Endocrine Examples of Phase Separation in Biology.

An estimated 10% to 20% of Food and Drug Administration-approved drugs modulate the NRs ( 140 ), with SHRs serving as major targets for cancers ( 77 , 141 ), inflammatory disorders ( 142 ), and cardiac pathologies ( 143 ). However, drug resistance and side effects remain longstanding challenges ( 144 - 146 ). Emerging evidence suggests condensate formation regulates SHR-mediated gene transcription. Therefore, understanding drug partitioning into these condensates paves the way not only for understanding and improving existing treatments but also for developing new therapeutic strategies to overcome resistance. Tamoxifen is widely used for ER-positive breast cancer to inhibit ER-mediated transcriptional activity ( 147 ). A recent study found that tamoxifen partitioned into MED1 condensates and dislodged ERα, enhancing its antiestrogen potency ( 77 ). This condensate partitioning also helps explain resistance associated with ERα mutations or MED1 overexpression in breast cancer ( 77 ). These alterations reduce tamoxifen concentration within condensates, diminishing its ability to displace ERα ( 77 ). Enhancing drug partitioning into specific condensates could open new avenues for more effective chemotherapeutics. Instead of directly targeting ERα, another emerging approach is to target Eleanor clouds, which have only been found in hormone therapy-resistant breast cancer cells ( 88 - 90 ). Both antisense oligonucleotides and the small molecule resveratrol ( Fig. 3B ) decreased Eleanor cloud formation, disrupted long-range interactions with the FOXO3 gene, and triggered FOXO3-mediated apoptosis ( 88 , 89 ). Interestingly, glyceollin I ( Fig. 3B ) reduced Eleanor clouds even in the absence of ERα and induced greater apoptosis in endocrine–resistant breast cancer cells while having less effect on normal fibroblast cells than ERα-dependent resveratrol ( 90 ). This suggests targeting Eleanor clouds may yield treatments with increased effectiveness and selectivity for hormone therapy-resistant cancer cells while reducing side effects. Drugs such as bicalutamide and enzalutamide are commonly used in the treatment of advanced prostate cancer. They bind to the AR LBD as antagonists, reduce condensate formation, and decrease AR-mediated transcription ( 99 , 141 ). However, in castration-resistant prostate cancer (CRPC), most patients express AR variations, which confer resistance to these drugs. For instance, an AR construct with a W742C mutation (AR W742C) alters the ligand binding pocket, converting bicalutamide to an agonist, thereby enhancing AR condensate formation and promoting gene transcription ( 146 , 148 - 150 ). The ARv7 isoform, which lacks the LBD and is constitutively active, forms condensates and drives gene transcription independently of ligand, rendering enzalutamide largely ineffective ( 150 - 152 ). Given these challenges, inhibiting phase separation has emerged as an alternative strategy. For example, an amphiphilic molecule EPI-001 disrupts transient helices in the AR NTD ( 153 ), and its derivative, ET-516, impairs the foci formation of ARv7 and AR W742C, decreases AR-mediated transcriptional activity, and reduces tumor size in mouse models ( 99 ). Another derivative, EPI-7386, is safe and well-tolerated in a phase 1b study for metastatic CRPC ( Fig. 3B ) ( 154 ). Additionally, covalent modifier UT-143 targeting Cys residues on the AR NTD has also shown promise in a preclinical study ( Fig. 3B ). It disrupts ARv7 foci formation, eliminating ARv7 FRAP recovery and reducing AR target gene transcription and tumor growth in mice ( 101 ). Further development of these LLPS-targeting compounds could advance treatments for CRPC.

Nuclear receptors (NRs) are pivotal in coordinating responses to endocrine cues by regulating gene expression programs for development, metabolism, and homeostasis. NRs are a super-family of TFs activated by a broad spectrum of signals, including endogenous hormones ( 60 ) and metabolic ligands ( 61 62 ). The steroid hormone receptors (SHRs) constitute an important NR subfamily and include the estrogen receptor (ER androgen receptor (AR), progesterone receptor (PR), glucocorticoid receptor (GR), and mineralocorticoid receptor (MR). Upon binding signaling ligands, SHRs enter the nucleus to bind specific response elements, recruit transcription coregulators, and drive context-specific gene activation or repression ( 63 ). NRs/SHRs are multidomain proteins containing 2 conserved, well-folded domains: a DNA-binding domain (DBD) that recognizes response elements and a C-terminal ligand binding domain (LBD) that binds to signaling molecules ( Fig. 3A ). They also have 2 less conserved regions: a disordered N-terminal domain (NTD) and an unstructured hinge region. The NTD contains the activation function-1, essential for full transcriptional activity. The activation factor −2 in the LBD interacts with conserved motifs in coregulators: coactivators possess LxxLL (L: leucine, x: any amino acid) motifs, while corepressors contain Lxxx(I/L)xxx(I/L) (I: isoleucine) motifs ( 64 ). Coregulators can be histone modification enzymes that regulate chromatin compactness or scaffolding proteins that mediate interactions with the general transcription machinery. Ligand-bound SHRs do not localize uniformly in the nucleus but instead form up to 4500 small foci. Groundbreaking studies suggest these foci are phase-separated transcriptional condensates, many of which are SEs or other enhancer clusters. For instance, diffusion coefficient (D) values derived from abundant SPT data match expected LLPS behavior ( 29 , 65 - 67 ). All SHRs experience 3 major diffusive environments: a lowest-mobility state (D 3 mm 2 /s) in the surrounding nucleoplasm ( 68 - 70 ). The 2 low-mobility states match trajectories of histone proteins, which also phase separate, and the second low-mobility state relies on the presence of IDRs ( 67 ). Within condensates, SHRs encounter boundaries resulting in either altered direction or change in D ( 65 - 67 ). Tracking several SHR particles simultaneously revealed fusion and growth to a threshold size ( 65 ). Collectively, these properties align with the LLPS behavior of SHRs. SHR condensates at gene loci spatially organize different interactions by enriching relevant coregulators while excluding similar but functionally distinct ones. Different SHR domains play distinct roles: the DBD directs genomic localization, while the NTD and LBD drive phase separation through interactions with other cofactors. The LBD mediates specific binding to the LxxLL motif of coregulators, whereas the NTD shapes condensate properties, including hydrophobic interactions and transitions between LLPS and LGPS. For example, an expansion of a specific region in AR NTD facilitates LLPS-to-LGPS transitions, which contributes to transcriptional dysregulation and is a key pathogenic mechanism of spinal and bulbar muscular atrophy ( 29 ). Here we focus on SHRs as prototypical NRs to understand the roles of phase separation in endocrinology. We review evidence of how multivalent interactions, LLPS, and LGPS mediate SHR function and contribute to SHR-related pathologies. Finally, we introduce how therapeutics targeting these transcriptional condensates provide new options for hormone therapy-resistant cancers. Both ERs, ERα and ERβ, recognize the primary endogenous estrogen 17β-estradiol (E2) and activate genes via estrogen response elements ( 63 ). While ERα regulates genes essential for the development and function of female reproductive tissues, its dysregulation is closely linked to hormone-dependent breast and endometrial cancers ( 71 , 72 ). In contrast, ERβ has antiproliferative roles and frequently counteracts ERα by preferentially engaging corepressors ( 71 ). While ERα-ERα, ERβ-ERβ, and ERα-ERβ dimers form foci in the nucleus in the presence of E2 ( 73 , 74 ), ERα homodimer-containing foci have been extensively characterized and are the focus here. Approximately 1000 estrogen-responsive enhancers are exceptionally active based on enhancer RNA transcription levels. These enhancers harbor assembly of ERα-dependent, large (megadalton-scale) protein complexes, known as “MegaTrans” complexes ( 75 , 76 ). These complexes contain clusters of TFs and coactivators, many of which have IDRs, such as GATA3 and MED1 ( 76 ). These proteins form condensates that possess many hallmarks of LLPS ( Fig. 1A ; Table 1 ). Namely, purified ERα forms spherical droplets in vitro that exhibit size scaling ( 76 ). ERα also forms droplets with GATA3 ( 76 ) or MED1 ( 77 ). FRAP analysis showed a t 1/2 of 25 seconds and recovery fraction of >0.9 ( 76 , 78 ). Additionally, 1,6-HD decreased the intensity and number of ERα foci in cells ( 76 ). LLPS at MegaTrans enhancers is functionally significant. In the presence of E2, the MegaTrans complex assembles at estrogen response elements, promotes chromatin unwinding, and upregulates transcription ( 75 , 76 , 79 ). However, both deletion of MED1/GATA3 IDRs and disruption of condensates with 1,6-HD markedly reduce E2-mediated transcription of MegaTrans target genes ( 76 ). Interestingly, MED1 contains 2 LxxLL coactivator motifs that bind the AF-2 site of the ERα-agonist complex ( 80 ). ERα is incorporated into in vitro droplets formed by a MED1 construct with LxxLL motifs, and this incorporation increases in the presence of E2 ( 78 ). These data suggest that multiple interactions govern ERα-MED1 co-phase separation, including a stable interaction at the AF-2 site and IDR-mediated interactions ( Fig. 1B ). LLPS at MegaTrans enhancers can coordinate long-range interactions among distant MegaTrans enhancers and between these MegaTrans enhancers and nuclear speckles ( Fig. 2 ) ( 76 ). Condensates form after E2 stimulation bring MegaTrans enhancers located megabases away into close proximity, and the majority of these condensates make close contact with nuclear speckles ( 76 ). Both types of long-range contacts increase transcriptional efficiency ( 76 ). Disrupting phase separation through 1,6-HD or altering nuclear speckle interactions by knockdown of splicing factors decreases transcription efficiency ( 76 ). These results are consistent with previous findings that ER coregulators such as CAPERα and RBM17 relocate to nuclear speckles and play roles in splicing ( 81 - 84 ). Indeed, analyses of published mass spectrometry and phase separation databases identify many coactivators in nuclear speckles, including DEAD/DEAH-box RNA helicases and scaffold attachment factors, although NRs themselves have not been found in speckles ( 6 , 85 - 87 ). While disrupting the interactions that drive phase separation at MegaTrans enhancers impairs their function, stabilizing these interactions through contacts over an extended time can lead to LGPS traits, which also compromises enhancer function ( 76 ). Prolonged exposure (90 minutes or longer) to E2 causes MegaTrans enhancer condensates to lose their sensitivity to 1,6-HD, exhibit reduced FRAP t 1/2 and recovery fractions, and become less effective at enhancing transcription ( 76 ). These findings suggest that optimal enhancer activity depends on maintaining a dynamic, liquid-like environment supported by short-lived interactions. Another membraneless compartment involved in ER signaling occurs at the locus of the gene ESR1, which codes for ERα it-self. ESR1 locus enhancing and activating noncoding RNAs (Eleanors) are long noncoding RNAs transcribed from a 0.7-Mb region encompassing ESR1 introns and adjacent un-translated regions ( 88 , 89 ). Eleanors form condensates known as “Eleanor clouds” at ESR1 promoter-enhancer loops ( 88 - 90 ). These condensates contribute to estrogen-therapy resistance in cellular studies and mark an increased chance of long-term recurrence in patients with breast cancer ( 88 , 91 ). Data from ATAC-seq, Hi-C, and 4C-seq indicate Eleanor clouds remodel chromatin at the ESR1 promoter and stabilize long-range interactions with other genes such as FOXO3, which plays a key role in apoptosis ( 88 , 89 , 92 ). Chromatin immunoprecipitation sequencing results revealed that Eleanor clouds recruit hypophosphorylated Pol II, leading to increased transcription of ESR1, evasion of FOXO3-mediated apoptosis, and enhanced cancer cell stemness ( 89 , 91 ). While the function and relevance of Eleanor clouds have been well established, the properties, composition, and mechanism of action of Eleanor clouds remain largely elusive. To date, no comprehensive biophysical study has clarified whether Eleanor clouds assemble via LLPS ( 93 ). Proposed mechanisms include RNA-mediated nucleosome destabilization, where RNA-DNA base pairing displaces histones ( 92 ) and protein-RNA condensate formation, where high local Pol II concentration facilitates its repeated promoter loading ( 92 , 93 ). These hypotheses are supported by decreased thermostability of nucleosomes in the presence of Eleanors ( 92 ) and the partitioning of Pol II and possibly TFs in Eleanor clouds ( 89 , 90 ). Characterization of the protein composition and protein–RNA interactions within Eleanor clouds would better elucidate their functions and support their development as therapeutic targets. AR binds testosterone and dihydrotestosterone (DHT) and regulates transcription primarily by binding to androgen response elements. Upon ligand binding and dimerization, AR activates genes involved in prostate development and seminal plasma maintenance, including the KLK family serine proteases such as KLK3 (prostate-specific antigen), which are also dysregulated in prostate cancer ( 94 ). Consequently, AR is a key therapeutic target in hormone-sensitive prostate cancer, and antiandrogens can reduce cancer cell proliferation and induce senescence ( 95 , 96 ). AR exhibits properties consistent with phase separation ( 29 , 97 ). For example, activation of GFP-labeled AR by DHT in LNCaP prostate cancer cells results in nuclear foci formation in most green fluorescent protein-positive cells, and the number of cells with AR foci decreases in the presence of 1,6-HD ( 29 ). Moreover, AR forms spherical droplets that exhibit size scaling in vitro, and FRAP recovery occurred on the order of seconds both in vitro and in vivo ( 29 ). However, FRAP of AR showed modest recovery fractions (~0.55), suggesting it may possess either gel-like properties or tight interactions with binding partners such as DNA or the nuclear matrix ( Table 1 ) ( 29 ). Phase separation of AR is optimal when both the NTD and LBD are present and involves the interactions between these domains, which induce transient α-helical structure formation in some NTD regions ( 97 , 98 ). However, the NTD is sufficient for phase separation both in vitro and in cells ( 29 , 99 ). AR has a particularly long NTD (559 amino acids), enriched in aromatic residues and cysteines and containing several low-complexity regions ( Fig. 3A ), including stretches of poly-Q (23 residues), poly-P (8 residues), and poly-G (23 residues) ( 100 ). Foci formation is virtually abolished in AR constructs lacking the NTD ( 29 ). Phase separation of the AR NTD increases in vitro by higher NaCl concentration or higher temperatures ( 70 , 97 , 101 ), suggesting an entropically driven process involving hydrophobic interactions ( Fig. 1A ) ( 102 - 104 ). Substituting Phe or Tyr residues with Ser reduces in vitro droplet size and the number of AR foci–positive cells, indicating the importance of the NTD aromatic residues in LLPS ( Fig. 1B ) ( 29 , 97 ). Many AR foci function as SEs, with AR and MED1 colocalizing at ~1000 SEs in the VCaP prostate cancer cells ( 105 ). DHT stimulation triggers corecruitment of MED1 and AR to these regions, but AR antagonist enzalutamide leads to a marked decrease in MED1 occupancy ( 105 ). Silencing of MED1 results in depletion of androgen-responsive gene expression ( 105 ). Mounting evidence suggests the roles of AR/MED1 SEs in transcription rely on condensate formation. For instance, both proteins form droplets in vitro individually and with each other ( 20 , 97 ). Treatment with 1,6-HD in cells diminishes MED1 recruitment to AR binding sites at SEs ( 98 ). ATAC-seq and RT-qPCR results indicated that AR condensate formation increased chromatin accessibility and transcription at AR enhancer sites, and these effects were greatly reduced when phase separation was disrupted by NTD deletion or treatment with 1,6-HD ( 29 ). In contrast, poly-Q expansion (from 23 to 69 or 92) in the NTD—characteristic of spinal and bulbar muscular atrophy—increases foci formation and droplet size but reduce AR mobility, FRAP recovery, and transcription, even though AR remains bound to chromatin, suggestive of increasing gel-like properties ( 29 ). This is associated with reduced MED1 recruitment, suggesting that the hardened interactions mediated by increased poly-Q repeats of AR are less functional than the dynamic, liquid-like interactions ( 29 ). In addition to transcription initiation, emerging evidence suggests that AR also contributes to elongation by interacting with coregulators such as P-TEFb subunits ( 106 ), which remain associated with elongating Pol II in nuclear speckles ( 51 , 107 , 108 ). PR has 2 functional isoforms: the shorter, less active PRα and full-length PRβ, which contains an additional 164 amino acids in its NTD. PRβ plays an integral role in development of mammary glands by activating genes such as RANKL, WNT4, and HAND2 ( 109 , 110 ). PRα regulates uterine and ovarian function and can repress PRβ in some circumstances ( 111 , 112 ). Disrupted PRα/PRβ levels are common in cancers ( 111 , 113 , 114 ) and endometriosis ( 115 ). Both PRs localize to foci when endogenous ligand progesterone levels are elevated. PRβ is the dominant isoform in foci, whereas PRα largely remains dispersed in the nucleus, indicating that the intact NTD is essential for PR foci formation in cells ( 116 ). Like other transcriptional condensates, PR foci are typically small and short-lived ( 65 ), but they become en-larged in endometrial cancers or upon chromatin disruption by a histone deacetylase inhibitor ( 117 ). Disruption of PRβ binding to the nuclear matrix impairs its localization to foci and gene transcription ( 118 ). This suggests the small size and localization patterns of PR foci are relevant to its functional role in cells. There are fewer studies on PR condensates compared to other SHRs. However, a recent investigation involving MCF7 cells confirmed that PR foci possess many of the LLPS hallmarks ( 65 ). SPT results indicated PR particles exhibited high- and low-mobility states, similar to other SHRs ( 65 ), and individual particles traveled between condensates ( 65 ). These condensates fused and were dispersed by 1,6-HD ( 65 ). Notably, MD simulations refined against SPT data suggest a mechanism underlying the small size of the PR condensates ( 65 ). While condensates generally associate, fuse, and merge into one larger structure following Brownian motion coalescence ( 65 , 119 ), PR condensate growth follows this model with a key modification: a unique “escape” term, whereby PR molecules exit condensates at a defined rate, causing condensates to grow until a critical size, after which condensate radius plateaus ( 65 ). Topological tension within the chromatin scaffold ( 120 ), along with the redistribution of coregulators into nuclear speckles ( 82 ), may provide mechanisms for this escape. Although this model was applied to a PR system with constant expression levels, it may help explain SHR size scaling patterns with increased expression. For example, increased AR expression leads to increased condensate numbers while condensate size remains small ( 98 ). However, further studies are needed to confirm the physiological relevance of this model. GR is activated by cortisol and regulates genes involved in metabolism, development, immune suppression, and stress response. GR binds to a canonical glucocorticoid response element (GRE) to activate target genes. However, GR is unique among the SHR family as it also recognizes negative GRE and NF-κB response element (κBRE) to drive gene repression, particularly during inflammation ( 121 , 122 ). GR is abundant in metabolically active tissues such as the liver and adipose tissue, where it regulates glucose and lipid metabolism. It is also highly expressed in the immune system, where it mediates the potent anti-inflammatory and immunosuppressive effects of glucocorticoid drugs, such as dexamethasone. These drugs are used to treat a wide range of autoimmune and inflammatory disorders, including rheumatoid arthritis and asthma ( 123 ). GR forms spherical droplets that size-scale in vitro and re-distribute between different droplets, indicating both internal rearrangement and external exchange ( 28 ). Upon treatment with cortisol or dexamethasone, GR enters the nucleus and forms foci, which exhibit roundness, fusion, and susceptibility to 1,7-heptanediol ( 124 , 125 ). FRAP recovery in cells occurs with a t 1/2 on the order of seconds and recovery fraction around 1, suggesting a liquid-like environment ( 68 , 125 ). More and brighter GR foci are formed as the NaCl concentration increases, a process that is reversed upon NaCl washout ( 68 ). Interestingly, a GR construct lacking the NTD forms fewer foci under high salt conditions ( 68 ), suggesting the key role of NTD-mediated hydrophobic interactions in condensate formation. Similar to other SHRs, GR condensates coincide with functionally significant long-range chromatin interactions and up-regulation of gene transcription ( 126 - 128 ). GR activation by dexamethasone leads to clustering of GR and enhancers to confined chromatin regions where diffusion coefficients are decreased in a manner consistent with LLPS ( Fig. 1A ) ( 67 , 125 , 127 ). Contacts between GR and the cohesin loader protein NIPBL enable long-range chromatin interactions and enhance gene transcription ( 127 ). Loss of the cohesin complex causes both enhancers and GR to disperse based on SPT results ( 127 ). Furthermore, cohesin mutations found in acute myelogenous leukemia patients with dexamethasone insensitivity cause reduced GR-mediated transcription in response to dexamethasone ( 127 ). This suggests GR condensates are key sites for functionally significant, cohesion-mediated long-range interactions and transcription. In addition to NIPBL, several coregulators, including MED1 ( 125 , 129 ), BRD4 ( 129 , 130 ), the histone methyltransferase G9a ( 28 ), and members of the NCoA family ( 28 , 131 ), are recruited into GR condensates. Multivalent interactions between IDRs of GR and these coregulators are crucial for recruitment, and stable LxxLL–AF-2 interactions further enhanced enrichment of NCoA3 into GR condensates ( 28 ). This indicates both IDR-mediated, multivalent interactions and stable, specific interactions at the AF-2 site contribute to the coregulator enrichment in GR condensates. Different DNA sequence motifs, including canonical GRE, negative GRE, and κBRE, modulate condensate formation, but these GR-DNA complexes are not miscible with each other, suggesting GR forms physically separated activating and repressive condensates ( 28 ). These condensates can recruit different coregulators, consistent with transcriptional output ( 28 ). For example, MED1 was recruited to condensates with the canonical GRE for activation, whereas G9a was enriched into repressive GR condensates to facilitate gene repression ( 28 , 132 ). This indicates that coactivators or corepressors can be selectively recruited to distinct condensates in a GRE-dependent manner to fine-tune the up- and downregulation of GR target genes. Condensate formation provides an additional layer to GR-mediated gene transcription. MR is abundantly expressed in epithelial tissues such as the kidney, where it regulates sodium and potassium homeostasis ( 133 ). It is also expressed in the heart and vascular endothelium, where it regulates blood pressure and cardiac remodeling ( 133 ). MR mediates the actions of both aldosterone and cortisol and activates target genes such as epithelial sodium channel subunits and serum and glucocorticoid-regulated kinase ( 134 ). Like GR, MR can bind canonical GREs, despite producing different downstream phenotypes ( 135 ). An early study on the subnuclear distribution of endogenous MR and GR in rat hippocampal CA1 neurons found that upon cortisol stimulation, both GR and MR localized to ~1000 discrete foci in the nucleus ( 136 ). While some foci exhibited colocalization of both receptors, the majority (>80%) contained only GR or MR ( 136 ). This suggests GR and MR are targeted to specific nuclear compartments, potentially facilitating coordinated regulation of gene expression in hippocampal neurons. Subsequent studies in CV1 cells found that overexpressed MR rapidly entered the nucleus upon aldosterone stimulation, forming up to 4500 foci ( 137 , 138 ). These foci associate with regions of euchromatin and the nuclear matrix, and they are excluded from nucleoli ( 137 ). Removal of the NTD alters the localization pattern of MR, resulting in perinucleolar enrichment and partial nucleolar localization ( 138 ). Importantly, the foci disperse following an MR antagonist treatment, consistent with the reversible nature of LLPS. In line with this, FRAP in vitro and in vivo revealed a t ½ on the order of seconds ( 68 , 139 ). A recent study showed that MR formed spherical droplets that fused and exhibited size scaling ( 139 ). These condensates enhance MR target gene transcription, particularly at cold temperatures during transplant organ cryopreservation, where excessive MR activation causes oxidative tissue damage ( 139 ). In contrast, inhibiting MR phase separation with canrenone mitigates this damage and improves organ viability ( 139 ). These findings support a model in which MR forms dynamic, liquid-like condensates that promote target gene transcription. Given the high sequence similarity between DBD and LBD in MR and GR, their distinct abilities to modulate LLPS likely stem from differences in their NTDs, which share less than 15% homology and recruit distinct coactivators and corepressors ( 135 ). Understanding MR LLPS could reveal a novel layer of regulatory control over shared MR/GR target genes, presenting an intriguing avenue for future investigation that could inform strategies to selectively target MR in disease.

Clear evidence distinguishing functional phase separation from aggregation or nonspecific clustering remains a central topic in the field. Traditional methods, such as FRAP, 1,6-HD, and fluorescence microscopy, are widely used to assess the diffusion, reversibility, circularity, and fusion properties of the condensates and provide reliable criteria to differentiate reversible phase separation from irreversible aggregation. However, these methods have limitations in distinguishing different types of phase separation or separating LLPS from simple binding events, particularly when they are applied to study transcriptional condensates. Skepticism of the model that transcriptional condensates form via LLPS stems from transient interactions between Pol II and MED 1 and the generally small, short-lived nature of transcriptional condensates ( 16 , 47 ). The short lifespans and sizes near the diffraction limit complicate in vivo FRAP measurements, make the observation of roundness potentially misleading, and elicit questions about size scaling ( 16 ). Complementing cellular studies with in vitro assays, where condensates are larger, helps to address some concerns but raises new questions about the physiological relevance of isolated systems, particularly when using nonphysiologically relevant protein concentrations and crowding agents ( 16 ). SPT overcomes many obstacles to distinguishing condensates formed by LLPS from clusters of multivalent interactions that fail to mature into a distinct phase ( 15 , 16 ). SPT was used to rule out phase separation in replication compartments, where Pol II showed no change in diffusion after formation of the compartment, instead continuing to exhibit only 2 major states: a free and DNA-bound state ( 155 ). Additionally, the relative population of these 2 states was unaffected by the length of the Pol II IDR ( 155 ). Instead, increasing nonspecific DNA-binding sites enhanced the compartment occupancy, supporting a model of multiple low-affinity interactions facilitating target search for the high-affinity binding site ( 155 ). This model, in which protein clusters engage in many nonspecific simple-binding events, represents a major alternative explanation for condensate formation ( Table 1 ) ( 15 , 47 ). SPT results for SHRs differ from those of replication compartments, with features such as an IDR-dependent second low-mobility state supporting expected behavior for LLPS-mediated condensates, as discussed earlier ( 67 ). However, histone H2B from chromatin also displays 2 distinct low-mobility states that overlap with those of SHRs, underscoring the importance of using alternative labeling strategies to rigorously evaluate nonstoichiometric interactions and viscoelastic properties ( 69 , 70 ). Inert probes that do not engage specific interactions with condensate components are valuable controls because their diffusion is affected solely by condensate boundaries ( 15 , 18 ). Such probes have been used to distinguish phase separation from simple binding within heterochromatin compartments ( 156 ). Open questions also remain about the shuttling of proteins from transcriptional condensates to nuclear speckles ( 6 , 7 ). Key uncertainties include how often coregulators move between these compartments and the functional consequences of such trafficking. Additionally, the degree of overlap between nuclear speckles with condensates involved in elongation, such as super elongation condensates, is still unclear ( 47 , 157 , 158 ). Tracking the trajectories of coregulators and splicing factors together with Pol II could help reveal which cofactors accompany Pol II during elongation, identify the condensates involved, and determine how frequently these shuttling events occur. An important direction for future research is to determine how phase separation contributes to nongenomic functions of the SHRs. In addition to their genomic actions in the nucleus, approximately 5% of SHRs also localize to the plasma membrane ( 159 , 160 ). For example, the G-protein-coupled estrogen receptor (GPER) mediates rapid, nongenomic responses to estrogen stimulation ( 161 ). While a definite link between condensate formation and membrane-localized SHRs has yet to be established, nonnuclear foci formation of AR and ER was observed in some cell lines ( 101 , 162 ). Recent evidence also indicates that GPER activation disrupts membraneless organelles known as stress granules ( 163 ). While stress granules mediated resistance to the chemotherapeutics cisplatin and paclitaxel in HeLa cells, and this GPER-mediated granule disruption restored drug sensitivity ( 163 ). Further exploration into the interplay between membrane-bound SHRs and membraneless organelles may provide deeper insights into the role of phase separation in SHR activity, particularly in the context of therapeutic treatments.

Phase separation is reshaping how we understand SHR functions in the endocrine system. Condensate formation at SHR target genes and gene loci (such as Eleanor clouds) brings distant chromatin regions into proximity, facilitates chromatin opening, concentrates Pol II, and promotes transcription initiation. Additionally, the dynamic distribution of SHR coregulators from transcriptional condensates to nuclear speckles enhances the efficiency of transcription elongation and pre-mRNA processing. While this review focuses on SHRs, condensate formation has been observed in other NRs and their coregulators, including thyroid hormone receptor α ( 61 ), peroxisome proliferator-activated receptors ( 164 ), retinoic acid receptors ( 165 ), retinoid X receptors ( 166 ), and the coactivator PGC-1α ( 164 ). Finally, this emerging field is transforming our fundamental understanding of NR function while providing new insights into endocrine dysregulation. These discoveries are laying the groundwork for compelling new therapeutics targeting endocrine therapy–resistant cancers, as well as inflammatory and cardiac disorders, with condensate-modifying compounds establishing new frameworks for clinical breakthroughs. As research on phase separation has expanded, the criteria for defining LLPS have continued to evolve. Studying SHR condensates with a critical yet integrative perspective offers a promising path to uncover new therapeutic strategies.

Phase separation is implicated in multiple aspects of gene regulation, ranging from chromatin remodeling ( 32 , 33 ) and transcription ( 9 , 10 ) to pre-mRNA processing ( 6 , 7 ) and nuclear export ( 34 , 35 ). As these processes often occur simultaneously, the relationship between transcription and LLPS is particularly complex ( 36 - 41 ). Distinct condensates regulate the transcription initiation and elongation stages ( 31 ), while some components transition between condensates as transcription progresses ( 42 ). Transcription is mediated by transcription factors (TFs) binding to DNA sequences known as response elements on promoters or enhancers. Promoters are sites upstream of the gene coding sequence where RNA polymerase II (Pol II) initiates transcription ( 43 , 44 ), while enhancers are regions that can be linearly distal from the gene and regulate initiation by spatially associating with promoters ( 45 ). Transcriptional condensates form at promoter-enhancer interfaces and typically enhance gene transcription significantly ( 46 , 47 ). However, some condensates can repress transcription by sequestering TFs or pausing Pol II ( 48 , 49 ). Super-enhancers (SEs) are transcriptional condensates consisting of large clusters of enhancers that collectively drive elevated transcription ( Fig. 2A ) ( 46 ). They are characterized by increased chromatin accessibility, marked by high levels of H3K27ac and high occupancy of the Mediator complex that facilitates interactions between TFs and hypophosphorylated Pol II during transcription initiation ( 46 ). Evidence suggests this dense clustering of proteins, many of which contain IDRs, results in condensates at SEs ( 9 , 10 ). For example, MED1 and histone-acetylation reader protein BRD4 form round condensates in vitro and in cells ( 10 ). SEs are highly transient, with lifetimes averaging ~11 seconds ( 9 ), which may result from the growing RNA strand destabilizing the condensates ( 50 ). SEs maintain a compact size, which may stem from the release of hyperphosphorylated Pol II during transcription elongation, when it loses affinity for MED1 and exits SEs ( 31 ), carrying with it TFs and coregulators ( 42 ), including BRD4 and the positive transcription elongation factor b complex ( 51 ). Another condensate involved in transcription is the nuclear speckle (ie, interchromatin granule or splicing condensate) ( Fig. 2B ). Nuclear speckles are heterogeneous in size (~0.3-3 mm) and contain ~350 TFs, coregulators, and splicing factors along with RNA ( 52 , 53 ). These components form multilayered condensates ( 54 , 55 ) that coordinate elongation, pre-mRNA splicing, and nuclear export ( 40 , 51 , 56 , 57 ). Hyperphosphorylated, elongating Pol II is enriched at the periphery of nuclear speckles ( 40 , 52 , 56 ), selectively partitioning with splicing factors such as SRSF1 and U1-70K ( 31 , 56 ). These splicing factors are best characterized for their roles in pre-mRNA splicing, but they also enhance transcript ( 58 , 59 ).

Biomolecular liquid–liquid phase separation (LLPS) occurs when proteins/nucleic acids have a higher affinity for each other than other components of the surrounding solution, separating into distinct aqueous phases ( 12 , 13 ). This process creates a membraneless partition that resists osmotic pressure and entropy penalties, providing the structural scaffold for many membraneless organelles ( 12 , 13 ). Condensates formed by LLPS possess liquid-like properties but can transition into a more rigid, gel-like state through liquid–gel phase separation (LGPS) ( Table 1 ) ( 14 ). They are round, scale in size with an increased number of component molecules, and fuse with similar condensates ( Fig. 1A ) ( 14 - 16 ). Additionally, interior biomolecules can move within the condensate (internal rearrangements) or between the condensate and the surrounding environment (external exchange) ( 14 - 16 ). These dynamic motions are best characterized by fluorescence recovery after photobleaching (FRAP) and single particle tracking (SPT) ( 14 - 16 ). A combination of both assays is recommended for reliable characterization of LLPS, as reviewed by extensive literature ( Fig. 1A ) ( 14 - 18 ). FRAP curves reflect net diffusive properties and plateau close to the prebleaching intensity (recovery fraction close to 1) if LLPS is present, indicating most molecules are mobile ( 14 , 17 , 18 ). LLPS is also characterized by a short FRAP recovery half-time (t 1/2 ), with most classic condensates displaying t 1/2 values on the order of seconds to minutes ( 19 - 21 ). SPT follows trajectories of single particles to obtain diffusion coefficient (D) values. Individual particles experience decreased D values while crossing a condensate boundary ( 15 ). Finally, another defining feature of LLPS is reversibility and sensitivity: condensates can form and dissolve in response to changes in the cellular environment, including salt, temperature, and small molecules ( Fig. 1A ) ( 22 , 23 ). For example, phase separation mediated by nonpolar interactions can be disrupted by aliphatic alcohol 1,6-hexanediol (1,6-HD) or 1,7-heptanediol ( 22 , 24 ). These properties define condensates formed by LLPS, distinguishing them from simple binding events, other types of phase separation, and protein aggregation. Protein intrinsically disordered regions (IDRs) lack stable secondary structures and often participate in multiple dynamic, weak interactions simultaneously ( 25 , 26 ). This property is referred to as multivalency and enables the formation of a fluid environment ( 26 ). Folded regions also play a significant role in condensates by providing specific, stable binding sites that enhance selectivity and functionality ( 26 ). These roles are not strictly defined, as IDRs can have regions that form modular interactions, and structured domains can be multivalent ( 27 ). For instance, the IDRs of many endocrine proteins contain both specific, stable interaction sites and multivalent, low-complexity tracts ( Fig. 1B ) ( 28 , 29 ). The forces driving phase separation vary between condensates, directing unique proteins to specific membraneless organelles ( 30 , 31 ).