Steroid 5 alpha-reductase 3 (SRD5A3) promotes tumor growth and predicts poor survival of human hepatocellular carcinoma (HCC)

SRD5A3 is highly expressed in various tumor tissues including HCC

We first employed the UALCAN web portal to explore the general expression of SRD5A3 in different cancers (Figure 1). Various cancers including HCC showed moderate to strong expression of SRD5A3 compared to their corresponding normal tissues. In order to verify whether this high expression of SRD5A3 in HCC tumor tissues is a common phenomenon, seven HCC datasets including GSE22058, GSE25097, GSE36376, GSE63898, GSE64041, TCGA(LIHC), and GSE54236 were analyzed (Figure 2). Bioinformatics analysis of these datasets revealed that the expression of SRD5A3 in HCC tissues was generally higher than that in non-HCC liver tissues (Figure 2A–2F) and corresponding adjacent liver tissues (Figure 2G–2I). Additionally, we performed quantitative reverse-transcription polymerase chain reaction (qRT-PCR) and Western blot analyses to verify this speculation again. Our data displayed that the mRNA and protein expression of SRD5A3 were significantly increased in the tumor tissues of 36 HCC patients compared with the matched adjacent non-HCC liver tissues (Figure 3A–3C), which was consistent with the results in public datasets. Immunohistochemistry (IHC) staining showed that SRD5A3 protein was predominantly located in the cytoplasm of HCC and liver cells (Figure 3D).

Figure 1. The expression of SRD5A3 in various carcinomas. Transcriptional expression of SRD5A3 in 24 types of cancer (the UALCAN web portal).

Figure 2. SRD5A3 is generally highly expressed in HCC tumor tissues compared to non-HCC liver tissues. SRD5A3 expression in HCC tumor tissues and non-HCC liver tissues were shown for the (A) GSE22058, (B) GSE25097, (C) GSE36376, (D) GSE63898, (E) GSE64041 and (F) TCGA(LIHC) datasets (unpaired t-test). SRD5A3 expression in HCC tumor tissues and adjacent non-HCC liver tissues (peri-HCC) were shown for the (G) GSE54236, (H) GSE64041 and (I) TCGA(LIHC) datasets ((unpaired t-test)). SRD5A3 mRNA level was shown as log2 copy number units. LIHC, liver hepatocellular carcinoma.

Figure 3. qRT-PCR, Western blot and IHC measurements of SRD5A3 in HCC and peri-HCC tissues. (A) The mRNA expression of SRD5A3 in 36 pairs of HCC and peri-HCC tissues was measured by qRT-PCR analysis. The SRD5A3 mRNA levels of peri-HCC tissues were normalized as ‘1’. (B, C) The protein level of SRD5A3 in 7 pairs of HCC and peri-HCC tissues was analyzed by Western blot. (D) The location of SRD5A3 protein was detected by IHC staining. T, HCC tumor tissue; N, peri-HCC tumor tissue.

Up-regulation of SRD5A3 is associated with the tumor progression of HCC

Next, the association between SRD5A3 expression and corresponding clinicopathological characteristics was evaluated. We found that the expression of SRD5A3 had a significant correlation with the clinical stage of HCC (P = 0.0222, F = 3.247, Figure 4A). Moreover, as vascular invasion intensified, the average expression of SRD5A3 seemed to increase; however, no significant statistical difference was detected among them (Figure 4B). Interestingly, we found that the expression of SRD5A3 was also associated with the patient's gender (Figure 4C). The average expression of SRD5A3 in female HCC patients tended to be higher than that of male HCC patients (Figure 4C).

Figure 4. The expression of SRD5A3 is correlated with the clinical stage and the gender of HCC patients. In the TCGA(LIHC) dataset, the correlation between (A) clinical stage, (B) vascular invasion, (C) gender, and SRD5A3 expression was analyzed in HCC patients with clinical data. LIHC, liver hepatocellular carcinoma (HCC).

Up-regulation of SRD5A3 is associated with the prognosis of cancers including HCC

To confirm the correlation between the SRD5A3 expression and the prognosis of HCCs, survival rates between the high and low SRD5A3 level groups were compared. We performed these analyses on the Kaplan–Meier plotter survival analysis platform, and found that HCC patients in the high SRD5A3 expression group had poorer overall survival (OS, HR = 1.91 (1.34 – 2.73), log-rank test P = 0.00029), relapse-free survival (RFS, HR = 1.59 (1.14 – 2.22), log-rank test P = 0.0055), progression-free survival (PFS, HR = 1.61 (1.20 – 2.17), log-rank test P = 0.0015) and disease specific survival (DSS, HR = 2.13 (1.35 – 3.39), log-rank test P = 0.00098) rates than that in the low SRD5A3 expression group (Figure 5A–5D). These were consistent with the data of HCC in the Human Protein Atlas database (Figure 5E), which again indicated that up-regulation of SRD5A3 expression predicts poor prognosis of HCC.

Figure 5. The association of the SRD5A3 expression and the prognosis of HCC. (A) Overall survival (OS), (B) relapse-free survival (RFS), (C) progression-free survival (PFS) and (D) disease specific survival (DSS) curve based on high and low expression levels of SRD5A3 from the Kaplan–Meier plotter survival analysis platform. (E) Prognostic significance of SRD5A3 expression in HCC in the Human Protein Atlas database. HR, hazard ratio.

Moreover, we also explored the prognostic significance of SRD5A3 in other cancers, and found that the high expression of SRD5A3 was associated with the poor OS rates of skin cutaneous melanoma (SKCM, P = 0.00034), kidney renal clear cell carcinoma (KIRC, P = 0.023), kidney renal papillary cell carcinoma (KIRP, P = 0.018), uveal melanoma (UVM, P = 0.0071), brain lower grade glioma (LGG, P = 0.037), breast invasive carcinoma (BRCA, P = 0.041), thyroid carcinoma (THCA, P = 0.036), and glioblastoma multiforme (GBM, P = 0.048) (Figure 6A–6H). These results suggested that high expression of SRD5A3 may be a common unfavorable prognostic factor for many different types of cancer.

Figure 6. The expression of SRD5A3 has prognostic value in various malignant cancers. In the UALCAN website portal, the effect of high and low expression levels of SRD5A3 on the overall survival of (A) skin cutaneous melanoma (SKCM), (B) kidney renal clear cell carcinoma (KIRC), (C) kidney renal papillary cell carcinoma (KIRP), (D) uveal melanoma (UVM), (E) brain lower grade glioma (LGG), (F) breast invasive carcinoma (BRCA), (G) thyroid carcinoma (THCA) and (H) glioblastoma multiforme (GBM) cancer patients.

Knockdown of SRD5A3 suppresses the growth of HCC in vitro and in vivo

To investigate the potential biological function of SRD5A3 in HCC tumorigenesis, we knocked down the expression level of SRD5A3 in HCC cell lines by infecting them with SRD5A3 shRNA (sh-SRD5A3) lentivirus or control vector (sh-NC). The knockdown efficiency was evaluated by Western blot and qRT-PCR analyses (Figure 7A–7E). Colony formation and cell viability assays were applied to reflect the proliferation inhibitory ability in the HCC cells lacking SRD5A3. As shown in Figure 7F–7J, the colony formation rate and the cell viability of HCC cell lines infected with sh-SRD5A3 were lower than that of cells infected with sh-NC. Furthermore, we inoculated HCC cells infected with sh-SRD5A3 or sh-NC in immunodeficient mice subcutaneously. The tumor growth curve, image and weight suggested that the down-regulation of SRD5A3 inhibited the growth of HCC (Figure 8A–8E).

Figure 7. SRD5A3 inhibition suppresses cell proliferation of HCC in vitro. SMMC-7721 and QGY-7703 cell lines were infected with sh-SRD5A3 or sh-NC lentivirus. The knockdown efficiency of SRD5A3 was confirmed by (A–C) Western blot and (D, E) qRT-PCR analyses. (F–H) Colony formation assays were performed and analyzed. (I, J) CCK-8 assays were conducted to measure cell viability (two-way ANOVA analysis). NC, negative control. *, P value < 0.05.

Figure 8. Down-regulation of SRD5A3 suppresses HCC cell growth in vivo. (A, B) Growth curve of the xenograft tumors (two-way ANOVA analysis). (C, D) Images and weight of xenograft tumors in sh-SRD5A3 and sh-NC groups (unpaired t-test). Three mice per group. NC, negative control. Error bar, mean with standard deviation. *, P value < 0.05.

Molecular mechanisms behind the cellular function of SRD5A3

Finally, the potential molecular mechanisms behind the effects of SRD5A3 in HCC were predicted. The GeneMANIA and STRING databases showed that SRD5A3 might interact with TECRL, SLC25A51, SRD5A1, SRD5A2, TECR, NUDT15, SLC25A25, SLC25A24, ADK, MSH4, SPTSSA, DNM3, C4BPA, CYP27B1, MFSD6, JTB, PON2, TMEM167A, TEFM, POLDIP2, SYP17A1, HSD17B3, DOLK, AKR1C3, SRD5A1 and etc. (Figure 9A, 9B). Moreover, the miRNACancerMAP web portal predicted several more reliable microRNAs upstream of SRD5A3 based on the TCGA(LIHC) dataset. These microRNAs included miR-125b-2-3p, miR-130a-3p, miR-126-5p, miR-450b-5p, miR-186-5p, miR-139-5p and miR-342-3p (Figure 9C). And the UALCAN web portal showed that SRD5A3 was hypomethylated in HCC tissues (Figure 9D), suggesting that the high level of SRD5A3 in HCC might be related to the hypomethylation status. Moreover, the GEPIA2 web portal showed that SRD5A3 was positively correlated with some genes such as PPIAP22 (R=0.56, P<0.05), EIF3K (R=0.56, P<0.05), KIAA1522 (R=0.54, P<0.05), MYL6 (R=0.54, P<0.05), PPIA (R=0.53, P<0.05), MRPS12 (R=0.53, P<0.05), SUMF2 (R=0.53, P<0.05), RRP36 (R=0.53, P<0.05), ORMDL2 (R=0.53, P<0.05) and etc. (Table 1). Some of them were genes related to the cell cycle, including CDK5 (R=0.42, P<0.05), CDK4 (R=0.41, P<0.05), CDK7 (R=0.40, P<0.05), CDKAL1 (R=0.40, P<0.05), CCNB1 (R=0.36, P<0.05), CDK1 (R=0.35, P<0.05) and PLK1 (R=0.32, P<0.05) (Figure 9E–9K and Table 1). Additionally, AR expression was found to have a negative correlation with SRD5A3 (Figure 9L and Table 1).

Figure 9. Potential molecular mechanisms behind the effects of SRD5A3. (A, B) Protein-protein interaction network between SRD5A3 and other protein-coding genes from the GeneMANIA and STRING databases. (C) Upstream microRNA of SRD5A3 from the miRNACancerMAP web-portal. (D) Methylation level of SRD5A3 from the UALCAN web portal. (E–L) Correlation analysis between CDK5, CDK4, CDK7, CDKAL1, CCNB1, CDK1, PLK1, AR and SRD5A3 in the GEPIA2 website portal. R, correlation coefficient.

Bioinformatics analysis

SRD5A3 expression in various cancers and its relationship with tumor progression were analyzed via employing the UALCAN web portal (http://ualcan.path.uab.edu/index.html), the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) and The Cancer Genome Atlas database (TCGA, https://cancergenome.nih.gov/). The impact of SRD5A3 expression on survival prognosis of different cancer patients was predicted by the Kaplan–Meier plotter survival analysis platform (http://kmplot.com/analysis/), Human Protein Atlas database (https://www.proteinatlas.org/) and UALCAN web portal. Methylation level of SRD5A3 was analyzed by UALCAN web portal. Potential interactions between SRD5A3 and other coding genes or microRNAs were predicted by using GeneMANIA (http://genemania.org/search/), STRING (https://string-db.org/cgi/input.pl), and miRNACancerMAP (http://cis.hku.hk/miRNACancerMAP/MAPQuest1.php) and GEPIA2 (http://gepia2.cancer-pku.cn/) website portals.

Clinical specimens

HCC tumor (HCC) and adjacent non-HCC liver tissues (peri-HCC) were obtained from 36 patients who underwent hepatectomy after signing informed consent. The specimens were immediately stored in liquid nitrogen until analysis. The clinical data and pathological characteristics of these HCC patients were summarized in Supplementary Table 1. The application of HCC specimens was approved by the Ethics Committee of Guangdong Provincial People’s Hospital.

RNA isolation and qRT-PCR

Total RNA was extracted from HCC tumor tissues and corresponding adjacent non-HCC liver tissues following the instruction of TRIzol reagent (Cat. no. T9109, TaKaRa, USA). The reverse transcription of mRNA was performed by using PrimeScript RT Master Mix Kit (Cat. no. RR036A, TaKaRa, USA). The cDNA generated above was used as a template to detect the mRNA expression of SRD5A3 with TB Green™ Premix Ex Taq™ detection kit (Cat. no. RR420A, TaKaRa, USA). β-actin gene acts as an internal control. Primer sequences were as follows: SRD5A3 (forward primer: 5′-TTTAATCAGGCCCTGTCTGC-3′, reverse primer: 5′-GGGGTATAGAAATGGAATGGAGA-3′); β-actin (forward primer: AGCGAGCATCCCCCAAAGTT, reverse primer: GGGCACGAAGGCTCATCATT). These primers were synthesized by RuiBiotech (China). The relative expression of SRD5A3 mRNA level was computed according to the 2^-ΔΔCт method.

Western blot analysis

Total protein was extracted from specimens or cell lines using RIPA lysis buffer (Cat. no. #FD008, Fdbio Science, China) supplemented with protease inhibitor cocktail (Cat. no. #FD1001, Fdbio Science, China) and phosphatase inhibitor cocktail (Cat. no. #FD1002, Fdbio Science, China). The protein supernatant was collected after centrifuging at 14000 g and 4° C. Then, the concentration of the protein supernatant was quantified using bicinchoninic acid assay kit (Cat. no. #FD2001, Fdbio Science, China), and boiled at 100° C for 10 minutes. Equal amounts of protein from different samples were separated by polyacrylamide gel electrophoresis and transferred onto 0.45 μm nitrocellulose filter membranes (GE Healthcare, United States). Subsequently, the membranes were probed with primary antibodies against human SRD5A3 (Cat. no. #NBP1-69612, Novus, United States) or β-actin (Cat. no. #4970, Cell Signaling Technology, United States) overnight at 4° C, followed by incubation with a secondary horseradish peroxidase (HRP)-conjugated antibody for 60 minutes at room temperature. Finally, protein blot images were detected using an enhanced chemiluminescence kit (Cat. no. #WBKLS0100, Millipore, United States), and were quantified by using ImageJ software (Ver. 1.46r; MD, United States).

Immunohistochemistry (IHC) staining

HCC tumor and adjacent non-HCC liver tissues embedded in paraffin were sectioned onto 3-μm-thick slices. After retrieving antigen by heating in citrate buffer (pH 6.0) and blocking endogenous peroxidase activity by H2O2, the slices were incubated with the primary antibody against SRD5A3 (Cat. no. #HPA027006, Sigma, United States) at a concentration of 1:50 overnight. Then, the slices were incubated with a secondary HRP-conjugated antibody. Subsequently, the slices were stained with diaminobenzidine (DAB) and counterstained with hematoxylin. Finally, the staining images were captured.

Cell culture

Human HCC cell lines QGY-7703 and SMMC-7721 were obtained from the Guangzhou Jennio Biotech Co.,Ltd. (GD, China). Cells were maintained in high glucose Dulbecco’ s modified Eagle’ s medium (DMEM, Gibco, MA, United States) supplemented with 10 % fetal bovine serum (FBS, Gibco, MA, United States) at 37° C in 5% CO2 condition, and were routinely checked for mycoplasma contamination.

Establishment of HCC cell lines with down-regulated SRD5A3

A lentiviral SRD5A3 interference vector was obtained from Hanbio Biotechnology Co., Ltd. (Shanghai, China). The sequence of SRD5A3 and negative control (NC) shRNA were as follows: AAATAAAGCAGGAGTGGTCATTC (sh-SRD5A3), TTCTCCGAACGTGTCACGT (sh-NC). Lentivirus production, viral infection and qRT-PCR were performed according to the manufacturer’s instructions. Positive and stable infectants were selected by puromycin.

Colony formation

Colony formation experiment was performed as previously described [34, 35]. In brief, 500 cells infected with sh-SRD5A3 or sh-NC were seeded in 6-well plates. After culturing in DMEM medium supplemented with 10% FBS for 7 to 14 days, the cell colonies were fixed with methanol for 2 minutes, then stained with crystal violet for 30 minutes at room temperature. After washing away excess dye, the number of cell colonies was counted by imageJ (Ver. 1.46r; MD, United States) software.

Cell viability measurement

3 × 103 cells were seeded per well in 96-well plate, then the CCK-8 (Dojindo, Japan) assay was performed every 24 hours for 4 consecutive days as previously described to measure cell viability [34, 36].

Tumor xenograft experiments

BALB/c female nude mice (four-week-old) were subcutaneously inoculated with 5×10⁶ HCC cells stably infected with sh-SRD5A3 or sh-NC. Tumor length and tumor width were measured every 3 days for a total of 18 days. The tumor volume was calculated as length × width²/2. The animal experiment was approved by the Animal Health and Ethics Committee. All procedures for this experiment were conducted in accordance with the guidelines for the care and use of laboratory animals.

Statistical analysis

All data were expressed as the mean ± standard deviation. Statistical comparisons were performed using unpaired or paired Student's t test in two-groups, or ANOVA (one-way or two-way) test among more than two groups. The OS, RFS, PFS and DSS survival curves of HCC patients were generated by Kaplan-Meier method, then analyzed via log-rank test. The analyses in this study were conducted through GraphPad Prism 6.0 software (United States). P value < 0.05 was defined as statistically significant.