RET Signaling in Prostate Cancer

Prostate cancer is the second-leading cancer cause of cancer in American men, with 27,540 deaths expected to occur due to prostate cancer in 2015 (1). Although the prognosis for early-stage prostate cancer is generally excellent, few effective therapeutic options exist for advanced prostate cancer. It has been appreciated for many years that the tumor microenvironment plays an important role in the initiation and progression of prostate and other cancers. One important component of this microenvironment is nerves. It is well known that prostate cancer has a propensity to grow in perineural locations, as do a number of other cancers such as pancreatic cancer. Perineural invasion (PNI) is defined as the presence of cancer infiltration in, around, and/or through the nerves (2) and is the result of reciprocal interactions between cancer cells and adjacent nerves (3). PNI is an adverse prognostic factor for many cancers, including prostate, pancreatic, head and neck, colon, skin, and salivary cancers (4–8). Although PNI per se is not predictive of aggressive disease in prostate cancer, large diameter perineural tumor is one of the most significant pathologic predictors of poor outcome (9) following radical prostatectomy. Furthermore, PNI is associated with poor outcomes following radiation therapy (10, 11), suggesting a prosurvival effect of prostate cancer cell interactions with nerves. These clinical observations show that the interactions between nerves and prostate cancer cells can have a significant impact on treatment outcomes in men with prostate cancer, which ultimately must be related to the underlying biology.

Recent functional studies in vitro and correlative studies in vivo have shown significant interactions between nerves and adjacent cancer cells that promote cell survival, proliferation, and migration of prostate cancer cells (2, 3, 12). For example, prostate cancer cells adjacent to nerves display increased proliferation and decreased apoptosis compared to cells away from nerves (12), indicating local microenvironmental influence on the cancer cells in this niche. Similar findings have been reported in other neurotrophic cancer such as pancreatic cancer (13). Studies in rats have shown that denervation of the prostate leads to almost complete loss of epithelium (14), indicating a strong trophic effect of nerves on normal prostate epithelium. Similarly, men with complete spinal cord injury had significantly smaller prostates than controls (15). Studies by Magnon and colleagues (16) have shown that chemical or surgical ablation of nerves inhibits tumorigenesis and metastasis in both xenograft and transgenic mouse models of prostate cancer, unequivocally establishing that nerve—prostate cancer cell interactions play a significant role in prostate cancer initiation and progression but the molecular basis of these interactions is still unclear.

We have carried out expression microarray analysis of laser captured prostate cancer reactive stroma (17) and shown that among the upregulated genes is glial cell line-derived neurotrophic factor (GDNF). Interestingly, GDNF levels are increased during androgen induced regrowth of the prostate after castration (18). GDNF is present in the peripheral nerves of normal prostate and in reactive stroma in prostate cancers, where it can be secreted and potentially interact with prostate cancer cells. Of course, GDNF is expressed in nerves in potential metastatic sites as well. Functional studies in pancreatic cancer implicate GDNF as a key factor promoting perineural migration in vitro in this disease (19, 20). It has also been shown in breast cancer that inflammatory cytokines can induce expression of GDNF by fibroblastic cells and tumor cells and GDNF increases proliferation and motility (21), indicating that GDNF is also expressed away from nerves in some contexts.

GDNF binds to RET, a transmembrane receptor tyrosine kinase, in conjunction with its co-receptor GFRα1 and activates cellular signaling (20). Robinson and colleagues have shown RET is expressed in the three prostate cancer cell lines tested (PC3, DU145, and LNCaP) and the CWR series of xenografts (22). Given that RET is present on at least some prostate cancer cells and GDNF is expressed in nerves and reactive stroma cells, we sought to determine the potential role of RET signaling in prostate cancer. We have found that GDNF is not only expressed in the tumor microenvironment but is also expressed as an autocrine factor by prostate cancer cell lines. Our data indicate that the rate limiting factor for RET signaling in most prostate cancer is GFRα1, which is secreted by nerves, creating a niche promoting RET signaling in the perineural space. Furthermore, RET signaling can enhance cellular phenotypes associated with cancer progression in vitro and tumorigenesis in vivo. We have found that PI3K and ERK signaling downstream of RET induce activation of p70 S6 kinase (p70S6K), which is required for RET induced proliferation, invasion, and soft agar colony formation. Finally, we have found that RET is expressed in subset of localized prostate adenocarcinomas as well as small-cell neuroendocrine cancers. Given that agents targeting RET signaling are already available in the clinic, our findings suggest that RET is a potential therapeutic target on both adenocarcinoma and small-cell carcinoma of the prostate.

Human prostate cancer cells LNCaP, 22Rv1, DU145, and PC3 cells, and PNT1A, an immortalized normal prostate cell line, were all maintained in RPMI1640 medium (Invitrogen) supplemented with 10% FBS (Invitrogen). LAPC4 cells were cultured in RPMI1640 medium with 10% FBS supplemented with 10 nmol/L R1881 (Sigma). VCaP and 293T cells were maintained in DMEM (Invitrogen) with 10% FBS. All cell culture medium contained 1× antibiotic-antimycotic (Gibco). PNT1A cells were obtained from the European Type Culture Collection. All other cell lines were obtained from the ATCC. Cells were obtained between 2001 and 2012, expanded, frozen, and stored as stocks in liquid nitrogen. Cell lines were tested monthly for mycoplasma contamination. All cell lines are authenticated by short tandem repeat (STR) analysis at MD Anderson Cancer Center Characterized Cell Line Core Facility at the time of use.

For GDNF and GFRα1 treatments, cells were cultured with FBS-free medium for 24 hours and then treated with recombinant human GDNF at 100 ng/mL (ProSpec) or recombinant human GFRα1 at 100 ng/mL (R&D Systems) for 30 minutes. For inhibitor treatments, cells were treated with the PI3K inhibitor LY294002 at 20 μmol/L (Cell Signaling Technology), the MEK inhibitor U0126 at 10 μmol/L (EMD Millipore), or vehicle for 1 hour and cell pellets collected for Western blot analysis.

Tissue samples were obtained from Baylor College of Medicine Human Tissue Acquisition and Pathology Core under an Institutional Review Board approved protocols.

For Western blot analysis, cell pellets were resuspended with T-PER tissues protein extraction reagent (Thermo Scientific) containing protease inhibitor cocktail (Sigma-Aldrich) and phosphatase inhibitor cocktail (PhosSTOP EASYpack, Roche Diagnostics), sonicated for 10 seconds and clarified by centrifugation. Extracted proteins were quantified by the Bio-Rad Protein Assay. Proteins were subjected to electrophoresis in PAGEr Gold Precast Protein Gels (Tris-Glycine; Lonza) and transferred to membranes. Membranes were then blocked with 5% nonfat dry milk for 1 hour at room temperature, incubated with primary antibody overnight at 4°C followed by a secondary antibody–conjugated horseradish peroxidase for 1 hour at room temperature. Protein–antibody complexes were detected by Pierce ECL Western Blotting Substrate or SuperSignal West Femto Maximum Sensitivity Substrate (both from Thermo Scientific). The following antibodies were used: GFRα1 (H-70; Santa Cruz Biotechnology); RET (Novus Biologicals); AKT, phospho-AKT Ser 473), p44/42 MAPK (ERK1/2), phospho-p44/42 ERK1/2, p70S6K, phospho-p70 S6K (T421/S424 #9204), β-actin (all from Cell Signaling Technology). Ponceau S staining was carried out on membranes by briefly rinsing in distilled water and then incubating the membrane in Ponceau S Staining Solution (Cell Signaling Technology) for 2 minutes at room temperature, washed in distilled water until distinct reddish-pink protein bands were visible, and a photograph of the stained membrane was taken.

To examine the mRNA expression of RET and GFRα1 in human prostate cancer cell lines, we carried out conventional RT-PCR and quantitative real-time RT-PCR (Q-RT-PCR). Total RNA was extracted using the RNeasy Kit (Qiagen). cDNA was synthesized using the amfiRivert cDNA synthesis Platinum Master Mix Kit (GenDEPOT). Primers used for conventional RT-PCR were: RET: 5′-ACA GGG GAT GCA GTA TCT GG-3′ (RET-2767F) and 5′-CTG GCT CCT CTT CAC GTA GG-3′ (RET-2920R) and GFRα1: 5′-TCC GGG TGG TCC CAT TCA TA-3′ (GFRα1-1039F) and 5′-AGC ATT CCG TAG CTG TGC TT-3′ (GFRα1-1333R). HPRT was used for endogenous control using primers: HPRT-F: 5′-GCAGACTTTGCTTTCCTTGG-3′ (HPRT-F) and 5′-TCAGGGATTTGAATCATGTTTG-3′ (HPRT-R). Quantitative PCR was performed according to standard TaqMan Fast protocol (Applied Biosystems). TaqMan primer/probe combinations for specific RET gene used was RET Hs01120030 (Life Technologies). β-Actin (ACTB Hs01060665; Life Technologies) was used for endogenous control.

Cells were cultured in FBS-free medium for 5 days. The conditioned medium was concentrated by Amicon Ultra-15 (UFC900324). ELISA was performed using GDNF E_max ImmunoAssay Systems (Promega) according to the manufacturer's instructions.

IHC was carried out on sections of formalin-fixed paraffin-embedded tissues by standard procedures on a Leica Bond III autostainer. Antigen retrieval was carried out with heat retrieval using Bond Epitope Retrieval Solution 2 (AR9640; Leica). A rabbit monoclonal anti-RET antibody (Cell Signaling Technology, clone E1N8X) was used at 1:100 dilution for 30 minutes at room temperature. Detection was carried out using Leica's Bond Polymer Refine Kit (DS9800) for 16 minutes followed by chromogen for 5 minutes. Counterstain was hematoxylin.

Cells were plated at 4 × 10³ cells/well in 100 μL medium in 96-well plates. The next day, cells were cultured with FBS-free medium. Cells were then treated with or without GFRα1 (100 ng/mL), recombinant human GDNF (100 ng/mL), anti-GFRα1 (2 μg/mL; Santa Cruz Biotechnology), anti-GDNF antibody (Santa Cruz Biotechnology), normal goat IgG at 2 μg/mL (Santa Cruz Biotechnology), or 10 μL of concentrated conditioned media (CM) with/without dorsal root ganglia (DRG) for 6 days, with media changed one time on day 3. Cell proliferation was then assessed by Cell Counting Kit-8 assay (Dojindo Molecular Technologies).

Cell invasion assay was performed using the BD BioCoat Matrigel Invasion Chamber (BD Biosciences). Cells were starved for 24 hours in serum-free medium prior to seeding and 1 × 10⁵ cells were then seeded into each insert in 0.5 mL of serum-free medium. Below the insert, 0.7 mL of serum-free medium with/without GFRα1 (100 ng/mL), GDNF (100 ng/mL), anti-GFRα1 (2 μg/mL; Santa Cruz Biotechnology), normal goat IgG (2 μg/mL; Santa Cruz Biotechnology), or 10 μL of concentrated CM with or without DRG was added into each well. After 24 hours, the invasion cells on the bottom surface of the membrane were fixed with 100% methanol and stained with 0.3% crystal violet in 2% ethanol for 20 minutes and cells counted under a microscope.

To evaluate anchorage-independent growth, 2.5 × 10³ cells/well were seeded in 0.4% agar on top of a base layer containing 0.6% agar in the 6-well plate. Cells were treated two times per week with 2 mL of cell culture media with/without GFRα1 (100 ng/mL), GDNF (100 ng/mL), anti-GFRα1 (2 μg/mL; Santa Cruz Biotechnology), normal goat IgG (2 μg/mL; Santa Cruz Biotechnology), or 50 μL of concentrated CM with/without DRG. The plate was incubated at 37°C at 5% CO₂ in a humidified incubator for 21 days. At the end of the experiment, colonies were fixed with methanol, stained with 0.5% crystal violet, and counted.

The RET inhibitor lenvatinib was obtained from Selleckchem (Catalog no. S1164). Cell invasion was assessed as above using 0.7 mL of 10% FBS-medium below the well in the absence or presence of lenvatinib (100 or 500 nmol/L). After 24 hours, the invasion cells on the bottom surface of the membrane were counted as described above. To evaluate anchorage-independent growth, 2.5 × 10³ cells/well were seeded in 0.4% agar on top of a base layer containing 0.6% agar in the six-well plate. Cells were treated two times per week with 2 mL of cell culture media in the absence or presence of lenvatinib (100 or 500 nmol/L) for 21 days. At the end of the experiment, colonies were fixed with methanol, stained with 0.5% crystal violet, and counted.

LNCaP or 22Rv1 cells were plated in 12-well plates at a density of 1 × 10⁵ cells/well. The next day, the growth media was removed and 1 mL of fresh medium with 8 μg/mL polybrene (Sigma) and 100 μL scrambled (catalog no. LVP015-G), GDNF (catalog no. iV008619b), or RET (catalog no. iV028816a) siRNA GFP Lentivirus (Applied Biological Materials) was added. The virus was removed after 24 hours of incubation and fresh medium was added to the wells. Stable transfectant populations of cells were selected by treatment with 2 μg/mL puromycin (Sigma) for 15 days.

Cells were transiently transfected with scrambled siRNA (catalog no. sc-37007), GFRα1 siRNA (catalog no. sc-35469), RET siRNA (catalog no. sc-36404; all from Santa Cruz Biotechnology) or p70S6K siRNA (catalog no. 6572, Cell Signaling Technology) using the Lipofectamine 2000 transfection reagent (Invitrogen). Twenty-four or 48 hours after transfection, cells were used for further analyses.

Dorsal root ganglia (DRG) were harvested from C57BL mice and were placed in growth factor-depleted Matrigel matrix (BD Biosciences) in a 100-mm dish with 10% FBS medium at 37°C at 5% CO₂ in a humidified incubator for 3 days. DRG were then cultured with 0.1% FBS medium for additional 3 days, and medium were collected and concentrated using centrifugal filter (Merck Millipore) to 200-μL volume. Control conditioned medium was obtained from identical conditions and procedure except there were no DRG in the Matrigel.

The in vitro model of nerve invasion is originally described by Ayala and colleagues (3) DRG were harvested from C57BL mice and placed in 6-well plates and covered with 20 μL of growth factor reduced Matrigel (BD Biosciences). After growth in RPMI1640 medium with 10% FBS for 7 days, 2 × 10⁵ cancer cells were added into the well. For GFRα1 blockage, anti-GFRα1 antibody (2 μg/mL) was added to media. Normal goat IgG (2 μg/mL) was used as a control. Twenty-four hours after the cancer cells were added, images were acquired using an Axiovert 40 CFL inverted microscope (ZEISS) and invasion cells were counted.

All experiments were carried out on 8- to 10-week-old male SCID mice in accordance with the IACUC-approved protocol. Tumor xenografts were established by subcutaneous injection over each flank using either LNCaP cells with stable RET knockdown (si-RET) or vector controls LNCaP cells (si-V; 5 × 10⁶ cells/site) mixed 1:1 with Matrigel (Becton Dickinson).The tumor size was measured twice weekly using calipers and the tumor volume in mm³ calculated as volume = (height × width × length)/2. Tumors were harvested 35 days after the cell inoculation and tumors were excised and weighed.

Numerical values were compared using Student t test, two-tailed with P < 0.05 considered significant when two groups were compared. For multiple groups, one-way ANOVA was used and when significant, pairwise comparison carried out with P < 0.05 considered significant for pairwise comparisons. The χ² test was used to compare the RET expression in benign prostate and cancer tissues.

Given that GDNF is secreted in the tumor microenvironment, we initially sought to determine if RET is expressed in a broad array of prostate cancer cell lines. RT-PCR of RNAs from all prostate cancer cell lines and the immortalized prostate epithelial cell line PNT1A shows that all prostate cancer cell lines and prostate epithelial cells express RET mRNA at variable levels (Fig. 1A). Western blot analysis confirmed RET expression at the protein level that was concordant with mRNA levels measured by qRT-PCR (Fig. 1B and C). Both GDNF and GFRα1 are released by nerves and have been previously shown to be able to induce cancer cell proliferation and invasion (23). We therefore treated prostate cancer cell lines with GDNF, GFRα1, or both and evaluated proliferation (Fig. 1D). Somewhat surprisingly, all cell lines responded to GFRα1 alone but only 22Rv1 cells responded to GDNF alone. Furthermore, combined treatment with GDNF and GFRα1 was no more potent than GFRα1 alone. Similar results were seen when invasion and soft agar colony formation were assessed (Fig. 1E). These findings implied that GDNF was present in the prostate cancer cell lines because all cell lines responded to exogenous GFRα1. Furthermore, only 22Rv1 cells can respond to exogenous GDNF, suggesting that they are the only prostate cancer cell line that expresses GFRα1. Indeed, as seen in Fig. 2A and B, by Western blotting and ELISA for GDNF, all prostate cancer cell lines expressed variable levels of GDNF protein. This indicates that GDNF can act as an autocrine factor in prostate cancer, in addition to its role as a potential paracrine factor from nerves. Furthermore, RT-PCR analysis showed that only 22Rv1 expressed significant amounts of GFRα1 (Fig. 2C) and Western blot analysis shows that 22Rv1 are the only prostate cancer cell line to express GFRα1 protein (Fig. 2D). It should be noted that even 22Rv1 responded to exogenous GFRα1, indicating that exogenous GFRα1 protein can enhance the response of prostate cancer cells to GDNF even in the presence endogenous GFRα1.

It is known that nerves can release soluble GFRα1 (24). We therefore examined the ability of DRG excised from mice to enhance cancer phenotypes in prostate cancer cells in vitro. As expected, GFRα1 is present in the CM of DRG in culture (Fig. 3A). CM from DRG was able to significantly enhance proliferation, invasion, and soft agar colony formation in both LNCaP and 22Rv1 cells (Fig. 3B). This was significantly blocked by neutralizing anti-GFRα1 antibody (Fig. 3C).

To evaluate the impact of GFRα1 on PNI, we used the DRG in vitro co-culture model of PNI. We found that LNCaP and 22Rv1 cancer cells invaded along the neurites and that PNI was significantly attenuated by anti-GFRα1 antibody (Fig. 3D), indicating that GFRα1 also plays an important role in the PNI. To confirm that PNI was increased by RET signaling, we carried out similar experiments using LNCaP and 22Rv1 cells with RET knockdown and showed significant decreases in PNI (Fig. 3E).

To confirm the role of RET in the biological effects of GFRα1, LNCaP and 22Rv1 cell lines were stably infected with siRNA lentivirus targeting RET or nontargeting control. qRT-PCR analysis confirmed that the expression RET mRNA was decreased by 81% in both targeted cell lines compared to scrambled control cells (Supplementary Fig. S1A). Knockdown of RET expression reduced the increased cell proliferation (Supplementary Fig. S1B), invasion (Supplementary Fig. S1C), and colony formation (Supplementary Fig. S1D) by GFRα1 in both LNCaP and 22Rv1 cell lines, indicating that RET mediates the effects of GFRα1 on these cell biological responses. These data were confirmed by transient transfections using a second siRNA (Supplementary Fig. S2).

To determine whether RET signaling can promote tumor growth in vivo, we inoculated LNCaP with stable RET knockdown or vector controls as subcutaneous xenografts in SCID mice. As can be seen in Fig. 4, RET knockdown significantly decreased tumor growth in vivo, confirming the oncogenic impact of RET signaling in vivo. Relevant to this observation, we have found that lenvatinib, which inhibits RET, can significantly inhibit both invasion and soft agar colony formation in both LNCaP and 22RV1 cells (Supplementary Fig. S3). Lenvatinib is a multikinase inhibitor that inhibits several other receptor tyrosine kinases in addition to RET (25). Thus, it is not clear that the effects of lenvatinib are attributable to RET inhibition alone, but these results, along with our in vivo studies, suggest that RET is potentially targetable in in vivo using known agents that are currently in clinical use for targeting RET in other cancers.

To determine the signaling pathways downstream of RET signaling, we treated LNCaP and 22Rv1 cells with GDNF, GFRα1, or both proteins and analyzed activation of key signaling pathways by Western blotting. In LNCaP cells, GFRα1 (and GDNF plus GFRα1) significantly stimulated phosphorylation of ERK1/2, but did not increase AKT phosphorylation, which is already high in these cells due to PTEN inactivation (Fig. 5A). In 22Rv1 cells, GDNF, GFRα1, and GDNF plus GFRα1 stimulated AKT phosphorylation, with GFRα1 and GDNF plus GFRα1 showing stronger effects (Fig. 5A). ERK1/2 phosphorylation was not increased in 22Rv1 cells, which have intrinsically high ERK activation. However, in both cell lines there was a marked increase on phosphorylation of p70S6K. Knockdown of RET confirmed that the effects of GFRα1 are mediated via RET signaling (Fig. 5B). This was further confirmed with a second RET siRNA (Supplementary Fig. S4A).

To dissect the role of AKT and ERK phosphorylation in the activation of p70S6K, we treated each cell lines with either the selective PI3K inhibitor, LY294002, or the selective inhibitor of the MAPK pathway, U0126. Both drugs showed the expected activities in both cell lines (Fig. 5C). Interestingly, treatment with either LY294002 or U0126 almost completely abolished activation of p70S6K in both cell lines. These results indicate that activities of both PI3K and MAPK pathways are required for activation of p70S6K, although the dominant pathway activated by GDNF/GFRα1/RET depends on the cellular context.

Transfection of p70S6K targeting siRNA in LNCaP and 22Rv1 cells was used to evaluate the role of p70S6K in the effect of GFRα1 on cell phenotypes. Knockdown of p70S6K almost completely abolished cell proliferation (Fig. 5D), invasion (Fig. 5E), and colony formation (Fig. 5F) in response to GFRα1 treatment. This was confirmed with a second p70S6k siRNA (Supplementary Fig. S4B and S4C). Moreover, inhibition of p70S6K also decreased PNI (Supplementary Fig. S5A–S5D). These findings indicate that p70S6K plays a pivotal role in the effect of GFRα1/RET on these cellular behaviors.

Our data shows that RET is expressed in all prostate cancer cell lines tested, all of which were established from advanced prostate cancers. To examine the expression of RET in clinically localized prostate adenocarcinomas, we examined a tissue microarray containing 325 clinically localized cancers from radical prostectomies that has been described previously (26) using IHC. Positive control tissue was adrenal medulla (Supplementary Fig. S6A). Intensity (0–3) and extent (0–3) of staining was scored to derive a multiplicative index ranging from 0 to 9, with 9 representing strong diffuse staining (Fig. 6A) as described previously (26). A total of 61 cancers (18.8%) showed staining. Some had weak staining (≤3, shown in Fig. 6B), but this staining is almost certainly biologically significant since LNCaP xenografts, which are inhibited by RET knockdown, show similar intensity of staining by IHC (Fig. 6C). Interestingly, staining was highly cancer specific, with only a single case showing staining in benign luminal epithelium (P < 0.001; χ²). Nerves and ganglia in the stroma of occasional tissue microarray cores were stained as expected as an internal positive control (Fig. 6D). Supporting these findings, examination of TCGA databases in cBioportal show that prostate cancer has the third highest RET mRNA levels after pheochromocytoma/paragangioma and breast cancer (Supplementary Fig. S7). Of course, pheochromocytoma and paraganglioma would be expected to express high levels of RET because their tissues of origin intrinsically express high levels of RET. It has also been shown that more than half of all breast cancers express RET by IHC (27).

Recent reports have shown that RET mRNA is expressed in lung small-cell carcinoma cell lines (28) and in small intestinal neuroendocrine tumors (29). We therefore examined whether RET protein is expressed in small-cell neuroendocrine cancers of the prostate, which are highly aggressive tumors that can be present at initial diagnosis or arise from adenocarcinomas after prolonged treatment of with androgen blockade. We examined a radical prostatectomy which contained extensive areas of small-cell carcinoma and high-grade prostate adenocarcinoma. The majority of the adenocarcinoma component was negative for RET protein while the neuroendocrine component stained strongly (Fig. 6E). Some foci were observed with clear cut glandular features and showed weaker RET staining (Fig. 6F). We also examined two “channel TURPs” from patients with small-cell neuroendocrine carcinoma. Both stained positively for RET in a heterogeneous pattern (Fig. 6G; Supplementary Fig. S6B). To determine whether RET expression was also present in neuroendocrine tumors from other sites, we also analyzed a primary small intestinal neuroendocrine tumor (carcinoid), a carcinoid metastatic to liver, and a primary lung small-cell carcinoma. The metastatic carcinoid stained strongly (Fig. 6H), whereas the primary small intestinal carcinoid and the pulmonary small-cell carcinoma stained heterogeneously (Supplementary Fig. S6C and S6D).

In this study, we demonstrate that RET is ubiquitously expressed in prostate cancer cell lines, which are derived from advanced prostate cancers. In addition, it is expressed in cancer cells in 18% of clinically localized adenocarcinomas as assessed by IHC of a tissue microarray. It should be noted that prostate cancer is heterogeneous disease and other important molecular alterations, such as SPOP mutation or SPINK1 expression, are present in less than 15% of cases (30, 31). A smaller study of 30 whole mount prostatectomy sections by Dawson and colleagues (32) showed RET expression in the majority of cases by IHC. It should be noted that examination of whole mount specimens is more likely to detect areas of RET expression if there is significant heterogeneity in its expression, because TMAs present only a fraction of the cancer present in any given prostate. Thus, our TMA study may underestimate the number of cases with at least focal staining in the radical prostatectomy specimen. Of note, we only observed a single case (of 325) which showed staining of benign luminal epithelial cells, so staining was significantly increased in prostate cancer compared with benign epithelium. Dawson and colleagues (32) observed no staining in benign luminal cells, although they did observe weak staining in basal cells, which we did not observe. Thus, RET is expressed at increased levels in a significant fraction of prostate adenocarcinomas, although determining the exact percentage in various cancer stages will require further investigation.

In addition, to prostate adenocarcinomas, RET was expressed in three of three small-cell neuroendocrine prostate cancers tested, although in a heterogeneous pattern. Drake and colleagues (33) have previously identified a small-cell carcinoma of the prostate with activation of RET signaling based on phosphoproteomics, so it is likely that the RET signaling is active in some or all small-cell carcinomas of the prostate. In addition, we observed RET protein expression in neuroendocrine cancers of other origins, consistent with mRNA studies by other groups (28, 29), so RET expression is almost certainly a more general property of neuroendocrine cancers. Recent studies have shown that the ASCL1 transcription factor plays a key role in pulmonary neuroendocrine cancers and RET is a downstream target of this gene (28).

Our studies show that RET signaling can enhance proliferation, invasion, and soft agar colony formation in vitro and tumor growth in vivo in prostate cancer. In GDNF–RET signaling pathway, GDNF first binds to GFRα1 and then binds to and activates RET. GDNF is expressed by nerves and we have previously shown that GDNF is expressed in prostate cancer stroma (17). Normal prostatic stroma is rich in nerve fibers and nerves are increased in the peritumoral stroma in prostate cancer (34). We have now shown that prostate cancer cell lines express and secrete GDNF. Thus, GDNF is available within prostate cancer tumors to stimulate RET signaling as an autocrine and/or paracrine growth factor if GFRα1 is present. Of the cell lines tested, only 22Rv1 cells express sufficient GFRα1 to support GDNF/RET signaling. Huber and colleagues (35) have previously noted that only prostate cancer cell lines making GFRα1 can respond to exogenous GDNF. Our data indicate that secreted GFRα1 from nerves can provide this limiting factor in a paracrine manner. The expression of both GDNF and GFRα1 by nerves creates a perineural niche in which RET signaling can occur for prostate cancers expressing RET. Given that 22Rv1 cells express GFRα1, it seems likely that a subset of prostate cancers express GFRα1. The relative contribution of nerves and prostate cancer cells to GFRα1 protein expression is not completely clear and will require further studies. In addition, prostate cancer cell lines express GDNF, again implying that at least a subset of advanced prostate cancers express this growth factor. Thus, in addition, to RET signaling in the perineural niche, there may be autocrine signaling or combined autocrine/paracrine signaling via RET in prostate cancer. Again, further studies are needed to define the extent of expression GFRα1 and GDNF and the cells expressing these factors in various stages of prostate cancer.

RET signaling can activate both the PI3K/AKT and MAPK/ERK signaling pathways through phosphorylation of tyrosine 1062 in RET (36). It has been well documented that PI3K/AKT/mTOR and MAPK/ERK signaling pathways play a fundamental role in the regulation of cell growth, division, survival, and migration in cancer. In this study, we found that in LNCaP cells, RET signaling stimulated ERK1/2 phosphorylation whereas in 22Rv1 cells, RET signaling stimulated AKT phosphorylation. Thus, the exact pathways stimulated by RET signaling are dependent on the cellular context. However, in both cell lines, p70S6K phosphorylation was markedly increased by RET signaling. Inhibitor studies showed that activation of both ERK and AKT signaling was needed for RET induced p70S6K phosphorylation. p70S6K is well known as a key downstream target of P13K/AKT/mTOR pathway. It is less appreciated the ERK signaling can also enhance p70S6K phosphorylation (37). Mechanistically, this has been shown to be due to phosphorylation of TSC2 by ERK, which abrogates TCS2 inhibition of mTOR (38). In addition, RSK, which is downstream of ERK, can also directly target mTORC1 and increase its kinase activity (37). Of note, our studies demonstrate that knockdown of p70S6K markedly decreased proliferation, invasion, and soft agar colony formation induced by RET signaling, indicating that it plays a critical role in promoting these phenotypes. Interestingly, Horii and colleagues (39) have shown a strong correlation of phospho-p70S6K with both phospho-ERK and phospho-4EBP1 (another mTOR target) and with recurrent disease in breast cancer. Whether phospho-p70S6K has prognostic significance in prostate cancer is unknown but will be important to determine.

Our studies suggest that RET may be a therapeutic target in the subset of prostate adenocarcinomas expressing RET. PNI is associated with poor outcomes following radiotherapy (10, 11, 40), suggesting a prosurvival effect of prostate cancer cell interactions with nerves during radiotherapy. Of note, decreasing phospho-p70S6K using an mTOR inhibitor has been shown to enhance radiation sensitivity in lung cancer (41). Both ERK and AKT phosphorylation have also been strongly implicated in resistance to radiotherapy (42, 43). Treatment with vandetanib, which inhibits RET, results in improved responses to radiation in head and neck cancer (44), which is also neurotropic. We hypothesize that prostate cancers expressing RET may be particularly resistant to radiotherapy due to prosurvival RET signaling in the perineural niche and that targeting RET signaling could potential improve outcomes of radiotherapy. Of note, Huber and colleagues (35) have shown that DNA damage by radiation or chemotherapy enhances GDNF secretion by fibroblasts from the prostate and bone, which could further enhance RET signaling in this context. In addition, men with advanced cancer, including small-cell cancers might also benefit from targeting RET signaling. There are a variety of agents in the clinic or under development that can target RET signaling (36) or RET (27), including lenvatinib, which has been approved for treatment of thyroid and renal cancer (25). Further studies are needed to explore this potential therapeutic approach in prostate cancer.

No potential conflicts of interest were disclosed.

Conception and design: K. Ban, L. Shao, M. Ittmann

Development of methodology: K. Ban, S. Feng, L. Shao, M. Ittmann

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Ban, S. Feng, L. Shao, M. Ittmann

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Ban, S. Feng, L. Shao, M. Ittmann

Writing, review, and/or revision of the manuscript: K. Ban, S. Feng, L. Shao, M. Ittmann

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

Study supervision: M. Ittmann

We gratefully acknowledge the skilled assistance of Julie Zhao with RET IHC.

This work was supported by grants from the Department of Defense Prostate Cancer Research Program (W81XWH-14-1-0505; to M. Ittmann), the Prostate Cancer Foundation (to M. Ittmann), the Department of Veterans Affairs Merit Review program (5-IO1 BX002560-02; to M. Ittmann), the National Cancer Institute to the Dan L. Duncan Cancer (P30 CA125123) supporting the Human Tissue Acquisition and Pathology Shared Resource and by the use of the facilities of the Michael E. DeBakey VAMC.

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.