In our study, the UPR signal pathway was identified by GSVA analysis as a highly activated channel in tumor cells of GC patient, which keeps consistence with former experiment.[36] It is reasonable that UPR is extremely upregulated as a result of much new antigen generated by GC cell. Previous studies has verified this upregulation is associated with several mechanism of GC such as the proliferation, migration, and treatment resistance.[37–39] However, no research focused on the prognostic value of UPR signal transduction pathway. In our study, a novel 6 genes signature selected from UPR gene set was constructed and validated to well predict OS of GC patients. Then, a nomogram was built to predict 5-years OS time of GC patients. To the best of our knowledge, this is the first UPR related prognostic gene signature for GC, indicating that UPR path play an important role in GC and these 6 UPR related genes (SLC1A4, FKBP14, TUBB2A, SEC11A, IFIT1, IGFBP1) are key molecules working in this process.
GC features by its highly heterogeneous genome variation and complex interplay in patients and environmental factors such Helicobacter pylori, virus infection, dietary habits, tobacco and others.[40] Accounting for carcinogenesis and progression of GC, widely genome aberrance and complex regulatory networks are acting and changing all the time.[41] Consequently, multi-molecule models have demonstrated better effect than single ones for diagnosis and prognosis of GC.
There are several studies having established multi-molecule biomarkers for GC prognosis including mRNAs, non-coding RNAs, DNA methylation, and so on.[42–48] All these models demonstrated good prediction effect for OS or RFS (Recurrence free survival) of GC. But seldom of them provided a detailed description of normalization process with GEO array express data. It is suggested that data normalization is pretty significant to eliminate the impact of batch effect which can intensely influence the analyzing result.[49] In our study, we presented the batch effect of GSE84437 and the normalized data after overcoming this disadvantage, making our prognostic signature more robust and reliable. Although the 1-,3-,5-year AUC of our 6-URGs prognostic signature are lower than that of former studies [42–48] (most AUC > 0.7), this doesn’t imply a lower prediction performance of it as a big sample size (406 and 431 in our case) can strongly affect the AUC of a model and AUC can even close to 0.5 when sample size is larger than 500.[50]
It is necessary to search the underlying mechanisms of the 6 UPR genes identified in our study. FKBP14, also called FK506-binding proteins 14,having been reported to be an oncogene in certain human cancers, can promotes colon cancer cells’ proliferation and migration through IL-6/STAT3 signaling pathway.[51] In GC patients, FKBP14 is uprregulated and stands for a worse survival in experiment[52],which is in line with our result. Further experiments are needed to explore the exact mechanism. TUBB2A, one major constituent of microtubules, are reported to be high-expressed in colon cancer and associates with poor survival outcome [53], without previous report of its relation with GC. SLC1A4, a neutral amino acid transporter, works as the common target of MYC, AR (androgen receptor) and mTOR signaling pathways to promote glutamine uptake and subsequent growth of prostate cancer.[54] The upregulation of SLC1A4 in pancreatic cancer helps to maintain the alanine concentration to stimulate cancer cell growth.[55] Interestingly, SLC1A4 is a protective factor in our prognostic signature, indicating its unknown effects and corresponding mechanisms in GC. SEC11A, one component of the microsomal signal peptidase complex, can encodes the SPC18 protein which has been proven to promote gastric and colon cancer proliferation.[56, 57] IFIT1 belong to IFIT (interferon-induced protein with tetratricopeptide repeats) gene family and are also a subset of IRDS (interferon-related DNA damageresistant signature) which contributes to tumor growth and drug resistance.[58] IFIT1 and IFIT3 can promote oral squamous cell carcinoma metastasis through EGFR path [59], but no description of its effect on GC is reported. IGFBP1 is one of the cysteine-rich proteins that bind to IGFs (insulin-like growth factor family) to modulate its function. It is considered that IGFBP1 can promote or inhibit IGF path action.[60] In the breast cancer cell line of MCF-7, IGFBP1 can inhibits IGF1-induced tumor growth.[61] However, it was viewed as a risk factor in our model, suggesting a complex underlying mechanism of IGFBP1 in GC. All in all, the 6 UPR related genes play significant roles in GC and may serve as potential biomarkers and therapeutic targets for GC treatment.