br numerous subsets including HER and


numerous subsets, including HER2- and intestinal. Moreover, CXCR1/2 were identified as independent prognostic factors associated with RFS, instead of OS. Nonetheless, the Adrucil of CXCR1/2 between high/ low risk groups in the prognostic signature of CXCR remains insignif-icant. In fact, our results complemented the prognostic values of CXCR1/2 in GC with multi-dimensional clues. The expression of CXCR4 displayed a stage-specific pattern, indicating the potential association between tumor progression and CXCR4 expression. Previously, Masuda et al. reported that the immunoreactivity of nuclear CXCR4 was an independent prognostic factor and associated with reduced 5-year OS [40]. Moreover, Yasumoto et al. reported that positive CXCR4 in GC significantly associated with the peritoneal metastases, indicating the metastatic role of CXCR4 in GC [41]. Consistently, the signature prognosis analysis also highlighted the distinct upregulated expression pattern of CXCR4 in high risk group compared to the low risk group. The low risk group, with lower expression of CXCR4, showed a sig-nificant favorable OS than high risk group (HR = 3.22, 95% CI = 2.21–4.69, p = 1.057e−09). It may seem contradictory to the favorable survival outcome of CXCR4 evidenced by KM plotter using optimal cutoff (HR = 0.7, 95%CI: 0.58–0.86, p = 0.00048). However, using multivariate analysis, CXCR4 was finally identified as an in-dependent risk factor in both OS (HR = 1.72, 95% CI = 1.19–2.48,
Fig. 7. Nomogram model for the RFS of CXCR members. (A, B) The calibration plots for the 3- and 5- year RFS; (C) nomogram model for 3- and 5-year associated RFS; (D) the decision curve analysis.
p = 0.004) and RFS (HR = 3.20, 95% CI = 1.72–5.96, p < 0.001). Several issues needed to be addressed. Firstly, the different datasets in KM plotter and SurvExpress (TCGA) may account for the contradictory facts. Secondly, different statistic strategies were involved. Log rank analysis was used for KM plotter with optimal cutoff value while the entire CXCR member were included for prognostic risk algorithm ana-lysis prior to log rank analysis in SurvExpress. Thirdly, this study used univariate and multivariate analysis to confirm the independent prog-nostic values of CXCR4. Noteworthy, the prognostic values of CXCR signature were not further analyzed either by multivariate method or external validation. CXCR7 is the latest chemokine receptor implicated in oncological field [42]. Previous study indicated that high CXCR7 mRNA expression was a poor prognostic indicator for gastric cancer in both Singapore and Japanese cohorts [21]. Consistently, in our study, high expression of CXCR7 showed significantly poor prognosis in different pathological stages, HER2+ subset and intestinal subset, respectively. High level of five methylation sites of CXCR7 showed significant favorable OS. Moreover, CXCR7 was also an independent prognostic factor in the multivariate OS analysis. Reasonably presume, the prognostic values of CXCR7 could be further characterized and validated in several subsets, including HER2+ and intestinal type. In terms of HER2, the association between CXCR7 and HER2 is yet to be clarified. Previous studies 
indicated that CXCR7 expression was significantly reduced in HER2 type compared to luminal type in breast cancer [43]. Mechanistically, CXCR7/ERK1/2 signaling pathway could be involved with the regula-tion of BRCA1 and OTUB1 and further impact upon the stabilization of estrogen receptor α (ERα) [43]. Moreover, CXCR7 depletion could re-duce the phosphorylation site of ERK1/2 and epidermal growth factor receptor (EGFR) (Tyrosine 1110) [42]. In fact, the correlation between CXCR7 and ErbB family (EGFR, HER2) in GC remains further in-vestigation.