Protein-mediated LLPS is highly dependent on protein concentration, ionic strength, pH, temperature, and the presence of different “clients” or “cargos”, such as other protein partners and/or nucleic acids (Elbaum-Garfinkle et al., 2015; Nott et al., 2015; Saha et al., 2016; Dignon et al., 2020). Proteins that drive LLPS are often characterized by intrinsically disordered regions (IDRs), which lack well-defined secondary and tertiary structure (Burke et al., 2015; Elbaum-Garfinkle et al., 2015; Kroschwald et al., 2018; Dignon et al., 2019). IDRs have a high degree of conformational flexibility and often exhibit multivalency, which is closely associated with the collective interactions between polypeptide chains (Pak et al., 2016). Accordingly, IDRs can drive phase separation via cation-pi, pi-pi, electrostatic attraction, and hydrophobic interactions (Tsang et al., 2020). It is worth noting that IDR-harboring proteins do not necessarily form condensates (Alberti et al., 2019). A subset of IDRs, called prion-like proteins or prion-like domains (PrD), are more frequently associated with phase separation, and are often enriched in asparagine and glutamine residues and bear sequence similar to yeast prion proteins. In addition to disordered proteins, LLPS can be initiated by folded domains such as coiled-coils (Ramšak et al., 2023)and via the formation of protein-RNA complexes mediated by RNA-recognition motifs (RRM) (Dignon et al., 2020; Ramšak et al., 2023).

Protein-mediated LLPS is commonly studied by diverse fluorescence-based techniques including Fluorescence Recovery After Photobleaching (FRAP) and Fluorescence Correlation Spectroscopy (FCS). FRAP is a method for determining the dynamics of fluorescently labelled proteins in a condensate. The fluorophore can be partially or fully photobleached and its fluorescence recovery curve examined. This analysis provides valuable insights into condensate dynamics, including the mobility and exchange of molecules within the condensates or between the condensate and the environment. Alternatively, FCS can analyze molecular dynamics in both in vitro and in vivo condensate phases, determining molecular concentration, oligomeric state, diffusion coefficient (hydrodynamic radius), and molecular interactions (Wang et al., 2022b). These techniques, while applicable in various research areas, are particularly instrumental in elucidating the behavior of biomolecular condensates, thus advancing our understanding of LLPS. Of note, based on FRAP and FCS experiments, many proteins, which initially form highly mobile liquid condensates, may become more visco-elastic and rigid as they “age” over time, forming gel-like states and fibrils (Molliex et al., 2015; Murakami et al., 2015; Patel et al., 2015; Wegmann et al., 2018; Ray et al., 2020). In animals, these fibrils have been linked to neurological diseases, including Alzheimer’s disease (Lin et al., 2015; Xiang et al., 2015; Pak et al., 2016). However, the physiological role of hydrogel/fibril formation in plants remains to be uncovered. However, the formation of gel and fibril states raises the intriguing question of whether or not this persistent state may act to establish a form of memory.

Environmental cues such as the prolonged exposure to low temperatures significantly influence the timing of flowering. The expression levels of the flowering repressor FLC time the onset of flowering in biennials and winter annual plants. The prolonged exposure to winter cold is required to repress FLC expression, allowing the plant to flower in the following spring. Vernalization, acting as an epigenetic memory of winter cold, determines the expression levels of FLC, synchronizing flowering to favorable spring conditions. High levels of FLC expression are maintained by the transcription factor FRI, resulting in suppression of flowering. Under low temperatures, the cold stabilization of FRI and its increased interaction with multiple factors drive the formation of FRI nuclear condensates, which requires its C-terminal IDR (Zhu et al., 2021). These nuclear condensates reduce FRI binding to the FLC locus via a sequestration mechanism, which consequently decreases FLC expression. Notably, FRI condensation is slowly reversible, dissipating within 5 hours upon returning to warmer conditions, suggesting that LLPS may act as a short-term memory of cold conditions (Zhu et al., 2021). Furthermore, noncoding RNAs may play a role in FRI LLPS, although there is some contradiction in the literature regarding their role in FRI condensation. Some evidence suggests that FRI condensation is promoted by cold-induced antisense RNA COOLAIR derived from the FLC locus (Zhu et al., 2021), while other studies indicate that condensation of FRI is independent of COOLAIR (Zhang et al., 2023).

The RNA-binding protein FLOWERING CONTROL LOCUS A (FCA), containing two PrDs, and the coiled-coil protein, FLL2, promote nuclear body formation with COOLAIR and antagonize the activity of FRI, resulting in reduced FLC transcription (Fang et al., 2019) ( Figure 1A ). In addition, VERNALIZATION 1 (VRN1), a two B3 domain DNA-binding protein that binds non-specifically to DNA, undergoes LLPS and is required for the stable repression of FLC (Levy et al., 2002). VRN1 condensates form on the FLC locus, inducing structural changes in the FLC chromatin and ensuring FLC silencing (Zhou et al., 2019) ( Figure 1A ). In summary, several LLPS events elicit the stable repression of FLC, contributing to the molecular memory of winter cold. Notably, FRI condensation, which is cold-dependent and slowly dissipating upon a shift to warmer conditions, suggests that LLPS may act as a direct mechanism for cold memory.

The transcriptional regulator ELF3 represents another example of LLPS in plant environmental response and possibly temperature memory establishment. ELF3 integrates information from the circadian clock, temperature, and light perception pathways to act as a key component in photoperiodic flowering (Ezer et al., 2017; Jung et al., 2020; Silva et al., 2020). It has no structured domains of known function and acts as a scaffold binding many TFs (Hicks et al., 2001; Nusinow et al., 2011; Herrero et al., 2012; Ezer et al., 2017; Silva et al., 2020). Upon exposure to higher ambient temperature, ELF3 undergoes reversible LLPS driven by its C-terminal PrD and fine-tuned by the length of polyglutamine regions in this domain (7 to 29 glutamines across 181 natural A. thaliana accessions) (Jung et al., 2020; Hutin et al., 2023). In vitro and in vivo FRAP experiments on ELF3 condensates demonstrate an aging process in which a low mobility hydrogel forms. In addition, in vitro temperature ramp experiments for ELF3 PrD reveal a delay in phase separation reversibility upon the return to lower temperatures, indicating a degree of hysteresis in the protein behavior (Hutin et al., 2023). This suggests that the system may maintain a memory of its past states. Indeed, in planta ELF3 forms nuclear condensates in response to increasing temperatures ( Figure 1D ). Their reversibility via cooling requires a more pronounced temperature shift or longer time compared to the warm temperature-induced LLPS. Consequently, as ELF3 remains in the inactive phase separated state, the regulation of its target genes is delayed, thereby potentially establishing a short-term memory of temperature (Murcia et al., 2022).

In response to hyperosmotic stress, the transcriptional regulator SEU forms condensates due to conformational changes within its N-terminal IDR. SEU condensation, which likely occurs via LLPS, is triggered by intracellular molecular crowding. It is vital for the stress response and the subsequent expression of osmotic stress-responsive genes (Wang et al., 2022a) ( Figure 1B ). Although SEU forms reversable condensates, FRAP analysis revealed their incomplete recovery, suggesting that they exhibit partial persistence with lower mobility or reduced turnover. This incomplete recovery hints at the possibility that SEU forms a gel-like state, indicating a potential mechanism for retaining a stress memory.

Additionally, FLOE1 is an PrD-containing protein that suppresses seed germination under unfavorable conditions (Dorone et al., 2021). It undergoes LLPS upon hydration, allowing the embryo to sense water ( Figure 1C ). Interestingly, its biological function is modulated by the biophysical state of its condensates, which is governed by the glutamine, proline, and serine rich (QPS) domain and the aspartic acid and serine rich (DS) domain. While the QPS domain is responsible for driving FLOE1 condensation and its function in water-sensing, the DS domain regulates the fluidity of its condensates. Natural variation in the DS domain across ecotypes promotes local adaptation by fine-tuning the fluidity of FLOE1 condensates (Dorone et al., 2021). Notably, FLOE1 condensates appear spontaneously and exhibit full reversibility through repeated hydration-dehydration cycles. However, deleting the DS domain leads to the formation of an irreversible hydrogel and promotes germination under stress, indicating that changes in the dynamics of FLOE1 condensates affect the stress response (Dorone et al., 2021).