High Expression of SRD5A3 in Human Hepatocellular Carcinoma (HCC) Predicts Poor Prognosis: A Study Based on TCGA Data


 Background: Hepatocellular carcinoma (HCC) is the fourth leading cause of cancer-related death worldwide. Steroid 5 alpha-reductase 3 (SRD5A3) was reported to be up-regulated in many types of cancer. However, its expression and role in HCC remains to be elucidated. We aim to evaluate the significance of SRD5A3 expression in HCC by using analysis of a public dataset from The Cancer Genome Atlas (TCGA).Methods: The relationship between clinical pathologic features and SRD5A3 were analyzed with the Kolmogorov‐Smirnov test and the logistic regression. Cox regression and the Kaplan-Meier method were used to assess the clinicopathologic characteristics associated with overall survival (OS) in TCGA patients. In addition, GSEA was used to predict potential hallmarks associated with different expression of SRD5A3 on transcriptional sequences from TCGA database.Results: SRD5A3 was highly expressed in HCC tumor tissue compared to normal tissue. A total of 184 upregulated DEGs (differentially expressed genes) and 58 downregulated DEGs were identified between high expression and low expression of SRD5A3. Among them, 22 hub genes mainly belonging to the keratin and MUC family demonstrated by connectivity degree in the PPI network were screened out. Kaplan-Meier method showed that HCC patients in the high SRD5A3 expression group had poorer overall survival (OS, HR=2.26(1.58-3.24), p<0.001). In addition, cell cycle mitotic, cell cycle checkpoints, mitotic nuclear division, Q-glycan processing, protein O-linked glycosylation were differentially enriched in the high SRD5A3 expression phenotype pathway. In addition, SRD5A3 expression level has significant correlations with infiltrating levels of Th17 (R = -0.238, p < 0.001), Cytotoxic cells (R = -0.234, p < 0.001) and Th2 cells (R = 0.258, p < 0.001) in HCC.Conclusions: High expression of SRD5A3 was significantly correlated with poor prognosis in HCC patients. It may be a potential biomarker in HCC.


Background
According to 2018 statistics from the World Health Organization, primary liver cancer is the sixth most common cancer and the fourth in mortality worldwide, causing about 840,000 new cases and more than 780,000 deaths annually [1]. Of all liver cancer cases, >90% are hepatocellular carcinomas (HCCs) [2]. Patients are often diagnosed with liver cancer in advanced stages, contributing to its poor prognosis [3]. In the past decade, there have been improvements in non-drug therapies and drug therapies for HCC treatment. Surgical resection is the primary treatment for primary hepatocellular carcinoma. However, since hepatocellular carcinoma usually occurs in the context of cirrhosis, residual liver resection still presents a speci c tumor microenvironment, and recurrence becomes the most serious problem. In recent years, cancer immunotherapy has become a research hotspot by arousing the immune system to respond to developing cancer, but its clinical application is still limited, and clinical trials for molecular therapies targeting different aspects of the hepatocellular carcinogenesis cascade are under way. To provide a highly accurate prediction of a patient' survival and/or response to individualized treatment therapy, new biomarkers has been conducted among patients.
Studies have con rmed that steroid 5α reductase (SRD5A3), as an important component of steroid metabolic cascade, irreversibly converts testosterone (T) to dihydrotestosterone (DHT), and binds to androgen receptor (AR) to form DHT-AR complex that speci cally acts on AR response elements and promotes the transcription of downstream target genes. Such as the Wnt/β-catenin signaling pathway, CD24, and vascular endothelial growth factor (VEGF), promote tumor proliferation, metastasis and angiogenesis [4]. To date, three types of human 5α-steroid reductase enzymes have been reported (types 1, 2 and 3). Both SRD5A1 and SRD5A2 are expressed in human hepatocytes, and SRD5A1 alone is expressed in adipocytes [5]. Steroid 5 alpha-reductase 3 (SRD5A3) mainly distributed in the brain, retina, heart, liver and other tissues, a protein-coding gene that irreversibly reduces steroid hormones and eventually becomes the corresponding metabolite with biological activity [6]. Recent studies have found that it is overexpressed in a variety of cancers. SRD5A3 overexpressed speci cally in HRPC cells and played important roles in HRPC growth and progression.is a promising molecular target for prostate cancer therapy [7,8]. In addition, SRD5A3 is validated as independent prognostic markers of EC and Triple-Negative Breast Cancer [9,10]. However, previous studies have focused on particular aspects of SRD5A3 in other diseases and thus lack a comprehensive exploration and description.
Thus, the objective of this study is to evaluate the prognostic value of SDR5A3 expression in HCC based on data obtained from TCGA. We investigated the expression levels of SRD5A3 and correlated it with survival in HCC patients, in the hope that our ndings could be helpful for understanding of pathological process and aggressiveness in HCC. Through immune in ltration analysis, gene enrichment analysis and protein-protein interaction (PPI) network, we further investigated the important role of SRD5A3 in the immune microenvironment and cell cycle, and further explored the potential value of SRD5A3 in immunotherapy and chemotherapeutic resistance therapy related to cell cycle mechanisms. It is expected to provide some guidance for the determination of new targets for precise treatment of HCC.

Materials And Methods
Data acquisition SRD5A3 expression and clinical data of TCGA pan-cancer data and GTEx were obtained from the UCSC Xena database [22] (https://xenabrowser.net/datapages/). The clinical data for the HCC cohort were obtained from the data from TCGA (https://cancergenome.nih.gov/). Patients who have insu cient clinicopathological data or "0" gene expression values were not included. Since the data come from these online databases, it is not necessary to get the study approval by the ethics committee. The RNA-Seq gene expression FPKM (Fragments Per Kilobase per Million) of 371 cases with HCC and clinical data were retained and further analyzed ( Table 1). The HTSeq-FPKM data were transformed to TPM (transcription per million reads) for the following analyses.
Analysis of differentially expressed genes (DEGs) for HCC between high and low SRD5A3 expression groups.
According to the median expression value of AAA in TCGA XXX samples, patients were classi ed into low-and high-expression groups. Expression pro les (HTSeq-TPM) were compared between high and low SRD5A3 expression groups to identify differentially expressed genes (DEGs) within the limma Package software by using R [23]. Genes with |logFC|>1.5 and FDR<0.05 were considered as DEGs.

Functional Enrichment analysis
The Metascape database (http://metascape.org/) is a free and credible gene-list analysis online tool for high-throughput functional analysis of genes. It can realize habitual and understand common and unique pathways within a group of orthogonal target-discovery studies. Metascape was applied to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. GO analysis is a common useful method for annotating genes and gene products and for identifying molecular function (MF), biological process (BP) and cellular component (CC) attributes for high-throughout genome or transcriptome data. KEGG is a collection of databases used for systematic analysis off gene functions and associating related gene sets with their pathways. In this study, we assessed the functions of SRD5A3 and their co-expression genes. The P value was set as 0.01, and the enrichment factor of >1.5 and minimum count of 3 were considered signi cant. We then selected a subset of representative terms from this cluster and converted them into a network layout. Molecular Complex Detection (MCODE) was supposed to recognize closely related protein components.

Gene Set Enrichment Analysis
GSEA was used to predict potential hallmarks associated with different expression of SRD5A3 on transcriptional sequences from TCGA database. A permutation test with 10000 times was used to identity the signi cantly changed pathways. Adj.P less than 0.05 and FDR less than 0.25 were identi ed as signi cantly related genes. Statistical analysis and graphical plotting were conducted using R package clusterPro ler (3.8.0) [24].

Immune in ltration analysis by ssGSEA
The immune in ltration analysis of HCC was done by ssGSEA (single-sample gene Set Enrichment Analysis) method using GSVA package [25] in R (3.8.0) for 24 types of immune cells in tumor samples. Based on the signature genes of the 24 types immunocyte in the literature [26], the relative enrichment score of every immunocyte was quanti ed from gene expression pro le for each tumor sample. The correlation between SRD5A3 and these immune cells was analyzed by Spearman correlation, and the in ltration of immune cells between the high and low-expression groups of SRD5A3 was analyzed by the Wilcoxon rank sum test.
Statistical analysis R (v.3.8.0) software was used to perform the statistical analysis. Because of normal distribution, the differences in the SRD5A3 expression levels between normal tissues and tumors in HCC cohorts were performed through the Wilcoxon signed-rank test. The Kolmogorov-Smirnov test and logistic regression were used to analyze the relationship between SRD5A3 expression and clinical pathologic characters. In this study, we compared SRD5A3 expression in patients with HCC tissues and adjacent normal tissues, and evaluated the discrimination ability of SRD5A3 in HCC by ROC analysis using pROC package. P-value < 0.05 was set up as the cut-off criterion.

Pan-cancer and HCC SRD5A3 expression analysis
We rst assessed SRD5A3 expression in pan-cancer data from TCGA and GTEx. The data revealed that  (Fig.1a). We further compared the expression of SRD5A3 in GTEX combined with TCGA normal samples and TCGA hepatocellular carcinoma samples, 50 paracancerous and 371 HCC samples from TCGA, 50 HCC samples and their corresponding paired paracancerous samples from TCGA. Finally, it was found that SRD5A3 was signi cantly overexpressed in HCC samples, and the results were statistically signi cant (P 0.001). (Fig.1b-d). In addition, the ROC curve with an area under the receiver operating characteristic curve value of 0.849 (95% CI 0.804 to 0.895, p< 0.001), which showed high diagnostic accuracy of SRD5A3 in HCC., as is shown in (Fig.1e).

Differentially expressed genes (DEGs) analysis
We performed a differentially expressed genes (DEGs) analysis using TCGA cohort data, patients with HCC were divided into the high expression group and the low expression group according to the expression level of SRD5A3. A total of 242 differentially expressed genes were found using screening, 184 were highly expressed and 58 were lowly expressed, as is shown in Fig.2a. We constructed a gene expression heatmap to get an overview for the top 10 genes with the greatest differences in expression in HCC (Fig.2c).
Association with SRD5A3 expression and clinicopathologic variables As shown in Table 1, 371 primary tumors with both clinical and gene expression data were downloaded from TCGA data. There were 250 male and 121 female with median age 61 years in the cohort, of the 371 specimens. Correlation analysis showed that SRD5A3 protein expression was signi cantly associated with race (p = 0.001), T stage (p < 0.001), pathologic stage (p < 0.001), tumor status (p = 0.008), and residual tumor (p = 0.017). No correlation was found between SRD5A3 expression and other clinicopathologic features, detail information was showed in table 1. Univariate analyses using logistic regression revealed that up-regulation of SRD5A3 in HCC is signi cantly associated with race (p < 0.001), BMI (p = 0.049), pathologic stage (p < 0.001), tumor status (p = 0.006) and residual tumor (p = 0.016).

As shown in
To con rm the correlation between the SRD5A3 expression and the prognosis of HCC, survival rates between the high and low SRD5A3 level groups were compared. The Kaplan-Meier survival analysis found that HCC patients in the high SRD5A3 expression group had poorer overall survival (OS, HR=2.26(1.58-3.24), p<0.001) (Fig. 4) Table 3.
Based on Cox proportional risk regression model, tumor status, TP53 status, T stage and the expression of SRD5A3 were included in the nomogram (Fig.5a). The C-index of the prognostic model was 0.692(95%CI:0.666-0.719). We constructed calibration plots evaluating the agreement between the predicted and the actual OS for the prognosis model, and the results showed that the predicted results of the nomogram were reliable (Fig. 5b). Gene sets enriched in SRD5A3 expression phenotype and protein-protein interaction The functions of SRD5A3 were predicted by analyzing GO and KEGG in Metascape. The top 20 GO enrichment items were classi ed into three functional groups: biological process group (11 items), cellular component group (3 items), and molecular function group (6 items) (Fig. 6a, b). Regarding to the KEGG pathway analysis as shown in Fig.6c. Mineral absorption, gastric cancer, Salivary secretion, Protein digestion and absorption, parathyroid hormone synthesis, secretion and action were the signi cant metabolic pathways.
In addition, to better understand the potential biological function of SRD5A3 in HCC, we performed a Metascape protein-protein interaction enrichment analysis. The protein-protein interaction network and MCODE components identi ed in the gene lists are shown in Fig. 6d Fig. 6e. These hub genes mainly belong to the keratin and MUC family.

Relationship between SRD5A3 expression and tumorin ltrating immune cells
Analysis and characterisation of proteins and genes involved in cancer development at the molecular level, could add to our knowledge of potential prognostic factors [11]. Numerous studies had now documented a link between the immune in ltrate in several human carcinoma types and prognosis and response to therapy [12]. Therefore, we tried to nd whether SRD5A3 expression was associated with immune in ltration in HCC. Spearman correlation was employed to show the association between the expression level (TPM) of SRD5A3 and immune cell in ltration level quanti ed by GSEA in the HCC tumor microenvironment. Additionally, our ndings strongly underline the signi cant role of SRD5A3 in immune in ltration. SRD5A3 was negatively correlated Th17(R = -0.238, p < 0.001) (Fig. 8a), Cytotoxic cells (R = -0.234, p < 0.001) (Fig. 8b), whereas it was positively correlated with Th2 cells (R = 0.258, p < 0.001) (Fig.  8c).

Discussion
SRD5A3 is a membrane protein, which is mainly located in cellular microsomes and nuclei, and belongs to the steroid 5α reductase family. As a microsomal enzyme, it participates in the reduction reaction of steroid compounds by reducing hydrogen provided by NADPH. Studies have shown that it can promote the occurrence of prostatic hyperplasia and prostate cancer by regulating androgen metabolism. Besides, SRD5A3 reduces polyprenols to polyterpenoids and participates in the synthesis of N-glycosylated oligosaccharide precursors, leading to congenital glycosylation disorders in patients [13,14]. What's more, SRD5A3 is involved in the occurrence of various malignant biological behaviors of human tumors, and is a key gene in the development of tumors. The role of SRD5A3 in a variety of malignancies has been widely reported [15]. However, there are few reports on the role of SRD5A3 in liver cancer.
In this study, bioinformatics analysis using high throughput RNA-sequencing data from TCGA demonstrated that SRD5A3 was signi cantly highly expressed in HCC tissues compared with paired normal tissues, indicating that SRD5A3 played a role in tumorigenesis and progression. In addition, the ROC analysis showed that the AUC was 0.849 in the diagnosis of HCC, suggesting that SRD5A3 might be a potential diagnostic biomarker. On this premise, the connection between the statement of SRD5A3 and clinicopathological boundaries was additionally examined, and it was discovered that high articulation levels of SRD5A3 protein expression was signi cantly associated with T stage (p < 0.001), pathologic stage (p < 0.001). tumor status (p = 0.008), race (p = 0.001) and residual tumor (p = 0.017). Despite numerous reports on the role of the SRD5A3 in the development of HCC, the prognostic value of SRD5A3 has rarely been investigated. We constructed a prognostic gene signature model based on SRD5A3 Kaplan-Meier curve indicated a good performance of this model for HCC survival prediction. We found that HCC patients in the high SRD5A3 expression group had poorer OS, PFI and DSS. Multivariate analysis showed that SRD5A3 was an independent factor affecting survival of HCC patients (P<0.05), which proves that it may become a biomarker of HCC.
A signi cant conclusion in our study concerns the SRD5A3 expression correlation with immune in ltration levels in HCC. Numerous studies have demonstrated that the hepatocellular carcinoma microenvironment contributes to the immunological changes during the progression of HCC. This nding suggested that SRD5A3 might play an important role in the immune system. Thus, the difference in immune cell in ltration between patients with high and low SRD5A3 expression was compared in this study. SRD5A3 was negatively Th17 (R = -0.238, p < 0.001), Cytotoxic cells (R = -0.234, p < 0.001), whereas it was positively correlated with Th2 cells (R = 0.258, p < 0.001). Tumor microenvironment, as an important structural function that mainly supports and promotes the occurrence, development, recurrence and metastasis of tumors, has been paid attention and recognized by many scholars in recent years [16,17].In the context of precision medicine, a comprehensive analysis of tumor in ltrating cells, cytokines/chemokines, gene expression and other aspects to develop personalized precision immunotherapy for liver cancer patients is of great signi cance for the effective evaluation and prediction of the e cacy of immunotherapy and the adoption of combined therapy strategy.
Another important aspect of this study revealed the potential value of SRD5A3 in the progression of HCC cell cycle. To further explore the key pathways in HCC, we performed GSEA using TCGA data, GSEA showed that cell cycle mitotic, cell cycle checkpoints, mitotic nuclear division, o glycan processing, mitotic rometaphase protein o linked glycosylation were differentially enriched in SRD5A3 high expression phenotype. Tumor was considered as a cell cycle disease. Researches reveal that abnormal cell cycle, cell apoptosis, and DNA repair play an important role in drug resistance of tumors. For example, abnormal cell cycle signal transducation pathway can restore effective circulation disorder of drugresistant tumor cells and return to mitosis, leading to resistance to chemotherapy drugs. With long-term use, chemotherapeutic drugs, such as sorafenib, have additional issues such as toxicity and/or drug ine cacy [18].Chemotherapy resistance has become one of the major obstacles to cancer remission.
In recent years, it has become a hot topic to study the mechanism of chemotherapeutic drug resistance from the perspective of genetics to guide clinical treatment practice. Recent studies have shown that small stress molecules such as arginine and anion small stress molecules acetylcholine chloride have the potential to regulate p53 mutation polymerization [19]. Robert J. Schleifer et al. suggested that decreased KLHL5 expression increased sensitivity to anticancer compounds, especially Akt/PI3K/mTOR inhibitors, which implies that inhibiting or regulating KLHL5 could be a potential method of sensitizing cancer cells to speci c drugs [20]. The therapeutic targeting of cyclin-dependent kinase inhibitors (CKI) with anticancer drugs or related genes increases chemotherapy resistance by reducing cyclin E levels or activity in tumor cells [21]. We tried to predict the potential connection between SRD5A3 and the cell cycle, and further provided a new idea for the exploration of chemotherapy drug resistance and the discovery of new biomarker therapeutic target. At the same time, the identi cation of molecular markers and the classi cation of HCC patients with high heterogeneity to determine whether they are suitable for speci c (targeted or untargeted) therapy will play an important role in guiding the individualized precise treatment of HCC.
There were some limitations to our study that need to be recognized. First, this is a preliminary study designed to investigate the role played by SRD5A3 in HCC. Further experimental veri cations are necessary to elucidate the biological functions of these predicted molecular mechanisms in HCC to deepen our understanding of the direct impact of SRD5A3 on HCC. Besides, the sample size in our study was relatively small, and larger sample sizes are needed to increase the reliability of our ndings. Last, the clinical information from the TCGA database was not comprehensive, and more clinical data concerning tumor progression and prognosis should be included to better evaluate the relationship between the SRD5A3 and HCC.

Conclusion
In this study, we conducted a comprehensive bioinformatics analysis and found that high expression of SRD5A3 was signi cantly associated with poor prognosis in HCC patients. It may be a potential biomarker for HCC. The important role of SRD5A3 in tumor immune microenvironment and chemotherapeutic drug resistance was further explored through immune in ltration analysis and GSEA enrichment analysis, so as to provide research ideas for precise and individualized treatment of liver cancer. However, the limitations of this study lie in the small sample size and lack of experimental veri cation. Therefore, more population-based studies with larger sample sizes and further functional studies are still necessary to con rm our ndings.    Survival analysis of SRDA3 in HCC. Overall survival curve of patients with high versus low SRD5A3 expression levels.