Background
Bladder cancer (BC) is the tenth most commonly diagnosed carcinoma, with an estimated 549,000 new cases and 200,000 deaths reported globally in 2018, and BC ranks the first in urinary malignant neoplasm among males [
1]. Therefore, it is crucial to developed accurate prognostic tools for predicting clinical results to help clinicians make decisions about treatment, drug therapy, and conservation options [
2].
Conventional signatures used to predict overall survival (OS) can range from tumor clinical parameters and tumor pathology to special mutated genes. For instance, the tumor node metastasis (TNM) classification system is the most frequently utilized to predict the prognosis of cancer patients [
3,
4]. Zhang et al. constructed a prediction tool based on clinical parameters to predict the survival of patients with BC [
5]. The most significant advantage of TNM is straightforward, but the inevitable disadvantage is not an individualized prediction for each patient [
6]. Besides, an increasing number of single signatures have been explored to predict the OS of BC patients, such as
OIP5 [
7],
B4GALT1 [
8],
ASPM [
9], and
HMGA2 [
10]. Xie et al. utilized the expression of
B4GALT1 to predict the prognosis of patients with muscle-invasive bladder cancer, and the expression of
B4GALT1 was correlated with OS of patients with BC [
8]. However, it is a challenge to predict the OS of patients with BC using a single signature, because of the impact of genetic heterogeneity [
11]. Therefore, it is essential to develop a comprehensive prognostic evaluation system that can improve the predictive accuracy of the prognosis of patients with BC.
Nowadays, gene-based prognostic signatures in conjunction with other clinical parameters have been explored extensively in predicting the OS of cancer patients [
12‐
14]. Song et al. identified signature combined immune-related genes and clinical characters to predict the OS of patients with BC, which suggested the signature was clinically useful for patients with BC [
15]. And a growing number of studies have shown that prognostic signatures dependent on gene expression levels have a strong potential to predict the prognosis of cancer patients [
16]. Therefore, in-depth analysis of gene expression databases may discover other prognostic genes and establish a robust prognostic signature, which can be a powerful tool for predicting cancer prognosis and individualized care [
13].
In our study, we developed a signature to predict OS of BC patients based on multiple prognostic genes and clinical parameters. The RNA-seq was downloaded from TCGA and GEO, and analyzed via DEGs analysis. Then we utilized univariate Cox regression, LASSO regression with tenfold cross-validation, and stepwise multivariate Cox regression to identify six candidate genes. And the gene-based risk score was calculated through the stepwise multivariate cox coefficient multiplied by the expression of the gene. Then a nomogram was constructed based on the risk score and clinical parameters, which was assessed by the calibration plot, decision curve analysis (DCA), and time-dependent ROC analysis. Finally, potential pathways of these candidate genes were analyzed via functional enrichment analysis, Gene Ontology (GO) enrichment, and Kyoto Encyclopedia of Genes and Genomes (KEGG). Bioinformatic methods “guilt by association” (GBA) [
17] and Gene set enrichment analysis (GSEA) were applied to explore the mechanism of candidate genes.
Discussion
The incidence of BC is a crucial neoplasm among men, with respective incidence and mortality rates of 9.6 and 3.2 per 100,000 in men: about 4 times those of women globally [
1]. It is necessary to screen potential prognostic biomarkers and construct satisfying tools to predict the survival of patients with BC.
In the previous study, numerous prognosis predictions of patients with BC are based on clinical information only [
5,
35,
36]. TNM staging system is commonly used to predict the prognosis of bladder cancer. However, as we mentioned above, the single clinical parameter has poor power of prognosis prediction [
3]. Therefore, combining other prognostic parameters would be the better way to boost the accuracy of prediction.
In our study, the DEGs between normal tissue and tumor were firstly obtained from three datasets. The intersected genes between DEGs and prognosis-related genes sifted out from the training set were analyzed with LASSO-penalized regression and stepwise multiple Cox regression to screened six candidate genes (SORBS2, GPC2, SETBP1, FGF11, APOL1, H1–2). As we are concerned, the method of screening candidate genes via intersecting DEGs and prognosis-related genes was not similar to most bladder cancer prediction model research. The six genes, except SORBS2, are significantly related to the overall survival of patients with bladder cancer.
GPC2, glypican 2, is a type of cerebroglycan related to oncoprotein. Bosse et al. showed that
GPC2 can be a candidate immunotherapeutic target in High-Risk neuroblastoma [
37]. Shou et al. showed that SETBP1 mutation is associated with a poor prognosis in patients with myelodysplastic syndromes [
38]. However, the role of
GPC2 and
SETBP1 in urothelial carcinoma is not certain due to the lack of sufficient studies.
FGF11, fibroblast growth factor 11, is a member of the fibroblast growth factor (FGF) family. Researchers reported that
FGF11 acts as a novel modulator of hypoxia-induced tumor progression [
39,
40].
APOL1, apolipoprotein L1, encodes a secreted high-density lipoprotein, which binds to apolipoprotein A-I. Some researches indicated
APOL1 is related to cardiovascular disease and renal disease [
41,
42].
H1–2, H1.2 linker histone, is also called
HIST1H1C. Li et al. reported that inhibition of H1.2 phosphorylation at T146 was related to the carcinogenic role of K-Ras-ERK1/2 signaling in bladder cancer [
43]. This aspect of
H1–2 was also verified in our analysis that the hazard ratio (HR) of H1–2 was significantly less than 1 (Fig.
3 E) and the patients with low
H1–2 expression had a high probability of death, which means the low expression of
H1–2 is related with progression and bad prognosis of patients with BC.
Among these five genes (GPC2, SETBP1, FGF11, APOL1, H1–2) related to the prognosis of patients with BC, there are no reports or experiments about these genes related to bladder cancer, except for H1–2. Based on our analysis, these genes may be a potential novel therapeutic target for patients with BC. The mechanism of these four genes is worth to be explored.
The KM survival analysis for the training set and risk stratification in patients with gender, age, race, AJCC stage, AJCC-T, AJCC-N, AJCC-M showed that the risk score had relatively median accurate OS prediction. As for the patients in the T0/1/2 group, low-risk group had worse OS than high-risk group. The reason was that the number of patients with T0/1/2 was probably insufficient, and the bias of this subgroup was enlarged. The time-dependent ROC indicated that the AUC of the nomogram was larger than that of the risk score, resulting from the combination with clinical parameters. It is reasonable that age is an essential risk factor in the progression and prognosis of patients. Some researchers also demonstrated that senescence was associated with a pathological process such as cancer [
44]. Therefore, the six-gene-based prognostic nomogram can assist clinicians in predicting the survival outcome of BC patients and provide a more reliable reference for therapy guidance than the single conventional clinical parameter. Besides, these six genes have not been previously studied as prognostic genes in BC patients. To some extent, it is necessary to conduct the following functional experiment exploration based on these six prognostic genes.
The limitations of this study are supposed to be discussed. Although we screened and identified six genes potentially related to the progression and prognosis of patients with BC via some statistical methods and we explored the potential pathways and mechanism of each gene, this study is lacking experiments (in vivo and in vitro validation) to validate the link between these genes and BC. Therefore, these analyses can be our follow-up studies.
Conclusion
In our current study, we screened six novel prognosis-related DEGs from the public database and constructed a six-gene-based prognostic nomogram that contained other clinical parameters, such as age, gender, pathological stage, to predict the 1-year, 3-year, 5-year OS of patients with BC. The estimation showed that the nomogram has relatively stable accuracy in the prediction of OS. That is to say, the six genes could be potential biomarkers in BC and, in clinical practice, the related gene-based nomogram could theoretically be utilized to predict the individual survival rate and facilitate the selection of individual treatment options.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.