Materials availability

All unique reagents generated in this study are available with a materials transfer agreement.

Experimental model and subject detailsCell culture

PC3 (ATCC; CRL-1435) and LNCaP clone FGC (ATCC; CRL-1740) cells were cultured in RPMI-1640 containing 4.5 g L–1 glucose (Glutamax, Gibco) supplemented with either 10% (vol/vol) charcoal-stripped serum (CSS, Thermo Fisher Scientific A3382101) or 5% FBS (vol/vol), as specified below, and antibiotics. Induction of transcriptional activation by the AR in experiments using 5% FBS cultured LNCaP cells (Fig. 6a–f and Extended Data Figs. 1a,d and 7a–e) was verified using high-resolution microscopy and quantitative reverse transcription PCR (qRT–PCR). HEK293T cells (ATCC; CRL-3216) and AR-eGFP Hela stable cells69 (a gift from the M. Pennuto lab) were maintained in DMEM containing 4.5 g L–1 glucose supplemented with 10% (vol/vol) charcoal-stripped FBS and antibiotics. LNCaP95 cells were obtained from S. R. Plymate (University of Washington) and cultured in phenol-red-free RPMI supplemented with 10% (vol/vol) charcoal-stripped FBS (Gibco) and antibiotics. Cells were cultured in a humidified atmosphere containing 5% CO2 at 37 °C. Cell cultures tested negative for mycoplasma contamination.

Human prostate cancer xenografts

All animal experiments adhere to regulatory and ethical standards and were approved by the University of British Columbia Animal Care Committee (A18-0077). Before any surgery, metaCAM (1 mg per kg body weight, 0.05 ml per 10 g body weight) was administered subcutaneously. Isoflurane was used as an anesthetic. CO2 was used to euthanize the animals. Six- to eight-week-old male mice (NOD-scid IL2Rgammanull) were maintained at the Animal Care Facility at the British Columbia Cancer Research Centre. Five million LNCaP cells were inoculated subcutaneously in a 1:1 volume of matrigel (Corning Discovery Labware). Tumors were measured daily using digital calipers, and the volume was calculated using the formula for ovoid volume: length × width × height × 0.5236. When xenograft volumes were approximately 100 mm3, the mice were castrated, and treatment dosing started one week later. Animals were dosed daily by oral gavage with 30 mg per kg body weight of 1ae, 10 mg per kg body weight enzalutamide, or vehicle (5% DMSO, 1.5% Tween-80, 1% CMC).

Cloning of constructs

The primers and synthetic genes used in this work are listed in Supplementary Data Table 6.

GFP-AR FL, V7, and ∆NLS cloning strategy

For peGFPC1-AR-∆NLS, the NLS sequence (RKLKK, corresponding to amino acids 629–633 of AR) of the eGFP-AR fusion protein70 was removed from peGFP-C1-AR (Addgene no. 28235) using the Q5 site-directed mutagenesis kit and primer design tools (New England BioLabs). Any clones found to have expansion or shrinkage of either the polyQ or polyG site in the AR were corrected by replacing the 1510-base-pair (bp) KpnI-KpnI fragment with that of the WT AR sequence from peGFP-C1-AR.

For peGFPC1A-V7, the V7 variant of AR was generated from peGFP-C1-AR using the Q5 site-directed mutagenesis kit and primer design tools (New England Biolabs). Any clones that were found to contain expansion or shrinkage of either the polyQ or polyG site in the AR were corrected by replacing the 1510-bp KpnI-KpnI fragment with that of the WT AR sequence from peGFP-C1-AR.

For monomeric eGFP (mEGFP) constructs, mEGFP was subcloned into vectors containing human AR (Addgene no. 29235) and AR-V7 (Addgene no. 86856) using Gibson assembly to create mEGFP-AR-FL and mEGFP-AR-V7 (referred to as AD+DBD+NLS in Fig. 1b and Extended Data Fig. 2a,b) mammalian expression vectors. AR-V7 contains a 16-amino-acid (aa) constitutively active NLS containing an exon that replaces the LBD exons in AR-FL71. The sequence downstream of the AR activation domain in AR-V7, containing the DBD and NLS, was subcloned into an mEGFP plasmid (Addgene no. 18696) using Gibson assembly to create the mEGFP-AR-V7-ΔAD (referred to as DBD+NLS in Fig. 1b and Extended Data Fig. 2a,b) expression vector.

AR tyrosine mutagenesis strategy Production of YtoS mutants for mammalian expression

The sequences were optimized for expression in human cells, synthesized and cloned into the pUC57 plasmid (high-copy AmpR) by GenScript Biotech. To enable simple excision from pUC57 and insertion into plasmids derived from peGFPC1-AR, two HindIII sites were included as flanks on the fragments. After digestion using HindIII, the resulting 1,722-bp fragments were excised from TBE agarose gels, purified using the E.Z.N.A. MicroElute Gel Extraction Kit (Omega Biotech) and ligated into HindIII-cut, calf intestinal alkaline phosphatase (CIP)-treated and gel-purified peGFPC1-AR, peGFPC1-AR ∆NLS or peGFPC1A-V7 plasmids to produce the YtoS mutants.

Production of YtoS mutants for bacterial expression

pDEST17 plasmids for bacterial recombinant production of AR AD YtoS mutants were synthesized by Thermo Fisher Scientific with open reading frame (ORF) sequences flanked with attB1 and attB2 sequences.

AR helix-breaking mutagenesis strategy pDONR221-AR-AD-WT

The DNA sequence corresponding to the 1,558-aa fragment of AR-AD was synthesized and encoded in a pDONR221 vector by Thermo Fisher Scientific (flanked with attB1 and attB2 sequences).

pDEST17-AR-AD-WT

pDONR221-AR-AD-WT was subcloned into a pDEST17 vector using the LP clonase reaction (Thermo Fisher Scientific).

pDEST17-AR-AD-WT*

L26P was introduced into a WT AR sequence (pDONR221-AR-AD-WT) using a Quickchange protocol with Pfu Turbo polymerase (Agilent), and the resulting plasmid with the L26P substitution (pDONR221-AR-AD-WT*) was subcloned into a pDEST17 vector using the LP clonase reaction (Thermo Fisher Scientific).

pDEST17-AR-AD-L56P*

L56P was introduced into the pDONR221-AR-AD-WT* (bearing L26P; described above) using a Quickchange protocol with Pfu Turbo polymerase (Agilent) to generate pDONR221-AR-AD-L56P*. The resulting plasmid with the L26P and L56P substitutions (pDONR221-AR-AD-L56P*) was subcloned into a pDEST17 vector using the LP clonase reaction (Thermo).

pDEST17-AR-AD-Tau-1*

The A186P, L192P and C238P substitutions were introduced in a step-wise manner into pDONR221-AR-AD-WT* (bearing L26P; described above) using a Quickchange protocol with Pfu Turbo polymerase (Agilent) to generate pDONR221-AR-AD-Tau-1*. The resulting plasmid with the L26P, A186P, L192P and C238P substitutions (pDONR221-AR-AD-Tau-1*) was subcloned into a pDEST17 vector using the LP clonase reaction (Thermo Fisher Scientific).

pDEST17-AR-AD-Tau-5*

The A356P, A398P and T435P substitutions were introduced in a step-wise manner into pDONR221-AR-AD-WT* (bearing L26P; described above) using a Quickchange protocol with Pfu Turbo polymerase (Agilent) to generate pDONR221-AR-AD-Tau-5*. The resulting plasmid with the L26P, A356P, A398P and T435P substitutions (pDONR221-AR-AD-Tau-5*) was subcloned into a pDEST17 vector using the LP clonase reaction (Thermo Fisher Scientific).

pDEST17-AR-AD-L56P+Tau-1+Tau-5*

The L56P, A186P, L192P and C238P substitutions were introduced in a step-wise manner into pDONR221-AR-AD-TAU-5* (bearing the L26P, A186P, L192P and C238P substitutions; described above) using a Quickchange protocol with Pfu Turbo polymerase (Agilent) to generate pDONR221-AR-AD-L56P+Tau-1+Tau-5*. The resulting plasmid containing the L26P, L56P, A186P, L192P, C238P, A356P, A398P and T435P substitutions (pDONR221-AR-AD-L56P+Tau-1+Tau-5*) was subcloned into a pDEST17 vector using the LP clonase reaction (Thermo Fisher Scientific).

eGFP-AR-ΔNLS-Δ21–35

A 507-bp fragment with deletion of residues 21–35 was amplified from pCMV5-FLAG-AR deltaFQNLF30 using KOD polymerase (Merck Millipore) and the supplied buffer no. 2. The resulting fragment was purified using AmPure XT (Beckman) before InFusion (Takara Bio) into SalI and AflII-cut and gel-purified peGFPC1-AR ∆NLS plasmid.

eGFP-AR-ΔNLS-Tau-1

The A186P, L192P and C238P substitutions were introduced in a step-wise manner into the WT AR sequence encoded in pDONR221-AR-AD-WT using a Quickchange protocol with Pfu Turbo polymerase (Agilent). A 755-bp fragment was amplified from the resulting clone, incorporating the A186P, L192P and C238P substitutions (pDONR221-AR-AD-TAU1), using KOD polymerase (Takara Bio). The resulting fragment was digested with DpnI to remove the template and purified using AmPure XT (Beckman) before InFusion into AflII-BstEII-cut and gel-purified peGFP-C1-ARΔNLS plasmid.

eGFP-AR-ΔNLS-Tau-5

The A356P, A398P and T435P substitutions were introduced in a step-wise manner into the WT AR sequence (pDONR221-AR-AD-WT) using a Quickchange protocol with Pfu Turbo polymerase (Agilent). A 1,544-bp fragment was then amplified from the resulting plasmid, incorporating the A356P, A398P and T435P substitutions (pDONR221-AR-AD-TAU-5), using KOD polymerase (Takara Bio). The resulting fragment was digested with DpnI to remove the template and purified using AmPure XT (Beckman) before InFusion into KpnI-cut and gel-purified peGFP-C1-ARΔNLS plasmid.

BioID plasmid-generation strategy

Constructs for expression of FLAG-MTID or its fusions with AR WT and 22YtoS were synthesized by Genscript and were either cloned into pcDNA3.1(–) and subsequently cloned into pLenti-CMV-MCS-GFP-SV-puro using XbaI and BamHI to replace GFP or cloned directly into pLenti-CMV-MCS-GFP-SV-puro by Genscript using the same sites. Sequences were codon optimized for mammalian expression and verified by sequencing. pLenti-CMV-MCS-GFP-SV-puro was a gift from P. Odgren (Addgene plasmid no. 73582).

Experiments in vitroExpression and purification of constructs

WT and mutant AR AD (1–558 aa) were recombinantly produced in E. coli Rosetta (DE3) cells that were transformed with pDEST17 plasmid encoding His-AR-AD, as described previously72. Briefly, cell cultures at an optical density of 600 nm (OD600) of 0.5 were induced with 0.1 mM IPTG at 22 °C overnight. Cells were lysed in PBS buffer and centrifuged. The pellet was solubilized overnight in Tris buffer (20 mM Tris, 500 mM NaCl, 5 mM Imidazole, pH 8) containing 8 M urea and 500 mM NaCl at pH 8. The protein was captured on Nickel columns (His Trap HP, GE Healthcare) and eluted with 500 mM imidazole. After urea removal by dialysis, the His-tag was cleaved by TEV protease at 4 °C overnight. Urea (8 M) was added to cleaved protein before reverse-nickel affinity chromatography to separate noncleaved protein and the His-tag. Protein in the flowthrough was concentrated, filtered and stored at −80 °C. To prevent protein aggregation or instability, an additional purification step was conducted, and the sample was run on a Superdex 200 16/600 column pre-equilibrated with AR AD buffer (20 mM sodium phosphate, 1 mM TCEP pH 7.4). Tau-5* (330–448 aa) was expressed and purified as previously described23, and an equivalent protocol was used to express and purify fragment AR AD (441–558 aa).

AR-LBD (663–919 aa) containing an amino-terminal His-tag and encoded in a pET15b plasmid (Addgene no. 89083) was expressed in Rosetta (DE3) cells with 1 mM IPTG at 16 °C overnight. Cells were resuspended in Ni-Wash buffer (25 mM HEPES, 500 mM NaCl, 10% glycerol, 1 mM DTT, 10 μM DHT, 1% Tween-20, 20 mM imidazole at pH 7.4), lysed and centrifuged. Soluble protein was captured by IMAC and eluted with 500 mM imidazole. During an overnight dialysis, His-tag was cleaved by thrombin (GE Healthcare), and the NaCl concentration was reduced to 100 mM. Cleaved protein was captured by cation exchange (GE Healthcare) and eluted with 1 M NaCl gradient. LBD was injected in a Superdex 200 16/600 column pre-equilibrated with 25 mM HEPES, 250 mM NaCl, 10% glycerol, 1 mM TCEP, 10 μM DHT, 1 mM EDTA and 0.5% Tween-20 at pH 7.2.

MED1 IDR (948–1573), encoded in a peTEC plasmid, was a gift from T. Graf. A 3C cleavage site was introduced by Q5 site-directed mutagenesis (New England Biolabs) between mCherry and the MED1 sequence, yielding peTEC-His-mcherry-3C-MED1-IDR plasmid. Protein was expressed in B834 (DE3) cells at 16 °C overnight with 0.1 mM IPTG. Upon cell lysis in Tris buffer with 100 mM NaCl, the soluble cell fraction was injected in a HisTrap HP column, and protein was eluted with 500 mM imidazole. The eluted protein was concentrated and separated by cation exchange chromatography. The collected fractions were cleaved by 3C protease, and MED1 IDR was separated from other protein fragments by size-exclusion chromatography (SEC) with Superdex 200 16/600 column pre-equilibrated with 20 mM sodium phosphate, 100 mM NaCl and 1 mM TCEP at pH 7.4.

RNAPII CTD (1592–1970) was produced in E. coli B834(DE3) cells transformed with the pDEST17 plasmid, which encodes H6-3C-RNAPII-CTD. The protein was expressed at 25 °C overnight with 0.1 mM IPTG and extracted from the insoluble cell fraction. The pellet was resuspended in Tris buffer with 8 M urea and loaded on a HisTrap HP column. Captured protein was dialyzed against 50 mM Tris-HCl, 50 mM NaCl and 1 M NaCl at pH 8 and was cleaved by 3C protease overnight at 4 °C. RNAPII CTD was injected in a Superdex 200 16/600 column pre-equilibrated with 20 mM sodium phosphate, 150 mM NaCl, 5% glycerol and 1 mM TCEP at pH 7.4.

AR-LBD, MED1-IDR and RNAPII-CTD fractions from SEC were concentrated, filtered and stored at −80 °C until further use.

Turbidity measurements

Protein samples were prepared in AR AD buffer (20 mM sodium phosphate, 1 mM TCEP pH 7.4), with the indicated protein and NaCl concentrations, on ice. Samples were centrifuged at 21,130 r.c.f. for 20 min at 4 °C, and the supernatant was transferred to a quartz cuvette. Phase separation Tc measurements of protein solutions were monitored by measuring the absorbance of the solutions at 340 nm as a function of temperature on a Cary 100 Multicell UV-vis spectrophotometer, equipped with a Peltier temperature controller, at a heating rate of 1 °C min–1. The Tc values were obtained as the maximum of the first-order derivative of the obtained curves from three independent samples.

Protein labeling

For in vitro condensation experiments, proteins were labeled with fluorescent dye instead of being tagged with fluorescent protein, to avoid nonspecific interactions in heterotypic condensates. AR AD and MED1 IDR were fluorescently labeled with Dylight 405 or Alexa Fluor 647, respectively, unless otherwise indicated in the figure legends. LBD and RNAPII-CTD were labeled with Oregon Green 488. In all cases, sulfhydryl-reactive dyes were used, which reacted to the naturally occurring cysteines of the protein, except for RNAPII-CTD in which an N-terminal Cys was added. Protein was labeled according to the manufacturer’s instructions for sulfhydryl-reactive dyes (Thermo Fisher Scientific). Briefly, protein and dye were mixed at a 1:20 ratio in each protein storage buffer, adjusted to pH 7.5 overnight at 4 °C. Then, 1 mM DTT was added to stop the reaction, and protein was separated from free dye with a pre-equilibrated PD-10 column. Protein was concentrated and filtered, and the concentration and conjugation efficiency were analyzed, following the manufacturer’s instructions for sulfhydryl-reactive dyes (Thermo Fisher Scientific).

Fluorescence microscopy of in vitro protein condensation

Each protein solution was prepared by adding approximately 1% of equivalent labeled protein. Solutions were stored on ice. Samples were prepared by mixing proteins at the indicated protein concentration with AR AD buffer (20 mM sodium phosphate, 1 mM TCEP pH 7.4) in low binding PCR tubes at RT. Once all proteins were mixed, the phase separation trigger was added: NaCl for AR samples, or Ficoll 70 for transcriptional component samples. Samples were homogenized, and 1.5 μl of sample was transferred into sealed chambers composed of a slide and a PEGylated coverslip sandwiching 3M 300 LSE high-temperature double-sided tape (0.34 mm). Coverslips were PEGylated according to the published protocol73. Images were taken using Zeiss LSM 780 Confocal Microscope with a Plan-ApoChromat ×63/1.4 Oil objective lens. Fluorescence recovery after photobleaching (FRAP) experiments were recorded using the same set-up on a 50 μM AR AD sample containing approximately 1% of protein labeled with DyLight 488 dye (Thermo Fisher Scientific) with 500 mM NaCl. The data were analyzed using the EasyFRAP software74 to extract the mobile fractions and recovery half-times.

NMR experiments Assignment strategy

All NMR experiments were performed at 5 °C (278 K) on either a Bruker 800 MHz (DRX or Avance NEO) or a Bruker Avance III 600 MHz spectrometer, both equipped with TCI cryoprobes, and versions 3.2 and 4.0.8 of TOPSPIN.

A 300 μM 15N,13C-double-labeled AR AD (441–558 aa) sample in NMR buffer (20 mM sodium phosphate (pH 7.4), 1 mM TCEP, 0.05% (wt:vol) NaN3) was used for backbone resonance assignment. The following series of three-dimensional (3D) triple resonance experiments were acquired: HNCO, HN(CA)CO, HNCA, HN(CO)CA, CBCANH and CBCA(CO)NH. Chemical shifts were deposited in the Biological Magnetic Resonance Bank (BMRB) (ID: 51476).

The assignment of AR AD (1–558 aa) was guided by the assignments obtained for the smaller AR fragments that were first studied here (residues 441–558) or previously reported (residues 1–151 (BMRB ID: 25607) and 142–448 (BMRB ID: 51479)). In addition, 3D HNCO and HNCA experiments were acquired for two 15N,13C-double-labeled AR AD (1–558 aa) samples (25 μM and 100 μM) dissolved in NMR buffer. For the 100 μM sample, additional 3D HN(CA)CO and HN(CO)CACB experiments were also recorded. Three-dimensional experiments were done using 25% non-uniform sampling. Chemical shifts were deposited in the BMRB (ID: 51480).

Backbone resonances of AR WT* were almost identical to those of AR AD (1–558 aa), with only local differences in residues around the position substituted (L26), which were assigned using non-uniform sampled 3D BEST-TROSY HNCO and HNCA experiments75 recorded on a 50 μM 15N,13C-double-labeled WT* AR AD sample dissolved in NMR buffer.

Site-specific and residue-type identification of oligomerization

The oligomerization of AR AD was monitored by recording the induced intensity changes on the two-dimensional 1H,15N correlation spectrum by adding increasing amounts of unlabeled sample on a 25 μM 15N-labeled AR AD to reach total concentrations of 57.5, 100.8, 122.5 or 155 μM. Spectra were recorded using 128 scans per increment (with an experimental time of 21 h per spectrum) to ensure that intensities in the regions with weaker signals were quantified properly. Throughout the article, the term oligomer refers to intermolecular complexes formed through weak, and therefore reversible, site-specific interactions between monomers.

Helicity studies upon TFE addition

The effect of TFE on 50 μM WT* AR AD and Tau-5* secondary structures were monitored by a series of 1H,15N correlation spectra and non-uniform sampled 3D BEST-TROSY, HNCO and HNCA experiments recorded in the presence of increasing TFE amounts (0%, 2.5% and 5%).

Binding studies

EPI-001 and 1aa binding to Tau-5* was studied by comparing 15N chemical shifts in 2D 1H,15N CP-HISQC76 spectra at 37 °C (310 K), using 60 μM Tau-5* in the absence or presence of 60 μM compounds (1:1 ratio). Samples contained NMR buffer (above) at pH 6.6 with 200 mM NaCl and 2% DMSO-d6. The CP-HISQC pulse sequence and the pH level of 6.6 were chosen to reduce water exchange of labile amide protons at 37 °C (310 K).

Data processing

Data processing was done using qMDD77 for non-uniform sampled data, and NMRPipe78 for all uniformly collected experiments. Data analysis was performed with CcpNmr Analysis79. Helix populations were extracted using the δ2D software67.

Peptides

FQNLFQ and FQNPFQ synthetic peptides were obtained as lyophilized powders with >95% purity from GenScript with amidated C and acetylated N termini. The lyophilized peptides were solubilized at a final concentration of 5 mM in DMSO. Immediately before each experiment, the stock solutions were diluted to 125 μM in 20 mM HEPES buffer, pH 7.5, with 150 mM NaCl. For aggregation assays, the samples were incubated overnight at 37 °C at 600 r.p.m. agitation. The term aggregate refers to the quasi-irreversible formation of fibrillar species stabilized by strong intermolecular interactions, involving a large conformational change.

Synchronous light scattering

Synchronous light scattering was monitored using a JASCO Spectrofluorometer FP-8200. The conditions of the spectra acquisition were: excitation wavelength of 360 nm, emission range from 350 to 370 nm, slit widths of 5 nm, 0.5-nm interval and 1,000 nm min–1 scan rate. The peptides were sonicated for 10 min in an ultrasonic bath (Fisher Scientific FB15052) before measurement.

Fourier transform infrared spectroscopy

Fourier transform infrared spectroscopy (FT-IR) experiments were performed using a Bruker Tensor 27 FT-IR spectrometer (Bruker Optics) with a Golden Gate MKII ATR accessory. Each spectrum consists of 16 independent scans, measured at a spectral resolution of 4 cm−1 within the 1,800–1,500 cm−1 range. All spectral data were acquired and normalized using the OPUS MIR Tensor 27 software. Data was afterwards deconvoluted using the Peak Fit 4.12 program. The buffer without peptide was used as a control and subtracted from the absorbance signal before deconvolution.

Transmission electron microscopy

The morphology of the aggregated FQNLFQ peptide was evaluated by negative staining using a JEOL JEM-1400Plus Transmission Electron Microscope. Five microliters of peptide solution was placed on carbon-coated copper grids and incubated for 5 min. The grids were then washed and stained with 5 μl of 2% wt/vol uranyl acetate for 5 min. Then, the grids were washed again before analysis. Images and videos were processed with ImageJ.

Cell imaingMicroscopy

PC3 cells were seeded in collagen-I-coated µ-slide four-well glass-bottom plates (Ibidi 80426) at 60% confluency 24 h before transfection. Then, 170 ng of expression vectors encoding WT AR tagged with eGFP (eGFP-AR) or mutant AR proteins were transfected per well using polyethylenimine (PEI) (Polysciences) at a ratio of 1 µg DNA to 3 µl PEI. Four hours after transfection, the medium was changed to RPMI supplemented with 10% charcoal-stripped FBS and cells were cultured for 16 h before imaging. Transiently transfected PC3 cells expressing eGFP-AR were imaged in 3D during 1 min, by taking one image every 15 s, to acquire a baseline readout of AR expression. Cells were then treated immediately with 1 nM of DHT and imaged during 1 h, also by taking an image every 15 s. Time-lapse imaging was performed in an Andor Revolution Spinning Disk Confocal with an Olympus IX81 microscope and a Yokogawa CSU-XI scanner unit. Images were acquired with an Olympus PlanApo N ×60/1.42 Oil objective lens. A stable temperature (37 °C) was maintained during imaging under CO2 in a temperature-regulated incubation chamber (EMBL). eGFP was excited with a 488 nm laser, and Z-stack images were acquired every 0.45 μm. Time-lapse images were compiled, processed and analyzed with Fiji (ImageJ)80. Intensity thresholds were set manually and uniformly to minimize background noise.

FLAG-MTID-AR-WT and PC3 FLAG-MTID-AR-WT-Y22toS cell lines were seeded in 24-well culture plates, on 12-mm sterilized coverslips. The next day, 50 μM biotin (or DMSO for a negative control) and 1 nM DHT were added for 2 h. The culture medium was removed and the cells were washed with PBS. Next, cells were fixed for 15 min with 4% paraformaldehyde. After fixation, cells were washed with PBS and then permeabilized with 0.1% Triton X-100 for 10 min. Coverslips were then washed and blocked with blocking buffer (3% BSA, 0.1%Tween, PBS) for 1 h at 37 °C. Coverslips were incubated with primary antibody—anti-AR (Abcam, ab108341, 1:100)—overnight. The next day, coverslips were washed with PBS, and secondary antibodies were added (1:500): anti-streptavidin antibody conjugated to Alexa Fluor 488 (Thermo Fisher Scientific, S11223) or Alexa Fluor 488-conjugated goat anti-rabbit-IgG (H+L) (Thermo Fisher Scientific, A11008). Fluorescence images were acquired using a Leica TCS SP8 confocal microscope. Images were taken with ×63 oil objectives, and standard LAS-AF software was used.

HEK293T cells in DMEM with 10% FBS were seeded at a density of 40,000 cells per well on glass-bottom chambered coverslips (Ibidi 80827). Sixteen hours later, wells were refreshed with 280 µl seeding medium and transfected with 50 ng of mEGFP expression plasmids, as shown in Extended Data Fig. 2a, using LipoD293 transfection reagent (SignaGen SL100668) according to the manufacturer’s protocol. Forty-eight hours later, wells were refreshed with medium spiked with 10 nM DHT or the equivalent DMSO control (vol/vol). Four hours after treatment, coverslips were imaged on a Zeiss LSM 880 confocal microscope with a Plan-ApoChromat ×63/1.4 Oil DIC objective lens in a CO2 incubation chamber set to 37 °C. Images were acquired across two biological replicates.

STED microscopy Sample preparation

HEK293T and HeLa eGFP-AR cells in DMEM with 10% FBS were seeded at a density of 40,000 cells per well on glass-bottom chambered coverslips (Ibidi 80827). Sixteen hours later, wells containing HEK293T cells were refreshed with 280 µl seeding medium and transfected with 50 ng of mEGFP expression plasmids, as shown in Fig. 1b, using LipoD293 transfection reagent (SignaGen SL100668), according to the manufacturer’s protocol. Forty-eight hours later, wells were refreshed with medium spiked with 10 nM DHT. Samples were imaged after 4 h of DHT treatment.

LNCaP cells (Clone FGC, ATCC CRL-1740) were seeded in RPMI-1640 5% FBS onto PLL-coated 18-mm no. 1.5 thickness glass coverslips (Sigma P4707, Roth LH23.1) at a density of 100,000 cells per coverslip on a 24-well plate. Sixteen hours later, the media was refreshed and cells were incubated further for another 24 h. For fixation, wells were washed with PBS, then fixed with 1 ml of 4% PFA in PBS for 20 min at room temperature. After a second wash in PBS, cells were permeabilized with 0.5% Triton X-100, PBS (vol/vol) (Sigma 93443) and then stained with anti-AR (AR 441, scbt 7305, 1:50) and STAR 635P secondary antibody (Abberior, ST635P-1001, 1:200). Nuclear translocation of the AR signal was validated by staining LNCaP cells grown in RPMI-1640 5% CSS (Gibco, A3382101), following the same protocol. DNA was counterstained with 1:2,000 Spy555-DNA (Spirochrome, SC201), and samples were mounted onto glass slides with vectashield (Biozol, VEC-H-1900-10).

Live-cell STED

HEK293T and HeLa cells were imaged on a Leica Stellaris STED DMI 8 microscope equipped with an okolab incubation chamber set to 37 °C and a constant concentration of CO2 (5%). eGFP imaging was performed using a 473-nm stimulation wavelength laser at 20% power and a 592-nm depletion laser at 20% power. Images were taken using a HC PL APO CS2 ×63/1.40 oil objective, with a final resolution of 23 nm pixel–1.

Stimulated emission depletion fluorescence-lifetime imaging microscopy

Fixed and stained LNCaP cells were imaged on a Leica Stellaris STED DMI 8 microscope. Abberior STAR 635P immunofluorescence imaging was performed using a laser with a stimulation wavelength of 633 nm at 5% power, and a 776-nm depletion laser at 5% power. Images were taken using a HC PL APO CS2 ×63/1.40 oil objective, with a final resolution of 48 nm pixel–1. Fluorescence-lifetime imaging microscopy cutoffs and fluorescence-lifetime stimulated emission depletion deconvolution strengths were determined automatically using Leica LAS-X software v2.5.6 to filter background photons with low fluorescence lifetimes (Extended Data Fig. 1d).

FRAP assay in live cells

PC3 cells were transfected and prepared for microscopy in identical conditions to those of the live-cell imaging experiments. Before performing FRAP assays, cells were treated with 1 nM DHT. FRAP data for each condition were acquired over the course of approximately 1 h after treatment, and results were combined for each condition because no trend was observed between FRAP data acquired at the beginning versus the end of the hour. FRAP measurements were performed on an Andor Revolution Spinning Disk Confocal microscope with a FRAPPA Photobleaching module and a iXon EMCCD Andor DU-897 camera. Images were taken using a ×100/1.40 Oil U Plan S-Apo objective lens. Pre-bleaching and fluorescence recovery images of the eGFP-AR were acquired using a laser power of 488 nm with an exposure time of 100 ms. Bleaching was done in a 10 × 10 pixel square region of interest (ROI) on top of a droplet, which was repeated five times; the maximum laser power intensity was 488 nm, and the dwell time for bleaching was 40 µs. Twenty pre-bleached images and 200 post-bleached images in total were taken at intervals of 180 ms. Post-bleached images were acquired immediately after the bleaching. Mean gray intensity measurements were quantified in three different ROIs in each FRAP experiment: a bleached region, a background region outside the cells and a region spanning the whole cell were drawn to allow normalization of the gray values. Fiji was used to measure it in each ROI using the plot z-axis profile function to extract the intensity data. Exported csv tables were normalized and fitted in EasyFrap software74 to extract kinetic parameters, such as T-half and mobile fraction. Double normalization was used to correct for fluorescence bleaching during imaging and for differences in intensity.

Drug-condensate interactionsDrug partition coefficient calculation

Concentrations of EPI-001 in the dense and light phases of WT* AR AD and 8YtoS AR AD were determined using the Agilent Technologies 1200 HPLC instrument, using a Jupiter analytical C4 column from Phenomenex. H2O and ACN:H2O (9:1) were used as mobile phases, containing 0.1% TFA.

Samples were prepared on ice in 20 mM sodium phosphate buffer (pH 7.4), 1 mM TCEP and 0.05% (wt/wt) NaN3. One equivalent of compound was added to 60 μM of protein from DMSO stocks. The final concentration of DMSO in all samples was 2%. Liquid–liquid phase separation of the protein was induced by adding 1.25 M NaCl, followed by incubation for 5 min at 37 °C and centrifugation at 2,000 r.p.m. for 2 min at 37 °C to separate the light and dense phases. The light phase was transferred to a new microcentrifuge tube, and the dense phase was diluted nine times by adding the buffer containing 4 M urea, which dissolvd the condensates. These solutions were injected in an HPLC system. The corresponding peaks of small molecules were integrated, and concentrations were determined using standard calibration curves that were obtained by measuring four concentrations for each compound.

Effect of compounds on AR AD phase separation in vitro

The effects of compounds on AR AD phase separation in vitro were assessed by turbidity (see ‘Turbidity measurements’) and microscopy (see ‘Fluorescence microscopy of in vitro protein condensation’). The samples contained 25 μM WT* AR AD with 1 molar equivalent of the indicated compounds in 20 mM sodium phosphate buffer (pH 7.4), 1 mM TCEP, 0.05% (wt/vol) NaN3, 1 M NaCl and 2% DMSO.

Experiments in cellsLuciferase reporter assay in HEK293T

HEK293T cells were co-transfected with an androgen-response element (ARE)-luciferase construct containing a luciferase reporter gene under the control of three AREs (kindly provided by the M. Pennuto lab), along with an empty vector, an AR-expression vector (pEGFP-C1-AR or AR V7) or different mutants in the presence or absence of DHT. HEK293T cells were maintained in DMEM with 10% charcoal-stripped FBS during the assay. Transfections were carried out using PEI, and cells were treated with vehicle or 1 nM DHT 24 h after transfection. Cell extracts were prepared 48 h after transfection, when eGFP-AR mutants are mostly localized to the nucleus, and assayed for luciferase activity using the Promega luciferase detection kit. Luciferase activities were normalized to co-transfected β-galactosidase activity81.

Luciferase reporter assays in LNCaP

PSA(6.1 kb)-luciferase, V7BS3-luciferase and AR-V7 plasmids and transfections of cells have been described previously24,52,53,82. PSA(6.1 kb)-luciferase reporter plasmid (0.25 μg well–1) was transiently transfected into LNCaP cells that were seeded in 24-well plates. Twenty-four hours after transfection, cells were pretreated with compounds for 1 h prior to the addition of 1 nM R1881 and incubation for an additional 24 h. For the V7BS3-luciferase reporter, an expression vector encoding AR-V7 (0.05 μg well–1) and a filler plasmid (pGL4.26, 0.45 μg well–1) were transiently co-transfected with V7BS3-luciferase reporter plasmid (0.25 µg well–1) into LNCaP cells in 24-well plates. After 24 h, the cells were treated with the indicated compounds for additional 24 hours. Transfections were completed under serum-free conditions using Fugene HD (Promega). Luciferase activity was measured for 10 s using the Luciferase Assay System (Promega) and normalized to total protein concentration determined by the Bradford assay. Validation of consistent levels of expression of AR-V7 protein was done using western blot analyses.

Proliferation assays

LNCaP cells (Clone FGC, ATCC CRL-1740) in RPMI-1640 with 5% FBS were seeded at a density of 4,000 cells well–1 into 96-well flat-bottom plates (Greiner, 655075) that had been pre-coated with poly-l-lysine (Sigma P4707). Sixteen hours later, triplicate wells were refreshed with 100 µl of seeding medium spiked with 7× serial dilutions of EPI-001 from 200 µM (Selleckchem lot no. S795502), 7× serial dilutions of 1ae from 50 µM, or DMSO control, at a constant DMSO concentration of 0.5% (vol/vol). Ninety-six hours later, wells were washed with 200 µl PBS and then fixed with 100 µl of 4% PFA in PBS for 20 min at room temperature. After fixation, LNCaP nuclei in each well were counterstained using 100 µl of Hoechst 33342 (Abcam ab228551), diluted to 1:4,000 in PBS, for 20 min at room temperature. After a final wash in PBS, plates were imaged using a Celldiscoverer 7 microscope equipped with a ×20 air objective. Twenty-five tile regions (5 × 5 tiles) were imaged for each technical replicate well (triplicate wells for each dose and compound). Data were acquired across two biological replicates performed on different weeks.

To compare the antiproliferative effects of 1ae and enzalutamide in LNCaP and LNCaP95 cells, LNCaP cells (5,000 cells well–1) were plated in 96-well plates in their respective media plus 1.5% dextran-coated charcoal (DCC)-stripped serum. LNCaP cells were pretreated with the compounds for 1 h before they were treated with 0.1 nM R1881 for an additional 3 d. Proliferation and viability were measured using the Alamar blue cell viability assay, following the manufacturer’s protocol (Thermo Fisher Scientific). LNCaP95 cells (6,000 cells well–1) were seeded in 96-well plates in RPMI plus 1.5% DCC for 48 h before the addition of compounds and incubation for an additional 48 h. BrdU incorporation was measured using BrdU Elisa kit (Roche Diagnostics).

Quantitative real-time polymerase chain reaction

Target primer sequences are listed in Supplementary Table 6. LNCaP cells (Clone FGC, ATCC CRL-1740) in RPMI-1640 with 5% FBS were seeded at a density of 300,000 cells well–1 in 6-well plates. Sixteen hours later, wells were refreshed with seeding medium spiked with either EPI-001 or 1ae at doses roughly corresponding to the IC50 and IC10 values calculated from proliferation assays, indicated in Extended Data Fig. 7a, and 0.5% vol/vol DMSO control. After 6 or 24 h, the medium was removed and cells were collected using 300 µl of TRIzol reagent (Thermo Fisher Scientific 15596026) in each well. RNA was then extracted using a Zymo DirectZol Micro kit (Zymo R2062), according to the manufacturer’s protocol. cDNA was synthesized using 1 µg of RNA, random hexamer primers, and the RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific K1622). cDNA collected from LNCaP cells treated with either EPI-001 or 1ae at each dosage, and time point, were then probed for transcript targets on 384-well plates using the SYBR Green master mix (Thermo Fisher Scientific A25777), and a QuantStudio 7 real-time qPCR machine. For calculation of the fold change (2−ΔΔCt method), Ct values from target regions were normalized to Ct values from control regions from the treatment sample, and were then normalized to the DMSO sample. Data were collected from three biological replicates performed on different days.

RNA-seq data generation

LNCaP cells (Clone FGC, ATCC CRL-1740) in RPMI-1640 with 5% FBS were seeded at a density of 300,000 cells well–1 into 6-well plates. Sixteen hours later, wells were refreshed with seeding medium spiked with either EPI-001 or 1ae at the doses indicated in Fig. 6c and 0.5% vol/vol DMSO control. After 6 or 24 h, medium was removed and cells were collected using 300 µl of TRIzol (Thermo Fisher Scientific 15596026) in each well. RNA was then extracted using a Zymo DirectZol Micro kit (Zymo R2062), according to the manufacturer’s protocol. Total RNA-seq libraries were then prepared using 1 µg of RNA from each sample and the KAPA RNA HyperPrep Kit with RiboErase (Roche KR1351), according to the manufacturer’s protocol, with ten amplification cycles. Libraries were sequenced on a NovaSeq 6000 with paired-end reads of 100 bp, with a read depth of 50 million fragments per library. Three libraries from three corresponding biological replicates were prepared for each treatment (time, dosage, and compound).

Western blot

To compare the levels of AR expression, cells were washed and collected in PBS ×1 and lysed in RIPA buffer ×1 (Thermo Fisher Scientific, 88900) containing phosphatase and protease inhibitors (Roche). Lysates were centrifuged at 15,000g to separate soluble and pellet fractions. Total protein was quantified using a BCA assay (Pierce Biotechnology). Proteins were resolved by 4–12% gradient Bis-Tris SDS–PAGE (Invitrogen NP0323), transferred to PVDF membranes and blocked with 5% non-fat milk in TBST for 1 h at room temperature with shaking. The membranes were incubated with the following antibodies: anti-GAPDH (Abcam, ab59164, 1:2,000) and anti-AR (Abcam, ab108341, 1:2,000) as well as RD-680-conjugated anti-mouse (LI-COR, 926-68072, 1:10,000) and CW-800-conjugated anti-rabbit (LI-COR, 926-32211, 1:10,000) secondary antibodies. Membrane fluorescence was read with the Odyssey CLx infrared imaging system (LI-COR).

To determine the effect of 1ae treatment on AR levels, LNCaP cells (ATCC, CRL-1740) were seeded in RPMI-1640 (Thermo Fisher Scientific, 11875093) supplemented with 5% FBS (Thermo Fisher Scientific, 1835030) and 1% penicillin–streptomycin (Thermo Fisher Scientific, 15140122) at a density of 150,000 cells well–1 into six-well plates (Thermo Fisher Scientific, 140685). Forty-eight hours later, cycloheximide (Sigma, C7698) was added to a final concentration of 50 µg ml–1 and incubated for 3 h before incubation with 1ae (at the indicated concentrations) for 21 h. Cells were washed in PBS (Thermo Fisher Scientific, 11835030) and lysed in RIPA buffer (Thermo Fisher Scientific, 88900) containing protease inhibitor (Abcam, ab274282). Lysates were centrifuged at 15,000g to separate soluble and insoluble fractions. Soluble protein was quantified using a BCA assay (Pierce Biotechnology, 23225). Protein extracts (10–70 µg) were electrophoresed in a Bolt 8% Bis-Tris gel (Invitrogen, NW00085BOX) and transferred with PVDF transfer stacks (Invitrogen, PB5210). Membranes were blocked with 3% non-fat milk in TBST for 1 h at room temperature with shaking. Afterwards, membranes were incubated with the following antibodies: anti-GAPDH (Abcam, ab59164, 1:1,000), anti-AR (Abcam, ab108341, 1:1,000), 800-CW-conjugated anti-rabbit (LI-COR, 926-32211, 1:10,000) and RD-680-conjugated anti-mouse (LI-COR, 926-68072, 1:10,000). Imaging was done using the Odyssey CLx, and protein-band intensity was quantified with ImageJ.

Lentiviral production for FLAG-BioID-AR cell lines

FLAG-MTID, FLAG-AR-WT-MTID or FLAG-22YtoS-MTID were subcloned from pcDNA3.1(–) (Genscript) into pLenti-CMV-MCS-GFP-SV-puro (Addgene no. 73582) by replacing GFP using XbaI-BamHI digestion. Vectors were co-transfected with lentiviral packaging plasmid vectors REV (cat. no. 12253), RRE (cat. no. 12251) and VSV-G (cat. no. 8454) into 293T cells with PEI (Sigma-Aldrich). Two days after transfection, virus-containing medium was collected and filtered through a 0.45-µm low-protein-binding filtration cartridge. The virus-containing medium was used to directly infect LNCaP/PC3 cells in the presence of polybrene (8 µg ml–1) for 48 h, before 1 μg ml–1 puromycin was introduced for 72 h to select for stable cell lines. pMDLg/pRRE was a gift from D. Trono (Addgene plasmid no. 12251; http://n2t.net/addgene:12251; RRID: Addgene_12251). pCMV-VSV-G was a gift from B. Weinberg (Addgene plasmid no. 8454; http://n2t.net/addgene:8454; RRID: Addgene_8454). pRSV-Rev was a gift from D. Trono (Addgene plasmid no. 12253; http://n2t.net/addgene:12253; RRID: Addgene_12253).

BioID–MS

Prior to BioID experiments, MTID-containing stable cell lines were generated through lentiviral infection and puromycin selection. They were subsequently grown in RPMI-1640 medium modified with l-glutamine without phenol red or biotin (United States Biological, R9002-01) with 10% (vol/vol) charcoal-stripped FBS for 48 h. Cells were seeded, and the next day, 50 μM biotin (IBA; 2-1016-002) and 1 nM DHT were added for 2 h. For small-molecule inhibitors, cells were pretreated for 1 h with either EPI-001 or 1ae, then for 2 h with DHT and biotin. For MS, cells were collected through trypsinization, washed two times in PBS and snap-frozen on dry ice. Cell pellets were lysed in modified RIPA buffer (1% TX-100, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.1% SDS, 0.5% sodium deoxycholate and protease inhibitors) on ice and treated with 250 U benzonase (Millipore), and biotinylated proteins were isolated using streptavidin-sepharose beads (GE Healthcare). Proteins were washed in ammonium bicarbonate and digested with trypsin. Mass spectrometry was performed in the IRB Barcelona Mass Spectrometry and Proteomics facility, as described previously83. Data were analyzed using SAINTq84.

Proximity ligation assay

Protein–protein interactions were studied using a Duolink In Situ Orange Starter Kit Mouse/Rabbit (Sigma, DUO92102), following the manufacturer’s protocol. Briefly, transduced prostate cancer cells were seeded in coverslips and cultured overnight. The next day, they were treated with 50 μM biotin and 1 nM DHT for 2 h or were pretreated initially with small-molecule inhibitors. Slides were washed with cold 1× PBS and fixed in 4% paraformaldehyde for 15 min, washed in PBS and permeabilized using 0.1% Triton X-100 for 10 min and washed then blocked with blocking buffer (3% BSA, 0.1% Tween in PBS) for 1 h at 37 °C. The coverslips were blocked with Duolink Blocking Solution in a pre-heated humidified chamber for 30 min at 37 °C. Primary antibodies to the following proteins were added and incubated overnight at 4 °C: androgen receptor (ER179(2)) (Abacam, no. ab108341, 1:200), Nup153 (Abacam, QE5, no. ab24700, 1:200), Med1 (Abacam, no. ab64965, 1:200) and ARID1a/BAF250A (Cell Signalling, no. 12354, 1/200). Coverslips were then washed with 1× wash buffer A and subsequently incubated with the two PLA probes (1:5, diluted in antibody diluents) for 1 h, then the ligation-ligase solution for 30 min, and the amplification-polymerase solution for 100 min in a pre-heated humidified chamber at 37 °C. Before imaging, slides were washed with 1× wash buffer B and mounted with a cover slip using Duolink In Situ Mounting Medium with DAPI. Fluorescence images were acquired using a Leica TCS SP8 confocal microscope. Images were taken with ×100 oil objectives, using standard LAS-AF software.

Gene expression analysis

To analyze tumor gene expression, tumors were flash frozen, and ~100-mg samples were pulverized under liquid nitrogen. Samples were added to 1 ml TRIzol (Invitrogen) and homogenized using a FastPrep-24 tissue homogenizer (MP Biomedicals). Total RNA was extracted using the RNeasy Micro Kit (Qiagen), cleaned using the DNase I Kit, amplification grade (MilliporeSigma), and reverse transcribed using the High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific), according to the manufacturers’ protocols. Diluted cDNA and Platinum SYBR Green qPCR SuperMix-UDG with ROX (Invitrogen) were combined with gene-specific primers. Transcript quantification was completed using a QuantStudio 6 RT–qPCR machine, and calculation of the mean normalized expression of target transcripts was done using the 2−ΔΔCt method using the housekeeping gene SDHA. To analyze gene expression in PC3 cells expressing AR V7, 2 × 105 cells were plated in duplicate in 6-well plates. After 48 h, total RNA was extracted using the SV Total RNA Isolation System (Promega), following the manufacturer’s instructions.

Determination of AR levels in LNCaP treated with 1ae and CHX

LNCaP cells (ATCC, CRL-1740) were seeded in RPMI-1640 (Thermo Fisher Scientific, 11875093) supplemented with 5% FBS (Thermo Fisher Scientific, 1835030) and 1% penicillin–streptomycin (Thermo Fisher Scientific, 15140122) at a density of 150,000 cells well–1 into 6-well plates (Thermo Fisher Scientific, 140685). Forty-eight hours later, cycloheximide (Sigma, C7698) was added to a final concentration of 50 µg ml–1 and incubated for 3 h before incubation with 1ae (at the indicated concentrations) for 21 h. Cells were washed in PBS (Thermo Fisher Scienti