Loss of PKCδ Induces Prostate Cancer Resistance to Paclitaxel through Activation of Wnt/β-Catenin Pathway and Mcl-1 Accumulation

Prostate cancer is the leading cause of cancer-related death among men in Western countries. Castration therapy remains the most widely used treatment for locally advanced prostate cancer. However, tumors progress into a castration-resistant prostate cancer (CRPC), which is prone to metastasize to distant sites. In this case, the treatment of the advanced disease is based on taxanes (1). Even though paclitaxel is active in most cases of CRPC, the first chemotherapeutic option is docetaxel combined with prednisone. Unfortunately, response rates of CRPC to chemotherapeutic drugs are low. As such, there is a real need for the development of new therapeutic strategies for CRPC (2, 3). Paclitaxel and other taxanes interact with β-tubulin, blocking normal mitotic spindle assembly and activating the spindle assembly checkpoint (SAC). When the SAC is active, BubR1 is bound to Cdc20, which prevents the APC/C-Cdc20–mediated degradation of cyclin B1 and Pttg1 and generates an anaphase-inhibitory signal that triggers mitotic arrest. After a prolonged mitotic arrest, cells either die or exit mitosis without cytokinesis into a tetraploid G₁ state through a mechanism named slippage, which is characterized by the degradation of cyclin B1 and Pttg1 despite SAC activation. Following slippage, cells either die, arrest in G₁, or initiate a new round of cell cycle. Cell death normally occurs via the intrinsic apoptotic pathway, where Bcl-2 family proteins with antiapoptotic (Bcl-2, Bcl-xL, Mcl-1) and proapoptotic (Bax, Bak) members play an essential role. It has been described that paclitaxel sensitivity is determined by the antiapoptotic proteins Mcl-1 and Bcl-xL (3–5).

The protein kinase C (PKC) family of Ser/Thr kinases regulates several cellular functions such as cell cycle, cell survival, malignant transformation, and apoptosis (6). It is generally believed that the activation of PKC isoforms contributes to cancer, but downregulation rather than activation of PKCδ has been associated with tumor progression (7). While some studies point to a role of PKCδ in cell survival and the antiapoptotic response, most studies indicate that the PKCδ isoform promotes suppression of cell proliferation and apoptosis induction. PKCδ is activated by genotoxins, oxidative stress, and death receptors, and its inhibition blocks apoptosis induced by a variety of stimuli in many cell types (6). In the case of prostate cancer, most reports indicate that PKCδ expression is required to induce apoptosis by anticancer drugs or death receptors (8–14). Concretely, in a previous report, we documented that PKCδ silencing induced apoptosis resistance to phenylethyl isothiocyanate and anti-Fas or paclitaxel treatments in PC3 and LNCaP cells, whereas the overexpression of PKCδ improved apoptosis induced by these treatments in PC3 cells. Moreover, we observed the loss of PKCδ in poorly differentiated human prostate cancers and suggested that this could be related to apoptosis resistance, tumor progression, and a less favorable prognosis (15). Thus, we sought to investigate how PKCδ absence impairs paclitaxel-induced apoptosis in prostate cancer cells and elucidate the signaling pathways involved. Also, we aimed to identify new therapeutic strategies to improve the sensitivity to taxanes in the absence of PKCδ.

PC3 and LNCaP prostate cancer cell lines validated by isoenzymes and HLA-DP beta, respectively, were purchased from the Interlab Cell Line Collection in the year 2013. Cells were initially grown and multiple aliquots were cryopreserved. For all experiments, cells were used within 6 months after resuscitation. Cell lines were routinely grown in RPMI1640 supplemented with 10% FBS, 50 U/mL penicillin, 50 μg/mL streptomycin, 10 mmol/L HEPES buffer, and 1 mmol/L glutamine in a 37°C, humidified incubator under 5% CO₂. Cells were harvested by trypsinization.

siRNA transfections were carried out using the DharmaFECT 2 Reagent (GE Dharmacon) according to the manufacturer's instructions. The PKCδ-specific siRNAs and the control-negative siRNA were validated pools from GE Dharmacon (ON-TARGET plus Nontargeting pool D-001810 and ON-TARGET plus Human PRKCD siRNA-SMART pool L-003524). siRNAs were used at 50 nmol/L. Cells were subjected to different treatments 24 hours after silencing.

Stock solutions of paclitaxel (Calbiochem), XAV939 (Selleck Chemicals), and LY294002 (Calbiochem) were prepared at 10 mmol/L in DMSO (Calbiochem) and stored frozen. DMSO was added to untreated cells. For sequential treatments, cells were treated with the first drug during 8 hours and then, the second drug was added during 40 hours. Caspase activation and cleavage of PARP were assessed by Western blotting.

Mouse monoclonal anti-human PARP (1:500) and rabbit polyclonal anti-Bax (1:3,000), anti-Bak (1:3,000), and anti-cyclin D1 (1:2,000) were available from BD Biosciences. Mouse monoclonal anti-Bcl-xL (1:200) and anti-Pttg1 (1:500), rabbit polyclonal anti-PKCδ (1:3,000), anti-Mcl-1 (1:1,000), anti-cyclin B1 (1:1,000), and anti-phospho-histone-H3^Ser10 (1:1,000) antibodies were from Santa Cruz Biotechnology. Mouse monoclonal anti-β-actin (1:20,000) and anti-c-Myc (1:500) were from Sigma. Rabbit polyclonal anti-cleaved caspase-9 (Asp 315; 1:500), anti-cleaved caspase-3 (Asp 175; 1:500), phospho-Mcl-1^Ser159 (1:500), anti-phospho-GSK3β^Ser9 (1:1,000), and anti-phospho-Akt^Ser473 (1:1,000) were from Cell Signaling Technology. Rabbit anti-BubR1 (1:3,000) antibody was from Bethyl Laboratories. Rabbit polyclonal anti-Aurora kinase A (AurKA; 1:5,000) was from Novus Biologicals. Rabbit polyclonal anti-β catenin (1:15,000) was from Lab Vision Neomarkers and mouse monoclonal anti-active β-catenin (1:500) was from Millipore.

Western blots were carried out as described previously (5, 15). The experiments were performed at least three times and densitometric analysis was performed using QuantiScan software (Biosoft). Arbitrary densitometric units of the proteins of interest were corrected for those of β-actin. Data comparing differences between two conditions were statistically analyzed when indicated, using paired Student t test. Statistical analyses were performed with Prism 6 (GraphPad). Differences were considered as significant when P < 0.05.

FISH was performed as described previously (5). At least 100 cells were counted to calculate the percentage of cells with normal ploidy and higher ploidy in each condition. Data comparing differences between two conditions were statistically analyzed using paired Student t test with Prism 6.

Eighty prostate carcinomas were selected from radical prostatectomies with the approval of the institutional ethical board. IHC was performed as described previously (15). Antigen retrieval was performed with 1 mg/mL trypsin, 15 minutes at 37°C for PKCδ; with 4 N HCl for 15 minutes at room temperature followed by 1 mg/mL trypsin for 15 minutes at 37°C for active β-catenin; or with 1 mmol/L EDTA (pH 9.0) in a microwave for AurKA, phospho-GSK3β^Ser9, and Mcl-1. Sections were incubated overnight at 4°C with the following primary antibodies: anti-PKCδ (Abcam, 1:400), anti-active β-catenin (Millipore, 1:200), anti-AurKA (Novus, 1:700), anti-phospho-GSK3β^Ser9 (Cell Signaling Technology, 1:25), and anti-Mcl-1 [Santa Cruz Biotechnology, (1:1,200)]. Immunostains were scored as low (<25%) or high expression (≥25%) according to the extent of positive cells. Tumors were graded using the Gleason score (2–10) and classified as low (2–6) or high (7–10) Gleason score tumors. Correlations between protein expression and Gleason score or between immunohistochemical expression of cited proteins were analyzed by χ² or Fisher exact test using SPSS software (IBM). Differences were considered as significant when P < 0.05.

To assess whether PKCδ has a role in cell-cycle distribution after paclitaxel treatment, PKCδ was silenced and cells were treated with 2.5 μmol/L paclitaxel during 48 hours. We then analyzed cyclin B1, Pttg1, and phopho-histone-H3^Ser10 proteins that accumulate during mitotic arrest as well as BubR1 protein, indicative of active SAC. We have shown previously (5) that paclitaxel-treated PC3 cells during 24 hours are arrested in mitosis. However, after 48 hours of paclitaxel treatment, most cells have exited mitosis without cytokinesis through slippage. In concordance with these results, we observed that a slippage process occurred in siRNA control PC3 cells following 48 hours of drug treatment, as showed by the decreased Pttg1, cyclin B1, and BubR1 protein levels. However, in paclitaxel-treated siRNA PKCδ PC3 cells, we observed an increased mitotic arrest as indicated by the higher levels of cyclin B1, Pttg1, and phospho-histone-H3^Ser10, and activation of the SAC, as indicated by increased BubR1 levels in comparison with paclitaxel-treated siRNA control PC3 cells (Fig. 1A). We carried out a FISH to confirm that PKCδ-silenced PC3 cells were not slipping out of the mitotic arrest imposed by paclitaxel treatment, whereas in paclitaxel-treated siRNA control PC3 cells, the percentage of cells with higher ploidy than the normal was 77.6%; in paclitaxel-treated siRNA PKCδ PC3 cells, this percentage diminished to 59.2%. Western blot analysis and FISH results indicate that absence of PKCδ prevents slippage and induces a stronger mitotic arrest in paclitaxel-treated PC3 cells (Fig. 1B).

In paclitaxel-treated siRNA control LNCaP cells, we observed increased levels of cyclin B1, Pttg1, phospho-histone-H3^Ser10, and BubR1. These results reproduce those of our previous report, where we described that paclitaxel treatment caused the arrest of LNCaP cells in mitosis (5). As for paclitaxel-treated siRNA PKCδ PC3 cells, we observed an accumulation of cell cycle–related proteins in paclitaxel-treated siRNA PKCδ LNCaP cells (Fig. 1C), showing that both prostate cancer cell lines increased the mitotic arrest upon drug treatment in the absence of PKCδ.

We investigated whether the robust mitotic arrest induced by paclitaxel treatment in the context of PKCδ knockdown was linked to changes in apoptosis induction, because an efficient mitotic arrest is not necessarily related to higher paclitaxel sensitivity (5). Paclitaxel-treated PKCδ-silenced cells showed lower PARP cleavage and decreased caspase-9 and caspase-3 activation in comparison with paclitaxel-treated siRNA control cells. These results show that lack of PKCδ abrogates paclitaxel-induced apoptosis in both cell lines (Fig. 2A and B). Moreover, we studied the influence of PKCδ silencing on expression levels of several proteins belonging to the Bcl-2 family. In both cell lines, we observed that levels of the proapoptotic proteins Bax and Bak remained unchanged upon paclitaxel treatment in siRNA control and siRNA PKCδ cells; however, siRNA PKCδ cells showed a decrease of Bak protein compared with siRNA control cells. We also observed that antiapoptotic Mcl-1_L and Bcl-xL proteins were downregulated after 48 hours of paclitaxel treatment independently of PKCδ silencing, but downregulation was higher in siRNA control cells than in siRNA PKCδ cells. Interestingly, DMSO-treated PKCδ-silenced PC3 and LNCaP cells showed higher levels of Mcl-1_L and Bcl-xL proteins in comparison with siRNA control cells in both cell lines (Fig. 2A and B). The higher levels of antiapoptotic proteins Mcl-1_L and Bcl-xL and the lower levels of proapoptotic protein Bak in cells silenced for PKCδ, as well as the decreased downregulation of both antiapoptotic proteins observed after paclitaxel treatment in PKCδ-silenced cells, are in concordance with the reduced sensitivity to apoptosis of PKCδ-silenced prostate cancer cells.

PKCδ negatively modulates the canonical Wnt/β-catenin pathway (16) and this pathway modulates Mcl-1 expression in several tumor types (17–19). As we observed a prominent increase of Mcl-1_L in PKCδ-silenced cells and this protein has an important role in paclitaxel-induced apoptosis, we examined whether PKCδ modulates Mcl-1 expression through the Wnt/β-catenin pathway in prostate cancer cells. Total β-catenin expression levels remained unchanged in PKCδ-silenced PC3 and LNCaP cells in comparison with siRNA control cells. However, active β-catenin levels were significantly higher in the absence of PKCδ in both cell lines (Fig. 3A and B). We studied the expression of β-catenin target genes c-Myc and cyclin D1 and we observed that both were significantly increased in PKCδ-silenced PC3 and LNCaP cells compared with siRNA control cells. We also examined another transcriptional target of β-catenin, AurKA (20) a kinase that phosphorylates Akt at Ser473 and has an important role in its activation (21, 22). As expected, expression levels of AurKA and phospho-Akt^Ser473 were higher in PKCδ-silenced cells. Moreover, we studied GSK3β which is part of the degradation complex of β-catenin and is phosphorylated at Ser9 by Akt and AurKA, causing its inactivation (23–25). According to the activation state of Akt and increased AurKA expression, phospho-GSK3β^Ser9 expression was increased following PKCδ silencing. GSK3β targets Mcl-1 for destruction by the proteasome through its phosphorylation at Ser159 (26). In the absence of PKCδ, GSK3β was less active and we observed a decrease in phospho-Mcl-1^Ser159 levels together with an increase in Mcl-1_L levels (Fig. 3A and B). All the differences were statistically significant in both cell lines.

We observed that the Wnt/β-catenin pathway was downregulated by paclitaxel treatment as indicated by the decrease of active β-catenin without changes in the total β-catenin levels. The Akt pathway was also inhibited after paclitaxel treatment, as shown by the diminished levels of phospho-Akt^Ser473. Downregulation of both pathways was stronger in siRNA control than in siRNA PKCδ PC3 and LNCaP cells (Fig. 3C and D and Supplementary Fig. S1A and S1B). In concordance with these results, paclitaxel treatment induced GSK3β activation in both cell lines independently of PKCδ silencing, but activation was higher in siRNA control cells than in siRNA PKCδ cells. The levels of phospho-Mcl-1^Ser159 were practically unchanged by paclitaxel treatment, although we observed the downregulation of Mcl-1_L after drug treatment. This downregulation was more prominent in siRNA control cells than in siRNA PKCδ cells in both cell lines (Fig. 3C and D and Supplementary Fig. S1A and S1B). AurKA is predominantly expressed during mitosis, where it is essential for the proper timing of mitotic entry and the formation of bipolar spindles (27). Concordantly, we observed an increased expression of AurKA after paclitaxel treatment in both cell lines, especially in siRNA PKCδ cells that were more blocked in mitosis as indicated by the levels of cyclin B1, Pttg1, or phospho-histone-H3^Ser10.

In both cell lines, LY294002 treatment resulted in the inhibition of Akt activity as indicated by lower levels of phospho-Akt^Ser473 and the concordant activation of GSK3β, as observed by decreased phospho-GSK3β^Ser9 levels upon LY294002 treatment. The downregulation of phospho-Akt^Ser473 and phospho-GSK3β^Ser9 by LY294002 promoted an increase of phospho-Mcl-1^Ser159 and consequent decrease of Mcl-1 levels as well as a decrease of active β-catenin without changes in total β-catenin levels (Fig. 4A and B). We corroborated that LY294002 is capable of inhibiting the signaling pathways induced in the absence of PKCδ. We then investigated whether LY294002 treatment would sensitize PKCδ-silenced PC3 and LNCaP cells to paclitaxel. LY294002 treatment only slightly induced the cleavage of PARP in LNCaP siRNA control cells, but did not induce caspase-9 or caspase-3 activation. According to previous results, PKCδ-silenced PC3 and LNCaP cells were more resistant to paclitaxel. Interestingly, LY294002 sensitized PC3 and LNCaP cells to paclitaxel-induced apoptosis even in PKCδ-silenced cells, as indicated by higher PARP cleavage and caspase-9 and caspase-3 activation in LY294002 followed by paclitaxel-treated cells than in cells treated with paclitaxel alone (Fig. 4C and D). All the differences were statistically significant in both PKCδ-silenced cell lines.

PC3 and LNCaP cells were treated with XAV939 to examine the effect of this β-catenin inhibitor on the pathways that were activated in the PKCδ silencing context and on the expression of Mcl-1 (Fig. 5A and B). XAV939 treatment had no effect on total β-catenin levels, but the expression levels of active β-catenin were significantly decreased in both cell lines upon XAV939 treatment. In concordance with a lower activation of β-catenin, Western blot analyses also showed that XAV939 treatment strongly decreased the expression of its transcriptional targets c-Myc, cyclin D1, and AurKA. As expected, the β-catenin inhibitor diminished the activation of the Akt pathway and consequently, the activation state of GSK3β was higher after XAV939 treatment as indicated by decreased levels of phospho-Akt^Ser473 and phospho-GSK3β^Ser9. Following XAV939 treatment, the levels of proteasome-targeted phospho-Mcl-1^Ser159 were higher and so the levels of Mcl-1_L were lower in XAV939-treated cells than in DMSO-treated cells in both cell lines. Therefore, XAV939 inhibits activation of the β-catenin and Akt pathways and induces the degradation of the antiapoptototic protein Mcl-1. We then investigated whether XAV939 treatment increased paclitaxel-induced apoptosis in PKCδ-silenced PC3 and LNCaP cells. XAV939 treatment did not induce apoptosis as indicated by the absence of PARP cleavage and caspase-9 and caspase-3 activation in both cell lines. According to previous results, siRNA control PC3 and LNCaP cells showed more sensitivity to paclitaxel than siRNA PKCδ PC3 and LNCaP cells. Interestingly, the combined treatment of XAV939 and paclitaxel induced more apoptosis than single treatment with either XAV939 or paclitaxel. We observed the highest PARP cleavage and caspase-9 and caspase-3 activation following the combined treatment of both PC3 and LNCaP cells, even when PKCδ was silenced where all described differences were statistically significant (Fig. 5C and D).

We performed an immunohistochemical analysis of PKCδ, active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1 proteins in tumor tissue sections from 80 patients with primary prostate cancer to study the clinical relevance of our results. Tumor cells showed different degrees of cytoplasmic immunostaining for PKCδ, AurKA, phospho-GSK3β^Ser9, and Mcl-1 proteins, whereas immunostaining for active β-catenin was nuclear. Forty-nine tumors (61.2%) were of low Gleason score and 31 (38.8%) of high Gleason score. Out of 49 low Gleason score tumors, 44 (89.8%) showed high PKCδ expression and 29 (65.9%) expressed low levels of active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1 proteins. On the other hand, out of 31 high Gleason score tumors, 25 (80.6%) showed low PKCδ expression and 15 (60.0%) showed high expression of active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1 proteins (Fig. 6A).

In this study, immunohistochemical expression indicated that 80.6% of high Gleason score tumors expressed low PKCδ, whereas only 10.2% of low Gleason score tumors exhibited low PKCδ expression (P < 0.0001). Interestingly, the expression of PKCδ also correlated with high active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1 expression (P < 0.001). Active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1 were highly expressed in 80.0%, 66.7%, 56.7%, and 76.7% of tumors expressing low PKCδ, and these percentages decreased to 12.0%, 12.2%, 2.0%, and 32%, respectively, in tumors expressing high levels of PKCδ. In relation to the Gleason score, statistical analysis also revealed a significant correlation between high Gleason score tumors and high expression of active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1 (P < 0.0001). Moreover, high active β-catenin expression correlated with high expression of AurKA, phospho-GSK3β^Ser9, and Mcl-1 (P < 0.0001); high expression of AurKA showed a strong correlation with high expression of phospho-GSK3β^Ser9 and Mcl-1 (P < 0.0001), and high phospho-GSK3β^Ser9 immunohistochemical expression was associated with high Mcl-1 expression (P < 0.001; Supplementary Table S1).

Most advanced prostate cancers are initially sensitive to hormone therapy; however, the disease often progresses to a hormone-refractory state. Taxane-based chemotherapy constitutes the first-line treatment for CRPC, but the majority of patients develop resistance to chemotherapy (1, 2). Paclitaxel stabilizes microtubules and induces prolonged mitotic arrest by the activation of the SAC, and cells die in mitosis or exit mitosis without cytokinesis by slippage (3). In this study and in a previous report (5), we demonstrated that paclitaxel-treated LNCaP cells maintained an efficient mitotic arrest and were more sensitive to the drug than PC3 cells, which exited from mitosis by slippage. Interestingly, in PKCδ-silenced cells, paclitaxel treatment induced a higher increase of BubR1, cyclin B1, Pttg1, and phospho-histone-H3^Ser10 proteins in LNCaP and PC3 cells, suggesting a maintained SAC activation and a more robust mitotic arrest in comparison with paclitaxel-treated siRNA control cells. In PC3 cells, this increased mitotic arrest was supported by the decrease in the percentage of cells with higher ploidy in paclitaxel-treated siRNA PKCδ cells, indicative of a minor mitotic exit after paclitaxel exposure. We previously demonstrated that PKCδ-silenced cells were less sensitive to paclitaxel than siRNA control cells (15). Thus, all these data support the idea that a robust mitotic arrest is not necessarily accompanied by a higher apoptotic induction; there are other blockades that can prevent apoptosis during mitosis (3–5). In this study, we corroborated that PKCδ silencing induced resistance to paclitaxel in prostate cancer cells as indicated by the decreased cleavage of PARP protein and diminished caspase-9 and caspase-3 activation. Moreover, several events occurred in PKCδ-silenced PC3 and LNCaP cells that could contribute to paclitaxel resistance. Drug exposure did not induce changes in the expression levels of proapoptotic proteins Bax and Bak, but downregulation of antiapoptotic proteins Mcl-1 and Bcl-xL was observed in siRNA control and siRNA PKCδ cells, although the downregulation was higher in siRNA control cells and in concordance with higher apoptosis induction. Interestingly, in siRNA PKCδ cells, Mcl-1_L and Bcl-xL were upregulated and proapoptotic Bak was downregulated with respect to siRNA control cells in both cell lines. Again these data are in concordance with the lower paclitaxel-induced apoptosis observed in the absence of PKCδ. Supporting these data, there are several studies that demonstrate that the expression levels of cited antiapoptotic proteins determine the sensitivity to apoptosis during mitotic arrest induced by antimitotic drugs. Concretely, the loss of Mcl-1 seems to be a key event to trigger death during mitotic arrest (3–5).

Our data point to the regulation of Mcl-1 protein levels by PKCδ. There are several reports that link PKCδ with the regulation of the Wnt/β-catenin pathway (16), and other reports link this signaling pathway with Mcl-1 modulation (17–19). On the basis of these articles, we propose that PKCδ would modulate Mcl-1 by regulation of the Wnt/β-catenin pathway in prostate cancer cells. The Wnt/β-catenin pathway regulates embryogenesis, proliferation, cell differentiation, or migration, and it is abnormally activated in different tumor types (18, 20, 23, 24, 28, 29). β-catenin is a strongly regulated factor with a central role in this pathway. In the absence of Wnt activation, a degradation complex formed by GSK3β between other proteins is active and GSK3β phosphorylates β-catenin targeting it for proteasome degradation. Wnt activation implies the dissociation of the degradation complex and β-catenin translocation into the nucleus, where it activates the transcription of its target genes c-Myc and cyclin D1 which are involved in proliferation, migration, and survival (20, 23, 24, 29, 30). In prostate cancer cells, the knockdown of PKCδ induced higher activation of β-catenin in comparison with siRNA control cells, as well as higher expression of its transcriptional targets. The increased AurKA levels observed in the absence of PKCδ were of particular interest given the link between the Wnt and Akt pathways (21, 22). This pathway is also involved in resistance to microtubule targeting drugs (29, 31) and, therefore, the activation of β-catenin was associated with the activation of Akt and consequent inactivation of GSK3β. This inactivation impaired the phosphorylation of Mcl-1 at Ser159 and phosphorylation of β-catenin, preventing their destruction, and favoring Mcl-1 accumulation and β-catenin activation. The abnormal activation of β-catenin in siRNA PKCδ cells, as well as high Mcl-1 levels lead to diminished paclitaxel-induced apoptosis in prostate cancer cells. These results are supported by other studies that associate these changes with resistance to chemotherapeutic drugs (23, 25, 29, 32).

In accordance with data published in glioma cells (24), we observed that LY294002 inhibited Akt resulting in GSK3β activation and consequent Mcl-1 and β-catenin degradation. Similar results were obtained with XAV939 that inhibited β-catenin and the expression of its transcriptional targets, including AurKA. As a consequence, Akt was less active and GSK3β could phosphorylate Mcl-1, targeting it for degradation. These inhibitors used in combined therapy with paclitaxel resulted in an increased apoptotic response in comparison with single inhibitor or paclitaxel treatment, even in PKCδ-silenced cells. Supporting our results, other authors showed that the inhibition of these pathways improved the apoptotic response to EGFR-targeting therapy in lung cancer (28), deguelin and microtubule-targeting drugs in prostate cancer (29, 31), paclitaxel and irinotecan in colon cancer (30), or tamoxifen in glioma cells (24).

We previously reported that high Gleason score tumors showed lower PKCδ expression than low Gleason score tumors and this could be associated with apoptosis resistance (15). Thus, we expected an association between our results in prostate cancer cell lines and expression patterns of studied proteins in human prostate cancers. In our series of patients, high Gleason score significantly correlated with low expression of PKCδ and high expression of active β-catenin, AurKA, phospho-GSK3β^Ser9, and Mcl-1. Several studies have associated aberrant expression or high nuclear and cytoplasmic expression of β-catenin with high Gleason score tumors (33–37), yet others did not find an association between β-catenin expression and the Gleason score (38–40). We are the first to show a strong correlation between high active β-catenin immunohistochemical expression and high Gleason score prostate tumors. Many studies have linked overexpression of AurKA with progression and malignant phenotype of prostate cancers (41–44), although there are controversial results with respect to its correlation with the Gleason score (45, 46). We observed a significant correlation between high expression of AurKA and high Gleason score tumors. Several articles reported that high cytoplasmic expression of GSK3β correlated with high Gleason score (47, 48), but these observations are controversial because GSK3β is considered a tumor suppressor and its phosphorylation at Ser9 by Akt or AurKA reduces the apoptotic response to different stimuli. We studied the activation state of GSK3β in prostate cancer and found that the inactivation of GSK3β occurred in high Gleason score tumors. Finally, we observed an association between high Mcl-1 expression and high Gleason score tumors. This was in concordance with a previous report, although the authors only found significant differences grouping prostatic intraepithelial neoplasias with low and moderate Gleason score tumors versus high Gleason score tumors with bone and lymph node metastases (49). In addition, we found that expression of each examined protein correlated in a significant way with the expression of the other tested proteins. According to a part of our results, it has been reported that there is a correlation between nuclear β-catenin expression and high phospho-GSK3β^Ser9 expression in prostate cancer (50).

In summary, we have demonstrated that PKCδ has a proapoptotic role, acting as negative regulator of the Wnt/β-catenin pathway, as silencing of PKCδ induced resistance to paclitaxel through activation of Wnt/β-catenin pathway and impaired Mcl-1 degradation and the proposed model is presented in Fig. 6B. In parallel with these results, we observed that loss of PKCδ expression in high Gleason score tumors was associated with Wnt/β-catenin pathway activation and increased AurKA levels, as well as GSK3β inactivation and high levels of Mcl-1. We have also shown that combined treatments with paclitaxel and inhibitors of Wnt/β-catenin or Akt pathways improved paclitaxel-induced apoptosis even in the absence of PKCδ. These findings suggest that targeting Akt or β-catenin pathways may increase the efficacy of taxane therapy in advanced human prostate cancers that have lost the expression of PKCδ.

No potential conflicts of interest were disclosed.

Conception and design: M.L. Flores, C. Castilla, R. Medina, F. Romero, M.A. Japón, C. Sáez

Development of methodology: M.L. Flores, C. Castilla, J. Gasca

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Castilla, J. Gasca, R. Medina, B. Pérez-Valderrama

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.L. Flores, C. Castilla, J. Gasca, B. Pérez-Valderrama, M.A. Japón, C. Sáez

Writing, review, and/or revision of the manuscript: M.L. Flores, R. Medina, F. Romero, M.A. Japón, C. Sáez

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.L. Flores

Study supervision: M.A. Japón, C. Sáez

The authors thank the Andalusian Public Health System Biobank (ISCIII-Red de Biobancos PT13/0010/0056) for the human specimens used in this study.

This work was supported by research grants from the Instituto de Salud Carlos III, FIS PI13/02282 (to C. Sáez), Ministerio de Economía y Competitividad, SAF2014-53799-C3-1/2-R (to F. Romero and M.A. Japón), Consejería de Salud, AI-0025-2015 (to M.A. Japón) and Consejería de Innovación, Ciencia y Empresa, Junta de Andalucía, P10-CTS-6243 (to C. Sáez). M.L. Flores was supported by a pre-doctoral grant from the Consejería de Innovación, Ciencia y Empresa, Junta de Andalucía, P10-CTS-6243. C. Sáez was supported by a contract from Nicolás Monardes Program, Consejería de Salud, Junta de Andalucía.

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.