Loss of CDCP1 triggers FAK activation in detached prostate cancer cells

Abstract

A major metastasis suppressing mechanism is the rapid apoptotic death of cancer cells upon detachment from extracellular matrix, a process called anoikis. Focal adhesion kinase (PTK2/FAK) is a key enzyme involved in evasion of anoikis. We show that loss of the Cub-domain containing protein-1 (CDCP1), paradoxically stimulates FAK activation in the detached state of prostate cancer cells. In CDCP1^low DU145 and PC3 prostate cancer cells, detachment-activation of FAK occurs through local production of PI(4,5)P₂. PI(4,5)P₂ is generated by the PIP5K1c-201 splicing isoform of PIP5K1c, which contains a unique SRC phosphorylation site. In the detached state, reduced expression of CDCP1 and an alternative CDCP1-independent SRC activation mechanism triggers PIP5K1c-pY644 phosphorylation by SRC. This causes a switch of Talin binding from β1-integrin to PIP5K1c-pY644 and leads to activation of PIP5K1c-FAK. Reduced CDCP1 expression also inactivates CDK5, a negative regulator of PIP5K1c. Furthermore, immersion of prostate cancer cells in 10% human plasma or fetal bovine serum is required for activation of PIP5K1c-FAK. The PIP5K1c induced detachment-activation of FAK in preclinical models sensitizes CDCP1^low prostate cancer cells to FAK inhibitors. In patients, CDCP1^High versus CDCP1^low circulating tumor cells differ in expression of AR-v7, ONECUT2 and HOXB13 oncogenes and TMPRSS2 and display intra-patient heterogeneity of FAK-pY397 expression. Taken together, CDCP1^low and CDCP1^high detached prostate cancer cells activate distinct cytoplasmic kinase complexes and targetable transcription factors, which has important therapeutic implications.

Introduction

Cell adhesion provides a critical survival signal for adherent cancer cells and detachment from extracellular matrix causes rapid cell death through a mechanism called anoikis [1,2]. In cancer, focal adhesion kinase (PTK2/FAK) is known as the principle kinase involved in anoikis-resistance. FAK activation regulates the turnover of focal adhesions [3] and is caused through the release of autoinhibition of the kinase domain by the FERM domain of FAK [4]. While mechanisms of FAK activation downstream of integrin engagement in focal adhesions have been extensively investigated [5,6], the mechanism of FAK activation to oppose anoikis in detached cells has not been investigated.

CDCP1 is a cell surface receptor that sequesters SRC and PKC-δ to plasma membrane subdomains [7]. CDCP1 forms a complex with β1-integrin that is required for tumor metastasis [8]. The shedding of the CDCP1 ectodomain through cleavage by plasmin is another mechanism critical for metastatic dissemination [9,10]. Reduced CDCP1 expression has been linked to cell detachment and anchorage-independent growth [11-13]. We demonstrated that in the detached state reduced CDCP1 expression abolishes the activity of CDK5, which leads to the loss of inside-out activation of β1-integrin [11]. In patients with lethal, metastatic castration-resistant prostate cancer (mCRPC), reduced CDCP1 was observed in circulating tumor cells (CTCs) and in approximately 50% of mCRPC tissues [11,14].

The first in vivo evidence of the pro-oncogenic phenotype resulting from loss of CDCP1 and its connection to FAK activation emerged from a genetically engineered PyMT mouse model of basal type breast cancer [15]. FAK activation was observed when explanted PyMT CDCP1 knockout cancer cells were detached from extracellular matrix (ECM) and cultured in suspension. Expression of CDCP1 prevented FAK activation in detached cells [16]. However, the mechanism leading to FAK activation downstream of CDCP1-loss remained unclear.

A potential mechanism of direct FAK kinase activation in the detached state and independent of integrin engagement is through PtdIns[4,5]P₂ (PI(4,5)P₂). Structural analysis demonstrates that PI(4,5)P₂ binds to the FERM domain of FAK [4,17,18]. The production of PI(4,5)P₂ occurs under tight spatial regulation by phosphatidylinositol 4-phosphate 5-kinases (PIP5K) [19,20]. The PIP5K1c-v2 (PIP5K1c-201) isoform is associated with focal adhesions and responsible for FAK activation. This PIP5K1c isoform is the only isoform containing Exon17, which puts the PIP5K1c under regulation by the c-SRC kinase [21]. Phosphorylation of PIP5K1c-Y644 by SRC leads to activation of PIP5K1c by inducing Talin binding and the switch of Talin binding from β1-integrin to PIP5K1c [22,23]. We hypothesized that this mechanism, which is independent of cell adhesion, may be responsible for detachment activation of FAK.

In this study we demonstrate how FAK is activated in CDCP1^Low PC cells upon detachment from ECM. Using CTCs from patients with mCRPC, a PC xenograft mouse model and normal human primary lymphocytes, we establish the molecular association between loss of CDCP1 and FAK activation. We demonstrate that FAK activation in the detached state occurs by the convergence of three pathways. We also demonstrate that loss of CDCP1 increases the sensitivity of detached PC cells to FAK inhibitors. Reduced expression of CDCP1 leads to a decline in β1-integrin activation, the loss of CDCP1 may participate in coordinating cell detachment from ECM and FAK activation. This mechanism would be able to overcome anoikis and allow PC cells to survive bottlenecks of tumor metastasis that involve cell detachment, such as vascular intravasation.

Materials and methods

Antibodies

The following antibodies were purchased from Cell Signaling Technologies: Rabbit anti-human CDCP1 (#4115), β1 Integrin (#4706), Phospho-SRC (#2101), SRC (#2108), Phospho-FAK (#8556), Phospho-FAK (#3281), Phospho-FAK (#3284), FAK (#13009), Phospho-p44/42 (ERK1/2) (#4377), p44/42 (ERK1/2) (#9102), Phospho-Akt (Thr308) (D25E6) XP^® Rabbit mAb #13038, Akt (#9272), GFP (D5.1) XP^® Rabbit mAb #2956, Anti-rabbit IgG, HRP-linked Antibody #7074. Activated β1-integrin (HUTS4, MAB2079Z) was obtained from Millipore (400 Summit Drive, Burlington, MA, 01803).

Pharmalogical inhibitors PF573228 (#PZ0117) and PD98059 (#P215) were purchased from Sigma-Aldrich. Saracatinib (#S1006) was purchased from Selleckchem (9330 Kirby Drive, STE 200, Houston, TX, 77054). All experiments were conducted with three technical and two biological replicates.

Human Plasma (#P9523) and Bovine Serum Albumin (#05470) were purchased from Sigma-Aldrich. Exogenous Phosphatidylinositol 4,5-bisphosphate diC8 (PI(4,5)P₂ diC8) (PI(4,5)P₂) (P-4508-100UG) was purchased from Echelon Biosciences, Inc. FAK plasmid pLV-neo-CD2-FAK (#37013) and PIP5K plasmid GFP-PIPK1-gamma 90 (#22299) were purchased from Addgene. Mutant constructs were generated with Agilent Technologies QuikChange Site-Directed Mutagenesis Kit (#200518).

Study approval

All experiments were conducted with approval from the institutional review board and the animal welfare committee of Cedars-Sinai Medical Center.

NanoVelcro CTC analysis in a cohort of metastatic PC

Patients provided permission for CTC analysis by signing a consent. CTCs were enriched and captured by the NanoVelcro CTC platform using similar protocols as described in [24]. The NanoVelcro CTC isolation system combines streptavidin-coated three-dimensional nano-structured substrates and an overlaid polydimethylsiloxane microfluidic chaotic mixer to capture CTCs. Venous blood was collected in acid-citrate dextrose-containing vacutainers (BD Bioscience, San Jose, CA, USA) from mCRPC patients under the existing protocols approved by the Cedars-Sinai Medical Center (CSMC) and processed within 24 hours. Following density gradient centrifugation, 1 mL of separated peripheral blood mononuclear cells (PBMC) were incubated with a capture agent (biotinylated anti-human EpCAM antibody, R&D Systems, Minneapolis, MN, USA). The sample was then loaded into the NanoVelcro chip, which contains an automated fluid handler at (0.5 mL/hr). Cells captured on the surface of the chip were fixed with 2% paraformaldehyde (PFA, Electron Microscopy Science) and stained by immunocytochemistry (ICC) with DAPI (Vector Laboratories, Burlingame, CA, USA), mouse anti-pan-cytokeratin (Agilent Technologies, Santa Clara, CA, USA), rat anti-CD45 (Life technologies, Carlsbad, CA, USA), and rabbit anti-FAK-pY379 (Cell Signaling Technology, Danvers, MA, USA). Cells were visualized with Alexa Fluor 488-conjugated anti-mouse (Invitrogen, Carlsbad, CA, USA), Alexa Fluor 555-conjugated anti-rabbit (Invitrogen, Carlsbad, CA, USA), and Alexa-647 donkey anti-rat (Abcam, Cambridge, MA, USA) antibodies.

Imaging of stained CTCs

An automated scan of the NanoVelcro chip was conducted (10X) for enumeration and images of cells were taken at 40X using an upright fluorescence microscope (Eclipse 90i, Nikon) with NIS-Element imaging software (Nikon) for visualization of nuclear and cytoplasmic staining. A multi-wavelength filter cube device permitted the separation of signals from fluorophores with different emission wavelengths (DAPI (nucleus), FITC (CK), CY5 (CD45) and TRITC (FAK-pY379). CTCs displayed positive DAPI and CK staining and were negative for CD45 staining. These cells were selected for quantification of FAK-pY379 expression. FAK-pY379 signal intensity was measured using ImageJ in the CK and/or DAPI positive pixel areas. The average staining intensity in the FAK-pY397 channel was determined and the background was subtracted. The resulting value of FAK-pY397 expression in each CTC is plotted on the Y-axis of the graph shown in Figure 1C.

Figure 1.

CDCP1 and ONECUT2 subgroups in metastatic castration-resistant prostate cancer (mCRPC). A. mRNA expression of CDCP1, ONECUT2 and 10 mCRPC-related genes in CTCs from 23 patients with mCRPC. Cases are ranked based on CDCP1 expression. The red arrow marks CDCP1^high PC3 cells and the blue arrow indicates CDCP1^Low PC3 cells. B. Boxplots comparing ARv7-CE3, HOXB13, ONECUT2 and TMPRSS2 gene expression levels in top versus bottom CDCP1 quartiles, **P < 0.001. C. Phospho-FAK-Y379 expression in CTCs from 8 patients with mCRPC. Each bar depicts an individual CTC. The mean, background adjusted immunofluorescent signal intensity of FAK-pY379 in individual CTCs is plotted on the y-axis. CTCs are from CDCP1^High (grey bar) and CDCP1^Low (orange bar) mCRPC patients and are identified by positive DAPI and pan-cytokeratin and negative CD45 staining (Figure S2).

Analysis of CDCP1 mRNA expression in CTCs and bulk tumor

CDCP1 mRNA expression in CTCs was measured on the Nanostring^® Encounter platform. Custom designed probes for quantification of individual genes were purchased from Nanostring^®. Genes were selected based on expression in PC and lack of expression in blood cells. Expression counts for each gene were normalized as described in [43]. Analysis of CDCP1, ONECUT2 and AR in bulk tumor was performed in a subgroup of PCs that displays the greatest variance in ONECUT2 activity based on expression levels of genes comprising the ONECUT2 cistrome. We previously published three subtypes of PC according to pathway activation signatures [31]. Amongst the three subtypes, subtype 3 (PCS3) exhibits the greatest dynamic range of the ONECUT2 activation signature [26]. Therefore, we performed the analysis of ONECUT2, AR and CDCP1 interactions in the PCS3 subtype.

Analysis of FAK in TCGA and PCTA

Data of gene alterations with the focal adhesion associated protein kinase (PTK2/FAK) gene were obtained from The Cancer Genome Atlas through cBioportal (https://www.cbioportal.org/). The Prostate Cancer Transcriptome Atlas (PCTA) [31] was used for analysis of FAK activity signatures procured from data published [29,30]. Genes differentially expressed in wild type and FAK^-/- murine squamous carcinoma cells (SCC) were identified through profiling on Affymetrix microarrays. Differentially expressed genes were separated into groups of reduced or elevated expression levels. Genes with increased expression in FAK^-/- compared to wt-FAK SCC were interpreted as regulated in a repressive mode by FAK (FAK^DOWN signature), while RNA expression levels of genes downregulated in FAK^-/- (FAK^UP signature) were interpreted as being induced by FAK. Using the gene sets as a surrogate to determine FAK activation we evaluated changes during PC progression in PCTA and detected a significant change only with the FAK^DOWN gene set. The FAK^UP signature did not change significantly (ANOVA P > 0.01) during PC progression.

Cell lines and mouse xenografts

Sublines from DU145 and PC3 cells and silencing of CDCP1 were previously described by our group [32]. Cells were authenticated specifically for this study as DU145 or PC3 cells by short tandem repeat profiling and tested free of mycoplasma contamination (University of Arizona Genetics Core). Mouse studies were conducted under an IACUC approved protocol from Cedars-Sinai Medical Center [11]. Investigators were not blinded to the data.

Cell suspension on HEMA-coated plates, Western Blot protocols, co-immoprecipitations and MTT assays are described previously [11].

Exogenous PI(4,5)P₂ treatment

Phosphatidylinositol 4,5-bisphosphate diC8 (PI(4,5)P₂ diC8) (PI(4,5)P₂) (P-4508-100UG) was purchased from Echelon Biosciences, Inc. and administered in micelles formed by addition of 0.1% Triton X-100 (Sigma-Aldrich) at the indicated concentrations. DU145 and PC3 cells were treated for 3 hours while suspended in HEMA coated plates.

Bioinformatics analyses of PIP5K1c

To measure mRNA abundance of the PIP5K1c splice variant that contains Exon 17, RNA-Seq reads were mapped to GRCh37 with Gencode [44] v24lift37 annotation using STAR (v2.5.3) [45]. The GeneCounts parameter was set to TranscriptomeSAM and the resulting bam file was used as input to the rsem-calculate-expression program (v1.3.0) [46]. RSEM isoforms levels of PIP5K1c splicing isoforms were used to calculate the Spearman correlations between PIP5K1C isoforms. The PIP5K1c-201 Ensemble designation (ENST00000335312.7) corresponds the splicing variant designated as PIP5K1c-v2 in functional studies of the splicing isoform [21]. Copy number alteration (CNA) analysis was used to determine the genomic integrity of the PTEN and TP53 genes using single nucleotide detection (SNP) on the OncoScan platform. CNA calls are from OncoScan FFPE Express 3.0 arrays as previously described [34]. Correlation of CNA loss with RNA expression levels was performed in data from the same cases.

FAK mutant constructs

FAK construct pLV-neo-CD2-FAK (#37013) was purchased from Addgene. pLV-neo-CD2-FAK was mutated to a FERM domain mutant according to [17] and a FAT domain mutant according to [47] using Agilent Technologies QuikChange Site-Directed Mutagenesis Kit (#200518) and transfected via electroporation using a Neon transfection system. 48 hours post-transfection, positive cells were selected by 5 days of G418 treatment.

PIP5K1c mutant constructs

The PIP5K1c construct was purchased from Addgene (75 Sidney Street, Suite 550A Cambridge, MA 02139). PIP5K, GFP-PIPK1 gamma 90 (#22299) was mutated to a D316A dominant negative mutant [21] using Agilent Technologies QuikChange Site-Directed Mutagenesis Kit (#200518) and transfected via electroporation using a Neon transfection system. 48 hours post-transfection, positive cells were selected by 5 days of G418 treatment.

Immunocytochemistry/immunofluorescence of lymphocytes

Peripheral blood lymphocytes from normal donors were suspended in HEMA-coated 60 mM tissue culture plates for 3 h in 10% FBS or 1% BSA. Cells were fixed for 10 min with 3.4% PFA and permeabilized/fixed for 30 min. Cells were spun at 1000 × g for 2 min and re-suspended in 1:1000 primary antibodies in 5% goat serum in PBS for 1 h. Cells were washed with PBS, then incubated with 1:2000 secondary antibodies in 5% goat serum in PBS for 30 min. Cells were washed with PBS, mounted with ProLong Gold Antifade (Life Technologies P36930) and Fisher microscope coverglass (Fisher #12-545-83), allowed to dry overnight, and imaged on a Nikon Eclipse TI inverted fluorescent microscope.

Discussion

Both overexpression and reduced expression of CDCP1 are observed in mCRPC [11,14,36,37]. It has been well established that CDCP1 activates β1-integrin [9], enhances cell adhesion, promotes anchorage independent growth and tumor metastasis [9,38] and leads to activation of the AKT pathway [8]. In addition to pro-oncogenic activities of CDCP1, there is evidence for a tumor suppressive function [37]. The CDCP1 gene is located at the 3p21.3 region of the genome, which is lost in many cancer types [39]. The CDCP1 deficient PyMT breast cancer model possesses enhanced cancer initiation and growth [15,40]. Exfoliated cells from these tumors exhibit increased growth factor receptor signaling, reduced adhesion and anchorage dependence.

In the current study, we demonstrate that in detached cells, reduced CDCP1 expression unleashes FAK activation by promoting local generation of PI(4,5)P₂ through PIP5K1c. In CDCP1^Low DU145 cells, PIP5K1c becomes activated by SRC phosphorylation and recruitment of Talin. An association between active FAK-pY397 and PIP5K1c complex formation is observed in CDCP1^Low PC cells and in primary lymphocytes, which are CDCP1 negative. Since reduced CDCP1 expression also inhibits β1-integrin activation, the loss of CDCP1 can potentially coordinate the activation of FAK survival pathways with detachment from the substratum (Figure 7).

Figure 7.

Loss of CDCP1 coordinates cell detachment and survival. (A) Loss of CDCP1 (pink color) can occur through TGFβ1 [11], promoter methylation [48] or palmitoylation [49] and (B) is associtated with an altered gene expression program in the detached state and an increase in FAK activation. (C) Normally, loss of adhesion rapidly results in anoikis. (D) To overcome anoikis, FAK is activated through localized production of PI(4,5)P₂ by PIP5K1c. PIP5K1c activation requires SRC phosphorylation and Talin binding and is negatively regulated by CDK5 phsophorylation. By coordinating cell detachment and FAK activation, the loss of CDCP1 facilitates cell survival during tumor metastasis.

We confirmed and expanded previous data related to the underlying mechanism of PIP5K1c activation in the detached state. Forced expression of PIP5K1c in HEK293 cells led to the first evidence of a PIP5K1c-SRC-Talin complex under anchorage-independent growth conditions [35]. Our study demonstrates that this mechanism is regulated by CDCP1 and associated with activation of FAK in the detached state in two PC cell lines and in primary lymphocytes. While SRC is required for activation of PIP5K1c, CDK5 inhibits PIP5K1c through phosphorylation [41]. Reduced expression of CDCP1 in detached cells not only promotes SRC phosphorylation of PIP5K1c, but also of the regulatory subunit of CDK5. This causes inactivation of CDK5 and prevents phosphorylation of PIP5K1c by CDK5, which is inhibitory for PIP5K1c activation [42]. In addition, we demonstrate the activation of PIP5K1c in detached cells required human plasma or FBS, however responsible plasma factor and cell surface receptor are currently unknown.

In addition to regulation of FAK, CDK5 and PIP5K1c kinase activities by CDCP1, we identified gene expression differences of targetable transcription factors in CDCP1^Low versus CDCP1^High CTCs from patients. The expression of CDCP1 mRNA correlates with expression of ONECUT2, a transcription factor that promotes cell survival, castration resistance and neuroendocrine differentiation and that can be targeted by a chemical compound [26]. Another interesting association that is observed only in CTCs and not in bulk tumors lies between CDCP1, AR-v7 and HOXB13 mRNA expression levels. As is the case for ONECUT2, AR-v7 and HOXB13 alter the AR cistrome and cause resistance to AR signaling inhibitors. Thus, in CDCP1^High CTCs, ONECUT2 and CDK5 are possible therapeutic targets, while in CDCP1^Low CTCs, FAK may be an actionable drug target.

Altogether, CDCP1 regulated cytoplasmic kinase networks and associated gene expression provides a starting point for novel treatment approaches to combat mCRPC.

Results

FAK-pY397 is expressed in CDCP1^low circulating tumor cells (CTCs) from patients with metastatic, castration-resistant prostate cancer (mCRPC)

In patients, approximately 50% of metastatic and castration resistant PCs (mCRPC) express high CDCP1, while 50% of mCRPC express reduced levels of CDCP1 protein [14]. CDCP1 has emerged as a new treatment target in mCRPC with PTEN deletion, but the strategy only applies to CDCP1^High mCRPC cases [14]. Druggable pathways in CDCP1^Low mCRPC have not been elucidated. To explore potentially druggable and pro-metastatic mechanisms in CDCP1^Low PC, we investigated the CDCP1-FAK pathway and the cytotoxic effects of FAK inhibition in non-adherent PC cells.

Circulating tumor cells (CTC) from patients with mCRPC were used as a source of non-adherent, detached cancer cells. CTCs were isolated from 31 blood samples of 23 patients using the Nanovelcro system [24] for measurement of mRNA expression. The Nanovelcro platform leads to a large enrichment of CTCs over white blood cells. Since blood lymphocytes and monocytes express negligible amounts of CDCP1 [25], the CDCP1 mRNA signal in our CTC preparation is primarily derived from cancer cells. Each sample provided multiple CTCs that were pooled for RNA extraction. We measured a hand-curated panel of genes associated with aggressive PC that included CDCP1 on the Nanostring^® platform. In concordance with published data [11,14], we observed subgroups of CTCs with high versus low CDCP1 expression (Figure 1A). The cut-off between CDCP1^high and CDCP1^Low CTCs is based on the midpoint of the mean-centered CDCP1 expression. The Nanostring^® panel was also used for analysis of mRNA from PC-3 and DU145 cell lines in which we silenced CDCP1 by expression of a CDCP1 shRNA (cell line validation in [11]). A supervised cluster analysis based on CDCP1 expression demonstrated that the CDCP1-knockdown cell lines (Figure 1A, blue arrow) co-clustered with CDCP1^Low CTCs while the control knockdown cell lines (Figure 1A, red arrow) co-clustered with CDCP1^High CTCs.

Next, we identified eleven genes that correlated with CDCP1 expression with a correlation coefficient, r > 0.3 and a p-value, P < 0.01 (Figure 1A). These include the AR-V7 splice variant (AR-CE3), HOXB13, ONECUT2, CELSR3, TPX2, and CENPF. These genes are known for their association with aggressive PC and are directly regulated by the androgen receptor (AR) [26]. While mRNA expression of AR-V7, HOXB13 and ONECUT2 is significantly increased in CDCP1^high CTCs (P < 0.01), expression of TMPRSS2 is significantly decreased (P < 0.01) (Figure 1B). We previously demonstrated how ONECUT2 represses the AR axis at multiple levels [26]. Since a recent study demonstrated that the AR-axis inhibits CDCP1 gene expression [14,27], we postulated that the suppression of the AR by ONECUT2 leads to increased CDCP1. To test this premise, we examined the relationship between CDCP1 and ONECUT2 activity (i.e. the z-score of the ONECUT2 cistrome) in bulk tumor tissue. As expected, we observed a positive correlation between CDCP1 gene expression and ONECUT2 activity (Spearman’s rho = 0.41, P = 4 × 10^-4) and a negative correlation between CDCP1 and ONECUT2 or AR transcription factor activities (Figure S1).

Next, we interrogated FAK activation in CDCP1^High and CDCP1^Low CTC populations of patients with mCRPC. We utilized cryopreserved CTCs from patients previously analyzed for CDCP1 mRNA expression (Figure 1A). While CDCP1 mRNA was measured in pooled CTCs from each case, FAK activation was measured through expression of FAK-pY397 in individual CTCs using fluorescent in-situ detection with the FAK-pY397 antibody. CTCs from three patients in the CDCP1^High and five patients in the CDCP1^Low groups were retrieved for analysis. Cases in the CDCP1^High group contained 4-7 CTCs for analysis, while CDCP1^Low cases only harbored 0-3 CTCs. Therefore, to analyze a comparable number of CTCs in both groups required more CDCP1^Low than CDCP1^High cases. CTCs were identified through positive cytokeratin and DAP1 and negative CD45 fluorescent staining and FAK-pY397 expression quantified in CTCs using a forth antibody-fluorophore label (Figure S2). In the CDCP1^Low group 3/5 (60%) patients demonstrated high FAK-pY397 expression in at least one CTC, while in the CDCP1^High group, 1/3 (33%) patients expressed high FAK-pY397 (Figure 1C). Because of antibody and imaging limitations, double staining for CDCP1 and FAK-pY397 was not feasible. Therefore, it is uncertain whether intra-patient variability of FAK-pY397 is related to heterogeneous CDCP1 expression across individual CTCs. A statistical analysis of FAK-pY397 data did not reveal a significant difference in FAK-pY397 expression between CDCP1^High and CDCP1^Low groups because of the low sample size. Nevertheless, there appears to be a trend between the association of reduced CDCP1 expression and FAK activation in the detached state.

To determine the association between FAK activity and PC aggressiveness and progression, we investigated FAK in bulk patient tumors. Gene amplification can result in kinase activation [5]. The FAK gene resides at 8q24.3 and is amplified in 6% of prostate adenocarcinomas and 38% of neuroendocrine PC (Figure 2A). In addition to its role in focal adhesion turnover [28], FAK can translocate into the nucleus to regulate gene expression [29]. Both, up (FAK^UP) and down (FAK^DOWN) regulated FAK-driven gene signatures have been identified in squamous cell carcinoma cell lines [30]. FAK activation signatures were analyzed in the Prostate Cancer Transcriptome Atlas (PCTA, n = 1352) [31], which is a cohort of cases that spans the entire PC disease course from benign prostate tissue to metastatic and castration-resistant PC. While the signature of genes positively regulated by FAK did not show a significant change during PC progression (Figure S3), the signature of genes that are suppressed by FAK revealed a remarkable decline along the PC disease course (Figure 2B). The consistent decline of the FAK^DOWN gene signature with increasing cancer grade and stage is suggestive of an increase in FAK activity during PC progression.

Figure 2.

FAK activation in prostate cancer cells. A. FAK gene amplifications (AMP), deletions (DEL) or mutations (MUT) in neuroendocrine PC (NEPC, n = 114) or prostate adenocarcinoma (TCGA, n = 494). B. A FAK-repressed gene signature increases with PC progression. A panel of FAK-repressed genes (Table S1) is analyzed in the Prostate Cancer Transcriptome Atlas (PCTA, n = 1352 [31]). The z-scores of the FAK signature in cases of benign prostate, low grade (Gleason Sum < 7), intermediate grade (Gleason Sum = 7), high grade (Gleason Sum > 7) and metastatic castration-resistant PC (mCRPC) are shown. The p-value is from the ANOVA test of z-scores of the FAK-repressed gene signature. C. Xenograft analysis of CDCP1^Low DU145 cell line xenografts. Whole cell lysates from xenografts were analyzed by Western blotting for expression of FAK-pY397 and activation state of β1-integrin. The HUTS4 antibody binds selectively to the active configuration of β1-integrin. D. CDCP1 loss promotes survival in the detached state. Images of adherent (upper panel) and detached (lower panel) CDCP1^low cultures. At indicated time points, cells were stained with Trypan blue and viable cells were counted. Growth curves of adherent and detached cultures. Absorbance data from MTT assays [31]. **P < 0.01.

CDCP1^low PC xenografts express activated FAK despite inactive β1-integrin

To investigate how FAK is activated in CDCP1^Low PC cells, we employed cell culture models of DU145 and PC-3 that we published previously [32]. In cells derived from these models, reduced expression of CDCP1 stimulates DU145 xenograft growth (Figure 2C and [11]). We generated five xenografts from the CDCP1^Low cell populations that were established by transfection of shCDCP1 without further clonal selection. As a result of heterogeneous CDCP1 expression in the inoculated cells, four of five xenografts did not express CDCP1, while one xenograft was CDCP1 positive (Figure 2C). FAK-pY397 expression was increased in the four CDCP1^Low xenografts compared to CDCP1^High xenograft. Interestingly, three of four xenografts with active FAK-pY397 expression revealed greatly diminished active β1-integrin as measured by the HUTS4 antibody which is specific for the active state of β1-integrin. These data are consistent with our previous study showing that β1-integrin is inactivated after silencing of CDCP1 and demonstrate that FAK activation can occur in cancers with inactive β1-Integrin. To determine whether reduced CDCP1 expression promotes tumor cell growth in the detached state, we employed a suspension culture system (Figure 2D). Growth curves of CDCP1^High and CDCP1^Low DU145 cells were identical in adherent culture, but differed significantly in suspension culture (Figure 2D right panel). The main difference between CDCP1^High and CDCP1^Low PC cells was observed during the first 24 hours in suspension culture, consistent with a pro-survival effect of reduced CDCP1 expression. After 24 hours in culture, those CDCP1^High cells that survived, proliferated at a rate similar to CDCP1^Low cells, as indicated by the parallel growth curves. Since FAK activation is known to stimulate cell survival in the detached state, we hypothesized that FAK activity in CDCP1^Low cells is the reason for the increased survival of CDCP1^Low PC cells.

Low CDCP1 sensitizes detached PC cells to killing by FAK inhibitors

To further pursue the mechanism underlying the potential inverse relationship of FAK and β1-integrin activation in cell line xenografts, we analyzed adhesion and suspension cultures of DU145 and PC3 cells (Figure 3). CDCP1^High and CDCP1^Low DU145 and PC3 cells were compared for expression of FAK-pY397 and HUTS4 (Figure 3A). Cells were suspended for 2 hours in media containing 10% FBS. In addition to phosphorylation of FAK-Y397, we observed phosphorylation of FAK-Y576/577 and FAK-Y925, which are SRC phosphorylation sites in the FAK protein. We did not examine the FAK-Y861 phosphorylation site, since its regulation by CDCP1 has been previously analyzed [33]. In addition to FAK activation, SRC, MAPK and AKT activities were increased in detached CDCP1^Low compared to CDCP1^High PC cells (Figure 3B). To determine whether extracellular factors are required for FAK activation, we tested human plasma as a source of factors for FAK activation in CTCs and primary human lymphocytes and fetal bovine serum (FBS) as a source of factors in detached cell cultures. In addition, we tested the sensitivity of FAK activation to trypsinization conditions. Cells were released from culture plates using either trypsin or EDTA. After a washing step, cells were suspended in media containing 10% human plasma, 10% FBS or 1% BSA. The cells were kept in a detached and floating state without adhesion to plastic on HEMA coated plates for the duration of the experiment. FAK-pY397 expression in DU145 and PC3 cells required FBS or human plasma (Figure 3C, 3D). Surprisingly, FAK phosphorylation was only observed after 2 hours in suspension after detachment by trypsin (Figure S4). In contrast, when cells were released from plastic dishes with EDTA, FAK phosphorylation occurred after 30 minutes (Figure 3D). FAK was not activated by treatment with a preparation of exosomes that were isolated from the serum-free conditioned medium of CDCP1^Low cells (Figure S5). Altogether, these data demonstrate that FAK activation in CDCP1^Low cells requires human plasma or FCS and a trypsin sensitive cell surface receptor.

Figure 3.

CDCP1 loss sensitizes DU145 PC cells to FAK Inhibitor treatment. A. Suspension-activation of FAK. Cell lysates from CDCP1^Low and CDCP1^High DU145 and PC3 cells were probed with anti-FAK-pY379 (FAK autophosphorylation site), anti-FAK-pY576/7 (SRC phosphorylation site) and anti-FAK-pY925 (SRC phosphorylation site). B. Detached CDCP1^Low or CDCP1^High PC cells were probed for active, phosphorylated FAK, SRC, MAPK and AKT and reprobed for total protein expression. C. FAK activation requires fetal bovine serum (FBS). Cells were cultured in suspension with 10% FBS (F) or 1% bovine serum albumin (BSA, B). Western blots were probed for FAK-pY397 and reprobed for total FAK. D. FAK activation by human plasma. CDCP1^Low DU145 and PC3 PC cells were cultured under detached conditions in 10% human plasma (squares) or BSA (circles) containing media for times indicated and probed for expression of FAK-pY397. Cells were released from culture dishes with either trypsin (red) or EDTA (yellow). The y-axis represents the relative amount of FAK-pY397 expression based on quantification of Western Blot bands in Figure S3. E, F and H. Growth inhibition of DU145 cells treated with single drug treatments of FAK inhibitor, PF573228, SRC inhibitor, AZD0530 (Saracatinib) or MEK inhibitor (PD98059) at concentrations indicated on the x-axis. Cells were preincubated with drugs before release from culture plates and drug treatments were continued in suspension for 24 hours. G, I and J. Growth inhibition of DU145 cells treated with drug combinations. DU145 cells were treated with 2 uM PF 573228 (FAK inhibitor) plus AZD0530 (Saracatinib, SRC inhibitor), PD98059 (MEK inhibitor) or AZD0530 and PD98059. X-axis labels indicate drug concentrations and Y-axis labels the percentages of viable cells. *P < 0.05, **P < 0.01.

Next, we questioned whether detached cells with increased FAK activation are more sensitive to FAK inhibitors. Cells were treated with FAK, SRC and MEK inhibitors and drug effects confirmed by Western blotting (Figure S6). We observed that loss of CDCP1 sensitized cells to FAK inhibitors. The GI50 for the FAK inhibitor, PF573228, in CDCP1^low cells was 3 uM compared to 6 uM in CDCP1^high cells. The greatest difference in cytotoxicity was observed reproducibly at the 4 uM concentration (Figure 3E). Adding the SRC inhibitor, Saracatinib, or a MEK inhibitor to 2 uM FAK inhibitor demonstrated an additive effect (Figure 3F-I). The combination of 3 drugs revealed lower cytotoxic concentrations of individual compounds, suggestive of an additive effect (Figure 3J). In contrast to the FAK kinase inhibitor, an inhibitor directed against the FAT domain of FAK or the expression of a FAT-domain mutant FAK did not demonstrate reduced FAK-pY397 expression in the detached state of CDCP1^Low DU145 cells (Figure S7). These results provide pre-clinical evidence that FAK inhibitors might be more effective in PC cells with low CDCP1 expression.

Detached, CDCP1^low PC cells activate FAK independent of β1-integrin

We previously demonstrated that CDCP1 maintains β1-integrin in an active configuration after cell detachment through CDK5-mediated phosphorylation of Talin and that reduced CDCP1 leads to loss of inside-out activation of β1-integrin [11]. Since FAK is activated downstream of β1-integrin in adherent cells, we questioned whether active CDK5/β1-integrin was a requirement for FAK activation in detached, non-adherent cells and controlled the activation state of β1-integrin in the detached state through silencing of CDCP1. Lysates from CDCP1^Low and CDCP1^High cells were used in co-immunoprecipitation experiments to examine β1-integrin and CDK5 complex formation with FAK. As expected, β1-integrin in detached CDCP1^High PC cells is complexed to pTalin/CDK5/CDKR1, which maintains the inside-out activation of β1-integrin (Figure 4A, green box). However, surprisingly, despite β1-integrin activaion, FAK kinase is inactive in the β1-integrin complex. In contrast, in detached CDCP1^Low DU145 cells, FAK is active while β1-integrin is inactive (Figure 4A). In addition, reduced CDCP1 expression in the detached state leads to dissociation of CDK5 from the β1-integrin complex. CDK5 in detached CDCP1^Low cells is inactive and exists in a complex with active FAK-pY397. Thus FAK-pY397 resides in two separate complexes, one with β1-integrin and the other with CDK5 (Figure 4A, 4B). Interestingly, Talin-pS425, which is found in the β1-integrin complex in CDCP1^High cells, moves into the CDK5 complex in CDCP1^Low cells (Figure 4A, 4B).

Figure 4.

FAK protein complexes are regulated by CDCP1 in the detached state. Comparison of protein complexes in detached CDCP1^High and CDCP1^Low DU145 cells. Protein lysates from CDCP1^High and CDCP1^Low DU145 cells were divided into two equal parts for analysis of either β1-integrin complexes or CDK5 complexes by co-precipitation. A. β1-integrin/Talin/FAK complex. β1-integrin was precipitated and the protein complex was probed for FAK-pY397/total FAK, Talin-pS425/Talin, CDK5 and CDK5R1. B. CDK5/FAK complex. CDK5 was precipitated and the complex probed for FAK-pY397/FAK, HUTS4 (active β1-integrin)/total β1-integrin and Talin-pS425/Talin. C. Schematic illustration of FAK complexes in CDCP1^High versus FAK-pY397 in CDCP1^Low cells. The serum dependence of complex formation is indicated by the color of the bar. C-control antibody precipitation, WCL-whole cell lysate, IP-immunoprecipitation.

Since FAK activation requires FBS, we analyzed the regulation of FAK complexes by FBS. Complex formation of inactive β1-integrin with active FAK-pY397 requires FBS, while FAK-CDK5 complexes form without FBS (Figure S8). A summary of regulation of FAK complexes by CDCP1 is shown in Figure 4C. Together the data support a model in which reduced expression of CDCP1 in the detached state leads to β1-integrin-independent FAK activation.

In the detached state, FAK is activated by PI(4,5)P₂

Since FAK activation (i.e. expression of FAK-pY397) in detached CDCP1^Low PC cells occurs in the absence of active β1-integrin, we examined alternative mechanisms of FAK activation. One mechanism of FAK activation is through the binding of PI(4,5)P₂ to the FERM domain (Figure 5A). To demonstrate that FAK can be activated by PI(4,5)P₂ in the detached state, we treated suspended cells with exogenous PI(4,5)P₂. PI(4,5)P₂ activation of FAK was concentration dependent and did not require FBS (Figure 5B, 1% BSA condition). Furthermore, exogenous treatment with PI(4,5)P₂ resulted in FAK-pY397 in CDCP1^High cells, demonstrating that PI(4,5)P₂ reverses the upstream inhibition of FAK activation by CDCP1 (Figure 5B).

Figure 5.

Suspension-activation of FAK by PtdIns(4,5)P₂ (PI(4,5)P₂). A. Binding to PI(4,5)P₂ releases the autoinhibition of the FERM domain. Activation of the FAK kinase results in FAK-pY397 (Y1) expression. B. Exogenous PI(4,5)P₂ activates FAK. Detached CDCP1^High DU145 cells, which remain FAK-pY397 negative in 10% FCS or 1% BSA were treated for 3 hours with PI(4,5)P₂ at concentrations specified above each column. Western blots were probed for FAK-pY397 expression. Detached CDCP1^Low DU145 cells (FAK-pY397 negative) were cultured in 1% BSA and treated with PI(4,5)P₂ at concentrations specified above each column. Western blots were probed for FAK-pY397 expression. C. Expression of PIP5K1c splicing isoforms. Spearman correlation coefficients between total PIP5K1c mRNA expression and expression of PIP5K1c-201 (Exon 17 positive) and PIP5K1c-205 (Exon 17 negative) splicing isoforms in intermediate risk PC (n = 144, [34]). D. Association of PIP5K1c-201 with PTEN and TP53 loss. Expression levels of PIP5K1c-201 mRNA, indicated as TPM (transcripts per million) are compared in cases with an intact PTEN or TP53 gene locus versus cases with PTEN or TP53 bi-allelic deletions. Deletions were detected by the Oncoscan SNP-chip arrays. Box and whisker plots with horizontal lines in boxes representing the 1st, 2nd and 3rd quartiles. Whiskers are indicated outside the box with limits of 1.53 of the interquartile range. **p = 2.36 X 10^-3; ++p = 8.18 X 10^-3. E. PIP5K1c-FAK complexes lack β1-integrin. PIP5K1c was precipitated from detached CDCP1^High and CDCP1^Low DU145 cells and probed for β1-integrin and FAK. F. Serum dependence of FAK-PIP5K1c complexes. G. PIP5K1c-D316A mutant. A GFP-fusion protein of a kinase-dead mutant PIP5K1c or of wild type PIP5K1c was expressed in detached CDCP1^Low DU145 cells. Cells were cultured in the detached state with 10% FCS for 3 hours before precipitation of PIP5K1c. Immunoprecipitated complexes were analyzed for FAK-pY397 and FAK. C-control antibody precipitation, WCL-whole cell lysate, IP-immunoprecipitation. H. Schematic representation of PIP5K1c-FAK complexes in CDCP1^High versus CDCP1^Low PC cells. Talin-PIP5K1c complexes are shown in detail in Figure 6.

In the detached state, reduced CDCP1 activates PIP5K1c/PI(4,5)P₂ through SRC phosphorylation and Talin binding

PI(4,5)P₂ is produced by three PIP5K isoforms that differ in subcellular localization. Only the PIP5K1c-201 isoform is associated with cell adhesion complexes. PIP5K1c-201 differs from other isoforms by the inclusion of Exon 17, which contains a unique SRC phosphorylation site [22]. Analysis of cases from a cohort of intermediate risk PC patients [34] revealed that PIP5K1c-201 strongly correlates with total PIP5K1c mRNA expression (r = 0.99) (Figure 5C). Further, PIP5K1c-201 was significantly reduced in tumors with PTEN and TP53 gene deletions, which are genetic alterations associated with aggressive PC behavior (Figure 5D).

To determine whether FAK activation through PIP5K1c is regulated by complex formation in detached CDCP1^Low PC cells, we immunoprecipitated PIP5K1c and probed for FAK and β1-integrin (Figure 5E). Although we observed the formation of PIP5K1c-FAK-pY397 complexes, β1-integrin was not part of these complexes. Further, the formation of PIP5K1c-with inactive FAK in CDCP1^Low PC cells occurred spontaneously without a requirement for FBS, but FAK activation (i.e. FAK-pY397) requires FBS (Figure 5F). To determine the role of PI(4,5)P₂ for FAK activation in detached cells, we expressed a kinase-dead PIP5K1c, PIP5K1c-D316A. PIP5K1c-D316A blocked the activation of FAK in a dominant negative manner by competing with wild type PIP5K1c for binding to FAK (Figure 5G). In summary, in the detached state, reduced expression of CDCP1 leads to PIP5K1c activation, local production of PI(4,5)P₂, changes of FAK protein complexes and activation of FAK (Figure 5H).

Next we investigated the mechanism leading to PIP5K1c activation in the detached state. PIP5K1c has previously been demonstrated to reside in a complex with SRC and Talin under anchorage independent growth conditions. Under these conditions, PIP5K1c requires SRC phosphorylation and Talin binding for activation [35]. To determine whether reduced expression of CDCP1 in the detached state leads to formation of Talin-PIP5K1c complexes, we compared Talin-PIP5K1c and Talin-β1-integrin complexes in CDCP1^High and CDCP1^Low cells. We precipitated β1-integrin from lysates of detached CDCP1^High DU145 cells and PIP5K1c from lysates of detached CDCP1^Low DU145 cells. As expected in detached CDCP1^High PC cells, Talin forms a complex with the active β1-integrin (Figure 6A blue arrow). However, upon reduced expression of CDCP1 (CDCP1^Low columns in Figure 6A), Talin no longer interacts with active β1-integrin. Instead, Talin forms a complex with PIP5K1c in CDCP1^Low but not in CDCP1^High cells (Figure 6A, red arrow). This complex formation requires 10% FBS (Figure S9). The switch of Talin from β1-integrin to PIP5K1c was also observed in CDCP1^Low xenografts (Figure 6B). Furthermore, in CDCP1^Low xenografts, FAK-pY397 is associated with Talin-PIP5K1c complexes (Figure 6C). Next, we determined whether SRC activity is needed for formation of Talin-PIP5K1c complexes in detached CDCP1^Low PC cells. PC cells were treated with 10 uM Saracatinib, a specific SRC kinase inhibitor during culture in suspension. Treatment with Saracatinib abolished Talin-PIP5K1c complexes (Figure 6D).

Figure 6.

Suspension-activation of PIP5K1c through a Talin switch. A. β1-integrin and PIP5K1c co-immunoprecipitation with Talin. CDCP1^High and CDCP1^Low DU145 and PC3 cells were used for precipitation of β1-integrin or PIP5K1c. Western blots of protein complexes were probed for Talin-pS425 and total Talin. The blue arrow points to Talin in complex with β1-integrin in CDCP1^High cells, while the red arrow points to Talin complexed to PIP5K1c in CDCP1^Low cells. B. Talin complexes in mouse xenografts. Co-immunoprecipitation of PIP5K1c and β1-integrin in CDCP1^High or CDCP1^Low DU145 xenografts. Xenografts #1 (CDCP1^High ) and #4 (CDCP1^Low) described in Figure 2C were used. The blue arrow points to Talin in complex with β1-integrin, while the red arrow points to Talin complexed to PIP5K1c. C. FAK-pY397-Talin complexes in CDCP1^Low DU145 cells. D. PIP5K1c-Talin complex formation requires SRC phosphorylation. PIP5K1c was immunoprecipitated from detached CDCP1^Low DU145 cells treated with 10 uM Saracatinib or control vehicle for 48 hours. Immunoprecipitates were probed for Talin. E. Primary, lymphocytes from 2 separate individuals were cultured in 10% FBS (F) or 1% BSA (B) and analyzed for expression of FAK-pY397 and total FAK protein by Western blotting. F. The same lymphocyte lysates were used for immunoprecipitation of PIP5K1c. PIP5K1c complexes were probed for expression of Talin-pS425 and total Talin. C-control antibody precipitation, WCL-whole cell lysate, IP-immunoprecipitation. G. Schematic illustration of Talin complexes in CDCP1^High versus CDCP1^Low PC cells.

To demonstrate the association between FAK activation (i.e. FAK-pY397 expression) and Talin-PIP5K1c complexes in a different system, we analyzed primary lymphocytes. Lymphocytes are naturally non-adherent and CDCP1 negative. Lymphocytes from two separate individuals were cultured in 10% FCS or 1% BSA and analyzed for expression of FAK-pY397 (Figure 6E). The same lysates were analyzed for PIP5K1c-Talin complexes by immunoprecipitation. Western blotting confirmed that 10% FCS induces FAK-pY397 and leads to formation of PIP5K1c-Talin complexes (Figure 6F), while culture in BSA does not. In summary, when cells are maintained in a detached state, SRC phosphorylates PIP5K1c, which causes the switch of Talin from β1-integrin to activate PIP5K1c (Figure 6G). In these cells, activation of PIP5K1c is associated with expression of FAK-pY397.

Disclosure of conflict of interest

None.