IntroductionTransactivation domains (TADs) of transcription factors are enriched in intrinsically disordered regions (IDRs) that lack a stable three-dimensional structure. The plasticity of an IDR permits for dynamic and multiple alterations of its conformation, an ensemble, thereby extending its repertoire of interacting molecules to regulate cellular and biological functions. The ensemble properties of an IDR depend upon both its primary sequence and its cellular milieu which are influenced by the concentration of interacting molecules and post-translational modifications.1 The sensitivity of IDR ensembles to changes in the cellular milieu enables fine tuning in a cell-specific manner and for a transcription factor, in a gene-specific manner. Recently long IDRs have been suggested to impart specificity and decrease the search time of a transcription factor for binding to DNA sequence motifs.2 While progress has been made in understanding how protein disorder regulates and tunes cellular function, the lack of small molecules that bind to IDRs has limited advances in drug development against these important therapeutic targets.The first small molecule that directly binds to an IDR to reach clinical trials was ralaniten acetate (NCT02606123). This molecule binds to the N-terminal TAD of androgen receptor (AR) within activation function-1 (AF-1).3,4,5 AR-TAD comprises approximately 556 residues and is predominantly disordered. Several scaffolds and unique structural classes of AR-TAD inhibitors (ARTADIs) have been discovered that include sintokamides and multiple ralaniten-like molecules with different structural motifs.3,6,7,8,9,10,11,12 Interest in the development of ARTADIs is driven by the fact that they are not affected by common resistance mechanisms that evolve following treatment with therapeutics targeting the folded ligand-binding domain (LBD) of AR. These resistance mechanisms include alterations to AR-LBD, such as gain-of-function mutations and loss of LBD to yield constitutively active splice variants such as AR-V7 (AR3).13,14The advancement of a second ARTADI into clinical trials for the treatment of prostate cancer (NCT04421222; NCT05075577) provided proof-of-concept that drugs targeting an IDR are feasible. Termination of the clinical trial for first-generation ralaniten-acetate was due to pill burden which emphasized that its drug-like properties needed to be optimized.15 We have analysed more than 560 ARTADIs to optimize potency (see supplemental references8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25). This library of compounds provides an opportunity to study the impact of specific chemical structures of ARTADIs against AR activities in terms of affinity, potency, specificity, gene expression, in vivo efficacy and impact on cellular function and pathways; providing a roadmap for developing next-generation ARTADI compounds. Here, we explore the feasibility of ARTADIs as therapeutics in the presence of androgens as well as in cells and xenografts driven by AR-V7.ResultsHalogens on the phenyl rings of ralaniten improve potencyInspiration to enhance potency while still maintaining specificity came from the imaging agent, (6) iodoralaniten/EPI-10000, that was ≈10-fold more potent compared to (1) ralaniten/EPI-002 at blocking full-length AR and AR-V7 driven transcriptional activities and cellular proliferation.16 Consistent with those data, replacement of the iodine substituent with a single bromine or chlorine atom in (4) EPI-20000 and (5) EPI-12,000, respectively, was ≈5-fold more potent than (1). Adding a second chlorine atom gave (8) BU1-2 that resulted in ≈10-fold better potency than (1), (Fig. 1a–c). Thus, subsequent analogs had dichlorinated bisphenol rings.Fig. 1The alternative text for this image may have been generated using AI.Full size imageImproving the potency of compounds. a Structures of compounds (1–10). b CLogPs and IC50s for blocking androgen-induced PSA-luciferase activities in LNCaP cells for compounds (1–8). c Dose response curves of inhibition of androgen-induced PSA-luciferase activity. d Chemical structures showing the route to BU130/BU170. e, f IC50s for pairs of compounds which differed only by the presence (right, red) or absence (left, blue) of the chlorohydrin group using the PSA-luciferase reporter in androgen-induced LNCaP cells. g Table showing the IC50s derived from the colony formation assays in LNCaP cells or (h) PC3 cells. i Dose response curves from the colony formation assay using LNCaP (solid lines) or PC3 (dashed lines) cells. Blue lines: compounds lacking the chlorohydrin functional group. Red lines: matched compounds with the chlorohydrin. n.s.: not significant. ND: not detected. N/A: not available. Error bars: mean ± SEM. See Supplementary Fig. 1Removing primary and secondary alcoholsWe next endeavored to resolve the metabolic liabilities of (1) ralaniten, which resulted in its failure in clinical trials. To this end we modified the primary alcohol on C1 which was identified in the Phase I clinical trial as a major source of oxidation.15 A series of compounds were generated (11-19) with various functional groups replacing the primary alcohol; imidazole [(11), BU4-12], morpholine [(12), BU6-2], triazole [(13), BU121; (14), BU100], methylether [(15), BU5-12], methyl sulfone [(16), BU2-12], ethyl sulfone (17, BU2-22) and methyl sulfonamide [(18), BU27; (19), BU3-12; Fig. 1d]. Supplementary Fig. 1a shows structures of control compounds enzalutamide and masofaniten (EPI-7386). While all of these compounds were quite potent with IC50s between 1–4 µM, the CLogP generally became less favorable (Supplementary Fig. 1b,c). Compounds (16-19) which had methyl sulfone, ethyl sulfone or methyl sulfonamide groups were superior. Of these, (19) BU3-12 showed the greatest potency with the best pharmacological profile with limited evidence of off-target effects and a favorable CLogP. (19) BU3-12, aka EPI-7170, has shown therapeutic potential in vivo in multiple prostate cancer models.8,9,17,18The secondary alcohols on C3 and C20 of ralaniten are potential sites of metabolism, with the latter shown to be a target of glucuronidation.15 Removal of these secondary alcohols was investigated with compounds (3) EPI-087, (18) BU27 and (20) BU81 which lacked the chlorohydrin alcohol and had the other secondary alcohol converted to a tertiary alcohol to potentially block oxidation and provide steric hindrance to glucuronidation and sulfation. These compounds showed reasonable potency (Supplementary Fig. 1b,c). Unfortunately, the loss of the secondary alcohols or addition of methyl groups to create tertiary alcohols increased their CLogPs [CLogPs: (3), 4.07; (18), 5.15; (20), 5.55]. ClogPs over a value of 5 become challenging to evaluate in vitro or administer to animals due to poor solubility.Calculations showed that adding a single fluorine atom to each of the bridging methyl groups of (18) BU27 and (20) BU81 should reduce their CLogP values. Therefore, the difluorinated compound (21) (BU87, CLogP = 3.70); (22) (BU86, CLogP = 2.43); (23) (BU108, CLogP = 3.99); (24) (BU104, CLogP = 4.87); (25) (BU130, CLogP = 4.10); and (26) (BU170, CLogP = 2.83) were prepared and evaluated (Supplementary Fig. 1c, d). No obvious trend was observed between potency and ClogP values.In vivo testing using castrated hosts bearing LNCaP xenografts was employed to test several of the more potent ARTADIs designed to be metabolically stable. (17) BU2-22 and (19) BU3-12 were the most efficacious when compared to equal molarity doses of (8) BU1-2 and (2) ralaniten-acetate (EPI-506)(Supplementary Fig. 1e). At equal doses of 30 mg/kg body weight, (19) BU3-12 was better than (8) BU1-2; (14) BU100; and (18) BU27 (Supplementary Fig. 1f). Compounds (8) BU1-2 dosed daily at 20 and 30 mg/kg body weight and (18) BU27 at 30 mg/kg body weight had no significant impact on tumor growth. From these studies, (19) BU3-12 was selected as a compound of interest for further investigation.Loss of the chlorohydrin groupThe chlorohydrin group of ralaniten and other ARTADIs may be essential for their binding mechanism to AR-TAD.4,5,12 Here we investigated the potency and specificity of novel compounds lacking this chemical group compared to its matching chlorohydrin-containing analogue; a difference of a single atom. We expected that destroying the chlorohydrin would reduce the ability of the compound to bind to the AR-TAD and inhibit activity. Initial testing using the PSA-Luc reporter was inconclusive. In two comparisons [(7), BU122 vs (8), BU1-2; and (18), BU27 vs (19), BU3-12] highly significant differences in IC50 were observed (Fig. 1e, f). Additionally, adding a methyl group to the chlorohydrin as seen in (23) BU108 reduced potency ~4-fold supporting this hypothesis (Supplementary Fig. 1c). Conversely, the comparisons of (3) EPI-087 vs (1) EPI-002; and (21) BU87 vs (22) BU86 indicated that analogs lacking the chlorohydrin were more potent. The remaining comparisons were not significant (Fig. 1e, f and Supplementary Fig. 1c, g).A stronger link was seen when measuring growth and survival of AR-dependent LNCaP cells. In this case, loss of the chlorohydrin tended to impact potency (e.g., BU3-12 vs BU27; Fig. 1g). In the case of (8) BU1-2, while the difference was not significant there was a trend of improved potency associated with the chlorohydrin (Supplementary Fig. 1h). The least potent compounds against LNCaP growth and survival were (3) EPI-087 and (9) BU162 that lacked a chlorohydrin (Fig. 1g and Supplementary Fig. 1h).AR-deficient PC3 cells are independent of AR for growth and survival making this a model for off-target effects. The only compounds which had any measurable effect on PC3 survivability were (13) BU121 and (21) BU87 analogs which lack the chlorohydrin functional group (Fig. 1h, i). Taken together, these results imply that the chlorohydrin is generally important for on-target activity and potency. An interesting exception was seen where (25) BU130 (no chlorohydrin) showed similar or slightly better potency compared to (26) BU170 (chlorohydrin) with minimal effect on PC3 growth. Enzalutamide has outstanding potency and specificity against the growth and survival of LNCaP cells as expected (Fig. 1g, h and Supplementary Fig. 1h). Masofaniten (EPI-7386) is tricyclic ARTADI that lacks a chlorohydrin and recently failed clinical trials (NCT04421222; NCT05075577). Several compounds showed improved potency compared to EPI-7386 in their ability to inhibit LNCaP survival while maintaining a therapeutic window estimated to be greater than 10-fold over PC3 survival; namely (10) BU85, (19) BU3-12, and (22) BU86. Of these, compound (19) was superior (BU3-12; 2.09 µM vs. EPI-7386; 3.10 µM; p = 0.0048, Fig. 1g–i and Supplementary Fig. 1h).Comparison of the in vivo efficacy of analogs with and without a chlorohydrin against the growth of LNCaP xenografts revealed that the chlorohydrin analog was significantly better when comparing (18) BU27 versus (19) BU3-12 (Fig. 2a). As before, there was only a slight/negligible improvement with the chlorohydrin when comparing the efficacy of (25) BU130 versus (26) BU170 (Fig. 2b). Therefore, (19) BU3-12, (25) BU130 and (26) BU170 were selected for further analysis.Fig. 2The alternative text for this image may have been generated using AI.Full size imageEfficacy and IC50s to block endogenous expression of genes transcriptionally regulated by full-length AR. a LNCaP xenografts grown in castrated mice dosed by gavage daily with vehicle (VEH), BU27 (30 mg/kg), its matching analogue BU3-12 (30 mg/kg), and enzalutamide (10 mg/kg). b LNCaP xenografts grown in castrated mice dosed by gavage daily with vehicle (VEH), BU130 (30 mg/kg), its matching analogue BU170 (30 mg/kg), and enzalutamide (10 mg/kg). c mRNA levels of androgen-induced genes normalized to housekeeping gene SDHA from LNCaP cells. Data presented as mean ± SEM and normalized to DMSO vehicle control in presence of R1881 (n = 3 independent experiments). d Table showing IC50 values for each inhibitor for genes shown in c. The sum-of-squares F test was performed to compare IC50’s for each inhibitor against EPI-002 (ralaniten). e Transcript levels of androgen-repressed genes normalized to levels of housekeeping gene SDHA from LNCaP cells treated as in c. Data presented as mean ± SEM, and analyzed by one-way ANOVA with Dunnet’s correction (n = 3 independent experiments). ND: not detected. *p