Get closer and make hotspots: liquid–liquid phase separation in plants

Abstract

Liquid–liquid phase separation (LLPS) facilitates the formation of membraneless compartments in a cell and allows the spatiotemporal organization of biochemical reactions by concentrating macromolecules locally. In plants, LLPS defines cellular reaction hotspots, and stimulus‐responsive LLPS is tightly linked to a variety of cellular and biological functions triggered by exposure to various internal and external stimuli, such as stress responses, hormone signaling, and temperature sensing. Here, we provide an overview of the current understanding of physicochemical forces and molecular factors that drive LLPS in plant cells. We illustrate how the biochemical features of cellular condensates contribute to their biological functions. Additionally, we highlight major challenges for the comprehensive understanding of biological LLPS, especially in view of the dynamic and robust organization of biochemical reactions underlying plastic responses to environmental fluctuations in plants.

Liquid‐liquid phase separation and the formation of biomolecular condensates facilitates rapid cellular responses to changing environmental conditions, which is of particular importance for sessile organisms like plants. This review discusses the fundamental principles and functional importance of condensate formation in plants.

Glossary

ADCP1

AGENT DOMAIN‐CONTAINING PROTEIN 1

ARF

AUXIN RESPONSE FACTOR PROTEIN

Arp2/3

ARABIDOPSIS RIBOSOMAL PROTEIN 2/3

CCE

CRY C‐terminal extension

cpTat

Chloroplast twin‐arginine translocation

Cry

Cryptochrome

CRL3

CULLIN‐RING UBIQUITIN LIGASE 3

CUL4

CULLIN 4

DCP5

DECAPPING 5

DDX

DEAD‐box ATPase

EBF1

EIN3‐BINDING F‐BOX 1

EC

Evening Complex

ELF

EARLY FLOWERING

EIL1

EIN3‐LIKE 1

EIN2

ETHYLENE‐INSENSITIVE 2

EPYC1

ESSENTIAL PYRENOID COMPONENT 1

ER

Endoplasmic reticulum

ETI

Effector‐triggered immunity

FLL2

FLX‐LIKE 2

HR

Hypersensitive response

HSP101

HEAT SHOCK PROTEIN 101

HY5

ELONGATED HYPOCOTYL 5

IDP

Intrinsically disordered protein

IDR

Intrinsically disordered region

LCD

Low‐complexity sequence domain

LD

LUMINIDEPENDENS

LHP1

LIKE HETEROCHROMATIN PROTEIN 1

LLPS

Liquid–liquid phase separation

MAPK

Mitogen‐activated protein kinase

NPR

NONEXPRESSOR OF PR GENES

PAMP

Pathogen‐associated molecular pattern

PcG

Polycomb group

Phy

Phytochrome

PHR

Photolyase‐homologous region

PIF3

PHYTOCHROME‐INTERACTING FACTOR 3

PRC

POLYCOMB REPRESSIVE COMPLEX

PTM

Post‐translational modification

PWO1

PWWP‐DOMAIN INTERACTOR OF POLYCOMBS 1

UBP1

OLIGOURIDYLATE‐BINDING PROTEIN 1

PAB

POLY (A)‐BINDING PROTEIN

PB

Processing body

PrD

Prion‐like domain

RBP

RNA‐binding protein

RRM

RNA recognition motif

SA

Salicylic acid

SAP3

STRESS‐ASSOCIATED PROTEIN 3

SAR

Systemic acquired resistance

SG

Stress granule

TAD

Topologically associating domain

TCP1

TCP DOMAIN PROTEIN 1

TZF

TANDEM ZINC FINGER

TZP

TANDEM ZINC KNUCKLE PROTEIN

VOZ2

VASCULAR PLANT ONE‐ZINC FINGER 2

In need of answers.

• How is LLPS spatiotemporally organized? How do cellular environments influence disordered proteins and thereby condensate formation?

• How is composition of condensates determined? How can we analyze the composition of biomolecules in condensates?

• How do different liquid condensates interact each other? Are there specific players that mediate interactions between condensates?

• Does LLPS define transcription hotspots related to epigenetic modification and 3D genome conformation? What molecular mechanisms are involved in the formation of those gene expression condensates in plants?

• Are there plant‐specific biological functions regulated by LLPS? Do plants have their own unique strategies for biogenesis, maintenance, and turnover of liquid condensates? Is there any evolutionary evidence providing insight into the unique features of LLPS in plants?

• How can we manipulate LLPS to artificially generate biochemical reaction hotspots in plant cells?

Introduction

Eukaryotes have evolved elaborate strategies to organize biochemical reactions in living cells in a spatiotemporal manner. Cells utilize membrane‐enclosed organelles to selectively isolate biochemical reactions into confined compartments, ensuring reaction efficiency and cell functionality. In contrast to membrane‐enclosed organelles, liquid‐like properties of membraneless condensates enable a rapid assembly, disassembly, and concentration of cellular components, facilitating the dynamic formation of local reaction centers with spatiotemporal specificity. The P granules, which function predominantly in RNA metabolism, were the first identified membraneless condensates with liquid‐like properties (Brangwynne et al, 2009). Subsequently, additional liquid‐like condensates have been discovered in both nucleus and cytoplasm, such as nucleoli (Montgomery, 1898; Brangwynne et al, 2011) and stress granules (Nover et al, 1983; Patel et al, 2015). The observation that also other membraneless condensates, such as processing bodies (PBs) (Sheth & Parker, 2003) and Cajal bodies (Cajal, 1903), exhibit liquid‐like properties suggests that liquid‐like condensate formation might be a general paradigm of cellular organization.

Liquid condensates are different from irreversible, non‐physiological, insoluble protein aggregates, which usually form because of protein misfolding and exhibit solid‐like properties (Kaganovich et al, 2008; Patel et al, 2015). Liquid condensates usually exhibit liquid‐like behavior both in vitro and in vivo (Brangwynne et al, 2011), whereby molecular constituents efficiently diffuse within the condensates (Phair & Misteli, 2000; Weidtkamp‐Peters et al, 2008), as shown in fluorescence recovery after photobleaching (Freeman Rosenzweig et al, 2017). Because of their liquid‐like properties, cellular condensates interact with themselves to efficiently coordinate many cellular functions. Condensates also form two‐dimensional pseudocondensates by physically associating with membranes of typical organelles (Banjade & Rosen, 2014), expanding the repertoire of cellular reactions and signaling. The dynamic nature of rapid assembly and disassembly of chemical reaction centers enables the fine‐tuning of biochemical reactions on demand.

The formation of cellular reaction hotspots is particularly important in biology. While the vast majority of research on this topic has been conducted using yeast and animal models, striking examples of plant condensates are starting to emerge. Based on its dynamic property, liquid–liquid phase separation (LLPS) is strongly linked to stimulus‐responsive biological functions in plants. Here, we review liquid‐like condensates in plants and illustrate how the reversible formation of condensates regulates plant responses to internal and external stimuli. Additionally, we discuss challenges facing the elucidation of molecular mechanisms underlying the organization and composition of cellular hotspots and elaborate future perspectives.

Molecular control of LLPS

Most liquid condensates are composed of a heterogeneous mixture of cellular macromolecules and depend on the local concentration of the condensate components that is determined by the number and affinity of multivalent molecules (Fig 1). A representative type of multivalent molecules is nucleic acid, such as RNA and DNA, which contains multiple nucleic acid‐ and protein‐binding regions, and thereby contributes to condensate formation (Strom et al, 2017; Roden & Gladfelter, 2020). Moreover, proteins containing multiple modular interaction domains and/or disordered regions, which provide multiple weakly adhesive sequence elements, also play a primary role in LLPS (Lin et al, 2017).

Figure 1. Molecular mechanism of liquid–liquid phase separation (LLPS).

The presence of intrinsically disordered regions (IDRs) attracts multivalent interactions and induces LLPS. Post‐translational modifications (PTMs), such as phosphorylation, acetylation, methylation, and deamidation, can influence LLPS by inhibiting electrostatic and other types of interactions, thus serving as on/off switches. Multiple environmental factors including ionic strength, pH, and temperature can also affect LLPS by inducing changes in various types of interactions among macromolecules.

A fascinating group of proteins involved in LLPS comprises the intrinsically disordered proteins (IDPs), which usually contain intrinsically disordered regions (IDRs), low‐complexity sequence domains, or prion‐like domains (PrDs) that exhibit a compositional bias for a small subset of polar and charged amino acids (Cascarina et al, 2020). IDPs are disordered and lack a well‐defined 3D structure either throughout their full length or within IDRs (Habchi et al, 2014). Conformational flexibility enables multivalent protein–protein and protein–nucleic acid interactions (Bah & Forman‐Kay, 2016).

Multivalent interactions decrease the solubility of condensate components because of entropy‐driven effects, promoting LLPS (Lin et al, 2015). At a critical threshold, proteins associate to form a highly viscous phase with ~10‐ to 100‐fold higher protein concentration than that outside this phase; however, when the protein concentration drops below the critical threshold, these compartments disassemble (Alberti, 2017).

Interactions between macromolecules that drive LLPS are based on weak chemical forces such as hydrophobic interactions (Murthy et al, 2019). Additional chemical forces are also involved in the sensitive control of LLPS, such as electrostatic interactions between positively charged amino acid residues (arginine, lysine, and histidine) in IDPs and negatively charged residues in other proteins or in the phosphate backbone of nucleic acids (DNA and RNA) (Murthy et al, 2019), and hydrogen bonds in proteins or accessible nucleotides (Murthy et al, 2019). These interactions are influenced by the environmental parameters of the cell, such as ionic strength, pH, redox state, and temperature (Nott et al, 2015; Onuchic et al, 2019; Adame‐Arana et al, 2020). Additionally, post‐translational modifications (phosphorylation, acetylation, methylation, and deamidation) of amino acid residues also alter the electrostatic interactions among macromolecules (Bah & Forman‐Kay, 2016), collectively determining the critical threshold for LLPS (Fig 1).

In biological systems, energy‐consuming, active processes are also associated with biogenesis, maintenance, and turnover of liquid condensates. For example, RNA‐dependent DEAD‐box ATPases (DDXs) regulate the assembly and turnover of RNA‐containing condensates in both prokaryotes and eukaryotes. DDXs in their ATP‐bound conformation act as scaffold proteins and recruit specific RNAs, establishing LLPS (Hondele et al, 2019). By contrast, ATP hydrolysis induces condensate turnover and RNA release (Hondele et al, 2019), demonstrating that DDXs catalyze reversible LLPS, depending on cellular needs. Together, multiple routes driven by physicochemical and biological activities underlie reversible LLPS possibly to facilitate the dynamic control of cellular metabolism and signaling.

Key proteins responsible for the formation of biological condensates in plants

Although IDPs have been extensively investigated in animal models, the IDPs are most likely conserved in living organisms. Transient, dynamic formation of condensates is particularly important for rapid cellular responses to changing environmental conditions. Hence, it has been suspected that as the sessile organisms, plants essentially require environment‐responsive LLPS for intuitively sensing the environmental changes and subsequent signal transduction (Rahman et al, 2013; Covarrubias et al, 2017; Fang et al, 2019; Jung et al, 2020).

A computational approach identified 474 genes encoding PrD‐containing proteins in the Arabidopsis thaliana genome (Chakrabortee et al, 2016). Gene ontology (GO) enrichment analysis showed that Arabidopsis PrD‐encoding genes are particularly enriched in biochemical processes, such as transcription control, RNA binding, and RNA metabolic processes, as well as cellular activities implicated in non‐membrane‐bound organelles. In addition, several biological processes, including flower development and reproduction, are presumably regulated by PrD‐containing proteins and thus potentially by LLPS (Chakrabortee et al, 2016).

Consistent with the multivalent nature of RNAs in LLPS, mRNAs can alter the biophysical properties of droplets, such as viscosity, fusion propensity, and condensate component exchange rates (Zhang et al, 2015). Furthermore, the N⁶‐methyladenosine modification of mRNAs further increases LLPS, indicating that RNAs play a fundamental role in the physiological condensate assembly (Ries et al, 2019). Accordingly, RNA‐binding proteins (RBPs) are rich in condensates and possibly contribute to the organization of cellular condensates in plants (Weber et al, 2008; Jain et al, 2016; Hubstenberger et al, 2017; Fang et al, 2019). A total of 2,701 potential RBPs with 836 reliable RNA‐binding domains (false discovery rate < 5%) were recently identified in Arabidopsis (Marondedze, 2020). GO enrichment analysis of Arabidopsis RBPs reveals that not only RNA processing and RNA‐binding activity (Marondedze, 2020), but also non‐membrane‐bound organelles category is overrepresented (Marondedze, 2020), indicating that plant RBPs are associated with LLPS.

Comparative analysis of publicly available genome sequences of plants, ranging from green algae to angiosperms, can be used to trace the evolution of LLPS. Analysis on the complete genome sequences of multiple species, including unicellular green algae, mosses, lycophytes, liverworts, and angiosperms, has shown that PrD‐containing proteins and RBPs are well conserved across diverse plant species (Alba & Pages, 1998; Kerner et al, 2011; Yruela et al, 2018; Nowacka et al, 2019). These observations suggest that LLPS is potentially conserved across plant lineages and might be generally implicated in cellular functions.

LLPS is pervasively observed in subcellular regions, including the nucleus, cytoplasm, and chloroplasts, and mediates versatile repertories of cellular processes. The dynamic nature of LLPS is particularly relevant to the stimulus‐responsive organization of cellular reaction hotspots in plants. The molecular composition of liquid‐like condensates also varies with the external environment, in terms of the number and identity of constituents (Decker & Parker, 2012; Cuevas‐Velazquez & Dinneny, 2018; Meyer, 2020). Therefore, here, we emphasize the involvement of LLPS in rapid responses to internal and external stimuli and provide a mechanistic basis for LLPS‐regulated biological processes in plants. We also cover several cellular foci, which are expected to potentially have liquid‐like property, and introduce their biological functions with the emphasis on molecular interactions among condensate components.

Phase separation in the cytoplasm

Stress granules (SGs)

SGs are eukaryotic condensates that comprise proteins and RNAs, and typically function in translational control through the storage, protection, or degradation of proteins and mRNAs, especially under stress conditions (Kosmacz et al, 2019). SGs are responsive to diverse environmental stresses, and different molecular factors are likely involved in stress‐specific SG formation. OLIGOURIDYLATE‐BINDING PROTEIN 1 (UBP1), an SG component with three RNA recognition motifs, is implicated in the reversible formation of hypoxia‐responsive SGs (Sorenson & Bailey‐Serres, 2014). UBP1C constitutively binds to a subpopulation of mRNAs with uracil‐rich 3'‐untranslated regions (3'‐UTRs) under normoxic conditions (Sorenson & Bailey‐Serres, 2014). However, during hypoxia, UBP1C associates more with non‐uracil‐rich mRNAs and aggregates into cytoplasmic SGs, globally repressing the translation of mRNAs to conserve energy. Upon reoxygenation, SGs rapidly disaggregate, returning the sequestered mRNAs to polysomes (Sorenson & Bailey‐Serres, 2014). Consistently, mutation of UBP1C interferes with seedling establishment and reduces plant survival under hypoxic conditions.

The formation of heat‐responsive SGs, which perform distinct molecular functions, is regulated by different genetic factors. Another component of SGs, UBP1b, localizes to cytoplasmic foci upon exposure to heat stress in Arabidopsis. UBP1b‐associated SGs prevent the degradation of mRNAs involved in heat shock and stress responses, such as DnaJ and stress‐associated protein 3 (SAP3) (Nguyen et al, 2016) (Fig 2A). Compared with the wild type, UBP1b overexpression lines exhibit increased heat tolerance and higher expression of a number of heat‐inducible genes, whereas ubp1b mutants exhibit greater sensitivity to heat stress (Nguyen et al, 2016). Another Arabidopsis SG component, VASCULAR PLANT ONE‐ZINC FINGER 2 (VOZ2), also participates in heat tolerance. VOZ2 is dispersed throughout the nucleus and cytoplasm under normal growth conditions; however, under heat stress, the nuclear‐localized VOZ2 is rapidly degraded via the ubiquitin/proteasome pathway, sequestering the VOZ2 protein to cytosolic SGs (Koguchi et al, 2017). Since VOZ proteins function as transcriptional repressors of DEHYDRATION‐RESPONSIVE ELEMENT‐BINDING PROTEIN 2A (DREB2A) in the nucleus (Koguchi et al, 2017), VOZ2 sequestration enhances heat tolerance and up‐regulates DREB2A expression (Koguchi et al, 2017).

Figure 2. Functional diagrams of cytosolic condensates in plants.

(A) Heat‐responsive stress granules (SGs). UBP1b inhibits the degradation of DnaJ and SAP3 mRNAs, activating gene expression. Heat shock recovery releases ribosomal protein (rProtein)‐encoding transcripts from the SGs, possibly through HSP101, and promotes translation. (B–D) Processing bodies (PBs). Dark‐induced PBs containing DCP5 attenuate the translation of mRNAs involved in photomorphogenesis (B). Upon ethylene perception, EIN2 associates with the 3'‐UTRs of EBF1 and EBF2 transcripts and sequesters them within PBs, derepressing ethylene responses (C). The PB component TZF9 is phosphorylated by MAPKs upon PAMP recognition, releasing defense‐related mRNAs from PBs to trigger the immune response (D). (E) NPR1 condensate. Salicylic acid (SA)‐induced NPR1 monomer forms cytosolic condensates, together with CUL3–E3 ligase complexes, and ubiquitinates effector‐triggered immunity (ETI) proteins, such as NB‐LRRs, to inhibit the hypersensitive response (HR).

During heat shock, mRNAs encoding ribosomal proteins are preferentially stored in SGs (Kosmacz et al, 2019). Upon heat shock recovery, efficient release of ribosomal protein‐encoding mRNAs from SGs is required for the rapid restoration of the protein translation machinery (Merret et al, 2017). HEAT SHOCK PROTEIN 101 (HSP101) mediates the release of mRNAs to produce ribosomes and thereby enhance translation (Merret et al, 2017), possibly through interaction with the 26S proteasome complex (McLoughlin et al, 2019) (Fig 2A). Consistently, the Arabidopsis hsp101 mutant shows defects in the recovery of protein translation and dissociation of SGs after heat shock (Merret et al, 2017), with an increase in the amount of ubiquitinated proteins (McLoughlin et al, 2019). Overall, a variety of mRNA metabolic processes are controlled via the formation of stress‐responsive condensates.

Processing bodies (PBs)

PBs contain specific molecular factors responsible for RNA processing, such as decapping enzymes, deadenylation enzymes, and exoribonucleases, and participate in the degradation and translational arrest of mRNAs (Decker & Parker, 2012). Several plant PB components have been characterized as crucial regulators of stimulus‐responsive condensate formation (Maldonado‐Bonilla, 2014). For example, an Arabidopsis PB protein, TANDEM ZINC FINGER 1 (TZF1), is induced by darkness and wounding stress, and establishes TZF1‐associated cytoplasmic foci in actively growing tissues and stomatal precursor cells (Pomeranz et al, 2010), allowing the processing of mRNAs containing AU‐rich elements in their 3'‐UTRs, probably to enhance stress responses.

Light‐dependent developmental programs also require PB activity. Dark‐inducible PBs attenuate the translation of mRNAs involved in photomorphogenesis, thereby promoting skotomorphogenesis (Jang et al, 2019). The DECAPPING 5 (DCP5) protein induces PB formation in darkness to attenuate the translation of specific mRNAs involved in photosynthesis and light‐responsive plant development, such as protochlorophyllide synthesis‐related transcripts and PIN‐LIKE 3 (PIL3) (Jang et al, 2019). This is supported by the phenotypes of the Arabidopsis dcp5‐1 mutant, which exhibits PB formation defects, compromised mRNA processing (Xu & Chua, 2009), and reduced plant fitness (Jang et al, 2019) (Fig 2B). The dark‐inducible PB formation is facilitated in part by CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) in dark‐grown seedlings (Jang et al, 2019). The dark‐grown cop1‐6 mutants exhibit open cotyledons and short hypocotyls, with a low number of PBs, similar to light‐grown wild‐type seedlings (Jang et al, 2019). Meanwhile, light perception by photoreceptors may repress the action of COP1 and reduce the accumulation of PBs, releasing the translationally stalled mRNAs for active translation (Jang et al, 2019).

The control of ethylene signaling also depends on PB function. The endoplasmic reticulum (ER) membrane‐bound ETHYLENE‐INSENSITIVE 2 (EIN2) protein and two F‐box proteins, EIN3‐BINDING F‐BOX 1 (EBF1) and EBF2, are implicated in PB‐dependent ethylene signaling. Upon ethylene perception, EIN2 associates with 3'‐UTRs of EBF1 and EBF2 mRNAs, together with PB components, EIN5 and POLY(A)‐BINDING PROTEINS (PABs), and induces PB localization of the EBF mRNAs. Consequently, the translation of EBF1 and EBF2 mRNAs is hindered, and the target substrates of EBF F‐box proteins, such as EIN3 and EIN3‐LIKE 1 (EIL1), are derepressed to stimulate ethylene responses (Li et al, 2015) (Fig 2C).

In addition, the reversible formation of PBs is linked to pathogen‐associated molecular pattern (PAMP)‐triggered defense mechanisms. TZF9, which comprises a ribonucleoprotein complex, sequesters specific transcripts within PBs to inhibit the translation of subsets of defense‐related mRNAs under normal conditions (Tab assum et al, 2020). Upon PAMP recognition, the PB component TZF9 is phosphorylated by two PAMP‐responsive mitogen‐activated protein kinases (MAPKs), MPK3 and MPK6. The phosphorylation of TZF9 by MAPKs reduces its RNA‐binding activity and concomitantly leads to the disassembly of TZF9‐associated PBs (Maldonado‐Bonilla et al, 2014; Tabassum et al, 2020). Consequently, the translationally arrested defense‐related mRNAs are liberated from PBs, triggering PAMP‐induced immune responses (Tabassum et al, 2020) (Fig 2D).

Cytoplasmic liquid condensates and foci, which are frequently driven by RNA molecules as well as RBPs, are mostly involved in mRNA metabolism, such as mRNA storage, degradation, decapping, turnover, and translational control (Protter & Parker, 2016; Luo et al, 2018; Garcia‐Jove Navarro et al, 2019). Consistent with their similar biochemical functions, communication between cytoplasmic condensates has also been demonstrated. PBs are functionally linked to other mRNA‐containing cytoplasmic condensates, such as SGs, to control mRNA fate (Stoecklin & Kedersha, 2013). In Arabidopsis, the PB component TZF1 is also associated with SGs (Pomeranz et al, 2010), and TZF4/5/6 co‐localize with both PB and SG markers (Berry et al, 2013; Bogamuwa & Jang, 2013). Additionally, TZF9, another PB component, interacts with and promotes the recruitment of PABs to the SGs not only under stress, but also under normal conditions (Tab assum et al, 2020). Different types of cytoplasmic condensates thus synergistically exert translational control and fine‐tune plant growth and development according to environmental conditions.

LATE EMBRYOGENESIS ABUNDANT (LEA) body

LEA proteins, which confer tolerance to water deficit, are highly conserved throughout the plant kingdom (Cuevas‐Velazquez et al, 2016). Notably, the Group 4 LEA proteins of Arabidopsis are disordered in aqueous solution, consistent with strongly disordered regions at their C‐terminal regions, and exhibit high potential to expose an α‐helix in vitro (Cuevas‐Velazquez et al, 2016). While under normal conditions, the conformational transition is inhibited by the N‐terminal region of LEA proteins, dehydration stress induces conformational changes in LEA proteins, thereby exposing the α‐helix, with the help of chaperone‐like activity (Cuevas‐Velazquez et al, 2016). Whether this conformational change is required for the function of LEA proteins in drought tolerance remains to be determined. However, the observation that the brine shrimp Artemia franciscana LEA6 protein promotes desiccation tolerance in a LLPS‐dependent manner (Belott et al, 2020) opens the possibility that structural plasticity of hydrophilin proteins could have been evolved to improve plant tolerance to environmental stresses (Cuevas‐Velazquez et al, 2016).

AUXIN RESPONSE FACTOR (ARF) body

A subset of ARF proteins, class A activating ARFs, are key signaling mediators of auxin responses. These ARFs contain an intrinsically disordered middle region (MR) and a PHOX and BEM1 (PB1) domain, both of which are potentially important for extensive non‐stoichiometric protein assembly (Powers et al, 2019). ARF7 and ARF19 form liquid condensates in the cytoplasm, especially in tissues with attenuated auxin responsiveness, such as mature sections of the upper root (Powers et al, 2019). By contrast, cytoplasmic assemblies are reduced, and ARFs are localized to the nucleus in actively growing cells near the root meristematic regions. The MR of ARF19 is sufficient for condensate formation, and its replacement with the MR of another ARF belonging to different classes disrupts the ARF body (Powers et al, 2019). Likewise, mutation of a single lysine residue within the PB1 domain also diminishes cytoplasmic assemblies, promoting ARF nuclear localization and altering auxin responsiveness in plants (Powers et al, 2019). These observations imply that phase separation determines the cellular competence for auxin signaling, although the molecular factors that determine the auxin‐dependent LLPS of ARFs remain elusive.

NONEXPRESSOR OF PATHOGENESIS‐RELATED GENES 1 (NPR1) body

Eukaryotic cells have evolved innate immune systems, and effector‐triggered immunity (ETI) is a representative response induced upon the recognition of pathogen virulence effectors at infection sites. During ETI, infected cells turn on programmed cell death (hypersensitive response [HR]) through the activation of the NB‐LRR‐EDS1/PAD4‐WRKY54/70 signaling cascade (Aarts et al, 1998). Simultaneously, an increase in salicylic acid (SA) content leads to the dephosphorylation of the transcriptional co‐activator NPR1, which converts the oligomeric form of NPR1 into a monomer. The NPR1 monomer then enters the nucleus, where it is sumoylated and consequently degraded by the Cullin‐RING‐ligase CRL3^NPR3/4 complex, promoting HR in infected cells (Spoel et al, 2009; Fu et al, 2012).

By contrast, in neighboring cells, where the cell survival program is promoted, NPR1 is suspected to undergo different post‐translational modifications regulated possibly by the low pathogen load. As a result, SA‐activated nuclear NPR1 interacts with TGA transcription factors to induce the expression of systemic acquired resistance genes. Moreover, NPR1 contains IDRs, further establishing cytosolic NPR1 condensates, along with the CULLIN 3 (CUL3)–E3 ligase complex, in neighboring cells. The NPR1 body sequesters and ubiquitinates HR regulators, such as NB‐LRRs, EDS1, and WRKY54/70 (Zavaliev et al, 2020), attenuating programmed cell death (Fig 2E). Overall, NPR1 is the main factor that determines cell death or cell survival in plant immunity by modulating its conformation (Zavaliev et al, 2020).

Phase separation in the nucleus

FLOWERING LOCUS CA (FCA) body

Several proteins involved in the autonomous pathway that controls flowering time are known to form condensates in the nucleus (Fang et al, 2019). Interestingly, luminidependens, FLOWERING LOCUS PA (FPA), FLOWERING LOCUS Y (FY), and FCA proteins contain PrDs and establish higher‐order polymer‐containing nuclear condensates (Chakrabortee et al, 2016). These proteins most likely work together, as FY‐ and FPA‐containing nuclear bodies fully overlap with FCA bodies (Fang et al, 2019). Based on a recent study on the FCA LLPS, a detailed working mechanism has been unraveled. The FCA body contains additional proteins including the PrD‐containing coiled coil protein FLX‐LIKE 2 (FLL2), as well as 3'‐RNA processing components (Fang et al, 2019). The FCA condensate promotes proximal polyadenylation, possibly through optimal organization of the polyadenylation hotspot at specific poly‐A sites (Fig 3A). FLL2 is particularly essential for the LLPS of the FCA body. Ectopic expression of FLL2 increases the size and number of FCA nuclear bodies and enhances proximal polyadenylation, whereas the FLL2 mutation (sof78) reduces FCA body formation without affecting the FCA protein level (Fang et al, 2019). Taken together, the FCA body is a collection of RBPs and 3'‐RNA processing machineries and facilitates nuclear RNA processing for gene expression control.

Figure 3. Nuclear condensates in plants.

(A) FCA body. The phase‐separated FCA body enhances polyadenylation by compartmentalizing 3'‐end processing factors. The FCA body is further regulated by FLL2 and other autonomous proteins. (B) Photobody. Left: Upon blue light exposure, the active forms of CRYs enter the nucleus and form a photobody, together with core light signaling proteins, such as ELF3, COP1, SPA1, and HY5. Right: The phyB‐associated photobody induces the phosphorylation and consequently the degradation of the PHYTOCHROME‐INTERACTING FACTOR 3 (PIF3) protein. (C) ELF3 body. At low temperature, ELF3 forms the Evening Complex, along with ELF4 and LUX, and acts as a transcriptional repressor. At high temperature, the PrD‐containing ELF3 is assembled into nuclear bodies. The temperature‐responsive LLPS of ELF3 prevents its binding to EC‐target genes, thus allowing gene expression. (D) PcG body. PRC2 interacts with PWO1, EMB1579, CUL4, DDB1, and MSI4 to form hotspots for establishing H3K27me3. The IDR‐containing LHP1 protein also binds to PcG‐deposited H3K27me3 and contributes to the formation of PcG body.

Photobody

Upon excitation by red/far‐red light, phytochromes are rapidly translocated from the cytoplasm to the nucleus (Sakamoto & Nagatani, 1996). Red light predominantly leads to the nuclear translocation of phytochrome B (phyB), whereas far‐red light induces phyA nuclear translocation (Yamaguchi et al, 1999; Kim et al, 2000). Soon after the nuclear import, photoreceptors, together with transcription factors and E3 ubiquitin ligases involved in photomorphogenesis, form subnuclear photobodies (Ang et al, 1998; Stacey & von Arnim, 1999; Seo et al, 2003; Bauer et al, 2004) (Fig 3B, right). The photobody stabilizes the active form of phytochromes and induces the phosphorylation and consequently the degradation of PHYTOCHROME‐INTERACTING FACTOR 3 (PIF3) (Chen et al, 2010). Furthermore, phyB recruits additional light‐responsive factors to the photobody, including TANDEM ZINC KNUCKLE PROTEIN (TZP), a key regulator of photoperiodic flowering (Kaiserli et al, 2015), ensuring prolonged light responses for the control of cell elongation, flowering time, and circadian oscillation (Chen, 2008; Kaiserli et al, 2015; Ronald & Davis, 2019).

Cryptochromes (CRYs) mediate photoperiodic promotion of floral initiation and blue light‐mediated inhibition of hypocotyl elongation. Photoexcitation disengages the disordered domain of the C‐terminal extension of CRY from the light‐sensing N‐terminal photolyase‐homologous region domain, inducing the unfolding of the tail to make the CRY proteins active (Gu et al, 2012). Upon exposure to blue light, the active conformation of CRYs forms nuclear photobodies that are distinct from phyB‐containing photobodies (Fig 3B, left) (Yu et al, 2009). In the nuclear body, CRY1/2 are stabilized, and the light‐induced disordered state of CRYs allows their interaction with other proteins such as ELONGATED HYPOCOTYL 5 (HY5), EARLY FLOWERING 3 (ELF3), COP1, and SUPPRESSOR OF PHYA‐105 1 (SPA1), to promote the blue light signal transduction pathway (Ang et al, 1998; Yu et al, 2008; Lian et al, 2011; Liu et al, 2011; Zuo et al, 2011). Consistently, several CRY‐interacting proteins also possess IDRs, as exemplified by the HY5 protein, which has disordered domains within the N‐terminal region (Yoon et al, 2006). Overall, photobodies are pivotal cellular hotspots that robustly induce light signal transduction by establishing light signaling clusters.

Temperature sensing condensates

Warm temperatures induce thermomorphogenesis through a signaling pathway that overlaps with shade avoidance signaling. Similar to the destabilization of phyB in darkness, high temperatures trigger phyB disassembly from photobodies (Hahm et al, 2020). The thermostability of photobodies depends on the photosensory N‐terminal module of the phyB N‐terminal domain, while the C‐terminal module of phyB is not sufficient for thermo‐responsiveness. At high temperature, phosphorylation of Ser‐86 residue in the N‐terminal domain of phyB enhances the thermal reversion rate of Pfr‐Pr heterodimer, inducing photobody disassembly (Medzihradszky et al, 2013; Viczian et al, 2020). Thus, the mechanism of early signaling to induce temperature responses may involve dynamic assembly/disassembly of photobodies (Hahm et al, 2020).

A more intuitive temperature sensing mechanism has been demonstrated in Arabidopsis. The PrD of ELF3 undergoes thermoresponsive LLPS and serves as a tunable thermosensor (Jung et al, 2020). At low temperature, ELF3 is diffused in the nucleus and forms a tripartite Evening Complex (EC) with ELF4 and LUX ARRHYTHMO (LUX), which represses the transcription of various genes involved in cell elongation, photosynthesis, stress responses, and the circadian clock (Nusinow et al, 2011; Chow et al, 2012; Mizuno et al, 2014; Silva et al, 2020). However, at high temperatures, ELF3 dissociates from the target chromatin and forms nuclear bodies in a PrD‐dependent manner (Jung et al, 2020). High temperature‐dependent LLPS of ELF3 can occur by itself. Purified ELF3 protein aggregates in vitro, and heterologous expression of ELF3 in the yeast model system that lacks EC homologs leads to LLPS at high temperatures (Jung et al, 2020), indicating an entropy‐driven ratiometric LLPS response of ELF3 to temperature (Fig 3C). The LLPS‐linked temperature sensing mechanism has evolved, depending on the natural habitat of the plant species. The ELF3 protein of Brachypodium distachyon, which is habituated to warmer climates, is not predicted to have a PrD region. ELF3 variants with undetectable PrDs are unable to respond to warm temperatures. By contrast, the ELF3 proteins of Arabidopsis and potato (Solanum tuberosum), which grow in temperate climates, possess PrDs for temperature responsiveness (Jung et al, 2020). These data suggest that the LLPS of ELF3 has evolved as an adaptive thermosensory mechanism, which might be conserved in other living organisms.

Nuclear condensates for landscaping the chromatin structure

There is increasing evidence that plant genomes possess unique local and global chromatin conformation. It is noteworthy that local chromatin configurations, especially closed chromatin structures caused by histone H3 lysine 9 methylation (H3K9me) and/or H3 lysine 27 methylation (H3K27me), involve LLPS. The Polycomb group (PcG) body is a representative condensate that inhibits gene expression via the formation of silenced chromatin conformation. The POLYCOMB REPRESSIVE COMPLEX 2 (PRC2) forms nuclear speckles, possibly reminiscent of the PcG body, and recruits a variety of proteins, including PWWP‐DOMAIN INTERACTOR OF POLYCOMBS 1 (PWO1) (Hohenstatt et al, 2018), EMBRYO DEFECTIVE 1579 (EMB1579), CUL4, DNA DAMAGE‐BINDING PROTEIN 1 (DDB1), and MULTIPLE SUPPRESSOR OF IRA 4 (MSI4), to efficiently establish the H3K27me3 mark (Zhang et al, 2020). The spatial organization of the PcG body may be further influenced by DNA‐binding factors. Consistently, the PcG‐interacting protein, VERNALIZATION 1 (VRN1), which contains two B3 DNA‐binding domains flanked by an IDR, binds to DNA sequences and undergoes LLPS (Zhou et al, 2019). Then, PcG‐deposited H3K27me3 is recognized by the IDR‐containing LIKE HETEROCHROMATIN PROTEIN 1 (LHP1), which also interacts with PRC1 and PRC2, so that the PcG target genes are stably repressed, probably in the PcG body (Xu & Shen, 2008; Veluchamy et al, 2016; Berry et al, 2017) (Fig 3D). In parallel, a functional homolog of the animal HETEROCHROMATIN PROTEIN 1, AGENT DOMAIN‐CONTAINING PROTEIN 1, acts as a multivalent H3K9me reader that recognizes and associates with H3K9me2‐marked heterochromatin, organizes heterochromatic chromocenters, and facilitates transposon silencing through H3K9me‐dependent LLPS (Zhao et al, 2019). Collectively, these observations indicate that heterochromatin formation in plants is facilitated at least in part by a conserved LLPS mechanism.

Plant genomes have extensive 3D long‐range interactions, which might result from the demixing of genomic compartments. The 3D interaction genomic unit that is conserved in eukaryotes, topologically associating domain (TAD), is frequently observed in nuclear speckles (Dong et al, 2017, 2019; Liu et al, 2017). Plant genomes are known to lack homologs of the CCCTC‐binding factor and insulators (Ong & Corces, 2009); instead, analysis of conserved sequence motifs at TAD borders has suggested that plant‐specific TCP DOMAIN PROTEINS (TCPs) presumably control 3D chromatin packing (Liu et al, 2017). Consistently, the TCP1 of Marchantia polymorpha has been proven to establish TAD structure (Karaaslan et al, 2020). Interestingly, a significant portion of plant TCPs contain IDRs (Pontvianne & Liu, 2020) and are located in nuclear speckles (Valsecchi et al, 2013; Kubota et al, 2017; Mazur et al, 2017), further supporting that LLPS contributes to the 3D genome configuration in plants.

Taken together, a variety of condensates organized possibly by LLPS are found in the nucleus. Nuclear condensates unequivocally regulate gene expression by controlling 3'‐RNA processing, transcription factor stability, transcription factor combinations, chromatin landscaping, and transcription factor–promoter–enhancer interactions. Further studies are needed to understand the mechanisms used within the nucleus to coordinate chromatin dynamics and genome activity in higher eukaryotes.

Phase separation in chloroplasts

Pyrenoid matrix

Phase separation assists in the formation of the pyrenoid matrix, a hotspot for CO₂ fixation found in the chloroplasts of algae and certain plants. The CO₂‐fixing enzyme Ribulose‐1,5‐bisphosphate (RuBP) carboxylase/oxygenase (Rubisco), which is involved in the first major steps in the Calvin cycle for the carboxylation of RuBP to produce 2 molecules of glycerate‐3‐phosphate, is packaged into highly stable heterocomplexes in the pyrenoid matrix with liquid‐like properties (Wunder et al, 2018). In the matrix, Rubisco proteins are regularly distanced to maximize their catalytic performance for carbon fixation in a variety of photosynthetic eukaryotes (Mackinder et al, 2016). Notably, the low‐complexity repeat protein, ESSENTIAL PYRENOID COMPONENT 1 (EPYC1), interacts with Rubisco and induces LLPS‐dependent pyrenoid biogenesis in C. reinhardtii (Freeman Rosenzweig et al, 2017) (Fig 4A). Five repeat regions of EPYC1 associate with the surface‐exposed α‐helices of the Rubisco small subunit to determine regular spacing and arrangement of Rubisco in the pyrenoid (Meyer et al, 2012; Mackinder et al, 2016; He et al, 2020). Consistently, EPYC1 colocalizes with Rubisco throughout the pyrenoid and is essential for normal pyrenoid size, number, and morphology, Rubisco content, and efficient carbon fixation at low CO₂ (Mackinder et al, 2016; Atkinson et al, 2020).

Figure 4. Liquid–liquid phase separation in chloroplasts.

(A) Pyrenoid of C. reinhardtii. The pyrenoid matrix exhibits liquid‐like properties through multivalent interactions between Rubisco and EPYC1. EPYC1 has five Rubisco‐binding regions, and Rubisco has eight binding sites for EPYC1. The multivalent interactions allow regular spacing of Rubiscos in pyrenoid matrix to maximize CO₂ fixation and photosynthetic efficiency. (B) STT body. STT1 and STT2 form heterodimers through C‐terminal ankyrin domain interaction. N‐terminal intrinsically disordered regions of STT proteins enable LLPS along with cpTat pathway substrates, such as OE23. After the STT body moves to thylakoid membrane, Hcf106 allows reverse phase separation and the release of OE23 to translocate across thylakoid membrane.

STT body

Phase separation emerged as a mechanism driving intra‐chloroplast cargo sorting based on their signal peptides (Ouyang et al, 2020). As one of thylakoid protein translocation pathways, chloroplast twin‐arginine translocation (cpTat) pathway comprises three membrane proteins, Tha4, Hcf106, and cpTatC, and translocates folded proteins into the lumen. Notably, cpTat substrates are prone to form condensates in the stroma based on their hydrophobicity (Celedon & Cline, 2013), which is required for further targeting to correct subcompartment within the chloroplast (Robinson et al, 2000).

Two plant‐specific ankyrin proteins, STT1 and STT2, are essential for the recognition and sorting of the cpTat pathway substrates and their subsequent transport to the thylakoid cpTat translocon. The STT1 and STT2 proteins form heterodimers based on an interaction between their C‐terminal ankyrin domains. Moreover, the STT complex has intrinsically disordered N‐terminal regions, facilitating interactions with the signal peptide of cpTat substrates. The formation of phase‐separated STT‐substrate bodies enables their rapid movement from the chloroplast envelope to the thylakoid membrane (Ouyang et al, 2020) (Fig 4B). Upon arrival at the thylakoid membrane, the loosely structured carboxyl terminus‐2 domain of HIGH CHLOROPHYLL FLUORESCENCE 106 (Hcf106), which has conserved negatively charged residues, interacts specifically with STT proteins, allowing reverse phase separation for cargo targeting and translocation across thylakoid membranes (Ouyang et al, 2020) (Fig 4B). In support, the stt knockdown mutant lines exhibit selectively compromised transport of the cpTat pathway substrates and display defects in chloroplast development and plant growth (Ouyang et al, 2020). Formation of liquid droplets is not limited to chloroplasts but may be broadly exploited as an effective strategy for cargo sorting in crowded compartments. Considering that key proteins involved in intra‐organelle protein sorting frequently possess IDRs (Bernstein et al, 1989; Schunemann, 2004; Bae et al, 2008; Grudnik et al, 2009; Skalitzky et al, 2011; Gonzalez‐Esquer et al, 2018; Barros‐Barbosa et al, 2019), LLPS might be a general way of protein translocation in plant organelles.

Phase separation at the membrane surface

Clustering of proteins into micrometer‐sized structures at membranes is observed in many signaling pathways (Cho & Stahelin, 2005). Although limited examples have been elucidated in plants, LLPS has been identified to promote actin assembly at lipid bilayers. The adhesion receptor Nephrin is attached to the membrane surface and forms micrometer‐sized protein clusters together with cytoplasmic partners, Nck and N‐WASP, through 2‐dimensional phase separation (Banjade & Rosen, 2014). Dynamics of the actin regulatory clusters are specifically modulated by interacting proteins. The ARABIDOPSIS RIBOSOMAL PROTEIN 2/3 (Arp2/3) complex promotes LLPS of the Nephrin/Nck/N‐WASP clusters and increases membrane residency time of the N‐WASP–Arp2/3 complex to increase actin assembly at the membrane (Case et al, 2019). Interactions between multivalent proteins could be a general mechanism employed by cytosolic adaptor proteins to organize membrane receptor clusters.

Biological relevance of LLPS

LLPS attracts multivalent constituents and instigates high concentrations of components in a confined region to ensure efficient chemical and biological reactions. Strikingly, the origin of life is presumably driven by the spontaneous assembly of phase‐separated open systems (Oparin, 1953). A coacervate is a droplet containing high concentrations of insoluble macromolecule complexes (Nott et al, 2015; Adame‐Arana et al, 2020), which allows unprecedented primitive metabolism and self‐replication of simple RNAs (Oparin, 1953).

Consistent with the observation of LLPS in primitive life, LLPS‐dependent control of biochemical reactions is well conserved in most living organisms, ranging from the ancient cyanobacteria to the most evolved mammalian cells (Wang et al, 2019). This is supported by the finding that molecular factors implicated in LLPS, such as IDPs and RBPs, are conserved across nearly all plant lineages, and similar working mechanisms are exploited to create cellular hotspots for efficient reactions. Indeed, crucial macromolecular metabolism, such as DNA replication, RNA transcription, mRNA processing and translation, and protein degradation, involves LLPS to maintain distinct biochemical properties of macromolecules from a surrounding phase (Decker & Parker, 2012; Boija et al, 2018; Fang et al, 2019; Kosmacz et al, 2019; Parker et al, 2019).

Given that plants need to rapidly adapt to changing environments, LLPS, which dynamically gives rise to reaction centers, becomes the central mechanism underlying plant responses to environmental changes. LLPS‐dependent formation of cellular hotspots enables intuitive sensing of the surrounding environment (Jung et al, 2020) as well as robust signaling and biochemical reactions in response to environmental changes (Zavaliev et al, 2020). Overall, LLPS provides an important clue about the pivotal cellular and molecular processes essential for the plant response to a given stimulus.

Future challenges and concluding remarks

LLPS provides the ability to organize diverse cellular processes by concentrating multivalent biomolecules into distinct compartments, when needed. Diverse cellular condensates are also orchestrated to generate appropriate cellular responses to internal and external cues. Given that plant LLPS is highly responsive to various environmental stimuli, such as light and mechanical forces, it is important to understand the plant‐specific mechanisms involved in condensate formation, maintenance, interactions, and turnover. More specifically, how and to what extent the cellular condensates act as sensors of environmental fluctuations will be interesting to elucidate.

The formation of liquid condensates is spatially controlled on demand. These condensates are concentrated in specific regions of the cell, as shown by the importance of positional enrichment of compartments in cell polarity and cell fate decision (Saha et al, 2016). The cellular components are likely distributed along a concentration gradient in the cell, leading to spatial asymmetry in condensate formation. Additionally, LLPS is used to sort cargo and transport substrates to the target subcellular position. Although several studies have shown that LLPS is spatially controlled in different cellular compartments, it remains unclear where the condensates are formed or deformed, and how the condensates are transported to specific locations; this is worthy of further exploration.

Cellular condensates carry out a variety of biochemical/biological functions. To gain insight into the biological functions of condensates, it is important to understand their molecular composition. In addition to classical approaches, such as antibody staining, mass spectrometry analysis, and fluorescent protein colocalization assays, proximity‐labeling approaches can address dynamic interaction networks of the plant proteome. Condensate‐enriched proteins can be fused to a catalytic enzyme that non‐specifically biotinylates adjacent proteins, which can then be purified by streptavidin‐based precipitation and analyzed by mass spectrometry (Markmiller et al, 2018). Analogous proximity‐ligation techniques can be used for interrogating the interactome in the vicinity of specific RNAs or genomic loci (Gao et al, 2018; Myers et al, 2018; Ramanathan et al, 2018).

Remarkably, unique features of LLPS provide fundamental insights into cellular organization, but also raise new fundamental questions and research opportunities in the field of plant biology. This research field is now beginning to grow, and many plant‐specific discoveries can demonstrate the novel biological relevance of condensates in living systems. This research area intrinsically requires extensive interdisciplinary approaches, and it will not only expand the plant research community but also significantly improve our understanding of plant‐specific LLPS.

Author contributions

PJS conceived the study. All authors participated in writing or revising the draft.

Conflict of interest

The authors declared that they have no conflict of interest.

EMBO reports (2021) 22: e51656.