Somatic variations in male and female GC patients
We first compared the frequency of somatic mutations from TCGA cohort between female and male GC patients. Intriguingly, only ATRX mutation were found to significantly occur more in female patients compared to male patients after adjusting the false discovery rate (FDR = 0.0108) (Figure 1A).
To avoid accidental bias caused by a single cohort, we further verified the sex biases of ATRX mutation in cohort of GACA-CN and GACA-JP. The mutation frequency of ATRX in GACA-CN and GACA-JP was 31.71% (39/123) and 6.32% (37/585) respectively. The proportions of female patients with ATRX mutation in GACA-CN (female 40.5% vs male 27.9%, Fishers exact test, P = 0.206) and GACA-JP (female 11.0% vs male 4.3%, Fishers exact test, P = 0.004) were both higher than that of male patients (Figure 1B). Subsequently, we attempted to examined whether the sex dipartites in ATRX mutation could be found in other cancer projects from ICGC. We excluded the sex-restricted tumors and cancer projects with ATRX mutation frequency less than 5%. Remarkably, half of cancer projects (50%, 17/34) showed significantly sex differences in ATRX mutation (Figure 1C, Additional file 2). In most cancer projects (88.2%, 30/34) the mutation frequency of ATRX was higher in female patients than that of male patients, although the difference in some cancer projects did not show statistical significance. In addition, we evaluated the potential impact of ATRX missense mutation on protein structure by PolyPhen-2 software. The deleterious missense mutations in female patients were predominant among all kinds of ATRX mutation in TGCA cohort (Additional file 3). Overall, these results indicate the ATRX mutation in gastric cancer patients might exist sex biases.
Correlation between ATRX mutation and clinical features
We first compared the distribution of clinical features between ATRX mutant type and wild type patients (Table 1). Remarkably, the majority of ATRX mutant patients were microsatellite instability-high (MSI-H) tumors compared to the ATRX wild type patients (Figure 2A). Then we proceeded to examine whether ATRX mutation affect the overall survival of GC patients. Log-rank test in Kaplan-Meier analysis demonstrated that ATRX mutant patients trended towards better overall survival than wild patients (Figure 2B). Kaplan-Meier analysis of patients from GACA-CN also showed the similar result (Figure 2C). Next, we evaluated the effect ATRX mutation on ATRX expression. We found that ATRX mutant tumors had lower ATRX expression, although the difference did not manifest statistical significance (Figure 2D). In summary, ATRX mutation was significantly correlated to MHS-H subtype and better over survival in GC patients.
ATRX mutation is associated with increased tumor mutation burden and higher expression of PDL1
Tumor mutation burden (TMB) and PDL1 expression were two common indicators of predicting response to immunotherapy. We calculated the TMB of each GC patient from TCGA, GACA-CN and GACA-JP respectively. As shown in Figure 3A, higher TMB was observed in ATRX mutant GC patients from cohort of TCGA (median TMB 16.61 vs 2.526, P < 0.0001), GACA-CN (median TMB 1.974 vs 1.211, P = 0.0001) and GACA-JP (median TMB 27.82 vs 2.421, P < 0.0001). We then compared the expression of PDL1 between GC patients with or without ATRX mutation in TCGA cohort. ATRX mutant patients harbored higher expression of PDL1 (median expression 2.027 vs 1.098, P = 0.0244) than that of wild patients (Figure 3B).
Given the potential sex differences in ATRX mutation, we further to explore the correlation of ATRX mutation with TMB and PDL1 expression stratified by sex (Sub1: female patients with ATRX mutation; Sub2: female patients without ATRX mutation; Sub3: male patients with ATRX mutation; Sub4: male patients without ATRX mutation). In cohort of GACA-CN and GACA-JP, both female and male patients with ATRX mutation had higher TMB than ATRX wild type patients (Figure 3C). However, in cohort of TCGA, only ATRX mutant type female patients showed higher TMB than other patients without ATRX mutation. There is no significant difference of TMB between male patients with or without ATRX mutation. Furthermore, in three cohorts no significant differences were observed between female and male patients with ATRX mutation. When comparing the PDL1 expression between the four subgroups in TCGA cohort, we found that the expression of PDL1 in ATRX mutant female patients was higher than patients without ATRX mutation (Figure 3D).
In general, these results suggested that patients with ATRX mutation, especially female GC patients, were significantly associated with higher TMB and PDL1 expression.
Female GC patients with ATRX mutation show enhanced anticancer immunity
We resorted to several proven immunotherapy-related factors, including cytolytic activity (CYT)26, IFN-γ signaling6, antigen presenting machinery (APM)27 and tumor infiltrating immune cells28, to characterize the anticancer immunity. We have used the method of ssGSEA to explore the association of ATRX mutation with anticancer immunity. Although the scores of anticancer immunity in patients with ATRX mutation were higher than that in patients without ATRX mutation, the difference exhibited no statistically significant (Figure 4A).
However, when assessing the influence of ATRX mutation on anticancer immunity stratified by sex, we found that female patients with ATRX mutation showed significantly higher scores of the immunotherapy-related factors than patients without ATRX mutation (Figure 4B). There were no significant differences of anticancer immunity scores between ATRX mutant and wild male patients. Furthermore, female patients with ATRX mutation manifested higher scores of immunotherapy-related indicators than ATRX mutant male patients.
Finally, we used CIBERSORT method to investigate the proportions of infiltrating immune cells among ATRX mutant female patients (Sub1), ATRX wild female patients (Sub2), ATRX mutant male patients (Sub3) and ATRX wild male patients (Sub4). T follicular helper cells, M1 macrophages and activated dendritic cells were significantly enriched in female patients harboring ATRX mutation, while the proportions of resting dendritic cells were significantly lower than other patients (Figure 4C). T follicular helper cells, M1 macrophages and activated dendritic cells are associated with antigen presentation machinery, which were consistent with the increased APM in female patients with ATRX mutation.
In conclusion, female GC patients with ATRX mutation showed stronger anticancer immunity than ATRX wild patients or mutant male patients. The enrichment of T follicular helper cells, M1 macrophages and activated dendritic cells in ATRX mutant female patients also supported the above result.
The effect of ATRX mutation on anticancer immunity in female patients is not affected by sex-based immune heterogeneity
In order to examine whether the effect of ATRX mutation on anticancer immunity in female patients was due to the sex-based immune heterogeneity, we further compared the above anticancer immunity factors between male and female patients. Only TMB displayed the sex differences, and the female patients had higher TMB than male patients (Additional file 4). However, the differences of TMB between women and men disappeared when we exclude the female samples with ATRX mutation (Additional file 4). These results revealed that the enhanced anticancer immunity in ATRX mutant female patients was not affected by sex-based immune heterogeneity.
Independent prognostic analysis of ATRX mutation
We first carried out the Kaplan-Meier analysis to determine the prognostic values of the above immunotherapy-related factors. The TMB and PDL1 expression were divided into high and low groups on the basis of the X-tile tool. As shown in Figure 5, the TMB and PDL1 expression exerted the influence on patients’ survival. The high group of TMB and PDL1 implied a better outcome. We then included TMB and PDL1 expression into the subsequent analysis to examine the independent prognostic value of ATRX mutation. Univariate and multivariate Cox regression analyses illustrated that ATRX mutation was the independent prognostic factor (Table 2).
Impact of ATRX mutation on DNA damage repair
We performed GSEA to analyze the functional context of ATRX mutation. KEGG pathway analysis showed ATRX mutation were mainly enriched in base excision repair (BER), nucleotide excision repair (NER) and homologous recombination repair (HRR) (Figure 6A). The biological function of ATRX mutation was significantly involved in DNA damage repair (DDR). We then used the ssGSEA to characterize the pathway of BER, NER and HRR between ATRX mutant and wild patients. As illustrated in Figure 6B, BER (median score 0.8734 vs 0.7909, P = 0.0002), NER (median score 0.4333 vs 0.3605, P = 0.044) and HRR (median score 0.7618 vs 0.6898, P = 0.0098) were significantly enriched in ATRX mutant patients. Furthermore, we continued to explore the effect of ATRX mutation on BER, NER and HRR pathways stratified by sex. Only female patients with ATRX mutation manifested higher scores of BER, NER and HRR than other patients without ATRX mutation (Figure 6C).
Moreover, we chose the core genes enriched in these three pathways based on GSEA results to examine the expression differences between ATRX mutant and wild patients. Interestingly, most core genes expressed highly in ATRX mutant type patients when comparing that of wild type patients (Figure 6D). After sex stratification, the expression levels of these DDR-related core genes were also higher in subgroups of female and male patients with ATRX mutation than other two wild type subgroups (Additional file 5).
Taken together, these results implied the possibly enhanced activation of DDR pathways in ATRX mutant patients. The higher TMB and activation of DDR system might indicate the likelihood of fierce mutagenesis and corresponding compensatory DDR activation in ATRX mutant patients.
ATRX mutation is associated with favorable clinical benefit to ICI
In order to investigate whether patients with ATRX mutation could benefit from ICI, we obtained the publicly available genomic and survival data of MSKCC ICI-treated cohort from cbioportal database15. The MSKCC ICI-treated cohort contains 1610 patients of various cancer types with mutation data including 54 gastric cancer patients. The mutation frequency of ATRX in the gastric or gastroesophageal junction cancer patients is 7.41% (4/54). We observed that patients with ATRX mutation had higher TMB than ATRX wild type patients (Figure 7A). In the patients of gastric cancer, ATRX mutation patients trended toward a longer overall survival (median, not reached vs 13 months, log rank P = 0.194) (Figure 7B). Similarly, the overall survival of patients with ATRX mutation in the whole cohort was also superior to that of ATRX wild patients (median, 30 months vs 18 months, log rank P = 0.075) (Figure 7B). We then divided the whole cohort into two groups (female and male) and compared the overall survival between ATRX mutation statuses. Interestingly, only female patients with ATRX mutation obtained significantly prolonged overall survival compared with wild type female patients (median, 30 months vs 14 months, log rank P = 0.023) (Figure 7C). Furthermore, we compared the prediction power of ATRX mutation in male and female patients. The HR of ATRX mutant type vs wild type for female patients (HR = 0.636, 95%CI = 0.455-0.890, P = 0.023) is better than the HR for male patients (HR = 0.929, 95%CI = 0.596-1.362, P = 0.7117) (Figure 7C).
Collectively, these data suggested that ATRX mutation especially in female patients was associated with favorable clinical benefit to ICI treatment.