SEC14-like condensate phase transitions at plasma membranes regulate root growth in Arabidopsis

Protein function can be modulated by phase transitions in their material properties, which can range from liquid- to solid-like; yet, the mechanisms that drive these transitions and whether they are important for physiology are still unknown. In the model plant Arabidopsis, we show that developmental robustness is reinforced by phase transitions of the plasma membrane-bound lipid-binding protein SEC14-like. Using imaging, genetics, and in vitro reconstitution experiments, we show that SEC14-like undergoes liquid-like phase separation in the root stem cells. Outside the stem cell niche, SEC14-like associates with the caspase-like protease separase and conserved microtubule motors at unique polar plasma membrane interfaces. In these interfaces, SEC14-like undergoes processing by separase, which promotes its liquid-to-solid transition. This transition is important for root development, as lines expressing an uncleavable SEC14-like variant or mutants of separase and associated microtubule motors show similar developmental phenotypes. Furthermore, the processed and solidified but not the liquid form of SEC14-like interacts with and regulates the polarity of the auxin efflux carrier PINFORMED2. This work demonstrates that robust development can involve liquid-to-solid transitions mediated by proteolysis at unique plasma membrane interfaces.


Introduction
Under certain conditions, biomolecules can separate from their bulk phase through liquid-liquid phase separation (LLPS), thereby attaining liquid-like properties, such as surface tension, which leads to highly circular condensates akin to droplets [1].LLPS determines the formation of many evolutionary conserved condensates, such as nucleoli, stress granules, and processing bodies.Starting as liquids, some condensates undergo transitions in their material properties that affect their viscosity, surface tension, and degree of penetrance by other molecules.For example, in Drosophila melanogaster, oskar ribonucleoprotein (RNP) condensates undergo a liquid-to-solid transition, which is important for the polar distribution of some RNAs in the cell [2].Whereas oskar RNP liquidity allows RNA sequestration, its solid phase precludes the incorporation of RNA while still allowing protein sequestration.Furthermore, although they are not delimited by membranes, condensates can interface with them or even engulf small vesicles [3].
The past few years have experienced tremendous progress in the evolution of a molecular grammar that underpins LLPS.Molecules such as proteins and RNAs are polymers with attractive groups known as "stickers" that form noncovalent and mainly weak interactions.At certain concentrations, which are determined by various factors (e.g., temperature, redox state, pH), interactions are enabled among intra-or intermolecular stickers.When reaching a system-specific threshold concentration, the whole system containing various proteins and/or RNAs undergoes LLPS.The stickers promote the attraction between charged residues, dipoles, or aromatic groups that are usually provided by the so-called "intrinsically disordered regions" (IDRs) [4].Stickers are connected by "spacers" that regulate the density transitions (i.e., liquid-to-solid transitions) by orienting stickers.The IDRs lack a defined structure and thus can easily expose their stickers.Furthermore, IDRs can increase the apparent size known as hydrodynamic radius adopted by the solvated, tumbling protein molecule [5].
In the model plant, Arabidopsis (Arabidopsis thaliana) LLPS condensates are involved in, for example, the internal chloroplast cargo sorting, transcriptional circuits modulating defence, RNA processing, and temperature sensing [6][7][8][9][10].Furthermore, plants form conserved condensates like stress granules and processing bodies [11][12][13].Recent evidence suggests that like their animal counterparts, plant condensates can interface with membranes.For example, condensates of the TPLATE, a plant-specific complex modulating endocytosis, can likely form on the plasma membrane [14].We have also shown that condensates of processing bodies form on membranes in Arabidopsis and can attain polarity (i.e., localizing asymmetrically at the plasma membrane) [13].However, the functional significance of condensates at the plasma membrane is unclear.
In plants, the few known polar plasma membrane proteins provide crucial information for robust development [15][16][17].We have previously discovered a link between development and a complex comprising the Arabidopsis caspase-like protease separase (also named EXTRA SPINDLE POLES [ESP]) and 3 Arabidopsis microtubule (MT)-based centromeric protein-Elike Kinesins 7 (KIN7), which belong to the so-called KIN7.3-clade (KIN7.1,KIN7.3, and KIN7.5).This complex (the kinesin-separase complex [KISC]) is recruited to MTs; the most abundant and important kinesin from the KISC is KIN7.3 [18].ESP is an evolutionarily conserved protein responsible for sister chromatid separation and membrane fusion in both plants and animals [19,20].ESP binds to the KIN7.3-cladeC termini (the so-called "tails"), inducing conformational changes that expose the MT-avid N-terminal motor domain of KIN7s, thereby increasing KISC binding on MTs.The KISC can also modulate polar domains of the plasma membrane (PM), as the temperature-sensitive radially swollen 4 (rsw4) mutant harbouring a temperature-sensitive ESP variant or KIN7.3-clade mutants display reduced delivery of polar auxin efflux carriers PINFORMED (PINs) at the PM [18].Yet, how the KISC acts upon PM polar domains to regulate development remains elusive.
Whether condensates interfacing with membranes can undergo liquid-to-solid transitions like cytoplasmic ones and if these transitions would have any significance is unclear.Here, we discovered that a previously uncharacterized SEC14-like lipid transfer protein that we named SEC FOURTEEN-HOMOLOG8 (SFH8) recruits KISC to the PM.The ESP part of KISC trimmed SFH8 protein removing an IDR, leading to the conversion of SFH8 from a liquid to a more solid filamentous phase that remains attached to the PM, an event that we could also reconstitute in vitro.This liquid-to-solid transition was associated with SFH8 polarization, interaction with PIN2, and robust root development.Remarkably, we showed how spatiotemporally confined proteolysis can yield changes in the material properties of proteins and how these underlie robust development.

The KISC associates with the lipid-transfer protein SFH8 at polar PM domains
As the KISC regulates processes that are relevant to the PM (e.g., PIN delivery), we aimed to survey an underlying molecular mechanism.We observed that in the distal meristem of the root (as defined below), ESP and KIN7.3, detected by native antibodies, decorated the PM at apical domains in the outermost layer, the epidermis and basal domains in the adjacent layer, the cortex (S1A and S1B Fig) .We obtained similar results with ESP and KIN7.3 fluorescent fusions, under an estradiol-inducible module driven by the KIN7.3 promoter (for ESP) or the meristem-specific promoter RPS5a (for KIN7.3) (S1C-S1E Fig) .As KISC proteins lack lipidbinding motifs, we postulated that the KISC associates with the PM via a protein tether, which we sought to identify by screening a yeast two-hybrid (Y2H) library using KIN7.3 as bait (S2A Fig) .Among the 5 clones identified, we focused on AT2G21520, as its encoded protein showed localization reminiscent of the PM when expressed transiently in Nicotiana benthamiana leaves (S2B Fig; S1 Text).The protein encoded by AT2G21520 is a SEC14-like protein (BLAST-P: pAU : PleasenotethatPvalues}1e À 49}hasbeenreformattedto}1 � 10 À 49}inthesentence}T = 1 × 10 −49 ) that was ascribed the symbol SFH8, bearing a C terminal "nodulin"like motif (aa 479-637) punctuated with positively charged lysine (K) residues (S2C Fig and S1 File) [21].This nodulin-like motif promoted the interaction with KIN7.3 tail (S2A Fig) .SFH8 is a genuine SEC14-like protein, as it could rescue the Saccharomyces cerevisiae sec14-1 temperature-sensitive mutant (S2D Fig; 34.5˚C) [22].We confirmed that GFP-KIN7.3likely interacts with FLAG-tagged SFH8 in stable Arabidopsis lines, as shown by co-immunoprecipitation (Fig 1A).Furthermore, SFH8 associates with KIN7.3 likely at the PM, as evidenced by a transient ratiometric bimolecular fluorescence complementation (rBiFC) assay in Arabidopsis root protoplasts (Fig 1B).Unlike conventional BiFC, which lacks an internal reference marker, rBiFC can distinguish weak interactions from background fluorescence levels [23].In rBiFC, we used as a positive control the KIN7.3 interaction with the N terminus of ESP (aa 1-791; "DomA"), while as negative the KIN7.3 interaction with the C terminus of ESP (aa 1,622-2,178; "DomC") [18].
The localization and functions of SFH proteins in Arabidopsis are unknown; genetic evidence suggests that SFH1 is essential for root hair development [21].We expressed SFH8-mNeon under the SFH8 promoter (SFH8pro) to explore its localization; the SFH8-mNeon signal was exclusively observed in the root meristem.To expedite our localization analyses, we defined 4 developmental root regions along the proximodistal axis: core meristem (1; stem cell niche); meristematic zone (2; proximal meristem); meristematic/transition zone (3); and late transition zone (distal meristem; root regions described in Fig 1C) (4).We ).Later, we discuss this polarization in more detail, but altogether, these results suggest that KISC proteins associate with SFH8 at polar domains of the PM in root cells.

SFH8 clusters recruit the KISC where ESP cleaves SFH8 creating filaments
The interactions between KISC-SFH8 prompted us to examine whether SFH8 is tethering KISC at the PM.To address this question, we identified 2 T-DNA insertion mutants in SFH8, designated sfh8-1 and sfh8-2 (Fig 2A).We continued further analyses with the sfh8-1 mutant (hereafter "sfh8") because as explained laterAU : Pleasecheckandconfirmthattheedittothesentence}Weco , sfh8-1 phenotype is similar to sfh8-2.In sfh8, GFP-KIN7.3displayed both a reduced PM localization and polarity compared to that in the wild type (Fig 2B).SFH8-mNeon tethering at the PM did not appear to depend on KISC, as SFH8 could still be tethered at the PM in the 2 partial loss-of-function KISC mutants, the KIN7.3-clade mutant kin7.1 kin7.3 kin7.5 (k135; [18] To further validate this result, we used an inducible system that leads to the overaccumulation of the KIN7.3C terminal tail with the ability to deactivate KISC (XVEpro>KIN7.3pro:HA-KIN7.3tail), as it titrates ESP out of the active KISC [18].Thus, this transient depletion led to a loss of SFH8 polarity and perturbed gravitropism within 2 days, as expected given the link of KISC to PINs that regulate auxin and gravitropism (Fig 2D-2F and [18]).Hence, KISC and SFH8 synergistically define their localization: SFH8 tethers KISC at the PM, and, in turn, KISC promotes SFH8 polarization.
Interestingly, in follow-up experiments aiming at studying in detail the localization of SFH8 in k135 and rsw4 backgrounds, we observed that the full-length SFH8 levels increased in these 2 mutants (Fig 3A, "FL" arrowhead).In particular, immunoblot analysis of lines expressing a construct encoding SFH8 with a hexahistidine-triple-flag (referred to as HF)-mScarlet tag at the C or N terminus (approximately 106 kDa) under the control of the RPS5apro showed increased full-length SFH8 abundance in k135 and rsw4 backgrounds (Fig 3A).We used the RPS5apro here as the SFH8pro could not lead to a detectable signal in immunoblots.Given the increased abundance in k135 and rsw4 backgrounds of SFH8 and considering that ESP is a protease, we decided to examine the possibility that KISC regulates SFH8 levels.The reduced abundance of SFH8 levels in the wild type compared to that in k135 and rsw4 associated with a presumptive approximately 40-kDa (or approximately 10-kDa excluding mScarlet) N-terminal cleavage product (Fig 3A, arrowhead with asterisk).When SFH8 was tagged C-terminally with HF, it produced a double band, consistent with the cleavage of SFH8 at the N terminus (Fig 3A, right blot).We, thus, speculated that in the presence of KISC, SFH8 is cleaved from ESP close to its N terminus, producing a 10-kDa product (hereafter, identified as "cleavage product").We followed up the putative SFH8 cleavage in vivo using lines with SFH8 tagged Cor N-terminally with the mNeon fluorescent protein (Fig 3B).We observed fluorescent cytoplasmic puncta that accumulated gradually from regions 1 to 4 in mNeon-SFH8 expressing lines in regions 3 and mainly in 4 that, as shown above, KISC shows strong colocalization with SFH8; these foci were absent from the C-terminally tagged mNeon lines (SFH8-mNeon), where the SFH8 signal was mostly on the PM (Fig 3B), suggesting that likely only the N-terminus of SFH8 is cleaved and released in the cytoplasm.These results altogether suggest that SFH8 is progressively cleaved during development by ESP at the N terminus part, creating a cleavage product in the form of cytoplasmic puncta.
To further dynamically follow SFH8 cleavage in vivo, we established a double-labelled Nterminal/C-terminal tagged fluorescent SFH8 (hereafter "cleavage biosensor").We speculated that mScarlet-SFH8-mNeon cleavage would disrupt the colocalization of mNeon and mScarlet signals.Indeed, we observed a lack of mNeon/mScarlet colocalization in region 3 and mainly in 4 (Fig 3C).High-resolution imaging at the PM defined a more clustered form of SFH8 in region 1 (where both signals colocalize, indicative of an intact cleavage biosensor) and a more filamentous form in regions 3 and 4 of the remaining C terminal part of the SFH8 (where no colocalization between mNeon/mScarlet is observed) (Fig 3C, detail and graph).We thus showed that SFH8 transitions from a cluster (full-length protein) to a filament (containing only the C terminal part) upon its cleavage from ESP (Fig 3C, compare "filaments" versus "clusters").We also aimed at defining the exact cleavage site within SFH8.Accordingly, we immunoprecipitated SFH8-mScarlet-HF using α-FLAG and quantified the abundance of SFH8 peptides via mass spectrometry (MS), resulting in the identification of a potential cleavage site right after the residue R84 (Fig 3D and 3E).The size of the predicted cleavage fragment was in good agreement with the immunoblots shown in Fig 3A (approximately 10 kDa).The I 80 EDVR 84 D sequence corresponded to the reported non-plant ESP cleavage consensus motif φEXXR [24,25], also found in other SFH8-like proteins (S1 File).By establishing an in vitro ESP cleavage assay, we confirmed that immunopurified ESP, mitotically activated through coexpression with Cyclin D [26], can cleave recombinant glutathione S-transferase (GST)-SFH8 at R84; we validated our assay by showing the cleavage of a cohesin (SYN4), the well- Lower panel (left): super-resolution imaging of clusterto-filament conversion (epidermis regions 2 and 4).Note the absence of mScarlet signal from filaments (denoted as "no colocalization").The experiment was replicated 5 times.Scale bars, 0.8 μm.(D) SFH8 protein architecture (IDR corresponding to aa 1-96; CRAL-TRIO: active site for SEC14 proteins).The φEXXR cleavage motif for ESP is also shown.(E) SFH8 IDR peptides identified in pRPS5a:SFH8-mScarlet-HF pull-down experiments coupled with LC-MS/MS.Right: the cleavage motif of ESP on SFH proteins, φEXXR is conserved (presented here for 3 SFH protein paralogs, SFH6/8/9).P1'-P1 correspond to residues R and D, respectively.Raw data can be found in the Supporting information section (S1 Data and S1 Raw Images).DAG, day after germination; FL, full-length; IDR, intrinsically disordered region; KISC, kinesin-separase complex; PM, plasma membrane; rsw4, radially swollen 4; SFH8, SEC FOURTEEN-HOMOLOG8; WT, wild type.https://doi.org/10.1371/journal.pbio.3002305.g003known target of ESP (S3A-S3D Fig) [27].Hence, the filamentous conversion of SFH8 to filaments likely depends on the cleavage at R84.

SFH8 forms transient clusters with liquid-like properties that exclude KISC
We further aimed to follow the colocalization of KISC with SFH8 and the cleavage of SFH8 in more detail.As the KISC binds MTs [18], and since SFH8 showed a filamentous localization in regions 3 and 4, we postulated that SFH8 and the KISC might copartition in MT filaments in proximity to the PM.Contrary to our expectations, the marker MAP4 MBD (MT-binding domain of MICROTUBULE-ASSOCIATED PROTEIN4) or β-tubulin showed only partial colocalization with KIN7.3 at the PM at less than 10% of the filaments (S4A Fig) .Furthermore, amiprophos-methyl (APM) that disassembles MTs (10 nM; [18]) did not significantly alter KIN7.3 localization at the PM, although a small part of KIN7.3 filaments and, in particular, their edges remained attached in some cases in bundled MTs (S4B Fig; approximately 10%).In Arabidopsis roots, SFH8 filaments were short (<0.5 μm) and insensitive to APM treatment; ESP decorated similar filaments as shown in root cells expressing GFP-tagged ESP under an estradiol-inducible promoter driving expression at KIN7.3 domains (KIN7.As the previous results suggested that the KISC and SFH8 coassemble in cytoskeleton-independent filaments, we aimed at deciphering KISC and SFH8 localization in detail at the PM.We thus examined the localization of KISC components and SFH8 in Arabidopsis roots by total internal reflection fluorescence (TIRF) microscopy, which is suitable for analyzing the PM due to the shallow illumination penetration.By focusing on lateral cell junction domains (Fig 4A; regions 3 and 4; 3 to 5 days after germination), we determined that SFH8-mNeon segregates into at least 2 major populations: (i) immobile filaments that colocalize with KIN7.3-tagRFP and (ii) mobile or immobile KIN7.3-tagRFP-independent cluster-like structures (Fig 4B -4E).These results are consistent with the above observation of SFH8 and KISC localization in clusters and filaments (Fig 3C).The mobile SFH8 clusters showed little diffusion at the PM, circularity, and occasionally fused (or underwent fission), properties that are reminiscent of cellular condensates that sometimes form through LLPS (Fig 4C and 4D and S1-S4 Movies).We also observed some small nondiffusing clusters with reduced circularity (S3 and S4 Movies) that may show intermediate phases between the cluster state (droplet-like) and the filamentous state.As a cautionary note here, we did not examine other parameters used for cytoplasmic condensates, such as dripping or saturation concentrations as membrane-bound condensates, deform through the physical interfacing with the underlying lipids (the process known as wetting; [29][30][31][32]).KIN7.3 and SFH8 colocalized in short filaments but not in SFH8-decorated clusters; these clusters showed variable residence times at the PM, unlike filaments that were permanently assembled at the PM (Fig 4F and 4G).We wished to determine why the KISC was excluded from the SFH8 clusters; we hypothesized that converting the polybasic charge of the SFH8 nodulin patch to a hydrophobic region would promote the clustered (condensed) state of SFH8.This hypothesis is based on the counterion negative charge at the PM that could attenuate repulsion of positively charged residues at the nodulin part of SFH8 (suggested previously for membrane-associated peptides [32]).Indeed, replacing 6 pertinent lysines (K) with alanines (A; SFH8 6KtoA ) in the nodulin patch artificially increased SFH8 clustering in N. benthamiana leaves (although also exhibiting a slightly reduced localization at the PM) and decreased its association with KIN7.3 in Y2H (S4E and S4F Fig) .In Arabidopsis, mNeon-SFH8 6KtoA showed reduced localization at the PM, reduced filaments, and lacked polarity (S4G Fig) .This result suggested that hindering the interaction between SFH8 and the KISC by reducing the accessibility to the SFH8 blocked SFH8 filamentous transition.Hence, filaments are produced through KISC where KISC-SFH8 remain associated.On the other hand, the liquid-like SFH8 clusters are not accessible by KISC.

The N-terminus of SFH8 defines its liquid-like properties
To address the link between the removal of the SFH8 N-terminus by KISC and changes in SFH8 structure at the PM (i.e., the filamentous transition), we first aimed at linking SFH8 lack of diffusion with SFH8 filaments (considering their likely permanent residence at the PM).We thus used fluorescence recovery after photobleaching (FRAP) to follow SFH8 diffusion.Owning to their transient association of SFH8 clusters with the PM as shown by TIRFM, we anticipated that PMs with liquid-like SFH8 clusters would show increased FRAP rates.Indeed, SFH8-mNeon showed recovery at the PM close to the meristem (regions 1 and 2), unlike the distal meristem (regions 3 and 4 5A, model).Hence, FRAP confirmed that SFH8 filaments are stably attached to the PM and that cleavage by KISC could be somehow involved in this cluster-to-filament transition.
Next, we asked whether SFH8 cleavage fragment removal associates with the cluster-to-filament transition.Through in silico predictions, we determined that the cleavage fragment is an IDR (hereafter SFH8 IDR ; Fig 5C).We established that this protein architecture is conserved throughout the evolution of SFH proteins, which implies the functional importance of this IDR (S2 File).As IDRs are usually enriched in proteins undergoing LLPS [33] and considering the liquid-like behaviour of SFH8 clusters (described in Fig 4D and 4E), we tested whether full-length SFH8 undergoes LLPS.SFH8 IDR was predicted as an inducer of LLPS through the catGranule algorithm, while the corresponding region of a close SFH8 homolog, SFH6 (SFH6 IDR ), was predicted to exhibit a reduced propensity to undergo LLPS (Fig 5C); we verified this prediction in N. benthamiana where SFH6 could not form PM-localizing clusters (Fig 5D).SFH8 IDR sequence composition is distinct from that of animal proteins that undergo phase transitions in the cytoplasm (with prion-like domains (PLDs)) but show an amino acid distribution like that of the average IDR profile for Arabidopsis (Fig 5E) [34].We further observed that puncta formed by the SFH8 IDR failed to colocalize with vesicular markers and endosomes (SNX1, PI3P, and FM4-64), tonoplast (TIP), cellulose synthase complex (CESA6), or mitochondria, in Arabidopsis roots (mitotracker; S5A and S5B Fig) .Furthermore, SFH8 IDR showed LLPS hallmarks such as droplet-like dynamic morphology with frequent fission, fusion, and interconnections (S5C Fig and S5 Movie).FRAP analysis of these produced mNeon-tagged SFH8 IDR puncta demonstrated a rapid signal recovery (t 1/2 approximately 10 s, mobile fraction approximately 40%) and sensitivity to 1,6-hexanediol, which blocks in many cases LLPS (S5D and S5E Fig) [35,36]; 1,6-hexanediol dissolved SFH8 clusters on the PM but not SFH8 filaments (S5E Fig), confirming their solid-like structure.
Since SFH8 clusters at the PM displayed properties akin to condensates, we speculated that they may also form by LLPS much like SFH8 IDR .In silico prediction, using PLAAC (prion-like amino acid composition) and CIDER (classification of intrinsically disordered ensemble regions) showed that SFH8 can adopt context-specific conformational states with an absolute value of net charge per residue (NCPR) of 0.014, which suggests that is a polyampholyte [32].This result suggested that the propensity of SFH8 to undergo LLPS may be sensitive to the environment (e.g., lipid species) and that SFH8 may represent an ensemble of conformers.The highly electronegative field of the PM where SFH8 accumulates could unbalance opposite charges upon IDR removal (IDR NCPR = 0.071) due to repulsive forces, which would support a filamentous structure.Indeed, LLPS of SFH8 relied on the N-terminal IDR, as swapping it with the corresponding region from the SFH6, reduced clustering at the PM and the formation of puncta in the cytoplasm in the presence of ESP (Fig 5D, SFH6IDR SFH8).Overall, our analyses suggest that SFH8 behaves like an LLPS polyampholyte at the PM with negatively charged lipids (e.g., phosphatidylinositols (PIs)), likely buffering repulsive charges that restrict condensation.
To further examine LLPS of SFH8, we established an in vitro LLPS assay with fluorescently labelled proteins using thiol-reactive maleimide dyes (see Materials and methods).Under conditions that promote phase separation (S6A Fig for protein purification), recombinant GST-tagged SFH8 or the uncleavable variant SFH8 R84A formed condensates at relatively high concentrations (5 μΜ), while SFH8 ΔIDR (for delta IDR, i.e., SFH8 without the IDR) formed filament-like assemblies in good agreement with the in vivo situation (S6B Fig) .Consistent with the in vivo data, the SFH8 ΔIDR filaments showed no recovery after photobleaching and reduced circularity compared to SFH8 condensates (S6C Fig).
As mentioned above, phase separation at the PM could be affected by the interfacing of the condensate with lipids.Hence, as the above tests of bulk phase separation could be less relevant to SFH8, we established a system to test SFH8 phase separation on membranes.We used SUPER templates (supported lipid bilayers with excess membrane reservoir) that contain lowtension membranes surrounding a silicon bead [37].GST-SFH8 formed large droplets on SUPER templates containing PI lipids (i.e., PI(4,5)P2, as the yeast Sec14 binds on these lipids), at lower concentrations compared to the bulk-phase experiments (S6D Fig; 0.1 μΜ versus �5 μM in the bulk phase).This result suggested that membranes promote LLPS of SFH8.It is worth noting that, consistent with our data, the threshold concentration for LLPS in 2D systems like the PM can be an order of magnitude lower than in the 3D bulk phase (for example, [38]).By contrast, SFH8 ΔIDR did not show similar behaviour in this setting and formed oligomers (within 1 h) in native polyacrylamide gel electrophoresis; this behaviour could be also observed for SFH8 in the presence of KIN7.3 and ESP, as ESP cleaves the IDR converting SFH8 to SFH8 ΔIDR (S6E Fig) .Hence, as suggested above, PIs may neutralize electrostatic repulsions via counterion-mediated charge neutralization along SFH8, as suggested for other proteins, thereby mediating LLPS [39].

SFH8 phase separation enables delivery of some polar proteins
SFH8 IDR may act as an entropic bristle through random movements around its attachment point on lipids, which could, in theory, exclude access of other proteins to PM regions where uncleaved SFH8 resides [42].This property would also reduce the probability of full-length SFH8 undergoing filamentous transition due to inter-or intramolecular stereochemical hindrance imposed by the IDR [42].We thus aimed to decipher the significance of SFH8 phase transition at polar domains.As a relevant readout here, we used PIN2 because KISC plays a role in PIN2 delivery [18], but this choice is not implying a strict link between SFH8/KISC to auxin signalling.We observed that the PIN2-GFP (or α-PIN2 by immunohistochemistry) signal is lower by about 50% at the PM of sfh8 or KISC mutants (Fig 6A and 6B), suggesting that SFH8/KISC are required for PIN2 delivery, stability, and/or maintenance on polar domains.Notably, PIN2 accumulated in endosome-like structures in sfh8, while KISC or sfh8 mutants showed a slightly reduced PIN2 polar delivery mainly in the cortex (Fig 6B and 6C).Furthermore, the uncleavable variant of SFH8 R84A could not rescue the PIN2 defects of sfh8 (Fig 6C).We also observed increased localization to endosomes and a reduced delivery or maintenance for PIN1 on PM in sfh8 (likely not for other PINs), but not for nonpolar proteins (the H + -ATPase 1 [AHA1] or PLASMA MEMBRANE INTRINSIC PROTEIN 2a [PIP2a]), discounting a general role for SFH8 in exocytosis (S8A and S8B Fig) .These results suggest that cleavage of SFH8 and, thus, its conversion to filaments is required for the establishment of some polar PM domains.
Next, we asked whether SFH8 promotes the delivery of polar proteins or their maintenance to the PM.To this end, we used the drug brefeldin A (BFA) to induce intracellular agglomerates of PIN2 in so-called BFA bodies (aggregate of trans-Golgi network [TGN] and Golgi).We calculated the endocytosis rate of PIN2 to BFA bodies and the delivery rate from PIN2-positive BFA bodies back to the PM after BFA washout [43].We further validated BFA experiments with FRAP to measure the rate of PIN2 delivery at the PM.Both assays confirmed that PIN2 delivery to the PM is compromised in sfh8 and KISC mutants, while PIN2 endocytosis was not, as PIN2-positive BFA bodies were produced at the same rate in the wild type and the sfh8 mutant; these effects were independent of de novo PIN2 synthesis, as short cycloheximide treatments did not affect delivery rates or dissolution of the PIN2 endosomes in sfh8 (Figs 6D-6F, S8A and S8B).Furthermore, SFH8 did not significantly colocalize with clathrin clusters at the PM and sfh8 did not show defects in endocytosis traced by the FM4-64 or the peptide PEP1, which is internalized by clathrin-mediated endocytosis (S9 Fig) [44].These results likely exclude the possibility that PIN2 removal from the PM is due to increased endocytosis in sfh8.Similarly, endocytosis was not affected in KISC mutants [28].
To address the mechanism by which SFH8 might affect PIN2 delivery, we checked SFH8 and PIN2 localization dynamics at the PM.In region 4, SFH8-mScarlet (and KIN7.3), but not mScarlet-SFH8 clusters, colocalized with apicobasal-localized PIN2 and showed similar polarity (Fig 7A -7C; PCC approximately 0.9).On the contrary, PIN2-GFP and mScarlet-SFH8 PM puncta showed an anticorrelation of localization, excluding each other in regions 1 and 2, as observed in a super-resolution setting (Fig 7D; insets).Thus, SFH8 IDR properties in SFH8 clusters may reduce the delivery of proteins like PIN2 at the PM in regions 1 and 2. To address whether the entropic bristle effect is responsible for this exclusion, we evaluated whether the formation of SFH8 ΔIDR and the transition to filaments might permit delivery of PIN2, which would likely be manifested as increased SFH8 ΔIDR (filaments) proximity to PIN2 (as the 2 proteins colocalize).Using a specific antibody against SFH8 (C terminus), we observed that SFH8 localization is not affected in a pin2 mutant, suggesting that SFH8 affects PIN2 delivery, and not the other way around (S10A Fig) .To test for interactions between PIN2 and SFH8 in root cells, we refined a quantitative proximity ligation assay (PLA; [45]).PLAs use complementary oligonucleotides fused to antibodies to determine the frequency with which proteins of interest find themselves nearby (Fig 8A).We observed positive interactions between PIN2 and SFH8 in regions 3 and 4 (i.e., when SFH8 is in the form of SFH8 ΔIDR ) in epidermis and ).Arrowhead denotes an SFH8 cluster formed on the LUV liposome (images after deconvolution).The experiment was replicated 3 times.Scale cortex, using 2 different settings: (i) in roots expressing PIN2-GFP (PLA antibody combination α-GFP/α-SFH8) or (ii) in roots expressing PIN2-GFP and a C-terminally HF-tagged SFH8 (PLA antibody combination α-GFP/α-FLAG).We observed significantly lower PLA signals between PIN2 and SFH8 in the N-terminally tagged HF-SFH8 lines (α-PIN2/α-FLAG) as the full-length SFH8 showed anticorrelation of localization with PIN2.We detected no PLA signal for (i) SFH8 and a PM aquaporin (PIP2a-GFP; α-GFP/α-SFH8), (ii) in the sfh8 mutant (α-PIN2/α-SFH8), and (iii) in the vasculature where PIN2 is absent (Figs 8B-8E and S10B-S10D).PLA also showed that KIN7.3 interacts with SFH8 at the PM (S10C Fig; region 3 onwards), confirming the result from the rBiFC in Fig 1B .To follow these results in a live imaging setting, we used Fo ¨rster resonance energy transfer (FRET) analyses, in which we detected high FRET efficiency for the SFH8-mScarlet/PIN2-GFP pair, indicative of interaction, in epidermis or cortex of regions 3 and 4 (Fig 8F and 8G).Collectively, these results suggest that SFH8 filamentous conversion (SFH8 ΔIDR ) allows the association of proteins like PIN2 with the PM.
We then asked whether the observed anticorrelation between SFH8 cluster signal and PIN2 might indeed imply stereochemical hindrance through the entropic bristle effect imposed by full-length SFH8 that could restrain delivery of proteins such as PIN2.We thus established an in vitro membrane fusion assay as a proxy of stereochemical hindrance (or entropic bristle) at membranes based on cholesterol-modified DNA zippers (lipid-DNA-zippers; Fig 8H).DNA zippers promote fusion in the absence of other proteins [46].To test SFH8 effect on fusion, we used lipid-DNA-zippers assays with low content of labelled phosphatidylethanolamine (PE)-Texas red to decorate the periphery of large unilamellar vesicles (LUVs) , which also contained the dye fluorescein (LUVs; 400 nm).Under our super-resolution settings, we resolved 3 events driven by DNA zippers: membrane tethering, hemifusion (lipid mixing), and fusion resulting in the unification of the lipid bilayer and the intermixing of the volumes (Fig 8H).As SFH8 ΔIDR converted to filaments in a few minutes, to ascertain that observed effects would not be due to differential binding of SFH8 ΔIDR on LUVs (or other surfaces) due to its rapid conversion to filaments, we used quartz crystal microbalance with dissipation (QCM-D) to monitor SFH8 or SFH8 ΔIDR binding on LUVs in real time (S10E and S10F Fig; method details in the figure legend).QCM-D can reveal the interaction dynamics between lipids/proteins and/or the sensor surface, translating differential binding of proteins in an observable real-time response.The detection is based on measurements that depend on real-time mass changes [5], i.e., how much protein would bind on the sensor surface.In QCM-D, both SFH8 and SFH8 ΔIDR show only basal affinities toward LUVs (and to KIN7.3 tail), thus excluding differential binding as a potential driver of the changes in fusion dynamics.In the SFH8 ΔIDR samples, the average diameter of liposomes was approximately 1 μm, in contrast to SFH8 (approximately 0.5 μm), which was below that of free GST samples (approximately 0.7 μm) (Fig 8I).Content mixing analyses showed that almost 30% of the SFH8 ΔIDR samples show semi-fuse/ fused LUVs (approximately 2-fold lower for full-length SFH8), while fusion/hemifusion events with SFH8 were even less than those with GST (Fig 8J and 8K).This result suggests that the N-terminal SDH8 IDR when on SFH8 exerts an entropic bristle effect blocking the delivery of proteins, while SFH8 ΔIDR allows or even promotes this delivery.bars, 2 μm.(K) Super-resolution micrographs showing liposome (LUVs) content mixing with lumen stained with fluorescein (green), and lipids stained with PE-Texas red (magenta).Scale bars, 2 μm.Insets (right) show a hemifusion event (upper), and a combination of hemifusion with a tethered LUV (lower inset).The arrowheads show content mixing (pseudo-coloured white).Lower: quantification of the distribution (%) of LUVs in fused, unfused, and hemifused in the presence of recombinant GST, GST-SFH8, or GST-SFH8 ΔIDR (data are means ± SD, N = 3 pooled experiments, n = 32-40 fields of view per experiment; p-values were calculated by Dunnett for "fusion" relative to GST).Raw data can be found in the Supporting information section (S1 Data).DAG, day after germination; FRET, Fo ¨rster resonance energy transfer; GST, glutathione S-transferase; LUV, large unilamellar vesicle; PLA, proximity ligation assay; PM, plasma membrane; SE, sensitized emission; SFH8, SEC FOURTEEN-HOMOLOG8; WT, wild type.https://doi.org/10.1371/journal.pbio.3002305.g008

SFH8 cleavage by KISC mediates developmental robustness
Next, we asked whether the KISC-SFH8 link has biological meaning.At the seedling stage, the sfh8 phenotypes resembled those of KISC mutants showing both reduced root growth, gravitropism (compared to pin2), and slower response to gravistimulation; these phenotypes were rescued by tagged SFH8 (Figs 9A-9C and S11A-S11F; note the similarity between sfh8-1 and sfh8-2).We did not observe additive phenotypes of seedlings with KISC/SFH8 mutant combinations (Fig 9D; rsw4 sfh8 and k135 sfh8), suggesting functional convergence between KISC and SFH8.Adult mutants also showed a shorter stature, decreased branching, and smaller cotyledons, or leaves (S11A-S11C Fig) .Yet, sfh8 produced shorter siliques and exhibited a more severe adult phenotype than the one reported for the KISC mutants (S11A-S11C Fig) , suggesting additional functions for SFH8.In sfh8, the apical meristem length was approximately 2-fold smaller than in the wild type, likely due to lower mitotic activity as defined by the Cell Cycle Tracking system (Fig 9E and 9F; [47]).Although SFH8 is a SEC14-like protein and would, therefore, be expected to be involved in lipid homeostasis, its loss of function (or of KISC) did not affect lipid levels at the PM (S11D Fig) , suggesting that sfh8 phenotypes did not relate to perturbations in lipid homeostasis.Noteworthy, all the SFH8 constructs used rescued the sfh8 seedling phenotype (S11E and S11F Fig).
We further examined whether the phase transitions of SFH8 (cluster-to-filament) are biologically significant.The uncleavable SFH8 variant (SFH8 R84A ), which cannot form filaments, failed to rescue sfh8, and the same was observed for SFH8 6KtoA (S12A and S12B Fig) .On the other hand, deletion of the IDR (SFH8 ΔIDR ) also led to a significant loss of SFH8 polarity in roots, a lack of root developmental robustness, and only partial sfh8 rescue (S12C and S12D

Discussion
Here, we identify a condensate undergoing phase transitions on membranes.Collectively, the mechanism of these transitions consists of (i) SFH8 LLPS at the PM mediated by an IDR, (ii) SFH8 interaction with KISC, and (iii) proteolytic cleavage of SFH8 by KISC, followed by a phase transition that allows interactions with polar proteins.This module contributes to developmental robustness by regulating PM domains.Intriguingly, SFH8 phase transition is induced by the highly conserved protease ESP.So far, ESP targets have mainly been linked to functions in dividing cells.Apart from expanding the targets of ESP, this module also uncovers a novel way to regulate LLPS via proteolytic processing.We further speculate that the released N-terminal SFH8 IDR retains features of the LLPS SFH8 state even after its cleavage, suggesting a structural memory for condensates.In this direction, further work will reveal whether proteolysis is a general regulator of phase transitions.
Condensation is crucial for polarity establishment; yet, our work might appear counterintuitive, as it starts challenging (or extending) these models by showing that LLPS may simply be a mechanism for reducing polarized secretion.In synapses, for example, condensation exerts an opposite effect to those described here [1].SFH8 liquid condensates prelude shifts in material properties (from LLPS to likely more solid filaments) that promote a functionality switch for SFH8: from blocker to fusion enhancer.Likewise, Dynamin-related proteins (DRPs) oligomerize to remodel membranes [48], while coronaviruses (e.g., SARS-CoV-2) co-opt host proteases for structural reconfiguration that prime activity of SPIKE resulting in the fusion between the virus and its host cell [49,50].Upon IDR removal, SFH8 ΔIDR was stabilized, likely due to reduced intramolecular stereochemical repulsion and charge attenuation, which allowed filamentous state conversion and fusion at polar domains.The stabilization effect of SFH8 filaments may have to do with a steric barrier imposed by the increased size of filaments, which would be more resistant to endocytosis.This proposition is in accordance with recent findings suggesting that glycosylation presents a steric barrier for endocytosis by increasing the size of proteins [51].Overall, this mechanism establishes an example of LLPS with importance for PM micropatterning that transcends to polarized patterns.
Furthermore, the dimensionality reduction caused by the PM binding of SFH8 promotes condensation.Cluster formation through LLPS could have a direct effect on PM properties.Condensates can potentiate lipid clustering [52], and SFH8 could in turn affect lipid clustering with important roles in signalling during development.We believe that these functions may also be relevant during stress, given the important roles of lipids in stress signalling [53].Furthermore, the conversion of SFH8 IDR that has microscale entropic bristle-like properties in mesoscale condensates in the cytoplasm (the SFH8 IDR -bodies) with large diameter suggests that SFH8 IDR can restrict interactions.At this level, these properties of SFH8 could exclude certain proteins from binding to the PM and can likely establish sites for vesicle exclusion.We reconciled this proposition in vitro by showing that the removal of SFH8 IDR by ESP locally promoted attraction and fusion by alleviating steric hindrance.Alternatively, SFH8 clusters may engulf diffraction-limited vesicles and promote their fusion upon the removal of the IDR.Because SFH8 is a SEC14-like protein, it may render lipids vulnerable to enzymatic modifications [54], regulating local lipid environments at the microscale or nanoscale.These lipid modifications could thus promote vesicular fusion.SFH8 ΔIDR may also reduce the energy barrier required for fusion through an increase in fluidity by its filamentous structure, as has been shown for MTs [55], or the reduction of the entropic bristle effect.We provide a model for SFH8 functions in Fig 10 .Addressing further how KISC/SFH8 functions is an important priority for our future research.The details of structural modifications for SFH8 especially upon the removal of the SFH8 IDR need further exploration.Intriguingly, as aforementioned, the IDR of the SFH8 is conserved throughout evolution (S1 and S2 Files), suggesting that proteins with similar features and functions should exist in other eukaryotes.Other pertinent questions are, 'How do SFH8 clusters form in the first place, and what is the function (if any) of the cytoplasmic "SFH8 IDR -bodies" condensate?' Our work thus provides insights relevant to condensates interfacing with membranes.

Plant growth conditions
Arabidopsis seedlings were surface sterilized and germinated on ½ strength Murashige and Skoog (MS) agar medium with sucrose, under long-day conditions (16-h light/8-h dark) and were harvested, and treated, or examined as indicated in the context of each experiment.In all experiments involving the use of mutants or pharmacological treatments, the medium was supplemented with 1% (w/v) sucrose, unless stated otherwise.Arabidopsis plants/lines for crosses, phenotyping of the aboveground part, and seed harvesting were grown on soil in a plant chamber at 22˚C/19˚C (or 28 to 30˚C for restrictive temperature treatments in the case of rsw4), 14-h/10-h light/dark cycle, and light intensity 150 μE m −2 s −1 .N. benthamiana plants were grown in Aralab or Percival cabinets at 22˚C (16-h/8-h light/dark cycle, and light intensity 150 μE m −2 s −1 .SFH8 (full-length) is recruited on the plasma membrane in a nonpolar manner and forms LLPS clusters at the PM (cell 1).These clusters seem to block the delivery of polar proteins to the PM (e.g., PIN2).Later, KISC is recruited on SFH8, and KISC proteolytic part (ESP) cleaves the N-terminal part of SFH8, which results in the production of SFH8 ΔIDR (SFH8 lacking the N-terminal IDR; cell 2).The released IDR floats in the cytoplasm in the form of LLPS puncta.Microtubules may direct the KISC at the PM; however, microtubules/KISC do not remain associated at the PM.Later, SFH8 ΔIDR ages to a filamentous form with clear polarity at the PM and interacts with other polar proteins, like, for example, PIN2 (cell 3).This phase transition of SFH8 and its polar localization are essential components of robust root development.The released IDR persists in the cytoplasm and has an unknown role.ESP, EXTRA SPINDLE POLES; IDR, intrinsically disordered region; KISC, kinesin-separase complex; LLPS, liquid-liquid phase separation; PM, plasma membrane; SFH8, SEC FOURTEEN-HOMOLOG8AU : AbbreviationlistshavebeencompiledforthoseusedinFigs1 À 10:Ple . https://doi.org/10.1371/journal.pbio.3002305.g010

Phenotypic analyses and drug treatments
For quantification of phenotypes, seeds were surface sterilized, plated on MS medium, and seedlings were grown vertically.Customized Smart Plate Imaging Robot (SPIRO) imaging was done with 15-min intervals of fully automated imaging acquisition (https://www.alyonaminina.org/spiro), in a growth cabinet (Aralab).Arabidopsis Col-0 was used as the wild-type control.To define root length, images were captured of the plates using a Leica DM6000 with a motorized stage and computationally compiled together.Root length or size was determined using Image J/Fiji (National Institute of Health).For 1,6-hexanediol treatments, a 10% (v/v) aqueous solution was used.XVE-driven expression was activated by transferring seedlings on ½ MS plates containing various estradiol (in ethanol; light-tight tube aliquots) concentrations, which were determined experimentally in trial experiments (2 to 100 μM).Cycloheximide, APM, latrunculin B, and BFA treatments were done in liquid ½ MS, as described previously [13,18,28].

Root gravitropism assays
To observe root gravitropism, seedlings were grown vertically on plates (½ strength MS agar medium under long-day conditions (16-h light/8-h dark) for 12 DAG.The root tip angle change was measured using Fiji.

Evolutionary relationships of taxa and sequences analyses
The evolutionary history of SFH8 was inferred using the Neighbor-Joining method [67].The optimal tree is shown.The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1,000 replicates) is shown next to the branches [68].The evolutionary distances were computed using the Poisson correction method [69] and are in the units of the number of amino acid substitutions per site.This analysis involved 201 amino acid sequences.All positions with less than 70% site coverage were eliminated, i.e., fewer than 30% alignment gaps, missing data, and ambiguous bases were allowed at any position (partial deletion option).There was a total of 611 positions in the final dataset.Evolutionary analyses were conducted in MEGA X [70].

Recombinant protein production and purification from E. coli
The pGAT4/PDEST15 constructs were transformed in BL21 (DE3) Rosetta or BL21 (DE3) Rosetta II E. coli cells.Bacterial cultures were grown in 800 mL of LB supplemented with 100 mg L −1 of ampicillin and 25 mg L −1 of chloramphenicol.Protein production was induced at OD 600 = 0.5 with 0.05 to 1 mM IPTG (isopropyl ß-D-1-thiogalactopyranoside).After 3 h, the cells were harvested by centrifugation at 2,500g for 20 min at room temperature (RT) and frozen overnight at −80˚C.Preparation of his-tagged recombinant proteins was performed according to manufacturer instructions (Qiagen).Preparation of GST-tagged recombinant proteins was performed according to manufacturer instructions, using Sepharose beads (GE Healthcare Life Sciences), while the pH of purification was 8.3.Expression levels of proteins were estimated by CBB staining in PAGE or by immunoblots.The proteins were dialyzed overnight in assay buffers (2 L).

Protein immunopurification
Constructs expressing various forms of KIN7.3, SYN4, SFH8, ESP, or TAP-GFP were infiltrated into N. benthamiana leaves.Three to four days later, leaves were ground in liquid nitrogen and resuspended in 10 volumes of buffer A (50 mM Tris-HCl [pH 7.5], 5% [v/v] glycerol, 10% [v/v] Ficoll, 0.1% [v/v] Triton X-100, 300 mM NaCl, 5 mM MgCl2, 1 mM EDTA, 1 mM EGTA, plant-specific protease inhibitor cocktail [Sigma], phosphatase inhibitors [Roche], and 1 mM PMSF) and centrifuged at 14,000g for 20 min at 4˚C.The supernatant was filtered through 4 layers of Miracloth (Calbiochem).For TAP-ESP capture, samples were mixed with immunoglobulin G beads and incubated at 4˚C for 1 h with gentle rotation.Beads were precipitated by centrifugation at 300g, washed 3 times with buffer A, and treated for 4 h with PreScission protease (GE Healthcare).The TAP-ESP beads were used directly, or the supernatant was incubated for 30 min with nickel beads and his-ESP was eluted with 250 mM imidazole containing buffer A. Protein was dialyzed against 0.1 M PIPES (pH 6.8), 5 mM EGTA, 2 mM MgCl 2 , and 20% (v/v) glycerol buffer.Protein levels were estimated by immunoblot.

SFH8 in vitro cleavage assays
GFP-ESP was extracted from infiltrated N. benthamiana leaves as above using GFP-TRAP (CromoTek).GFP-ESP protein lysates were mixed with GST beads carrying GST-SFH8 or the corresponding mutants.The samples were left agitating at 37˚C for 1 to 2 h.

Protein labelling for LLPS
The dyes used for labelling were Alexa Fluor 647 C2 Maleimide and Alexa Fluor 555 C2 Maleimide (Thermo Fisher Scientific).The dyes were dissolved in water-free DMSO (dimethyl sulfoxide) at a concentration of 1 mM.After the protein isolation with GST-affinity chromatography, the proteins were dialyzed against 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl.Proteins were partially labelled in solution in a ratio of 1 mol protein to 0.1 mol of dye (1:10 molar ratio labelled/unlabeled) for 15 min on ice.The labelled proteins were kept at 4˚C and used within the next 3 d.

Preparation of surfaces for suspended liposomes
Before the experiments, Au-coated AT-cut crystals (QSX-301 Biolin Scientific, Q-Sense, Sweden) were immersed in 2% (v/v) Hellmanex solution and rinsed with double distilled water.The gold sensors were further dried using nitrogen flow and cleaned using UV-ozone cleaner (E511, Ossila, Sheffield, UK) for 30 min.The QCM-D experiments were performed using the Q-Sense E4 (Biolin, Q-Sense, Sweden) instrument and AT-cut quartz disks (5 MHz).All the experiments were performed in a buffer solution under a constant flow rate of 50 μL/min at 25˚C.Briefly, the gold surface was equilibrated with buffer, followed by neutravidin adsorption (200 μL, 0.2 mg/mL protein solution).5 0 -biotinylated, 3 0 -cholesterol-modified DNA was then used (200 μL, 0.075 pmol/μL) as a binding anchor to liposomes of different lipid compositions as previously described [71].All liposome solutions were used at a concentration of 0.2 mg/ml and a final volume of 70 μL.Finally, protein solutions of various concentrations were added at a volume of 500 μL.

Liposome DNA-mediated fusion assay and SUPER template setting
To establish DNA-zippers, the concentrations of oligomers-cholesterol were adjusted, and various temperatures (4 to 25˚C) were used to achieve DNA zipper-driven fusion at a percentage lower than approximately 10% (noise).Eventually, all liposomes irrespective of the conditions used, given enough time fused to a percentage of approximately 50%.The liposomes used were unilamellar vesicles (LUVs) of 400 nm diameter, with the following composition: DOPC 71.5%, DOPS 27.5%, PE-Texas Red conjugated 0.5% and PI(4,5)P 2 0.5% (PS was included for SUPER template experiment to provide required membrane flexibility).We also used large LUVs without PE-Texas Red lipid.Instead, they carried the fluorophore fluorescein (cf, 100 mM fluorescein in Tris-HCl [pH 7.5]).For the fusion of the 2 different liposome populations, we incubated each with a DNA primer for 45 min at RT.The sequence of the single-stranded DNA molecules can be found in S1 Table .For every liposome, 100 DNA molecules were used [72].SUPER templates were prepared, as described previously [37].Silica beads (monodispersed 4.9 μm in diameter; CorpuscularAU : PleasenotethatasperPLOSstyle; donotuseInc:; Ltd:; etc:exce , Cold Spring, NY) were added to a premixed solution of liposomes in LLPS buffer in a total volume of 100 μL in a 1.5-mL clear polypropylene centrifuge tube.Glass coverslips (Fisher Scientific, Pittsburgh, PA) were cleaned with piranha solution (concentration H 2 SO 4 /30% H 2 O 2 4:1, v/v) for 1 h at room temperature and washed extensively with boiling water.The coverslips were allowed to attain RT, stored underwater, and air-dried before use.To coat the coverslips with PEG-silane, piranha-cleaned coverslips were first dried in an oven at 200˚C for 3 h and then treated at RT for 30 min with PEG-silane (2%, [v/v]; GelestAU : PleasenotethatasperPLOSstyle; donotuseInc:; Ltd:; etc:exceptasappropriateinthea , Morrisville, PA) in acetone.The coverslips were later washed extensively with water and air-dried before use.The samples were observed with a confocal LIGHTNING SP8 module (high speed and resolution of 120 nm) using an observation chamber and an Apochromat 63× objective (NA = 1.4).Images were deconvoluted by Leica's built-in software.

Phase separation assays
The slides and coverslips used for confocal microscopy were treated overnight with 1 to 2 mg/ mL PLL-PEG, after cleaning with 2% (v/v) Hellmanex for approximately 2 h, at RT.The LLPS buffer used was the following: 10 mM HEPES (4-(2-hydroxyethyl)-1-piperazine ethane sulfonic acid), 150 mM NaCl, 0.1 mM EDTA, 2 mM DTT (pH 7.4), with and without 10% (w/v) polyethylene glycol (PEG 3000) as a crowding agent.The amount of the added labelled protein was 1 to 2 μM for 3 h at RT.Some of our samples contained LUVs consisting of DOPC 0.5%: DOPS 99%: DOPE-Texas Red 0.5%.Phase separation was imaged via confocal microscopy.

Transient assays in Nicotiana benthamiana
For transient assays, A. tumefaciens strain GV3101 was used.Infiltrations were done as described previously [73].During treatments, plants were kept in defined growth conditions in an Aralab growth chamber.

Ratiometric bifluorescence complementation and protoplast transformation
The rBiFC assay was done in 5 DAG Arabidopsis root protoplasts via polyethylene glycol transformation as described previously [74].After a 16-h incubation following transformation, protoplasts were observed with a Zeiss LSM780 laser scanning confocal microscope using 40×/ 1.2 W C-Apochromat in multitrack channel mode.Excitation wavelengths and emission filters were 514 nm/band-pass 530 to 550 nm for YFP, and 561 nm/band-pass 600 to 630 nm for RFP.

Quantification of fluorescent intensity, FRAP, and FRET
Fluorescence was measured as a mean integrated density in regions of interest (ROIs) with the subtraction of the background (a proximal region that was unbleached and had less signal intensity than the signal of the ROI region).FRAP mode of Zeiss 780 ZEN software was set up for the acquisition of 3 pre-bleach images, 1 bleach scan, and 96 post-bleach scans (or more).Bleaching was performed using 488, 514, and 561 nm laser lines at 100% transmittance and 20 to 40 iterations depending on the region and the axial resolution (iterations increased in deeper tissues to compensate for the increased light scattering).In FRAP, the width of the bleached ROI was set at 2 to 10 μm.Pre-and post-bleach scans were at minimum possible laser power (0.8% transmittance) for the 458 nm or 514 nm (4.7%) and 5% for 561 nm; 512 × 512 8-bit pixel format; pinhole of 181 μm (>2 Airy units) and zoom factor of 2.0.The background values were subtracted from the fluorescence recovery values, and the resulting values were normalized by the first post-bleach time point and divided by the maximum point set maximum intensity as 1.The objective used was a plan-apochromat 20× with NA = 0.8 M27 (Zeiss).For FRET, the sensitized emission (SE) module of the SP8 Leica confocal microscope was used, with standard modules and internal calibration.Fluorescence was detected using a water-or oil-corrected 40× objective.For SE-FRET, the correction factors β, α, γ, and δ were calculated with the donor-and acceptor-only reference samples, as described for TMK1-AHA1 interaction [75].

TIRFM imaging and tracking analyses
TIRF microscopy images were acquired using MetaMorph software on an Olympus IX-81 microscope.The system was maintained at 37˚C during imaging.A DV2 image splitter (MAG Biosystems) was used to separate GFP and RFP emission signals.Time-lapse movies were obtained at 100-ms intervals.For MSD analysis, 30-s-long movies with 100-ms intervals and 200-ms exposure were used.Particle tracking was limited to the amount of time that PM remained in a single focal plane; the median track length was 2,000 frames, corresponding to 3.5 s of imaging.The tracking of particles was performed with the Mosaic suite or Nano-TrackJ/TrackMate of Fiji, using the following typical parameters: radius 3 of fluorescence intensity, a link range of 1, cutoff of 0.1%, and a maximum displacement of 8 px, assuming Brownian dynamics.

Super-resolution imaging
Confocal SP8 Leica confocal microscope LIGHTNING module was used for super-resolution on-the-fly imaging.Confocal images were obtained when indicated at maximum scanning speed (40 frames per second) using a 63× water immersion objective with a theoretical x/yaxial resolution of 120 nm upon deconvolution.Post-acquisition, images were deconvoluted using the LIGHTNING algorithm, and water correction was set in the algorithm as the mounting medium.Imaging was done at RT in an inverted microscope setting.

LC-MS/MS lipids and proteins analyses and overlay assays of lipids
Lipids were extracted from roots after inactivating the tissue with boiling water or from microsomes [76].Lipid classes were purified by solid-phase extraction, and phospholipids, glycolipids, diacylglycerol, triacylglycerol, and total fatty acids were measured by quadrupole time of flight mass spectrometry (Q-TOF MS/MS) or gas chromatography-flame ionization detection (GC-FID), respectively, as previously described [77].For proteomic analyses of FLAG-tagged SFH8 immunoprecipitates, samples were analyzed by LC-MS using Nano LC-MS/MS (Dionex Ultimate 3000 RLSCnano System) interfaced with Eclipse Tribrid mass spectrometer (Thermo Fisher Scientific).Samples were loaded onto a fused silica trap column Acclaim Pep-Map 100, 75 μm × 2 cm (Thermo Fisher Scientific).After washing for 5 min at 5 μL/min with 0.1% (v/v) trifluoroacetic acid (TFA), the trap column was brought in line with an analytical column (Nanoease MZ peptide BEH C18, 130 A, 1.7 μm, 75 μm × 250 mm, Waters) for LC-MS/MS.Peptides were fractionated at 300 nL/min using a segmented linear gradient 4% to 15% B in 30 min (where A: 0.2% formic acid, and B: 0.16% formic acid, 80% acetonitrile), 15% to 25% B in 40 min, 25% to 50%B in 44 min, and 50% to 90% B in 11 min.Solution B then returned at 4% for 5 min for the next run.The scan sequence began with an MS1 spectrum (Orbitrap analysis, resolution 120,000, scan range from M/Z 350 to 1,600, automatic gain control (AGC) target 1E6, maximum injection time 100 ms).The top S (3 589 s) and dynamic exclusion of 60 s were used for the selection of Parent ions for MS/MS.Parent masses were isolated in the quadrupole with an isolation window of 1.4 591 m/z, AGC target 1E5, and fragmented with higher-energy collisional dissociation with a normalized collision energy of 30%.The fragments were scanned in Orbitrap with a resolution of 30,000.The MS/MS scan range was determined by the charge state of the parent ion, but the lower limit was set at 100 amu.

TAMRA-PEP1 and FM4-64 internalization assays
The peptide PEP1 (ATKVKAKQRGKEKVSSGRPGQHN) was labelled with 5 0 -801 carboxy tetramethylrhodamine at the N-terminus (TAMRA-PEP1) with an HPLC purity of 95.24% and molecular weight of 2,905.24(EZBiolab).The peptide was dissolved in water to obtain 1 mM peptide stocks.Further dilutions were done with ½ MS medium.Five-day-old seedlings were dipped into 1 mL ½ MS medium containing 100 nM TAMRA-PEP1 for 10 s, washed 5 times, and kept in 24-well plates with ½ MS medium for 40 min.Epidermal cells at the meristematic zone were imaged with a Zeiss LSM710 inverted laser scanning confocal equipped with 40×/1.2W C-Apochromat M27 objective at a zoom factor of 3.5.TAMRA-PEP1 was excited at 559 nm, and fluorescence emission was captured between 570 and 670 nm, GFP was excited at 488 nm, and fluorescence emission was captured between 500 and 540 nm.The FM4-64 experiments were as described in [28].In brief pulse labelling with FM4-64 (2 μM FM464, Molecular Probes; made from a 2-mM stock in DMSO) was done for 5 min (time 0) and then analyzed for 18 min at RT in Zeiss LSM 780 with excitation at 488 nm and fluorescence emission captured between 540 and 670 nm.After the pulse with FM4-64, the roots were washed 2 times in an ice-cold MS medium.

Quantification and statistical analyses
Statistical analysis was performed in R studio (R-project.org) or GraphPad Prism (version 9.2.0).Each data set was tested for normal distribution (Gaussian) by the Shapiro-Wilk, D' Agostino-Pearson, Anderson-Darling, and Kolmogorov-Smirnov normality tests.Lognormal versus Gaussian distributions were also evaluated (not used though herein).For tests involving pairwise comparisons, Student t test, or Wilcoxon tests (or as indicated) were mainly used to define whether differences were statistically significant.The significance threshold was set at p < 0.05 (significance claim-level alpha), and the exact values are shown in graphs (p-values <0.0001 are indicated as such).For tests involving multiple comparisons, 1-way ANOVA, or the Kruskal-Wallis test (nonparametric analogue of ANOVA) was used followed by Dunnet or Dunn multiple comparison tests to define whether differences were statistically significant (or as indicated).Graphs were generated by using Microsoft Excel, R, or GraphPad Prism 9.2.0.Details of the statistical tests applied, including the choice of statistical method, and the exact number of "n" is indicated in the corresponding figure legend or directly on the graph.In violin plots, upper and lower dotted lines represent the first and third quantiles, respectively, horizontal lines mark the mean, and edges mark the highest and lowest values.Plots were depicted as truncated or untruncated for aesthetic reasons."N" corresponds to biological replicate and "n" to technical.

Image analyses
Image analyses and intensity measurements were done using Fiji v. 1.49 software (rsb.info.nih.gov/ij).The intensity of the fluorescence signal was measured as the Integrated Density in an ROI.PMs of individual cells were selected with the brush tool with a size of 10 px as well as the intracellular space with the polygon selection tool.The average intensity of the top 100 highest pixels for both the plasma membrane and the intracellular space was used to obtain a ratio between intracellular and PM fluorescence.The dwell time rate of tagged proteins in FRAP experiments was calculated by the single exponential fit equations as described previously [78].Colocalization was analyzed using Pearson statistics (Spearman or Manders analyses produced similar results) [79].Images were prepared by Adobe Photoshop v. 2021 (Adobe).Statistical analyses were performed with JMP v. 9 or 11 (www.jmp.com),GraphPad, or R. Curve fitting was done as we have described in detail for PCs [80].Time series movies were compressed, corrected, and exported as.avi extension files.The unspecific fluorescence decay was corrected using Fiji v. 1.49 software and default options using the bleaching correction tool.Videos were digitally enhanced with Fiji-implemented filters, correcting noise using the Gaussian blur option and pixel width set to 1.0.Mean intensities in FRAP have been normalized (1 corresponds to pre-bleach signal intensity in arbitrary units).[81], and ESP in the presence of LUVs (made from phosphatidylcholine and containing 10 mol %phospatidylserine).The beads carrying GFP-tagged ESP variants in the ±LUVs were coincubated with GST-SFH8 at 37˚C for 1 h.Note that proteolytically inactive ESP failed to cleave SFH8, while cleavage is sustained in the presence of liposomes (ESP/LUV).The produced GST-SFH8 IDR is shown (band below 50 kDa).As the observed compromised full-length SFH8 transfer from the SDS-PAGE to the membrane, the immunoreactive signal increment of the SFH8 IDR /SFH8 ratio upon cleavage was not proportional to the full-length SFH8 depletion.Note that the band approximately 50 kDa was detected in all samples and thus could correspond to a nonspecific

Fig 4 .
Fig 4. SFH8 forms PM liquid-like clusters that lack association with KISC.(A) TIRFM setting for visualization of SFH8 at the PM.The model is showing the region used for imaging, and an example TIRFM micrograph of SFH8-mNeon (lower right; SFH8pro).Scale bar, 2 μm.In TIRFM imaging, the focal plane is restricted to the outermost tissues, and, therefore, epidermis of region 1 or 2 is inaccessible (see Fig 1C for a root model showing that the epidermis in this region is encapsulated by the root cap).(B) Example of a dual-channel TIRFM of lines expressing SFH8-mNeon (SFH8pro) and KIN7.3-RFP (RPS5apro).The experiment was replicated 5 times.Scale bar, 0.3 μm.(C) Kymograph showing laterally diffusing (*) and nondiffusing (**) clusters of SFH8.Arrowheads indicate the spatial offset of the diffusing cluster (lateral displacement on the PM plane is around approximately 200 nm).The arrows (2 μm and 1 s) show the spatiotemporal resolution.(D) Quantification of SFH8 clusters (SFH8pro) in 3 different stages, fusion, fission, and "stable" (i.e., not undergoing fission or fusion).The circularity of clusters is also shown (right).Data are means ± SD (N = 3 pooled experiments, n = 4-6 fields with percentages per experiment; the p-values were calculated by 1-way ANOVA).(E) Examples of SFH8-mNeon clusters fusing on the PM.Note that similar sizes and fusion dynamics of clusters were observed with 2 promoters (SFH8pro and RPS5apro), suggesting independence of these parameters from expression levels (higher for RPS5apro).Scale bars, 0.3 μm.(F) Dualchannel TIRFM of SFH8-mNeon/KIN7.3-RFPcoexpressing line showing SFH8 clusters and the formation of filaments that do not diffuse.Note the lateral diffusion of SFH8 clusters and the lack of filaments motility (circles).The experiment was replicated 3 times.Scale bars, 0.3 μm.(G) Kymographs show clusters with low (left) and high (right) dwelling times at the PM.Right: pie graphs showing quantifications of KIN7.3 and SFH8 colocalization percentages in clusters or filaments (N = 3 pooled experiments, n = as indicated; p-values were calculated by Wilcoxon).The arrows (1 μm and 0.2 s) show the spatiotemporal resolution.Raw data can be found in the Supporting information section (S1 Data).KISC, kinesinseparase complex; PM, plasma membrane; SFH8, SEC FOURTEEN-HOMOLOG8; TIRF, total internal reflection fluorescence microscopy.https://doi.org/10.1371/journal.pbio.3002305.g004 ) in which SFH8 lacked recovery (Fig 5A).As this filamentous transition of SFH8-mNeon was also reduced in the KISC mutants (Fig 5B; >3-fold), these findings further genetically confirm that KISC mediated the conversion of SFH8 clusters to solid-like filaments.These filaments are more stably attached to the PM as they do not show recovery in FRAP (Fig 5A), which is consistent with the lack of mobility in TIRFM (see above, Fig 4), and can retain an association with the KISC (Figs 4 and

Fig) ,
suggesting that the initial clustering of SFH8 is functionally important.Furthermore, SFH8 ΔIDR only partially rescued rsw4 and the SFH6IDR SFH8 (showing little condensation) showed a moderate rescue of sfh8 (S12D and S12E Fig), highlighting the importance of SFH8 liquid-to-solid transitions.

Fig 10 .
Fig 10.Model for SFH8 functions during development.Proposed model for the role of SFH8 in root development.SFH8 (full-length) is recruited on the plasma membrane in a nonpolar manner and forms LLPS clusters at the PM (cell 1).These clusters seem to block the delivery of polar proteins to the PM (e.g., PIN2).Later, KISC is recruited on SFH8, and KISC proteolytic part (ESP) cleaves the N-terminal part of SFH8, which results in the production of SFH8 ΔIDR (SFH8 lacking the N-terminal IDR; cell 2).The released IDR floats in the cytoplasm in the form of LLPS puncta.Microtubules may direct the KISC at the PM; however, microtubules/KISC do not remain associated at the PM.Later, SFH8 ΔIDR ages to a filamentous form with clear polarity at the PM and interacts with other polar proteins, like, for example, PIN2 (cell 3).This phase transition of SFH8 and its polar localization are essential components of robust root development.The released IDR persists in the cytoplasm and has an unknown role.ESP, EXTRA SPINDLE POLES; IDR, intrinsically disordered region; KISC, kinesin-separase complex; LLPS, liquid-liquid phase separation; PM, plasma membrane; SFH8, SEC FOURTEEN-HOMOLOG8AU : AbbreviationlistshavebeencompiledforthoseusedinFigs1 À 10:Ple .