Transcription factor condensates and signaling driven transcription

ABSTRACT

Transcription Factor (TF) condensates are a heterogenous mix of RNA, DNA, and multiple co-factor proteins capable of modulating the transcriptional response of the cell. The dynamic nature and the spatial location of TF-condensates in the 3D nuclear space is believed to provide a fast response, which is on the same pace as the signaling cascade and yet ever-so-specific in the crowded environment of the nucleus. However, the current understanding of how TF-condensates can achieve these feet so quickly and efficiently is still unclear. In this review, we draw parallels with other protein condensates and share our speculations on how the nucleus uses these TF-condensates to achieve high transcriptional specificity and fidelity. We discuss the various constituents of TF-condensates, their properties, and the known and unknown functions of TF-condensates with a particular focus on steroid signaling-induced transcriptional programs.

Keywords:

• Transcription
• chromatin
• histones
• super-enhancers
• condensates

Introduction

The cell is a multi-compartment system composed of various interlinked units, essential for segregating the entire chemical pool into restricted domains of defined activity. One of the mechanisms by which it is achieved is through membrane-less organelles. They are cellular compartments that work by spatially sequestering of macromolecules. In this process of locally enriching macromolecules, the otherwise homogenous nucleoplasm undergoes demixing to form two phases: a macromolecule-rich dense phase and the other rarer dilute phase. This physicochemical process is called phase separation [1,2]. The dense phase exists in the cell as dynamic structures, often referred to as biomolecular condensates, with molecules constantly shuttling between the condensates and the surrounding medium [3]. Of late, there has been tremendous growth in the field of biomolecular condensates, and they have been shown to perform a wide array of functions inside the nucleus ranging from rRNA processing and sorting [4], mRNA splicing, processing and transport [5–8], DNA replication [9,10], DNA damage repair [11,12] and transcription activation [13–16]. We focus this review on the latter type of condensates involved in transcriptional activation, which consists of transcription factors (TF) and co-regulators.

TF-condensates are dynamic structures, which constantly form and dissolve in the 3D nuclear space to regulate gene expression and chromatin architecture [17,18]. Recent studies have reported that cells use these TF-condensates to mount an appropriate response to cellular cues [15,16,19–22]. Although the current understanding of the mechanism through which they function is still rudimentary, we speculate that the nucleus uses these condensates to functionally segregate different transcriptional programs. It can be used to achieve biological reactions at high fidelity and to ensure that multiple signaling cascades and biological pathways are transcriptionally independent yet linked via the nuclear pool. Further, in our view, the presence of TF-condensates inside the nucleus and its ability to alter the local chromatin topology provides a new mechanism of how signaling cascades via TF-condensates shape the chromatin architecture, giving rise to specific gene expression programs.

In this review, we will use the term ”condensate” to discuss the TF-bodies, irrespective of whether they result from liquid-liquid phase separation or formed by other known and unknown physical principles. We will start by introducing the players involved in the formation and regulation of TF-condensates in the nucleus. Then, we will highlight the regulatory potential of condensates in the nucleus and share our speculations on the potential role(s) in shaping gene transcription and co-regulation. Finally, we discuss the different aspects of signaling, which can be explained by virtue of TF-condensates, the diseases arising once the condensate formation goes haywire. Finally, we discuss some of the still unknown aspects on TF-condensate functions.

Composition of TF-condensates

Signal transduction is a coordinated series of events that allows the cell to modulate its internal environment in response to external cues. TFs acting at the nuclear-cytoplasmic junction translate the cellular signal to changes in gene transcription. Hence, the spatiotemporal control of TFs is vital in regulating the specificity of signaling response. The spatial control of TF is accomplished by its ability to bind to DNA at specific motifs and protein – protein interactions, which restrict the number of genes activated/repressed by the TF. In addition, the temporal activation of TF depends on the nature and timing of the signal. The sheer number of signaling pathways in the cell, along with the diversity of molecules and their interactome, make the task of spatiotemporal control quite formidable. The compartmentalization of molecules into restricted domains, as in biomolecular condensates, provides a mechanism for controlling the TFs activity inside the cell.

A TF-condensate is potentially a hub of different types of biomolecules, i.e., proteins, RNA, and DNA. These biomolecules undergo extensive inter/intramolecular crosslinking. Based on the extent of interactions with the rest of the components, the different elements of TF-condensate can be grouped into two major categories. The first is a multivalent polymer that forms a broad range of interactions with other molecules and hence forms the central core of the condensates, called a scaffold. In contrast, the other category of accessory molecules is called clients [23]. Clients are low-valency molecules recruited to the condensates via interactions with the multivalent scaffold. Although the distinction of which molecule acts as a scaffold or a client is not always clear, majorly in TF-condensates, RNA/DNA can serve as scaffold hosting binding sites for multiple client TFs and regulatory proteins [13,14,24]. Client TFs bound to the scaffold can also recruit other co-regulators like MED1, P300/CBP, BRD4, and polymerase to the site [25,26]. The scaffold and client molecules of the TF-condensate work together to govern the nature and subsequent functions of these condensates [27,28]. Below, we will discuss the roles and properties of proteins, RNA, and DNA in the TF-condensates.

Proteins in TF-Condensates

TFs and cofactors like coactivators/repressors, polymerase complexes, and looping machinery are basic protein elements of TF-condensates [13,29–31]. TFs consist of two major interaction domains: (i) Structured Domains (SDs) with well-defined interaction surfaces that can interact with molecules at high affinities, like the Oligomerization Domains involved in protein-protein interaction or DNA/RNA binding domains (ii) Intrinsically Disordered Regions (IDRs) that do not adopt a stably folded structure [32,33]. The transactivation domain of TF, vital for recruiting co-regulatory machinery to the active transcribing region, contains stretches of IDR involved in low affinity promiscuous inter-molecular interactions. Since some of the characteristics of TF-condensates are not well studied in this regard, we will take examples of other condensates in the nucleus and speculate the similarity with TF-condensates.

SDs, as the name suggests, are structured/folded segments of a protein that can adopt secondary, tertiary or quaternary structures. SDs act as interaction modules for specific high-affinity interactions with molecules of the same proteins as in dimerization/oligomerization domains or other proteins and nucleic acid (RNA/DNA). For example, the DNA repair protein SSB (Single-Strand Binding Protein) recruits around 18 proteins involved in DNA repair to the damaged site by its C-Terminal Peptide CTP [34]. CTP-mediated protein recruitment is essential for the phase separation of SSB at the single-stranded DNA [34]. An example of the oligomerization domain facilitating phase separation of protein is in the case of ph proteins that belong to the Polycomb Group (PcG) family in Drosophila. Ph forms condensate through extensive homopolymer interactions via its SAM oligomerization domain [35]. Similarly, TFs also rely on oligomerization domains and form dimers and oligomers upon transcription activation [36]. The homo/heterodimers of TF formed as a result of dimerization bind to DNA, RNA, or proteins at different affinities compared to the monomer [37,38]. Hence, the varied interaction potential of TF-dimers/oligomers could in turn, affect the biochemical nature of the condensate.

IDRs are low-complexity domains of the protein composed of repeating units of a few charged/polar amino acids that do not adopt a unique three-dimensional structure under physiological conditions but are nevertheless functional domains [39,40]. These regions are involved in weak promiscuous binding with other proteins, which is majorly charge-dependent and is often supplemented by pi interactions of aromatic amino acids as in cation-pi interaction. These low-affinity interactions control the dynamic nature of biological condensates [41,42]. For example, NICD, Nephrin intracellular domain, is a protein-rich in negatively charged IDR. It forms phase-separated condensates via complex coaccervation, where positively charged proteins in the nucleus act as a buffer to decrease electrostatic repulsion between other NICD molecules [43]. The opposite of this is true for some TFs. Positively charged IDR generally surrounds TFs structured DNA binding domains, which can potentially be phase separated with negatively charged DNA/RNA/proteins in the nucleus [44]. In addition to the charge-based interactions, IDRs also contain short stretches (3–10 amino acid) of ubiquitous interaction modules called SLiMs (Short Linear Motifs), through which it interacts with the SDs of the proteins [45]. Few examples are The PXXP motif present in the IDR of Tau protein is responsible for its ability to interact with the SH3 Domain of Fyn [46]. The LxxLL motif present in the IDR of MED1 which also recruits glucocorticoid receptor (GR) to the MED1 condensates [47]. MED1 selectively partitions RNA polymerase II together with its positive allosteric regulators, while excluding negative regulators providing specific transcriptional activation and repression [48].

Post-Translational Modifications (PTMs) and amino acid mutations provide another layer on top of sequence determinants that can alter the valency of the protein and, therefore, its ability to phase separate. PTMs and mutations involving charged amino acids, such as lysine, arginine, serine, etc., work by their ability to modify the effective charge of the protein. IDRs are rich in amino acids that are post-translationally phosphorylated, making proteins with IDRs reaction substrates for both phosphatases and kinases [49–51]. The same PTMs can work differently in different proteins. For example, in the case of phosphorylation, FUS, and TDP-43 show decreased propensity to phase separate upon phosphorylation, whereas the Tau and HP1a phosphorylation favors droplet formation [52–55].

RNA in TF-condensates

Nuclear space is densely packed with a large number and variety of RNAs. RNA molecules act as a scaffold recruiting multiple-client TFs, leading to the formation of heterogenous RNA-TF condensates [56,57]. The information content of the RNA, i.e. the secondary structure, the sequence of RNA, the level inside the condensate and the RNA modification, are essential in determining the properties of the RNA-TF condensates like the size, viscosity, etc. [58–64]. Further, once the TF-condensate is formed, it can also regulate the level of RNA in the nucleus by controlling the rates of RNA transcription [21,65]. It creates a feedback system for RNA and TF-condensate, with RNA maintaining TF-condensate quality and quantity and its level being controlled by TF.

RNA interaction domains

Like protein-protein interaction, TF-RNA interactions are facilitated via structured and disordered domains in the protein. RNA Recognition Motif (RRM), K homology domain, Zinc finger domain, or double-stranded RNA binding motif are a few structured domains interacting with RNA at high affinity [66]. Similarly, positively charged IDRs in the TFs can interact with RNAs via weak charge-based interactions [61,63,67].

The FUS family of proteins is a great example, where the RNA binding domain consisting of the structured RRM and IDR governs the propensity of the protein to phase separate [60,68]. FUS is a family of around 30 proteins associated with neurodegenerative diseases like ALS and Alzheimer’s. It comprises two major domains: the PLD (Prion Like Domain), rich in IDR, and the RBD (RNA Binding Domain), which contains portions of IDR and structured RNA recognition motif. The interaction between the two domains of FUS is vital for the phase separation of the protein. The two domains alone form droplets in vitro at a high concentration (120uM), while the full-length FUS forms droplets at a physiologically relevant concentration (5uM) [68]. It is due to the balance of cation-pi interactions between aromatic amino acids Tyr and positively charged amino acid Arg. Further, the spacing of positive charges in the IDR, i.e. the RGG motifs, influence the properties of FUS droplets where the amino acid Gly in RGG is necessary for the flexibility of the IDR chain and makes the FUS droplets less viscous [68].

RGG or related arginine-rich motifs are low-complexity domains in the protein consisting of repeating units of Arg [69]. These motifs are enriched in RNA-binding proteins [70]. RGG motif-containing proteins interact with negatively charged RNA in a charge-dependent manner [60]. TFs containing arginine-rich motifs like ERα, SOX2, KLF4, and STAT3 interact with RNA at similar efficiencies as known RNA binding proteins like hnRNPA1 and U2AF2 [71]. These arginine-rich motifs surround the DNA binding domains of TFs, which might potentially supplement the DNA binding ability of TF due to the presence of transcribing RNA at the loci [71]. The amino acids of the RGG motif can also undergo post-translational modifications like methylation [69]. The arginine methylation in the RGG motif decreases the effective charge on the protein, hence its ability to bind to negatively charged RNA. It showed a marked decrease in the propensity of FUS to form phase-separated droplets [72].

RNA regulating TF-condensates and their specificity in the nucleus

RNA forms an integral unit of the TF-condensates as it holds information regarding the biochemical properties of the TF-condensate, which we refer to as RNA ID. The RNA ID can help segregate the condensate formed by one RNA from another non-interacting RNA, which keeps the unrelated transcriptional program separate. It also insulates the effect of one signaling event from the other and simultaneously gives opportunities for signaling cross-talk if the condensates contain similar RNAs.

One way RNA ID can work is through spatial segregation of condensates in the 3D nuclear space so that even the condensate with similar proteins cannot undergo fusion due to its separation. RNA can recruit proteins containing associated RBDs at various spatial locations inside the nucleus. It is seen by artificially localizing RNA molecules using LacO arrays, which is enough to make interacting proteins to get recruited at the site. For example, tethering of H2B-mRNA at a genomic locus led to the accumulation of NPAT transcription factor at the site, which is involved in forming HLBs (Histone Locus Bodies). Similarly, the tethering of NEAT1 lncRNA also caused the formation of paraspeckles at the genomic locus [62]. Enhancers are one such site with active RNA transcription that produces cis-regulatory RNA called eRNAs or enhancer RNAs. At these enhancer sites, TFs are recruited directly via DNA binding domain or interaction with other proteins/RNA (Indirect binding) [25]. Once recruited, nascently transcribing eRNA can act as a molecular gum trapping multiple TFs at the site along with other looping factors like YY1, CTCF, cohesins, and coactivators like Med1, p300 and CBP, etc. [73–75]. The binding of these wide varieties of proteins creates a ribonucleoprotein hub at the enhancer. Ultimately, the interactions between proteins, RNA, and DNA polymers result in a phase separation of the enhancer complex [13,14,16,24,76,77].

Once the condensate is formed, RNA can also control the fusion of one condensate with another. RNA-ID can influence coalescence and hence the growth of RNA-TF condensates. The linear sequence is an essential determinant for RNA-RNA interactions. Base pairing interactions between RNAs support the fusion of condensates. Homopolymers of base pairing RNA, poly rA + poly rU, form coexisting aggregates containing both polymers. While a mix consisting of homopolymers of poly rA + poly rC and poly rU + poly rC, which are non-base pairing in nature, exist as segregated droplets enriched with one of the two polymers but never both [64,78].

Besides the sequence, the information in the secondary structure of RNA, like the hairpin loops, stem, bulge, etc., is another factor that affects how proteins/RNA interact with the other RNAs. Condensate composed of RNAs with interacting secondary structures can coalesce into one condensate while preventing the fusion of condensate with non-interacting RNA. Whi3 is an RNA binding protein containing polyQ motif and RRM [58,61]. It interacts with different RNAs in a secondary structure-dependent manner. The specific binding with each type of RNA allows it to form separate condensates that do not coalesce [58]. We believe the ncRNAs, like eRNAs, potentially form specific condensates by virtue of their secondary structure being recognized by transcription factors or cofactors like p300, CBP, MED1 etc.

RNA also undergoes different post-transcriptional modifications like m6A methylation. It is involved in a wide range of functions, including nuclear transport of RNA, translation regulation, RNA stability, etc. [79]. These various functions result from the protein interactions of m6A methylated RNA [79,80]. For example, the m6A binding protein YTHDF1/2/3 forms droplets with m6A-modified RNA but not with non-methylated RNAs [81]. Similarly, eRNAs get m6A methylated and recruit readers like YTHDC1 to the active enhancer sites for the phase separation of these enhancer complexes [82].

At low RNA/Protein ratios, RNA supports condensate formation by lowering the C^sat (Saturation Concentration, i.e. the minimum concentration of the protein needed to phase separate) of the interacting protein [68,83]. However, high levels of RNA cause the shift from a growth supplementary to an inhibitor of TF-condensates [60,83]. The level of RNAs with respect to the protein is essential in controlling the condensate size. High RNA levels lead to a smaller size condensate [60,83,84]. It is seen in the MED1 condensate, where increasing levels of eRNA beyond a threshold decreases the size of the MED1 condensates [85,86]. This effect of RNA is also evident in mESCs cells when the reduction of RNA synthesis using chemical inhibitors (actinomycin-D and DRB) increases the size of MED1 and MED19 and CTCF condensates [30,83].

To summarize the role of RNA-ID in phase separation, it brings in multiple TFs and cofactors within the RNA binding domain at the transcribing site. It leads to the increase in the local concentration of proteins and subsequent-phase separation of the RNA-protein complex. Further, the newly formed TF-condensate also supplements RNA production, increasing the local RNA level at the transcribing site. The increased RNA levels, in turn, lead to a decrease in condensate size. Hence, this creates a feedback loop of RNA creating and RNA dissolving the condensate.

DNA and chromatin in TF-condensates

Distal and proximal regulatory regions like promoters and enhancers regulate the transcription of its associated genes. Enhancers interact with the cognate promoters and regulate their expression by recruiting TFs and cofactors [87–89]. They work as computers by taking input from multiple signaling cues and integrating them into regulatory output for gene expression [90–92]. The regulatory enhancers consist of three major elements: (1) TFs with DNA binding Domains, (2) The DNA motif that the TF bind and (3) The histone modifications and TF-dependent indirect recruitment of cofactors to the site.

Basic chromatin components of TF-condensates

TF-condensates primarily consist of TFs and cofactors. TF recognizes DNA with the help of domains, which form energetically favorable interactions with the DNA. At the same time, the cofactors, even though they lack DNA binding domain, are recruited to the site in association with TFs. The DNA binding domains contain elements, which exhibit specific binding to the major groove of the DNA. The major groove has enough info in the form of hydrogen donor, acceptor, or methyl group to differentiate between different bases. It gives sequence specificity to the DNA-binding domain interactions. There is a large variety of DNA binding domains in the protein world. Some well-known examples are (1) Helix-turn-Helix domain, (2) Zinc Fingers, and (3) Leucine Zipper Domains [93]. In these domains, structured elements like α-helix or β-sheet interact with DNA to regulate DNA-protein interaction. In addition to Structured DNA binding domains mentioned above, IDRs also interact with DNA and help the TF scan the DNA for its cognate-binding motif [94]. Positively charged amino acids in the IDR allows it to act as a clamp that binds on negatively charged DNA. It creates a temporary bridge in which two DNA strands are attached to the protein, one with the help of an alpha helix or other recognition domain and the other using IDR. Followed by the temporary bridge, the TF jumps from one strand to the other like someone on a monkey bar. Hence, it is called the monkey bar mechanism of DNA search by IDR [94]. These molecular interactions on DNA are reflected as various diffusion states for glucocorticoid receptors. GR contains both a structured DBD domain and an N-terminal IDR, which helps in its binding to DNA and IDR-IDR interaction slows the GR in a sub-diffusive state for it to interact with DNA. Therefore, GR undergoes different movements during motif search: Diffusive, i.e. random motion, Sub-Diffusive or Confined State and Bound State. The removal of N-Terminal IDR perturbs its confined state, resulting in the reduced condensate size of GR inside the nucleus [95].

Inside the nucleus, the DNA is often packed into nucleosomes composed of DNA and histone proteins. The nucleosome arrays of multiple nucleosomes behave much like solid aggregates in vitro [96]. The solid-like attributes of chromatin make it less amenable to changes because of the slow diffusion rate. However, various TF factors like pioneering factors and histone remodelers inside the nucleus can influence the nucleosome by evicting histone, a possible mechanism for making chromatin more liquid like [97,98]. Some intrinsic features of nucleosomes that make them more susceptible to changes are: (1) The Linker DNA length: Distances between nucleosomes are biased toward 10n +5 spacing, which is favorable to its phase separating behavior, whereas the linker lengths with 10n spacing disfavor droplet formation [96]. (2) Linker Histone: Histone H1 can phase separate on its own with DNA, but in association with nucleosome, it condenses the chromatin and slows the dynamics of the already viscous chromatin [99,100]. (3) Post-Translation Modification of histones: The histone proteins form the central core of the nucleosome complex with a disordered tail protruding outward. Histone tails are accessible by a wide range of histone readers, writers, and erasers that post-translationally modify the tail, subsequently leading to the recruitment of activation or repressive machinery. Acetylation, methylation, phosphorylation, sumoylation and ubiquitination are the major PTMs found in the nucleosomes. These histone marks are essential in segregating the active chromatin marked by H3K27ac and H3K4me1/2/3 from repressed, marked by H3K27me3 and H3K9me3 [101]. The chromatin polymer defined by these states allows it to form the domains for distinct functions in the nucleus. For example, acetylation of histone decreases the propensity of chromatin to phase separate, but the presence of BRD4 allows it to form condensates [96]. However, these BRD4 condensates do not mix with the unacetylated histone condensate, possibly due to differences in viscosity and the presence of non-interacting protein partners [96]. Similarly, in the case of methylation, H3K9me3 marked constitutive heterochromatin forms a condensate along with its reader HP1 and its writer SUV39H1 [102]. In addition, the SUV39H1 and HP1 condensates do not overlap with general transcription factors like TFIIB, hence keeping the region repressed. This selective compartmentalization of active/inactive marks into different condensates points toward the ability of chromatin to robustly segregate active and inactive genomic regions [47,102].

Clustered enhancers and TF-condensates

As already discussed, the cooperative binding of TFs with DNA binding domains, histone modifiers, and the cofactors bound at a single enhancer can potentially cause it to phase separate. However, co-regulatory enhancers in the genome occur as clusters with successive-binding events within kilobases from each other [16,85,86]. These clustered enhancers are characterized by unusually high binding to transcriptional coactivators and high transcriptional activity, and are often referred to as super-enhancers [85,86]. The cooperative nature and high density of TF binding at the super-enhancers or, for that matter, at clustered enhancers make it an ideal candidate to study phase separation. Oct4, Nanog, and Sox2 are transcription factors with highly active super-enhancers that overlap with phase-separated condensates around genes involved in maintaining pluripotency [13,86]. Simulations comparing super-enhancers and typical enhancers have shown that it requires a smaller number of molecules per enhancer to show the same transcriptional activity as typical enhancers. Further, a higher concentration of TFs at clustered enhancers increases the transcriptional activity of the associated gene by increasing the transcription burst frequency [76]. Phase separation of MED1 and BRD4 on Nanog super-enhancer validate these in silico findings [13].

Polymerase II in transcriptional condensates

The TF-condensate components are at equilibrium with the surrounding nuclear environment, with molecules constantly shuttling from the condensate to the surrounding media. The concentrated nature of the TF-condensates provides a suitable environment for the transcription reaction. These regions can act as transcription factories regulating the transcriptional programs and impact the state of chromatin surrounding these condensates such that chromatin is opened further for more TF to bind [103–107].

RNA Polymerase II, central to these transcription factories, forms a phase-separated condensate in vitro and inside the nucleus [18,108,109]. Since the Pol II condensates are active transcription sites, cells must keep their phase separation in check to prevent spurious transcription. One way is through the C-Terminal Domain of polymerase II (CTD). CTD consists of a low-complexity domain, which contains multiple [YSPTSPS] heptapeptide repeats. The heptapeptide repeats are evolutionarily conserved in eukaryotes, and their number varies from 52 in humans and 26 in yeast, scaling inversely with the gene density [110]. The number of repeats influences the ability of the polymerase to phase separate, where polymerase with shorter CTD (25× repeats) forms a smaller and lesser number of condensates relative to longer CTDs (52× repeats) [109]. CTD recruits TF and cofactors in a length-dependent manner, which leads to different-sized condensates in shorter v/s longer CTDs and relatively weaker vs more robust transcriptional activation, respectively [111,112]. Further, the shorter CTDs also exhibit delayed response to transcription activation signals, which might be due to more time needed for eRNA to reach levels to elicit a response [112].

CTD heptapeptide repeats are phosphorylated at the 5^th and 2^nd serine during transcription initiation and elongation, respectively. Therefore, PolII forms two types of condensates. The first is the initiation condensate that recruits multiple general transcription factors to the PIC [18,109]. After transcription initiation, kinases like CDK7/9 phosphorylate the Ser-2 of the CTD, which marks its exit from initiation condensates [108]. The Ser-2 phosphorylated CTD gets recruited to the second type of condensate called the splicing factor condensates composed of splicing factors [113] suggesting, the phosphorylation of PolII triggers its switching from transcriptional to splicing condensate.

Condensates regulating transcription - Insights into signaling-dependent TF-condensates

Unlike development, the transcriptional program during signaling is dynamic: genes cycle between induced and basal states during signaling. Surprisingly, how enhancers govern these dynamic signaling responses is not well established. Since the TF-condensates on enhancers can regulate gene transcription, as signaling peaks, new condensates are formed to support the cognate genes and are dissolved as soon as the signaling decays. Therefore, signaling-induced transcription and associated condensates provide an opportunity to understand the biological functions of condensates, the required proteins, RNAs, and their relationship with genome organization around these condensates. Below, taking examples of nuclear receptors under hormone signaling, we discuss how enhancers upon signaling operate and how they provide reproducible signaling response with the help of phase separation. We believe that the findings from nuclear receptor signaling can be generalized to other ligands and signaling cues.

TF-condensates upon signaling

Research over the decades has shown that several signaling-dependent TFs form distinct intra-nuclear foci in a ligand-dependent manner. The list includes but is not limited to STAT3, p65, and nuclear receptors like GR, AR, and ERα. Recently, some of these have been shown to be phase-separated condensates as well [16,20,114,115]. Ligand stimulation alters the TF concentration inside the nucleus and modifies them post-translationally. It can affect their ability to dimerize or multimerize and also modulate their interaction with coactivators/co-repressor [116]. These changes then dictate the binding of TFs and cofactors on specific chromatin regions to shape the signaling-driven transcriptional response.

The phase separation of TF upon ligand stimulation is a robust multi-step process that begins with initiation or nucleation, followed by the growth of the nucleated condensate, culminating in the assembly of a stable condensate. As mentioned above, for the nucleation event, the critical concentration of proteins is crucial [114]. We speculate that nucleation starts with the clustering of pioneering factors along with limited molecules of co-regulatory proteins on enhancers. Upon ligand addition, the signaling-induced TF translocates into the nucleus and utilizes these seeds (nucleated condensates on enhancers), so that a bigger condensate can be formed (Figure 1). The growth of condensate is the cumulative result of; (a) specific binding of cofactors, remodelers, readers and RNA polymerase complex on enhancers that TFs bind. Most of the TFs and cofactors are rich in IDRs, which results in the formation of droplets on DNA, (b) several enhancers, and promoters interact in 3D space, bringing their smaller condensates together to form larger ones, (c) resultant nascent RNA (eRNA or genic RNA) can then increase the valency further to recruit more client proteins leading to a vast entangled mesh of crosslinked DNA-RNA-proteins as stated above.

Figure 1. Nucleation and growth of TF condensates on enhancers: in absence of transcriptional stimuli, the pioneering factors with least critical concentration nucleate condensate by binding on their cognate-binding site at enhancers. Upon stimulation, transcription factors are recruited at these condensates which intern recruit cofactors, mediator and polymerase complexes. These proteins exhibit low and high affinity interactions by the virtue of their IDRs and activation domain. An active transcription complex triggers RNA expression which in turn enhances the valency of complex by many folds resulting in highly crosslinked complex forming TF condensates.

For example, hyperosmotic stress in the cell promotes YAP condensate formation for the activation of stress response genes. The YAP condensate consists of PolII along with TEAD1, a transcriptional coactivator [22,65]. Similarly, ERα, a steroid hormone receptor, forms condensate on active enhancer sites to activate estrogen-dependent genes upon hormone stimulation [16,77]. ERα condensates are megadalton complexes with multiple TFs like GATA-3, FOXA1, MED1, etc. [16,25,77]. Similarly, other nuclear receptors like AR and GR also form nuclear condensates in a hormone-dependent manner [19,47,117].

Apart from forming condensates, signaling can also modulate the transcriptional program by dissolving existing condensates. TAZ condensates activate gene expression through cofactors like BRD4, MED1, and PolII [21,118]. Phosphorylation of TAZ upon activation of hippo signaling causes the dissolution of TAZ condensates, altering the expression of the hippo signaling responsive gene [21]. Similarly, the overexpression of SOX2 causes perturbation of GR condensates [119].

As aforementioned, the local concentration of TFs, cofactors like BRD4, MED1, and polymerases is higher on clustered enhancers. These properties, along with the large surface of chromatin on clustered enhancers, are ideal environments for condensates to form. In this regard, even the singleton enhancers from various genomic regions across the genome can form condensates if they interact in 3D nuclear space. This may explain why singleton enhancers may give transcriptional output that is similar to clustered enhancers [120].

Role of basal signaling in condensate specificity

The cell ensures reproducible and less noisy transcriptional response to stress and cyclic signaling (cytokines and steroid hormones, such as estrogen, androgen, progesterone, and glucocorticoids). Under cyclic signaling, nucleation of condensate at specific enhancer and promoter as a seed under basal signaling guarantees a less noisy and reproducible signaling output [16,121]. Though most TF molecules bind the DNA after ligand stimulation, some TF molecules bind to key enhancers unliganded by unknown mechanisms [16]. For example, an unliganded ERα along with FOXA1 binds to the future functional enhancers under basal signaling to seed the future condensates [16] (Figure 2). The ERα condensates are preferably formed on these bookmarked-seed enhancers for faithful signaling response. At the end of the estrogen signaling response, the condensates dissolve but leave the seed enhancer nucleated by ERα and FOXA1 behind for the next round of ligand stimulation. Surprisingly, seed enhancers are less prone to degradation due to the presence of nucleated condensate.

Figure 2. Basal signaling pre-seeds TF condensates on enhancers: ERa and pioneering TF, FOXA1 bind on functional enhancers under basal signaling. Upon ligand stimulation, liganded ERa recuirts its trasncriptional machinary to these pre-seeded enhancers and froms condensates to activate estrogen regulated genes.

Downfall of signaling response; dissolution of condensates

Since the transcriptional outcome is correlated with condensates on enhancers, the regulation of condensates governs the transcriptional dynamics of a signaling response. For example, estrogen signaling response in terms of induced gene expression peaks at 1 h and declines from 3 h onwards. Concurrently, the ERα condensate formation also peaks at 1-h post signaling, and it starts to disappear at 3 h, accompanied by a concurrent drop in induced gene expression. The disappearance of these condensates could be due to several factors, including ERα turnover, eRNA, and RNA accumulation, post-translational modifications of ERα and its associated proteins or modification of eRNAs, which may change the valency of the proteins in the complex, thus, directly affecting the phase separation of ERα. As signaling declines, perhaps TFs escape profoundly from condensates, thereby dissolving the condensates. For example, the size of progesterone receptor (PR) condensates has been shown to be regulated by the escape of PR molecules [122]. However, this departure of TFs still leaves the nucleated condensates on enhancers serving as a bookmark for the future rounds of the active transcriptional cycle.

Condensates as the hubs of signaling cross-talk

The functional enhancers under estrogen signaling bind with ERα and its cofactors and other TFs like STAT1, RARα, RARγ in ERα-dependent manner. These diverse TFs that respond to different ligands at the same enhancer pose an interesting possibility of these enhancers acting as the hubs for different ligand-dependent signaling.

Enhancer elements act as a scaffold housing multiple DNA motifs for client TF molecules. Two general enhancer action mechanisms have been proposed: (1) The Enhanceosome model, where TFs bind in a highly synergistic manner that leads to a profound effect on Transcription than either factor binding alone [123,124]. Nanog is a pluripotency transcription factor that forms condensates on active enhancer elements along with MED1 [14,125]. In the genome, the Nanog motif shows a fixed periodicity, with subsequent motifs occurring after every 10.5 bp. The recurring motifs act cooperatively to support Nanog’s binding on active enhancers, hence its phase separation [126]. Recruitment of TFs like GATA3, FOXA1, and ERα to enhancer regions to activate cognate gene promoters upon E2 stimulation is another example of this model [25,77]. (2) the Billboard Model in which the entire enhancer element acts additively. In this case, the enhancer element is divided into several small subsets with single or multiple TFs binding motifs. Here, the enhancer element acts as an integration site for multiple signals, with the overall effect of the enhancer being the sum of all the elements [124]. Although the Billboard Model is not well characterized for TF-condensates undergoing phase separation, it can play a major role in understanding the cross-talk between different signaling pathways. Since a single enhancer element can recruit a different set of TFs depending on the signaling state, creating a platform for multiple signaling condensates fusing at enhancer sites. Further, TFs that bind stably on DNA can act as a scaffold than client. In both models, motif syntax, the spatial arrangement of motifs, and their orientation with respect to each other play an important role in regulating the binding of TF [126]. The overlapping binding can either perturb or help in the binding of TF [127,128].

Condensates to support weak promoters and gene co-transcription

In the dense nuclear space, 3D proximity and not the linear distances determine the promoter regulation by cognate enhancers. Owing to this, multiple enhancers and their respective genes can be proximal in 3D nuclear space and thus regulated by the same TF-condensate (Figure 3). Such hubs may utilize LLPS for enhanced transcriptional fitness despite the relatively weaker transcriptional strength of individual constituent enhancers and promoters in these hubs. This implies that these coregulated multi-enhancer transcriptional hubs in 3D may provide phenotypic robustness against biological perturbations [129,130]

Figure 3. cis and trans chromatin interactions in condensates: Condensates can harbor chromatin from cis or trans regions. Cis regions are from same stretch of chromatin with multiple regulatory elements placed at short distances like multiple enhancers in a super enhancer with different TF binding sites or clustered enhancers with same TF binding sites. The trans regions are the regulatory elements from longdistances on same chromosome or from different chromosomes. Whether chromatin proximity defines the specificity of these condensates or the act of phase separation brings these regulatory elements in 30 proximity remains unknown. Further, one or multiple genes may be involved in each condensate, allowing the co-regulation of these genes.

The presence of multiple genes in the same condensate can work efficiently even if there is a mutation in a single enhancer/gene other than the nucleating enhancer or in a co-regulator. Similarly, single or multiple-signaling inputs can simultaneously regulate constituent genes within the same condensate. It will be interesting to test the minimum regulatory elements/factors required for the expression robustness of a condensate. But we still do not understand the features of constituent enhancers or promoters and whether they are specific to a particular type of TF/cofactors/stimuli.

Condensates and diseases

The mutations in TFs alter their target DNA sequence specificity, thus increasing or decreasing their affinity to DNA, allowing TF to become promiscuous and acquire new-binding sites in the genome. This redistribution may trigger the nucleation of condensates at new DNA sites resulting in the dysregulation of multiple new genes in the 3D chromatin context. For example, FOXA1 mutations are frequent in breast and prostate tumors, resulting in the loss or gain of FOXA1 binding at specific and new sites. This impacts the binding of ERα and AR upon ligand stimulation; thus, these mutations impact hormonal therapy [131]. Similarly, the repeat expansion in TFs alters their phase separation propensity in the associated pathophysiologies [132]. Single nucleotide variation in the non-coding genome may create the binding of a TF, but for it to nucleate the condensate, the overall density of that chromatin region has to be supportive. If the gain increases the overall density of the TF binding sites above the threshold in that region, then these sites can act as a nucleation point and grow, as the cofactors with low affinity interact with these factors, which results in an emerging condensate [133]. These speculations are based on the fact that pathophysiologies involve new super-enhancers formed by virtue of single nucleotide polymorphism [134,135]. Mutations in DNA may modulate the transcribed RNA in a manner that changes their affinity and ability to bind with proteins, and since most condensates involve RNA: protein interactions, we speculate the growth and stability of the condensate may also change. Apart from mutations in proteins and DNA, the small-molecule inhibitors also that alter the properties of condensates and, therefore, provide a unique strategy to modulate the disease outcome (Figure. 4).

Figure 4. Nuclear condensates and mutations/Diseases/chemical perturbations: The small molecule inhibitors/drugs and the disease causing mutations in regulatory elements, TF/cofactors, alter the valency and affinity of these polymers. These alterations change the their propensity to form or dissolve the condensates thus, affect the biological functions. Importantly, such features can be harnessed to develop therapies.

Outstanding questions

Condensate existence is well accepted currently. However, the molecular details of the mechanism of action of these TF-condensates is not very well understood. The following are some of the outstanding questions in the field:

Although the signaling cascade is a highly dynamic system, and phase separation of TFs might play a role in increasing the local concentration of these factors at the transcription site but does this makes the system more robust and resistant to perturbations, even the mutation?

What is the identity of regulatory elements sharing a single condensate? Is it static or dynamic in response to cues?

Are TADs the result of phase separation supporting cis interactions among the regulatory elements within the condensate but not outside the condensate?

Do phase-separated condensates have the ability to sense and modify chromatin around them to shape the chromatin architecture?

Can condensates be used as a multi-enhancer hub for transcription [134]?

Author contributions

RM and DN wrote the manuscript.

Data availability statement

No new data was generated in the review.

Acknowledgments

The authors thank DN lab members and Amanjot Singh for the discussions and constructive inputs. We thank Sakshi Gorey and Ishfaq Ahamd Pandith for preparing schematic for the manuscript. We apologize to colleagues whose work is not cited.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The work was supported by Department of Atomic Energy, Government of India, under project no.12-R&D-TFR-5.04-0800 and intramural funds from NCBS-TIFR (to DN), Welcome-IA (IA/1/14/2/501539) (to DN) and DST core grant (CRG/2019/005714) (to DN). DN acknowledges the support from the EMBO Young Investigator program. RM is supported by the TIFR-NCBS graduate program.