Comprehensive role of prostate‐specific antigen identified with proteomic analysis in prostate cancer

Abstract Current treatments including androgen deprivation fail to prevent prostate cancer (PrCa) from progressing to castration‐resistant PrCa (CRPC). Accumulating evidence highlights the relevance of prostate‐specific antigen (PSA) in the development and progression of PrCa. The underlying mechanism whereby PSA functions in PrCa, however, has yet been elucidated. We demonstrated that PSA knockdown attenuated tumorigenesis and metastasis of PrCa C4‐2 cells in vitro and in vivo, whereas promoted the apoptosis in vitro. To illuminate the comprehensive role of PSA in PrCa, we performed an isobaric tag for relative and absolute quantitation (iTRAQ)‐based proteomic analysis to explore the proteomic change induced by PSA knockdown. Among 121 differentially expressed proteins, 67 proteins were up‐regulated, while 54 proteins down‐regulated. Bioinformatics analysis was used to explore the mechanism through which PSA exerts influence on PrCa. Protein‐protein interaction analysis showed that PSA may mediate POTEF, EPHA3, RAD51C, HPGD and MCM4 to promote the initiation and progression of PrCa. We confirmed that PSA knockdown induced the up‐regulation of MCM4 and RAD51C, while it down‐regulated POTEF and EPHA3; meanwhile, MCM4 was higher in PrCa para‐cancerous tissue than in cancerous tissue, suggesting that PSA may facilitate the tumorigenesis by mediating MCM4. Our findings suggest that PSA plays a comprehensive role in the development and progression of PrCa.

screening and a decreased risk of PrCa mortality, 4 which implies a crucial role of PSA in promoting the development and progression of PrCa. 3 Accumulating evidence highlights the role of PSA in the initiation and progression of PrCa. Free PSA has been found to be linked with expression variations of tumour-promoter genes and tumour-suppressor genes in PrCa cell lines. 5 Veveris-Lowe and his coworkers showed that PSA and hk4 co-functioned in the progression of PrCa via inducing the epithelial-mesenchymal transition (EMT). 6 In addition, the enzymatic activity of PSA correlates with the tumour growth of PrCa. 3 Saxena et al 7  In the present study, we investigated the role of PSA in PrCa using RNA interference in PrCa C4-2 cells and explored the proteome-wide changes induced by PSA knockdown using isobaric tag for relative and absolute quantitation (iTRAQ)-based LC-MS/MS analysis. Our findings provide an opportunity to advance our knowledge of the mechanistic role of PSA in PrCa, and provide novel evidence for targeting PSA-relevant signalling pathway to treat PrCa. and 1% penicillin-streptomycin (Gibco, USA) at 37°C in a humidified incubator with 5% CO 2 .

| Generation of PSA-specific shRNA-expressing C4-2 cells and selection of stably infected monoclonal colonies derived from C4-2
A shRNA oligonucleotides sequence targeting a region of the PSA cDNA was designed by MISSION ™ TRC-Hs 1.0 library (Sigma-Aldridge, St. Louis, MO). 3 A shRNA sequence non-specific to the region of PSA cDNA was designed by Genechem Corporation (Shanghai, China) as a control. The shRNA sequence was as follows: 5′-GCCTGGAGACATATCACTCAA-3′; the non-specific negative control sequence: 5′-TTCTCCGAACGTGTCACGT-3′. The shRNA lentivirus expression vector comprising the specific or non-specific sequence was designed and subsequently packaged in lentiviral particles which were isolated from HEK293T cells (Genechem Corporation, Shanghai, China). Then, the lentiviral particles were utilized to infect C4-2 cells according to the vendor's recommendations. Briefly, 1000 of C4-2 cells were planted in a well of 96-well plate in 200 μL of complete media. After 24 hours, the media was replaced with fresh media. Lentiviral particles and polybrene in certain volumes were added to the cells at a final multiplicity of infection (MOI) of 100. After 12 hours of infection, the media containing particles were replaced with fresh media. After culturing for 72 hours, the cells were digested with trypsin and washed twice with PBS for subsequent suspension in fresh media. The cell concentration was counted, and then, 100 cells were transferred to 10 mL RPMI 1640 media supplemented with 20% FBS in a 100 mm dish (Corning) for further maintaining in the humidified incubator. After about 20 days, when monoclonal colonies derived from C4-2 cells emerged and were visible by naked eyes, monoclonal colonies expressing GRP among them were transferred to wells of 24-well plate, respectively, with media supplemented with 20% FBS. When the confluence reached to 90%, those monoclonal colonies were transferred to wells on 6-well plates filled with fresh compete media for subsequent characterization.

| Western blot assay
Cells were digested with trypsin and rinsed twice with PBS and then suspended in radio immunoprecipitation assay buffer (RIPA buffer) (Beyotime Biotechnology, Shanghai, China) mixed with protease inhibitors phenylmethanesulphonyl fluoride (PMSF) (Beyotime Biotechnology) and another protease inhibitor cocktail (Servicebio Corporation, Wuhan, China) for at least 30 minutes on ice. After lysing, cell lysates were further crushed using Ultrasonic Cell Crusher prior to centrifugation in a microcentrifuge at 12 000g for 10 minutes at 4°C and collection of supernatant. A protein quantification assay was performed using bicinchoninic acid (BCA) protein assay kit (Beyotime Biotechnology) to determine the protein concentration.  and 72 hour, 10 μL WST reagents were added per well, and the cells were incubated for 2 hours. Then, the absorbance was scaled at 450 nm using a multilabel counter (PerkinElmer, Singapore).

| Cell migration and invasion assays
Migration assay was performed using 24-well Transwell plates with

| Apoptosis assay
Apoptosis assay was performed using Phycoerythrin (PE) Annexin V Apoptosis Detection Kit I (Cat. No. 559763, BD Pharmingen ™ , USA) following the recommended procedure provided by the manufacturer. Briefly, cells were seeded in 6-well plates first and once attachment, the medium was replaced with FBS-free medium.
After starvation for 24 hours, the cells were digested with EDTAfree trypsin and neutralized by complete medium. All cells in supernate and in attachment were collected via centrifugation in a microcentrifuge at 2500 g for 5 minutes. Cell pellets were rinsed twice with cold PBS and resuspended in 1X Binding Buffer. Then, 100 μL of cells at a concentration of 1 × 10 6 cells/mL were transferred to a 5-mL culture tube. Five microlitres of PE Annexin V and five microlitres of 7-aminoactinomycin D (7-AAD) were added into the tube, and then, the cells were gently vortexed and incubated for 15 minutes at 25°C in the dark. Finally, the tube was mixed with 400 μL of 1X Binding Buffer and the cells were analysed by a flow cytometry (BD FACSCalibur ™ , USA).

| In vivo tumorigenesis and metastasis assays
To evaluate the effect of decreased PSA expression on tumorigenesis and metastasis in murine models, in vivo tumorigenesis and metastasis assays were performed as described previously. 3

| Sample preparation for iTRAQ labelling
A total of 2 × 10 7 C4-2-derived shNC cells or one monoclonal colony (in which PSA expression was most extremely inhibited, n = 3) was collected in a tube. The samples were resuspended and thoroughly solubilized by lysis buffer and then sonicated in ice. The samples were bathed in boiling water for 15 minutes, followed by centrifugation at 14 000g for 15 minutes. Debris was discarded, and protein concentration was measured using BCA assay.
A total of 30 μL of protein solution of each sample was mixed with dithiothreitol (DTT) at a final concentration of 100 mΜ, bathed in boiling water for 5 minutes and then chilled at room temperature.
After addition of 200 μL UA buffer (8Μ urea, 150 mΜ Tris-HCl, pH 8.5), the proteins were enriched using 30-kDa centrifugal filter (Sartorius, Germany) at 14 000g for 15 minutes, which was repeated twice. Addition of 100 μL iodoacetamide (IAA) buffer (100 mΜ IAA in UA) was performed for protein alkylation prior to a vortex for one minute. After being incubated in dark for 30 minutes, the samples were centrifuged at 14 000g for 15 minutes. Addition of 100 μL UA buffer was followed by centrifugation at 14 000g for 15 minutes, and this process was repeated twice. After addition of 100 μL diluted dissolution buffer (AB SCIEX, USA) for ten times, the samples were centrifuged at 14 000g for 15 minutes and repeated twice. The samples were added 40 μL trypsin buffer (4 μg trypsin in 40 μL dissolution buffer) and then vortexed for 1 minute. After standing for 16 hours, the samples were centrifuged at 14 000g for 15 minutes for filtrate. After addition of 40 μL of dissolution buffer with ten times of dilution, the samples were centrifuged at 14 000g for 15 minutes.

| iTRAQ Labelling and high pH RP fractionation
The peptides of 100 μg taken from each sample were labelled using iTRAQ reagents (AB SCIEX, USA) following the manufacturer's protocol. The labelled peptides were mixed, and then, chromatography

| Bioinformatics analysis
To visualize the overall proteome-wide expression differences between shNC cells and PSA-silenced cells, hierarchical cluster analysis was performed using Multiple Experiment Viewer (MeV) software, which could generate a heatmap.

Gene Ontology (GO) functional annotation and Kyoto
Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of 121 differentially expressed proteins were performed using Cytoscape with the ClueGO 9 and CluePedia 10 plugins based on GO and KEGG databases. To discover the biological protein-protein associations, 121 proteins were subjected to formation of interaction networks using Search Tool for the Retrieval of Interacting Genes (STRING, https://strin g-db.org/) database, 11 and afterwards, the generated STRING protein-protein interaction (PPI) network was imported into Cytoscape with the cytoHubba 12 plugin containing one topological analysis method named Maximal Clique Centrality (MCC) to predict essential proteins from the existing STRING PPI network.

| Immunohistochemistry
Expression level of MCM4 in tissue specimens was measured by IHC following the IHC protocol developed by R&D Systems (https:// www.rndsy stems.com) Briefly, tissue sections were deparaffinized with xylene and alcohol and then rehydrated by buffer water.
Endogenous peroxidase activity was quenched using peroxidase blocking reagent and retrieval solution was used for retrieving antigen. Primary antibody applied was MCM4 Rabbit monoclonal antibody (1:100 dilution, Cat. No. 12973, CST, US). Staining of tissue was visualized under a microscope.

| Statistical analysis
All experiments were performed in three biological replicates at least. Data were showed as mean ± standard deviation (SD). Oneway ANOVA or Student's t test was applied for the PSA-induced phenotype experiments in vitro and in vivo using SPSS 23.0 (IBM, USA). Wilcoxon rank test was used in the statistical analysis of clinicopathological samples. Statistical analyses of bioinformatics were based on built-in methods in databases. The level of statistical significance was defined as P < .05.

| Verification of PSA knockdown efficiency in C4-2-derived monoclonal colonies
By infecting C4-2 cells with a packaged lentivirus expressing PSAtargeting shRNA, we acquired nine monoclonal colonies. The

| Exogenous PSA promotes the proliferation of C4-2 cells in vitro
We further evaluated the effects of exogenous PSA on C4-2 cell proliferation using recombinant PSA protein (AbD Serotec, USA). 13 As shown in Figure 2B, cells in groups under treatment with 500 ng/ mL of PSA and 1000 ng/mL of PSA were accelerated in growth at 72 hours and 96 hours, compared with cells in control group. The data suggest that exogenous PSA is sufficient to promote PrCa cell proliferation.

| Loss of PSA expression attenuates migration and invasion of C4-2 cells in vitro
A previous study reported that transfection with PSA promoted migration rather than invasion of PrCa PC-3 cells. 6  The results indicate that PSA is required for the migration and invasion of C4-2 cells. Monoclone9 ((6.55 ± 2.09)%) (P = .004) was significantly higher, suggesting that PSA knockdown promotes cell apoptosis.

| Loss of PSA expression slowed down tumorigenesis and inhibited the metastasis of C4-2 cells in vivo
To evaluate the impact of reduced PSA on PrCa in vivo, we established murine models mirroring tumorigenesis and metastasis of C4-2 cells in vivo. As shown in Figure 3, Figures S1 and S2, Monoclone9 groups showed minor tumour volumes (493.16 ± 385.51 mm 3 ; P < .001) and smaller metastatic tumour area (P < .001) than shNC groups respectively, suggesting that loss of PSA expression delayed growth of C4-2 cells in vivo and distant spreading to mice lung.

| iTRAQ-based LC-MS/MS and hierarchical cluster analysis
To understand the proteomic change induced by PSA alteration, three samples of C4-2-derived shNC (control group) or Monoclone9 (test group) cells were subjected to iTRAQ-based LC-MS/MS analysis. Totally, 6822 proteins (Table S1) were Three replicates in either control group or test group were clustered first, indicating biological replicate samples were homologous.

| GO and KEGG enrichment analysis
The 121 proteins differentially expressed were subjected to annotation and functional enrichment using ClueGO 9 and CluePedia 10 plugins within Cytoscape based on GO and KEGG databases.
As shown in Figure 5A and Table 1, the proteins were predominantly enriched into GO BP terms including DNA conformation change, depolymerization, secretion, maintenance or regulation of location. The enriched GO MF terms included DNA-related enzyme activity, phospholipase inhibitor activity, molecule binding and transporter activity ( Figure 5B and Table 2). Figure 5C and Table 3 indicated the enriched KEGG pathways with relevant proteins. There were four KEGG pathways being substantially enriched: glycolysis/gluconeogenesis, valine, leucine and isoleucine degradation, butanoate metabolism and mineral absorption, suggesting a potential regulation of genome stability and metabolism by PSA.

| Protein-protein interaction (PPI) network by STRING and cytoHubba analyses
To explore the association of 121 differentially expressed proteins and further discover potential candidate proteins that may function in the biological process of PrCa, STRING database and cytoHubba were sequentially utilized in data mining. The PPI network was firstly generated as shown in Figure 6A, one node denoting one protein and one edge denoting one potential correlation between every two proteins. After importing the STRING PPI network into Cytoscape, top 20 nodes were further selected and ranked by MCC in cyto-Hubba, which were mapped into a subnetwork ( Figure 6B). These 20 proteins were considered as the most important part of the STRING network according to MCC topological analysis. Among them, 6 proteins were down-regulated, and the other 14 proteins were upregulated. Five proteins were identified.

F I G U R E 3
Effects of decreased PSA on tumorigenesis and metastasis of C4-2 cells in vivo. A, Inhibited PSA expression obviously slowed down tumorigenesis (P < .001). n = 7. Scale bar = 1 cm. B, Interfered PSA shrank metastatic tumour area in mice lungs (P < .001), with treatment of HE staining. n = 7. Scale bar = 200 μm. ** shows P < .01

| D ISCUSS I ON
The failure of current therapeutic approaches to prevent PrCa from recurrence or evolving into a castration-resistant state with a high mortality rate promotes various research programmes exploring the pathological mechanism of the initiation and development of PrCa.  The hierarchical cluster analysis indicated that the quantified data were replicable and reduced PSA induced a wide change of proteins. Combination with the PPI subnetwork with GO/KEGG analysis further contributes to a better insight into changes induced by PSA. Using these methods, we chose five candidate proteins which may be under regulation by PSA, including POTEF, EPHA3, RAD51C, MCM4 and HPGD. POTEF, a membrane glycoprotein, was speculated that the enhanced galactosylation of it may contribute to the early step of TNBC metastasis. 15 In combination with the alteration of metastasis in PSA-silenced cells, it is speculated that PSA may regulate the expression or modification of TNBC. As a receptor tyrosine kinase and frequently overexpressed on the tumour-initiating cell population in glioma, EPHA3 modulates mitogen-activated protein kinase signalling to maintain glioblastoma cells in a less differentiated state, and once it is knocked down, tumorigenic potential of tumour cells is weaken. 16 Charmsaz et al 17 discovered that anti-EPHA3 monoclonal antibody (mAb) could attenuate the growth of EPHA3-expressing leukaemic xenografts. These results are consistent with the finding in the present study. RAD51C is essential for DNA repair by homologous recombination. 18 RAD51C deficiency accumulates somatic mutation, which leads to genomic instability, and finally promotes cancer. 19 In addition, loss of RAD51C accelerates tumour progression lacking of Trp53. 20 These studies suggest that up-regulated RAD51C by knockdown of PSA may attenuate CRPC progression. MCM4 is a component of the minichromosome maintenance complex (MCM2-7 helicase complex) essential for normal DNA replication and genome stability 21 which is involved in initiating cancer, 22 and is required for terminal NK cell maturation, 23 while NK cells shows antitumoural activity in the mouse model. 24   In addition of the down-regulation of these five proteins, the present study also found some down-regulated proteins.
HIST3H2BB is a member of the histone H2B family. S14 of histone H2B can be bound and phosphorylated by HIPK2, while loss of H2B-S14 P induced by HIPK2 depletion can prevent cell cleavage and result in tetra-and polyploidization. Cytokinesis defects and cell proliferation could be rescued by restoration of wild-type HIPK2 activity or expression of a phosphomimetic H2B-S14D derivative, suggesting H2B-S14 P is essential for a faithful cytokinesis. 26 The present study found that PSA silencing decreased HIST3H2BB expression and proliferation of C4-2 cells suggesting that PSA may exert an influence on proliferation of PrCa via modulating cytokinesis.
Phosphofructokinase, platelet (PFKP), a major isoform of cancer-specific phosphofructokinase-1 (PFK-1) enriched in glycolysis/ Hydroxymethylglutaryl-CoA synthase-1 (HMGCS1), an enzyme involved in lipid metabolic process, 30 is classified into two KEGG pathways: valine, leucine and isoleucine degradation, and butanoate metabolism in KEGG pathway. HMGCS1 was reported to be over twofold higher in a non-small-cell lung cancer Abraxane-resistant cells than Abraxane-sensitive, which indicated involvement of HMGCS1 in drug resistance. 30 In the present study, HMGCS1 was decreased after reduction of PSA, which implies that PSA may participate in the resistance to androgen deprivation therapy.

| CON CLUS IONS
We have confirmed that PSA exerts considerable influence on tumorigenesis and metastasis of PrCa via mediating proliferation, migration, invasion and apoptosis of PrCa cells. Besides, an iTRAQbased LC-MS/MS was performed to detect the comprehensive molecular mechanism of PSA elucidating its role in tumorigenesis and metastasis of PrCa. In detail, PSA would modulate EMT, cell cleavage, glucose metabolism or lipid metabolism, especially genomic instability via mediating POTEF, EPHA3, HPGD, RAD51C, and especially MCM4. Further research should be performed to explore the mechanism of PSA involved in tumorigenesis and metastasis of PrCa, supporting the therapy target of PSA.

ACK N OWLED G EM ENT
We thank Jianping Li and Junmou Hong for their brilliant technical assistance, and Zhihua Wang for his contribution to this article.

CO N FLI C T S O F I NTE R E S T
There were no potential conflicts of interest to be disclosed.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used to support the findings of this study are available from the corresponding author upon request.