PI5P4Kα supports prostate cancer metabolism and exposes a survival vulnerability during androgen receptor inhibition

Abstract

Phosphatidylinositol (PI)regulating enzymes are frequently altered in cancer and have become a focus for drug development. Here, we explore the phosphatidylinositol-5-phosphate 4-kinases (PI5P4K), a family of lipid kinases that regulate pools of intracellular PI, and demonstrate that the PI5P4Kα isoform influences androgen receptor (AR) signaling, which supports prostate cancer (PCa) cell survival. The regulation of PI becomes increasingly important in the setting of metabolic stress adaptation of PCa during androgen deprivation (AD), as we show that AD influences PI abundance and enhances intracellular pools of PI-4,5-P₂. We suggest that this PI5P4Kα-AR relationship is mitigated through mTORC1 dysregulation and show that PI5P4Kα colocalizes to the lysosome, the intracellular site of mTORC1 complex activation. Notably, this relationship becomes prominent in mouse prostate tissue following surgical castration. Finally, multiple PCa cell models demonstrate marked survival vulnerability following stable PI5P4Kα inhibition. These results nominate PI5P4Kα as a target to disrupt PCa metabolic adaptation to castrate resistance.

Fig. 1.. PIP4K2A expression correlates with low androgen receptor signaling.

(A) Schematic comparing known kinase activity of noncanonical (left) and canonical (right) phosphoinositide kinase pathways. PI5P4K generates PI-4,5-P₂ from a PI-5-P lipid instead of PI-4-P. PI5P4K functions at internal organelles of the endomembrane system as opposed to the exterior cell membrane–like canonical pathway. The importance of PI3K signaling has been established in prostate cancer (PCa), but the question remains how PI5P4K affects prostate biology and androgen receptor (AR) signaling. (B) Analysis of PIP4K2A transcript expression shows inverse correlation with AR signature score (28) on available transcript data from The Cancer Genome Atlas (PRAD-TCGA; N = 333, P = 4.9 × 10⁻¹⁶) and (C) Stand Up to Cancer datasets (SU2C; N = 149, P = 5.1 × 10⁻⁶). (D) RNA sequencing (RNA-seq) transcript abundance for established PCa cells lines for PIP4K2A, PIP4K2B, and PIP4K2C isoforms. (E) Highest levels of PIP4K2A transcript expression occurs in AR and KLK3 negative models (ARpos mean, 1.127; ARneg mean, 2.941; t test, P = 0.048). (F) Western blot analysis with mouse embryonic fibroblast (MEF) double knockout (negative control) and BT474 (positive control) confirms protein expression of PI5P4Kα in PCa cell lines. (G) PIP4K2A expression further characterized in a panel of human PCa organoids using RNA-seq (R² = 0.2146, P = 4.5 × 10⁻²). t test values: *P < 0.05; **P < 0.01; and ***P < 0.001.

Fig. 2.. PIP4K2A has a functionally inverse relationship with AR in PCa cell models.

(A) Stable overexpression of AR in LNCaP model corresponds to decrease in PI5P4Kα protein by Western blot, which is quantified with (B) protein densitometry (N = 3, 0.48-fold of control, P = 6.0 × 10⁻³). GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (C) Following a 48-hour treatment of AR inhibitor, 10 μM APA does not alter PIP4K2A transcript or (D) PI5P4Kα protein expression (CA, cells alone; NT, nontargeting control). (E) Protein densitometry of AR following siPIP4K2A knockdown in androgen depletion (AD) (n = 3, 2.49-fold of control, P = 2.9 × 10⁻²). (F) AR signature transcript expression is boosted following 48-hour siRNA transfection of LNCaP cells and 18-hour DHT stimulation in AD. (G) Western blot of stable PIP4K2A-targeted LNCaP clones using CRISPR-Cas9 lacking protein expression and (H) shows up-regulated AR relative to wild-type (WT) control clones and mutant clones (PIP4K2A^Cas9mut). Immunoblot quantification values shown following housekeeper and control normalization (relative to bolded sample). Red values highlight experimental change. t test values: *P < 0.05; **P < 0.01.

Fig. 3.. PI5P4Kα is nonredundant to the PI3K pathway.

(A) A relationship has been established between AR and the PI3K (type 1) pathway. Graphical summary of suggested feedback of AR when PI3K/AKT or mTORC1 is inhibited. (B) LNCaP cells in androgen deprivation (AD) were treated with siRNA targeting PIP4K2A (siPIP4K2A) for 48 hours and/or 10 μM BKM120 for 24 hours. Changes in AR transcript expression and AR signature genes (KLK3 and TMPRSS2) by quantitative real-time polymerase chain reaction (qRT-PCR) are represented. (C) Western blot analysis also compares protein-level changes following siPIP4K2A and/or 10 μM BKM120. (D) Signaling changes from siPIP4K2A treatment of LNCaP appear to phenocopy treatment with rapamycin. Phospho-target changes are represented using Western blot comparison of 1 nM rapamycin compared to dimethyl sulfoxide (DMSO) control for 24 hours. Immunoblot quantification values shown following housekeeper and control normalization (relative bolded sample). Red values highlight experimental change. t test values: *P < 0.05.

Fig. 4.. PI5P4Kα affects mTORC1 signaling.

(A) Whole-exome transcript expression was measured with bulk RNA-seq. Statistically significant differential GSEA pathways between nontargeting siRNA control (NT) and siRNA specific for PIP4K2A (siPIP4K2A) treatments are represented. Pathways shown all with P_adj significant P < 0.05. (B) Western blotting confirms changes in AR pathway targets following a knockdown in PI5P4Kα protein expression. (C) Further analysis of patient data from TCGA (PRAD-TCGA) shows a positive association with PIP4K2A (R = 0.17, P = 9.8 × 10⁻⁵) and an mTOR activation score (17). In addition, PIP4K2B (R = 0.27, P = 2 × 10⁻⁹) but not PIP4K2C (R = 0.0044, P = 0.92). (D) LNCaP cells infected with pLKO.1 lentiviral PIP4K2A shRNAs (shPIP4K2A) show reduced phospho-S6^235/236 48 hours after GFP sorting. Immunoblot quantification values shown following housekeeper and control normalization (relative bolded sample). Red values highlight experimental change. (E) LNCaP cells in androgen deprivation (AD) were treated with siPIP4K2A for 48 hours. Changes of AR transcript expression and TFEB signature genes (CTSD, NEU1, HEXA, and CTSF) by qRT-PCR are represented. Data representative of three independent experiments, mean ± SEM. t test values: *P < 0.05; **P < 0.01; and ***P < 0.001.

Fig. 5.. AR inhibition dysregulates PI signaling at the endolysosome.

(A) LysoTracker Red IF staining of LNCaP cells treated with 10 μM APA or DMSO for 48 hours. Scale bars, 10 μm. (B) Relative quantification of LysoTracker Red IF intensity in cell populations following 48 hours androgen deprivation (AD) or APA treatment. Data representative of three independent experiments. (C) LNCaP cultured in complete RPMI 1640 + 5% FBS compared to AD conditions. Total PI lipids were measured in LNCaP cells transfected with siRNA targeting PIP4K2A (siPIP4K2A) for 24 hours and then switched into AD conditions for an additional 48 hours. Relative PI lipid content was measured using lipidomic liquid chromatography–tandem mass spectrometry (LC-MS/MS) methodology. CA, cell alone in complete medium unless indicated with AD; NT, nontargeting siRNA control. Increases between CA and CA in AD (0.32-fold, P = 6.04 × 10⁻⁹) and NT and siPIP4K2A (0.15-fold, P = 3. × 10⁻⁴) are noted. (D) qRT-PCR shows suppression of AR target gene KLK3 in paired experiments where (E) intracellular pools of PI-4,5-P₂ are detected with IF between LNCaP treated for 48 hours of 10 μM APA or DMSO control. Scale bars, 10 μm. (F) Image quantification of 50 cells per group showed elevated intensity of PI-4,5-P₂ pool staining intensity in APA treatment. DAPI, 4′,6-diamidino-2-phenylindole. t test values: *P < 0.05; **P < 0.01; and ***P < 0.001.

Fig. 6.. Lysosome localized PI5P4Kα is linked to castration adaptation of prostate tissue.

(A) C57BL/6J mice were aged to 8 weeks, surgically castrated or used for control sham surgery procedure. Tissues were harvested 10 weeks after surgery. (B) PI5P4Kα detection was optimized and characterized in normal adult mouse prostate. Examples of positive 3,3′-diaminobenzidine (DAB) staining by immunohistochemistry are shown (scale bars, 20 μm) and (C) quantified between sham and castrated tissue groups as a percentage of positive DAB cells per number of cells counted per gland (P = 5.18 × 10⁻⁹). (D) Representative images of normal mouse prostate compared to castrated tissue when stained with LAMP1 endolysosome marker with IF-IHC. Scale bars, 10 μm. (E) Quantification of Lamp1 detection in castrated versus sham surgery mouse prostate tissue (Welch’s t test, P = 0.0193). DAPI nuclear stains are used as positive cell detection marks for IF assays. (F) IF staining of normal adult mouse prostate for colocalization of LAMP1 (green) and PI5P4Kα (red). Yellow indicates areas of colocalization. Scale bars, 10 μm. See fig. S7 for Zeiss software colocalization statistics. DAPI nuclear stains are used as positive cell detection marks for IF assays. t test values: *P < 0.05; and ***P < 0.001.

Fig. 7.. PI5P4Kα affects prostate gland regression during androgen deprivation.

(A) Prostate-specific GEM model with R26eYFP reporter gene. PB-driven Cre recombinase system is used to delete Pip4k2a^fl/fl alleles at time of mouse sexual maturity. An eYFP fluorophore is permanently “turned on” from Cre recombinase activation. WT, wild type. (B) Graphical summary of in vivo androgen deprivation experiment using mouse castration. PB-Cre⁺;R26eYFP control and PB-Cre⁺;R26eYFP;Pip4k2a^fl/fl mice were aged to sexual maturity to activate Cre recombinase system, which can be detected with eYFP Cre reporter. Male mice were castrated at 8 weeks of age and harvested 10 weeks following surgery. (C) Representative images of GFP/eYFP detection with IF-IHC staining of castrated tissues. PB-expressing luminal cells show diffuse cytoplasmic GFP/eYFP staining of glandular epithelium. Scale bars, 100 μm. (D) Quantification of gland area of eYFP⁺ glands following surgical castration. Data are represented as median with Mann-Whitney test for significance (P = 0.0002). DAPI nuclear stains are used as positive-cell detection marks for IF assays. (E) Castrated UG mass relative to total mouse body weight shows an average of 0.26% for eYFP controls (PB-Cre⁺;R26eYFP; n = 9) compared to 0.32% for Pip4k2a^fl/fl animals (PB-Cre⁺;R26eYFP Pip4k2a^fl/fl; n = 8, one outlier, t test, P = 0.0004). t test values: **P < 0.01; and ***P < 0.001.

Fig. 8.. Pip4k2a is elevated in GEM cancer models, and deletion reduces prostate cell metabolic stress survival.

(A) Graphical representation of PB-Cre⁺;R26eYFP;Pip4k2a^fl/fl prostate luminal cells with and without Cre recombinase activation and eYFP reporter expression in PB-expressing cells. (B) Light phase microscopy image of PB-Pip4k2a^fl/fl organoids cells in 2D culture demonstrates vacuole morphologic features. Relative to eYFP⁺ control cells. Scale bars, 400 μm. (C) qRT-PCR gene expression shows that Pip4k2a ^fl/fl cells have less Pip4k2a transcript (0.23-fold, P = 1.32 ×10⁻⁶) and have an increase in Ar (1.83-fold, P = 0.002). (D) PB-Pip4k2a^fl/fl cells were treated with 1, 2, 5, or 10 μM of 2DG or H₂O control for 72 hours. Differences in cell confluency are represented compared to eYFP⁺ only control cells from three independent experiments. (E to F) Transcript expression was measured from mouse anterior prostate from animals with normal prostates, genotypes, and histology indicating early progression of prostatic intraepithelial neoplasia (PIN) and tumors from PB-Cre⁺; Pten^fl/fl; Trp53^fl/fl prostates. Ar signature genes, Nkx3.1 (0.01 tumor, P = 2.55 ×10⁻⁸) and Fkbp5 (7.2-fold in PIN, P = 0.01; 5.4-fold in tumor, P = 6.19 × 10⁻⁵) increase with early tumor development but are lost in advanced-stage disease. Pip4k2a is elevated in prostate tumors (32-fold in tumor, P = 2.52 ×10⁻⁷). t test values: *P < 0.05; **P < 0.01; and ***P < 0.001.

Fig. 9.. Acute depletion of PIP4K2A exposes survival vulnerability in models of PCa.

Human PCa cell lines are impaired in proliferation when infected with stable PIP4K2A-targeted shRNA (shPIP4K2A) construct relative to control pLKO.1 lentivirus. PI5P4Kα knockdown confirmed by Western blots and cell proliferation changes measured by changes in confluence over time shown for (A) LNCaP, (B) C4-2, (C) 22RV1, and (D) DU145, which represents a model of castrate-resistant PCa (CRPC). (E) Changes in fold growth over time shown for the 3D Neuroendocrine PCa (NEPC) organoid line PM154. t test values: *P < 0.05 . Data are represented as means ± SD. (F) Graphical representation of the proposed mechanism. PI5P4Kα is an important regulator of metabolic organelles from the endolysosome system, which regulates the nutrient sensing and metabolic stress pathway driven by mTORC1. In PCa, AR and mTORC1 are established to have a dynamic interplay that is critical in prostate cell growth and survival. (i and ii) We show that the depletion of PIP4K2 increases PCa cell stress and results in the up-regulation of AR and suggest that this is mitigated via the down-regulation of mTORC1. (iii) Patient data suggests an increase in PIP4K2A in AR-independent PCa cases (CRPC). PI5P4Kα may become critical in maintaining CRPC growth by promoting mTORC1 and thereby exposes a vulnerability in the cancer metabolism.