VEGF/Neuropilin-2 Regulation of Bmi-1 and Consequent Repression of IGF-IR Define a Novel Mechanism of Aggressive Prostate Cancer

Understanding mechanisms that drive the initiation and progression of prostate cancer is essential to improving the clinical management of this disease, which is the most common cancer in men in the United States and Western Europe (1). Although early-stage prostate cancer can be treated reasonably well clinically, advanced forms of this disease are very aggressive and difficult to manage with existing therapies. As a consequence, these tumors are associated with a high degree of morbidity and mortality. Our approach to understanding the biology of aggressive prostate cancer is rooted in the hypothesis that tumor cells express receptors for VEGF and that VEGF signaling in tumor cells contributes to tumor formation and progression (2–5). Surprisingly, however, little is known about the expression and function of VEGF receptors on prostate carcinoma cells and their contribution to this disease. This problem is significant because these receptors and the signaling pathways that they regulate are prime targets for therapeutic intervention.

The neuropilins (NRP) are one class of VEGF receptors that are particularly interesting with respect to cancer biology. NRP1 and NRP2 were identified initially as neuronal receptors for semaphorins, which are axon guidance factors that function primarily in the developing nervous system (6, 7). The seminal finding by Soker and colleagues that neuropilins can also function as VEGF receptors and that they are expressed on endothelial and tumor cells launched studies aimed at understanding their functional contribution to angiogenesis and tumor biology (8). Neuropilins have the ability to interact with and modulate the function of tyrosine kinase VEGF receptors (VEGFR1 and VEGFR2) as well as other receptors (9, 10). There is also evidence that neuropilins can function independently of other receptors (11) and that they are valid targets for therapeutic inhibition of angiogenesis and cancer (12–14).

This study shows that the expression of NRP2 in prostate cancer cells is induced by PTEN deletion and that its expression correlates with Gleason grade. Given that the inactivation of PTEN is one of the most common genetic lesions in prostate cancer and its frequency increases with more aggressive disease (15), our discovery led to the hypothesis that VEGF/NRP2 signaling in tumor cells has a key role in prostate carcinogenesis. Indeed, the data obtained reveal that VEGF/NRP2 signaling contributes to the expression of Bmi-1, a Polycomb group transcriptional repressor that has been implicated in the etiology of prostate cancer induced by PTEN deletion (16). We also discovered that Bmi-1 represses transcription of the insulin-like growth factor-I receptor (IGF-IR), which is relevant because IGF-IR signaling contributes to tumor growth (17) and the IGF-IR is a potential therapeutic target for prostate cancer (18). We conclude from our data that aggressive prostate cancer cells exhibit VEGF/NRP2 signaling, Bmi-1 expression, and reduced IGF-IR expression. The functional and therapeutic implications of these findings were evaluated, and the results highlight a novel role for NRP2 as a biomarker for predicting response to IGF-IR therapy and as a therapeutic target itself in combination with IGF-IR inhibition.

Although aggressive prostate cancer is characterized by elevated VEGF expression in tumor cells (19, 20), therapeutic approaches based on VEGF inhibition have been largely unsuccessful. Bevacizumab, a VEGF antibody, did not have an impact on patients with metastatic, castration-resistant prostate cancer, and the trial was halted because it did not meet predefined response goals (21). Bevacizumab, however, does not inhibit the interaction of VEGF with the VEGF receptor NRP2 (22), which is noteworthy because we observed that NRP2 expression is induced in prostate cancer and that it is correlated with Gleason grade (Fig. 1A and B). Specifically, there is no significant NRP2 expression in normal, immortalized prostate cell lines, normal prostatic epithelium, or prostatic intraepithelial neoplasia [(PIN) Fig. 1A–C]. The expression of NRP2 in prostate cancer cell lines correlates with their invasive potential (Fig. 1A; Supplementary Fig. S1A). Robust NRP2 expression is seen in high-grade prostate cancer (Gleason grade 5) and metastases compared with low-grade prostate cancer (Gleason grade 3; Fig. 1B and C; Supplementary Fig. S1B–S1D). Importantly, NRP2 expression in prostate cancer is correlated significantly with grade (P < 0.001) and clinical outcome (P = 0.047; Supplementary Fig. S2A–S2D). Prostate carcinoma cells also express VEGFR2 but lack expression of VEGFR1 (Supplementary Fig. S2E; refs. 23, 24).

Given that PTEN loss has a causative role in the genesis of aggressive prostate cancer and correlates with Gleason grade (15, 25), we assessed the association of NRP2 expression with PTEN loss and observed a striking correlation. Relatively high NRP2 expression is seen in high Gleason grade tumors and metastases, which exhibit frequent PTEN loss (Fig. 1B; refs. 15, 25). Interestingly, however, there is heterogeneity in NRP2 expression in Gleason grade 3 prostate cancer specimens, in which approximately 12.5% of the cases expressed high levels of NRP2 (expression levels 4 and 5; Fig. 1B; Supplementary Fig. S2A–S2D). Gleason grade 3 tumors that harbored a PTEN deletion were positive for NRP2 expression, but those tumors that expressed PTEN were negative for NRP2 (Fig. 2A). These data are consistent with the finding that PTEN loss occurs in approximately 15% to 20% of low-grade prostate cancer (15, 25). The association between PTEN loss and NRP2 expression was substantiated by an analysis of prostate tumors in Pten^loxp/loxp; PB-Cre⁺ (referred to as Pten^pc−/−) mice, which are invasive and aggressive (26). These tumors display relatively high NRP2 expression compared with the mouse prostatic epithelium (Fig. 2B). NRP2 expression is also high in PIN lesions in the Pten^pc−/− mice, consistent with the hypothesis that PTEN loss induces NRP2 (Fig. 2B). This consequence of PTEN loss is specific to NRP2 because NRP1 is expressed in both normal prostate epithelium and PTEN^pc−/− tumors (Fig. 2B; Supplementary Fig. S3A). These results imply that PTEN represses NRP2 but not NRP1 expression. To test this hypothesis, PTEN was expressed in PC3 and C4-2 cells, which are PTEN-null and express NRP2. PTEN expression ablated NRP2 expression and the activity of a NRP2 promoter reporter construct but it had no effect on NRP1 expression (Fig. 2C–E). These results indicate that NRP2 is associated with aggressive, PTEN-null prostate cancer. Importantly, NRP2 expression is not a function of tumor differentiation because of the heterogeneity of its expression in grade 3 tumors, which are all differentiated, and our observation that NRP2 is not expressed in VCaP cells, which are poorly differentiated but PTEN-positive (Fig. 2D). DU145 cells, which are PTEN-positive, also lack NRP2 (Fig. 2D). Additional evidence to support the conclusion that PTEN loss or inactivation induces NRP2 was obtained by evaluating AKT activation because it is a read-out of PTEN deletion and mutation (27). Indeed, a significant correlation between phosphorylated AKT (p-AKT) and NRP2 expression was observed in prostate tumors (Supplementary Fig. S3B and S3C).

To investigate the mechanism behind PTEN-mediated NRP2 suppression, we analyzed the role of c-Jun because it is activated upon PTEN deletion in prostate tumors and contributes to tumorigenesis (28, 29). A significant downregulation of c-Jun occurred upon PTEN expression in PC3 and C4-2 cells (Fig. 2E and F), and PTEN expression inhibited the activity of a c-Jun promoter and AP1 reporter constructs significantly (Fig. 2G). However, PTEN expression did not affect Coup-TFII, a transcription factor previously shown to regulate NRP2 in lymphatic vessel development (Fig. 2F; refs. 30). We established the role of c-Jun in the regulation of NRP2 using c-Jun short hairpin RNA (shRNA) and TAM67 (a dominant-negative c-Jun). Expression of either c-Jun shRNAs or TAM67 reduced NRP2 protein and mRNA levels, as well as NRP2 promoter activity (Fig. 3A and B). Also, an increase in NRP2 promoter activity upon c-Jun overexpression was observed in PC3 and C4-2 cells (Fig. 3C). These results indicate that PTEN represses NRP2 expression by inhibiting c-Jun. We assessed this hypothesis by showing that c-Jun expression rescued NRP2 expression in PC3 and C4-2 cells that had been transfected with a PTEN construct (Fig. 3D). Moreover, our analysis of recently published gene expression data derived from PTEN induction in Pten^−/− cells (29) revealed an inverse correlation between c-Jun and PTEN and between NRP2 and PTEN expression (Fig. 3E).

To investigate the regulation of NRP2 by c-Jun more rigorously, we carried out chromatin immunoprecipitation (ChIP) and observed that c-Jun binds directly to distinct sites in the NRP2 promoter (Fig. 3F and G). More specifically, the NRP2 promoter contains 2 AP1-binding sites located within NP3 and NP7 (Fig. 3F) and ChIP confirmed binding to these sites (Fig. 3G). In addition, mutating these sites inhibited NRP2 promoter activity (Supplementary Fig. S4A). As further proof that PTEN represses NRP2 expression by a c-Jun–dependent mechanism, we observed that PTEN significantly inhibited the binding of c-Jun to the NRP2 promoter (Fig. 3H). Given that c-jun-NH₂-kinases (JNK) are the upstream kinases that regulate c-Jun activity (31), we found that either a dominant-negative JNK or a JNK1 shRNA, but not a JNK2 shRNA, inhibited NRP2 expression (Supplementary Fig. S4B–S4D).

To evaluate the role of NRP2 in prostate tumorigenesis, we generated PC3 cells with diminished NRP2 expression using shRNAs. Loss of NRP2 had no effect on morphology or proliferation on tissue culture plates (data not shown), but it did reduce the ability of these cells to grow in soft agar (Fig. 4A, left). The NRP2-depleted cells also grew more slowly than control cells when implanted subcutaneously in mice (Fig. 4A, right). Although significant, the impact of NRP2 loss on growth in soft agar and tumor formation in vivo was modest, suggesting a potential compensatory mechanism. In this context, we observed a marked increase in IGF-IR mRNA expression concomitant with NRP2 loss but no change in the expression of either EGF receptor (EGFR) or insulin receptor (IR) mRNA expression (Fig. 4B). This effect of NRP2 loss on the induction of IGF-IR expression was also evident at the protein level in both PC3 and C4-2 cells (Fig. 4C). This effect is specific to NRP2 because no change in IGF-IR expression was observed upon NRP1 loss (Fig. 4D). The repression of IGF-IR expression by NRP2 in PC3 cells was maintained in vivo as evidenced by the observation that loss of NRP2 induced a significant increase in IGF-IR in 3 independent xenograft tumors (Fig. 4E). Also, immunostaining of human prostate cancer specimens revealed a positive correlation between p-AKT and NRP2 expression but a negative correlation between NRP2 and IGF-IR (Supplementary Fig. S3B and S3C). The possibility existed that FOXO transcription factors contribute to the regulation of IGF-IR expression based on the recent report that phosphoinositide 3-kinase activation represses receptor tyrosine kinases including the IGF-IR by feedback inhibition through inactivation of FOXO transcription factors (32). However, no change in FOXO activation was observed upon NRP2 depletion, and downregulation of FOXO factors did not affect the induction of IGF-IR expression caused by NRP2 (Supplementary Fig. S5A–S5C). Thus, the mechanism we propose appears to be distinct from this interesting study because inactivation of FOXO factors affected a number of receptors including the insulin receptor, EGFR, and c-MET. In contrast, NRP2 represses only the IGF-IR (Fig. 4B).

To investigate the role of NRP2 on IGF-IR signaling and function further, we expressed NRP2 in p69 noncancerous prostate epithelial cells, which express high levels of IGF-IR and less c-Jun than more aggressive cell lines (33). p69 cells are NRP2-null, and expression of NRP2 reduced IGF-IR levels (Fig. 4F). These cells also express endogenous VEGF (Supplementary Fig. S5D). Consistent with reduced IGF-IR expression, NRP2 expression inhibited IGF-1–stimulated cell proliferation (Fig. 4G, left) and insulin receptor substrate 1 (IRS-1) phosphorylation (a substrate of IGF-IR tyrosine kinase; Fig. 4G, right). On the basis of these results, it is expected that PTEN-positive tumors, which express low levels of NRP2, have high IGF-IR levels. This hypothesis was confirmed by analyzing gene expression derived from human prostate cancer specimens (n = 128; ref. 34). We found a positive correlation between IGF-IR and PTEN expression (P = 2 × 10⁻⁵; Fig. 4H).

The compensatory relationship between NRP2 and IGF-IR expression suggested that inhibition of both receptors would be more effective at impeding tumor growth than inhibition of either receptor individually. To test this hypothesis, we investigated the effect of an IGF-IR inhibitory antibody on the ability of control PC3 cells or cells in which NRP2 expression had been ablated to grow in soft agar or in immunocompromised mice. A significant effect of combined NRP2 downregulation and IGF-IR inhibition was evident in both soft agar growth in vitro (Fig. 4I) and tumor growth in vivo (Fig. 4J; P < 0.01). Indeed, combined treatment resulted in complete inhibition of tumor growth (Fig. 4J). The inhibition of NRP2 and IGF-IR did not change the expression of the insulin receptor (Supplementary Fig. S6A), discounting the possibility of compensation by this receptor. Moreover, combined treatment significantly increased the cleavage of caspase-3, suggesting activation of cell death (Supplementary Fig. S6B). Addition of A12 (the clone name of the IGF-IR inhibitory antibody) blocked IGF-1–stimulated tyrosine phosphorylation of IRS-1, validating the functionality of this antibody (Supplementary Fig. S6C). Further evidence in support of our findings was provided by the effect of A12 on PTEN-null (PC3) and PTEN-postitive (VCaP) prostate cancer cell lines. In agreement with our hypothesis, A12 inhibited cell proliferation of VCaP but not PC3 cells (Supplementary Fig. S6D).

To establish the mechanism by which NRP2 represses the IGF-IR, we focused on our novel finding that NRP2 is required to maintain expression of Bmi-1, a Polycomb group transcriptional repressor that has been implicated in PTEN deletion–induced prostate cancer (16). Specifically, NRP2 loss in both PC3 and C4-2 cells diminished Bmi-1 protein expression markedly (Fig. 5A), as well as Bmi-1 mRNA expression in PC3 cells (Fig. 5B, left). This effect of NRP2 loss on Bmi-1 expression was also observed in PC3 xenografts (Fig. 5B, right). These data linking NRP2 to Bmi-1 further imply that NRP2 has a role in tumor formation. This effect was manifested in soft agar assays because overexpression of Bmi-1 rescued the impairment in soft agar growth caused by NRP2 downregulation (Fig. 5C). Moreover, analysis of the gene expression data from Taylor and colleagues (34) revealed a positive correlation between expression of Bmi-1 and c-Jun (Fig. 5D; P = 1 × 10⁻⁶).

Given that Bmi-1 is a transcriptional repressor and our finding that NRP2 represses expression of the IGF-IR, we examined the ability of Bmi-1 to repress IGF-IR transcription. Initially, we observed that depletion of Bmi-1 expression in PC3 cells using shRNAs increased IGF-IR mRNA levels (Fig. 5E). Loss of Bmi-1 also increased expression of the IGF-IR protein significantly, confirming the importance of Bmi-1 in repressing IGF-IR expression (Fig. 5F). To investigate the regulation of IGF-IR by Bmi-1, we carried out ChIP using a Bmi-1 antibody and discovered that Bmi-1 binds directly to the IGF-IR promoter (Fig. 5G). More specifically, Bmi-1 appears to bind to a region between −74 and −756 in the promoter (Fig. 5G). As further proof that NRP2 represses IGF-IR expression by a Bmi-1–dependent mechanism, we infected PC3 cells that had been depleted of NRP2 with a lentivirus-expressing Bmi-1 and assessed IGF-IR expression. We observed that Bmi-1 represses IGF-IR expression in the absence of NRP2 (Fig. 5H).

Recently, we reported that VEGF/NRP2 signaling activates focal adhesion kinase (FAK) in concert with the α6β1 integrin (35, 36), and we observed that NRP2 and α6β1 interact in prostate carcinoma cells (Supplementary Fig. S6E). To investigate whether NRP2-mediated Bmi-1 expression involves FAK signaling, we evaluated FAK activation in tumor samples in vivo. Tumor xenografts from NRP2-depleted cells showed reduced FAK activation (Fig. 6A), and treatment of PC3 cells with a FAK inhibitor reduced expression of Bmi-1 (Fig. 6B). In addition, we also assessed FAK activation in tumor samples by immunostaining. We found that NRP2^high (PTEN^low) tumors express high levels of activated FAK at the membrane compared with NRP2low (PTEN-positive) tumors (Fig. 6C). These data were confirmed by overexpression of constitutively active FAK (CA-FAK; K38A) in NRP2-depleted cells, which resulted in increased expression of Bmi-1 (Fig. 6D). To analyze the effect of c-Jun expression on FAK activation, we used LNCaP cells expressing low levels of NRP2 and observed that c-Jun expression increases NRP2, as well as FAK activation (Fig. 6E). Similarly, downregulation of c-Jun reduced FAK activation and Bmi-1 expression, and led to increased IGF-IR levels (Fig. 6F).

A critical issue is whether the observed effects of NRP2 on Bmi-1 and IGF-IR expression involve VEGF, an NRP2 ligand. To address this issue, we downregulated VEGF expression in PC3 cells using either siRNA or shRNA and observed that decreased VEGF expression is associated with decreased Bmi-1 and increased IGF-IR expression (Fig. 7A and B; Supplementary Fig. S6F). Moreover, exogenous VEGF increased Bmi-1 expression in a dose-dependent manner (Supplementary Fig. S6G). Our data also imply that VEGF expression should be higher in prostate tumors with PTEN loss than in those with PTEN-positive tumors and that it should positively correlate with NRP2 and inversely correlate with IGF-IR expression. To assess this hypothesis, quantitative PCR was carried out using RNA isolated from 12 prostate tumors (6 PTEN^low and 6 PTEN^high). The PTEN^low tumors exhibited relatively high expression of NRP2 and VEGF but low expression of the IGF-IR compared to the PTEN^high tumors (Fig. 7C). The expression pattern of VEGF is consistent with our finding that VEGF expression in tumor cells correlates with Gleason grade (20). To exclude the possibility that VEGF regulation of Bmi-1 is mediated by other VEGF receptors, especially VEGF tyrosine kinase receptors, we used bevacizumab, which inhibits the interaction of VEGF with these receptors but not with NRP2 (22). An NRP2-blocking antibody inhibited VEGF induction of Bmi-1 expression, but bevacizumab had no significant effect (Fig. 7D).

Collectively, the data generated in this study indicate that NRP2 expression in tumor cells is a potential biomarker to predict the efficacy of IGF-IR inhibitors, which are currently in clinical trials. To test this latter possibility, we used 2 LuCaP xenograft models: LuCaP-35 and LuCaP-86.2. LuCaP-35 is sensitive to the IGF-IR inhibitor A12 and tumor growth is substantially inhibited upon A12 treatment. In contrast, LuCaP-86.2 is insensitive to A12 (Fig. 7E). We detected significantly reduced expression of IGF-IR and increased expression of NRP2 and Bmi-1 in LuCaP-86.2 (PTEN-null) compared with LuCaP-35 (PTEN-positive) tumors. Conversely, LuCaP-35 tumors expressed significantly higher levels of IGF-IR than NRP2 and Bmi-1 (Fig. 7F). These data were further strengthened by the analysis of an additional 15 A12 responders and 8 A12 nonresponders (Fig. 7G; Supplementary Fig. S7). As shown in Fig. 7G, we detected 150-fold higher NRP2 expression in the nonresponders than in responders and only a 10-fold increase in IGF-IR expression in the responders compared with the nonresponders. To further substantiate our finding that NRP2 loss can enhance sensitivity to IGF-IR inhibition, we ablated NRP2 expression in LuCaP-86.2 cells and observed a significant increase in the response to A12 as assessed by cell proliferation (Fig. 7H). Importantly, this response is associated with a significant decrease in Bmi-1 expression and increase in IGF-IR expression (Fig. 7H).

Our data highlight a novel role for VEGF/NRP2 signaling in prostate cancer cells that impacts our understanding of the biologic nature of this disease, especially aggressive prostate cancer, and our approach to therapy. Specifically, we conclude that the VEGF receptor NRP2 is induced by PTEN loss and that VEGF/NRP2 signaling regulates the expression of Bmi-1, a key effector of prostate tumorigenesis induced by PTEN deletion (16). The consequences of this mechanism are significant for several reasons, including our finding that Bmi-1 represses transcription of the IGF-IR and abrogates IGF-IR signaling. We exploited these results to show that NRP2 is a novel biomarker for high-grade prostate cancer that can be used to predict response to IGF-IR therapy. Importantly, our work indicates that combined NRP2 and IGF-IR inhibition is an effective approach for impeding tumor growth that overcomes the resistance caused by inhibiting either receptor independently.

A major conclusion from our work is that NRP2 expression and function are linked directly to oncogenic transformation associated with loss of PTEN function. More specifically, we show that PTEN loss induces NRP2 expression by activating JNK1/c-Jun and that an inverse correlation between PTEN and NRP2 expression is evident in both the Pten^pc−/− mice and in human prostate cancer. These expression data were substantiated by our finding that VEGF/NRP2 signaling regulates the expression of Bmi-1, a Polycomb group transcriptional repressor implicated in prostate tumorigenesis (16). Although a previous study had foreshadowed a role for NRP2 in the xenograft growth of colorectal carcinoma cells (37), our study is the first to show that an oncogenic pathway, that is, c-Jun activation induced by PTEN loss, regulates NRP2. Moreover, our finding that NRP2 expression increases with Gleason grade is consistent with our previous finding that VEGF expression in cancer cells also increases with Gleason grade (20), supporting the hypothesis that autocrine VEGF/NRP2 signaling is characteristic of high-grade, aggressive prostate cancer. We do not exclude the possibility that this signaling pathway involves other VEGF receptors in addition to NRP2. Specifically, prostate carcinoma cells express VEGFR2 but lack expression of VEGFR1 (Supplementary Fig. S2E). The absence of VEGFR1 is consistent with previous reports and may result from hypermethylation of the FLT1 gene encoding VEGFR1 in prostate cancer (23, 24). Nonetheless, the possibility that VEGFR2 contributes to NRP2 signaling is mitigated but not discounted by our finding that bevacizumab, which inhibits the interaction of VEGF with tyrosine kinase VEGF receptors but not with NRP2 (22), did not inhibit the ability of VEGF to regulate Bmi-1.

Unexpectedly, we identified NRP2 as a novel c-Jun target that is regulated by PTEN. The implication of these findings is that VEGF/NRP2 signaling is a component of the mechanism by which PTEN loss induces prostate cancer. Our key findings that VEGF/NRP2 signaling regulates Bmi-1 expression and that there is a strong correlation between NRP2 and Bmi-1 in human prostate cancer buttress this conclusion. The significance of these data is that they provide a causal link between VEGF/NRP2 signaling and a mechanism of prostate tumorigenesis and they build upon the seminal report that Bmi-1 regulates the self-renewal of prostate stem cells and malignant transformation (16). Our data provide a specific mechanism (autocrine VEGF/NRP2 signaling) that contributes to Bmi-1 regulation in prostate cancer. Of note, this mechanism is consistent with the hypothesis that tumor stem cells secrete growth factors in an autocrine manner to protect themselves from differentiation and to support their self-renewal (38).

An important issue that arises from our data is how VEGF signals through NRP2, given that NRP2 lacks intrinsic signaling function. We hypothesize that the specific interaction between NRP2 and the α6β1 integrin (CD49f) potentiates this signaling. This hypothesis is based on our previous finding that NRP2 facilitates the α6β1-mediated activation of FAK and Src (36) and our observation here that NRP2 and α6β1 interact in prostate carcinoma cells. Interestingly, the association of NRP2 with α6β1 does not require the cytoplasmic domain of NRP2 (Supplementary Fig. S6E), indicating that this association occurs in the membrane, perhaps in microdomains enriched for these receptors. We stress that our findings do not discount a role for tyrosine kinase VEGF receptors in promoting NRP2 signaling. Numerous studies have shown that the neuropilins can enhance the function of these receptors (6, 7). It is important to note, however, that even if such VEGF receptors contribute to the signaling pathway we describe, NRP2 is essential and a prime target for therapeutic intervention. Moreover, the data presented in Fig. 4J indicate that targeting NRP2 and IGF-IR is sufficient to block tumor growth in the absence of any other VEGF receptor inhibition.

We implicate FAK as a downstream effector of VEGF/NRP2 signaling that regulates Bmi-1 expression. This ability of FAK to regulate Bmi-1 is novel and it adds to our understanding of how FAK contributes to tumorigenesis. FAK expression is increased in prostate cancer cells compared with prostatic epithelium, and there is evidence that FAK participates in prostate tumorigenesis (39). FAK is also necessary for mammary tumorigenesis and for the function of mammary tumor stem cells (40), but the mechanism by which FAK functions in this capacity has not been elucidated. Clearly, the ability of FAK to regulate Bmi-1 provides one such mechanism. As mentioned above, it is likely that the α6β1 integrin is a component of this mechanism (35, 36).

Our data support the conclusion that IGF-IR expression and signaling are diminished in high-grade, aggressive prostate cancer, most likely as a result of PTEN loss. Although this conclusion is unexpected because IGF-IR signaling is thought to promote tumor proliferation in prostate and other cancers (41), this receptor can also promote differentiation (42, 43) and this function might need to be impeded by VEGF/NRP2 signaling for the genesis of high-grade, aggressive cancer. This possibility is supported by the report that abrogation of functional IGF-IR expression actually promoted earlier emergence of more aggressive, less-differentiated carcinomas in a p53-compromised model of prostate cancer (44).

The data presented have significant implications for prostate cancer therapy, especially because the clinical use of IGF-IR inhibitors is feasible (17) and NRP2 inhibitors are available (13). On the basis of our data, most of the ongoing IGF-IR clinical trials will not succeed because these trials recruited many patients with high-grade, aggressive prostate cancer, which expresses NRP2 but lacks IGF-IR and PTEN expression. Inhibition of the IGF-IR may reduce the burden of low-grade tumors, but such tumors likely contain a small population of NRP2-positive cells that are resistant to IGF-IR inhibition and will cause relapse unless this VEGF receptor is also targeted. NRP2 inhibition could be a very effective therapy for aggressive, high-grade tumors provided that the IGF-IR is also inhibited to reduce tumor burden caused by compensatory signaling. Furthermore, this study identifies NRP2 as a novel biomarker for predicting response to IGF-IR therapy. A critical finding in this context is that NRP2 is a much more robust biomarker than the IGF-IR itself. This finding is significant because the IGF-IR is also expressed in normal glands, in contrast with NRP2. Importantly, the data presented provide a rationale for initiating clinical trials that combine inhibitors of both NRP2 and IGF-IR.

Tissue samples of defined Gleason grade, lymph node, and bone metastases were obtained from the University of Massachusetts Medical School (UMASS) Cancer Center Tissue Bank with approval of the Institutional Review Board (IRB) of UMASS Medical School (Worcester, MA). The IRB granted a waiver for obtaining patients' consent in accordance with NIH guidelines because these are pre-existing, de-identified specimens. Formalin-fixed, paraffin-embedded sections either from primary tumors, lymph node metastases, or bone metastases were stained using either NRP2 or control IgG antibody as described before (20). Frozen sections were microdissected by laser capture microscopy (Arcturus PixCell 2) as described elsewhere (20) to obtain pure populations of tumor cells of defined Gleason grades. RNA was isolated from these microdissected samples using the RNeasy Kit (Qiagen) and prepared cDNA using Superscript II reverse transcriptase (Invitrogen). Quantitative PCR was carried out using the TaqMan Assay Kit (Applied Biosystems): PTEN (Hs02621230_s1), NRP2 (Hs00187290_m1), VEGF (Hs00900055_m1), IGF-IR (Hs00609566_m1), and GAPDH (Hs9999905_m1). Tissue microarrays containing 52 primary prostate cancer tissues were also used.

Bmi-1 (Cell Signaling), actin (Sigma), IGF-IR (Cell signaling), NRP2 (C-9, H-300, Santa Cruz and R&D), p-FAK (Y397, Invitrogen) a-VEGF (Calbiochem), insulin receptor (Santa Cruz), cleaved caspase-3, PTEN (Cell Signaling), p-Tyr (PY99), NRP1 (Santa Cruz), p-AKT (Ser473), pS6K, pFOXO1 (Ser256), FOXO1 (L27), AKT, S6K, p-c-Jun (S73, Cell Signaling), c-Jun (Santa Cruz), IRS-1 (Bethyl), and CD49f (AA6A; kindly provided by Dr. A. Cress, University of Arizona Cancer Center, Tucson, AZ). Bevacizumab was obtained from the pharmacy of UMASS Memorial Medical Center. A12, cixutumumab, was kindly supplied by Dr. Dale Ludwig from ImClone Systems Incorporated. Immunohistochemistry for p-AKT was conducted using D9E antibody (Cell Signaling). IGF-IR antibody α-IR3 was purchased from Abcam and PF573228 (FAK inhibitor) from Tocris.

The PTEN plasmid was provided by Dr. Alonzo Ross (UMASS Medical School); Bmi-1-ires-eGFP cloned into FUGW was obtained from Addgene; c-Jun expression constructs (HA-tagged) were provided by Dr. Dirk Bohmann (University of Rochester Medical Center, Rochester, MN); CA-FAK (K38A) construct was provided by Jun-Lin Guan (University of Michigan, Ann Arbor, MI). Lentiviruses vectors expressing NRP2 shRNAs (Open Biosystems; clone ID TRCN0000063308, TRCN0000063309 or TRCN0000063312), VEGF shRNA (TRCN0000003343), Bmi-1 shRNA (TRCN0000020154, TRCN0000020156, TRCN0000012565), or c-Jun shRNAs (provided by S. Huang, Georgia Health Sciences University, Augusta, GA) or a GFP control (Open Biosystems, RHS4459) were used.

Prostate cancer cells, LNCaP, PC3, VCaP, DU145 (American Type Culture Collection), C4-2 (UroCor, Inc.), or normal prostate immortalized cell lines RWPE-1 (provided by Dr. SK Batra, University of Nebraska Medical Center, Omaha, NE), PNT1A (provided by Dr. M. Ittmann, Baylor College of Medicine, Houston, TX), and PZ-HPV7 (provided by Dr. T.C. Chen, Boston University, Boston, MA) were used. PC3-M cells (45) were kindly provided by Dr. Paul Mak (UMASS Medical School). These cell lines were not authenticated. Lentiviruses containing NRP2 shRNAs, VEGF shRNA, Bmi-1 shRNA, or c-Jun shRNAs or a GFP control were generated, titrated according to the manufacturer's instructions, and used to infect cells following standard protocols. Stable cell transfectants were generated by puromycin selection (2 mg/mL). In some experiments, cells were transfected with VEGF siRNA (Smartpool, Dharmacon) or scrambled control siRNA.

Growth in soft agar was assayed by scoring the number of colonies formed in medium [Dulbecco's Modified Eagle's Media (DMEM) supplemented with 10% FBS] containing 0.3% agarose, with a 0.5% agarose medium underlay as described (46). PC3-parental, PC3-GFP-sh, or PC3-NRP2-sh (NRP2-sh-1 or NRP2-sh-2) cells were seeded on 60-mm diameter plates in triplicate. Cells were fed with 1.5 mL of suspension medium (DMEM supplemented with 10% FBS) every 3 days. The number of colonies larger than 100 μm was counted after 14 days.

The Pten^pc−/− mice (26) was backcrossed more than 10 times to obtain a BALB/c background. Specimens of PIN (6 weeks) or invasive prostate tumors (20 weeks) from these mice and mice with wild-type Pten were stained for NRP2 or NRP1 expression using immunohistochemistry. All animal experiments were in accordance with institutional guidelines and were approved by the Institutional Animal Care and Use Committee.

Cells were transfected with the luciferase construct and Renilla luciferase construct to normalize for transfection efficiency. Relative light units were calculated as the ratio of firefly luciferase to Renilla luciferase activity (normalized luciferase activity). The protocol used for transfection and measurement of luciferase activity has been described previously (20). We used an NRP2 promoter luciferase construct (−3,000/+195) or the same construct with AP1 mutations. A full-length wild-type c-Jun promoter (−1,780/+731) luciferase construct was provided by Dr. Wayne Vedeckis (LSU Health Sciences Center, New Orleans, LA) and used to study the effect of PTEN expression.

ChIP assays were conducted according to our published protocol (47). All ChIP experiments were carried out at least 3 times and the variation was less than 20%. The primers used to amplify the IGF-IR and NRP2 promoters, as well as other genes analyzed in this study, are provided in Supplementary Fig. S8.

Tumor fragments (20–30 mm³) from the human prostate tumor xenografts (LuCaP-35 or LuCap-86.2; established by co-author R.L. Vessella) were implanted s.c. into 6-week-old–castrated severe combined immunodeficient mice as previously described (48, 49). Once the implanted tumor reached a volume less than 100 mm³, animals were randomized into 2 groups (10 mice per group). Group 1 animals received human IgG treatment and were designated as controls. Group 2 animals received A12 treatment at a dose of 40 mg/kg. All treatments were given intraperitoneally (i.p.) 3 times a week for 4 weeks. Tumors were measured twice weekly and tumor volume was estimated by the formula: volume = L × W²/2. Single-cell suspensions of LuCaP-86.2 cells were prepared by resecting the s.c. xenografts and cutting them into small pieces in DMEM/F12 containing 10% FBS. The small pieces were mechanically dissociated, and cell viability was determined by trypan blue counting as described (50). Cells were cultured in DMEM/F12 containing 20% FBS and 5 ng/mL EGF. PC3 transfectants (GFP-sh or NRP2-sh) implanted in immunocompromised mice were injected intraperitoneally with either A12 or IgG, once the tumor volume reached approximately 150 mm³. Xenograft growth was measured on alternate days. All animal experiments were in accordance with institutional guidelines and were approved by the Institutional Animal Care and Use Committee.

Cell proliferation was measured using MTT assay. Cell invasion assays were conducted using Matrigel-coated Transwell plates.

No potential conflicts of interest were disclosed.

Conception and design: H.L. Goel, S.R. Plymate, A.M. Mercurio

Development of methodology: H.L. Goel, C. Chang, B. Pursell, S.R. Plymate

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.L. Goel, C. Chang, B. Pursell, S. Lyle, H. Adisetiyo, P. Roy-Burman, P.S. Nelson, R.L. Vessella, R.J. Davis, S.R. Plymate

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.L. Goel, H.S. Xi, C.-C. Hsieh, I.M. Coleman, A.M. Mercurio

Writing, review, and/or revision of the manuscript: H.L. Goel, S. Lyle, C.-C. Hsieh, P. Roy-Burman, R.L. Vessella, R.J. Davis, S.R. Plymate, A.M. Mercurio

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.L. Goel, I.M. Coleman, S.R. Plymate

Study supervision: H.L. Goel, I. Leav, S. Lyle, A.M. Mercurio

The authors thank Drs. S.K. Batra, Azad Bonni, T.C. Chen, M. Ittmann, A. Ross, L. Shaw, V. Sharma, C. Cheng, and P. Mak for reagents, cell lines, technical assistance, and manuscript discussion; and Michael Roche and Dr. Cherie Taglienti from UMASS Tissue Bank for their help with human tissue specimens.

This work was supported by DOD Prostate Cancer grant PC111410 (A.M. Mercurio); NIH grants R01CA80789 and CA89209 (A.M. Mercurio), R01CA59705 (P. Roy-Burman), PO1 CA85859 and the Pacific Northwest Prostate Cancer SPORE P50 CA097186 (S.R. Plymate, R.L. Vessella, and P.S. Nelson); the Veterans Affairs Research Program (S.R. Plymate); and an award from the Prostate Cancer Foundation (P.S. Nelson).