PKD2 and PKD3 promote prostate cancer cell invasion by modulating NF-κB- and HDAC1-mediated expression and activation of uPA

Although tremendous efforts have been made to improve the effectiveness of treatment, prostate cancer remains the second leading cause of cancer-related deaths among men in the USA. Morbidity and mortality primarily result from the growth of metastatic tumors in distant organs. Multiple steps are involved in the metastatic process. In the initial step, the extracellular matrix (ECM) surrounding the tumor is remodeled, allowing tumor cells to invade across the basement membrane and into circulation. One of the most important molecules involved in the ECM remodeling process is urokinase plasminogen activator (uPA), a serine protease that initiates the activation of ECM-degrading enzymes and, through association with its receptor (uPAR), modulates various biological functions, including cell migration, angiogenesis, differentiation and wound healing (Dass et al., 2008). However, the mechanisms that regulate uPA expression during metastasis have not been fully defined.

Many studies have suggested that nuclear factor kappa B (NF-κB) has a pivotal role in promoting the invasive and metastatic potential of prostate cancer. In contrast to LNCaP cells, a less invasive prostate cancer cell line, PC-3 and DU145 cells exhibit constitutive activation of the NF-κB DNA-binding activity (Palayoor et al., 1999). In addition, overexpression of NF-κB in LNCaP and DU145 cells significantly increases the expression of prostate-specific antigen (PSA), which is closely associated with the progression of prostate cancer (Chen and Sawyers, 2002). Conversely, inhibition of NF-κB prevents tumor angiogenesis, invasion and metastasis in PC-3M xenografts in nude mice (Huang et al., 2001). Nevertheless, the signaling pathways upstream of NF-κB that lead to prostate cancer invasion and metastasis have not yet been illustrated.

Protein kinase D (PKD) belongs to a family of serine/threonine protein kinases that is now classified as a subfamily of the Ca²⁺/calmodulin-dependent kinase (CaMK) superfamily. Three members of the PKD family have been identified to date, namely, PKD1 (PKCμ), PKD2 and PKD3 (PKCν) (Wang, 2006). PKD has been implicated in various biological processes, such as cell proliferation (McEneaney et al., 2010; Papazyan et al., 2008; Wang, 2006), survival (Arun et al., 2011; Song et al., 2009), cell migration (Eiseler et al., 2010; Peterburs et al., 2009), differentiation (Jadali and Ghazizadeh, 2010; Kleger et al., 2011), inflammation (Kim et al., 2010b; Park et al., 2009) and vesicle transportation (Chen et al., 2009; Hausser et al., 2005; Lu et al., 2007). The role of PKDs in cancer invasion/metastasis has been increasingly recognized, and, of the three PKD isoforms, PKD1 has been the focus of most studies in this context. In breast cancer cells, the expression of PKD1 has been shown to be inversely correlated with invasive potential, impeding this process through downregulation of the expression of the matrix metalloproteinases (MMPs) MMP2, MMP7, MMP9, MMP10, MMP11, MMP13 and MMP14 (also known as membrane-type 1 MMP, or MT1-MMP) (Eiseler et al., 2009). Furthermore, overexpression of PKD1 increases aggregation, but reduces motility and invasiveness of prostate cancer cell lines through interaction with and phosphorylation of E-cadherin (Jaggi et al., 2005; Syed et al., 2008). In our previous study, we found that PKD3 levels were substantially increased in human prostate tumors compared with those in normal tissues, as revealed by immunohistochemistry. Moreover, PKD3 exhibited a marked increase in nuclear localization in tumor tissues, correlating with tumor grade. Increased protein expression and nuclear accumulation of PKD3 were observed in the more invasive and metastatic PC-3 and DU145 cell lines compared to the less aggressive LNCaP cell line, and this was shown to contribute to the growth and survival of prostate cancer cells (Chen et al., 2008). However, the exact role of PKD2 and PKD3 in tumor cell invasion has not been characterized. Thus, the primary goal of this study was to investigate the function and regulatory mechanisms of PKD2- and PKD3-mediated pathways in prostate cancer cell invasion.

Our previous study demonstrated that PKD3 promotes prostate cancer cell growth and survival (Chen et al., 2008). In addition, we found that PKD3 expression was increased in more invasive prostate cancer cell lines (i.e. expression of PKD3 in PC-3 and DU145 cells is higher than in LNCaP cells), suggesting that there is a positive role for PKD3 in regulating the aggressiveness of prostate cancer cells. However, the involvement of PKD3 in prostate cancer cell invasion has not been determined. Here, we analyzed whether PKD2 and PKD3, the two predominant PKD isoforms expressed in PC-3M, PC-3 and DU145 cells, have a role in the migration and invasion of prostate cancer cells. In DU145 and PC-3M cells, small interfering RNA (siRNA)-mediated depletion of PKD2 or PKD3 significantly decreased cell migration, with depletion of PKD3 resulting in a more substantial decrease than depletion of PKD2 (Fig. 1A,B). Analysis of cell invasion using Matrigel-coated filters in transwell chambers demonstrated that knockdown of PKD2 or PKD3 inhibited the invasion of PC-3M cells. Moreover, depletion of both PKD2 and PKD3 resulted in greater inhibitory effects (Fig. 1C). To demonstrate the absence of possible off-target effects, we used two siRNAs for each isoform, targeting different regions of the PKD2 or PKD3 mRNA, and monitored the effects of these siRNAs on the invasion of DU145 cells. Consistent with our above results, the invasion potential of DU145 cells was significantly reduced by two different siRNA targeting PKD2 or PKD3 (Fig. 1D). These results suggest that both PKD2 and PKD3 contribute to prostate cancer cell migration and invasion.

Given that PKD2 and PKD3 were shown to be crucial for prostate cancer cell migration/invasion, we sought to identify the genes responsible for PKD2- and PKD3-induced invasion. Our results demonstrated that depletion of PKD2 (siPKD2-1), PKD3 (siPKD3-1), or both PKD2 and PKD3 (siPKD2+3) in PC-3M cells led to significant downregulation of activin A, MT1-MMP, uPA and uPAR, and upregulation of plasminogen activator inhibitor-2 (PAI-2) at the mRNA level compared with a nontargeting siRNA control (siCTL); no significant difference in the expression of osteopontin or vascular endothelial growth factor (VEGF) was observed (Fig. 2A). Among these invasion/metastasis-related genes, we focused on those involved in the uPA-uPAR and MMP systems and further investigated changes in the expression of these targets at the protein level by western blotting. Protein levels of uPA, uPAR and MT1-MMP were significantly downregulated in PC-3M (Fig. 2C) and DU145 cells (supplementary material Fig. S2A) after silencing of PKD2 or PKD3 using two different siRNAs, also verifying the absence of off-target effects for these PKD2 or PKD3 siRNAs. Because phorbol 12-myristate 13-acetate (PMA) has been shown to activate PKC and PKD family members (Wang, 2006) and upregulate uPA-, uPAR- and MT1-MMP-encoding mRNA expression in PC-3M cells (supplementary material Fig. S1), we used PMA to induce the expression PKD and the uPA system proteins. Transfection with siPKD2, siPKD3 or siPKD2+3 resulted in decreased expression of uPA, uPAR and MT1-MMP and increased expression of PAI-2 compared with transfection with siCTL in PC-3M (Fig. 2B) and DU145 (supplementary material Fig. S2C) cells, both with and without PMA stimulation. Moreover, given that epidermal growth factor (EGF) has been shown to upregulate uPA-uPAR signaling (Festuccia et al., 2005; Amos et al., 2010) and initiate prostate cancer cell invasion (Jarrard et al., 1994), and tumor necrosis factor (TNF)-α (pivotal in inflammation and invasion interactions) (Stock et al., 2008) has been shown to upregulate expression of the gene encoding uPA (PLAU, hereafter referred to as the uPA gene) (Kim et al., 2010a; Guerrini et al., 1996), we used EGF and TNF-α treatments to determine whether PKD2 or PKD3 regulate uPA, uPAR and MT1-MMP expression through these two physiological agonists. As shown in Fig. 2D and supplementary Fig. S2B, TNF-α significantly induced the expression of uPA and MT1-MMP, and this effect was dramatically reduced by silencing of PKD2 or PKD3 in both cell lines. Similarly, EGF-induced MT1-MMP expression was also reduced in both cell lines by PKD2 or PKD3 knockdown, although there was little effect on uPA expression. Both EGF and TNF-α did not substantially alter uPAR expression in either cell line, but baseline uPAR expression was significantly downregulated by PKD2 or PKD3 knockdown.

Activation of uPA-uPAR signaling has been shown to activate a cascade of MMPs, subsequently modulating tumor invasion (Dass et al., 2008). To further confirm the effects of PKD2 and PKD3 on uPA and MMP activity, we analyzed uPA activity using a chromogenic approach and assayed the gelatinase activity of MMP9 by zymography. Knockdown of either PKD2 (siPKD2-1, siPKD2-2) or PKD3 (siPKD3-1, siPKD3-2) significantly reduced uPA activity in the culture medium of PC-3M cells; knockdown of both PKD2 and PKD3 (siPKD2+3: siPKD2-1 + siPKD3-1) resulted in an even greater reduction in uPA activity (Fig. 3A, left). Similar results were obtained in DU145 cells in the presence or absence of PMA, and maximal inhibition of uPA activity was obtained with depletion of PKD3 (Fig. 3A, right). Moreover, although no detectable baseline MMP9 activity was observed, PMA-induced MMP9 gelatinase activity was greatly reduced after PKD2 or PKD3 depletion in both DU145 and PC-3M cells (Fig. 3B,C).

Because several studies have demonstrated that NF-κB might have a central role in the transactivation of uPA and uPAR (Killeen et al., 2009; Sliva et al., 2002; Tsunoda et al., 2005), we explored whether PKD2- and/or PKD3-dependent expression and activation of uPA could be attributed to upregulation of NF-κB signaling. As shown in Fig. 4A, the baseline and PMA- (Fig. 4A, left) or TNF-α-induced (Fig. 4A, right) DNA-binding activity of NF-κB to a uPA-specific promoter sequence (uPA-luc) declined dramatically in PC-3M cells in response to PKD2, PKD3 or combined PKD2 and PKD3 depletion. Similarly, the baseline and PMA- (Fig. 4B, left) or TNF-α-induced (Fig. 4B, right) DNA-binding activity of NF-κB to a consensus promoter sequence (2×NF-κB) declined substantially in all PKD-knockdown cells. PMA-induced uPA transactivation was also significantly reduced by PKD2 and/or PKD3 depletion in DU145 cells (Fig. 4C, left). Moreover, silencing of PKD2 or PKD3 with different siRNAs significantly inhibited uPA transactivation in DU145 cells cultured with 10% fetal bovine serum (FBS) (Fig. 4C, right).

To further confirm the possibility that downregulation of uPA transactivation resulting from PKD2 and/or PKD3 depletion was caused by reduced binding of NF-κB to the uPA promoter sequence, we utilized chromatin immunoprecipitation (ChIP). As shown in Fig. 4D, the baseline and PMA-induced binding of p65 NF-κB to the uPA promoter region were dramatically reduced in response to PKD2 or PKD3 depletion. As a negative control, binding of p65 NF-κB to the 11th exon of uPA, which has been shown to not bind p65 NF-κB, was also monitored (Fig. 4D, right); no binding of this negative control over the level of background (which can be attributed to nonspecific binding of the beads) was observed, indicating that the IP and real-time PCR-based amplification of the uPA promoter sequence was specific. Taken together, these results indicate that PKD2 and PKD3 have a significant role in NF-κB-mediated transactivation of uPA.

Although our data suggested that PKD2 and PKD3 were essential for NF-κB-dependent transcription of uPA, the exact mechanisms remained unclear. Because recent studies have reported the occurrence of translocation-independent NF-κB activity, we utilized a subcellular fractionation approach, combined with immunofluorescence, to analyze the influence of PKD2 and PKD3 on the PMA-induced nuclear translocation of p65 NF-κB. As shown in Fig. 5A, PMA-induced p65 NF-κB nuclear translocation was substantially reduced by PKD2 depletion, but only slightly reduced by PKD3 depletion. These differences could not be attributed to different knockdown efficiencies of the PKD2 and PKD3 siRNAs given that both PKD2 and PKD3 protein levels were sufficiently reduced. Consistent with these results, the amount of cytosolic p65 NF-κB decreased significantly upon PMA stimulation in siCTL-transfected cells, but remained constant in PKD2- or PKD3-knockdown cells, both with and without PMA treatment. Additionally, overexpression of PKD2, but not PKD3, caused a significant increase in PMA-induced p65 nuclear translocation in PC-3M cells (Fig. 5A, right).

The role of PKD2 and PKD3 in PMA-induced p65 nuclear distribution was further confirmed by immunofluorescence staining and confocal microscopy (Fig. 5B). As shown in Fig. 5C, PMA treatment induced p65 nuclear translocation in ∼78.3% of siCTL-transfected cells, whereas only 6.7%, 43% and 6.7% of cells exhibited a nuclear distribution of p65 in siPKD2-, siPKD3-, and siPKD2+3-transfected cells, respectively. Taken together, these results indicate that PMA-induced p65 nuclear translocation was primarily dependent on PKD2 rather than PKD3.

The above results suggest that PKD2 and PKD3, but especially PKD2, have an important role in p65 nuclear translocation. Here, we sought to further elucidate the possible mechanisms underlying p65 nuclear translocation. Because the phosphorylated IκB kinase, phosphorylated inhibitor of NF-κB, IκB (pIKK-pIκB-IκB) degradation cascade is the classical pathway that mediates p65 nuclear translocation, we monitored potential changes in this pathway in response to PKD2 and/or PKD3 knockdown using western blotting. In PC-3M cells, PMA stimulation reduced IKK and IκB phosphorylation to a greater extent in PKD2-knockdown cells than in PKD3-knockdown cells, and greater inhibition of IκB degradation was observed when PKD2 was silenced (Fig. 6A). Similar, albeit weaker, results were obtained when PC-3M cells were treated with TNF-α (Fig. 6B). Collectively, these data suggest that PKD2 is the primary regulator of pIKK-pIκB-mediated IκB degradation in this context. More importantly, we found that endogenous PKD2 or PKD3 co-immunoprecipitated with IKKβ in PC-3M cells, and this interaction could be significantly enhanced by PMA stimulation (Fig. 6C). In contrast, PKD2 and PKD3 did not directly interact with p65 with or without PMA treatment (supplementary material Fig. S3).

In addition to nuclear translocation, phosphorylation at Ser536 or Ser276 of p65 has been reported to be involved in regulating the interaction of p65 with its co-activator CBP/p300 and subsequent transcriptional activation of its target genes (Hu et al., 2004; Jeong et al., 2005; Oeckinghaus and Ghosh, 2009; Zhong et al., 2002). Thus, we evaluated the effects of PKD2 and PKD3 on p65 phosphorylation at Ser276 and Ser536 and subsequent uPA transactivation triggered by PMA and TNF-α. Interestingly, PMA- and TNF-α-induced phosphorylation of p65 at Ser276 was reduced to a greater extent by PKD2 depletion than PKD3 depletion, whereas phosphorylation at Ser536 was reduced mainly by PKD3 silencing in PC-3M cells (Fig. 6A,B).

To further confirm that phosphorylation of p65 at Ser536 was involved in PKD3-mediated uPA transactivation, we analyzed whether Ser536 activation could rescue the transcriptional deactivation of uPA caused by PKD3 depletion by co-transfecting a mutant p65 plasmid with PKD3 siRNA. Ser536 of p65 was replaced with either alanine (S536A, dominant-inactive mutant) or aspartic acid (S536D, constitutively active mutant) (Sasaki et al., 2005). As shown in Fig. 7A, overexpression of wild-type p65 (p65WT) or constitutively active p65 (S536D) reversed the downregulation of uPA transactivation observed with PKD3 depletion, and a maximum rescue effect was obtained upon expression of p65 S536D. In contrast, expression of constitutively inactive p65 (S536A) showed little rescue of uPA transactivation.

Next, we investigated whether the reduced tumor cell invasion observed with PKD2 or PKD3 silencing could be rescued by overexpression of p65. As expected, overexpression of p65 rescued DU145 cell invasive potential reduced by knockdown of PKD2 or PKD3 (Fig. 7B).

Given that histone deacetylase 1 (HDAC1) binds to the uPA promoter and negatively regulates uPA transcriptional activity in LNCaP cells (Pulukuri et al., 2007), we analyzed the effects of silencing PKD2 or PKD3 on HDAC1 expression in prostate cancer cells. Knockdown of PKD3 significantly enhanced nuclear HDAC1 expression in the absence or presence of PMA stimulation, whereas knockdown of PKD2 upregulated HDAC1 expression only in the presence of PMA treatment (Fig. 8A). As shown in Fig. 8B, PKD3 knockdown enhanced HDAC1 expression at all time points during PMA stimulation, suggesting that PKD3 suppresses the constitutive expression of HDAC1. Likewise, in DU145 cells, depletion of PKD3, much more so than depletion of PKD2, enhanced HDAC1 expression without any stimulation (Fig. 8C). Similarly, overexpression of PKD3 suppressed HDAC1 expression in HEK293 cells (supplementary material Fig. S4). Although the exact mechanism of HDAC1 suppression by PKD3 is not yet clear, PKD3, but not PKD2 (data not shown), constitutively interacted with HDAC1 in PC-3M cells (Fig. 8D).

To further investigate the possibility that PKD3 enhances uPA transactivation through removal of HDAC1 from the uPA promoter, we measured the ability of HDAC1 to bind to the uPA promoter using ChIP followed by real-time PCR. Knockdown of PKD3 enhanced HDAC1 binding to the uPA promoter by ∼3.2-fold in the absence of PMA treatment and ∼11.8-fold in the presence of PMA. Knockdown of PKD2 did not significantly change HDAC1 binding in the absence of PMA, but amplified it by nearly 2.1-fold in the presence of PMA (Fig. 8E). Although acetylation at Lys9 and/or Lys14 of histone H3, which binds to the uPA promoter, did not change significantly upon PKD3 depletion (supplementary material Fig. S5), we cannot exclude the possibility that PKD3 regulates histone acetylation through HDAC1. Collectively, these data demonstrate that PKD3 has a significant role in the regulation of HDAC1 gene expression and might activate the uPA promoter by suppressing HDAC1 expression.

To test the hypothesis that PKD2 or PKD3 contributes to uPA transactivation by suppressing HDAC1 expression and its subsequent binding to the uPA promoter, we co-transfected HDAC1 and PKD3 siRNAs into PC-3M cells and analyzed uPA transactivation with or without TNF-α stimulation. As expected, knockdown of HDAC1 alone significantly increased baseline uPA transactivation (P<0.01 versus siCTL) and this effect was reduced by TNF-α treatment. Moreover, in contrast to single silencing of PKD3, concomitant depletion of HDAC1 and PKD3 did not significantly decrease uPA transactivation compared with that in the control cells (P>0.05; Fig. 8F).

In the present study, we showed that both PKD2 and PKD3 contributed to prostate cancer cell invasion by activation of uPA signaling via the NF-κB pathway. Whereas PKD2 served mainly to promote nuclear translocation and Ser276 phosphorylation of p65, PKD3 contributed more to the regulation of Ser536 phosphorylation. Importantly, PKD3 and, to a lesser extent, PKD2 caused reduced expression of nuclear HDAC1 and decreased its binding to the uPA promoter. Conversely, either overexpression of p65 or silencing of HDAC1 could rescue the decrease in uPA transactivation resulting from PKD3 depletion. In summary, PKD2 and PKD3 might play partially overlapping, yet largely distinct, functional roles in prostate cancer cell invasion.

As shown in the current study, double knockdown of PKD2 and PKD3 diminished prostate cancer cell invasion to a greater extent than single knockdown of either protein, indicating that PKD2 and PKD3 have a cooperative role in tumor cell invasion. Moreover, data from RT-qPCR, western blotting, and gelatinase zymography demonstrated that both the expression of uPA signaling proteins, and the activity of uPA and MMP-9 were regulated by PKD2 and PKD3, albeit to different degrees, further implying that PKD2 and PKD3 cooperate to mediate prostate cancer cell invasion through a uPA-uPAR-MMP pathway, which has been reported to be pivotal for prostate cancer invasion (Li and Cozzi, 2007).

Although the PKD1-mediated NF-κB pathway has been shown to regulate cell survival in response to stimuli, such as H₂O₂ and UVB, in a variety of studies (Arun et al., 2011; Song et al., 2009; Storz and Toker, 2003), the current study provides the first evidence that PKD2- and PKD3-mediated NF-κB pathway can promote activation of uPA signaling to regulate prostate cancer cell invasion. PKD2 and PKD3 both appeared to be important for general NF-κB activation and transactivation of uPA by NF-κB. We further investigated the precise mechanisms through which PKD2 and PKD3 mediated NF-κB-dependent uPA transactivation. As expected, PKD2 was primarily responsible for p65 NF-κB nuclear translocation, a process that was only partially blocked by PKD3 depletion. This is plausible because PKD2 is mainly localized within the cytosol, whereas PKD3 shows nuclear distribution; perhaps only cytosolic PKD2 and PKD3 are responsible for the degradation of IκB and subsequent translocation of p65 to the nucleus. Consistent with these results, PKD2 was primarily responsible for IKK phosphorylation and subsequent IκB degradation, indicating that PKD2 has a major role in classical IKK-IκB-mediated NF-κB nuclear translocation. However, whereas the interaction between IKKβ and PKD2 was enhanced by PMA in the current study, elucidation of the mechanism through which PKD2 mediates IKK phosphorylation (directly or indirectly) will require further investigation. Although PKD3 also interacted with IKKβ, and this interaction was dramatically enhanced by PMA treatment, it remains to be determined whether nuclear or cytosolic PKD3 (or both) interacted with IKKβ. Given that PKD3 is primarily localized in the nucleus, the interaction of cytosolic PKD3 with IKKβ might have a similar, but less significant role, to that of PKD2, and the interaction of nuclear PKD3 and IKKβ could play a quite different, but important role(s), within the nucleus. Given that slight inhibition of p65 nuclear translocation does not appear to be sufficient to explain the dramatically decreased transactivation of uPA and subsequent uPA protein expression in response to PKD3 depletion, additional PKD3-mediated mechanisms of uPA transactivation might exist that are independent of p65 translocation.

The current study has also shown that PKD2 was primarily responsible for p65 Ser276 phosphorylation, whereas PKD3 was responsible for p65 Ser536 phosphorylation. Moreover, PKD3-mediated p65 Ser536 phosphorylation is important for the transactivation of uPA, which is strongly supported by the fact that constitutive activation of Ser536 rescues the downregulation of uPA transactivation caused by PKD3 silencing (Fig. 6C). However, Ser276 and/or Ser536 may not be directly phosphorylated by PKD2 or PKD3 given that the amino acid sequences surrounding Ser276 and Ser536 of p65 do not contain a classical PKD substrate motif ([V/I/L]xx[K/R]x[S/T]). Because our previous study has indicated that PKD3 might activate the p38 mitogen-activated protein kinase (MAPK) or phosphoinositol 3-kinase (PI3K)-Akt pathway in prostate cancer cells (Chen et al., 2008), it is plausible that PKD3 might phosphorylate p65 at Ser536 via p38 MAPKs and/or phosphoinositol 3-kinase (PI3K) and Akt. Therefore, the outcome of PKD2- and PKD3-mediated p65 phosphorylation in prostate cancer cell invasion remains to be determined. Most importantly, phosphorylation at Ser536 has been reported to orchestrate p65 activation independently of IκB degradation, because nuclear translocation of phosphorylated p65 (specifically at Ser536), but not total p65, has been shown to be unaffected by MG-132-dependent inhibition of IκB depletion (Sasaki et al., 2005). Collectively, this suggests that, whereas PKD2 might regulate canonical NF-κB activation through sequential activation of the pIKK-pIκB-IκB degradation p65 translocation cascade and Ser276 phosphorylation, PKD3 may mediate noncanonical NF-κB activation through induction of Ser536 phosphorylation. This could explain why depletion of PKD3 alone dramatically decreased uPA transactivation and expression and reduced the invasive potential of prostate cancer cells, with little effect on p65 nuclear translocation.

Because individual phosphorylation of Ser276 or Ser536 is responsible for transactivation of some, but not all, NF-κB target genes (Dong et al., 2008; Smale, 2010), it is possible that PKD2- or PKD3-dependent p65 phosphorylation at Ser276 and/or Ser536 might play a unique role in the transactivation of different genes (such as those encoding uPA, uPAR, or MT1-MMP).

Interestingly, our evidence that PKD3 suppressed the binding of HDAC1 to the uPA promoter might shed light on our understanding of an unidentified role of PKD3 within the nucleus. HDAC1, a member of the class I HDAC subfamily, has been reported to negatively regulate androgen receptor (AR)-dependent transcription of PSA (Gaughan et al., 2005; Glozak and Seto, 2007; Logan et al., 2006) and to suppress transactivation of NF-κB on the uPA promoter in LNCaP cells. Despite a variety of recent studies demonstrating that nuclear exclusion of class IIa HDACs is triggered by PKD-directed phosphorylation of multiple residues (Jensen et al., 2009; Matthews et al., 2006), the effects of PKDs on HDAC1, especially in the process of prostate cancer invasion, remain unknown. The current study has clearly demonstrated that PKD3 could constitutively interact with HDAC1 and suppress HDAC1 expression. The constitutive nuclear localization of PKD3 in PC-3M and DU145 cells might facilitate its interaction with HDAC1, which is exclusively found in the nucleus, potentially explaining why PKD3 has a greater effect than PKD2 on HDAC1 in untreated cells. Although the exact mechanisms of PKD2- and PKD3-mediated HDAC1 depletion remain to be determined, HDAC1 has been shown to be regulated by ubiquitin-proteasome-mediated degradation in response to multiple kinds of stimuli (TNF-α, IL-1, oxidants, etc.) (Vashisht Gopal and Van Dyke, 2006a; Vashisht Gopal et al., 2006b). Given that IKKβ has been reported to mediate TNF-α-induced HDAC1 degradation, and, in the current study, PKD3 interacted with both IKKβ and HDAC1 in PC-3M cells, PKD3 might act as a functional link between IKKβ and HDAC1, directing HDAC1 ubiquitylation and depletion through IKKβ. Ongoing studies in our laboratory will verify the interaction between PKD3 and HDAC1 in vitro and characterize additional partner(s) involved in this potential PKD3–IKKβ–HDAC1 complex. We also cannot exclude the possibility that PKD3 phosphorylates E3 ligases, such as CHFR, thus promoting HDAC1 degradation (Oh et al., 2009).

PKD3 not only suppressed the expression of HDAC1 in whole cell lysates and nuclear fractions, but also decreased the binding of HDAC1 to the uPA promoter region. Moreover, simultaneous silencing of HDAC1 and PKD3 rescued the downregulation of uPA transactivation caused by PKD3 depletion. These results clearly indicate that PKD3 might activate uPA transcription by decreasing HDAC1 protein expression and binding to the uPA promoter. Additionally, PKD3 might be pivotal for shifting the balance of uPA transactivation between NF-κB and HDAC1 at the epigenetic level, thus promoting the invasion and metastasis of prostate cancer cells. Our evidence for a negative role of HDAC1 in prostate cancer invasion is in accordance with that reported by Pulukuri et al. (Pulukuri et al., 2007), but conflicts with reports by Glozak and Seto (Glozak and Seto, 2007) and Kim et al. (Kim et al., 2011), which indicate that HDAC1 might have positive roles in prostate cancer migration and invasion through suppression of E-cadherin expression. However, our evidence demonstrating that PKD3 mediates HDAC1 depletion and the corresponding reduction in HDAC1 binding to the uPA promoter does not exclude the possibility of increased HDAC1 binding to the E-cadherin promoter. Consequently, the use of ChIP-seq technology to monitor alteration of HDAC1 binding to promoter regions of different genes (either tumor-promoting or tumor-repressing) could have predictive value, given that imprudent utilization of HDAC inhibitors to treat cancers might cause unpredicted side effects.

In conclusion, we have demonstrated that PKD2 might serve as a switch to turn on p65 nuclear translocation and activation through Ser276 phosphorylation, and that PKD3 might function to activate p65 via Ser536 phosphorylation while deactivating HDAC1 in highly invasive prostate cancer cells. These processes might all contribute to shifting of the epigenetic balance between NF-κB and HDAC1, initiating transactivation of uPA, and promoting cell invasion. While PKD2 regulates p65 activation through the canonical NF-κB pathway, PKD3 can also trigger its activation in a noncanonical manner (see schematic in supplementary material Fig. S6). Further studies are required to determine whether additional NF-κB targets are also controlled by this cooperative PKD2- and PKD3-mediated mechanism. Nevertheless, the current study provides novel insights into the partially redundant, but largely distinct, roles of different isoforms of the same kinase family in cancer cells.

PMA and all ultrapure reagents were from Promega (Madison, WI, USA) and Sigma (St Louis, MO, USA). uPA activity assay kits were purchased from Millipore (Billerica, MA, USA). Prolong Antifade Reagent was from Molecular Probes (Invitrogen, Carlsbad, CA, USA). Lipofectamine 2000 was from Invitrogen, and HilyMax transfection reagent was from Dojindo (Kamimashikigun, Kumamoto, Japan). The PKD2 Validated Stealth RNAi DuoPak was from Invitrogen (catalog no. 45-3192), and Silencer Selected Validated SiRNA PKD3 was from Ambion (Invitrogen, catalog no. 4390824). BioCoat Matrigel chambers were from BD Biosciences (Franklin Lakes, NJ, USA). The ChIP-IT Express Enzymatic Chromatin Immunoprecipitation Kit was from Active Motif (Carlsbad, CA, USA). All-in-One First-Strand cDNA Synthesis Kit and All-in-One qPCR Mix were from GeneCopoeia (Rockville, MD, USA). Antibodies used for immunoblotting and IP assays were as follows: the HDAC1 primary antibody for chromatin IP was purchased from Millipore; primary antibodies against PKD2 were from Bethyl Laboratories (Montgomery, TX, USA); uPA and MT1-MMP antibodies were from Lab Vision, Thermo Scientific (Kalamazoo, MI, USA); antibodies targeting PKD3, p65, phosphorylated p65 at Ser276 and Ser536, acetylated p65 at Lys310, IKKβ, phosphorylated IKKα/β, IκB, phosphorylated IκB at Ser32 and Ser36, and acetylated histone H3 at Lys9 and Lys14 were purchased from Cell Signaling Technology (Danvers, MA, USA); antibodies targeting uPAR, p65, and all unconjugated secondary antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA, USA); and Alexa-488- and 594-conjugated secondary antibodies were from Molecular Probes (Invitrogen).

PC-3M and DU145 cells were from the American Type Culture Collection and were cultured according to the manufacturer's recommendations. The siRNAs were transfected into cells using Lipofectamine 2000 according to the manufacturer's instructions. When HDAC1 siRNAs were transfected, both strands were pooled together. pEGFP-C2, pEGFP-PKD2, and pEGFP-PKD3 plasmids were transfected into cells using HilyMax according to the manufacturer's protocol. The following siRNA sequences were used: PKD2-1, 5′-CCUGAGUGUGGCUUCUACGGCCUUU-3′ (Invitrogen validated siRNA, catalog no. 45-3192); PKD2-2, 5′-AAUGACCUUAACUGCCACGUCCCGG-3′ (Azoitei et al., 2010); PKD3-1, 5′-GAACGAGUCUUUGUAGUAAtt-3′ (Silencer Selected Validated siRNA, catalog no. 4390824); PKD3-2, 5′-GAAAGUUCCACACACAUUU-3′ (Chen et al., 2011); HDAC1, 5′-CAGCGACUGUUUGAGAACCTT-3′ and 5′-CUAAUGAGCUUCCAUACAATT-3′ (Senese et al., 2007).

For the migration assay, we used confluent cultures (80–90%) in 12-well plates. Twenty-four hours after transfection with the indicated siRNAs, the cell monolayer was scratched with a 10-µl pipette tip, and the length of the cell-free area was monitored at the indicated times using light microscopy. Data are expressed as a percentage of the initial length at time zero. In vitro invasion assays were carried out in BD BioCoat Matrigel chambers according to the manufacturer's protocol.

Total RNA from each cell line was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. Reverse transcription and quantitative PCR (qPCR) were carried out using the All-in-One First-Strand cDNA Synthesis Kit and All-in-One qPCR Mix (GeneCopoeia) according to the manufacturer's protocol. The following RT-qPCR primers were used: MT1-MMP forward, 5′-CCTTGGACTGTCAGGAATGAG-3′ and reverse, 5′-GAGGGGTCACTGGAATGCT-3′; osteopontin forward, 5′-CGCAGACCTGACATCCAGT-3′ and reverse, 5′-GGCTGTCCCAATCAGAAGG-3′; uPA forward, 5′-TGACCCACAGTGGAAAACAG-3′ and reverse, 5′-TTGTCCTTCAGGGCACATC-3′; uPAR forward, 5′-GGTGACGCCTTCAGCATGA-3′ and reverse, 5′-CCCACTGCGGTACTGGACAT-3′; activin A forward, 5′-CGGGTATGTGGAGATAGAGG-3′ and reverse, 5′-GTCTTCCTGGCTGTTCCTGAC-3′; VEGF forward, 5′-TGCAGATTATGCGGATCAAACC-3′ and reverse, 5′-TGCATTCACATTTGTTGTGCTGTAG-3′; ubiquitin-C (internal control) forward, 5′-ATTTGGGTCGCGGTTCTTG-3′ and reverse, 5′-TGCCTTGACATTCTCGATGGT-3′.

Subcellular fractionation was conducted using the Qproteome Cell Compartment Kit following the manufacturer's recommendations. Western blotting and IP were performed as previously described (Lu et al., 2007). Conditioned medium was concentrated using Amicon Ultra-4 Centrifugal Filter Units (Millipore) and was subjected to SDS-PAGE for western blot analysis of secreted proteins.

DU145 and PC-3M cells were transiently transfected with the indicated siRNAs or plasmids, cultured for an additional 24 hours, serum starved for 24 hours, and stimulated with PMA for an additional 24 hours. The resulting conditioned medium was directly assayed using the Chemicon uPA Assay Kit following the manufacturer's protocol or concentrated using Amicon Ultra-4 Centrifugal Filter Units (Millipore) and separated on 10% SDS polyacrylamide gels containing 0.1% (w/v) gelatin under nonreducing conditions. Zymography was then performed as previously described (Cheng et al., 2007).

Cells were seeded on coverslips in 6-well plates and allowed to attach overnight. Immunofluorescent staining was performed as previously described (Lu et al., 2007). Fixed and permeabilized cells were incubated overnight at 4°C with primary antibodies, followed by incubation with Alexa-488- or Alexa 594-conjugated secondary antibodies for 1 hour at room temperature. Finally, coverslips were mounted onto glass slides with Prolong Gold Antifade reagent (after staining the nuclei with DAPI), and stained cells were imaged under a laser scanning confocal fluorescent microscope.

PC-3M and DU145 cells were co-transfected with the indicated siRNAs, uPA promoter reporter plasmid (uPA-luc) (Li et al., 2005) and internal control plasmid (pGL4.74[hRluc/TK], from Promega). Twenty-four hours after transfection, the cells were serum starved for 12 hours and subsequently stimulated with PMA (100 nM) for 16 hours. Firefly and Renilla luciferase activities of the stimulated cells were measured using the Dual-Luciferase Reporter Assay System (Promega).

ChIP was performed using the ChIP-IT Express Enzymatic Chromatin Immunoprecipitation Kit (Active Motif) with chromatin isolated from 1×10⁶ cells, according to the manufacturer's protocol. Quantitative PCR of co-immunoprecipitated genomic DNA fragments was performed with the following promoter-specific primers: uPA-NF-κB sense, 5′-GAGGGGGCGGAAGGGGAGAA-3′ and uPA-NF-κB antisense, 5′-TGTGGTCAGTTTTGTTTGGATTTG-3′ (Cicek et al., 2009); and nonpromoter sequence primers of the uPA-encoding gene: uPA-EXO11 sense, 5′-TTGTATCTTTGGCGTCACAGG-3′ and uPA-EXO11 antisense, 5′-CATTCTCTTCCTTGGTGTGAC-3′ (Ibañez -Tallon et al., 2002).

All statistical analyses were conducted using GraphPad Prism V software. A P-value less than 0.05 was considered statistically significant.

The authors appreciate the help of those who provides us with plasmids: the 2×NF-κB luciferase reporter was a kind gift from Dr Yong Jiang at Southern Medical University; the uPA-luc reporter plasmid was a kind gift from Dr Shuang Huang at Georgia Health Sciences University; and p65WT, S536D, and S536A were kind gifts from Dr Carl Y. Sasaki at the Laboratory of Immunology, NIA, National Institutes of Health. The authors declare no conflicts of interest.