DNA‐binding domain as the minimal region driving RNA‐dependent liquid–liquid phase separation of androgen receptor

Abstract

Androgen receptor (AR) is a nuclear hormone receptor that regulates the transcription of genes involved in the development of testis, prostate and the nervous system. Misregulation of AR is a major driver of prostate cancer (PC). The primary agonist of full‐length AR is testosterone, whereas its splice variants, for example, AR‐v7 implicated in cancer may lack a ligand‐binding domain and are thus devoid of proper hormonal control. Recently, it was demonstrated that full‐length AR, but not AR‐v7, can undergo liquid–liquid phase separation (LLPS) in a cellular model of PC. In a detailed bioinformatics and deletion analysis, we have analyzed which AR region is responsible for LLPS. We found that its DNA‐binding domain (DBD) can bind RNA and can undergo RNA‐dependent LLPS. RNA regulates its LLPS in a reentrant manner, that is, it has an inhibitory effect at higher concentrations. As RNA binds DBD more weakly than DNA, while both RNA and DNA localizes into AR droplets, its LLPS depends on the relative concentration of the two nucleic acids. The region immediately preceding DBD has no effect on the LLPS propensity of AR, whereas the functional part of its long N‐terminal disordered transactivation domain termed activation function 1 (AF1) inhibits AR‐v7 phase separation. We suggest that the resulting diminished LLPS tendency of AR‐v7 may contribute to the misregulation of the transcription function of AR in prostate cancer.

1. INTRODUCTION

A transformative recent discovery in molecular cell biology is that besides well‐known and amply studied membrane‐bound organelles, such as the mitochondrion and the nucleus, cellular material is also organized in complex ways by organelles that are not limited by lipid membranes. ¹ , ² Several of these membraneless organelles (MLOs), such as the nucleolus in the nucleus or the stress‐granule in the cytoplasm, have been known to exist for a long time, but it was only recently recognized that they arise by liquid–liquid phase separation (LLPS), which results in dynamic, liquid‐like cellular bodies. MLOs are thought to carry out specific cellular functions, such as localizing biochemical pathways, sequestering cellular material, targeting biochemical reactions, buffering protein and RNA components, generating force, sensing environmental changes, and many more. ³ , ⁴

Cellular MLOs often contain many specific protein and RNA components, which play different roles in LLPS and MLO formation. Roughly, they can be divided to four categories: (i) Drivers, which drive LLSP on their own, (ii) co‐drivers which need another partner to phase separate, (iii) clients, which localize to droplets generated by LLPS but have no influence on the process itself, and (iv) regulators, most often modifying enzymes, which promote or inhibit LLPS, but do not take part in the final droplet formed. ⁵ Due to the explosive expansion of the field, the regulation and function of different MLOs is the subject of intense scrutiny. Owing to the complexity of the underlying regulatory relations, the exact roles and mechanisms of LLPS‐related proteins can only be ascertained by a combination of in vitro and in vivo experimental approaches. It is of special relevance that LLPS proteins have abundant structural disorder, and multivalent interactions required for mediating phase separation are often realized by a combination of various types of interactions in which intrinsically disordered regions (IDRs) play key roles. ⁶ The importance of LLPS is underscored not only by its prevalence in cell biology, but also by its involvement in disease. Of these, neurodegenerative disorders have been mostly in the focus, but possible roles in other diseases, such as cancer have also been implicated. ⁷ A particular area where LLPS might be involved in both physiological and pathological processes is that of transcription regulation. ⁸ , ⁹

Mediating and regulating the complex process of gene transcription requires the recruitment of a vast number of proteins around the genomic loci. According to the current mechanism, transcription factors bind to the enhancer or promoter elements and recruit coactivators, RNA polymerase II and the Mediator complex to facilitate gene transcription. However, the exact mechanism of regulation has remained elusive. Many of the proteins involved in the gene transcription contain long disordered domains with low sequence complexity which has been shown, in many cases, to promote LLPS. ¹⁰ In addition, in the presence of their natural ligand the heterogeneous distribution of steroid dependent nuclear receptors—known as foci formation—at the genomic loci has been observed decades ago. ¹¹ , ¹² However, for a long time studying the nature of these objects in the nucleus was cumbersome owing to their highly dynamic behavior and small size and, therefore, their role remained unclear. Due to the emergence of new high‐resolution and single molecule microscopy techniques, these questions can now be addressed. The crucial role of LLPS in mediating transcriptional activity has recently been demonstrated in the case of Super Enhancers in embryonic stem cells. ¹³ MED1, a subunit of the Mediator complex that contains IDRs can phase separate on its own as well as with many TFs, including GR and ER. Recently, it was shown that androgen receptor (AR) forms dynamic AR‐rich, liquid‐like foci in cellular prostate cancer (PC) models in a MED1‐dependent manner. ¹⁴ AR phase separation was also observed to be promoted by a SPOP, a cullin3‐RING ubiquitin ligase (CRL3) adaptor of tumor suppressor activity. ¹⁵

AR is a multidomain transcription factor of 919 amino acids. Its N‐terminal domain (NTD) (amino acid 1‐555) is an IDR, followed by the DNA‐binding domain DBD (555‐623) and ligand binding domain (LBD) (665‐919), which are folded (globular) domains. ¹⁶ Conventionally, drugs have been used to target the LBD, however, these are often ineffective in late—castration resistant—stages of PC, primarily because of the appearance of splice variants such as AR‐v7 that lack the ligand‐binding domain. ¹⁷ Interestingly, AR‐v7 does not undergo foci formation, ¹⁴ suggesting an internal interlay between AR domains inhibitory to LLPS.

Here, we used bioinformatics and domain deletions to uncover the role of various domains in the LLPS of AR. We make the intriguing observation that its DBD mediates RNA‐dependent LLPS, which is downregulated by activator function 1 (AF1) within its NTD. RNA‐driven LLPS shows a reentrant behavior and is dependent on DNA, which suggests a complex interplay of AR foci formation at transcriptionally active regions. In all, we suggest that the observed diminished LLPS tendency of AR‐v7 might be due to a loss of interaction between AF1 and LBD ¹⁸ and subsequent increased inhibitory effect of AF1 on AR‐v7 LLPS, which may be an important component of the misregulation of the transcription function of AR in PC.

2. RESULTS

2.1. Sequence‐based indications for the phase separation propensity of AR‐v7

The LLPS propensity of AR is underscored by that many nuclear hormone receptors have been observed to drive phase separation both in vitro into liquid‐like droplets and in vivo into MLOs (Table 1).

TABLE 1.

Similarity of AR‐v7 to other phase‐separating hormone‐activated transcription factors

Protein	Condensate	Source	BLAST e‐value	Reference publication
Human androgen receptor (1–559)	In vitro droplet	LLPSDB	0.0	15
Human estrogen receptor	In vitro droplet	LLPSDB	3.63e‐23	8
Human retinoic acid receptor α	In vitro droplet	LLPSDB	9.24e‐16	8
Human estrogen receptor	Enhanceosome (scaffold)	PhaSePro	1e‐22	19
Human estrogen receptor	In vitro droplet (regulator)	DrLLPS	7e‐22	8
Human retinoic acid receptor α	In vitro droplet (scaffold)PML nuclear body (scaffold)	DrLLPS	2e‐14	8, 20
Human thyroid hormone receptor α	Stress granule (regulator)	DrLLPS	4e‐13	21
Human peroxisome proliferator‐activated receptor α	Centrosome (client)	DrLLPS	1e‐10	22
Human glucocorticoid receptor	PML nuclear body (client)	DrLLPS	1e‐32	23
Mouse steroidogenic factor	Nuclear speckle (client)	DrLLPS	6e‐18	24

Abbreviation: AR, androgen receptor.

In addition, the NTD of AR was shown to phase separate at 100 μM protein concentration with 8% ficoll 70 and 60 mM NaCl. ¹⁵ The presence of 15 μM Speckle‐type POZ protein (SPOP), a transcriptional repressor, significantly lowered the saturation concentration (c _sat) of AR‐NTD to 40 μM under the same conditions. ¹⁵ The estrogen receptor (ER), the closest relative of AR, has been shown to scaffold enhanceosomes in vivo, while it has been demonstrated to undergo LLPS in vitro at c _sat of 40 μM in presence of 10% PEG‐8000 and 125 mM NaCl. ⁸ Similarly to AR, c _sat of ER is also lowered to 10 μM by a partner MED1. ⁸

Based on these previous results, first we investigated which regions of AR, in particular within AR‐v7, could mediate LLPS under physiological protein concentrations. To this end, we used sequence‐based LLPS prediction tools (PScore ²⁵ and catGRANULE ²⁶ ) to generate hypotheses which regions of AR‐v7 are LLPS‐prone (Figure 1(a), (b)): Both tools predicted region G425‐D496. This region is located at the C‐terminal end of NTD, preceding DBD, and will be termed Fragment 5 (F5) herein.

FIGURE 1.

Sequence‐based bioinformatics predictions for AR‐v7. Various prediction tools were used to assess potential LLPS‐related features of the AR‐v7 sequence (UniProt: P10275‐3). For LLPS propensity, we used (a) PScore ²⁵ [and (b) catGRANULE ²⁶ direct LLPS‐propensity predictions. (c) Charge distribution at pH 7.4 was smoothed with a 20‐aa sliding window, whereas for RNA‐binding propensity (d) PPRINT ⁴² RNA‐binding scale was used. AR, androgen receptor; LLPS, liquid–liquid phase separation

That AR (and potentially AR‐v7) could undergo LLPS in the cell is further indicated by its high number of known partner proteins that are known drivers of LLPS (Table 2). SPOP and DAXX are not the only examples of co‐scaffolding partners for AR, ¹⁵ among the others two interaction partners NONO and PSF were shown to drive paraspeckle formation, ²⁷ whereas MED1 and BRD4 are partners in enhanceosome formation. ²⁸

TABLE 2.

Interaction partners of AR also undergo phase‐separation

AR region involved in the binding	Reference for binding	Binding partner	Reference for the LLPS of the partner
LBD	29	NCOA3 (SRC‐3)	30
Full‐length	31	FUS	32
NTD	33	MED1	28
NTD	34	BRD4	28
NTD	35	DAXX	15
Hinge region	36	SPOP	15
NTD	37	GRB2	38
NTD	39	NONO	27
NTD	39	PSF	27

Abbreviations: AR, androgen receptor; LBD, ligand binding domain; NTD, N‐terminal domain.

Apart from assessing the LLPS propensity of AR alone, we also wanted to address if it could co‐scaffold the condensate formation with nucleic acids, because both RNA and DNA are well‐known co‐drivers of LLPS in various well‐established systems of biomolecular condensates. ⁵ , ⁴⁰ As AR and AR‐v7 can bind DNA via their DBD, DNA can potentially co‐scaffold the phase‐separated droplet formation. Due to DNA binding regions often also contributing to binding RNA ⁴¹ we first asked if AR has putative RNA‐binding region(s) that would enable it to phase separate in an RNA‐dependent manner. The charge distribution of AR‐v7 already hinted that the DBD has a positive net charge (Figure 1(c)), a prerequisite for binding the negatively charged RNAs. When predicting the RNA‐binding propensity of the protein by PPRINT, a position‐specific scoring matrix (PSSM) and machine learning‐based (SVM) predictor, ⁴² a clear signal within region D565‐G644 (overlapping with the DBD) for RNA binding appears (Figure 1(d)).

2.2. DBD is an RNA‐binding domain

Therefore, first we set out to experimentally address the RNA‐binding propensity of AR DBD. To this end, we carried out microscale thermophoresis (MST) measurements with Cy5‐labeled U30 RNA and unlabeled DBD. Although data points for saturation concentration were not reached, the results unambiguously confirmed the RNA‐binding of DBD, with an approximate Kd of 150 ± 70 μM (Figure 2(a)). In comparison, we also measured the interaction of AR DBD with the RNA mimic polyuridylic acid (poly U), tRNA and specific DNA‐recognition regions of AR (Androgen response element ARE1 and ARE2) by biolayer interferometry (BLI). The Kds determined (Figure 2(b)–(e)) show a preference for DNA binding but only slightly weaker binding of different RNA variants. DBD‐poly U interaction was also confirmed by MST (Figure S1).

FIGURE 2.

Interaction between DNA‐binding domain (DBD) and RNA/ARE (DNA). (a) The interaction of AR‐v7's DBD was measured with Cy5‐labeled U30 at concentration of 300 nM in Tris buffer (50 mM Tris, 150 mM NaCl, 0.05% Tween 20, pH 7.5) by MST. Interaction of immobilized (b) biotinylated poly U; (c) tRNA; (d) ARE‐1; (e) ARE‐2 on streptavidin sensors and DBD in solution measured by BLI. ARE, androgen response element; BLI, biolayer interferometry; DBD, DNA‐binding domain; LLPS, liquid–liquid phase separation; poly U, polyuridylic acid

2.3. DBD undergoes LLPS with RNA

As full‐length AR appears to be required for LLPS, ¹⁴ we tested its different regions for LLPS in the presence of 0.05 μg/μl poly U RNA mimic by following turbidity of the solution (optical density 600 nm OD600). The C‐terminal region of NTD (F5) predicted as a potential LLPS‐driving region (Figure 1(a), (b)), did not show an increase in turbidity, however, DBD and a longer construct F5‐DBD exhibited a transient increase in OD600 (Figure 3(a), Figure S2). This result suggests that DBD acts as a minimal region for co‐driving phase separation with RNA. Without a small amount of RNA (poly U concentration of 0.05 μg/μl) DBD could not drive LLPS alone, as observed in both turbidity (Figure 3(a)) and dynamic light scattering (DLS) (Figure 3(b)) measurements. DLS showed that the hydrodynamic radii of droplets steadily increased in the first hour after mixing, then reached a plateau at an average radius of approximately 750–800 nm. Droplet formation and an increase in droplet size was also observed by fluorescence microscopy by Dylight 488‐DBD and Cy5‐poly U (Figure 3(c)). The inclusion of F5 to the DBD construct (F5‐DBD) increased the droplet size of the biomolecular condensates co‐scaffolded by F5‐DBD and poly U, suggested by both the turbidity and DLS measurements (Figure S3B–D). The hydrodynamic radii of these droplets plateaued after 1 h at an average radius of around 1200 nm.

FIGURE 3.

Phase separation of DNA‐binding domain (DBD) with polyuridylic acid (poly U). (a) OD 600 at different concentrations of DBDin presence of 0.05 μg/μl of poly U (b) Hydrodynamic radius of DBD measured by dynamic light scattering (DLS) in the presence (filled circles) and absence of 0.05 μg/μl poly U (filled squares). (c) Dylight® 488‐labeled DBD(green) was used at 50 μM and mixed with Cy5‐labeled poly U (red) for final concentration of 0.05 μg/μl of poly U. Panels C show the droplets (upper snapshots) right after phase separation, and (lower snapshots) after 1 h

We have further analyzed details of LLPS driven by the DBD‐RNA interaction. First, we checked various constructs to see the additional effects of Ar‐v7 regions on DBD‐driven LLPS (Figure 4(a)). AF1 or F5 itself did not phase separate, whereas F5 fused to DBD had a slight (although not significant) effect. Apparently, AF1 inhibits DBD‐driven LLPS, as neither AF1‐DBD nor AR‐v7 showed any tendency to phase separate. To check if this tendency can be reversed by crowding, we repeated LLPS experiments with various RNA variants in the presence of 10% dextran (Figure S4), but observed only very minor increase.

FIGURE 4.

Phase separation of different constructs of AR‐v7 in the presence of RNA. (a) OD 600 for different constructs of AR‐v7 (as defined in the text, AF1: activation function 1, F5: fragment 5, DBD: DNA‐binding domain, F5‐DBD: F5 + DBD fused, AF1‐DBD: AF1 + DBD fused, and AR‐v7: AR splice variant v7) at a concentration of 25 μM, in the presence of 0.05 μg/μl poly U with buffer (50 mM phosphate, 50 mM NaCl, 0.5 mM TCEP, pH 7.0). Control is measured in the absence of RNA. (b) DBD with 0.05 μg/μl of different types of RNA (tRNA: transfer RNA, rRNA: ribosomal RNA, control: in the absence of RNA); (c, d, e) DBD with different concentrations of salt, hexanediol and poly U. ARE, androgen response element; DBD, DNA‐binding domain; poly U: polyuridylic acid, poly A: polyadenylic acid, poly G: polyguanylic acid

To see if other types of RNA can also productively interact with DBD, we tested four other RNA constructs: poly A, poly G, rRNA and tRNA. With DBD concentrations up to 50 μM, OD600 increase indicated LLPS with poly A and poly G (to an extent similar to poly U), but not with rRNA and tRNA (Figure 4(b), Figure S2). These results suggest that only unstructured RNA can co‐scaffold LLPS with DBD. In addition, based on the high positive charge of DBD and high negative charge of RNA, we assumed that the LLPS observed is primarily driven by electrostatic interactions. To verify this, turbidity measurements were carried out in presence of 0–1000 mM NaCl. To assess the possible contribution of hydrophobic interactions, we also tested the effect of 0%–25% 1,6‐hexanediol. Our results agree with electrostatics, as NaCl above 150 mM disrupted condensate formation (Figure 4(c)), whereas hexanediol up to 25% concentration had no effect (Figure 4(d)). We also tested the RNA concentration dependence of LLPS with poly U (Figure 4(e)). Interestingly, turbidity increases to a poly U concentration up to 0.5 μg/μl, after which it shows a rapid decline, indicative of a “reentrant” behavior in promoting LLPS.

2.4. DNA and RNA colocalize in phase‐separated droplets

An intriguing question about the RNA/DBD biomolecular condensate formation was centered around the colocalization of DNA and RNA, as the original observation of AR LLPS demonstrated chromatin localization, ¹⁴ primarily compatible with LLPS centered around interaction with DNA. As we observed RNA‐driven phase separation, we next examined whether DNA and RNA can compete for the same binding site, and maybe form exclusive condensates (i.e., separate DNA‐DBD and RNA‐DBD droplets). First, we used Dylight® 488‐labeled DBD and Cy5‐labeled DNA (androgen response element [ARE, corresponding to ARE1], for details, see Methods Section 4) and found that the poly U‐induced phase separation resulted in droplets containing both DBD and DNA (Figure 5(a)). Higher concentrations of DBD resulted in more droplets in the given volume (Figure S5). Microscopy on droplet formation with Cy5‐labeled poly U confirmed that RNA is also part of the condensates formed (Figure 5(a), Figure S5). Droplets were found to fuse and grow in size (as a consequence of fusion), progressively reaching the size of about five times that of the original by 1 h after the induction of LLPS, which also agree with the results of DLS measurements (Figure 3(b)). Very similar trends were observed for the condensates formed by F5‐DBD and poly U RNA (Figure S3A, C). However, when the DBD‐ARE complex reached 1:1 ratio, LLPS was abolished (Figure 5(b), (e)). In order to test whether the saturation of DBD with ARE affected only the LLPS or the binding of RNA as well, the interaction between the DBD‐ARE (1:1) complex and poly U was measured. Two different techniques (MST and BLI) concluded the same results, that is, that upon complete DNA binding both the LLPS and the binding of poly U were largely inhibited (Figure 5(c), (d)).

FIGURE 5.

Colocalization of DNA‐binding domain (DBD) and androgen response element (ARE) DNA. Dylight® 488‐labeled DBD (green) mixed with DBD for final concentration of (a) 25 μM, then Cy5‐labeled (red) ARE‐1 was added at 2.5 μM in buffer (50 mM phosphate, 50 mM NaCl, 0.5 mM TCEP, pH 7.0). (b) 12.5 μM of ARE‐1 was mixed with DBD (2.5 μM of Cy5‐labeled ARE‐1 with 10 μM of unlabeled ARE‐1). Phase separation was induced by 0.05 μg/μl poly U. Scale bar is 10 μm. Microscopy snapshots were taken ~ 1 min after polyuridylic acid (poly U)‐induced liquid–liquid phase separation (LLPS). (c) Competition assay for DBD between DNA and RNA measured by MST. Cy5‐labeled poly U was used as fluorophore at a stable concentration and the binding to DBD as well as to DBD‐ARE‐1 and DBD‐ARE‐2 complex (1:1) was measured in buffer (50 mM Tris, 150 mM NaCl, 0.05% Tween 20, pH 7.5. (d) Competition assay for DBD between DNA and RNA measured by biolayer interferometry (BLI). RNA mimic poly U was immobilized on sensors and dipped in 10 μM of DBD or DBD‐DNA complex (1:1) in buffer (50 mM Tris, 150 mM NaCl, 0.05% Tween 20, pH 7.5. (e) OD 600 for DBD‐poly U with different concentrations of ARE‐1 and ARE‐2 in buffer (50 mM phosphate, 50 mM NaCl, 0.5 mM TCEP, pH 7.0)

3. DISCUSSION

AR is a transcription factor that plays important regulatory roles in cell physiology, and is involved in a causative manner in PC. A particular feature of AR is that in the late—castration resistant—phase of PC, shorter splice variants of AR form, which lack the ligand‐binding domain and are thus not subject to androgen (testosterone) regulation. These forms, such as AR‐v7, are primarily responsible for the insensitivity of PC for therapies relying on androgen deprivation. In accord, AR has been in the focus of active research addressing its function and possible targeting in PC. ¹⁶

Recently, a new angle in AR research has taken shape, by the recognition that AR forms transcriptionally active foci in cells, possibly by a mechanism of LLPS in a Med1‐ ¹⁴ or SPOP‐ ¹⁵ dependent manner. LLPS of AR fits into the general trend of the formation of super enhancers by transcription factors around transcriptionally active sites in the chromatin, and may provide novel mechanistic insight leading to successful targeting of AR in PC. The analysis of the molecular determinants of LLPS have led to somewhat contradicting results, as in Reference 14 it was found that only full‐length AR (none of its regions) can phase separate on its own. In Reference 15, on the other hand, it was found that the NTD carries multiple putative SPOP‐binding regions, and it can undergo SPOP‐dependent LLPS. Our results detail these results further.

First, we verified that the DNA binding domain of AR is also capable of RNA binding, which is not unusual as DNA binding regions of proteins frequently contribute to RNA binding. ⁴¹ More intriguing, however, is that the transcription factor AR, which is supposed to phase separate in super enhancers around particular DNA loci, undergoes LLPS in an RNA‐dependent manner. Interestingly, disordered RNA mimics (poly U, poly A, and poly G) promoted LLPS, whereas ordered RNA (tRNA or rRNA) or DNA (AREs) did not. As DNA is localized into RNA‐driven droplets, but does not drive LLPS of AR on its own, we suggest that mRNA products of AR‐driven transcription may promote local LLPS of AR in a reentrant manner. As transcription activity appears to depend on the ability of AR to form active foci, ¹⁴ the promotion of LLPS by the mRNA produced may represent a new element of transcription activation. In this regard, our observation fits perfectly with the RNA‐mediated feedback model of transcriptional regulation recently proposed. ⁴³ In this concept, modeling of the potential roles of RNA in controlling transcription suggested a nonequilibrium feedback control mechanism, in which low levels of RNA promote condensates formed by electrostatic interactions, while higher levels promote the dissolution of condensates, thereby downregulating transcription. This model is also consistent with our observation of a primarily electrostatic interaction between RNA and our AR constructs, and also fits with the general reentrant nature of RNA‐dependent LLPS often observed in other cases. ⁴⁴

Another interesting aspect of our results is the region of AR driving RNA‐dependent LLPS. In an in vivo study ¹⁴ it was found that only full‐length AR can form Med1‐dependent foci in cells, not even the two halves of the molecule (NTD and the rest) can do it. In the other relevant study, ¹⁵ it was found that the NTD with multiple potential SPOP‐interacting motifs is sufficient to drive LLPS of AR. Our observation that the primary LLPS driver within AR is the DBD, is not necessarily in contradiction with these results. The NTD has been shown to specifically interact with the ligand‐binding domain of AR, that is, there is a long‐range regulatory communication between the two ends of AR. ¹⁸ This interaction has to be relieved for transcription activation of AR. If this interaction competes with LLPS, phase separation of AR provides the mechanism of activation of AR not only by localizing the transcription factor on its target sites, but also by relieving the inhibitory interaction between the two ends of the protein molecule. By the very nature of transcription activity, it has to increase the accessibility of DBD, which, if responsible for driving LLPS, can further increase LLPS in a positive feedback manner. LLPS has then to occur as a result of a cooperative interplay between different regions—maybe NTD, DBD and LBD all contributing, which would explain the diminished LLPS tendency of AR‐v7 that lacks LBD. As the unfolding mechanism can hold the key to AR misregulation in cancer, further studies into the mechanistic details of AR LLPS are imminent in the near future.

4. MATERIALS AND METHODS

4.1. Constructs and proteins

The coding gene of AR‐v7 (UniProt P10275‐3) has been ordered from MWG Eurofins in a carrier pET15b plasmid, optimized for E. coli expression. This plasmid has been used as a template to create all AR variants by mutagenesis: AR‐v7 (Uniprot: P10275‐3 aa. M1‐C649), AF1 (aa. V144‐P488), DBD with the C‐terminus of AR‐v7, referred to as DBD (aa. C561‐C649), AF1‐DBD (aa. V144‐C649), Fragment 5 (aa. G473‐T560) and F5‐DBD (aa. G473‐649C). Due to their long disordered regions, which prone to proteolytic degradation AR‐v7, AF1 and DBD‐AF1 were subcloned into a plasmid created in the lab, containing TEV cleavable N‐terminal GST‐tag as well as C‐terminal His‐tag.

4.2. Nucleic acids

Polyuridylic acid potassium salt (poly U), polyguanylic acid (poly G) and polyadenylic acid (poly A) have been ordered from Sigma Aldrich, while ribosomal RNA from bioWORLD and U30 RNA from Eurofins. Yeast tRNA has been ordered from Thermo Fisher Scientific.

Both labeled (Cy5, Biotin) and unlabeled DNA oligos were ordered from Eurofins genomics. Labeled oligos were ordered using HPLC purification. Double stranded ARE sequences (see below for details) were reconstituted in water by annealing using thermocyclers.

Biotinylated tRNA and poly U were obtained by performing labeling using T4 RNA ligase enzyme (Thermo Fisher Scientific) and biotinylated Cytidine (Bis)phosphate (Thermo Fisher Scientific). Cy5‐labeled poly U and U30 were also labeled using T4 RNA Ligase and Cy5‐conjugated Cytidine (Bis)phosphate (Jena Bioscience).

4.3. C3(1)‐ARE (“ARE‐1”)

The C3(1) gene codes for a component of the prostatic binding protein and contains an ARE in the first intron. Due to it being bound by glucocorticoid receptor and AR with a similar affinity it is considered to be a nonspecific ARE. ⁴⁵

>C3(1)‐ARE + flanking (10).

AGCTTACATAGTACGTGATGTTCTCAAGGTCGA.

4.4. ARE‐containing DNA fragments in PSA's promoter (“ARE‐2”)

An ARE in the promoter of the prostate specific antigen (PSA, gene name: KLK3) has been described by directed low‐throughput studies: AGAACAGCAAGTGCT. ⁴⁶ , ⁴⁷ This matches the more recent refinement study on the ARE consensus sequence motif AGAACANNNTGTTCT ⁴⁸ by just two nucleotide differences AGAACAGCAaGTgCT.

PSA's promoter sequence (AGAACAGCAp) was found 178 nucleotides upstream from the KLK3 gene in NCBI Nuccore. The KLK3 gene sequence was retrieved with the promoter region included (−800 bases) from NCBI Nuccore (NC_000019.10:50854115‐50860764), and the flanking positions (+/− 50 bases) and (+/− 5 bases) of the ARE were retrieved:

>PSA‐ARE + flanking (50).

GTCTCCATGAGCTACAGGGCCTGGTGCATCCAGGGTGATCTAGTAATTGCAGAACAGCAAGTGCTAGCTCTCCCTCCCCTTCCACAGCTCTGGGTGTGGGAGGGGGTTGTCCAGC.

>PSA‐ARE + flanking (5).

ATTGCAGAACAGCAAGTGCTAGCTC.

4.5. Protein expression and purification

Fragment 5 (F5), DBD, and F5‐DBD were expressed in BL21 Star (DE3)™ (Thermo Fisher Scientific) with His‐Smt3 tag on N‐terminus by using pHYRSF53 plasmid (Addgene construct: https://www.addgene.org/64696/) with kanamycin resistance gene. Expression was induced at OD 0.6 with 1 mM isopropyl β‐d‐1‐thiogalactopyranoside (IPTG) at 28°C for 4 h. Afterward, cells were collected, and pellet was resuspended in lysis buffers (100 mM Tris, 500 mM NaCl, 5% glycerol, 1 mM benzamidine, 0.5 mM TCEP, pH 8.0). Cells were lysed by sonication (on a Sonica VCX‐70 vibra cell) for 10 min (5 s pulse on, 5 s pulse off, 60% amplification) after sonication DNase I (5 μg/ml) was added with 5 mM MgCl2, cell debris were removed by centrifugation at ×40,000 g for an hour.

Supernatant was filtered through 0.45 μm filter and loaded on a 5 ml nickel‐charged IMAC column (HisTrap HP ™ GE healthcare). After loading, the column was washed with a buffer (50 mM Tris, 1 M NaCl, 0.5 mM TCEP, pH 8.0) and sample was eluted with 500 mM imidazole. Afterward, sample was buffer exchanged to 50 mM Tris, 150 mM NaCl, pH 8.0 with column (HiTrap 26/10 Desalting column‐GE Healthcare). To cleave the smt3 tag, the sample was incubated with ULP1 protease (purified in our lab) in 1:100 ratio in a cooling cabinet at 4°C for 1 h, followed by reverse IMAC to collect protein in flow through and smt3 in elution.

After concentrating the flow‐through, it was loaded on a gel‐filtration column (S75 16/100 GE Healthcare) equilibrated with 50 mM Tris‐base, 150 mM NaCl, 0.5 mM TCEP, pH 8.0. a pure sample was collected, concentrated, flash frozen and stored at − 80°C.

AR‐v7, AF1, and AF1‐DBD were expressed in BL21 Star (DE3) ™ using a plasmid created in our lab, containing TEV cleavable N‐terminal GST‐ and C‐terminal His‐tag with ampicillin/carbenicillin selection marker. The proteins were expressed in Terrific Broth media. The expression was induced at 16°C and OD 1.2 with 0.5 mM IPTG for overnight. After collecting the cells they were resuspended in the lysis buffer (100 mM Tris, 500 mM NaCl, 10 mM IPTG, 10% glycerol, 1 mM TCEP, pH 8.0) supplemented with 1 mM benzamidine, 1 mM PMSF as well as cOmplete ULTRA EDTA‐free Protease Inhibitor cocktail (Roche, 1 tablet/50 ml). The entire purification process was performed on ice or using a cooling cabinet.

Cells were lysed by sonication (on a Sonica VCX‐70 vibra cell) for 10 min (5 s pulse on, 5 s pulse off, 60% amplification) after sonication DNase I (5 μg/ml) was added along with 5 mM MgCl2 and the cell debris were removed by centrifugation at ×40,000 g for an hour.

The supernatant was filtered through a 0.45 μm filter and loaded on a 5 ml HisTrap HP™ column (GE healthcare). After loading, a washing buffer was applied (50 mM Tris, 500 mM NaCl, 500 mM KCl, 10 mM ATP, 20 mM MgCl2, 1 mM TCEP, pH 8.0), then the elution was performed by the elution buffer (50 mM Tris, 300 mM NaCl, 1 M Imidazole, 1 mM TCEP, pH 7.0). After collecting the sample, the pH was adjusted to 7.3 and it was applied to a 5 ml GSTrap FF ™ (GE healthcare) followed by a washing step (50 mM Tris, 300 mM NaCl, 1 mM TCEP, pH 7.3), then eluted by the elution buffer (50 mM Tris, 10 mM reduced Glutathione, pH 8.0). Subsequently TEV protease containing a His‐tag (purified in our lab) was added in a 1:100 ratio and the sample was dialyzed (SERVA, MWCO 12,000–14,000) against 50 mM Tris, 300 mM NaCl, 10 mM IPTG, 10% glycerol, 1 mM TCEP, pH 7.3 overnight at 4°C. Then a reverse IMAC then a reverse GST purification was performed. The pure flowthrough was collected, concentrated, flash frozen and stored at − 80°C.

4.6. Phase separation

Phase transition of protein sample (AR‐v7, AF1, AF1‐DBD, DBD, F5‐DBD, Fragment 5) at different concentrations was induced by polyuridylic acid (poly U) at 0.05 μg/μl in 50 mM phosphate, 50 mM NaCl, 0.5 mM TCEP, pH 7.0. Protein without RNA was used as control.

Further experiments were carried out at 25 μM of protein with 0.05 μg/μl of poly U. Phase transition of DBD was studied in presence of different concentration of polyadenylic acid (poly A) sodium chloride, 1,6‐hexanediol and 0.05 μg/μl of ribosomal RNA (rRNA), transfer RNA (tRNA), polyadenylic acid (poly A), and polyguanylic acid (poly G).

4.7. Turbidity assay

Turbidity of 200 μl sample was measured in 96 well black transparent bottom plates (Greiner bio‐one, chimney well, μclear®). After mixing the samples, the plate was covered with transparent film VIEW seal (Greiner). Absorbance of the sample was measured at 600 nm for 12 h on a BioTek SynergyTM Mx plate reader at 25°C. Each experiment was performed at least in the form of duplicates.

4.8. Dynamic light scattering

DLS measurements were performed on a DynaPro NanoStar (Wyatt) instrument. A disposable cuvette (WYATT technology) was filled with 100 μl of 25 μM DBD or F5‐DBD with or without 0.05 μg/μl of poly U. Sides of the cuvette were filled with buffer 50 mM phosphate, 50 mM NaCl, 0.5 mM TCEP at pH 7.0. Intensity of scattered light was recorded at 25°C on 95°C for 2 h, collecting 10 acquisitions (8 s each). Measurements were repeated at least three times.

4.9. Microscopy

Phase separation of DBD and F5‐DBD was monitored by fluorescence microscopy, protein was fluorescently labeled on lysine residues using Alexa fluorescent Dylight® 488 dye (Thermo Scientific) according to manufacturer's protocol. Dylight® 488‐labeled protein was mixed with 200 times excess unlabeled protein to final concentration indicated in legend. Small amount of Cy5‐labeled poly U or Cy5‐labeled androgen response element (ARE‐1) was added to the sample and phase separation was induced by adding 0.05 μg/μl of poly U.

Microscopy was performed on a Leica DMi8 microscope equipped with a Leica DFC7000 GT camera. Samples of 10–20 μl were loaded onto glass slides and covered with coverslip and imaged immediately or after indicated time. Droplets were 100 × oil‐immersion objectives.

4.10. Microscale thermophoresis

The Monolith™ NT.115 instrument (NanoTemper) was used to study interaction between Cy5‐labeled U30 and poly U and DBD. Experiment was performed in a buffer (50 mM Tris, 150 mM NaCl, pH 7.5, 0.05% Tween 20) using Monolith™ NT.115 MST Standard coated capillaries (NanoTemper).

In the case of U30, an experiment 300 nM of labeled U30 was titrated with increasing concentration of DBD. Experiments were carried out at 40% MST power and 50% LED power. Dissociation constant was obtained by using MO. Affinity Analysis software to fit triplicates.

In the case of poly U, due to the heterogeneity of the sample the concentration could not be determined. Therefore, an optimal signal strength was adapted to 50% LED power and the MST power was 50%. The measurements were carried out using fixed concentration of Cy5‐labeled poly U titrated with DBD or DBD complexed with two different sets of androgen responsive elements (ARE‐1, ARE‐2).

4.11. Biolayer interferometry

BLI was used to study the interaction between immobilized biotinylated (ARE‐1, ARE‐2, poly U, and tRNA) and AR‐v7 DBD. Experiments were carried out in a buffer (50 mM Tris, 150 mM NaCl, 0.05% Tween 20 pH 7.5) on OCTET RED 96.

Biotinylated ligand was loaded onto Streptavidin biosensors (FORTEBIO), followed by baseline, association, and dissociation of DBD at 25°C with a shaking speed of 1000 rpm.

Baseline signal was obtained in a buffer for 60 s, followed by loading of biotinylated ligand for 400 s. Afterward, baseline signal was obtained for 120 s followed by DBD association and dissociation for 500 and 900 s, respectively. Each experiment was performed at least in duplicates. Dissociation constant (Kd) was estimated by fitting response (nm) with DBDand ARE‐1 at (2:1) ratio by Octet data Analysis Software 9.0. Final graph was generated by GraphPad Prism.

Competition experiments were performed by immobilization of poly U on sensors and measurements of the interaction were carried out by dipping sensors into DBD and DBD–ARE‐1 (1:1).

4.12. Bioinformatics and statistical analysis

4.12.1. Similarity of AR to other phase‐separating transcription factors

Homology search for in vitro phase separating proteins was performed in LLPSDB's ⁴⁹ BLAST‐based sequence similarity search tool using the canonical full‐length AR‐v7 (UniProt: P10275‐3) sequence. The search returned with two hits for experiments including AR (AR & AR + SPOP), ¹⁵ and six other experiments with two LLPS drivers (ESR1, RARA) described in the same publication. ⁸ Both hits had highly significant BLAST e‐values, 3.63e‐23 and 9.24e‐16, respectively.

Homology search for in vivo phase separating proteins was performed in DrLLPS' ⁵⁰ BLAST‐based sequence similarity search tool using the canonical full‐length AR‐v7 (P10275‐3) sequence. The query resulted in six hits (ESR1, RARA, THRA, PPARA, NR3C1, NR5A1) with e‐value <1e‐9 (Table 1). In case of orthologous hits only the human protein was considered.

For ESR1, additional supporting evidence was added from PhaSePro. ⁵¹

4.12.2. Charge distribution and prediction of LLPS and RNA‐binding propensities for AR‐v7

PScore ²⁵ (http://abragam.med.utoronto.ca/~JFKlab/Software/psp.htm) and catGRANULE ²⁶ (http://service.tartaglialab.com/new_submission) profiles for LLPS propensities were generated using their webserver interfaces. Thresholds to define LLPS‐prone regions were chosen as PScore <4.00 and catGRANULE score <0.00, as suggested by the default options in the webserver interface.

Charge distribution for AR‐v7 was plotted at pH 7.4 smoothened with a 20‐aa sliding window using VOLPES (https://homepage.univie.ac.at/lukas.bartonek/testserver/).

PPRINT ⁴² RNA‐binding site prediction tool was used via its web‐interface (http://osddlinux.osdd.net/raghava/pprint/) to predict putative RNA‐binding regions for AR‐v7. Propensities were calculated using the default SVM threshold of − 0.2.

AUTHOR CONTRIBUTIONS

Junaid Ahmed: Conceptualization; investigation. Attila Meszaros: Conceptualization; investigation. Tamas Lazar: Conceptualization; formal analysis. Peter Tompa: Conceptualization; project administration.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information

Supplementary Figure S1 Binding experiment of DBD to polyU measured by MST. Cy5‐labeled polyU was used as fluorophore at a stable concentration and the binding to DBD was measured in buffer (50 mM Tris, 150 mM NaCl, 0.05% Tween 20, pH 7.5).

Supplementary Figure S2 Turbidity of different concentrations of DBDin presence (filled circle) and absence (filled square) of 0.05 μg/μl (A) polyU, (B) polyA, (C) polyG, and (D) tRNA.

Supplementary Figure S3 AR‐v7's F5‐DBD phase separates in presence of polyU.

Supplementary Figure S4 Turbidity measurements of 25 μM AF1 in presence of different types of RNA and 10% dextran.

Supplementary Figure S5 Colocalization of DBD and androgen response element (ARE) DNA.

Ahmed J, Meszaros A, Lazar T, Tompa P. DNA‐binding domain as the minimal region driving RNA‐dependent liquid–liquid phase separation of androgen receptor. Protein Science. 2021;30:1380–1392. 10.1002/pro.4100