Direct Metabolic Interrogation of Dihydrotestosterone Biosynthesis from Adrenal Precursors in Primary Prostatectomy Tissues

Androgen deprivation therapy by medical or surgical castration is a mainstay of treatment for advanced prostate cancer. The primary action of androgen deprivation is attributable to the disruption of androgen signaling by depleting gonadal testosterone (T), a modest androgen receptor (AR) agonist, and canonical precursor to the more potent agonist dihydrotestosterone (DHT; ref. 1). Although initially effective, androgen deprivation therapy is invariably followed by the onset of castration-resistant prostate cancer (CRPC). CRPC is largely typified by an inappropriate restoration of the AR signaling axis (2–4), and in this setting, intratumoral biosynthesis of DHT from adrenal precursor steroids plays a major role to sustain AR activation (4–6). Thus, a detailed interrogation of androgen metabolism and enzymology is imperative toward understanding the pathophysiology of this disease.

In the absence of gonadal T, one route of DHT biosynthesis involves enzymatic modification of dehydroepiandrosterone/dehydroepiandrosterone sulfate (DHEA/DHEA-S), a primarily adrenal-derived androgen and the most abundant 19-carbon steroid in serum (5, 7). Three enzymatic steps are required to convert DHEA to DHT, consisting of (1) 3β-hydroxyl oxidation and Δ⁵→Δ⁴ isomerization by 3β-hydroxysteroid dehydrogenase/isomerase (3β-HSD), (2) 17-keto reduction by 17β-hydroxysteroid dehydrogenase (17β-HSD), and (3) 5α-reduction by 5α-reductase (7, 8). In normal male physiology, this sequence chiefly occurs within the testes and first requires the conversion of DHEA to its Δ⁴-3-keto derivative androstenedione (AD), which is subsequently 17-keto reduced to T (8). T is then secreted into circulation, where it may undergo irreversible 5α-reduction to DHT in the prostate (Fig. 1A, left). This pathway (DHEA→AD→T→DHT) was also surmised to be the metabolic route of DHT biosynthesis from circulating DHEA by CRPC, in which T serves as an obligate precursor. Notably though, AD and T share the same Δ⁴-3-keto structure that is amenable to 5α-reduction (9, 10). We have previously shown that AD is the preferred substrate relative to T for 5α-reductase expressed in CRPC (11). This results in generation of a 5α-reduced metabolite, 5α-androstanedione (5α-dione), which can be 17-keto reduced to DHT to bypass T via a different metabolic route (Fig. 1A, right). Surprisingly, this “5α-dione pathway” (DHEA→AD→5α-dione→DHT) appears to be the predominant directionality of DHEA metabolism in the majority of metastatic CRPC cell lines, as well as in sampled metastatic CRPC patient tissues (11–14). We therefore sought to determine whether the same metabolic phenotype is operational in the eugonadal setting of clinically localized prostate cancer, where both gonadal T and adrenal steroids are potential precursors of DHT.

A number of studies have previously attempted to characterize androgen synthesis in clinically localized disease. These efforts have relied mostly on steroidogenic enzyme expression to infer patterns of metabolism or sampling of androgen content in frozen tissues, which provides a snapshot at one point in time (15–21). These methods yield informative observations but cannot definitively ascertain the directionality of metabolic flux from precursors into their respective metabolites. Thus, the true metabolic pathway of adrenal steroid metabolism in primary prostatic tissue remains unclear. Stable isotopes are valuable in the study of metabolism because isotopic tagging of a substrate permits direct identification of isotope-labeled enzymatic products over time, thus offering unambiguous and temporally encoded information on pathway accessibility (22). Here, we report on a novel approach to assay androgen metabolism ex vivo in human prostatectomy tissues harboring localized cancer. Using liquid chromatography-mass spectrometry (LC-MS/MS) and capitalizing on stable isotopic tracing to follow the concurrent metabolism of two potential precursor steroids of DHT under mixed substrate conditions, our method demonstrates that primary prostate tissues harboring cancer metabolically utilize both AD and T as substrates of 5α-reductase. Consistent with observations in CRPC, the 5α-dione pathway appears to be the major directionality of flux from adrenal steroids to DHT in this context. However, AD and T appear to be comparably metabolized by 5α-reductase in primary prostate tissues, contrasting the preference exhibited by CRPC in which AD is clearly favored over T. This finding provides direct evidence that the transition of eugonadal primary disease to eventual CRPC may coincide with a metabolic switch toward AD as the preferred substrate of 5α-reductase to thereby augment adrenal steroid utilization in the biosynthesis of DHT and enable therapeutic resistance.

Human prostate cancer cell lines LNCaP and 22Rv1 were purchased from the American Type Culture Collection (ATCC) and maintained in RPMI 1640 with 10% FBS. VCaP and DU145 cells were purchased from the ATCC and maintained in DMEM with 10% FBS. MDA-PCa-2b cells were purchased from the ATCC and cultured in BRFF-HPC1 (Athena ES) containing 20% FBS. RWPE-1 cells were obtained from the ATCC and maintained in Keratinocyte Growth Medium with 0.05 mg/mL bovine pituitary extract and 5 ng/mL human recombinant epidermal growth factor according to ATCC recommendations. LAPC4 cells were generously provided by Charles Sawyers (Memorial Sloan Kettering Cancer Center, New York, NY) and were cultured in Iscove's Modified Dulbecco's Medium with 10% FBS. C4-2 cells were generously provided by J. T. Hsieh (University of Texas Southwestern, Dallas, TX) and maintained in 10% FBS. All cell lines were incubated in a 5% CO₂ humidified incubator, and 1% penicillin/streptomycin (P/S) was added to all culture media. FBS and charcoal:dextran-stripped FBS were purchased from Gemini Bio Products. All cell lines were authenticated prior to use and routinely tested for mycoplasma contamination.

Institutional Review Board approval and patient written informed consent were obtained at the Cleveland Clinic prior to procurement of deidentified human tissue samples from surgical specimens of study participants undergoing radical prostatectomy for clinically localized prostate cancer. Following surgical removal of the prostate, the fresh surgical specimen was transported to pathology for processing. One 5 mm diameter core (approximately 0.03–0.10 g) was taken by punch biopsy from the peripheral and transition zone of the whole gland specimen, placed in separate sterile containers on ice, and brought to the laboratory by tissue procurement personnel for immediate ex vivo experimentation. The remaining prostate gland was then fixed in formalin and paraffin-embedded, slide-mounted, and assessed by the collaborating expert prostate cancer pathologist (C. Magi-Galluzzi). The region surrounding the site of the core biopsy was reviewed by the pathologist for any identifiable prostate cancer. Clincopathologic data of study patients were retrospectively obtained by review of the electronic medical record.

Mouse studies were performed under a protocol approved by the Institutional Animal Care and Use Committee at the Cleveland Clinic Lerner Research Institute (protocol number 2015-1549). NSG male mice (6–8 weeks old) were purchased from the Jackson Laboratory. A total of 10⁷ LNCaP cells (resuspended in 50 μL of matrix gel) were subcutaneously injected into mice (n = 3). As previously described, mice were then surgically orchiectomized and implanted with DHEA pellets to mimic human adrenal DHEA production in patients with CRPC (11, 13, 23). When tumors reached 300 to 500 mm³, mice were sacrificed for harvesting of tumor tissue.

Cells were seeded into 12-well plates at 400,000 cells/well in triplicate and incubated overnight in respective phenol red–free media containing 10% charcoal:dextran-stripped FBS and 1% P/S prior to the experiment. LAPC-4 cells and MDA-PCa-2b cells were seeded specifically on plates coated with poly-DL-ornithine (Sigma-Aldrich). On the day of the experiment, phenol red–free, serum-free media were spiked with 1 μmol/L of ¹³C₃-[2,3,4]-androstenedione (Cerilliant) and 1 μmol/L of unlabeled testosterone (Steraloids) dissolved in 100% ethanol. Cells were incubated in 1 mL of the steroid-spiked media over 48 hours, with aliquots of media obtained at indicated times. Samples were subsequently treated with 1,000 units of β-glucuronidase (Helix Pomatia; Sigma-Aldrich) at 37°C in a 5% CO₂ incubator for 2 hours to deconjugate all steroids.

Following procurement, peripheral and transition zone tissue samples were grossly minced and manually homogenized using razor blades in a 10 cm petri dish under sterile conditions and divided into four replicates per core of approximately equal weight. Given limited tissue availability, more replicates could not be performed. Each replicate was transferred to a 12-well plate, and two replicates were subsequently incubated in 1 mL of phenol red–free, serum-free DMEM with 1% P/S spiked with 1 μmol/L of ¹³C₃-[2,3,4]-androstenedione and 1 μmol/L of testosterone dissolved in 100% ethanol, whereas the other two replicates were incubated in 1 mL of serum-free DMEM spiked with 1 μmol/L of ¹³C₃-[2,3,4]-androstenedione alone. Samples were incubated at 37°C in a 5% CO₂ humidified incubator for 48 hours. The feasibility of culturing primary prostate tissues ex vivo and longevity of measurable enzyme activity during a time course experiment were previously established (24). Aliquots of media were obtained at indicated times and frozen at −20°C until the day of sample analysis.

After removal of necrotic regions, xenograft tumor tissues (n = 3, 0.4–0.5 g each) were pooled, grossly minced, and manually homogenized using scissors in a 10 cm petri dish under sterile conditions and divided into triplicates. Each replicate was transferred to a 12-well plate and incubated in serum-free media spiked with 1 μmol/L of ¹³C₃-[2,3,4]-androstenedione and 1 μmol/L of testosterone or ¹³C₃-[2,3,4]-androstenedione alone as previously described, with aliquots of media obtained at indicated times. Samples were subsequently treated with β-glucuronidase and analyzed as previously described.

Aliquot samples were thawed from storage at −20°C and subjected to centrifugation at 5,000 RPM for 5 minutes at 4°C to spin down insoluble debris, after which an equal volume (200 μL) of media supernatant from each aliquot was transferred to a new 1.5 mL Eppendorf tube. Samples were spiked with 20 μL of internal standard (50 ng/mL D₄-[9,11,12,12]-cortisol; [D₄]-cortisol; CDN Isotopes), briefly vortexed, and transferred to glass test tubes for liquid–liquid extraction. After addition of 2 mL of methyl-tert-butyl ether (MTBE, Across), glass tubes were vortexed for 5 minutes using a multi-tube vortexer (Fisher) and centrifuged for 5 minutes at 2,500 RPM at 4°C. Samples were placed on dry ice for 20 minutes, and the top organic solvent with extracted steroid was transferred to a new glass tube. This fraction was evaporated under nitrogen gas at 40°C and then reconstituted in 110 μL of 50% methanol/water (v/v) and briefly vortexed again. The reconstituted samples were transferred into 1.5 mL microcentrifuge tubes, centrifuged at 13,000 RPM for 10 minutes at 4°C, and the supernatants were collected for mass spectrometry injections in high-performance liquid chromatography (HPLC) vials. To analyze glucuronidated steroids separately, the aqueous phase of the liquid–liquid extraction was separately collected, washed twice with MTBE extraction again to remove residual free steroids, and subjected to an overnight dry period to evaporate remaining MTBE. Samples were treated with β-glucuronidase as previously described. The samples were analyzed by LC-MS/MS to calculate the conjugated fraction of steroid metabolites.

Extracted steroids from samples were quantified using LC/MS/MS. Briefly, the LC-MS/MS system consists of an ultra-pressure liquid chromatography system (UPLC; Shimadzu Corporation), comprising of two LC-30AD pumps, a DGU-20A5R vacuum degasser, a CTO-30A column oven, SIL-30AC autosampler, and a system controller CBM-20A and coupled with a Qtrap 5500 mass spectrometer (AB Sciex). Extracted steroids were injected onto the Shimadzu UPLC system, and the steroids were separated through a C18 column (Zorbax Eclipse Plus C18 column 150 mm × 2.1 mm, 3.5 μm, Agilent) using a gradient starting from 20% solvent B [acetonitrile/methanol (90/10, v/v) containing 0.2% formic acid] over 4 minutes and then to 65% solvent B over 15 minutes, followed by 90% solvent B for 3 minutes. Steroids were quantified on a Qtrap 5500 mass spectrometer using ESI in positive ion mode and multiple reaction monitoring (MRM) using characteristic parent → daughter ion transitions for specific molecular species monitored. D₄-[9,11,12,12]-cortisol was used as an internal standard for calibration of steroids in each sample. Data acquisition and processing were performed using Analyst software (AB Sciex, version 1.6.2). The peak area ratio of the analyte over the internal standard was used for quantification purposes. Each sample run included calibration curves with standards for data quantification using the analyte/internal standard peak area ratio. The specific transitions for molecular species were as follows: ¹³C₃-[2,3,4]-androstenedione [290.1>112.1], AD [287.1>109.0], ¹³C₃-[2,3,4]-testosterone [292.1>112.1], testosterone [289.1>109.0], ¹³C₃-[2,3,4]-5α-dione [292.2>274.2], 5α-dione [289.2>271.2], ¹³C₃-[2,3,4]-DHT [294.1>162.1], and DHT [291.1>255.2].

To calculate the rate of metabolic flux, a baseline precursor steroid concentration was first obtained from a sample draw of media spiked with 1 μmol/L of ¹³C₃-[2,3,4]-androstenedione and 1 μmol/L of testosterone prior to tissue incubation. Steroid metabolite concentrations in media samples drawn at different time points of tissue incubation were normalized against this initial precursor concentration to derive a percent conversion value. Duplicate values for each target metabolite were averaged, and the SD was determined for each individual patient sample. For pooled analyses across the entire patient cohort, target metabolite concentrations for individual patient samples were averaged, and SEM was reported.

Following procurement, tissues were minced, homogenized, and transferred into a 12-well plate in triplicate, as previously described. Samples were incubated in serum-free DMEM spiked with [³H]-androstenedione dissolved in 100% ethanol (300,000–600,000 cpm, PerkinElmer) and 100 nmol/L of unlabeled AD (Steraloids) at 37°C in a 5% CO₂ incubator over 48 hours. Aliquots of media were obtained, stored, and treated with β-glucuronidase, as previously described.

Steroids were extracted with 1:1 ethyl acetate:isooctane and concentrated by evaporation of the solvent under nitrogen gas in glass tubes. Dried samples were reconstituted in 50% methanol and injected on a Breeze 1525 system equipped with a model 717 plus autoinjector (Waters Corp.) and a Kinetix 100 × 2.1-mm, 2.6 μmol/L C₁₈ reverse-phase column (Phenomenex) and methanol/water gradients at 30°C. The column effluent was analyzed using a β-RAM model 3 in-line radioactivity detector (IN/US Systems, Inc.) using Liquiscint scintillation mixture (National Diagnostics). Steroid metabolites were identified using tritiated standards of AD, T, DHEA, and DHT (PerkinElmer). The expected retention time for 5α-dione was previously validated by synthesizing tritium-labeled 5α-dione (11).

Seventeen eugonadal patients presenting with clinically localized prostate cancer were consented for fresh tissue collection and subsequently underwent open or robotic-assisted radical prostatectomy. The first 12 samples were analyzed by LC-MS/MS, whereas the last 5 samples were used for independent validation through a radiotracing method. Clinicopathologic features of patients are reported (Table 1). The majority of patients had organ-confined disease, although final pathology revealed 3 of 17 patients with nodal metastases discovered upon pelvic lymph node dissection. Medication lists were also obtained from electronic medical records to determine whether any patients were on treatment with 5α-reductase inhibitors (5-ARI). Two patients were documented to be on finasteride and 1 patient on dutasteride at the time of surgery. Although these patients were included in the final patient series, data points from these cases are separately reported and were excluded from all pooled analyses to avoid any potential confounding from perturbed metabolism. Pathology slides were also assessed to determine whether sample cores obtained were involved with cancer. If the tissue surrounding the original core biopsy was found to harbor prostate cancer on review of histology, the core was designated as “possible cancer.” In total, 10 of 17 peripheral zone cores and 4 of 17 transition zone cores were deemed to potentially involve cancerous tissue, whereas the rest contained non-neoplastic tissue (Table 1).

In order to simultaneously monitor the metabolism of both AD and T, which may compete for 5α-reductase and share common downstream metabolites, we employed a stable isotopic labeling method to definitively distinguish the enzymatic conversion of each Δ⁴-3-keto precursor steroid by prostatic tissue into their respective steroid products. To do this, prostate tissue samples were incubated in media spiked with an equimolar mixture of an ¹³C isotopologue of AD (¹³C₃-[2,3,4]-androstenedione; [¹³C₃]-AD) and unlabeled T (T^unl). The generation of nascent metabolites in media was followed over time using LC-MS/MS. The origin of these metabolites was determined based on enrichment or absence of an incorporated M+3 ¹³C₃ fragment on MRM transitions (Fig. 1C–E). To confirm the validity of this approach, we verified the concordance of our measurements with a previously validated radiotracing method (11) in three different prostate cancer cell lines (LAPC4, C4-2, DU145; Supplementary Fig. S1).

Metabolic tracing revealed that the majority of [¹³C₃]-AD and T^unl was enzymatically converted by primary prostate tissues to downstream metabolites within 48 hours. The principal metabolite that was detected arising from [¹³C₃]-AD was [¹³C₃]-5α-dione (Fig. 2A), whereas T^unl was converted to DHT^unl as well as AD^unl to a lesser degree (Fig. 2B). Furthermore, [¹³C₃]-DHT originating from [¹³C₃]-AD was detected in media as early as 7 hours of incubation in all incubated samples and was more abundant than [¹³C₃]-T throughout the time course experiment (Fig. 2C). We additionally confirmed these findings via radiotracing experiments (Fig. 2D and E). Following treatment of prostate tissue samples with [³H]-AD, both [³H]-5α-dione and [³H]-DHT were successfully separated by HPLC and identified via radio flow detector. Notably, no [³H]-T signal was detectable (Fig. 2F). Taken together, these findings suggest that the 5α-dione pathway is likely the preferred metabolic route to DHT from the adrenal source. Interestingly, we did not observe a difference in 5α-dione pathway utilization between tissue samples possibly harboring cancer and non-neoplastic samples (Fig. 3). Thus, the 5α-dione pathway of DHT biosynthesis from adrenal steroids is probably the metabolic phenotype of prostatic tissue under normal physiologic conditions.

To investigate whether the pharmacologic effect of orally administered 5-ARIs in patients is appreciable in primary prostate tissues cultured ex vivo several hours after radical prostatectomy, we similarly incubated tissue samples from 3 patients confirmed to be on a 5-ARI at time of surgery. Following treatment of these tissues with [¹³C₃]-AD and T^unl, blockade of both [¹³C₃]-AD→[¹³C₃]-5α-dione and T^unl→DHT^unl was observed, consistent with the expected inhibitory effect of a 5-ARI (Fig. 4A and B). Furthermore, inhibition of 5-α-reductase enhanced the conversion of T^unl→AD^unl, which subsequently accumulated in media. However, in contrast to total blockade of T^unl→DHT^unl, [¹³C₃]-AD→[¹³C₃]-5α-dione conversion was reduced but not completely restricted, leading to residual [¹³C₃]-DHT formation (Fig. 4C and D). Furthermore, despite accumulation of [¹³C₃]-AD from partial blockade of 5-α-reductase, only a marginal increase in alternative flux from [¹³C₃]-AD to [¹³C₃]-T was observed (Fig. 4E), suggesting that 17-keto reduction of AD→T is overall limited in prostatic tissue. This observation, as well as the fact that AD can be converted to DHT despite total pharmacologic blockade of T→DHT, provides further evidence that the primary metabolic pathway from adrenal steroids to DHT may not require conversion of AD→T→DHT but rather involves the 5α-dione pathway. These findings also indicate that the dominantly accumulating Δ⁴-3-keto precursor steroid from blockade of 5α-reductase is AD rather than T, given the propensity of T to be oxidized back to AD in this context.

Because AD was observed to undergo relatively robust 5α-reduction, we also sought to determine whether AD is favored over T by 5α-reductase in primary prostate tissues. The conversion efficiency of [¹³C₃]-AD→[¹³C₃]-5α-dione was thus compared against T^unl→DHT^unl under equimolar mixed substrate conditions, permitting a quantitative assessment of precursor utilization by 5α-reductase within the same biological system. Although AD was capably 5α-reduced, T appeared to be a comparable substrate in primary prostate tissue samples (Fig. 5A). Because the 5α-reduced products of AD and T can be further metabolized by 17β-HSD via a second-step reaction to potentially confound conclusions drawn from tracking only the immediate metabolite, a secondary analysis was also performed by totaling the signal intensity from both the immediate 5α-reduced product and subsequent 5α-reduced, additionally C17-modified metabolite (Fig. 5B). This comparison further suggested that T is a comparable, if not slightly preferred, substrate to AD in primary prostate tissue samples. We also considered the possibility that conjugation of DHT by uridine 5′-diphospho-glucuronosyltransferase (UGT) enzymes at its 17-position might potentially bias the ratio of free [¹³C₃]-5α-dione/DHT^unl detected by our method. To exclude this, we separately analyzed the conjugated steroid content in media but found no evidence of UGT activity in incubated primary prostate tissues (Supplementary Fig. S2). This observation of 5α-reductase slightly favoring T was generally consistent across individual malignant and non-neoplastic samples, as reflected by a [¹³C₃]-5α-dione/DHT^unl generation ratio of <1 (Fig. 5C and D).

Remarkably, this phenotype in primary prostate tissues contrasts with that of the majority of human CRPC cell lines. Our prior radiotracing experiments have shown in CRPC that AD is 5α-reduced more expeditiously than T (11). To directly compare AD and T utilization by 5α-reductase between primary prostate tissues and CRPC, we re-applied our mixed substrate tracing method using eight available CRPC cell lines, LNCaP xenograft tumor tissue, as well as an immortalized benign prostate epithelial cell line (RWPE-1). In all CRPC cell lines and xenograft tissue, [¹³C₃]-AD→[¹³C₃]-5α-dione was more efficient than T^unl→DHT^unl ([¹³C₃]-5α-dione/DHT^unl generation ratio of >1), confirming the clear preference for AD over T by 5α-reductase expressed in CRPC (Fig. 5C and D). This metabolic pattern is also consistent with observations in tissues previously obtained from a metastatic CRPC patient, suggesting that this difference is not simply experimental artifact (11). Interestingly, the substrate preference of RWPE-1 aligned more closely with primary prostate tissue samples, supporting the notion that T may be the preferred substrate in non-neoplastic prostatic tissue. Overall, these findings indicate that the preference of Δ⁴-3-keto steroid utilization for DHT biosynthesis in metastatic CRPC differs from that of primary prostatic tissue.

Because AD and T are potentially competing substrates of 5α-reductase, we also considered the possibility that the enzyme kinetics observed under mixed substrate conditions might differ from that of monosubstrate conditions. To investigate potential effects of mixing alternative substrates, two conditions were established in which prostate tissues were incubated with [¹³C₃]-AD, with or without an equimolar concentration of T^unl. However, no appreciable difference was observed in the rate of [¹³C₃]-AD metabolism when T^unl was added to the substrate mix (Supplementary Fig. S3). The directionality of metabolic flux from AD is thus unlikely to be affected by competition for 5α-reductase under multiple substrate conditions in prostate tissues.

Despite castrate levels of serum T after androgen deprivation therapy, intraprostatic androgen levels generally remain incompletely depleted, even prior to the onset of CRPC (25–28). Furthermore, recent studies strongly suggest that targeting the adrenal contribution in castration-naïve prostate cancer is a viable and effective therapeutic strategy (29, 30). Although the modification of circulating adrenal steroids has long been recognized as a putative source of intraprostatic DHT requiring only a limited complement of steroidogenic enzymes expressed in the prostate (8, 26), this metabolic pathway has not been rigorously defined. In this study, we show that the 5α-dione pathway, which was previously determined to be the major metabolic pathway to DHT from circulating adrenal steroids in metastatic CRPC, is also the predominant sequence occurring in primary prostate tissues from eugonadal patients with clinically localized disease. Notably, this pathway circumvents T as a precursor, resulting in two different and distinctly nonredundant metabolic routes to derive DHT via circulating steroids present in a eugonadal state (gonadal T→DHT vs. adrenal DHEA→AD→5α-dione→DHT). This observation deviates from the general view that 5α-reduction of T is the principal enzymatic step to generate intraprostatic DHT, regardless of the source (8). The absence of an appreciable difference in metabolism between non-neoplastic and cancerous samples in our study further suggests that this pattern is probably the metabolic phenotype assumed by physiologic prostatic tissue in the eugonadal setting.

Although the 5α-dione pathway appears to be the consistent directionality of adrenal steroid metabolism in both primary tissue and CRPC, we also observe that AD and T are comparable substrates of 5α-reductase in prostatic tissue, juxtaposing findings in CRPC where AD is the clearly preferred substrate. This is noteworthy because the conversion of either gonadal T or adrenal DHEA into DHT, although sequentially distinct, must both engage 5α-reductase, albeit via different substrates (Fig. 5E). Thus, this shift toward AD as the more favorable substrate may represent a critical metabolic switch in CRPC that heralds a compensatory increase in adrenal steroid utilization for DHT biosynthesis. This is supported by other independent observations illustrating the impact of enhanced adrenal steroid metabolism in CRPC—for instance the clinical consequence of a gain-of-function missense variant in 3β-HSD isoenzyme-1, which dramatically increases the rate of DHEA→AD conversion (13, 31, 32). Remarkably, this metabolic switch toward AD also appears to coincide with a conspicuous loss in the relative ability of 5α-reductase to harness T. In fact, the majority of prostate cancer cell lines and sampled patient metastases demonstrate very low capacity to convert T→DHT and thus exhibit an idiosyncratic dependence on adrenal precursor steroids (11).

The 5α-dione pathway is probably the major route of adrenal steroid metabolism in the prostate due to the strong catalytic activity of 5α-reductase, which dictates the overall directionality of metabolic flux from AD. The notion that AD can be 17-keto reduced to T originally derives from observations in gonadal tissue, where the expression of 17β-HSD isoenzyme-3 is essential toward driving this reductive process. This is evidenced by genetic deficiencies in 17β-HSD isoenzyme-3, which manifest in the characteristic metabolic restriction of AD→T (33). In contrast to the testes, the prostate is devoid of this particular enzyme but instead expresses 17β-HSD isoenzyme-5 (AKR1C3). AKR1C3 also demonstrates the capacity to reduce 17-keto steroids, including both AD and 5α-dione (34–37), but in the prostate, 17-keto reduction and 5α-reduction are competing reactions due to the abundant expression of 5α-reductase. We observe that the relatively high rate of 5α-reduction in prostatic tissue appears to considerably exceed that of 17-keto reduction, so that the majority of AD is rapidly converted by 5α-reductase to 5α-dione prior to 17-keto reduction.

Like 17β-HSD, more than one isoenzyme of 5α-reductase has been confirmed in humans. 5α-reductase isoenzyme-2 (SRD5A2) is dominantly expressed in the prostate and is primarily responsible for the robust 5α-reduction of T to DHT (19, 38–40). However, prior cell-free enzyme kinetic studies suggest that this isoenzyme actually exhibits a similar kinetic profile for AD, explaining why the 5α-dione pathway is also observed in non-neoplastic tissue prostate samples in our study (10). In contrast, 5α-reductase isoenzyme-1 (SRD5A1) is expressed to a lesser degree in non-neoplastic prostatic tissue but appears to be the major isoenzyme found in CPRC. Unlike SRD5A2, the kinetic profile of SRD5A1 strongly favors AD (10) and appears to be main enzyme responsible for adrenal steroid metabolism in CRPC cell lines (11). Aberrantly high expression of SRD5A1 with relatively low expression of SRD5A2 is a fairly consistent pattern seen across human CRPC cell lines and sampled tissue from CRPC patients (11, 15, 17, 21, 41). It is probably this combination together that determines the overall reliance on adrenal steroids rather than T to generate DHT.

Our effort to determine the actual metabolic phenotype of prostate tissue samples obtained from patients adds direct evidence to pre-existing data from gene expression and cell-free enzyme kinetics studies. Importantly, this approach goes a step beyond inference, which should be acknowledged because enzyme transcript and protein expression levels do not always correlate well with extent of metabolic activity (13), and because kinetic information may not capture the overall metabolic picture within a complex, multiple substrate/multiple enzyme biological system. Furthermore, previous studies in this area have unfortunately been limited by the lack of available in vitro models which properly recapitulate the biology of primary disease. Likewise, primary disease of the prostate generally persists in a eugonadal state, where the interaction of several circulating steroids as potentially competing substrates is challenging to recapitulate. Our mixed substrate approach with stable isotopic labeling and mass spectrometry enables a direct comparison of substrate preference for 5α-reductase when both steroids are introduced into the same ex vivo biological system.

It is important to note that our study does not aim to ascertain the absolute contribution of adrenal steroids to DHT, which has been previously explored from in vivo studies on intraprostatic DHT concentrations in castrated men (42). Our study rather attempts to define the directionality of adrenal versus gonadal DHT biosynthesis and the comparative utilization of one metabolic pathway versus the other. However, although our approach serves as a proof-of-concept for the determination of androgen metabolism, we cannot conclude with certainty whether clinically localized cancer within our primary samples metabolically resembles the non-neoplastic phenotype. It is possible that steroid metabolites detected in media are not necessarily representative of localization of steroids across different tissue compartments. Our study also does not tease apart more nuanced differences in metabolism between epithelium and stroma within the tissue parenchyma or any possible paracrine effects which may contribute to regional differences in metabolism. Despite technical limitations, we nevertheless attempted to classify samples into tumor or non-neoplastic categories based on the pathology of the surrounding tissue, which did not reveal any appreciable differences in metabolism.

In summary, our findings posit that adrenal steroid metabolism to DHT in prostatic tissue harboring localized prostate cancer is dictated in large part by the catalytic activity of 5α-reductase to drive conversion via the 5α-dione pathway. This results in two distinctly nonredundant metabolic pathways through which gonadal T and adrenal DHEA are converted to DHT, both requiring 5α-reductase, albeit via different substrates. Under physiologic conditions of the eugonadal state where both T and DHEA are present, adrenal steroids are probably a comparatively inefficient source for DHT requiring multiple enzymatic modifications; however, pathway accessibility may increase in CRPC, in part due to a metabolic switch toward AD as the preferred substrate of 5α-reductase. This change is likely driven by differential expression of two 5α-reductase isoenzymes found in humans, which in the process creates an idiosyncratic dependence on adrenal steroids, while eschewing T as the major source of DHT. Ultimately, these findings underscore the fundamental need to better define steroid metabolic pathways of prostate cancer in its various stages, in the effort to develop more effective AR-directed therapies.

A.J. Stephenson reports receiving speakers bureau honoraria from Genomic Health. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Dai, Y.-M. Chung, E. Kovac, E.A. Klein, N. Sharifi

Development of methodology: C. Dai, Y.-M. Chung, E. Kovac, Z. Zhu, N. Sharifi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Dai, Y.-M. Chung, E. Kovac, C. Magi-Galluzzi, A.J. Stephenson, N. Sharifi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Dai, Y.-M. Chung, E. Kovac, A.J. Stephenson, E.A. Klein, N. Sharifi

Writing, review, and/or revision of the manuscript: C. Dai, Y.-M. Chung, E. Kovac, Z. Zhu, J. Li, C. Magi-Galluzzi, A.J. Stephenson, E.A. Klein, N. Sharifi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Dai, Y.-M. Chung, Z. Zhu, E.A. Klein, N. Sharifi

Study supervision: C. Dai, N. Sharifi

This work was supported by funding from the Howard Hughes Medical Institute Medical Fellows Program (to C. Dai), a Howard Hughes Medical Institute Physician-Scientist Early Career Award (to N. Sharifi), a Prostate Cancer Foundation Challenge Award (to N. Sharifi), and National Cancer Institute grants (R01CA168899, R01CA172382, and R01CA190289) to N. Sharifi.

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.