Targeted inhibition of ERα signaling and PIP5K1α/Akt pathways in castration‐resistant prostate cancer

In the present study, we unrevealed the effect of tamoxifen on ERα and PIP5K1α/AKT that are responsible for prostate cancer growth and invasion. Tamoxifen in combination treatment with PIP5K1α inhibitor, ISA‐2011B, led to tumor regression in xenograft mice. We found that depletion of tumor‐associated macrophages using clodronate sensitized the inhibitory effect of tamoxifen on prostate cancer.

Selective ERa modulator, tamoxifen, is well tolerated in a heavily pretreated castration-resistant prostate cancer (PCa) patient cohort. However, its targeted gene network and whether expression of intratumor ERa due to androgen deprivation therapy (ADT) may play a role in PCa progression is unknown. In this study, we examined the inhibitory effect of tamoxifen on castration-resistant PCa in vitro and in vivo. We found that tamoxifen is a potent compound that induced a high degree of apoptosis and significantly suppressed growth of xenograft tumors in mice, at a degree comparable to ISA-2011B, an inhibitor of PIP5K1a that acts upstream of PI3K/AKT survival signaling pathway. Moreover, depletion of tumor-associated macrophages using clodronate in combination with tamoxifen increased inhibitory effect of tamoxifen on aggressive prostate tumors. We showed that both tamoxifen and ISA-2011B exert their on-target effects on prostate cancer cells by targeting cyclin D1 and PIP5K1a/ AKT network and the interlinked estrogen signaling. Combination treatment using tamoxifen together with ISA-2011B resulted in tumor regression and had superior inhibitory effect compared with that of tamoxifen or ISA-2011B alone. We have identified sets of genes that are specifically targeted by tamoxifen, ISA-2011B or combination of both agents by RNAseq. We discovered that alterations in unique gene signatures, in particular estrogen-related marker genes are associated with poor patient disease-free survival. We further showed that ERa interacted with PIP5K1a through formation of protein complexes in the nucleus, suggesting a functional link. Our finding is the first to suggest a new therapeutic potential to inhibit or utilize the mechanisms related to ERa, PIP5K1a/AKT network, and MMP9/VEGF signaling axis, providing a strategy to treat castration-resistant ER-positive subtype of prostate cancer tumors with metastatic potential.

Introduction
Treatment options for metastatic prostate cancer (PCa) remain limited. PCa often inevitably progressed to castration-resistant disease after initial androgen deprivation therapy. Thus, using agents to target androgen and its receptor AR lack the effectivity for castration-resistant PCa and render the tumor to develop resistance to anti-androgen-related therapies [1,2]. It is an urgent need to develop new treatment strategies to effectively target subtypes of PCa which do not respond to anti-androgen therapies.
Estrogen agonist/antagonist, tamoxifen (1-[pdimethylaminoethoxyphenyl]-1,2-diphenyl-1-butene) has been shown to inhibit growth of PCa cells in a few preclinical studies [3]. Clinically, high-dose tamoxifen therapy was well tolerated in a heavily pretreated patient cohort with hormone-refractory prostate cancer as reported in clinical trial studies [4]. However, the precise mode of action and specific targeted network factors by tamoxifen in primary tumors of PCa patients or in PCa cell lines have not been reported. Tamoxifen is the gold standard for the adjuvant treatment of breast cancer in pre-and postmenopausal patients with estrogen receptor alpha (ERa)-positive tumors and is the first selective estrogen receptor modulator [5,6]. Tamoxifen exerts its effect either by disrupting estrogen biosynthesis or inhibiting ERa activity [7]. However, for patients with breast cancer or PCa who suffer bone metastasis, alternative tailed treatment to eradicate metastatic lesions in the skeleton of patients with bone metastasis are needed [8]. Thus, understanding of the mechanisms that underlying targets of tamoxifen in PCa is essential for development of new therapy.
Estrogen (17b-estradiol, E2) exerts its effects by binding to specific estrogen receptors, and it is converted from androgen by aromatase that is dramatically increased in metastatic PCa tissue compared with primary tumors [9][10][11]. Expression of ERa has been shown to be abnormally increased in castration-resistant PCa (CRPC) and metastatic tumors [12,13]. ERamediated signaling can be broadly classified as either ligand-dependent or ligand-independent which trigger distinct signaling pathways to regulate a broad range of estrogen responsive genes [9,11,[14][15][16]. ERa-induced estrogen effects were suggested to promote epithelial versus mesenchymal transition (EMT) by regulating the activity of NOTCH1 signaling pathway. Further, E2 promoted tumor formation and metastasis, which were inhibited by tamoxifen treatment in mouse models [17]. It has also been reported that estrogen/ ERa axis plays an important role in promoting EMT-associated events during PCa development [18].
PIP5K1a is a lipid kinase similar to PI3K [19]. PIP5K1a acts upstream of PI3K/Akt by producing PIP 2 , a major lipid substrate for triggering PI3K activity. PI3K converts PIP 2 into PIP 3 , a key molecule for activation of Akt family of serine/threonine kinases [20]. PIP5K1a promotes PCa progression by regulating PI3K/Akt and AR pathways [21]. Further, PIP5K1a acts on PI3K/Akt and ERa pathways in promoting ERa-positive breast cancer invasion. Silencing of PIP5K1a sensitizes breast cancer cells to tamoxifen treatment, which in turns blocks ERa signaling [22].
We have developed ISA-2011B, a novel selective inhibitor of PIP5K1a, that exhibits specific inhibitory effect on advanced PCa by inhibiting elevated levels of PI3K/Akt and AR [21,23,24]. As PIP5K1a acts upstream of PI3K/Akt/PTEN and AR, a better understanding of the consequences of inhibiting these pathways and its effects on estrogen signaling pathways will have a significant impact on refining therapeutic strategies and improving clinical outcome. It is therefore of great interests to investigate whether tamoxifen alone or in combination with ISA-2011B may be used as effective treatment strategies for targeted therapy for subgroups of PCa.

Patient cohort, tissue specimens, and tissue microarrays
Paraffin-embedded tissues including benign prostate hyperplasia (BPH) (n = 96) and primary PCa (n = 113), and metastatic lesions in lymph nodes (LN) (n = 10), bone (n = 16), and lung (n = 4) were used to generate tissue microarrays (TMAs). The study was approved by the Ethics Committees, and the Helsinki Declaration of Human Rights was strictly observed.

The Cancer Genome Atlas (TCGA) program
Raw RNA-seq counts and relevant clinical data for TCGA prostate adenocarcinoma (TCGA-PRAD) provisional cohort of 498 prostate cancer patients were extracted from Genomic Data Commons (GDC) data portal. In addition to that, clustered heatmaps of normalized RNA-seq counts along with selected RNA expression covariate data were acquired and adapted from Next-Generation Clustered Heatmaps (NG-CHM) viewer (software version: 2.14.2, map version: 1.1.0). Clustered data was established using iCluster package that uses multiple genomic data types (e.g., copy number, gene expression, DNA methylation) in order to fit a regularized latent variable model based clustering that generates an integrated cluster assignment based on joint inference across data types. A total of 1 lg of RNA/sample was used for library preparation, and RNA sequencing was performed using Illumina PE150 sequencer (Novegene, Beijing, China).

Mapping and bioinformatics analysis
Clean PE150 reads were mapped on human reference genome, GRCh38.p12, acquired from Gencode, using HISAT2 (V2.1.0) aligner. Mapped reads were then sorted and indexed using samtools (v1.9) and counted using htseq (v0.9.1). All bioinformatics analysis of raw read counts were performed on R (v3.5.1), using different gene expression analysis package, DESeq2 (v1.22.2). Gene set enrichment analysis of biological process on gene ontology was performed using the Gene Ontology (GO) stats R package, and the conditional parameter of preranked gene lists was performed using GSEA (v4.0.3) [25]. This was done to identify pathways that were more represented in the set of significantly differentially expressed genes than would be expected by chance alone. All P values were also adjusted for the fact that many genes were being simultaneously analyzed by controlling the false discovery rate (FDR) with the Benjamini-Hochberg method [26][27][28][29]. Enrichment maps using enrichment phenotypes with false discovery rate (FDR) cutoff of 0.1, P value cutoff of 0.005, and overlap coefficient of 0.5 were visualized using Enrichment Map (v3.3.0) on Cytoscape (v3.8.0) [30,31].

Immunohistochemistry
Immunohistochemistry on tumor tissue arrays was performed as previously described [32]. The staining procedure was performed using a semiautomatic staining machine (Ventana ES; Ventana Inc.). The staining intensity was scored as 0 (negative), 1 (weakly positive or positive), 2 (moderate positive), or 3 (strongly or very strongly positive).

Proliferation assay
The effects of 17b-Estradiol on PCa cell lines were determined using the nonradioactive MTS proliferation assay (Promega Biotech) and/or Resazurin (R&D Systems, Abingdon, UK) assays were used according to the manufacturer's protocol. The detailed information is described in the Supporting information Appendix S1.

Mouse models of xenograft PCa tumors and treatment
The animal studies were approved by the Swedish Regional Ethical Animal Welfare Committee. The animal welfare and guidelines were strictly followed. Athymic NMRI nude male mice (Charles River Biotechnology, MA, USA) aged 6 weeks and weighing 25-29 g were used. 1 9 10 6 PC-3 cells/mouse were subcutaneously implanted into mice. After tumor were established, tumor diameters were measured using calipers, and volumes were calculated using the equation (a 9 b 2 /2), where a and b represent the larger and smaller diameters, respectively. Mice were randomized and were treated with vehicle (control), tamoxifen (40 mg/ kg), ISA-2011B (40 mg/kg) or combinations of tamoxifen (20 mg/kg), and ISA-2011B (20 mg/kg) by intraperitoneal injection every second day (n = 5 mice/ group).

Mouse models for macrophage-depletion and tumor spheroid formation assays
Subcutaneously implanted PC-3 tumors were grown into approximately 400 mm 3 and were randomized into 6 groups (3-4 mice per group). Mice from vehicle control, Tamoxifen, and ISA-2011B treatment groups were randomized and administered either Encapsome Ò or Clodrosome Ò . Mice were treated with EncapsomeÒ or Clodronate via intratumoral injection in between the treatments with vehicle control, Tamoxifen, or ISA-2011B, which was administered intraperitoneally. For xenograft tumor-derived spheroid formation assays, 5 9 10 3 single cell suspension were cultured in suspension in modified tumor spheroid semi-solid medium for 20 days and then counted.
2.8. Immunoblot, Immunoprecipitation analysis, and subcellular fractionation Subcellular fractionation, immunoblot, and immunoprecipitation analysis were performed as described previously [21]. For immunoprecipitation analysis, antibody against ERa was used to pull down the immune complexes, and antibody to IgG (Thermo Fisher Scientific, Sweden) was used as a negative control.

Immunofluorescence analysis
The PCa cells were seeded on glass coverslips and were subsequently treated with the different agents. Cells were fixed with 4% paraformaldehyde in PBS. The images were viewed and taken under an Olympus AX70 fluorescent microscope and software ACT2U (ACT2U version. 1.5, Stockholm, Sweden) was used as described in [21].

Induced overexpression via transfections
Transfections were performed using TransIT-X2 transfection reagent, according to the manufacturer's protocol (Mirus Bio LLC, Madison, WI, USA). For transient or stable transfection, control vector containing full-length human ERa cDNA or empty control vector pLX-304 were used. Cells overexpressing ERa or control vector were transient transfected into PC-3 cells. The phenol-red free medium containing 10% charcoal-stripped serum was used for culturing the cells. All plasmid vectors were purchased from Plas-mID Repository, Harvard Medical School.

Flow cytometry (FACS)-based cell cycle analysis and apoptosis assay
Apoptosis of breast cancer cells after treatment with ISA-2011B were performed using FACS-based assays. FITC or PE-conjugated Annexin V and 7-AAD was used for apoptosis assay according to the manufacturers' Protocol (BD Biosciences, San Jose, CA, USA). Data was analyzed using FCS Express (DeNovo Software, CA, USA), FLOWJO (Tree Star, Inc., OR, USA), and CytExpert (Beckman Coulter, FL, USA) softwares.

Holomonitor analysis
The digital Holomonitor M4 live cell imaging system and softwares were applied for measuring cell morphological changes and motility in response to drug treatment (PHI technologies, Lund Sweden). The microscope is placed in incubator at 37°C and 5% CO 2. The images of live cells were captured every 10 min during the entire treatment period of 48 h.

Statistical analysis
Tukey test, t-test, Kruskal-Wallis/ANOVA, and Spearman rank correlation tests were performed. The immunohistochemistry H-scores were log-transformed to overcome the non-normal distribution of H score data. The mean is the average value of all samples. The standard deviation (SD) is an indication of variability of all samples. The precision of the sample mean is indicated by standard error. Confidence levels are expressed using 95% confidence interval (CI). All statistical testes were two-sided, and P values less than 0.05 were considered to be statistically significant. Data presented are representative of at least three independent experiments. Statistical software, Social Sciences software (SPSS, version 21, Chicago, IL, USA), was used. For in vitro experiments, at least three independent experiments were performed. Detailed statistical calculations were described in relevant figure text.

The effects of tamoxifen on estrogen signaling pathway in PC-3 cells
To study the inhibitory effect of tamoxifen on castration-resistant PCa in vitro and in vivo, and to examine whether it has an on-target inhibition on ERa-mediated cancer growth and survival pathways, we treated castration-resistant PCa cell lines including VCaP and PC-3 cells, respectively, with tamoxifen at various doses. Tamoxifen at 15, 20, 25, and 30 lM significantly reduced proliferation rate of VCaP cells by 69%, 25%, 17%, and 20% of vehicle-treated controls, and decreased proliferation rate of PC-3 cells by 80%, 44%, 19%, and 3% of vehicle-treated controls (for treatment of VCaP cells with 15 lM tamoxifen, p = 0.001, for doses of 20 lM, 25 lM and 30 lM, P < 0.001, Fig. 1A; for treatment of PC-3 cells with 15 lM tamoxifen, P = 0.014, for doses of 20 lM, 25 lM and 30 lM, P < 0.001, Fig. 1B). We decided to use tamoxifen at 20 lM, this concentration was the IC-50 dose as determined for the in vitro analyses. Since ISA-2011B is a selective inhibitor of PIP5K1a, a key regulator that acts upstream of PI3K/AKT to promote proliferation and survival of PCa cells [21], we therefore used ISA-2011B alone or in combination with tamoxifen to assess the effect of tamoxifen on PIP5K1a/PI3K/AKT pathways in PCa cells (Fig. 1C). We observed that tamoxifen treatment led to a dramatic reduction in cell volume and motility of PC-3 cells as determined by using a live cell holomonitor imaging analysis (Fig. 1C). The tamoxifen effect on reducing the proliferative activity of PC-3 cells was similar to what was achieved by ISA-2011B treatment (Fig. 1C), which was reported in our previous studies [21]. Treatment of PC-3 cells with tamoxifen in combination with ISA-2011B led to remarkable reduction of cell numbers and altered cell morphology (Fig. 1C).
To examine whether tamoxifen treatment may induce apoptosis in PC-3 cells, we subjected PC-3 cells that were treated with tamoxifen, ISA-2011B alone or together to flow cytometry-based AnnexinV-7AAD apoptosis assays. Tamoxifen treatment resulted in a significant increase in early apoptosis as represented by the population of Annexin V-positive and 7AAD-negative cells (P < 0.001, Fig. 1D). The rate of late apoptosis and necrosis as represented by the Annexin V-positive and 7AAD-positive populations was also increased in PC-3 cells after Tamoxifen treatment compared with controls (P < 0.001, Fig. 1D). ISA-2011B at 25 µM also led to significant increase in both early and late apoptosis/necrosis compared with the controls (for both early and late apoptosis, P < 0.001, Fig. 1D). Treatment of PC-3 cells with tamoxifen together with ISA-2011B led to a significantly higher proportion of early apoptosis and late apoptosis relative of controls (for both early and late apoptosis, P < 0.001, Fig. 1D).
To investigate whether tamoxifen has on-target effect in PCa cells, and to further identify the new targets of tamoxifen, ISA-2011B, and combination of both agents in PCa cells, we subjected total RNA that were isolated from PC-3 cells treated with tamoxifen (10 µM), ISA-2011B (25µ) or combination of both agents to the next-generation RNA sequencing analysis (RNA-seq). RNA-seq analysis is a highly sensitive method. To avoid unspecific changes resulted from the high rate of apoptosis, we used doses of tamoxifen at 10 µM), ISA-2011B at 25µ in treatment experiments for RNA-seq analysis. Bioinformatics analysis of the transcriptomic data revealed that MMP9 was mostly negatively affected by tamoxifen treatment (P < 0.001, Fig. 1E), while PIP5K1A and CCND1 were uniquely down-regulated, and CDKN1B was upregulated by ISA-2011B (P < 0.001, Fig. 1E). The key marker genes that are involved in proliferation, survival, and invasion of PCa, including AKT1, FLT1, FLK1, CDK1, CCNE1, and CCNB1, were down-regulated by tamoxifen or ISA-2011B alone (P < 0.001, Fig. 1E). Significantly, these key marker genes were down-regulated by tamoxifen and ISA-2011B combined together at higher degrees compared with that of tamoxifen or ISA-2011B alone (P < 0.001, Fig. 1E). Taken together, tamoxifen alone or together with ISA-2011B may exert their effects on PCa cells by specifically targeting a panel of genes that control cancer cell proliferation, survival, and invasion.
Consistent with previous reported studies, ISA-2011B treatment led to significant decrease in expression of both MMP9 and VEGF (for MMP9, P = 0.014; for VEGF, P = 0.026; Fig. 1I). Combination of tamoxifen and ISA-2011B had significantly inhibitory effects on MMP9 expression (P < 0.025, Fig. 1I), and on expression of VEGF (P < 0.031, Fig. 1I). Taken together, the data from immunoblot analysis validated majority of our findings from RNAseq data, and further suggest that tamoxifen and ISA-2011B may have synergistic inhibitory effect on proliferation, survival, and invasive pathways of PCa cells.

Tamoxifen alone or in combination with
PIP5K1a inhibitor ISA-2011B suppresses tumor growth in xenograft mouse models Consistent with the data mentioned above, we found that a final list of 3667 candidate genes that were affected by tamoxifen, and 5712 candidate genes affected by ISA-2011B, and 9074 genes affected by combination of tamoxifen and ISA-2011B ( under PDR cutoff, P < 0.05 Fig. 2A, Table S1). Gene ontology enrichment analysis of biological process further confirmed a robust relationship between candidate gene network and the effects of tamoxifen, ISA-2011B, and combination of both agents ( Fig. 2A). We found that the common gene networks that are responsible for cell cycle, ATP synthesis, metabolisms were significant affected by tamoxifen and ISA-2011B. Interestingly, gene networks that responsible for apoptosis, DNA and protein synthesis appeared to be preferentially affected by tamoxifen, while endoplasmic reticulum (ER) stress network was preferentially affected by ISA-2011B ( Fig. 2B and Table S1). Combination of tamoxifen and ISA-2011B showed synergistic effect on cell cycle, protein synthesis, ATP synthesis, ER stress, and cell death (Fig. 2B). Thus, we identified the unique and common gene network that were targeted by tamoxifen, ISA-2011B and combination of both agents in PCa cells.
To further test the biological and clinical relevance of tamoxifen, combination of tamoxifen and ISA-2011B together for targeted treatment of castration-resistant PCa, we employed xenograft mouse model bearing castration-resistant tumor PC-3 with elevated level of ERa and constitutively high levels of PIP5K1a/pSer-473 Akt. The mice were randomized into four groups and treated accordingly with vehicle control, tamoxifen (40 mg/kg), ISA-2011B (40 mg/kg), or tamoxifen and ISA-2011B at half dose of each via intraperitoneal administration, every second day for a total of 48 days (Fig. 2C). At the end of treatment, the mean volumes of tumors treated with tamoxifen, ISA-2011B or combination of tamoxifen and ISA- Student's t-test was used. *P < 0.05 and **P < 0.01 are indicated 2011B respectively were only 7.9%, 12.5%, and 1% of the mean volume of tumors that were treated with vehicle control (mean volume for vehicle-treated control = 406.8 mm 3 , 95% CI: 323.3-490.3; mean volume for tamoxifen-treated = 32.2 mm 3 , difference = 377.0 mm 3 ; 95% CI: 2.4-62.0 mm 3 , P < 0.001, n = 5; mean volume for ISA-2011B-treated = 51.0 mm 3 , difference = 351.5 mm 3 ; 95% CI: À4.2-106.3 mm 3 , P < 0.001, n = 5; mean volume for combinationtreated = 4.1 mm 3 , difference = 398.8 mm 3 ; 95% CI; 3.9-12.1 mm 3 , P < 0.001, n = 4; Fig. 2C). Tumors from two mice out of the combination treatment group disappeared (Fig. 2C, D), suggesting that tamoxifen in combination with ISA-2011B had pronounced inhibitory effect on tumor growth as compared with that of single agent. Mice were well tolerated to tamoxifen treatment alone or in combination with ISA-2011B.
Immunohistochemical analysis was further performed to examine the effects of tamoxifen, ISA-2011B alone or in combination on the targeted proteins including ERa, PIP5K1a, pSer-473 Akt, VEGF, and VEGFR2 in tumors from each treatment group. ERa expression was significantly decreased in tumors treated with tamoxifen, ISA-2011B or combination of tamoxifen and ISA-2011B, compared with that of controls (P < 0.001, Fig. 2E,F). As expected, PIP5K1a expression was significantly decreased in tumors treated with ISA-2011B (P = 0.016, Fig. 2G,H). Expression of pSer-473 Akt was markedly reduced in tumors treated with tamoxifen, ISA-2011B or combination of tamoxifen and ISA-2011B as compared with that of controls (for tamoxifen-treated P = 0.003; for ISA-2011B-treated P = 0.020; and combination treated, P = 0.03 Fig. 2I,J). Similar to ISA-2011B, tamoxifen treatment reduced expression of VEGF and its receptor VEGFR2 in xenograft tumors, so as the combination treatment (for VEGF, all treatment showed P < 0.001, and for VEGFR2, all treatment showed P < 0.001; Fig S1). Taken together, tamoxifen exerts its effect to suppress growth of PCa tumor in vivo, probably by inhibiting ERa and AKT pathways. Combined treatment using tamoxifen together with ISA-2011B led to regression of tumor growth in mice, probably by synergistically blocking elevated ERa and PIP5K1a/AKT pathways. Taken together, our in vivo data is in agreement with what was obtained from cell line-based studies mentioned above.

ESR1 mRNA expression in distinct subgroup of prostate cancer patients and its association with patient outcome
As reported in our previous studies, we found that clinically relevant gene profiles that are associated with PIP5K1a/AKT in subgroups of patients, and that elevated level of PIP5K1a was associated with poorer patient outcome [16] We therefore examined expression profiles of the target genes of tamoxifen identified by RNA-seq as mentioned above by using primary tumors from PCa patients. To this end, we extracted hierarchical clustering data analyzed by iCluster analysis using primary cancer tissues from 498 PCa patients from The Cancer Genomic Atlas (TCGA), publicly available in cBioPortal databases [33][34][35]. The gene signatures allowed to segregate the primary tumors into four molecularly distinct subclusters (Fig. 3A). Further, ESR1, MMP9, AR, progesterone receptor (PGR), AKT3, CDK1, AKT2, AKT1, VEGFA, and CCND1 were among the genes that exhibited differential expression signatures that divided primary tumors into four distinguished molecular subtypes (Fig. 3A,B,  Fig. S2). Statistical analysis further revealed that genes that were targeted by tamoxifen or ISA-2011B   Box plots show that ESR1, CCND1, MMP9 and AR are significantly differentially expressed among the four subclusters/subgroups. The statistical program ANOVA was used. P values indicate significant differences among the four subclusters/subgroups are shown. (G) Kaplan-Meier survival curve revealed that patients from subcluster/subgroup 1 (n = 183) suffered poorer disease-free survival (DFS) as compared to those from subclusters/subgroups 2-4 (n = 307). Differences in disease-free survivals between two groups were calculated by using SPSS log-rank statistical analysis. P value is indicated, P = 0.018.
including ESR1, CCND1, MMP9, AR showed differential expression in subgroup 1, which was significantly different compared with remaining subgroups (for these mentioned genes, all P values are < 0.001; Fig. 3C-F). This suggests that clinically relevant gene profiles that are targeted by tamoxifen and ISA-2011B are expressed uniquely by subgroups of patients. To examine whether the subgroup 1 with differential expression pattern of ESR1, CCND1, MMP9, AR and other key marker genes might be associated with worse clinical outcome compared with subgroups 2, 3 and 4, we performed Kaplan-Meier survival analysis. We found that patients who had primary tumors belonging to subgroup 1 (n = 183) suffered significantly poorer disease-free survivals (DFS) as compared to those from subgroups 2, 3 and 4 (n = 307), and this difference was statistically significant (P = 0.018) (Fig. 3G). These results indicate that clinically relevant subgroups of PCa patients may have unique gene signatures within the targeted frames of tamoxifen, ISA-2011B or combination of both agents.

ERa protein expression and its correlation with PIP5K1 in primary and metastatic PCa from patient cohorts
We next examined protein expression of the candidate genes as mentioned above by using primary and metastatic lesions from our own patient cohorts. Immunohistochemical analysis revealed that ERa protein was highly expressed in primary tumor tissues and in metastatic lesions from lymph nodes, lung, and bone marrow of PCa patients (Fig. 4A). PIP5K1a, pAKT and AR was also highly expressed in primary tumor tissues and metastatic lesions from PCa patients in these cohorts. Interestingly, we found a statistically significant correlation between ERa and PIP5K1a expression in primary cancer tissues and metastatic lesions (r 2 = 0.513, P < 0.001, Fig. 4B,C). The correlation between the expression of ERa and PIP5K1a in tumor tissues was also illustrated using the correlation scatter plot (Fig. S4). Given that PIP5K1a is a key factor that contributes to progression of PCa, this finding suggests that ERa may be functionally involved in PCa progression and serves as a target for tamoxifen treatment of PCa.

The effect of ERa overexpression on PCa cancer cell lines
We found that ESR1 mRNA was expressed in PC-3 cells by using targeted sequencing analysis (Fig. S3). Next, we investigated whether elevated expression of ERa might be associated with increased growth and survivals of PCa cells. We subjected VCaP cells to 17b-Estradiol treatment at increasing doses ranging from 0.1 nM up to 50 nM for 48 h. The proliferation rate was increased by 63% in VCaP cells treated with 17b-Estradiol at 10 nM relative of controls, and the effect of 17b-Estradiol on VCaP cell proliferation was dose-dependent (P = 0.001, Fig. 5A). Consistent with its positive effect on proliferation, 17b-Estradiol treatment significantly increased cyclin D1 expression by 39% relative of controls (P = 0.019, Fig. 5B,C). This was similar to the effect of DHT, where cyclin D1 expression increased by 123% after DHT treatment, compared with that of controls (P < 0.001, Fig. 5C). Constitutive activation of PIP5K1a/PI3K/Akt pathways are involved in progression and poor response to androgen deprivation therapy in PCa [21,36], we therefore examined effect of 17b-Estradiol treatment on expression of PIP5K1a and phosphorylated Ser-473 AKT. 17b-Estradiol treatment led to a significant increase in PIP5K1a and an increase in phosphorylated Ser-473 AKT, by 194% increase relative of the controls (for PIP5K1a, P = 0.002; for Ser-473 Akt, control mean expression value = 0.12, 95% CI: 0.07-0.17; 17b-Estradiol-treated mean value = 0.33, difference = 0.21, 95% CI 0.30-0.36, P = 0.032, Fig. 5D and e). 17b-Estradiol also markedly increased MMP9 and VEGF expression compared with controls (for VEGF, P = 0.019, Fig. 5F). These data show that 17b-Estradiol treatment is able to increase proliferative ability of PC-3 cells, and leads to upregulation of ERa and its associated cyclin D1, PIP5K1a, phosphorylated AKT, and MMP9/VEGF in PCa cells.
To further study a functional role of ERa as a transcriptional factor in mediating downstream target genes involved in growth, survival and invasion in PCa cells, we induced overexpression of ERa in PC-3 cells that lack AR expression, and treated the cells with 17b-Estradiol to further stimulate the activation of ERa. ERa overexpression alone was sufficient to upregulate cyclin D1 expression in PC-3 cells compared with controls (P < 0.001, Fig. 5G). ERa overexpression also enhanced expression of PIP5K1a in PC-3 cells (Fig. 5G). However, 17b-Estradiol treatment had no significant effect on expression of cyclin D1 and PIP5K1a (Fig. 5G). Since PC-3 cells were cultured in phenol-red free medium containing 10% charcoalstripped serum, in which the steroid hormones, eg, estrogens and androgens were depleted, thus this finding suggests that ERa and its associated pathways may be independent of 17b-Estradiol in PC-3 cells that lack functional AR. Interestingly, immunoprecipitation assays revealed that ERa formed protein-protein complexes with PIP5K1a exclusively in the nuclear compartment of PC-3 cells (Fig. 5H). Immunofluorescence analysis was performed to assess the subcellular localization of ERa and PIP5K1a expression in PC-3 cells that were transfected with ERa or control vectors. We found that ERa was predominantly localized in the nucleus, and ERa expression. PIP5K1a expression was observed in both nucleus and cytoplasmic compartments in PC-3 cells and PIP5K1a expression was markedly enhanced in ERa-transfected cells as compared with that of controls (Fig. 5I). ERa overexpression also led to a significant increase in MMP9 expression, a key marker of invasiveness by 70% relative of controls (control mean expression value = 0.26, 95% CI: 0.11.0-42; mean value in ERa overexpressing cells = 0.45, difference = 0.18, 95% CI: 0.42-0.48, P = 0.036, Fig. 5J).
To elucidate the functional consequences of ERa overexpression in PCa cells, we carried out dual-luciferase (Luc) assays by using control and (ERE) reporter plasmids. ERa overexpression induced ERE reporter luciferase activity by 167.1% as compared with that of controls in PC-3 cells (P = 0.006, Fig. 5K). Thus, ERa is able to mediate the activity of the key factors that contribute to proliferation, survival and invasion in castration-resistant PCa cells.

Depletion of tumor-associated myeloid cells sensitizes tamoxifen effect
It is known that estrogen is converted from androgen by aromatase. Depletion of androgen by androgen deprivation therapy, also leads to depletion of estrogen in PCa patients [37]. However, we previously found that bone marrow-derived myeloid cells produce high level of estrogen are capable of supporting growth of PC-3M cells, thus suggesting that ERa in PCa cells may utilize its ligands from bone marrow-derived cells [11]. To this end, we hypothesized that depletion of tumor-associated bone marrow-derived cells in particular macrophages by Clodronate may sensitize inhibitory effect of tamoxifen. Clodronate is a bisphosphonate that has been shown to improve survival of breast cancer patients [38]. Intriguingly, clodronate was shown to improve overall survival in men with metastatic PCa who were starting hormone therapy [39]. To test our hypothesis, we established aggressive PC-3 xenograft tumors models by allowing subcutaneously implanted tumors to grow into large size with the mean volume of approximately 400 mm 3 in mice. We then used clodronate to selectively deplete tumor-associated macrophages via intratumor injection in combination with intraperitoneal administration of tamoxifen or ISA-2011B in tumor-bearing mice. After two weeks of treatment, the tumors were removed and were made into single cell suspensions. Equal amounts of cells from each treatment group were subjected to tumor spheroid formation assays in 3D models (Fig. 6A). We found that there were significantly less tumor spheroids that were formed from cells extracted from tumors of mice treated with tamoxifen as compared with vehicle controls (P = 0.001, Fig. 6B,C). Similarly, there were also significantly less numbers of tumor spheroids from tumors of ISA-2011B-treated mice group compared with that of controls (P < 0.001, Fig. 6B,C). Interestingly, clodronate treatment alone reduced the ability of tumor cells to form tumor spheroids compared with that of controls (P < 0.001, Fig. 6B,C). Strikingly, there were barely tumor spheroids formation from tumors that were treated with both tamoxifen and clodronate, and this was statistically significant compared with controls (P = 0.003, Fig. 6B,C). In contrast, the numbers of tumor spheroids remained similar between ISA-2011B-treated alone compared with clodronate and ISA-2011B in combination (Fig. 6B,C). These data suggest that tamoxifen may be used together with clodronate to effectively inhibit growth of PCa (Fig. 6D).

Discussion
The castrate-resistant PCa represents an incurable and ultimately fatal stage, with a few treatment options. In this study, we applied novel therapeutic approaches to suppress growth of PCa by using tamoxifen alone or in combination with PIP5K1a inhibitor ISA-2011B to targeting estrogen/ERa and PIP5K1a/Akt pathways in cell line and in mouse models. Pathways analysis of tamoxifen treatment effects have been performed using breast cancer cell lines, and cell cycle and cholesterol biosynthesis were the targets of tamoxifen [40]. Here, we uncovered previously unrealized diversity in gene expression relating to treatment effects of tamoxifen, ISA-2011B or combination of tamoxifen and ISA-2011B on PCa cells. By analyzing gene signatures in PCa cells that have been treated with tamoxifen, ISA-2011B and the combination of both agents, we suggested a potential new approach to treat advanced CRPC.
In the present study, we analyzed multilayer expression signatures of marker genes in primary cancer tissues from 498 PCa patients by using iCluster analysis. We identified four distinguish molecular subtypes of primary tumors based on differential expression signatures such as ERa, AR, and PI3K/Akt signaling pathways. Our findings suggest that many of patients who had elevated expression of tumor-specific signatures such as ESR1, AR, VEGF, AKT1/2/3, CCND1, CDK1, and MMP9 had significantly poorer diseasefree survivals compared with the remaining subgroups.
Estrogen production is increased throughout the body during androgen deprivation therapy (ADT). Since ADT, nonsteroidal anti-androgens will block AR, which in turn through a feedback loop increase the secretion of luteinizing hormone (LH), leading to increased estrogen production, and elevated expression of ERa. Thus, as a direct side-effect, patients suffer gynecomastia or/and breast pain due to ERa expression in breast tissues. Enzalutamide, an androgen receptor signaling inhibitor was shown to be associated with a 49% rate of gynecomastia within 2 years. Tamoxifen has been used as drug intervention to block the ERa in breast tissues and the data from 8 clinical randomized trials revealed that tamoxifen is well tolerated by PCa patients [41].
However, given the facts that ERa overexpression can occur due to the disturbed balance between estrogens and androgens, it remained unexplored regarding to ERa expression, its role and underlying mechanisms in PCa progression. As illustrated by our proposed model in Fig. 6D, we have identified tumor-intrinsic ERa as tumor-promoting factor for PCa growth. Our model suggests that ERa signaling promotes growth by increasing downstream cyclin D1 and pAKT, PIP5K1a, and the effectors of pAKT including CDK1 and P27. We show that ERa and PIP5K1a formed protein complexes in the nuclear compartment, and affected PIP5K1a and phosphorylated AKT in PCa cells. These findings suggest that ERa acts as a transcriptional factor on its target genes such as cyclin D1, but ERa may also be able to mediate function of the proteins such as PIP5K1a and AKT at post-translational level in PCa cells. In the in vivo xenograft mouse models, we used PC-3 cells model, as these cells lack AR, this allowed us to specifically study the ERa-mediated effect in PCa cells. Our findings further revealed the mechanisms suggesting that ERa is a key target that mediates therapeutic response of tamoxifen in PCa. In this study, we observed that PC-3 cells that lack functional androgen receptor (AR) responded to 17b-Estradiol in different ways as compared to that of VCaP cells that express functional AR. Our findings suggest that AR may play an important role in mediating the effects of 17b-Estradiol in PCa cells. It will be interesting to apply VCaP cells to conduct in vivo studies to gain deeper understanding of the underlying mechanisms in the near future.
In the present study, we unrevealed the effect of tamoxifen on panel of genes and network that are associated with ERa and PIP5K1a/AKT that are responsible PCa growth and invasion. In the present study, treatment of xenograft mice bearing PCa tumors with tamoxifen combined with ISA-2011B resulted in regression of tumor growth. A synergistic effect of tamoxifen and ISA-2011B on induction of apoptosis and inhibition of tumor growth was observed. Combination treatment greatly inhibited ERa, PIP5K1a, pAKT, VEGF and VEGFR2 expression in tumors, which reversely correlated with the elevated levels of these key factors in aggressive tumors. The deeper understanding of the effect of therapeutic response to tamoxifen, ISA-2011B and combination of both agents may suggest a new approach in PCa therapy. Our data herein provide the rationale for development of these agents for castration-resistant PCa.
Data from the co-culture experiments in our reported studies have provided evidence suggesting that ERa expression in PCa can be stimulated by estrogen from the bone marrow cells. Thus, bone marrow cells promote growth of ERa-positive PCa cells by providing estrogen [11]. In this study, we depleted tumor-associated macrophages by using intratumor administration of clodronate given to xenograft micebearing aggressive and large tumors, as it is well known that macrophages produce high level of estrogen. We found that depletion of tumor-associated macrophages sensitized the effect of tamoxifen and sufficiently blocked ability of PCa cells to form tumor spheroids. Treatment of tumor-bearing xenograft mice with tamoxifen and clodronate in combination showed greater inhibitory effect as compared to the monotreatment. Our results highlighted an important role of tumor-associated macrophages/tumor microenvironment in PCa growth. This finding also suggests that tamoxifen may be used together with clodronate to effectively inhibit growth and progression of PCa. It will be interesting to systematically identify the target genes of tamoxifen and clodronate by performing RNA-seq in using both cell line systems and tumors in a side-by-side manner in the near future.

Conclusions
In relation to personalized medicine, we integrate preclinical and functional studies with the metagenomics and bioinformatics from clinical studies to better predict the drug and targets-mediated effects. Our study suggests that tamoxifen alone or in combination with ISA-2011B may provide new therapeutic opportunities, and the targets that we discovered and validated can be used as biomarkers to identify patients at risk of metastasis and aid in early stratification of candidates. and ASSK performed data analysis. JS, TW, and JLP contributed to manuscript writing. All authors contributed to writing, editing, and final approval of the manuscript.

Data availability statement
All data analyzed for this study are included in this published article and its supplemental information files. The raw data can be provided upon request to corresponding authors.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig S1. The effect of tamoxifen and ISA-2011B alone or in combination on VEGF and VEGFR2 expression in PC-3 xenograft tumors in mice. Fig S2. Clinical importance and gene expression signatures of estrogen-associated signaling pathways in primary cancer tissues from PCa patients. Fig S3. The presence of ESR1 mRNA in PC-3 cells. Fig S4. The correlation scatter plot. Table S1. Lists of functiona pathways that are significantly affected by tamoxifen-treatment alone, ISA-2011B-treatment alone, or combination treatment using tamoxifen together with ISA-2011B in PC-3 cells.