Neural Cell Adhesion Protein CNTN1 Promotes the Metastatic Progression of Prostate Cancer

Prostate cancer is the most frequently diagnosed malignancy and the second/third leading cause of cancer-associated deaths for men in the developed world (1). The disease progresses from high-grade prostatic intraepithelial neoplasia (HGPIN), to locally invasive carcinoma and finally to metastatic cancer. Prostate cancer predominantly metastasizes to the bone (2). Although local tumors can be well controlled through watchful waiting, surgery, and radiation, metastatic prostate cancer remains incurable. Therefore, identification of key molecules involved in prostate cancer metastasis is of an utmost importance for prognostic and therapeutic purposes. The capacity of primary tumor cells to metastasize relies on the integration of a complex network of signals, including those of the environment. As a result of signal integrations, cellular alterations occur that promote prostate cancer progression and metastasis. The typical changes for epithelium-originated tumors include the downregulation of E-cadherin leading to a reduction of cell–cell adhesion, a characteristic event of epithelial–mesenchymal transition (3, 4).

In addition to the reduction of E-cadherin, elevation of cell adhesion molecules (CAM), especially Ig-like neural CAMs plays an important role in cancer progression and metastasis. The Ig-like CAMs include contactins, the neural CAM (NCAM), and L1CAMs comprising of CAM L1-like (CHL1), neuronal CAM (NrCAM), and neurofascin (5, 6). Increases in L1CAM are associated with poor prognosis in patients with ovarian, uterine, colon, breast, and lung cancers (7–10). Interestingly, in addition to L1CAM, increased expression of contactin 1 (CNTN1 and F3/contactin) was detected in lung carcinoma. CNTN1 promoted lung cancer metastasis (11, 12) and its expression associated with poor prognosis for patients with lung carcinoma, esophageal, and oral squamous cell carcinomas (11, 13, 14). Nevertheless, to the best of our knowledge, the role of CNTN1 in prostate cancer progression and metastasis is unknown. CNTN1 is a neural cell adhesion protein consisting of 6 N-terminal Ig domains followed by 4 fibronectin (FN)-like repeats (15) and plays important roles in the development of the central nervous system, including lineage commitment and precursor proliferation (16, 17). CNTN1 also promotes axon elongation in the cerebellum, the formation of the septate-like junctions between axons and myelinating glial cells and the formation of the neuromuscular junction (5, 18, 19). Collectively, it appears that the physiologic functions of the Ig-like CAMs in neuron guidance and migration are commonly hijacked by tumors for metastasis.

The above observations would suggest contributions of an Ig-like CAM to prostate cancer metastasis. To pursue this possibility, we have taken advantage of our recently established conditions to isolate prostate cancer stem-like cells (PCSC), a cell population that plays a major role in cancer metastasis (20, 21). Specifically, we were able to culture PCSCs as suspension spheres from DU145 cells. These sphere cells are likely PCSCs because of their elevated (100-fold) ability to initiate tumors in NOD/SCID mice (22). With the availability of PCSCs and their isogenic non-PCSCs, we have profiled the respective gene expression, and intriguingly, CNTN1 was specifically identified in DU145 cell-derived PCSCs and in PC3 cells cultured under PCSC conditions. On the basis of the reported role of CNTN1 in promoting lung cancer metastasis (11, 12), we proceeded to determine CNTN1's ability to stimulate prostate cancer invasion and metastasis. Finally, we examined the association of CNTN1 expression with prostate cancer progression and biochemical recurrence (BCR)-free survival using clinical specimens.

DU145, PC3, and LNCaP cells were from the ATCC and cultured accordingly in Minimum Essential Medium, F12 Medium and RPMI-1640 Medium supplemented with 10% FBS (Sigma) and 1% Penicillin–Streptomycin (Life Technologies). The androgen-independent LNCaP derivative C4-2 cell line was kindly provided by Dr. Martin Gleave at The University of British Columbia, British Columbia, Canada (23). All cell lines were thawed fresh every 2 months; no further authentication was performed. CNTN1 shRNA and control shRNA were from Santa Cruz Biotechnology. CNTN1 shRNA is a pool of three different plasmids with the following short hairpin sequences (5′-3′): (A) GATCCGGGTGATAATTGAATGCAATTCAAGAGATTGCATTCAATTATCACCCTTTTT (B) GATCCCAAGAGCAGTGGACTTAATTTCAAGAGAATTAAGTCCACTGCTCTTGTTT TT (C)GATCCGCATCCTTGTCTACTTGGATTCAAGAGATCCAAGTAGACAAGGATGCT TTTT. The sequence for the scrambled control shRNA is proprietary and could not be provided by the company. CNTN1 isoform 3 cDNA was from Open Biosystems (GE Healthcare).

DU145 spheres were isolated as described previously (22). DU145 monolayer cells were individualized and resuspended at a density of 5,000 cells/mL in serum-free (SF) media supplemented with 0.4% BSA (Bioshop Canada Inc.) and 0.2 × B27 minus Vitamin A (4 mL/L, Life Technologies).

Hairpin shRNAs were expressed by a retroviral-based shRNA vector as described previously (24). Briefly, a gag-pol, a rev and an envelope-expressing vector (Stratagene) were transiently cotransfected with a designed retroviral plasmid into HEK293T cells. The viral vector-containing medium was harvested 48 hours later, filtered through a 0.45μm filter and centrifuged at 20,000 × g for 120 minutes to concentrate the viral vectors. The quality and titers of viral vectors were not determined. However, by visual inspection we consistently had less than 30% cell death following transduction and selection. Cells were selected for stable integration with puromycin (1 μg/mL; Sigma).

RNA was isolated with TRizol (Life Technologies). Gene expression was examined using the Affymetrix Human Gene 1.0 ST microarrays by the University Health Network Microarray Center in Toronto. Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-4118. Functional analysis of gene expression was performed using Ingenuity Pathway Analysis (Ingenuity).

Western blots were performed as described previously (24). The antibodies used were: anti-CNTN1 (1:200; R&D Systems), Anti-AKT Ser473 phosphorylation (1:1,000; Cell Signaling Technology), anti-AKT (1:1,000; Santa Cruz Biotechnology), anti–E-cadherin (1:2,500; BD Biosciences), anti-actin (1:1,000; Santa Cruz Biotechnology), anti-goat (1:3,000; Santa Cruz Biotechnology), anti-mouse (1:3,000; GE Healthcare), and anti-rabbit (1:3,000; GE Healthcare).

Prostate tissues were collected from patients who underwent biopsies or radical prostatectomy at St. Joseph's Hospital in Hamilton (Ontario, Canada) under the approval from the local Research Ethics Board (REB#11-3472) and with consent from patients.

IHC was performed as previously described (24). Primary antibody specific for CNTN1 (1:10; Sigma) and prostatic acid phosphatase (PAP; 1:300; Abcam) was incubated with the sections overnight at 4°C. Images were taken with a light microscope (Olympus).

Tissue microarray (TMA) 2 and TMA5 were obtained from the Cooperative Prostate Cancer Tissue Resource (CPCTR). TMA2 was organized according to Gleason score and consisted of 1,128 prostate tissue cores derived from 250 prostate cancer tumors, 32 nontumor and 58 HGPIN cases (Supplementary Table S1). TMA5 was organized according to the BCR that was defined by CPCTR as an increase in serum PSA levels >0.6 ng/mL (single value) or consecutive rise in serum PSA levels between 0.4 and 0.6 ng/mL after radical prostatectomy (25). TMA5 contained 1,616 prostate cancer tissue cores derived from 404 patients based on their BCR (Supplementary Table S1). IHC was performed as follows. Primary antibody specific for CNTN1 (1:3,000; Sigma) was incubated with the sections overnight at 4°C, followed by the use of a TSA Plus Biotin kit (PerkinElmer) according to the manufacturer's protocol. TMA slides were scanned using a ScanScope and analyzed using the ImageScope software (Aperio). All spots were also manually examined and scored. The scores obtained using the ImageScope software were representative of the scores obtained manually. The former scores were converted to H-Scores using the formula [(H-Score = % positive X (intensity+1); ref. 26]. Corrected H-Scores were achieved by subtracting the H-Scores obtained from the stroma regions with H-Scores obtained from the prostate glands for each core. Scores were assigned to a scale of 0 to 3 (0-negative, 1-weak, 2-moderate, and 3-strong staining). Positive staining for CNTN1 was classified as moderate to strong staining with an H-score ≥10.

Staining was performed as previously described using a goat anti-CNTN1antibody (1:50; R&D Systems; ref. 24).

Reverse transcription was performed using superscript III (Life Technologies). PCR primers used are: CNTN1 (forward) 5′-CAATAGTGCAGGGTGTGGAC-3′ and (reverse) 5′-TGGCTAGGAGGTGCTTTCTT-3′. Actin (forward) 5′-ACCGAGCGCGGCTACAG-3′ and (reverse) 5′-CTTAATGTCACGCACGATTTCC-3′. Real-time PCR was performed using the ABI 7500 Fast Real-Time PCR System (Applied Biosystems) in the presence of SYBR-green according to the manufacturer's instructions. All samples were run in triplicate.

DU145 EV and CNTN1 (1,000/well) cells were seeded into a 96-well plate. Proliferation was daily measured using the WST-1 Cell Proliferation Assay Kit (Millipore) according to the manufacturer's instructions. Absorbance readings were measured with a plate reader at 420 nm.

Cells were seeded at the indicated number in triplicates and cultured for 1 week. Cells were stained with crystal violet (0.5%). Quantification was carried out by taking images of four quadrants per well and pixel count for crystal violet staining was determined using ImageScope.

Modified Boyden chambers consisted of inserts with an 8-μm-pore membrane coated with Matrigel (BD Biosciences) placed in a 24-well plate. Assays were performed as previously described (24).

10⁶ DU145 EV (n = 4) and CNTN1 (n = 5) as well as 10⁴ DU145 shCTRL (n = 4) and shCNTN1 (n = 6) sphere cells were resuspended in media per Matrigel mixture (1:1 volume; 0.1 mL), followed by subcutaneous implantation in flanks of 8-week-old male NOD/SCID mice (The Jackson Laboratory). Tumor volumes were determined using a caliper according to the standard formula: L × W² × 0.52, where L and W are the longest and shortest diameters, respectively. For the generation of lung metastasis, 10⁶ DU145 EV (n = 4) and CNTN1 (n = 5) cells were resuspended into 0.3 mL of PBS and injected through the tail vein of NOD/SCID mice. Lungs were harvested at 16 weeks after injection. Tumor volumes were photographed and measured using ImageJ. All animal work was carried out according to experimental protocols approved by the McMaster University Animal Research Ethics Board.

Statistical analysis was performed using the Student t test unless otherwise specified. Two-way ANOVA was used to determine significance for tumor growth. For comparison between H-score intensity and prostate cancer Gleason scores, a one-way ANOVA was performed. For comparison between CNTN1 staining for noncancer versus cancer and between prostate cancer Gleason a chi-square test was performed. A log-rank test was performed to assess statistical significance between survival curves. All tests were two tailed. A P value of <0.05 was considered statistically significant.

To examine whether DU145 sphere cells (PCSCs) were associated with unique prostate cancer–promoting factors, we profiled the gene expression in DU145 monolayer (non-stem cells) and PCSCs. Among the set of genes with differential expression CNTN1 was prominent. Its expression was increased approximately 10-folds in sphere cells compared with monolayer cells (Fig. 1A). The upregulation was confirmed by both real-time PCR (Fig. 1B) and Western blot analysis (Fig. 1C). Consistent with CNTN1 as a cell surface adhesion protein in neural cells, the surface presence of CNTN1 in DU145 sphere cells was apparent under both immunofluorescence (Fig. 1D) and IHC (data not shown) staining. To further confirm the CNTN1 upregulation is unique to PCSCs, we were able to show that both LNCaP and PC3 cells did not express CNTN1 when cultured in the presence of 10% FBS (Fig. 1C). In contrast, a significant increase in CNTN1 mRNA was detected in PC3 cells cultured under serum-free (PC3-SF) PCSC-culturing conditions (Fig. 1E). PC3-SF cells do not proliferate at a slower rate than PC3 cells cultured in a FBS-containing medium (data not shown). In addition, PC3-SF cells adopted a different morphology appearing more rounded and less elongated, compared to PC3 cells cultured in FBS-containing medium (Supplementary Fig. S1A). PC3-SF cells formed significantly larger colonies (Supplementary Fig. S1B–S1C), suggesting that PC3-SF cells have acquired PCSC characteristics. Under SF-PCSC culture conditions, LNCaP cells did not survive (data not shown). Collectively, the above observations reveal upregulation of CNTN1 in PCSCs.

The cell surface localization of CNTN1 in DU145 spheres suggests CNTN1 being functional. To address this, ectopic CNTN1 was expressed in DU145 monolayer cells (Fig. 1F). In comparison with DU145 empty vector (EV) cells, ectopic CNTN1 did not alter cell proliferation (Supplementary Fig. S2A), but significantly enhanced DU145 cell invasion (Fig. 1G). To further support these observations, a CNTN1 expressing stable line was also constructed in LNCaP C4-2 cells (Supplementary Fig. S2B). Compared with the EV cells, ectopic CNTN1 did not affect the cell's ability to form colonies (Supplementary Fig. S2C), but promoted LNCaP C4-2 cell's invasion ability (Supplementary Fig. S2D). Collectively, we provide evidence that CNTN1 promotes prostate cancer cell invasion.

One of the major pathways affecting cell invasion is the PI3K–AKT pathway (27). To examine the impact of CNTN1 on AKT activation, we demonstrated an enhancement of AKT activation in DU145 CNTN1 and LNCaP C4-2 CNTN1 cells in comparison with their respective EV cells in response to serum stimulation (Fig. 2A, Supplementary Fig. S2F), despite ectopic CNTN1 not affecting the basal levels of AKT activation in both lines (Supplementary Fig. S2E). Conversely, knockdown of CNTN1 in DU145 PCSCs reduced AKT activation (Fig. 2B). To further determine the relevance of CNTN1-affected AKT activation during prostate tumorigenesis, we examined the status of AKT activation in xenograft tumors produced by DU145 cells, in which CNTN1 expression was modulated (see next section for details regarding xenograft tumor formation). In comparison with DU145 EV cell–derived xenograft tumors, an elevation of AKT activation was detected in DU145 CNTN1 monolayer cell–produced xenograft tumors (Fig. 2C, top), although the differences were not significant, which could have been the result of multiple factors (see discussion for details). However, a significant reduction of AKT activation in xenograft tumors produced by CNTN1-knockdown DU145 sphere cells was observed (Fig. 2C, bottom). In summary, these observations support the concept that CNTN1 plays a role in AKT activation in prostate cancer cells.

Another key event leading to increase invasive ability of epithelial cell–originating tumors is the loss or reduction of E-cadherin expression (3, 4). CNTN1 may reduce E-cadherin levels, as CNTN1 activates AKT, which has been shown to play a role in E-cadherin downregulation (28). In support of this hypothesis, overexpression of CNTN1 reduces E-cadherin levels in comparison with DU145 EV cells (Fig. 2D); conversely, knockdown of CNTN1 in DU145 sphere cells leads to E-cadherin upregulation (Fig. 2E). Furthermore, xenograft tumors derived from DU145 CNTN1 cells express lower levels of E-cadherin than xenograft tumors produced by EV cells (Fig. 2F), whereas DU145 shCNTN1 cells produce xenograft tumors with elevated levels of E-cadherin (Fig. 2F). Taken together, we demonstrate that CNTN1 reduces E-cadherin expression concurrently with the upregulation of AKT activation.

CSCs are defined by their ability to initiate tumors in immunocompromised mice. The specific expression of CNTN1 in DU145-derived PCSCs and its role in promoting prostate cancer cell invasion and AKT activation as observed above collectively suggest a role of CNTN1 in DU145 PCSCs-associated tumor initiation. To examine this hypothesis, we knocked down CNTN1 in DU145 sphere cells (Fig. 2B). CNTN1 downregulation did not have an apparent effect on the cell's ability to form spheres (data not shown). However, in comparison with DU145 Ctrl shRNA sphere cells, knockdown of CNTN1 significantly reduced the cell's ability for tumor formation (Fig. 3A, top). CNTN1 expression was significantly lower in xenograft tumors produced by CNTN1 knockdown sphere cells (Fig. 3A, bottom). In addition, overexpression of CNTN1 in DU145 monolayer cells enhanced xenograft tumor formation (Fig. 3B, top). CNTN1 levels remained higher in DU145 CNTN1 cell–produced xenograft tumors than DU145 EV cell-derived tumors (Fig. 3B, bottom). CNTN1 was clearly detected at the cell surface in DU145 CNTN1– and DU145 shCTRL–produced xenograft tumors (Fig. 3A and B, bottom). Interestingly, despite DU145 EV cells expressing no detectable levels of CNTN1 in vitro, clusters of CNTN1-positive cells with membrane localization were clearly detected (Fig. 3B, bottom). On the other hand, CNTN1-negative clusters of cells were also present in DU145 CNTN1 monolayer and DU145 Ctrl shRNA sphere cell–produced xenograft tumors (Fig. 3B; Fig. 3A, bottom; see discussion for details). Collectively, the above results demonstrate that despite the heterogeneity in CNTN1 expression in xenografts, CNTN1 has an important role in tumor formation.

The observation that CNTN1 promotes DU145 cell invasion (Fig. 1G) indicates a role of CNTN1 in prostate cancer metastasis. To examine this hypothesis, DU145 EV and CNTN1 cells were administrated into NOD/SCID mice via the tail vein. Although DU145 EV cells formed lung metastases, DU145 CNTN1 produced more and larger lung nodules (Fig. 4A, Supplementary Fig. S3A). On average, the number of lung nodules generated from the EV cells and CNTN1 cells was 3 and 11 per mouse, respectively. For one mouse injected with EV cells, lung metastasis could not be clearly detected (Supplementary Fig. S3A, see the image without arrows). We thus obtained an average of 3 nodules/mouse for 3 mice with a total of 10 lung nodules for the EV group and 11 nodules/mouse for 5 mice for a total of 54 lung metastases for the CNTN1 group (Fig. 4B). Because the nodule volumes are spread out in a large range, to better capture the dynamics of lung metastases produced by both cell types, metastatic tumors were divided into two groups (0–15 mm³ and 16 mm³+) based on the median tumor volume of the EV group, so that each group contained five tumors (Fig. 4B). Not only did CNTN1 cells generate more tumors (Fig. 4B), they produced larger lung nodules especially in the ≥16 mm³ tumor group (Fig. 4C). As expected, CNTN1-positive cell clusters were detected in both EV and CNTN1-produced lung metastases, but the latter contained more CNTN1-positive cells with higher levels of CNTN1 expression (Fig. 4A, bottom).

To further characterize the role of CNTN1 in promoting prostate cancer tumorigenesis, we examined its expression in primary prostate cancer tissues. An anti-CNTN1 antibody specifically recognized tumor-associated CNTN1 in IHC staining, evidenced by no detectable staining with control IgG (data not shown) and the positive signals could be competed out with a CNTN1 peptide (Supplementary Fig. S3B). In a limited number of patients from our cohort consisting of three samples with Gleason Score (GS) 6 to 7, six GS 8 to 10, 3 pairs of local and lymph node metastases, and 9 bone metastases; CNTN1 intensity was noticed as either negative or at very low levels in normal prostate glands and high (positive) in advanced (GS8-10) prostate carcinomas (Supplementary Fig. S3C). The cell surface presence of CNTN1 was clearly observed in both the primary tumor and the matched lymph node (Fig. 5A, Supplementary Fig. S3D) and bone (Fig. 5A, Supplementary Fig. S4) metastases. The bone metastases were confirmed as originating from prostate cancer, as evidenced by the positive staining for PAP (Fig. 5A, Supplementary Fig. S4). Subsequently, we examined CNTN1 expression using two TMAs. Typical staining for nontumor tissue, HGPIN, GS6, and GS9 carcinomas showed a trend for increased CNTN1 expression with prostate cancer progression (Fig. 5B). Quantification of CNTN1 staining was then measured in 637 patients using H-scores. The results revealed increased levels of CNTN1 expression with prostate cancer progression from GS5/6 to GS7, and to GS8/9 carcinomas (Fig. 5C). The number of CNTN1-positive cases was also increased following the progression from nontumor prostate tissues to HGPINs and to carcinomas (Table 1). Among carcinomas of different Gleason scores, the percentage of CNTN1-positive tumors increased from GS5 to GS9 carcinomas (Table 1). Furthermore, Kaplan–Meier survival analysis revealed that patients with CNTN1-positive prostate cancer, which is classified as moderate to strong staining with an H-Score of >10 were associated with decreased BCR-free survival (Fig. 5D; P < 0.05). Collectively, we suggest that CNTN1 expression associates with clinical prostate cancer progression and reduction of BCR-free survival.

Prostate cancer progression and metastasis is a complex process involving numerous factors. Our observations that CNTN1 enhanced tumor initiation and metastasis and associated with prostate cancer progression and BCR suggest that CNTN1 affects the expression of tumorigenesis-relevant genes. In line with this possibility, analysis of gene expression using the Affymetrix platform revealed the alterations of numerous gene expression by ≥1.5-fold following overexpression of CNTN1 in DU145 cells and knockdown of CNTN1 in DU145 spheres (Fig. 6A and B; Supplementary Table S2). The expected changes of CNTN1 in the overexpression and knockdown cells were confirmed (Fig. 6A and B; Supplementary Table S2). Knockdown of CNTN1 in spheres predominantly affected genes functioning in tumorigenesis, whereas overexpression of CNTN1 in DU145 cells altered genes involved in cancer, immunologic disorders, and inflammation (Fig. 6C, top); the latter two processes are well known to impact tumorigenesis (29, 30). The main physiologic processes affected by these candidate genes are organism and organ development (Fig. 6C, middle). Consistent with CNTN1 as a neural cell adhesion protein, these genes contribute to neurologic disorders (Fig. 6C, top right) and nervous system development (Fig. 6C, middle left). Knockdown of CNTN1 in DU145 spheres affected the expression of genes functioning in cell–cell communications and cellular function maintenance, which is consistent with CNTN1 being a cell surface protein and in agreement with PCSCs' role in maintaining a tumor mass (Fig. 6C, bottom left). The overexpression of CNTN1 in DU145 monolayer cells changed the expression of genes functioning in cell movement (Fig. 6C, bottom right) supports its role in promoting DU145 cell's invasion (Fig. 1G). In summary, knockdown of CNTN1 in DU145 spheres (PCSCs) affected genes largely functioning in tumorigenesis, whereas overexpression of CNTN1 in DU145 monolayer cells altered genes involved in multiple processes, including tumor cell invasion (Supplementary Fig. S5). Collectively, these observations support a general role of CNTN1 in promoting prostate cancer progression.

Accumulating evidence reveals a critical role of cancer stem cells (CSC) in cancer progression and metastasis. Tumors consist of heterogeneous cell populations (31, 32), in which CSCs are prone to survival resulting from selection pressure (33, 34). Prostate stem cells (PSC) have been identified in both humans and mice (35, 36). Although PSCs regenerate the prostate, PCSCs produce recurrent castration-resistant prostate cancer (CRPC; ref. 37). In addition, the signatures of PCSCs are associated with prostate cancer bone metastasis and poor prognosis (38, 39).

Despite this knowledge, the mechanisms governing PCSCs-mediated metastasis remain poorly understood. In a genome-wide analysis of 94 metastatic PCs derived from 30 patients for copy number changes, the gene copy alterations in metastases at different sites were clustered at the level of individual patients rather than metastatic site-specific, that is, no recurrent metastatic copy-number alterations were identified, supporting a major role of epigenetic alterations in driving prostate cancer metastasis (40). This theme is also seen in other cancer types (41). Accordingly, unique epigenetic alterations are likely contributing to the acquisition of neural cell adhesion proteins CNTN1 and L1CAM in a variety of cancer types (11, 12, 42).

In accordance with PCSCs being critical in prostate cancer progression and metastasis, it is thus possible that epigenetic alterations can cause CNTN1 induction in DU145 PCSCs following the culture of non-stem DU145 cells under SF-PCSC conditions. This possibility is supported by the reduction of CNTN1 following culturing PCSCs under 10% FBS (data not shown), but more importantly by the respective production of CNTN1-positive and -negative cells in CNTN1-negative DU145 EV cell and CNTN1-positive sphere cell–derived xenograft tumors (Fig. 3), and by the acquisition of CNTN1 in PC3 cells when cultured under SF PCSC conditions (Fig. 1C). It is possible that other mechanisms are also in action to regulate CNTN1 expression, such as transcription factors and miRNAs. In addition, for the PC3 cells, despite the SF PCSC conditions robustly elevating CNTN1 mRNA, the CNTN1 protein was rather low compared with DU145 sphere cells (data not shown), suggesting that CNTN1 expression is also regulated at the post-mRNA level. As PC3 and DU145 cells were derived from bone and brain metastasis, respectively, the induction of CNTN1 in DU145 cell-derived PCSCs may also suggest a specific role in brain metastasis, as a similar role has been reported for L1CAM in breast cancer brain metastasis (10). However, this does not exclude the possibility that CNTN1 exhibits a general role in promoting prostate cancer progression and metastasis, evidenced by CNTN1's contribution to DU145 PCSC-associated tumor initiation, and its association with prostate cancer progression and BCR (Fig. 3 and Fig. 5). This possibility is also supported by the emerging concept that the “driver” mutants promote cancer progression and metastasis, by the common induction of L1CAM in multiple cancer types (42), and by the L1CAM-mediated lung cancer metastasis to bone and kidney in mice (9). In addition, modulations of CNTN1 positively affected AKT activation, a major pathway stimulating cell invasion. This conclusion was based on the observed significant decrease in AKT activation in vitro (Fig. 2B) and in vivo (Fig. 2C) following CNTN1 knockdown in DU145 PCSCs as well as the trend of increasing AKT activation in vitro and in vivo in response to CNTN1 overexpression (Fig. 2A and C). Several factors can contribute to the elevation of AKT activation not being significant in vivo (xenograft tumors; Fig. 2C) following CNTN1 expression. (i) There may be a maximal level of AKT activation that can be achieved in xenograft tumors; (ii) the method used to quantify the differences might not be sufficiently sensitive to capture the small differences; and (iii) the level of CNTN1 was overexpressed to 3.5-fold in DU145 cells in comparison with EV cells based on microarray analysis, whereas the endogenous CNTN1 was knocked down by 9-fold in DU145 cell-derived PCSCs compared with shCTRL PCSCs.

A critical role of CNTN1 in contributing to PCSC-derived tumor initiation is intriguingly supported by the predominant effects of CNTN1 knockdown on genes that function in tumorigenesis (Fig. 6C, Supplementary Fig. S5). On the other hand, overexpression of CNTN1 in DU145 non-stem (monolayer) cells altered genes that function in an array of physiologic and cellular processes, which affect tumorigenesis (Fig. 6C, Supplementary Fig. S5). These observations are in line with our understanding of PCSCs in cancer progression and metastasis, but also suggest a role of non-stem cancer cells in supporting tumorigenesis via the integration of a series of complex signaling networks. Although our attention is focused on PCSCs in general, we cannot neglect the importance of the bulk non-stem cancerous cells. The physiologic relevance of these cells in tumorigenesis was clearly reflected by the emergence of CNTN1-null cells following the process of tumorigenesis in mice by CNTN1-positive DU145 PCSCs (Fig. 3 and Fig. 4). A similar role of non-stem cells in tumorigenesis in other cancer types should also be considered.

CNTN1's role in prostate cancer progression was supported by the observations that CNTN1 knockdown in DU145 PCSCs and its ectopic expression in DU145 monolayer cells affected lipid metabolism (Fig. 6C, the bottom). Prostate cancer tumorigenesis is intimately connected with lipid metabolism. Androgen regulates lipid metabolism and cholesterol is the precursor of androgen biosynthesis (43). It is thus tempting to speculate a role of CNTN1 in CRPC development.

Although it remains essentially unknown regarding what cellular mechanisms lead to CNTN1 expression in PCSCs, it is certain that these alterations also affect the expression of other important tumorigenic genes. It is likely that similar changes might also be in place to cause the expression of other neural CAMs in other cancer types. Research into this mechanism will not only advance our understanding of tumorigenesis, but also enables developing better means in cancer diagnosis and therapy.

We report the expression of CNTN1 in DU145 cell-derived PCSCs and in PC3 cells cultured under PCSC conditions. CNTN1 enhances cell invasion and AKT activation, and reduces E-cadherin expression. In addition, CNTN1 promotes xenograft tumor formation and lung metastasis concurrently with the upregulation of AKT activation and the downregulation of E-cadherin. In prostate cancer specimens, CNTN1 is detected in high-risk primary disease as well as in lymph node and bone metastases, and is associated with prostate cancer progression and a reduction in BCR-free survival. Thus, our research provides the first evidence of the neural CAM, CNTN1 in promoting prostate cancer tumorigenesis and metastasis.

J. Yan and D. Ojo have ownership interest (including patents) in Patent. J.H. Pinthus has received speakers bureau honoraria and is a consultant/advisory board of Ferring. D. Tang has ownership interest (including patents) in 30%. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Yan, J.H. Pinthus, G. Wood, D. Tang

Development of methodology: J. Yan, X. Lin, T. Aziz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Yan, D. Ojo, A. Kapoor, F. Wei, N. Wong, J. De Melo, H. Peng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Yan, D. Ojo, A. Kapoor, T. Aziz, F. Wei, J. De Melo, G. Wood, H. Peng, D. Tang

Writing, review, and/or revision of the manuscript: J. Yan, D. Ojo, A. Kapoor, J.H. Pinthus, T.A. Bismar, N. Wong, J.-C. Cutz, P. Major, G. Wood, D. Tang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Yan, A. Kapoor, T.A. Bismar, F. Wei

Study supervision: J. Yan, T. Aziz, J.-C. Cutz, D. Tang

Other (did selection of pathology material; bone biopsies with metastatic disease; and selection of cases of immunohistochemistry study): T. Aziz

The authors are grateful to Drs. Ying Xia and Alison Allan at The University of Western Ontario for examining CNTN1-positive circulating prostate cancer cells. We would like to dedicate this work to a great mother Ms. Guorui Zeng.

This work is supported in part by funds from Prostate Cancer Canada, CIHR (RMS79-71, MOP-843861), and McMaster University and St. Joseph's Hospital (GAP funding to D. Tang).

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.