Vulnerabilities of PTEN–TP53-Deficient Prostate Cancers to Compound PARP–PI3K Inhibition

There is an urgent need for the development of novel successful strategies for advanced prostate cancer to improve patient outcomes (1, 2).

The loss of at least one allele of phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is observed in more than 40% of prostate cancers (3, 4). As a lipid phosphatase, PTEN suppresses the activation of the PI3K–AKT signaling cascade that is a central proto-oncogenic signaling hub in prostate cancer development (5–9). Notably, PTEN also exerts tumor-suppressive functions in the nucleus by regulating genomic instability, and DNA-repair defects driven by PTEN loss have been proposed as a rationale for the sensitivity to PARP inhibitors comparable with the enhanced sensitivity of BRCA-deficient cancers (5, 10–12). On this basis, we therefore aimed to test whether PARP inhibition, alone or combined with PI3K inhibition, would represent effective therapeutic modalities in prostate cancer.

We first aimed to determine whether common genetic alterations in prostate cancer respond toward olaparib. To mimic different stages of prostate cancer progression, we selected Pten^+/− (early stage), Pten^Lx/Lx (advanced hormone sensitive), and Pten^Lx/Lx;Trp53^Lx/Lx mouse embryonic fibroblasts (MEF; advanced hormone insensitive; Supplementary Fig. S1A) to perform a cell proliferation assay upon treatment with 10 μmol/L olaparib. Interestingly, MEFs of all genotypes (Fig. 1A–C) showed a significant growth inhibition upon exposure to olaparib. Surprisingly, in contrast with Pten^Lx/Lx;Trp53^Lx/Lx MEFs, Pten^+/− and Pten^Lx/Lx MEFs showed a robust senescence response (Supplementary Fig. S1B and S1C) that was significant and dose dependent in Pten^+/− (Fig. 1D) as well as Pten^Lx/Lx (Fig. 1E) compared with wild-type MEFs. Strikingly and surprisingly, further analysis revealed that in contrast to Pten^Lx/Lx MEFs, Pten^Lx/Lx;Trp53^Lx/Lx MEFs showed a significant increase in apoptosis rather than senescence (Supplementary Fig. S1C and 1F) as analyzed by Annexin V staining (Supplementary Fig. S1D and S1E) and detection of cleaved caspase-3/7 (Fig. 1G).

Western blot analysis of Pten^Lx/Lx and Pten^Lx/Lx;Trp53^Lx/Lx MEFs treated with increasing concentrations of olaparib revealed that whereas Pten^Lx/Lx MEFs showed a further increase in p53 protein levels, both Pten^Lx/Lx and Pten^Lx/Lx;Trp53^Lx/Lx MEFs showed increased DNA damage as visualized by γH2AX staining (Fig. 1H). This analysis demonstrates that the senescence response in Pten^Lx/Lx is likely driven by the induction of p53, as previously described (13). However, the concomitant loss of p53 induces increased DNA damage that in turn morphs this phenotype into an apoptotic response.

To validate our findings in vivo, we next studied the sensitivity of genetically engineered mouse models (GEMM) of prostate cancer toward PARP inhibition. Similar to our in vitro analysis, we enrolled Pten^+/−, Pten^Lx/Lx;Probasin-Cre (referred to as Pten^pc−/−) and Pten^Lx/Lx;Trp53^Lx/LxProbasin-Cre (referred to as Pten^pc−/−;Trp53^pc−/−) mice. In line with the data observed in MEFs, pharmacologic inhibition of PARP induced a strong and significant induction of senescence in Pten^+/− (Supplementary Fig. S2A) and Pten^pc−/− (Fig. 2A and B) models compared with vehicle-treated controls. In Pten^pc−/− mice, the senescence response was accompanied by increased DNA damage as analyzed by γH2AX staining of treated prostate tumors (Supplementary Fig. S2B). Histologic analysis of Pten^pc−/− tumors treated with olaparib revealed a modest decrease in high-grade prostatic intraepithelial neoplasia (HGPIN; Fig. 2C). However, this trend did not reach statistical significance.

Next, we tested whether Pten^pc−/−;Trp53^pc−/− mice showed a similar apoptotic response upon treatment with olaparib as observed in vitro. In line with the in vitro data, olaparib treatment increased γH2AX in dorsolateral prostate (DLP) tumors of Pten^pc−/−;Trp53^pc−/− mice (Supplementary Fig. S2C). Surprisingly, macroscopic analysis and cytokeratin 8 staining (luminal cells) of olaparib-treated prostates revealed that more glands were lined by a single layer compared with the vehicle control (Fig. 2D and Supplementary S3A). In addition, analysis of cytokeratin 14 showed a reduction of the intermediate basal cell population in single-layered glands, suggesting a certain degree of normalization after treatment (Supplementary Fig. S3B). Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) and caspase-3 staining further revealed a significant increase in apoptotic cells upon olaparib treatment (Fig. 2E and F and Supplementary S2D). However, similar to Pten^pc−/− mice, histologic analysis of Pten^pc−/−;Trp53^pc−/− tumor reduction after drug treatment did not reach statistical significance (Fig. 2G), suggesting that single-agent olaparib treatment is not sufficient to induce a robust antitumor response in these models. Interestingly, mass spectrometry analysis of olaparib in prostates revealed that only approximately 2 μmol/L of the drug is delivered into the individual lobes, an amount that is significantly lower when compared with the dose used in our in vitro studies (Supplementary Fig. S3C). This marked difference in drug concentration may in turn provide one possible explanation for the limited overall tumor response in vivo.

On the basis of our data in mice, we sought to determine whether olaparib could trigger a similar response in human prostate cancer cell lines. To this end, we engineered LNCaP cells that lack PTEN to additionally lose p53 function by overexpressing either a p53 short hairpin construct (Supplementary Fig. S4A) or dominant-negative p53 (p53-DN; Supplementary Fig. S4B). Similar to MEFs, olaparib induced senescence in a p53-proficient setting that was blunted upon loss of p53 (Fig. 3A). In contrast, LNCaP cells expressing p53-DN did not show differential sensitivity to olaparib treatment. Furthermore, only high olaparib concentrations (10 μmol/L) induced a strong growth inhibition overall (Fig. 3B), suggesting mechanisms limiting olaparib efficacy beyond the limited uptake observed in vivo.

We therefore investigated whether classical survival signaling such as the PI3K–AKT pathway might be superactivated upon olaparib exposure and therefore analyzed AKT activation upon olaparib treatment in vitro (7, 14–16). Indeed, Pten^Lx/Lx;Trp53^Lx/Lx MEFs (Fig. 3C) and LNCaP cells (Fig. 3D) hyperactivated AKT upon exposure to olaparib, suggesting that AKT could affect the response to olaparib. Indeed, knockdown of AKT1, the major AKT isoform, significantly increased growth inhibition upon PARP inhibition (Supplementary Fig. S4C and 3E) and induced apoptosis, as shown by increased PARP cleavage (Fig. 3F).

To investigate the mechanism of AKT hyperactivation, we made use of the dual PI3K inhibitor BKM120 that efficiently blocked AKT activation in LNCaP cells (Fig. 3F). Strikingly, PI3K inhibition was able to restore pAKT to DMSO control levels in olaparib-treated cells, suggesting that the major activator of AKT upon PARP inhibition is indeed PI3K (Fig. 3G). Importantly, analysis of cleaved PARP demonstrated that the combination of PARP and PI3K inhibition shows synergistic apoptosis induction, suggesting efficacy for a combinatorial treatment strategy in prostate cancer (Fig. 3G).

On the basis of our findings, we next aimed to investigate whether a combination treatment including PARP and PI3K inhibition would be synergistic in vitro. To this end, we treated Pten^Lx/Lx;Trp53^Lx/Lx MEFs and human prostate cancer cells with either single or combination therapy (Supplementary Fig. S4D–S4G). Indeed, the combination significantly increased growth suppression compared with the single treatments in all tested cell lines.

These promising in vitro data prompted us to next evaluate the efficacy of the combinatorial inhibition of PARP and PI3K in our Pten^pc−/−;Trp53^pc−/− model of highly aggressive prostate cancer (Supplementary Fig. S5A). In vivo efficacy of control, single agent, and the combination arm was assessed by MRI analysis for the dorso-lateral prostate (DLP) and the anterior prostate (AP; Supplementary Fig. S5B and S5C), followed by the analysis of survival and histopathologic differences.

To assess the efficacy of PARP and PI3K inhibition in vivo, we first followed tumor progression by MRI (Fig. 4A). Interestingly, treatment with olaparib or BKM120 showed tumor stabilization but eventually tumors regrew due to potential resistance mechanisms. Strikingly, the combination of the two compounds potentiated tumor stabilization and induced robust tumor regression. These data indicate that the combinatorial treatment of PI3K and PARP inhibition is synergistic to the single-compound arms in vivo. In contrast with the DLP, only slight growth suppression was seen in APs that was potentiated in the combination, suggesting a similar cooperation as seen in the DLP albeit with a much lower efficacy (Supplementary Fig. S6A).

We further analyzed tumor specimens of all treatment arms after 1 to 4 weeks by histopathology (Fig. 4B and C and Supplementary S6B and S6C). Surprisingly, despite the initial response toward olaparib observed at 1 week, histopathologic analysis and cytokeratin 8 staining revealed no difference in the occurrence of HGPIN after 4 weeks compared with the vehicle control. In contrast, the BKM120 treatment arm triggered a clear decrease in HGPIN. Importantly, the combinatorial treatment significantly potentiated this effect, clearly showing cooperation between the two compounds. Furthermore, cytokeratin 14 staining in the DLPs treated with the combination showed a clear reduction in the intermediate basal cell population that is not apparent in the vehicle- or single-arm–treated tumors. These data indicate a trend toward normalization of the glands specifically in the combination treatment.

In line with these data, progression-free survival (PFS) was markedly improved in the combinatorial treatment arm compared with the single and control arms (Fig. 4D and Supplementary S7A and S7B). However, and coherent with the lack of response in the AP, overall survival (OS) analysis did not reveal a striking difference between the single and combination treatment arms (Supplementary Fig. S7C).

Finally, to determine whether the data obtained in our GEMMs could be confirmed in human cell lines, we performed xenograft studies using PC3 prostate cancer cell lines (Fig. 4E). Similar to our GEMMs, PC3 xenografts displayed a significant and synergistic response when treated with the combination compared with the single agents. Collectively, these data strongly support the notion of a marked cooperative effect between the two drugs for the treatment of prostate cancer.

This study allowed us to reach a number of important conclusions:

First, we demonstrated, in vitro as well as in vivo, that PTEN deficiency triggers sensitivity toward PARP inhibition with differential responses. In a p53-proficient setting, loss of PTEN triggers a senescence response upon PARP inhibition that is morphed into apoptosis upon loss of p53.

Second, a robust response toward PARP inhibition was observed only at very high concentrations of olaparib in cell lines, whereas mouse DLPs treated with olaparib displayed no significant difference in HGPIN grades. In line with these observations, we find that olaparib reaches only suboptimal concentration in vivo in the prostate.

Third, we found that olaparib triggers the activation of AKT, a classical cell survival mechanism in response to cellular stresses (7, 14–16). These data demonstrated that olaparib treatment on its own is insufficient to trigger a significant tumor response in the prostate due to a number of reasons.

Fourth, we showed that PI3K inhibition by BKM120 significantly increased sensitivity toward olaparib both in GEMMs and xenograft models. Importantly, PC3 xenografts showed a clear synergistic effect of growth inhibition in the combination arm in vivo, validating the efficacy of the combination in a human setting. Intriguingly, these data, especially in PC3 xenografts, are in clear contrast with the in vitro data that rather suggested an additive effect of the combination. These results could be explained by the profound differences in nutrients and growth factors in the two experimental settings, or by the fact that the tumor microenvironment/metabolism may also affect the efficacy of the treatments, as previously suggested (17–19).

Finally, despite these striking findings, combinatorial PARP and PI3K inhibition prolonged PFS and OS significantly only when compared to vehicle, but not to single-agent–treated animals. However, these data are potentially influenced by the general characteristics of Pten^pc−/−;Trp53^pc−/− prostate tumors (13). Cancers in the APs in this model are intrinsically resistant to therapy, including androgen deprivation, and showed only a slight growth inhibition in all of the treatment arms (20). However, the rapid growth of the APs strongly influences the survival of Pten^pc;Trp53^pc−/− mice and, thus, strongly affects the assessment and interpretation of OS.

In conclusion, we show that PTEN-deficient prostate cancer is sensitive to PARP inhibition. Because of AKT hyperactivation and poor pharmacokinetics, the in vivo efficacy of olaparib is modest, but concomitant suppression of PARP and PI3K results in a cooperative antitumor effect. Thus, the combination of PI3K and PARP inhibitors represents a promising therapeutic approach for the treatment of PTEN-deficient prostate tumors.

Pten^+/−, Pten^Lx/Lx, and Pten^Lx/Lx;Trp53^Lx/Lx MEFs were prepared as previously described. For the excision of target genes, MEFs were retrovirally infected with pMSCV–Cre–PURO–IRES–GFP (13).

Human cell lines were obtained from the ATCC and cultured according to the manufacturer's instructions. Cell lines were tested for mycoplasma (MycoAlert; Lonza), but not further authenticated.

To study proliferation, cells were plated in a 12-well plate (MEFs at 1.7 × 10⁴ cells/well and prostate cancer cells at 1 × 10⁵ cells/well). After 2 to 9 days, cells were washed with PBS, fixed with 10% paraformaldehyde, and stained with crystal violet. Crystal violet was extracted with 10% acetic acid and absorbance was detected at 595 nm.

To determine senescence, cells were plated in a 6-well plate (MEFs at 3.4 × 10⁴ cells/well and LNCaP at 1.7 × 10⁴ cells/well). SA-β-Gal was detected after 4 days with the Senescence Detection Kit (Calbiochem) following the manufacturer's instructions. For quantification, more than 200 cells per sample were counted.

For prostate tissue, frozen sections (6 mm) were stained for SA-β-Gal (Calbiochem) following the manufacturer's instructions. For quantification, slides were digitalized and analyzed with the Aperio Spectrum ImageScope software.

Apoptosis was determined by the Apo-ONE caspase-3/7 assay (Promega) and additionally analyzed by a dual Annexin V–APC/7-AAD (BD Biosciences) staining using flow cytometry following the manufacturer's instructions.

For in vivo apoptosis, samples were analyzed with the In Situ Cell Death Detection Kit (Roche) following the manufacturer's instructions. For the quantification of positive TUNEL staining, a total of 500 cells were counted from five different fields.

For Western blotting, cell lysates were prepared with RIPA buffer or NP40 buffer (Boston BioProducts) supplemented with protease (Roche) and HALT phosphatase inhibitor cocktail (Thermo Scientific), and subsequently subjected to SDS–Gel separation (Invitrogen) and Western blotting. The following antibodies were used from Cell Signaling Technology: anti-γH2AX, anti-AKT1, anti-mouse PTEN, anti-mouse p53, anti-AKT, anti-phospho threonine 308 and anti–phospho-serine 473 anti-PARP, anti–cleaved PARP or human p53 (DO1; Santa Cruz Biotechnology), anti-PAR (BD Biosciences), and anti–β-actin (AC-74; Sigma). Western blots were quantified using ImageJ software.

For IHC, tissues were fixed in 4% paraformaldehyde and embedded in paraffin in accordance with standard procedures. Hematoxylin and eosin (H&E) staining of sections was performed by the Histology Core at the Beth Israel Deaconess Medical Center (BIDMC; Boston, MA).

For in vitro studies, drugs were resuspended in DMSO (olaparib; LC Laboratories, 10⁻² mol/L; BKM120, Novartis Pharma AG, 10⁻² mol/L). For in vivo studies, olaparib in DMSO (50 mg/mL) was diluted in vehicle solution [10% 2-hydroxyl-propyl-βCyclodextrine (Sigma) in PBS, 10 μL/g body] and administered daily by i.p. injection at a dose of 50 mg/kg. BKM120 was resuspended in methylcelullose solution at a final concentration of 6 mg/mL and administered daily at a dose of 30 mg/kg by oral gavage.

All mice were maintained in the animal facilities of BIDMC/Harvard Medical School in accordance with institutional rules and ethical guidelines for experimental animal care. All animal experiments were approved by the BIDMC Institutional Animal Care and Use Committee protocol 066-2011 (8, 13, 21).

For drug treatments, Pten^+/− (30-week-old) and Pten^pc−/− (10-week-old) mice were treated with olaparib over 2 weeks. Pten^pc−/−;Trp53^pc−/− (4-month-old) mice were treated with olaparib over 1 week.

For the combination treatment of olaparib and BKM120, Pten^pc−/−;Trp53^pc−/− (4-month-old) mice were evaluated by MRI for the presence of prostate tumors and subsequently enrolled in the following treatment arms: vehicle, olaparib, BKM120, or combination at the same doses. Mice were imaged with MRI before treatment, monitored by MRI until tumor progression, and followed for survival analysis. Tumor progression was defined as an increase of >50% in tumor volume.

Xenografts were performed with PC3 cells. A total of 1 × 10⁶ cells were injected subcutaneously in the flanks NCr nude mice (Taconic). Mice were enrolled at a tumor size of 300 m³ in the following treatment arms: vehicle, olaparib, BKM120, or combination at the same doses. Mice were euthanized after 2 weeks of treatment, and tumors were subjected to further analysis.

Mouse prostate images were acquired on an ASPECT Model M2 1T tabletop MRI scanner (ASPECT Magnet Technologies Ltd.) and imaged as previously described (20). Tumor volume quantification was performed as previously described (22).

For all statistical analyses, GraphPad Prism 6 software was used, and values of P < 0.05 were considered statistically significant. Survival was determined by Kaplan–Meier curves. The Mantel–Cox test was used to determine significance between survival curves. All other statistical analyses were done by an unpaired Student t test.

Scoring of HGPIN status was performed as previously described (23). At least three samples were analyzed at the accorded time points. For the preclinical assessment of the combinatorial treatment in Pten^pc−/−;Trp53^pc−/− mice, at least three mice were analyzed after 4 weeks of treatment except the vehicle control, in which mice were pooled from 1- and 4-week treatment.

L.C. Cantley has received a commercial research grant from GlaxoSmithKline and is a consultant/advisory board member for Novartis. No potential conflicts of interest were disclosed by the other authors.

The Editor-in-Chief of Cancer Discovery (L.C. Cantley) is an author of this article. In keeping with the AACR's Editorial Policy, the paper was peer reviewed and an AACR journal editor not affiliated with Cancer Discovery rendered the decision concerning acceptability.

Conception and design: E. González-Billalabeitia, N. Seitzer, A. Patnaik, C. Nardella, L.C. Cantley, P.P. Pandolfi

Development of methodology: E. González-Billalabeitia, A. Patnaik, X.-S. Liu, A. Papa, M. Chen, M. Loda, J.M. Asara

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. González-Billalabeitia, A. Patnaik, R.M. Hobbs, M. Chen, C. Ng, K.A. Webster, M. Loda, J.M. Asara, C. Nardella, J.G. Clohessy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. González-Billalabeitia, N. Seitzer, M.S. Song, A. Patnaik, R.M. Hobbs, S. Signoretti, M. Loda, J.M. Asara, J.G. Clohessy, P.P. Pandolfi

Writing, review, and/or revision of the manuscript: E. González-Billalabeitia, N. Seitzer, M.S. Song, A. Patnaik, A. Lunardi, M. Loda, J.M. Asara, L.C. Cantley, P.P. Pandolfi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. González-Billalabeitia, M.S. Song, A. Patnaik, M.T. Epping

Study supervision: E. González-Billalabeitia, N. Seitzer, A. Patnaik, J.G. Clohessy, P.P. Pandolfi

Carried out experiments: S.J. Song

Provided histology and immunohistochemistry: K.A. Webster

The authors thank the current members of the Pandolfi laboratory for critical discussion, Thomas Garvey for insightful editing, and Min Yuan and Susanne Breitkopf for help with mass spectrometry experiments.

This work has been supported through a Mouse Models of Human Cancers Consortium/NCI grant (RC2 CA147940-01), R01 GM41890, and P01 CA089021 (to P.P. Pandolfi and L.C. Cantley), and NIH grant 5P01CA120964-04 and NIH Dana-Farber/HCC Cancer Center Support Grant 5P30CA006516-46 (to J.M. Asara). E. González-Billalabeitia has been supported by a Bolsa de ampliación de estudios (BAE) from the Instituto de Salud Carlos III (Spain).