Paxillin Regulates Androgen- and Epidermal Growth Factor-induced MAPK Signaling and Cell Proliferation in Prostate Cancer Cells*

Although transcriptional effects of androgens have been extensively studied, mechanisms regulating transcription-independent (nongenomic) androgen actions are poorly understood. Previously, we have shown that paxillin, a multidomain adaptor protein, is a critical regulator of testosterone-induced MAPK-signaling during Xenopus oocyte maturation. Here we examine the nongenomic effects of dihydrotestosterone (DHT) in prostate cancer cells, focusing on how paxillin mediates Erk signaling and downstream physiologic actions. We show that in LnCAP cells DHT functions as a growth factor that indirectly activates the EGF-receptor (EGFR) via androgen receptor binding and matrix metalloproteinase-mediated release of EGFR ligands. Interestingly, siRNA-mediated knockdown of paxillin expression in androgen-dependent LnCAP cells as well as in androgen-independent PC3 cells abrogates DHT- and/or EGF-induced Erk signaling. Furthermore, EGFR-induced Erk activation requires Src-mediated phosphorylation of paxillin on tyrosines 31/118. In contrast, paxillin is not required for PKC-induced Erk signaling. However, Erk-mediated phosphorylation of paxillin on serines 83/126/130 is still needed for both EGFR and PKC-mediated cellular proliferation. Thus, paxillin serves as a specific upstream regulator of Erk in response to receptor-tyrosine kinase signaling but as a general regulator of downstream Erk actions regardless of agonist. Importantly, Erk-mediated serine phosphorylation of paxillin is also required for DHT-induced prostate-specific antigen mRNA expression in LnCAP cells as well as EGF-induced cyclin D1 mRNA expression in PC3 cells, suggesting that paxillin may regulate prostate cancer proliferation by serving as a liaison between extra-nuclear kinase signaling and intra-nuclear transcriptional signals. Thus, paxillin may prove to be a novel diagnostic or therapeutic target in prostate cancer.

(PCa) development (4,9), we chose to examine the regulatory role of paxillin in androgen-induced Erk signaling and downstream physiologic actions. Androgen and epidermal growth factor (EGF) signaling are critical regulators of PCa development and progression (4,9). In fact, paxillin association with focal adhesion molecules may be up-regulated in metastatic prostate carcinoma (14), and in PCa cell lines paxillin potentiates AR trans-activation by functioning as an AR co-activator (15). Here we report that paxillin is important for kinase signaling in response to multiple extra-nuclear signals in PCa cells, functioning as an upstream mediator of Erk activation and a downstream regulator of Erk signaling. Furthermore, we provide evidence that paxillin plays an important role in orchestrating cross-talk between extra-nuclear kinase signaling and intra-nuclear transcriptional events.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-LnCAP and PC3 cell lines were obtained from ATCC and cultured in RPMI 1640 medium (Invitrogen) containing 10% FBS and 1% penicillin-streptomycin. For experiments involving pharmacological inhibitors, cells were treated overnight with serum-free, phenol red-free RPMI 1640 media. Thereafter, cells were treated with vehicle (0.1% DMSO) or inhibitors Galardin, PP2, AG1478 (Calbiochem), flutamide, or erlotinib (Sigma) for 30 min before stimulation with 0.1% ethanol (vehicle) or 25 nM DHT for 30 min.
EGFR Transactivation Assay-A431 cells were used to detect DHT-mediated release of EGFR ligands from LnCAP cells. A431 cells (ATCC) were cultured in DMEM/F-12 (1:1) medium (Invitrogen) containing 10% FBS and 1% penicillin-streptomycin, serum-starved overnight, and then stimulated with medium from DHT-, DHT ϩ galardin-, or vehicle (0.1% ethanol)treated LnCAP cells for 60 min. As controls, A431 cells were stimulated with DHT or media alone. Thereafter, A431 cells were isolated for Western blot analysis to detect phosphorylated and total EGFR. Transient Transfection-PC3 or LnCAP cells were treated with non-targeting siRNA pool (ThermoFisher Scientific) or paxillin-specific siRNA according to manufacturer's instructions. Two sets of human paxillin siRNAs were used: 1) human paxillin siRNA (Santa Cruz Biotechnology) containing three target-specific 20 -25 nucleotide siRNAs or 2) human paxillin siRNA ON-TARGET plus SMARTpool (ThemoFisher Scientific) containing four siRNAs targeting the paxillin mRNA. The latter was used for all experiments here, although results were similar with both pools. For experiments involving constitutively active (ca) Raf (William Walker, University of Pittsburgh) or MEK (Melanie Cobb, University of Texas Southwestern Medical Center), cells were co-transfected with paxillin or nonspecific siRNAs and cDNAs encoding caRaf or caMEK. After 72 h, cells were treated overnight with serum-free, phenol redfree RPMI 1640 media and stimulated with 0.1% ethanol/ DMSO (vehicle), 25 nM DHT (Steraloids), 20 ng/ml EGF (BD Biosciences), or 100 nM PMA (Sigma) for the times indicated for Western blots or 24 h for MTT assays.
Paxillin Rescue Experiments-PC3 cells were transfected with paxillin siRNA as described above. After 96 h, media was removed, and cells were transfected with Lipofectamine (Invitrogen) or Lipofectamine plus WT, S83A/S126A/S130A, or Y31A/Y118A paxillin cDNA. After 48 h, cells were treated overnight with serum-free, phenol red-free RPMI 1640 media and stimulated with the indicated ligands for 30 min for Western blots or 24 h for MTT assays.
MTT Assay-MTT assays were performed using a colorimetric assay cell proliferation kit (Roche Applied Science) according to the manufacturer's instructions.
Cell Migration/Invasion Assay-Cell migration/invasion assays were performed using a colorimetric QCM Cell invasion assay kit containing 24-well Boyden chambers with extracellular matrix-coated 8-m pore size membranes (Millipore) according to the manufacturer's instructions.
Statistical Analysis-Results for the MTT assay, cell migration-invasion assay, and real-time PCR were analyzed using Student's t test. A value of p Յ 0.05 was considered significant.

DHT-induced Erk Signaling Occurs via Matrix Metalloproteinase (MMP)-mediated Transactivation of the EGFR-Be-
cause androgens induce Erk1/2 signaling in PCa cell lines (3, 10 -12), we examined the underlying mechanism by which DHT activates Erk1/2 in LnCAP cells. DHT treatment of LnCAP cells for 30 min significantly induced Erk1/2 phosphor-ylation/activation (Fig. 1A, lane 2). Notably, saturating concentrations of DHT (25 nM) were used in these studies to maximize the significance of the inhibitor effects; however, lower concentrations (1-10 nM) promoted a similar magnitude response (not shown). DHT levels in the prostate have not been accurately determined but are reported to be at least 10 -20 nM (19,20). Pretreatment with the AR antagonist flutamide (Fig.  1A, lane 3) as well as EGFR inhibitors AG1478 and Erlotinib (Fig. 1A, lanes 4 and 5) blocked Erk1/2 phosphorylation, indicating that DHT induces EGFR trans-activation and subsequent Erk1/2 signaling via classical ARs. The concentrations of AG1478 (20 M) and erlotinib (5 M) used were based on concentration gradient experiments (supplemental Fig. S3, A and B) and previous studies in PCa cells (21)(22)(23)(24)(25) and other cell lines (18, 26 -29). Of note, AG1478 at 20 M specifically blocks EGFbut not FGF-induced Erk activation in LnCAP (supplemental Fig. S3C) and MEK activation in MLTC (18) cells, thus, demonstrating the specificity of AG1478 to EGFR inhibition and ruling out off target effects on the Ras/Raf/MEK/Erk pathway. Finally, the inhibitors alone (in absence of DHT) had no effect on Erk signaling (data not shown).
The ability of the MMP inhibitor galardin to block DHTinduced Erk1/2 phosphorylation (Fig. 1, lane 7) and its rescue by EGF treatment (supplemental Fig. S1) suggests that DHT activates the EGFR through MMP activation, possibly by release of membrane-associated EGFR ligands (30,31). In fact, medium from DHT (Fig. 1B, lane 2)-but not vehicle-treated (Fig. 1B, lane 1) LnCAP cells increased EGFR phosphorylation in A431 cells (known to express very high levels of EGFR), indicating that EGFR ligands were being released into the medium from DHT-treated LnCAP cells. Moreover, the addition of galardin to DHT-treated LnCAP cells (lane 5) blocked the ability of the medium to activate EGFR signaling in the A431 cells, demonstrating that the release of EGFR ligands is indeed MMP-dependent. Finally, DHT alone (lane 4) did not promote EGFR activation in the A431 cells.

Paxillin Regulates DHT-and EGF-induced ERK Activation in
PCa Cells-Next we investigated whether paxillin regulated Erk1/2 activation in PCa cells. In androgen-dependent LnCAP cells, both DHT ( Fig. 1D) and EGF ( Fig. 1E) induced Erk1/2 activation from 15 to 120 min. Knockdown of paxillin abrogated DHT-induced Erk1/2, but not Akt, activation (Fig. 1D). Furthermore, paxillin knockdown in either LnCAP or androgen-independent PC3 cells inhibited EGF-induced Erk1/2 ( Fig. 1E) but not Akt (not shown) activation. Similar effects were observed in DHTtreated LAPC-4 PCa cells (not shown) and FGF or EGF-treated HEK-293 cells (supplemental Fig. S2A). These observations suggest that paxillin is an important regulator of growth factor receptor-induced Erk1/2 signaling regardless of cell type or how the growth factor receptor is activated (EGFR either indirectly by DHT or directly by EGF, or FGFR directly by FGF). Furthermore, EGF-mediated Erk activation was attenuated in mouse embryonic fibroblasts from paxillin null mice (from Dr. Sheila Thomas, Harvard University), providing genetic confirmation of the siRNA experiments that paxillin is important for Erk1/2 signaling (supplemental Fig. S2B).
Paxillin Acts Downstream of the EGFR but Upstream of Raf/MEK-To investigate where paxillin functions in EGFRinduced signaling, we examined DHT-or EGF-induced activa- The caMEK used has mutations substituting glutamic and aspartic acid for Ser-218 and Ser-222, significantly increasing the basal activity of MEK over the unphosphorylated wild-type enzyme (32,33). The caRaf used is a fusion protein of the membrane localization signal of Ras to the carboxyl terminus of Raf that is constitutively activated independent of cellular Ras (34). Finally, knockdown of paxillin minimally affected DHT-or EGF-induced EGFR phosphorylation (Fig. 2D). Together these results demonstrate that paxillin functions downstream of the EGFR but upstream of Raf and MEK to regulate Erk1/2 activation.
We next used an in vitro cell migration/invasion assay consisting of a 24-well Boyden chamber with an extracellular matrix-coated membrane to investigate the importance of paxillin for EGF-induced cell migration and invasion of PC3 cells. PC3 cells migrated through the membrane and invaded the matrix in response to EGF, and paxillin knockdown abrogated this physiological process. Fig. 3B demonstrates quantitative analysis of migration/invasion by measuring absorbance after staining invading cells (upper panel) as well as qualitative images of the extracellular matrix-coated membrane underside containing migrated and invaded cells (lower panel). Collectively, these data demonstrate that paxillin is critical for proliferation, migration, and invasion of prostate cancer cell lines.
Paxillin Specifically Regulates Receptor-tyrosine kinase-but Not PKCinduced Erk Activation-Because our data indicated that paxillin was a critical regulator EGFR/Src-induced Erk1/2 activation, we next determined whether paxillin was a universal modulator of Erk. As an alternative, receptor-tyrosine kinase/Src-independent means of activating Erk1/2, we used PMA to promote PKC-mediated Erk1/2 signaling. Paxillin-siRNA-treated LnCAP cells were stimulated with 0.1% DMSO (vehicle), EGF, or PMA for 30 min and Erk1/2 phosphorylation measured. Knockdown of paxillin abrogated EGF (Fig. 5A, lane 6)-but not PMA (Fig. 5A, lane 5)-induced Erk1/2 activation, demonstrating that the regulation of Erk activation by paxillin may be relatively specific to the receptor-tyrosine kinase/Src signaling pathway.
Erk-mediated Phosphorylation of Paxillin Is Required for Normal Transcription in LnCAP and PC3 Cells-Some studies suggest that extra-nuclear kinases activated by steroids or growth factors may regulate transcription (10, 11, 44 -50). To examine the role of Erk and paxillin in regulating transcription in PCa cells, we first studied DHT-induced expression of PSA mRNA in LnCAP cells. Inhibition of Erk signaling by the MEK inhibitor U0126 or the EGFR inhibitor AG1478 as well as knockdown of paxillin expression abrogated DHT-induced expression of PSA mRNA (Fig. 6A). These data suggest that, in LnCAP cells, extra-nuclear DHT-mediated Erk1/2 activity (via EGFR and paxillin) is essential for normal intra-nuclear DHT-mediated transcription. Surprisingly, reduction of paxillin expression in PC3 cells similarly reduced EGF-mediated expression of cyclin D1 mRNA (Fig. 6B). Cyclin D1 mRNA expression could be rescued by re-expression of wild-type paxillin but not S83/126/20A-paxillin. These results, which mirror the proliferation data in Figs. 4 and 5, suggest that paxillin may regulate Erk-induced proliferation in part by enhancing Erk1/2-mediated transcription. Thus, paxillin may help mediate cross-talk between cytoplasmic kinase and nuclear transcriptional signaling.

DISCUSSION
This study reveals a remarkable conservation of paxillin function from frog germ cells to human PCa cells. However, although the general concepts of paxillin actions on Erk signaling are similar in lower versus higher vertebrates, this study in somatic cells highlights several novel regulatory roles of paxillin in Erk signaling that ultimately control important physiological functions in PCa cells such as transcription and proliferation.
To summarize our data, we propose the following model to describe extra-nuclear AR-mediated signaling in PCa cells (Fig. 6). Androgens bind to classical ARs, most likely located at or near the cell surface (3,12,51), to promote activation of MMPs and release of membrane-associated EGFR ligands. Although we have not identified specific EGFR ligands being released in LnCAP cells, previous studies implicate heparin bound-EGFs as meditators of Gprotein-coupled or steroid receptor cross-talk with EGF receptors (30,31). These ligands bind to and activate the EGFR, which then activates Src, Akt, and MEK/Erk1/2 (Fig. 1). Notably, prior co-immunoprecipitation studies in LnCAP cells suggested that DHT-induced  Erk1/2 activation (10, 11) might be mediated by extra-nuclear steroid receptors directly binding to and activating Src (52) followed by EGFR phosphorylation (3,12). However, here EGFR inhibition blocked DHT-induced Src phosphorylation, whereas Src inhibition had minimal effect on DHT-induced EGFR phosphorylation (Fig. 1B). Thus, similar to EGFR signaling in other cell types (53)(54)(55)(56)(57)(58), Src actions appear downstream of EGFR activation in DHT-stimulated LnCAP cells.
Irrespective of the underlying mechanism, the rapid and robust trans-activation of the EGFR by DHT highlights the novel concept that, outside of the nucleus, androgen actions are just like EGF with respect to activation of cytoplasmic kinase cascades (3,4,44). EGFR-induced kinase pathways are known to modulate steroid receptor-mediated transcriptional signaling by altering both receptor and co-regulator activities (10,11,44,45,49,50,59,60). Thus, indirect activation of the EGFR by DHT similarly leads to "outside-inside" cross-talk whereby rapid activation of extra-nuclear kinases enhances intra-nuclear transcriptional signaling. Data that MEK and EGFR inhibition block DHT-induced PSA mRNA expression (Fig. 6A) support this model.
What mediates this outside-inside cross-talk between extranuclear kinases and intra-nuclear transcription? Because paxillin knockdown abrogated both EGF-and DHT-induced Erk1/2 activation (Fig. 1, C and D) as well as DHT-induced PSA mRNA (Fig. 6A) and EGF-induced cylin D1 mRNA expression (Fig.  6B), paxillin appears to be at least one key regulator of outsideinside signaling in response to both direct (EGF) and indirect (DHT) activation of the EGFR. Our data further suggest that paxillinmediated regulation of extra-nuclear kinases and intra-nuclear transcription in turn controls PCa cell proliferation, invasion, and migration. Thus, paxillin is a critical regulator of multiple EGFR/Erk-regulated processes in PCa cells.
How does paxillin mediate these EGFR/Erk-regulated functions? One mechanism is at the level of EGFR/ Src-induced activation of the Raf/ MEK/Erk signaling pathway (Fig.  6C). Prior studies suggested that Src-mediated phosphorylation of tyrosines 31/118 on paxillin was important for Erk activation in response to FAK or integrin-mediated signaling (6,40,41). Our paxillin knockdown and rescue experiments with Y31A/Y118A-paxillin in PC3 cells unequivocally confirm that tyrosine phosphorylation of paxillin at these residues is required for EGFR-mediated Erk1/2 activation and downstream proliferation (Fig. 4). Previous studies (40,41,61) also suggested that paxillin might function as a scaffold to hold Raf, MEK, and Erk1/2 in a signaling complex. Most of these studies (17, 40 -42, 61, 62) focused on the association of paxillin with focal adhesion molecules, using overexpression and co-precipitation studies to show interactions. In our study, the morphology of PCa cells with reduced paxillin expression appeared grossly normal, although we did not specifically examine cellcell adhesions. However, in paxillin knockdown cells, expression of caRaf or caMEK was sufficient to promote Erk1/2 signaling (Fig. 2, B and C). Thus, although paxillin may form a complex with Raf, MEK and Erk, these interactions are not necessary for signaling by or downstream of Raf. In fact, paxillin appears to function between EGFR and Raf (Fig. 6C), as EGFR phosphorylation is minimally affected by paxillin knockdown (Fig. 2D).
Interestingly, in androgen-induced maturation of Xenopus oocytes, paxillin also functions just upstream of MOS, the germ cell homologue of Raf, again demonstrating the remarkable conservation of paxillin function from lower to higher vertebrates. However, the requirement for initial Src-mediated tyrosine phosphorylation of paxillin, the ability of paxillin to regulate downstream Erk functions regardless of the agonist, and the ability of paxillin to mediate cross-talk between cytoplasmic kinase and nuclear transcriptional signaling are all specific to somatic cells, as they are not seen in frog oocytes.
Our study suggests that paxillin is a relatively specific regulator of receptor-tyrosine kinase/Src-mediated Erk1/2 activation in PCa cells, as paxillin knockdown had no effect on PKCmediated Erk activation in PC3 cells (Fig. 5A). In contrast, with nonspecific (Nsp)-or paxillin (Pax)-specific siRNA for 96 h followed by transfection with WT or mutated paxillin (S83A/S126A/S130A) before stimulation with either media or 20 ng/ml EGF for 24 h. All data are represented as the mean Ϯ S.E. (n ϭ 3) and normalized to GAPDH levels. *, Student's t test, p Յ 0.05, stimulus versus media. All experiments were performed at least three times with similar results. C, shown is a proposed model describing the paxillin role in nongenomic AR or EGFR signaling in PCa cells.
paxillin appears to be a general regulator of Erk-mediated cellular processes such as transcription or proliferation, irrespective of the stimulus, as it was required for both PKC and EGFR-mediated proliferation in PC3 cells (Fig. 5). Importantly, serine phosphorylation of paxillin appears critical for these Erk-mediated processes, as in paxillin-knockdown PC3 cells S83A/S126A/S130A-paxillin was unable to rescue proliferation or induction of cyclin D1 mRNA in response to either EGF-or PKC-mediated Erk1/2 activation (Figs. 5 and 6). In fact, our results (Figs. 3C and 5B) and previous evidence (5,16,17,40,41) indicate that Erk itself may directly phosphorylate paxillin at these serine residues.
To summarize, paxillin regulates Erk-mediated processes by two means (Fig. 6C); 1) by specifically regulating receptor-tyrosine kinase-mediated Erk1/2 activation via Src-mediated tyrosine phosphorylation and 2) more broadly by regulating of Erk-mediated downstream processes like intra-nuclear transcription and proliferation via Erk-mediated serine phosphorylation. Thus, paxillin can be both an affector and an effector of Erk signaling in PCa cells, depending upon the stimulus. How paxillin act as an effector to regulate transcription in PCa cells is unknown; however, one possibility is that paxillin constitutively binds to inactive Erk1/2 to keep it sequestered. Consistent with this hypothesis, the binding affinity of inactive Erk1/2 to paxillin appears higher than that of activated Erk1/2 (41,50). Irrespective of how it is triggered, activated Erk1/2 might then promote serine phosphorylation of paxillin, releasing Erk from paxillin and permitting Erk-mediated transcription, cell proliferation, migration, and invasion. Further work is needed to investigate the details of this pathway. However, these studies underscore the importance of paxillin in regulating proliferation in both androgen-dependent and -independent PCa regardless of the stimulus (steroids or other growth factors). In fact, understanding how paxillin regulates Erk-mediated proliferation may lead to novel therapeutic targets for cancer treatment as well possible diagnostic markers for tumor aggressiveness.