Prognostic values of signal transducers activators of transcription in gastric cancer

The signal transducers and activators of transcription genes family (STATs) have been well studied as prognostic predictors for various solid tumors, but their prognostic values in gastric cancer (GC) patients have not been fully elucidated. The ‘Kaplan–Meier plotter’ and multiple public available databases were used for the characterization of the prognostic roles of STATs family in GC. The results indicated that high mRNA expression of all individual STATs, except STAT3 and STAT6, were significantly associated with favorable overall survival (OS) in GC. Moreover, the prognostic values of STATs were further characterized in subtypes, including HER2 status, Lauren’s classification, differentiation, and clinical stages. Moreover, the prognostic value of STATs signature was also characterized. Low risk group displayed a significantly favorable OS than high risk (HR: 1.71; 95% CI: 1.09–2.66, P=0.0184). In addition, STATs showed distinct expression between GC and normal groups. Meanwhile, comparable high correlation between STATs and tumor immune infiltrating cells (TIICs) was also observed. STAT4 displayed highest correlation with dendritic cells (correlation = 0.716, P=1.63e-59) and CD8+ T cells (correlation = 0.697, P=5.02e-55). In conclusion, our results suggest that all individual STATs, except STAT3 and STAT6, may act as prognostic markers in GC.


Introduction
Gastric cancer (GC) is the fourth most common cancer and second leading cause of cancer-related deaths in the world, accounting for approximate 9% of total cancer deaths [1,2]. Although the 5-year overall survival (OS) can reach more than 90% in early GC by early diagnosis and multi-disciplinary therapeutic strategies, the prognosis of advanced or metastatic GC patients remain largely unsatisfactory, with median survival period being around 1 year [3,4]. Therefore, novel prognostic biomarkers for GC could contribute to the identification of those risky cases and maximize the OS benefits.
Signal transducers and activators of transcription (STATs) are a gene family of cytoplasmic transcription factors, consisting of seven members, STAT1-STAT4, STAT5a, STAT5b, and STAT6 [5,6]. STATs play important roles in numerous biological processes, including cell proliferation, differentiation, apoptosis, and survival [7]. STATs are activated via tyrosine phosphorylation, a process which occurs either through KIT-based interaction or cytokine-induced JAK pathway [8]. STATs can also be activated by constitutively activated non-receptor protein tyrosine kinases (PTKs), including c-Src and Bcr-Abl. Activated STATs rapidly translocate into nucleus, and bind to the promoter region of target genes, serving as transcription regulators. Increasing evidence have indicated that the STATs, particularly STAT1, STAT3, and STAT5 play critical roles in various cancer progressions and have been identified as potential therapeutic targets [9][10][11][12][13]. Nevertheless, reports focusing on the relationship between GC and STATs remain limited [14][15][16]. In fact, the prognostic values of STATs family in GC patients are yet to be fully characterized.

Prognostic values of STATs signature via the SurvExpress platform
The prognostic values of STATs signature were explored via the SurvExpress platform (http://bioinformatica. mty.itesm.mx:8080/Biomatec/SurvivaX.jsp) with the stomach adenocarcinoma (STAD) of the cancer genome atlas (TCGA) selected as input dataset (n=352) [18]. High/low risk groups were categorized based on the default prognostic risk algorithm [18].

Oncomine database analysis
The mRNA expression of STATs members was investigated via the Oncomine database, which was a comprehensive cancer genomic platform [19]. Briefly, the mRNA expression of each STAT gene was compared between cancer and normal groups with P-value <0.05 set as statistically significant. Only significant results were displayed.

Prognostic values of STATs in all GC patients
The prognostic values of STATs in all GC were investigated via the KM plotter. Among seven STATs members, five were significantly associated with prognosis for all GC patients ( Figure 1A

Prognostic values of STATs in HER2 subtypes
Next, the prognostic values of STATs associated with the HER2 subtypes in GC patients were explored (Figure 2). High mRNA expression of STAT1, STAT5a, and STAT5b were associated with better OS in HER2-negative GC patients ( Figure 2). Meanwhile, high expression of STAT4 and STAT6 were associated with better OS in HER2-positive GC patients. STAT5a was modestly associated with favorable OS in HER2-positive GC patients ( Figure 2). The rest STAT members were not significantly correlated with prognosis.

Prognostic values of STATs in GC patients with differentiation subtypes
High mRNA expression of STAT1 and STAT2 were associated with unfavorable prognosis in moderately differentiated type GC, but with favorable prognosis in poorly differentiated type GC ( Figure 4A-D). High expression of STAT6 was associated with unfavorable OS in poorly differentiated type GC ( Figure 4E).

Prognostic values of STATs in different GC clinical stages
Only high expression of STAT4 was associated with worse prognosis in stage I in GC patients (Table 1). In stage III, high expression of STAT1, STAT2, and STAT4 were associated with favorable OS, while STAT5b was associated with poor OS ( Table 1). None of the STAT members were found to be correlated with OS in stages II and IV. In addition, high expression of STAT6 was associated with worse prognosis in lymph node-negative GC. For lymph node-positive GC, STAT1 and STAT3 were associated with worse survival, whereas STAT4 was associated with better OS (Table 1). High level of STAT1 mRNA expression was associated with worse prognosis both in M0 and M1 stage GC patients whereas STAT3 was associated with poor OS in M0 stage ( Table 1).

The prognostic value of STATs signature
Intriguingly, low risk group displayed a significantly favorable OS than high risk (HR: 1.71; 95% CI: 1.09-2.66, P=0.0184) ( Figure 5C), indicating potential prognostic role of STATs signature in GC. Of note, STAT5B showed   the most up-regulated expression in high risk compared with low risk (Figure 5A, B).

The mRNA expression analysis of STATs in GC
The expression levels of STATs have been investigated via the Oncomine database. Intriguingly, compared with normal group, STATs showed distinct expression not only in general GC, but also in Laurens subtypes (intestinal, mixed, and diffuse types) and pathological types (papillary adenocarcinoma and tubular adenocarcinoma) ( Table 2).  (Figure 6).

Discussion
STATs gene family have participated in diverse roles in carcinogenesis and tumor progression. In the present study, the prognostic roles of STATs members in GC patients were assessed through the KM plotter. High mRNA expression of STAT1, STAT2, STAT4, STAT5a, and STAT5b were significantly correlated to a favorable OS in GC patients. Moreover, inverse prognostic values of STAT1, STAT2, and STAT6 were found between Lauren's subtypes. Meanwhile, we also observed inverse prognostic values of STAT1 and STAT2 between the moderately and poorly differentiated subtypes. Several reasons may hold accountable. First, comparable smaller sample size in subtype analysis may be a confounding bias factor in prognosis evaluation, particularly in mixed subtype of STAT1 (n=25), STAT6 (n=25). Second, both Lauren's classification and differentiation represent key pathological features of GC, in which molecular features are yet to be fully disclosed. Distinct prognostic values between each subtype in the present study may contribute to deeper knowledge of the pathological keys in GC. Third, the inverse prognostic values open up a new question as whether STATs may have opposite roles in each subtype.
Several studies had reported that STAT1 plays a role in gastric inflammation and tumorigenesis in mice model [27,28]. But so far, studies about STAT1 and prognosis in GC are limited. Deng et al. [29] indicated that STAT1 mRNA was related to favorable prognosis in GC. Our results revealed that high expression of STAT1 mRNA was associated with a better OS in all GC patients. This result is in line with previous result. Stage specifically, we observed that high mRNA expression of STAT1 indicated a better OS in stage III. Collectively, our study suggested that STAT1 may be a prognosis predictor especially for late stage and poorly differentiated GC patients.
The role of STAT2 in GC remains limited. In the current study, we found that high STAT2 mRNA level was associated with favorable OS. Similar to STAT1, high STAT2 mRNA expression was associated with better OS in poorly differentiated and stage III GC patients. STAT3, activated by tyrosine phosphorylation in response to growth factors and cytokines, was mainly involved in the oncogenesis of several human cancers, including GC [30,31]. Previous studies showed the relationship between STAT3 mRNA expression and the prognosis in GC patients. However, controversies remained. Some studies [14,[32][33][34][35][36][37] indicated that elevated STAT3 mRNA expression was associated with poor outcomes in GC patients. Nonetheless, Woo et al. [37] reported that the overexpression of STAT3 was correlated with favorable outcome of   patients with GC. Nevertheless, Xiong et al. [38] and Lee et al. [39] showed no correlation between high STAT3 mRNA expression and OS. Interestingly, our data found there was no significant correlation between STAT3 mRNA expression and OS of GC patients. This can be attributed to different study design, race diversity, clinical stage, sample size, and cutoff definition. Moreover, Chatterjee et al. [34] showed that STAT3 expression was associated with poor prognosis in the intestinal subtype, consistent with our findings. Furthermore, Deng et al. [35] and Kim et al. [16] demonstrated that high STAT3 expression was significantly associated with lymph node invasion, which was consistent with our results that high STAT3 mRNA level was significant associated with unfavorable OS. The present study also indicated that STAT3 mRNA expression was correlated to poor prognosis in M0 stage GC patients. Collectively, although STAT3 showed no effect on OS, but it remained a potential prognostic predictor in subtypes of GC. Similar to STAT2, studies about STAT4 and its prognosis in GC patients are limited. Nishi et al. [40] indicated that high STAT4 expression was associated with better disease free survival in GC. Our results revealed that high STAT4 mRNA expression was significantly associated with better OS for all GC patients, subtypes in HER2-positive, diffuse subtype, stage III, and lymph node-positive GC.
STAT5 is consisted of two highly homologous genes, STAT5a and STAT5b [41]. In GC, Kim et al. [16] reported that STAT5 had no statistical significance in the analysis of survival. However, our results suggested that high mRNA expression STAT5a and STAT5b were significantly associated with better OS.
So far, studies focusing on the prognostic value of STAT6 expression in malignancies, especially GC, remain limited. In the present study we found that high STAT6 mRNA level was not significantly associated with OS.
The present study has some limitations, including lack of experimental and clinical validations to confirm the prognostic values of STATs in GC, as well as comparably small sample size in some subtypes analysis. More studies with large sample size are warranted to validate the prognostic value of STATs family.

Conclusion
The present study showed that high mRNA expression of STATs except STAT3 and STAT6 were significantly correlated to favorable OS in GC patients.