for GC patients with liver metastases. According to our results, we recommend evaluating PD-L1 expression in GC patients who develop liver metastases, especially aged male patients, which may provide useful information for evaluating immunotherapy in these patients.

AGING
The PD-L1 CPS score was developed using the 22C3 assay as a companion diagnostic for immunotherapy in GC, which was approved by the U.S. Food and Drug Administration [14]. Most trials regarding the use of immunotherapy in GC are designed for metastatic GC patients. A randomized, multicenter, double-blind, placebo-controlled phase 3 study, ATTRACTION-05 (ONO-4538-38/BMS CA209844), investigating the role of Nivolumab in combination with adjuvant AGING Whole-genome sequencing demonstrated that males had more somatic structural variants than females in GC, but the underlying reason and clinical significance for this result is unclear [15]. Chromosome conformation capture (Hi-C) methods have identified subchromosomal structures of higher-order chromatin interactions called topologically associated domains (TADs) that are separated from each other by boundary regions [16,17]. By subdividing the genome into discrete regulatory units, TADs restrict the contacts that enhancers establish with their target genes [18][19][20]. One of the major TADs related to gender is sex-determining region Y-related high-mobility group box 9 (SOX9) [17]. It was reported that SOX9 regulated CEACAM1 (carcinoembryonic antigen cell adhesion molecule 1) expression and immune resistance in melanoma [21]. Furthermore, PD-L1 expression appears to be directly or indirectly regulated by several X-linked miRNAs [22]. According to TargetScan 7.1 (http://www.targetscan.org), three miRNAs localized on chromosome X, including miR-106b, miR-20b, and miR-513, are repressors of PD-L1 by direct binding to their 3ʹUTRs. PD-L1 transcription is induced by hypoxia-inducible factor-1α (HIFα) and signal transducer and activation of transcription-3 (STAT3) factors, directly acting on its promotor [23]. miR-20b can target both HIF-1a and STAT3 transcription factors [24], thus repressing PD-L1 expression. Therefore, there might be a gene-to-gene enhancing coalition effect among sex and PD-L1 expression. Further investigation to understand the mechanisms underlying the gender disparity and aging in the treatment response for GC is still needed.
In the study of Sasao et al. [7], for patients aged 40 or younger, males were more likely to have tumor located in the lower-third of the stomach than females (30% vs. 5%). In the present study, as shown in Table 1, females were more likely to have tumors in the middle-third stomach than males before propensity-score matching.
For the subgroup analysis for younger and older patients, because the number of patients aged 40 or younger is too small for analysis, we choose an age of 65 years or older to categorize aged patients according to the World Health Organization. In the present study, for patients aged younger than 65 years, females were more likely to have tumor located in the middle-third of the stomach than males (46.5% vs. 37.6%), while tumors in the lower-third of the stomach were more common in males than females (44.3% vs. 37.6%). For patients aged 65 or older, males were more likely to have tumor located in the upper-third of the stomach (20.6% vs. 10.8%). Based on our results and others, it seems that the tumor location of GC was associated with both sex and age.
Regarding the physiological gastrointestinal motility, for liquid diet, there was no significant difference between sexes; however, for solid diet, the gastric emptying diet is significantly longer in females than males [25]. Pre-menopausal females have significantly longer gastric emptying time than males for solid diet [26]. For post-menopausal females, the gastric emptying time decreases and becomes similar to that in males [27]. The gastric emptying time increased with age in the women, while the gastric emptying time remained almost unchanged with age in the males [28]. In addition to gender, age, and the type of meal, patient statue and the size of stomach and diabetic mellitus may also affect gastric motility. The possible impact on the gastric emptying time should be evaluated in separate male and female subpopulations.
In our hospital, there has been a case manager in charge of contacting gastric cancer patients for more than 30 years. Consequently, most patients receive regular follow-up at our hospital. Our previous study demonstrated that approximately 90% of gastric cancer recurrences occurred within 3 years after surgery [29].
In the present study, only 5.8% of our patients had follow-up time shorter than 3 years, we believe that the small percentage of patients have little impact on the tumor recurrence rate in the present study.
Because this study is a retrospective study with single center data, this study has limitations. Consequently, we performed propensity-score matching to minimize selection bias. GC, a multifactorial disease, is easily associated with dietary culture and economical levels. Diet with high fat/salt/nitrogen, a history of infection with Helicobacter pylori or Epstein-Barr virus, tobacco use, immune system, and genetic mutations caused by chronic inflammation are risk factors for cancer development [30][31][32][33][34]. In the present study, we could not analyze the influence of dietary and economical levels in our GC patients because the data is not available in most patients. Although our results demonstrated higher PD-L1 expression in males than females, especially in stage III GC, none of our enrolled patients received immunotherapy for tumor recurrence. We could not evaluate the efficacy of immunotherapy in this group of patients. Further randomized controlled trials investigating the use of immunotherapy for adjuvant therapy are required to validate our results and may provide convincing evidence regarding GC treatment in the future.
In conclusion, males were associated with more tumor recurrence and higher PD-L1 expression than females, especially in stage III GC. Examining PD-L1 expression is recommended for GC patients with liver metastases, especially aged male patients; immunotherapy may be beneficial for this group of GC patients. Further randomized controlled studies are required to validate our results.

Patients and sample collection
Between January 1992 and December 2013, in total, 2673 GC patients with adenocarcinoma who underwent curative surgery (R0 resection) were enrolled in the present study. Propensity-score matching was performed to minimize selection bias in the comparison of the clinicopathological features between the sexes.
Tumor and normal gastric mucosa tissues were collected and stored in a biobank at our institution. The genetic alterations of 438 GC patients with available tumor and normal gastric mucosa tissues were analyzed. The pathological staging of the GC was performed according to the 8 th edition of the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC) TNM classification system [35].

Follow-up
Follow-up examinations, including physical examinations, blood tests with measurements of tumor markers, chest radiography, and sonography or computerized tomography scans of the abdomen, were performed every 3 months during the first 3 years after surgery and every 6 months thereafter.

DNA extraction and analysis of microsatellite instability and genetic mutations
The DNA extraction from the tissue specimens was performed using a QIAamp DNA Tissue Kit and a MinElute Virus Kit (Qiagen, Valencia, CA, USA) according to a previous report [36].
As described in a previous study [12], five reference microsatellite markers, namely D5S345, D2S123, D17S250, BAT25 and BAT26, were used to determine MSI. MSI-high (MSI-H) was defined as ≥ 2 loci of instability with 5 markers, while MSI-low/stable (MSI-L/S) was defined as one unstable locus or no MSI loci.

Immunohistochemical (IHC) staining for PD-L1
IHC staining was performed to detect PD-L1 expression using PD-L1 IHC 22C3 pharmDx kit on a Dako ASL48 platform [13]. The combined positive score (CPS) was calculated, and a CPS score >1 was defined as positive expression of PD-L1.

Propensity-score matching strategy
As shown in Table 1, to minimize selection bias, propensity-score matching was performed based on logistic regression modeling of seven covariates (age, tumor location, gross appearance, Lauren's classification, lymphovascular invasion, adjuvant chemotherapy, and pathological TNM stage) to balance the potential confounders between the males and females. A 1:1 ratio matching of males and females was applied. A specific caliper width equal to 0.1 standard deviation was used.

Statistical analysis
IBM SPSS Statistics 25.0 (IBM Corp., Armonk, NY, USA) was used for the statistical analyses. A χ 2 test AGING with Yates correction or Fisher's exact test was used to compare the categorical data. The OS was defined from the date of surgery to the date of death or the last follow-up, while DFS was defined as the length of time after surgery during which the patient was alive without GC recurrence. The Kaplan-Meier method was used for the survival analysis of the OS and DFS. A multivariate analysis with Cox proportional hazards models was performed to analyze the independent prognostic factors of the OS and DFS. A P value < 0.05 was defined as statistically significant.

AUTHOR CONTRIBUTIONS
Li-Wen Hsu and Wen-Liang Fang carried out the statistical analysis and drafted the manuscript. Chien-Hsing Lin carried out the molecular genetic studies, including the MassARRAY analysis. Wen-Liang Fang and Kuo-Hung Huang collected patients' records and samples. Wen-Liang Fang conceived of the study and participated in its design and coordination as well as complete drafting of the manuscript. All authors read and approved the final manuscript.