Main

MicroRNAs (miRNAs) are short, non-coding RNAs that negatively regulate gene expression at the post-transcriptional level.1 By hybridizing to at least partially complementary regions on target mRNAs, miRNAs can induce mRNA degradation or translation inhibition, thus finely tuning protein expression in a variety of biological processes.1 Consequently, aberrant miRNA expression and function have been linked to the pathogenesis of human diseases, including cancer, where specific miRNAs have been proven to act as oncogenes or tumor suppressors.2

We previously showed that miR-205 is downregulated in prostate cancer (PCa) compared with adjacent non-neoplastic tissue.3 This finding was then confirmed by several independent studies (reviewed in Gandellini et al.4; Schaefer et al.5), and miR-205 recognized as the best single miRNA able to correctly distinguish prostate tumor from normal tissue.6 We also reported that miR-205 acts as a tumor suppressor in human prostate, as its reintroduction in PCa cells reverts epithelial-to-mesenchymal transition (EMT),3 thus suggesting that miR-205 reduction may drive the progression toward a cell phenotype with enhanced invasive properties and favor metastasis. Accordingly, tumors from patients with lymph node dissemination show lower miR-205 expression than those from node-negative patients.3 However, evidence of a downregulation of the miRNA in clinically localized carcinomas4 suggests that loss of miR-205 in PCa may anticipate disease progression. To gain insight into this early loss of the miRNA and into the mechanisms of PCa development, we investigated the physiological role of miR-205 in normal prostate.

Prostatic epithelium is characterized by three different cell layers: (i) an outer, androgen-independent basal layer, lying on a basement membrane (BM) and characterized by the expression of high molecular weight cytokeratins K5 and K14, and the transcription factor p63; (ii) an inner secretory/luminal layer expressing androgen receptor, prostate-specific antigen, and low molecular weight cytokeratins K8 and K18; (iii) a minor cell population of neuroendocrine cells rich in serotonin-containing granules, located between secretory cells.7

It has been reported that in murine prostate miR-205 expression is higher in basal than luminal cells,8 whereas the expression pattern of miR-200 family members, despite a similar role in regulating EMT,9 is inverse. In addition to suggesting that functions of miR-205 and miR-200 family may not completely overlap, the finding indicates that miR-205 downregulation in PCa may not only reflect the induction of EMT but also loss of the basal layer of epithelium.

To explain the enriched presence of miR-205 in normal basal cells, we investigated whether any transcription factor exclusively expressed by basal but not luminal cells could regulate miRNA expression and focused our attention on p63. TP63 gene encodes for six different protein products, resulting from alternative splicing – α, β and γ – at the 3′ end of two different pre-mRNA transcripts – amino-terminal transcriptional activation domain-containing (TAp63) and deleted p63 (ΔNp63) – generated from two distinct promoters.10 The ΔNp63α isoform is selectively expressed at high levels in the basal cells of stratified and glandular epithelia, including prostate, and its expression decreases with differentiation. Strikingly, p63 is not expressed in most prostate carcinomas, reflecting a loss of the basal layer, and the absence of staining for p63 is almost definitively diagnostic for PCa.11

In this study, the presence of an interplay between ΔNp63α and miR-205 in basal cells was investigated to elucidate the physiological role of miR-205 and the mechanism by which its loss may contribute to PCa development.

Results

Coregulation of miR-205 and genes involved in the maintenance of the basal phenotype

We interrogated a public dataset of gene and miRNA expression profiles (hereafter named Ambs dataset) obtained on prostatic normal and tumor tissue specimens12 to determine the genes correlated with miR-205. Positively correlated genes are supposed to be coexpressed in specific cell types and eventually participate in common pathways. Among genes positively correlated with miR-205, we found a significant enrichment of gene families related to the regulation of cell morphogenesis and ectodermal development, cell–cell junctions, and cell polarity (Supplementary Table S1). Besides factors related to EMT (CDH1, CDH3), there were basal cytokeratins K5, K14, K15 and K17, and additional genes involved in epithelial organization, such as integrins and laminins (Supplementary Table S2). Interestingly, not only basal cytokeratins were correlated with miR-205, but miR-205 was the miRNA most strongly correlated with K5 and the fifth most correlated with K14. The miRNAs most correlated with E-cadherin were instead miR-200 family members, which however did not correlate with basal cytokeratins. Overall, the findings corroborated the hypothesis that, despite a similar involvement in EMT, miR-205 could have additional functions related to maintenance of the basal compartment of epithelium.

In accord with this hypothesis, the probe set most correlated with miR-205 was 211194_s_at, which recognizes p63 (Supplementary Table S2). A significant positive correlation was also found for four other p63 probe sets (Figure 1a). In addition, miR-205 was the miRNA most correlated with the p63 probe sets 207382_at, 209863_s_at, and 211194_s_at. P63, and miR-205 were both expressed at higher levels in normal than in tumor samples, thus reflecting loss of the basal cell layer from tumors. Consistently, we found a significant direct correlation between ΔNp63 mRNA and miR-205 expression in a series of paired tumor and non-neoplastic tissue specimens (Figure 1b), where the levels of both factors were generally higher in normal samples than in matched cancer tissues (Figure 1c) and correlated with the percentage of normal tissue in each sample (Supplementary Figure S1). Enrichment of miR-205 expression in basal cells of normal epithelium and loss of miR-205 in PCa were then confirmed by in situ hybridization on tissue cryosections (Figure 1d).

Figure 1
figure 1

MiR-205 and p63 expression in prostate tissues. (a) Spearman’s correlation (rS) between miR-205 and the probe sets for p63 within Ambs dataset of clinical normal prostate (green) and prostate tumor specimens (red). (b) Correlation between p63 and miR-205 expression, as measured by qRT-PCR in our in-house series of paired normal and tumor specimens from 18 PCa patients subjected to radical prostatectomy. (c) qRT-PCR showing higher expression of both miR-205 and ΔNp63 in normal samples with respect to matched cancer tissues. Data represent average ΔNp63 or miR-205 expression levels ±S.D. *P<0.05; **P<0.01. (d) In situ detection of miR-205 in a representative normal prostate tissue (left) and a tumor specimen (right), showing enriched expression of the miRNA in the basal cell layer of normal acini and loss of expression in cancer. Original magnification for insert is × 20

Full size image

miR-205 and ΔNp63α interplay in prostate basal cells

Given the strong correlation found between p63 and miR-205, we wondered whether the miRNA could be transcriptionally regulated by p63 in prostate basal cells. Owing to the lack of cell lines showing a purely basal phenotype, functional investigations were performed on RWPE-1 cells. As they are derived from normal prostate tissue,13 they express both basal and luminal cytokeratins and are characterized by high levels of miR-205 and ΔNp63α (Supplementary Figure S2).

To guide the search for p63 sites in prostate cells, we made use of our chromatin immunoprecipitation-sequencing (ChIP-Seq) data obtained in human HaCaT keratinocytes,14 where miR-205 expression is fundamental.15 We noticed two peaks of p63, at a −13 Kb and +2 Kb from the recently identified16 miR-205 transcriptional start site (TSS; Figure 2a). Search for p63 motifs in these regions highlighted three highly predicted consensus sequences in the former, and two flanking the +2 Kb site (Figure 2a). Furthermore, the p63 peaks were within regions enriched in positive histone modifications, that is, H3K4 monomethylation and H3K27 acetylation, in different cell types (Figure 2a). ChIP was performed in RWPE-1 cells with two anti-p63 antibodies and confirmed that p63 binding to the −13 Kb peak was very strong (Figure 2b), less so for the +2 Kb peak. Functionally, transcriptional regulation of miR-205 by p63 was verified by silencing and overexpressing ΔNp63α in RWPE-1 cells, which resulted, respectively, in a concomitant decrease and increase in the levels of miR-205 primary transcript and mature form (Figure 2c).

Figure 2
figure 2

p63 binds miR-205 promoter and enhances miR-205 transcription in RWPE-1 cells. (a) Binding of p63 to the miR-205 locus in ChIP-Seq experiments of HaCaT keratinocytes (top). Ovals represent in silico-predicted p63 binding sites. Gray lines identify the regions at about −13 Kb and +2 Kb from miR-205 TSS that were amplified in ChIP experiments as including or contiguous to the predicted p63-responsive elements. Positive histone modifications (H3K4me1, H3K4me3, H3K27ac) are visualized in the miR-205 locus, with respect to p63 peaks and the TSS (data from the ENCODE consortium, obtained on K562, HUVEC, H1-hESC, NHEK, GM12878, HepG2, HSMM, HMEC cell lines) (bottom). (b) Chromatin from RWPE-1 cells was precipitated with two specific anti-p63 antibodies, the mouse monoclonal 4A4 (Mab) and a rabbit polyclonal (Pab), then qRT-PCR was run to amplify regions putatively bound by p63. Flag antibody and amplification for a centromeric region were used as negative controls to normalize data. Quantification by qRT-PCR shows strong p63 enrichment in the miR-205 −13 Kb region, less so for +2 Kb, as normalized for anti-flag immunoprecipitants. (c) qRT-PCR showing the levels of the primary transcript and the mature form of miR-205 in RWPE-1 cells overexpressed (using ΔNp63α expression vector) or silenced (using a specific siRNA, si-p63) for ΔNp63α expression. Data are reported as log2-transformed relative quantity with respect to the relative controls, empty vector or control siRNA (si-ctrl) (top). Western blotting showing ΔNp63α expression levels in RWPE-1 cells upon transfection. Ratios of band intensities in treated samples versus controls (both normalized for β-actin as equal loading control) are reported (bottom). Overexpression of p63 resulted in enhanced transcription of miR-205, as evidenced by increased expression levels of its primary transcript, which was paralleled by an increase in the mature form. Accordingly, silencing of p63 resulted in a marked downregulation of miR-205 primary transcript starting from 1 day after transfection, which was followed by a reduction of mature miRNA levels at day 3. *P<0.05; **P<0.005

Full size image

It is known that most miRNAs work in feedback loops where a transcription factor and a miRNA are mutually regulated.17 In this context, we wondered whether miR-205 was able to directly or indirectly regulate ΔNp63α. Silencing of miR-205 in RWPE-1 cells resulted in a marked upregulation of ΔNp63α protein (Figure 3a), although its mRNA levels appeared to be slightly down-modulated (Figure 3b). Consistent with this finding, overexpression of miR-205 induced an appreciable reduction in the amount of ΔNp63α protein (Figure 3a), regardless of increased mRNA expression levels (Figure 3b). Such data suggested the existence of two main mechanisms by which miR-205 may reciprocate its transcriptional activation by ΔNp63α, specifically a weak enhancement of p63 transcription accompanied by a major negative feedback loop effect consisting of post-transcriptional repression.

Figure 3
figure 3

MiR-205 can transcriptionally enhance but post-transcriptionally repress p63 in prostatic basal cells. (a) Western blotting showing modulations of ΔNp63α protein amount upon transfection of RWPE-1 cells with a LNA-modified antisense oligomer (LNA205) directed against miR-205, miR-205 synthetic precursor (pre205), and siRNA targeting ZEB1 (siZEB1). (b) Quantification of ZEB1 and ΔNp63 mRNA expression levels following miR-205 inhibition, overexpression, or ZEB1 silencing. Data are reported as relative transcript expression in LNA205-, pre205- or siZEB1-transfected cells over relative controls (LNA-Scr, preNeg, and si-ctrl). *P<0.05; **P<0.005. (c) Western blotting showing up-modulation of ZEB1 protein upon miR-205 silencing in RWPE-1 cells. (d) Treatment with a subtoxic concentration of bortezomib (1 nM) prevents ΔNp63α protein reduction by pre205. (e) Overexposure of p63 blotting showing appearance of high molecular weight p63 forms both upon miR-205 overexpression and bortezomib treatment. (f) Representative IP experiment showing increased ubiquitination of ΔNp63α by miR-205. α-Ub: anti-ubiquitin antibody. (g) Western blotting showing that NQO1 may be a potential target of miR-205. In fact, NQO1 levels are increased or reduced in RWPE-1 cells upon miR-205 silencing or overexpression, respectively

Full size image

To characterize the former effect, we focused on the transcription factor ZEB1, which is implicated in EMT, tumorigenesis, and metastasis.18 ZEB1 has been validated as a miR-205 target in canine MDCK cells, where it mediates the effect of the miRNA on EMT.9 ZEB1 has been also shown to repress ΔNp63 transcription.19 In RWPE-1 cells, silencing of miR-205 increased the amount of ZEB1 protein, thus confirming the ability of the miRNA to control ZEB1 expression (Figure 3c). In addition, silencing of ZEB1 (Supplementary Figure S3) induced up-modulation of ΔNp63 transcript levels, as also obtained by overexpressing miR-205 (Figure 3b). Such findings suggested that the slight enhancement of p63 transcription by miR-205 could be attributed to relief of ZEB1 repression on p63 promoter.

As regards the post-transcriptional repression of p63 by miR-205, it was surprising to observe that none of the nine tested computational algorithms for miRNA target prediction found putative miR-205 binding sites in ΔNp63α 3′ untranslated region (UTR), suggesting that miR-205 limited the amount of ΔNp63α protein mainly through an indirect mechanism, presumably by enhancing its degradation. To investigate this aspect, we exposed RWPE-1 cells to the proteasome inhibitor bortezomib and found that ectopic miR-205 expression failed to decrease ΔNp63α protein levels (Figure 3d). Rather, upon uncoupling from the proteasomal pathway, pre205-transfected cells expressed even higher ΔNp63α protein amounts than preNeg-transfected cells (Figure 3d), thus making it more evident that miR-205 may enhance p63 transcription. In addition, miR-205 induced the appearance of a ladder of bands of molecular weight above 98 kDa, which also accumulated with increasing doses of bortezomib (Figure 3e), thus representing poly-ubiquitinated forms of ΔNp63α. Consistently, increased ubiquitination of ΔNp63α by miR-205 was confirmed by probing anti-p63 immunoprecipitants with an anti-ubiquitin antibody (Figure 3f). Notably, the silencing of ZEB1 did not phenocopy miR-205 overexpression, as it induced an increase in ΔNp63α protein (Figure 3a), consistent with its transcriptional effect (Figure 3b).

Overall, such findings suggested that p63 suppression by miR-205 was ascribable to increased protein turnover, likely due to the ability of the miRNA to target genes that prevent p63 proteasomal degradation. In this context, putative miR-205 binding sites were predicted in the 3′UTR of several de-ubiquitinating enzymes, such as ubiquitin specific peptidase (USP)6, 22, 32, 35, 37, and 38, and in NAD(P)H:quinone oxidoreductase 1 (NQO1), recently shown to stabilize p63 protein by sequestering and protecting it from proteasomal degradation.20 Additional investigation would be required to discriminate which miR-205 targets are actually involved in the negative feedback loop from miR-205 to p63, however, preliminary evidence highlights the ability of miR-205 to modulate NQO1 expression (Figure 3g).

To validate the physiological relevance of the unique network involving p63 and miR-205, RWPE-1 cells were grown in different culture conditions to obtain shifts in p63 and cytokeratin expression. Strikingly, modulations of basal markers were paralleled by correlated variations in miR-205 expression, thus supporting both the functional relationship between ΔNp63 and miR-205 and the actual involvement of miR-205 in maintenance of the basal phenotype (Figure 4).

Figure 4
figure 4

Coordinate modulation of miR-205 and p63 during in vitro differentiation of RWPE-1 cells. (a) Representative phase-contrast micrographs of RWPE-1 cells grown in different culture conditions known to induce shifts in cytokeratin expression. High: 5 ng/ml epidermal growth factor (EGF) and 50 μg/ml bovine pituitary extract (BPE), low: 1.25 ng/ml EGF and 12.5 μg/ml BPE. (b) Quantification of ΔNp63, K14, and miR-205 expression levels by qRT-PCR in RWPE-1 cells cultured in media. Data are reported as relative expression with respect to cells grown in standard conditions (high EGF and BPE concentration). The table on the right shows calculated correlation coefficients between ΔNp63, K14, and miR-205. Specifically, lowering of EGF and BPE concentration in keratinocyte serum-free (KSF) medium resulted in increased ΔNp63 and K14 expression levels, whereas addition of serum or culturing in high calcium RPMI medium reduced them

Full size image

Functional role of miR-205 in basal cells

The specific role of miR-205 was investigated by profiling gene expression upon its depletion in RWPE-1 cells. A gene ontology enrichment analysis conducted on differentially expressed genes showed over-representation of functional families related to the extracellular region, cell cortex, and proteinaceous extracellular matrix (ECM) (Supplementary Figure S3), suggesting that miR-205 could mainly regulate the interaction between basal cells and the BM. In this context, we focused on laminin-332, the major adhesive component of prostatic BM. Composed of three subunits (α3, β3, and γ2), it is a ligand for at least two integrins expressed by epithelial cells, α3β1 and α6β4.21, 22

In Ambs dataset, expression of LAMA3, LAMB3, LAMC2, and ITGB4 was directly correlated with that of miR-205, suggesting a functional relationship. Accordingly, mRNA levels of all components of the laminin-332/integrin-β4 complex decreased in RWPE-1 cells silenced for miR-205, as evidenced by gene expression profiles and quantitative reverse transcriptase PCR (qRT-PCR; Figure 5a). Consistent with this observation, the amounts of laminin-β3, laminin-γ2, and integrin-β4 proteins in the whole extracts (Figure 5b) and in the media (Figure 5c) obtained from RWPE-1 cells were reduced following miR-205 depletion. Conversely, miR-205 overexpression increased laminin mRNA and protein levels, thus confirming that the miRNA is involved in the regulation of BM deposition by prostate basal cells (Figures 5a–c). These results were validated by laminin-γ2 immunofluorescence (Figure 5d).

Figure 5
figure 5

miR-205 is able to regulate the expression of laminin-332 and its receptor integrin-β4. (a) qRT-PCR assessment of mRNAs levels of LAMA3, LAMB3, LAMC2, and ITGB4 in RWPE-1 cells upon modulation of miR-205 expression or silencing of ZEB1. Data are reported as log2-transformed relative quantity with respect to the relative controls. *P<0.05; **P<0.005; ***P<0.0005. (b) Western blotting showing laminin-β3, laminin-γ2, and integrin-β4 expression levels in the whole extracts from RWPE-1 cells modulated for miR-205 expression or silenced for ZEB1. (c) Western blotting showing laminin-β3, laminin-γ2, and integrin-β4 expression levels in the media from RWPE-1 cells modulated for miR-205 expression or silenced for ZEB1. (d) Immunofluorescence showing the effect of modulating either miR-205 or ZEB1 on laminin-γ2 deposition by RWPE-1 cells. The effect on laminin-γ2 production by DU145 cancer cells restored for miR-205 expression is also reported. (e) Western blotting showing that miR-205 ectopic expression is sufficient to induce the expression of the components of laminin-332/integrin-β4 complex in PCa cells. However, silencing of ZEB1 can only up-modulate laminin-β3 expression, besides downregulating laminin-γ2 and integrin-β4

Full size image

As ZEB1 has been shown to repress the transcription of LAMA3, LAMB3, LAMC2, and ITGB4,23, 24 we checked whether miR-205 might enhance laminin production by down-modulating ZEB1. Strikingly, silencing of ZEB1 relieved transcriptional repression on only LAMA3 and LAMB3 genes and instead induced downregulation of LAMC2 and ITGB4 mRNAs and proteins (Figures 5a–d), as previously reported by Spaderna et al.24 Overall, these findings suggested that ZEB1 may not be the crucial mediator of miR-205 effect on laminins.

Given the reciprocal interplay between miR-205 and p63, we wondered whether ΔNp63α could also stimulate BM deposition and whether it could be specifically mediated by miR-205. In RWPE-1 cells, ΔNp63α overexpression increased protein (Figure 6a) and mRNA levels (Figure 6b) of all laminin-332/integrin-β4 complex subunits, although to a lesser degree than miR-205 overexpression. This effect appeared to be at least in part dependent on p63-induced up-modulation of miR-205 (Figure 2c), as it was reverted by locked nucleic acid (LNA)205 (Figure 6).

Figure 6
figure 6

Induction of laminin-332/integrin-β4 complex by ΔNp63α is at least in part dependent on miR-205. (a) Western blotting showing laminin-β3, laminin-γ2, integrin-β4 and p63 expression levels in the whole extracts from RWPE-1 or PC-3 cells overexpressed for ΔNp63α in combination or not with miR-205 inhibitor. (b) qRT-PCR assessment of mRNAs levels of LAMA3, LAMB3, LAMC2 and ITGB4 in RWPE-1 cells overexpressed for ΔNp63α in combination or not with miR-205 inhibitor. Data are reported as log2-transformed relative quantity with respect to the cells transfected with the combination of empty vector and LNA-Scr. *P<0.05; **P<0.01 (P-values calculated with respect to empty+LNA-Scr for ΔNp63α+LNA-Scr-transfected cells and with respect to ΔNp63α+LNA-Scr for ΔNp63α+LNA205-transfected cells)

Full size image

miR-205 stimulates BM deposition by PCa cells

PCa is characterized by disruption of the acinar architecture of epithelium, with loss of the basal layer.10 As a consequence, PCa cells have a predominantly luminal phenotype and express neither p6310 nor miR-205.3 We wondered whether the capability of miR-205 to regulate BM deposition could be at least partially retained in p63-deficient PCa cells. We found that restoration of miR-205 in DU145 and PC-3 cells was sufficient to increase the expression levels of all the components of the laminin-332 complex (Figures 5d and e and Supplementary Figure S4), even though p63 mRNA and protein levels remained undetectable (data not shown). Conversely, ΔNp63α increased protein (Figure 6a) and mRNA levels (twofold) of only integrin-β4, an effect that was not reverted by LNA205. Notably, miR-205 expression remained undetectable in ΔNp63α-expressing PC-3 cells (data not shown), thus confirming the strong epigenetic repression of the miRNA found in these cells25 and showing that, in the absence of miR-205, p63 cannot enhance production of the whole laminin-332 complex. The retained ability of p63 to induce integrin-β4 may rely on a direct transcriptional activation, as reported by Carroll et al.26 In this context, integrin-β4 expression levels were the least affected by miR-205 silencing also in ΔNp63α-overexpressed RWPE-1 cells (Figure 6).

To assess the effect of laminin re-expression by miR-205 on the growth of PCa cells in a three-dimensional landscape, invasive PC-3 cells were grown in growth factor-reduced Matrigel. PreNeg-transfected cells mainly grew as non-polarized masses, which turned into a stellate-invasive phenotype from day 7 (Figure 7a), as also described by Härmä et al.27 However, miR-205 transient re-expression inhibited the formation of invadopodia and allowed PC-3 cells to organize as acinar structures surrounded by an intact BM, as assessed by microscopy and immunofluorescence (Figure 7). Such findings suggested that the enhancement of laminin-332/integrin-β4 production by miR-205 was sufficient to restore a normal-like epithelial organization of PCa cells and inhibit invasion through the ECM. Again, silencing of ZEB1 in PC-3 cells (Supplementary Figure S3) did not completely phenocopy the rearrangements induced by miR-205, as it failed to induce up-modulation of all the components of the laminin complex (Figure 5e). Accordingly, suppression of ZEB1 only partially reverted the invasive phenotype of PC-3 cells grown in Matrigel, as it just retarded the sprouting of invadopodia, without evidence of acinar organization (Figure 7).

Figure 7
figure 7

miR-205 allows prostate cancer cells to reorganize into normal-like 3D structures. (a) Time course assessment of three-dimensional growth of PCa cells in Matrigel. Micrographs are provided for cells overexpressed for miR-205 or silenced for ZEB1, with relative controls. (b) Immunofluorescence on 3D cultures showing that miR-205 ectopic expression is sufficient to revert invasive phenotype of PC-3 cells into a normal-like acinar morphology, where cell/ECM contacts stain for laminin receptor integrin-β4. However, silencing of ZEB1 does not restore normal morphology

Full size image

Discussion

Forming a continuous layer in human prostate, basal cells are the presumed progenitor cells of prostatic epithelium and exclusively express ΔNp63,7 which is required for key events in prostate development, including lineage commitment and differentiation.28, 29 ΔNp63 is also involved in regulating deposition of the BM, a layer of specialized ECM that surrounds normal glands and separates epithelial cells from underlying mesenchymal tissues, thus providing a biophysical support as well as a hub to regulate cell–cell and cell–protein interactions.21, 29, 30, 31

Starting from the evidence that miR-205 expression is enriched in murine prostatic basal cells,8 we identified an interplay between ΔNp63α and miR-205 in human basal cells, where ΔNp63α is able to enhance miR-205 transcription, whereas the miRNA can post-transcriptionally limit the amount of p63 protein (Figure 8). Apparently, this negative feedback loop relies on differential p63 proteasomal degradation rather than on a canonical miRNA/target interaction, as no binding sites for miR-205 were predicted in the 3′UTR of ΔNp63α. However, when we extended the search to other regions in ΔNp63α mRNA, a putative miR-205 site was found in 5′UTR, thereby suggesting that miR-205 might also induce ΔNp63α silencing by the classical miRNA pathway, even through a non-canonical interaction.

Figure 8
figure 8

Theoretical miR-205/p63/laminin network in prostate basal cells. The drawing summarizes the network involving ΔNp63 and miR-205, where p63 can enhance miR-205 transcription (A), miR-205 transcriptionally regulates p63 by repressing ZEB1 (B), and miR-205 indirectly reduces ΔNp63 protein amount by increasing its proteasomal degradation (C). Functionally, miR-205 is able to regulate laminin-332 deposition, thus playing a pivotal role in the maintenance of tissue integrity (D)

Full size image

The bi-univocal regulation between miR-205 and p63 guarantees their coordinate expression in prostate basal cells, as evidenced by the analysis of clinical series (Figure 1) and functionally validated in a differentiation assay (Figure 4). It is noteworthy that coexpression of miR-205 and p63 is not specific for prostatic tissue, as miR-205 has been also detected in the basal/myoepithelial layers of human mammary gland32 and in p63-positive tumors, such as squamous cell lung cancer. In this context, high expression of miR-205 has been proposed as a marker of squamous histology compared with adenocarcinoma,33 but an oncogenic role for the miRNA has not been proved. Overall, the widespread coexpression of miR-205 and p63 in basal cells from glandular and stratified epithelia underscores the participation of miR-205 in p63-regulated pathways that are fundamental for the maintenance of epithelial structure.

We found that depletion of only miR-205 did not convert RWPE-1 cells into full-blown luminal cells, thus suggesting that additional arrangements (such as the concomitant downregulation of p63 and up-modulation of miR-200 family members) are necessary to complete transdifferentiation of basal into secretory cells. However, silencing of miR-205 was sufficient to reduce the secretion of all subunits of laminin-332 (the first ECM protein to be synthesized during embryogenesis and thought to have a crucial role in morphogenesis and organogenesis22) and integrin-β4. As laminin-332/integrin α6β4 complexes are critical to connect the ECM to intermediate filaments within hemidesmosomes,22 miR-205 may be involved in stably anchoring prostatic basal cells to basal lamina, thus controlling normal prostate morphogenesis and tissue integrity.

Loss of cell polarity and cell:BM adhesion has been observed in many carcinomas including invasive PCa and advanced prostatic intraepithelial neoplasia (PIN), which is generally accepted as a precursor of invasive cancer.10, 11, 31 One of the hallmarks of PIN is indeed the focal attenuation of the basal cell layer, which is completely lost in even low-grade invasive carcinoma. As the lack or discontinuity of BM structures is a prerequisite for tumor cell invasion into the interstitial space and metastatic dissemination, BM maintenance represents a barrier against cancer development and progression. On the basis of our data, pathological loss of miR-205 in basal cells may result in a reduced expression of BM components and pave the way for the development of PIN and invasive cancer, which are indeed characterized by loss of integrin-β4 and laminin-332,22, 31 both regulated by the miRNA. In this context, it is worth mentioning that the PCa origin is still unclear.34 The evidence that PCa is characterized by a predominantly luminal phenotype convinced many scientists that tumor cells might arise as a consequence of mutations in luminal cells. Low expression of miR-205 (and p63) in PCa cells may hence just reflect their luminal origin. However, Lawson et al.35 recently demonstrated that a phenotypically luminal PCa may also arise from basal cells upon constitutive activation of oncogenic signaling pathways. Starting from this evidence, we cannot exclude that pathological miR-205 loss in prostate basal cells may favor malignant transformation. In our hands, however, the sole depletion of miR-205 in RWPE-1 cells did not result in enhanced cell growth in vitro (data not shown).

miR-205 overexpression in cancer cells was sufficient to enhance expression of different subunits of laminin-332, thus indicating that its capability to modulate BM assembly is at least in part independent of ΔNp63α. This aspect is of particular interest in the cancer context, as it would suggest that miR-205 replacement therapy can be envisaged not only to increase cell–cell adhesion and reduce intrinsic invasion of PCa cells,3 but also to counteract the loss of the basal cell layer. Strikingly, in 3D-culture experiments, miR-205 indeed restored normal morphogenesis of invasive PCa cells, likely due to increased deposition of BM components. Accordingly, Wang et al.36 showed that invasive MDA-MB-231 breast cancer cells can be induced to form morphologically normal structures, a phenomenon called ‘phenotypical reversion’, upon treatment with inhibitors of integrin-β1 or epidermal growth factor receptor.

Overall, our data show that miR-205 participates in a network involving ΔNp63α and ZEB1, which seems to be essential for maintenance of the basal cell layer of prostatic epithelium. Specifically, miR-205 is able to modulate BM deposition (Figure 8), thus being presumably required for prostate morphogenesis and to guarantee tissue integrity. As a consequence, pathological loss of miR-205 may favor tumorigenesis by creating discontinuities in the BM. From a therapeutic point of view, miR-205 replacement in PCa cells may both revert EMT and increase BM deposition, thus ultimately hampering PCa progression.

Materials and Methods

See Supplemental Materials and Methods for the description of experimental models, qRT-PCR, microarray analysis, primers, and antibodies used.

Ambs dataset analysis

Normalized miRNA and mRNA data were downloaded from GEO repository (http://www.ncbi.nlm.nih.gov/geo), with GEO submission accession number GSE8126 and GSE6956, respectively. Gene expression data were not further processed, whereas miRNA data were filtered, discarding mouse probes, and multiple probes were averaged. Data were analyzed by using R software (http://www.R-project.org) to determine the genes correlated with miR-205. Positively correlated genes were functionally classified using DAVID (http://david.abcc.ncifcrf.gov/) public software, and gene ontology enrichment analysis was performed using Affymetrix (Santa Clara, CA, USA) gene chip HG-U133A 2.0 as background.

In situ hybridization

In situ hybridization for miR-205 on human prostate tissue cryosections (4 μm) was performed using miRCURY LNA microRNA ISH Optimization kit (#90000, Exiqon, Vedbaek, Denmark), according to the manufacturer’s instructions. Briefly, 40 nM of 5′-biotin labeled miRCURY LNA miRNA detection probe for human miR-205 (#18099-03, Exiqon) was hybridized onto glass slides at 55 °C for 2 h. Dako REAL Detection System Peroxidase/DAB+ kit (K5001, Dako, Glostrup, Denmark), designed for the detection of biotinylated probes, was then used according to the manufacturer’s instructions. Nuclei were counterstained with hematoxylin. Images were acquired at room temperature in oil immersion by a Nikon Eclipse Ti equipped with Hamamatsu ORCA-R2 camera and Nikon PlanFluor 40 × /1.30 oil or Nikon PlanAPO VC 20 × /0.75, using Volocity 5.4 software (PerkinElmer, Walthan, MA, USA), then processed with Adobe Photoshop Image Reader 7.0.

Chromatin immunoprecipitation

The procedure for ChIP was essentially as described in Testa et al.37 A total of 3–5 × 106 RWPE-1 cell equivalents of chromatin were immunoprecipitated with 10–15 μg of anti-p63 or anti-flag control antibodies. After reversal of formaldehyde cross-linking, RNase A and proteinase K treatments, input and IP-enriched DNAs were used to perform qRT-PCR to amplify regions putatively bound by p63. A genomic region (centromeric satellite 11) was also amplified as a negative control. A Biorad (Hercules, CA, USA) MyIQ single color thermal cycler and a SYBR Green PCR Master mix were used for qRT-PCR. Specificity of products was monitored with a heat dissociation curve. For ChIP, fold enrichment was calculated with the 2−ΔΔCt method, where the Ct represented the threshold cycles of the input, the specific antibody and the negative antibody. A detailed description of antibodies and region-specific primers used in ChIP experiments can be found in Supplemental Materials and Methods.

The presence and position of putative p63 binding sites within precipitated regions was predicted using the CONSITE algorithm (http://www.phylofoot.org/), employing the matrix reported in Ortt and Sinha38 and setting a cutoff ≥0.85.

Cell-based experiments

Cells seeded at the appropriate density were transfected for 4 h with 100 nM LNA-modified antisense oligomers, 20 nM miRNA precursors, 25 nM siRNAs or 0.5 ng/μl ΔNp63α expression/empty vector using Lipofectamine2000 (Invitrogen/Life Technologies, Carlsbad, CA, USA), according to the manufacturer’s instructions, and processed for 48–72 h after transfection. For siRNA sequences, and LNA and miRNA precursors refer to Supplemental Materials and Methods.

miRNA target prediction

The presence of putative miR-205 binding sites in a given mRNA was predicted using MiRWalk (http://www.ma.uni-heidelberg.de/apps/zmf/mirwalk), which performs a comparison of the information produced by nine established miRNA target prediction programs, and PITA (http://genie.weizmann.ac.il/pubs/mir07/index.html) online tools. Only targets simultaneously predicted by at least five programs were considered as reliable.

Immunoblotting analyses

For immunoblotting, 30 μg of cell lysates or 12.5 μl of cultured media (after removing of cell debris and concentration with Amicon Ultra-4, Centrifugal Filter Device, Millipore, Billerica, MA, USA) were fractioned by SDS-PAGE and transferred onto Hybond nitrocellulose membranes (GE Healthcare Europe GmbH, Pollards Wood, UK). Equal protein loading was verified by Ponceau staining. Filters were blocked in PBS-Tween-20/5% skim milk and probed with specific antibodies, which were visualized by SuperSignal West PICO chemiluminescent detection system (Pierce Biotechnology, Rockford, IL, USA). Densitometric analysis has been carried out using Vision Works LS software (UVP, Upland, CA, USA) and β-actin was used as equal protein loading control for cell lysates.

Immunoprecipitation

For assessment of the ubiquitinated form of p63, precleared, detergent-solubilized cell extracts (100 μl) were immunoprecipitated with the anti-p63 4A4 antibody (200 ng/ml) for 16 h at 4 °C, with precipitation of the immune complex by addition of 50 μl of a 50 : 50 protein A slurry.

Immunofluorescence on 2D cultures

For immunofluorescence analyses, cells grown on glass coverslips were fixed with 4% formaldehyde, permeabilized with 70% cold ethanol, then probed with specific primary antibodies for 1 h and with Alexa Fluor488-labeled (Life Technologies) secondary antibodies for 1 h at room temperature. Nuclei were counterstained with DAPI. Images were acquired at room temperature in oil immersion by a Nikon Eclipse Ti equipped with Hamamatsu ORCA-R2 camera and Nikon PlanFluor 40 × /1.30 oil using Volocity 5.4 software (PerkinElmer) , then processed with Adobe Photoshop Image Reader 7.0.

3D cultures

3D cultures were prepared as described by Tyson et al.,39 with few modifications. Briefly, 10 μl of thawed growth factor-reduced Matrigel (E6909, Sigma, St. Louis, MO, USA) was added to each well of μ-slides angiogenesis plates (Ibidi, Munchen, Germany). After a 1-h incubation at 37 °C, 1000, cells/well were added to 50 μl of culture medium containing 2% Matrigel. Medium was then replaced every 3 days and cell morphology was checked in a time-lapse humidified chamber at 37 °C under a Nikon Eclipse Ti equipped with Hamamatsu ORCA-R2 camera and Nikon PlanAPO VC 20 × /0.75 using Volocity 5.4 software. 3D reconstitutions were performed by Volocity 5.4 software without setting alterations. For immunofluorescence analyses, 3D cultures were fixed in 4% paraformaldehyde, then probed overnight with primary antibodies and 4 h with Alexa Fluor488-labeled (Life Technologies) secondary antibodies. Images were acquired as described above.

Statistical analyses

Spearman’s coefficients with associated P-values were calculated for correlation analyses. Two-sided one-sample or two-sample Student's t tests were used to analyze the differences in miRNA and gene expression. P<0.05 was considered statistically significant. Data are reported as mean values±S.D. of at least three independent experiments.