ETV4 promotes late development of prostatic intraepithelial neoplasia and cell proliferation through direct and p53-mediated downregulation of p21

The rat probasin promoter (PB) was excised from the ARR2PBCAT plasmid [51] and cloned within the previously described TME vector [34] to generate the pPB-ETV4 vector that express the TMPRSS2-ETV4 fusion cDNA under the control of PB promoter (Fig. 1a). A 3.96 kb fragment—containing PB, TMPRSS2-ETV4, and SV40 polyA sequences—was excised from pPB-ETV4 vector (Fig. 1a) and used for pronuclear injections into FVB mouse fertilized eggs that were implanted into pseudo-pregnant females at the LiGeMA Facility of the University of Florence, Italy. Potential founder animals were screened by PCR using primers specific for human ETV4 (hETV4) and the murine β-actin (Table S1), and characterized for construct integration by southern blot analyses (Fig. S1). This animal study was performed in compliance with relevant regulatory standards, and it was approved by the Institutional Animal Care and Use of University of Florence, Italy, and Italian Ministry of Health.

Hek-293T, PC3, (IST “Cell Bank and Cell Factory,” Genoa, Italy), PNT1A, and RWPE (American Type Culture Collection) cell lines were cultured according to cell-bank instructions and transfected as previously described [34].

Murine prostate tissues were fixed in 10% buffered formalin and embedded in paraffin. For histology, sections were stained with hematoxylin and eosin. Pathologic diagnosis was performed following the recommendations of the “Mouse Models of Human Cancer Consortium Prostate Pathology Committee” and the reference classification of mouse prostatic intraepithelial neoplasia (mPIN) in genetically modified animals [52, 53].

For immunohistochemical examination, 4 μm sections were deparaffinized and incubated with the relevant antibodies (Table S2) and visualized using the biotin–streptavidin complex (Thermo Scientific, Rockford, IL, USA) and diaminobenzidine (Dako, Glostrup, Denmark) as chromogen. Slides were then counterstained with hematoxylin. Bromodeoxyuridine (BrdU) incorporation has been evaluated on prostate tissues obtained from 2 transgenic and 2 control mice 12 h after intraperitoneal injection of 100 mg/kg body weight of BrdU and processed as described above. The slides, after antigen retrieval and HCl treatment, were stained with anti BrdU antibody (Abcam, UK).

RNA extraction and qRT-PCR were performed as previously described [34] with SsoFast EvaGreen Supermix and the CFX96 thermocycler (Bio-Rad Hercules, CA, USA). Expression level of each gene was analyzed by 2ΔΔC(T) method using either murine or human glyceraldeide 3-phosphate dehydrogenase as housekeeping gene. The primers are reported in Table S1. Each experiment has been performed at least three times in triplicate.

Proteins from mouse prostate, after tissue disruption with Tissue lyser II (Qiagen, Germantown, MD, USA), and from cell lines were extracted in RIPA buffer. Western blot analyses were performed as previously described [34] by using antibodies listed in Table S2. Horseradish-peroxidase-conjugated secondary antibody signals were detected using ECL (Pierce, Rockford, IL, USA) and Chemidoc XRS plus (Bio-Rad). Densitometric quantifications have been performed by using the software ImageJ (https://​imagej.​nih.​gov/​ij/​) [54].

RWPE cells were transiently transfected with either pCMV-3Tag-3A empty vector (Clontech, Mountain View, CA, USA) or with its derivative vector, pETV4-3Flag, expressing flagged ETV4. Magna ChIP A/G Chromatin Immunoprecipitation Kit (Mercks-Millipore, Billerica, MA, USA) has been used according to the manufacturer’s instructions. In brief, transfected RWPE cells were fixed with 1% formaldehyde and lysed. DNA was sonicated and diluted with ChIP buffer, and one-tenth of lysate was collected as input control. Chromatin was incubated overnight at 4 °C with G/A magnetic beads pre-conjugated with either anti-Flag antibody (Cell Signaling) or a non-specific IgG control. DNA was purified, and measured by quantitative PCR (qPCR) using the primers reported in Table S1. A fragment of the COX2 promoter, a well-known ETV4 regulated gene, was used as positive control. A region of G6PD gene was used as negative control. ChIP assays were performed at least three times.

Dual luciferase reporter experiments were performed using Dual-Glo Luciferase Assay System and the GloMax 20/20 Luminometer (Promega, Madison, WI, USA). Each firefly luciferase pGL4.20 reporter vector (listed below) was used in combination with renilla luciferase pRL-TK reporter vector (Promega) (ratio 10:1) to normalize luciferase activity. The relative luciferase activity (Firefly/Renilla ratio) was measured in (i) RWPE cells transfected with an ETV4-expressing vector (FL-ETV4) [34] normalized to those transfected with an empty vector, and in (ii) PC3 cells transfected with vectors containing shRNA against ETV4 and normalized to those transfected with a scrambled shRNA (siRNA anti-ETV4 and scrambled siRNA—Dharmacon, Lafayette, CO, USA—have been used in some experiments). Each experiment has been performed at least three times in quadruplicate.

A set of luciferase pGL4.20 reporter vectors, in which luciferase is driven by putative ETV4 responsive elements from the human CDKN1A promoter (GenBank Accession #NC_000006.12), were generated as follows. The 845-bp (from 36677905 to 36678749) and the 1656-bp (from 36677084 to 36678739) fragments were amplified using the primers listed in Table S1. These fragments, containing putative ETV4-binding sites, were cloned within the luciferase pGL4.20 reporter vector to obtain the ETV4 responsive luciferase vectors ETV4-BS-A and ETV4-BS-AB, respectively. Derivative vectors were obtained mutating the putative ETV4-binding sites with the QuikChange II site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA): the normal sequence “CCGGAAGC” of the ETV4 BS A (Figs. 5a, c and 6a) was replaced with the mutated sequence “CCGATATC”; the normal sequence “AGAGGAAGAA” of ETV4 BS-B (Figs. 5a and 6a) was replaced with the mutated sequence “AACCGAAGAA.” In addition, also a p53-responsive luciferase reporter vector (Addgene) [55] was used.

All data are expressed as mean ± sem. Student’s t test or one-way ANOVA (followed by Bonferroni correction), as suitable, were performed using GraphPad Prism v.5.0 for Windows (GraphPad Software, La Jolla, CA, USA). Statistical significance was accepted for P ≤ 0.05.

In the fusion transcript TMPRSS2-ETV4, found in prostate cancer patients, a sequence upstream of the TMPRSS2 gene is juxtaposed to the last 9 bp of intron 2 of ETV4 gene [25]. This results in a protein lacking the first 39 amino acids of normal ETV4, as the start codon becomes an ATG in ETV4 exon 4. To assess in vivo the pathogenic role of ETV4 in prostate, we engineered a vector that expresses this ETV4-encoded shortened protein [34] under the control of a modified rat probasin promoter (PB) (Fig. 1a, b) that is able to drive an androgen-inducible prostate-specific expression [56]. By using this PB-ETV4 vector, we have obtained six founders in the FVB strain: three of these were able to transmit the transgene to their progeny and two of them expressed the exogenous hETV4 in the prostate of the derived mouse lines (ETV4-A, ETV4-B) (Fig. 1c, d). Very low levels of hETV4 were detected in the seminal vesicles but not in other tissues. The levels of murine etv4 were not affected by the expression of the hETV4 (data not shown). By southern blot analysis of founders and offspring we determined that in every mouse of these two lines the transgene was present at a single integration site (Fig. S1): this ensures uniform genetic transmission of the transgene.

The expression of the hETV4 transgene was different in the distinct lobes that constitute the mouse prostate with the highest levels in the ventral lobe. Indeed, the increase in total ETV4 expression in our transgenic mice compared with wild-type mice was 18.6 ± 2.4 in the ventral prostate (VP, P < 0.0001); 5.9 ± 1.3 in the dorso-lateral prostate (DLP, P < 0.009) and 7.4 ± 1.9 in the anterior prostate (AP, P < 0.03). ETV4 expression was increased in both transgenic mouse lines but it was higher in the ETV4-A line (Fig. 1c), the difference between the two lines was statistically significant only in the AP (P < 0.001).

Neither gross nor microscopic prostate lesions were found in FVB wild-type (n = 10) or transgenic mice (n = 8) before 6 months of age. At 10–11 months, focal atypical lesions of the prostate epithelium were seen in 13 out of 18 (72%) transgenic mice from line A and in 4 out of 11 (36%) transgenic mice from line B (Table 1). These lesions were characterized by crowding and stratification of luminal cells (Fig. 2b, c) with variable degrees of nuclear atypia in the form of nuclear enlargement, pleomorphism, and hyperchromasia (Fig. 2d); they were identified as mouse prostatic intraepithelial neoplasia (mPIN). Immunohistochemical analysis showed loss of p63-positive basal cells, a major diagnostic criterion of PIN in humans (Fig. 2e). The average number of these lesions was 1.6 ± 0.2 in line A and 1.5 ± 0.3 in line B. The majority of lesions were classified as mPIN1; however, six mice from line A also showed the more severe phenotype of mPIN2 (Fig. 2f; Table 1). There was no evidence of invasive growth.

ETV4 has an important role in cell motility and in the invasiveness of prostate, breast, and colon cancer cells through the regulation of matrix metalloproteinases (MMPs) that cause degradation of extracellular matrix [30, 57]. We have previously shown in two human prostate cellular models (cancerous PC3 and normal RWPE cell lines) that ETV4 expression regulates migration in the wound-healing assay and invasion in Matrigel through the transcriptional regulation of some MMPs [34]. In ETV4 mice (n = 7), we found no variations of MMPs expression in AP; however, we found increased expression of MMP2, MMP7, and MMP9 mRNA in VP (3.0 ± 0.7, 3.7 ± 0.7, and 2.5 ± 0.7 fold respectively; P < 0.05) and in DLP (8.6 ± 3.5; 4.7 ± 1.8, and 1.8 ± 0.3 folds respectively; P < 0.08 not significant), suggesting that ETV4 regulates MMPs expression also in vivo.

The role of ETV4 in proliferation has not been broadly studied; however, we have previously demonstrated in human prostate cell lines that ETV4 overexpression have a role in cell proliferation [34], and this effect has been associated with the modulation of a set of cell cycle-regulating genes. In order to verify whether ETV4 increases the proliferation rate also in vivo, we measured the percentage of the Ki67+ cells in the prostate (Fig. 2g). In 5-month-old ETV4 mice, the percentage of proliferating (Ki67+) prostate cells was significantly increased in comparison with wild-type mice (P < 0.05; Fig. 2h), and this was confirmed also by PCNA staining (Fig. S2a,b) and BrdU incorporation (Fig. S2c, d). Thus, we analyzed two cell-cycle regulatory proteins of Cip/Kip family of cdk inhibitors, the cyclin-dependent kinase inhibitor 1A (Cdkn1a) and cyclin-dependent kinase inhibitor 1B (Cdkn1b) [58] in the prostate of ETV4 mice. We found a reduced expression of Cdkn1a mRNA in both ETV4 mouse lines compared with wild-type mice (Fig. 3a): the reduction was statistically significant in VP (P < 0.001) and AP (P < 0.01), but not yet in DLP (because the large amount of non-prostatic tissues in DLP or, alternatively, because prostatic lobes may differ in the gene expression pattern). No variation of Cdkn1b mRNA level was observed in ETV4 mice (data not shown). At variance, western blot (Fig. 3b, c) and immunohistochemical staining (Fig. 3d) showed that both p21 (encoded by Cdkn1a) and p27 (encoded by Cdkn1b) proteins were reduced.

In order to confirm that ETV4 regulates negatively CDKN1A and CDKN1B also in human prostate cells, we have tested the expression levels of these genes and of their encoded proteins (p21/WAF1/CIP1 and p27/KIP1, respectively) in two cellular models: in PC3 cells in which the high endogenous level of ETV4 expression was reduced by specific shRNAs and in RWPE cells in which we overexpressed ETV4. In keeping with the in vivo data and confirming our previous observations, we found that in both these specular cellular models ETV4 modulated the expression of p21 at both mRNA and protein level (Fig. 4), whereas p27 expression was modulated at protein level (Fig. 4b, c) but only slightly and non-significantly at mRNA level (Fig. 4a).

PC3 cells, in addition to ETV4, express also low amounts of ETV1 [59]. ETV4 silencing does not significantly affect ETV1 levels and vice versa (Fig. S3). In addition, ETV1 silencing does not significantly affect p21 and p27 levels (Fig. S3). Thus, the co-expression of ETV4 and ETV1 in this experimental setting is not complementary and does not invalidate our results.

To determine whether CDKN1A expression was regulated directly by ETV4, we studied the binding of ETV4 to the promoter of human CDKN1A by chromatin immunoprecipitation (ChIP) assays. First, by using the TESS-Transcription Element Search System software (http://​www.​cbil.​upenn.​edu/​tess) and the Tfsitescan/dynamicPlus server (http://​www.​ifti.​org/​cgi-bin/​ifti/​Tfsitescan.​pl) (Fig. 5a), we identified several ETV4-binding sites (BSs) in a 2.5 kb region upstream (sequence NG_009364 from GenBank) to the transcriptional start site (TSS).

Next, we performed ChIP in RWPE cells transiently transfected with either an ETV4-Flag expressing vector or a Flag-only control vector. The lysates from transfected cells were immunoprecipitated with either anti-flag antibody or an IgG control and the precipitated DNA fragments were PCR-amplified by using a pair of primers complementary to sequences of the CDKN1A promoter surrounding the ETV4 BSs (Table S1, Fig. 5a). As positive control, we used the sequence containing the ETV4 BS in the promoter of COX2 [60], a gene that is directly regulated by ETV4. As negative control, we used a sequence in the G6PD gene. The quantitative PCR (q-PCR) results suggested that ETV4 binds to only one of the putative BSs in CDKN1A promoter: the proximal ETV4 BS A (BS-A) at -676/-671 from TSS (Fig. 5b).

The effect of the binding of ETV4 to the BS-A site was analyzed by dual luciferase reporter assay, using a vector in which the firefly luciferase is under the control of a 845 bp fragment of the CDKN1A promoter that includes the ETV4 BS-A (Fig. 5c). In both RWPE and PNT1A (two immortalized non-cancer human prostate cell lines), the overexpression of ETV4 reduced the luciferase expression from the CDKN1A promoter (Fig. 5d). This effect was abrogated when a mutation was introduced into BS-A (Fig. 5d). As a counter-proof of the effects of ETV4 overexpression, we tested a human cancer prostate cell line, PC3, that has high basal levels of ETV4. In these cells, ETV4 silencing by vectors expressing ETV4-specific shRNA (shETV4a or shETV4b) increased the luciferase expression driven by the CDKN1A promoter (Fig. 5e). Again, no variation in luciferase expression was observed when the ETV4 BS-A was mutated (Fig. 5e). These results were confirmed also when ETV4 expression was silenced by a specific siRNA against ETV4 (Fig. 5e). These mirror experiments (ETV4 overexpression in PNT1A and RWPE cells; ETV4 silencing in PC3 cells) concur in suggesting that ETV4 negatively regulates CDKN1A by interacting with the proximal BS-A in the promoter of this gene.

In order to further confirm this conclusion, we performed dual luciferase reporter assay in PC3 and RWPE cells transfected with a vector containing the firefly luciferase under the control of a larger section of the CDKN1A promoter, a 1656 bp fragment that includes the ETV4 BS-A and the putative ETV4 BS-B that had not been confirmed by ChIP (Fig. 6a). Again, we also used derivatives of this vector in which one or both of these ETV4 BSs had been mutated (Fig. 6a). As expected, in PC3 cells, the reduction of ETV4 expression by two shRNA resulted in the increase of luciferase expression, compared with control cells, only with the wild-type ETV4 BS-A but not when it was mutated (Fig. 6b). In RWPE cells, the overexpression of ETV4 reduced luciferase expression with wild-type ETV4 BS-A but, surprisingly, also with the mutated ETV4 BS-A (Fig. 6c).

This last unexpected result raised the possibility that the 1656 bp fragment of the CDKN1A promoter might contain one or more elements responsive to a gene potentially regulated by ETV4 rather than ETV4 itself. Since the PC3 cells are p53 null, while the RWPE cells are p53 competent, p53 itself seemed like a good candidate because p53 is known to regulate p21 and a p53 BS is present in the CDKN1A promoter near to the ETV4 BS-B [55]. By luciferase assay with a vector containing only the p53 biding site (Fig. 7a) of the CDKN1A promoter [55], we found in the PNT1A and RWPE cell lines that transient overexpression of ETV4 reduced the luciferase expression compared to control cells (Fig. 7a).

Since these experiments suggest that ETV4 might regulate p53 expression, we tested the effect of ETV4 expression on p53 levels. In human RWPE, prostate cell line ETV4 overexpression does not modify the levels of p53 mRNA (data not shown); however, the levels of p53 protein are reduced (Fig. 7b, c). Although we do not yet know the mechanism, the same is true in the prostate of ETV4 mice (Fig. 7d, e).