Hypoxic Tumor Kinase Signaling Mediated by STAT5A in Development of Castration-Resistant Prostate Cancer

In this study, we hypothesized that androgen-deprivation therapy (ADT) in prostate cancer, although initially efficient, induces changes in the tumor kinome, which subsequently promote development of castration-resistant (CR) disease. Recognizing the correlation between tumor hypoxia and poor prognosis in prostate cancer, we further hypothesized that such changes might be influenced by hypoxia. Microarrays with 144 kinase peptide substrates were applied to analyze CWR22 prostate carcinoma xenograft samples from ADT-naïve, androgen-deprived (AD), long-term AD (ADL), and CR disease stages. The impact of hypoxia was assessed by matching the xenograft kinase activity profiles with those acquired from hypoxic and normoxic prostate carcinoma cell cultures, whereas the clinical relevance was evaluated by analyzing prostatectomy tumor samples from patients with locally advanced disease, either in ADT-naïve or early CR disease stages. By using this novel peptide substrate microarray method we revealed high kinase activity mediated by signal transducer and activator of transcription 5A (STAT5A) in CR prostate cancer. Additionally, we uncovered high STAT5A kinase activity already in regressing ADL xenografts, before renewed CR growth was evidenced. Finally, since increased STAT5A kinase activity also was detected after exposing prostate carcinoma cells to hypoxia, we propose long-term ADT to induce tumor hypoxia and stimulate STAT5A kinase activity, subsequently leading to renewed CR tumor growth. Hence, the study detected STAT5A as a candidate to be further investigated for its potential as marker of advanced prostate cancer and as possible therapeutic target protein.


Introduction
The initial growth of a malignant prostate tumor is stimulated by androgens [1], and standard first-line treatment of patients with advanced prostate cancer includes androgen-deprivation therapy (ADT) [2]. Although initially effective, tumors inevitably recur in a castration-resistant (CR) state. Metastatic, CR prostate cancer is still remaining the most apprehensive aspect in prostate cancer management, defying prolonged treatment effects. Hence, improved understanding of the driving forces behind CR prostate cancer is warranted to enable development of more efficacious treatment strategies for this group of patients.
Hypoxia is a common microenvironmental factor of solid tumors, promoting tumor growth as well as angiogenesis, metastasis, and therapy resistance [3,4]. In prostate cancer, tumor hypoxia has indeed been correlated to poor prognosis [5][6][7], although skepticism remains as to its role and routine clinical importance. In vitro studies have shown that hypoxia increases the activity and sensitivity of the androgen receptor [8], which in many CR tumors may lead to a selection of adapted phenotypes [9].
It has for many years been well known that disease progression and development of treatment resistance in cancer generally are characterized by altered expression and activity of key mediators in cellular signal transduction, regulating processes such as proliferation, invasion, and angiogenesis [10]. The significance and potential use of such changes to improve cancer treatment are increasingly being recognized, specifically as aberrant protein kinases are playing significant roles in malignant tumor progression [11][12][13]. Of the human kinome (all kinases), approximately half of the tyrosine kinase complement is implicated in cancer [14]. Identification of aberrant tyrosine kinases has therefore emerged as an attractive approach in the quest for improved detection of disease aggressiveness and novel therapeutic targets.
We hypothesized that ADT in prostate cancer, although initially efficient, induces changes in the activity of the tumor kinome that subsequently lead to renewed, CR tumor growth.
To examine this hypothesis, microarrays with kinase peptide substrates, mainly representing tyrosine residues, were applied to analyze samples from an in vivo preclinical prostate cancer model reflecting different stages in development of CR disease. The impact of hypoxia was assessed by matching this model's kinase activity profiles with those generated from hypoxic and normoxic prostate carcinoma cell cultures. Clinical relevance was evaluated by analyzing prostatectomy tumor samples from patients with locally advanced disease, either in ADT-naïve or early CR disease stages.

Ethics Statement
All animal experiments were performed in accordance with protocols approved by the Animal Care and Use Committee at Department of Comparative Medicine, Oslo University Hospital (Permit Number: , in compliance with guidelines on animal welfare of the Norwegian National Committee for Animal Experiments. The clinical study protocol was approved by the Regional Committee for Medical and Health Research Ethics (REC South East, Permit Number: S-08607b) and was in accordance with the Helsinki Declaration. Written informed consent was required for participation.
The course of development of CR disease following ADT is illustrated in Figure 1A, using the preclinical model to elicit different disease stages. First, xenografts representing ADT-naïve disease were recruited when their shortest diameter reached 8 mm. Second, ADT was performed when the shortest xenograft diameter reached 12 mm, by surgical castration and simultaneous discontinuation of exposure to the testosterone pellet. Androgendeprived (AD) xenografts were recruited when their diameter was reduced to 8 mm (about 1 month post-castration). Third, longterm AD (ADL) xenografts were recruited when their diameter was reduced to 4 mm (about 4 months post-castration). Last, CR xenografts with renewed growth were included when their diameter reached 8 mm, being 5.460.4 months after castration. Xenograft diameters were measured by a caliper twice weekly from implantation until castration and once weekly after castration until the end of the experiment.
Blood samples for prostate-specific antigen (PSA) analysis were withdrawn from all animals' femoral veins right before euthanasia. The samples were allowed to coagulate before being centrifuged and stored at 280uC until all samples were analyzed simultaneously. Free and total serum PSA levels were assayed by the fluoroimmunonometric AutoDELFIA ProStatus TM PSA Free/ Total kit (PerkinElmer Life and Analytical Sciences, Wallac Oy, Turku, Finland). Euthanasia of animals was performed by cervical dislocations and immediately followed by excision of tumor tissue, which directly was snap-frozen in liquid nitrogen, and stored at 280uC.

Patients
For the purpose of validating preclinical data, four prostate cancer patients that had undergone radical prostatectomy with clear resection margins were included in the study ( Table 1). The patients were recruited at the Norwegian Radium Hospital, Oslo University Hospital. Histopathologic tumor and nodal status (pTN), tumor Gleason score, and preoperative serum PSA levels were recorded as part of standard clinical procedures. Patient 4 had received prolonged (,2 years) three-monthly s.c. goserelin at a dose of 10.8 mg until his PSA level started to increase following the ADT nadir, and he was referred to definitive surgery.

Tissue Preparation
Xenografts and surgical biopsies from the prostatectomy specimens, fresh-frozen immediately following resection, were first evaluated for tumor cell content, after being sectioned using a cryostat microtome and stained with hematoxylin and eosin. Any normal tissue present in tumor samples was roughly removed by dissection guided by the stains. All tumor specimens used had tumor cell content higher than 80%. The specimens were subsequently cut in 10 mm sections, continuously kept frozen and immediately re-stored at 280uC. A volume of ,3 mm 3 was cut of all xenograft samples, the required number of sections to cut calculated after measuring the samples' width and length. By identical procedures, ,5 mm 3 was sectioned of each prostatectomy biopsy (normal and tumor) specimen.

Data Analysis of Prostate Carcinoma Xenografts' Kinase Activity Profiles
Procedures for preparation of protein lysates and generation of kinase activity profiles are described in Text S1. Protein lysates from three biological replicates from each of the four experimental groups (ADT-naïve, AD, ADL, CR) were analyzed to generate the xenograft kinase activity profiles.
After visual check and background subtraction, endpoint signal intensities (the average signal intensity of the three last kinetic cycles) for each spot, i.e. for each kinase peptide substrate per array, were calculated by BioNavigator (PamGene International B.V.). A small number of negative signal intensities were managed by subtracting the 1% quantile of all data and setting remaining signal intensities less than 1 to 1. Subsequently, all signal intensities were log 2 -transformed, before the mean replicate signal intensity within each experiment was calculated for each peptide substrate. To correct for possible experiment-to-experiment variation, the mean of the four experimental groups' signal intensities for each peptide substrate was subtracted from each group's peptide substrate signal intensity. The microarray data are submitted to ArrayExpress (http://www.ebi.ac.uk/arrayexpress/); accession number E-MTAB-1572.
Statistical analysis was subsequently performed in PASW Statistics 18.0 (IBM, Somers, NY), with p,0.05 considered as statistically significant. ANOVA was applied to identify peptide substrates with significantly different phosphorylation levels between the experimental groups (ADT-naïve, AD, ADL, and CR). The Dunnett's many-to-one (AD, ADL, and CR versus ADT-naïve) post-hoc test adjusted for multiple comparisons. The peptide substrates' protein and gene names, including information about the cellular processes they are involved in, were retrieved using UniProt (http://uniprot.org; May 15, 2012) and HUGO Gene Nomenclature Committee (HGNC) (http://www. genenames.org; May 15, 2012) [16].

Kinase Activity Profiling of Prostatectomy Biopsy Specimens
Three technical replicates (of both tumor and normal samples) were analyzed from each patient, on one Tyrosine PamChipH96 plate, at the PamGene International B.V. location in 's-Hertogenbosch, The Netherlands. Sample running was performed as previously described [17]. From the resulting data set, calculated after visual check and background subtraction, data were log 2 -transformed after setting a few data points with signal intensities less than 1 to 1.

Hypoxic Exposure of Prostate Carcinoma Cells
The 22Rv1, PC3, and DU145 human prostate carcinoma cell lines were obtained from the American Type Culture Collection (ATCC). All cell lines were authenticated by short tandem repeat (STR) profiling based on 16 polymorphic markers in use by ATCC and paternity testing, and confirmed to be mycoplasmafree (MycoAlert, BioNordika, Oslo, Norway). All cell lines were maintained in RPMI 1640 medium (Sigma-Aldrich, Oslo, Norway) supplemented with 10% fetal calf serum, 1% L-Glutamine (Invitrogen, Oslo, Norway), and 1% penicillin/streptomycin (Invitrogen). The cells were routinely grown as a monolayer in 75 cm 2 flasks at 37uC in 95% air/5% CO 2 , and subcultured twice weekly to maintain exponential growth.
To investigate changes in kinase activity induced by hypoxia, exponentially growing cells at ,60% confluence, with fresh medium, were transferred to a hypoxia chamber for 24 hours (Invivo2 200; Ruskinn Technologies, Leeds, UK). Moderate hypoxia (1% O 2 ) was achieved by 5% CO 2 and residual N 2 . Severe hypoxia (anoxia) was obtained by 5% CO 2 and a mixture of 10% H 2 and 90% N 2 in combination with a catalyst that catalyzed the conversion of H 2 with any O 2 left to H 2 O. All normoxia and hypoxia experiments were performed thrice in order to generate three biological replicates of all experimental conditions.

Kinase Activity Profiling of Prostate Carcinoma Cells
Preparation of protein lysates is described in Text S1. Kinase activity profiling was done identically as for the xenografts, using 10 mg of total protein from all samples. Protein lysates from three biological replicates from each of the three experimental groups (normoxia, moderate hypoxia, severe hypoxia) were analyzed to  generate the kinase activity profiles. In the post-hoc Dunnett's test, the control condition was represented by the normoxic cells. An evaluation of the technical reproducibility of the arrays (interchip and intrachip correlations following analysis of the same cell culture lysate in all arrays) is presented in Text S2.

Cobalt Chloride Treatment
To simulate hypoxic exposure and expression of the hypoxiainducible factor-1a (HIF-1a) cells were exposed to 100 mM of the hypoxia-mimetic agent cobalt chloride (CoCl 2 ) (Sigma Aldrich) for 4 hours before protein lysates were made. Treatment with CoCl 2 induces HIF-1a expression by binding to the PAS domain, which blocks the HIF-1a pVHL binding, thereby giving HIF-1a stabilization [18].

Validation of the Preclinical Model
The CWR22 prostate carcinoma xenografts exhibited volume response patterns following castration similar to those seen in patients, and developed CR xenografts after 5.460.4 months ( Figure 1A). When accounting for the about five times faster metabolism in mice, this corresponds to around 2-3 years in humans. The clinically relevant disease progression of the xenograft was further confirmed by serum PSA quantifications from the animals' blood samples ( Figure 1B). The reduction in PSA from ADT-naïve xenografts to AD and ADL xenografts were significant (p = 0.032 and p = 0.042, respectively). In CR xenografts, pre-ADT PSA levels were restored.

Kinase Activity in Xenografts Following ADT
From analysis of the kinase activity profiles generated by the xenografts, phosphorylation of 35 of the kinase peptide substrates was found to be significantly affected (reduced or increased) after ADT by castration ( Table 2). These substrates represented proteins involved in several cellular processes, such as angiogenesis (e.g. PDGFR, VEGFR, EPOR), survival, proliferation, and progression (e.g. MAPK, RAF, STAT), and adhesion, migration, and invasion (e.g. FER, FES, MET).
Compared to ADT-naïve xenografts, some substrate phosphorylation levels generated by AD xenografts were weakly decreased, while others were increased. In Figure 2A the mean log 2 ratio and s.e.m. for all 35 significant substrates are shown for the three treatment groups versus ADT-naïve xenografts, with blue color indicating negative log 2 ratio and red color indicating positive log 2 ratio, respectively. In Figure 2B Volcano plots (-log 10 p-values versus log 2 ratios) are visualizing the kinase peptide substrates with the most significant and/or highest log 2 ratios between treated and ADT-naïve tumors' phosphorylation levels.
In accordance with the significant tumor growth inhibition following long-term ADT the level of substrate phosphorylation by ADL xenografts was strongly decreased for 32 of the 35 kinase peptide substrates, compared to the respective phosphorylation levels originating from exponentially growing ADT-naïve xenografts. Two of the three remaining kinase peptide substrates, representing the signal transducer and activator of transcription 1 and 5A (STAT1 and STAT5A), demonstrated significantly increased phosphorylation by ADL xenografts, with a log 2 ratio versus ADT-naïve xenografts of 1.560.3 for STAT1 (p = 0.004) and 1.560.7 for STAT5A (p = 0.019). Phosphorylation by CR xenografts was for the majority of the substrates restored at the same levels as for ADT-naïve xenografts. However, also from the CR xenografts, the phosphorylation of the STAT1 and STAT5A substrates was high, for STAT5A being significant compared to ADT-naïve xenografts (p = 0.040; log 2 ratio = 1.360.8).

Kinase Activity in ADT-naïve Prostatectomy Samples
To assess clinical relevance, kinase activity profiles of prostatectomy samples from prostate cancer patients were generated. A kinase activity signature was retrieved from the 100 kinase peptide substrates with highest phosphorylation levels, for xenografts and patients' tumor samples. For xenografts, the signature was represented by the mean of all ADT-naïve xenografts. For clinical samples, the signature was expressed by the mean of the three ADT-naïve tumor samples (cases 1, 2, and 3 in Table 1). The comparison of the two resulting signatures revealed 74 shared substrate phosphorylations ( Figure 3A and B), whereas each of them comprised 26 exclusively phosphorylated substrates ( Figure 3C). The 74% overlap indicates that the xenograft model used in the current study is a biologically relevant model of the clinical situation.

STAT1 and STAT5A Substrate Phosphorylation by Prostatectomy Samples
After uncovering increased phosphorylation of the STAT1 and STAT5A substrates by ADL and CR xenografts compared to by ADT-naïve xenografts, the corresponding substrate phosphorylation levels from prostatectomy samples were investigated. To allow comparisons between different patients, the substrate phosphorylations by each of the three ADT-naïve patient's tumor sample was normalized to the respective phosphorylations by the contralateral lobe normal tissue sample. One patient (case 4, Table 1) had received long-term ADT before early signs of CR disease were detected (rising serum PSA) and referral to radical prostatectomy. Since this patient's normal tissue also was androgen-deprived, normalization of substrate phosphorylation generated by the tumor sample to that from the normal tissue sample was precluded. Thus, for this patient, substrate phosphorylations by the tumor sample were normalized to the mean of the normal tissue substrate phosphorylations by the three ADT-naïve patients. The STAT1 substrate phosphorylation was low and similar for all patients. However, whereas the STAT5A substrate phosphorylation by tumor versus normal samples in all of the ADT-naïve patients was low; log 2 ratio = 0.761.2 (n = 3), it was substantially higher in the early CR patient; log 2 ratio = 2.6 (n = 1) ( Figure 4A).

Hypoxia-induced Increase in STAT5A Substrate Phosphorylation by Prostate Carcinoma Cells
From the results achieved in xenografts and prostatectomies, combined with the recognition of the potential role of hypoxia in development of CR disease [5], it was hypothesized that hypoxia might be involved in the altered STAT5A kinase activity. To Table 2. Kinase peptide substrates with significantly affected (reduced or increased) phosphorylation levels by samples from androgen-deprived (AD), long-term AD (ADL), or castration-resistant (CR) prostate carcinoma xenografts, versus androgendeprivation therapy (ADT)-naïve xenografts.  assess if the hypoxic state was present in the prostatectomy samples, Western immunoblot analysis of HIF-1a was performed on the prostatectomy tumor lysates. HIF-1a is the master regulator of oxygen homeostasis, known to be activated and stabilized under hypoxic conditions. Figure 4B shows that expression of HIF-1a was found in the lysate from the patient with early CR disease, as reflected by rising PSA following long-term ADT (case 4), whereas no HIF-1a expression was seen in the lysates from ADT-naïve tumors.
To further inspect this hypothesis, kinase activity profiles were generated from normoxic 22Rv1 prostate carcinoma cells, as well as from cells exposed to moderate (1% O 2 ) and severe (,0.02% O 2 ) hypoxia for 24 hours. The 22Rv1 cell line is androgenresponsive and originally derived from the relapsed CWR22 xenograft [19], expressing both the androgen receptor and PSA, Figure 3. Kinase activity in androgen-deprivation therapy (ADT)-naïve xenografts versus ADT-naïve patient tumors. An ADT-naïve kinase activity signature was defined as the 100 kinase peptide substrates with highest phosphorylation levels. For prostate carcinoma xenografts, the signature was represented by the mean of all ADT-naïve substrate phosphorylation levels. For prostate cancer patients, a top 100 signature was produced by calculating the mean of the three ADT-naïve (case 1, 2, and 3; Table 1)  thus representing a model of advanced prostate cancer. The phosphorylation levels of 22 kinase peptide substrates were significantly increased or decreased by hypoxic cell lysates (moderate or severe hypoxia) compared to by normoxic cell lysates. These included the STAT5A substrate, which presented an increasing phosphorylation level with increasing hypoxia ( Figure 5A, blue bars), although the mean phosphorylation level for all the 22 significantly affected kinase peptide substrates was low ( Figure 5A, grey bars). The difference in mean phosphorylation level at moderate versus severe hypoxia was not significant. For STAT5A, the log 2 ratios compared to normoxic cells were 0.4760.27 (p.0.05) for the moderately hypoxic cells and 0.6360.31 (p = 0.020) for the severely hypoxic cells. A list of the 22 kinase peptide substrates significantly affected by hypoxia is provided in Figure 5B.
A Western immunoblot correspondingly confirmed an increased expression level of phospho-STAT5A/B (Y694/699) in severely hypoxic 22Rv1 cells, compared to normoxic 22Rv1 cells ( Figure 5C). The expression of the total STAT5A protein was also found to increase under hypoxia, but to a lesser degree than the increase in phospho-STAT5A/B expression. To exclude that the finding of hypoxia-induced increased STAT5A activity was cellline specific we analyzed two additional cell lines, the PC3 and DU145 prostate cancer cell lines, revealing similar results as in the 22Rv1 cell line ( Figure 5C).
To elucidate whether STAT5A activation was directly induced by HIF-1a activation we treated the cells with CoCl 2 , a chemical inducer of HIF-1a. As seen in Figure 5D, CoCl 2 induced increased expression of both HIF-1a and phospho-STAT5A/B in all cell lines. The STAT5A protein expression was increased in one cell line and unchanged in the two other cell lines. Since HIF-1a specifically induced an increase in STAT5 phosphorylation these results supports that hypoxia is the cause of increased STAT5 activity, through a HIF-1a-mediated mechanism.

Discussion
The current study identified increased phosphorylation level of the STAT5A peptide substrate by ADL and CR prostate carcinoma xenografts, compared to by ADT-naïve xenografts. Similarly, in a limited population of prostate cancer patients studied, low phosphorylation levels of the STAT5A substrate by tumors from ADT-naïve patients' were revealed, whereas the STAT5A phosphorylation level generated by the tumor from an early CR disease case was substantially higher. Thus, an association between high kinase activity mediated by STAT5A and development of CR disease was uncovered both in xenografts and patient tumor samples. Further, we detected higher phosphorylation levels of the STAT5A substrate by hypoxic prostate carcinoma cells than by the normoxic counterpart. Hypoxia-mediated activation of STAT5A signaling has to our knowledge not been reported in prostate cancer.
STAT5 belongs to the STAT gene family of transcription factors [19]. STAT5 comprises two highly homologous isoforms; the 94 kDa STAT5A and the 92 kDa STAT5B, being encoded by separate genes [20]. STAT5A and B (STAT5A/B) are acting both as cytoplasmic signaling proteins and as nuclear transcription factors. STAT5A/B is activated by phosphorylation of a tyrosine residue in the carboxy-terminal domain (Y694 for STAT5A and Y699 for STAT5B) upon stimulation of various mediators, often by a tyrosine kinase of the Janus kinase (JAK) protein family [20,21]. After phosphorylation, STAT5A and STAT5B homodimerize or heterodimerize and translocate into the nucleus, where they bind to the STAT response element of target genes to For the androgen-deprivation therapy (ADT)-naïve prostate cancer patients, the log 2 ratios between the substrate phosphorylation levels generated by the prostatectomy tumor sample versus the respective contralateral lobe normal tissue sample were calculated. For the patient that received long-term ADT before radical prostatectomy, the log 2 ratio between the substrate phosphorylation levels from the tumor sample versus the mean substrate phosphorylation levels from the ADT-naïve patients' normal tissue samples were calculated (since this patient's normal tissue also was exposed to ADT). (B) Western immunoblot analysis of the prostatectomy tumor samples' lysates with an antibody against hypoxia-inducible factor-1a (HIF-1a), the main protein known to be activated and stabilized under hypoxic conditions, revealing increased HIF-1a expression in the lysate from the early CR tumor (case 4). No HIF-1a expression was seen in lysates from ADT-naïve tumors. doi:10.1371/journal.pone.0063723.g004 Figure 5. Hypoxia-induced increase in STAT5A substrate phosphorylation by prostate carcinoma cells. (A) 22Rv1 cells were exposed to moderate (1% O 2 ) and severe (,0.02% O 2 ) hypoxia for 24 hours in a hypoxia chamber. Protein lysates from three biological replicates from each of the three experimental groups (normoxia, moderate hypoxia, severe hypoxia) were analyzed to generate the kinase activity profiles. Kinase activity profiling of hypoxic and normoxic cell lysates detected increased STAT5A substrate phosphorylation with increasing hypoxia (blue bars), whereas the mean phosphorylation level for all 22 kinase peptide substrates with significant changes was low (grey bars). *The increase from normoxia to severe regulate specific gene transcription [20,22]. The STAT transcription factors are involved in regulation of several biological responses, including differentiation, proliferation, and migration [22].
Previous studies have identified STAT5A/B to be a critical survival factor in prostate cancer cells [23], and have further demonstrated increased STAT5A/B immunostaining in highgrade and CR clinical prostate tumors [24,25], that active STAT5A/B in primary prostate cancer is predictive of early CR disease [26], and that transcriptional synergistic interaction between STAT5A/B and the androgen receptor inhibits apoptosis and promotes growth of prostate cancer cells in vitro [25]. Moreover, STAT5 is activated in a high percentage of clinical metastatic tumors, and activation of STAT5 in DU145 prostate carcinoma cells has been shown to cause substantially increased formation of experimental metastatic lung tumors [27]. Recently, these results were confirmed by the demonstration that STAT5 knockdown delayed development of CR disease in an in vivo preclinical model [28]. Our study verified the increased STAT5Amediated kinase activity in CR prostate cancer xenografts. Further, we showed that phosphorylation of the STAT5A peptide was heightened already in ADL xenografts.
In liver cancer cells, hypoxia-induced activation of STAT5B was shown to regulate expression of insulin-like growth factors [29], which are known to be involved in cell survival and differentiation. In a study of microvascular endothelial cells, the JAK/STAT axis was shown to be activated by hypoxia and related to inhibition of a cellular pro-apoptotic pathway, and the results indicated that the JAK/STAT pathway plays an important role in endothelial cell survival during tissue hypoxia and angiogenesis [30]. Although the JAK/STAT signaling cascade often has been associated with immune function, recent studies have evidenced a role of JAK/STAT in mediating angiogenesisrelated responses. For example, activation of STAT3 by the vascular endothelial growth factor (VEGF) was required for endothelial cell formation in a human in vitro model [31], and STAT5 was involved in interleukin-3-stimulated endothelial cell migration [32].
Hence, several studies propose a captivating link between the STAT proteins and survival of tumor endothelium during angiogenesis. Interestingly, in our study, several substrates representing kinases involved in angiogenesis were reactivated, i.e. presenting significantly higher level of substrate phosphorylation by CR xenografts compared to by ADL xenografts. These included PECAM1, several VEGFRs, PDGFRs, PIK3R1, and EPOR ( Figure 2). Signaling commenced by VEGFRs is involved in the formation and development of endothelial cells, whereas PDGFR-signaling is engaged in stabilization of immature vessel walls by formation of pericytes [33,34]. PDGFR, VEGFR and EPOR are also known to be activated by tumor hypoxia [4,35]. Intriguingly, since hypoxia is suggested to promote development of CR disease in prostate cancer, STAT5; as was found to be significantly activated already in ADL xenografts in this study, may represent a key hypoxia-induced factor involved in subsequent reactivation of angiogenesis-related kinase signaling, which consequently stimulates renewed, CR tumor growth.
In summary, by using a novel peptide substrate microarray method, we were able to generate kinase activity profiles of preclinical and clinical samples from different stages in development of CR prostate cancer. From these profiles, previously published observations of high STAT5A activity in CR prostate cancer were confirmed. In addition, we also detected high STAT5A kinase activity already in regressing xenografts after long-term ADT. Since increased STAT5A kinase activity was generated by exposing prostate carcinoma cells to hypoxia, we propose that long-term ADT may induce tumor hypoxia and stimulate STAT5A kinase activity, which subsequently may lead to renewed, CR tumor growth. Hence, the study detected STAT5A as a candidate to be further investigated for its potential as marker of advanced prostate cancer and as possible therapeutic target protein.

Supporting Information
Text S1 Materials and methods.