Protein Kinase D1 Suppresses Epithelial-to-Mesenchymal Transition through Phosphorylation of Snail

A large body of evidence suggests that the epithelial-mesenchymal transition (EMT) is a key neoplastic program that is often activated during cancer invasion and metastasis (1). During transient EMT process, epithelial cells acquire mesenchymal cell properties, including morphologic changes, expression of mesenchymal markers, and, most importantly, the ability to migrate and invade. A variety of extracellular signals, including transforming growth factor-β (TGF-β), epidermal growth factor (EGF), fibroblast growth factors, hepatocyte growth factor, and insulin-like growth factor, have been shown to induce EMT during tumor progression in an autocrine or paracrine manner (2). Recent studies show that induction of EMT in immortalized human mammary epithelial cells or transformed cells results in the acquisition of mesenchymal traits and in the gain of epithelial stem cell–like properties (3, 4), suggesting that EMT in tumor progression may contribute to generation of cancer stem cells with tumorigenic potential and tumor heterogeneity in addition to migration and invasion.

E-cadherin is a well-recognized pivotal molecule in maintenance of epithelial cell-cell adhesion. E-cadherin binds to β-catenin to form a protein complex that links E-cadherin to actin and microtubule cytoskeleton and regulates nuclear availability of β-catenin for its transcriptional activity. Seminal studies have shown antiproliferation, anti-invasion, and antimetastasis roles for E-cadherin (5, 6). In cancer, loss of E-cadherin function has been implicated in progression and metastasis of numerous malignancies (7). Multiple mechanisms for loss of E-cadherin expression in malignant tumors have been proposed, including transcription repression, genetic mutation, epigenetic silencing, and proteolytic processing (5). Several transcription factors, including Snail (8, 9) and Twist (10), have been implicated in the repression of E-cadherin and induction of EMT (7). The repression activity of Snail is regulated by multiple signal pathways. Glycogen synthase kinase 3β (GSK3β) phosphorylates six serine residues of Snail and regulates Snail protein stability and subcellular localization (11, 12). Inflammation-activated NF-κB signaling also stabilizes Snail protein and promotes cell migration and invasion (13). In addition, p21-activated kinase 1 (PAK1) is able to phosphorylate on Ser²⁴⁶ and increase Snail repression activity (14).

Protein kinase D1 (PKD1), a serine/threonine kinase, was originally described as a novel μ isoform of the protein kinase C (PKC) family, as it shares two cysteine-rich domains (C1a and C1b) that bind phorbol esters and diacylglycerol as in the PKC family. Unlike other members of the PKC family, PKD1 also has a unique pleckstrin homology (PH) domain (15), and the catalytic domain of PKD1 is most closely related to calcium calmodulin–dependent kinases (16). Homozygous PKD1 deletion using a CAG-Cre transgene in mice caused embryonic lethality (17), suggesting that the PKD1 function is vital for mouse development, and the other two closely related PKD family members, PKD2 and PKD3, cannot replace PKD1 function. PKD1 has been shown to be downregulated in prostate, breast, and stomach cancers (18–20). Epigenetic methylation of PKD1 was found as a major cause of PKD1 inactivation in gastric cancer (19). PKD1 is capable of influencing major fundamental characteristics of cancer cells, including proliferation, motility, and invasion (19–22). Here, we describe the inhibitory role of PKD1 in cancer growth and metastasis by influencing EMT through Snail phosphorylation.

The full-length Snail gene was amplified by PCR from Snail cDNA (accession no. BC012910) with primers AAAAGGGATCCCTATGCCGCGCTCTTTCCTCGTC (forward) and AAAAGCTCGAGGCGGCGGGGACATCCTGAGCAGC (reverse). The PCR product was inserted into pEGFP-N3 (BglII-SalI), pGAD-T7 (BamHI-XhoI; Clontech), and pGEX-5X-1 (BamHI-XhoI; GE Healthcare) vectors in frame. Ser¹¹ to valine (S11V) and Ser¹¹ to glutamic acid (S11E) mutants were made via QuikChange kit (Stratagene) with primers (forward only) GCGCTCTTTCCTCGTCAGGAAGCCCGTCGACCCCAATCGGAAGCCTAACTACAGCG (S11V) and GCGCTCTTTCCTCGTCAGGAAGCCCGAGGACCCCAATCGGAAGCCTAACTACAGCG (S11E). The sequence used for RNA interference targeting of Snail is CCAGGCTCGAAAGGCCTTCAA. The sequences for targeting Twist (siTwist3 and siTwist5) were the same as described by Yang and colleagues (10). The sequences for targeting E-cadherin and PKD1 were described previously (22). All these oligos were inserted into a homemade short hairpin RNA (shRNA) expression vector, KS-U6 (23). HA-tagged 14-3-3σ, 14-3-3ϵ, and 14-3-3ζ isoforms were acquired from Addgene (http://Addgene.org).

Prostate cancer DU145 and LNCaP cell lines were purchased from the American Type Culture Collection (ATCC) in 2001. Prostate C4-2 cell line was purchased from Urocor, Inc. in 2001. We authenticated LNCaP and C4-2 cells using microsatellite markers D1S 1612, D2S 1399, and D2S 1363. The microsatellite analysis also confirmed that C4-2 cells are derived from LNCaP cells. MCF7 and 184A1 were obtained in 2007 from ATCC and used within 6 months after purchase. Prostate primary epithelial cell line RWPE1 was purchased from ATCC in 2008 and used within 6 months after purchase. RWPE1 was grown in keratinocyte serum-free medium (Invitrogen) supplied with bovine pituitary extract (Invitrogen) and human recombinant EGF (2 ng/mL; PeproTech). Mammary primary epithelial cell line 184A1 was grown in HuMEC Ready medium (Invitrogen). The prostate cancer cell lines LNCaP, which has less metastatic potential, and metastatic cell lines DU145 and C4-2 (derived from LNCaP cell line) were cultured in DMEM plus 10% fetal bovine serum (FBS). Because the C4-2 expresses low level of PKD1 (18), a stable C4-2 cell line that expresses PKD1-GFP (green fluorescent protein) was established by G418 selection (C4-2/PKD1) and cultured in DMEM plus 10% FBS and sodium pyruvate supplement (24).

A synthetic peptide consisting of the first 15–amino acid residues with a phosphoserine at position 11 of human Snail was used to generate the phosphoserine-specific antibody (pS11) by a commercial contractor. Other antibodies used in the present study are listed in Table 1, which appears below.

Glutathione S-transferase–tagged wild-type (WT) or S11V mutant of Snail was expressed and purified from Escherichia coli. Purified PKD1 (0.2 μg; Cell Signaling Technology) was mixed with 5 μg of purified Snail protein in the manufacturer's recommended buffer plus 10 μmol/L ATP and 10 μCi [γ-³²P]ATP (3,000 Ci/mmol) to a volume of 50 μL. Reaction was carried out at room temperature for 1 hour.

The C4-2 and C4-2/PKD1 cells were transfected by firefly luciferase for in vivo bioluminescence imaging. Following approval of animal experimental protocol by Institutional Animal Care and Use Committee, C4-2 or C4-2 PKD1 cells (1 × 10⁶ in 0.2 mL PBS) were injected s.c. into groin of 8-week-old male NCR nude mice (n = 12 each group). Tumor growth was monitored weekly by bioluminescence imaging and tumor size measurements to 12 weeks.

Knockdown of E-cadherin by shRNA induces EMT and invasiveness in breast cancer cells (25). In spite of low expression of E-cadherin in metastatic DU145 cells, knockdown of E-cadherin increases expression of mesenchymal markers N-cadherin, fibronectin, and vimentin (Fig. 1A1). Meanwhile, knockdown of PKD1 in DU145 cells generates an EMT protein expression pattern similar to knockdown of E-cadherin (Fig. 1A1). The data corroborate previous evidence that invasiveness of prostate cancer cell lines is also increased when PKD1 or E-cadherin is knocked down (22, 26). In contrast, the expression of these mesenchymal markers is inhibited in metastatic C4-2 cells that stably expressed PKD1 (C4-2/PKD1 cells), whereas the expression of E-cadherin is upregulated (Fig. 1A2). Interestingly, β-catenin protein is increased following upregulation or downregulation of PKD1 or E-cadherin in DU145 and C4-2 cells. Similar results have also been reported in colon cancer cells (27), indicating complicated mechanisms of regulation of β-catenin. Immunofluorescence studies confirm the Western blotting results (Fig. 1B). Vimentin (Fig. 1B1) and N-cadherin (Fig. 1B2) are expressed exclusively in C4-2 cells, and E-cadherin (Fig. 1B3) is expressed only in C4-2/PKD1 cells. Meanwhile, membranous β-catenin accumulates in C4-2/PKD1 cells with minor expression in C4-2 cells (Fig. 1B4). Prostate cancermetastasis repressor KAI1 (CD82) is downregulated during metastasis (28). Overexpression of PKD1 restores KAI1 expression (Fig. 1B6). Finally, the mRNA level of E-cadherin is significantly elevated in C4-2/PKD1 cells, indicating that PKD1 increases E-cadherin expression at a transcriptional level (Fig. 1A3). A luciferase reporter assay (8) shows that E-cadherin promoter activity is upregulated in a dosage-dependent manner by PKD1 (Fig. 1C).

PKD1 has been previously shown to have antiproliferation and antimotility properties in vitro (22, 24, 26). To confirm the tumor suppressor properties of PKD1 in vivo, metastatic C4-2 or derivative C4-2/PKD1 cells were injected s.c. into groin of nude mice. Ectopic PKD1 expression greatly repressed tumor incidence and metastasis (Fig. 1D); 3 of 12 (25%) mice developed tumor in C4-2/PKD1 group compared with 11 of 12 (92%) mice in the C4-2 control group with one metastasis incidence (Fig. 1D). Overall, the data suggest that PKD1 represses EMT, tumor growth, and metastasis in prostate cancer cells.

A few transcription factors have been known to repress the expression of E-cadherin and to induce mesenchymal markers (7). Twist (29) and Snail (30, 31) have been reported to play a role in prostate cancer metastasis. As a first step, we explored whether knockdown of Snail or Twist could reproduce an EMT protein expression pattern similar to overexpression of PKD1. As seen in Fig. 2A, knockdown of Snail or Twist by shRNA individually only partially represses or induces expression of fibronectin or E-cadherin, respectively, whereas combining the two shRNAs achieves a greater repression or induction, suggesting that expression of fibronectin and E-cadherin is controlled by both Snail and Twist. In contrast, the expression of N-cadherin and vimentin, both repressed by PKD1 (Fig. 1A2), is solely controlled by Snail. The finding is consistent with results cited in published literature indicating that knockdown of Twist does not alter expression of vimentin in prostate cancer (29). Based on these experiments, we concluded that inhibition of Snail, but not Twist, generates a similar protein expression pattern as overexpression of PKD1. Therefore, Snail is the most likely target for PKD1 in prostate cancer cells.

PKD1 does not inhibit Snail at transcription level (Fig. 1A3). To detect protein interaction between PKD1 and Snail, individual domains of PKD1 (i.e., C1a, C1b, and PH) and catalytic domains were tested for interaction with full-length Snail using yeast two-hybrid assay. Snail strongly binds to PKD1 catalytic and weakly to the C1a domain (Fig. 2B). In addition, the two proteins are present in the same protein complex that can be coprecipitated (Fig. 3B). A search of PKD1 consensus phosphorylation motif LxRxxS/T (32, 33) found a conserved sequence at the NH₂ terminus of Snail (₆-LVRKPSDP-₁₃) as potential PKD1 phosphorylation site (Fig. 2C). WT, but not S11V mutant, Snail can be phosphorylated by PKD1 in vitro (Fig. 2C). These results strongly suggest that S11 on Snail is phosphorylated by PKD1.

A phosphospecific antibody (pS11) that recognizes the phosphorylated Ser¹¹ on Snail was developed and used to identify Ser¹¹ phosphorylation in vivo. The pS11 antibody detects a moderate signal intensity when WT Snail is expressed in 293T cells (Fig. 2D, lane 2), suggesting that intrinsic kinase(s) can phosphorylate Ser¹¹. Coexpression of PKD1 strongly enhances the pS11 signal intensity (Fig. 2D, lane 3). In contrast, coexpression of S11V mutant with PKD1 totally abolishes the pS11 detection, confirming the specificity of the pS11 antibody (Fig. 2D, lane 4). When WT Snail is coexpressed with kinase-dead (KD) PKD1, the pS11 antibody signal is greatly diminished (Fig. 2D, compare lanes 3 and 5). These results strongly suggest that the Snail S11 is an in vivo phosphorylation site for PKD1.

To study the consequence of PKD1 phosphorylation on Snail, GFP-tagged Snail nonphosphomimetic and phosphomimetic mutants (S11V and S11E, respectively) were transfected into the normal mammary epithelial cell line 184A1 (Fig. 3A) or the normal prostate epithelial cell line RWPE1 (Supplementary Fig. S1A). The WT Snail mainly localizes to the nucleus. Adding proteasomal inhibitor MG132 stabilizes Snail in the cytoplasm, suggesting that Snail undergoes proteolysis in cytoplasm (11). S11E and S11V are distributed in the cytoplasm and nucleus, respectively, regardless of treatment with MG132, indicating that the S11 phosphorylation affects Snail nuclear localization and protein stability. However, when transfected into prostate cancer cell lines LNCaP and C4-2 (Supplementary Fig. S1B), breast cancer cell line MCF7, and colon cancer cell line SW480 (data not shown), the three Snail isoforms predominately localized in the nuclei, suggesting that tumor cells may have altered nuclear export mechanisms.

The Snail S11 phosphorylation also creates a potential 14-3-3 binding motif of RxxS*xP (where S* is phosphorylated serine residue; Fig. 2C; ref. 34). To investigate possible involvement of 14-3-3 proteins in Snail nuclear export, each of the seven mammalian 14-3-3 isoforms was transiently cotransfected with Snail and PKD1 into 293T cells. His-tagged 14-3-3 proteins were pulled down using Ni²⁺ beads. Snail is coprecipitated with five 14-3-3 isoforms (γ, η, σ, θ, and ζ), and PKD1 is effectively pulled down by three 14-3-3 isoforms (γ, σ, and θ; Fig. 3B1). Because the 14-3-3 isoforms are highly conserved and have redundant functions (35), it is not surprising that multiple 14-3-3 isoforms are able to bind to Snail. The σ isoform is of particular interest because it is specifically expressed in epithelial cells (36–38), and its expression is frequently lost during breast and prostate cancer progression (39, 40). Based on the binding study and an established role in prostate cancer, we used 14-3-3σ as a model to study Snail binding. As seen in Fig. 3B2, in the presence of 14-3-3σ, WT Snail is selectively coprecipitated by Ni²⁺ beads with WT PKD1 (lane 2), but not with KD PKD1 (lane 3). In contrast, Snail-S11V mutant is poorly coprecipitated by Ni²⁺ beads even in the presence of WT PKD1 (lane 4). These data strongly suggest that S11 phosphorylation is critical for binding to 14-3-3σ and PKD1 is able to phosphorylate S11 in vivo. Snail protein is also pulled down with 14-3-3η independent of PKD1 (Fig. 3B1). Note that PKD1 has nonspecific binding to Ni²⁺ beads (Fig. 3B2, lane 6). Hence, we suggest that PKD1 band in the η lane is due to nonspecific binding or that other factors may be also involved in Snail binding to 14-3-3 proteins.

The 14-3-3 proteins are known to shuttle phosphorylated proteins among different subcellular locations (35). To study whether nuclear export of Snail by PKD1 is mediated by 14-3-3 proteins, C4-2 cells were transfected by GFP-tagged WT Snail with either 14-3-3σ, 14-3-3ϵ (which was not coprecipitated with Snail; Fig. 3B1), or 14-3-3ζ (which was coprecipitated with Snail, but not efficiently with PKD1; Fig. 3B1). 14-3-3ϵ largely localizes to the cytoplasm, 14-3-3ζ is present in both the cytoplasm and nucleus, whereas Snail is nuclear, regardless of the presence of WT or KD PKD1 (Fig. 3C). 14-3-3ζ and Snail colocalize to the nucleus, with a minor portion of Snail seen in the cytoplasm. In contrast, the WT Snail is present in the cytoplasm and nucleus with 14-3-3σ in the presence of WT PKD1, but not KD PKD1. On the other hand, the S11V mutant is not exported from the nucleus even in the presence of 14-3-3σ and WT PKD1, suggesting that S11 phosphorylation by PKD1 may be a critical event in nuclear export of Snail. Compared with nuclear localization of S11E mutant in the absence of 14-3-3σ (Supplementary Fig. S1B), cotransfection with 14-3-3σ alters Snail-S11E mutant to the cytoplasm and nucleus, which is dependent on PKD1 kinase activity (Fig. 3C). The data clearly show that PKD1 phosphorylation of Snail at Ser¹¹ results in Snail nuclear export by 14-3-3σ. In addition, activation of PKD1 by bryostatin 1 can increase the colocalization of Snail and 14-3-3σ in the cytoplasm, particularly in the perinuclear areas (white arrow in Supplementary Fig. S1C), a pattern similar to the accumulation of activated PKD1 (41).

Next, we compared the subcellular localizations of endogenous Snail in C4-2 and C4-2/PKD1 cells. Immunofluorescence staining confirms that Snail in C4-2 cells predominantly locates to the nucleus (Fig. 3D1, arrowheads), whereas Snail in C4-2/PKD1 cells distributes more diffusely. Western blots using subcellular fractions confirm that the majority of Snail localizes in the nuclei of C4-2 cells or in the cytoplasm of C4-2/PKD1 cells (Fig. 3D2).

A previous study showed that exogenous WT Snail was not able to efficiently repress endogenous E-cadherin expression in breast cancer MCF7 cells and failed to induce expression of EMT markers (11). We selected this system to evaluate Snail-WT, Snail-S11E, and Snail-S11V functions. Stable transfection of S11V mutants is able to more effectively repress E-cadherin expression than WT or S11E mutants as shown by immunofluorescence study (Fig. 4A1), which is further qualitatively confirmed by Western blots (Fig. 4B) and quantitatively measured by E-cadherin promoter reporter assay (Fig. 4C). Induction of expression of vimentin is seen only in S11V cells, but not in WT or S11E cells (Fig. 4A2). These data suggest that the nonphosphorylated Snail is a more effective repressor for E-cadherin expression and inducer for EMT.

Mani and colleagues (3) showed that induction of an EMT in human mammary epithelial cells results in the acquisition of mesenchymal traits as well as stem cell–like properties. The MCF7 cells carrying the S11V mutant show fibroblast-like morphology and remain scattered instead of being closely aggregated as seen in the WT or S11E cells (Fig. 4A3). To further characterize these cell lines, we used flow cytometry to analyze cell surface markers; CD44^hiCD24^lo population is regarded as stem cell–like (3). Expression of S11V mutant dramatically shifts the CD44^loCD24^hi population to CD44^hiCD24^lo (Fig. 4D). The data strongly suggest that regulation of S11 phosphorylation is critical for efficient induction of EMT, and support the concept that cancer cells undergoing EMT share stem cell–like properties.

Because the reduction in PKD1 expression correlated with the downregulation of E-cadherin expression in the cell line–derived EMT model, we asked whether such a correlation exists in tumor specimens from prostate cancer patients. We searched publicly available DNA microarray database derived from human prostate cancers at the National Center for Biotechnology Information's Gene Expression Omnibus. Two data sets, GDS1439 and GDS2545, contain both PKD1 and E-cadherin in normal/benign, primary, and metastatic samples. In GDS1439 data set, concurrent downregulation of PKD1 and E-cadherin is observed in metastatic prostate cancer samples (Fig. 5A1). The difference in expression of PKD1 and E-cadherin between benign/primary and metastatic samples is statistically significant (P < 0.05). The expression of PKD1 correlates directly with expression of E-cadherin (r = 0.57, R² = 0.76). In the other data set, GDS2545 (Fig. 5B), E-cadherin and PKD1 show a trend toward lower expression in metastatic samples, although the difference is not statistically significant compared with normal, normal adjacent to tumor, and primary tumor samples. Interestingly, in normal (GDS2545) or benign (GDS1439) samples, expression of PKD1 is restricted within a narrow range, suggesting that PKD1 expression level is tightly controlled. In metastatic tumors, about 68% to 83% (17 of 25 in GDS2545 and 5 of 6 in GDS1439) samples have lower PKD1 expression than the average PKD1 level in normal samples. In contrast, expression of E-cadherin in normal samples is spread over a wide range (Fig. 5A2), implying that alteration of E-cadherin alone may not be sufficient to cause cancer (42) and expression of E-cadherin can be independent of PKD1. This is consistent with the role of PKD1 in regulation of E-cadherin through repression of Snail, which is presumed to be inactive in normal prostate luminal cells (43).

We describe that PKD1 controls cell-cell adhesion in epithelial cells through positive regulation of E-cadherin transcription via inhibitory phosphorylation of transcription factor Snail. We have dissected the exact molecular mechanisms by which PKD1 regulates Snail activity. We found that Ser¹¹ of Snail is phosphorylated by PKD1 and induces 14-3-3σ binding to the phosphorylated Snail, resulting in nuclear export. The Ser¹¹ residue locates at the end of SNAG domain, which is known for the recruitment of Sin3A–histone deacetylase 1/2 complex to repress E-cadherin (44). We found that the phosphorylation status of S11 on Snail is important for Snail function as an E-cadherin repressor. The nonphosphorylated Snail is sufficient to induce EMT and to confer cancer cells with stem cell–like traits. The amino acid sequence around S11 consists of a consensus PKD1 phosphorylation motif, which is conserved only in mammalian Snail1 family, but not in Snail2/Slug family, suggesting that the two Snail families have different functions and regulatory controls. Our results also suggest a molecular function of 14-3-3σ in tumor metastasis. As a tumor suppressor, 14-3-3σ is a p53-responsive gene and loss of 14-3-3σ protein frequently occurs in prostate and breast cancer (45). We show that 14-3-3σ, unlike 14-3-3ϵ or 14-3-3ζ, has specific affinity to bind and export phosphorylated Snail from the nucleus (Fig. 3), and thereby prevents Snail from repressing E-cadherin expression. A cartoon depicting Snail S11 phosphorylation and function is shown in Fig. 5C.

Phosphorylation of Snail by other kinases, including Wnt/GSK3β (11, 12), PAK1 (14), and PKD1, at different sites also results in alteration of Snail subcellular locations and protein stability. Thus, Snail, as a major inducer of EMT, is at the crossroads of many signaling pathways. It is yet unclear how these kinases coordinate under physiologic conditions. PKD1 and GSK3β seem to share similar functions and regulated cellular proliferation and motility. Besides phosphorylation and nuclear export of Snail, both of them regulate β-catenin (26), which induces cell migration, proliferation, and oncogenesis in a coordinated fashion. We speculate that PKD1, GSK3β, and other signaling pathways crosstalk to each other to coordinate cellular behavior.

The inhibitory role of PKD1 in EMT may involve mechanisms other than Snail phosphorylation. Drosophila has a single PKD homologue, whose expression during late embryogenesis is restricted to ectodermal derivatives (46). Vein phenotypes caused by overexpression of KD PKD1 in wing discs are indistinguishable from those caused by hyperactive dpp (47), the mammalian homologue of TGF-β, a major signal leading to EMT, suggesting a possible interaction of the PKD and TGF-β pathway. The EMT is characterized by loss of apical-basal cell polarity, which is essential for epithelial cell function. Mammalian Par-1 (partitioning defective) kinase family is essential for determining asymmetrical cell division and polarized cell growth, which is phosphorylated by PKD1 at a conserved serine residue and excluded from cell polarity of the Par-3/Par-6/aPKC complex in epithelial cells (48). Onder and colleagues (25) showed that β-catenin is necessary for mediation of loss of E-cadherin–induced EMT. Because PKD1 interacts with and inhibits β-catenin–mediated transcription (26), this interaction may also play a role in the EMT process. A recent elegant study showed that androgen signaling axis induces EMT and invasive phenotypes of prostate cancer cells (49). The cross-talk between PKD1 and androgen receptor is known (50), suggesting complex protein-protein interplay among PKD1, β-catenin, and androgen receptor, leading to EMT. Based on established or putative complex molecular interaction and in vivo data, our results strongly suggest that PKD1 is a novel tumor and metastasis suppressor, which at least partly functions through phosphorylation of Snail.

No potential conflicts of interest were disclosed.

We thank Drs. L.R. Languino (University of Massachusetts Medical School), H. Fu (Emory University), D. Cantrell (University of Dundee, Dundee, United Kingdom), and A. García de Herreros (Universitat Pompeu Fabra, Barcelona, Spain) for reagents.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.