Small-Molecule Activators of Protein Phosphatase 2A for the Treatment of Castration-Resistant Prostate Cancer

Depletion of circulating androgens is the standard first-line treatment for metastatic prostate cancer and results in tumor regression and symptomatic improvement in the majority of patients. However, metastatic prostate cancer inevitably progresses despite castrate levels of serum testosterone, a disease state known as castration-resistant prostate cancer (CRPC). Studies in model systems, and in patients, have confirmed that the androgen receptor (AR) signaling axis remains a key therapeutic target in CRPC, and novel approaches to androgen biosynthesis inhibition and androgen receptor blockade have resulted in improved patient outcomes (1, 2). Still, such treatments are not curative and prostate cancers ultimately develop resistance via a variety of mechanisms including ligand-independent AR signaling or alternative pathway activation, highlighting the need for new therapeutic approaches (3).

Protein phosphatase 2A (PP2A), a serine/threonine phosphatase and bona fide tumor suppressor, has been implicated in the pathogenesis of CRPC. PP2A dephosphorylates a large number of critical oncogenic proteins including AKT1, ERK, MYC, BCL2, and others (4). Indeed, PP2A has been shown to directly bind and dephosphorylate the AR (5). PP2A is frequently decreased or functionally inactivated in CRPC and decreased expression in human tumor specimens has been associated with inferior outcomes and enzalutamide resistance (6–8). Nonpharmaceutical activators of PP2A have demonstrated anticancer effects in prostate cancer model systems (9–15). Together, these findings point to therapeutic activation of PP2A as a novel strategy for the treatment of prostate cancer.

Tricyclic neuroleptics have been reported to activate PP2A in cells through direct binding of the PP2A Aα subunit, though clinical development of such molecules for cancer therapy is limited by central nervous system toxicity (16, 17). We have developed first-in-class small-molecule activators of PP2A (SMAP) by repurposing and reengineering FDA-approved tricyclic neuroleptics (18). By replacing the basic amine with a neutral polar functional group, the central nervous system effects were abrogated and further chemical derivatization has improved anticancer potency (18). Through binding studies using a tritiated version of SMAPs, in silico docking calculations, photo-affinity labeling experiments, and hydroxyl radical footprinting studies, we have shown that these SMAPs directly bind to the PP2A Aα subunit (19). The postulated mechanism of action is that SMAP binding promotes allosteric conformational changes, leading to the activation of the phosphatase and dephosphorylation of key substrates (19). Given the critical role of the AR, a known PP2A substrate in CRPC, as well as the role of PP2A in the pathogenesis of prostate cancer, we explored the effects of SMAPs in cell culture and in vivo model systems of CRPC.

All compounds were synthesized in the laboratory of Dr. Michael Ohlmeyer at the Icahn School of Medicine at Mount Sinai (New York, NY). Compounds were stored at room temperature.

LNCaP (catalog no. CRL-1740, lot no. 59453491) and 22Rv1 (catalog no. CRL-2505, lot no. 60437301) cell lines were purchased from the ATCC and authenticated by ATCC (STR DNA profiling). The LNCaP/AR cell line was a generous gift from Dr. Charles Sawyers (Memorial Sloan Kettering Cancer Center, New York, NY) and was authenticated by ATCC STR DNA Profiling Authentication Services (ATCC) with the FTA Sample Collection Kit for Human Cells (catalog no. 135-XV3). All cell lines included in the Oncopanel 240 Anti-Cancer Drug Profiling Assay (Eurofins Panlabs, Inc.) were authenticated through STR DNA profiling by Genetica DNA Laboratories. Mycoplasma testing was performed routinely with Lonza MycoAlert Mycoplasma Detection Kit as per the manufacturer's protocol (catalog no. NC9922140, Thermo Fisher Scientific). LNCaP, 22Rv1, and LNCaP/AR cells were cultured in RPMI1640 medium (Thermo Fisher Scientific) with 10% fetal bovine serum (HyClone) and 0.5% penicillin–streptomycin (Thermo Fisher Scientific). All cell lines were maintained at less than 80% confluence and 25 passages. Cells were maintained at 37°C with 5% CO2. SMAPs (dissolved in DMSO) were diluted to a stock concentration of 80 mmol/L. Dilutions to the needed concentrations were made in RPMI1640 (Thermo Fisher Scientific).

Cells were treated with SMAPs and screened for cell viability through the MTT assay using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma-Aldrich). For clonogenic assays, cells were plated at a low density in 6-well plates. After 48 hours, cells were treated with DMSO (Simga-Aldrich) or increasing concentrations of SMAPs for 10–12 days. Drug medium was refreshed every 48 hours. Cells were fixed and stained with 1% crystal violet solution (Sigma-Aldrich). Quantification was performed through the cell counter function on ImageJ (imagej.nih.gov/ij/).

Cell protein was isolated with RIPA Lysis and Extraction Buffer (Thermo Fisher Scientific). Isolated protein was quantified, normalized via Bio-Rad assay (Bio-Rad), run on a 12% SDS-PAGE (Invitrogen, Life Technologies), and transferred onto nitrocellulose membranes (Bio-Rad). The membrane was blocked with 5% nonfat milk (LabScientific Inc.) in Tris-buffered saline–Tween 20 buffer. Antibody information is provided in Supplementary Materials and Methods.

Annexin V staining was performed using Annexin V conjugate Alexa Fluor-488 (Invitrogen, Life Technologies) or 7-Aminoactinomycin D (7-AAD; Thermo Fisher Scientific) and Annexin binding buffer (catalog no. V13246, Invitrogen, Life Technologies), according to the manufacturer's protocol. Cells were also stained with propidium iodide (Roche) to ascertain the DNA content and as a marker of cell death. Each experiment was performed in triplicate.

Quantitative global phosphorylation studies were performed with LNCaP cells using an unfractionated label-free LC/MS-MS workflow. Cells were treated with DMSO or 30 μmol/L SMAP for 6 hours. After incubation, cells were harvested, pelleted, and washed with PBS. Samples were lysed with 2% SDS solution containing both protease (catalog no. P8465, Sigma Aldrich) and phosphatase inhibitors (PhosphoSTOP, Roche). All samples were quantified by the Bicinchoninic acid assay (BCA assay) and normalized on the basis of total protein concentration before processing for global phosphoproteomics. Detergent removal was performed on 200 μL of the cell lysate using the FASP cleaning procedure (20). Eight-hundred micrograms of each sample was digested using by a 2-step Lys-C/trypsin proteolytic cleavage. Each digest was equally split into two 400-μg samples to provide two technical replicates. Each technical replicate was subsequently enriched for phospho-peptides using commercially available TiO2 enrichment spin tips (Thermo Fisher Scientific). The entirety of the samples was phospho-enriched, with no residual amount saved for parallel unenriched LC/MS-MS. A 3-hour LC/MS-MS method was performed using a UPLC system (NanoAcquity, Waters) that was interfaced to Orbitrap ProVelos Elite MS system (Thermo Fisher Scientific) to perform the LC/MS-MS data collection. Clustering of all the peptide precursor ions across all the chromatographic analyses was performed using the peak alignment algorithm of the Rosetta Elucidator software. Automated differential quantification of phospho-peptides was accomplished using downstream quantitative analysis modules of Elucidator. MS/MS peak lists were generated and subsequently searched by Mascot version 2.4.0 (Matrix Science). The database used was human UniProt (538,585 sequences). Search settings were as follows: trypsin enzyme specificity; mass accuracy window for precursor ion, 8 ppm; mass accuracy window for fragment ions, 0.6 Daltons; variable modifications including carbamidomethlylation of cysteines, phosphorylation of serine, threonine, and tyrosine and 1 missed cleavage. Peptide and protein identifications were integrated from the protein database search engine output with these quantifications. Fold changes were calculated from the mean peptide intensity of SMAP over the mean intensity of DMSO. Statistical significance of abundance changes was determined using Welch t test.

Data preprocessing was performed using R. An unfiltered list of phosphoproteins identified from phosphoproteomics was searched against the HPRD binary protein–protein interaction database, Release 9. Hits and their interactions were compiled into a .sif file and imported into Cytoscape 3.0.2 for downstream processing and image generation. For proteins with multiple phosphosites detected, the residue with the lowest P value was selected for P value and fold change illustrations. The MCODE plugin was used to extract clusters using the default settings except for a Node Score Cutoff = 0.5. Node degree was calculated using the NetworkAnalyzer tool.

Calculations and plotting were performed using R. Please refer to the original publication (21) for details on the formulas. The authors reported three variations on the algorithm; our analysis was based on the third method described in their Materials and Methods. We assigned kinase–substrate links based on the Kinase Substrate Dataset from PhosphoSitePlus (October 2015 release), search restricted to human proteins. The calculated P values were adjusted for multiple hypotheses testing using the p.adjust function in R (method=“fdr”) and reported in the Supplementary Table S1.

Studies were conducted after Institutional Animal Care and Use Committee approval at the Icahn School of Medicine at Mount Sinai [protocol: small-molecule activators of PP2A (SMAP) for prostate cancer therapy, LA13-00005]. Animal use and care was in strict compliance with institutional regulatory standards and guidelines. Eight-week-old male SCID/NCr (BALB/c background) noncastrated and castrated mice (strain no. 01S11, NCI-Frederick Mouse Repository, Frederick, MD) were used for treatment studies. Studies were nonblinded. MDV3100 (enzalutamide, catalog no. S1250) was purchased from Selleck Chemicals. LNCaP/AR cells (5 × 10⁶) were suspended in 200 μL of 1:1 RPMI1640/Matrigel (BD Biosciences) and were injected subcutaneously into the right flank of the mice. When tumor volumes reached an average of 100–250 mm³ mice were randomized to treatment groups. Tumor volumes and body weights were measured every other day throughout the study. Tumor volume was assessed by a digital caliper and determined using the formula: length × width²/2. Percentage of mice body weights during treatment were calculated as: weight at each time-point/initial weight × 100. Mice were treated by oral gavage with vehicle control, MDV3100 (100 mg/kg once daily), SMAP (100 mg/kg or 400 mg/kg twice daily), or SMAP-2 (30 mg/kg or 100 mg/kg twice daily). MDV3100, SMAP, and SMAP-2 for efficacy studies were formulated in a homogenous suspension with 0.1% Tween-80 (lot no. MKBP0682V, Sigma-Aldrich)/0.5% NaCMC (lot no. SLBF3845V, Sigma-Aldrich) in diH₂O. SMAP-2 for the pharmacodynamic study was prepared in a N,N-Dimethylacetamide (DMA)/Kolliphor HS-15 (Solutol)/diH₂0 solution. Animals were observed for signs of toxicity (i.e., mucous diarrhea, abdominal stiffness, and weight loss). Blood and tumor tissue was harvested 2 hours post-final dose of treatment. Tumors were formalin-fixed for IHC or snap frozen in liquid nitrogen for immunoblotting. Serum was submitted for toxicology testing at IDEXX Laboratories.

Isolated tissue was fixed in 10% buffered formalin phosphate (Thermo Fisher Scientific, catalog no. SF100-4), transferred to 70% ethanol, and blocked in paraffin. Serial tissue sections (5-μm thickness) were cut from the paraffin-embedded blocks and placed on charged glass slides. Tumor sections were stained with hematoxylin and eosin (H&E) and PCNA (Abcam, catalog no. ab92729). Sections were deparaffinized with xylene, rehydrated through graded alcohol washes, and followed by antigen retrieval in a pressure cooker (Dako/Agilent Technologies) in citrate buffer (10 mmol/L, pH 6.0, Vector Laboratories). Slides were incubated in hydrogen peroxide–methanol and incubated in normal goat serum/PBS. Primary antibody was applied overnight at 4°C. DAB substrate was applied followed by counterstaining in hematoxylin. The ApopTag Fluorescein In Situ Apoptosis Detection Kit (TUNEL; catalog no. S7110, Millipore-Sigma) was used according to the manufacturer's protocol. Vectashield Mounting Medium with propidium iodide (catalog no. H-1300), Vector Laboratories) was used for counterstaining. Bright-field and fluorescent images were captured using a Zeiss Axioplan 2IE microscope. Imaging was performed at the Icahn School of Medicine at Mount Sinai Microscopy CORE. ImageJ software with cell counter function was used to quantify stained cells (imagej.nih.gov/ij.).

Total RNA was extracted using the QiaShredder Kit (Qiagen) and the RNeasy Kit (Qiagen). cDNA synthesis was carried out using the iScript cDNA Synthesis Kit (Bio-Rad) as per the manufacturer's instructions. Sequences can be found in Supplementary Materials and Methods. Real-time PCR was performed with SYBR green PCR Master Mix (Applied Biosystems) on the Applied Biosystems 7900HT Fast Real-Time PCR System.

GraphPad Prism 6 or 7 software (GraphPad Software Inc.) were used for statistical analysis. The results are presented as a mean ± SD. In vitro experiments were performed in triplicate for three biological replicates. One-way ANOVA was utilized to compare mean values and Tukey or Dunnett test were applied where appropriate. P values ≤ 0.05 were considered statistically significant.

To investigate the broad utility of our reengineered tricyclics toward various cancer cell lines, an Oncopanel 240 anticancer drug-profiling screen was performed by Eurofins Pathlab, Inc. The tool compound, TRC-382, displayed growth-inhibitory properties in several tumor types, including prostate cancer (Fig. 1A; Supplementary Fig. S1). Although TRC-382 exhibited broad activity across a variety of cancer types, which is perhaps not surprising given the range of oncogenic proteins that serve as PP2A substrates, prostate cancer cell lines were among the most sensitive. Furthermore, among the prostate cancer cell lines, the AR-expressing cell lines, 22Rv1 and LNCaP, exhibited the greatest sensitivity to TRC-382 (Fig. 1A), whereas growth was not inhibited in the immortalized but nontumorigenic cell line, BPH-1. On the basis of these findings and prior research indicating that nonpharmaceutical activators of PP2A have anti-prostate cancer activity (9–15) and literature indicating that several PP2A substrates are implicated in the pathogenesis of prostate cancer (7, 22–25), further functional studies were pursued in prostate cancer. Subsequent studies focused on AR-expressing prostate cancer and were performed with SMAP and SMAP-2, structurally similar variants of TRC-382 containing a hydroxylated linker that confers improved bioavailability (Supplementary Fig. S2; ref. 18). LNCaP and 22Rv1 cells were treated with increasing concentrations of SMAP for 48 hours. SMAP treatment decreased viability in both cell lines with IC₅₀s of 16.9 μmol/L (LNCaP) and 14.1 μmol/L (22Rv1; Fig. 1B). As a pharmacologic control, these cells were also treated with increasing doses of TRC-766, which is structurally similar to SMAP but biologically inactive. TRC-766 lacks a key N-H sulfonamide hydrogen bond donor function likely necessary for the interaction with PP2A's catalytic subunit (Supplementary Fig. S2). While TRC-766 still binds PP2A, it does not activate the phosphatase and therefore it was postulated to not affect cell viability. Consistent with this hypothesis, biologically inactive TRC-766 displayed negligible effects on cell viability (Fig. 1B). Next, clonogenic assays were performed on LNCaP and 22Rv1 cell lines. Cells plated at low density were treated with increasing doses of SMAP or TRC-766 for 12 days. Consistent with its effects on viability, SMAP significantly inhibited cell survival, which was dose-dependent, whereas TRC-766 had minimal to no effect (Fig. 1C; Supplementary Tables S2 and S3). Subsequently, Annexin V staining of LNCaP and 22Rv1 cells was performed with increasing doses of SMAP for 24 hours. SMAP induced an increase in Annexin V positivity in LNCaP and a statistically significant increase in 22Rv1 cells, suggesting that treatment with increasing doses of SMAP can induce cell death (Fig. 1D; Supplementary Table S4). These results were confirmed by probing for cleaved PARP and caspase-3 cleavage, which were detected by Western blot analysis in both LNCaP and 22Rv1 cells at a dose of 30 μmol/L of SMAP as early as 6 hours posttreatment (Fig. 1E).

PP2A dephosphorylates and inactivates numerous substrates. To define perturbations in phosphorylation resulting from SMAP treatment, an analysis of the global phosphoproteome was performed. LNCaP cells were treated with 30 μmol/L of SMAP for 6 hours versus control. Extracted proteins were digested and phospho-enriched prior to LC/MS-MS analysis under the label-free protocol. A total of 3,051 unique phosphosites (p-sites), which mapped to 2,735 phosphopeptides and 1,496 phosphoproteins, were identified. Overall, 927 p-sites (mapping to 651 phosphoproteins) met a P < 0.05 Welch t test criterion and thus were deemed significantly perturbed by drug treatment as compared with control. From this list, 673 p-sites were downregulated upon treatment, and 254 were upregulated (Fig. 2A). This count suggests that SMAP treatment predominantly led to a decrease in phosphorylated peptides.

From this dataset, we then applied two bioinformatics tools to extract biological patterns in a relatively unbiased manner. First, we utilized a network model to identify candidate master regulators that are altered by SMAP treatment. We overlaid all our identified phosphoproteins onto a protein–protein interaction (PPI) network built from relationships documented in the Human Protein Reference Database, Release 9 (26). The resulting PPI contained 613 unique phosphoproteins linked by 982 interactions. We then reduced this large PPI into a more concise subnetwork using MCODE, a tool designed to enrich for tightly interconnecting proteins that are functionally related (27). We then profiled members of the top-scoring cluster to find those with the highest number of interactions (degree). These so-called “hub proteins” are often essential nodes of regulation within the cell (27). Interestingly, AR had the highest degree within our subnetwork (Fig. 2B) and across the original network (Supplementary Tables S5 and S19). Given this finding and the observation that this protein was significantly downregulated upon SMAP treatment, we hypothesized that our compound's activity disrupts AR-related processes.

In addition to the network analysis, we also sought to profile differential signaling fluxes in response to SMAP treatment. Kinase-Substrate Enrichment Analysis (KSEA) scores each kinase's relative activity output based on the collective phosphorylation fold change of its identified substrates (28). A negative value indicates that the majority of the kinase's substrates had decreased phosphorylation upon treatment, thereby suggesting that its overall signaling output is decreased compared with control. This approach allowed indirect profiling of proteins that may not be detected by mass spectroscopy but that are potentially regulated by SMAP treatment. We were able to score 101 unique kinases with at least one identified substrate from our phosphoproteomics experiment (Supplementary Tables S1 and S6). Interestingly, 3 of the 4 significantly downregulated kinases (negative score, P < 0.05, 3+ substrates) are documented regulators of AR (Fig. 2C): CDK1 (29), CDK5 (30), and SRC (31). Moreover 75% of all significantly scored kinases that are known to act on AR (negative + positive scores, P < 0.05) had decreased output (Fig. 2D). Altogether, these findings are consistent with inhibition of many upstream AR regulators.

As the bioinformatics results converged onto AR signaling, we subsequently decided to probe into SMAP's effects on the protein itself. Phosphoproteomics revealed significantly low levels of the AR-derived S308 phosphopeptide compared to control, with P = 0.028 and -4.1 log₂ fold change (Fig. 2A). This particular residue is a previously documented dephosphorylation site of the catalytic subunit of PP2A (PP2A-C; ref. 5). The phosphorylation status of this site can affect AR target gene transcription and AR-mediated cell growth (32). Importantly, the phosphoproteomic pipeline does not measure changes in total protein levels and therefore does not permit distinction between a decrease in p-AR due to dephosphorylation versus total AR protein loss. Given this limitation, and in the context of prior studies that demonstrated that PP2A can promote degradation of its substrates (e.g., MYC), we next probed the effects of SMAP on AR in cell culture.

LNCaP and 22Rv1 cells were treated with 0, 10, 20, and 30 μmol/L SMAP and harvested at 1, 3, 6, 12, and 24 hours after drug treatment. SMAP induced a time and dose-dependent decrease in AR protein expression in both LNCaP (Fig. 3A and B) and 22Rv1 cells (Fig. 3C) as well as a dose and time-dependent decrease in PSA protein expression in LNCaP cells (Fig. 3A). Antibodies specific to either the N- or the C-terminal domains of the AR were used for Western blots. Intriguingly, SMAP induced degradation of both AR WT and the splice variant AR (AR-v7) in 22Rv1 cells. The AR-v7 splice variant lacks the ligand-binding domain and has been shown to confer resistance to enzalutamide in VCaP and 22Rv1 cells (33). Moreover, the presence of AR-v7 in circulating tumor cells from metastatic CRPC patients correlates with clinical resistance to enzalutamide and abiraterone (34). AR is regulated at multiple levels, with numerous posttranslational modifications (PTM) including phosphorylation, acetylation, sumoylation, ubiquitination, and methylation (35, 36) with AR phosphorylation occupying a predominant role (37). PP2A binds and dephosphorylates AR at five phosphosites all located on the N-terminal domain, including Ser81, Ser94, Ser256, Ser308, and Ser424 (5, 22). Among the most well characterized of these phosphosites is Ser81, which regulates AR stability, transcriptional activity, and cellular localization (22, 30, 38, 39). Given Ser81's role in regulating AR stability as well as it being the most stoichemetrically favored phosphosite on the AR (32), we studied its relevance in mediating SMAP-induced AR degradation. Western blot analysis of LNCaP cells treated with 30 μmol/L of SMAP and harvested at 1, 3, 6, 12, and 24 hours showed time-dependent Ser81 dephosphorylation occurring at time-points prior to AR degradation (Fig. 3A and B). In LNCaP cells, AR expression for each time-point evaluated in Fig. 3A was quantified and the ratio of phosphorylated to total AR levels was calculated (Fig. 3B; Supplementary Table S7). This result suggests that AR degradation may occur as a result of PP2A-mediated dephosphorylation of Ser81. Although loss of AR and p-AR appears to begin to occur at the 10–20 μmol/L dose of SMAP at 6 hours, a dose and time-point where we did not observe pronounced evidence of apoptosis, the loss of AR and p-AR was most notable at higher doses of SMAP and/or later time-points. Thus, we cannot completely exclude that the observed effects of SMAP on AR and p-AR expression were related to apoptosis.

We next evaluated AR mRNA in dose and time-course studies. qRT-PCR showed that the decrease in AR protein expression was not accompanied by a simultaneous decrease in AR mRNA in LNCaP or 22Rv1 cells (Fig. 3D), implying that the decrease in AR levels occurs at the protein level rather than as a consequence of a decrease in AR mRNA (e.g., as a result of protein degradation). To define the downstream effects of SMAP treatment on the AR pathway, we next evaluated the transcription of a panel of AR-regulated genes (AR targets were chosen after Clegg and colleagues; ref. 40) upon SMAP treatment. AR target gene mRNA expression was measured by qRT-PCR and confirmed altered expression of multiple AR-regulated genes, including PSA and TMPRSS2 in LNCaP and 22Rv1 cells (Fig. 3D; Supplementary Table S8). These effects occurred in a time-dependent manner, following a pattern consistent with loss of AR transcriptional control. In addition, SMAP treatment induced changes in AR target gene expression (Fig. 3D; Supplementary Table S8).

To determine whether the effects of SMAP on AR protein were a result of AR degradation, we next evaluated AR half-life. LNCaP and 22Rv1 cells were preincubated for 3 hours with SMAP or vehicle control before treating with cycloheximide (μg/mL) for 1, 3, 6, and 9 hours to inhibit protein translation. A consistent decrease in AR half-life in LNCaP and 22Rv1 cells (Fig. 4A) was observed in the SMAP-treated samples. To determine whether AR degradation induced by SMAP was proteasome-mediated, LNCaP and 22Rv1 cells were treated for 6 hours with vehicle control, 30 μmol/L SMAP, or 30 μmol/L SMAP + proteasome inhibitor, bortezomib (100 nmol/L for LNCaP and 1 μmol/L for 22Rv1). Treatment with bortezomib impaired SMAP-induced AR degradation (Fig. 4B; Supplementary Table S9), suggesting that the proteasome pathway was mediating, in part, SMAP-induced AR degradation. To further define the role of PP2A in AR degradation, coimmunoprecipitation of AR protein complexes was performed in LNCaP cells treated with DMSO or 30 μmol/L of SMAP. Significant increased PP2A binding to the AR after 3 and 6 hours of SMAP treatment in LNCaP cells was observed by Western blot analysis (Fig. 4C; Supplementary Fig. S4); an IgG-negative control demonstrated a lack of nonspecific binding of AR or PP2A-C.

To ascertain whether PP2A was mediating SMAP-induced AR degradation, LNCaP cells were stably transduced with the Simian virus 40 (SV40) small T antigen (ST), a potent oncoprotein and specific inhibitor of PP2A (Fig. 5A). The ST alters PP2A's activity by interacting with the PP2A Aα subunit and displacing regulatory B subunits from the dimer. This displacement perturbs the function of PP2A and its activity toward multiple substrates (41, 42). Overexpression of ST in LNCaP cells resulted in an attenuation of the biological activity and target engagement of SMAP as shown by Western blot analysis of cleaved PARP, p-AR (Ser81), and AR (Fig. 5A; Supplementary Table S10), providing further evidence that PP2A is mediating the effects of SMAP on AR degradation and downstream signaling. As AR posttranslational modifications may be influenced by ligand binding, and treatment of CRPC occurs in patients in an androgen-depleted state, we next studied SMAP's ability to degrade AR protein (Fig. 5B; Supplementary Table S11) and mRNA (Fig. 5C; Supplementary Table S12) in LNCaP cells grown in charcoal-stripped FBS (CSS) media and in the presence of an AR agonist (R1881). AR transcriptional and translational degradation was enhanced in LNCaP cells grown in CSS with 30 μmol/L of SMAP when compared with LNCaP cells in FBS media with 30 μmol/L of SMAP, for 3 and 6 hours at the mRNA and protein level (Fig. 5B and C). This enhanced AR mRNA and AR protein degradation in the CSS media was not overcome by the addition of 1 nmol/L of R1881 to SMAP-treated cells (Fig. 5B and C).

To evaluate the antitumor activity of SMAP as a single agent in vivo, we used the LNCaP-AR xenograft (parental LNCaP cells overexpressing WT-AR to model the clinical scenario; ref. 43) model in SCID mice. MDV3100 (enzalutamide) was utilized as a comparator in these studies to assess SMAP activity relative to a current standard of care in the disease. Mice were dosed for 38 days in the following treatment groups: vehicle control (n = 6), 100 mg/kg SMAP twice daily (n = 6), 400 mg/kg SMAP twice daily (n = 6), and 100 mg/kg MDV3100 once daily (n = 5). Treatment in all groups was orally administered. Two doses of SMAP were used to assess whether there was a relationship between dose and antitumor activity. SMAP delivered at 400 mg/kg twice daily resulted in durable stasis (Fig. 6A; Supplementary Table S13) and mice were assessed for signs of toxicity every other day for the duration of the study. No significant weight loss or signs of toxicity (i.e., diarrhea, abdominal stiffness, lethargy) were noted (Fig. 6B). The antitumor activity of the 400 mg/kg SMAP treatment group was comparable with enzalutamide (Fig. 6A–C; Supplementary Table S13). In parallel drug development efforts, a second-generation compound was prepared that exhibited greater potency than SMAP. These compounds, exemplified by SMAP-2, contain a cyclic linker variation (Supplementary Fig. S2; ref. 19). It should be noted that restricting rotatable bonds can increase bioavailability and on-target potency while decreasing off-target effects. Exhibiting similar bioavailability to SMAP, SMAP-2 demonstrated increased potency in several cell lines (Supplementary Fig. S3) prompting its evaluation in mouse models of CRPC. SMAP-2 was tested in two in vivo efficacy studies in both castrated and noncastrated models of prostate cancer in SCID mice harboring LNCaP/AR xenograft tumors. In the noncastrated efficacy study, mice were randomized to two treatment groups based on initial tumor volumes: Control (n = 11) or SMAP-2 100 mg/kg twice daily (n = 8) and treated for 28 days. Tumor volumes and body weights were measured every other day. SMAP-2 demonstrated significant activity as defined by fold change in tumor volume (final tumor volume/initial tumor volume) between the control and SMAP-2 (7.4 vs. 3.6; *, P< 0.01) treated groups without significant weight loss in the drug-treated mice compared with vehicle control–treated mice after one month of dosing (Fig. 6D–F). In the castration study, which was utilized to mimic the clinical scenario in which CRPC patients receive treatment in an androgen-depleted state, male SCID mice were castrated 2 weeks before beginning treatment on either vehicle control (n = 11) or 100 mg/kg SMAP-2 twice daily (n = 9). The SMAP-2 group demonstrated significant activity as defined by fold change in tumor volume as compared with control-treated group (1.4 vs. 6.7, respectively; ***, P< 0.0001; Fig. 6G–I). These studies further established the safety and efficacy of SMAP-2 in prostate cancer in vivo models; no significant weight loss (Fig. 6H), clinical chemistry abnormalities, or behavioral signs were observed in the drug-treated mice after one month of dosing (Fig. 6B).

Upon establishing the safety and efficacy of SMAP-2 in the LNCaP/AR model in vivo, we subsequently explored the pharmacodynamic effects of treatment. Castrated LNCaP/AR tumor-bearing mice were treated for 6 days with SMAP-2, at 30 mg/kg (n = 7) or 100 mg/kg twice daily (n = 8), and vehicle control (n = 12). Remarkably, when the drug formulation was switched from suspension to solution SMAP-2 demonstrated significant activity (Supplementary Fig. S5A). Tumor regressions and no significant weight loss was observed in both the 30 mg/kg and 100 mg/kg SMAP-2 treatment groups after only 6 days of treatment (Fig. 7A–C; Supplementary Table S14), and greater antitumor activity correlated with better exposure in serum analyzed from drug-treated mice (Supplementary Fig. S5B and S5C). Posttreatment serum chemistry panels displayed no adverse effects on liver, kidney, or pancreatic function in SMAP-2–treated mice. Moreover, SMAP-2 treatment decreased proliferation and induced apoptosis as determined by PCNA and TUNEL staining of tumor samples from drug-treated animals, respectively (Fig. 7D–F; Supplementary Tables S15 and S16). Effects on tumor volume correlated strongly with downstream markers of target engagement in vivo as measured by Western blot analysis of AR and PSA protein expression in representative tumor samples (Fig. 7 G–I; Supplementary Tables S17 and S18).

Despite an expanded armamentarium over the past decade for the treatment of CRPC driven by a better understanding of the underlying pathogenesis of the disease, resistance develops to all available therapies contributing to over 27,000 deaths each year in the United States. The AR, including splice variant forms lacking the ligand-binding domain, remains a central therapeutic target in treatment-resistant CRPC though several additional targets have been identified. Novel approaches capable of eliminating AR signaling, while simultaneously inhibiting complimentary oncogenic signaling networks, are critically needed to further improve the outcomes of men with CRPC.

Protein phosphatase 2A is a bona fide tumor suppressor and is implicated specifically in the pathogenesis of CRPC (14). Indeed, dephosphorylation of key oncogenic proteins, such as MYC, AKT1, BCL2, and AR likely explains the critical role of PP2A in prostate cancer. Unlike the majority of tumor suppressors, which are inactivated in cancer due to loss-of-function genetic mutations, mutations of components of the PP2A heterotrimer are relatively rare in human malignancy, with the exception of endometrial cancer. Rather, PP2A function is most commonly decreased as a consequence of decreased subunit expression and/or overexpression of endogenous inhibitors such as CIP2A and SET raising the possibility of therapeutic activation as a novel treatment strategy. Consistent with this hypothesis, knockdown of SET has recently demonstrated anticancer effects in enzalutamide-resistant prostate cancer cell lines and mouse models (44). Protein phosphatases themselves, however, have been largely ignored for drug development because of their perceived “undruggable” nature. We have developed a novel class of molecules, which directly bind PP2A, promote conformational changes, and thereby lead to phosphatase activation (19).

Here, we show that our SMAPs induce anticancer effects in model systems of prostate cancer while modulating known PP2A substrates. In cell culture and in vivo models of prostate cancer, SMAP treatment led to induction of apoptosis. Given the large number of known PP2A substrates important in oncogenesis, we employed a global phosphoproteomic approach to begin to probe the mechanistic basis for the observed anticancer effects.

These studies revealed that the AR was among the most significantly dephosphorylated proteins by SMAP treatment and that members of its signaling axis were downregulated. We focused subsequent experiments on AR, given its central role in CRPC, demonstrating that SMAP induces dephosphorylation, and proteasome-mediated degradation, of the AR (including the splice variant AR-v7) in cell culture. The effects of our SMAP on the AR were confirmed to be PP2A dependent, as expression of the ST abrogated these effects. Finally, a decrease in AR protein expression correlated with the growth-inhibitory effects of SMAPs in prostate cancer xenograft models. The development of therapeutic approaches capable of degrading full-length and splice variant AR could be potentially transformative in the treatment of CRPC.

The large number of possible posttranslational modifications of the AR, including phosphorylation/dephosphorylation of several distinct phosphosites, has complicated an understanding of the global role of these events and, to date, precluded their therapeutic exploitation. Here we show that SMAP treatment results in degradation. Indeed, phosphorylation of this site has previously been shown to regulate AR stability, transcriptional activity, and cellular localization (22, 30, 38, 39). Importantly, all of the known PP2A-AR phosphosites are localized to the N-terminal domain of the AR, potentially explaining the degradation of both full-length and splice variant AR observed in cell culture with SMAP treatment (36). Although the role of SMAP-mediated AR dephosphorylation in inducing AR degradation is supported by our time-course studies and prior studies exploring the functional consequences of AR phosphorylation, the mechanism by which these events target the AR for proteasome-mediated degradation, and the specific phosphosites involved, require further investigation.

In addition, despite the critical role of the AR in driving prostate cancer growth, further studies are required to establish the direct link between AR degradation and the anticancer effects induced by SMAPs. Studies employing AR phosphosite mutants are currently planned to shed additional light on these issues. Indeed, SMAPs may also serve as important tools to advance our general understanding of the functional consequences of posttranslational modifications of the AR.

There are potential limitations to our study. Our phosphoproteomic studies involved a single cell line and a single time-point. The optimal time-point for phosphoproteomic analysis was determined from time course studies performed in cell culture to identify a time at which reproducible changes were detected in phosphorylation yet prior to induction of apoptosis. This approach facilitated the determination of direct drug-mediated phosphorylation changes while limiting the potential secondary changes in the phosphoproteome resulting from the induction of cell death. In spite of this, it is possible that the PP2A substrates identified by such analyses are dependent on the particular PP2A regulatory (B) subunits expressed by a cell line and the timing of the binding and dephosphorylation events. Furthermore, the phosphoproteomic pipeline does not permit distinction between loss of p-AR due to dephosphorylation versus decrease in total AR expression. Still, these studies did identify AR as among the most significantly affected proteins. Decreased p-AR and AR protein expression was observed with SMAP treatment in a dose- and time-dependent manner in LNCaP and 22Rv1 cells; given the induction of apoptosis observed with SMAP treatment, we cannot exclude that the loss of p-AR and AR was related to cell death rather than dephosphorylation and degradation though additional studies with cycloheximide or bortezomib were supportive of the latter. Demonstration of target engagement in our in vivo studies focused on events downstream of PP2A, specifically AR and PSA. While one might consider functional assays of PP2A in tumor or tissue extracts as a means to demonstrate target engagement, such an approach has multiple limitations that preclude accurate and reproducible results including (i) the need to separate and recover PP2A from other phosphatases given the lack of specificity of phosphatase assays and the (ii) high probably of dissociation of SMAPs from PP2A during preparation of extracts given noncovalent bonding.

However, our prior data demonstrating that our SMAPs directly bind and activate PP2A (19), coupled with our current study demonstrating that PP2A inhibition (with small T antigen) abrogated the AR-degrading effects of these molecules in cell culture, does provide support that the pharmacodynamic effects observed in vivo were indeed a result of PP2A activation. The activity with SMAPs in cell culture experiments occurs in the micromolar range as compared with kinase inhibitors, which tend to exert activity in the nanomolar range. There are several reasons for these differences in relative potency. PP2A is a very abundant cellular enzyme with an intracellular concentration of approximately 250 nmol/L and the cell-free–binding affinity of these SMAPs at PP2A A-alpha is 235nmol/L and the stoichiometry of binding is 1:1 (19). Given that protein binding for these compounds is approximately 96%, the free drug concentration (which based upon our studies drives the drug activity) is approximately 1–2 μmol/L for a 10-μmol/L drug dose in 10% serum containing medium. Thus, given the mechanism of action (allosteric agonism), the abundance of the cellular target PP2A, and the protein binding for this drug series, cellular potencies of less than 1 μmol/L are most likely not achievable.

Importantly, PP2A substrates other than the AR may be contributing to the anticancer effects observed with SMAP treatment, including MYC. MYC is one of the best characterized PP2A substrates and PP2A has previously been shown to prime MYC for proteasome-mediated degradation through dephosphorylation of serine 62 (45). In breast cancer models, depletion or inhibition of the endogenous PP2A inhibitors, CIP2A or SET, reduced MYC expression and activity and decreased tumorigenic potential (46). Chromosome 8q, including the MYC gene, is amplified in approximately 30% of human prostate cancers and MYC overexpression recapitulates human prostate carcinogenesis in genetically engineered mouse models (47, 48). Prior reports have shown that MYC is further upregulated in CRPC and contributes to independent prostate cancer cell growth (22–25). Furthermore, MYC has been shown to be a key androgen ligand-independent AR target gene and MYC depletion was demonstrated to reduce prostate cancer cell survival in androgen ligand-depleted conditions (49). Together, these findings raise the hypothesis that simultaneously cotargeting AR and MYC for degradation is responsible for the anticancer effects observed with SMAPs in prostate cancer model systems. Additional studies are planned to define the effect on MYC induced by SMAP treatment. Similarly, the role of other known PP2A substrates previously implicated in CRPC pathogenesis, including AKT1, BCL2, and others warrant further investigation.

D.B. Kastrinsky has ownership interest (including patents) in Dual Therapeutics. D.L. Brautigan reports receiving a commercial research grant from Bristol–Myers Squibb and is a consultant/advisory board member for Dual Therapeutics and Bristol-Myers Squibb. M. Ohlmeyer is a consultant at Dual Therapeutics LLC, reports receiving a commercial research grant from Dual Therapeutics LLC, has ownership interest (including patents) in Dual Therapeutics, and is a consultant/advisory board member for Dual Therapeutics LLC. M.D. Galsky has ownership interest (including patents) in Dual Therapeutics. The Icahn School of Medicine at Mount Sinai, on behalf of G. Narla and M. Ohlmeyer, has filed patents covering composition of matter on the small molecules disclosed herein for the treatment of human cancer and other diseases (International Application Numbers: PCT/US15/19770, PCT/US15/19764; and US Patent: US 9,540,358 B2). RAPPTA Therapeutics LLC has licensed this intellectual property for the clinical and commercial development of this series of small-molecule PP2A activators. G. Narla, M. Ohlmeyer, and M.D. Galsky have an ownership interest in RAPPTA Therapeutics LLC. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. McClinch, R.A. Avelar, D. Callejas, M. Cooper, D. McQuaid, A.C. Levine, M. Ohlmeyer, G. Narla, M.D. Galsky

Development of methodology: K. McClinch, R.A. Avelar, D. Callejas, S. Izadmehr, J. Sangodkar, D. Schlatzer, M. Cooper, R.C. Sears, M. Ohlmeyer, G. Narla, M.D. Galsky

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. McClinch, R.A. Avelar, S. Izadmehr, A. Perl, J. Sangodkar, D. Schlatzer, M. Cooper, J. Kiselar, A. Stachnik, S. Yao, D. Hoon, S.R. Plymate, A.C. Levine, A. DiFeo, G. Narla, M.D. Galsky

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. McClinch, R.A. Avelar, D. Callejas, S. Izadmehr, D. Wiredja, A. Perl, J. Sangodkar, D. Schlatzer, M. Cooper, A. Stachnik, S.R. Plymate, A.C. Levine, A. Kirschenbaum, J.P. Sfakianos, G. Narla, M.D. Galsky

Writing, review, and/or revision of the manuscript: K. McClinch, R.A. Avelar, S. Izadmehr, D. Wiredja, A. Perl, J. Sangodkar, D.B. Kastrinsky, M. Cooper, A. Stachnik, D. McQuaid, D.L. Brautigan, S.R. Plymate, C.C.T. Sprenger, W.K. Oh, A.C. Levine, J.P. Sfakianos, A. DiFeo, G. Narla, M.D. Galsky

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. McClinch, S. Izadmehr, J. Sangodkar, D.B. Kastrinsky, S. Yao, Y. Gong, G. Narla, M.D. Galsky

Study supervision: R.A. Avelar, Y.A. Ioannou, G. Narla, M.D. Galsky

Other [involved in chemical synthesis to generate small-molecule activators of protein phosphatase-2A (SMAP)]: N. Zaware

Other (design and synthesis of small-molecule PP2A activators that are the subject of paper): M. Ohlmeyer

We gratefully acknowledge the New York City Investment Fund (Bioaccelerate Prize) and Dual Therapeutics/BioMotiv for continued support. G. Narla received a Howard Hughes Medical Institute (HHMI) Physician-Scientist Early Career Award and a Case Comprehensive Cancer Center Pilot Award. S. Izadmehr is a LRP awardee (NCATS/NIH). This work was partially supported by an NCI/NIH R01 grant 1R01CA181654-01A1 (G. Narla, M.D. Galsky, M. Ohlmeyer, and A.C. Levine), DoD grant W81XWH-15-1-0596 (M.D. Galsky), and Prostate Cancer Foundation Young Investigator Award (M.D. Galsky).

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.