S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E BIOCHEMISTRY Copyright © 2023 The Authors, some A hotspot for posttranslational modifications on the rights reserved; exclusive licensee androgen receptor dimer interface drives pathology American Association for the Advancement and anti-androgen resistance of Science. No claim to original U.S. Government Works. Distributed Andrea Alegre-Martí1,2†, Alba Jiménez-Panizo1,2†, Adrián Martínez-Tébar3†, Coralie Poulard4, under a Creative M. Núria Peralta-Moreno5, Montserrat Abella1,2, Rosa Antón6, Marcos Chiñas3,7, Ulrich Eckhard8, Commons Attribution Josep M. Piulats3, Ana M. Rojas9, Juan Fernández-Recio10, Jaime Rubio-Martínez5, NonCommercial Muriel Le Romancer4, Álvaro Aytes3‡*, Pablo Fuentes-Prior6‡*, Eva Estébanez-Perpiñá1,2‡* License 4.0 (CC BY-NC). Mutations of the androgen receptor (AR) associated with prostate cancer and androgen insensitivity syndrome may profoundly influence its structure, protein interaction network, and binding to chromatin, resulting in altered transcription signatures and drug responses. Current structural information fails to explain the effect of pathological mutations on AR structure–function relationship. Here, we have thoroughly studied the effects of selected mutations that span the complete dimer interface of AR ligand–binding domain (AR-LBD) using x-ray crystallography in combination with in vitro, in silico, and cell-based assays. We show that these variants alter AR-dependent transcription and responses to anti-androgens by inducing a previously unde- scribed allosteric switch in the AR-LBD that increases exposure of a major methylation target, Arg761. We also corroborate the relevance of residues Arg761 and Tyr764 for AR dimerization and function. Together, our results reveal allosteric coupling of AR dimerization and posttranslational modifications as a disease mechanism with implications for precision medicine. INTRODUCTION Another EM structure of the AR core has recently been published The androgen receptor (AR; AR/NR3C4) is a member of the steroid in three distinctive configurations (7), further corroborating the subfamily of nuclear receptors essential for male development and noncanonical AR-LBD dimerization mode (3). These EM struc- tissue homeostasis (1–2). The AR protein is composed of an intrin- tures evidence that AR is a highly dynamic protein. AR physiolog- sically disordered N-terminal region (NTD) followed by a “core” ical functions are inherently linked to not only the correct folding of that contains the actual DNA- and ligand-binding domains (DBD DBD and LBD modules but also to their relative orientations (e.g., and LBD, respectively) (Fig. 1A) (2–3). The atomic details of full- N-terminal and C-terminal (N/C) interaction), oligomerization, length AR (FL AR) still remain elusive, but a low-resolution struc- and coregulator binding to control transcription (2, 8). Besides, ture by cryo–electron microscopy (cryo-EM) complexed to DNA AR protein structure and function relationship are regulated by and a coregulator has provided a model of its overall architecture posttranslational modifications (PTMs) such as methylation, phos- (4). This EM structure features a noncanonical head-to-head phorylation, ubiquitination, acetylation, and sumoylation, adding dimer (i.e., centered on H5 helices, in contrast to the canonical, an extra level of complexity that has only been scarcely studied H10-centered arrangement [see (3) for a detailed comparison of (9–19). In particular, no structural data have been reported on the the dimerization modes], in line with previously reported x-ray mechanistic consequences of these covalent modifications to date. structures of dimeric DBD (5) and LBD isolated moieties (6). Tight spatial and temporal regulation of AR activity is critical for the proper functioning of tissue-specific signaling cascades. There- 1 Structural Biology of Nuclear Receptors, Department of Biochemistry and Molec- fore, dysfunctions of the AR-regulated pathways have profound ular Biomedicine, Faculty of Biology, University of Barcelona (UB), 08028 Barcelona, pathophysiological consequences (2, 8, 20–22). Up-regulated recep- Spain. 2Institute of Biomedicine of the University of Barcelona (IBUB), University of tor activity is the dominant driver in one of the leading cancers in Barcelona (UB), 08028 Barcelona, Spain. 3Programs of Molecular Mechanisms and adult men, prostate cancer (PCa; 1–2, 8, 21–22), making AR a Experimental Therapeutics in Oncology (ONCOBell) and Cancer Therapeutics Re- sistance (ProCURE), Catalan Institute of Oncology, Bellvitge Institute for Biomedical central therapeutic target for PCa treatment (2). On the other Research, 08908 Barcelona, Spain. 4Cancer Research Center of Lyon, CNRS hand, impaired response to androgens leads to an X chromo- UMR5286, Inserm U1502, University of Lyon, 69000 Lyon, France. 5Department some–linked disorder of sex development in genetic males termed of Materials Science and Physical Chemistry, Faculty of Chemistry and Institut de Recerca en Química Teorica i Computacional (IQTCUB), 08028 Barcelona, Spain. androgen insensitivity syndrome (AIS) (23–25). Many point muta- 6 Biomedical Research Institute Sant Pau (IIB Sant Pau), 08041 Barcelona, Spain. tions in the AR gene associated with either resistance to PCa ther- 7 Universidad Nacional Autónoma de México, Centro de Ciencias Genómicas, Cuer- apeutic drugs (anti-androgens; 26–28) and/or with AIS have been navaca, 61740 Morelos, Mexico. 8Department of Structural and Molecular Biology, Molecular Biology Institute of Barcelona (IBMB-CSIC), 08028 Barcelona, Spain. reported (29) (table S1 and Fig. 1A). Knowledge of the structure 9 Computational Biology and Bioinformatics, Andalusian Center for Developmental of the wild-type (WT) protein does not allow to predict the struc- Biology (CABD-CSIC), 41013 Sevilla, Spain. 10Instituto de Ciencias de la Vid y del tural and functional impact of point mutations and therefore their Vino (ICVV-CSIC), CSIC-UR-Gobierno de La Rioja, 26007 Logroño, Spain. pathogenicity and clinical implications in patients (3, 6–7). Whereas †These authors contributed equally to this work. ‡These authors contributed equally to this work. mutations causing AIS occur in the germ line, PCa-linked muta- *Corresponding author. Email: evaestebanez@ub.edu (E.E.-P.); fuentespriorpablo@ tions are mostly somatic and occur before endocrine treatment or gmail.com (P.F.-P.); aaytes@idibell.cat (A.A.) Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 1 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 1. Mutations at the AR-LBD dimer interface affect cellular phenotypes, transcriptional activ- ity, and response to antiandro- gens. (A) Cartoon of AR domain organization. Physiologically rele- vant interaction sites [top; activation function-1 (AF-1), AF-2 and binding function-3 (BF-3)], mutated AR-LBD residues (red; middle), and second- ary structure elements [bottom; helices (H, cylinders) and β strands (S1 to S4, triangles)] are indicated. Neighboring residues of major PTM sites are shown [methylation (M), phosphorylation (P), and ubiquiti- nation (U)]. (B) Three-dimensional structure of AR-LBD (gray). Note that disease-linked residues (red spheres) and PTM sites (gray spheres) high- lighted in (A) form an extended path on the AR-LBD dimer interface. (C) Relative AR transcriptional activity in CNT, WT, and mutant AR-transduced PC3 cells (mean ± SD, n = 3). Differ- ences against WT were calculated using a t test and considered sig- nificant at P values <0.1. (#P < 0.1 *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). The same P value guidelines and asterisk sig- nificance will be used through this manuscript. (D) Clonogenicity assay quantification in transduced PC3 cells (mean ± SD, n = 3). Represen- tative crystal violet–stained cultures are shown in fig. S1A. (E) Time course of transduced PC3 cells proliferation (mean ± SD, n = 3). Proliferation is completely suppressed with WT AR or V785A but no other mutants. (F) Relative cell counts at day 7 after transduction. (G and H) Effect of anti-androgens on the relative AR transcriptional activity of transduced PC3 cells (mean ± SD, n = 5). (I and J) Viability fold change in response to anti-androgens in transduced cells relative to the CNT cells (mean ± SD, n = 3). Differences against CNT (D and F) between anti-androgen treated and nontreated cells (G and H) or against WT (I and J) were cal- culated using t tests. ns, not significant. are induced by the selective pressure of anti-androgen therapies (2– (table S1). These oncogenic mutations (W742L, H875Y, and 3). While few mutations in PCa have been described as oncogenic or T878A) map to the internal ligand-binding pocket (LBP) and are drivers (causative of the disease), most are considered passenger or responsible for rendering the anti-androgens cyproterone acetate, silent mutations, although a contribution to PCa development and/ flutamide, and bicalutamide ineffective, respectively. Comparison or progression cannot be excluded (29–32). of the WT AR-LBD and these mutant structures did not reveal To date, only three point mutants of the AR-LBD have been any overall conformational changes. However, these mutations structurally characterized out of the over 350 reported variants change the volume and shape of the LBP allowing the receptor to Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 2 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E remain active even in the presence of antagonists, thus explaining (36, 39, 42–48), or Gln799Glu (Q799E) (39, 42, 44–45, 48–54) tumor escape from therapeutic control (2, 26–28). On the other (Fig. 1, A and B). As expected, WT expression resulted in a 24.5- hand, there are a wide variety of phenotypes associated with AR fold increase in AR transactivation as measured by luciferase report- gene mutations in patient with AIS, ranging from mild symptoms er expression under the control of a synthetic AR promoter. All four (MAIS), to partial (PAIS) or completely feminized phenotypes mutants displayed a significantly lower AR transactivation activity (CAIS) (23–25, 33–34). Expectedly, about half of the CAIS-linked compared to WT (Fig. 1C). Further, expression of WT and V758A mutations map to the AR-LBD, but only a few of them can be resulted in a nearly complete inhibition of colony formation easily rationalized as directly affecting hormone binding (3, 29). (Fig. 1D and fig. S1A) and proliferation (Fig. 1, E and F) compared AR-LBD pathogenic mutations outside the LBP may affect AR- to CNT cells that were highly statistically significant. Dihydrotestos- dependent gene signatures through modulation of the monomer- terone (DHT) treatment of AR-transduced DU145 (fig. S2, B to D) oligomer equilibrium, accessibility of PTM sites, and/or by altering and of AR-sensitive lymph node carcinoma of the prostate (LNCaP) dynamic interactions with coregulators and chromatin. These pro- cells (fig. S2, E to G) resulted in comparable growth-suppressive cesses are likely to be intricately intertwined and coupled via allo- effects that were not observed in AR-negative parental DU145 steric signaling transmission both within the AR monomer and cells (fig. S2, B to D), in a castration-resistant PCa cell line, 22rv1 across its dimerization interface (3–7). Here, we present a multidis- (fig. S2, H to J), or in AR-insensitive, nontumor human prostate ep- ciplinary investigation of the impact of several mutations of solvent- ithelial cells,RWPE1 (fig. S2, K to M). These observations confirm exposed AR-LBD residues identified in patients with PCa and/or previous reports on the growth-suppressive effect of AR stimulation AIS (Fig. 1A; see table S2 for major information on the studied mu- in AR-responsive cells (55, 56). In notable contrast, F755V and tations). Transcriptomic and in vitro cell–based assays demonstrate Q799E had proliferation rates not significantly different from that these variants affect PCa proliferation and response to anti-an- those of the CNT cells, while proliferation was only modestly im- drogens. Biochemical and biophysical investigations including paired by Y764C (Fig. 1, E and F). Similarly, colony forming capac- high-resolution x-ray structures of mutant AR-LBD showed ity was unchanged for F755V and partially, albeit significantly, reduced thermal stability of the isolated domains as well as reduced in Glu799- and Cys764-expressing cells (Fig. 1D and changes in the flexibility in functional sites of the receptor distal fig. S1A). from the mutation site and unexpectedly revealed a previously un- Next, we asked whether the differences in proliferation and clo- observed rearrangement of the loop-featuring residue Arg761. The nogenicity displayed by the AR variants would translate into dis- structural reorganization of this loop increases the degree of tinct responses to anti-androgenic treatments. WT or mutant AR- Arg761 exposure, which correlated with an apparent increase in expressing cells were treated with enzalutamide (Fig. 1, G and I), the extent of its methylation in cells. Together, our findings point abiraterone (Fig. 1, H and J), or apalutamide (fig. S1B), and AR to an intricate interplay between networks of local structural transactivation was measured. Transactivation was significantly in- changes at and around the dimerization interface and global AR ter- hibited by enzalutamide and abiraterone in both WT and mutant tiary and quaternary structures that are coupled to PTM events. In AR cells (Fig. 1, G and H). Reduction in AR transactivation was turn, this network has a definite impact on AR transcriptional ac- not a consequence of impaired cell viability upon anti-androgen tivity and gene programs that are affected by specific mutations. Our treatment (Fig. 1, I and J, and fig. S1B). Specifically, V758A cells results may serve to predict the structural and functional impact of showed an 11.7, 20.6, and 16.1% increased viability compared to AR pathogenic mutations and open unexplored avenues for struc- WT when treated with enzalutamide, apalutamide, or abiraterone, ture-guided modulation of its function in PCa tumors and patients respectively. On the contrary, F755V, Y764C, and Q799E mutant with AIS. cells displayed a significantly increased sensitivity to all three anti-androgens (Fig. 1, I and J, and fig. S1B). Together, these results indicate that expression of ligand-bound RESULTS WT AR in AR-null PCa cells induces a quiescent cell phenotype in We previously presented the crystal structure of the transcription- AR-responsive cells without affecting the response to anti-androgen ally active, isolated dimeric AR-LBD (6) also corroborated in the treatment. The data also demonstrate that, overall, point mutations multidomain protein (4, 7). The structural analysis revealed that at the AR dimerization surface negatively impact its ability to sup- the dimerization interface is a hotspot for mutations identified in press proliferation and clonogenicity and mostly sensitize cells to patients with PCa and/or AIS [see (3) and table S1]. There was no anti-androgens (Fig. 1 and figs. S1, A and B, and S2). clear-cut separation between PCa- and AIS-linked mutations, and some of these variants were linked to both conditions, albeit occur- Mutations at the AR-LBD dimer interface profoundly ring in different developmental stages. These observations prompt- deregulate androgen-dependent transcription programs ed us to perform a thorough characterization of the structural Having demonstrated that dimerization surface mutations in the changes and functional impact of AR point mutants at the dimeri- AR-LBD have an important impact on cellular functions, we next zation interface (Fig. 1B). investigated whether these mutations alter gene expression patterns (Fig. 2 and fig. S1). Transcriptomic data were generated for WT and Mutations at the AR-LBD dimer interface differently affect each mutant AR and compared to nontransduced cells as control. cell proliferation and response to anti-androgens As expected, AR and canonical AR target genes such as Transmem- To assess the functional impact of point mutations that affect the brane protease, serine 2 (TMPRSS2), FK506-binding protein 5 AR-LBD dimer interface, we generated AR-null PC3 cell lines (FKBP5), and TCDD-inducible poly [ADP-ribose] polymerase (CNT) stably expressing either WT AR or mutants Phe755Val (TIPARP) were significantly up-regulated in both WT and mutant (F755V) (35–37), Val758Ala (V758A) (38–42), Tyr764Cys (Y764C) AR-expressing cells (Fig. 2A and fig. S1C), which was orthogonally Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 3 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 2. The transcriptional profiles of AR-LBD dimerization surface mutants differ from that of the WT receptor. The results of RNA-seq experiments conducted in triplicate for each cell line are summarized. (A) Differential ex- pression analysis between CNT and AR-transduced cells. Genes with |log2(fold change)| ≥ 1 and P ≤ 0.05 are shown in red. (B) GSEA demonstrates similar signatures of WT and mutant AR. (C) Differen- tially expressed genes overlap between WT AR and its mutants. The bar plot (top) indicates the number of genes in the intersec- tion between signatures. Vertical lines (bottom) connect the corre- sponding overlapping signatures. For each WT and mutant AR, sig- natures are separately given for up- (U) and down-regulated genes (D), and the size of the gene set is indicated (right). (D and E) Princi- pal component analysis (PCA) and heatmap analysis of transcription profiles identifies three groups: (i) CNT, (ii) V785A and WT, and (iii) F755V, Y764C, and Q799E. (F) Volcano plots of AR target genes for WT and mutant AR. Canonical repressed (blue) and activated (red) AR targets are labeled. Note that repression of genes such as MYC, SOX4, FOXA1, and CDH2 is lost in Q799E, Y764C, and F755V but not in V785A. (G) GSEA of the mutant AR signatures against WT AR on prostate cancer gene sets. Note that the V758A signature shows very limited overall enrich- ment indicating its strong similar- ity with WT. (H) GSEA of WT and mutant AR against a previously defined PCa malignancy signature (58). Enrichment for up- and down-regulated targets of the malignancy signature drivers fork- head box protein M1 (FOXM1), Centromere Protein F (CENPF), or both are shown separately. Only the F755V, Y764C, and Q799E sig- natures are enriched in the FOXM1 and/or CENPF regulons. Normal- ized enrichment score (NES) scale and false discovery rate (FDR) P value thresholds are indicated. validated in quantitative polymerase chain reaction (qPCR) assays the AR regulon was conserved in all mutants, there were significant (fig. S1, G and H). In agreement with the functional evidence de- differences between the four variants. V758A showed a dynamic scribed above, the V758A mutant displayed AR target gene expres- gene expression profile most similar to WT, while the differentially sion patterns most similar to WT AR (Fig. 2, C to E). Accordingly, expressed gene (DEG) sets of F755V, Y764C, and Q799E were much gene set enrichment analysis (GSEA) showed that all four mutant smaller and shared fewer target genes compared to WT AR or AR signatures were similar to the WT signature (Fig. 2B). Although V758A (Fig. 2C and fig. S1C). Variance assessment and clustering Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 4 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E of expression profiles further confirmed the similarity of WT and only 1°C lower than the WT AR-LBD protein, demonstrating V758A transcriptomes, on the one hand, and those of F755V, WT-like stability (fig. S3, A and B). Energy estimations using Y764C, and Q799E, on the other. These two clusters were identified FoldX based on previously reported structures of monomeric and both by a principal components analysis (PCA) (Fig. 2D) and by dimeric AR-LBD also suggested that the studied missense muta- clustering analysis computing sample-to-sample distances using tions have an important impact on the domain stability (fig. S3C), variance-stabilizing transformation (VST) data (Fig. 2E). Further, both in its monomeric and putative dimeric conformations. Muta- gene expression in the AR regulon as defined by the DoRothEA col- tions of the core interface, F755V/L and Y764C, are predicted to be lection of human regulons (57) showed that the set of activated AR particularly deleterious for homodimeric AR-LBD (fig. S3C). target genes in WT and mutants remains largely conserved AR-LBD is a highly basic protein domain (isoelectric point: 8.94) (Fig. 2F). However, only WT and V758A maintained a repressed with 14 arginines (fig. S4A) and 14 lysines outbalancing 11 negative- gene set that included cancer drivers Forkhead Box A1 (FOXA1) ly charged aspartate or glutamate residues. To assess whether the and MYC proto-oncogen (MYC) as well as epithelial-to-mesenchy- decrease in thermal stability of AR-LBD point mutants also mani- mal transition drivers Snail Family Transcriptional Repressor 2 fests as an increased sensitivity to proteolysis, we incubated highly (SNAI2/SLUG), Cadherin-2 (CDH2, encoding N-cadherin), and purified samples of WT AR-LBD and its point mutants with the en- SRY-Box Transcription Factor 4 (SOX4) (Fig. 2F and fig. S1I). doproteinase, Arg-C, which cleaves peptide bonds at the C-termi- This is consistent with the important reduction in proliferation nus of Arg residues (fig. S4, B to E). and clonogenicity described above (Fig. 1, D to F, and fig. S1A). Again, the resulting cleavage pattern of V758A is most similar to GSEA analysis of the four AR mutants versus WT signatures evi- that of WT AR-LBD (fig. S4, B, C, E, and F). Although the thermal denced again that V758A was not enriched in any of the PCa signa- stabilities of the two mutants are similar (fig. S3, A and B), the pro- tures in the Molecular Signature Database (MSigDB), as opposed to teolytic patterns of V758A and Q799E differ notably (fig. S4, C to F the significant enrichment demonstrated for F755V, Y764C, and and H). Mutants F755L (fig. S4, B, E, and G), F755V (fig. S4, C, E, Q799E (Fig. 2G). Accordingly, only these three mutant signatures and I), and Y764C (fig. S4, D, E, and H) have enhanced proteolytic were found positively enriched in a previously defined up-regulated susceptibility. In summary, point mutations of residues at the AR- gene signature strongly associated with poorer outcome in patients LBD homodimerization interface appear to have an unanticipated with PCa (Fig. 2H) (58). global impact on the overall domain stability, as evidenced by Last, querying the collection of manually drawn Kyoto Encyclo- notably lower Tm and enhanced sensitivity to proteolytic attack pedia of Genes and Genomes (KEGG) database evidenced an unex- (figs. S3 and S4). pected enrichment for the Y764C mutant in biological pathways associated with pathogen infection and immune response including AR-LBD dimer interface mutants exhibit enhanced coronavirus disease 2019 (fig. S1D), which was supported by the flexibility and distinct structural rearrangements both significant regulon enrichment in the mediator of type I interferon locally and at distant AF-2 and BF-3 sites signaling, STAT2 (fig. S1E), and the activation of the Janus kinase/ The results of the experiments described above demonstrate a large signal transducers and activators of transcription (JAK/STAT) impact of AR-LBD dimer interface mutants on AR transcriptional pathway (fig. S1F) using the DoRothEA and the Pathway RespOn- function, stability, and sensitivity to proteolytic attack. To assess sive GENes (PROGENy) algorithms, respectively. Together, these whether these differences are reflected by noticeable structural rear- data suggest that the observed differences in cellular phenotype rangements, we expressed crystallized and refined at high resolution between the different AR mutants are at least in part the result of the x-ray crystal structures of the V758A, Y764C, F755V, F755L, and a differentially reprogramed transcriptome in which WT and Q799E AR-LBD variants (Figs. 3 and 4). All mutant proteins crys- V758A show a less oncogenic and undifferentiated molecular phe- tallized as monomers in the orthorhombic space group (P212121) notype compared to F755V, Y764C, and Q799E mutants. (Figs. 3A and 4A and fig. S3D; see fig. S3E and Fig. 4B for compar- ison; Table 1). The “helical sandwich” fold typical of the nuclear re- AR-LBD interface residue mutations do not compromise ceptor superfamily is conserved (3), and the hormone (DHT) is domain folding but lead to reduced thermal stability and bound inside the LBP essentially as in WT AR-LBD (Figs. 3A and increased sensitivity to proteolysis 4A). The root mean square deviations (RMSDs) for the equivalent To study the possible structural impact of point mutants located at Cα atoms between the WT structure [Protein Data Bank (PDB) the AR-LBD dimerization interface, we produced and purified to 1T7T] and the V758A, Y764C, F755V, Q799E, and F755L homogeneity WT AR-LBD and its point mutants F755L, F755V, mutants are 0.29, 0.33, 0.47, 0.52, and 0.54 Å, respectively. V758A, Y764C, and Q799E. Next, we characterized these recombi- Despite this overall conservation of the global protein fold, nant proteins both biophysically using differential scanning fluo- superimposition of the current mutant structures on previously rimetry (DSF) and biochemically through limited proteolysis solved structures of WT AR-LBD reveals important local structural assays (figs. S3 and S4). The results of these experiments demon- changes on four specific distant sites of the domain (see Fig. 3A for strated that all mutants are properly folded but have substantially an overall representation of the domain with these sites highlighted lower thermal stability compared to the WT protein, as indicated and Fig. 3, B to E for close-ups of these areas): the H5-S1-H7 area by the reduced melting temperatures (Tm) (fig. S3, A and B). In (from now on termed “R761 zone,” described in detail below) line with the RNA sequencing (RNA-seq) results presented above, (Fig. 3B), the L1-3 loop, H6, and the C-t end of H11 (Fig. 3C), core dimer interface mutants F755V, F755L, and Y764C exhibited a two major functional sites, the BF-3 pocket comprising H9 and remarkable drop in Tm compared to the WT protein (between 5° L9-10 (Fig. 3D), and the AF-2 pocket featuring the L3-4 loop and and 6°C), the latter being the most thermolabile of all the studied H12 (chaperone and coactivator binding sites, respectively) variants. By contrast, the Tm of mutants V758A and Q799E was (Fig. 3E). In particular, several side chains of charged AF-2 residues Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 5 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 3. High-resolution crystal structures of AR-LBD dimer interface mutants reveal local and long-range conformational changes. (A) Superimposition of current crystal structures of mutant AR-LBD on WT monomeric (1T7T) and dimeric (5JJM) forms of the domain. Secondary structure elements identical in all structures are de- picted as a gray cartoon, with DHT as teal spheres and the AF-2-bound peptide (from 5JJM) in red. Large main- and/or side chain conformational changes cluster in four areas (B to E), highlighted by the major secondary structure elements (coral sticks). (B) Dimer interface core lined by H5 and H7, which, in addition to the studied point mutations [Phe755 and Val758 (H5), Tyr764 (S1), and Gln799 (H7)] features mostly residues with nonpolar/aromatic side chains, along with the positively charged Arg761. (C) The more distal part of the dimer interface formed by H3 and L1-3 [Tyr774 and His777 from (B) are shown for orientation]. In this area, several polar residues exhibit noticeable conformational changes. (D) BF-3 pocket, where multiple residues exhibit conformational changes, most notably those from H9. (E) AF-2 pocket, where in addition to the charge-clamp residues, Lys721 and Glu894, known to stabilize bound coregulators at this interaction site, the side chains of both charged (Lys718, Arg727, Lys823, and Glu898) and aliphatic residues (Met735 and Met895) display conformational variability. (F) Sector 1 (teal) comprises 17 residues in and around H5, H7, H8, H9, and H10-11 (Met743, Phe748, Gly751, Arg753, Leu798, Ile800, Thr801, Met808, Leu811, Phe814, Glu838, Ile842, Thr851, Tyr858, Thre861, Lys862, and Leu864), whereas sector 2 (green) features 20 residues mostly from H3, H5, H7, H8, S3, and H10-11 (Arg711, Leu713, Trp719, Ala720, Lys721, Phe726, Leu729, Gln734, Tyr739, Trp742, Gly744, Met746, Ala749, Trp752, Ser754, Leu791, Lys809, Leu813, Asp820, and Arg856). Last, three residues (Gly725 and Ile738 at the AF-2 groove and Phe805 at H7) belong to both sectors (yellow). Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 6 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 4. AR-LBD dimer interface mutants display local flexibility differences. (A to G) Visualization of protein mobility for WT and AR-LBD mutant proteins. Proteins are represented as ribbons sized by temperature factors (B-factors) and colored according to the average B-factors of each residue (i.e., areas with larger atomic displacements appear thicker and are depicted with warmer colors (red, orange, and yellow), whereas those with lower thermal motion are shown thinner and with cold colors (green, blue, and violet). DHT molecules bound inside the LBP pockets are shown as gray spheres with oxygen atoms in red. WT AR-LBD·DHT monomer (A, PDB 1T7T) and dimer (B, PDB 5JJM) are shown for comparison. Note that all mutants exhibit a larger mobility than WT AR-LBD. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 7 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Table 1. X-ray crystallography data collection, refinement statistics, and model quality. V758A Y764C F755V F755L Q799E PDB code 7ZU1 7ZTZ 7ZTX 7ZTV 7ZU2 Wavelength (Å) 0.979 0.979 0.979 0.979 0.979 66.17– 65.84– 65.82– 56.2– 44.55– Resolution range 1.68 (1.71–1.68) 1.40 (1.42–1.40) 1.89 (1.93–1.89) 1.94 (1.99–1.94) 1.74 (1.77–1.74) Space group P212121 P212121 P212121 P212121 P212121 Cell constants a, b, c (Å) / α, β, 55.62, 66.17, 70.95 / 56.39, 65.84, 72.82 / 54.57, 65.82, 70.37 / 56.18, 65.80, 72.31 / 56.51, 66.04, 72.42 / γ (°) 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 Molecules in the ASU 1 1 1 1 1 Total reflections 176567 (4168) 247697 (9727) 143549 (9454) 84880 (3942) 147708 (8184) Unique reflections 29533 (1229) 53810 (2576) 20910 (1309) 18026 (1027) 28510 (1559) Multiplicity 6.0 (3.4) 4.6 (3.8) 6.9 (7.2) 4.7 (3.8) 5.2 (5.2) Completeness (%) 96.6 (80.4) 99.6 (98.2) 99.9 (100.0) 88.1 (79.8) 100.0 (100.0) Mean I/sigma(I) 10.3 (0.7) 18.8 (0.5) 8.8 (0.7) 10.4 (2.8) 12.3 (1.5) Wilson B-factor 27.1 27.0 38.2 22.8 26.1 R-meas 0.084 (2.026) 0.030 (1.692) 0.116 (3.399) 0.093 (0.720) 0.089 (2.398) R-pim 0.033 (1.015) 0.014 (0.841) 0.045 (1.257) 0.042 (0.363) 0.039 (1.042) CC1/2 0.998 (0.364) 1.000 (0.395) 0.998 (0.364) 0.989 (0.854) 0.999 (0.337) Reflections used in refinement 28174 51127 19842 17086 27029 Reflections used for R-free 1235 (4.2) 2619 (4.9) 1020 (4.9) 888 (4.9) 1433 (5.0) R-work 0.21 0.19 0.20 0.22 0.18 R-free 0.25 0.22 0.25 0.26 0.22 Total number of atoms / Protein 2135 / 2008 / 21 / 81 2161 / 2010 / 21 / 120 2041 / 1973 / 21 / 32 2066 / 1997 / 21 / 44 2080 / 1985 / 21 / 64 / DHT ligand / Water RMS (bonds) 0.007 0.011 0.008 0.007 0.009 RMS (angles) 1.420 1.828 1.606 1.500 1.581 Ramachandran favored (%) 98.76 98.74 98.32 98.33 99.16 Ramachandran allowed (%) 1.24 1.26 1.26 1.25 0.84 Clashscore 5.32 1.23 3.47 2.47 3.7 Average B-factor / Protein 40.4 / 40.0 (36.9, 42.8) 35.6 / 35.1 (32.0, 38.1) 48.2 / 48.1 (44.2, 51.9) 37.5 / 37.6 (34.3, 40.8) 37.3 / 37.2 (33.3, 40.9) (main, side chains) / DHT ligand / 29.1 / 44.4 / 25.4 / 42.8 / 34.3 / 47.3 / 25.3 / 36.3 / 26.0 / 41.9 / Water exhibit conformational differences, most notably the so-called Residues with particularly higher B-factors in the mutants are charge-clamp residues, Lys721 and Glu894, which are known to sta- Asn693 and Gln694 (L1-3), Arg727 (L4-5), Asn759, and Arg761, a bilize bound coregulators at this protein-protein interaction site residue strictly conserved in AR from fish to humans, but not (Fig. 3E). Regarding the BF-3 pocket, and in addition to charged/ found in other steroidal receptors (H5 and loop H5-S1; fig. S3, F polar residues such as Arg841 and Thr851, also the aromatic Phe827 and G), Asp820 and Lys823 (L9-10), His886 (H6), and Glu894 (H12) exhibits a large flexibility (Fig. 3D). Noteworthy, Phe827 and Arg841 (Fig. 4). Overall, mutants V758A and Y764C are most similar to the undergo major conformational changes upon small-molecule WT protein in this regard, while Q799E, F755L, and F755V exhibit binding to this surface-exposed pocket (63). Loop L1-3 also exhibits increased flexibility in H5, L1-3, and L9-10 (Fig. 4, A to G). side-chain conformational differences in several residues (Tyr774, Our previous statistical coupling analysis (SCA) (64) revealed His777, and Lys778), which we had previously identified as part of that two groups of correlated residues or “sectors” define the inter- a secondary dimer interface in AR-LBD crystals to generate a tetra- nal circuits linking major interaction sites in the LBD of the related meric arrangement of the receptor (Fig. 3C) (6). glucocorticoid receptor (GR). To analyze whether the AR-LBD di- These structural variations are also reflected by an overall en- merization surface is allosterically coupled to other functional hanced flexibility of all mutants compared to WT AR-LBD regions, we also performed SCA analysis, which identified 40 resi- [average temperature factor (B-factor) of 21.2 for the monomeric dues grouped in two sectors in the upper half of the domain (Fig. 3, AR-LBD; PDB 1T7T], as demonstrated by the higher B-factors of F and G). Sector 1 residues cluster around the domain N-terminus both main and side chain atoms in all variable regions (Fig. 4). (except the DHT ligand, Arg752), while sector 2 comprises LBP and Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 8 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E AF-2 residues. Further, three residues common to both sectors are Phe765-Leu769 stretch (boxes D1 and D2 in Fig. 5C, respectively). strategically located to cross-connect the LBP, AF-2, and BF-3 and Mutant V758A presents fewer changes in the average distances in dimerization interface (Fig. 3, F and G). Together, our results the same regions, some of which are positive in the first one (box suggest that both sectors link functionally relevant regions thus cou- D1 in Fig. 5C). For the Y764C mutant, the first group of distances pling, e.g., ligand binding to dimerization, coregulator binding, or becomes completely positive, while the second one practically dis- chaperone docking/release. appears. No noticeable changes were detected for mutants F755V To further explore the LBD dynamics and possible allosteric and Q799E. In summary, residue pair distance analysis shows that couplings between AR-LBD functional areas, we performed four the different interface mutations induce local conformational 1-μs-long molecular dynamics (MD) simulations for both WT changes in AR-LBD. These changes are located far from the AR-LBD and all the mutants (Fig. 5). The analysis of the time evo- mutated residues in line with an allosteric mechanism. Moreover, lution of the RMSD for all the Cα atoms using the experimental although some changes are common to all the mutations, markedly structure of WT AR-LBD as reference confirms that the overall different local responses were observed in specific AR regions. fold of the domain is stable against these point mutations, and all structures show global structural stability. Moreover, global confor- The noncanonical AR-LBD dimerization mode is conserved mations do not vary remarkably for any of the mutants, and their upon single point mutations of dimer interface residues final RMSDs oscillate at around 2 Å (Fig. 5A). To evaluate the flex- Because the studied AR mutations are located at the noncanonical ibility of the mutants in more detail, we analyzed how a given dimerization interface, we next wondered whether they might residue fluctuates around a specific position with time. To this impair homodimer formation (Fig. 6). We had previously shown end, we quantified the fluctuations for all residues using the aver- that AR-LBD dimerization in solution could be demonstrated aged root mean square fluctuation (RMSF) value, averaged for the with the homo-bifunctional cross-linker, bis(maleimido)ethane four MD replicas (Fig. 5B). As expected, more flexible regions cor- (BMOE), which cross-links residue Cys687 from two monomers respond to those with higher B-factors in the current structures (6). To verify whether mutant proteins are able to dimerize as (compare Figs. 4 and 5). To analyze how specific mutations affect WT AR-LBD, we incubated purified samples of the recombinant the dynamics of the different regions, we calculate the RMSF differ- proteins with either BMOE (spacer arm length: 8.0 Å) or a similar ences between WT AR and the studied variants (Fig. 5B). The cross-linker with a slightly larger spacer arm, 1,4-bis(maleimido)- stretch Lys884-Val888 (H10-H11) presents, for all the mutants, a butane (BMB; 10.9 Å) and separated the reaction products by markedly decreased flexibility compared to WT (Figs. 5B and SDS–polyacrylamide gel electrophoresis (SDS-PAGE). The results 3C). On the other hand, the L1-3 loop (residues Cys687-Asn693) is of these experiments revealed similar patterns of AR-LBD homo- notably more flexible in the V758A and F755L mutants. This behav- dimer formation in all cases, indicating that receptor homodimeri- ior is also observed in the Q799E mutant, although less pro- zation is not compromised by point mutations of interface nounced, but neither in Y764C nor in F755V mutants. These two residues (Fig. 6A). regions are located far from the mutated residues, suggesting that Further, and to assess the possible impact of the studied mutants mutations in the AR-LBD dimerization site induce conformational on the dimerization potential of AR-LBD mutants, we performed changes in the AF-2 pocket and thus that dimerization and coregu- docking calculations using normalized interface propensity (NIP) lator binding are allosterically coupled. (65) and optimal docking area (ODA) (66) (fig. S5). The computed Next, and to complement the dynamic description of the impact binding energy of dimeric WT AR-LBD (PDB 5JJM; Fig. 4B; chains of the different mutations, we quantified local conformational B and C, removing residues 845 to 851; Fig. 4B) was similar to that changes by calculating the differences in the average Cα-Cα distanc- of V758A and Y764C mutants (assuming the same orientation as es between spatially neighboring residue pairs [extended up to the WT) and more favorable than that of the other mutants. Actually, second coordination shell (SCS); ~9.1 Å]. The results of these ΔRij the electrostatic energy contribution was much larger for WT than analyses are presented in Fig. 5C. Overall, all mutant structures tend for any mutant, while the desolvation energy showed the opposite to have larger inter-residue distances compared to the WT protein, trend (fig. S5A). The size of the dimer interface as estimated from and some specific groups of residues show remarkable changes in the buried surface area (BSA) was larger for WT than for the their average pair distances. This is particularly true for residues mutants, being V758A the one closest to WT (fig. S5A). Lys884-His886 and Val888-Phe892 at the C-terminal end of H10- We also built alternative dimer arrangements by docking, based H11 and the L11-12 loop (box A in Fig. 5C). In line with the exper- on WT (PDB 5JJM, chain B without residues 845 to 851) and imental findings, the R761 zone shows important changes in all the current mutant monomeric structures (fig. S5B). The top ranked mutants (residues Asn759 to Tyr764, L5-S1 loop). This region pre- model in WT docking matched the x-ray dimer orientation (5.6 Å sents two different groups of moving residues, consisting of the RMSD), with slightly better binding energy (fig. S5B). For all stretch from Arg789 to Phe795 (H7) and Ser754-Val758 of H4-H5 mutants, we also obtained similar orientations to the experimental (boxes B1 and B2 in Fig. 5C, respectively). Furthermore, in all dimer (RMSD < 10 Å), although not as the lowest-ranking models mutants, changes in average distances of residues Leu729-Asp733 (fig. S5B). Binding energies for V758A and Y764C docking dimers (L3-4 loop and N-terminal end of H4-H5) and Gly821-Asn824 (S3- were similar to those obtained by superimposition on the x-ray 9 loop) are coupled (box C in Fig. 5C). However, in this case, dimer, while in the rest of the mutants, the binding energy of the changes are negative for mutations Q779E and F755V, both nega- docking dimer was more favorable (although showing slightly tive and positive for Y764C and V758A mutants, or positive for higher RMSD). NIP-based interface hotspot residues predicted F755L. Last, residues Cys687 to Gln694 (L1-3) behave differently in from docking (red residues in fig. S5C) were located at the x-ray each mutant. For F755L, this region shows negative changes in their dimer interface in WT and all mutants (except in F755V, which average distances to residues Ala688-Arg711 (L1-3 and H3) and the shows an additional NIP region around Asp829, Arg832, and Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 9 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 5. Molecular dynamics study of the WT and mutant AR-LBD structures demonstrate overall domain stability coupled with local changes in flexibility. (A) Evolution of the RMSD for WT AR- LBD and the studied mutants along four independent MD runs. All structures converge demonstrating global structural stability. (B) Aver- aged RMSF of the four MD runs cal- culated for each structure: WT AR- LBD (black), V758A (blue), Y764C (purple), Q799E (green), F755L (violet), and F755V (red) (top plot). The remaining plots show the diff- erences between the WT and mutant RMSF values. Residues with RMSF values higher than +2σ or lower than −2σ (indicating de- creased or increased flexibility com- pared to WT, respectively) are labeled. Only changes with values greater than twice their SDs were considered significant (red lines). Major changes in flexibility around the mutation site were detected only for the V758A mutant. Mutant F755V only shows a small change in this region, while Q799E does not present any changes near the mutated residue but around Asn759 instead, in line with the experimen- tal structure. Note that flexibility in the L5-S1 loop is remarkably higher in V758A, F755V, and Q799E com- pared to WT (blue). On the contrary, the C-terminal end of H10-11 shows lower relative flexibility in all the mutants. (C) Plots of the differences in average pairwise distances ðΔRij Þ, with ΔRij ¼ Rij;WT Rij;mut , being Rij;WT and Rij;mut the average dis- tances between Cα atoms of resi- dues i and j along the MD trajectories of the WT and mutated structures, respectively. Red and blue represent positive and negative ΔRij values, corresponding to residue pairs whose average dis- tances decrease or increase due to the mutation, respectively. Only residue pairs with ΔRij > 2.50 or < −1.40, which represent the most significant distance changes, are shown. Ile836), with residue at position 755 appearing among the highest predicted dimer interfaces from docking experiments and ODA NIP values in all cases but in F755V. Desolvation areas predicted are consistent with the x-ray dimer orientation (6). by ODA surface analysis were mostly located at the x-ray dimer in- terface in WT and all mutants (fig. S5D), indicating that a large part of the expected dimeric interface has favorable desolvation energy, especially the region around residue position 755. In general, Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 10 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 6. Point mutations at the AR-LBD dimer interface do not impair homodimer formation, which is even enhanced in the Tyr764Cys mutant. (A) SDS-PAGE analysis of WT and mutant AR-LBD samples treated with the bifunctional cross-linkers, BMOE or BMB. Note similar intensities of bands corresponding to dimeric AR-LBD in all cases, indicating that homodimer formation is not compromised by any of the studied point mutants. (B) Representative MS/MS spectra identifying BMB-cross-linked tryptic peptides between Cys764 residues from two LBD monomers. (C) Three-dimensional structure of the AR-LBD dimer (PDB 5JJM) in surface representation. Selected interface residues of both monomers [Cys687, Tyr764, and Phe755 (as reference)] are shown as sticks. The DHT molecule is depicted as salmon spheres. (D) SDS-PAGE analysis of AR-LBD(Y764C) behavior under reducing (lanes 2 and 4, +DTT) and nonreducing conditions (lanes 3 and 5, −DTT). Note the spontaneous dimerization in solution when the protein is incubated under nonreducing conditions. (E) Representative MS/MS spectra demonstrating formation of a disulfide bridge between mutant Cys764 residues from two LBD monomers. m/z stands for mass/charge ratio. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 11 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E The recurrent point mutation Y764C highlights the ability H). This extended conformation is stabilized by the side chain of of AR-LBD dimer interface to adopt different relative Asn759, which accepts an H-bond from the main chain N atom of orientations in AR homodimers Arg761. The proline-like conformation adopted by the preceding AR variant Tyr764Cys was first identified in a PAIS family and has residue, Ser760, is also noteworthy (Fig. 7G). Important contacts since been repeatedly reported in AIS and PCa individuals world- with a crystal neighbor constrain H5-S1 residues from larger dis- wide (36, 39, 42–48). Because this recurrent mutation introduces a placements (Fig. 7, C to H). We reasoned that, in solution, these solvent-exposed cysteine at the AR-LBD dimerization interface rearrangements could lead to even more extensive unfolding of (Fig. 6C), we reasoned that the mutant protein would be more re- the R761 switch zone. Important residue movements in this active, favoring formation of disulfide-bridged homodimers in sol- region were identified in our current MD simulations, some of ution. Y764C but neither WT AR-LBD nor other mutants tested which are larger than the deviations between mutant and WT in rapidly forms covalent dimers in solution at basic pH as assessed experimental structures (Fig. 5B and fig. S6). by nonreducing SDS-PAGE (Fig. 6D). To directly prove that resi- dues Cys764/Cys764′ from each monomer are responsible for disul- Full-length AR is preferentially monomethylated in vivo by fide bridge-mediated dimerization in solution, bands the methyltransferase, PRMT5, and methylation of Arg761 is corresponding to the AR-LBD(Y764C) dimer were excised from affected by AR-LBD dimer interface disease-linked the gel and subjected to enzymatic digestion with endoproteinases mutations trypsin and GluC. Tandem mass spectrometry (MS/MS) analysis of Our structural analysis of AR-LBD point mutants that affect the these digests allowed indeed the identification of peaks correspond- dimer interface revealed important structural rearrangements, ing to Cys764–cross-linked peptides (Fig. 6E and table S3). The which parallel the differences in the transcription profiles of these dimeric AR-LBD conformation previously identified by x-ray crys- variants (see above). These observations prompted us to explore tallography (Fig. 4B) (6) would not support reaction of neighboring possible functional consequences of the studied mutations. Cys764 residues (distance between Cβ atoms: 11 Å; Fig. 5C), but the Residue Arg761 adopts a markedly different, more exposed confor- recent structure of multidomain AR reveals the plasticity of this in- mation in several mutants (Fig. 8, A to G), and it has been previous- terface (7). Therefore, and to provide additional evidence of the ly reported as a substrate of a major type II arginine plasticity of the AR-LBD homodimer, we analyzed by MS/MS the methyltransferase with both mono- and symmetrical dimethylation band corresponding to BMB–cross-linked, dimeric AR- activity, Protein arginine methyltransferase 5 (PRMT5), with im- LBD(Y764C) (Fig. 6B). The results corroborated the presence of a portant functional implications (18). To assess direct physical inter- BMB-mediated linkage between Cys764 Sγ atoms from two mono- actions between FL AR and endogenous PRMT5 in living cells, we mers (Fig. 6B and table S4). performed proximity-ligation assays (PLA). The results of these ex- periments demonstrated strong AR-PRMT5 interactions in PC3- The R761 zone undergoes large conformational WT and PC3-Q799E cells, which were significantly reduced in rearrangements in AR-LBD point mutants PRMT5-silenced cells (Fig. 9, A and B; note the 50% loss of AR- In the AR-LBD homodimer, the major interface is formed by polar PRMT5 interactions for both WT and Q799E). Furthermore, PLA and hydrophobic residues from H5, β strand S1, and their connect- experiments demonstrated arginine monomethylation of FL AR, as ing loop, with additional contributions made by residues from H1 well as weaker symmetric dimethylation, which was reduced 50% and H7-H9 and L1-3 (Fig. 7, A and B). The core dimer interface upon incubation with the PRMT5-specific inhibitor, GSK595 features residues Val758-Arg761, along with the aromatic residues (Fig. 9, D to F). In all PLA experiments, we observed heterogeneous Trp752, Phe755, and Tyr764, which are involved in an intricate patterns, with some nuclei showing strong AR-PRMT5 or methyl- network of hydrogen bond (H-bond) and van der Waals (VdW) in- ation signals while others were practically devoid of these signals teractions. Most notably, stacked residues Trp752 and Phe755 rigidify (Fig. 9, A and C). We also note that levels of Q799E mutant were the H5-H5′ dimer interface, which is strengthened by strong VdW remarkably lower in these experiments. contacts of the latter with Pro802 as well as by H-bonds between the Western blot analysis of immunoprecipitated AR samples veri- guanidinium group of Arg761 and the main chain carbonyl oxygen fied that Arg residues of both WT and mutant AR proteins were atoms of Glu679, Ala680, and/or Glu682 from the neighboring monomethylated in PC3 cells (Fig. 8H). Similar to the results of monomer (Fig. 7B). Unexpectedly, point mutations F755V/L PLA experiments for the Q799E variant (Fig. 9C), the concentra- (Fig. 7, F and G) and Q799E (Fig. 7H) resulted in a fully different tions of mutant proteins were lower, and comparison of the relative conformation of H5-S1 loop residues compared to WT (Fig. 7C). intensities of bands corresponding to total and methylated AR frac- Although residues Val755/Leu755 occupy essentially the same posi- tions revealed that mutants were methylated to a lower extent. This tion as the WT Phe755, loss of stabilizing interactions mediated by feature is probably due to enhanced degradation of the mutant pro- the Phe755 phenol ring ultimately result in the exposure of the teins, compared to WT AR (figs. S3, A and B, and S4) rather than downstream Asn758-Arg761 sequence, which deviates by up to 4 Å reduced expression. In this regard, recombinant WT AR-LBD and from the more compact conformation commonly observed in struc- its mutants are produced at similar levels in a heterologous system tures of AR-LBD, and in which residue Ser760 is essentially buried (Fig. 8I). In light of the structural information presented above, (Fig. 7, C to H). Most notably, the side chain of Arg761 extends into these findings strongly suggest that residue Arg761 is preferentially bulk solvent and exhibits various degrees of freedom in the different monomethylated by PRMT5 in WT and AR-LBD mutants (Figs. 3B and 7, A to G). These residues are best defined by mutants (Fig. 8H). electron density in the F755L and Q799E crystals, in which the gua- In the absence of structural information on the mechanism of nidinium group of Arg761 is sandwiched between the carboxylates recognition and processing of globular, macromolecular substrates of its “own” Glu773 and Asp829 from a crystal neighbor (Fig. 7, G and by the methylosome (the hetero-octameric PRMT5·MEP50 Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 12 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 7. The R761 zone undergoes large conformational rearrangements in AR-LBD point mutants. (A) Surface and cartoon representation of the AR-LBD·DHT dimer (PDB 5JJM). Residues comprising the core dimer interface are shown as color-coded sticks (oxygen, red; nitrogen, blue; carbon, teal or sky blue). (B) Close-up of the core dimer interface. Interface residues are shown as sticks and labeled. Major interdomain hydrogen bonds are represented as orange dotted lines. (C to H) Close-ups of the R761 zone in the crystal structures of mutant AR-LBD (teal-colored cartoons). The closest crystal neighbor is also shown in all cases (colored dark gray), to highlight major interactions due to the crystal packing. Residues involved in the crystal contacts are represented as color-coded sticks. Note the tighter contacts involving the Arg761 side chain in mutants F755V (F), F755L (G), and Q799E (H) compared to WT (C), V758A (D), and Y764C (E), most notably salt bridges with the Asp829 carboxylate from the neighboring monomer. Distances between the posttranslationally modifiable residues, Arg761 and Ser792, and from both of them to the reference residue, Trp797, are also given. complex) and to explore possible binding modes of AR, we per- S1 loop in the AR-LBD would fit into the active site of the methyl- formed docking experiments using the current structure of AR- transferase, supported by additional interactions of residues from LBD Q799E mutant and previously reported structures of the the neighboring PRMT5 monomer (Fig. 9, G and H). Further, in- PRMT5·MEP50 complex (Fig. 9, G and H) (67–70). Similar inter- spection of the docking solutions suggests that the AR-PRMT5 actions could be expected for WT AR and other point mutants. The complex might be stabilized by additional interactions of AR- results of these docking experiments suggest how the extended H5- DBD with the N-terminal subunit of the transferase, while the Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 13 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 8. Arg761 is more solvent exposed in pathological AR mutations affecting on arginine methylation of the full-length receptor. (A to F) Surface representation of the AR-LBD pocket spanned by residues Trp752 and Phe755-Val758 (H5), Tyr764 (S1), Arg789-Gln793, Trp797, and Gln799 (H7) and Asn759-Arg761 (loop H5-S1). All residues are shown as teal spheres superimposed by a gray surface to highlight the pocket topography. Note that conformational changes of Arg761 (labeled in red) and surrounding residues remodel this pocket thus increasing solvent accessibility of Ser792 (salmon), a key phosphorylation site by AKT kinase nested at its bottom. (G) Simplified cartoon representation of the Arg761 zone depicting the major displacement of Arg761 triggered by the dimerization mutants (different shades of red) relative to the WT position (gray). The pathological mutations induce a dislodgement of loop H5-S1, best appreciated by the different conformations adopted by the Arg761 side chain: from an intermediate state in V758A and Y764C (dark red, facing the viewer) to a right-oriented, more solvent-exposed position in F755V, F755L and Q799E (in salmon). (H) Western blot analysis of total and monomethylated FL AR fractions in PC3 cell lines expressing transduced WT or point mutants of the receptor. L, total lysate; IP, im- munoprecipitated fraction. A monoclonal antibody that specifically recognizes monomethylated arginine residues was used for WB. Monomethylated AR bands were double normalized against the total IP AR and against the monomethylated WT AR band (= 1.00), so that monomethylation levels of AR mutants are not biased by the amount of immunoprecipitated proteins. (I) Relative expression levels of WT and mutant AR-LBD recombinant domains in BL21 Escherichia coli cells. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 14 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Fig. 9. AR is preferentially monome- thylated in living cells by the PRMT5·MEP50 complex. (A) Represen- tative PLA images demonstrating AR·PRMT5 physical interactions (red dots). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; blue). PRMT5-silenced cells (si-PRMT5) were compared to nonsilenced cells (si- NS). (B) Quantified PRMT5-AR interac- tions are shown (mean ± SEM, n = 3). Differences between si-PRMT5 and si-NS cells were calculated using a t test. (C) Whole-cell extracts from (A) were ana- lyzed and quantified for AR, PRMT5, and tubulin expression by immunoblot. (D) Representative images from AR methyl- ation assessed using PLA with anti-AR and pan-methyl antibodies recognizing mono- (MMA) or symmetrically dime- thylated arginine residues (SDMA). Red dots correspond to detected MMA and SDMA WT AR, and nuclei counterstained with DAPI (blue). A PRMT5-specific in- hibitor was used to verify that the methylation signal was PRMT5-depen- dent. (E) The number of detected MMA/ AR and SDMA/AR is shown (mean ± SEM, n = 3). Differences between catalytically active and inhibited PRMT5 cells were calculated using a t test. (F) Whole-cell extracts from (D) were analyzed and quantified for total AR, MMA-, and SDMA-protein patterns and tubulin ex- pression by immunoblot. Note that, while SDMA can be mediated only by PRMT5 and PRMT9, MMA can be cata- lyzed by type I and II PRMTs. This ex- plains the larger impact of the PRMT5- specific inhibitor on the SDMA fraction. (G) Model of AR-LBD(Q799E) approach- ing the active site of the PRMT5·MEP50 methylosome. For simplicity, only a heterotetrameric (PRMT5)2·(MEP50)2 complex is shown (PDB 4GQB). Note that folded substrates such as AR-LBD would interact with two neighboring catalytic subunits, in addition to NTD-MEP50 and DBD-PRMT5 contacts. (H) Diagram of putative interactions between AR domains and the methylosome. (I) Summary of PTMs affecting the AR-LBD. Note that PRMT5 regulates the cascade at different levels, including epidermal growth factor receptor (EGFR/HER1), protein kinase B (AKT), and AR. Phosphorylation (yellow), methylation (blue), and ubiquitination (green) pathways are shown: insulin-like growth factor 1 (IGF-I), insulin-like growth factor 1 receptor (IGF-1R), epidermal growth factor (EGF), epidermal growth factor receptor (HER1, 2 or 3), phosphoinositide 3-kinase (PI3K), mitogen-activated protein kinase 4 (MAPK4), aurora kinase A (AURKA), clonal hematopoiesis of indeterminate potential (CHIP), ring finger protein 6 (RNF6), checkpoint kinase 2 (CHK2), proto-oncogene serine/threonine-protein kinase (PIM-1) and E3 ubiquitin-protein ligase (Siah2). Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 15 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E MEP50/WDR77 component could recognize methylated Arg resi- On the other hand, introduction of a cysteine in a solvent- dues within the AR-NTD (Fig. 9, G and H). Synergy between these exposed position at the dimer interface (Y764C) increases the pro- different interactions would ultimately provide the necessary energy pensity to nonproductive dimer formation, particularly in the to enforce structural rearrangements required to insert the nongly- crowded nuclear environment and likely further conformational cine side chains of Ser760 and Met762 in the corresponding subsites changes in the FL protein. In this regard, the recently presented of the enzyme, thus allowing methylation of residue Arg761. low-resolution cryo-EM structures of the core AR have revealed, in addition to an “entrenched” arrangement, which would corre- spond to our previously presented structure of dimeric AR-LBD DISCUSSION (6), a more open or “splayed” assembly of AR monomers (7). Although missense mutations of the AR gene have repeatedly been Our previous MD simulations of the V758A mutant indicated associated with conditions such as PCa and AIS (29), the molecular allosterically transmitted conformational changes in the AR-LBD mechanisms of disease causation and the affected pathways remain (41). The current crystal structures of AR-LBD point mutations poorly understood to date. Previous structural analyses of AR point that affect the dimerization interface confirm and extend these ob- mutants have been limited to internal residues that line the LBP, servations. These replacements induce conformational changes at which are particularly relevant because of their association with re- areas located in the diametrically opposed pole of the domain, sistance to anti-androgen therapy in patients with PCa (e.g., most notably the AF-2 binding pocket (Figs. 1B and 3E). This is W742L, H875Y, and T878A) (26–28). Here, we present a thorough the first time that a mutation in the AR-LBD dimerization site multidisciplinary analysis of the structural and functional impact of has been linked to conformational changes in the AF-2 pocket, sug- several mutations that are buried at the AR-LBD dimer interface (6) gesting that dimerization and coregulators binding are allosterically and which have been repeatedly linked to either AIS and/or PCa coupled. Careful inspection of the crystal structures supported by (F755L/V, V758A, Y764C, and Q799E). Although being essentially MD and SCA analyses suggests allosteric pathways connecting the solvent-exposed residues in monomeric AR and thus unlikely to dimer interface with the LBP and the AF-2 and BF-3 sites in compromise the overall folding of the protein, we show that these AR-LBD. disease-linked variants have significant structural and functional Previous investigations have revealed several functional defects implications. Most notably, transcriptomic analysis demonstrated in the studied mutants, including a lower affinity for agonists, that, despite overall conservation of AR programs, notable differ- which, at least in the case of Y764C, was shown to result from a ences exist at the transcriptional level between WT AR and most more rapid dissociation of ligands without significant changes in mutants, including key drivers of PCa progression such as FOXA1 the LBP, a significant to near complete loss of NTD-LBD (N/C) in- and MYC. In general, mutations at the AR-LBD dimer interface im- teractions, and a largely reduced transcriptional activity, particular- paired AR’s ability to halt proliferation and colony formation while ly at physiological DHT concentrations (summarized in table S2). retaining sensitivity to AR targeting drugs, consistent with a role for Together, these functional findings are in line with our current the nuclear factor as master regulator of differentiation (58). The structural observations of increased mobility of AR-LBD mutants growth-suppressive role for ligand-bound AR had previously been and have important consequences for the structure-function rela- shown (55, 56) and has been the basis for the proposed supraphy- tionship of the receptor. First, the overall decrease in domain stabil- siologic testosterone treatment in patients with PCa that is being ity together with subtle displacements of AF-2 residues might evaluated in clinical settings as part of the bipolar androgen impair both intra- (N-/C-terminal) and intermolecular interactions therapy (71). Mechanistically, AR signaling has been shown to be (with coactivators and corepressors). Conversely, the presence of dichotomous, promoting or repressing growth at low or high activ- the NTD would seem to stabilize the LBD upon N-/C- interactions, ity levels, respectively (72), and distinct transcriptional outputs have as suggested by the lower thermostability of a double mutant with a been identified for patients responding to supraphysiologic testos- shortened polyQ stretch in the NTD in addition to the Y764C point terone (73). mutation (43). This is also in line with the cryo-EM structures of FL These large differences in transcriptional programs were paral- AR, which revealed a compact conformation with a central LBD leled by a largely reduced stability of the mutant AR-LBD modules, surrounded by the NTD (4, 7), and with our recent identification as indicated by notably lower Tm of up to 6°C and increased sensi- of NTD-LBD interactions (74–75). Second, even mutations that tivity to proteolysis in F755L/V and Y764C. Our current high-res- do not affect the LBP directly can have a profound impact on olution crystal structures of the mutant domains reveal, in addition DHT affinity, by facilitating its dissociation from the internal to changes in the targeted dimerization interface, an increased pro- binding site. In this regard, hormone trafficking in and out the pensity to side chain movements of residues that cluster in previous- LBP is likely to involve displacements of residues in the H5-S1 ly described functional areas of the receptor: the exposed AF-2 and loop similar to the ones observed in our current crystal structures BF-3 pockets but also in the minor dimerization site identified in or perhaps of a larger magnitude. N/C interactions of the Q799E our previous crystal structure of multimeric AR-LBD and which mutant were normal, however (48), highlighting the unique molec- evokes the tetrameric assemblies identified in living cells (6). ular fingerprint of each mutation. Similar movements were also identified in MD simulations of The PCa-associated transcription factor, estrogen-responsive mutant proteins (Fig. 5 and fig. S6). Structural rearrangements gene (ERG), was previously shown to recruit protein arginine meth- are particularly noteworthy in variants F755L/V and Q799E, in yltransferase PRMT5 to AR-target genes (18). PRMT5 then meth- which residues of the R761 zone (H5-S1 loop) adopt a previously ylates AR at position 761, ultimately down-regulating AR not observed, much more solvent-exposed conformation, which de- transcriptional activity. Here, we corroborate and extend these find- viates by up to 4 Å from that observed in all previously reported ings by showing a direct physical interaction of endogenous PRMT5 structures of AR-LBD bound to both agonists and antagonists. with both WT and the Q799E mutant of FL AR in living cells. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 16 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Further, we verified that AR is both mono- and, to a lesser extent, Another important PRMT5 substrate, AKT kinase, is also meth- symmetrically dimethylated in cells in a PRMT5-dependent ylated at a noncanonical sequence that resembles the AR-LBD mod- manner (Fig. 9, A to F). The presence of nonglycine residues both ification site (Gln-Arg-Leu) (88). The side chain of the methylated before and after Arg761 would seem to make its methylation less fa- arginine residue, Arg391, is fully buried in the AKT protein core, in- vorable than canonical Gly-Arg-Gly motifs, in light of current dicating that large conformational changes are not uncommon in structural evidence for PRMT5 (68–69) and other methylases PRMT5 substrates. This finding is also relevant because AKT has (76). However, a recent systematic analysis of the human methyl- been shown to phosphorylate AR at two canonical RXRXX(S/T) ome has revealed a large number of substrates with bulky side sites, Ser213 in the NTD and Ser792 in the LBD (12–15). Noteworthy, chains preceding and/or following the modified arginine (77). Ser792 is located close to Arg761, and it is more exposed in the Our docking experiments suggest that Arg761 processing by the oc- mutants with an open H5-S1 loop. However, we have not detected tameric (PRMT5)4·(MEP50)4 heterocomplex known as methylo- significant differences in the degree of phosphorylation between some would depend on secondary interactions of residues in the WT and mutant AR. immediate vicinity of the methylated arginine, on the one side, Inspection of the AR-LBD 3D structure reveals that the dimeri- and exposed residues of the PRMT5 catalytic domain and the N-ter- zation interface comprises, in addition to Arg761 and the nearby minal subunit of a neighboring molecule in the methylosome, on Ser792, several residues that have been previously reported as the other. Substrate recognition might be assisted by additional in- targets of PTMs (Figs. 1A and 9I). This extended patch includes teractions with the highly acidic noncatalytic subunit of the meth- in particular the pair of lysine residues at positions 846 and 848, ylosome, MEP50/WDR77, which have been shown to orient which are ubiquitinated by RNF6 to modulate coactivator recruit- substrate histones toward the buried active site of PRMT5 (78). A ment and AR transcriptional activity (16). Last, residue Thr851 is summary of predicted AR interactions with the methylosome that phosphorylated by the PIM-1L isoform of this Ser/Thr kinase, might ultimately determine specific recognition and processing of which stabilizes AR by recruiting RNF6 (17). The phosphomimetic Arg761 is schematically presented in Fig. 9 (G and H). Thr851Asp mutation resulted in increased transcriptional activity Although the more exposed, protruding conformation of Arg761 due to enhanced recruitment of coactivators at AF-2 (89), mirroring in AR-LBD mutants would favor recognition and processing by the the impact of several point mutations at the dimer interface charac- methylosome, we have actually observed a decrease in the total and terized in the current work. Together, this AR surface emerges as a methylated fractions of the studied AR mutants in cells. Point major hotspot for PTMs (a summary of AR-LBD PTMs and their mutants of the AR-LBD are unlikely to affect protein expression, functional consequences is given in Fig. 9I). These findings under- and we have observed similar expression levels of recombinant score the realization that PTMs regulating DNA accessibility and WT AR-LBD and its mutants. Therefore, this apparent contradic- transcription not only are not limited to histones but also affect tion points to an enhanced degradation of mutant AR protein. AR is also other elements of the transcription machinery (extended polyubiquitinated by several E3 ligases (Fig. 9I), followed by its deg- histone code) (90). Future investigations should depend on the radation in the proteasome. The tumor suppressor, C terminus of cross-talk between PTMs and AR quaternary structure including in- Hsp70-interacting protein (CHIP), appears to be particularly rele- terdomain motions and allosteric pathways connecting the different vant in this regard, as it couples the chaperone machinery to AR functional areas of the nuclear factor and their impact on its inter- degradation, inducing mitotic arrest (79, 80). Of particular clinical actions with specific coregulatory proteins and its proteasome-me- relevance, CHIP has been shown to modulate sensitivity to anti-an- diated degradation. drogen therapy in advanced PCa (81). These findings are also im- In summary, the results of our multidisciplinary approach reveal portant with regard to the current development of proteolysis a complex landscape of structure and function alterations induced targeting chimeras as pharmacological alternatives for PCa treat- by mostly conservative mutations that affect the dimerization inter- ment (82, 83). face of the AR-LBD, which opens various avenues for structure- Multiple additional connections between the PRMT5·MEP50 guided modulation of AR function. The characterization of AR mu- methylosome, AR pathophysiological activity, and PCa etiology tations presented here through a combination of complementary have been reported. PRMT5 functions as an epigenetic activator structural (x-ray crystallography coupled with docking experiments of AR transcription in PCa (84), while the noncatalytic methylo- and MD simulations) and functional assays including transcrip- some component, MEP50 (also termed androgen coactivator tome analysis could be straightforwardly extended to other patho- p44), has been previously identified as an AR interactor linked to logical AR mutants including other oncogenic forms of the PCa progression, which enhances AR-driven gene expression inde- receptor. Early prognosis and precision pharmacology of AR- pendently of PRMT5 catalytic activity. Accordingly, MEP50 expres- linked diseases would benefit from this integration of structural sion levels are higher in PCa tissue than in adjacent, normal prostate and cell biology studies under consideration of the interplay tissue (85), and it has been associated with androgen-independent between AR mutations and PTM-driven pathology. Eventually, it PCa (86). Further, MEP50 is required for homeostasis of prostate might be possible to comprehensibly link all AR mutations to spe- epithelial cells by controlling the proper expression of AR-target cific PCa subtypes, as has been achieved for the related estrogen re- genes, and its translocation from the nucleus into the cytoplasm ceptor in breast cancer (21–22), as an important step toward truly in PCa cells results in exacerbated proliferation of prostate epithelial personalized treatment of patients with PCa. cells (87). Together, the close relationship between methylosome Our results open various avenues for structure-guided modula- and AR activity at the light of our current observations provokes a tion of AR function. The future of precision pharmacology to tackle systematic analysis of AR methylation in patients with PCa and sug- AR-linked diseases and also for the prognosis and treatment of pa- gests avenues for therapeutic intervention, in particular in tients with PCa needs to integrate structural biology with studies of TMPRSS2:ERG-positive PCa (18). cellular effects of mutations and PTM-driven pathology. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 17 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E MATERIALS AND METHODS Functional cell assays Cell lines and plasmids Colony formation and proliferation assays PC3, DU145, LNCaP, 22rv1, and RWPE-1 cell lines were purchased Approximately 30,000 PC3 cells were seeded in six-well plates and from and authenticated by ATCC (American Type Culture Collec- grown in RPMI 1640 media supplemented with 10% CSS, in two tion). For maintenance, cells were cultured in RPMI 1640 or Dul- independent replicates for each experiment. The next day (day 1), becco’s modified Eagle’s medium (DMEM), respectively, cells were stimulated with 20 nM DHT. The stimulation was repeat- supplemented with 10% fetal bovine serum (FBS). For functional ed every 2 days. Colonies were grown until day 9 and stained with assay, cells were depleted of androgens using 10% charcoal-stripped crystal violet [0.5% (w/v) in 10% ethanol and 90% formaldehyde]. serum (CSS; Thermo Fisher Scientific). Site-directed mutagenesis After 30 min of incubation and several washes with water, colonies of a commercial lentiviral AR plasmid (pLENTI6.3/AR-GC- were scanned and quantified using ImageJ software (https://imagej. E2325, plasmid #85128, Addgene) was conducted to generate vari- nih.gov/ij/). Colony formation capacity was normalized to control ants AR-F755V, AR-V758A, AR-Y764C, and AR-Q799E. For lenti- cells. Statistical analysis was performed using t test, showing the dif- virus production, human embryonic kidney 293FT (HEK293-FT) ferences with respect to PC3 WT AR. For colony formation assays in cells were cultured in DMEM (Thermo Fisher Scientific), and the other cell lines, cells were seeded on the basis of their proliferation lentiviral packaging vectors psPAX2 and pCMV-VSV-G were co- rate (DU145: 8000 cells per well; RWPE-1: 20,000 cells per well; transfected in polyethylenimine (1 mg/ml) containing Opti-MEM 22Rv1: 30,000 cells per well; LNCaP: 80,000 cells per well). While (Thermo Fisher Scientific). After viral transduction using superna- DU145 and RWPE cells were cultured in CSS for this assay, tants supplemented with polybrene (8 μg/ml), PC3 cells were incu- 22Rv1 and LNCaP were cultured in FBS to prevent them from stop- bated for 48 hours and selected with blasticidin (5 μg/ml) to ping proliferation. In all cases, media were supplemented with in- generate control, WT AR, and mutants cell lines. creasing concentrations of DHT. Proliferation assays Luciferase reporter assay Cells were trypsinized and counted (with trypan blue stain to dis- To assess AR transcriptional activity, a luciferase-based assay was criminate living cells) every 2 days, starting at day 3 and finishing at performed in PC3 cells, which were seeded (15,000 cells per well) day 7, to follow the time course growth. To compare proliferation at in white with clear flat-bottom 96-well plates (RPMI 1640 media day 7, the number of cells of each PC3 variant was normalized to supplemented with 10% CSS). After 24 hours, 50% confluent cells their corresponding nonstimulated control (without DHT). Statis- were transfected using a Lipofectamine 2000 transfection reagent tical analysis was performed using t test, showing the differences (Thermo Fisher Scientific), following the manufacturer’s instruc- with respect to PC3 WT AR. tions. Renilla luciferase expression is controlled by a constitutive MTT cell viability assays promoter for normalization (pRL-TK Renilla luciferase, Approximately 4000 cells (n = 6) were seeded in 96-well plates in Promega). On the other hand, Firefly luciferase plasmid (pARE- RPMI 1640 media containing 10% CSS. Cells were supplemented 4X-Luciferase) contains four androgen response elements (AREs) the next day with DHT combined with abiraterone acetate (Memo- cloned in tandem into pGL3 (Promega) (91). Transfection was per- rial Sloan Kettering Cancer Center), enzalutamide (MDV3100; formed using 100 ng of pARE-4X-Luciferase and 10 ng of pRL-TK Sigma-Aldrich), or apalutamide (Janssen) at different concentra- Renilla luciferase. Both luciferases are monomeric and neither re- tions. After 72 hours, cells were treated for 3 hours with 10 μl per quires posttranslational processing, so they can function as well of MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium genetic coreporters immediately upon translation. bromide; 5 mg/ml], a yellow tetrazole that is reduced to purple for- Transfection mixes were prepared in Opti-MEM I (Thermo mazan crystals by living cells. Formazan crystals were dissolved with Fisher Scientific). Six hours after transfection, the medium was as- 100 μl of solubilization buffer (10% SDS and 10 mM HCl) for 3 pirated and replaced with RPMI 1640 media supplemented with hours. Plates were read at optical density = 560 nm in a spectropho- 10% CSS. For the assay testing the transactivation level of AR tometer, and then the percentage of viable cells compared to the mutants, cells were stimulated 24 hours after transfection for 4 untreated wells was determined. In the case of abiraterone, the in- hours with vehicle or 20 nM DHT. For luciferase assay testing AR hibitory concentration was calculated using nonlinear regression fit mutants’ resistance to PCa treatment drugs, cells were treated for in GraphPad Prism 8.0 software, normalizing to untreated cells. 30 min with abiraterone (40 or 80 μM) or enzalutamide (50 or Statistical analysis was performed using parametric paired t test 200 μM) before 20-nM DHT stimulation for 4 hours. Luciferase ac- comparing each PC3 AR-mutant regression curve versus PC3 WT tivity was measured after cell incubation with the corresponding AR. Because of drug resistance of PC3 cells against enzalutamide compounds with the Dual-Glo Luciferase Assay System and apalutamide, data did not fit nonlinear regression models. To (Promega) using an EnSpire Alpha plate reader (PerkinElmer). compare enzalutamide and apalutamide resistance, normalized cell After background subtraction, the ratio of firefly to Renilla lucifer- viability (at 50 μM enzalutamide or 100 μM apalutamide) was ex- ase activity was calculated. Assays were performed in three indepen- pressed as a fold change percentage with respect to WT AR. Statis- dent experiments with five internal replicates each. GraphPad Prism tical analysis was performed using a t test, showing the differences 8.0 software was used to perform the statistical analyses using t test, between the PC3 WT AR and its mutants. For viability assays of once confirmed the normal distribution of the values. other cell lines, cells were seeded to reach confluence (90%) at the end of the experiment (DU145: 2500 cells per well; LNCaP, 22Rv1, and RWPE-1: 5000 cells per well). While DU145 and RWPE were cultured in CSS for this assay, 22Rv1 and LNCaP were cultured in FBS to prevent them from stopping proliferation. In both cases, media were supplemented with increasing concentrations of DHT. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 18 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E Immunoprecipitation and Western blotting (CPM). To define DEGs, we considered an absolute log2 fold For immunoprecipitation (IP) of FL AR, PC3 cells were scraped off change ≥ 1 and an adjusted P value or false discovery rate (FDR) the plate and transferred to a microcentrifuge tube using lysis buffer ≤ 0.05. The following DE analyses were performed: PC3 control (LyB) containing 50 mM tris, 150 mM NaCl, 0.2% NP-40, and 10% versus WT AR and point mutants F755V, V758A, Y764C, and glycerol supplemented with protease inhibitor cocktail (Roche), 1% Q799E, as well as WT AR versus PC3, F755V, V758A, Y764C, phosphatase inhibitor cocktail 3 (Sigma-Aldrich), and 0.5% phenyl- and Q799E. To identify the intersection of genes differentially ex- methylsulfonyl fluoride (Sigma-Aldrich) (in following, inhibitor pressed between contrasts, we used the function UpSet from the mix). Samples were sonicated on ice three times for 5 s each, 30% package Complex heatmaps (97). Volcano plots were generated amplitude, and centrifuged. IP of the collected supernatant was per- using the Enhanced Volcano R package on DESeq2. AR targets formed using magnetic beads (Protein G and Protein A Mag Se- were recovered from DoRothEA (57). pharose beads, Thermo Fisher Scientific). AR XP (1:50; Cell For PCA, raw counts were filtered to have at least three samples Signaling, 5153) and immunoglobulin G (IgG; 1:50; rabbit IgG with a CPM ≥ 5. Counts were transformed using the VST function polyclonal–isotype control, Abcam) antibodies were preincubated (98). The PCA was calculated with the plotPCA function in DESeq2 for 1 hour with 25 μl of A + G beads (1:1) at 4°C, and cell lysates and visualized with the ggplot2 package. For the VST data, we cal- (1.6 mg) were precleared with 5 μl of A + G beads (1:1). Beads were culated the Euclidean distances between samples using the R func- removed from the precleared lysates and then transferred into AR or tion dist. We plotted the sample distance matrix with the rows/ IgG tubes for IP at 4°C overnight in a rotating wheel. Afterward, columns arranged by the distances, manually providing sampleD- beads from the negative control (IgG tubes) and the immunoprecip- ists to the clustering distance argument of the pheatmap function. itated samples were washed twice with LyB and five times with LyB lacking NaCl. Last, beads were eluted with 70 μl of Laemmli buffer Enrichment analysis containing 10% β-mercaptoethanol for 5 min at 98°C, and samples GSEA (99) was performed to assess the enrichment of the different were analyzed by Western blotting. mutant AR signatures against a reference signature of DEGs For Western blot analysis of total protein extracts, PC3 cells were between control versus WT AR ranked according to the log2 fold lysed with 1× radioimmunoprecipitation assay (RIPA) buffer [10 change value and the clusterProfiler R package (100) on DESeq2 mM tris-HCl (pH 7.5), 0.1% SDS, 1% deoxycholate sodium salt, results. GSEA was also used to ascertain the enrichment in selected 1% triton X-100, 0.15 M NaCl, and 1 mM EDTA] supplemented gene sets that match against the search for “prostate cancer” in with inhibitor mix. Proteins (30 μg per lane) were resolved by MSigDB or the entire Biological Pathways collection in the KEGG SDS-PAGE and transferred onto a polyvinylidene difluoride mem- database. We performed GSEA against PCa malignant signatures brane (GE Healthcare, Amersham), which was blocked with PBS-T FOXM1, CENPF, and FOXM1 + CENPF for each DESeq2 result (phosphate-buffered saline and 0.1× Tween 20) containing 5% between PC3 control versus WT AR, F755V, V758A, Y764C, or bovine serum albumin (Sigma-Aldrich). Primary antibodies for Q799E. Signatures recovered are defined as top 200 up- and Western blot were incubated at 4°C overnight: anti-AR (1:500; down-regulated DEGs (58). Abcam), anti-mono methyl arginine (R*GG) (1:500; Cell Signal- Transcription factor activity was estimated with the DoRothEA ing), and anti-rabbit IgG horseradish peroxidase–linked antibodies R package (57) using results from the limma package (96) between (1:10,000; Cell Signaling). HRP-conjugated secondary antibody was PC3 versus WT AR or F755V, V758A, Y764C, or Q799E mutants. incubated for 1 hour at room temperature. The ECL Plus Western We estimated activity for pathways involved in cancer with Blotting Detection Kit (Thermo Fisher Scientific) was used for de- PROGENy (101). DEG results between PC3 control versus WT tection in a ChemiDoc MP Imaging System (Bio-Rad). AR, F755V, V758A, Y764C, or Q799E using as input the list of genes ranked by the log2 fold change. Transcriptomic analysis For transcriptomic analysis, approximately 500,000 PC3 cells were AR target genes expression analysis by RT-qPCR seeded in 10-cm plates in RPMI 1640 media supplemented with For qPCR, approximately 300,000 PC3 cells were seeded in six-well 10% CSS (hormone depletion conditions) (n = 3). After 48 hours, plates in duplicates in RPMI 1640 media supplemented with 10% cells were stimulated with 20 nM DHT for 4 hours and collected. FBS. After 48 hours, 70% confluent cells were hormone-depleted RNA was purified using the Maxwell RSC simplyRNA Tissue Kit by using RPMI 1640 media supplemented with 10% CSS. After 24 (Promega), following the manufacturer’s instructions. RNA hours, cells were stimulated with 20 nM DHT for 4 hours and col- quality control, library preparation, 2 × 100 bp, >25 M paired-end lected. RNA was extracted using a TRIzol reagent according to the reads, and stranded mRNA sequencing (RNA-seq) were performed manufacturer’s instructions. After DNA digestion (deoxyribonu- using an Illumina HiSeq2500 platform. RNA-seq lanes were joined clease, Amplification Grade, Invitrogen), cDNA was obtained and reads were mapped with STAR (92) to the primary assembly of using the High-Capacity cDNA Reverse Transcription Kit the human reference genome (GRCh38) in conjunction with the (Applied Biosystems). Total cDNA was diluted 1:10, qPCR was per- complete gene annotation file available in GENCODE (93). The formed using PowerUp SYBR Green Master Mix (Applied Biosys- gene read count was performed with the STAR parameter - quant- tems), and plates were read with a LightCycler 480 (Roche). The Mode GeneCounts. Results indicate, on average, 90% of reads following reverse transcription (RT)–qPCR primers were used to uniquely mapped in all samples; around 80% of the reads are over- amplify glyceraldehyde phosphate dehydrogenase (GAPDH; 5′- lapping genes. CATCTTCCAGGAGCGAGATC; 5′-GCTGTTGTCATACTTCT- We used the IDEAmex service (94) to perform differential ex- CATGG), TMPRSS2 (5′-CCTGACGCAGGCTTCCAAC; 5′- pression analysis using DESeq2 (95) and limma methods (96). GGTCAAGGTGATGCACAGTG), FKBP5 (5′-GCGAAGGAGAA- The count matrix was filtered considering 5 counts per million GACCACGACAT; 5′-TAGGCTTCCCTGCCTCTCCAAA), and Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 19 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E FOXA1 (5′-GCAATACTCGCCTTACGGCT; 5′-TACA- GAGTTGACATTGGTGAAGGATCG; V758A: 5′- TTCAC- CACCTTGGTAGTACGCC). Results were normalized to GADPH CAATGcCAACTCCAGGATGCTC, 5′- GGATCGCCAGCC- expression level, and each sample was quantified in triplicate. Rela- CATGGC; F755V: 5′- CTGGCGATCCgTCACCAATGT, 5′- tive quantification considers GAPDH internal standard and control CCCATGGCAAACACCATGAGC; F755L: 5′- CTGGCGATCCT- conditions (PC3-CNT without DHT) to normalize and compare TaACCAATGT, 5′- CCCATGGCAAACACCATGAGC; and samples. GraphPad Prism 8.0 software was used to perform the stat- Q799E: 5′- TGGATGGCTCgAAATCACCCC, 5′- istical analyses using t test, which confirmed the normal distribu- AACTCTTGAGAGAGGTGC). tion of the values. Differential scanning fluorimetry Proximity ligation assay Thermofluor experiments were performed in an iQ5 Multicolor PLA experiments were performed using reagents from the Naveni- Real-Time PCR Detection System (Bio-Rad) using 96-well plates flex MR Kit (Navinci) essentially as previously described (102). (Hard-Shell High-Profile Semi-Skirted PCR Plate, Bio-Rad) and a Briefly, cells were seeded onto coverslips in 12-well plates, fixed in 25-μl total volume for each reaction essentially as described in methanol for 2 min, and then washed twice in 1× phosphate-buff- 104. Melting curves were acquired from eight replicates to deter- ered saline. Fixed cells were stored at 4°C for subsequent staining or mine the average Tm. AR samples (0.5 mg/ml) were prepared in saturated with the blocking solution for 1 hour at 37°C. All further 20 mM Hepes (pH 8.0), 150 mM Li2SO4, 10% glycerol, 50 mM im- steps were performed at this temperature. Cells were then incubated idazole, 1 mM dithiothreitol (DTT), and 50 μM DHT and centri- with different pairs of primary antibodies: (i) anti-AR (Cell Signal- fuged 5 min at 14,000 rpm. SYPRO Orange dye (Sigma-Aldrich) ing, #5153) and anti-PRMT5 (Santa Cruz Biotechnology, sc- was first prepared at 80× in the protein buffer, starting from a 376937), (ii) anti-AR (Santa Cruz, sc-7305) and anti-MMA (Cell 5000× commercial dilution. The final concentration of SYPRO Signaling, #8711), or (iii) anti-AR (Santa Cruz, sc-7305) and anti- Orange dye in each 25-μl reaction was 5×. The plates were sealed SDMA (Cell Signaling, #13222) for 1 hour. After three washes in with optical quality sealing film (Microseal B Seals, Bio-Rad) and tris-buffered saline with 0.1% Tween® 20 detergent (TBS-T), centrifuged at 2000g for 30 s. Samples were equilibrated for 60 s mouse and rabbit secondary antibodies conjugated with comple- and analyzed using a linear gradient from 16° to 95°C in increments mentary oligonucleotides (“Navenibodies”) were added and incu- of 1°C/min, recording the SYPRO orange fluorescence throughout bated for 1 hour. Again, cells were washed three times in TBS-T the gradient in the iQ5 Optical System Software 2.0. Values were and subsequently incubated with reaction reagents A, B, and C to fitted using the online tool JTSA with the Boltzmann four-param- activate the oligonucleotides, to allow the interaction of Navenibod- eter logistic equation, and the calculated fluorescence shift mid- ies that are in close proximity and the DNA hybridization and am- points were compared via unpaired t test for equal variances in plification of the proximity fluorescent probes (ATTO647), GraphPad Prism 8. respectively. Last, cells were washed 2 min in 1× tris-buffered saline (TBS) and incubated for 15 min with 0.1× TBS at room tem- Proteolytic digestion perature. The samples were mounted using Duolink in situ mount- Samples of recombinant WT AR-LBD and all studied mutants were ing medium containing 4′,6-diamidino-2-phenylindole (DAPI; buffered-exchanged into 25 mM tris (pH 7.6), 10 mM CaCl2, and 1 Sigma-Aldrich, #DUO82040). The edges of the coverslips were mM DTT using a PD-10 column and concentrated to 0.5 mg/ml sealed using nail polish. Slides were then stored in the dark at 4°C using a centrifugal filter device. Sequencing-grade Arg-C protease for a short term or visualized under a Zeiss fluorescence micro- (Clostripain, P5214 Abnova) was added [50/1 (w/w), protein/prote- scope, and interactions were counted for at least 300 cells using ase], and mixtures were incubated at 37°C with shaking. Aliquots of ImageJ software (103). In some experiments, PRMT5 was down- the proteolytic reactions were taken at indicated times, immediately regulated using ON-TARGETplus Human SMARTpool small in- mixed with reducing Laemmli buffer, and heated at 95°C for 5 min. terfering RNA (siRNAs; Dharmacon, #L-015817-00-0005). The un- Reaction products were separated on SDS-polyacrylamide gels and related siRNA ON-TARGETplus Non-targeting Control Pool stained with Coomassie Brilliant Blue. (Dharmacon, #D-001810-10-05) was used as control in these siRNA experiments. Alternatively, PRMT5 activity was blocked by Crystallization and structure determination preincubating the cells for 48 hours with the specific inhibitor, Purified, concentrated DHT-bound AR-LBD mutants (F755L, GSK595 (0.5 μM; Sigma-Aldrich). Cells incubated with the same F755V, V758A, Y764C, and Q799E) were combined with a volume of dimethyl sulfoxide were used as reference in these twofold molar excess of UBA3 peptide and incubated for 1 hour experiments. at room temperature. Drops of the AR-LBD mixture were equili- brated against 0.1 M tris-HCl (pH 8.0) containing either 0.2 M Cloning and mutagenesis of the hAR-LBD construct sodium chloride, 2.0 M ammonium sulfate, 0.1 M ammonium Recombinant human AR-LBD (hAR-LBD; amino acids 641 to 920) acetate, or 2.5 M sodium formate using the sitting drop vapor-dif- was cloned into a modified pGEX vector as previously published fusion method. Diffraction data were collected at 100 K at the (63). The resulting fusion protein consists of a histidine tag se- XALOC beamline of the ALBA-CELLS synchrotron (www.cells. quence followed by a glutathione S-transferase and a Tobacco etch es/en/beamlines/bl13-xaloc) and processed using MOSFLM virus (TEV) cleavage site before the corresponding sequence of the (www.mrc-lmb.cam.ac.uk/harry/mosflm/) and CCP4 (www.ccp4. hAR-LBD. The pGEX vector template encoding WT hAR-LBD was ac.uk/). The crystal structures were solved and refined using mutated using the Q5 Site-Directed Mutagenesis Kit (New England MOLREP (www.ccp4.ac.uk/html/molrep.html#references), Biolabs) to generate AR-LBD mutant vectors with the following REFMAC5 (www2.mrc-lmb.cam.ac.uk/groups/murshudov/ primers (Y764C: 5′-CAGGATGCTCTgGCTTCGC, 5′- content/refmac/refmac.html), and COOT (www2.mrc-lmb.cam.ac. Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 20 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E uk/personal/pemsley/coot/) from the CCP4 package. Crystal Docking experiments and dimeric interface analysis packing was analyzed using PISA (www.ebi.ac.uk/), model quality The dimerization energy of the AR-LBD dimeric conformation was assessed with MolProbity (http://molprobity.biochem.duke. found in the x-ray structure (PDB 5JJM, chains B:C) as well as edu/), and structure figures were prepared with PyMOL (www. that of the modeled mutant dimers (built by superimposition of pymol.org). the mutant monomeric structures onto the corresponding subunits of the crystallographic dimer) was computed with the pyDock Cross-linking experiments bindEy module (105) based on energy terms previously optimized Purified recombinant WT and mutant AR-LBD proteins were incu- for rigid-body docking. The pyDock binding energy is basically bated with fourfold molar excess of BMOE or BMB cross-linkers for composed of Lennard-Jones VdW, Coulombic electrostatics, and 2 hours at 37°C following the manufacturer’s instructions. Samples implicit desolvation based on accessible surface area (ASA) with of the reaction mixtures were boiled in the presence of reducing atomic solvation parameters optimized for protein-protein interac- Laemmli sample buffer and resolved by SDS-PAGE. tions (106). To avoid excessive penalization from possible clashes derived from the rigid-body approach, VdW total contribution Y764C covalent dimerization in solution was weighed by a factor of 0.1. For the same reason, VdW and elec- Purified recombinant Y764C (3 mg/ml) was incubated in 20 mM trostatic values per intermolecular atom pair were restricted to a Hepes (pH 8.0), 150 mM Li2SO4, 10% glycerol, 50 mM imidazole, maximum of +1.0 kcal/mol or to a range between −1.0 and +1.0 1 mM DTT, 50 μM DHT, or in the same buffer without DTT after kcal/mol, respectively. BSA values of WT and mutant dimers were desalting using a PD-10 column (GE Healthcare). Y764C samples computed as the difference in ASA between the dimer and the incubated under reducing and nonreducing conditions were boiled unbound molecules, using ICM-Browser (www.molsoft.com). In in the presence of reducing Laemmli sample buffer and resolved in a addition, homodimeric docking models of WT AR-LBD and its 10% SDS-polyacrylamide gel. mutants were built using the pyDock docking and scoring method (105). First, protein models were prepared by removing Nano–liquid chromatography–MS/MS experiments all cofactors and heteroatoms, and missing side chains were Bands corresponding to dimeric, BMB–cross-linked Y764C were modeled with SCWRL 3.0 (105). Then, the fast Fourier trans- excised from the gels and subjected to in-gel digestion following form–based docking programs FTDock (with electrostatics and the standard protocols. Briefly, excised bands were reduced (10 0.7 Å grid resolution) (www.sbg.bio.ic.ac.uk/docking/ftdock.html) mM DTT) in 50 mM bicarbonate buffer (pH 8.0) for 45 min at and ZDOCK 2.1 (https://zdock.umassmed.edu/) were used to gen- 56°C, alkylated (50 mM iodoacetamide in 50 mM ammonium bi- erate 10,000 and 2000 rigid-body docking poses, respectively. These carbonate buffer for 30 min at 25°C), and digested with trypsin were merged in a single pool for subsequent pyDock scoring, as de- overnight at 37°C in 100 mM ammonium acetate buffer (pH 8). (Se- tailed above. From the resulting docking poses, NIP values were ob- quencing-grade endoproteases were from Promega). tained for each residue with the built-in patch module of pyDock, The following procedures have been performed as previously implementing the pyDockNIP algorithm (65). An NIP value of 1 published (64). Tryptic peptides were diluted in 1% formic acid indicates that the corresponding residue is involved in all predicted (FA) and loaded onto a 180 μm by 20 mm C18 Symmetry trap interfaces of the 100 lowest energy docking solutions, while a value column (Waters) at a flow rate of 15 μl/min using a nanoAcquity of 0 means that it appears as expected by random chance. Last, a Ultra Performance LCTM chromatographic system (Waters). Pep- negative NIP value implies that the residue appears at the low- tides were separated using a C18 analytical column (BEH130 C18, energy docking interfaces less often than random. Usually, residues 75 mm by 25 cm, 1.7 μm; Waters) with a 120-min run, comprising with NIP ≥ 0.2 are considered as hotspot residues. In addition, three consecutive linear gradients: from 1 to 35% B in 100 min, from ODA was obtained by computing surface patches with optimal des- 35 to 50% B in 10 min, and from 50 to 85% B in 10 min (A = 0.1% olvation energy generated from each surface residue (66). Residues FA in water, B = 0.1% FA in CH3CN). The column outlet was di- with low ODA values, usually less than −10.0 kcal/mol, indicate rectly connected to an Advion TriVersa NanoMate fitted on an regions with favorable desolvation energy upon interaction with a Linear trap quadrupole furier transform (LTQ-FT) Ultra mass spec- partner protein. Residues from Lys845 to Ser851 were removed from trometer (Thermo Fisher Scientific), which was operated in positive WT for all energy and docking calculations, for consistency in the mode using the data-dependent acquisition mode. Survey MS scans comparative analysis, since these residues are missing in all mutant were acquired in the Fourier transform ion cyclotron resonance structures. Additional docking experiments were performed using (FT-ICR) cell with the resolution (defined at 400 mass/charge the LZerD protein docking web server (https://lzerd.kiharalab.org/) ratio) set to 100,000. Up to six of the most intense ions per scan with the structure of a hetero-tetrameric (PRMT5)2·(MEP50)2 were fragmented and detected in the linear ion trap. The ion complex as receptor (PDB 7L1G) and the current structure of count target value was 1,000,000 for the survey scan and 50,000 AR-LBD(Q799E) as ligand. Solutions that satisfied the constraints for the MS/MS scan. Target ions already selected for MS/MS were (distance of Arg761-flanking residues to the PRMNT5 catalytic ma- dynamically excluded for 30 s. Spray voltage in the NanoMate chinery, <6 Å) were visually inspected. source was set to 1.70 kV. Capillary voltage and tube lens on the LTQ-FT were tuned to 40 and 120 V, respectively. The minimum Computational analysis of the estimated energetic impact signal required to trigger MS to MS/MS switch was set to 1000, of AR point mutations and activation Q value was set at 0.25. Singly charged precursors We estimated the impact of the studied mutations on the overall were rejected for fragmentation. protein stability with the FoldX empirical force (https://foldxsuite. crg.eu/) (107). Ten iterations were conducted for each mutation and later averaged. Computed free energy differences between mutant Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 21 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E and WT proteins (estimated ΔΔG) < 1 kcal/mol were considered of 4 fs (117), and the SHAKE algorithm (118) was used to constrain not significant, those between 1 and 2, 2 and 4, and >4 kcal/mol the bonds involving the hydrogen atoms. as slightly, mildly, and strongly destabilizing, respectively. (FoldX RMSD and RMSF has an estimated error of ~0.7 kcal/mol). RMSD along the simulation time was computed using the CPPTRAJ module from AMBER18 (https://amber-md.github.io/ Statistical coupling analysis cpptraj/CPPTRAJ.xhtml) for all the MD trajectories to assess the Following previously published protocol [(64), DOI: 10.5281/ structural stability of the systems along time. RMSD was computed zenodo.7274365], we identified two different sectors by using the with respect to all the Cα atoms, using the not minimized x-ray SCA5 8/2011 version (108) and the updated version pySCA (109– structure of the WT as reference for all the systems. The resulting 110) with an alignment of 880 sequences. The stability of the iden- superimposed trajectories were used to calculate the RMSF for tified sectors was assessed with a statistical test based on hypergeo- each of the residues of the protein to obtain information about metric calculations of the groups of residues belonging to given their conformational flexibility. sectors between pairs of alignments. P values were adjusted using Residue pair distance FDR. Next, specific residues from the significant sectors were ex- To quantify the effect of the different mutations on the distances tracted and selected according to their rank. between AR residue pairs, we implemented the algorithm described by Vatansever and coworkers (119). This algorithm is based on the Molecular dynamic simulations concept of first coordination shell defined in the Gaussian network System preparation model but extended to a SCS with a radius of ~9.1 Å to better The crystal structure of monomeric AR-LDB·DHT (PDB 1T7T) was account for the contribution of nonbonded pairs (120). Briefly, used to represent the WT receptor, while the x-ray mutant struc- for each Cα,i atom, the time-averaged distance to any other Cα,j tures used were those determined in this work. In a first step, hydro- atom that belongs to its SCS was obtained for the WT (Rij;WT ) gens were added to all protein residues in their corresponding and the studied mutants (Rij;mut ). Then, the difference (ΔRij ) protonation states at pH 7.0, and side chain orientations were ad- between Rij;WT and Rij;mut was calculated, where justed using the Protein Preparation Wizard workflow included in ΔRij ¼ Rij;WT Rij;mut . To this end, we first calculated for all resi- Maestro v.10.0 software package (www.schrodinger.com/products/ dues the changes in the time-averaged distance between residue i maestro). Missing residues in all the mutant structures (Lys845 to and all j residues that belong to its SCS for the WT structure Ser851) were obtained from the WT structure by superposition. The parameters for DHT were obtained using the general amber (Rij;WT ) and for all mutants (Rij;mut ). The obtained ΔRij differences force field (gaff2) (111), and charges were generated with the re- indicate the degree of distortion introduced by a specific mutation. strained electrostatic potential (112) at the HF/6-31G(d,p) level Residue pairs (i and j) that have the largest positive (ΔRij > 2.50 Å) using the Antechamber module of AMBER18 (https://ambermd. and negative distances (ΔRij < −1.40 Å) between WT and each org/). Next, each system was placed in a cubic periodic box filled mutant were highlighted. with Optimal Point Charge water molecules (113), setting a minimum distance of 15 Å between the solute and the box walls. Water molecules closer than 1.0 Å to any complex atom were Supplementary Materials removed. Then, counter ions were added to neutralize the system This PDF file includes: Figs. S1 to S6 following a grid-shaped procedure for mapping the electrostatic po- Tables S1 to S4 tential surface. All calculations were done using the ff19SB force References field (114) with a cutoff of 10 Å for noncovalent interactions, apply- ing the PME (Particle Mesh Ewald) method (115) for the treatment View/request a protocol for this paper from Bio-protocol. of the electrostatic interactions. Molecular dynamics Before starting the MD calculations, the structures were first relaxed References and Notes to eliminate possible steric clashes with a 5000-step minimization 1. W. D. Tilley, C. M. Wilson, M. Marcelli, M. J. McPhaul, Androgen receptor gene expression procedure using the steepest descent method. Then, the systems in human prostate carcinoma cell lines. Cancer Res. 17, 5382–5386 (1990). 2. E. Estébanez-Perpiñá, C. L. Bevan, I. J. McEwan, Eighty years of targeting androgen re- were stepwise heated to 300 K at a rate of 30 K every 20 ps, fixing ceptor activity in prostate cancer: The fight goes on. Cancers 13, 509 (2021). the main atoms of the proteins with a harmonic positional restric- 3. A. Jiménez-Panizo, P. Pérez, A. M. Rojas, P. Fuentes-Prior, E. Estébanez-Perpiñá, Non-ca- tion of 5 kcal/mol Å−2, using the Langevin thermostat algorithm nonical dimerization of the androgen receptor and other nuclear receptors: Implications with a collision frequency of 3 ps−1 under the NVT ensemble for human disease. Endocr. Relat. Cancer 26, R479–R497 (2019). (heating). Subsequently, 100-ps simulations were performed at 4. X. Yu, P. Yi, R. A. Hamilton, H. Shen, M. Chen, C. E. Foulds, M. A. Mancini, S. J. Ludtke, constant pressure (NPT ensemble) keeping fixed the main atoms Z. Wang, B. W. O’Malley, Structural insights of transcriptionally active, full-length andro- gen receptor coactivator complexes. Mol. Cell 79, 812–823.e4 (2020). with the same harmonic positional restrictions for density equili- 5. P. L. Shaffer, A. Jivan, D. E. Dollins, F. Claessens, D. T. Gewirth, Structural basis of androgen bration (density equilibration). Last, conventional MD of 1 μs for receptor binding to selective androgen response elements. Proc. Natl. Acad. Sci. U.S.A. all the mutants and 1.2 μs for the WT protein were carried out 101, 4758–4763 (2004). within the NVT ensemble in quadruplicate to increase the confor- 6. M. Nadal, S. Prekovic, N. Gallastegui, C. Helsen, M. Abella, K. Zielinska, M. Gay, M. Vilaseca, mational sampling of the system (116). The ParmEd program M. Taulès, A. B. Houtsmuller, M. E. van Royen, F. Claessens, P. Fuentes-Prior, E. Estébanez- Perpiñá, Structure of the homodimeric androgen receptor ligand-binding domain. Nat. (https://parmed.github.io/ParmEd/html/index.html) was used to Commun. 8, 14388 (2017). introduce hydrogen mass repartitioning to allow integration times Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 22 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E 7. E. V. Wasmuth, A. Vanden Broeck, J. R. LaClair, E. A. Hoover, K. E. Lawrence, N. Paknejad, 25. I. A. Hughes, J. D. Davies, T. I. Bunch, V. Pasterski, K. Mastroyannopoulou, J. Macdougall, K. Pappas, D. Matthies, B. Wang, W. Feng, P. A. Watson, J. C. Zinder, W. R. Karthaus, M. J. de Androgen insensitivity syndrome. Lancet 380, 1419–1428 (2012). la Cruz, R. K. Hite, K. Manova-Todorova, Z. Yu, S. T. Weintraub, S. Klinge, C. L. Sawyers, 26. C. E. Bohl, W. Gao, D. D. Miller, C. E. Bell, J. T. Dalton, Structural basis for antagonism and Allosteric interactions prime androgen receptor dimerization and activation. Mol. Cell. 82, resistance of bicalutamide in prostate cancer. Proc. Natl. Acad. Sci. U.S.A. 102, 2021–2031 (2022). 6201–6206 (2005). 8. K. E. Knudsen, T. M. Penning, Partners in crime: Deregulation of AR activity and androgen 27. C. E. Bohl, Z. Wu, D. D. Miller, C. E. Bell, J. T. Dalton, Crystal structure of the T877A human synthesis in prostate cancer. Trends Endocrinol. Metab. 21, 315–324 (2010). androgen receptor ligand-binding domain complexed to cyproterone acetate provides 9. Y. Wen, M. C.-T. Hu, K. Makino, B. Spohn, G. Bartholomeusz, D. H. Yan, M. C. Hung, HER-2/ insight for ligand-induced conformational changes and structure-based drug design. neu promotes androgen-independent survival and growth of prostate cancer cells J. Biol. Chem. 282, 13648–13655 (2007). through the Akt pathway. Cancer Res. 60, 6841–6845 (2000). 28. J. S. Sack, K. F. Kish, C. Wang, R. M. Attar, S. E. Kiefer, Y. An, G. Y. Wu, J. E. Scheffler, 10. H. K. Lin, S. Yeh, H. Y. Kang, C. Chang, Akt suppresses androgen-induced apoptosis by M. E. Salvati, J. Krystek, R. Weinmann, H. M. Einspahr, Crystallographic structures of the phosphorylating and inhibiting androgen receptor. Proc. Natl. Acad. Sci. U.S.A. 98, ligand-binding domains of the androgen receptor and its T877A mutant complexed with 7200–7205 (2001). the natural agonist dihydrotestosterone. Proc. Natl. Acad. Sci. U.S.A. 98, 11. H. K. Lin, Y. C. Hu, L. Yang, S. Altuwaijri, Y. T. Chen, H. Y. Kang, C. Chang, Suppression versus 4904–4909 (2001). induction of androgen receptor functions by the phosphatidylinositol 3-Kinase/Akt 29. B. Gottlieb, L. K. Beitel, A. Nadarajah, M. Paliouras, M. Trifiro, The androgen receptor gene pathway in prostate cancer LNCaP cells with different passage numbers. J. Biol. Chem. mutations database: 2012 update. Hum. Mutat. 33, 887–894 (2012). 278, 50902–50907 (2003). 30. Z. Culig, A. Hobisch, M. Erdel, G. Bartsch, H. Klocker, Studies on androgen receptor mu- 12. S. S. Taneja, S. Ha, N. K. Swenson, H. Y. Huang, P. Lee, J. Melamed, E. Shapiro, tations and amplification in human prostate cancer. Methods Mol. Med. 81, M. J. Garabedian, S. K. Logan, Cell-specific regulation of androgen receptor phosphory- 267–275 (2003). lation in vivo. J. Biol. Chem. 280, 40916–40924 (2005). 31. E. R. Hyytinen, K. Haapala, J. Thompson, I. Lappalainen, M. Roiha, I. Rantala, H. J. Helin, 13. K. Xu, H. Shimelis, D. E. Linn, R. Jiang, X. Yang, F. Sun, Z. Guo, H. Chen, W. Li, H. Chen, O. A. Jänne, M. Vihinen, J. J. Palvimo, P. A. Koivisto, Pattern of somatic androgen receptor X. Kong, J. Melamed, S. Fang, Z. Xiao, T. D. Veenstra, Y. Qiu, Regulation of androgen re- gene mutations in patients with hormone-refractory prostate cancer. Lab. Invest. 82, ceptor transcriptional activity and specificity by RNF6-induced ubiquitination. Cancer Cell 1591–1598 (2002). 15, 270–282 (2009). 32. M. J. McPhaul, M. Marcelli, S. Zoppi, C. M. Wilson, J. E. Griffin, J. D. Wilson, Mutations in the 14. D. E. Linn, X. Yang, Y. Xie, A. Alfano, D. Deshmukh, X. Wang, H. Shimelis, H. Chen, W. Li, ligand-binding domain of theandrogen receptor gene cluster in two regions of the gene. K. Xu, M. Chen, Y. Qiu, Differential regulation of androgen receptor by PIM-1 kinases via J. Clin. Invest. 90, 2097–2101 (1992). phosphorylation-dependent recruitment of distinct ubiquitin E3 ligases. J. Biol. Chem. 33. C. Radmayr, Z. Culig, A. Hobisch, S. Corvin, G. Bartsch, H. Klocker, Analysis of a mutant 287, 22959–22968 (2012). androgen receptor offers a treatment modality in a patient with partial androgen in- 15. D. Gioeli, B. M. Paschal, Post-translational modification of the androgen receptor. Mol. sensitivity syndrome. Eur. Urol. 33, 222–226 (1998). Cell. Endocrinol. 352, 70–78 (2012). 34. T. R. Brown, D. B. Lubahn, E. M. Wilson, F. S. French, C. J. Migeon, J. L. Corden, Functional 16. T. van der Steen, D. J. Tindall, H. Huang, Posttranslational modification of the androgen characterization of naturally occurring mutant androgen receptors from subjects with receptor in prostate cancer. Int. J. Mol. Sci. 14, 14833–14859 (2013). complete androgen insensitivity. Mol. Endocrinol. 4, 1759–1772 (1990). 17. Y. Koryakina, H. Q. Ta, D. Gioeli, Androgen receptor phosphorylation: Biological context 35. J. M. Lobaccaro, S. Lumbroso, R. Ktari, R. Dumas, C. Sultan, An exonic point mutation and functional consequences. Endocr. Relat. Cancer 21, T131–T145 (2014). creates a MaeIll site in the androgen receptor gene of a family with complete androgen 18. Z. Mounir, J. M. Korn, T. Westerling, F. Lin, C. A. Kirby, M. Schirle, G. McAllister, G. Hoffman, insensitivity syndrome. Hum. Mol. Genet. 2, 1041–1043 (1993). N. Ramadan, A. Hartung, Y. Feng, D. R. Kipp, C. Quinn, M. Fodor, J. Baird, M. Schoumacher, 36. Y.-H. Liu, C. Li, S.-F. Zhou, Prediction of deleterious functional effects of non-synonymous R. Meyer, J. Deeds, G. Buchwalter, T. Stams, N. Keen, W. R. Sellers, M. Brown, R. A. Pagliarini, single nucleotide polymorphisms in human nuclear receptor genes using a bioinfor- ERG signaling in prostate cancer is driven through PRMT5-dependent methylation of the matics approach. Drug Metab. Lett. 3, 242–286 (2009). androgen receptor. eLife 5, e13964 (2016). 37. R. Tadokoro-Cuccaro, T. I. Bunch, J. W. Schwabe, I. A. Hughes, J. C. Murphy, Comparison of 19. L. Malbeteau, H. T. Pham, L. Eve, M. R. Stallcup, C. Poulard, M. Le Romancer, How protein the molecular consequences of different mutations at residue 754 and 690 of the an- methylation regulates steroid receptor function. Endocr. Rev. 43, 160–197 (2022). drogen receptor (AR) and androgen insensitivity syndrome (AIS) phenotype. Clin. Endo- 20. Q. Wang, W. Li, X. S. Liu, J. S. Carroll, O. A. Jänne, E. K. Keeton, A. M. Chinnaiyan, K. J. Pienta, crinol. 71, 253–260 (2009). M. Brown, A hierarchical network of transcription factors governs androgen receptor- 38. M. Marcelli, M. Ittmann, S. Mariani, R. Sutherland, R. Nigam, L. Murthy, Y. Zhao, dependent prostate cancer growth. Mol. Cell 27, 380–392 (2007). D. Diconcini, E. Puxeddu, A. Esen, J. Eastham, N. L. Weigel, D. J. Lamb, Androgen receptor 21. D. Robinson, E. M. Van Allen, Y. M. Wu, N. Schultz, R. J. Lonigro, J. M. Mosquera, mutations in prostate cancer. Cancer Res. 60, 944–949 (2000). B. Montgomery, M. E. Taplin, C. C. Pritchard, G. Attard, H. Beltran, W. Abida, R. K. Bradley, 39. X. B. Shi, A. H. Ma, L. Xia, H. J. Kung, R. W. De Vere White, Functional analysis of 44 mutant J. Vinson, X. Cao, P. Vats, L. P. Kunju, M. Hussain, F. Y. Feng, S. A. Tomlins, K. A. Cooney, androgen receptors from human prostate cancer. Cancer Res. 62, 1496–1502 (2002). D. C. Smith, C. Brennan, J. Siddiqui, R. Mehra, Y. Chen, D. E. Rathkopf, M. J. Morris, 40. X. E. Zhou, K. M. Suino-Powell, J. Li, Y. He, J. P. MacKeigan, K. Melcher, E. L. Yong, H. E. Xu, S. B. Solomon, J. C. Durack, V. E. Reuter, A. Gopalan, J. Gao, M. Loda, R. T. Lis, M. Bowden, Identification of SRC3/AIB1 as a preferred coactivator for hormone-activated androgen S. P. Balk, G. Gaviola, C. Sougnez, M. Gupta, E. Y. Yu, E. A. Mostaghel, H. H. Cheng, receptor. J. Biol. Chem. 285, 9161–9171 (2010). H. Mulcahy, L. D. True, S. R. Plymate, H. Dvinge, R. Ferraldeschi, P. Flohr, S. Miranda, 41. S. Grosdidier, L. R. Carbó, V. Buzón, G. Brooke, P. Nguyen, J. D. Baxter, C. L. Bevan, P. Webb, Z. Zafeiriou, N. Tunariu, J. Mateo, R. Perez-Lopez, F. Demichelis, B. D. Robinson, E. Estébanez-Perpiñá, J. Fernández-Recio, Allosteric conversation in the androgen re- M. Schiffman, D. M. Nanus, S. T. Tagawa, A. Sigaras, K. W. Eng, O. Elemento, A. Sboner, ceptor ligand-binding domain surfaces. Mol. Endocrinol. 26, 1078–1090 (2012). E. I. Heath, H. I. Scher, K. J. Pienta, P. Kantoff, J. S. De Bono, M. A. Rubin, P. S. Nelson, 42. C. W. Hay, I. J. McEwan, The impact of point mutations in the human androgen receptor: L. A. Garraway, C. L. Sawyers, A. M. Chinnaiyan, Integrative clinical genomics of advanced Classification of mutations on the basis of transcriptional activity. PLOS ONE 7, prostate cancer. Cell 161, 1215–1228 (2015). e32514 (2012). 22. D. A. Quigley, H. X. Dang, S. G. Zhao, P. Lloyd, R. Aggarwal, J. J. Alumkal, A. Foye, V. Kothari, 43. M. J. McPhaul, M. Marcelli, W. D. Tilley, J. E. Griffin, R. F. Isidro-Gutierrez, J. D. Wilson, M. D. Perry, A. M. Bailey, D. Playdle, T. J. Barnard, L. Zhang, J. Zhang, J. F. Youngren, Molecular basis of androgen resistance in a family with a qualitative abnormality of the M. P. Cieslik, A. Parolia, T. M. Beer, G. Thomas, K. N. Chi, M. Gleave, N. A. Lack, A. Zoubeidi, androgen receptor and responsive to high-dose androgen therapy. J. Clin. Invest. 87, R. E. Reiter, M. B. Rettig, O. Witte, C. J. Ryan, L. Fong, W. Kim, T. Friedlander, J. Chou, H. Li, 1413–1421 (1991). R. Das, H. Li, R. Moussavi-Baygi, H. Goodarzi, L. A. Gilbert, P. N. Lara, C. P. Evans, 44. I. Murono, B. B. Mendonca, I. J. Arnhold, A. C. Rigon, C. J. Migeon, T. R. Brown, Human T. C. Goldstein, J. M. Stuart, S. A. Tomlins, D. E. Spratt, R. K. Cheetham, D. T. Cheng, K. Farh, androgen insensitivity due to point mutations encoding amino acid substitutions in the J. S. Gehring, J. Hakenberg, A. Liao, P. G. Febbo, Genomic hallmarks and structural vari- androgen receptor steroid-binding domain. Hum. Mutat. 6, 152–162 (1995). ation in metastatic prostate cancer. Cell 174, 758–769.e9 (2018). 45. C. A. Quigley, A. De Bellis, K. B. Marschke, M. K. El-Awady, E. M. Wilson, F. S. French, An- 23. D. B. Lubahn, T. R. Brown, J. A. Simental, H. N. Higgs, C. J. Migeon, E. M. Wilson, F. S. French, drogen receptor defects: Historical, clinical, and molecular perspectives. Endocr. Rev. 16, Sequence of the intron/exon junctions of the coding region of the human androgen 271–321 (1995). receptor gene and identification of a point mutation in a family with complete androgen 46. E. Langley, J. A. Kemppainen, E. M. Wilson, Intermolecular NH2−/carboxyl-terminal in- insensitivity. Proc. Natl. Acad. Sci. U.S.A. 86, 9534–9538 (1989). teractions in androgen receptor dimerization revealed by mutations that cause androgen 24. O. Hiort, G. H. Sinnecker, P. M. Holterhus, E. M. Nitsche, K. Kruse, The clinical and molecular insensitivity. J. Biol. Chem. 273, 92–101 (1998). spectrum of androgen insensitivity syndromes. Am. J. Med. Genet. 63, 218–222 (1996). Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 23 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E 47. K. F. Melo, B. B. Mendonca, A. E. C. Billerbeck, E. M. Costa, M. Inácio, F. A. Silva, A. M. Leal, domain explains its manifold physiological activities. Nucleic Acids Res. 50, A. C. Latronico, I. J. Arnhold, Clinical, hormonal, behavioral, and genetic characteristics of gkac1119 (2022). androgen insensitivity syndrome in a Brazilian cohort: Five novel mutations in the an- 65. S. Grosdidier, J. Fernández-Recio, Identification of hot-spot residues in protein-protein drogen receptor gene. J. Clin. Endocrinol. Metab. 88, 3241–3250 (2003). interactions by computational docking. BMC Bioinformatics 9, 447 (2008). 48. I. Jääskeläinen, A. Deeb, J. W. Schwabe, N. P. Mongan, H. Martin, I. A. Hughes, Human 66. J. Fernandez-Recio, M. Totrov, C. Skorodumov, R. Abagyan, Optimal docking area: A new androgen receptor gene ligand-binding-domain mutations leading to disrupted inter- method for predicting protein-protein interaction sites. Proteins 58, 134–143 (2005). action between the N- and C-terminal domains. J. Mol. Endocrinol. 36, 361–368 (2006). 67. L. Sun, M. Wang, Z. Lv, N. Yang, Y. Liu, S. Bao, W. Gong, R. M. Xu, Structural insights into 49. J. Batch, D. Williams, H. Davies, B. Brown, B. A. Evans, I. A. Hughes, M. Patterson, Androgen protein arginine symmetric dimethylation by PRMT5. Proc. Natl. Acad. Sci. U.S.A. 108, receptor gene mutations identified by SSCP in fourteen subjects with androgen insen- 20538–20543 (2011). sitivity syndrome. Hum. Mol. Genet. 1, 497–503 (1992). 68. S. Antonysamy, Z. Bonday, R. M. Campbell, B. Doyle, Z. Druzina, T. Gheyi, B. Han, 50. C. L. Bevan, B. B. Brown, H. R. Davies, B. A. Evans, I. A. Hughes, M. N. Patterson, Functional L. N. Jungheim, Y. Qian, C. Rauch, M. Russell, J. M. Sauder, S. R. Wasserman, K. Weichert, analysis of six androgen receptor mutations identified in patients with partial androgen F. S. Willard, A. Zhang, S. Emtage, Crystal structure of the human PRMT5:MEP50 complex. insensitivity syndrome. Hum. Mol. Genet. 5, 265–273 (1996). Proc. Natl. Acad. Sci. U.S.A. 109, 17960–17965 (2012). 51. Q. Wang, F. J. Ghadessy, A. Trounson, D. de Kretser, R. McLachlan, S. Ng, E. Yong, Azoo- 69. M. C. Ho, C. Wilczek, J. B. Bonanno, L. Xing, J. Seznec, T. Matsui, L. G. Carter, T. Onikubo, spermia associated with a mutation in the ligand-binding domain of an androgen re- P. R. Kumar, M. K. Chan, M. Brenowitz, R. H. Cheng, U. Reimer, S. C. Almo, D. Shechter, ceptor displaying normal ligand binding, but defective trans-activation. J. Clin. Endocrinol. Structure of the arginine methyltransferase PRMT5-MEP50 reveals a mechanism for Metab. 83, 4303–4309 (1998). substrate specificity. PLOS ONE 8, e57008 (2013). 52. O. Hiort, P.-M. Holterhus, T. Horter, W. Schulze, B. Kremke, M. Bals-Pratsch, G. H. G. Sin- 70. K. M. Mulvaney, C. Blomquist, N. Acharya, R. Li, M. J. Ranaghan, M. O’Keefe, D. J. Rodriguez, necker, K. Kruse, Significance of Mutations in the Androgen Receptor Gene in Males with M. J. Young, D. Kesar, D. Pal, M. Stokes, A. J. Nelson, S. S. Jain, A. Yang, Z. Mullin-Bernstein, Idiopathic infertility. J. Clin. Endocrinol. Metab. 85, 2810–2815 (2000). J. Columbus, F. K. Bozal, A. Skepner, D. Raymond, S. LaRussa, D. C. McKinney, Y. Freyzon, 53. N. Kalfa, P. Philibert, R. Werner, F. Audran, A. Bashamboo, H. Lehors, M. Haddad, J. M. Guys, Y. Baidi, D. Porter, A. J. Aguirre, A. Ianari, B. McMillan, W. R. Sellers, Molecular basis for R. Reynaud, P. Alessandrini, K. Wagner, J. Y. Kurzenne, F. Bastiani, J. Bréaud, J. S. Valla, substrate recruitment to the PRMT5 methylosome. Mol. Cell 81, 3481–3495.e7 (2021). G. M. Lacombe, M. Orsini, J. P. Daures, O. Hiort, F. Paris, K. McElreavey, C. Sultan, Minor 71. M. T. Schweizer, E. S. Antonarakis, H. Wang, A. Seun Ajiboye, A. Spitz, H. Cao, J. Luo, hypospadias: The “Tip of the Iceberg” of the partial androgen insensitivity syndrome. M. C. Haffner, S. Yegnasubramanian, M. A. Carducci, M. A. Eisenberger, J. T. Isaacs, PLOS ONE 8, e61824 (2013). S. R. Denmeade, Effect of bipolar androgen therapy for asymptomatic men with castra- 54. J. Qi, M. Tripathi, R. Mishra, N. Sahgal, L. Fazil, S. Ettinger, W. J. Placzek, G. Claps, tion-resistant prostate cancer: Results from a pilot clinical study. Sci. Transl. Med. 7, L. W. Chung, D. Bowtell, M. Gleave, N. Bhowmick, Z. A. Ronai, The E3 ubiquitin ligase siah2 269ra2 (2015). contributes to castration-resistant prostate cancer by regulation of androgen receptor 72. M. D. Nyquist, L. S. Ang, A. Corella, I. M. Coleman, M. P. Meers, A. J. Christiani, C. Pierce, transcriptional activity. Cancer Cell 23, 332–346 (2013). D. H. Janssens, H. E. Meade, A. Bose, L. Brady, T. Howard, N. De Sarkar, S. B. Frank, 55. R. Kumar, J. Mendonca, O. Owoyemi, K. Boyapati, N. Thomas, S. Kanacharoen, M. Coffey, R. F. Dumpit, J. T. Dalton, E. Corey, S. R. Plymate, M. C. Haffner, E. A. Mostaghel, P. S. Nelson, D. Topiwala, C. Gomes, B. Ozbek, T. Jones, M. Rosen, L. Dong, S. Wiens, W. N. Brennen, Selective androgen receptor modulators activate the canonical prostate cancer androgen J. T. Isaacs, A. M. D. Marzo, M. C. Markowski, E. S. Antonarakis, D. Z. Qian, K. J. Pienta, receptor program and repress cancer growth. J. Clin. Invest. 131, e146777 (2021). D. M. Pardoll, M. A. Carducci, S. R. Denmeade, S. K. Kachhap, Supraphysiologic testos- 73. X. Qiu, L. G. Brown, J. L. Conner, H. M. Nguyen, N. Boufaied, S. A. Alaiwi, J. H. Seo, T. El Zarif, terone induces ferroptosis and activates immune pathways through nucleophagy in C. Bell, E. O’Connor, B. Hanratty, M. Pomerantz, M. L. Freedman, M. Brown, M. C. Haffner, prostate cancer. Cancer Res. 81, 5948–5962 (2021). P. S. Nelson, F. Y. Feng, D. P. Labbé, H. W. Long, E. Corey, Response to supraphysiological 56. H. M. Lam, H. M. Nguyen, M. P. Labrecque, L. G. Brown, I. M. Coleman, R. Gulati, B. Lakely, testosterone is predicted by a distinct androgen receptor cistrome. JCI Insight 7, D. Sondheim, P. Chatterjee, B. T. Marck, A. M. Matsumoto, E. A. Mostaghel, M. T. Schweizer, e157164 (2022). P. S. Nelson, E. Corey, Durable response of enzalutamide-resistant prostate cancer to 74. V. Sheikhhassani, B. Scalvini, J. Ng, L. W. H. J. Heling, Y. Ayache, T. M. J. Evers, E. Estébanez- supraphysiological testosterone is associated with a multifaceted growth suppression Perpiñá, I. J. McEwan, A. Mashaghi, Topological dynamics of an intrinsically disordered N- and impaired DNA damage response transcriptomic program in patient-derived xeno- terminal domain of the human androgen receptor. Protein Sci. 31, e4334 (2022). grafts. Eur. Urol. 77, 144–155 (2020). 75. S. A. Mosure, P. Munoz-Tello, K.-T. Kuo, B. M. Tavish, X. Yu, D. Scholl, C. C. Williams, 57. L. Garcia-Alonso, C. H. Holland, M. M. Ibrahim, D. Turei, J. Saez-Rodriguez, Benchmark and T. S. Strutzenberg, J. Bass, R. Brust, A. A. Deniz, P. R. Griffin, D. J. Kojetin, Structural basis of integration of resources for the estimation of human transcription factor activities. interdomain communication in PPARγ (2022); www.biorxiv.org/content/10.1101/2022. Genome Res. 29, 1363–1375 (2019). 07.13.499031v1.abstract. 58. A. Aytes, A. Mitrofanova, C. Lefebvre, M. J. Alvarez, M. Castillo-Martin, T. Zheng, 76. X. Zhang, X. Cheng, Structure of the predominant protein arginine methyltransferase J. A. Eastham, A. Gopalan, K. J. Pienta, M. M. Shen, A. Califano, C. Abate-Shen, Cross- PRMT1 and analysis of its binding to substrate Peptides. Structure 11, 509–520 (2003). species regulatory Network Analysis identifies a synergistic interaction between FOXM1 77. D. Musiani, J. Bok, E. Massignani, L. Wu, T. Tabaglio, M. R. Ippolito, A. Cuomo, U. Ozbek, and CENPF that drives prostate cancer malignancy. Cancer Cell 25, 638–651 (2014). H. Zorgati, U. Ghoshdastider, R. C. Robinson, E. Guccione, T. Bonaldi, Proteomics profiling 59. H. Takahashi, M. Furusato, W. C. Allsbrook, H. Nishii, S. Wakui, J. C. Barrett, J. Boyd, of arginine methylation defines PRMT5 substrate specificity. Sci. Signal. 12, Prevalence of androgen receptor gene mutations in latent prostatic carcinomas from eaat8388 (2019). japanese men. Cancer Res. 55, 1621–1624 (1995). 78. E. S. Burgos, C. Wilczek, T. Onikubo, J. B. Bonanno, J. Jansong, U. Reimer, D. Shechter, 60. W. Weidemann, B. Linck, H. Haupt, B. Mentrup, G. Romalo, K. Stockklauser, Histone H2A and H4 N-terminal tails are positioned by the MEP50 WD repeat protein for A. O. Brinkmann, H. U. Schweikert, K. D. Spindler, Clinical and biochemical investigations efficient methylation by the PRMT5 arginine methyltransferase. J. Biol. Chem. 290, and molecular analysis of subjects with mutations in the androgen receptor gene. Clin. 9674–9689 (2015). Endocrinol. 45, 733–739 (1996). 79. P. Chymkowitch, N. Le May, P. Charneau, E. Compe, J. M. Egly, The phosphorylation of the 61. J. Thompson, F. Saatcioglu, O. A. Jänne, J. J. Palvimo, Disrupted amino- and carboxyl- androgen receptor by TFIIH directs the ubiquitin/proteasome process. EMBO J. 30, terminal interactions of the androgen receptor are linked to androgen insensitivity. Mol. 468–479 (2011). Endocrinol. 15, 923–935 (2001). 80. S. Sarkar, D. L. Brautigan, S. J. Parsons, J. M. Larner, Androgen receptor degradation by the 62. O. A. O’Mahony, M. P. Steinkamp, M. A. Albertelli, M. Brogley, H. Rehman, D. M. Robins, E3 ligase CHIP modulates mitotic arrest in prostate cancer cells. Oncogene 33, Profiling human androgen receptor mutations reveals treatment effects in a mouse 26–33 (2014). model of prostate cancer. Mol. Cancer Res. 6, 1691–1701 (2008). 81. C. Liu, W. Lou, J. C. Yang, L. Liu, C. M. Armstrong, A. P. Lombard, R. Zhao, O. D. V. Noel, 63. E. Estébanez-Perpiñá, L. A. Arnold, P. Nguyen, E. D. Rodrigues, E. Mar, R. Bateman, P. Pallai, C. G. Tepper, H. W. Chen, M. Dall’Era, C. P. Evans, A. C. Gao, Proteostasis by STUB1/HSP70 K. M. Shokat, J. D. Baxter, R. K. Guy, P. Webb, R. J. Fletterick, A surface on the androgen complex controls sensitivity to androgen receptor targeted therapy in advanced prostate receptor that allosterically regulates coactivator binding. Proc. Natl. Acad. Sci. U.S.A. 104, cancer. Nat. Commun. 9, 4700 (2018). 16074–16079 (2007). 82. M. L. Mohler, A. Sikdar, S. Ponnusamy, D. J. Hwang, Y. He, D. D. Miller, R. Narayanan, An 64. A. Jiménez-Panizo, A. Alegre-Martí, T. T. Tettey, G. Fettweis, M. Abella, R. Antón, overview of next-generation androgen receptor-targeted therapeutics in development T. A. Johnson, S. Kim, R. L. Schiltz, I. Núñez-Barrios, J. Font-Díaz, C. Caelles, A. F. Valledor, for the treatment of prostate cancer. Int. J. Mol. Sci. 22, 2124 (2021). P. Pérez, A. M. Rojas, J. Fernández-Recio, D. M. Presman, G. L. Hager, P. Fuentes-Prior, 83. S. Ha, G. Luo, H. Xiang, A comprehensive overview of small-molecule androgen receptor E. Estébanez-Perpiñá, The multivalency of the glucocorticoid receptor ligand-binding degraders: Recent progress and future perspectives. J. Med. Chem. 65, 16128–16154 (2022). Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 24 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E 84. X. Deng, G. Shao, H. T. Zhang, C. Li, D. Zhang, L. Cheng, B. D. Elzey, R. Pili, T. L. Ratliff, 105. T. M. K. Cheng, T. L. Blundell, J. Fernandez-Recio, PyDock: Electrostatics and desolvation J. Huang, C. D. Hu, Protein arginine methyltransferase 5 functions as an epigenetic ac- for effective scoring of rigid-body protein-protein docking. Proteins Struct. Funct. Genet. tivator of the androgen receptor to promote prostate cancer cell growth. Oncogene 36, 68, 503–515 (2007). 1223–1231 (2017). 106. J. Fernández-Recio, M. Totrov, R. Abagyan, Identification of protein-protein interaction 85. K. Hosohata, P. Li, Y. Hosohata, J. Qin, R. G. Roeder, Z. Wang, Purification and identification sites from docking energy landscapes. J. Mol. Biol. 335, 843–865 (2004). of a novel complex which is involved in androgen receptor-dependent transcription. Mol. 107. R. Guerois, J. E. Nielsen, L. Serrano, Predicting changes in the stability of proteins and Cell. Biol. 23, 7019–7029 (2003). protein complexes: A study of more than 1000 mutations. J. Mol. Biol. 320, 86. Y. Peng, F. Chen, J. Melamed, L. Chiriboga, J. Wei, X. Kong, M. Mcleod, Y. Li, C. X. Li, A. Feng, 369–387 (2002). M. J. Garabedian, Z. Wang, R. G. Roeder, P. Lee, Distinct nuclear and cytoplasmic functions 108. N. Halabi, O. Rivoire, S. Leibler, R. Ranganathan, Protein sectors: evolutionary units of of androgen receptor cofactor p44 and association with androgen-independent prostate three-dimensional structure. Cell 138, 774–786 (2009). cancer. Proc. Natl. Acad. Sci. U.S.A. 105, 5236–5241 (2008). 109. S. W. Lockless, R. Ranganathan, Evolutionarily conserved pathways of energetic con- 87. L. Zhou, H. Wu, P. Lee, Z. Wang, Roles of the androgen receptor cofactor p44 in the growth nectivity in protein families. Science 286, 295–299 (1999). of prostate epithelial cells. J. Mol. Endocrinol. 37, 283–300 (2006). 110. O. Rivoire, K. A. Reynolds, R. Ranganathan, Evolution-based functional decomposition of 88. S. Yin, L. Liu, C. Brobbey, V. Palanisamy, LE. Ball, S. K. Olsen, M. C. Ostrowski, W. Gan, proteins. PLOS Comput. Biol. 12, e1004817 (2016). PRMT5-mediated arginine methylation activates AKT kinase to govern tumorigenesis. 111. J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, Development and testing of a Nat. Commun 12, 1–13 (2021). general Amber force field. J. Comput. Chem. 25, 1157–1174 (2004). 89. C. Helsen, T. Nguyen, T. Vercruysse, S. Wouters, D. Daelemans, A. Voet, F. Claessens, The 112. C. I. Bayly, P. Cieplak, W. D. Cornell, P. A. Kollman, A well-behaved electrostatic potential T850D phosphomimetic mutation in the androgen receptor ligand binding domain en- based method using charge restraints for deriving atomic charges: The RESP model. hances recruitment at activation function 2. Int. J. Mol. Sci. 23, 1557 (2022). J. Phys. Chem. 97, 10269–10280 (1993). 90. I. Maksimovic, Y. David, Non-enzymatic covalent modifications as a new chapter in the 113. S. Izadi, R. Anandakrishnan, A. V. Onufriev, Building water models: A different approach. histone code. Trends Biochem. Sci. 46, 718–730 (2021). J. Phys. Chem. Lett. 5, 3863–3871 (2014). 91. H. S. Soifer, N. Souleimanian, S. Wu, A. M. Voskresenskiy, F. K. Collak, B. Cinar, C. A. Stein, 114. C. Tian, K. Kasavajhala, K. A. A. Belfon, L. Raguette, H. Huang, A. N. Migues, J. Bickel, Direct regulation of androgen receptor activity by potent CYP17 inhibitors in prostate Y. Wang, J. Pincay, Q. Wu, C. Simmerling, Ff19SB: Amino-acid-specific protein backbone cancer cells. J. Biol. Chem. 287, 3777–3787 (2012). parameters trained against quantum mechanics energy surfaces in solution. J. Chem. 92. A. Dobin, C. A. Davis, F. Schlesinger, J. Drenkow, C. Zaleski, S. Jha, P. Batut, M. Chaisson, Theory Comput. 16, 528–552 (2020). T. R. Gingeras, STAR: Ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013). 115. T. Darden, D. York, L. Pedersen, Particle mesh Ewald: An N⋅log(N ) method for Ewald sums 93. A. Frankish, M. Diekhans, A. M. Ferreira, R. Johnson, I. Jungreis, J. Loveland, J. M. Mudge, in large systems. J. Chem. Phys. 98, 10089–10092 (1993). C. Sisu, J. Wright, J. Armstrong, I. Barnes, A. Berry, A. Bignell, S. Carbonell Sala, J. Chrast, 116. J. J. Perez, M. S. Tomas, J. Rubio-Martinez, Assessment of the sampling performance of F. Cunningham, T. Di Domenico, S. Donaldson, I. T. Fiddes, C. García Girón, J. M. Gonzalez, multiple-copy dynamics versus a unique trajectory. J. Chem. Inf. Model. 56, T. Grego, M. Hardy, T. Hourlier, T. Hunt, O. G. Izuogu, J. Lagarde, F. J. Martin, L. Martínez, 1950–1962 (2016). S. Mohanan, P. Muir, F. C. P. Navarro, A. Parker, B. Pei, F. Pozo, M. Ruffier, B. M. Schmitt, 117. C. W. Hopkins, S. Le Grand, R. C. Walker, A. E. Roitberg, Long-time-step molecular dy- E. Stapleton, M. M. Suner, I. Sycheva, B. Uszczynska-Ratajczak, J. Xu, A. Yates, D. Zerbino, namics through hydrogen mass repartitioning. J. Chem. Theory Comput. 11, Y. Zhang, B. Aken, J. S. Choudhary, M. Gerstein, R. Guigó, T. J. P. Hubbard, M. Kellis, B. Paten, 1864–1874 (2015). A. Reymond, M. L. Tress, P. Flicek, GENCODE reference annotation for the human and 118. J. P. Ryckaert, G. Ciccotti, H. J. C. Berendsen, Numerical integration of the cartesian mouse genomes. Nucleic Acids Res. 47, D766–D773 (2019). equations of motion of a system with constraints: Molecular dynamics of n-alkanes. 94. V. Jimenez-Jacinto, A. Sanchez-Flores, L. Vega-Alvarado, Integrative differential expres- J. Comput. Phys. 23, 327–341 (1977). sion analysis for multiple experiments (IDEAMEX): a web server tool for integrated RNA- 119. S. Vatansever, B. Erman, Z. H. Gümüş, Oncogenic G12D mutation alters local conforma- seq data analysis. Front. Genet. 10, 279 (2019). tions and dynamics of K-Ras. Sci. Rep. 9, 11730 (2019). 95. M. I. Love, W. Huber, S. Anders, Moderated estimation of fold change and dispersion for 120. A. R. Atilgan, P. Akan, C. Baysal, Small-world communication of residues and significance RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). for protein dynamics. Biophys. J. 86, 85–91 (2004). 96. M. E. Ritchie, B. Phipson, D. Wu, Y. Hu, C. W. Law, W. Shi, G. K. Smyth, limma powers 121. Y. C. Yang, C. A. Banuelos, N. R. Mawji, J. Wang, M. Kato, S. Haile, I. J. McEwan, S. Plymate, differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids M. D. Sadar, Targeting androgen receptor activation function-1 with EPI to overcome Res. 43, e47 (2015). resistance mechanisms in castration-resistant prostate cancer. Clin. Cancer Res. 22, 97. Z. Gu, R. Eils, M. Schlesner, Complex heatmaps reveal patterns and correlations in mul- 4466–4477 (2016). tidimensional genomic data. Bioinformatics 32, 2847–2849 (2016). 122. N. G. R. D. Elshan, M. B. Rettig, M. E. Jung, Molecules targeting the androgen receptor (AR) 98. S. Anders, W. Huber, Differential expression analysis for sequence count data. Genome signaling axis beyond the AR-Ligand binding domain. Med. Res. Rev. 39, 910–960 (2019). Biol. 11, R106 (2010). 99. A. Subramanian, P. Tamayo, V. K. Mootha, S. Mukherjee, B. L. Ebert, M. A. Gillette, Acknowledgments: We thank J. Lüders (IRB) for critical comments and suggestions on the A. Paulovich, S. L. Pomeroy, T. R. Golub, E. S. Lander, J. P. Mesirov, Gene set enrichment manuscript. We thank M. Taulès (CCiTUB), M. Vilaseca (MS facility), R. El Bizri, L. Franco (Idibell), analysis: A knowledge-based approach for interpreting genome-wide expression profiles. and K. Kahlem (CNAG) for technical support. We thank ALBA-Cells synchrotron Xaloc team for Proc. Natl. Acad. Sci. U.S.A. 102, 15545–15550 (2005). beamline support. Funding: E.E.-P. thanks G. E. Carretero and A. T. Hagler for sponsorships, 100. T. Wu, E. Hu, S. Xu, M. Chen, P. Guo, Z. Dai, T. Feng, L. Zhou, W. Tang, L. Zhan, X. Fu, S. Liu, CaixaImpulse 2018-00018 La Caixa, and 2018-Llavor with 2017SGR1089 (to P.F.-P.; Generalitat X. Bo, G. Yu, clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. de Catalunya). E.E.-P. thanks 2021SGR01433, BFU-Retos2017-86906-R, PDC2021-121688-100, Innovation 2, 100141 (2021). and SAF2017-71878-REDT (Red Nacional de Receptores Nucleares, NurCaMeIn) with SAF2019- 101. M. Schubert, B. Klinger, M. Klünemann, A. Sieber, F. Uhlitz, S. Sauer, M. J. Garnett, REDT-Uroncomol (to A.A.; Red Nacional de Mecanismos Moleculares del Cáncer Urológico, N. Blüthgen, J. Saez-Rodriguez, Perturbation-response genes reveal signaling footprints Uroncomol), RTI2018-101500-B-I00 (to P.F.-P.), PID2019-110167RB-I00/AEI/10.13039/ in cancer gene expression. Nat. Commun. 9, 20 (2018). 501100011033 (to J.F.-R.), and RTI2018-096735-B-100 (to A.R.; MINECO). A.A. thanks PI19/00342 102. C. Poulard, J. Jacquemetton, T. H. Pham, M. Le Romancer, Using proximity ligation assay and CPII20/00024 (Instituto de Salud Carlos III), W81XWH-18-1-0193 (Department of Defense), to detect protein arginine methylation. Methods 175, 66–71 (2020). and 2017SGR448 (Generalitat de Catalunya). U.E. acknowledges support from the Spanish 103. J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, Ministry of Science and Innovation (RYC2020-029773-I and PID2021-128682OA-I00). M.N.P.-M. C. Rueden, S. Saalfeld, B. Schmid, J. Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, thanks Predocs-UB2020 5760700. J.R.-M. thanks grants 2021SGR00350 and CEX2021-001202-M. P. Tomancak, A. Cardona, Fiji: An open-source platform for biological-image analysis. Nat. C.P. and M.L.R. were supported by financial support from ITMO Cancer of Aviesan within the Methods 9, 676–682 (2012). framework of the 2021-2030 Cancer Control Strategy, on funds administered by Inserm. Author 104. R. Gustafsson, U. Eckhard, W. Ye, E. D. Enbody, M. Pettersson, P. Jemth, L. Andersson, contributions: E.E.-P., P.F.-P., and A.A conceived and designed the initial project and supervised M. Selmer, Structure and characterization of phosphoglucomutase 5 from atlantic and its overall execution, provided financial support, and share overall responsibility for the baltic herring—An inactive enzyme with intact substrate binding. Biomolecules. 10, presented results and final approval of the article. A.A.-M., A.J.-P., M.A., and R.A. expressed and 1631 (2020). purified recombinant proteins. A.A.-M., A.J.-P., and M.A. performed crystallization trials. A.A.-M., A.J.-P., R.A., and E.E.-P. performed site-directed mutagenesis and protein analyses. A.A.-M. and U.E. conducted and analyzed DSF experiments. A.M.R. performed sequence and SCA analyses with A.J.-P. P.F.-P. and E.E.-P. interpreted sequence and SCA analyses. J.F.-R. performed docking, Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 25 of 26 S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E NIP, and ODA calculations. J.F.-R., A.A.-M., P.F.-P., and E.E.-P. interpreted docking, NIP, and ODA the paper and/or the Supplementary Materials. The atomic coordinates and structure factors calculations. A.A.-M., A.J.-P., R.A., P.F.-P., and E.E.-P. performed MS experiments and interpreted reported here have been deposited in the PDB with the accession codes 7ZTZ (Y764C), 7ZU2 results. A.A.-M., A.J.-P., P.F.-P., and E.E.-P. collected x-ray crystallography diffraction data, P.F.-P. (Q799E), 7ZU1 (V758A), 7ZTX (F755V), and 7ZTV (F755L). The PDB accessibility has been interpreted diffraction data and solved and refined 3D structures. P.F.-P., A.A.-M., A.J.-P., and E.E.- designed “for immediate release on publication.” The MS proteomics data have been deposited P. analyzed and interpreted the structures. A.A.-M., A.M.-T., M.C., and A.A. performed and to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset supervised RNA-seq experiments with created stable PCa cell lines and performed transcription identifier PXD028039. The transcriptomic data have been deposited at the NCBI Gene activity, anti-androgens, and cell proliferation assays. C.P. and M.L.R. performed analysis of AR- Expression Omnibus (GEO) repository under the accession code GSE210393. Bioinformatics PRMT5 interactions and AR methylation in living cells using PLA assays. A.A. and J.M.P. advised data and code for the clustering and sector analyses are available at Zenodo (DOI: 10.5281/ on anti-androgen treatments and discussions on prostate cancer implications. J.R.-M. and zenodo.7274365). M.N.P.-M. performed MD simulations and interpreted results with A.A.-M., J.F.-R., E.E.-P., and P.F.- P. E.E.-P., P.F.-P., and A.A. drafted the article. A.A.-M., A.J.-P., A.A., P.F.-P., and E.E.-P. critically Submitted 2 August 2022 reviewed the manuscript. All authors discussed the results and commented on the manuscript. Accepted 15 February 2023 Competing interests: The authors declare that they have no competing interests. Data and Published 15 March 2023 materials availability: All data needed to evaluate the conclusions in the paper are present in 10.1126/sciadv.ade2175 Alegre-Martí et al., Sci. Adv. 9, eade2175 (2023) 15 March 2023 26 of 26