Rational optimization of a transcription factor activation domain inhibitor

The study addresses whether the intrinsically disordered activation domain (AD) of the androgen receptor (AR) can be rationally targeted with small molecules by leveraging its phase-separation properties. Transcription factors (TFs) are pivotal in oncogenesis but are often deemed undruggable due to disordered regions critical for activity. In prostate cancer, anti-androgens target the AR ligand-binding domain (LBD), yet castration-resistant prostate cancer (CRPC) frequently emerges with constitutively active AR splice variants (e.g., AR-V7) that lack the LBD and are insensitive to such therapies. Intrinsically disordered regions (IDRs) in TFs can drive liquid–liquid phase separation and partitioning into transcriptional condensates with effectors like Mediator (MED1) and RNA polymerase II, and mutations that alter phase separation can modulate transcriptional activity. The authors hypothesize that deciphering the sequence and structural determinants of AR AD phase separation could enable rational optimization of small molecules to inhibit AR-driven oncogenic transcription, including LBD-independent variants relevant in CRPC.

Prior work established that IDRs can mediate liquid–liquid phase separation and formation of biomolecular condensates, including TF condensates with Mediator and RNAPII. For several TFs, mutations affecting phase separation correlate with altered transcriptional activity. AR forms nuclear speckles and condensates with coactivators; resistance to LBD-targeting anti-androgens is associated with AR splice variants lacking the LBD. EPI-001, a small molecule identified by phenotypic screening, targets AR AD (Tau-5) but has limited potency. Structural studies and simulations indicated that EPI-001 forms dynamic complexes with partially helical Tau-5, stabilized by interactions with aromatic residues. These insights suggest targeting transient structures within disordered ADs can be feasible for drug development.

- In vitro biophysics: Recombinant AR activation domain (AD, residues 1–558) and fragments (including Tau-5) were expressed and purified. Solution NMR (1H–15N spectra; triple-resonance 3D experiments) mapped residues engaged in transient interactions and assessed helical propensities, including TFE titrations. Turbidity assays measured cloud point temperatures (Tc; LCST behavior) across salt, temperature, and concentration to quantify phase separation propensities. - Mutagenesis: Systematic Tyr-to-Ser (Y→S) substitutions generated 8YtoS, 14YtoS, and 22YtoS AD variants. Proline substitutions (helix-breaking) were introduced within/adjacent to predicted helical motifs in polyQ-flanking regions, Tau-1, and Tau-5 to assess contributions of transient helices to condensation. - In vitro condensate assays: Fluorescently labeled AR AD, MED1 IDR, and RNAPII CTD were used to form droplets. Partitioning and mixing in heterotypic condensates were imaged by confocal microscopy; FRAP characterized dynamics. - Cell imaging and functional assays: eGFP-tagged full-length AR, AR-V7, and mutants (including ΔNLS AR-ΔNLS to stabilize cytosolic condensates) were transiently expressed in PC3 cells for live-cell imaging of condensate formation and nuclear translocation kinetics upon DHT treatment. Granularity (nuclear clustering) was quantified. - Interactome profiling: Proximity-dependent BioID-MS (Mini-TurboID fusions) in PC3 and LNCaP cells mapped AR interactomes (pre/post DHT), analyzed with SAINTq; STRING-based enrichment; validation via proximity ligation assays (PLA) for AR interactions with MED1, ARID1A, and NUP153. - Transcriptional readouts: Luciferase reporter assays (ARE-driven and AR-V7-specific V7BS3) quantified transcriptional activity for AR/AR-V7 and mutants in HEK293T/LNCaP cells. - Small-molecule design and characterization: EPI-002 analogs were rationally synthesized by modifying the linker between aromatic rings (single/double/triple bonds) and adding substituents (R1/R2) to optimize aromaticity/hydrophobicity. Potency assessed via PSA-luciferase assays; ChromLogD measured hydrophobicity. NMR CSPs on Tau-5* probed binding; REST2 MD simulations quantified contact probabilities and simulated Kd values. HPLC quantified partition coefficients of EPI-001 into AD condensates. - Compound mechanistic assays: Effects of compounds (EPI-001; optimized analog lae/1ae) on AR AD phase separation (turbidity, droplet size distributions), AR interactome (BioID-MS), and AR–coactivator proximity (PLA) were measured. - Transcriptomics and proliferation: RNA-seq of LNCaP cells treated with compounds (IC10/IC50, 6 h and 24 h) with DESeq2 analysis; GSEA (MSigDB hallmarks) for pathway perturbations. Viability/proliferation assays in LNCaP/LNCaP95. - In vivo efficacy: Human CRPC xenografts (LNCaP and AR-V7-high LNCaP95-D3) in castrated NOD-scid IL2Rγnull mice treated daily by oral gavage with lae (30 mg/kg), enzalutamide (10 mg/kg), or vehicle. Tumor volumes/body weights monitored; qPCR assessed on-target gene expression in tumors.

- Aromatic residues drive AR AD condensation: - NMR intensity analyses identified hydrophobic and aromatic residues, especially clustered near the 23FQNLF27 motif and in Tau-5, as hotspots for transient intermolecular interactions. - Tyr-to-Ser substitutions reduced phase separation: under conditions where WT* AD phase-separated with Tc ≈ 34 °C, 8YtoS and 14YtoS mutants required higher concentration/ionic strength to phase-separate (Tc ≈ 31 °C and 48 °C at higher conditions), and 22YtoS did not phase separate below 60 °C; overall, aromatic depletion markedly decreased LLPS capacity. - WT* AD partitioned into MED1 IDR and RNAPII CTD droplets; 22YtoS partitioning was reduced. In MED1 IDR+RNAPII CTD biphasic droplets, addition of WT* AD blended phases into one, whereas 22YtoS preferentially partitioned into MED1 IDR, indicating aromaticity-dependent mixing. - Phase separation links to AR function: - In PC3 cells, WT AR formed nuclear condensates and rapidly translocated to the nucleus upon DHT. Y→S mutants (8YtoS, 14YtoS, 22YtoS) failed to form condensates and showed reduced/slower nuclear translocation; 22YtoS remained cytosolic at 60 min. - AR-V7 granularity (nuclear clustering) decreased with Y→S mutants; replacing tyrosines with phenylalanines (22YtoF) restored AR-V7-driven gene expression to WT levels. - BioID-MS: After DHT, WT AR interactors enriched for transcriptional machinery (including known AR partners). 22YtoS showed fewer interactors, enriched for nuclear transport/nucleoporins; PLA confirmed loss of WT interactions with MED1 and ARID1A and increased 22YtoS association with NUP153. - Luciferase assays: Y→S substitutions reduced transcriptional activity of full-length AR and AR-V7. - Transient helices enhance condensation: - NMR identified seven regions with helical propensity (polyQ-flanking, Tau-1 LKDIL, Tau-5 WAAAAAQ and WHTLF motifs). TFE increased helicity and decreased AD Tc (−12 °C at 2.5% TFE; −35 °C at 5% TFE), indicating helices promote LLPS. - Helix-breaking proline substitutions in Tau-5 increased Tc by ~10 °C (reduced phase separation) and reduced number/size of cytosolic AR-ΔNLS condensates in cells; Tau-1 helix mutations had smaller effects. - Androgen-induced oligomerization promotes condensation: - AR activation creates N/C interaction and LBD dimerization, increasing valency and lowering LLPS threshold. In vitro, mixing AD with hormone-bound LBD induced droplets containing both domains; deleting FQNLF reduced droplet size/number. - Small-molecule optimization and mechanism: - EPI-001 partitions strongly into AD condensates: partition coefficient PWT_EPI-001 ≈ 55; P8YtoS_EPI-001 ≈ 32 (≈40% lower), linking partitioning to AD aromaticity. - Linker engineering between aromatic rings (triple bond 1aa, double-bond isomers 2aa/3aa, single bond 4aa) improved potency versus EPI-002 in PSA-luc assay: EPI-002 IC50 8.44 ± 0.42 μM; 1aa 5.08 ± 0.38 μM; 2aa 6.56 ± 0.61 μM; 3aa 14.57 ± 1.62 μM; 4aa 10.42 ± 0.68 μM. - Substitutions modulating hydrophobicity/aromaticity further improved activity: 1ab (R2=CH3) IC50 1.07 ± 0.03 μM; 1ba (R1=CH3) 1.32 ± 0.09 μM; 1bb (R1=R2=CH3) 0.49 ± 0.08 μM; 1af (R2=t-Bu) 0.22 ± 0.03 μM; 1ae (R2=Ph; lae) 1.54 ± 0.11 μM. - NMR CSPs on Tau-5* were larger with 1aa than EPI-001; REST2 MD showed higher contact probabilities and lower simulated Kd for 1aa (1.4 ± 0.1 mM) versus EPI-002 (5.2 ± 0.4 mM), indicating stronger binding to aromatic/helix regions. - lae (1ae) inhibited AR-V7-driven transcription dose-dependently; EPI-002 required higher concentrations; enzalutamide had no effect on AR-V7. - Compounds reduced AD LLPS in vitro: both EPI-001 and lae decreased AD T1/2 (turbidity onset) and droplet sizes. - Impact on AR interactome and transcriptional programs: - BioID-MS in LNCaP cells: lae (5 μM) caused a stronger decrease than EPI-001 (10 μM) in AR interactions, with significant reduction in known AR interactors; Mediator subunit MED1 was most reduced. - PLA confirmed time-dependent decreases in AR–MED1 and AR–ARID1A proximity with both inhibitors. - RNA-seq: 24 h treatment with 25 μM EPI-001 altered 64 genes; 5 μM lae altered 231 genes. GSEA showed significant depletion of androgen response targets for both, with lae producing a more pronounced reduction across downregulated DEGs. At 5 μM lae, AR protein levels were unchanged. - Proliferation: lae reduced LNCaP viability with lower IC50 than EPI-001; lae inhibited AR-V7-driven proliferation where enzalutamide was ineffective. - In vivo efficacy: - In castrated mouse xenografts, daily lae (30 mg/kg) modestly but significantly reduced tumor volumes in both LNCaP (AR-FL) and LNCaP95-D3 (AR-V7-high) models versus vehicle; in the AR-V7-driven model, lae outperformed enzalutamide. No overt toxicity observed (stable body weight). Tumor gene expression confirmed on-target inhibition of androgen-induced genes in LNCaP and AR-V7 targets in LNCaP95-D3; housekeeping gene ALAS1 unaffected.

The findings link AR activation domain phase separation—driven by aromatic residues and enhanced by transient helices—to essential AR functions: nuclear translocation, condensate formation with transcriptional machinery, and transcriptional activation. Disrupting aromatic content or helical propensity impairs LLPS, diminishes AR transit through the nuclear pore (with increased nucleoporin engagement), reduces interactions with coactivators (e.g., MED1), and suppresses transcription. Androgen-induced oligomerization (N/C interaction and LBD homodimerization) increases valency and promotes AR condensation, rationalizing hormone-triggered condensates. These mechanistic insights guided optimization of AD-targeting small molecules. Compounds partition into AR condensates via hydrophobic/aromatic interactions and stabilize a transient helical pocket within Tau-5, thereby reducing AD LLPS and blocking coactivator engagement, which translates to suppression of AR-dependent transcriptional programs. Enhancing compound aromaticity/hydrophobicity improved partitioning and binding, yielding markedly higher potency than the progenitor EPI-001, including efficacy against AR-V7-driven transcription and tumor growth where LBD-directed antiandrogens fail. The work supports a broader paradigm: transient secondary structures within disordered, phase-separating regions can confer ‘druggability’ within biomolecular condensates, enabling rational design/optimization of inhibitors of ‘undruggable’ TF activation domains.

This study elucidates that AR activation domain condensation is primarily driven by aromatic residues and aided by short transient helices, underpinning nuclear translocation and transcriptional activation. Leveraging these principles, the authors rationally optimized EPI-001 to generate more potent analogs (e.g., lae/1ae), which bind more strongly to Tau-5, partition into AR condensates, disrupt AR–coactivator interactions, suppress AR/AR-V7 transcriptional programs, and reduce tumor growth in CRPC xenografts, including models resistant to LBD-targeting drugs. The work establishes a general framework for targeting intrinsically disordered activation domains of oncogenic TFs by exploiting the physicochemical principles of condensate formation and transient structural motifs. Future research could expand structure–activity relationships to further enhance potency/selectivity, evaluate pharmacokinetics and safety profiles, explore combination therapies (e.g., with LBD antagonists or CDK inhibitors), and apply this strategy to other TFs with disordered, condensation-prone activation domains.

- Protein constructs for in vitro biophysics employed an AR AD L26P variant (WT*) to mitigate aggregation, which may not fully replicate native behavior. - Full-length AR containing Tyr-to-Ser alterations could not be purified in sufficient quality/quantity for in vitro assays, limiting some mechanistic studies to AD fragments. - In vivo antitumor effects of lae, while significant, were modest in the tested xenograft models; optimization of efficacy, dosing, and pharmacokinetics remains necessary. - Transient overexpression systems in cell-based assays may not fully recapitulate endogenous expression levels and complex regulatory contexts.