Abstract
Genetic alterations that potentiate PI3K signaling are frequent in prostate cancer, yet how different genetic drivers of the PI3K cascade contribute to prostate cancer is unclear. Here, we report PIK3CA mutation/amplification correlates with poor survival of patients with prostate cancer. To interrogate the requirement of different PI3K genetic drivers in prostate cancer, we employed a genetic approach to mutate Pik3ca in mouse prostate epithelium. We show Pik3caH1047R mutation causes p110α-dependent invasive prostate carcinoma in vivo. Furthermore, we report that PIK3CA mutation and PTEN loss coexist in patients with prostate cancer and can cooperate in vivo to accelerate disease progression via AKT–mTORC1/2 hyperactivation. Contrasting single mutants that slowly acquire castration-resistant prostate cancer (CRPC), concomitant Pik3ca mutation and Pten loss caused de novo CRPC. Thus, Pik3ca mutation and Pten deletion are not functionally redundant. Our findings indicate that PIK3CA mutation is an attractive prognostic indicator for prostate cancer that may cooperate with PTEN loss to facilitate CRPC in patients.
Significance: We show PIK3CA mutation correlates with poor prostate cancer prognosis and causes prostate cancer in mice. Moreover, PIK3CA mutation and PTEN loss coexist in prostate cancer and can cooperate in vivo to accelerate tumorigenesis and facilitate CRPC. Delineating this synergistic relationship may present new therapeutic/prognostic approaches to overcome castration/PI3K–AKT–mTORC1/2 inhibitor resistance. Cancer Discov; 8(6); 764–79. ©2018 AACR.
See related commentary by Triscott and Rubin, p. 682.
This article is highlighted in the In This Issue feature, p. 663
Introduction
Prostate cancer is the second most common cause of male cancer-related death worldwide, emphasizing the failure of mainstay therapeutic regimens to treat advanced disease (1). A pivotal constraint for prostate cancer research is the lack of diverse in vivo models that accurately reflect the clinic. Expanding the range of prostate cancer models that display key clinicopathologic characteristics is vital to (i) delineate the complex molecular mechanisms underpinning prostate cancer, (ii) identify novel prognostic markers and therapeutic targets, and (iii) accurately establish the efficacy of novel therapies that are urgently needed in the clinic.
Class 1A PI3Ks are heterodimers consisting of a regulatory subunit encoded by PIK3R1 (p85α/p55α/p50α), PIK3R2 (p85β), or PIK3R3 (p85γ), and a catalytic subunit encoded by PIK3CA (p110α), PIK3CB (p110β), or PIK3CD (p110δ; refs. 2, 3). Upon activation of receptor tyrosine kinases (RTK), G-protein–coupled receptors (GPCR), or RAS, PI3K is recruited to the membrane. Here, PI3K catalyzes the generation of the second messenger phosphatidylinositol(3,4,5)trisphosphate (PIP3), which recruits the serine threonine kinases AKT [protein kinase B (PKB)] and PDK1 to the membrane, as well as a plethora of other PIP3 binding proteins (4). The first identified, and most well studied, PIP3 effector is AKT (4). PDK1 phosphorylates AKT at Thr308, leading to phosphorylation of downstream targets, including TSC2 that activates mTORC1 to promote proliferation, survival, and migration (2, 3). Phosphorylation of AKT at both Ser473 and Thr308 is required for the full activation of AKT and has been linked to coactivation of the mTORC1 and mTORC2 pathways (5). Key substrates of mTORC2 include AKT at Ser473 (5), and SGK1, which phosphorylates and inactivates the metastasis suppressor N-Myc–downregulated gene-1 (NDRG1/DRG1/CAP43; ref. 6). AKT dephosphorylation by PHLPP at Ser473 and PP2A at Thr308 deactivate AKT (5). The fine-tuning of AKT phosphorylation levels mediates AKT pathway activity and subsequent cellular events. In addition, the tumor suppressor PTEN serves to negatively regulate the PI3K cascade by catalyzing the dephosphorylation of PIP3 to PIP2. PTEN also mediates lipid phosphatase–independent tumor-suppressor activities via its protein phosphatase domain (2).
PI3K pathway hyperactivation is invariably associated with prostate cancer progression in the clinic, thus presenting an attractive therapeutic target. Indeed, loss of PTEN, a negative regulator of the PI3K pathway, is estimated to occur in 40% to 50% of patients with prostate cancer (7, 8). However, PI3K pathway hyperactivation can occur via a range of mechanisms (e.g., PIK3CA oncogenic mutation) that can independently influence downstream signaling events. We sought to determine if genetic drivers of the PI3K pathway that are present in prostate cancer, but have not been investigated previously, can also contribute to prostate cancer initiation/progression. To this end, we generated a new, clinically relevant genetically modified mouse model harboring a heterozygous activating mutation in Pik3ca specifically within prostate epithelial cells, and compared prostate histopathology with the well-characterized Pten-deleted mouse model of prostate cancer (9). Overall, our findings emphasize the prognostic value of PI3K genetic drivers to better inform personalized therapy design.
Results
PIK3CA Mutation/Amplification Correlates with Advanced Prostate Cancer Progression
PIK3CA oncogenic mutation and amplification, which may increase p110α PI3K catalytic activity, are frequently detected in human cancers (10–13). To better understand the frequency of PIK3CA alterations in prostate cancer, we analyzed nine prostate cancer genomic datasets for PIK3CA mutations and gene amplification (14). Our analysis shows that PIK3CA mutations occur in up to 4% of patients with prostate cancer, whereas PIK3CA copy-number gain/amplification occurs in as many as 62% of cases (Fig. 1A; Supplementary Table S1). PIK3CA mutations were predominantly nucleotide missense substitutions (87.5%; Supplementary Fig. S1A) within the helical (44.1%) and kinase (20.6%) domains, and previously reported hotspot mutations in exons 9 (E542K, E545K) and 20 (H1047R; refs. 10–12, 15) were most frequent (Fig. 1B). Notably, the majority of PIK3CA mutations observed (83%, 19/23) have been previously detected in prostate or nonprostate malignancies (12, 13, 15, 16), and are reported to increase p110α activity (refs. 12, 15–18; Fig. 1B).
PIK3CA somatic mutation and amplification frequency in prostate cancer. A, Histogram displaying PIK3CA mutation and copy-number amplification/gain frequency across nine prostate cancer genomic datasets. Pie charts show the distribution of primary and metastatic samples for each genomic dataset. B, Schematic diagram of p110α illustrating mutation frequency across the nine prostate cancer datasets analyzed in relation to the core functional domains. Codons with frequent missense mutations at common hotspots are labeled magenta. To our knowledge, the four genetic alterations in light gray text have not been previously reported in other human malignancies, and their impact on p110α function is unknown. p85, PI3K p85 regulatory subunit binding domain; C2, calcium-dependent phospholipid-binding domain; Helical, PI3K helical domain; Kinase, PI3/4-kinase domain; aa, amino acid. C, Kaplan–Meier plot comparing the survival probability of PIK3CA mutation and/or amplification/gain carriers with PIK3CA unaltered patients within the TCGA provisional prostate cancer patient dataset. PIK3CA age-adjusted COXPH HR: 0.55.
PIK3CA somatic mutation and amplification frequency in prostate cancer. A, Histogram displaying PIK3CA mutation and copy-number amplification/gain frequency across nine prostate cancer genomic datasets. Pie charts show the distribution of primary and metastatic samples for each genomic dataset. B, Schematic diagram of p110α illustrating mutation frequency across the nine prostate cancer datasets analyzed in relation to the core functional domains. Codons with frequent missense mutations at common hotspots are labeled magenta. To our knowledge, the four genetic alterations in light gray text have not been previously reported in other human malignancies, and their impact on p110α function is unknown. p85, PI3K p85 regulatory subunit binding domain; C2, calcium-dependent phospholipid-binding domain; Helical, PI3K helical domain; Kinase, PI3/4-kinase domain; aa, amino acid. C, Kaplan–Meier plot comparing the survival probability of PIK3CA mutation and/or amplification/gain carriers with PIK3CA unaltered patients within the TCGA provisional prostate cancer patient dataset. PIK3CA age-adjusted COXPH HR: 0.55.
To determine if PIK3CA mutation/amplification correlates with prostate cancer progression, we analyzed The Cancer Genome Atlas (TCGA) provisional prostate cancer dataset (Supplementary Table S2). Our analysis revealed that PIK3CA mutation and copy-number gain/amplification frequency significantly correlates with poor prostate cancer survival, regional lymph node metastasis, and a higher primary tumor category and Gleason grade (Fig. 1C; Supplementary Table S3), resembling PTEN loss (Supplementary Fig. S1B; Supplementary Table S4).
Coexpression analysis within the 9 prostate cancer datasets analyzed revealed that 39.4% (13/33 patients) of PIK3CA mutation carriers also harbored PTEN mutation or copy-number loss, indicating that patients with PIK3CA-mutant prostate cancer have a high frequency of coexistent PTEN deleterious genetic alterations, consistent with ovarian, breast, endometrial, and colorectal cancer studies (11, 19, 20). Interestingly, 47.5% (96/202) of patients with prostate cancer with PIK3CA amplification/gain also carried a PTEN mutation or copy-number loss. Moreover, statistical analysis of the larger TCGA provisional dataset revealed a significant tendency toward PIK3CA mutation/amplification/gain and PTEN mutation/loss co-occurrence in patients with prostate cancer (P < 0.001, Fisher exact test, Log OR = 0.916). Together, these findings suggest that PIK3CA mutation/amplification/gain play an oncogenic role during prostate cancer and indicate that p110α gain of function and PTEN loss may cooperate to promote prostate cancer growth.
Pik3caH1047R Mutation in Mouse Prostate Epithelium Causes Locally Invasive Prostate Carcinoma
To delineate the oncogenic role of a clinically relevant PIK3CA mutation within the prostate, we intercrossed mice that harbor a conditional latent H1047R mutation in Pik3ca to the PBiCre transgenic line. Using an exon-switch Cre-LoxP approach, expression of Pik3caH1047R was driven specifically within the prostate following PBiCre-mediated excision of the floxed wild-type (WT) Pik3ca exon 20 and subsequent expression of a latent downstream mutant exon 20 (11, 21). Recombination was confirmed by sequencing and allele-specific PCR analysis of cDNA isolated from the prostate glands of PBiCre+/−;Pik3ca+/+ and PBiCre+/−;Pik3ca+/Lat-H1047R mice, hereafter referred to as WT and Pik3ca+/HR, respectively (Supplementary Fig. S2A and S2B). Histologic analysis of WT and Pik3ca+/HR prostate lobes revealed no gross phenotype in WT mice, whereas Pik3ca+/HR cohorts displayed a progressive malignant phenotype. Pik3ca+/HR mice developed multifocal simple and/or cribriform hyperplasia in all prostate lobes by 100 days and homogeneous, locally invasive prostate carcinoma by 300 to 400 days (dorsolateral lobe, Fig. 2A and B; ventral/anterior lobes, Supplementary Fig. S2C–S2E and Supplementary Table S5). Invasion was confirmed by the absence of smooth muscle actin (SMA) staining by immunohistochemistry (IHC; Supplementary Fig. S2F). Pik3ca+/HR prostate carcinomas were predominantly dysplastic/mucinous, and reactive stroma and immune infiltrate were evident (Fig. 2A; 300–400 days; Supplementary Fig. S2C). Taken together, these data demonstrate that heterozygous Pik3caH1047R oncogenic mutation is sufficient to cause invasive prostate cancer in mice and to our knowledge is the first example of a monoallelic mutation driving invasive prostate cancer growth in vivo.
Heterozygous Pik3caH1047R oncogenic mutation causes invasive prostate cancer in mice that does not phenocopy Pten deletion. A, Representative hematoxylin/eosin images of WT, Pik3ca+/HR, and Ptenfl/fl dorsolateral prostate epithelium (scale bar, 100 μm). B, Histogram displaying phenotype incidence in WT, Pik3ca+/HR, and Ptenfl/fl dorsolateral prostate. DLP, dorsolateral prostate. C, Bar graph displaying total prostate weight normalized to body weight for WT, Pik3ca+/HR, and Ptenfl/fl mice. n = as indicated (N). Error bars, SEM, *, P < 0.05 compared with WT, one-way ANOVA with Tukey multiple comparison test.
Heterozygous Pik3caH1047R oncogenic mutation causes invasive prostate cancer in mice that does not phenocopy Pten deletion. A, Representative hematoxylin/eosin images of WT, Pik3ca+/HR, and Ptenfl/fl dorsolateral prostate epithelium (scale bar, 100 μm). B, Histogram displaying phenotype incidence in WT, Pik3ca+/HR, and Ptenfl/fl dorsolateral prostate. DLP, dorsolateral prostate. C, Bar graph displaying total prostate weight normalized to body weight for WT, Pik3ca+/HR, and Ptenfl/fl mice. n = as indicated (N). Error bars, SEM, *, P < 0.05 compared with WT, one-way ANOVA with Tukey multiple comparison test.
To determine if the genetic driver of the PI3K cascade influences prostate tumorigenesis and/or malignant progression, we compared our novel Pik3ca+/HR prostate cancer model with the well-characterized Pten-deleted mouse model of prostate cancer (9). To this end, we generated age-matched cohorts of PBiCre+/−;Ptenfl/fl mice (denoted Ptenfl/fl) deficient for both copies of Pten within prostate epithelial cells and compared the phenotype on the same genetic background. In contrast to Pik3ca+/HR mice, we observed early onset of hyperplastic lesions at 56 days and rapid tumor progression from prostate intraepithelial neoplasia (PIN) to locally invasive carcinoma in Ptenfl/fl mice by 200 days (Fig. 2A and B; Supplementary Fig. S2C–S2E; Supplementary Table S5). Furthermore, Ptenfl/flprostate tumor burden was significantly greater and more heterogeneous than the Pik3ca+/HR model (Fig. 2A and C), as carcinosarcomas were also present by 300 days in 29% of the cohort (2/7). However, metastasis to the liver, lung, lymph nodes, or bone was not detected in either model. Ptenfl/fl mice were also prone to seminal vesicle neoplasia, urethra neoplasia, and adrenal pheochromocytoma that were rare, or absent, in Pik3ca+/HR mice (Supplementary Table S5). These findings indicate that relative to the Pik3ca+/HR model, early disease onset and potentially accelerated progression contribute to the earlier emergence of invasive carcinoma in Ptenfl/fl mice.
To investigate if Pten biallelic loss accelerates prostate tumor growth compared with Pik3ca+/HR-driven prostate cancer, we performed and quantitated IHC to detect the S-phase proliferation marker PCNA. Pik3ca+/HR prostate disease (100–400 days) showed a significant increase in the number of PCNA-positive proliferating cells compared with WT controls; however, a clear proliferation advantage was evident in Ptenfl/fl mice between 56 and 200 days (Fig. 3A and B). Furthermore, significantly more PCNA-positive cells were detected in Ptenfl/fl prostate hyperplasia compared with Pik3ca+/HR hyperplastic lesions (Supplementary Fig. S2G), indicating this is an early phenomenon during disease progression. These data show that increased proliferation in Ptenfl/fl mice facilitates accelerated prostate cancer progression compared with Pik3ca+/HR mice. Notably, apoptosis evasion is not likely to contribute to accelerated disease progression in Ptenfl/fl mice, as the number of cleaved caspase-3 (CC3)–positive apoptotic cells was largely increased in Ptenfl/fl prostate epithelium compared with age-matched Pik3ca+/HR mutants (Supplementary Fig. S3A and S3B). IHC analysis to detect prostate basal and luminal cell markers, cytokeratin-5 (CK5) and cytokeratin-8 (CK8), respectively, revealed that, unlike Ptenfl/fl prostate tumors, Pik3ca+/HR-driven tumors are predominantly comprised of luminal epithelial cells and rarely display expansion/mislocalization of the CK5-positive basal cell population (Supplementary Fig. S3C). Taken together, these data show that although a single Pik3caH1047R activating mutation predisposes to murine prostate cancer, like biallelic loss of Pten, these two genetic drivers of the PI3K cascade do not completely phenocopy. Overall, these findings suggest that Pik3ca oncogenic mutation and Pten loss may drive prostate tumor phenotypes via distinct molecular mechanisms, which could present novel therapeutic targets.
Pten deletion triggers mTORC2 signaling to facilitate rapid prostate cancer progression relative to Pik3caH1047R mutation. A, IHC to detect the proliferation marker PCNA in WT, Pik3ca+/HR, and Ptenfl/fl prostate carcinoma at 400 days of age. Scale bars, 50 μm. B, Quantitation of PCNA-positive nuclei in WT, Pik3ca+/HR, and Ptenfl/fl prostate epithelium (n = 3, *, P < 0.05 compared with WT, or as indicated, one-way ANOVA with Tukey multiple comparison test; error bars, SEM). C, Representative IHC images to detect PTEN, mTORC1 signaling components (pAKT Thr308, pRPS6 Ser235/236, and p4E-BP1 Thr37/46), and mTORC2 substrates (pAKT Ser473 and pNDRG1 Thr346) in WT dorsolateral prostate and Pik3ca+/HR and Ptenfl/fl prostate carcinoma at 400 days of age (n = 3; scale bar, 50 μm; inset scale bar, 10 μm). IHC quantitation for (D) pAKT Thr308, (E) pRPS6 Ser235/236, (F) p4E-BP1 Thr37/46, (G) pAKT Ser473, and (H) pNDRG1 Thr346 in WT dorsolateral prostate and Pik3ca+/HR and Ptenfl/fl prostate carcinoma at 400 days of age (n = 3; error bars, SEM; *, P < 0.05 compared with WT, or as indicated, one-way ANOVA with Tukey multiple comparison correction).
Pten deletion triggers mTORC2 signaling to facilitate rapid prostate cancer progression relative to Pik3caH1047R mutation. A, IHC to detect the proliferation marker PCNA in WT, Pik3ca+/HR, and Ptenfl/fl prostate carcinoma at 400 days of age. Scale bars, 50 μm. B, Quantitation of PCNA-positive nuclei in WT, Pik3ca+/HR, and Ptenfl/fl prostate epithelium (n = 3, *, P < 0.05 compared with WT, or as indicated, one-way ANOVA with Tukey multiple comparison test; error bars, SEM). C, Representative IHC images to detect PTEN, mTORC1 signaling components (pAKT Thr308, pRPS6 Ser235/236, and p4E-BP1 Thr37/46), and mTORC2 substrates (pAKT Ser473 and pNDRG1 Thr346) in WT dorsolateral prostate and Pik3ca+/HR and Ptenfl/fl prostate carcinoma at 400 days of age (n = 3; scale bar, 50 μm; inset scale bar, 10 μm). IHC quantitation for (D) pAKT Thr308, (E) pRPS6 Ser235/236, (F) p4E-BP1 Thr37/46, (G) pAKT Ser473, and (H) pNDRG1 Thr346 in WT dorsolateral prostate and Pik3ca+/HR and Ptenfl/fl prostate carcinoma at 400 days of age (n = 3; error bars, SEM; *, P < 0.05 compared with WT, or as indicated, one-way ANOVA with Tukey multiple comparison correction).
Pik3ca Mutation and Pten Loss Stimulate mTORC1 Signaling to Promote Prostate Tumorigenesis and Facilitate Malignant Progression in Mice
Given that both Pten deletion and Pik3ca oncogenic mutations can activate PI3K signaling, the phenotypic differences observed between the genetic drivers could reflect differential activation of the PI3K cascade. To determine if PTEN tumor-suppressive function is maintained in the Pik3ca+/HR model, we performed IHC to detect PTEN. We observed uniform membranous PTEN staining in WT prostate epithelium and Pik3ca+/HR tumors, whereas PTEN-positive cells were absent in Ptenfl/fl prostate tumors, consistent with biallelic Pten ablation (Fig. 3C). To determine if PTEN impairs effector cascades downstream of PI3K in the Pik3ca+/HR model to delay prostate tumorigenesis and progression, we evaluated AKT activation and the status of the PI3K downstream mTOR effector cascades. mTOR is a serine/threonine kinase that forms part of two distinct complexes, mTORC1 and mTORC2. We show that the number of prostate epithelial cells displaying activation of AKT via Thr308 phosphorylation at the cell membrane is significantly increased in Pik3ca+/HR and Ptenfl/fl prostate carcinoma compared with WT controls (Fig. 3C and D), indicating that both PI3K genetic drivers stimulate mTORC1 signaling to promote tumor growth. In support, the proportion of cells displaying phosphorylation of well-known downstream mTORC1 targets, namely the ribosomal protein S6 (RPS6) at Ser235/236 that regulates cell size and proliferation and 4e-binding protein 1 (4E-BP1) at Thr37/46 that mediates translational machinery, were also significantly elevated in both models (Fig. 3C, E–F). Our analysis of the number of pAKT Thr308-, pRPS6-, and p4E-BP1–positive cells in Ptenfl/fl and Pik3ca+/HR hyperplastic lesions also revealed that mTORC1 signaling upregulation occurs premalignancy (Supplementary Fig. S3D–S3G). Of note, the number of pRPS6- and p4E-BP1–positive cells was comparable in Pik3ca-mutated and Pten-deleted hyperplastic lesions and advanced tumors, despite significantly more pAKT Thr308-positive cells being detected in the Ptenfl/fl model, suggesting that pAKT Thr308–independent phosphorylation of RPS6 and 4E-BP1 may occur, or that the partial increase in pAKT Thr308 in Pik3ca+/HR tumors is sufficient to sustain pRPS6 and p4E-BP1 signaling. Taken together, these data indicate that both Pik3caH1047R oncogenic mutation and Pten biallelic loss stimulate mTORC1 signaling to facilitate prostate tumor formation and progression, and that PTEN-mediated tumor-suppressive functions do not impair mTORC1 downstream signaling in the context of Pik3ca mutation.
Relative to Pik3ca Mutation, Pten Deletion Augments mTORC2 Signaling to Further Promote Prostate Tumor Formation and Progression in Mice
Because activation of the mTORC1 downstream targets was comparable between Pik3ca+/HR and Ptenfl/fl prostate cancer models, we reasoned that the early onset and accelerated progression observed in the Ptenfl/fl model may be attributable to mTORC2 signaling. To investigate this, we performed and quantitated IHC in Pik3ca+/HR and Ptenfl/fl prostate carcinomas to detect the phosphorylation of two mTORC2 targets: AKT at Ser473 and NDRG1 at Thr346. We show that the number of pAKT Ser473– and pNDRG1 Thr346–positive cells was significantly increased in Ptenfl/fl prostate tumors compared with the Pik3ca+/HR model and WT controls (Fig. 3C, G–H). Similar results were observed when comparing hyperplastic lesions (Supplementary Fig. S3D, S3H, and S3I). Thus, mTORC2 signaling presents a direct mechanism whereby Pten homozygous deletion can promote tumor onset/progression relative to Pik3caH1047R oncogenic mutation in this setting.
Pik3caH1047R Mutation Causes p110α-Dependent Prostate Cancer
Recent reports have demonstrated that p110α and p110β isoforms of the PI3K catalytic subunit play distinct cellular functions and are regulated independently by differential binding partners (22–25). For instance, in vitro assays have established that RAS subfamily members can directly bind to the RAS binding domain (RBD) of p110α (and not p110β) to activate p110α kinase activity, and p110β RBD:RAC1 interactions have been shown to be required for GPCR-mediated p110β signaling (22–24). Moreover, Pten-deleted prostate cancers are considered to preferentially activate the p110β isoform, and p110β blockade has been shown to activate p110α owing to relief of feedback inhibition (e.g., via IGF1R; refs. 25–28). Thus, we sought to determine if the phenotypic difference between Pten loss and Pik3ca oncogenic mutation reflects differential activation of PI3K catalytic isoforms. To this end, we performed IHC to detect pERK, a downstream target of the RAS cascade, and activation of RAC1 GTPase in Pik3ca+/HR and Ptenfl/fl prostate tumors to distinguish activation of RAS–p110α and RAC1–p110β signaling axes, respectively. We find that pERK expression is markedly elevated in Pik3ca+/HR and Ptenfl/fl prostate tumors compared with age-matched WT controls, indicating p110α signaling is activated in both models (Fig. 4A). However, only Ptenfl/fl prostate tumors displayed Active RAC1–GTP staining (Fig. 4B), indicating that activation of p110β signaling may promote prostate cancer growth induced by Pten deletion.
Pik3ca+/HR prostate cancer is p110α-dependent, whereas Ptenfl/fl prostate cancer is p110α and p110β codependent. Representative IHC images to detect (A) pERK Thr202/Tyr204 and (B) Active RAC1-GTP in WT dorsolateral prostate and Pik3ca+/HR and Ptenfl/fl prostate carcinoma at 400 days of age (n = 3; low-magnification scale bar, 100 μm; high-magnification scale bar, 10 μm). Bar graph indicating total prostate weight normalized to body weight for Pik3ca+/HR mice (C) and Ptenfl/fl mice (D) with prostate carcinoma administered with either vehicle, p110α-specific inhibitor (A66), p110β-specific inhibitor (TGX-221), pan-PI3K inhibitor (BKM120), or A66 + TGX-221 for 4 weeks compared with age-matched WT controls. n = as indicated (N). Error bars, SEM; *, P < 0.05 compared with vehicle, one-way ANOVA with Tukey multiple comparison correction; ns, not significant. E, Histogram displaying phenotype incidence for dorsolateral prostate from Pik3ca+/HR and Ptenfl/fl mice treated with either vehicle, p110α-specific inhibitor (A66), p110β-specific inhibitor (TGX-221), pan-PI3K inhibitor (BKM120), or A66 + TGX-221 for 4 weeks.
Pik3ca+/HR prostate cancer is p110α-dependent, whereas Ptenfl/fl prostate cancer is p110α and p110β codependent. Representative IHC images to detect (A) pERK Thr202/Tyr204 and (B) Active RAC1-GTP in WT dorsolateral prostate and Pik3ca+/HR and Ptenfl/fl prostate carcinoma at 400 days of age (n = 3; low-magnification scale bar, 100 μm; high-magnification scale bar, 10 μm). Bar graph indicating total prostate weight normalized to body weight for Pik3ca+/HR mice (C) and Ptenfl/fl mice (D) with prostate carcinoma administered with either vehicle, p110α-specific inhibitor (A66), p110β-specific inhibitor (TGX-221), pan-PI3K inhibitor (BKM120), or A66 + TGX-221 for 4 weeks compared with age-matched WT controls. n = as indicated (N). Error bars, SEM; *, P < 0.05 compared with vehicle, one-way ANOVA with Tukey multiple comparison correction; ns, not significant. E, Histogram displaying phenotype incidence for dorsolateral prostate from Pik3ca+/HR and Ptenfl/fl mice treated with either vehicle, p110α-specific inhibitor (A66), p110β-specific inhibitor (TGX-221), pan-PI3K inhibitor (BKM120), or A66 + TGX-221 for 4 weeks.
To directly test p110α and p110β isoform dependency in Pik3ca-mutant and Pten-deleted prostate cancers, we administered isoform-specific inhibitors (A66, a p110α-specific inhibitor, or TGX-221, a p110β-specific inhibitor) or a pan-PI3K inhibitor (BKM120) to cohorts of Pik3ca+/HR and Ptenfl/fl mice with prostate carcinoma for 4 weeks. Pik3ca+/HR tumor burden regressed significantly in response to A66 and BKM120, whereas TGX-221 had no effect, indicative of p110α dependency (Fig. 4C). In contrast, Ptenfl/fl tumor burden was not reduced upon single isoform-specific inhibitor treatment but did respond to BKM120 or combined A66 and TGX-221 therapy, suggesting Pten-deleted tumors are p110β/p110α codependent (Fig. 4D). Histopathologic analysis of prostate lobes confirmed tumor regression inA66- and BKM120-treated Pik3ca+/HR mice and BKM120- and A66+TGX-221-treated Ptenfl/fl mice (Fig. 4E; Supplementary Fig. S4A–S4C). These data suggest that p110β-mediated signaling events could facilitate Pten-deleted prostate cancer but not Pik3caH1047R-mutated prostate cancer, and support previous work showing that combined p110α and p110β blockade improves therapeutic outcome in PTEN-deficient prostate cancers compared with PI3K isoform–specific monotherapy (22, 26, 28). Indeed, PI3K pathway inhibitors on their own have been shown to have limited efficacy in the clinic due to multiple feedback loops, PI3K-independent pathways, and/or additional oncogenic mutations, and can cause side effects (e.g., hyperglycemia; refs. 22, 26, 28, 29). Thus, treatment approaches that combine PI3K pathway inhibitors with other therapeutic agents are currently being explored to improve outcomes for patients with prostate cancer.
Pten Deletion and Pik3ca+/HR Mutation Cooperate to Accelerate Prostate Cancer Progression in Mice
Given that we have found that PIK3CA mutation and PTEN loss are not mutually exclusive events in patients with prostate cancer, we sought to generate a new clinically relevant model of prostate cancer and to test if PI3K genetic drivers can cooperate to facilitate prostate cancer growth. Hence, we crossed Pik3ca+/HRmutants with Ptenfl/fl animals to develop PBiCre+/−;Pik3ca+/HR;Ptenfl/fl compound mutants (termed Pik3ca+/HR;Ptenfl/fl) that harbor Pik3ca+/HR mutation and biallelic Pten loss in prostate epithelial cells. At 56 and 100 days, we observed aggressive, locally invasive carcinoma with 100% incidence in all Pik3ca+/HR;Ptenfl/fl prostate lobes (Fig. 5A and B; Supplementary Fig. S5A and S5B; Supplementary Table S5). IHC analysis revealed that Pik3ca+/HR;Ptenfl/fl prostate tumors resemble Ptenfl/fl tumors, where the CK5+ basal cell population is expanded/mislocalized and the CK8+ luminal cells are predominant (Supplementary Fig. S5C). Local invasion was confirmed by the absence of SMA staining (Supplementary Fig. S5C). Tumor burden was also significantly greater in compound mutants than age-matched single mutants (Supplementary Fig. S5D). Visceral metastases were not detected by 100 days of age, and the development of nonprostate malignancies reflecting leaky PBiCre-mediated recombination (predominantly benign buccal mucosal/cutaneous papillomas and penile prolapse) prevented further aging of Pik3ca+/HR;Ptenfl/fl mice.
Pik3ca mutation and Pten loss cooperate to accelerate prostate cancer progression in mice by upregulating proliferation and mTORC1/2 signaling. A, Representative IHC images of Pik3ca+/HR;Ptenfl/fl prostate carcinoma at 56 and 100 days of age. Scale bars, 100 μm. B, Phenotype incidence histogram for WT, Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl dorsolateral prostate at 56 and 100 days of age. C, IHC to detect PCNA in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas. Scale bars, 50 μm. IHC quantitation for (D) PCNA, (E) pAKT Thr308, (F) pRPS6 Ser235/236, (G) p4E-BP1 Thr37/46, (H) pAKT Ser473, and (I) pNDRG1 Thr346 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas (n = 3; *, P < 0.05 compared with Pik3ca+/HR or as indicated, one-way ANOVA with Tukey correction; error bars, SEM). J, RNA in situ hybridization (ISH) analysis of Pik3ca and Pik3cb transcripts in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas (n = 3; scale bars, 50 μm; inset scale bars, 5 μm). Quantitation of (K) Pik3ca and (L) Pik3cb mRNA molecules detected by in situ hybridization in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas (n = 3; *, P < 0.05 compared with WT, one-way ANOVA with Tukey correction; error bars, SEM).
Pik3ca mutation and Pten loss cooperate to accelerate prostate cancer progression in mice by upregulating proliferation and mTORC1/2 signaling. A, Representative IHC images of Pik3ca+/HR;Ptenfl/fl prostate carcinoma at 56 and 100 days of age. Scale bars, 100 μm. B, Phenotype incidence histogram for WT, Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl dorsolateral prostate at 56 and 100 days of age. C, IHC to detect PCNA in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas. Scale bars, 50 μm. IHC quantitation for (D) PCNA, (E) pAKT Thr308, (F) pRPS6 Ser235/236, (G) p4E-BP1 Thr37/46, (H) pAKT Ser473, and (I) pNDRG1 Thr346 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas (n = 3; *, P < 0.05 compared with Pik3ca+/HR or as indicated, one-way ANOVA with Tukey correction; error bars, SEM). J, RNA in situ hybridization (ISH) analysis of Pik3ca and Pik3cb transcripts in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas (n = 3; scale bars, 50 μm; inset scale bars, 5 μm). Quantitation of (K) Pik3ca and (L) Pik3cb mRNA molecules detected by in situ hybridization in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl stage-matched prostate carcinomas (n = 3; *, P < 0.05 compared with WT, one-way ANOVA with Tukey correction; error bars, SEM).
To investigate the mechanism underpinning cooperation between Pik3ca mutation and Pten loss, we determined the number of proliferative and apoptotic cells in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl locally invasive prostate carcinomas by PCNA and CC3 IHC, respectively. We show that compound-mutant tumors have significantly more PCNA-positive proliferating cells than single mutants (Fig. 5C and D), whereas CC3-mediated apoptosis is unaltered (Supplementary Fig. S5E and S5F). These findings indicate that Pik3ca oncogenic mutation and Pten loss synergize to accelerate prostate cancer progression by increasing proliferation, but not survival.
To ascertain if the increased proliferation in Pik3ca+/HR;Ptenfl/fl mice reflects further activation of mTORC1/2 signaling, we performed IHC to detect pAKT Thr308 that leads to mTORC1 activation, as well as the phosphorylation of known mTORC1/2 downstream signaling targets. Quantitation of IHC staining revealed that the number of cells expressing membranous pAKT Thr308 is significantly increased in Pik3ca+/HR;Ptenfl/fl prostate carcinomas compared with stage-matched single mutants (Fig. 5E; Supplementary Fig. S5G). In accordance, the mTORC1 downstream targets pRPS6 and p4E-BP1 positively correlated with pAKT Thr308 activation (Fig. 5F and G; Supplementary Fig. S5G), indicating increased mTORC1 signaling accelerates prostate cancer growth in Pik3ca+/HR;Ptenfl/fl mutants. Phosphorylation of the mTORC2 downstream targets pAKT Ser473 and pNDRG1 was also significantly increased in compound mutants compared with single mutants (Fig. 5H and I; Supplementary Fig. S5G). Taken together, these findings suggest that further potentiation of mTORC1 and mTORC2 signaling, which correlates with superactivation of AKT at Thr308/Ser473, contributes to the cooperative relationship between Pik3ca mutation and Pten loss during prostate cancer in this setting.
Previous work has shown that amplification/overexpression of Pik3ca and Pik3cb increases oncogenicity (22, 30, 31), and amplification frequently correlates with poor patient outcome in multiple malignancies (22). To establish if Pik3ca/b transcripts are expressed at physiologic levels in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl prostate carcinomas, we performed RNA in situ hybridization and quantitated RNA molecules relative to WT controls. We show that Pik3ca and Pik3cb mRNA is significantly increased in double-mutant prostate tumors compared with Ptenfl/fl and Pik3ca+/HR single-mutant tumors and age-matched WT controls (Fig. 5J–L; Supplementary Fig. S5H). However, the functional consequences of Pik3ca and Pik3cb mRNA upregulation remain elusive.
Pik3ca+/HR- and Ptenfl/fl-Induced Prostate Cancers Acquire CRPC in Mice
PTEN loss is widely reported to correlate with resistance to androgen deprivation therapy in patients with prostate cancer and mice (9, 32, 33). To examine if Pik3ca+/HR-driven prostate cancer also confers castration-resistant disease, we aged cohorts of Pik3ca+/HR-mutant mice until invasive prostate carcinoma had developed (300 days) and assessed the early and long-term response to surgical castration. Although we observed a significant reduction in Pik3ca+/HR total prostate weight at 2 and 10 weeks after castration (Fig. 6A), histopathologic analysis revealed that prostate tumors were still present in Pik3ca+/HR mice, indicating the development of acquired castration-resistant prostate cancer (CRPC; Fig. 6B; Supplementary Fig. S6A). These findings are in keeping with partial androgen sensitivity and the latent acquisition of CRPC, mirroring homozygous deletion of Pten (Fig. 6A and B; Supplementary Fig. S6A), as previously reported (33, 34). IHC to detect androgen receptor (AR) confirmed the reduction of androgens after castration, as cytoplasmic AR was detected in Pik3ca+/HR and Ptenfl/fl prostate epithelial cells following castration, whereas uncastrated controls displayed active nuclear AR (Supplementary Fig. S6B).
Pik3ca+/HR and Ptenfl/fl prostate cancers acquire CRPC, whereas Pik3ca+/HR;Ptenfl/fl compound mutants display innate resistance to castration. A, Bar graph displaying total prostate weight normalized to body weight for Pik3ca+/HR and Ptenfl/fl mice 2 and 10 weeks after castration relative to age-matched/uncastrated controls. n = as indicated (N); error bars, SEM; *, P < 0.05, one-way ANOVA with Tukey correction. B, Representative hematoxylin/eosin (H&E) images of Pik3ca+/HR- and Ptenfl/fl-uncastrated dorsolateral prostate, and 2 and 10 weeks after castration. Scale bar, 100 μm. C, Representative H&E images of Pik3ca+/HR;Ptenfl/fl-uncastrated dorsolateral prostate and 2 weeks after castration. Scale bar, 100 μm; n = 5. D, Bar graph displaying total prostate weight normalized to body weight for Pik3ca+/HR:Ptenfl/fl mice 2 weeks after castration relative to age-matched/uncastrated controls (error bars, SEM; P = 0.3394, unpaired, two-tailed t test, n = 5). IHC quantitation for (E) PCNA and (F) CC3 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl dorsolateral prostate 2 weeks after castration compared with uncastrated/age-matched controls. Error bars, SEM; *, P < 0.05, one-way ANOVA with Tukey correction, n = 3. Mice were castrated when prostate carcinoma was prevalent; Pik3ca+/HR = 400 days old; Ptenfl/fl = 200 days old; and Pik3ca+/HR;Ptenfl/fl = 100 days old.
Pik3ca+/HR and Ptenfl/fl prostate cancers acquire CRPC, whereas Pik3ca+/HR;Ptenfl/fl compound mutants display innate resistance to castration. A, Bar graph displaying total prostate weight normalized to body weight for Pik3ca+/HR and Ptenfl/fl mice 2 and 10 weeks after castration relative to age-matched/uncastrated controls. n = as indicated (N); error bars, SEM; *, P < 0.05, one-way ANOVA with Tukey correction. B, Representative hematoxylin/eosin (H&E) images of Pik3ca+/HR- and Ptenfl/fl-uncastrated dorsolateral prostate, and 2 and 10 weeks after castration. Scale bar, 100 μm. C, Representative H&E images of Pik3ca+/HR;Ptenfl/fl-uncastrated dorsolateral prostate and 2 weeks after castration. Scale bar, 100 μm; n = 5. D, Bar graph displaying total prostate weight normalized to body weight for Pik3ca+/HR:Ptenfl/fl mice 2 weeks after castration relative to age-matched/uncastrated controls (error bars, SEM; P = 0.3394, unpaired, two-tailed t test, n = 5). IHC quantitation for (E) PCNA and (F) CC3 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl dorsolateral prostate 2 weeks after castration compared with uncastrated/age-matched controls. Error bars, SEM; *, P < 0.05, one-way ANOVA with Tukey correction, n = 3. Mice were castrated when prostate carcinoma was prevalent; Pik3ca+/HR = 400 days old; Ptenfl/fl = 200 days old; and Pik3ca+/HR;Ptenfl/fl = 100 days old.
To determine if Pik3ca heterozygous oncogenic mutation sensitizes preneoplastic prostate epithelium to CRPC transition, we examined the short-term and long-term response of Pik3ca+/HR mice to castration at 100 days, when only hyperplastic disease is present. Prostate epithelial regression was detected 2 weeks after castration and correlated with a reduction in prostate weight, yet small prostate hyperplastic and dysplastic tumors, resembling uncastrated Pik3ca+/HR mutants, had developed by 42 weeks after castration in 100% (6/6) and 67% (4/6) of cases, respectively (Supplementary Fig. S6C and S6D). These data demonstrate that Pik3ca-mutated prostate epithelium possesses an inherent ability to acquire CRPC, similarly to Pten loss (9).
Pik3ca Oncogenic Mutation and Pten Loss Synergize, Predisposing to De Novo CRPC
Next, we castrated Pik3ca+/HR;Ptenfl/fl mice at 100 days of age when invasive carcinoma was present, to test if Pik3ca mutation and Pten loss can also cooperate to promote CRPC growth. At 2 weeks after castration, castrated compound mutants phenocopied intact controls, and no appreciable difference in tumor burden was detected (Fig. 6C and D). These findings contrast the partial regression observed in the single mutants and indicate that Pik3ca oncogenic mutation and Pten homozygous deletion cooperate to promote de novo CRPC in vivo. In support, single mutants displayed a significant reduction in the percentage of PCNA-positive proliferative cells and elevated CC3-positive apoptotic cells 2 weeks after castration, which were unaltered in compound mutants (Fig. 6E and F; Supplementary Fig. S7A and S7B). These data indicate that de novo CRPC in Pik3ca+/HR;Ptenfl/fl mice is attributable to both the sustained level of proliferation and castration-induced apoptosis evasion. In accordance with de novo CRPC, IHC to detect AR also revealed that a noticeable proportion of Pik3ca+/HR;Ptenfl/fl prostate epithelial cells displayed AR activation (i.e., nuclear translocation) 2 weeks after castration, which was not apparent in single mutants at this early time point (Supplementary Fig. S6B). Of note, Nkx3.1 and Pbsn, AR transcriptional target genes, are significantly reduced in Pik3ca+/HR prostate carcinomas relative to WT prostate, and levels were further diminished in Ptenfl/fl and Pik3ca+/HR;Ptenfl/fl tumors (Supplementary Fig. S7C and S7D). These findings support previous work indicating that PI3K activation perturbs AR-mediated signaling (29) and indicate that Pbsn and Nkx3.1 transcription is not likely to facilitate de novo CRPC in this setting.
The molecular mechanisms underpinning the emergence of CRPC are largely unknown. Pten-deleted CRPC acquisition has been previously associated with elevated AKT signaling, suggesting that further activation of the AKT cascade contributes to CRPC transition (9). In support, we observed a significant increase in the percentage of cells positive for mTORC1 signaling components pAKT (Thr308), pRPS6 (Ser235/236), and p4E-BP1 (Thr37/46) and the mTORC2 target pAKT (Ser473) in both the Pik3ca+/HR and Ptenfl/fl models just 2 weeks after castration (Fig. 7A–D). Notably, phosphorylation of NDRG1 was not altered in either model after castration (Fig. 7E). Thus, Pten-deleted and Pik3ca-mutated prostate epithelial cells appear to hyperactivate AKT upon castration, which elevates mTORC1 signaling downstream targets to facilitate CRPC transition. Nevertheless, we do not exclude the possibility that additional molecular events may also contribute to CRPC transition in these models, including PTEN and/or AKT signal transduction independent of PI3K (2, 35).
De novo CRPC in Pik3ca+/HR;Ptenfl/fl double transgenic animals correlates with NDRG1 inactivation. Quantitation of IHC to detect mTORC1 signaling components (A) pAKT Thr308, (B) pRPS6 Ser235/236, (C) p4E-BP1 Thr37/46, (D) pAKT Ser473, and (E) pNDRG1 Thr346 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl prostate tissue 2 weeks after castration compared with uncastrated, age-matched controls (n = 3; error bars, SEM; *, P < 0.05, one-way ANOVA with Tukey multiple comparison correction). F, Representative IHC images of pNDRG1 Thr346 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/flprostate tissue 2 weeks after castration compared with uncastrated, age-matched controls. Scale bars, 50 μm; n = 3. G, RPPA analysis was performed on lysates from Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl prostate tissue 2 weeks after castration, and compared with uncastrated (UC), age-matched controls. Heat map represents Log2-normalized and median-centered data (means of duplicates, n = 3 per cohort). Mice were castrated when prostate carcinoma was prevalent; Pik3ca+/HR = 400 days old; Ptenfl/fl = 200 days old; and Pik3ca+/HR;Ptenfl/fl = 100 days old.
De novo CRPC in Pik3ca+/HR;Ptenfl/fl double transgenic animals correlates with NDRG1 inactivation. Quantitation of IHC to detect mTORC1 signaling components (A) pAKT Thr308, (B) pRPS6 Ser235/236, (C) p4E-BP1 Thr37/46, (D) pAKT Ser473, and (E) pNDRG1 Thr346 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl prostate tissue 2 weeks after castration compared with uncastrated, age-matched controls (n = 3; error bars, SEM; *, P < 0.05, one-way ANOVA with Tukey multiple comparison correction). F, Representative IHC images of pNDRG1 Thr346 in Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/flprostate tissue 2 weeks after castration compared with uncastrated, age-matched controls. Scale bars, 50 μm; n = 3. G, RPPA analysis was performed on lysates from Pik3ca+/HR, Ptenfl/fl, and Pik3ca+/HR;Ptenfl/fl prostate tissue 2 weeks after castration, and compared with uncastrated (UC), age-matched controls. Heat map represents Log2-normalized and median-centered data (means of duplicates, n = 3 per cohort). Mice were castrated when prostate carcinoma was prevalent; Pik3ca+/HR = 400 days old; Ptenfl/fl = 200 days old; and Pik3ca+/HR;Ptenfl/fl = 100 days old.
Our analysis of Pik3ca+/HR;Ptenfl/fl prostate tumors before and after castration revealed that the high proportion of pAKT (Thr308)-, pRPS6 (Ser235/236)-, and pAKT (Ser473)-positive cells is maintained at a superactivated state, and that the percentage of p4E-BP1 (Thr37/46) and pNDRG1 (Thr346) cells is increased even further (Fig. 7A–E). Despite an increase in p4E-BP1 in Ptenfl/fl- and Pik3ca+/HR;Ptenfl/fl-castrated tumors compared with Pik3ca+/HR-castrated animals, p4E-BP1 was not significantly elevated in compound mutants compared with the Ptenfl/fl model, signifying 4E-BP1 phosphorylation at Thr37/46 and subsequent inactivation are not likely to promote de novo CRPC formation. However, our findings suggest that NDRG1 inactivation may contribute to de novo CRPC. In addition to increased phosphorylation of NDRG1 after castration, androgen deprivation in Pik3ca+/HR;Ptenfl/fl mice also positively correlated with pNDRG1 nuclear localization (Fig. 7F). Nevertheless, the precise role of NDRG1 inactivation during de novo CRPC remains to be determined and warrants further investigation. Taken together, these results infer that a high threshold of AKT hyperactivation prior to castration, and/or NDRG1 inactivation, may prove to be useful biomarkers of intrinsic CRPC in the clinic.
To explore potential mechanisms underpinning the synergistic relationship between Pik3ca mutation and Pten deletion, and during castration-resistant disease formation, we performed a reverse-phase protein array (RPPA) on protein lysates isolated from Pik3ca+/HR, Ptenfl/fl and Pik3ca+/HR;Ptenfl/fl stage-matched uncastrated prostate carcinomas, and 2 weeks after castration (Fig. 7G; Supplementary Tables S6 and S7). RPPA data analysis revealed that Pik3ca-mutated and Pten-deleted prostate tumors display distinct RPPA profiles, supporting the contention that Pik3ca oncogenic mutation and Pten loss may mediate distinct signaling events to facilitate prostate cancer growth. For instance, compared with Pik3ca+/HR tumors, Ptenfl/fl tumors displayed enhanced signal intensities for PI3K cascade phosphoproteins (e.g., pAKT Thr308, pAKT Ser473, pFOXO3A Ser318/321, pGSK-3β Ser9, and pNDRG1 Thr346), whereas tyrosine kinase–mediated (pEGFR Tyr1173, pSHP-2 Tyr542, pSRC family Tyr416) and MAPK (pERK1/2 Thr202/Tyr204) phosphoproteins were elevated in Pik3ca+/HR tumors (Fig. 7G; Supplementary Table S7). Interestingly, Pik3ca+/HR and Ptenfl/fl prostate carcinomas did not display significant differences in RPPA signal intensities for the senescence markers p21 or p27 (Fig. 7G; Supplementary Table S7), suggesting that the observed phenotypic differences are not due to changes in senescence. A significant increase in p53 signal intensity was observed in Ptenfl/fl tumors relative to Pik3ca+/HR tumors. As p21 and p27 are unaltered, these findings indicate that the observed changes in p53 are not regulating senescence, but may instead be mediating other cellular functions, such as apoptosis. This correlates with our observations in CC3 and phosphorylated p53 (Ser15; Fig. 7G; Supplementary Fig. S3B; Supplementary Table S7). RPPA results for pAKT Thr308 and pAKT Ser473 were confirmed by western blotting (Supplementary Fig. S7E).
RPPA profiles for uncastrated and castrated compound mutants were strikingly similar. Indeed, only four targets were significantly altered: pSHP2 (Tyr542) and pSRC family (Tyr416) signals were increased, and YAP and CK2α signals were decreased (Fig. 7G; Supplementary Table S7). This result contrasts single mutants that acquired CRPC and is consistent with the general lack of effect of castration on the prostate tumors in the double-mutant mice (Fig. 6C–F). However, it should be noted that the RPPA was not sensitive enough to detect elevated pNDRG1 (Thr346) in Pik3ca+/HR;Ptenfl/fl mutants after castration that was observed by IHC (Fig. 7E and F), presumably owing to tumor heterogeneity and/or stromal content, which may also be a contributing factor in the lack of significant difference in other proteins as well. Nevertheless, distinct differences between the Pik3ca+/HR and Ptenfl/fl models were detected after castration (Fig. 7G; Supplementary Table S7). For instance, Pik3ca-mutated tumors displayed a significant increase in IGF1Rβ and pNF-kB p65 (Ser 536) signal intensities after castration, which were not altered in Pten-deleted tumors after castration. In addition, castrated Pik3ca+/HR tumors displayed elevated JAK/STAT and MAPK signaling relative to castrated Ptenfl/fl tumors. Taken together, our findings suggest that Pik3ca oncogenic mutation and Pten loss may mediate distinct signaling events to facilitate prostate cancer growth and resistance to castration.
Discussion
We report that PIK3CA mutation/amplification positively correlates with poor prostate cancer patient prognosis and overall survival. Our findings are the first to demonstrate that the PIK3CAH1047R oncogenic mutation is sufficient to cause invasive prostate cancer in vivo and that concomitant loss of PTEN and PIK3CA mutation, which frequently occurs in the clinic, can cooperate to accelerate prostate cancer growth in mice.
Our data support the hypothesis that different genetic drivers of the PI3K cascade are not functionally redundant, but instead drive prostate tumorigenesis via distinct signaling events. We show that relative to p110α-dependent Pik3ca+/HR-induced prostate cancers, Ptenfl/fl prostate tumors are p110α/β codependent and exhibit accelerated tumor formation and progression owing to AKT hyperactivation, elevated mTORC2, and RAC1–p110β signaling. The failure to induce robust AKT signaling in Pik3ca+/HR epithelium is probably attributable to the maintenance of PTEN tumor-suppressive function that reduces PIP3 levels, AKT membrane recruitment, and subsequent activation of AKT, as previously reported (35). In corroboration, Pten loss has been shown to positively correlate with disease progression in mice, as Pten LOH is required for prostate cancer growth in Pten heterozygous prostate epithelium (9). We speculate that PTEN function is also likely to be conserved in transgenic mice expressing myristoylated/activated AKT or p110β in prostate epithelial cells, as only low-grade prostate epithelial neoplasia develops that does not progress to carcinoma with ageing (36, 37). Taken together, these observations suggest that additional mTORC2/RAC1/p110β-independent cooperative events are likely to facilitate malignant progression to an invasive state in Pik3ca+/HR mutants that express PTEN. Indeed, PIK3CA mutations have been shown to potentiate a PDK1–SGK3, AKT-independent signaling axis in various human cancer cell lines that express PTEN (35), and PDK1–SGK1, AKT-independent signaling has been shown to cause resistance to p110α inhibition by directly phosphorylating TSC2 to activate the mTORC1 pathway (38).
Guertin and colleagues have previously shown that RICTOR, a key regulatory component of mTORC2, is required for PC3 PTEN-null human prostate cancer cells to form tumor xenografts and that biallelic deletion of Rictor prevents prostate cancer formation driven by Pten loss in mice by reducing proliferation and AKT phosphorylation at Ser473 (39). We show that Ptenfl/fl mice displayed early prostate tumor formation and accelerated progression relative to Pik3ca+/HR mutants, reflecting elevated mTORC2 signaling and subsequent AKT phosphorylation at Ser473 in the context of Pten loss. Thus, our findings support the notion that mTORC2 signaling plays a critical role during prostate tumorigenesis and progression, and strengthen the rationale for mTORC2-targeted therapy in PTEN-deleted prostate cancer.
The absence of pAKT Ser473 phosphorylation in Pik3ca+/HR-mutant prostate cancer may be attributable to reduced PIP3 levels and/or distinct AKT regulation in Pten-null and Pik3ca-mutant prostate cancers, as AKT phosphorylation is dependent on a plethora of AKT protein kinases and phosphatases (5). Of note, the mechanism of AKT regulation may also depend upon the type of PIK3CA mutation (i.e., helical vs. kinase) and tissue context, as several human cancer cell lines with PIK3CA kinase mutations have been shown to express high levels of pAKT Ser473 and Thr308 in the presence of PTEN (35). PTEN is also reported to play a broader AKT-independent tumor-suppressive role via protein- and lipid-phosphatase activities to mediate p53, cell-cycle arrest and integrin, insulin and focal adhesion kinase signaling, reviewed in ref. 40. Thus, developing our combined understanding of AKT regulation, p110 PI3K isoform signaling, and PTEN mode of action during prostate cancer is vital to determine optimal therapeutic approaches that inhibit the PI3K signaling network and subsequently prostate cancer growth and progression.
Although p110α and p110β isoforms have been shown to form mutually exclusive signaling complexes with RAS and RHO family (RAC1/CDC42) small GTPase protein superfamily members, respectively (22), the molecular mechanisms underpinning their different modes of action are poorly understood. This study provides additional data that underline a distinct role for the RAC1–p110β signaling axis in Pten-deleted prostate cancer, and raise the possibility that RAC inhibition may show therapeutic efficacy against PTEN-deleted prostate cancer in the clinic, as recently demonstrated for a Pten-null, p110β-dependent mouse model of myeloid neoplasia (41). By taking this approach, PI3K-independent functions of PTEN and AKT may be advantageously cotargeted, as RAC1 activation is mediated by PI3K-dependent (e.g., PREX1/TIAM/mTORC2) and PI3K-independent (e.g., SRC/p130CAS) signaling (25, 42).
We have generated a new clinically relevant transgenic mouse model of advanced prostate cancer driven by concomitant Pik3ca heterozygous oncogenic mutation and Pten homozygous deletion. We show that these two oncogenic drivers cooperate to promote rapid progression to invasive prostate cancer, characterized by the synergistic elevation of mTORC1/2 signaling, AKT superactivation, and increased Pik3ca/b mRNA transcript expression. These data provide direct evidence that Pik3ca mutation and Pten deletion coordinate independent oncogenic signaling events during prostate cancer, in corroboration with the distinct RPPA profiles observed. Furthermore, our findings emphasize that the coexistence of mutated PIK3CA and PTEN loss may prove to be an important prognostic indicator for rapid prostate cancer progression and de novo resistance to androgen deprivation therapy in the clinic.
Currently, the cause and consequence of upregulated Pik3ca/b transcription are poorly understood. Theoretically, increased Pik3ca/b gene expression could promote prostate cancer progression in Pik3ca+/HR;Ptenfl/fl mice by increasing p110α/β protein levels, and thus total PI3K activity, as PIK3CA amplification is thought to do in ovarian cancer cells (43). FOXO3A, NF-kB, YB1, and p53 have been shown to promote PIK3CA transcription (reviewed in ref. 44); however, PIK3CB transcriptional regulators remain to be identified. Further investigation is needed to determine the underlying mechanism by which increased p110 catalytic activity and loss of PTEN phosphatase activity cooperate to upregulate Pik3ca and Pik3cb transcription, and to establish the functional significance of this observation.
Despite numerous phenotypic differences, we report that both Pten-null and Pik3ca+/HR-driven prostate cancers are partially sensitive to androgen withdrawal and acquire CRPC in association with augmented PI3K signaling. These data signify that both p110α and p110β PI3K catalytic isoforms can induce PI3K signaling in response to androgen deprivation, supporting previous in vitro work in the PTEN-deficient human prostate cancer LNCaP cell line that showed PI3K signaling induced by an AR inhibitor is diminished by p110α or p110β inhibition (26). Because AKT hyperactivation and augmented mTORC1/2 are consistent features of intact Pik3ca+/HR;Ptenfl/fl prostate tumors, it is tempting to speculate that AKT hyperactivation may be a prerequisite for de novo CRPC in this setting. In addition, innate CRPC in Pik3ca+/HR;Ptenfl/fl compound mutants was associated with increased phosphorylation of the mTORC2–SGK substrate NDRG1 at Thr346, suggesting that NDRG1 inactivation may facilitate de novo CRPC. Significantly, NDRG1 has been shown to function as a metastasis suppressor in mouse xenograft models of prostate cancer by reducing ATF3 transcription, and NDRG1 mRNA downregulation correlates with Gleason score and worse prostate cancer survival (45). Stein and colleagues have also reported that NDRG1 is a p53 transcriptional target that is required for p53-mediated apoptosis (46). Given that Pik3ca+/HR; Ptenfl/fl prostate tumors evade castration-induced apoptosis and proliferation arrest, it will be important for future studies to determine if NDRG1 inactivation contributes to CRPC transition. However, because oncogenic PI3K/AKT signaling has been linked to increased genomic instability (47), we do not exclude the possibility that Pik3ca+/HR;Ptenfl/fl prostate tumors create an environment capable of inducing additional oncogenic mutations that promote CRPC formation.
It is becoming clear that approaches inhibiting multiple targets within the PI3K network, either simultaneously or sequentially, are necessary to enhance therapeutic efficacy. Thus, further characterization of p110α/β-mediated signaling, PI3K-independent PTEN tumor-suppressive functions, AKT-independent signaling, and AKT regulation is required to improve our understanding of how to target the PI3K network and identify mechanisms of therapeutic resistance to improve our management of prostate cancer in the clinic. Future work addressing how to personalize treatment for tumors driven by diverse PI3K genetic drivers is paramount and is likely to entail the coinhibition of PI3K-dependent and PI3K/AKT-independent signaling pathways.
Methods
Experimental Animals
PBiCre transgenic mice that express Cre recombinase under the control of the Probasin promoter and Ptenfl/fl mice have been described previously (21, 48). Pik3caH1047R-mutant mice were generated in-house (11). All mice were maintained on a pure FVB/NJ background. Mice were genotyped from DNA isolated from toe biopsies, as described previously (11, 49). Age-matched males were randomly assigned to uncastrated/castrated cohorts. Castration experiments involved the surgical removal of the testis and epididymis. Animal experiments followed the National Health and Medical Research Council (NHMRC) Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and were approved by the Animal Experimentation Ethics Committee at Peter MacCallum Cancer Centre.
Tissue Isolation and Histology
Tissue was harvested and fixed for 16 to 24 hours in 10% neutral-buffered formaldehyde at 4°C before being paraffin embedded and sectioned at 4 μm. Sections were stained with hematoxylin/eosin for histologic analysis by a certified pathologist (P. Waring) blinded to genotype/treatment. Defining characteristics for prostate disease were based upon the pathologic classification of mouse prostate disease outlined in ref. 50.
Immunohistochemistry
Staining was carried out as described previously (49) on formalin-fixed, paraffin-embedded (FFPE) sections. Primary antibodies: Active RAC1–GTP 1:800 (#26903; NewEast Biosciences), AR 1:300 (#sc-816; Santa Cruz Biotechnology), PCNA 1:400 (#610665; BD Biosciences Pharmingen), and Cell Signaling Technology antibodies: Cleaved caspase-3 1:300 (#9664), pAKT (Ser473) 1:400 (#4060), pAKT (Thr308) 1:400 (#13038), pERK (Thr202/Tyr204) 1:200 (#4376), PTEN 1:300 (#9559), pRPS6 (Ser235/236) 1:400 (#2211), pNDRG1 (Thr346) 1:800 (#5482), and p4E-BP1 (Thr37/46) 1:200 (#2855). IHC scoring represents the mean percentage of positive cells counted from 8 to 10 images/mouse (×200 magnification, BX-51 Olympus microscope, n = 3/genotype).
PI3K Inhibitor Administration
Cohorts of male Pik3ca+/HR or Ptenfl/fl mice were treated at 400 and 200 days old, respectively: A66 (p110α-specific inhibitor, 100 mg/kg, daily p.o.), TGX-221 (p110β-specific inhibitor, 30 mg/kg, daily p.o.), and BKM120 (pan-PI3K inhibitor, 40 mg/kg, daily p.o.). Inhibitors were dissolved in filter-sterilized 20% hydroxyproyl-beta-cyclodextrin (Sigma), sonicated for 10 minutes, and dosed immediately (4 weeks; 5 days on, 2 days off). No appreciable toxicity was observed (i.e., >20% weight loss). A66 and TGX-221 were generated in house by P.R. Shepherd (University of Auckland, New Zealand), and BKM120 was obtained from SYNkinase.
RNA Isolation and qRT-PCR Analysis
Total RNA extraction from mouse prostate tissue and qRT-PCR were performed according to standard methods described in the Supplementary Methods.
RNA In Situ Hybridization
FFPE mouse prostate tissue sections were probed using the RNAscope 2.5 high-definition red detection kit (#322350; Advanced Cell Diagnostics). Slides were counterstained with hematoxylin. Scoring represents the average number of RNA molecules per 50 cells/mouse (×400 magnification, BX-43 Olympus microscope, n = 3/genotype).
RPPA
Protein lysates were prepared from snap-frozen tissue homogenized in CLB1 buffer (Zeptosens, Bayer) and quantified using a Pierce Coomassie Plus (Bradford) Protein Assay Kit (n = 3/cohort). Using a Sciclone/Caliper ALH3000 liquid handling robot (Perkin Elmer), samples were serially diluted in 10% CLB1:90% CSBL1 buffer (Zeptosens, Bayer) and spotted onto ZeptoChips (Zeptosens) in duplicate using a Nano-plotter-NP2.1 noncontact microarray system (GeSim). Chips were blocked under noncontact conditions for 1 hour with BB1 buffer (Zeptosens) and incubated with prevalidated primary antibodies (1:500, 20 hours) and Alexa Fluor 647 anti-rabbit secondary antibody (1:1,000, 4 hours; #Z-25308; Thermo Fisher Scientific). Chips were read on a Zeptosens instrument, and software version 3.1 was used to calculate the relative fluorescence intensity. All samples were normalized to the background values reported in the secondary antibody-only negative control. Pearson's correlation was calculated to confirm replicate pairs were adequately correlated (correlation coefficient >0.9). Data were Log2-normalized, median centered and rescaled between 0 and 1 using the formula: represents a vector of antibody responses for a given sample. The RPPA heat map was generated in R using pheatmap. For details on western blotting validation, see the Supplementary Methods.
Analysis of Genomic Datasets
Analysis of PIK3CA gene mutation/amplification was performed on prostate cancer patient datasets with sequencing and copy-number alteration (CNA) data using the cBioPortal platform (14). The TCGA provisional dataset was downloaded from the TCGA data portal (https://tcga-data.nci.nih.gov/); PIK3CA segment mean Log R-Ratio ≥ 0.135. To minimize CNA noise, probe number was filtered to ≤10. Silent mutations were excluded.
Statistical Analysis
Prostate weight and IHC scoring were analyzed using a one-way ANOVA with Tukey correction or an unpaired t test (95% confidence interval) as indicated using GraphPad Prism_7.03 software. Kaplan–Meier plots were generated, and age-adjusted Cox proportional hazard regression ratio was calculated using R software. For RPPA, an unpaired two-tailed t test with Welch's correction was calculated using R software. P < 0.05 was considered statistically significant.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: H.B. Pearson, P.O. Humbert, W.A. Phillips
Development of methodology: H.B. Pearson, J. Li, A.A. Macpherson, K.J. Simpson
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.B. Pearson, V.S. Meniel, C.M. Fennell, P. Waring, K.G. Montgomery, R.J. Rebello, A.A. Macpherson, L. Furic, C. Cullinane, K.J. Simpson, T.J. Phesse, P.R. Shepherd, P.O. Humbert, W.A. Phillips
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.B. Pearson, J. Li, K.J. Simpson, W.A. Phillips
Writing, review, and/or revision of the manuscript: H.B. Pearson, J. Li, C.M. Fennell, K.G. Montgomery, R.J. Rebello, L. Furic, C. Cullinane, R.W. Clarkson, M.J. Smalley, K.J. Simpson, T.J. Phesse, P.O. Humbert, O.J. Sansom, W.A. Phillips
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.B. Pearson, J. Li, S. Koushyar, W.A. Phillips
Study supervision: H.B. Pearson, O.J. Sansom, W.A. Phillips
Acknowledgments
This work was generously supported by a Prostate Cancer Foundation of Australia concept grant (#CG 1611) and an NHMRC project grant (#1080491) awarded to W.A. Phillips, and a Peter MacCallum Cancer Foundation grant (#1520) awarded to H.B. Pearson. H.B. Pearson is supported by a Marie Skłodowska Curie Actions/Sêr Cymru II/Horizons 2020 COFUND fellowship (#663830-CU-041). T.J. Phesse is supported by a Capital Medical University/Cardiff University Fellowship. L. Furic is supported by the Department of Health and Human Services acting through the Victorian Cancer Agency (MCRF16007). P.O. Humbert is supported by an NHMRC Senior Research Fellowship (#1079133). The Victorian Centre for Functional Genomics (VCFG) and the RPPA platform (K.J. Simpson) is funded by the Australian Cancer Research Foundation (ACRF), the Australian Phenomics Network (APN) through funding from the Australian Government's National Collaborative Research Infrastructure Strategy (NCRIS) program, the Peter MacCallum Cancer Foundation, and the University of Melbourne Collaborative Research Infrastructure Program. The authors wish to thank the animal, bioinformatics, VCFG-RPPA, microscopy, and histology core facilities at the Peter MacCallum Cancer Centre for supporting this project, and the histology departments at the Beatson Institute of Cancer Research and the European Cancer Stem Cell Research Centre. We also thank Nathan Crouch (VCFG-RPPA) for bioinformatics analysis of RPPA data, as well as Samantha McIntosh, Kerry Ardley, Susan Jackson, Lauren Dawes, Stephanie Le, Katherine Papastratos, and Qerime Mundrea at the Peter MacCallum Cancer Centre, Rachel Ridgway at the Beatson Institute for Cancer Research, and Derek Scarborough at The School of Biosciences, Cardiff University, for their technical assistance.
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.