Identification of Anabolic Selective Androgen Receptor Modulators with Reduced Activities in Reproductive Tissues and Sebaceous Glands

Androgen replacement therapy is a promising strategy for the treatment of frailty; however, androgens pose risks for unwanted effects including virilization and hypertrophy of reproductive organs. Selective Androgen Receptor Modulators (SARMs) retain the anabolic properties of androgens in bone and muscle while having reduced effects in other tissues. We describe two structurally similar 4-aza-steroidal androgen receptor (AR) ligands, Cl-4AS-1, a full agonist, and TFM-4AS-1, which is a SARM. TFM-4AS-1 is a potent AR ligand (IC50, 38 nm) that partially activates an AR-dependent MMTV promoter (55% of maximal response) while antagonizing the N-terminal/C-terminal interaction within AR that is required for full receptor activation. Microarray analyses of MDA-MB-453 cells show that whereas Cl-4AS-1 behaves like 5α-dihydrotestosterone (DHT), TFM-4AS-1 acts as a gene-selective agonist, inducing some genes as effectively as DHT and others to a lesser extent or not at all. This gene-selective agonism manifests as tissue-selectivity: in ovariectomized rats, Cl-4AS-1 mimics DHT while TFM-4AS-1 promotes the accrual of bone and muscle mass while having reduced effects on reproductive organs and sebaceous glands. Moreover, TFM-4AS-1 does not promote prostate growth and antagonizes DHT in seminal vesicles. To confirm that the biochemical properties of TFM-4AS-1 confer tissue selectivity, we identified a structurally unrelated compound, FTBU-1, with partial agonist activity coupled with antagonism of the N-terminal/C-terminal interaction and found that it also behaves as a SARM. TFM-4AS-1 and FTBU-1 represent two new classes of SARMs and will allow for comparative studies aimed at understanding the biophysical and physiological basis of tissue-selective effects of nuclear receptor ligands.


Androgen replacement therapy is a promising strategy for the treatment of frailty; however, androgens pose risks for unwanted effects including virilization and hypertrophy of reproductive organs. Selective Androgen Receptor Modulators (SARMs) retain the anabolic properties of androgens in bone and muscle while having reduced effects in other tissues. We describe two structurally similar 4-aza-steroidal androgen receptor (AR) ligands, Cl-4AS-1, a full agonist, and TFM-4AS-1, which is a SARM. TFM-4AS-1 is a potent AR ligand (IC 50 , 38 nM) that partially activates an AR-dependent MMTV promoter (55% of maximal response) while antagonizing the N-terminal/C-terminal interaction within AR that is required for full receptor activation. Microarray analyses of MDA-MB-453 cells show that whereas
Cl-4AS-1 behaves like 5␣-dihydrotestosterone (DHT), TFM-4AS-1 acts as a geneselective agonist, inducing some genes as effectively as DHT and others to a lesser extent or not at all. This gene-selective agonism manifests as tissue-selectivity: in ovariectomized rats, Cl-4AS-1 mimics DHT while TFM-4AS-1 promotes the accrual of bone and muscle mass while having reduced effects on reproductive organs and sebaceous glands. Moreover, TFM-4AS-1 does not promote prostate growth and antagonizes DHT in seminal vesicles. To confirm that the biochemical properties of TFM-4AS-1 confer tissue selectivity, we identified a structurally unrelated compound, FTBU-1, with partial agonist activity coupled with antagonism of the N-terminal/C-terminal interaction and found that it also behaves as a SARM. TFM-4AS-1 and FTBU-1 represent two new classes of SARMs and will allow for comparative studies aimed at understanding the biophysical and physiological basis of tissue-selective effects of nuclear receptor ligands.
Androgens, primarily testosterone (T) 7 and its more potent derivative, 5␣-dihydrotestosterone (DHT), induce male reproductive physiology and secondary sexual traits such as facial hair and deepened voice. Additionally, in both genders androgens regulate bone and muscle anabolism, adipose mass, lipoprotein metabolism, and behavior (1)(2)(3). Androgens decline with age in both men and women (4), which contributes to age-related bone and muscle loss and increases in fat mass (5). Several studies report low testosterone as a risk factor for age-related diseases including osteoporosis (6), sarcopenia (7), atherosclerosis (8), type II diabetes/metabolic syndrome and obesity (9), cognitive impairment (10), and depression (11). Restoring androgens to youthful levels could thus slow unfavorable changes in body composition and improve mood, motivation, and general health. Unfortunately, current androgens induce male secondary sexual traits such as acne and hirsutism, an effect known as virilization, (12) and pose concerns related to unwanted effects in the prostate and other reproductive organs (13)(14)(15). Therefore, androgens are limited by concerns over safety and tolerability.
Androgens exert their physiological effects by activating the androgen receptor (AR), a nuclear receptor expressed in reproductive tissues and other organs. Once bound by ligand, AR dissociates from chaperones in the cytoplasm and translocates to the nucleus where it induces target genes by binding to DNA sequences called androgen response elements (AREs) present in promoter/enhancer regions of responsive genes (16,17). AR also represses transcription by binding and inhibiting certain transcription factors (18 -21). Based on precedent with the estrogen receptor ␣ (ER␣), for which tissue-selective estrogen receptor modulators (SERMs) have been developed into effective medicines, there has been an effort to discover tissue-selective androgen receptor modulators (SARMs). Ideally, SARMs would produce anabolism in bone and muscle with limited effects on uterus, skin, and prostate (22,23). The properties of several experimental SARMs such as BMS-565929, S-40503, S-1, S-4, Ostarine, LGD-2226, and LGD-2941, have been reviewed (23). All of these compounds exhibit different doseresponse curves in androgen-responsive tissues relative to a full agonist, they vary in their in vitro and in vivo properties, and it remains to be determined which will be suitable for therapeutic use. Furthermore, the molecular mechanisms by which tissue selectivity occurs are not known, impairing rational design of improved SARMs.
We hypothesized that all AR ligands capable of significant transactivation would produce anabolism, but SARMs would create a unique receptor conformation that could be detected in other biochemical experiments. AR has a unique feature in which full activation requires physical interaction between its N-terminal and C-terminal domains (N/C interaction). Human mutations associated with partial androgen insensitivity reduce this interaction (24 -31). Because these individuals are often incompletely virilized, we hypothesized that those AR ligands that do not promote the conformation required for the N/C interaction would have reduced virilizing effects. Using this approach, we identified two new distinct anabolic SARMs, TFM-4AS-1 and FTBU-1, which display agonism in bone and muscle while having reduced effects on reproductive organs and sebaceous glands. We provide evidence suggesting that the molecular basis of these differential responses is gene-selective agonism.
N/C Interaction-The N/C interaction of rhAR was evaluated by a mammalian two-hybrid assay in CV1 cells. The Gal4-DNA binding domain was fused with the ligand binding domain (LBD) of rhAR (amino acids 637-895); a VP16 construct was fused with amino acids 1-513 of rhAR. Both plasmids were co-transfected with a luciferase reporter under the control of multiple Gal4-DBD binding sites. The N/C interaction was detected as a ligand-mediated increase of luciferase activity.
Animal Studies-Animal procedures were performed as described (38 -40). Briefly, bone, body composition, and uterine studies were performed in ovariectomized (OVX) or shamoperated rats at age 6 -9 months, 3 months post-surgery. Animals were also treated with the bone resorption inhibitor alendronate unless noted (5.6 g/kg/week). Animals were randomized into groups of equal weights (n ϭ 10 -16) and compounds in 3% benzyl alcohol in sesame oil (vehicle) were given by subcutaneous (sc) injection for 24 days. Uteri were dissected at the cervix and weighed wet. Sebaceous gland area in dorsal skin sections was measured using BioQuant. Femurs were analyzed as described (38 -41). The primary measurement, bone formation rate (BFR), was assessed by histological analysis of fluorochrome double-labeling at the periosteal surface of the distal femur. Calcein (10 mg/kg, SC) was administrated 12 and 3 days prior to the termination of the study. Fat and lean body mass composition changes were assessed by dual energy x-ray absorptiometry (DEXA). Statistical analysis was performed by Kruskal-Wallis non-parametric ANOVA followed by Student-Neuman-Keuls post-hoc testing for intergroup differences. Prostate and seminal vesicles were studied in 3-4 month-old 250 -300 g rats after orchidectomy (ORX) or sham operation. Nine days following surgery, animals were injected (sc) daily with test compounds for 7 or 14 days. At the indicated times, animals were euthanized by CO 2 , and ventral prostates were weighed. Data were analyzed by Fisher's PLSD and ANOVA (Statview, version 5.0).
Microarray and RNA Analyses-Prostate microarray studies were performed as described (38). For cell culture studies, total RNA was extracted using TRIzol (Invitrogen, Grand Island, NY) from duplicate 10-cm dishes of MDA-MB-453 cells treated 18 h. Microarray analysis was performed on 5 g of total RNA (38,41). Data were normalized to achieve identical median fluorescence intensity of each array. For a transcript to be considered DHT-regulated, the probe must have corresponded to a gene annotated at Entrez Gene, and the hybridization signals must have tested different from vehicle control (p Ͻ 0.05, Rosetta error model (42) and differed from vehicle controls by Ն1.5-fold in both duplicate 200 nM DHT samples. Gene expres-

TFM-4AS-1 Is a Novel Selective Androgen Receptor Modulator
sion data are the mean of the duplicates Ϯ S.D. For quantitative RT-PCR, total RNA was collected and analyzed as described after 18 h of treatment (41). For studies involving cycloheximide, MDA-MB-453 cells in 10-cm dishes were pretreated for 30 min with 10 g/ml cycloheximide, and the AR ligand was added at indicated concentrations for an additional 6 h.

TFM-4AS-1 Displays Partial Agonism in Transcriptional
Assays-To identify AR ligands with distinct transcriptional properties, we mined the Merck chemical library for compounds with binding affinities lower than 300 nM for AR and counterscreened them for binding glucocorticoid, progesterone, mineralocorticoid (GR, PR, MR), and both ERs. Compounds selective for AR were tested for MMTV promoter transactivation mediated by endogenous AR in MDA-MB-453 cells. This cell line was selected because agonists (DHT, testosterone, R1881, mibolerone, and danazol) and antagonists (cyproterone acetate, hydroxyflutamide, and bicalutamide) exhibited activities that matched their actions in human and animal studies. The MMTV promoter screen produced two hits we selected for further study. The structurally related compounds Cl-4AS-1 and TFM-4AS-1 had potent AR binding activities in radioligand binding assays with IC 50 values of 12 and 30 nM, respectively (43); IC 50 values for GR, PR, MR, or ER were greater than 5000 nM. Both molecules were synthesized during the development of finasteride, a 5␣ reductase inhibitor, so in addition they are also inhibitors of rat 5␣-reductase enzymes type I (6 and 2 nM, respectively) and type 2 (10 and 3 nM, respectively) (Ref. 43 and Fig. 1A).
In binding studies, DHT exhibited similar apparent affinity (K d values of 0.3-0.46 nM) for AR that is natively expressed in MDA-MB 453 cells and to the rhAR LBD that was expressed as GST fusion protein in yeast (Fig. 1B). In contrast, TFM-4AS-1 had a 40-fold reduction in the affinity to the rhAR LBD relative to the AR aporeceptor natively expressed in MDA-MB-453 cells. This change in potency was not related to species differences because TFM-4AS-1 had the expected binding affinity of ϳ30 nM to full-length rhAR transfected into cells (data not shown). Therefore, unlike DHT, TFM-4AS-1 requires a mature AR aporeceptor complex for high affinity binding.
In transactivation assays using the MMTV promoter, Cl-4AS-1 maximal fold stimulation was similar to the full agonist R1881 ( Fig. 2A). Its maximal transactivation relative to 100 nM R1881 reached an agonistic activity of 135 Ϯ 23% (mean of n ϭ 7 S.E.) with an inflection point (IP) value of 2.8 Ϯ 2.1 nM compared with 0.3 nM for R1881. In contrast, TFM-4AS-1 produced only partial stimulation of the MMTV promoter that was maintained through 2-3 log concentrations at ϳ50% ( Fig. 2A). Relative to 100 nM R1881, its maximal transactivation was 55 Ϯ 11% with an IP value of 28.4 Ϯ 19.9 nM (average of n ϭ 244). TFM-4AS-1 antagonizes 1 nM R1881 as does the anti-androgen bicalutamide, but unlike bicalutamide, TFM-4AS-1's antagonist activity did not exceed ϳ50% up to 1 M (Fig. 2A). These data established that the partial activation of the MMTV promoter by TFM-4AS-1 represents a true submaximal effect that stems from functional activity of the ligand on the receptor and confirms that TFM-4AS-1 is an AR ligand with mixed agonist and antagonist activities.
We next examined AR-dependent repression of the phorbol ester-activated MMP-1 promoter in 22Rv1 human prostate cancer cells. Cl-4AS-1 suppressed promoter activity much like DHT (Fig. 2B). In contrast, TFM-4AS-1 behaved like AR antagonists such as hydroxyflutamide and did not decrease MMP-1 promoter activity. Thus, in contrast to MMTV promoter transactivation, TFM-4AS-1 did not suppress the MMP-1 promoter where transrepression is mediated via AP-1 binding.
To compare the ability of these compounds to induce the AR N/C interaction, a mammalian two-hybrid assay in CV1 cells was employed. The Gal4-DNA binding domain (DBD) was fused with the LBD of rhAR and transfected with a VP16 construct containing the rhAR NTD and a Gal4-DBD luciferase reporter. Upon agonist binding to the LBD, a conformational change occurs and the NTD/VP16 fusion protein is recruited, resulting in luciferase transcription. Cl-4AS-1 (10 M) effectively promoted the AR N/C interaction (Fig. 2C), with an average (n ϭ 184) maximal activity of 35.3% Ϯ 5.3% relative to 1 nM R1881. In contrast, TFM-4AS-1 (10 M) showed an activity that  (43). B, ligand displacement assays with native AR (apo-AR) and truncated rhARLBD from yeast (AR-LBD). Binding activities from reactions in the absence of unlabeled ligand competition were identified as maximal binding. The nonspecific binding was determined in the presence of 500-fold excess unlabeled ligand (35). DECEMBER 25, 2009 • VOLUME 284 • NUMBER 52 was only slightly higher than that of hydroxyflutamide or bicalutamide (Fig. 2C). On the average, its maximal activity was only 5.7 Ϯ 1.9% (n ϭ 179) of 1 nM R1881. Moreover, 1 M TFM-4AS-1 antagonized the N-C interaction induced by 10 nM CL-4AS-1 (Fig. 2C). These data suggest that Cl-4AS-1 activates the N-C interaction much like an agonist (albeit to a lesser degree), but that TFM-4AS-1 does not.

TFM-4AS-1 Is a Novel Selective Androgen Receptor Modulator
TFM-4AS-1 Differentially Modulates the Expression of Native Androgen-responsive Genes-We then examined native gene regulation by TFM-4AS-1 in MDA-MB-453 cells. Cells were treated for 18 h and using microarrays, 294 DHT-responsive genes were identified (supplemental Table S1). The results were displayed on a heatmap where the order of experiments is fixed on the y-axis, and genes are clustered on the x-axis by correlation coefficient (Fig. 3A). Next, for each gene the effect of 200 nM DHT was set to 100%, and all other conditions were normalized to this value to calculate potency and allow statistical comparisons. DHT dose-dependently altered the expression of genes with an average EC 50 of 0.97 nM (Fig. 3B). Cl-4AS-1 (1 M) behaved like DHT (Fig. 3, A and B), with transcript changes on average equaling 90.6 Ϯ 37.1% of 200 nM DHT. TFM-4AS-1 also regulated many genes similarly to DHT (Fig. 3A) with an EC 50 of 10.85 nM, close to its half-maximal effects in binding assays and MMTV transactivation (Fig. 3B). However, at 1 M, ϳ100-fold greater than its EC 50 , TFM-4AS-1 transcript changes were on average 74.6 Ϯ 45.5% the effect of
This analysis indicated that TFM-4AS-1 was less effective than DHT and Cl-4AS-1 at inducing this gene set. To assess whether this sub-effective response was due to all genes being inefficiently induced, or rather to some genes responding fully and others not responding, we compared the variation in gene response between treatments. We first compared the effect of 1   DECEMBER 25, 2009 • VOLUME 284 • NUMBER 52 was the ligand, while another ϳ25% are less responsive. To illustrate this result, we selected the genes UGT2B7 and FGF18 (Fig. 3D). Both genes were induced ϳ10-fold by DHT. For FGF18, the response to the three ligands was indistinguishable. In contrast for UGT2B7, TFM-4AS-1 only induces the RNA 2.7-fold at 1000 M. The responses of UGT2B7 and FGF18 to DHT were insensitive to cycloheximide, which blocks transcriptional effects requiring translation of another gene product (Fig.  3E) but were blocked by bicalutamide, suggesting direct AR involvement (Fig. 3E). Sequence analysis detected potential AREs in the 5Ј-regions of both genes (data not shown). Thus FGF18 and UGT2B7 are potentially both direct transcriptional targets of AR, supporting the hypothesis that TFM-4AS-1 has selective effects on native AR target genes within a uniform cellular context depending on promoter context.

Distinct Effects of TFM-4AS-1 and CL-4AS-1 on Bone Formation and Uterus
Weight-We used double calcein labeling to quantify changes in BFR in the periosteum, the external bone surface that in rats is highly androgen-sensitive. Alendronate was included to isolate the anabolic effect of andro-gens by eliminating any periosteal bone formation occurring secondary to catabolic increases in bone resorption as these processes are coupled. BFR measurements in the femora revealed dose-dependent stimulatory effects of both compounds on the periosteum. TFM-4AS-1 dosed sc for 24 days at 10 mg/kg/day increased the periosteal double-labeled surface, mineral apposition rate, and bone formation rate to levels similar to 3 mg/kg/day sc DHT, the lowest fully effective DHT dose as determined in pilot experiments (Fig. 4, A-C).
In rats, DHT has a pronounced trophic effect on uterus. Whereas DHT increased uterine weight, TFM-4AS-1 produced little or no effect at doses that fully induced bone formation (Fig. 4D). In contrast, Cl-4AS-1 produced significant increases in uterine weight at all the doses that stimulated bone formation, indicating TFM-4AS-1 has tissue-selective effects in vivo.
Effects of TFM-4AS-1 on Body Composition-We then compared the effects of TFM-4AS-1 to DHT on lean body mass (LBM) and fat mass (FM) in OVX rats. Treatments for 6 weeks with 3 mg/kg/day DHT sc or 10 mg/kg/day sc TFM-4AS-1 significantly increased LBM by 16.5 grams (118%) and 11.4 grams (81%) above that observed in vehicletreated rats, respectively (Fig. 4E). The LBM increase caused by TFM-4AS-1 was significantly different than vehicle group but not from the DHT-treated group. DHT significantly decreased FM, whereas the reduction by TFM-4AS-1 was not significant (Fig. 4F). DHT again caused a ϳ4-fold increase in uterus weight whereas TFM-4AS-1 did not. Thus TFM-4AS-1 showed anabolic activity without uterotrophic activity.
Effects of TFM-4AS-1 on Sebaceous Gland Formation-To evaluate the potential of TFM-4AS-1 to stimulate the pilosebaceous unit, histomorphometric measurements of sebaceous gland area in dorsal skin were performed in aged OVX female rats after 24 days of treatment with the maximally anabolic doses of 10 mg/kg/day for TFM-4AS-1 and 3 mg/kg/day DHT. In pilot experiments in this OVX rat model, DHT increased sebaceous gland area with a maximal response occurring within 4 days, whereas the anti-androgen cyproterone acetate produced little or no increase in sebaceous gland size after 28 days. 8 Similar to the previous experiment (Fig. 4), DHT and TFM-4AS-1 significantly increased bone formation rate by 204 and 308%, respectively (Fig. 5A). In the same animals, DHT increased sebaceous gland mean area by 108% and uterus weight nearly 400% (Fig. 5, B and C). In contrast, 10 mpk TFM-4AS-1 increased gland mean area by 33% and did not increase uterus weight. These data indicate that TFM-4AS-1 has reduced effects on the pilosebaceous unit and the uterus at anabolic doses.
Differential Effects of TFM-4AS-1 and Cl-4AS-1 on Prostate and Seminal Vesicle Growth-The above data suggested that TFM-4AS-1 acts as a SARM; however, it remained possible that this property results from limited tissue distribution or 5␣-reductase inhibition. Therefore, we compared prostate or seminal vesicle weight in three groups of rats: sham castrated, castrated, or castrated/DHT-supplemented treated with vehicle, TFM-4AS-1, or Cl-4AS-1. Sexually mature males were castrated (orchiectomized, ORX) or sham-operated the day before treatment and dosed with TFM-4AS-1 or Cl-4AS-1 (10 mg/kg/day sc) for 7 days. In shamoperated rats, TFM-4AS-1 significantly reduced ventral prostate weights by 50%, demonstrating its antagonism of endogenous androgens (Fig. 5D). In castrated rats, TFM-4AS-1 caused a small (10%) but significant increase in ventral prostate weight. Treatment for 17 days did not further increase prostate weight (data not shown). In contrast, Cl-4AS-1 produced no significant reduction in prostate weight in intact animals and in castrated rats caused a significant increase of ventral prostate weight to nearly 70% of controls. These data indicate that Cl-4AS-1 behaves like DHT in stimulating the prostate gland, whereas TFM-4AS-1 has reduced effects.
We then compared TFM-4AS-1 to the potent AR antagonist bicalutamide on seminal vesicle weight in ORX males supplemented with DHT pellets. DHT is the product of 5␣-reductase, so providing it eliminates any contribution of 5␣-reductase inhibition. ORX male rats were implanted with DHT pellets and treated daily for 14 days with the indicated doses. Castration reduced seminal vesicle weight by 95% compared with sham-operated rats (Fig.  5E). In ORX rats, the 2.5-mg DHT pellet restored seminal vesicle weight to 96% of sham-treated intact males. Combined treatment of DHT with TFM-4AS-1 resulted in a dose-dependent inhibition of the DHT-induced increase in seminal vesicle weight (55% reduction by 30 mg/kg/ day TFM-4AS-1). Higher doses of TFM-4AS-1 did not achieve greater plasma concentrations and were not tested. In contrast treatment with the antagonist bicalutamide at 10 mg/kg/day resulted in a 97% reduction of the DHT effect. These results confirm that TFM-4AS-1 is a partial agonist in seminal vesicles.
Gene Selectivity in Prostate Tissue-To understand the basis of the reduced effects of TFM-4AS-1 in the male reproductive tract, microarray analyses were performed on the rat ventral prostate 6 and 24 h after a single injection of 3 mg/kg DHT or 10 mg/kg TFM-4AS-1. These time points were chosen to enrich for direct transcriptional events (38). One-dimensional agglomerative clustering of the DHT-regulated RNAs (38) revealed that TFM-4AS-1 generally did not induce or repress OVX rats were dosed for 24 days with DHT or TFM-4AS-1 at fully anabolic exposures: A, bone formation rate determined from double-calcein labeling; B, mean sebaceous gland area determined by quantitative histomorphometry of dorsal back skin (nϾ3 fields per specimen, 10 -16 per group); and C, uterine wet weight. All values are Ϯ S.E., n ϭ 10 -16. *, different from vehicle (p Ͻ 0.05, Kruskal-Wallis). D, effects of Cl-4AS-1 and TFM-4AS-1 on rat prostate. Intact or castrated (ORX) rats treated for 7 days with the indicated compound (10 mpk) and prostate wet weights were measured and expressed as percent of body weight (mean of 8 animals, Ϯ S.E.). Note that Cl-4AS-1 is less effective at reducing prostate wet weight, and more effective at restoring prostate weight in ORX rats than is TFM-4AS-1. E, TFM-4AS-1 and bicalutamide antagonize the stimulations of SV growth by DHT. Seminal vesicle weight in ORX or mock-castrated (SHAM) animals treated for 14 days with a DHT pellet designed to provide DHT at a constant level regardless of gonadotropins or 5␣-reductase inhibition. Bicalutamide fully inhibited the effects of DHT whereas TFM-4AS-1 partially inhibited DHT at 30 mg/kg/day. All values are Ϯ S.E., n ϭ 9. *, indicates different from control (p Ͻ 0.05, Kruskal-Wallis). Control represents vehicle-treated ORX animals with only DHT pellets. DECEMBER 25, 2009 • VOLUME 284 • NUMBER 52 these transcripts to the same extent as DHT. However, closer inspection suggested that some genes are regulated similarly by both (Fig. 6A). To explore this gene selectivity at an individual transcript level, we selected four DHT-responsive transcripts expressed in the prostate, uterocalin (lipocalin 2, Lcn2), fibroblast growth factor receptor 4 (Fgfr4), insulin-like growth factor-1 (IGF-1), and cyclin D1, and analyzed them in a time course experiment (days 0, 1, 4, and 7) where ORX rats were treated once daily. Total prostate RNA was collected at each time point and analyzed by quantitative RT-PCR. Uterocalin and Fgfr4 were similarly responsive to both DHT and TFM-4AS-1 over the course of this study (Fig. 6B). In contrast, IGF-1 and cyclin D1 transcripts were induced only by DHT, with no effect of TFM-4AS-1 at any time point. These data further illustrate the gene selective activity of TFM-4AS-1 relative to DHT.

Identification of a Non-steroidal SARM Devoid of 5␣-Reductase
Activity-Whereas these data support the hypothesis that partial transactivation activity coupled with limited ability to induce the N/C-terminal interaction would be hallmarks of SARMs, TFM-4AS-1 is a steroidal compound that in principal could be affecting any of the number of proteins that recognize the steroidal structure (in addition to 5␣-reductase). Thus we wanted to confirm the observations with TFM-4AS-1 using a non-steroidal SARM with no 5␣-reductase inhibition. A high-throughput screen identified a class of AR ligands with partial agonist activity; structural optimization yielded FTBU-1 (Fig.  7A). FTBU-1 has an AR binding IC 50 of 38 nM, shows no significant binding activity toward GR, PR, MR, or ER, and is devoid of 5␣-reductase inhibition. This compound exhibited partial agonist activity in the MMTV transactivation assay (81% of maximal activity) and did not fully stimulate the N/C interaction (5% of maximal activity, Fig. 7B). A separate microarray study in MDA-MB-453 cells was conducted comparing FTBU-1 to DHT. As suggested by the intermediate partial agonist effects in transcription assays, 1 M FTBU-1 regulated the 294 previously identified DHT-sensitive transcripts in MDA-MB-453 cells to a mean 83.2 Ϯ 29.4% relative to 200 nM DHT, a significant difference (p Ͻ 0.0001, Student's t test) and an intermediate value between those observed with TFM-4AS-1 and Cl-4AS-1. Nearly one-third of all RNAs were Ͼ2 S.D. less than the mean for 20 nM DHT (Fig. 7C). Based on these data, FTBU-1 was judged to be a partial agonist like TFM-4AS-1, although with slightly more agonistic activity, and thus was tested in OVX rats. The lowest FTBU-1 dose tested, 10 mg/kg/ day sc for 24 days, produced an anabolic response in the periosteum equivalent to that of 3 mg/kg/day sc DHT, but had no significant effect on uterus weight (Fig. 7D). At 10 mg/kg/day, the exposure area under the curve was 77 M.h (0 -24 h). Dosing 9-fold higher, 90 mg/kg/day sc, produced an exposure ϳ8-fold higher (600 M⅐h) and also provided a fully osteoanabolic stimulus, but significantly increased uterus weight by 233% (Fig. 7D), about half the effect of DHT (477%). Therefore, at exposures that were at least 8-fold greater than required for osteoanabolism, FTBU-1 still produced less uterotropic effects than DHT.  Fig. 2) showing the gene expression effects of DHT and TFM-4AS-1 in prostate tissue from castrated males 6 and 24 h after a single injection. Note that TFM-4AS-1 alters the expression of some but not all DHT-regulated genes. Gene expression values are the mean of three measurements each from pooled RNA from three specimens. B, actions of DHT and TFM-AAS-1 on gene expression in prostate was measured by quantitative RT-PCR data for four selected genes, uterocalin (lipocalin 2, Lcn2), FGFR4, IGF-1, and cyclin D1. The activities were normalized to cyclophilin. Nine days after surgery, ORX rats were daily treated with vehicle, 3 mg/kg DHT, or 10 mg/kg TFM-4AS-1 (n ϭ 6/group). Prostate samples were collected at 0.25, 1, 4, and 7 days of dosing. Note that over time TFM-4AS-1 altered uterocalin and FGFR4 RNAs similarly to DHT, whereas it had no effect on cyclin D1 and IGF-1 at any time point. Values are normalized to cyclophilin levels measured within the same reaction and are the mean of six measurements from individual animals Ϯ S.D.; *, different than vehicle treatment values (p Ͻ 0.05 ANOVA).

DISCUSSION
For androgen replacement to become widely used it will be necessary to identify androgens with anabolic activities but reduced propensity to induce male secondary sexual traits and stimulate reproductive organs. Here we describe TFM-4AS-1, an anabolic AR ligand with limited effects on reproductive tissues and sebaceous glands. The structurally similar AR ligands, Cl-4AS-1 and TFM-4AS-1, exhibited the profiles of an agonist and a partial agonist, respectively. Cl-4AS-1 fully transactivated the MMTV promoter and repressed the activity of MMP1 promoter. In contrast TFM-4AS-1 only partially (ϳ50%) transactivated the MMTV promoter and antagonized 50% of the activity of a full agonist; thus it can be characterized as a ligand with mixed agonist and antagonist activities. Similar to an AR antagonist, TFM-4AS-1 does not repress the AP-1-sensitive MMP-1 reporter or induce the N/C interaction. These differences in transcriptional activity were correlated with effects on native gene transcription in MDA-MB-453 cells and in rat models where Cl-4AS-1 behaved like DHT whereas TFM-4AS-1 exhibited gene and tissue selectivity.
Cl-4AS-1 and TFM-4AS-1 are highly similar structures but cause significant differences in receptor function. Because they do not promote the N/C interaction equally, they likely produce differences in LBD conformation; crystallography data would provide specific insight into the structural bases of their biochemical differences. Mutations clustered in the AR LBD that influence the structure formed by helices 3, 4, 5, and 12 and that inhibit the N/C interaction were identified in individuals with reduced virilization (24 -31). By not supporting the N/C interaction, TFM-4AS-1 mimics these mutations. These data suggest that the molecular mechanism for TFM-4AS-1 selectivity, at least in part, is due to inducing an AR structure that does not support the N/C interaction.
Analysis of native gene regulation in MDA-MB-453 cells, which endogenously express AR, revealed that TFM-4AS-1 regulates a similar set of genes as DHT, but differs quantitatively in terms of maximal effect. For example, TFM-4AS-1 affects the expression of some genes, such as UGT2B7, much like DHT. In contrast, FGF18 is significantly less induced by TFM-4AS-1 than DHT. These data are intriguing because many models of tissue-selective actions of nuclear receptor ligands invoke the presence of tissue-specific transcriptional cofactors. In these models, activity in one tissue accompanied by lack of activity in another is explained by the selective expression of permissive cofactors. However, TFM-4AS-1 acts as a SARM within a uniform cellular context. Similar data were observed in the prostate, where TFM-4AS-1 regulated uterocalin and FGFR4 similarly to DHT whereas the DHT-responsive genes IGF-1 and cyclin D1 were unaffected. Thus, we propose that TFM-4AS-1 exhibits tissue selectivity because within each cell type it regulates a subset of AR-responsive genes: in some tissues this subset is sufficient to generate a physiological response and in others it is not. In this regard, SARMs should provide valuable insight into molecular requirements for androgenic effects.
Our experiments in castrated rats demonstrated that TFM-4AS-1 is a partial agonist that partially antagonizes both endogenous androgens and co-dosed DHT. However, poor solubility prevented us from testing whether it maintains tissue-selectivity at higher exposures. We identified a non-steroidal SARM, FTBU-1, which has no 5␣-reductase activity and improved solubility. This compound closely mimics the in vitro transcriptional profile of TFM-4AS-1, albeit with higher agonistic activity (81% versus 55% MMTV transactivation). Like TFM-4AS-1, FTBU-1 had little effect on the uterus at anabolic doses and at exposures Ն8-fold above those required for anabolism FTBU-1 exhibited 50% less uterotrophic activity than DHT. While based on our experience with related compounds, we suspect that this uterotrophic effect is caused by the higher agonism. 9 It remains  DECEMBER 25, 2009 • VOLUME 284 • NUMBER 52 possible that at high doses, tissue-selective SARMs could cause unwanted effects.

TFM-4AS-1 Is a Novel Selective Androgen Receptor Modulator
The transcriptional profile of our SARMs is distinct from that of the SERMs. Raloxifene, an osteoprotective ER ligand that lacks the agonistic activities of estradiol in breast and uterus (44 -47), and is an ER antagonist in transactivation assays but represses ER-controlled AP-1 binding sites (46 -48). In contrast, TFM-4AS-1 is an agonist in transactivation assays and does not inhibit AP-1-mediated transcription of the MMP-1 reporter. Unlike the significant body of information describing the clinical properties of SERMs, little clinical information is available regarding the actions of SARMs. However, in a 12-week study in healthy postmenopausal subjects, an AR ligand, MK-0773, which was selected based on similarity to TFM-4AS-1, exhibited SARM-like properties by increasing LBM without affecting markers of skin virilization or endometrial proliferation (49). Thus the properties of SARMs described here might translate into patients and apply broadly to the discovery of new therapeutic androgens.