Formation of subcellular compartments by condensation-prone protein OsJAZ2 in Oryza sativa and Nicotiana benthamiana leaf cells

OsJAZ2 protein has a propensity to form condensates, possibly by multivalent interactions, and can be used to construct artificial compartments in plant cells. Eukaryotic cells contain various membraneless organelles, which are compartments consisting of proteinaceous condensates formed by phase separation. Such compartments are attractive for bioengineering and synthetic biology, because they can modify cellular function by the enrichment of molecules of interest and providing an orthogonal reaction system. This study reports that Oryza sativa JAZ2 protein (OsJAZ2) is an atypical jasmonate signalling regulator that can form large condensates in both the nucleus and cytosol of O. sativa cells. TIFY and Jas domains and low-complexity regions contribute to JAZ2 condensation, possibly by multivalent interaction. Fluorescence recovery after photobleaching (FRAP) analysis suggests that JAZ2 condensates form mostly gel-like or solid compartments, but can also be in a liquid-like state. Deletion of the N-terminal region or the TIFY domain of JAZ2 causes an increase in the mobile fraction of JAZ2 condensates, moderately. Moreover, JAZ2 can also form liquid-like condensates when expressed in Nicotiana benthamiana cells. The recombinant JAZ2 fused to the green fluorescent protein (GFP) forms condensate in vitro, suggesting that the intermolecular interaction of JAZ2 molecules is a driving force for condensation. These results suggest the potential use of JAZ2 condensates to construct artificial membraneless organelles in plant cells.


Introduction
Eukaryotic cells contain membraneless organelles consisting of proteins and other biomolecules, which are also called biomolecular condensates, coacervates, bodies, granules, paraspeckles, or droplets (Courchaine et al. 2016;Banani et al. 2017;Shin and Brangwynne 2017). Membraneless organelles form various compartments, including nucleoli, stress granules, and processing bodies, which allow specific molecule enrichment, efficient biochemical reaction, biomolecule storage, inhibitory substance compartmentation, and protein turnover regulation (Courchaine et al. 2016;Banani et al. 2017;Shin and Brangwynne 2017;Franzmann et al. 2018;Pancsa et al. 2019). These compartments are often formed by liquid-liquid phase separation, depending on the concentration, pH, and temperature. Moreover, some condensates formed by liquid-liquid phase separation can also adopt solid or gel-like states (Kroschwald et al. 2015;Lin et al. 2015;Banani et al. 2017;Shin and Brangwynne 2017;Woodruff et al. 2017Woodruff et al. , 2018Alberti et al. 2019;Bose et al. 2022). Membraneless organelles will be potentially useful in bioengineering and synthetic biology, because they create platforms to modify cell function by providing an orthogonal reaction system or native protein sequestration Communicated by Attila Feher.
Jasmonate ZIM domain (JAZ) proteins are involved in jasmonate (JA) signalling repression in plants (Wasternack and Hause 2013;Chini et al. 2016;Huang et al. 2017;Howe et al. 2018). JA is a plant hormone that plays versatile roles in development and stress responses, particularly in defence responses (Wasternack and Hause 2013;Huang et al. 2017;Howe et al. 2018). JAZ proteins directly or indirectly repress transcription factors that activate JAinduced gene expression, such as bHLH transcription factors, including MYC2 and MYC3 of JA signalling, by recruiting a repressive complex (Toda et al. 2013;Chini et al. 2016;Huang et al. 2017;Howe et al. 2018). JA signalling is activated when plants are attacked by insects or pathogens, injured, or exposed to environmental stresses such as drought or salt damage (Wasternack and Hause 2013;Kurotani et al. 2015;Howe et al. 2018;Ogawa et al. 2021). Under these conditions, JA is synthesised and converted into jasmonoyl-Ile (JA-Ile), an active form of JA. In turn, JA-Ile binds to the COI1 receptor, a substrate recognition subunit of ubiquitin ligase, in association with the co-binding of a JAZ factor to COI1. This ubiquitinates JAZ, thereby degrading it via the 26S proteasome and derepressing JA-induced genes (Wasternack and Hause 2013;Chini et al. 2016;Huang et al. 2017;Howe et al. 2018).
Besides these JAZ factors, atypical JAZ factors, exemplified by Arabidopsis JAZ7 and JAZ8 (AtJAZ7 and AtJAZ8), lack the COI1 binding site in the Jas domain, which is required for JA signalling-dependent JAZ degradation (Shyu et al. 2012). AtJAZ8 is induced by JA and involved in the negative feedback regulation of JA signalling (Shyu et al. 2012). OsJAZ2 (or OsTIFY5; Os07g0153000), Oryza sativa JAZ2, belongs to this class of JAZ proteins and lacks the COI1 binding site in the Jas domain (Toda et al. 2013;Hori et al. 2014). OsJAZ2 might also influence JA signalling feedback regulation, since OsJAZ2 expression levels are low under normal growth conditions but inducible by JA, based on the public expression database (RiceXpro; https:// ricex pro. dna. affrc. go. jp). However, how the inhibitory effects of these atypical JAZ factors on JA-induced expression are attenuated remains unknown.
This study reports that OsJAZ2 tends to form condensates in the nucleus and cytosol when fused to the enhanced cyan fluorescent protein (eCFP) or enhanced yellow fluorescent protein (eYFP) markers, and is highly expressed in O. sativa and Nicotiana benthaminana cells transiently. OsJAZ2 condensates form compartments resembling those formed by proteins, which causes phase separation. Moreover, OsJAZ2 often forms large condensates in O. sativa cytosol. Based on the characterisation of JAZ2 condensates in these plant cells, as well as in vitro, we discuss the possible roles of JAZ2 condensates and their potential use in constructing artificial membraneless organelles in plants.

Plant materials
O. sativa L. cv. 'Nipponbare' seeds were surface sterilised and germinated on a sterile 1/2 × MS medium (Murashige and Skoog basal medium, 0.35% gellan gum, pH 5.7), in a set of plant boxes (60 × 60 × 100 mm 3 , CULJAR300; Iwaki, Tokyo, Japan) combined with two open tops. N. benthamiana seeds, generous gifts from Dr. H. Yoshioka of Nagoya University, were sown on soil in plant pots. O. sativa and N. benthamiana plants were grown at 27 °C (day) and 25 °C (night) in a chamber under 16 h-light (4500 lx) and 8 h-dark cycles. Protoplasts were isolated from the leaves of O. sativa seedlings harvested 6-7 days after sowing. For Agrobacterium infiltration, 3-to 5-week-old N. benthamiana leaves were used.

O. sativa protoplast isolation and transient gene expression
O. sativa leaves (parts that are upper than the coleoptiles and lower than the lamina joint of the second leaves, except the leaf blade of the second leaves) were sliced into 0.5-1 mm pieces using a new razor blade, soaked in 0.6 M mannitol, and incubated in dark for 10 min. After removing the mannitol solution, 20 mL enzyme solution (0.6 M mannitol, 10 mM MES-KOH [pH5.7], 1.5% Cellulase R-10 [Yakult Pharmaceutical Industry, Tokyo, Japan], 0.75% Macerozyme R-10 [Yakult Pharmaceutical Industry], 0.1% bovine serum albumin [BSA], 10 mM CaCl 2 ) was added and vacuum infiltrated into the tissues using a pump (DA-20D, ULVAC KIKO, Miyazaki, Japan), followed by incubating in dark for 5 h with gentle agitation (75 rpm) at 25 °C. After enzymatic digestion of the cell walls, 20 mL W5 solution (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, 2 mM MES-KOH [pH5.7]) was mixed, filtered through an autoclaved polyester mesh cloth (PET73), and centrifuged at 500 × g for 2 min. The precipitates were suspended in 2 mL W5 solution, filtered, and re-centrifuged. The supernatants were filtered and centrifuged similarly once or twice. Finally, all precipitates were suspended in 2 mL W5 solution and re-centrifuged. The final precipitate was suspended in 0.5-2 mL MMg solution (0.4 M mannitol, 15 mM MgCl 2 , 4 mM MES-KOH [pH5.7]). Isolated protoplasts in the solution were counted using a haemocytometer under a microscope and diluted with the same buffer solution to a cell density of 2 × 10 6 -2 × 10 7 cells/mL.

Agrobacterium infiltration of N. benthamiana
Agrobacterium infiltration of N. benthamiana using pSoupand pGeenII-derived plasmids was performed as previously described (Hellens et al. 2000;Asai et al. 2008). The plasmids were introduced into Agrobacterium tumefaciens (strain GV3101) by electroporation. The transformed Agrobacterium cell cultures were infiltrated into N. benthamiana leaves and incubated in the plant chamber for 2-3 days. For microscopic observation, leaf segments (1 × 2 cm 2 ) were mounted upside-down on glass slides (S2441, Matsunami Glass Ind.,) with coverslips and sterile water.

Microscopic observation and fluorescence recovery after photobleaching (FRAP) analysis
The fluorescence of reporter proteins was observed under a confocal laser scanning microscope (CLSM; FV 1000, Olympus, Tokyo, Japan) at 440/460-500, 515/530-545 nm, and 559/575-675 nm (excitation/emission wavelengths) for CFP, YFP, and red fluorescent protein (RFP), respectively. Green fluorescent protein (GFP) fluorescence of the recombinant proteins was observed using CLSM at 473 or 488 nm (excitation wavelengths) and 491-535 nm (emission wavelengths). For observing O. sativa cells, monomeric RFP1 (mRFP1)-expressing protoplasts were selected by introducing the pDH51-mRFP1 plasmid carrying the CaMV35S promoter-driven mRFP1 gene. To visualise the nucleus, a pUGW42-based plasmid carrying the CaMV35S promoterdriven eYFP-OsbHLH094 gene (Os07g0193800; Toda et al. 2013) was introduced into O. sativa protoplasts. For control experiments with O. sativa cells expressing eCFP alone, pUGW45 was used. To distinguish the localisation of condensates inside and outside of the nucleus, observation was carried out by shifting the focal plane. We used a 20 × objective lens for normal observation, a 60 × objective lens for FRAP analysis, and an 100 × objective lens for observation as in Fig. 1C and Supplementary Fig. 3. For FRAP analysis of condensates formed in vivo, circular areas (radius: 0.8-1 μm for O. sativa and 1.6 μm for N. benthamiana) were bleached for 0.5 s, using tornado bleach pulse, and CFP fluorescence recovery was monitored continuously without interval setting. For FRAP analysis of GFP-JAZ2 condensates formed in vitro, GFP foci were photo-bleached Fig. 1 OsJAZ2 frequently forms condensates in Oryza sativa protoplasts. A Top, schematic domain structure of OsJAZ2. TIFY and Jas domains, EAR motif, and five low-complexity regions (LCRs) are depicted. Bottom, prion-like domain (PrLD) and intrinsically disordered regions (IDRs) are predicted by 'Prion-like Amino Acid Composition' (PLAAC; http:// plaac. wi. mit. edu/) and 'Database of Disordered Protein Prediction' (D 2 P 2 ; https:// d2p2. pro/) algorithms, respectively. B Fluorescence microscopy of O. sativa protoplasts expressing eCFP-JAZ2, eYFP-bHLH094, and mRFP1. CFP, YFP, and RFP images are shown with pseudo-colours. Subcellular localisation of eCFP-JAZ2 (I) without or (II-V) with CFP foci in five representative cells is shown. The JAZ2 condensate and nucleus positions are indicated by white and orange arrowheads, respectively. I, no JAZ2 condensate but nuclear localisation of JAZ2. II and III, JAZ2 condensates present in the vicinity of the nucleus. IV and V, large JAZ2 condensates in the vicinity of the nucleus and in the cytosol, respectively. eYFP-bHLH094 and mRFP1 were co-expressed for visualisation of the nucleus (YFP) and both the nucleus and cytosol (RFP), respectively. In images I and II, cells with large vacuoles (vac) are shown. The position of JAZ2 foci shown in II is devoid of RFP signals. Scale bars, 5 μm. C Z-stack image set of the nucleus in the cell shown in B (panel II) and Supplementary Fig. 3B. Images of CFP (top) and YFP (middle) of the same nucleus are merged (bottom). White arrowheads indicate typical eCFP-JAZ2 condensates formed in the nucleus. The position of the focal plane on the z-axis is shown in the top panels. The images were obtained with an 100 × objective lens. Scale bars, 1 μm (color figure online) with a 488 nm laser pulse (3 frames, 25%), and GFP fluorescence recovery was monitored at 1 s intervals for 4 min. Fluorescence intensity of the region of interest (ROI) was measured using FV1000 software or FIJI/ImageJ. Normalised fluorescence intensity (I nor ) was calculated as follows; where I t and BG t are the fluorescence intensity and background at time point t, and I pre and BG ave are the average intensity and background of 50 frames before bleaching, respectively.

Recombinant protein expression and purification and in vitro condensation
MBP-fused proteins were expressed in E. coli strain BL21 (DE3). E. coli cells were cultured in a liquid medium (rich broth supplemented with glucose) at 37 °C with shaking at 105 rpm until the culture absorbance at 600 nm reached 0.4-0.5. The cells were cooled on ice and cultured in the presence of 0.5 mM isopropyl ß-D-1-thiogalactopyranoside (IPTG) at 18 °C with shaking overnight.
MBP-fused proteins were purified using amylose-resin (E8021S, New England Biolabs Japan Inc., Tokyo, Japan). The cells were harvested by centrifugation at 4000 × g at 4 °C for 10 min, resuspended in 5 mL column buffer (20 mM Tris-HCl [pH 7.4], 200 mM NaCl, 1 mM EDTA [pH 8.0], cOmplete, EDTA-free [Roche, Merck KGaA], 1 mM dithiothreitol [DTT]), frozen at − 20 °C, thawed in a water bath at 37 °C until only a few ice pieces remained, and the remaining ice pieces were thawed by gentle tumbling and mixing. The cells were sonicated in an ice-cold water bath in ten cycles of 10 s bursts with 60 s cooling intervals using a sonicator (BRANSON SONIFIER 450, Branson Ultrasonics Corp, CT, USA) at output control 3 and 30% duty cycle. The cell lysate was centrifuged at 9000 × g for 20 min at 4 °C and separated into soluble and insoluble fractions. The soluble fractions were aliquoted into several microtubes, frozen in liquid nitrogen, and stored at − 80 °C.
The soluble fraction (70-140 µL) was incubated with 30-35 µL (bed volume) pre-washed amylose resin in a 1.5 mL tube for 2 h at room temperature, with gentle agitation using a seesaw shaker (BC-700, BIO CRAFT, Tokyo, Japan). The resin-bound MBP-fused protein was collected by centrifugation at 3300 × g for 1 min. After removing the supernatant, the MBP-fused protein was washed with 1 mL column buffer, collected by centrifugation, eluted by incubation with 80 µL column buffer supplemented with 10 mM maltose for 10 min at room temperature with gentle agitation, recovered in the supernatant by centrifugation at 3300 × g for 1 min, and stored at 4 °C. The MBP-tag was separated from the protein of interest by cleavage with TEV protease (12575-015, Invitrogen) at 30 °C for 2 h after adding an equal volume of 2 × TEV cleavage buffer (100 mM Tris-HCl pH 8.0, 1 mM EDTA) and 1 mM DTT final concentration. NaCl concentration in the buffer for cleavage by the TEV protease was increased to examine condensation in the presence of increased NaCl concentration. IPTG induction, purification using amylose resin, and MBP-tag cleavage of proteins were confirmed by SDS-PAGE analysis. The amount of protein after MBP-tag cleavage was estimated from the staining of bands on SDS−PAGE with BSA as a standard. For microscopic observation, 8 μL protein solution was placed on a glass slide (S024410, Matsunami Glass Ind.) and overlaid with a coverslip. The protein samples were incubated at 23-25 °C for 19 h after the TEV protease reaction at 30 °C for 2 h for observing long time-dependent condensate formation.
For treatment of GFP-JAZ2 condensates with 1,6-HD, 50% 1,6-HD dissolved in H 2 O, together with the TEV protease reaction buffer premix containing the column buffer, was added to the protein solution after the TEV protease reaction to a final concentration of 10% 1,6-HD. The protein solution was then placed on a glass bottom dish (D141400, Matsunami Glass Ind.) and immediately used for observation of GFP at 488/500-515 (excitation/emission wavelengths) under a CLSM (FV 3000, Olympus, Tokyo, Japan) with a 60 × objective lens.

eCFP-JAZ2 expression causes condensate formation in O. sativa protoplasts
Subcellular localisation of several JAZ factors and other nuclear proteins in O. sativa indicated that eCFP-fused OsJAZ2 (eCFP-JAZ2) frequently formed condensates when transiently expressed in O. sativa protoplasts (Fig. 1). The CFP fluorescence of eCFP-JAZ2 was detected in not only the nucleus, but also the cytosol with strong foci. JAZ2 condensates in the cytosol were frequently observed in the vicinity of the nucleus (Fig. 1, Supplementary Fig. 1), possibly reflecting the cytosolic condensation-dependent nuclear localisation inhibition. Extensive observations showed that eCFP-JAZ2 also formed condensates in the nucleus (Fig. 1C). Such foci were scarcely observed with other nuclear factors, such as for eCFP-fused JAZ9 (or OsTI-FY11a) ( Supplementary Fig. 2), a typical JAZ protein that can bind to COI1 and is degraded in a JA signalling-dependent manner (Toda et al. 2013;Wu et al. 2015), and for eYFPfused bHLH094 that was used for nucleus visualisation in this study (Toda et al. 2013) (Fig. 1, Supplementary Figs. 1,  2). In both cases, eCFP and eYFP signals were detected predominantly in the nucleus (Fig. 1, Supplementary Figs. 1, 2). CFP fluorescence of eCFP-JAZ9 was weaker than that of eCFP-JAZ2 in most cells, probably due to JAZ9 degradation. JAZ2 condensates were detected in 45-100% of the observed cells, depending on the experiments (mean ± s.d., 86 ± 16%, n = 20). Notably, eCFP-JAZ2 occasionally forms extremely large condensates in the cytosol, which is nearly comparable to the nucleus in size (Fig. 1, Supplementary Fig. 1). Moreover, signals of the fluorescent protein, mRFP1, another marker protein used for both nucleus and cytosol visualisation, were often less visible at the site of eCFP-JAZ2 condensates (Fig. 1, Supplementary Figs. 1, 3). This implies that eCFP-JAZ2 condensates form subcellular compartments. eYFP-bHLH094 was occasionally excluded from JAZ2 condensates, but sometimes combined with JAZ2 condensates.

eCFP-JAZ2 forms stable condensates in O. sativa
The eCFP-JAZ2 condensates observed in O. sativa cells varied in size and shape; condensates observed in the nucleus were relatively small and spherical or ellipsoidal (0.5-1.5 μm diameter), whereas those in the cytosol were often large and non-spherical or non-ellipsoidal (up to approximately 7 μm diameter; Fig. 1, Supplementary Figs. 1,  3). In some cases, eCFP-JAZ2 condensates in the nucleus and cytosol appeared to form a connection, possibly through the nuclear pores ( Supplementary Fig. 3). Protoplast burst by changing osmotic pressure immediately diffused cytosolic mRFP1, but not eCFP-JAZ2, signals in the vicinity of the nucleus ( Supplementary Fig. 4A). Treatment of protoplasts with 1,6-hexanediol, which is widely used to distinguish between liquid and solid condensates (Kroschwald,et al. 2017;Alberti et al. 2019), had little effect on eCFP-JAZ2 condensates ( Supplementary Fig. 4B). These observations suggest that eCFP-JAZ2 forms stable condensates.
In this study, protoplasts were incubated in a nutrient-poor solution after plasmid DNA introduction. Thus, whether cell starvation might cause eCFP-JAZ2 condensate formation, similar to yeast starvation-induced proteasome storage granule, solid-like Cdc19 aggregate, or Sup35 condensate formation (Laporte et al. 2008;Saad et al 2017;Franzmann et al. 2018), should be studied. However, eCFP-JAZ2 condensed even when protoplasts were incubated in a nutrient-rich R2P medium ( Supplementary Fig. 5A).
When eCFP alone was expressed, CFP foci reflecting protein condensation were not detected ( Supplementary  Fig. 5B). Since GFP-derived proteins, such as CFP, can form dimers, we examined JAZ2 fusion to eCFP A207K , which carries a mutation to avoid dimerisation of GFP derivative makers (Zacharias et al. 2002;Segami et al. 2014). eCFP A207K -JAZ2 developed CFP foci, indicating condensation, although the condensate formation frequency depended on the presence or absence of YFP-bHLH094 (Supplementary Fig. 5C). JAZ2-fused eYFP also produced eYFP foci ( Supplementary Fig. 5D), indicating that JAZ2-mediated foci are not necessarily linked to CFP fluorescence. Taken together, these results suggest that JAZ2 is a condensationprone protein. This was further confirmed by mutational analyses of JAZ2 and in vitro condensation experiments, as described below.

JAZ2 low-complexity regions are involved in condensate formation
OsJAZ2 has IDRs with five low-complexity regions (LCRs), in addition to an EAR motif and the TIFY and Jas domains, which are highly conserved among JAZ factors (Fig. 1A), whereas OsJAZ9 possesses only two LCRs (Supplementary Fig. 2). Moreover, the existence of a PrLD was predicted in OsJAZ2 at the second LCR (LCR2) (Fig. 1A), but not OsJAZ9 (Supplementary Fig. 2). Deletions of regions containing LCR1, LCR2, or LCR3 and LCR4 decreased eCFP-JAZ2 condensate formation frequency, although the extent of their effects varied between experiments ( Supplementary Fig. 6). Deletion of a region containing LCR3-LCR5 resulted in a more drastic decrease in condensation frequency (Supplementary Fig. 6). These results suggest that LCRs are involved in eCFP-JAZ2 condensation. It should be noted that LCR3-LCR4 deletion weakened the overall CFP signals, and finding cells with sufficient CFP signals for observation was relatively difficult.

TIFY and Jas domains contribute to JAZ2 condensate formation
Next, whether the conserved domains among JAZ factors are involved in JAZ2 condensate formation were examined because they contain amino acid sequences predicted to generate amyloid-like structures using the FoldAmyloid program (Garbuzynskiy et al. 2010) (Fig. 2A). The TIFY domain contains two β-strands forming an antiparallel β-sheet (Fig. 2B), raising the question whether these cause weak interactions between the molecules. EAR motif and TIFY and Jas domain deletions revealed that the Jas domain was required for eCFP-JAZ2 condensation, in addition to its function in nuclear localisation (Withers et al. 2012) (Fig. 3 and Supplementary Fig. 7). The TIFY domain positively affected JAZ2 condensate formation, whereas the contribution of the N-terminal region containing the EAR motif differed depending on the experiment (Fig. 3 and Supplementary Fig. 7). Additional deletion analysis showed that not only the N-terminal part containing the two β-strands, but also the C-terminal part of the TIFY domain could be involved in JAZ2 condensate formation and most amino acids forming the two β-strands, including two aromatic residues (Phe and Tyr), were not necessary for JAZ2 condensation ( Supplementary Fig. 8).
The Jas domain of JAZ2 is rich in basic amino acids, such as Lys and Arg (K, R; Fig. 2, Supplementary Fig. 9), which potentially cause π-cation or electrostatic interactions (Gomes and Shorter 2019;Wang et al. 2018). Deletion analysis showed that a region with basic amino acids (BAA) in the Jas domain is required for frequent eCFP-JAZ2 condensation ( Supplementary Fig. 9). In the case of the FUS protein, Arg is more important than Lys in molecular interactions that cause phase separation in terms of interaction with Tyr and Phe, because the chemical structure of the cationic side chain of Arg delocalises electron clouds, leading to strong directional preferences with aromatic moieties (Wang et al. 2018). Arg-to-Lys substitutions in the Jas domain (Jas RtoK ) decreased eCFP-JAZ2 condensation, in the context of size and frequency, although the secondary structure of the Jas domain might not be affected (Supplementary Figs. 9,10). In contrast, substituting Lys and Arg with Ala in the Jas domain (Jas K,RtoA ) formed apparent and even larger condensates in the cytosol (Supplementary Figs. 9, 10). These results suggest that condensation with the Jas domain depends on molecular interactions involving Arg in the Jas domain, but these can be replaced by interactions other than π-cation or electrostatic interactions.

JAZ9 can be incorporated into JAZ2 condensates
Since the TIFY and Jas domains are involved in JAZ2 condensation, possibly by interactions between JAZ2 molecules, or JAZ2 and other possible JAZ2 condensate components, if any, we wondered if JAZ9 carrying the conserved domains (Toda et al. 2013;Hori et al. 2014) might be associated with JAZ2 condensates, even though JAZ9 itself does not form condensates as described above. To test this possibility, eCFP-JAZ9 was co-expressed with eYFP-JAZ2 in O. sativa protoplasts. This resulted in CFP foci of eCFP-JAZ9, which co-localised with eYFP-JAZ2 condensates in both the nucleus and cytosol (Fig. 4A). Such eCFP-JAZ9 and eYFP-JAZ2 co-condensation was observed in most cells with eYFP-JAZ2 condensates (90%, n = 10). Thus, JAZ9 may be incorporated into the JAZ2 condensates. Moreover, the CFP signals of eCFP-JAZ9 were more prominent when expressed with eYFP-JAZ2, suggesting that JAZ9 association with JAZ2 condensates represses JAZ9 degradation.
In contrast, eYFP-LCR5-Jas or eYFP-TIFY-LCR2 scarcely formed condensates when expressed alone (Supplementary Figs. 11B,C), suggesting that the TIFY or Jas domain alone does not sufficiently account for the condensate-forming propensity of JAZ2. Moreover, the combined expression of eYFP-LCR5-Jas and eYFP-TIFY-LCR2 did not result in the formation of condensates ( Supplementary Fig. 11D). Together with the results of the deletion analyses, these results suggest that the TIFY or Jas domains more likely contribute to JAZ2 condensation cooperatively with other regions. Fig. 2 Prediction of amyloidogenic regions and eCFP-JAZ2 3-D structure. A Amino acid sequence of the eCFP-JAZ2 fusion protein was analysed by the FoldAmyloid program (http:// bioin fo. protr es. ru/ fold-amylo id/). Amino acids marked in blue-violet indicate sequences prone to amyloidlike aggregation. Amino acid sequences of eCFP and JAZ2 are boxed. EAR motif, TIFY, and Jas domain are indicated by blue, red, and yellow underlines, respectively. The sequence predicted as JAZ2-PrLD (see Fig. 1) is indicated by a green underline. B Prediction of the three-dimensional structure of eCFP-JAZ2 by Alphafold2 (https:// colab. resea rch. google. com/ github/ sokry pton/ Colab Fold/ blob/ main/ Alpha Fold2. ipynb) (Mirdita et al. 2021;Jumper et al. 2021). Blue and red indicate regions of high and low confidence, respectively. The black arrow indicates the position of the TIFY domain of JAZ2. The secondary JAZ2 structure was predicted as also shown in Supplementary  Fig. 10A (color figure online)

JAZ2 forms condensates in N. benthamiana
We expressed eCFP-JAZ2 in N. benthamiana, to find out if JAZ2 forms condensates depending on the cellular environment and experimental conditions. We also thought that if JAZ2 could form condensates in N. benthamiana, which is often used for the production of proteins of interest, we would be able to use condensates for engineering in the future. eCFP-JAZ2 foci were found in N. benthamiana leaf epidermal cells after transient eCFP-JAZ2 expression by Agrobacterium infiltration methods, similar to that in O. sativa protoplasts (Fig. 5). In contrast, no foci were detected with eYFP, which was localised in the nucleus and cytosol (Fig. 5), and the control eCFP without fusion to JAZ2 (Supplementary Fig. 12). In cells co-expressing eCFP-JAZ2 and eYFP, YFP signals were often excluded at the site of JAZ2 foci, indicating that interaction between the GFP derivatives (eCFP and eYFP) is not necessary for JAZ2 condensate formation. These results suggest that JAZ2 condensation does not require components specific to O. sativa leaf cells (other than OsJAZ2 itself) and experimental procedure for protoplast transient assays, including PEG-mediated transfection of DNA. Interestingly, however, the JAZ2 condensates differed between the two experimental systems in terms of size and localisation. The diameter of eCFP-JAZ2 condensates found in N. benthamiana leaf epidermal cells were 1-4 μm. The smaller (up to 3 μm) condensates were mostly spherical or ellipsoidal, whereas larger condensates (approximately 4 μm) often formed indefinite shapes. They were mostly localised in the nucleus and occasionally in the vicinity of the nucleus (Fig. 5). However, unlike O. sativa protoplasts, JAZ2 condensates in the cytoplasm were not observed at sites distant from the nucleus.

Different properties of JAZ2 condensates in O. sativa and N. benthamiana
Next, eCFP-JAZ2 molecule fluidity in the JAZ2 condensates was examined by fluorescence recovery after photobleaching (FRAP), which assessed unbleached eCFP-JAZ2 redistribution into the bleached area. In most cases, CFP fluorescence was only slightly recovered after bleaching in O. sativa protoplasts, indicating the low mobility of eCFP-JAZ2 molecules ( Fig. 6A and Supplementary Fig. 13A). These results suggest that JAZ2 mostly forms gel-like or solid condensates in the O. sativa cells. However, in some cases, moderate but apparently higher fluorescence recovery was observed, indicating an increase in the mobile fraction of the condensates (Figs. 6B-C). Noticeably, after bleaching of a portion of the JAZ2 condensate, fluorescence increases in the bleached area occurred concomitantly with fluorescence decrease in the unbleached area of the same condensate (Fig. 6C), suggesting the exchange of eCFP-JAZ2 molecules between these areas. Therefore, it is most likely that JAZ2 condensates can also adopt a liquid-like state. The difference in the degree of fluorescence recovery seems to be cell dependent, but not size or condensate localization related (Figs. 6D-E).
We expected that the material properties of the JAZ2 condensate might change if the interactions contributing to condensation are compromised. We, therefore, examined the effects of deletion or substitution of potential interacting domains or regions on the molecular fluidity of JAZ2 condensates. Interestingly, fluorescence recovery was moderately increased when condensates formed by eCFP-JAZ2ΔNterm and eCFP-JAZ2ΔTIFY were bleached (Fig. 6F, Supplementary Fig. 13B). Unlike this, fluorescence 1 3 recovery between the condensates formed by eCFP-JAZ2 and other variants, such as JAZ2-Jas RtoK and JAZ2-Jas K,RtoA , was similar ( Supplementary Fig. 13C).
In contrast, JAZ2 condensates exhibited substantial levels of fluorescence recovery in N. benthamiana leaf cells, indicating improved eCFP-JAZ2 mobility (Fig. 6G). This suggests that JAZ2 forms liquid-like compartments in N. benthamiana cells. The liquidity of JAZ2 condensates might be caused by lower levels of eCFP-JAZ2 expression in N. benthamiana cells than in O. sativa protoplasts. We, therefore, examined FRAP of eCFP-JAZ2 condensates in O. sativa protoplasts transfected with lower amounts of plasmid DNA, carrying CaMV 35S promoter::eCFP-JAZ2. The formation of JAZ2 condensates was observed when the amount of plasmid used for the transfection was reduced from 5 μg to 2 μg, but not with 1 μg (Supplementary Fig. 14A). Even under the condition of 2 μg, JAZ2 condensates did not exhibit significant fluorescence recovery in O. sativa protoplasts (Supplementary Fig. 14B). Therefore, it seems unlikely that the liquidity of JAZ2 condensation is caused simply by low levels of JAZ2 expression. Thus, the properties of JAZ2 condensates differ between O. sativa and N. benthamiana.

JAZ2 forms condensates in vitro
Recombinant GFP-JAZ2 was prepared to test whether JAZ2 itself has the propensity to form condensates. The RRM domain and PrLD of the Arabidopsis FCA protein (Fang et al. 2019) were used for the control experiments. PrLD, but not the RRM domain of FCA, undergoes phase separation both in vitro and in vivo (Fang et al. 2019). GFP-JAZ2, GFP-FCA (RRM), GFP-FCA (PrLD), and GFP alone were expressed in E. coli., as soluble proteins by fusion with MBP-tag, and purified by specific binding of MBP-tag to amylose resin and subsequent elution by maltose (Supplementary Fig. 15). GFP-JAZ2 condensation began within 1 h after MBP-tag removal by digestion using TEV protease ( Fig. 7A-D, Supplementary Fig. 15). Approximately, 0.1-0.3 μM GFP-JAZ2 caused condensation in vitro (Fig. 7,  Supplementary Fig. 15). Under the same experimental conditions, GFP condensation was observed with GFP-FCA (PrLD) even at low levels, but was scarcely or not detected with GFP-FCA (RRM) and GFP alone (Fig. 7, Supplementary Fig. 15). Thus, JAZ2 were able to form condensates outside of plant cells, although the recombinant proteins were not completely pure in these analyses (Fig. 7C, Supplementary Figs. 15B, C). These results suggest that JAZ2 is condensate prone.
The fluidity of GFP-JAZ2 molecules in JAZ2 condensates in vitro was examined using FRAP. The GFP signals were not recovered after GFP-JAZ2 condensate photobleaching, suggesting that JAZ2 forms gel-like or solid condensates under the experimental conditions (Fig. 7E). Consistent with this, increased incubation of recombinant GFP-JAZ2 in vitro formed large, non-spherical, or nonellipsoidal condensates and occasionally formed massive intricate condensates (Supplementary Figs. 16A, B). In the presence of 0.5 M NaCl, JAZ2 condensate formation drastically decreased (Supplementary Figs. 16C, D), which is consistent with the possibility that electrostatic and/or π-cation interactions are important for JAZ2 condensation. In contrast, treatment with 1,6-hexanediol, which inhibits weak hydrophobic interactions (Kroschwald et al. 2017), had little effect on the stability of the JAZ2 condensates formed in vitro ( Supplementary Fig. 16E), as in the experiments using O. sativa protoplasts.

Discussion
This study showed that OsJAZ2 is condensation prone and eCFP-JAZ2 expression causes condensation in O. sativa and N. benthamiana leaf cell nucleus and cytosol, in addition to typical protein localisation in the nucleus. JAZ2 condensates formed subcellular compartments that excluded mRFP1 and eYFP markers in O. sativa and N. benthamiana, respectively. Several properties of JAZ2 are similar to those of proteins that play pivotal roles in biomolecular condensate formation via phase separation. Proteins that drive phase separation are often rich in IDR, allowing several conformations and undergoing multivalent interactions among molecules  Supplementary Fig. 13B. G A representative image of eCFP-JAZ2 condensate in N. benthamiana leaf cell nucleus before and after photobleaching (top), and normalised FRAP intensity (bottom). CFP fluorescence intensity at 100 s is shown (mean ± SD, n = 3). Yellow and white arrowheads indicate the photobleaching site and the nucleus. Time 0 indicates the time point of the bleaching pulse. Scale bar, 5 μm. An asterisk indicates significance as evaluated by Welch's t test (P < 0.05). n.s. means 'not significant'. The image acquisition speed in these FRAP analyses was 2 μs/ pixel, except for the analysis in (C), which was 4 μs/pixel. Accordingly, the data in (C) is not included in the comparison in D-F (color figure online) ◂ (Banani et al. 2017;Uversky 2017;Peran and Mittag 2020). Similarly, the JAZ2 protein contains IDRs, including PrLD/ LCRs, and can form condensates in vitro. Deletion analyses of JAZ2 indicated that the TIFY and Jas domains, as well as LCRs, were involved in condensate formation. The TIFY-LCR2 and LCR5-Jas regions alone were associated with JAZ2 condensates, although each region did not show sufficient proficiency for stable JAZ2 condensate formation. Therefore, these regions may cause multivalent interactions to form JAZ2 condensates.
In this study, JAZ2 frequently formed condensates with varied size, shape, localisation, and molecular fluidity. This might reflect that the structures of the JAZ2 protein and its condensate are metastable, possibly because of multivalent interactions, and can be affected by cellular conditions. Notably, JAZ2 condensates are often larger than Arabidopsis FCA condensates, which are spherical with up to approximately 2.5 μm diameter in Arabidopsis and tobacco cells (Fang et al. 2019). In contrast, JAZ2 often forms nonspherical condensates that reach 4-7 μm in diameter in O. sativa and N. benthamiana cells. Such differences in size are likely related to the differences in the nature of the FCA and JAZ2 condensates; FCA forms liquid droplets in Arabidopsis and tobacco cells (Fang et al. 2019), whereas JAZ2 forms highly viscous (gel-like or solid) condensates mostly in O. sativa protoplasts.
JAZ2 condensate formation frequency was substantially reduced by deleting the Jas domain and adjacent sequence, and substituting four Arg residues in the Jas domain with Lys residues. However, the JAZ2 condensates with the Arg-to-Lys substitution were still solid. This suggests that intermolecular interactions involving Arg residues in the Jas domain, such as π-cation interactions, contribute to JAZ2 condensation process; however, JAZ2 condensate solidification or stability are supported by interactions involving other amino acids. In contrast, BAA (Arg and Lys) substitution with Ala in the Jas domain did not reduce JAZ2 condensate formation frequency, instead forming large condensates. This suggests that an increase in Ala residues in the Jas domain causes other molecular interactions, such as hydrophobic interactions, rather than π-cation or electrostatic interactions, thus enhancing JAZ2 condensate formation. Interestingly, an Alarich sequence in the H1 helix of the Syrian hamster prion protein, Ala-Gly-Ala-Ala-Ala-Gly-Ala, form a β-sheet in an infectious conformation and cause interactions that lead to amyloid formation, which can be mimicked by an Ala tripeptide in vitro (Lundberg et al. 1997;Bauer et al. 2011). Likewise, Ala-rich sequences in the mutated Jas domain may have more effect than hydrophobic interactions.
Amyloids such as amyloid-β and α-synuclein, or those found in prion proteins, are formed by cross-β structures (Lundberg et al. 1997;Balbirnie et al. 2001;Bauer et al. 2011;Nelson et al. 2005;Tuttle et al. 2016;Wälti et al. 2016). Although whether JAZ2 forms an amyloid-like structure in O. sativa and in vitro is unclear, as in the case of Xvelo, a PrLD-containing IDP that constitutes Balbiani bodies in Xenopus (Boke et al. 2016), β-sheets may be involved in solid JAZ2 condensate formation. Structural prediction showed that the TIFY domain of JAZ2 contains a region that forms an antiparallel β-sheet. Moreover, when most of the TIFY domain was deleted, the JAZ2 variant appeared to form more liquid-like condensates in O. sativa protoplasts ( Fig. 6F and Supplementary Fig. 13B). Another possible amyloid-like structure-forming sequence is the Gln-and Asn-rich PrLD/LCR2 of JAZ2 (Fig. 2), since polar Gln-and Asn-rich sequences form stacked β-sheet structures (Balbirnie et al. 2001;Nelson et al. 2005). Notably, LCR2 deletion largely impaired JAZ2 condensation, in addition to that TIFY-LCR2 was efficiently incorporated into JAZ2 condensates. It would be interesting to know if such sequences, together with multivalent interactions, contribute to the solid state of JAZ2 condensation.
Unlike condensates in O. sativa protoplasts, eCFP-JAZ2 formed liquid-like condensates in N. benthamiana cells. In FRAP analysis, eCFP-JAZ2 fluorescence recovery after bleaching was detected within 5 s and maintained continuously for 60 s (Fig. 6G). This is almost comparable to the time required to recover eGFP-or YFP-fused FCA in Arabidopsis and tobacco cells (recovery detected within 5 s and continued for 40 to 70 s; Fang et al. 2019). Interestingly, the fluidity of the eCFP-JAZ2 molecules in the JAZ2 condensates differed between O. sativa and N. benthamiana probably due to the differences in intracellular environments, such as cell volume, or factors leading to post-translational modifications. Noticeably, FRAP analysis indicated that JAZ2 forms solid condensates in vitro, as reported in FUS and RBM14 (Hennig et al. 2015). Recombinant GFP-Arabidopsis FCA condensates in vitro do not show high FCA-GFP fluidity, though FCA forms liquid-like bodies in vivo, suggesting the requirement of additional factors for the liquidity of FCA bodies formed in vivo (Fang et al. 2019). Likewise, JAZ2 condensates in N. benthamiana cells may contain an additional molecule that increases the molecular fluidity of JAZ2.
As discussed by Alberti and Hyman (2021), it is not easy to distinguish whether high-ordered protein assembly is caused by phase separation or reflects insoluble protein aggregation. Protein aggregates are usually referred to as those with irreversible intermolecular interactions. In contrast, however, a small portion of JAZ2 molecules are mobile in not all but some of the JAZ2 condensates (Fig. 6B, C). Considering that JAZ2 condensates can adopt not only solidlike but also liquid-like states in the cells, it is plausible that JAZ2 shares some propensities with several scaffold proteins that form both dynamic and non-dynamic biomolecular condensates. Interestingly, some of those condensates are solidified rapidly from liquid-like states (Woodruff et al. 2018). Assembles of C. elegans SPD5, the key scaffold protein of the pericentriolar material (PCM), are liquidlike droplets 2 min after formation in vitro but converted to solid-like structures 15 min after formation (Woodruff et al. 2017). In vitro reconstitution of oskar ribonucleoprotein granules, which behave as solids and function in embryonic development in Drosophila, also shows liquid-tosolid transition within 30 min (Bose et al. 2022). These observations suggest that the in vivo solid condensates are derived from liquid-phase condensates. At present, we cannot distinguish whether solid-like JAZ2 condensates may also arise from liquid-like assemblies or they are formed separately, although small JAZ2 condensates are observed with sphere shapes (e.g. in Fig. 7B), which is a criterion to assess phase separation-mediated condensates (Alberti et al. 2019). Another important point is that the material properties (such as liquid or solid state) of condensates formed by multivalent interactions depend on the pattern of their constituents and the valency of interactions (Alberti et al. 2019). For instance, tethering of FUS LCD to the oskar granule causes changes in its material properties; from solid to liquid states (Bose et al. 2022). In contrast, duplication of the RNA-binding domain of the scaffold protein G3BP1 results in less dynamic stress granules (Yang et al 2020). Notably, in our study, partial deletion of the N-terminal region or TIFY domain of JAZ2 increases the mobile fraction of JAZ2 condensates, even moderately ( Fig. 6F and Supplementary  Fig. 13B). This is consistent with the hypothesis that the JAZ2 condensate formation is also driven by multivalent interaction.
The physiological significance of the JAZ2 condensate in O. sativa is currently ambiguous. JAZ2 lacks a conserved COI1 binding site and is unlikely to be degraded depending on the JA signal. Therefore, JAZ2 condensates may play a role in JAZ2 inactivation concentration dependently. Interestingly, JAZ9 was incorporated into the JAZ2 condensates, thus stabilising JAZ9. In addition, TIFY and Jas domains conserved among JAZ factors are involved in JAZ2 condensation, and TIFY domains mediate dimerisation of some JAZ proteins (Chini et al. 2009), raising the possibility that condensation with JAZ2 would be a mode of storage and/ or another way of inactivating JAZ9 and possibly other JAZ factors, similar to storage granules reported in yeast (Laporte et al. 2008;Saad et al. 2017;Franzmann et al. 2018). JAZ factor sequestration may also serve as a stress memory, making cells sensitive to repetitive stress that activates JA signalling. However, JAZ2 expression levels might have been fairly high in our transient assay, whereas the expression levels of JAZ2 in O. sativa plants were relatively low, even though JAZ2 was induced by JA (RiceXpro; https:// ricex pro. dna. affrc. go. jp). Thus, JAZ2 accumulation levels might not be sufficient to form condensates under physiological conditions. Alternatively, the opposite situation may also be possible; such low expression levels could have allowed the condensation-prone propensity of JAZ2, without causing potentially detrimental effects by condensation. It is also noteworthy that YFP-fused AtJAZ1 and AtJAZ9 form subnuclear bodies, when expressed in Nicotiana tabacum leaves through Agrobacterium-mediated transient expression (Withers et al. 2012). However, the function of such subnuclear bodies has not yet been elucidated. Further studies should describe the physiological roles of JAZ protein condensation.
Transient gene expression using protoplasts and Agrobacterium infiltration-mediated gene expression in N. bethamiana has been used for several assays in plants. Our results suggest that the condensation-prone propensity of a protein, such as JAZ2, should decrease substantial concentration in cells or localised areas, and therefore affect the evaluation of activity, for example effector activity in the transcriptional activation assay. The same is true for proteins with mutations that would cause or enhance condensation-prone propensity. This possibility must be considered when creating variant proteins, for example, by genome editing.
It would be tempting if artificial subcellular compartments could be newly designed and constructed using condensation-prone proteins in plant cells. Such synthetic compartments may be applied for orthogonal reaction introduction, storage, and molecule sequestration, filtering, or sensing, as reported in mammals and yeast (Shin and Brangwynne 2017;Pancsa et al. 2019), thereby creating cells with new functions (Hastings and Boeynaems 2021). Designing the desired condensates requires a better understanding of controlling condensation-prone protein properties. For this purpose, JAZ2 with unique features is potentially useful for a model study in plant cells, considering that JAZ2 forms relatively large and stable liquid-like to solid-like condensates depending on the cell state or type. In particular, JAZ2 often produces extremely large condensates in the protoplast cytosol. Moreover, Ala substitution of Arg or Lys in the Jas domain formed even larger condensates, suggesting that the JAZ2 condensate size can be easily modulated. Current studies are focusing on amino acid sequences responsible for controlling not only JAZ2 and/or other protein condensate size, but also their subcellular localisation and molecular fluidity. This would enable the modification of such properties of other membraneless organelles, thereby extending the potential uses of artificial membraneless organelles in plant cells in the future.
Author contributions All authors contributed to the study conception and/or design. Material preparation, data collection, and analysis were performed by YK, YJ, TA, YY, HG, NH, and HK. OsJAZ2 condensation in Oryza sativa cells was first reported by HK. Data interpretation was performed by TH and ST in addition to the authors described above. The first draft of the manuscript was written by YK and ST and improved by the other authors' suggestions. All authors read and approved the final manuscript.
Funding This work was partially supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI (JP16K08140 and JP22K05427) and Nagoya University-National Institute of Advanced Industrial Science and Technology (NU-AIST) alliance project. YK was supported by the Japan Science and Technology Agency (JST) SPRING (JPMJSP2125) and "Graduate Program of Transformative Chem-Bio Research" in Nagoya University, supported by The Ministry of Education, Culture, Sports, Science and Technology (MEXT) (WISE Program).

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.