**3. Post-Translational Modifications of Intrinsically Disordered Domains Can Govern LLPS**

The lack of secondary structure makes IDRs especially susceptible to post-translational modifications (PTMs) [24]. In fact, IDRs are disproportionately modified post-translationally relative to the entire proteome [25–27]. A variety of PTMs can alter IDR charge, hydrophobicity, size, and structure. These changes may occur through additions of functional groups (e.g., phosphoryl, methyl, acyl, glycosyl, alkyl, etc.), or subtler chemical changes such as oxidation, deimidation, and deamidation [28].

There are many biological examples of PTMs serving as on/off switches, where they regulate a cellular event, such as protein signaling, localization, and degradation. In the case of IDRs and phase separation, PTMs can similarly have on/off functions by altering the nature of intermolecular contacts that support MLO formation or dissolution [29] (Table 1). Here, we discuss examples of PTMs and IDRs in proteins that undergo functional phase separation in cells. We evaluate the hypothesis that the combination of IDR multivalency and the capacity to be extensively modified results in reversible networks of interactions that can be regulated by specific cellular cues (Figure 1).

**Table 1.** Examples of post-translational modifications (PTMs) of intrinsically disordered regions (IDRs) altering the liquid–liquid phase separation (LLPS) of proteins. The underlined proteins have multiple PTMs that affect the phase separation propensity. Arrows indicate if PTMs promote (↑) or inhibit (↓) LLPS. C-terminal domain (CTD), N-terminal domain (NTD), prion-like domain (PrLD), arginine-glycine-glycine (RGG), stress granules (SGs), amyotrophic lateral sclerosis (ALS), and frontal temporal dementia (FTD).


**Figure 1.** Liquid–liquid phase separation (LLPS) of biopolymers, such as proteins and RNA, is a mechanism by which cells organize their contents into specific functional structures called membraneless organelles (MLOs). Post-translational modifications (PTMs) of intrinsically disordered proteins can influence LLPS and thus regulate the formation and dissolution of MLOs. The figure depicts different patterns of PTMs favoring dispersed or condensed states. Changes in the material properties of liquid-phase separated granules are hypothesized to cause some neurodegenerative diseases. According to this hypothesis, droplets lose their liquid (reversible) properties and adopt more rigid (less reversible) internal structures, which may be glass-like, or in some cases, may have solid amyloid-like structures. These irreversible phase states may have gain-of-function toxicity to neurons.

### *3.1. Serine*/*Threonine*/*Tyrosine Phosphorylation*

Phosphorylation is the covalent attachment of a phosphoryl group to an amino acid hydroxyl group. The phosphoryl group is negatively charged, so its addition changes a polar, uncharged residue to a negatively charged amino acid. Serine is the most commonly phosphorylated residue, followed by threonine and tyrosine [30]. The addition of charges to macromolecules may promote certain charge–charge interactions that drive complex coacervation (phase separation of oppositely charged polymers) [31]. Alternatively, addition of phosphates may cause charge repulsion or steric hindrance, thus inhibiting phase separation [2]. Depending on the protein context, the phosphate modification of amino acids can either favor or disfavor phase separation.

Serine/threonine phosphorylation has been shown to promote phase separation of IDPs such as fragile X mental retardation protein (FMRP) [32], TIA-1/TIAL RNA binding protein homolog (TIAR-2) [33], and microtubule-binding protein tau [34]. FMRP has 12 serine residues within its C-terminal IDR (aa 445–632) that have been identified as targets of casein kinase II (CKII). In vitro phosphorylation by CKII results in an increase in the negative charge densities throughout this IDR, increasing the propensity for multivalent electrostatic interactions and promoting phase separation [32]. TIAR-2 also contains a C-terminal intrinsically disordered PrLD that facilitates its LLPS into cytosolic granules [33]. The PrLD of TIAR-2 was shown to be serine phosphorylated when expressed in mechanosensory neurons. Ten serine residues in the intrinsically disordered PrLD were predicted as phospho-sites using NetPhos3.1 [35]. Expression of a non-phosphorylatable (S→A) TIAR-2 mutant (at 10, 8, or 2 serine residues) showed significantly less granule formation in the axons of neurons when compared to wild type. Alternatively, phosphomimetic (S→E) TIAR-2 mutants showed similar levels of granule formation when compared to wild type. These data suggest phosphorylation of serine residues promotes LLPS and formation of TIAR-2 positive granules in neurons of *C. elegans* [33].

PTMs also affect virally encoded proteins and their LLPS capabilities in cells. Measles virus phosphoprotein is a 507-amino acid virally encoded protein composed of multiple IDRs. Measles virus phosphoprotein and nucleoprotein undergo LLPS to form IBs. Phosphoprotein is phosphorylated at multiple sites, but Serine 86 and Serine 151—both of which are in IDRs—have been identified as regulatory sites for IB formation. Mutation or inhibition of phosphorylation at these two sites results in irregular and small IBs [12].

Examples of serine/threonine phosphorylation that disrupt LLPS include maternal-effect germline proteins (MEGs) in P granules of *C. elegans* [36], fused in sarcoma (FUS) [37–39] and TAR DNA-binding protein 43 (TDP-43) [40]. Proper segregation of P granules in zygotes of *C. elegans* requires the expression of MEG proteins [41]. Interestingly, two of the MEG family proteins (MEG-1 and MEG-3) are phosphorylated within their IDRs by a regulatory kinase (MBK-2) [36]. MBK-2 activity and counteractive phosphatase (PPTR-1) activity on MEGs is required for P granule disassembly and formation, respectively [36].

FUS and TDP-43 are frequently studied proteins because their phase separation in vivo has been linked to amyotrophic lateral sclerosis (ALS) [42]. A current hypothesis is that MLOs containing these proteins may promote their stochastic conversion into solid, pathological aggregates [43]. FUS contains an intrinsically disordered N-terminal PrLD, which is necessary and sufficient to drive LLPS [5]. The ~160 amino acid PrLD has 32 putative phosphorylation sites, 12 of which have been identified as PIKK family kinase consensus sites [44]. Phosphomimetic substitution (S/T→E) at 6 or 12 PIKK consensus sites diminishes FUS's ability to phase separate and form fibrillar aggregates in vitro [37]. In cells, a decrease in cytoplasmic aggregation is also observed upon increase in phosphomimetic substitution, suggesting a potential therapeutic target for disrupting pathological aggregate formation [37]. TDP-43 has a C-terminal PrLD, which is multiphosphorylated and aggregated in ALS motor neurons [45]. Two phosphorylation sites in the PrLD, Serine 409 and 410, identified in samples from frontotemporal lobar dementia patients were shown to regulate TDP-43 cytoplasmic granule formation. Phosphomimetic substitution (S→D) at Serine 409 and 410 showed a significant reduction in the number of cells containing TDP-43 puncta [46]. Interestingly, TDP-43 phase separation is regulated by PTMs in both an IDR and a structured domain. A single phosphorylation event in its N-terminal structured domain at Serine 48 is sufficient to suppress its LLPS in vitro and in cells [40]. Serine 48 is conserved in most species evaluated, including flies, mice, and humans [40].

There are two kinases, SKY1 and DYRK3, that have the ability to phase separate into SGs and, in the stages of recovery following a stress response, phosphorylate proteins containing IDRs, resulting in dissolution of the granules [47,48]. SKY1 is a yeast protein kinase with a PrLD that enables its recruitment into SGs. In SGs, SKY1 phosphorylates NLP3 at Serine 441, which is located in its serine-arginine rich C-terminal IDR [49]. This phosphorylation event promotes SG dissolution [48]. Similarly, DYRK3 (human homolog of MBK-2) was shown to phase separate into SGs via its intrinsically disordered N-terminal domain. Aside from regulation of SGs, DYRK3 was identified as a factor that controls phase separation and dissolution of several condensates containing IDPs during mitosis [50]. This kinase is interesting because it has broad-specificity and is generally proline-directed. Some proteins sensitive to DYRK3 inhibition, all of which contain IDRs, include splicing-speckle marker SC35, SG marker PABP, and pericentriolar-material protein PCM1. DYRK3 expression results in dissolution of these granules during mitosis, whereas a kinase-dead mutant or inhibition of DYRK3 results in the persistence of granules [50].

For some proteins, phosphorylation can be a driver or inhibitor of condensate formation, but there are instances where it is not clearly binary. Tau441 contains numerous IDRs and putative serine/threonine phosphorylation sites throughout the protein [51]. In experiments performed with bacterially produced recombinant full-length tau441 and molecular crowding agents, LLPS was driven mostly by electrostatic intermolecular interactions. There was no requirement for phosphorylation [52]. However, in a different study, phosphorylation was found to be required to initiate tau441 LLPS in vitro [34]. Of importance, there are 22 phospho-sites analyzed in this study, 15 of the sites are located in IDRs. Interestingly, LLPS of p-tau was dependent on hydrophobic interactions [34], whereas unphosphorylated tau LLPS was more dependent on ionic interactions [52]. These examples suggest the biophysical mechanism driving LLPS can be altered by the post-translational state of the protein.
