Dual targeting of the androgen receptor and PI3K/AKT/mTOR pathways in prostate cancer models improves antitumor efficacy and promotes cell apoptosis

2.1. Compounds

R1881 (synthetic androgen, also known as metribolone) and copanlisib were manufactured in‐house. Darolutamide was synthesized at Orion Corporation (Espoo, Finland). Apitolisib (S2696), dactolisib (S1009), capivasertib (S8019), and camptothecin (S1288) were from Selleckchem (Planegg, Germany). Alpelisib (T1921), duvelisib (T1988), and idelalisib (T1894) were from TargetMol (Wellesley Hills, MA, USA). Ipatasertib (HY‐15186) was from MedChemExpress (Monmouth Junction, NJ, USA).

2.2. Cell lines

The VCaP (RRID:CVCL_2235), LNCaP (RRID:CVCL_0395), 22Rv1 (RRID:CVCL_1045), PC‐3 (RRID:CVCL_0035), DU145 (RRID:CVCL_0105), and HeLa (RRID:CVCL_0030) cancer cell lines were acquired from the ATCC (American Type Culture Collection, Manassas, VA, USA). DNA fingerprinting was performed at the DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen, Braunschweig, Germany) for authentication. Mycoplasma contamination testing was carried out using MycoAlert (Lonza, Cologne, Germany).

2.3. Determination of target binding affinity and kinetic parameters

The principle of the homogeneous kinetic probe competition assay (kPCA) has been described [42], and detailed protocols for PI3Kα reported [43]. For the surface plasmon resonance (SPR) assay, PI3Kα was captured on a Biacore streptavidin‐CM5 chip and compound binding at increasing concentrations was analyzed in real time with a Biacore T200 SPR instrument (both from GE Healthcare Life Sciences, Uppsala, Sweden). Compound association and dissociation traces were fitted to a 1 : 1 Langmuir model to obtain the k _on and k _off kinetic parameters of the interaction. Steady‐state resonance unit signals from the same experiment were plotted against the compound dose to calculate the equilibrium affinity using a 1 : 1 binding model derived from the law of mass action.

The intracellular inhibitor potency for PI3Kα was evaluated using a probe displacement assay based on the NanoBRET technology [44]. HeLa cells transiently expressing NanoLuc‐labeled PI3Kα and cotransfected with the p85 regulatory subunit were equilibrated with increasing compound concentrations in the presence of a fixed amount (40 nm) of the Kinase Tracer 03 (Promega Corporation, Madison, WI, USA). Following the addition of the compounds and tracer, the cells were incubated for 3 h before the addition of Complete NanoBRET Nano‐Glo substrate plus extracellular NanoLuc inhibitor (both from Promega Corporation). Subsequent bioluminescence resonance energy transfer measurements were made on a GloMax Discover instrument (Promega Corporation). Competitive displacement curves for the test compound were fitted to the sigmoidal dose–response function to calculate apparent intracellular IC₅₀ values at each tracer concentration. The apparent intracellular IC₅₀ values were replotted as a function of tracer concentration, and the data were fitted to the linearized Cheng–Prusoff equation to determine the intracellular inhibition constant (K _i). Washout experiments were performed following a previously described protocol in the presence of 40 nm Kinase Tracer 03 [43].

2.4. In vitro antiproliferative impact and apoptosis induction

Proliferation experiments were performed in RPMI without phenol red, 10% charcoal‐stripped fetal bovine serum, and in the presence of 0.1 nm R1881, except for LNCaP cells where 1 nm R1881 was added. A dose range of inhibitor concentrations was tested and cell viability measured after 4 days (PC‐3, DU145) or 6 days (VCaP, LNCaP, 22Rv1) using the CellTiter‐Glo luminescent assay (Promega Corporation). For the determination of the combination index (CI), viability assays were performed in starved VCaP cells subsequently treated for 6 days with 0.1 nm R1881 and different ratios of darolutamide and copanlisib or apitolisib. The respective IC₅₀ values were calculated and used to determine the CI.

Apoptosis was evaluated following a 48‐h treatment with a dose range of inhibitor concentrations by measuring caspase 3/7 activation (CaspaseGlo 3/7 assay; Promega Corporation) or PARP cleavage (PARP Cleaved 214/215 ELISA assay; Invitrogen/ThermoFisher, Waltham, MA, USA).

The expression of the Bcl‐2‐binding component 3 (BBC3) gene involved in apoptosis was determined using RNA extracted from treated VCaP cells using the TaqMan Fast Advanced Master Mix 4444557 (Applied Biosystems, Waltham, MA, USA) followed by real‐time PCR with primers specific for BBC3 (Hs00248075_m1; Life Technologies, Carlsbad, CA, USA) in a 7900 HT Fast Real‐Time PCR system (Applied Biosystems). Human cyclophilin A was used as endogenous control (4326316E; ThermoFisher).

2.5. Sandwich ELISA

H2A.X phosphorylation at serine 139 was determined by sandwich ELISA. Briefly, VCaP cells were grown in RPMI without phenol red, 10% charcoal‐stripped fetal bovine serum for 24 h, and then treated for 48 h with 0.03 nm R1881 and two different concentrations of copanlisib and darolutamide, alone or in combination. Cells were then washed with ice‐cold phosphate‐buffered saline and incubated for 5 min with lysis buffer, scraped, and transferred into sonication tubes. Sonication was performed on ice for 2 min twice, followed by centrifugation. The supernatants were analyzed by sandwich ELISA using the PathScan Phospho‐Histone H2A.X (Ser139) kit #50929, following the manufacturer's instructions (Cell Signaling Technology, Danvers, MA, USA). The chromogenic substrate 3,3′, 5,5′‐tetramethylbenzidine was used to determine peroxidase activity, which was measured at 450 nm absorbance in a Tecan Infinite microplate reader (Tecan Group, Männedorf, Switzerland).

2.6. Transcriptomic data analysis

VCaP cells were treated for 24 or 48 h with dimethyl sulfoxide (DMSO) only, 0.1 nm R1881, 0.1 nm R1881 and 1 μm darolutamide, 0.1 nm R1881 and 100 nm copanlisib, or 0.1 nm R1881, 1 μm darolutamide and 100 nm copanlisib. Three samples were profiled for each treatment and time point. RNA isolation and sequencing was performed as previously described [45].

FASTQ reads were mapped via STAR aligner to the human genome GRCh38 and quantified with RSEM (v1.3.0; University of Wisconsin‐Madison, Madison, WI, USA). Only protein‐coding genes with at least five transcripts per million (TPM) in three samples or more were used for further differential gene expression analysis. Differentially expressed genes were identified with DESeq2. Benjamini–Hochberg correction for multiple hypothesis testing was used to define false discovery rate (FDR). Genes with at least an expression fold‐change of 2 and FDR below 10% were considered differentially expressed. Single‐sample gene set enrichment analysis (ssGSEA) was performed with the Gene Set Variation Analysis package from R Bioconductor using TPM values and the hallmark gene sets collection from MSigDB [46].

For comparison with human expression, data from the TCGA Prostate Adenocarcinoma study [47] were downloaded from the Genomics Data Commons portal and differential expression between tumor and normal tissue was assessed with the Wilcoxon test.

2.7. In vivo efficacy study

Fox Chase SCID mice (CB17/Icr‐Prkdcscid/IcrIcoCrl) were obtained from Charles River Laboratories (Wilmington, MA, USA). They were maintained in a Specific Pathogen‐Free vivarium throughout the study. For the efficacy study, 5–8‐week‐old male mice were implanted subcutaneously with LuCaP 35 tumor bits, as previously described [41]. Once tumor volume reached 100 mm³, mice were randomized and treatment was started with darolutamide (200 mg·kg⁻¹, p.o.; 5‐day on, 2‐day off), copanlisib (14 mg·kg⁻¹, i.v.; 2‐day on, 5‐day off) or with both compounds combined. Vehicle and castration groups were used as controls. Subcutaneous tumor growth was measured twice weekly using a digital caliper and tumor volume was calculated using the formula: length × width × height × 0.5236. Animals were euthanized when tumors exceeded 1000 mm³, if body weight loss reached 20%, or if animals became otherwise compromised as defined in the University of Washington's Institutional Animal Care and Use Committee protocol 3202‐01. Statistical analysis was performed by ANOVA followed by Dunn's test.

At the end of the in vivo phase, tumors were harvested, fixed in formalin and embedded in paraffin. Immunohistochemistry (IHC) analysis was performed on tumor microarrays with 1.5 mm cores which were sectioned (3 μm) and stained with monoclonal antibodies specific for cleaved caspase 3 (clone 5A1E, #9664, 66 ng·mL⁻¹; Cell Signaling Technology, Cambridge, UK) or BBC3 (clone E2P7G, #98672, 500 ng·mL⁻¹; Cell Signaling Technology).

All animal procedures used in this study were approved by the Institutional Animal Care and Use Committee at the University of Washington in accordance with NIH guidelines. Approved animal license number for the study: University of Washington IACUC protocol 3202‐01.

2.8. Data generation and statistical analysis

Biochemical and cell‐based binding studies were performed in at least two independent experiments (N ⩾ 2) with two technical replicates each (n = 2). Cellular assays were conducted with at least two biological replicates (N ⩾ 2) and representative results are shown. Measurement for each treatment condition was done in three different wells (n = 3) for the apoptosis and real‐time PCR assays, and in two different wells for the sandwich ELISA (n = 2). Statistical analysis was performed using one‐way ANOVA and Dunnett's multiple comparisons test. Differential gene expression of TCGA Prostate Adenocarcinoma data was assessed by the Wilcoxon test.

3.1. PI3K/AKT/mTOR inhibitors with different selectivity profiles inhibit proliferation and induce apoptosis of androgen‐sensitive prostate cancer cell lines

Previous gene expression analysis of VCaP cells treated with the androgen R1881 revealed that beside the expected expression stimulation of androgen target genes, there was also a marked upregulation of the hallmark gene signature of the PI3K/AKT/mTOR pathway [45, 48]. We therefore determined how inhibitors of this pathway with different selectivity profiles affected prostate cancer cell proliferation. PI3K inhibitors, dual PI3K/mTOR inhibitors, and AKT inhibitors were evaluated to have a broad overview. Isoform‐selective effects of the different PI3K inhibitors have previously been described but detailed, comparative studies of binding affinities and kinetics for the essential PI3Kα isoform have so far not been reported. PI3Kα plays a crucial function in many cancer types including prostate cancer, especially in mCRPC where the corresponding gene is frequently amplified [13]. Biochemical kPCA and biophysical SPR assays, and a cellular target engagement assay were applied to establish the PI3Kα binding affinity and kinetic map for selected inhibitors. The steady‐state compound affinities measured with biochemical and biophysical methods were in excellent agreement (Fig. S1A). Also, the comparison between steady‐state K _i and kinetic K _D values showed a strong correlation (Fig. S1A). The six compounds investigated bound to PI3Kα with k _on association rates between 10⁵ and 10⁸  m ⁻¹·s⁻¹ and k _off dissociation rates between 10⁻³ and 10⁻¹·s⁻¹, with residence times between 10 s and 17 min (Fig. S1B). Independently of the method used, copanlisib was one of the fastest equilibrating and slowest dissociating compounds, and it ranked among the most potent PI3Kα binders. Next, we determined the intracellular PI3Kα inhibitor potency in living cells. We used the NanoBRET cellular probe displacement assay and found K _i values between about 3 and 500 nm for the compounds evaluated, with copanlisib being the most potent one with a K _i of 3.07 ± 1.18 nm (Fig. S1C). Comparison with the biochemical affinities revealed the cellular affinities to be 2‐ to 10‐fold weaker, most likely due to parameters such as cell permeability, cellular ATP concentrations and intracellular regulatory mechanisms which were not captured in the kPCA or SPR assays. The cellular kinetic displacement assay was then performed in washout mode, and we found that PI3Kα engagement by copanlisib and dactolisib lasted significantly longer than expected from the respective binding rate constants of the compounds (Fig. S1D), suggesting a differentiated mechanism of protracted residence time in living cells. Further investigations of this discrepancy between biochemical and cellular target occupancies led to the discovery of unique micropharmacokinetic properties of copanlisib (not shown) that are currently being further explored.

Copanlisib and dactolisib, and two additional compounds targeting AKT, were then assessed in five prostate cancer cell lines with a different PTEN gene status: wild‐type for VCaP and 22Rv1, heterogenous point mutation for DU145, homozygous deletion for LNCaP and PC‐3 [49, 50, 51]. These cells also exhibit different AKT phosphorylation levels with LNCaP and VCaP cells showing comparatively higher serine 473 phospho‐AKT to AKT signal ratios than 22RV1, PC‐3, or DU145 cells [51, 52, 53]. The pan‐PI3K inhibitor copanlisib and the dual PI3K/mTOR inhibitors apitolisib and dactolisib showed potent antiproliferative effects in all cell lines evaluated (Fig. 1). The AKT inhibitors capivasertib and ipatasertib had a strong impact mainly in androgen‐sensitive cell lines. The dual PI3Kδ/γ inhibitor duvelisib had marked effects in LNCaP cells, whereas the α and δ isoform‐selective compounds alpelisib and idelalisib were generally less active. As a positive control we evaluated the topoisomerase inhibitor camptothecin which inhibited all prostate cancer cell lines tested, as expected (Fig. 1). The IC₅₀ values calculated for each compound in each of the five cell lines are shown in Table S1.

We next investigated the effects of the compounds on apoptosis by determining caspase 3/7 activity in the treated cells (Fig. 2). Here also we found that the pan‐PI3K inhibitor copanlisib and the dual PI3K/mTOR inhibitors apitolisib and dactolisib had the strongest effects in the androgen‐sensitive cell lines VCaP and LNCaP. The AKT inhibitors capivasertib and ipatasertib, and the dual PI3Kδ/γ inhibitor duvelisib showed a comparable pro‐apoptotic impact in LNCaP cells and a slightly weaker effect in VCaP cells. The isoform‐selective PI3Kα inhibitors alpelisib and idelalisib were slightly less potent. Apoptosis induction was observed in all cell lines evaluated at the highest camptothecin concentration used (Fig. 2).

3.2. Combining PI3K/AKT/mTOR inhibitors with the AR inhibitor darolutamide has pro‐apoptotic effects in androgen‐sensitive prostate cancer cell lines

As the VCaP cell line showed the most pronounced response, we chose it for an in‐depth evaluation of the impact of the combination treatment with PI3K/AKT/mTOR and AR inhibitors. When adding darolutamide to copanlisib or ipatasertib, a dose‐ and time‐dependent increase of caspase 3/7 cleavage was observed, with the 48‐h regimen leading to the most pronounced induction effects (Fig. S2). We then combined darolutamide with different PI3K/AKT/mTOR pathway inhibitors and compared apoptosis induction at 48 h (Fig. 3A). Pronounced effects were observed when combining darolutamide with low copanlisib, apitolisib or dactolisib concentrations (0.01, 0.1 and 0.1 nm, respectively). The co‐treatment with alpelisib, duvelisib or idelalisib also had a significant apoptotic impact, but at higher concentrations (at least 1 nm). This was also the case for the AKT inhibitors capivasertib and ipatasertib where 0.3 nm and higher concentrations were necessary for a significant effect. We performed similar experiments using the LNCaP cell line by combining darolutamide with copanlisib or ipatasertib, as representative PI3K and AKT inhibitors, respectively, and also observed induction of apoptosis (Fig. 3B). These results were confirmed by the evaluation of PARP cleavage in VCaP and LNCaP cells treated with darolutamide and copanlisib (Fig. 3C,D). We then examined a range of genes involved in apoptosis and found that expression of the pro‐apoptotic BBC3 gene was upregulated by copanlisib treatment, and this was most pronounced upon combination with darolutamide in androgen‐stimulated VCaP cells (Fig. S3).

3.3. In vitro combination treatment with copanlisib and darolutamide has superior antiproliferative activity and comprehensively reverses the androgen response and mTOR transcriptional programs

We then determined the antiproliferative impact of different copanlisib and darolutamide combinations on VCaP cells by measuring viability after 6 days (Fig. S4). Moderate synergistic effects with a CI of 0.82–0.84 were calculated. Comparable results were also obtained when combining the PI3K/mTOR inhibitor apitolisib with darolutamide (Fig. S4).

Next, we performed a transcriptomic study of the impact of copanlisib and darolutamide as single agents or in combination, at two different time points, on the strongly androgen‐sensitive VCaP cell line. The results of principal component analysis (PCA) are outlined in Fig. 4A. A clear effect of androgen stimulation was visible along the first principal component, with a strong difference between DMSO‐treated and R1881‐stimulated samples. The combined R1881 and copanlisib treatment was only marginally different from R1881, while the R1881 and darolutamide as well as the R1881, darolutamide and copanlisib samples were closer to the DMSO control along the first principal component. The combined R1881, darolutamide and copanlisib treatment was the closest to the DMSO control, indicating the most efficient suppression of R1881‐induced transcriptional changes, which is consistent with the superior effect on proliferation observed. The difference between the time points was aligned along the second principal component for all the treatments. Pairwise comparison of gene expression changes between the different conditions revealed that the combined treatment with R1881, darolutamide and copanlisib led to the lowest number of genes being differentially expressed compared with DMSO (Fig. 4B). This number decreased over time, suggesting strong time‐dependent inhibition of androgen‐induced transcription (Fig. 4B).

In order to understand the pathways affected, we performed ssGSEA on all the samples using the hallmark gene sets collection from MSigDB. We selected gene sets with enrichment above 1.5‐fold in at least one sample for further visualization and exploration. The enrichment of selected hallmark gene sets was visualized in a heatmap (Fig. 5A). The results show that the androgen response hallmark was clearly upregulated by R1881 treatment as expected. The addition of copanlisib alone had only a moderate effect on androgen response gene set enrichment, while darolutamide and the darolutamide plus copanlisib cotreatment notably decreased it. Hierarchical clustering of the expression of genes from the hallmark androgen response gene set revealed a clear clustering of the copanlisib‐treated condition with the androgen‐stimulated one, while the darolutamide and the darolutamide plus copanlisib combination conditions assembled near the DMSO control, which demonstrates the efficient inhibition of classical androgen target genes by either of these treatments (Fig. 5B). We also observed a notable regulation of the hallmark gene set of mTORC1 signaling. It was upregulated by androgen; however, the addition of darolutamide alone did not suppress it, while copanlisib alone or in combination with darolutamide efficiently downregulated it (Fig. 5A). This was also evident upon hierarchical clustering of the expression of genes from this set (Fig. 5C). This analysis revealed that despite most of the mTORC1 pathway genes being regulated by androgen, only the combination of copanlisib and darolutamide reverted this feature close to the profile seen in the DMSO control, whereas the individual darolutamide or copanlisib treatment had less impact, highlighting the additional effect of the combination.

The described effects were exemplified by comparing the expression levels of individual genes in the treated groups. Androgen response gene set representatives encoding acyl CoA synthetase long‐chain family member 3 (ACSL3), FK506‐binding protein prolyl isomerase 5 (FKBP5), transmembrane protease, serine 2 (TMPRSS2) and mTORC1 pathway representatives encoding elongation of very long chain fatty acid protein 6 (ELOVL6), emopamil‐binding protein (EBP), and 7‐dehydrocholesterol reductase (DHCR7) are shown (Fig. S5A,B).

Notably, gene expression of mTOR itself and of its direct effectors, the translation initiation factors 4 [54], was upregulated by androgen and strongly repressed by darolutamide and copanlisib combination treatment (Fig. S6A). We also examined the expression of mTOR and of the eukaryotic translation initiation factors (EIF) 4E2 and 4G1 in clinical samples and found it to upregulated in prostate cancer compared with normal prostate tissue (Fig. S6B), which further supports the role of this pathway in prostate cancer oncogenesis. Concerning AR expression, similar transcript levels were observed in the androgen plus darolutamide, and also in the androgen plus darolutamide and copanlisib group, slightly lower than those seen in the DMSO only group (Fig. S5C).

3.4. In vitro combination treatment with copanlisib and darolutamide stimulates H2A.X serine 139 phosphorylation

AR function is directly linked to faithful DNA repair via regulation of DNA damage response genes [55, 56, 57]. As we observed a marked impact of the copanlisib and darolutamide combination on cell viability, we additionally determined the impact of the treatments on DNA double‐strand breaks by measuring H2A.X serine 139 phosphorylation (Fig. 6). The results show that individual copanlisib application had no impact on H2A.X phosphorylation. Conversely, darolutamide treatment increased H2A.X phosphorylation, especially at the higher concentration used. The effect was significantly enhanced when darolutamide was combined with copanlisib, indicating that DNA double‐strand breaks are involved in the antiproliferative effects we observed.

3.5. In vivo combination treatment with darolutamide and copanlisib enhances antitumor efficacy and leads to apoptosis

We then set out to investigate whether the synergistic effects observed translated to the in vivo situation. First, small studies were conducted with multiple prostate cancer PDX models in order to identify those most likely to respond to the combination treatment (data not shown). We selected the LuCaP 35 PDX model due to its moderate response to individual copanlisib or darolutamide treatment. It expresses wild‐type AR and harbors a biallelic PTEN deletion [58]. Single‐compound regimen had only a minor impact on tumor growth (Fig. 7A). The darolutamide and copanlisib cotreatment showed significantly superior efficacy to monotherapy from 5 weeks after dosing start (Fig. 7A). Immunohistochemistry analysis of tumor samples taken at the end of the experiment revealed increased levels of cleaved caspase 3 and BBC3 in the combination group, compared with the vehicle group (Fig. 7B).