Basic design of artificial membrane-less organelles using condensation-prone proteins in plant cells

Abstract

Membrane-less organelles, formed by the condensation of biomolecules, play a pivotal role in eukaryotes. Artificial membrane-less organelles and condensates are effective tools for the creation of new cellular functions. However, it is poorly understood how to control the properties that affect condensate function, particularly in plants. Here, we report the construction of model artificial condensates using the condensation-prone proteins OsJAZ2 and AtFCA in a transient assay using rice (Oryza sativa) cells, and how condensate properties, such as subcellular localization, protein mobility, and size can be altered. We showed that proteins of interest can be recruited to condensates using nanobodies or chemically induced dimerization. Furthermore, by combining two types of condensation-prone proteins, we demonstrated that artificial hybrid condensates with heterogeneous material properties could be constructed. Finally, we showed that modified artificial condensates can be constructed in transgenic Arabidopsis thaliana plants. These results provide a framework for the basic design of synthetic membrane-less organelles in plants.

**Subject ** Plant cell biology, Molecular engineering in plants

Introduction

Eukaryotic cells contain a variety of membrane-less compartments called membrane-less organelles (MOs) or biomolecular condensates^1–3^. They are formed by the phase separation caused by the self-assembly of condensation-prone proteins, which are often associated with other proteins and/or nucleic acids^1–3^. These membrane-less compartments can form either liquid or solid phases and support diverse cellular functions by concentrating specific molecules to facilitate enzymatic reactions, storage, or molecular sequestration^1–5^. The synthesis of artificial membrane-less organelles (AMOs) is of great interest in the field of synthetic biology because it enhances our knowledge of the fundamental aspects of MOs and their potential to create cells that adopt new functional modules. Thus far, various attempts have been made to construct AMOs with novel functions in bacterial and animal cells or in vitro^6–13^. However, studies on AMOs in plant cells specifically are scarce.

Proteins that drive phase separation have domains involved in multivalent, weak, non-covalent interactions between biomolecules, as well as intrinsically disordered regions (IDRs), low-complexity regions (LCRs), or prion-like domains (PrLDs) that allow conformational changes, thereby facilitating the assembly of molecules^1–3,14–21^. Multivalent interactions involved in phase separation include electrostatic interactions, π-cation and π–π interactions, dipole-dipole interactions, hydrophobic interactions, and interactions between β-sheets, causing cross-β amyloid-like structures^14,19–26^. The drivers of biomolecular condensates are called scaffolds, as the molecules essential for condensate formation can be a single protein, a combination of multiple proteins, or a mixture of proteins and nucleic acids^2,27^. Scaffolds can interact with molecules called clients that are dispensable for condensate assembly but can be included in the condensates^2,27^.

The properties of biological condensates affect their functions^12,22,24,28,29^. For example, the molecular dynamics of condensates, that is, the mobility of condensate constituents, modulate the rates of biochemical reactions inside or outside the condensate^12,22,28^. Such material properties range from liquid-like properties with high mobility and exchange of proteins between the dilute and concentrated phases to solid-like properties, including those of Maxwell glasses (glass-like) and cross-linked gels (gel-like) with negligible exchange between the two phases^7,22–24,29,30^. These properties depend on the mode of intermolecular interactions in the condensate constituents^25,31^. Other condensate properties, such as size, number, shape, and subcellular localization, should also be determined by interactions among molecules within the condensate, as well as the interaction between molecules on the surface of the condensate and those in the external area in contact with the condensate^32,33^. However, condensate component proteins are often subjected to complex cellular regulation through domain-domain interactions with various other proteins and post-translational protein modifications^28,34^. Accordingly, the contributions of individual molecules to the properties of condensates in living cells are not yet fully understood^7^. This is particularly true for condensates composed of multiple types of scaffold proteins^28^. By studying the relationship between intermolecular interactions and condensate properties, the properties of newly created condensates can be predicted. However, owing to the limited number of available studies, only biased models can be developed.

Membrane-bound condensates are another important type of biological condensates. They not only form membrane-associated compartments that regulate various membrane-mediated cellular functions, such as neuronal synapse^35,36^, T cell receptor (TCR) signaling clusters^37,38^, tight junction formation^39,40^, autophagosome assembly^41^ and endocytosis^42^, but also mediate membrane remodeling^43–45^. The phenomenon of ‘wetting’ has been proposed to contribute to the behaviours of condensates and membranes, depending on the states of condensate-membrane surface interactions^43–45^.

Increasing evidence suggests that many condensation-prone proteins in plant cells undergo phase separation and form membrane-less compartments. They are not limited to homologs of proteins that cause phase separation in animal cells but also include plant-specific proteins^23^. In Arabidopsis thaliana, these are exemplified by the proteins FLOWERING CONTROL LOCUS A (FCA) and FLX-like 2 (FLL2) that form FCA nuclear bodies and compartmentalize 3’-end RNA-processing factors^46^, FLOE1 that undergoes hydration-dependent phase separation and regulates seed germination^47^, the auxin response factor family proteins ARF7 and ARF19, forming cytoplasmic assemblies that limit auxin responsiveness^48^, a component of the evening complex, EARLY FLOWERING 3 (ELF3), that forms speckles under a high-temperature condition^49^, and CRYPTOCHROME 2 (CRY2) and TEOSINTE BRANCHED1-CYCLOIDEA-PCF 22 (TCP22) that form photo-bodies in a blue light-dependent manner^50^. For proteins involved in the formation of membrane-associated condensates, the DIX domain-containing protein SOSEKI as a cell polarity determinant^51^, the endocytic TPLATE complex subunit AtEH1 (Eps15 homology protein1)^52^, and SEC FOURTEEN-HOMOLOG8 (SFH8), which regulates the polarity of the auxin efflux carrier PINFORMED2, likely depending on the cleavage of SFH8 IDR^53^, were documented.

Our previous study showed that rice (Oryza sativa) JAZ2 (OsTIFY5), an atypical JAZ (JASMONATE ZIM DOMAIN protein) repressor of jasmonate signaling, forms condensates in both the nucleus and cytosol when transiently expressed^54^. JAZ proteins are nuclear factors that contain conserved TIFY (ZIM) and Jas domains; TIFY are involved in transcriptional repression complex formation through their interaction with novel interactor of JAZ (NINJA) and mediate the homo- or hetero-dimerization of some JAZ proteins^55^, whereas Jas domains interact with the COI1 receptor in response to jasmonate perception, leading to the degradation of JAZ proteins^56^. Several JAZ factors, including OsJAZ2, have their own EAR motifs^56,57^. Contrary to typical JAZ factors, the COI1-binding site is not conserved in OsJAZ2^57–59^. In rice cells, OsJAZ2 overexpression causes the formation of solid-to-gel-like condensates, both in the nucleus and cytosol, which are often larger, particularly within the cytosol^54^. The conserved TIFY and Jas domains, together with the LCRs, contribute to the formation and properties of the JAZ2 condensate. The Jas domain is required for JAZ2 condensation, whereas the N-terminal EAR motif-containing region and TIFY domain contribute to the solidification of OsJAZ2 condensates^54^. This raises the hypothesis that condensation-prone JAZ protein(s) might incorporate other JAZ factors into the condensates, thereby tuning jasmonate signaling by decreasing the stability and net concentration of JAZ factors in the nucleoplasm depending on the cellular conditions. Consistent with this, OsJAZ9 (a typical JAZ factor), but not OsJAZ9-interacting OsbHLH094^57^, can be incorporated into JAZ2 condensates^54^. In Nicotiana benthamiana, OsJAZ2 forms condensate with high protein mobility^54^. This implies that the properties of OsJAZ2 could be modified by other components.

We have previously shown that another intrinsically disordered protein (IDP) of rice, RSS1 (RICE SALT SENSITIVE1), has properties opposite to those of condensation-prone proteins^60,61^. RSS1 is a hydrophilic protein that is required for meristematic activity and, thereby, for plant growth and viability under environmental stress conditions, such as salinity, cold, and dehydration^60^. The precise molecular function of RSS1 is unclear, although it interacts with protein phosphatase 1 and is degraded in a cell cycle phase-dependent manner^60,61^. Intriguingly, recombinant RSS1 does not aggregate, even after exposure to high temperatures of 80–100 °C^61^. This inspired us to examine whether a chimeric protein between OsJAZ2 and RSS1 forms condensates with altered properties or whether it even shows a tendency to resist condensation.

To utilize AMOs as a platform for functional modules in cells, their design and customization are necessary according to the functions intended to be added^34^. Therefore, in this study, we generated various types of artificial condensates in plant cells, as illustrated in Fig. 1. We investigated methods to control the subcellular localization and protein mobility of AMOs in rice protoplast by transient assay (Fig. 1a, b) using condensate-prone proteins, such as OsJAZ2^54^ and AtFCA^46^, as model scaffold proteins. Moreover, we showed that the size and number of JAZ2 condensates (Fig. 1c), in addition to their localization, could be altered by the insertion of a portion of RSS1^60,61^ into the structurally flexible region of JAZ2. We also demonstrated the recruitment of proteins of interest (Fig. 1d, e) via nanobody^62^ binding to a specific antigen, rapamycin-mediated binding of FKBP (FK506-binding protein) and FRB (FKBP12-rapamycin-binding domain)^63^, as well as by direct fusion to the condensate-prone scaffold proteins themselves. Furthermore, we successfully constructed hybrid condensates that combined two condensate properties using two types of scaffold proteins (Fig. 1f). We further demonstrated that some of these AMOs could be constructed in transgenic Arabidopsis plants. These results provide a framework for the construction of artificial membrane-less condensates for plant cells.

Fig. 1 Schematic design of artificial membrane-less organelles (AMOs) in this study.The properties and functionality of artificial condensates (drawn as orange circles or cloud shapes) aimed to be controlled in this study are shown. a Subcellular localization of the condensates. Upper two in the cytosol (left) and nucleus (right). Lower two anchored at the plasma membrane (PM) (left) and conjugation of PM-anchored condensates (dark orange) and cytosolic condensates (light orange) (right). b Dynamicity of the condensate constituents. Condensate constituents with high (top) and low (bottom) molecular mobility are shown. Condensation-prone proteins are shown with smaller circles. c Condensate size. Large (top) and small (bottom) condensates in the nucleus are shown. d Enzymatic activity assigned to the condensates. e Recruitment of proteins of interest (yellow circles) to the condensates. In the right panel, recruitment is induced by a chemical (triangle). f A condensate with heterogeneous properties consisting of condensation-prone proteins with high (light orange) and low (dark orange) molecular mobility. Exchangeable condensation-prone proteins are shown with smaller orange circles.

Results

Selectable subcellular localization of artificial condensates

We examined whether the subcellular localization of AMOs could be altered in plant cells. For this analysis, we used a condensation-prone protein from rice, OsJAZ2. Nuclear localization of JAZ2 fused with the enhanced cyan fluorescent protein (eCFP-JAZ2) and its condensates depends on the nuclear localization signal (NLS) in the Jas domain, the truncation of which (eCFP-JAZ2ΔJas-2) results in condensate formation occurring predominantly in the cytosol (Fig. 2a, b and Supplementary Fig. 1a) and the deletion of which causes a decrease in the formation of the JAZ2 condensate overall as reported previously^54^. Upon addition of synthetic NLS^64^, modified JAZ2 (eYFP-JAZ2ΔJas-2-NLS-FLAG) formed condensates that were localized exclusively in the nucleus (Fig. 2b and Supplementary Fig. 1b). Since Jas domains are required for efficient condensate formation^54^, these results suggest that intermolecular interactions via the Jas domain perturb nuclear transport of the JAZ2 fusion proteins and that addition of synthetic NLS promotes condensation in the nucleus exclusively.

Fig. 2 Localization of the artificial condensates can be controlled.a Schematic drawing of OsJAZ2 and its deletion constructs used for the condensation assay. TIFY and Jas domains, the EAR motif, and low-complexity regions (LCRs) are depicted. a.a., amino acids. b Localization patterns of condensates formed by eCFP-JAZ2, eCFP-JAZ2-ΔJas-2, and eYFP-JAZ2-ΔJas-2-NLS-FLAG in O. sativa protoplasts. eYFP-bHLH094 (upper and middle) and bHLH094-eCFP (lower) visualize the nuclei. mRFP1 visualizes the cytosol and nucleus. c Condensation patterns of SCAMP1ΔN118-fused eCFP-JAZ2 in a representative O. sativa protoplast expressing eYFP-bHLH094 and mRFP1 markers. CFP condensates are observed as foci at two different focal planes (upper and lower panels) as illustrated. d Co-condensation of eCFP-JAZ2-ΔLCR5Jas and SCAMP1ΔN118-eYFP-JAZ2 in O. sativa protoplasts. Top, fluorescence patterns of eCFP-JAZ2-ΔLCR5Jas. Bottom, representative images of co-condensation of eCFP-JAZ2-ΔLCR5Jas and SCAMP1ΔN118-eYFP-JAZ2 in an O. sativa protoplast co-expressing both the constructs. The cytosol and nucleus are visualized by mRFP1. CFP and YFP foci in the same cell at two different focal planes are indicated as in (c) (upper and lower images). The rate of cells with CFP-YFP co-condensates is shown (n, the number of cells examined). 5 μg of plasmids were introduced into the protoplasts to express the respective constructs, except for the 10 μg to express SCAMP1ΔN118-eYFP-JAZ2 (d) to detect obvious SCAMP1ΔN118-eYFP-JAZ2 condensates. Typical JAZ2 condensates and the nucleus are indicated by arrowheads (white and orange, respectively). Vac, vacuole. Scale bar, 5 µm.

Next, we explored whether the artificial condensates could be anchored to the plasma membrane (PM). To this end, we fused eCFP-JAZ2 with OsSCAMP1ΔN118, a truncated version of the transmembrane protein OsSCAMP1 (Oryza sativa SECRETORY CARRIER MEMBRANE PROTEIN1) that is localised to the early endosomal compartment, trans-Golgi network, and PM^54,65,66^ (Supplementary Fig. 1c). OsSCAMP1ΔN118-eCFP and OsSCAMP1ΔN118-eYFP were observed throughout PM without condensation, although they were unevenly distributed (Supplementary Fig. 1d, e). In contrast, OsSCAMP1ΔN118-fused eCFP-JAZ2 and eYFP-JAZ2 formed condensates that were frequently observed as foci in PM but also in the internal region of rice protoplasts (Fig. 2c and Supplementary Fig. 1e, f), probably reflecting the localization of OsSCAMP1^65,66^. We further tested whether cytosolic JAZ2 could be incorporated into PM-anchored condensates. We expressed eCFP-JAZ2-ΔLCR5-Jas that cannot form condensates alone, together with OsSCAMP1ΔN118-eYFP-JAZ2. This resulted in the co-localization of eCFP and eYFP foci (Fig. 2d), indicating the assembly of cytosolic JAZ2 and OsSCAMP1-fused JAZ2. The co-condensates were observed not only in the vicinity to PM but also in the internal region of the cells (Fig. 2d and Supplementary Fig. 1g, h). Formation of co-condensates in the PM vicinity was observed slightly more often, when compared to co-condensates formed by co-expression of eCFP-JAZ2-ΔLCR5-Jas and eYFP-JAZ2.

Similar experiments were performed with another condensation-prone protein, OsNup98. Nup98 is a well-conserved nuclear pore protein that contains FG domains rich in Phe-Gly^67,68^. Nup98 forms a hydrogel through interactions between regions containing FG repeats^69,70^. When overexpressed in rice and N. benthamiana leaf cells, both OsNup98 and the N-terminal FG domain of OsNup98, designated Nup98-FG1, formed condensates in the nucleus and cytosol (Supplementary Figs. 2a–c, 3a). Upon fusion with OsSCAMP1ΔN118, the Nup98-FG1 condensate was frequently associated with PM in rice cells (Supplementary Fig. 4), similar to the manner of the JAZ2-based construct. Thus, we demonstrated that membrane-anchored condensates containing cytosolic proteins can be created. Taken together, these results suggest that the subcellular localization of AMOs can be controlled in plant cells.

Control of protein mobility of artificial condensates

Next, we investigated methods to control the protein mobility of AMOs in plant cells. The fluorescence recovery after photobleaching (FRAP) technique^71^ was used to monitor condensate constituent dynamics. In FRAP experiments, fluorescent molecule-labeled proteins in a defined condensate region are irreversibly bleached, and the recovery of the fluorescent signal in the bleached area by the exchange of bleached and non-bleached proteins is monitored^71^. A high recovery of the fluorescence signal is observed with liquid-like properties of the condensates, whereas a lower recovery is associated with solid-like properties. Previous FRAP analyses have shown that JAZ2 primarily forms solid-like condensates in rice protoplasts and liquid-like condensates in N. benthamiana cells^54^. Similarly, while Nup98-FG1 formed solid-like condensates in rice protoplasts, both solid and liquid condensates were formed in N. benthamiana cells (Supplementary Figs. 2d, 3b). Liquid condensates are often spherical owing to the principle of interfacial energy minimisation, whereas condensates that exhibit more solid-like properties and do not grow isotropically tend to be nonspherical^3^. Consistent with this, liquid-like condensates formed by JAZ2^54^ and Nup98-FG1 in N. benthamiana cells were nearly spherical, whereas solid-like JAZ2^54^ and Nup98-FG1 condensates formed in rice protoplasts and N. benthamiana cells were often non-spherical (Supplementary Figs. 2, 3). Thus, JAZ2 and Nup98-FG1 condensates share similar properties. In contrast, another condensation-prone polypeptide, the region containing the PrLDs of AtFCA (FCA-PrLD), formed spherical, liquid-like condensates in rice (Fig. 3a–c), similar to that seen in Arabidopsis and tobacco cells^46^. Therefore, we examined the liquidity of the artificial condensates formed using the chimeric proteins JAZ2 and FCA-PrLD.

Fig. 3 Dynamicity of condensates is altered by chimeric proteins of FCA-PrLD and JAZ2.a Schematic structure of FCA, JAZ2-FCA-PrLD, JAZ2N85-FCA-PrLD, and its variants. FCA has two RNA Recognition Motifs (RRMs) on the N-terminus and two PrLDs on the C-terminus. For the chimeric construct JAZ2-FCA-PrLD, the FCA-PrLD region, as highlighted in light blue, was inserted between the LCR2 and LCR3 of JAZ2. b Representative fluorescence images of O. sativa protoplasts expressing eCFP-fused FCA-PrLD and chimeric constructs of FCA-PrLD and JAZ2 (left). The nucleus and cytosol are visualized by expression of eYFP-bHLH094 and mRFP1. CFP condensates, subjected to FRAP, and the nucleus are indicated by arrowheads (white and orange, respectively). Vac, vacuole. Scale bar, 5 µm. FRAP analyses of condensates formed by the eCFP-fused constructs after photo-bleaching at the position indicated by yellow arrowheads (right). Time 0, the time point of the bleaching pulse. Scale bar, 1 μm. c Normalized CFP fluorescence intensity of condensates formed by FCA-PrLD, JAZ2-FCA-PrLD, and JAZ2N85-FCA-PrLD after photo-bleaching as shown in (b) (mean ± SD; n = the number of condensates). d FRAP analyses of the deletion constructs of JAZ2N85-FCA-PrLD. Normalized CFP fluorescence intensities for respective constructs are shown by distinct colors as indicated (mean ± SD; n = 18 condensates for ΔEAR-TIFY, 14 for ΔEAR, 11 for ΔNterm, 15 for ΔTIFY, 4 for ΔLCR1). The representative images of condensates before and after bleaching are shown in Supplementary Fig. 5a.

We generated a construct in which FCA-PrLD was inserted between LCR2 and LCR3 in the structurally flexible region of JAZ2 fused with eCFP (Fig. 3a). Since the LCR2 of JAZ2 only overlaps with the region with a high PrLD score^54^, insertion at this site was not expected to inhibit JAZ2 condensation itself. The resulting eCFP-JAZ2-FCAPrLD formed mostly solid condensates with low molecular mobility (Fig. 3b, c), as there was little indication of exchange between the bleached and unbleached molecules. Notably, when only the 85 amino acids of the N-terminal of JAZ2 (JAZ2N85) containing the TIFY domain were fused to FCA-PrLD, solid condensates were formed (Fig. 3a–c and Supplementary Figs. 5, 6). This solidification effect of JAZ2N85 was compromised by deletion of the EAR motif and TIFY domain of JAZ2, whereas deletion of LCR1 and LCR2 exhibited the opposite effect (Fig. 3a, d and Supplementary Fig. 5). Consistent with this, eCFP-JAZ2N85-FCA-PrLDΔEAR-TIFY showed higher molecular mobility than eCFP-FCA-PrLD **(**Fig. 3c, d). Thus, the dynamics of AMOs can be altered by a combination of protein domains. Interestingly, the effect of deletions appeared to differ between the EAR motif and TIFY domains. In the case of TIFY domain deletion alone, CFP fluorescence rapidly recovered after photobleaching, whereas EAR motif deletion alone caused slower fluorescence recovery (Fig. 3a, d and Supplementary Fig. 5a). This suggests that the intermediate states between the liquid- and solid-type condensates can be adjusted.

Alteration of JAZ2 condensate size and number by insertion of the OsRSS1 sequence

IDR sequences promote or inhibit condensate formation; sticky IDRs often promote condensation, whereas other IDRs prevent condensation^26^. We examined whether a sequence of an IDP, such as OsRSS1, when inserted into JAZ2, affected the properties of JAZ2 condensates (Fig. 4a). RSS1 is a highly hydrophilic protein that contains both acidic and basic regions (Supplementary Fig. 7). We used truncated RSS1 (RSS172-243) that is devoid of the N-terminal region containing D-boxes, through which RSS1 is degraded in a cell cycle-dependent manner^60^. We generated the construct eCFP-JAZ2-RSS172-243, in which RSS172-243 was inserted between LCR2 and LCR3 of JAZ2 (Fig. 4a). JAZ2 forms condensates not only in the nucleus but also in the cytosol in rice cells, often forming large condensates as previously reported^54^ (up to 4 μm in diameter in the nucleus and 7 μm in diameter in the cytosol) (Fig. 4b, c, Supplementary Fig. 8a–c). In contrast, chimeric JAZ2-RSS1 formed condensates more frequently in the nucleus than in the cytosol (Fig. 4b, c, Supplementary Fig. 8a–c). Moreover, JAZ2-RSS1 often formed clusters of small condensates in the nucleus (Fig. 4b, d, Supplementary Fig. 8c, d). No large JAZ2-RSS1 condensates (>2 μm in diameter) were observed either inside or outside the nucleus (Fig. 4b). These results indicate that RSS1-IDR insertion alters the size and localization of condensates formed by JAZ2. Considering that JAZ2, though not RSS1, has an NLS^54,60^, these results suggested that the insertion of RSS1-IDR prevented the condensation of JAZ2 in the cytosol, consequently promoting its active transport to the nucleus.

Fig. 4 Localization, size, and number of condensates are altered by the insertion of RSS1 to JAZ2.a Schematic drawing of RSS1 and JAZ2-RSS1. Intrinsically disordered regions (IDRs) of RSS1 predicted by PLAAC are indicated above the domain structure of RSS1. The DEN-box, D-box and D-box-like (the target sequences for degradation by the 26S proteasome), WAGE motif, C-terminal D and E-rich region^60^ are depicted. In JAZ2-RSS1, the region of RSS172-243, indicated by light blue highlight, is inserted between LCR2 and LCR3 of JAZ2. b Representative fluorescence patterns of eCFP-JAZ2 and eCFP-JAZ2-RSS1 in Oryza sativa protoplasts. The nucleus and cytosol are visualized by eYFP-bHLH094 and mRFP1 markers. Typical JAZ2 condensates and the nucleus are indicated by arrowheads (white and orange, respectively). Top two, large condensates formed by eCFP-JAZ2 in the vicinity of the nucleus (upper) and within the nucleus (lower) are shown. Bottom two, two (upper) and many (lower) small condensates formed by eCFP-JAZ2-RSS1 in the nucleus are shown. The images of the cell with JAZ2 condensates in the nucleus (top, lower panels) were obtained with a 60× objective lens, while all other cell images were obtained with a 20× objective lens. Vac, vacuole. Scale bars, 5 μm. c Proportion of cells in which condensates were observed only inside the nucleus and only outside the nucleus or both inside and outside (mean ± SD; n = 3 pooled experiments). A total of 4 to 10 cells for each construct per experiment were examined in each experiment. The numbers above the bars indicate the number of cells observed in the classified localization. d Comparison of condensate number per cell (mean ± SD). n means total number of the cells examined in three experiments. For (c) and (d), the data were obtained from the cells exhibiting condensation in the experiments as in (b). The relationship between the construct and localization (c) or number (d) of the condensates was assessed with Fisher’s exact test and Brunner-Munzel test, respectively. The cell with JAZ2 condensate in the nucleus in (b) was observed in an independent experiment, thus not counted in (c) and (d).

To evaluate the possibility whether the differences in fluorescence patterns would be due to differences in the protein expression levels, we quantified whole-cell CFP fluorescence intensity as an indicator of protein levels in each cell. In the same experiment, there was no significant difference in the expression level of eCFP-JAZ2 between the cells that formed condensate in the [nucleus only] and cells that formed condensate in the [nucleus and cytosol] or [cytosol only] (Supplementary Fig. 8a, right). The same was true for eCFP-JAZ2-RSS1 (Supplementary Fig. 8b, right). However, in the experiment with relatively low expression levels, JAZ2 formed condensates in both the nucleus and cytosol, whereas JAZ2-RSS1 condensates were detected in the nucleus only (Supplementary Fig. 8a, left). In contrast, in the experiment with relatively higher expression levels, cells expressing JAZ2 did not exhibit condensates that formed exclusively in the nucleus, whereas JAZ2-RSS1 condensates were observed throughout the nucleus and cytosol (Supplementary Fig. 8b, right). Therefore, the localization pattern of condensates might also be affected by protein expression levels, depending on the experimental conditions.

We further compared the cells that exhibited condensation only in the nucleus. The number of condensates was significantly increased in the cells expressing eCFP-JAZ2-RSS1 compared to the cells expressing eCFP-JAZ2 (Supplementary Fig. 8d, centre), while the fluorescence levels were comparable between these cells (Supplementary Fig. 8d, right). This increase was observed in conjunction with increased protein partitioning or enrichment in the condensates (Supplementary Fig. 8e-f). Therefore, in the nucleus, JAZ2-RSS1 exhibited a higher propensity to condense than JAZ2 did. It is also noteworthy that in the plotted graph in Supplementary Fig. 8d (left), for both eCFP-JAZ2 and eCFP-JAZ2-RSS1, condensate formation was more frequently observed in cells where CFP fluorescence intensity exceeded a certain threshold level. This is consistent with that a condensation-prone protein begins to form condensates when it reaches the saturation concentration^25,31^. This also supports that the quantified florescence intensity indeed reflects the fluorescent protein levels.

We investigated whether the altered properties of JAZ2 due to the RSS1 sequence insertion were coupled with changes in condensate molecular dynamics. FRAP analysis showed that JAZ2-RSS1 formed solid-like condensates, where bleached eCFP-JAZ2-RSS1 molecules were scarcely replaced by unbleached molecules, regardless of their size or location (Supplementary Fig. 9). Thus, the molecular mobility of the JAZ2 condensates was decreased by RSS1 insertion.

Creation of artificial condensates containing functional proteins

To evaluate AMOs as a platform for assembling and operating functional modules in plant cells, we designed a model condensate composed of FCA-PrLD proteins fused to a Tobacco Etch Virus (TEV) proteaseΔ220-242^72^ and a marker substrate eYFP-eCFP, respectively (Fig. 5a; FCA-PrLD-TEV proteaseΔ220-242 and eYFP-[TEV cleavage site]-eCFP-FCA-PrLD). The substrate contained a TEV protease cleavage site between eYFP and eCFP, enabling the monitoring of eYFP separation from eCFP-FCA-PrLD by observing a reduction in the level of colocalized YFP and CFP fluorescence in response to the TEV protease reaction. Before testing the monitoring system, we confirmed that the fusion of the TEV protease with FCA-PrLD did not perturb liquid-like condensate formation in rice cells (Supplementary Fig. 10).

Fig. 5 The TEV protease reaction is promoted by fusion of the enzyme and substrate with FCA-PrLD.a Fluorescence patterns of the cells with or without TEV protease activity. The substrate construct eYFP-eCFP-FCA-PrLD, containing a TEV protease cleavage site between eYFP and eCFP-fused FCA-PrLD, was expressed in Oryza sativa protoplasts with TEV proteaseΔ220-242 either with or without fusion to FCA-PrLD. Upper, a representative fluorescence pattern of a cell expressing eYFP-eCFP-FCA-PrLD and mRFP1 marker proteins, classified as [class1]. CFP and YFP fluorescence are observed exclusively at the FCA-PrLD condensates. Middle, a representative fluorescence pattern of a cell expressing eYFP-eCFP-FCA-PrLD, TEV proteaseΔ220-242, and mRFP1, classified as [class2]. While CFP fluorescence is observed exclusively at the FCA-PrLD condensates, YFP fluorescence is observed in part at the FCA-PrLD condensates, however, more widely in the cytosol and the nucleus. Lower, a representative fluorescence pattern of a cell expressing eYFP-eCFP-FCA-PrLD, FCA-PrLD-TEV proteaseΔ220-242, and mRFP1, classified as [class3]. While CFP fluorescence is observed exclusively at the FCA-PrLD condensates, YFP fluorescence is not, however, is instead widely detected in the cytosol and the nucleus. Condensates are indicated by white arrowheads. Vac, vacuole. Scale bar, 5 μm. b The rate of cells exhibiting fluorescence patterns is classified as in (a) (mean ± SD, n = 3 pooled experiments). A total of 7 to 13 cells per protease construct were examined in each experiment. The number of cells observed in the classified pattern is shown on the right. Differences in categorical variables were assessed with Fisher’s exact test. c Relationship between the proportion of condensate CFP fluorescence per whole cell CFP fluorescence (PCFP foci) and the degree of reaction progress. The difference between PCFP foci and PYFP foci (the proportion of condensate YFP fluorescence per whole cell YFP fluorescence) was calculated as an indicator of the reaction progress (See “Methods” for details). d Relationship between the amount of substrate expressed and the degree of reaction progress. The substrate amount is indicated with whole-cell CFP fluorescence intensity (arbitrary units, a.u.) in each cell. For (c) and (d), data from Experiment 3 (n = 12 cells for each) are shown with a correlation coefficient (r). Data from the other experiments are shown in Supplementary Fig. 11c, d.

We examined the localization patterns of the marker proteins by co-expressing the designed FCA-PrLD fusion proteins. Without the enzyme construct, the substrate eYFP-eCFP-FCA-PrLD frequently formed condensates, resulting in overlapping YFP and CFP foci classified as the pattern [class1] (Fig. 5a, b and Supplementary Fig. 11a, b). When co-expressed with the TEV proteaseΔ220-242, fluorescence patterns classified into [class2], where YFP signals were less co-localized with CFP foci (FCA-PrLD condensates) and increased outside of the condensates, were frequently observed. When co-expressed with TEV proteaseΔ220-242 fused to FCA-PrLD, most cells exhibited no YFP signals at the CFP foci, as depicted in the pattern [class3], indicating efficient eYFP separation. Consistent with this, the difference between proportions of fluorescence in CFP condensates and YFP condensates to the whole-cell fluorescence of CFP and YFP, respectively, showed a decrease in YFP condensate fluorescence by the expression of FCA-PrLD-fused TEV proteaseΔ220-242 (Fig. 5c, d and Supplementary Fig. 11c, d). The release of eYFP was less prominent in cells in which CFP fluorescence was not enriched in the condensates (Fig. 5c and Supplementary Fig. 11c). These results suggest that the co-localization of an enzyme with a specific substrate to create artificial condensates promotes successful enzymatic reactions.

In this system, because of the liquidity of the FCA-PrLD condensates, FCA-PrLD-fused proteins are expected to be exchangeable between the inside and outside of the condensates. This raises the possibility that eCFP-FCA-PrLD separated outside the condensates can also be incorporated into the condensates. In such a case, one may argue that the increased enzymatic reaction might reflect a higher association between the FCA-PrLD-fused substrate and the enzyme and not necessarily occur in the condensates. To verify this, we hypothesised that if all or most of the enzymatic reactions occurred outside the condensate, the reaction progress in a cell should correlate with the expression levels of the substrate. The degree of reaction progress was assessed as the difference between the proportion of CFP intensity in the CFP foci (PCFP foci) and that of YFP intensity in the YFP foci (PYFP foci). The expression levels of the substrate were estimated from the total CFP intensity in a cell (see “Methods” for details). Consistent with the hypothesis, the expression of non-fused TEV proteaseΔ220-242 resulted in a strong correlation (Fig. 5d and Supplementary Fig. 11d). Conversely, by the expression of TEV proteaseΔ220-242 fused to FCA-PrLD, this score was not necessarily well correlated with the whole cell CFP intensity (Fig. 5d and Supplementary Fig. 11d), suggesting that some reaction occurred inside the condensates.

Recruitment of proteins of interest to artificial condensates via nanobodies and FKBP

To develop a system to recruit a wide variety of proteins, we generated eCFP-JAZ2 and eCFP-FCA-PrLD constructs that were fused to an anti-GFP nanobody (GFP-binding protein, GBP), which binds specifically to GFP and YFP, not to CFP^73^ (Fig. 6a, b; GBP-eCFP-JAZ2 and eCFP-FCA-PrLD-GBP). These fusion proteins retained the ability to form condensates in rice cells (Fig. 6c, d). When co-expressed, eYFP co-localized with the respective condensates (Fig. 6a–d), indicating that the proteins of interest were recruited via the nanobody. Although GBP fusion occasionally caused a decrease in the frequency of condensate formation, this was restored by the co-expression of eCFP-JAZ2 and eCFP-FCA-PrLD that were not fused to GBP (Fig. 6c, d). The dynamics of the GBP-eCFP-JAZ2 condensates were similar to those of the eCFP-JAZ2 condensates (Supplementary Fig. 12a). Co-expression of eYFP slightly increased the mobile fraction of GBP-eCFP-JAZ2 condensates, as observed in the CFP-YFP co-condensates; however, this was compromised by the additional expression of non-GBP-fused eCFP-JAZ2 (Supplementary Fig. 12b, c). In contrast, the molecular dynamicity of the eCFP-FCA-PrLD-GBP condensates was reduced (Supplementary Fig. 12d) compared to that of the eCFP-FCA-PrLD condensates (Fig. 3a–c). However, even in this case, the dynamicity of the eCFP-FCA-PrLD-GBP condensates was restored by co-expression of eYFP (as observed in the CFP–YFP co-condensates) with or without non-GBP-fused eCFP-FCA-PrLD (Supplementary Fig. 12e, f). Thus, the properties of the co-condensates of eYFP and GBP-fused scaffold proteins appear to be similar to those of the GBP-free condensates. Similar tendencies were also observed in experiments using eCFP-JAZ2-GBP, in which GBP was fused to the opposite end of eCFP-JAZ2 (Supplementary Fig. 13). These results indicated that a nanobody-based system for recruiting specific proteins to AMOs can be constructed using condensation-prone proteins that form solid- or liquid-type condensates.

Fig. 6 Proteins of interest can be recruited to the JAZ2 and FCA-PrLD condensates using nanobody or rapamycin-mediated induction systems.a Fluorescence patterns of eYFP co-expressed with (top) eCFP-JAZ2 or (bottom) GBP-eCFP-JAZ2 in Oryza sativa protoplasts. The nucleus and cytosol are visualized by expression of mRFP1. CFP condensates are indicated by white arrowheads. Co-condensation of YFP and CFP fluorescence is observed by co-expression of eYFP with GBP-eCFP-JAZ2 but not with eCFP-JAZ2. Note that similar fluorescence patterns are observed between RFP and YFP, except at the site of co-condensation of CFP and YFP. Scale bar, 5 μm. b Fluorescence patterns of eYFP co-expressed with (top) eCFP-FCA-PrLD or (bottom) eCFP-FCA-PrLD-GBP in O. sativa protoplasts. The nucleus and cytosol are visualized by expression of mRFP1. CFP condensates are indicated by white arrowheads. Co-condensation of YFP and CFP fluorescence is observed by co-expression of eYFP with eCFP-FCA-PrLD-GBP but not with eCFP-FCA-PrLD. Scale bar, 5 μm. c, d Frequency of CFP condensate and CFP-YFP co-condensate formation. The rate of cells with CFP condensate and CFP-YFP co-condensate formation by expression of the indicated constructs with mRFP1 is shown (n = 3 pooled experiments). Ten cells were examined for each construct set in each experiment. Different letters above the bars indicate statistically significant differences, determined using one-way ANOVA with Games-Howell post hoc test. All p-values were shown in Supplementary Table 2. e Time-lapse images of an O. sativa protoplast expressing eCFP-JAZ2, FKBP-eCFP-JAZ2, and eYFP-FRB after treatment with 1 μM rapamycin. The nucleus and cytosol are visualized by mRFP1 marker. The images were obtained every 2 min. JAZ2 condensates are indicated by white arrowheads. Co-condensation of YFP in CFP condensates began to be observed 5 min after rapamycin treatment. Scale bar, 5 µm.

We also examined rapamycin-mediated FKBP-FRB interaction^63,74^. Co-expression of FKBP-eCFP-JAZ2 and eYFP-FRB resulted in the rapid co-condensation of CFP and YFP after rapamycin treatment, whereas treatment with DMSO alone did not induce co-condensation (Fig. 6e and Supplementary Fig. 14). These results indicate that the recruitment of proteins of interest to AMOs can be induced by the FKBP-rapamycin-FRB system as well.

Construction of artificial hybrid condensates with two different material properties

Next, we constructed AMOs that combined two condensate-prone proteins, eCFP-JAZ2 and eYFP-FCA-PrLD, using the affinity between GBP and eYFP. These proteins did not form co-condensates when co-expressed (Supplementary Fig. 15a). The expression of either GBP-eCFP-JAZ2 or eCFP-JAZ2-GBP, forming solid-type condensates, with eYFP-FCA-PrLD, forming liquid-type condensates, resulted in the co-localization of CFP and YFP foci (Supplementary Fig. 15b, c), indicating co-condensation of both combinations of proteins. The resulting co-condensate was considered to be almost a single miscible phase because the CFP and YFP signals overlapped. Due to the higher frequency of co-condensation of eCFP-JAZ2-GBP and eYFP-FCA-PrLD (Supplementary Fig. 15b, c), we further analyzed these co-condensates by FRAP (Fig. 7). This revealed that the observed condensates exhibited different properties depending on the ratio of the amount of plasmid used for expression. When the amount of the plasmid expressing eCFP-JAZ2-GBP was higher than that of the plasmid expressing eYFP-FCA-PrLD, both the CFP and YFP fluorescence of the formed condensates did not recover or only slightly recovered after bleaching (Fig. 7a, c, d and Supplementary Fig. 16a; designated as Type I condensate). In contrast, when the eYFP-FCA-PrLD-expressing plasmid was introduced in excess, a substantial portion of the condensates showed markedly high recovery of YFP fluorescence after bleaching, while CFP fluorescence remained relatively low (Fig. 7b–d, and Supplementary Fig. 16b, c; designated as Type II condensate). These results indicate that nanobody-mediated binding of the two scaffold proteins made it possible to construct artificial hybrid condensates that possess two different properties in terms of their molecular dynamics.

Fig. 7 Artificial condensates with two different material properties are constructed using nanobodies to recruit condensation-prone proteins.a, b Representative FRAP of co-condensates of eCFP-JAZ2-GBP and eYFP-FCA-PrLD. Left (top), co-condensation of CFP and YFP observed in Oryza sativa protoplasts expressing eCFP-JAZ2-GBP and eYFP-FCA-PrLD, visualized by mRFP1 marker. The amount of plasmids introduced into the protoplasts was indicated. Condensates indicated by white arrowheads were subjected to FRAP analysis. Scale bar, 5 μm. Left (bottom), representative images of CFP–YFP co-condensates before and after photobleaching. Images with no recovery of either CFP or YFP fluorescence are shown in (a). Images with prominent recovery of YFP but not CFP fluorescence are shown in (b). Yellow arrowheads indicate the bleaching site. Scale bar, 1 μm. Right, the normalized FRAP intensity of (top) CFP and (bottom) YFP corresponds to the left FRAP images. c Rate of cells classified into two types by the FRAP experiments as performed in (a, b). Type I, recovery of CFP and YFP fluorescence in normalized FRAP intensity is highly overlapping. Type II; recovery of YFP fluorescence in normalized FRAP intensity is obviously higher than that of CFP fluorescence. The numbers above the bars indicate the number of Type I and Type II condensates, respectively. Differences in categorical variables were assessed with Fisher’s exact test. d Normalized intensity at 30 ± 1 s after bleaching of eCFP and eYFP condensates in the FRAP experiments performed as in (a, b) (mean ± SD). n, the number of CFP-YFP co-condensates. P-values were determined by two-tailed Welch’s t-test (left) and Brunner-Munzel test (right). e A model of type I and type II condensates. Assuming stronger multivalent interactions between JAZ2 molecules and weaker multivalent interactions between FCA-PrLD molecules in addition to the strong interaction between GBP and YFP, the different compositions of eCFP-JAZ2-GBP and eYFP-FCAPrLD could result in differences in the molecular dynamics of the condensates. In type I condensates, abundant, strong interactions between eCFP-JAZ2-GBP molecules results in low mobility of both eCFP-JAZ2-GBP and eYFP-FCAPrLD. In type II condensates, abundant eYFP-FCA-PrLD molecules increases interactions between eYFP-FCA-PrLD molecules that are not bound to eCFP-JAZ2-GBP, leading to a mobile fraction of eYFP-FCA-PrLD but not eCFP-JAZ2-GBP molecules.

We hypothesized that the properties of the resulting condensates, which exhibit differing patterns of molecular dynamics, reflect the strength of the interactions between the assembled molecules. The interactions between the JAZ2 molecules that form solid-type condensates were expected to be stronger than those between the FCA-PrLD molecules that form liquid-type condensates, as illustrated in (Fig. 7e). Under conditions where JAZ2 is more abundant than FCA-PrLD in the same condensate, the interactions between JAZ2 and FCA-PrLD might cause primarily immobile interactions (Type I condensate) (Fig. 7e). Conversely, when FCA-PrLD was more abundant than JAZ2, some FCA-PrLD molecules could form mobile interactions (Type II condensate) (Fig. 7e). We hypothesized that a similar situation would occur when eCFP-JAZ2N85-FCAPrLD and eYFP-FCAPrLD were co-expressed. Consistent with this finding, we observed co-condensates with lower mobility for eCFP-JAZ2N85-FCAPrLD and with higher mobility for eYFP-FCA-PrLD (Supplementary Fig. 17a, b). Similar results were observed when the amount of eCFP-JAZN85-FCA-PrLD was significantly higher than that of eYFP-FCA-PrLD (Supplementary Fig. 17a, b). This was likely due to the increased proportion of FCA-PrLD in the co-condensates, facilitating the mobility of the eYFP-FCA-PrLD molecules (Supplementary Fig. 17c).

Utilization of a hybrid condensate with two different material properties

We reasoned that the formation of hybrid condensates would enable us to examine whether the TEV protease fused to FCA-PrLD can cleave substrates in the solid phase of the condensates. Therefore, we designed mRFP1 (monomeric red fluorescent protein1)-[TEV cleavage site]-eCFP fused to OsJAZ2-GBP as a substrate (Fig. 8a). Co-expression of this substrate with eYFP-FCA-PrLD resulted in co-condensation with RFP, CFP, and YFP foci (Fig. 8b, c). Moreover, with an increased ratio of the FCA-PrLD construct, we observed hybrid condensates with higher eYFP mobility and lower mRFP1 and eCFP mobility (Supplementary Fig. 18). We assumed that the eYFP-FCA-PrLD could function as an assembler to recruit the FCA-PrLD-fused TEV proteaseΔ220-242 to the hybrid condensates (Fig. 8a). Expression of these three constructs resulted in co-condensation with CFP and YFP foci, whereas RFP fluorescence was diffused, suggesting that mRFP1 was cleaved from the substrate (Fig. 8b, c and Supplementary Fig. 19a). This mRFP1 diffusion was more prominent by expression of FCA-PrLD-fused TEV proteaseΔ220-242 than non-fused TEV proteaseΔ220-242, indicating that the cleavage was enhanced by the fusion of the enzyme. In addition, the mRFP1 release by FCA-PrLD-TEV proteaseΔ220-242 tended to be generally higher in cells where the substrate was more enriched in the condensates, although the effect varied between the experiments (Fig. 8d and Supplementary Fig. 19b).

Fig. 8 TEV protease reaction is promoted in the hybrid-type condensates.a Experimental design to monitor TEV protease activity using hybrid-type condensates. mRFP1-eCFP-JAZ2-GBP (substrate), containing a TEV protease cleavage site (tev site) between mRFP1 and eCFP-fused JAZ2-GBP, is co-expressed in Oryza sativa protoplasts with eYFP-fused FCA-PrLD (assembler), and with TEV proteaseΔ220-242 either without or with fusion to FCA-PrLD. This leads to the formation of a hybrid condensate containing the mRFP1-eCFP labeled substrate with lower mobility and the eYFP-labeled assembler with higher mobility. FCA-PrLD-fused TEV proteaseΔ220-242 is preferentially recruited to the condensate. After digestion of the substrate, mRFP1 separated from the condensate is observed. b Fluorescence patterns of the cells with or without TEV protease activity after the introduction of the indicated amounts of plasmids. Upper, a representative fluorescence pattern of a cell expressing the substrate and the assembler, classified as [class1]. RFP, CFP, and YFP fluorescence are observed almost exclusively at the condensates. Middle, a representative fluorescence pattern of a cell expressing TEV proteaseΔ220-242 with the substrate and the assembler, classified as [class2]. While CFP and YFP fluorescence is observed exclusively at the FCA-PrLD condensates, RFP fluorescence is observed in part at the condensates, however, more widely in the cytosol and the nucleus. Lower, a representative fluorescence pattern of a cell expressing FCA-PrLD-fused TEV proteaseΔ220-242 with the substrate and the assembler, classified as [class3]. While CFP and YFP fluorescence is observed exclusively at the FCA-PrLD condensates, RFP fluorescence is not, however, is instead widely detected in the cytosol and the nucleus. Condensates are indicated by white arrowheads. Scale bar, 5 μm. c The rate of cells exhibiting fluorescence patterns is classified as in (b) (mean ± SD, n = 3 pooled experiments). A total of 10 to 15 cells per protease construct were examined in each experiment. The number of cells observed in the classified pattern is shown on the right. Differences in categorical variables were assessed with Fisher’s exact test. d Relationship between the proportion of condensate CFP fluorescence per whole cell CFP fluorescence (PCFP foci) and the degree of reaction progress. The difference between PCFP foci and PRFP foci (the proportion of condensate RFP fluorescence per whole cell RFP fluorescence) was calculated as an indicator of the reaction progress (See Methods for details). e Relationship between the amount of substrate expressed and the degree of reaction progress. The substrate amount is indicated with whole-cell CFP fluorescence intensity (arbitrary units, a.u.) in each cell. For (d) and (e), data from Experiment 1 (n = 10 cells for each) are shown with a correlation coefficient (r). Data from the other experiments are shown in Supplementary Fig. 19b, c.

We investigated the relationship between the degree of reaction progress and the expression levels of the substrate (Fig. 8e and Supplementary Fig. 19c), based on the aforementioned hypothesis that these should be well correlated if most of the enzymatic reactions occur outside the condensate. With the hybrid condensates, the expression of non-fused TEV proteaseΔ220-242 exhibited lower correlation coefficient values (Fig. 8e and Supplementary Fig. 19c) when compared with the experiments using the liquid-like condensates (Fig. 5d and Supplementary Fig. 11d). This was conspicuous, showing a negative correlation in Experiment 1, where the expression levels of the substrate construct containing eCFP were lower than those in other experiments (Fig. 8e). These results likely reflect that the lower accessibility of the enzyme outside the condensate to the substrate, which is prone to forming a solid phase in the hybrid condensates, causes decreased reactivity. By the expression of the FCA-PrLD-fused TEV proteaseΔ220-242, the correlation coefficient values also showed the tendency to be lower in the experiments with the hybrid condensates (Fig. 8e and Supplementary Fig. 19c) than with the liquid-like condensates (Fig. 5d and Supplementary Fig. 11d). Together, these results suggest that the hybrid condensate system suppresses reactions outside the condensates.

Construction of AMOs in transgenic Arabidopsis plants

We constructed AMOs with modified properties in transgenic Arabidopsis plants by expressing the genes encoding eYFP-fused FCA-PrLD and JAZ2N85-FCA-PrLD under the control of the CaMV 35S promoter. We observed that both proteins formed spherical condensates in leaf cells, such as epidermal pavement (Supplementary Fig. 20a) and stomatal guard cells (Fig. 9a and Supplementary Figs. 20b, 21a). They are most frequently observed in the cytosol, near or away from the nucleus, but also in the nucleus. For subsequent FRAP and quantitative analyses, we focused on stomatal guard cells due to the high frequency of observation of condensates and their uniform cell size and developmental stage. A comparison of the ratio of condensate fluorescence density to nucleoplasm fluorescence density (outside the condensates, if any) showed that JAZ2N85-FCA-PrLD had significantly higher scores than FCA-PrLD (Fig. 9b, c), suggesting that JAZ2N85-FCA-PrLD was prone to form a denser condensate. Moreover, FRAP analyses showed that FCA-PrLD and JAZN85-FCA-PrLD formed liquid- and solid-like condensates, respectively (Fig. 9d and Supplementary Figs. 20c, d, 21), as observed in the transient assay (Fig. 3). Thus, the addition of the JAZN85 sequence modified the molecular dynamics of FCA-PrLD condensates in an analogous manner in both transient expression and stable transformation systems.

Fig. 9 Condensates formed by FCA-PrLD and JAZ2N85-FCAPrLD expression in transgenic Arabidopsis plants.a Representative YFP fluorescence images of the stomatal guard cells of transgenic Arabidopsis lines expressing eYFP-fused FCA-PrLD (line 27) and JAZ2N85-FCAPrLD (line 2). Differential interference contrast (DIC) image, chlorophyll auto-fluorescence, and their merged images with YFP fluorescence are shown. Condensates and the nucleus are indicated by arrowheads (white and orange, respectively). Scale bar, 5 μm. Cells were observed 57–58 d after sowing. b, c Comparison of YFP fluorescence density (AU μm^−2^) of condensate and nucleoplasm. Intensity of YFP fluorescence in the guard cells (n = 27 for eYFP-FCA-PrLD, 22 for eYFP JAZ2N85-FCAPrLD) observed as in (a) was quantified. For each cell, a condensate that gave the highest average fluorescence signals was selected for quantification. Ratio of the density (condensate/nucleoplasm) is shown in (c). P value was determined by Brunner-Munzel test. d FRAP analyses of FCA-PrLD and JAZ2N85-FCAPrLD condensates formed in the guard cells (mean ± SD). n = 11 and 6 condensates from the respective transgenic lines. Time 0, the time point of the bleaching pulse. AU arbitrary units.

We also examined the expression of eYFP-fused OsJAZ2 proteins with and without RSS1-IDR insertion in transgenic Arabidopsis plants. We observed the formation of small condensates predominantly in the nucleus, almost exclusively in stomatal guard cells expressing either OsJAZ2 or OsJAZ2-RSS1 (Fig. 10a and Supplementary Fig. 22a, d). The number of nuclear condensates was significantly higher in cells expressing OsJAZ2 than in cells expressing OsJAZ2-RSS1, while the YFP fluorescence intensity in the nucleus was comparable between these cells (Fig. 10b and Supplementary Fig. 22e). Moreover, this was associated with a significantly higher distribution and moderate enrichment of OsJAZ2-RSS1 in the condensates (Fig. 10c, d and Supplementary Fig. 22f, g), suggesting that JAZ2-RSS1 has a higher propensity for condensation than JAZ2. Furthermore, FRAP analysis showed that both OsJAZ2 and OsJAZ2-RSS1 formed solid-like condensates (Supplementary Fig. 22a–c). Thus, the property of the OsJAZ2-RSS1 condensates formed in the transgenic plants was comparable to that of the condensates observed in rice protoplasts (Fig. 4, and Supplementary Fig. 9), except that condensation in the cytosol and the formation of enlarged condensates of OsJAZ2 were not obvious in the leaf cells. Collectively, these results suggest that the modification of AMOs in the transient assay can be applied, in part, for the design and construction of AMOs in transgenic plants.

Fig. 10 Condensates formed by JAZ2 and JAZ2-RSS1 expression in transgenic Arabidopsis plants.a Representative images of YFP fluorescence in the stomatal guard cells in the transgenic lines expressing eYFP alone (line 9), eYFP-JAZ2 (line 1), and eYFP-JAZ2-RSS1 (line 4). Condensates and the nucleus are indicated by arrowheads (white and orange, respectively). Differential interference contrast (DIC) image, chlorophyll autofluorescence, and their merged images with YFP fluorescence are shown. The dotted line shows the outline of a stomata. Cells were observed 59 (eYFP), 21 (eYFP-JAZ2), and 64 d (eYFP-JAZ2-RSS1) after sowing. Scale bar, 5 μm. b Number of condensates in the nucleus of the guard cells. c The ratio of YFP fluorescence intensity in condensates to those in the nucleoplasm. d Enrichment of YFP signals in the condensates in the nucleus. In (b–d), condensate number, the ratio of YFP fluorescence intensity and fold enrichment relative to YFP fluorescence intensity of the nucleus are also shown. In (b) (right), YFP fluorescence intensity of the nucleus is shown as a bar graph. Data are from two observations (mean ± SD, n = cells). The significance differences were evaluated using Brunner-Munzel test (b-centre, d) or Welch’s t-test (b-right, c). Fluorescence intensity is shown in arbitrary units (a.u.).

Discussion

Recent studies have revealed that molecular condensates function not only on a molecular scale involving the regulation of biochemical reactions but also on a larger scale involving the directional organisation of intracellular structures with specific locations relevant to their activity and regulation^28^. From a synthetic biology perspective, we investigated several methods to control the properties of AMOs in model studies.

We demonstrated the creation of membrane-anchored condensates. In the case of autophagosome formation, the early pre-autophagosomal structure (PAS) is tethered to the vacuolar membrane through interactions between Atg13 (a component of Atg1 complexes that undergo phase separation) and Vac8 (a vacuolar membrane protein)^41^. In the TCR cluster, the transmembrane protein linker for the activation of T cells (LAT) and its binding partners coalesce at the plasma membrane upon TCR activation^37^. Conversely, in the present study, the membrane protein SCAMP1 was directly fused to condensation-prone proteins. A simple strategy using chimeric proteins may facilitate the use of AMO for functional studies of membrane-associated condensates and their future applications, for example, creating synthetic membrane domains that assemble functional membrane proteins, such as TCR signaling clusters^37^ or postsynaptic density^36^.

The localization of JAZ2 was often altered by protein condensation (Fig. 4) or fusion with GBP at the C-terminal site (Supplementary Fig. 13a). When designing AMOs, such effects should be considered depending on the position of the localization signal and the nature of the fused proteins. Contrarily, the insertion of RSS1-IDR suppressed extranuclear JAZ2 condensation in our transient assays, suggesting that RSS1-IDR prevents the appearance of condensation-prone structures that frequently form before nuclear transport. As RSS1-IDR is highly hydrophilic and has both acidic and basic regions, it may affect the electrostatic and/or π-cation interactions involved in JAZ2 condensation^54^. This chaperone-like function of RSS1 is reminiscent of that of the polyD/E protein DAXX, which de-condenses molecules independent of ATP^75^. Determining whether the function of RSS1 in counteracting condensation is related to its role under stressful conditions, as reported, will be interesting^60,76^.

To design AMOs, selecting more appropriate material properties, such as liquid and solid phases, according to their assigned functions, is preferable. The liquid phase is preferred for specific or orthogonal reactions, whereas the solid phase protects specific molecules from degradation. Several lines of evidence support the idea that the molecular dynamics of condensates formed by multivalent intermolecular interactions depends on the mode and strength of the interactions themselves. For example, while aromatic and positively charged amino acids that cause cation-π interactions, particularly Tyr and Arg, govern the saturation concentration of the phase separation of FUS protein, Gly maintains liquidity and Glu and Ser contribute to hardening^25^. Studies using the PrLD of heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1), a component of stress granules, and Velo1, a mediator of Balbiani body assembly, suggest that clusters of aromatic or aliphatic residues promote solid-like condensate formation, whereas distributed aromatic residues with hydrophilic spacers promote liquid-like condensate formation^77,78^. Moreover, the recruitment of FUS LCD to oskar granules alters the state of the condensates from solid to liquid^24^. In contrast, duplication of the RNA-binding domain of G3BP1 reduced the dynamicity of stress granules^31^. In this study, the properties of FCA-PrLD to form liquid condensates were altered to form solid condensates by JAZ2N85. The solidification required the EAR and TIFY regions that are predicted to be amyloidogenic regions and form β-strands (Supplementary Fig. 6), consistent with that molecular mobility of JAZ2 condensates is increased by deleting the N-terminal and TIFY domains^54^. In addition, LCR1 and LCR2 exhibit opposite effects on solidification. Thus, the properties of the target condensates can not only be altered by direct fusion but also be further tuned by modifying the fused protein at the amino acid or domain level.

The size and number of condensates are important for proper biological function. The condensate size is restricted by the interfacial tension of the condensate, which minimises the total condensate surface area^33^. The reduction in the total interfacial area was caused by the coalescence of condensates and Ostwald ripening^33,79^. In our study, by RSS1-IDR insertion into JAZ2, the condensate size decreased, and the number of condensates increased. Given that the JAZ2 condensates were mostly solid-like^54^ and the RSS1-IDR insertion further solidified the condensates, coalescence was unlikely to occur frequently. Thus, RSS1-IDR insertion may suppress Ostwald ripening. Notably, amphiphilic surfactant-like proteins are known to influence condensate size^80^. Co-expression of the RGG domain of LAF-1 with an amphiphilic protein that fuses the RGG domain with maltose-binding protein (MBP) results in the MBP region being located on the outside, reducing the size of the condensate formed in vitro^80^. The amphiphilic protein localized on the surface of condensates is interpreted to function as a surfactant and alter the interfacial properties^80^. Similarly, RSS1-IDR, which is rich in hydrophilic amino acids, might cause a thermodynamically stable structure when located near the outside of the condensate, whereas the JAZ2 region, which contributes to multivalent interactions, could be typically oriented inward, possibly resulting in changes in interfacial properties and smaller-sized condensates.

In nature, there are multiphase condensates with internal subcompartments, as exemplified by the nucleoli and stress granules with multilayered structures^81,82^. Notably, both the nucleolus and stress granules contain proteins with different mobile fractions^81,83^. Importantly, we demonstrated that hybrid AMOs with two different molecular dynamics could be created by the nanobody-based association of two types of condensation-prone proteins, as well as using chimeric scaffold proteins. The resulting heterogeneous properties relied on the interaction strength and molecular ratio of the constituent proteins, as illustrated by the simplified model (Fig. 7e and Supplementary Fig. 17c). This basic concept can also be applied to other, more complex systems, although our hybrid condensates exhibited a single miscible phase. Further control of the protein interaction networks and possibly the surface organisation would contribute to the construction of multiphase AMOs^33,45,84–86^.

We showed that OsJAZ2 forms condensate in transgenic Arabidopsis plants. This implies that JAZ2 can be used as a model condensation-prone protein in plants. Moreover, modification of condensate properties by the fusion of JAZ2N85 with FCA-PrLD and RSS1-IDR insertion into JAZ2, observed in rice protoplasts, was reproduced in transgenic plants, although not completely. These results demonstrate the robustness of the modifications. In general, experiments using transformed plants allow stable gene expression and long-term analyses; however, they can be time-consuming. Conversely, transient expression systems allow observation at various expression levels and provide rapid results. Thus, transient systems to efficiently evaluate the effects of many protein variants are useful, especially in the initial stages of creating artificial condensates and changing their properties, as in this study.

Our results provide a general framework for constructing AMOs using various protein domains in plant cells. The heterogenous scaffold system can be combined with functional proteins, as demonstrated in this study (see Supplementary Note 1 for details). By combining modification methods with other systems, such as inducible phase separation systems^10,11,87^, more complicated AMOs with multiple properties can be constructed. However, the deletion or addition of protein domains may affect the function of the original protein, and, in particular, the biological function of JAZ2 and its condensates remains ambiguous. JAZ2 condensation may be involved in the storage or inactivation of JAZ factors^54^, causing changes in jasmonate sensitivity. In Arabidopsis, AtJAZ1 physically interacts with AtMYC2 and both proteins form co-condensates when transiently co-expressed in Nicotiana tobacum leaf cells^88^ raising the possibility that the AtJAZ1 condensate could affect AtMYC2-dependent gene expression. However, OsJAZ2 did not form co-condensates with OsbHLH94 (eg. in Fig. 4b, #2) but OsJAZ9^54^, which is a repressor of jasmonate signaling^57^. Further modification of the protein domain or replacement of the interaction domains would be helpful in minimising the potential side effects.

The design and practical use of AMOs in plant cells, for example, to produce medically effective substances (including nanobodies) or biomaterials, is highly anticipated in the future because plant cells have the advantage of avoiding the risk of infection by animal viruses. Another attractive goal is to create various plants that will benefit agricultural sustainability threatened by global climate change^89^. Under more practical conditions and systems, the application of AMOs can provide an avenue for adding another value to plants.

Methods

Plant materials

Seeds of rice (O. sativa L. cv. “Nipponbare”) were surface sterilized and germinated on 1/2 ×Murashige and Skoog basal medium (pH 5.7) containing 0.35% gellan gum in the plant box set (height 20 cm)^54^. N. benthamiana seeds were sown in plant pots. O. sativa and N. benthamiana plants were grown at 27 °C (day) and 25 °C (night) under 16 h-light (4500 lux) and 8 h-dark cycles. O. sativa protoplasts were isolated from the leaves at 6–7 d after sowing. Agrobacterium infiltration was performed using leaves of 3–5-week-old N. benthamiana plants.

Seeds of Arabidopsis (Arabidopsis thaliana) Columbia-0 (col-0) were surface-sterilised and germinated on MS medium (pH 5.7) containing 20% sucrose and 0.3% gellan gum. Seedlings were grown at 22 °C in a growth chamber under 16-h light (3000 lx) /8-h dark cycles for 10 to 14 d on the same plates and then grown on rock fibre (Grodan). Transformation of Arabidopsis was carried out by the Agrobacterium-mediated floral-dip method^90^, using Agrobacterium tumefaciens (GV3101) carrying pCAMBIA1301_based plasmids. Transgenic Arabidopsis plants (T1) were observed under a microscope, except for the observation of cells of T2 plants expressing eYFP-FCA-PrLD, as shown in Fig. 9 and Supplementary Fig. 20. Since eYFP-JAZ2 overexpression lines were sterile and eYFP-JAZ2-RSS1 overexpression plants showed low fertility, we could not perform the experiment using T2 plants for these lines.

Transient gene expression using O. sativa leaf protoplast and N. benthamiana leaves

Protoplast isolation and PEG-mediated plasmid introduction were performed basically as described^54^. Protoplasts were prepared from leaves of O. sativa seedlings by digestion of the cell walls. Sliced leaf tissues pieces (0.5-1 mm wide) were pre-incubated in 0.6 M mannitol in dark for 10 min, and then incubated with 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 CaCl2) in dark for 5 h with gentle agitation (75 rpm) at 25 °C. After washing out the enzyme solution, protoplasts were suspended in 0.5–2 mL of a solution (0.4 M mannitol, 15 mM MgCl2, and 4 mM MES-KOH [pH 5.7]) to a cell density of 2 × 10^6^–2 × 10^7^ cells/mL. Plasmid DNA solution (10–20 μL) (5 μg DNA for each construct, unless otherwise mentioned) was mixed gently with 100 μL of protoplast solution and equal volume (to DNA and protoplast solution) PEG solution, containing 40% PEG4000 (SIGMA 81420, Merck KGaA, Darmstadt, Germany], 0.1 M CaCl2, and 0.2 M mannitol. After incubation for 10–20 min, protoplasts were mixed gently with four times the PEG solution volume of W5 solution (154 mM NaCl, 5 mM KCl, 125 mM CaCl2, and 2 mM MES-KOH [pH 5.7]), precipitated by centrifugation at 500 × g for 2 min, and suspended in 300 μL of WI solution (0.5 M mannitol, 20 mM KCl, and 4 mM MES-KOH [pH 5.7]). After incubation at 22 °C in the dark for 16–22 h, protoplasts were used for observation^34^. For treatment with rapamycin (R0161, WAKO, Japan), 10 mM of rapamycin stock solution dissolved in DMSO was diluted in water to 40 μM, added to the protoplast solution to final concentrations of 2 μM, and incubated at ~25 °C for 20 min. For the time-lapse experiments, rapamycin was added to the protoplast solution to a final concentration of 1 μM.

Transformation of Agrobacterium tumefaciens (strain GV3101) with pSoup- and pGeenII-derived plasmids by electroporation and Agrobacterium infiltration of N. benthamiana leaves were performed basically as described^54^. N. benthamiana leaves were infiltrated with the transformed Agrobacterium cell cultures and incubated in a plant chamber for 2–3 days.

Microscopic observation and FRAP analysis

Microscopic observation and FRAP analysis were performed as previously described^54^. The fluorescence of CFP, YFP, and RFP was observed under a confocal laser scanning microscope (CLSM; FV 1000, Olympus, Tokyo, Japan) at 440/460–500, 515/530–545, and 559/575–675 nm (excitation/emission wavelengths), respectively. FV10-ASW Viewer software (Ver.4.2b, Olympus, Tokyo, Japan) was used to acquire and analyze images. O. sativa protoplasts were selected with expression of mRFP1, using a plasmid carrying the CaMV35S promoter-mRFP1 gene, pDH51_mRFP1. For visualization of the nucleus in O. sativa protoplasts, pUGW42- and pUGW45-based plasmids carrying the CaMV35S promoter-driven eYFP-OsbHLH094 gene and OsbHLH094-eCFP gene^57^ (Os07g0193800) were used. For the experiments using nanobodies, pUGW42 was used to express eYFP. For normal observation, a 20× objective lens was used, with a confocal aperture (C.A.) set to 80 μm. For acquiring an image of the entire rice cell, the C.A. was set to 800 μm to expand the depth of focus. For larger cells, images were acquired in two separate sessions. To observe transgenic Arabidopsis cells, leaf segments (1 × 2 cm^2^) were mounted on glass slides with coverslips and sterile water. We used a 60× objective lens with a C.A. set to 300 μm. Cells that formed YFP condensates were randomly selected to count the number of condensates.

For FRAP analysis using a 60× objective lens, the region of interest (ROI) (circular 0.8–1 μm radius for O. sativa or Arabidopsis and 1.6 μm for N. benthamiana) were bleached for 0.5 s, using a tornado bleach pulse, and fluorescence recovery was monitored continuously without interval setting. After fluorescence measurements by FV1000 software or FIJI/ImageJ, the normalized fluorescence intensity (Inor) was calculated by the following

where It and BGt are the fluorescence intensity and background at time point t, respectively, and Ipre and BGave are the average intensity at ROI and background of the 50 frames before bleaching, respectively^54^. The unbleached regions were selected for the background setting. In some cases, condensates move in protoplasts for protoplasmic flow or other unknown reasons. To obtain the FRAP results condensates that shifted from the Z-axis (confocal plane) or deviated significantly from the z-axis during monitoring were excluded from the analysis. When the analysed condensate drifted slightly from the bleaching area but was still on the focal plane (XY plane), ROI was corrected in Fiji/ImageJ according to the position of the condensate. This is illustrated in Supplementary Fig. 3b (i), (iii).

Plasmid construction

Construction of plasmids used in this study to express the genes of interest under the CaMV 35S promoter is described below. After verifying the inserted DNA and junction regions of the respective constructs by DNA sequencing, the plasmids were purified using a QIAGEN Midi prep kit (QIAGEN K.K., Tokyo, Japan). Primer sequences and template DNA used in this study are listed in Supplementary Data 1. All plasmids used in this study are listed in Supplementary Table 1.

Plasmids used for the transient assay of O. sativa protoplasts were constructed as below. The pUGW45-based plasmids (pUGW45_eCFP-JAZ2, pUGW45_eCFP-JAZ2-del-jas-2, and pUGW45_eCFP-JAZ2-del-LCR5jas) and plasmids carrying eYFP-bHLH94 and mRFP1, respectively, were used for CaMV 35S promoter-mediated expression as described previously^54^. The NLS-FLAG sequence was amplified via PCR using specific primers. The NLS-FLAG sequence was sub-cloned into pENTR/D-TOPO (Invitrogen, Thermo Fisher Scientific K.K., Tokyo, Japan) and inserted by LR reaction into the pUGW42 vector^91^, resulting in pUGW42_eYFP-NLS-FLAG. pUGW42_eYFP-JAZ2-del-jas-2-NLS-FLAG was constructed by insertion of the JAZ2Δ-ljas-2 sequence into pUGW42_eYFP-NLS-FLAG using the SLiCE method^92^ and specific primers.

The cDNA fragments of OsNup98 (XM_015762965.2) and OsSCAMP1 (XM_015789328) were amplified by PCR from an O. sativa basal shoot region-derived cDNA mixture using specific primers and PrimeSTAR GXL DNA polymerase (Takara Bio, Shiga, Japan). After subcloning into pENTR/D-TOPO, OsNup98 cDNA was inserted by LR reaction into pUGW42. To express eYFP-Nup98-FG1, site-directed mutagenesis was performed by PCR amplification of pUGW42_Nup98 with the mutated primers and KOD FX-Neo DNA polymerase (TOYOBO, Osaka, Japan), followed by DpnI digestion and transformation into E. coli DH5-α. The XbaI fragment of the amplified cDNA encoding OsSCAMP1ΔN118 was inserted into the XbaI sites of pUGW45, pUGW42, pUGW45_eCFP-JAZ2, or pUGW42_eYFP-Nup98-FG1, resulting in pUGW45_-SCAMP1ΔN118-eCFP, pUGW42-SCAMP1ΔN118-eYFP, pUGW45_SCAMP1ΔN118-eCFP-JAZ2, and pUGW42_SCAMP1ΔN118-eYFP-Nup98-FG1. The pUGW42_eYFP-FCA-PrLD and pUGW45-based plasmids to express eCFP-FCA-PrLD, eCFP-JAZ2N85-FCA-PrLD, and eCFP-JAZ2-RSS1 were constructed using the SLiCE method after PCR amplification of the respective template DNA using the specific primers. To prepare plasmids for expressing eCFP-JAZ2-FCA-PrLD, the stop codon at the C-terminus of FCA-PrLD in pUGW45_eCFP-JAZ2N85-FCA-PrLD was removed by site-directed mutagenesis using the mutated primers. The pUGW45-based plasmid to express eCFP-JAZ2N85-FCA-PrLD variants carrying various deletions was prepared by site-directed mutagenesis using the designed primers.

A plasmid expressing eYFP-[TEV protease cleavage site (tev site)]-eCFP-FCA-PrLD was prepared using the SLiCE method with specific primers and template DNA. To obtain the inserted DNA encoding the eYFP-[tev site] region, the eYFP and [tev site] regions were amplified from pUGW42_YFP-JAZ2 and pETL8-GFP-JAZ2^54^ using PCR, and the PCR products were used as templates for the second PCR amplification. The plasmids to express TEV proteaseΔ220-242 and FCA-PrLD-TEV proteaseΔ220-242 under the control of the CaMV 35S promoter (pUGW_TEV proteaseΔ220-242 and pUGW_FCAPrLD-TEV proteaseΔ220-242) were prepared by SLiCE. A TEV protease gene (GenBank: DQ516974.1) with an optimized codon usage for expression in O. sativa was synthesized and subcloned by Eurofin Genomics Inc. (Tokyo, Japan). From the resulting plasmid, the TEV protease sequence was amplified and inserted into pUGW45_eCFP-FCA-PrLD, resulting in pUGW45_eCFP-FCAPrLD-TEV protease. From this construct, the eCFP region was deleted, resulting in pUGW_FCAPrLD-TEV protease. To prepare the plasmid pUGW_TEV protease carrying CaMV 35S-promoter:TEV protease, the eCFP-JAZ2 region of pUGW45_eCFP-JAZ2 was replaced with TEV protease. Finally, pUGW_TEV proteaseΔ220-242 and pUGW_FCAPrLD-TEV proteaseΔ220-242 were prepared from pUGW_TEV protease and pUGW_FCAPrLD-TEV protease, respectively, by site-directed mutagenesis.

For the experiment with nanobody-mediated recruitment, the GBP gene with an optimized codon usage for expression in O. sativa was synthesized by gBlock® (IDT®). The pUGW45-based plasmids to express GBP-eCFP-JAZ2, eCFP-JAZ2-GBP, and eCFP-FCA-PrLD-GBP were prepared using the SLiCE method.

Plasmids containing FKBP and FRB were generously provided by Dr. Kazuhiro Aoki of the National Institute for Basic Biology. To construct pUGW42-eYFP-FRB, the FRB gene amplified by PCR was inserted into pENTR/D-TOPO and subsequently into pUGW42 by the LR reaction of the Gateway system. pUGW45_FKBP-eCFP-JAZ2 was prepared by inserting the FKBP sequence into pUGW45_eCFP-JAZ2 using the SLiCE method.

A plasmid for the expression of mRFP1- [TEV protease cleavage site (tev site)]-eCFP-JAZ2-GBP was prepared using the SLiCE method. To obtain the inserted DNA encoding the mRFP1-[tev site] region, the mRFP1 and [tev site] regions were amplified from pDH51-mRFP1 and pETL8-GFP-JAZ2, and the PCR products were used as templates for the second PCR amplification. For the transient assay in N. benthamiana leaves, GFP in pGreenII-2x35s-TL-GFP^54^ was substituted with eCFP and eYFP-Nup98-FG1 using the SLiCE method after amplification of the DNA fragments.

To prepare the constructs used for Arabidopsis transformation, the YFP-JAZ2 region of pUGW42_eYFP-JAZ2 was substituted with eCFP-JAZ2N85-FCAPrLD and eCFP-JAZ2-RSS1 using SLiCE after DNA fragment amplification. For the eYFP control construct, the ccdB region was removed from pUGW42 using PCR, resulting in pUGW42_eYFP-del-ccdB. pUGW42-based plasmids containing eYFP, eYFP-JAZ2, eYFP-FCAPrLD, eYFP-JAZ2N85-FCA-PrLD, and eYFP-JAZ2-RSS1 were digested with AfiII and SbfI, and the DNA fragments were inserted between the AfiII and SbfI site of pCAMBIA1301vector. The resulted pCAMBIA1301_based plasmids carrying CaMV 35S-promoter:eYFP, 35S-promoter:eYFP-FCA-PrLD, 35S-promoter:eYFP-JAZ2N85-FCA-PrLD, 35S-promoter:eYFP-JAZ2, or 35S-promoter:eYFP-JAZ2-RSS1 were used for Agrobacterium transformation by electroporation.

Quantitative analysis of images

All images were analysed using FIJI/ImageJ. To quantify the fluorescence intensities of the condensate, nucleus, and whole cell, the sum of pixels × fluorescence intensity in the ROI was measured and the background was subtracted.

To compare the ratio of CFP fluorescence inside and outside the condensates formed by eCFP-JAZ2 and eCFP-JAZ2-RSS1 in O. sativa protoplasts, the total fluorescence intensity of the nuclear condensates was divided by [whole-cell fluorescence minus condensate fluorescence]. To calculate fold enrichment, the average fluorescence density of the CFP condensates (total condensate CFP intensity/total condensate area) was divided by the average fluorescence density of CFP in a cell (whole-cell intensity/cell area). Cells with multiple nuclei were excluded from quantitative analysis.

To evaluate the classified fluorescence patterns of the cells in the TEV protease assay, the fold enrichment of YFP or RFP in foci (degree of YFP or RFP fluorescence enrichment in condensates) was calculated as the average fluorescence density of YFP or RFP condensates (total condensate YFP or RFP intensity/total condensate area) was divided by the average fluorescence density of YFP or RFP in a cell (whole cell intensity/cell area).

To verify TEV protease activity in the cell, YFP released from the YFP–CFP co-condensate formed by eYFP-[tev site]-eCFP-FCA-PrLD was monitored from the images of the entire cell. The ROI was set at the area of the CFP condensate, and the fluorescence intensities of YFP and CFP in the condensates and whole cell were quantified for each protoplast. The proportion of the total intensity of YFP and CFP fluorescence in the condensates in a cell compared to the whole-cell intensity of YFP and CFP was designated as the PYFP foci and PCFP foci, respectively. The difference between PYFP foci and PCFP foci [PCFP foci minus PYFP foci] was used as an indicator of TEV protease activity. Similarly, RFP released from RFP–CFP–YFP co-condensates formed from mRFP1-[tev site]-eCFP-JAZ2-GBP and YFP-FCA-PrLD was monitored. The ROI was set in the area of the YFP condensate because the YFP condensates were most clearly observed and well co-localised with the CFP–RFP condensates. After quantification of RFP and CFP fluorescence in the condensates and whole cell, PRFP foci and PCFP foci were calculated and the difference between PRFP foci and PCFP foci [PCFP foci minus PRFP foci] was used as an indicator of TEV protease activity in the cell. In cells with no obvious YFP (or RFP) condensates owing to the release of YFP (or RFP) by the TEV protease reaction, YFP (or RFP) fluorescence in the region of the CFP (or YFP) condensates was quantified. In this case, the quantified fluorescence might contain fluorescence in the cytosol along the Z-axis where condensates were observed because the whole-cell image was used for quantification. Additionally, the detection of mRFP1 and chlorophyll was not completely separated under the wavelength setting.

For the quantification of condensate fluorescence intensity in FRAP analyses of transgenic Arabidopsis cells, as shown in Fig. 9 and Supplementary Figs. 20–22, the average fluorescence intensity from 50 frames of time-lapse images before bleaching was used. For other quantitative analyses of stomatal guard cells in transgenic Arabidopsis, YFP fluorescence was quantified from maximum intensity projection (MIP) images. Cytosolic fluorescence in stomatal guard cells was excluded from the quantification because high levels of autofluorescence were frequently detected. To compare the YFP fluorescence density of the condensate and nucleoplasm in guard cells expressing YFP-fused FCA-PrLD and JAZ2N85-FCA-PrLD, the cytosolic condensate that showed the highest average fluorescence signals was selected for quantification. To calculate the density ratio (condensate/nucleoplasm) plotted in Fig. 9, the fluorescence density of the YFP condensate in the cytosol (YFP signals of the condensate/area of the condensate) was divided by the fluorescence density of the nucleoplasm in the cell ([fluorescence of the nucleus minus total fluorescence of the nuclear condensates]/[area of the nucleus minus the total area of the nuclear condensates]). To calculate the ratio of the YFP fluorescence intensity of eYFP-JAZ2 and eYFP-JAZ2-RSS1 in the condensates to that in the nucleus, the total fluorescence of the condensates in the nucleus was divided by the total fluorescence in the nucleus. To determine the fold enrichment of YFP signals in the condensates in the nucleus, the average fluorescence density of the condensates (total condensate fluorescence/area of condensates) was divided by the average fluorescence density of the nucleus (total fluorescence in the nucleus/area of the nucleus).

Statistics and reproducibility

For statistical analysis, categorical data, such as classified localization pattern and types of condensates based on the FRAP results, were analyzed using Fisher’s exact test. Data normality was assessed with the Shapiro-Wilk test. For comparisons between two groups, when the distribution is assumed to be normal, Welch’s t-test was applied. If the distribution is not assumed to be normal, the Brunner-Munzel test was applied. For the data in Experiment 2 in Supplementary Fig. 8c, the Mann-Whitney U test was used because the Brunner-Manzel test was not applicable. For the data in Fig. 6c, d, one-way ANOVA with Games-Howell post-hoc test was used as an alternative to a non-parametric test due to skewed data distribution, although some groups did not satisfy the normality assumption. All statistical analyses were performed using R version 4.4.1, except that Pearson’s r was calculated by Excel. The sample sizes and number of biological replicates are indicated in the figures or figure legends. In the data for FRAP analyses, n means the number of condensates. For fluorescence/condensation pattern analyses, n means the number of cells, except that in Figs. 4c, 5b, 6c, d, 8c and Supplementary Figs. 1a-b, 2c, 10a, n means the number of independent experiments. P values are indicated in Figure or Supplementary Table 2. We consider that p-value < 0.05 is significant. For the data in Fig. 5c, d, 8d, e, Supplementary Fig. 11c, d, Supplementary Fig. 19b, c, Pearson’s r (correlation coefficient) serves as an effect size index. All statistical tests were two-sided.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Acknowledgements

We thank Drs. M. Matsuoka for helpful suggestions with regards to transient assays using rice protoplasts, H. Shibata and T. Kojima for useful suggestions on FRAP and SLiCE techniques, respectively, and H. Yoshioka for suggestions and materials on Agrobacterium infiltration of N. benthamiana leaves. We thank T. Nakagawa and C. Dean for gifting us the gateway-based vectors pUGW42 and pUGW45 and plasmids carrying FCA-PrLD, respectively, K. Aoki for the plasmids used for the rapamycin-induced system, K. Maeo for the Agrobacterium strain GV3101 for Arabidopsis transformation. Finally, we thank T. Hattori, C. Ueguchi, and N. Ishii for the technical advice and helpful discussions. This work was partially supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI (JP16K08140 and JP22K05427) and the 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 the Graduate Program of Transformative Chem-Bio Research at Nagoya University, supported by The Ministry of Education, Culture, Sports, Science and Technology (MEXT) (WISE Program).

Author contributions

All authors contributed to the conception and design of the study. Material preparation, data collection, and analyses were performed by Y.K., Y.J., T.A., Y.Y., H.K., and K.T. Data interpretation was performed by S.T. in addition to the authors described above. The manuscript was written by Y.K. and S.T. and improved based on the other authors’ suggestions. All authors read and approved the manuscript.

Peer review

Peer review information

Communications Biology thanks Panagiotis Moschou, Peijin Li, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: David Favero. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information and Supplementary Data, or from the corresponding author upon reasonable request. Primer sequences and plasmid information are provided in Supplementary Data 1 and Supplementary Table 1, respectively. The source data behind the graphs in this paper can be found in Supplementary Data 2. Gene accessions/IDs are as follows; OsJAZ2/OsTIFY5 (Os7g05830/Os07g0153000): NM_001421968, eCFP: AB626686 (in pUGW45), AB626685 (in pUGW44), eYFP: AB626684 (in pUGW42), mRFP1: N9844 (Nottingham Arabidopsis Stock Centre; NASC) (in pDH51_mRFP1), OsSCAMP1: XM_015789328, OsNup98: XM_015762965.2, OsbHLH98 (Os07g0193800): XM_015790337.2, AtFCA (AT4G16280): NM_148347.2, OsRSS1(Os02g0606700): AB753859, TEV DQ516974.1, GBP: #49172 (addgene plasmid collection) (in pOPINE GFP nanobody), FKBP: M34539.1, FRB: OP056761.1., pCAMBIA1301: AF234297.

Competing interests

The authors declare no competing interests.

Footnotes

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-024-07102-8.

References

Associated Data

Supplementary Materials

Data Availability Statement

All data supporting the findings of this study are available within the paper and its Supplementary Information and Supplementary Data, or from the corresponding author upon reasonable request. Primer sequences and plasmid information are provided in Supplementary Data 1 and Supplementary Table 1, respectively. The source data behind the graphs in this paper can be found in Supplementary Data 2. Gene accessions/IDs are as follows; OsJAZ2/OsTIFY5 (Os7g05830/Os07g0153000): NM_001421968, eCFP: AB626686 (in pUGW45), AB626685 (in pUGW44), eYFP: AB626684 (in pUGW42), mRFP1: N9844 (Nottingham Arabidopsis Stock Centre; NASC) (in pDH51_mRFP1), OsSCAMP1: XM_015789328, OsNup98: XM_015762965.2, OsbHLH98 (Os07g0193800): XM_015790337.2, AtFCA (AT4G16280): NM_148347.2, OsRSS1(Os02g0606700): AB753859, TEV DQ516974.1, GBP: #49172 (addgene plasmid collection) (in pOPINE GFP nanobody), FKBP: M34539.1, FRB: OP056761.1., pCAMBIA1301: AF234297.