Pharmacologic Inhibition of Jak2–Stat5 Signaling By Jak2 Inhibitor AZD1480 Potently Suppresses Growth of Both Primary and Castrate-Resistant Prostate Cancer

Prostate cancer (PC) cells initially require androgens and androgen receptor (AR) signaling for sustained growth and survival, a dependency exploited by androgen deprivation as a first-line therapy for advanced prostate cancer (1–4). The median duration of response to androgen deprivation of advanced prostate cancer is less than 3 years, after which castrate-resistant prostate cancer (CRPC) emerges (1–4). Additional hormonal manipulations and conventional chemotherapy may be used but these have limited impact on overall survival in CRPC (4) and, therefore, new and more efficient therapies for CRPC are needed. Development of CRPC has been attributed to numerous molecular mechanisms, including: (i) somatic mutations of AR resulting in increased affinity for ligands (5); (ii) amplification of the AR gene locus (6); (iii) biosynthesis of androgens within prostate cancer cells from adrenal steroids and cholesterol (7); (iv) expression of constitutively active alternative splice variants of AR, which do not require ligand to support prostate cancer growth (8–11); and (v) promotion of prostate cancer cell growth and survival by protein kinase pathways, which act through stimulation of AR (12, 13).

Distinct from these AR-dependent pathways, it is recognized that protein kinase signaling pathways are also capable of stimulating CRPC growth independently of AR signaling. Identification of new therapeutic targets in signaling pathways that bypass AR may provide a novel strategy for treatment of advanced prostate cancer. Stat5a/b has been identified and validated as a potential therapeutic target protein in prostate cancer (14–22). A rationale for the clinical use of Stat5a/b inhibitors is provided by the finding that Stat5a/b promotes prostate cancer growth and tumor progression through both AR-dependent and -independent mechanisms (12, 14–17, 20). Although Stat5a/b increases transcriptional activity of AR and proliferation of AR-positive prostate cancer cells (12), Stat5a/b is also critical for viability of prostate cancer cells that are negative for AR (17), suggesting that targeting Stat5a/b may provide a dual strategy to inhibit growth of CRPC by suppressing AR-dependent as well as AR-independent pathways. Activated Stat5a/b and autocrine prolactin (Prl) levels are associated with prostate cancers of high histologic grades (21, 23), CRPCs (21, 23) and distant metastases (18, 23), while absent in normal prostate epithelium (14). Stat5a/b is involved in progression to advanced prostate cancer, as evidenced by the ability of Stat5a/b to promote metastasis of prostate cancer cells as well as enhance hallmarks of the epithelial-to-mesenchymal transition that precedes metastasis (18). Activated Stat5a/b expression in prostate cancer was recently identified as a significant predictor of both early-disease recurrence and prostate cancer–specific death (21, 22), implicating Stat5a/b involvement in growth and progression of metastatic CRPC in patients. Finally, the Stat5a/b gene locus has been shown to undergo amplification in 29% of distant CRPC metastases (24).

Stat5a/b belongs to the Stat family of transcription factors, and is composed of two highly homologous isoforms, 94-kDa Stat5a and 92-kDa Stat5b (25). Stat5a/b are latent cytoplasmic proteins that become activated by phosphorylation of a conserved tyrosine residue in the carboxy-terminal domain. In prostate cancer cells, the predominant tyrosine kinase that activates Stat5a/b is Jak2 (26), which is recruited to the membrane-proximal region of ligand-activated cytokine receptors and is activated through autophosphorylation. Activated Jak2 phosphorylates Stat5a/b on conserved tyrosine residues Y694Stat5a and Y699Stat5b, leading to Stat5a/b dimerization and translocation to the nucleus, where the dimers bind to specific Stat5a/b response elements of target genes for transcriptional regulation (25). Thus, Stat5a/b signaling can be targeted at the level of the transmembrane receptor, further downstream by inhibition of Jak2 kinase or by direct blockade of Stat5a/b activity.

In this study, we investigated the efficacy of pharmacologic targeting of the Jak2–Stat5a/b signaling pathway to block primary and castrate-resistant growth of prostate cancer. Jak2-induced phosphorylation and activation of Stat5a/b was inhibited by AZD1480, a potent ATP-competitive small-molecule inhibitor of Jak2 kinase (27). We show that AZD1480 effectively suppressed Stat5a/b signaling and decreased viability of not only human prostate cancer cell lines and xenograft tumors, but also of clinical prostate cancers ex vivo in organ explant cultures. Most importantly, AZD1480 blocked the growth of primary CWR22Pc tumors, the emergence of castrate-resistant tumors and the growth of established CRPCs in the CWR22Pc xenograft tumor model.

Human prostate cancer cell lines CWR22Rv1, PC-3, DU145, LNCaP (American Type Culture Collection), and CWR22Pc were cultured in RPMI-1640 (Mediatech) containing 10% FBS (Quality Biological) and penicillin/streptomycin (Mediatech, Inc.; 50 IU/mL and 50 μg/mL, respectively). LNCaP and CWR22Pc cells were cultured in the presence of 0.5 and 0.8 nmol/L dihydrotestosterone (DHT; Sigma), respectively. Normal human prostate epithelial cells RC165N (28) were cultured in keratinocyte-serum–free (Gibco) medium supplemented with EGF, bovine pituitary extract (Gibco), and l-glutamine. AZD1480 and bicalutamide were provided by AstraZeneca and docetaxel (20 mg/mL) was purchased from Sanofi-Aventis.

CWR22Rv1, CWR22Pc, DU145, and PC-3 cells were solubilized and immunoprecipitations and immunoblottings were conducted as described previously (14–18). Antibodies used for immunoprecipitation and immunoblotting are described in the Supplementary Materials and Methods.

Generation of Stat5a constructs and the dimerization assay are described in the Supplementary Materials and Methods.

PC-3 cells were transfected with pStat5a-Flag and pPrlR, serum-starved for 16 hours, pretreated with AZD1480 or vehicle for 2 hours, stimulated with 10 nmol/L of human Prl (hPrl) for 30 minutes, fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and incubated with mouse anti-Flag polyclonal antibody (pAb; 1:200; Genomics), followed by goat anti-mouse fluorescein immunoglobulin G (IgG; 1:200; Vector Laboratories). Immunofluorescence staining was detected using a Zeiss LSM 510 laser-scanning microscope with an Apochromat X63/1.4 oil immersion objective.

COS-7 cells were transfected with plasmids (3 μg of each) expressing Prl receptor (PrlR; pPrlR) and Stat5a (pStat5a) or Stat5b (pStat5b) using FuGENE6, serum-starved for 10 hours, and pretreated with AZD1480 or vehicle for 2 hours, followed by stimulation with 10 nmol/L hPrl for 30 minutes. Nuclear extracts were prepared and a gel electromobility shift assay (EMSA) was conducted as described previously (16, 29, 30).

PC-3 cells (2 × 10⁵) were transiently cotransfected with 0.25 μg of pStat5a or pStat5b, pPrlR (prolactin receptor), 0.5 μg of pStat5a/b-luciferase (β-casein-Luc), and 0.025 μg of pRL-TK (Renilla luciferase). After 24 hours, cells were serum-starved for 20 hours, pretreated with AZD1480 at indicated concentrations for 1 hour, and stimulated with 10 nmol/L hPrl for 16 hours. The lysates were assayed for firefly and Renilla luciferase activities using the dual-luciferase reporter assay system (Promega) as described previously (12).

Gene delivery and expression of dominant-negative (DN) Stat5a/b (DNStat5a/b) and DNStat3 in prostate cancer cells were achieved using an adenoviral (Ad) vector. Generation of adenoviral constructs is described in the Supplementary Materials and Methods.

Cell viability was analyzed by the CellTiter 96 AQueous Assay Kit (Promega) according to the manufacturer's protocol.

CWR22Rv1 (data not shown) or CWR22Pc cells were treated with AZD1480 or vehicle for 24, 48, and 72 hours. Cells were fixed with 70% ethanol at 4°C overnight and washed with cold PBS twice before staining with propidium iodide (PI) and RNase A (Sigma). PI fluorescence intensity was analyzed by a flow cytometer using FL-2 channel.

Caspase-3 activity was determined by a fluorometric immunosorbent enzyme assay (Roche) as described in the Supplementary Materials and Methods.

CWR22Rv1 tumor xenografts were grown in male CB-17 severe combined immunodeficient (SCID) mice purchased from Charles River Laboratories. Mice were maintained under specific pathogen-free conditions and were used in compliance with protocols approved by the Institutional Animal Care and Use Committee of AstraZeneca, which conform to institutional and national regulatory standards on experimental animal usage. Studies were carried out using CWR22Rv1 cells implanted subcutaneously or luciferase-tagged CWR22Rv1 cells implanted orthotopically.

For studies with CWR22Pc tumors (31), castrated male athymic mice (Taconic), cared for according to the institutional guidelines of Thomas Jefferson University (Philadelphia, PA), were implanted with sustained-release DHT pellets (60-day release, 1 pellet/mouse; Innovative Research of America) 3 days before prostate cancer cell inoculation. CWR22Pc cells (1.5 × 10⁷) in 0.2 mL of Matrigel (BD Biosciences) were inoculated subcutaneously into flanks of nude mice (1 tumor/mouse). AZD1480 was dissolved in 0.1% Tween 80 (Sigma)/0.5% Methyl Cellulose (HPMC, K4M prep; Dow Chemical), bicalutamide in 0.5% Tween 80/PBS, and docetaxel in PBS (1 mg/mL).

Details of drug administration and tumor measurements for CWR22Rv1 and CWR22Pc tumor studies are provided in the Supplementary Materials and Methods.

For organ cultures, prostate cancer specimens were obtained from clinical patients (Table 1) with localized or locally advanced prostate cancer undergoing radical prostatectomy and bilateral iliac lymphadenectomy. The use of the de-identified excess tissue specimens for research purposes was approved by the Thomas Jefferson University Institutional Review Board. Details of the ex vivo organ culture conditions and methodology, which we have described previously (19, 29, 32–37), are provided in the Supplementary Materials and Methods.

Immunohistochemical staining of CWR22Pc xenograft tumors and clinical human prostate cancers grown as ex vivo organ cultures was conducted as described previously (19, 21, 22, 29, 30, 38). Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) assay was conducted using the In Situ Cell Death Detection Kit (Roche), as described previously (32).

Viable cells, active nuclear Stat5a/b, Stat3, or apoptotic cells with fragmented DNA versus total number of cells (viable and dead) were counted for three views/tumor and one view/organ culture explant (20–25 explants per treatment group per patient) and expressed as percentages, as shown previously (32). All percentages within each treatment group (tumor or organ culture explants) were averaged.

Statistical analysis of prostate cancer xenograft tumors and ex vivo organ cultures of clinical prostate cancers are provided in the Supplementary Material and Methods.

To determine the efficacy of the Jak2 inhibitor AZD1480 in suppressing Stat5a/b–driven growth of prostate cancer, we selected the CWR22 cell line/tumor system as our model because it mimics the clinical course of human prostate cancer when grown as a xenograft tumor in mice (31). CWR22 tumors were originally derived by Pretlow and colleagues (39) from a Gleason score 9 prostate cancer, and the original tumors could be maintained only by serial regrafting. CWR22Pc is a cell line, established from the original androgen-regulated primary CWR22 tumors (31), the growth of which is strongly upregulated by androgens (31) and downregulated by antiandrogens in culture (Supplementary Fig. S1). CWR22Pc cells form tumors in nude mice with 100% tumor take in the presence of androgens (31). Importantly, the tumors regress in response to androgen deprivation but eventually recur as castrate-resistant tumors (31). CWR22Rv1 is an androgen-independent prostate cancer cell line established from one of the original castrate-resistant CWR22R tumors (40).

Given that Jak1/2 are the primary kinases that phosphorylate both Prl-activated Stat5a/b (19, 23) and interleukin-6 (IL-6)–activated Stat3 (41, 42), we first evaluated the activation status of Stat5a/b versus Stat3 in exponentially growing CWR22Pc and CWR22Rv1 cells. Stat5a/b was expressed and constitutively activated, whereas Stat3 was not activated in CWR22Rv1 and CWR22Pc cells (Supplementary Fig. S2). In comparison, DU145 cells showed high activation of Stat3 as previously shown (ref. 17; Supplementary Fig. S2). CWR22 tumors and the cell lines derived from them are known to express autocrine Prl (23, 31). AZD1480 inhibited Prl-induced phosphorylation of Stat5a at 15 and 18 nmol/L concentration by approximately 50% (IC₅₀) in serum-starved CWR22Rv1 and CWR22Pc cells, respectively, whereas the IC₅₀ for Prl-induced Stat5b phosphorylation was 48 nmol/L for CWR22Rv1 and 70 nmol/L for CWR22Pc (Fig. 1A). The efficacy (IC₅₀) of AZD1480 in inhibiting constitutive activation of Stat5a in exponentially growing CWR22Rv1 and CWR22Pc cells was 16 and 7 nmol/L, respectively, whereas the IC₅₀ for Stat5b phosphorylation was 65 nmol/L for both cell lines (Fig. 1A). In summary, AZD1480 effectively reduced both ligand-induced and constitutive activation of Stat5a/b in a dose-dependent manner in human prostate cancer cells.

To investigate whether AZD1480 disrupts dimerization of Stat5a/b, we generated Flag- and Myc-tagged Stat5a constructs and cotransfected them into PC-3 cells (negative for endogenous Stat5a/b and AR; ref. 43). The cells were serum-starved and treated with AZD1480 or vehicle for 2 hours before induction of Stat5a dimerization by hPrl stimulation for 30 minutes. The lysates from the cells were immunoprecipitated with anti-Myc antibodies and immunoblotted with anti-Flag or anti-Myc antibodies (Fig. 1B). Protein input levels were analyzed by immunoblotting of whole-cell lysates (WCL) with anti-Flag, anti-Myc, and anti-actin antibodies. Stimulation of cells by hPrl induced Stat5a/b dimerization, as expected, and this was potently inhibited by AZD1480 treatment (Fig. 1B).

Because AZD1480 inhibited both phosphorylation and dimerization of Stat5a/b, AZD1480 was also expected to disrupt nuclear translocation of Stat5a/b in prostate cancer cells. To test this hypothesis, PrlR and Stat5a/b were expressed in PC-3 cells using replication-deficient adenovirus as an expression vector. The cells were serum-starved for 16 hours, treated with AZD1480 at indicated concentrations (Fig. 1C) for 2 hours, and stimulated with hPrl (10 nmol/L) for 20 minutes. In the absence of hPrl, Stat5a/b was predominantly localized to the cytoplasm of PC-3 cells (Fig. 1C, top), with hPrl stimulation inducing nuclear translocation of Stat5a/b (Fig. 1C, second from the top). Following pretreatment of cells with 200 nmol/L AZD1480, there was no ligand-induced nuclear immunostaining of Stat5a/b, indicating that AZD1480 inhibited nuclear translocation of Stat5a/b (Fig. 1C, bottom).

To determine whether AZD1480 inhibits DNA binding of ligand-induced phosphorylated Stat5, we used EMSA analysis. The Stat5a/b response element of the β-casein gene promoter was used as a probe and the ability of anti-Stat5a/b antibody to supershift Stat5a/b was verified (Fig. 1D). COS-7 cells transfected with PrlR and Stat5a were starved and pretreated with AZD1480 for 2 hours at indicated concentrations followed by hPrl-stimulation (10 nmol/L) for 30 minutes and EMSA analysis of the nuclear extracts. AZD1480 inhibited binding of Stat5a/b to DNA by more than 50% at a concentration of 25 nmol/L compared with cells treated with vehicle (Fig. 1D). To further assess whether AZD1480 is capable of inhibiting transcriptional activity of Stat5a/b, PC-3 cells transiently transfected with a β-casein-luciferase reporter gene were cotransfected with PrlR, Stat5a, or Stat5b. Cells were serum-starved for 16 hours and treated with AZD1480 for 1 hour, followed by stimulation with hPrl (10 nmol/L) in serum-free medium for 16 hours. As shown in Fig. 1D, AZD1480 inhibited transcriptional activity of Stat5a and Stat5b with IC₅₀s of 150 and 250 nmol/L, respectively. In summary, the results of the experiments shown in Fig. 1 indicate that AZD1480 inhibits transcriptional activity of Stat5a/b through disruption of Stat5a/b phosphorylation, dimerization, nuclear translocation, and DNA binding.

Inhibition of Stat5a/b by various methodologic approaches (antisense oligonucleotides, RNA interference, and adenoviral expression of a dominant-negative mutant of Stat5a/b) has been shown to inhibit proliferation and induce apoptotic death of human prostate cancer cells (14–17, 20). CWR22Rv1, CWR22Pc, PC-3, DU145, and normal prostate epithelial cells (RC165N; refs. 28, 44) were cultured for 72 hours with increasing concentrations of AZD1480 to determine whether Jak2 kinase inhibition suppresses cell growth (Fig. 2A). The number of viable cells decreased by 50% (GI₅₀) at concentrations of 482 nmol/L for CWR22Pc cells and 438 nmol/L for CWR22Rv1 cells. The GI₅₀ was significantly higher for PC-3 (1,755 nmol/L), DU145 (3,517 nmol/L), and normal prostate epithelial cells (2,083 nmol/L). Cell death induced by more than 1,000 nmol/L concentration of AZD1480 is considered to be caused by off-target effects of AZD1480 (27, 44). PC-3 cells do not express Stat5a/b (43) or Stat3 (43), and the kinase that activates Stat5a/b in DU145 cells is not Jak2 (18). Furthermore, RC165N normal prostate epithelial cells express low levels of Stat5a/b, which are not constitutively activated, and RC165N cells do not undergo cell death in response to Stat5a/b inhibition as previously shown (17), consistent with the lack of responsiveness to AZD1480. To verify that reduction of viable CWR22Rv1 and CWR22Pc cells in response to AZD1480 was caused by inhibition of Stat5a/b rather than Stat3, we inhibited Stat5a/b and Stat3 in parallel by adenoviral expression of dominant-negative Stat5a/b (AdDNStat5a/b) and Stat3 (AdDNStat3) at multiplicity of infection (MOI) 5 (Supplementary Fig. S3). In line with the results shown in Supplementary Fig. S2 showing lack of Stat3 activation in CWR22Rv1 and CWR22Pc cells, neither cell line responded to AdDNStat3, whereas AdDNStat5a/b decreased cell number by approximately 40% (Supplementary Fig. S3).

To characterize the growth inhibition of CWR22Rv1 (data not shown) and CWR22Pc cells, we conducted cell-cycle analysis at 24, 48, and 72 hours after treatment of cells with AZD1480 versus vehicle (Fig. 2B). AZD1480 did not alter the fraction of cells in G₁ or G₂–M, whereas the fraction of dead cells (sub-G₁) was increased at all time points in AZD1480-treated CWR22Pc cells. AZD1480 (800 nmol/L) treatment increased caspase-3 activation by 200% in CWR22Rv1 cells at 72 hours and by 150% in CWR22Pc cells at 48 hours (Fig. 2C). To determine whether introduction of active Stat5a/b is able to rescue cells from AZD1480-induced apoptosis, constitutively active Stat5a/b (17), which is not dependent on phosphorylation by Jak2 kinase for activation, was overexpressed using an adenoviral vector in CWR22Rv1 and CWR22Pc cells 6 hours before treatment of cells with AZD1480 (800 nmol/L). As shown in Fig. 2D, expression of constitutively active Stat5a/b prevented AZD1480-induced caspase-3 activation in both cell lines. Moreover, AZD1480 (800 nmol/L) induced caspase-3 activation to the same extent as genetic knockdown of Jak2 or Stat5 by RNA interference in CWR22Rv1 cells (Supplementary Fig. S4). In summary, the results shown in Fig. 2 show that AZD1480 induces apoptotic cell death in CWR22Rv1 and CWR22Pc cells via inhibition of Stat5a/b.

Because targeting Stat5a/b by Jak2 inhibition induced apoptosis of prostate cancer cells in culture, we next investigated whether AZD1480 affects prostate cancer xenograft tumor growth in vivo. Mice inoculated subcutaneously or orthotopically with CWR22Rv1 cells (Fig. 3A) were treated daily with AZD1480 (10 and 30 mg/kg) or vehicle. At the end of the treatment, growth of subcutaneous tumors was suppressed by 72% following treatment with 30 mg/kg AZD1480 (P < 0.0001) and by 52% following 10 mg/kg AZD1480 (P = 0.0002), whereas orthotopic tumor growth was suppressed by 54% following treatment with 30 mg/kg AZD1480 (P = 0.004) and by 6% following 10 mg/kg AZD1480 (P = 0.42; Fig. 3A). Immunoblotting of subcutaneous xenograft tumor lysates showed significant inhibition of Stat5a/b phosphorylation 2 hours following treatment with 30 mg/kg AZD1480 (Fig. 3B, top). Immunohistochemical analysis of tumors harvested after 11 days of drug treatment showed a 70% decrease in nuclear Stat5a/b in tumor tissues relative to the control group; very little expression or modulation of nuclear Stat3 was observed (Fig. 3B, bottom).

To investigate whether AZD1480 is capable of suppressing both primary and castrate-resistant growth of prostate cancer in a more disease-relevant model system, we used the CWR22Pc cell line/tumor model (31). CWR22Pc tumors mimic the clinical course of prostate cancer in humans, based on formation of androgen-dependent primary prostate cancer tumors, which regress upon androgen deprivation but recur as castrate-resistant prostate cancer (31). CWR22Pc cells were inoculated subcutaneously into the flanks of nude mice that had been castrated and supplied with androgen pellets to normalize serum androgen levels. To evaluate the efficacy of AZD1480 in suppressing growth of androgen-dependent primary CWR22Pc tumors, mice were treated with AZD1480 (30 mg/kg; n = 10), bicalutamide (50 mg/kg; n = 10), or vehicle (n = 10) starting on day 12 for 21 days (Fig. 3C, treatment window 1). AZD1480 significantly inhibited primary CWR22Pc tumor growth by 73% (P = 0.0004), whereas bicalutamide suppressed tumor growth by only 30% at a nonsignificant level (P = 0.78).

In the next set of experiments, mice carrying CWR22Pc tumors subcutaneously were androgen-deprived on day 32 by removal of DHT pellets. To evaluate whether AZD1480 is able to suppress the emergence of castrate-resistant CWR22Pc tumors (Fig. 3C, treatment window 2) or growth of established castrate-resistant CWR22Pc tumors (treatment window 3), the mice were treated for 95 days with AZD1480 (30 mg/kg; n = 10), docetaxel (5 mg/kg; n = 10), or vehicle (n = 10) starting on day 13 (treatment window 2) or 23 (treatment window 3) after the onset of androgen deprivation. AZD1480 significantly blocked not only the emergence of castrate-resistant CWR22Pc tumors (Fig. 3C, treatment window 2; P = 0.036), but also significantly blocked growth of established castrate-resistant CWR22Pc tumors (Fig. 3C, treatment window 3; P < 0.0001). Although docetaxel suppressed growth of tumors in treatment window 2 to a similar degree as AZD1480, docetaxel-treated mice were noticeably more cachectic and lethargic in comparison with AZD1480-treated mice, which maintained healthy body weight and overall activity in both treatment windows 2 and 3 (Supplementary Fig. S5) and better overall survival (Fig. 3D). Notably, treatment of established castrate-resistant CWR22Pc tumors in treatment window 3 with AZD1480 was more effective than docetaxel (Fig. 3C), and AZD1480-treated tumors showed slower regrowth posttreatment than docetaxel-treated tumors (Fig. 3C, posttreatment). Collectively, these data show that pharmacologic inhibition of Jak2 effectively inhibits not only primary androgen-dependent prostate cancer growth but also castrate-resistant growth of CWR22Pc tumors in nude mice.

CWR22Pc tumors harvested at the end of the treatment in all treatment windows were evaluated for nuclear Stat5a/b (Fig. 4A) and Stat3 (Supplementary Fig. S6) by immunohistochemistry (IHC), as well as for viable versus apoptotic cells by analysis of cell morphology (Fig. 4C) and presence of fragmented DNA (Supplementary Fig. S7). AZD1480 decreased nuclear Stat5a/b levels in all treatment windows significantly (Fig. 4A), whereas nuclear Stat3 levels were too low in this tumor model to show significant modulation (Supplementary Fig. S6). As expected, bicalutamide or docetaxel had no effect on nuclear Stat5a/b or Stat3 (treatment windows 2 and 3). Consistent with the inhibition of Stat5a/b nuclear localization by AZD1480, Western blot analyses of immunoprecipitated Stat5a/b from tumor lysates showed marked inhibition of Stat5a/b phosphorylation by AZD1480 (Fig. 4B). Importantly, AZD1480 reduced the fraction of viable tumor cells in all treatment windows (Fig. 4C) and, at the same time, increased the proportion of apoptotic cells (Supplementary Fig. S7). Moreover, AZD1480 downregulated the Stat5 target gene Bcl-XL expression in CWR22Rv1 cells in culture (Supplementary Fig. S8) and CWR22Pc tumors in vivo (Supplementary Fig. S9). AZD1480 also upregulated cell surface E-cadherin expression in CWR22Pc tumors (Supplementary Fig. S9) recapitulating the effects shown previously using genetic manipulation of Stat5 activity (15, 18). Finally, immunoblotting of tumor lysates with anti-AR antibody shows that AZD1480 did not affect the levels of the full-length AR, the AR splice variants or tissue prostate-specific antigen (PSA; Fig. 4D).

To investigate the efficacy of AZD1480 in inhibiting Stat5a/b activation and inducing cell death in clinical prostate cancers, we exploited a previously described ex vivo organ explant culture system of prostate cancers obtained from radical prostatectomies (19, 29, 32–37). We have extensively characterized this experimental model system previously for hormone/autocrine/paracrine regulation of normal and malignant prostate tissue at transcriptional and posttranscriptional levels (19, 29, 32–37). The advantages of explant organ cultures over cell lines are that all tissue components of prostate cancer are present in this ex vivo system and therefore the interactions of epithelium and stroma are maintained, which is critical for tissue-specific functions and for predicting the behavior of an individual tumor and its therapeutic response.

To determine responsiveness of clinical prostate cancers to AZD1480 in ex vivo organ explant cultures, 12 prostate cancers (Table 1) were cultured for 7 days in the presence of AZD1480 or vehicle at the indicated concentrations (Fig. 5). At the end of each culture period, viable and apoptotic carcinoma cells were assessed, as previously described (32), and Stat5a/b and Stat3 activation were analyzed by IHC (14, 30). Nine of 12 prostate cancers responded to AZD1480 by extensive loss of viable epithelium starting at a concentration of 25 μmol/L (Fig. 5A, left; P < 0.0001). At the same time, three of 12 prostate cancers did not respond to AZD1480 by reduced epithelial cell viability (Fig. 5A, right). It is important to note that a 25 μmol/L concentration of hormones or pharmacologic agents in organ explant culture of prostate cancer is equivalent to a concentration approximately 100 times lower (250 nmol/L) in cell culture, as has been established previously (19, 29, 32–37). This is because diffusion of hormones or small-molecule pharmacologic agents occurs much less efficiently in the tissue compartments of prostate cancer explants compared with single-layer cell culture, where cells are fully immersed in culture medium. Stat5a/b immunostaining of the explants showed that nuclear Stat5a/b levels were significantly decreased (P < 0.0001) in both AZD1480 treatment groups for the nine prostate cancers showing responsiveness to AZD1480 by reduced epithelial cell viability (Fig. 5B, left). In contrast, nuclear Stat5a/b levels remained constant in the nonresponsive prostate cancers (Fig. 5B, right). Importantly, AZD1480 induced apoptotic death of epithelial cells in cancer acini of the nine AZD1480-responsive prostate cancers compared with explants cultured in the presence of vehicle (Fig. 5C; 25 μmol/L, P = 0.011; 50 μmol/L, P = 0.023). This was evidenced by accumulation of dead cells in acinar lumens and an increase in apoptotic cells shown by in situ end labeling of fragmented DNA. Collectively, these data suggest that a majority of clinical prostate cancers from patients are responsive to pharmacologic Jak2-inhibition, showing decreased Stat5a/b activation and cell viability with increased apoptosis following AZD1480 treatment.

Therapeutic options for patients with advanced prostate cancer are limited. Current therapeutic strategies for metastatic prostate cancer are directed against AR signaling or use nontargeted cytotoxic regimens (1, 4). In this work, we show that prostate cancer growth in preclinical models can be effectively suppressed by targeting Stat5a/b signaling using pharmacologic inhibition of Jak2. The pharmacologic Jak2 inhibitor AZD1480 disrupted phosphorylation, dimerization, nuclear translocation, DNA binding, and transcriptional activity of Stat5a/b in prostate cancer cells, and induced apoptotic death of Jak2–Stat5a/b–driven prostate cancer cells in culture. Most importantly, AZD1480 inhibited not only primary androgen-dependent prostate cancer growth, but also castrate-resistant growth of recurrent prostate cancer, which emerged after androgen deprivation–induced regression of the original tumors in mice. Finally, we show that pharmacologic targeting of Stat5a/b in clinical prostate cancers ex vivo in organ explant cultures induced extensive epithelial cancer cell death.

Targeting Stat5a/b by AZD1480 inhibited growth of Jak2–Stat5a/b–driven prostate cancer cells in culture. We have shown previously that Stat5a/b is critical for the viability of human prostate cancer cells in vitro and for human prostate cancer tumor growth in nude mice using multiple alternative methods to genetically disrupt Stat5a/b action in prostate cancer cells (adenoviral delivery of DNStat5a/b, antisense oligonucleotides, or Stat5a/b siRNA; refs. 14–17, 20). These observations have been validated and extended to the TRAMP mouse prostate cancer model (45) and C4-2 model (20) by others. In this work, we show for the first time that pharmacologic inhibition of Jak2–Stat5a/b signaling inhibited prostate cancer cell viability. AZD1480 induced cell death through the mechanism of apoptosis and viability was rescued by overexpression of constitutively active Stat5a/b, which are results consistent with the type of cell death previously shown to be induced by genetic disruption of Stat5a/b activity in prostate cancer cells (14, 15).

Pharmacologic inhibition of Stat5a/b by AZD1480 effectively suppressed both primary androgen-dependent prostate cancer growth as well as CRPC growth in nude mice after androgen deprivation–induced regression of the original tumors. Stat5a/b has been shown previously to be involved in progression of prostate cancer to castrate-resistant disseminated disease. Notably, high nuclear Stat5a/b expression in clinical prostate cancers was identified as a significant predictor of both early recurrence and prostate cancer–specific death (21, 22), which both typically result from development of castrate-resistant metastatic disease. Moreover, Stat5a/b has been shown to promote development of disseminated prostate cancer in in vitro and in vivo models (18). In this work, we have shown that inhibition of Stat5a/b signaling suppressed CRPC growth in vivo. AR transcriptional activity is believed to recover in CRPC despite pharmacologic AR blockade and depletion of circulating and local androgens. Reactivation of AR in CRPC has been attributed to various molecular mechanisms, yet none completely account for progression to CRPC. Recently, we introduced the novel concept of a biologically significant, reciprocal physical interaction between Stat5a/b and AR, with each protein enhancing the other's nuclear localization and transcriptional activity in AR-positive prostate cancer cells (12). Consistent with our findings, a recent report showed that knockdown of Stat5a/b by antisense oligonucleotides reduced stability of the AR protein in prostate cancer cells (20). However, in recurrent castrate-resistant CWR22Pc tumors AZD1480 did not suppress transcriptional activity of AR, as evidenced by unaltered expression of PSA in the AZD1480-treated tumors versus control group. Furthermore, AZD1480 did not downregulate expression of full-length AR or AR splice variants, which have been suggested to mediate castrate-resistant growth of prostate cancer (8–11). Collectively, these results suggest that AZD1480-induced suppression of primary and castrate-resistant CWR22Pc tumor growth was not mediated through Jak2–Stat5a/b regulation of the AR. In line with this finding, we have previously shown that Stat5a/b inhibition induced rapid death of DU145 prostate cancer cells, which do not express AR (17). Finally, AZD1480 is a potent inhibitor of not only Jak2-induced Stat5a/b activation, but also Stat3 activation in prostate cancer cells. However, as we intended to focus our study on determining the importance of targeting Stat5a/b in CRPC growth by AZD1480, we selected the CWR22Pc cell line/tumor system as our experimental model because of low Stat3 activation levels in CWR22Pc cells and tumors. The results of the present study suggest that AZD1480 inhibited both primary and castrate-resistant growth of CWR22Pc tumors independently of Stat3. Future work will focus on identifying Stat5-regulated, AR/Stat3-independent molecular mechanisms governing regulation of prostate cancer cell viability.

AZD1480 induced apoptotic cell death and loss of viable epithelial cells in the majority of clinical prostate cancers ex vivo in organ explant cultures. In addition to the Prl–Jak2–Stat5 pathway, Stat5a/b in prostate cancer is potentially activated by multiple kinases other than Jak2 such as Src (17) and potentially by the EGF receptor family (46–49). Analysis of Stat5a/b activation status in prostate cancer samples before organ culture testing revealed that some of the AZD1480 nonresponsive prostate cancers had high levels of active Stat5a/b. This may be a result of kinases other than Jak2 activating Stat5a/b in those given nonresponsive prostate cancers. This suggests that direct inhibition of Stat5 by small-molecule Stat5 inhibitors may target larger fraction of prostate cancers with active Stat5. In future studies, it will also be important to determine whether expression of components of the Prl–Jak2–Stat5a/b autocrine loop, such as Prl and Jak2, will predict responsiveness of clinical prostate cancers to pharmacologic Jak2 inhibition. Clinical data available for Jak2 inhibitors in the treatment of myelofibrosis suggest that the adverse effects on hematopoiesis are manageable (50).

In summary, the findings presented in this work show, for the first time, responsiveness of primary and CRPC to pharmacologic targeting of the Jak2–Stat5a/b signaling pathway. These findings are important and of high translational value as they pave the way for clinical efficacy studies assessing the pharmacologic targeting of Stat5a/b in prostate cancer.

M. Zinda and D. Huszar have ownership interest (including patents) in AstraZeneca. L.G. Gomella is a consultant/advisory board member of Astellas, Dendreon, Bayer, and Janssen. M.T. Nevalainen has a sponsored research contract from AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Gu, A. Dagvadorj, M. Zinda, D. Huszar, M.T. Nevalainen

Development of methodology: L. Gu, A. Dagvadorj, S. Gupta, P. McCue, M.T. Nevalainen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Gu, Z. Liao, A. Dagvadorj, S. Blackmon, P. Talati, C.D. Lallas, E.J. Trabulsi, P. McCue, L. Gomella, M.T. Nevalainen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Dagvadorj, S. Gupta, S. Blackmon, E. Ellsworth, P. Talati, B. Leiby, E.J. Trabulsi, M.T. Nevalainen

Writing, review, and/or revision of the manuscript: L. Gu, D. Hoang, S. Gupta, S. Blackmon, E. Ellsworth, P. Talati, M. Zinda, C.D. Lallas, E.J. Trabulsi, D. Huszar, M.T. Nevalainen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Gu, D. Hoang, A. Dagvadorj, E. Ellsworth, L. Gomella, M.T. Nevalainen

Study supervision: L. Gu, M.T. Nevalainen

The authors thank Geraldine Bebermitz, Shenghua Wen, Denis Hughes, and Corinne Remer at AstraZeneca for their contributions.

This work was supported by grants from National Cancer Institute (2RO1CA11358-06), AstraZeneca, and Pennsylvania Department of Health (to M.T. Nevalainen). Shared Resources of Kimmel Cancer Center (KCC) are partially supported by NIH grant CA56036-08 (Cancer Center Support Grant, to KCC).

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.