The Efflux Transporter ABCG2 Maintains Prostate Stem Cells

This article is featured in Highlights of This Issue, p. 115

The ATP binding cassette sub-family G member 2 (ABCG2) transporters belong to the ATP binding cassette (ABC) superfamily of transporters and use ATP hydrolysis to catalyze the transport of a wide range of substrates ranging from chemotherapeutic drugs, steroids, and xenobiotics from the intracellular to the extracellular milieu. The side population assay is a flow cytometric method that isolates stem cells based on the dye-efflux properties of the ABC transporters and ABCG2 is the molecular determinant of the stem cell–enriched side population phenotype in the hematopoietic, retinal, cardiac, limbal, and skeletal muscle system (1–5). The importance of ABCG2 to stem cell maintenance can be adjudged from the observations that ABCG2 expression is lost as stem cells undergo differentiation (1, 2, 6). In fact, overexpression of ABCG2 has been shown to promote stemness while the loss of ABCG2 expression has been shown to promote lineage commitment in the retinal, hematopoietic, and cardiac side population cells (1, 2, 7). While the importance of ABCG2 in stem cell maintenance has been shown, the precise mechanism by which ABCG2 prevents stem cells from undergoing differentiation is unknown.

An ABCG2-expressing side population is present in normal and cancerous prostate tissues and cell lines (8). ABCG2 expression is found in approximately 1% of the cells in the prostate basal compartment (9, 10). ABCG2-expressing cells are present in biopsies following androgen deprivation therapy (ADT)(9) and are upregulated upon androgen blockade in culture (9, 11). Although, there are reports of ABCG2-effluxing bicalutamide (12), there have been no studies that show enzalutamide efflux by ABCG2. Moreover, ABCG2-expressing side population in the prostate demonstrates multipotency and self-renewal properties, thus suggesting an enrichment of stem cells in this population (13). Transcriptome analysis of ABCG2-expressing prostate cells confirmed the expression of stem cell–associated genes such as CD133, Oct4, and nestin and the lack of androgen receptor (AR) expression which is typical of prostate stem cells (PSC; refs. 9, 10, 14). Although, ABCG2 transporters are suspected to play a role in maintaining the prostate stem cell phenotype, the mechanistic significance of ABCG2 to stem cell maintenance is largely unknown. Elucidating the mechanisms by which PSCs maintain their stemness will provide opportunities to target them in prostate cancer therapy. PSCs are implicated in the etiology of castration-resistant prostate cancer (CRPC) and are resistant to ADT (15). In fact, ABCG2-expressing prostate cells have been shown to not only survive ADT but are amplified in CRPC (9, 11), thus suggesting a role for the transporters in conferring ADT-resistance to PSCs and in the etiology of CRPC, in addition to stem cell maintenance.

Previous studies showed that mitoxantrone-resistant rat progenitor epithelial (Mx-RPE) cells, which have been selected for based on high ABCG2 expression, had elevated retention of radiolabeled DHT when treated with the ABCG2 inhibitors, novobiocin and fumitremorgin C, thus, suggesting that ABCG2 can regulate androgen levels (9). Several studies have demonstrated that the exogenous introduction of AR in prostate cells that either lack or have very low levels of endogenous AR such as PC3, HPr-1, and PSCs forces cells to undergo differentiation to a luminal phenotype (14, 16, 17). Thus, we hypothesized that ABCG2 plays a critical role in maintaining PSCs in an undifferentiated state via the efflux of androgens and inhibiting the ABCG2-mediated androgen efflux would force the ABCG2-expressing PSCs to undergo an AR-modulated differentiation to a luminal phenotype. In the current study, we investigate the effects of pharmacologic inhibition of ABCG2 on cellular differentiation in the ABCG2-expressing HPr-1-AR (nontumorigenic prostate epithelial cell line) and the CWR-R1 (tumorigenic CRPC cell line derived from CWR22 xenografts) cells by sphere-formation assays and immunoblots. In vivo differentiation is assessed by treating nude mice implanted with ABCG2-expressing CWR-R1 cells with either placebo or Ko143. Inducing luminal differentiation in PSCs and sensitizing them to ADT is the basis of differentiation therapy. Our findings provide a rationale for targeting ABCG2 for differentiation therapy in prostate cancer.

The HPr-1-AR cells were obtained from Dr. Dominic Smiraglia and cultured in keratinocyte-serum-free media (Thermo Fisher Scientific) supplemented with 1× penicillin-streptomycin. CWR-R1 cells were obtained from Dr. James Mohler and were cultured in Improved MEM supplemented with 2% FBS (FBS), 1× penicillin–streptomycin, 0.2× insulin-transferrin-selenium (ITS; Life Technologies), 0.9 mg/L linoleic acid (Sigma-Aldrich), and 100 μg/mL EGF (Thermo Fisher Scientific). Our preliminary data shows that prolonged charcoal stripping leads to the downregulation of ABCG2 membrane expression in mitoxantrone-resistant CWR-R1 cells (data not shown). Thus, the ABCG2-expressing CWR-R1 cells were cultured in FBS-supplemented media. All cells were maintained in an atmosphere of 5% CO₂ at 37°C. All experiments were done before cells reached passage 65. Treatments were administered in the appropriate media such that DHT (Sigma) was used at a concentration of 0.1 nmol/L for the HPr-1-AR cells and 1 nmol/L for the CWR-R1 cells. Ko143 (Axon chemicals) and enzalutamide (Selleck Chemicals) were administered at a concentration of 1 μmol/L each.

Cells were grown to 75% confluency and the HumanSep FITC Selection Kit (StemCell Technologies) was used in conjunction with the anti-BCRP antibody, clone 5D3, FITC-conjugated (Millipore) to isolate the ABCG2-expressing cells as per the manufacturer's protocol.

Cells were seeded onto coverslips, fixed with 10% formalin, and permeabilized with 0.1% Triton X. Cells were blocked with 1% BSA and immunostained with the anti-AR antibody (1:200, Millipore). The goat anti-rabbit antibody, Alexa Fluor 594 conjugate (1:1,000, Invitrogen) was used as a secondary antibody. Coverslips were mounted onto slides with Vectashield mounting medium containing DAPI (Vector Laboratories) for analysis under an immunofluorescent microscope.

Intracellular androgens from cells treated with [³H]DHT (Perkin Elmer) were extracted using methyl tert butyl ether (Sigma-Aldrich) and reconstituted in a mixture of 9:1 chloroform:acetone. The reconstituted mixture along with steroid standards (Steraloids, Inc) were loaded onto 20 cm × 20 cm silica-coated TLC plates (Millipore). The 9:1 chloroform:acetone mixture was used as the mobile phase to separate the steroids. The standards were visualized on the TLC plate by spraying the plate with p-Anisaldehyde stain. Zones corresponding to the reference steroid standards were scraped into vials containing the BetaMax Scintillation Cocktail (MP Biomedicals) and tritium levels were measured using a liquid scintillation counter.

RIPA buffer (Invitrogen) was used to prepare cell lysates and protein concentration was measured by the BCA Protein Assay Kit (Thermo Scientific). Cell lysates (20 μg) were loaded in the wells of Novex NuPAGE 4%–12% Bis-Tris Gels (Thermo Scientific). Electrophoresed proteins from gels were transferred onto polyvinylidene difluoride membranes. Membranes were blocked and stained with anti-AR (1:250, Millipore), anti-Sox2 (1 μg/mL, Millipore), p21 (1:200, C-19; Santa Cruz Biotechnology), PSA (1:500, Dako), Cytokeratin-8 (1:500, Covance), and β-actin (1:10,000, BioVision Inc.) antibodies and counterstained with the appropriate horseradish peroxidase–conjugated secondary antibodies (GE Healthcare). Signal from the secondary antibodies was detected using an enhanced chemiluminescence detection system (Pierce).

Cells were plated at a density of 2,000 cells per well or 4,000 cells per well (after xenograft tumor digestion) in 24-well ultralow attachment plates (Sigma-Aldrich) coated with Matrigel (Thermo Fisher Scientific). Cells in media were suspended in Matrigel at a ratio of 2:3 (cells in media volume: Matrigel volume) and this suspension was plated onto the edges of the wells of the plates and allowed to solidify at 37°C for 45 minutes. Upon solidification, treatments were administered and the number and sizes of spheres were quantitated 10–14 days later. Sphere area was measured using the Basic SPOT imaging software. The CWR-R1 spheres were maintained in improved MEM, original growth media, whereas the HPr-1-AR spheres were cultured in prostate epithelial cell growth medium (Lonza).

RNA was isolated using the RNeasy Mini Kit (Qiagen) and cDNA was synthesized using the SuperScript III Reverse Transcriptase Kit (Invitrogen) and oligodT primers. Gene expression was determined using SYBR Green PCR mix (Bio-Rad). Primer sequences are enlisted in Supplementary Table S1. qRT-PCR was performed on the Applied Biosystems 7300 machine. The following conditions were used to carry out the RT-PCR reactions: 300 seconds at 95°C; 50 cycles of 10 seconds at 95°C; 20 seconds at 60°C and 20 seconds at 72°C.

Twenty-six male nude mice (Harlan Sprague Dawley Inc.) aged 4–5 weeks were castrated and implanted with a testosterone silastic tubing to normalize the serum testosterone levels to 4 ng/mL. Mice were subcutaneously injected with 1 × 10⁶ ABCG2-expressing CWR-R1 cells suspended in Matrigel (1:1 ratio). Seven days postinoculation; mice were subdivided into two groups of 13 mice each. The first group of mice received placebo while the second group of mice received 15 mg/kg of Ko143 by oral gavage. Treatments were administered three times a week and tumors were measured at the time of drug delivery. Tumor volume was measured by digital calipers and calculated using the formula L₁×(L₂)²×0.523. Once tumors reached 1,500 mm³ volume, mice were euthanized and tumors were excised for further analysis.

Three mice from each group with multiple tumors originating from the same injection site or whose tumors failed to reach a volume of 250 mm³ on day 45 of cell inoculation were excluded from the study. Linear Mixed Model analysis was used to compute differences in the tumor growth rates between control and Ko143-treated animals. Data obtained from in vitro experiments was analyzed either by one-way or two-way ANOVA tests. All data analyses were generated using SAS/STAT software, Version 9.4.

IHC was done on one section of the paraffin-embedded tumors. Tissue specimens were immunostained with primary antibodies: anti-Cytokeratin-8 (1:1000, Covance), anti-BCRP antibody, clone BXP-21 (1:500, Millipore) and anti-AR antibody (1:200, Millipore). Biotinylated horse anti-mouse IgG antibody (1:1000, Dako) and biotinylated goat anti-rabbit (1:1,000, Abcam) were used as the secondary antibodies as appropriate. Vectastain Elite ABC HRP kit (Vector Laboratories) was used to visualize the immunostained tissue sections. Staining was performed on Dako Omni. Images were scanned using the Aperio scanner.

On the basis of previous studies showing the increased retention of radiolabeled DHT in Mx-RPE cells upon incubation with Novobiocin and Fumitremorgin C (9), we propose that inhibiting ABCG2 in the presence of DHT would increase the nuclear AR expression due to increased intracellular retention of androgens. Studies were conducted in the HPr-1-AR and the CWR-R1 cell lines. The HPr-1-AR and CWR-R1 cell lines were chosen due to the presence of functional ABCG2 in a subpopulation and an androgen-responsive phenotype. In addition, the HPr-1-AR and the CWR-R1 cells offer comparison between nontumorigenic and tumorigenic cells respectively. Optimal DHT concentrations for treatment were determined by assessing growth curves of the unsorted HPr-1-AR and CWR-R1 cells in response to DHT (Supplementary Fig. S1A and B) and TLC analysis of the intracellular androgens extracted from unsorted HPr-1-AR and CWR-R1 cells treated with increasing concentrations of [³H]DHT (Supplementary Fig. S1C and D). TLC analysis was performed to determine whether DHT is metabolized by hydroxysteroid dehydrogenase or aldo-keto reductase expressed in prostate cells. On the basis of these studies, a concentration of 0.1 nmol/L DHT was used for the HPr-1-AR cells (Supplementary Fig. S1A and C) and 1 nmol/L DHT was used for the CWR-R1 cells (Supplementary Fig. S1B and D) in all further experiments unless otherwise mentioned as these were the lowest concentrations of DHT at which there was both, a retardation in the growth response as well as an observable change in the amount of intracellular androgens. Using DHT concentrations greater than the ones used would overwhelm the ABCG2 transporters and lower concentrations would be undetectable.

To determine the effects of inhibiting the ABCG2-mediated androgen efflux on the expression and cellular localization of AR, ABCG2-expressing HPr-1-AR and CWR-R1 cells were treated with the vehicle control, DHT, Ko143, or a combination of DHT and Ko143 and subjected to immunofluorescent staining. ABCG2-negative cells treated with DHT served as the positive control. In our study, ABCG2 was only pharmacologically inhibited as approximately 1% of the cell population express ABCG2, thus genetic silencing of ABCG2 expression in such a small population is technically challenging due to low cell yield and genetic silencing of ABCG2 expression in a mixed population of cells would lead to undetectable changes.

Our results indicate that in both, the HPr-1-AR and the CWR-R1 cells, a combination of DHT and Ko143, significantly increased the nuclear AR expression; the nuclear AR signal was comparable with the ABCG2-negative cells treated with DHT (Fig. 1). The vehicle control and Ko143-treated ABCG2-expressing HPr-1-AR cells showed the presence of cytoplasmic AR due to the transfected AR vector (Fig. 1A). Upon addition of DHT, cytoplasmic AR translocated to the nucleus; however, the nuclear AR signal was very weak (Fig. 1A and C). The control and Ko143-treated ABCG2-expressing CWR-R1 cells showed very weak AR signal, which is in agreement with previous studies that have demonstrated undetectable AR expression in PSCs (10, 14). The addition of DHT to the ABCG2-expressing CWR-R1 cells slightly increased the nuclear AR expression compared with the control and Ko143-treated cells; however, a combination of DHT and Ko143, led to a significant increase in the nuclear AR expression compared with DHT alone (Fig. 1B and D). To confirm that the increased nuclear AR expression upon treatment with a combination of DHT and Ko143 is due to increased intracellular androgen levels, the efflux of androgens by ABCG2 and the consequent increase in the intracellular androgen levels upon inhibiting the ABCG2 transporters was confirmed in the steroidogenic, HPr-1-AR, and CWR-R1 cells and the nonsteroidogenic Hek293 cells stably transfected with pcDNA3.1ABCG2 vector (Supplementary Fig. S2). Thus, our results indicate that inhibiting the ABCG2-mediated androgen efflux increases the nuclear translocation of AR.

To determine whether elevated nuclear AR expression leads to increased AR transcriptional activity, the expression of AR target genes was quantitated by qRT-PCR. To determine the change in the expression of AR target genes upon inhibiting the ABCG2-mediated androgen efflux, ABCG2-expressing HPr-1-AR and CWR-R-1 cells were incubated with either the vehicle control, DHT, Ko143 or DHT in combination with Ko143 for 24 hours. The AR antagonist, enzalutamide, was added to evaluate the role of AR in the upregulation of AR target genes. Expression of AR target genes was analyzed by qRT-PCR. Different genes were quantitated in the HPr-1-AR (MME, GHRL1, AQP3, and TMEM) and the CWR-R1 cells (TIPARP, MME, FKBP5, and GADD45B) as AR targets different sets of genes in the two cell lines (18, 19). qRT-PCR analysis revealed that both cell lines treated with a combination of DHT and Ko143 showed a significant increase in the expression of AR target genes compared with cells treated with the vehicle control and DHT (Fig. 2) and that enzalutamide antagonized the ABCG2-mediated upregulation of AR target genes (Fig. 2). Although the AR target gene expression showed a maximal increase upon treatment with a combination of DHT and Ko143, we observed an upregulation of AR target genes upon treatment with DHT alone (Fig. 2); this could be due to the incomplete efflux of DHT by ABCG2 due to saturation of the transporters with DHT. In addition, we did not observe any change/downregulation in the AR target gene expression upon treatment with enzalutamide (Fig. 2), possibly due to the low levels of nuclear AR in the ABCG2-expressing cells (Fig. 1). Our results confirm that inhibiting the ABCG2-mediated androgen efflux leads to an upregulation of AR target genes and suppression of AR target gene upregulation by enzalutamide confirms the direct role of AR in upregulating the AR target genes when ABCG2-mediated androgen efflux is inhibited.

Exogenous introduction of AR in cells that lack endogenous AR leads to a delayed cell growth response and cell differentiation (14, 16, 17). Thus, we propose that increased nuclear AR expression in the HPr-1-AR and the CWR-R1 cells, upon inhibiting the ABCG2-mediated androgen efflux, would lead to a delayed cell growth response due to increased luminal differentiation of cells. Cell growth response was assessed in the ABCG2-expressing HPr-1-AR and CWR-R1 cells treated with increasing concentrations of DHT with/without Ko143, by the sphere-forming assay, which serves as a direct indicator of cell growth response of a single sphere-forming cell in a 3D setting. In addition, the sphere-forming assay can also serve as an indicator of the relative number of stem and differentiated cells in tumor samples. While the proliferative stem cells possess the ability to multiply and form spheres, the differentiated cells are unable to grow as spheres. The sphere-forming efficiency was computed by measuring the sphere sizes and assessing sphere number. Sphere number serves as a direct indicator of the proliferative potential of cells capable of forming spheres, whereas a change in sphere size indicates the eventual loss/gain of proliferative potential thus leading to smaller/larger spheres, respectively. We observed that in both, the HPr-1-AR and the CWR-R1 cells, inhibiting the ABCG2-mediated androgen efflux, significantly impaired the sphere formation efficiency as evidenced by a decrease in sphere size and sphere number (Fig. 3A, B, E, and F)). Enzalutamide rescued the impaired sphere formation efficiency of the ABCG2-expressing HPr-1-AR and CWR-R1 cells, thus confirming that the delayed cell growth response upon inhibiting the ABCG2-mediated androgen efflux is modulated by AR (Fig. 3C, D, G, H).

Previous studies have shown that the exogenous introduction of AR in PSCs drive differentiation to a luminal phenotype (16). To investigate whether inhibiting the ABCG2-mediated androgen efflux leads to an AR-modulated cell differentiation, ABCG2-expressing cells, treated with different combinations of DHT, Ko143, and enzalutamide for 5 days were analyzed for the expression of differentiation and stem cell markers. We observed that ABCG2-expressing cells treated with the vehicle control lacked the expression of prostate luminal markers. However, DHT in combination with Ko143, increased the expression levels of the prostate luminal markers, CK8 and CK18, in both the cell lines (Fig. 3I and J). In the ABCG2-expressing HPr-1-AR cells, treatment with a combination of DHT and Ko143 led to an upregulation of PSA and the downregulation of Sox2 (stem cell marker; Fig. 3I), whereas in the ABCG2-expressing CWR-R1 cells, treatment with a combination of DHT and Ko143 led to an increase in AR and phosphorylated p21 (cell-cycle arrest marker) expression (Fig. 3J). Different markers were analyzed in both the cell lines as (i) we could not detect phosphorylated p21 in the ABCG2-expressing HPr-1-AR cells and Sox2 in the ABCG2-expressing CWR-R1 cells (data not shown) (ii) In the CWR-R1 cells, PSA expression is not altered by DHT (19) and (iii) HPr-1-AR cells have a stably transfected AR vector, such that AR is cytosolic in the absence of androgens and nuclear upon treatment with androgens. As the analysis was done on whole-cell lysates, the intracellular AR expression in the ABCG2-expressing HPr-1AR cells would remain unchanged. Of note, the CWR-R1 cells are known to have both, mutated AR as well as AR variants (Supplementary Fig. S3B; ref. 20). However, we did not observe AR variants in the ABCG2-expressing cells upon treatment with DHT and Ko143 (Fig. 3J). The reason for this observation could be that AR variants are generated in response to ADT; however, PSCs are intrinsically resistant to ADT (21, 22).

Interestingly, high levels of CK8 and CK18 were also observed in the Ko143-treated ABCG2-expressing CWR-R1 cells (Fig. 3J). However, the Ko143-treated CWR-R1 cells showed the absence of phosphorylated p21 (p-p21) expression (Fig. 3J). The ABCG2-expressing CWR-R1 cells were cultured in FBS-supplemented media because growth in charcoal-stripped FBS-supplemented media leads to the downregulation of ABCG2-expressing CWR-R1 cells. The HPr-1-AR cells are routinely cultured in a serum-free media, thus obviating the need for charcoal-stripped serum. Serum contains a wide range of steroids such as testosterone, progesterone, and cholesterol. Furthermore, prostate cells can synthesize DHT from the steroids through the intact, cholesterol, and the backdoor pathway (23). Pump inhibition with Ko143 could have trapped the steroidal androgens from the serum in the cells thus eliciting a weak AR response. Thus, inhibiting ABCG2, albeit in the absence of exogenous DHT, was sufficient to induce a very weak AR expression and consequently upregulate CK8 and CK18 but not p-p21 (Fig. 3J). The immunoblot results suggest that steroidal androgens from serum can partially induce the expression of some luminal differentiation markers but not induce cell growth arrest. These results are in accordance with the observations from the sphere-forming assay where Ko143-treated ABCG2-expressing CWR-R1 cells showed no change in the sphere formation efficiency compared with cells treated with the vehicle control (Fig. 3E and F). The role of AR in the forced differentiation of ABCG2-expressing cells was confirmed by enzalutamide (Fig. 3I and J). Thus, our results suggest that inhibiting the ABCG2-mediated androgen efflux forces prostate cells to undergo AR-regulated luminal differentiation.

On the basis of our in vitro observations, we proposed that inhibiting ABCG2 in vivo would slow tumor progression and increase the overall survival in mice. To determine whether Ko143 slows tumor progression, tumor volume was monitored in testosterone-supplemented nude mice subcutaneously implanted with ABCG2-expressing CWR-R1 cells treated with either placebo or 15 mg/kg Ko143. Survival was computed by the time it took for tumors to reach a volume of 1,500 mm³. Our results show that mice treated with Ko143 showed significantly slower tumor growth rates. Linear mixed model analysis that took the individual tumor volume trajectories into account revealed statistically significant differences between the two cohorts (P < 0.01; Fig. 4A–C).

In addition to slowing down tumor progression, mice treated with Ko143 demonstrated a trend for increased survival. The median survival in the control group was 56 days as opposed to 70 days in the treated cohort. We observed a strong tendency of Ko143 to increase the overall survival; however, the log-rank analysis revealed that the differences were not statistically significant (Fig. 4D).

Results from our in vivo data show that Ko143 slows the tumor growth rate. Slower tumor growth rates are a direct indicator of impaired cellular proliferation. To confirm the results of our in vivo observation that Ko143 leads to a delayed cell response, we isolated cells from the tumors of mice by protease digestion and subjected the isolated cells to sphere-formation assay. Sphere formation assay revealed that cells isolated from tumors of Ko143-treated group formed significantly smaller and fewer spheres (Fig. 5). On the basis of our sphere formation data, we can infer that the tumors of mice treated with Ko143 had a significantly higher percentage of differentiated cells, which are unable to grow into spheres, whereas tumors of mice in the placebo group had comparatively greater number of stem cells.

On the basis of our observations from in vitro experiments, showing that the inhibition of the ABCG2-mediated androgen efflux forces prostate stem cells to differentiate to a luminal phenotype, we hypothesized that Ko143-treated mice would have more differentiated tumors compared with mice from the control group. Luminal differentiation was assessed by immunostaining the paraffin-embedded tumor sections for CK8 and AR. We observed that tumors from the Ko143-treated group had a higher percentage of CK8-expressing cells (∼86.4% in the Ko143 group vs. ∼63.75% in the control group) and AR-expressing cells (∼74% in the Ko143 group vs. ∼57% in the control group) compared with the control group (Fig. 6A–D). Also, we observed a general overlap between CK8- and AR-expressing cells. In addition, tumor sections from the Ko143 cohort showed a 2.5-fold reduction in the number of ABCG2-expressing cells compared with tumors from the control group. In the control group, ∼87.5% of the tumors contained at least one foci (areas with ≥1 ABCG2-expressing cell) of ABCG2-expressing cells as opposed to only 55.5% of the tumors from the treated cohort (Fig. 6E and F). Our results thus suggest that the Ko143-treated group had more differentiated tumors compared with the control group as evidenced by an increased percentage of luminal and decreased percentage of ABCG2-expressing cells. The reduction in tumor growth rate and increased survival in the treated group could be attributed to increased differentiation.

Elucidating the mechanisms by which stem cells maintain their undifferentiated phenotype is important to improve the efficacy of cancer therapy due to their role in the etiology of cancer and their intrinsic resistance to conventional therapies. Although, ABCG2 is suspected to play an important role in stem cell maintenance, the precise mechanism is unknown. Our study dissects the mechanism by which ABCG2 protects PSCs from undergoing differentiation. Using the HPr-1-AR and CWR-R1 cell lines, we have shown that ABCG2 maintains PSCs in an undifferentiated state by effluxing androgens. Inhibiting the ABCG2-mediated androgen efflux leads to an AR-modulated luminal differentiation and a delayed cell growth response (Fig. 7). Our observations are consistent with previous reports, which show that inducing AR expression in PSCs leads to a delayed cell growth response and luminal differentiation rather than cell proliferation (14, 16). In accordance with our in vitro observations, we observed that Ko143 treatment reduced the tumor growth rate, increased survival, and led to more differentiated tumors in mice transplanted with ABCG2-expressing CWR-R1 cells. Most of the studies on ABCG2 and its relevance to stem cells have shown that ABCG2 maintains the stem cell phenotype either by effluxing cytotoxic substrates or by acting as a downstream effector of stem cell–specific pathways such as the Wnt, Notch, and Hedgehog pathways (2, 24, 25). This is the first study to show that ABCG2 plays a role in stem cell maintenance by effluxing differentiation-inducing substrates. In addition, our study provides a plausible reason for the lack of AR expression in PSCs. Previous studies show that PSCs have low AR levels due to the methylation of the AR promoter CpG islands (14). However, we detected AR transcripts in ABCG2-expressing HPr-1-AR and CWR-R1 cells (data not shown). Previous reports have confirmed the presence of AR transcripts and shown that AR is unstable and degraded in the absence of androgens in PSCs (26, 27). Our study suggests that androgen efflux by ABCG2 may facilitate AR degradation, whereas inhibiting the ABCG2-mediated androgen efflux may lead to stabilization and the nuclear translocation of AR in the PSCs.

The important role played by prostate cancer stem cells in the etiology of CRPC have sparked a lot of interest in differentiation therapy as an adjuvant to ADT as reviewed in ref. 28. However, initial attempts at applying differentiation therapy in prostate cancer management failed to yield significant results (29, 30). The failure of differentiation therapy could be attributed to the focus on bulk tumor response, not the prostate cancer stem cell population. In addition, most of the studies evaluated differentiation therapy in the LnCaP cell line, which is primarily comprised of luminal-differentiated cells (31, 32). In contrast, our study was conducted in the ABCG2-enriched undifferentiated subpopulation derived from both normal and tumorigenic prostate cell lines. Future studies will test the direct role of ABCG2 in delaying the progression to CRPC in mouse models such as such as the CWR22 and LuCaP35 xenografts. These models can progress from an androgen-dependent to an androgen-independent phenotype, thus mimicking CRPC disease progression in human.

As we could successfully induce luminal differentiation in the HPr-1-AR and CWR-R1 cell lines and CWR-R1 xenografts, our approach could be applied in the clinical setting, in both hormone naïve as well as castration recurrent scenario. Our proposed model relies on restored AR activity to induce differentiation, while this seems counterintuitive to the current standard of care, which focuses on inhibiting AR activity, we intend to highlight the dual role of AR, in inducing differentiation and growth retardation in PSCs and inducing proliferation in AR-expressing cancer cells. Ko143 is proposed to restore AR activity and induce luminal differentiation only in the small population of ABCG2-expressing PSCs, while the bulk of the ABCG2-negative cancer cells will not be affected. In addition, we demonstrated that inhibiting the ABCG2-mediated androgen efflux in the CWR-R1 cells increases only the full-length AR expression (which subsequently causes luminal differentiation) despite the presence of AR variants in the ABCG2-negative cell population (Supplementary Fig. S3). Blocking AR activity after Ko143 administration is proposed to target the newly differentiated stem cells as well as the pre-existing AR-expressing cancer cells (Fig. 7). The presence of androgens is important for this approach. Hence, in the hormone-naïve setting, Ko143 could be administered during intermittent ADT, when the androgen sensitivity is restored, to delay the progression to CRPC or prior to enzalutamide in the CRPC setting, when androgens are synthesized by intratumoral steroidogenesis to improve survival. However, further investigation is needed to fully understand the significance of targeting ABCG2 for differentiation therapy. Although ABCG2 expression enriches for PSCs, it is likely that a subpopulation of stem cells lack ABCG2 expression, thus it is important to evaluate the percentage of ABCG2-negative stem cells in the prostate and assess their contribution to the etiology of CRPC. In addition, it would be interesting to study the expression profile of ABCG2 as prostate cancer progresses from hormone naïve to CRPC and the effects of ADT and AR antagonists on ABCG2 expression.

The initial discovery of ABCG2 and its role in multidrug resistance sparked a lot of interest in the ABC transporters. However, the subsequent lack of clinical benefit and neurotoxicity associated with ABCG2 inhibitors led to a lot of disappointment (33). Despite the lack of clinical benefit associated with first- and second-generation ABCG2 inhibitors, Ko143 has shown immense potential in preclinical studies. In addition to being the most specific ABCG2 inhibitor to date, it is also well tolerated in mice at extremely high dosages (34) and for a long period of time, as confirmed by our study. Most of the studies conducted on clinically targeting ABCG2 have focused on targeting drug resistance by administering ABCG2 inhibitors in combination with chemotherapeutic drugs to increase the bioavailability of drugs (35–37). However, combination therapy is associated with increased toxicity as ABCG2 is expressed in organs that influence drug excretion. We propose a novel use of Ko143 in differentiation therapy, which entails administering Ko143 as a single agent in the presence of endogenous androgens, thus potentially minimizing the toxicity associated with combination therapy. In conclusion, we provide a strategy to induce luminal differentiation in the PSCs by targeting the ABCG2 transporters and sensitize them to ADT.

No potential conflicts of interest were disclosed.

Conception and design: N.G. Sabnis, M.A. Titus, W.J. Huss

Development of methodology: N.G. Sabnis, M.A. Titus, W.J. Huss

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.G. Sabnis, W.J. Huss

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.G. Sabnis, A. Miller, M.A. Titus, W.J. Huss

Writing, review, and/or revision of the manuscript: N.G. Sabnis, A. Miller, M.A. Titus, W.J. Huss

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.G. Sabnis

Study supervision: N.G. Sabnis, W.J. Huss

We thank Drs. Barbara Foster, Michael Moser, James Mohler, Kalyan Gangavarapu, Mugdha Samant, and Vineet Dhiman for insightful discussions about the in vivo experiments, sphere formation assays, and cell line. We thank Ellen Karasik and Bryan Gillard for their technical assistance with the in vivo experiments. We thank the following cores at RPCI supported by NCI Cancer Center Support Grant: Laboratory Animal Resources (LAR) Bioinformatics and Biostatistics and the Mouse Tumor Model Resource (MTMR) Core.

This study was supported by the NIHRO1CA095367 (to W.J. Huss), Pharmacology & Therapeutics Departmental Funding (to W.J. Huss), Mark Diamond Research Fund (to N.G. Sabnis) and the NCI Cancer Center Support Grant (P30CA016056) to RPCI supported Mouse Tumor Model Resource and Laboratory Animal Resources.

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.