Phase Separation of a Nodulin Sec14-like protein Maintains Auxin Efflux Carrier Polarity at Arabidopsis Plasma Membranes

Abstract

Robust development can be modulated by local protein accumulation which is sometimes manifested as polarity patterns; yet, the mechanisms that lead to these patterns are largely unknown. Using the model plant Arabidopsis, we report that auxin-induced polar patterns can be reinforced by the phase transition of a SEC14-like lipid-binding protein at the plasma membrane. Using imaging, genetics, and in vitro reconstitutions, we show that the SEC14-like phase transition is promoted outside the stem cell root niche through its association with the caspase-like protease separase and conserved microtubule motors in a plasma membrane interface. Separase cleaves SEC14-like to promote its transition from liquid-like clusters to solid-like filaments. Filaments self-amplify and potentiate the robustness and maintenance of plasma membrane domains through associations with polar proteins. This work uncovers that robust cell patterns involve proteolysis-mediated phase transitions at unpreceded plasma membrane interfaces.

Synopsis This study provides an example of a protein that undergoes liquid-like phase separation at the plasma membrane in plants. Importantly, this protein known as SFH8 is a lipid-like transferase and can undergo proteolytic-induced solidification, which is an important event in the maintenance of polarized delivery of PINFORMED proteins.

• The complex comprising of the conserved caspase-like protease separase and the kinesin Kin7.3 (known as “KISC”) associates with a SEC14-like homolog, the SFH8 protein at apicobasal polar domains

• Separase cleaves SFH8 to yield a carboxy-terminal proteoform that undergoes a liquid-to-solid transition producing filamentous assemblies at the plasma membrane

• The KISC/SFH8 constitutes a feedback loop with auxin that potentiates the interaction, fusion, and maintenance of proteins with polar domains

INTRODUCTION

The few known polar proteins in plants provide crucial information for robust development (Houbaert et al., 2018; Marhava et al., 2019; Platre et al., 2018). Polar protein accumulation may depend on diffusible signals such as that of the hormone auxin. At high auxin levels, for example, the crowding of negatively charged lipids dictates plasma membrane localisations of signaling proteins (Platre et al., 2018). In turn, the polar efflux auxin carriers PINFORMED (PINs) can regulate auxin levels. For example, PIN2 in the root localizes at apical or basal plasma membrane (PM) domains and is required for the formation of the auxin concentration gradient (Friml et al., 2004). Auxin is then perceived by cells in a concentration-dependent manner as a permissive or restrictive signal (Marhava et al., 2019). Auxin signaling can be modulated by the nucleo-cytoplasmic partitioning of the prion-like transcription factors known as AUXIN RESPONSE FACTOR (ARF). During development, ARF7 and ARF19 undergo condensation in the cytoplasm which blocks their entry into the nucleus, thereby attenuating auxin signaling (Powers et al., 2019).

In Arabidopsis, the caspase-like protease separase (extra spindle poles, “ESP”) interacts with three centromeric protein-E-like Kinesin 7 (Kin7) homologs, the so-called “Kin7.3-clade” (Kin7.1, Kin7.2, and Kin7.3). This complex (the kinesin-separase complex, “KISC”) is recruited on microtubules (MTs), and the most abundant protein of the complex is Kin7.3 (Moschou et al., 2016). ESP is an evolutionarily conserved protein responsible for sister chromatid separation and membrane fusion in both plants and animals (Bembenek et al., 2010; Liu et al., 2020). ESP binds the Kin7 C-termini, inducing conformational changes that expose the MT-avid N-terminal “motor domain” of the Kin-7, thereby increasing KISC binding atop MTs. KISC modulates PIN2 delivery to the plasma membrane (PM). This is supported by the finding that the temperature-sensitive radially swollen 4 (rsw4) mutant harboring a temperature-sensitive ESP variant or the Kin7.3-clade mutants have compromised PIN2 delivery and polarity at the PM (Moschou et al., 2016). Yet how KISC acts upon PM polar domains remains elusive.

Here we discover that KISC is recruited to the PM through the hitherto uncharacterized SEC14-like lipid transfer protein SFH8. Under the premise that the KISC interacted with SFH8, we examined the biological consequences of this interaction. ESP trimmed SFH8 protein that led to the conversion of SFH8 from a liquid to a more rigid filamentous phase at the PM. This transition is associated with SFH8 polarization and interaction with PIN2. Auxin promoted SFH8 clustering, and liquidity, halting SFH8 cleavage by KISC. Remarkably, our study shows how spatiotemporally confined proteolysis can lead to changes in the material properties of proteins and how these properties can be translated to developmental outcomes.

RESULTS

KISC Associates with the Lipid-transfer protein SFH8 at the Plasma Membrane

As KISC regulates PIN2 at the PM, we aimed at surveying potential associations between the KISC and the PM. We found that fluorescently-tagged or native ESP/Kin7.3 proteins decorated PM; in the distal meristem (as defined below) both proteins decorated basal domains in the cortex and apical domains in the epidermis (Figure S1A-F). As KISC proteins lack lipid-binding motifs, we postulated that KISC associates with the PM via a protein tether. To identify this tether we screened a yeast two-hybrid (Y2H) library against Kin7.3. Among the five proteins identified, we followed At2G21520, as it localized at the PM in Nicotiana benthamiana (Figure S1G; Supplemental Info). At2G21520 is a SEC14-like protein (BLAST: p= 1e^-49), with a C-terminal “nodulin” positively-charged patch and ascribed symbol “SFH8” (Figure S1H; File S1) (Vincent et al., 2005). In Y2H, although both SFH8 termini bound Kin7.3 C-terminal tail (but not the N-terminal motor domain), the SFH8 C-terminus was more avid (Figure S1I). We confirmed the Kin7.3-SFH8 interaction by co-immunoprecipitations or glutathione S-transferase (GST) pull-down assays (Figure 1A-C and S1K for protein purifications). Furthermore, through ratiometric bi-fluorescence complementation (rBiFc) in Arabidopsis root protoplasts, we showed that SFH8/Kin7.3 associate mainly at the PM (Figure 1D; see also Figure S6D).

Figure S1. KISC/SFH8 Polarity localization in live-cell imaging.

A. Apical/basal/lateral polarization nomenclature and polarity index determination.

B. Representative confocal micrographs of α-ESP/α-Kin7.3 localizations at similar polar domains (counterstained with α-β-tubulin; region 4, experiment replicated multiple times). Scale bars, 5 μm.

C. Representative confocal micrograph of the ESP estradiol inducible line (20 μM estradiol, 16 h), driven by the Kin7.3 promoter (pKin7.3>XVE>GFP-ESP). ESP showed polarization only at the distal meristematic region, while in the proximal meristem region ESP was apolar; Images were obtained in low resolution (512x512 with minimal exposure and no averaging), due to photobleaching and low expression levels of ESP (experiment replicated multiple times with variable expression). Scale bar, 50 μm.

D. Digitally zoomed-in enhanced micrographs (deconvoluted) from root tip cells of the pKin7.3>XVE>GFP-ESP. Merist., proximal meristematic cell (note the diffused cytoplasmic signal). Cap, lateral root cap. A cell plate decorated by ESP is also shown (bottom right). Note the MT-binding of ESP in the lateral root cap cell at which MTs are highly bundled (as also described in Moschou et al., 2013; 2016). Scale bars, 5 μm.

E. Representative confocal micrographs of 5 DAG seedlings expressing pKin7.3:Kin7.3-mNeon in the K135 background at the indicated regions. Right: representative confocal micrographs of epidermal cells of root tip 5 DAG from seedlings expressing pRPS5a:Kin7.3-tagRFP/K135 (complementation line). CP, cell plate; PH, phragmoplast. Scale bars, 20 μm.

F. Complementation of K135 root phenotype by various Kin7.3 fusions: pKin7.3:GFP-Kin7.3, pRPS5a:Kin7.3-tagRFP, and pRPS5a:Kin7.3-mNeon; quantification of root length (right). Data are means ± SD of five independent experiments each containing five measurements (2-tailed t-test).

G. Y2H between binding domain (BD)-Kin7.3 and four putative interactors (PP2CA1, DNAJ, SPC25, and PORcino; see also Supplemental Information for a description of the Kin7.3 interactors which were not followed herein). AD, activation domain; DDO, double dropout (control); QDO, quadruple dropout (selection). Negative controls: p53 and lamin (Lam). Y2H was replicated four times. Right: representative confocal micrograph (maximum intensity projection) showing p35S:SFH8-GFP localization in N. benthamiana leaves. Scale bar, 20 μm. The micrograph is from a single representative experiment replicated multiple times.

H. Proposed model of SFH8 architecture (up). Numbers indicate aa truncations used throughout the paper; SD1 corresponds to amino acid residues 1-478. IDR, intrinsically disordered region (1-96 aa; see also Figure 4D). The CRAL-TRIO domain binds small lipophilic molecules and is named after the cellular retinaldehyde-binding protein and TRIO guanine exchange factor; phylogenetic analysis of At2G21520 full-length sequence (lower). Numbers indicate bootstrap values with a confidence cutoff of 70 (STAR METHODS).

I. Y2H between Kin7.3 tail or Kin7.3 motor fused to the binding domain (BD) and different SFH8 truncations fused to the activation domain (AD). DDO, double dropout (growth control); QDO, quadruple dropout (selection medium for interaction). In QDO, the second column (dilution 1:100) corresponds to QDO supplemented with X-a-Gal that is metabolized by yeast cells to produce blue color when interactions between AD-BD fusions are strong (supplemented with aureobasidin, “AU” to further enhance selection). As a positive control, the cohesin loader complex components Mau2/Scc2 were used (Minina et al., 2017) (denoted with the plus symbol). The experiment was replicated four times. The spots from X-a-Gal/AU are from different plates.

J. Yeast temperature-sensitive sec14-1^ts loss-of-function mutant complementation by full-length SFH8 or SD1 (see H). Note that at 37°C the complementation was moderate, likely due to the instability of the protein at elevated temperatures (37°C is much beyond the physiological Arabidopsis growth temperature). Dilution series:0-10^-4. The experiment was replicated three times.

K. Representative denaturing SDS-PAGE gels counterstained with coomassie Brillant blue (CBB) and fixed showing various proteins used in the assays described in this paper. h, 6xhis-tag; G, GST tag. The images are representative, while variations in protein integrity were evident among different purifications for SFH8 (e.g. see Figure S4; overall, SFH8 is a very unstable protein).

In budding yeast, Sec14p localizes at the trans-Golgi network (TGN) where it promotes vesicle formation presumably by stimulating phosphatidylinositol-4-phosphate (PI4P) synthesis (Schaaf et al., 2008). SFH8 rescued the temperature-sensitive budding yeast sec14-1 mutant suggesting that SFH8 is a genuine Sec14-like protein (Figure S1J). SFH8 showed broad phosphatidylinositol (PI) and unexpected phosphatidylserine (PS) binding (Figure 1E), unlike other Sec14-like proteins which cannot bind PS (Hertle et al., 2020; Sugiura et al., 2019).

Localization and functions of SFH proteins are largely unknown, while genetic evidence suggests that SFH1 is essential for root hair development (Vincent et al., 2005). We expressed SFH8-mNeon under its native promoter (pSFH8) to address SFH8 localization in 4 developmental root regions: core meristem (1; stem cell niche), meristematic zone (2; “proximal meristem”), meristematic/transition zone (3), and late transition zone (“distal meristem”; Figure 1F) (4). SFH8-mNeon showed signals in all root cells at the PM, decorating apical PM domains in the epidermis and basal domains in the cortex/vasculature in distal meristem cells (Figure 1G). SFH8 associated with the cell plate but was excluded from the leading edge, much like the KISC proteins (Moschou et al., 2016; Moschou et al., 2013); SFH8 colocalized at the PM with Kin7.3 in filamentous-like patches, and colocalization increased during development (Figure 1H, and below). Furthermore, SFH8 and KISC proteins attained polarity at regions 3-4 (Figures 1G, I, J, S1E, and below). The KISC co-localization/interactions with SFH8 prompted us to examine possible biological consequences of this interaction.

SFH8 recruits KISC at the PM

Next, we examined a putative mutual targeting potentiation of SFH8/KISC at the PM. We thus isolated two T (transfer)-DNA insertions in SFH8, the sfh8-1, and sfh8-2; RNAseq showed that sfh8-1 harbored a truncated sfh8 mRNA (Figure 2A). Both mutants had defects in root growth ((Figures 2A) and gravitropism (Figure S2A) at early stages implying compromised auxin-signaling, with a shorter stature, decreased branching, smaller cotyledons, and leaves, Figure S2B). At the seedling stage, the sfh8 phenotypes resembled those of the KISC mutants (i.e. rsw4 and the partial loss-of-function Kin7.3-clade mutant kin7.1kin7.3kin7.5 (“K135”); with the readouts used sfh8 phenotypes were evident 3 days post-germination (Figure 2B-D). We did not observe additive phenotypes at early developmental stages in KISC/SFH8 mutant combinations, implying some functional convergence between KISC/SFH8 (Figure 2B). Yet, sfh8-1 had reduced silique length and a more severe adult phenotype than the one reported for the KISC mutants (Figure S2A-C). Due to the phenotypic similarity between sfh8-1 and -2, we carried out further analyses with sfh8-1 (hereafter “sfh8”).

Figure S2. Isolation and characterization of sfh8 mutants and phenotypic analyses of KISC/sfh8.

A. Adult plant phenotype (upper, 21 DAG, 4 plants per pot) and representative photos of WT, sfh8-1, and sfh8-2, and pSFH8:SFH8-mNeon/sfh8 lines at 10 DAG (lower left, “SFH8/sfh8”). Note also the reduced expansion of cotyledons (see also Figure 2). Hereafter, sfh8-1 is denoted as “sfh8”, and due to phenotypical resemblance with sfh8-2 allele is used throughout the paper. Bottom: circular plots showing moderate gravity perception defects of KISC, and sfh8 mutants grown vertically on plates at 7 DAG. Data are from a single representative experiment replicated multiple times.

B. SFH8 mutation affects auxin-related phenotypes. Quantification of branch number (three independent experiments; n=12, ordinary 1-way ANOVA).

C. Representative images of adult WT and sfh8 plants in short-day (10h light /14h dark) condition (upper), and silique lengths (lower). Length bar, 0.8 cm.

D. 5 DAG-old meristems from WT, sfh8, and SFH8/sfh8 (sizes were 210±12 for WT, 122±22 for SFH8, and 213±17 for SFH8/sfh8; mean meristem size±SD (n=10; stained with propidium iodide). Scale bars, 20 μm. The arrowhead in the inset denotes slight cell division plane orientation perturbations in sfh8. In general, 1-3 perturbations per root were observed. Similar phenotypes were observed for rsw4 but not K135 which is a hypomorphic mutant (Moschou et al., 2016). Inset scale bars, 20 μm.

E. LC-MS/MS lipid species quantitative analysis in WT, sfh8, K135, and SFH8/sfh8 complementation line (pSFH8:mNeon-SFH8/sfh8) and rsw4 mutant (treated for 24h at 28°C). MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; SQDG, sulfoquinovosyl diacylglycerol; PA, phosphatidic acid; PS, phosphatidyl-serine; PI, phosphatidyl-inositols; PG, phosphatidylglycerol; PE, phosphatidylethanolamine; PC, phosphatidylcholine. Data are means±SD of five independent experiments (n=5, significance was tested with various tests; see also File S3 for datasets).

F. Representative images of seedlings 10 DAG expressing various fusions of SFH8 in the sfh8 mutant background. The representative confocal micrographs show localization of mNeon-cSFH8/sfh8 (“c” for “complementary DNA”), or KN/KC, RN/RC (5 DAG, region 3 epidermal cells). Scale bar, 3.5 μm.

Figure 1. KISC and the lipid-binding protein SFH8 associate at the PM.

A. Kin7.3/SFH8 co-immunoprecipitation from N. benthamiana. Free-GFP was also present in the GFP-Kin7.3 samples, due to non-specific cleavage. WBs were replicated three times.

B. Kin7.3/SFH8 co-immunoprecipitation from Arabidopsis (5 days after germination (DAG)); quantification of the interaction (chart; n=4, 2-tailed t-test).

C. GST pull-down assays of SFH8 truncated forms with 6xhis-GFP-Kin7.3 (S1K: protein purification). WBs were replicated three times.

D. rBiFC assays showing Kin7.3/SFH8 interaction at the PM in Arabidopsis root protoplasts. Controls: Kin7.3-nYFP with ESP-DomA or -DomC (Moschou et al., 2016). Mean YFP/RFP signal ratios±SD (n=20) indicated on images. Scale bars, 6 μm.

E. GST-SFH8 lipid binding specificity on protein-lipid blot overlay. A representative blot is shown replicated twice.

F. Root micrograph showing the “4 root regions” examined herein. Vertical arrow: proximodistal axis.

G. SFH8-mNeon tissue-specific expression and subcellular localization. Numbers denote root regions according to the root model in F. White and yellow arrowheads: mature and expanding cell plates, respectively. Plasmolysis experiment: confirmation of SFH8 signal exclusion from the cell wall (white arrowhead; root region 2). Scale bars, 20 μm (5 μm for cell plate detail counterstained with FM4-64). Note the lack of SFH8 signal from the leading edge (denoted by the vertical yellow line.

H. Kin7.3/SFH8 localization at regions 2-4. Scale bars, 10 μm. Right top: high-resolution signal of Kin7.3/SFH8 at the PM (regions 2 and 4). For region 4, plot profile of signal intensity across a straight line of 0.87 μm (lower). The colocalization analyses using Pearson Correlation Coefficient (PCC) is also shown at the indicated regions of interest (ROIs; rectangular of ∼0.3 μm). The overall PCC values for regions 2 and 4 are also shown (ROIs: whole image). AUs, arbitrary units (five independent experiments from 3 adjacent cells±SD; n=15).

I. Example of α-ESP/α-Kin7.3 colocalization and polarization (counterstained with α-β-tubulin; region 3). Scale bars, 5 μm.

J. Quantification of SFH8 (data derived from G), Kin7.3, and ESP polarity in the 4 regions (values >1 represent polarized proteins and calculations are described in Figure S1A; five independent experiments; n≥25, Dunnett).

Figure 2. SFH8 modulates early development cooperatively with KISC.

A. T-DNA insertion sites for sfh8-1 and -2, RNAseq IGV representation of deleted exons in sfh8-1, and phenotypes of young WT, sfh8-1, and complementation lines (10 DAG on plates or one- month-old).

B.-D. Root length of various mutants (7 DAG) (B), and kinematic root growth quantification using “SPIRO” (see STAR METHODS; 3 DAG=time 0; at least five measurements (n≥25; 1-sided Dunnett); C), or in a time-frame of days (D). Asterisks indicate statistical significance (*, 0.05; ***, <0.001; ns, no significant: WT vs. sfh8).

E. Root meristem micrographs (5 DAG; propidium iodide staining); quantifications of WT, sfh8, or KISC mutants meristem sizes (n=19; ordinary ANOVA). Scale bars, 50 μm.

F. Micrograph of the “Cycle Tracking in Plants” marker (CyTRAP) in WT and sfh8 (7 DAG); quantification of S and M phase (three independent experiments, n=5 in each; paired t-test). Scale bars, 50 μm.

G. GFP-Kin7.3 PM localization in WT or sfh8 (5 DAG, region 3). The polarity index of Kin7.3 is also shown on the images (three pooled experiments from region 3±SD; n=40; differences were significant at p<0.0001, Dunnett). Scale bars, 5 μm; quantification of cytoplasmic to PM signal (right; a single representative experiment replicated multiple times; n=25; 2-tailed t-test). Arrowheads in the insets show apical or lateral localization of Kin7.3.

H. SFH8-mNeon localization in WT, rsw4, and K135 (5 DAG, region 3). Images are representative of an experiment replicated multiple times for polarity. The polarity index of SFH8 is also shown (three pooled experiments from region 3±SD; n=40; “*”: p<0.0001 to WT, Dunnett). Scale bars, 5 μm.

I. Perturbed gravitropism and growth of lines over-expressing transiently Kin7.3 tail (“t”; 24 h, 2 μM estradiol). Scale bars, 8 μm. Right: SFH8 polarity loss in Kin7.3t lines (region 3, epidermal cells). Images are representative of an experiment replicated multiple times for polarity. Scale bars, 10 μm. SFH8 polarity index is also shown on the images (three pooled experiments from region 3±SD; n=40; “*”: p<0.0001 to WT, Dunnett).

We ascertained the link between sfh8 phenotype with SFH8, as various N- and C-terminally SFH8 fusions (with various tags or fluorophores; showing similar localizations, see below) driven by several promoters (pSFH8, pKin7.3, and the meristematic pRPS5a; Figures 2A-E, S2A, B, D, E, F) could efficiently complement sfh8. In sfh8, the apical meristem was reduced (∼2-fold) showing lower mitotic activity (defined by “Cell Cycle Tracking” (Yin et al., 2014) and occasionally perturbed cell division planes (Figure 2F, S2D). Although SFH8 is a Sec14-like protein its deletion (or of KISC) did not affect lipid levels (Figure S2E, File S4), suggesting that the observed phenotypes did not relate to perturbations in lipid homeostasis.

When expressed in the sfh8 background, GFP-Kin7.3 showed both reduced PM localization and polarity (Figure 2G). While SFH8-mNeon tethering at the PM did not substantially depend on KISC, SFH8-mNeon in the corresponding mutants had reduced polarity (Figure 2H). We corroborated these results through transient KISC depletion through Kin7.3 C-terminal “tail” overexpression (pXVE>pKin7.3>HA-Kin7.3tail). This inducible Kin7.3 tail overaccumulation disrupts KISC, as it titrates ESP out of the active complex (Moschou et al., 2016). Consistently, Kin7.3 tail accumulation but not the full-length led to SFH8 polarity loss (Figure 2I). Hence, KISC/SFH8 synergistically define their localization: SFH8 tethers KISC at the PM, and in turn, KISC promotes SFH8 polarization.

SFH8 Forms PM Liquid-like Clusters that Restrict Association with KISC

As KISC binds on MTs, we postulated that SFH8/KISC interact atop MT filaments next to the PM. Contrary to our expectations, the standard MT marker MAP4^MBD (or β-tubulin) showed very partial colocalization with Kin7.3 at the PM, while amiprophos-methyl (APM) that disassembles MTs (10 nM; Moschou et al., 2016) did not alter PM localization of Kin7.3 (Figure S3A, C). We observed that KISC/SFH8 decorated the same filaments in N. benthamiana, while KISC proteins were excluded from SFH8-positive clusters structures at the PM (Figure S3B). In Arabidopsis roots, SFH8 filaments were short (<0.5 μm) and insensitive to APM (Figure S3C). Likewise, ESP and SFH8 decorated similar filaments (Figure S3D; root region 3). Hence, KISC/SFH8 co-assemble in MT-independent filaments.

Figure S3. SFH8 clusters exclude the KISC.

A. Kin7.3 partially colocalized with MTs and is excluded from PM MT-bundles. Representative confocal images of pKin7.3:GFP-Kin7.3/p35S:MAP4^MBD expressing lines (5 DAG, region 3 epidermal cells) counter-stained with FM4-64, ±APM (10 nM, 1 h; left). Scale bars, 12 μm. Right: representative confocal images of GFP-Kin7.3 at cell contours (5 DAG, region 3 epidermal cells), and cell contour localization of pKin7.3:GFP-Kin7.3/p35S:MAP4^MBD. The grey box at the top right shows the Pearson Correlation Coefficient (PCC) at the indicated with the arrowheads regions, which in the case of MAP4^MBD denote PM-attached MT bundles. Scale bar, 1 μm.

B. KISC is restricted from SFH8 clusters but colocalizes with SFH8 filaments. Representative confocal micrograph of N. benthamiana leaves transiently co-expressing (4 days post-infiltration) p35S:CFP-Kin7.3 (cyan)/p35S:mScarlet-SFH8/p35S:GFP-ESP. The experiment was replicated multiple times. Scale bars, 10 μm.

C. KISC filaments do not show significant signal colinearity with MTs. Representative confocal micrograph showing the partial colocalization of GFP-Kin7.3 filaments with MAP4^MBD in region 3 (defined in Figure 1) ±APM (10 nM, 1 h). Note that Kin7.3 filaments associated with bundled-MT remnants (reminiscent of clusters; insets, APM treatment). Lower: SFH8 filaments and their resistance upon APM treatment (10 nM, 1 h; region 3); colocalization between MAP4^MBD/α-tubulin signals. Note that MAP4 protein induces significant MT bundling. Scale bars, 0.1 μm (for filaments) or 3.4 μm for colocalization (MAP4/tubulin). Right: quantification of Kin7.3/SFH8 filaments’ length at the PM (six pooled independent experiments; n=40, Wilcoxon). Note the high density of SFH8 length counts around the mean, suggesting that SFH8 filaments are mostly short compared to MT filaments (note the MT length distribution) with little variability. Kin7.3 showed more dispersion, as shown by the violin plot (and the micrograph on the left), indicative of Kin7.3 association with MT filaments/bundles, as well. Scale bars, 1 μm.

D. ESP is also restricted from SFH8 clusters in Arabidopsis roots. Representative confocal micrographs of the line co-expressing pKin7.3>XVE>GFP-ESP/pRPS5a:mScarlet-SFH8, after estradiol induction (20 μM, 16-24 h; 5 DAG, region 2 epidermal cells, PM surface at the apicobasal junction). Images were obtained at maximum scanning speed (40 frames per second) using the LIGHTNING super-resolution module of the SP8 Leica confocal microscope (63x W with an x/y-axial 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. Similar settings were used in other images as well. Scale bar, 10 μm.

E. The nodulin motif modulates SFH8 clustering propensity most likely through the attenuation of repulsion from lysine (K) residues. Upper: representative confocal micrographs showing the propensity to form clusters of N-terminally RFP-tagged SFH8 variants (SFH8^6KtoA and SFH8^2KtoA, under the p35S) in N. benthamiana transient expression system. Insets (right) show details of clusters. The experiment was replicated two times. Scale bars, 20 μm. Right: representative confocal micrographs showing colocalization analyses of RFP-SFH8, -SFH8^2KtoA, and -SFH8^6KtoA with GFP-Kin7.3 (p35S:GFP-Kin7.3) in N. benthamiana transient expression system (3 days post-infiltration). Note that in cells expressing SFH8^6KtoA, GFP-Kin7.3 retained a cytoplasmic localization (insets and details on the right), and also the lack of GFP-Kin7.3 colocalization with RFP-SFH8^6KtoA clusters. Scale bars, 50 μm. The experiment was replicated 2 times.

F. Kin7.3 tail fails to interact efficiently with SFH8^6KtoA. Y2H between SFH8, SFH8^2KtoA, or SFH8^6KtoA with Kin7.3 tail. Kin7.3 tail was fused to the binding domain (BD) and the SFH8 variants to the activation domain (AD). DDO, double dropout (growth control); QDO, quadruple dropout (selection medium for interaction). Two dilution series are shown (0 and 0.1). The experiment was replicated three times.

G. Note the partial complementation of the sfh8 mutant background by the SFH8^6KtoA mutant variant. Images showing partial complementation of sfh8 by pRPS5a:mNeon-SFH8^6KtoA in 4 individually transformed lines (10 DAG). Two complementation lines with SFH8 are also shown.

H. Representative confocal micrographs showing the reduced levels of SFH8^6KtoA and lack of polarity of mNeon-SFH8^6KtoA (green; pRPS5a:mNeon-SFH8^6KtoA/sfh8; epidermal cells 5 DAG, region 3) counterstained with the dye FM4-64 (magenta; PM and endosomes staining) from line #2 shown in G. Scale bar, 25 μm. Right: quantification of root growth from WT, sfh8, and four lines (presented in G) expressing pRPS5a:mNeon-SFH8^6KtoA in the sfh8 background (two independent experiments; n=30, ordinary 1-way ANOVA). Note that the cytoplasmic puncta observed correspond to cellular aggregates of SFH8^6KtoA as they show different properties and sizes; in many cases, these puncta were found attached on the PM and were not motile (see Figure S4).

We further followed KISC/SFH8 localization in Arabidopsis roots by total internal reflection fluorescence (TIRF)-Microscopy. TIRFM excels at looking at cell surface processes due to the shallow illumination penetration (decay constant ∼100 nm), offering a high signal-to-noise ratio. We focused on lateral cell junction domains (Figure 3A; regions 3-4; 3 days post-germination). Our analysis showed that SFH8-mNeon segregated into two major populations: (i) immobile filaments that co-localized with Kin7.3-tagRFP (see Figure S1F for fusion functionality), and (ii) mobile or immobile Kin7.3-tagRFP-independent clusters (Figure 3B). The mobile SFH8 clusters showed restricted lateral diffusion at the PM and occasionally fused, properties reminiscent of cellular “biomolecular condensates” that sometimes form through a process known as liquid-liquid-phase separation (LLPS) (Movies S1-3, Figure 3C, D). Under certain conditions, proteins or other molecules can demix from the bulk phase through LLPS, thereby retaining liquid-like properties such as surface tension leading to high circularity akin to droplets (Milovanovic et al., 2018). We also observed some non-diffusing less droplet-like small clusters (Movies S2, 3), that could represent intermediate phases between the cluster-state (droplet-like) and the filamentous state.

Figure 3. SFH8 polarization depends on a droplet-to-filamentous transition.

A. Model showing the region used for imaging in TIRFM, and an SFH8 localization overview (5 DAG, region 3). Scale bar, 2 μm.

B. Dual-channel TIRFM of SFH8-mNeon/Kin7.3-RFP. Scale bar, 0.3 μm.

C. Kymograph showing diffusing (*) and non-diffusing (**) clusters. Arrowheads indicate the offset of the diffusing cluster (∼200 nm).

D. Examples of SFH8-mNeon clusters fusing on the PM. Scale bar, 0.3 μm.

E. Dual-channel TIRFM of SFH8-mNeon/Kin7.3-RFP co-expressing line showing SFH8 clusters and filaments. Note the lateral diffusion of SFH8 clusters and the lack of filaments motility (circles). Scale bars, 0.3 μm.

F. Kymographs show diffusing clusters with low and high dwelling times at the PM. Right: charts show Kin7.3/SFH8 colocalization percentage in clusters or filaments (three pooled experiments; 5 fields of view, n, as indicated; Wilcoxon).

G. SFH8-mNeon immobilization along the proximodistal meristem root axis determined by FRAP. Data are means ± SD from a single representative experiment with 10 roots at each point (n=10). CPi, immature cell plate; CPm, mature cell plate. Bottom: clusters and filaments in two root regions determined by super-resolution (120 nm, 40 frame per second) midsection confocal microscopy (regions 1 and 4, left and right, respectively) and a model showing the SFH8 clusters-to-filaments conversion.

H. SFH8-mNeon localization in WT, K135, and rsw4 root cells (5 DAG, region 3, TIRFM; arrowheads indicate filaments). Scale bar, 0.5 μm; quantification of SFH8-mNeon clusters and filaments in WT, K135, and rsw4 (right; n=28, 5 fields of view; Wilcoxon) and a model showing the dependency of SFH8 clusters-to-filaments conversion on KISC.

Kin7.3/SFH8 interacted prominently in short filaments but not in SFH8 clusters, which showed high or low dwelling times at the PM (Figure 3E). The lack of KISC-SFH8 association in clusters in which SFH8 presumably should show increased avidity for KISC due to the increased concentration of SFH8 was counterintuitive. We thus set out to determine why KISC was excluded from the SFH8 clusters. We assumed that reducing the repulsion of the polybasic charge of nodulin could promote the clustered state of SFH8, further increasing the concentration of the protein. In support of our assumption, charge attenuation by mutations of the six pertinent lysines (K) to alanine (A) in the nodulin artificially increased SFH8 clustering at the PM in N. benthamiana, while a mutation in two K exerted a moderate effect (Figure S3E). In good agreement, Kin7.3 interacted and colocalized with the WT variant but not with SFH8^6KtoA (Figure S3F). In Arabidopsis, mNeon-SFH8^6KtoA showed a lack of PM polarity and filaments (Figure S3G, H). This result implied that hindering interaction with KISC by reducing the accessibility of the C-terminus which is the preferred interaction interface with KISC, completely blocked SFH8 polarization.

To better understand the dynamics of SFH8 at the PM, we used fluorescence recovery after photobleaching (FRAP). We assumed that SFH8 clusters have increased liquidity and reduced stability at the PM, compared to filaments, as they showed features of LLPS which could imply a rapid exchange of SFH8 molecules with the cytoplasm. Although FRAP is not a decisive feature for testing LLPS, liquid-like clusters can display rapid internal reorganization and a corresponding rapid fluorescent recovery on the order of seconds, whereas more solid-like condensates such as non-dynamic filaments may show little recovery (Gutierrez-Beltran et al., 2015). FRAP showed that SFH8-mNeon was highly dynamic at the PM close to the meristem, unlike the distal meristem in which SFH8 lacked dynamicity (Figure 3G). Using super-resolution dynamic imaging, we observed that at stages at which KISC showed little colocalization with SFH8, SFH8 was in the clustered state while later was converted to filaments (e.g. region 2 to 4; Figure 3G). As the shift in SFH8 FRAP rate associated well with the cluster-to-filament transition, this suggested a role for KISC in this transition. In accordance, filamentous formation of SFH8-mNeon was significantly reduced in the KISC mutants (Figure 3H; >3-fold). Altogether, these findings suggest that condensed clusters of SFH8 abolish KISC-SFH8 interactions, but later during development KISC mediates the conversion of SFH8 clusters to solid-like filaments.

KISC Trimming of SFH8 Promotes the Cluster-to-Filamentous Transition

The SFH8 filamentous transition dependency on KISC prompted us to examine the underlying mechanism. We postulated that KISC may promote structural changes on SFH8 that enable the filamentous transition. Interestingly, western blots from lines expressing pRPS5a-driven SFH8 with a hexahistidine-triple-flag (referred to as “HF”)-mScarlet tag at the C or N terminus (∼106 kDa; complementation in Figure S1F) showed a decreased abundance of tagged-SFH8 with age (3-7 days) and increased SFH8 full-length abundance in K135 or rsw4 (Figure 4A). We also observed a presumptive ∼40 kDa (or ∼10 kDa excluding mScarlet) N-terminal cleavage product in WT, which was almost absent in K135 or rsw4 (Figure 4A). We confirmed that SFH8 is cleaved using N- and C-terminally HF-tagged SFH8 lines (Figure 4A; right panel). We also observed that pSFH8:mNeon-SFH8 lines had fluorescent cytoplasmic puncta that accumulated gradually during development, which were absent from pSFH8:SFH8-mNeon lines (Figure 4B). This result suggested that cleaved SFH8 N-termini coalesce in puncta, and implied a link between KISC-SFH8 interaction and their formation. Puncta were also observed in lines expressing SFH8 cDNA (complementing sfh8, Figure S2F), confirming that SFH8 fragments are produced by cleavage and not transcriptional mechanisms (e.g. splicing).

Figure 4. ESP cleaves SFH8 to promote SFH8 droplet-to-filamentous transition.

A. WB showing the reduction of SFH8 full-length levels from 3 to 7 DAG (upper left). Lower left: cleavage and release of an N-terminal fragment (black arrowhead at ∼40 kDa; FL, full-length) in WT, K135, or rsw4 (24 h at the restrictive temperature 28°C, two lines; lower). Right: WB showing the remaining C-terminal SFH8 fragment (asterisk, an additional truncated product of low abundance). WBs are representative of experiments replicated multiple times.

B. Micrographs of the SFH8 N- or C-terminally tagged with mNeon (3 regions, epidermis; 4^th region slightly below the transition zone in this case). Scale bars, 4 μm.

C. SFH8 cleavage biosensor expression (scale bar, 50 μm), details (right, mid-plane focus, scale bars, 4 μm), and relative signal intensity of cytoplasmic versus PM signal (chart). Data are means ± SD (n=10). Bottom: super-resolution imaging (120 nm) of cluster-to-filament conversion (regions 2 to 4). Note the absence of mScarlet signal from filaments.

D. SFH8 protein architecture (IDR, intrinsically disordered region, CRAL-TRIO, and NOD: nodulin motif; see also Figure S1H for annotations).

E. SFH8 IDR peptides identified in pRPS5a:SFH8-mScarlet-HF pull-down experiments coupled with LC-MS/MS. The cleavage motif of ESP on SFH proteins, φEXXR is conserved (presented here for three SFH protein paralogs, SFH6/8/9; upper left).

F. In silico predictions of IDRs by PONDR, and phase separation propensity determined by catGRANULE (Vernon and Forman-Kay, 2019) for SFH8 and SFH6. Note the shorter IDR length and lower LLPS propensity score for SFH6.

G. IDR defines the phase separation properties of SFH8. Reduced puncta formation in a chimeric protein of SFH6^IDR and the C-terminal SFH8 in N. benthamiana leaves, in the presence of ESP/CyclinD; see also Figure S4). Scale bars, 20 μm.

H. Similar experiment as in G, in the absence of ESP. Note the increased clustering of SFH8 and the lack of clusters in an ^SFH6IDRSFH8 chimera. Scale bars, 2 μm.

I. Lack of SFH8^R84A mutant cleavage in N. benthamiana leaves in the presence of ESP/CyclinD. Scale bars, 20 μm.

J. Comparative analysis of the SFH8 IDR composition. Each amino acid is assigned to one of six groups on the x-axis, and the fraction of grouped amino acids is shown. For comparison, model “condensators” are shown (ARF19 to FUS). SFH8 IDR is similar to the average IDR signature of disordered proteins in Arabidopsis (Powers et al., 2019). The lengths of the IDRs were determined by the fIDPnn (Hu et al., 2021).

K. Dependency of KISC for SFH8 polarity and cleavage. Reduced puncta formation and increased mNeon-SFH8 PM signal in the K135 or sfh8 (7 DAG, region 3). Scale bars, 5 μm. Lower: puncta quantification (single representative experiment; n=27 cells from region 3; ANOVA). Polarity indexes of SFH8-mNeon are also shown (“***”: p<0.0001 to WT, Dunnett).

L. Reduction of sfh8 complementation by mNeon-SFH8^R84A. Three-individual lines are shown.

M. Localization of SFH8^R84A. Scale bars, 2 μm. Right: persistence of PM SFH8 condensates in SFH8^R84A/sfh8 lines (TIRFM; scale bar, 0.2 μm). Bottom: quantification of mNeon-SFH8 or mNeon-SFH8^R84A clusters and filaments (four independent experiments; n as indicated, 2-tailed t-test).

Figure S4. KISC produces SFH8^IDR which does not colocalize with membrane markers, and phase separates.

A. The mitotic cohesin subunit SYN4 is cleaved by ESP in vivo in root cells. Representative images of root cells expressing pSYN4:CFP-SYN4-YFP in WT or rsw4 (at the restrictive temperature 28°C for 72 h, 5 -7 DAG). Insets show chromosomal bridges in rsw4 due to the presumptive lack of the mitotic SYN4 kleisin subunit cleavage by ESP (epidermal cells); the white arrowhead denotes a chromosomal bridge. The experiment was replicated multiple times (with 1-3 bridges evident per root). Scale bars, 40 μm. Under our experimental settings, CFP did not yield a robust signal.

B. SYN4 is cleaved by ESP in vitro confirming that purified ESP is active. Right: in vitro cleavage of SYN4 on beads (p35S:SYN4-myc) purified from N. benthamiana, in the presence of immunopurified from N. benthamiana aTAP-ESP immobilized on beads (aTAP carries c-myc and hexahistidine tags), mitotically pre-activated by cyclinD (see main text and STAR METHODS for details on purification). Incubation of the cleavage assay was done for 1 h at 37°C. The experiment was replicated three times. Right: increasing incubation time (2-4 h) led to the formation of additional cleavage fragments and smearing of SYN4-myc, suggesting that ESP may cleave SYN4 at multiple sites with differential preference. Asterisks indicate non-specific bands. The experiments were replicated twice.

C. Proteolytically inactive ESP fails to cleave SFH8, and cleavage is supported in the presence of membranes. In vitro GST-SFH8 cleavage by immunopurified from N. benthamiana GFP-tagged variants of ESP, ESP^PD (protease dead (Liu and Moschou, 2018), and ESP in the presence of large unilamellar vesicles (LUVs; made from PC and containing 10 mol %PC and PS; see STAR METHODS). The cleaved GST-SFH8^IDR is shown (band below 50 kDa). The beads carrying GFP or GFP-tagged ESP variants in the ±LUVs were co-incubated with GST-SFH8 at 37°C for 1 h. As the SFH8^ΔIDR was converted to filaments (see below) and due to the observed compromised full-length SFH8 transfer from the SDS-PAGE to the membrane, the immunoreactive signal increment of the SFH8^ΔIDR/SFH8 (α-GST) ratio upon cleavage was not proportional to the SFH8^IDR production. The experiment was replicated four times. Note that the band around 50 kDa was detected in all samples and thus could correspond to a non-specific cleavage product in the GFP-trap setting.

D. Confirmation of ESP cleavage site on SFH8. GST-SFH8^R84A is not cleaved by aTAP-ESP (similar experimental setting as in B). An additional product is due to non-specific cleavage during purification (above the specific cleavage product around 50kDa). The experiment was replicated three times.

E. The IDR region of SFH8 is biologically important. Representative confocal micrographs from meristematic epidermal root cells expressing pRPS5a:mNeon-SFH8^ΔIDR in the sfh8 mutant background, showing details of localization with lack of filaments, polarization, and cytoplasmic puncta formation. Note also the reduced robustness of PM localization, uneven SFH8 levels, and perturbations of growth realized as reduced growth anisotropy (details; middle right cell denoted by arrowhead). The Scale bar on the top, 40 μm and all others, 20 μm. Right: quantification of root length of three lines expressing pRPS5a:mNeon-SFH8^ΔIDR in the sfh8 background (two independent pooled experiments; n=18; ordinary 1-way ANOVA).

F. SFH8 likely does not colocalize with diffraction non-limited endomembranes and vesicles. Representative confocal micrographs from pRPS5a:mNeon-SFH8 expressing lines counterstained with FM4-64 (endosomes and plasma membrane) or mitotracker (mitochondria) dyes; lines co-expressing pRPS5a:mNeon-SFH8 with p35S:TIP-RFP (tonoplast intrinsic protein), or p35S:SNX1-RFP (sorting nexin 1; trans-Golgi network). In some cases, the duration of tracking was limited to the amount of time that particles remained in a single focal plane, as the required acquisition rate did not permit the collection of z-stacks (voxels); the median track length was 35 frames, corresponding to 20’’ of imaging. PCC analyses failed to show colocalization of SFH8 cytoplasmic puncta with any of the markers/dyes used (negative values). Scale bars, 2 μm.

G. SFH8 although it can bind to PtdIns(3)P (see also Figure 1E) it does not colocalize with relevant PtdIns(3)P-positive structures in vivo. Representative confocal micrograph of lines co-expressing mScarlet-SFH8-mNeon (green; pRPS5a:6xhis-3xFLAG(HF)-mScarlet-SFH8-mNeon; see also Figure 4C) with PtdIns(3)P (tagged with CFP and expressed under pUBI10). The inset on the right is digital magnification, showing a lack of colocalization between PtdIns(3)P and mScarlet-SFH8^IDR. Right: mNeon-SFH8 does not associate with cellulose synthase complex-puncta. Representative confocal micrograph of lines coexpressing pRPS5a:mNeon-SFH8 with Tdt-CESA6 (CELLULOSE SYNTHASE 6 isomer under the p35S). Note the lack of colocalization between the two proteins. Scale bar, 4 μm. The experiments were replicated three times for various developmental stages (2-7 DAG), and root tissues (regions 1-4).

H. SFH8 cytoplasmic puncta show features of LLPS proteins. Upper: representative confocal micrographs showing the FRAP signal recovery (normalized integrated density) as a fraction of time in lines expressing p35S:GFP-DCP1 or pSFH8:mNeon-SFH8. The regions of interest were set on mobile cytoplasmic puncta (at the midsection of the cell; 7 DAG, region 3-4). Note that the duration of tracking was limited to the amount of time that the observed particles remained in a single focal plane, as the required acquisition rate did not permit the collection of z-stacks (voxels). Middle: confocal time-lapse imaging (2 sec time interval) showing that mobile puncta can freely flow and deform around surfaces of other structures, as well as undergo tubulation, fusion, or fission. Arrowheads denote the coalescing puncta. Lower: FRAP signal recovery (normalized integrated density) as a fraction of time. The red faded band parallel to the Y-axis indicates laser iterations time. Data are means ± SD of three independent experiments with one replicate each (n=3). Scale bars, 0.4 μm.

I. SFH8 puncta and clusters show features of LLPS proteins. Micrographs from 10% (v/v) 1 h 1,6-hexanediol treatment of pDCP1:DCP1-GFP, pSFH8:mNeon-SFH8 or pFLOT1:FLOT1:GFP expressing lines (5 DAG cell surface and midsections, epidermal root cells regions 2-3). Note that in the case of DCP1 and SFH8 cytoplasmic puncta were converted to a diffused cytoplasmic signal. DCP1 decorates the LLPS compartment known as processing bodies and was used as a positive control for 1,6-hexanediol efficiency under our conditions. FLOT1 decorates microdomains at the PM (Jarsch et al., 2014). The insets are from the PM surface (TIRFM), showing that 1,6-hexanediol dissolved most of the SFH8 clusters. Scale bars, 5 μm.

J. In vitro LLPS assays for SFH8 variants. Representative confocal micrographs of in vitro recombinant GST, GST-SFH8, GST-SFH8^ΔIDR, GST-SFH8^R84A proteins stained with Alexa638 in the presence or absence of the PEG3000 crowding agent in LLPS conditions visualized in fluidic chambers (see STAR METHODS). Note that in the presence of PEG, SFH8 and SFH8^R84A switch to agglomerate-like states, while GST-SFH8^ΔIDR induced filamentous transition to GST-SFH8. Scale bars, 0.2 μm. Images are from a single representative experiment replicated three times. Bottom: representative confocal high-speed micrographs showing droplet and aggregate signal recovery, and quantification of FRAP signal recovery (normalized integrated density) as a fraction of time of GST-SFH8^R84A droplets and aggregates. Scale bars, 0.2 μm. The red faded band parallel to the Y-axis indicates laser iterations time. Data are mean ± SD of five or more independent experiments with one replicate each (n=5-8). Right: GST-SFH8 ^R84A droplet fusion in the same experiment.

K. SFH8 can phase separate on model membranes. Representative high-speed super-resolution (120 nm) images from a SUPER template experiment showing that GST-SFH8 labeled with Alexa638 (0.1 μM of GST-SFH8 in the assay) can form liquid-like droplets on membranes. Furthermore, GST-SFH8 showed an increased propensity to undergo LLPS in the presence of SUPER templates (see arrowhead denoting droplet; some droplets were released in the bulk phase as they do not attach on SUPER templates). Scale bar, 0.8 μm. Images are from a single representative experiment replicated three times.

L. SFH8 conversion to high molecular weight assemblies. Native gel electrophoresis of SFH8^ΔIDR and full-length SFH8. Western blots with α-SFH8 showed the rapid conversion of recombinant GST-SFH8^ΔIDR to agglomerates (1-2 h); very few agglomeration propensities were detected for GST-SFH8. Note that in the presence of Kin7.3/ESP GST-SFH8 converted faster to high-molecular-weight assemblies (1 h). Right: detection of 6xhis-GFP-Kin7.3 using α-his, showing that in the presence of SFH8, Kin7.3 also formed agglomerates (not that evident in GST samples). Western blots are from a single representative experiment replicated twice.

M. Nanoscale observation of filaments and clusters of SFH8. Negative staining-TEM of GST-SFH8^ΔIDR filaments and GST-SFH8. Note that in this setting filaments/clusters varied in size between several nm to 0.5 μm. Scale bars are indicated on the images. Images are from a single representative experiment replicated twice.

N. The C-terminally tagged SFH8 shows a different diffusion rate than the N-terminally tagged (see also relevant figure Figure 3G). Representative confocal micrograph showing FRAP signal recovery (normalized integrated density) as a fraction of time of the N or C-terminally tagged SFH8 with mNeon on the PM (under the pSFH8), implying increased diffusion (due to cleavage, liquidity, and dwelling time of clusters; note also that the curve in O of mNeon-SFH8 did not follow single exponential curve suggesting more complex diffusion dynamics). The white rectangular denotes the region of interest that was bleached. Scale bars, 5 μm.

O. The N-terminally tagged SFH8 diffuses from the PM. Quantification of FRAP signal recovery as a fraction of time from cells expressing pSFH8:mNeon-SFH8 or pSFH8:SFH8-mNeon lines (in sfh8 background) on the PM (5-7 DAG, epidermal cells of region 3; upper graph), Percentages indicate immobile fractions for N- and C-terminally tagged SFH8 with mNeon or N-tagged SFH8. The red faded band parallel to the Y-axis indicates laser iterations time. Data are means ± SD of three independent experiments with one replicate each (n=3). Asterisks indicate statistical significance at p<0.0001 (Wilcoxon).

To ascertain the association between SFH8 puncta and KISC, we established lines expressing an N/C-tagged fluorescent SFH8 (hereafter “cleavage biosensor”; pRPS5a:mScarlet-SFH8-mNeon). We assumed that if SFH8 was cleaved in vivo then signals of mNeon and mScarlet would not colocalize. In good agreement, we observed a lack of signal colinearity between mNeon and mScarlet in distal meristem regions but not earlier (region 3), coinciding with KISC-SFH8 interaction (Figure 4C). Furthermore, SFH8 clusters showed both mNeon/mScarlet signals, unlike the filaments which were only decorated by mNeon, suggesting that filaments represent a cleaved SFH8 form (Figure 4C). This developmentally “late cleavage” is consistent with the increment of cleavage with the plant age (Figure 4A-D), as 3 days after germination the root core meristem (region 1; around QC, see also Figure 1F) in which SFH8 is uncleaved, would comprise a large proportion of the root.

But why does the cleaved fragment of SFH8 form cellular puncta? To answer this question, we aimed first at identifying the structure of this fragment, and to do that with high precision we had to define the exact position at which cleavage takes place. We thus immunoprecipitated SFH8-mScarlet-HF and performed SFH8 proteomic peptide coverage abundance. Our results indicated a potential cleavage site right after the arginine R⁸⁴, which coincided with the reported in non-plants ESP cleavage φEXXR consensus sequence I⁸⁰EDVR⁸⁴D (Hellmuth and Stemmann, 2020; Liu et al., 2017) which is also conserved in other SFH8-like proteins (Figure 4D, E, File S3). We further established an in vitro ESP cleavage assay in which we confirmed that immunopurified ESP, mitotically activated through co-expression with cyclinD (Xu et al., 2020), could cleave recombinant GST-SFH8 at R⁸⁴; we confirmed the specificity of our assay by showing the cleavage of the well-known target of ESP in non-plants, the cohesin (Figure S4A-D) (Minina et al., 2017).

To define the structure of the 1-84 aa SFH8 N-terminal fragment, we turned to in silico predictions and found that it is intrinsically disordered (IDR; hereafter “SFH8^IDR”; Figure 4F), suggesting that is unlikely to be defined by persistent secondary and/or tertiary structures. This architecture is conserved throughout evolution for SFH-like proteins (File S3). Furthermore, deletion of this IDR region led to significant polarity loss of SFH8, lack of developmental robustness, and reduced sfh8 complementation (Figure S4E), suggesting that the controlled removal of the IDR is important for development.

As IDRs are enriched in proteins undergoing LLPS (Banani et al., 2017), we tested whether SFH8^IDR is associated with the liquid-like features of SFH8 at the PM. SFH8^IDR was predicted through catGranule as an inducer of LLPS, while for the corresponding region of a close SFH8 homolog, the SFH6 (SFH6^IDR) catGranule predicted reduced propensity to undergo LLPS (Figure 4F). We further observed that during imaging, “SFH8^IDR” puncta lacked colocalization with vesicular markers (Figure S4F, G), and showed LLPS hallmarks such as dynamic morphology with frequent splitting, fusion, and interconnections (Movie S4, Figure S4H). FRAP of mNeon-tagged SFH8^IDR puncta showed a rapid signal recovery (t1/2∼10 s, mobile fraction 40%) and sensitivity to 1,6-hexanediol which may block LLPS (Figure S4H, I) (Cuevas-Velazquez et al., 2021; Kroschwald et al., 2017); 1,6-hexanediol led also to the dissolution of SFH8 clusters on the PM but not of filaments (Figure S4I).

Since SFH8 clusters showed droplet-like morphology, sensitivity to 1,6-hexanediol, rapid FRAP implying liquidity, and fused akin to cellular condensates (Figure 3D), we assumed that SFH8 clusters may form by LLPS. In silico prediction using the algorithms, PLAAC and CIDER showed that SFH8 lack prion-like domains, but can adopt context-specific conformational states with an absolute value of net charge per residue of 0.014 (NPCR, “Janus” conformer). This result suggests that SFH8 propensity to LLPS may be sensitized by microenvironments and that SFH8 may be represented by an ensemble of conformers. The highly electronegative field of the PM where SFH8 accumulates could lead to (un)balancing of opposite charges upon IDR removal (IDR NCPR=0.071), and structural expansion due to repulsive forces, which could explain the filamentous transition.

LLPS of SFH8 relied on the N-terminal IDR, as swapping of SFH8 IDR with that of SFH6 reduced clustering at the PM and puncta formation in the presence of ESP (Figure 4G-I). SFH8^IDR sequence composition is distinct to known prion-like proteins that undergo phase transitions in the cytoplasm but shows similar amino acid distribution to the average IDR profile for Arabidopsis (Figure 4J; (Powers et al., 2019). CIDER predicts SFH8^IDR as a strong polyampholyte and thus may act as an “entropic bristle” through random movements around its attachment point and exclude the contact with other proteins, reducing the probability also of SFH8 to undergo expansion and filamentous transition. Overall, our analyses suggest that SFH8 behaves like a polyampholyte undergoing LLPS at the PM with the negatively charged lipids (e.g. PI) likely buffering repulsive charges.

To experimentally determine some of the SFH8 LLPS properties, we established an in vitro LLPS assay with fluorescently labeled proteins. Under phase separation conditions (see STAR METHODS), SFH8 or the uncleavable variant SFH8^R84A rapidly formed condensates at relatively high concentrations (5 μΜ), with a propensity for fusion, while SFH8^ΔIDR formed filament-like assemblies or aggregates as expected (Figure S4J). Interestingly, in the presence of SFH8^ΔIDR, SFH8 could also form filamentous structures suggesting that the filamentous conversion of only a few SFH8 molecules can transduce this conformation to non-filamentous SFH8. We thus speculate that cleavage of only a few SFH8 molecules by the low amounts of ESP in a cell can be sufficient in the initial stages of SFH8 filamentous development. In full consistency with the in vivo data, the in vitro filaments did not recover in FRAP experiments and showed reduced circularity compared to SFH8 condensates (Figure S4J).

One point of criticism in our LLPS assays could be their physiological relevance, as we tested LLPS in a bulk phase at which SFH8 could freely diffuse and explore a 3D space. We thus established a system to test whether SFH8 can phase-separate atop model membranes. To this end, we used SUPER templates (supported lipid bilayers with excess membrane reservoir) that contain low-tension membranes surrounding a glass bead (Pucadyil and Schmid, 2010). We showed that GST-SFH8 could form large circular clusters atop SUPER templates rich in PI lipids (Figure S4K). Surprisingly, atop SUPER templates, SFH8 could show LLPS at low concentrations (0.1 μΜ) than without SUPER templates (≥5 μM) suggesting that membrane attachment of SFH8 can promote its LLPS (Figure S4K). On the contrary, GST-tagged SFH8^ΔIDR did not show similar behavior and was oligomerized in native gels (Figure S4L). Furthermore, electron negative staining microscopy confirmed that SFH8^ΔIDR formed solid filaments (∼400 nm) in vitro, unlike the full-length SFH8 (Figure S4M). Hence, PIs may neutralize electrostatic repulsions via counterion-mediated charge neutralization along SFH8, as suggested for other proteins (Pak et al., 2016), thereby mediating LLPS and acting as multivalent scaffolds. Noteworthy, in 2D systems (e.g. PM), the threshold concentration for LLPS can be an order of magnitude lower than in 3D (Banjade and Rosen, 2014). We further speculate that the released N-terminal SFH8^IDR retains features of the LLPS SFH8 state even after its cleavage, implying a “structural memory” for condensates.

Next, we queried the biological importance of SFH8^IDR removal. In lines expressing the uncleavable variant pRPS5a:mNeon-SFH8^84RtoA in sfh8 background, cells lacked the cytoplasmic fluorescent puncta, while SFH8^84RtoA was apolar much like SFH8 in KISC mutants, and did not convert to filaments (Figure 4K-M). As expected, pSFH8:mNeon-SFH8 showed higher FRAP rates on the PM (due to cleavage and liquidity of clusters), unlike the corresponding C-terminally tagged SFH8-mNeon; in WT mNeon-SFH8 showed higher diffusion as determined by FRAP on the PM consistent with the increased cleavage and release from the PM (Figure S4N). Altogether, these results suggest that, SFH8 clusters show LLPS properties and that upon cleavage SFH8 (initially in LLPS) released two proteolytic proteoforms, the C-terminal SFH8^ΔIDR (converted to solid-like filaments) and the N-terminal SFH8^IDR (cytoplasmic puncta). These transitions of SFH8 are important, as the uncleavable variant of SFH8 (or the non-accessible SFH8^84RtoA) did not rescue the sfh8 mutant.

Conversion to SFH8 Filaments Promotes Fusion of Vesicles with the PM

Next, we aimed at further deciphering the biological significance of SFH8 phase transitions. Longitudinal sections in Transmission Electron Microscopy (TEM) of rsw4K135, and sfh8 root region 3 revealed significantly more vesicles compared to WT that although were efficiently tethered at the PM remained unfused (Figure S5A, B). To substantiate the previous, we followed PIN2 dynamic delivery at PM polar domains in the corresponding mutants. We chose PIN2 as KISC and SFH8 mutants showed phenotypic similarities at early developmental stages, and KISC plays a role in PIN2 delivery (Moschou et al., 2016). Accordingly, we observed that the PIN2-GFP (or α-PIN2) signal was 2-fold reduced at the PM of sfh8/KISC mutants (Figure 5A, B). At the same time, PIN2 accumulated in endosome-like structures in sfh8, while KISC or sfh8 mutants showed reduced PIN2 polarity (Figure 5B, C; at regions 3-4). Endosomes were also observed for PIN1 protein in sfh8 but not to non-polar proteins (H⁺-ATPase - AHA1 or plasma membrane intrinsic protein 2a - PIP2a), speaking against a general role for SFH8 in exocytosis (Figure S5C, D).

Figure S5. SFH8 effects on fusion and endocytosis.

A. SFH8-dependent fusion at the PM at a nanoscale resolution using TEM. TEM micrographs focusing on the PM (region 3) showing proper tethering but reduced vesicle fusion in sfh8, and the KISC mutant K135rsw4 (24 h at the restrictive temperature 28°C; this double mutant phenocopies rsw4 as shown in Moschou et al., 2016). Arrowheads denote vesicles that did not fuse with the PM and may imply that SFH8/KISC regulates fusion. CW, cell wall. Scale bars, as indicated on the images. Images are from a single representative experiment replicated two times (n=50, with various magnifications). Note that image inspection did not reveal perturbations in the ultrastructure of organelles in the mutants. We cannot rule out though subtle differences at the ultrastructure.

B. Quantification of fusion events from TEM images at the PM and the cell plate in WT, sfh8, and K135rsw4 (four independent pooled experiments; n=19 counts in each experiment, Wilcoxon).

C. and D. SFH8 does not seem to be a general regulator of the secretory pathway; the non-polar AHA1 and the polar PIN2 recovery rate at the PM are not affected by de novo protein synthesis. Quantification of FRAP signal recovery (normalized integrated density) as a fraction of time from p35S:GFP-AHA1 (upper; inset graph without CHX) or pPIN2:GFP-PIN2 (lower) expressing lines. In the lower quantification, GFP-PIN2 retention in sfh8 and K135 endosomes is shown (30 min CHX treatment, right). Note that CHX did not lead to the dissolution of PIN2 endosomes. The red faded band parallel to the Y-axis indicates laser iterations time (five or more independent experiments with one replicate each; n=5-8). Ιmages (right D) are representative of an experiment replicated two times.

E. SFH8 is affecting a subset of polarly localized PIN proteins and likely is not a general regulator of exocytosis or secretion. Representative confocal micrographs showing PIN1, PIN3, PIN4, and PIN7 delivery at polar domains in WT, sfh8, and rsw4(left) expressing pPIN1:PIN1-GFP, pPIN3:PIN3-GFP, pPIN4:PIN4-GFP, and pPIN7:PIN7-GFP, 5 DAG. PIN1 levels were also reduced in the mutants, while endosomes could also be observed. Note that slight level perturbations were also observed for PIN3, 4, and 7 in the sfh8 and K135rsw4 (24 h at the restrictive temperature 28°C; 5 DAG), but due to somewhat different patterning of the columella cells in these mutants these are hard to follow-up (see also discussion on patterning in Moschou et al., 2013). On the other hand, no endosomes were observed for these proteins in the mutant backgrounds. Bottom: localization details for PIN1; note that in some cases PIN1-GFP showed endosomes accumulating in the sfh8 (mostly) and K135rsw4, suggesting that PIN1 delivery was also compromised albeit to a lesser extent (as also discussed in Moschou et al., 2013 for the rsw4 mutant). Scale bars, 20 μm. Images are representative of an experiment replicated multiple times.

F. Representative confocal micrographs of AHA1-GFP in WT and sfh8 on the PM of epidermal root rip cells (5 DAG, region 3) expressing p35S:GFP-AHA1, or p35S:PIP2a-GFP. Scale bars, 20 μm. Images are representative of an experiment replicated multiple times.

G. Western blots showing lack of binding between immunopurified clathrin complexes containing CLC2-GFP from lines expressing pCLC2:CLC2-GFP (seedlings of 7 DAG), with recombinant GST-SFH8 or GST-SFH8^ΔIDR. As a control, a TAP-tagged GFP was used (producing a signal also due to secondaries that could efficiently bind the immunoglobulin of the TAP-tag; *IgG). Western blots are representative of an experiment replicated three times. Right: representative dual-channel TIRFM micrographs showing the lack of colocalization between mScarlet-SFH8 (magenta) and CLC2-GFP (green; mock) from pSFH8:SFH8-mNeon/pCLC2:CLC2-GFP expressing lines. Scale bar, 0.2 μm. Images are representative of an experiment replicated multiple times, irrespective of the root region examined.

H. Lack of colocalization between SFH8-mNeon and TAMRA-PEP1 (left; stains TGN and PM clusters), and quantification of the rate graphs showing TAMRA-PEP1 and FM4-64 uptake in WT, sfh8, and K135 (violin plot and rate plot, respectively; three independent experiments; n=9, Wilcoxon-other tests produced similar results; experimental setting in STAR METHODS). Scale bars, 2 μm.

Figure 5. SFH8 can affect the delivery of polar proteins like PIN2.

A. PIN2-GFP signal intensity (color-coded) in WT and sfh8 and quantification (right) of PIN2-GFP on PM of WT, sfh8, K135, and rsw4 (7 DAG, region 3; two independent experiments; n=164-203, multiple comparisons Dunnett). Scale bars, 50 μm.

B. PIN2 localization in WT and sfh8 (color-coded; counterstaining with a-β-tubulin to show focal plane; magenta; 5 DAG, region 3). Arrowheads indicate PIN2 accumulation maximum, and in the sfh8 slight polarity offset. Note the high number of PIN2-positive endosome-like structures in the sfh8. The sfh8 signal intensity was adjusted to normalize signal intensity between sfh8 and WT. Scale bars, 10 μm. Right: quantification of endosomes above the confocal diffraction limit (∼200 nm) in WT and sfh8 under normal conditions (four independent experiments; n=123, ordinary ANOVA).

C. α-PIN2 localization (midsection epidermis and cortex; 7 DAG, region 3) in WT,sfh8, and sfh8 with SFH8^R84A, or HF-SFH8 (“F”). Yellow arrowheads denote PIN2 polarity. Right: PIN2 polarity index and cells with proper polarity in WT, sfh8 (expressing also SFH8^R84A or HF-SFH8), K135, and rsw4 (for the index: means±SD; four independent experiments; n=10, 1-way ANOVA, “*”, p<0.0001; for the number of cells: four independent experiments; n=118, Kruskal-Wallis). Scale bars, 5 μm.

D. PIN2-GFP localization in WT and sfh8 (epidermis and cortex; 7 DAG, region 3) treated with 50 μm BFA (±CHX) for 1 h (left) and after BFA washout for 30 min. CHX was added to a final concentration of 30 μM (1 h pre-treatment and retained throughout the experiment). Right: BFA bodies quantification in WT and sfh8 (three independent pooled experiments; n=15, 5 fields of view; paired 2-tailed t-test between WT/SFH8 in the presence of BFA). Scale bars, 5 μm.

E. FRAP micrographs from polarized PIN2 (region 3) in WT and sfh8. Note the offset of PIN2 polarity (yellow arrowheads) in sfh8. The rectangular denotes the bleached region of interest (ROI). Scale bars, 3 μm.

F. FRAP PIN2 signal recovery (normalized integrated density). Data are means ± SD of three independent pooled experiments (n=15). The red faded band parallel to the Y-axis indicates laser iterations time. Numbers next to the genotype, denote recovery half-time±SD (n=15, paired 2-tailed t-test).

G. Representative images of co-localization of N- or C-terminally-tagged SFH8-mScarlet and HF-Kin7.3-RFP with PIN2-GFP in regions 1-4. Scale bars, 5 μm.

H. PCC colocalization quantifications between PIN2 and SFH8 or Kin7.3 (apicobasal or lateral domains, for Kin7.3; three independent pooled experiments; n=36, nested 1-way ANOVA, p<0.05, *; p<0.001, **; p<0.0001, ***, compared to region 1).

I. Polarity index for SFH8, Kin7.3, and PIN2 in 3 root regions (see also Figure 1; three independent pooled experiments; n=36, nested 1-way ANOVA).

J. High speed (40 frames per second) and super-resolution (120 nm) confocal images with insets showing details of HF-mScarlet-SFH8 cluster/PIN2 exclusion. Note that the two proteins colocalized in non-clustered PM domains (see also Figure 6 and corresponding text). PCC values represent colocalization analyses between filamentous patches of Kin7.3 or SFH8 with PIN2, while clusters of SFH8 (region 2: to observe the signal at the PM) showed anti-correlation (denoted also by the arrowhead). PCC±SD values are from three independent pooled experiments (n=120).

Next, we asked whether SFH8 can promote the delivery/stability of PIN2 at polar domains. We thus used the drug brefeldin-A (BFA) to form intracellular agglomerates of PIN2 in the so-called “BFA-bodies” (aggregate of TGN and Golgi). We were able to calculate the uptake of PIN2 to BFA-bodies from the PM, and the delivery rate from PIN2-positive BFA-bodies to the PM following washout of the BFA (Paponov et al., 2019). We further combined BFA experiments with FRAP to measure the PIN2 delivery rate at the PM. Both assays confirmed that in sfh8/KISC mutants PIN2 delivery to the PM was compromised, while PIN2 endocytosis was not compromised, as PIN2-positive BFA-bodies were produced in both WT and sfh8 mutants at the same rate; these effects were independent of de novo PIN2 synthesis as cycloheximide (CHX) treatments did not affect delivery rates or dissolve the PIN2 endosomes in sfh8 (Figure 5D-F, S5E, F).

To address the mechanism by which SFH8 could affect PIN2 delivery we checked whether SFH8 stabilized PIN2 at the PM. At root region 3, SFH8-mScarlet (and Kin7.3) but not mScarlet-SFH8 showed similar polarity and colocalized with PIN2 (Figure 5G-I; PCC ∼0.9), while in regions 1-2 PIN2-GFP and mScarlet-SFH8 PM puncta excluded each other as shown in a super-resolution setting (Figure 5J; insets). This anticorrelated localization implied PIN2 removal from the PM through endocytosis. Yet, SFH8 is unlikely to promote PIN2 internalization via clathrin-mediated endocytosis as it ii) did not affect PIN2 endocytosis (see above BFA experiment), and did not interact with Clathrin Light Chain (CLC2)-GFP or affected endocytosis of the dye FM4-64 or the peptide PEP1 (Figures 5D, S5G, H). Thus full-length PM SFH8 clusters may reduce access to certain polar proteins, perhaps through the “entropic bristle” effect exerted by the IDR that could establish “fusion-depletion zones” at early developmental stages, while it may exert a positive effect on fusion at later stages (see below).

As PIN2 showed colocalization at the PM with regions where SFH8^ΔIDR had formed, we tested whether SFH8^ΔIDR could associate with PIN2. Using a custom-specific antibody against SFH8 (C-terminus), we observed that in pin2 mutants SFH8 was not affected suggesting that SFH8 likely recruits PIN2, and not the other way around (Figure S6A and 6C for antibody specificity). To test direct PIN2/SFH8 interaction we refined a quantitative proximity ligation assay (PLA; (Teale et al., 2021)). PLAs use complementary oligonucleotides fused to antibodies to determine the frequency with which proteins of interest find themselves nearby (Figure 6A). In PLA, we observed positive interactions in meristems mainly of region 3 using two different settings: i) in GFP-PIN2 expressing roots (PLA antibody combination α-GFP/α-SFH8), or ii) in PIN2-GFP with C-terminally HF-tagged SFH8 (PLA antibody combination α-GFP/α-FLAG). Significantly reduced PLA signals were observed between PIN2/SFH8 in the N-terminally tagged HF-SFH8 lines, while no signals were observed for i) SFH8 and a PM aquaporin (PIP2a), ii) sfh8 (PIN2/α-SFH8), and iii) in the vasculature where PIN2 is absent (Figures 6B-E, S6B-D). PLA showed that Kin7.3 interacted exclusively at the PM with SFH8 (Figure S6C; region 3 onwards). We corroborated PLA results using Fösters Resonance Energy Transfer (FRET) analyses, showing that high FRET efficiency for the SFH8-mScarlet/PIN2-GFP pair, indicative of interaction, was evident at the root regions 3-4 (Figure 6F, S6). We verified also the PM interaction of ESP/Kin7.3 in the same region, using as FRET readout the N/C-terminus intramolecular interaction of Kin7.3 in the rsw4 background; Kin7.3 folded over on the PM in rsw4 background as previously suggested for MT binding (Figure S6B; Moschou et al., 2016). Collectively, these results show that SFH8 filaments associate with PIN2 at the PM.

Figure S6. SFH8/PIN2 interactions and assays control for PLA; liposome fusion assay controls.

A. PIN2 is not recruiting SFH8 at the PM. Representative confocal micrographs showing α-SFH8 signal in WT, and pin2 mutant background (5 DAG, region 3 cortex cells), and quantification of α-SFH8 signal intensity in WT, and pin2 mutant (lower; three independent experiments with signal calculations from cortex cells from regions 3-4 with three roots each; n=9, Wilcoxon). Scale bars, 5 μm.

B. Control experiments showing the specificity of the PLA approach (PIN2/SFH8 assay). SFH8 did not interact with PIP2a, while SFH8-HF (α-FLAG) interacted extensively with PIN2 (α-GFP) at region 3 onwards at the PM. Note the significantly reduced PLA-positive signal in HF-SFH8/PIN2 PLA experiment (bottom panel; PLA signal was observed at the cell plate as well). The sfh8 mutant was used as a negative control for the PLA approach (specificity of α-SFH8 in a PLA setting; sometimes a nuclear non-specific signal was observed in negative controls). Scale bars, 50 μm. Images are representative of experiments replicated five times.

C. SFH8/Kin7.3 interact significantly with each other at regions 3-4. PLA of α-mNeon/SFH8 and inset (right) showing the increased interaction strength towards region 4 (for example in the vasculature; upper images). The “no -α-SFH8” corresponds to a negative control (bottom) in which only α-mNeon was used and thus did not produce a positive PLA signal. Scale bars, 50 μm (for “region 3” panel, 10 μm). Images are representative of an experiment replicated five times.

D. FLAG epitope stainings confirm SFH8 localizations. α-FLAG signal in HF-SFH8 or SFH8-HF expressing lines in the sfh8 mutant (5 DAG; Figure S2F for complementation). Insets denote the localization at the corresponding region. Images are representative of an experiment replicated twice. Scale bars, 50 μm.

E. Kin7.3/ESP likely interact mainly at the PM. SE-FRET experiments at the PM show that N- and C-termini of Kin7.3 fold over in the absence of ESP (rsw4; see also Moschou et al., 2016 for a similar situation upon MT binding), suggesting also that ESP/Kin7.3 interact at the PM. Scale bar (micrograph), 5 μm. “control”: p35S:sGFP-tagRFP expressing lines (yielding high FRET). Data are means±SD of three independent experiments with three roots each (n=9, Wilcoxon).

F. Control experiments show that SFH8/SFH8^ΔIDR do not differ in their affinity for LUVs. As liposome co-flotation assays take significant incubation time and thus can be biased towards the increased molecular weight and sedimentation propensity of SFH8^ΔIDR due to its filamentous conversion and reduced hydrodynamic radius leading to erroneous affinity determination, we designed a semi-quantitative in vitro assay based on Quartz Crystal Microbalance with Dissipation monitoring (QCM-D). We used the QCM device as a sensitive mass sensor monitoring the frequency response (Δf) of the acoustic wave during binding events on the surface of a model membrane (Mateos-Gil et al., 2016). The device surface was first covered with neutravidin (layer 1) which was used to bind specifically a 5’-biotinylated DNA (layer 2). This DNA was also modified at the 3’-end with a cholesterol moiety, further employed to anchor a liposome (layer 3). We prepared 50 nm diameter liposomes with two different lipid compositions: DOPC:PI(4,5)P2 (99:1, n/n) and DOPS:PI(4,5)P2 (99:1, n/n). Finally, proteins (e.g. GST-SFH8) were infused (layer 4) on top of layer 3. Bottom graph: SFH8 binding on membranes is independent of curvature in QCM-D setting. The size of liposomes did not significantly affect SFH8 binding, excluding the possibility that SFH8 required curvature (convex or concave) or flat surfaces for binding. Binding of SFH8 on 50/200 nm on QCM-D coupled liposomes with PS:PI(4,5)P2 (99:1). Note that the small difference in Δf observed could be due to the larger surface of 200 nm liposomes, incorporating more protein. The schematic diagram shows the differences in curvature between liposomes with a diameter of 200 and 50 nm. Note the increased surface accessible to the tangent, which represents the interface between liposomes and proteins. Data are representative of an experiment replicated two times.

G. Confirmation of results in F using imaging. Representative confocal images showing that the hexahistidine tagged recombinant mScarlet-SFH8/mScarlet-SFH8^ΔIDR (proteins purified from E. coli) showed binding in the solution of PC:PI(4,5)P2 containing SUV liposomes (99:1; liposomes stained with fluorescein in their lumen) under the fusion assay conditions. Scale bars, 400 nm. Images are representative of an experiment replicated two times.

H. Determination of Kin7.3 binding on GST-SFH8 and GST-SFH8^ΔIDR. Again, due to the filamentous conversion of GST-SFH8^ΔIDR as described above (see F), we designed a real-time binding assay with a supported lipid bilayer approach for Kin7.3 tail affinity determination: GST-SFH8 and GST-SFH8^ΔIDR protein binding on QCM-D coupled liposomes. Lower: real-time curves showing dynamic binding through the frequency response of QCM-D obtained upon infusion of full-length SFH8 (upper panel) or GST-SFH8^ΔIDR (bottom panel) proteins on supported lipid bilayers carrying hexahistidine-tagged Kin7.3 tail immobilized on the supported lipid bilayer via DGS-NTA(Ni) lipid anchor. Only slight, insignificant differences were observed in the binding between the two protein variants. As of note, SFH8/SFH8^ΔIDR differ in molecular mass and thus SFH8 should give a slightly higher response in Δf as shown. Data are representative of an experiment replicated twice.

Figure 6. Removal of the SFH8 IDR promotes interaction with PIN2 and membrane fusion.

A. Proximity ligation (PLA) principle.

B. PLA-positive signal (resembling puncta) produced by α-SFH8/α-GFP in pPIN2:PIN2-GFP in WT or sfh8. Scale bars, 5 μm.

C. Micrographs showing α-SFH8 signal in WT (“sfh8” is a negative control). Scale bars, 5 μm.

D. Details of PLA-positive PIN2/SFH8 signal puncta at the PM. Scale bar, 1 μm.

E. Quantification of PLA-positive cells in 4 regions (five independent experiments; n=20, nested 1-way ANOVA).

F. Sensitized emission (SE)-FRET efficiency between PIN2-GFP (pPIN2) and SFH8-mScarlet (pRPS5a). Scale bar, 50 μm. Right: signal quantification of SE-FRET between the indicated combinations (five independent experiments; n=50, paired 2-tailed t-test).

G. Pipeline of the DNA-zippers approach to test membrane fusion.

H. DNA-zipper assay with GST-SFH8^ΔIDR or -SFH8 (full length; lumen labeled with fluorescein only). The enlarged micrographs (top) show a time series of the tethering/fusion of two liposomes that converted to GUVs due to SFH8^ΔIDR. Quantifications of fusions are also shown (three pooled independent experiments, with n=225; means indicated with vertical lines; Wilcoxon). Scale bars, 2 μm.

I. Fusion blockage by fluorescently-labeled SFH8 clusters on liposomes (stained with PE-texas red (magenta). Arrowhead denotes an SFH8 cluster formed atop the labeled liposome (images after deconvolution in a super-resolution setting with 120 nm resolution). Scale bars, 2 μm.

J. Liposome content mixing with lumen stained with fluorescein (green), and lipids stained with PE-texas red (magenta). Scale bars, 2 μm. Insets show a hemifusion event (upper), and a combination of hemifusion with a tethered LUV (lower inset). The arrows show content mixing (pseudocolored white). Lower: distribution (%) of LUVs among different treatments (three pooled independent experiments; n as indicated, Dunnett for “fusion”).

Next, we asked whether the SFH8/PIN2 anti-correlation related to stereochemical hindrance imposed by SFH8 clusters that restrained fusion (the aforementioned entropic bristle hypothesis). We thus established an in vitro membrane fusion assay based on cholesterol-modified DNA zippers (“lipid-DNA-zippers” (Stengel et al., 2008) (Figure 6G). This minimalistic approach lacked other proteins, allowing to focus on direct SFH8 effects. To test SFH8-mediated membrane fusion, we used assays with low content of labeled phosphatidylethanolamine (PE)-texas red (peripheral labeling), and fluorescein-containing large unilamellar vesicles (LUVs; 400 nm). Under our super-resolution settings, we could resolve three events driven by DNA zippers: membrane tethering, hemifusion (lipid mixing), and fusion which resulted in the lipid bilayer unification and the intermixing of the volumes. As SFH8^ΔIDR converted to filaments in a few minutes, and could thus precipitate in lipid co-flotation or co-sedimentation assays, to ascertain affinities and exclude non-specific effects on fusion by the phase behavior of the used proteins, we used Quartz Crystal Microbalance with Dissipation monitoring to monitor in real-time binding on LUVs, confirming validity of our system (Figure S6F-I). SFH8/SFH8^ΔIDR showed similar affinities for the LUVs, excluding differential binding in this setting as a potential driver of fusion (Figure S6F-I). In the SFH8^ΔIDR samples, the average diameter of liposomes was ∼1 μm, in contrast to SFH8 (∼0.5 μm) which was below that in the free GST samples (∼0.7 μm) used as controls (Figure 6H). Content mixing analyses showed that almost 30% of the SFH8^ΔIDR samples had semi-fused or fused LUVs (2-fold lower for the full-length SFH8; Figure 6H-J), while fusion/hemifusion events in SFH8 were less than those of GST. This result suggests that SDH8^IDR blocks fusion, while SFH8^ΔIDR may reduce the energy barrier required for fusion through the increase in fluidity by its filamentous structure (Carlini et al., 2020) or the reduction of the entropic bristle effect. Although the exact mechanism underlying SFH8 fusogenic activity is still an open question, our results suggest that SFH8^IDR can restrain vesicles from reaching membranes.

KISC/SFH8 Define a Positive Feedback Loop on Auxin Transport

We asked whether KISC/SFH8 phase transitions have a developmental impact, by examining whether defects of PINs in sfh8 are sufficient to drive changes in auxin signaling. Through the ratiometric auxin input R2D2 and the DR5 auxin output sensors (R2D2 principle is detailed in Supplemental Info), we could confirm auxin defects in sfh8 and KISC mutants with occasional ectopic auxin input or output in distal meristem regions, consistent with perturbations in auxin signaling robustness (Figure 7A, B). qRT-PCR showed reduced expression robustness of selected auxin-related genes at root meristems of sfh8 and K135 (Figure S7A). Using imaging, we confirmed that the least robust gene out of the ones tested, yucca2 showed surprising expression variation in the sfh8 mutant among root cells (Figure S7B). Yet, gene expression patterns of auxin-related genes were rather moderately affected in sfh8 or KISC mutants suggesting that auxin signaling buffers to some extent perturbations of polar auxin transport in a cell-autonomous manner. Overall, the auxin signaling changes are consistent with a model in which SFH8/KISC-mediated defects in PINs affect the robustness of auxin signaling.

Figure S7. SFH8 affects auxin signaling robustness.

A. Real time-quantitative-PCR (q-RT-PCR) setting for evaluation of changes in auxin-signaling genes (left). Two-millimeter root meristems were cut and processed for RNA extraction. Right: relative expression of the indicated genes normalized against Actin-7 and PP2A (mean±SD from three independent experiments with two technical replicates; 2-tailed t-test to test robustness). Note the high variability of gene expression in the mutants (expanded SD). Asterisks: significance at p<0.05 (*) or p<0.0001 (***); t-test.

B. Confirmation of reduced robustness of expression in sfh8 of the auxin biosynthetic gene yucca 2. Representative confocal micrographs showing pYUCCA2:GFP expression (5 DAG root meristem cells). Images are representative of experiments replicated five times. Scale bars, 20 μm.

C. Quantification of pYUCCA2-GFP signal distribution (signal integrated density), in WT and sfh8 (three independent experiments with three roots each; n=1,492 points across the root; 2-tailed t-test).

D. Plot profile showing relative signal intensity along the root in WT and sfh8. QC, quiescent center (single representative experiment, replicated three times). Note the aberrant intensity peaks of pYUCCA2-GFP in sfh8.

Figure 7. SFH8/KISC comprises a feedback loop with auxin.

A. R2D2 auxin input sensor (cartoon: transcriptional units DII-Venus (“DII”) and mDII-Tm (“mDII”)) in WT, sfh8, and KISC mutants. The arrowhead denotes a region with perturbed auxin input. Scale bars, 20 μm. Bottom: “DII” and “mDII” auxin input perturbations in sfh8 (right). The chart shows DII/mDII ratio quantification (QC to 21^st cell towards the shoot; left), and the heat map shows DII/mDII ratio in WT, sfh8, K135, and rsw4 (QC(1)-21^st cell). Data are mean ± SD from three different roots (n=21 cells per root).

B. Auxin output sensor DR5 in WT and sfh8. Arrowhead denotes an auxin ectopic maximum in sfh8. Yellow outlines denote the roots. Scale bars, 20 μm.

C. and D. mNeon-SFH8 localization ± 100 nM IAA (15 mins, 5 DAG epidermal cells; scale bars, 10 μm); quantification of PM/cytoplasmic signal (bottom)(C). GFP-Kin7.3 localisation ± 100 nM IAA (15 mins); quantification of PM/cytoplasmic signal (right; four pooled independent experiments; n=64, Kolmogorov-Smirnov)(D). Scale bars, 20 μm.

E. TIRFM of Kin7.3/SFH8 ±IAA (1.5 h, region 3). Images are representative of an experiment replicated three times. Scale bars, 0.2 μm.

F. and G. Diffusion and mobility distribution of SFH8 clusters in the presence or absence of auxin (NAA or IAA) from TIRFM (time interval 100 msec). The percentage of immobile and mobile molecules in each graph represents clusters with a trajectory below and above the log(D) of - 1.75, respectively (D, diffusion coefficient). Smaller inset graph: individual trajectories of mobile clusters in ± auxin (left and right, respectively; 500 frames, n=120). Quantification of SFH8 cluster size (±100 nM NAA/IAA 5 min, n=120) (G).

H. Reduction of FLAG-SFH8 or HF-mScarlet-SFH8 cleavage by IAA (1.5 h). Arrowheads denote cleaved SFH8^IDR (HF-IDR or HF-mScarlet-IDR). WBs were replicated three times. Right: quantification of SFH8/SFH8^IDR bands (HF-mScarlet-SFH8 experiments; n=4, 2-tailed t-test showed difference at p<0.05 between SFH8/SFH8^IDR)

I. Model for SFH8 function and IDR-dependent fusion blockage by SFH8^IDR.

We further postulated that reciprocally auxin could affect SFH8/KISC localization, as at cells with auxin levels SFH8 was in the clustered state (e.g. region 1). PM-resident mNeon-SFH8 signal increased in the presence of exogenous auxin (IAA; 15 min 100nM; Figure 7C). On the contrary, under the same conditions, GFP-Kin7.3 showed a slight decrease at the PM, and reduced overall colocalization with SFH8-mScarlet clusters in TIRFM (Figure 7D, E). Furthermore, IAA (or NAA) increased clusters of SFH8 and their size, while it did not increase SFH8 filament occurrence, and diffusion at the PM (Figure 7F, G). Yet, when diffusing SFH8 condensates were nearby pre-established KISC filaments, they showed a “hop” diffusion consistent with sterical entrapment on MTs. Consistently, prolonged IAA treatment led to SFH8 cleavage reduction (Figure 7H). Thus, auxin could induce SFH8 clustering at the PM and perhaps entrap SFH8 in lipid patches where SFH8 undergoes LLPS.

DISCUSSION

Through the discovery of the KISC-SFH8 link, we identify a developmental module incorporating protein material properties switches. Collectively, this mechanism comprises an array of events summiting from i) auxin-controlled SFH8 LLPS at the two-dimensional space of the PM, ii) reduction of SFH8 clustering when auxin drops, and consequent interaction with KISC/SFH8 and ii) proteolytic SFH8 cleavage accompanied by structural modifications enabling its functional switching that reinforces maintenance but not initial polarity, e.g. of PIN2. This reinforcement contributes to the robustness of auxin signaling, although we believe that this pathway has extended functions. Interestingly, the phase transition of SFH8 is induced by limited proteolysis executed by the highly conserved protease ESP, thereby expanding the repertoire of ESP functions. This developmental module also uncovers a novel way to regulate protein phase transition by showing that limited proteolysis is a post-translational modification involved in LLPS. PS patches form at the PM when auxin levels are high, thereby, stabilizing ROP6 to amplify signaling outputs (Platre et al., 2019) or reinforce polar domains in leaves (Pan et al., 2020). In contrast to the ROP6, SFH8 recruits KISC at low auxin levels, while it is incompetent to interact with KISC when it is in auxin-induced clusters. Because SFH8 interacts with PS and shows reduced clustering in the distal meristem, we suggest that SFH8 condensation could be promoted by PS or PIs patches. These lipids could buffer the highly positive localized charge of SFH8, especially at the nodulin domain thereby enabling the condensation required for LLPS.

It has long been understood that molecular condensation is crucial for out of equilibrium polarity establishment in biological systems. At first glance, our work seems counterintuitive as it starts challenging these models by showing that LLPS may simply be a mechanism for reducing polarized secretion, in contrast for example to what has been observed in synapses where protein condensation can increase localized secretion (Milovanovic et al., 2018). On the contrary, SFH8 becomes functionally important for polarity at a less “condensed” state upon ESP-mediated removal of the N-terminal IDR at the PM. This event preludes further structural reconfigurations and material properties shifts (from LLPS to solid filaments) that promote a functionality switch of SFH8: from fusion-blocker to fusion enhancer. Likewise, Dynamin-related proteins (DRPs) oligomerize into high order macro-molecular structures for membrane remodeling (Ramachandran, 2018), while coronaviruses (e.g. SARS-CoV-2) opt host proteases for structural reconfiguration that prime fusogenic activity of SPIKE that results in the viral-host cell fusion (Earnest et al., 2015; Tang et al., 2020). Upon IDR removal, SFH8^ΔIDR is stabilized likely due to reduced intramolecular stereochemical repulsion and charge attenuation enabling filamentous state conversion and fusion at polar domains. Overall, this mechanism establishes an example of LLPS at the PMs in plants with importance for PM micropatterning that transcends to polarized patterns.

Furthermore, the dimensionality reduction caused by the PM binding of SFH8 promotes condensation, which would be otherwise impossible at low SFH8 levels. Cluster formation through LLPS could have a direct effect on PM properties. The microscale “entropic bristle”-like properties of the SFH8^IDR when forming mesoscale condensates in the cytoplasm implies that when bound to the PM SFH8^IDR can restrict the accessibility of certain proteins to the PM at defined domains. We could reconcile this proposition in vitro using specialized fusion assays. Thereby, the removal of this IDR by ESP could locally promote vesicular fusion by removing steric hindrance effects or by promoting structural reconfigurations. Alternatively, SFH8 clusters may engulf diffraction-limited vesicles and promote their fusion upon IDR removal. Because SFH8 is a SEC14-like protein, it may render lipids vulnerable to enzymatic modifications (Kf de Campos and Schaaf, 2017), regulating local lipid environments at a micro- or nanoscale.

Addressing further how KISC/SFH8 properties are involved in the maintenance of polar domains will be an important priority for our future research. Intriguingly, the IDR region at the N-terminus is conserved throughout evolution (File S2), suggesting that similar mechanisms may apply in other eukaryotes. Thus, our work provides insights relevant to the burgeoning field of biomolecular condensates interfacing with membranes.

METHODS

Arabidopsis backgrounds and ecotypes

All the plant lines used in this study were in the Arabidopsis thaliana Columbia-0 (Col-0) ecotype except the ones as indicated individually, and a detailed description can be found in Materials Table. Primers used for genotyping, qRT-PCR, and cloning can be found in Table S1. The following mutants and transgenic lines used in this study were described previously: rsw4 mutant (Wiedemeier et al., 2002), K135 (Moschou et al., 2016), pro35S:mCherry-MAP4^MBD (Moschou et al., 2016), p35S:smRS-GFP-TUB6, pro35S:smRS-GFP-TUA6 (Nottingham Arabidopsis Stock Center-NASC; N6550), p35S:GFP-TUB9 (NASC; M84706), pPIN1:PIN1-GFP(Benková et al., 2003), pPIN2:PIN2-EGFP (Xu and Scheres, 2005); pPIN3:PIN3-EGFP, pPIN4:PIN4-EGFP and pPIN7:PIN7-EGFP (Blilou et al., 2005), DR5v2-ntdTomato / DR5-n3GFP (R2D2) (Liao et al,2015), Cytrap: pHTR2:CDT1a(C3)-RFP/pCYCB1:CYCB1-GFP (Yin et al., 2014). proSNX1:SNX1-mRFP (Jaillais et al., 2006), pTIP1:TIP1-GFP (Ma et al., 2004), p35S:GFP-PIP2a (Cutler et al., 2000), p35S:GFP-AHA1 (Hashimoto-Sugimoto et al.,2013), WAVE22, WAVE27 (Geldner et al., 2009). DR5rev:erTagRFP/pTCS:nGFP (Siligato et al., 2016), In Wassilewskija background: pCLC2:CLC2-EGFP (Konopka et al., 2008). The following lines were ordered from GABI or NASC: sfh8-1: GABI_55IF03, sfh8-2: SALK_006862.

Arabidopsis plants were transformed according to (Clough and Bent, 1998) using Agrobacterium tumefaciens strain GV3101. In all experiments, plants from T1 (co-localization experiments), T2, or T3 (for physiology experiments) generations were used. The Arabidopsis fluorescence marker lines were crossed with sfh8-1 and corresponding SFH8 transgenic lines to avoid the gene expression levels differences caused by positional effects; F1 (co-localization experiments) and F2 were used in experiments.

Plant growth conditions

Arabidopsis seedlings were sterilized and germinated on ½ strength Murashige and Skoog (MS) agar media under long-day conditions (16h-light/8h-dark) and were harvested, treated, or examined as indicated in the context of each experiment. In all experiments involving the use of mutants or pharmacological treatments, media was supplemented with 1% (w/v) sucrose. Arabidopsis plants/lines for crosses, phenotyping of the above-ground part, and seed harvesting were grown on soil in plant chamber at 22°C/19 °C, 14/10 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 (or 28-30°C for restrictive temperature treatments in the case of rsw4), 16/8 h light/dark cycle, and light intensity 150 µE m^-2 s^-1.

Plant Material, phenotypic analysis, and drug treatments

For quantification of phenotypes, seedlings were sterilized and 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). For a given genotype, differential contrast interference (DIC) images were captured on a Leica DM2500, Leica DM6B, or Leica DM6000. 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, 10% (v/v) aqueous solution was used. XVE-driven expression was activated by transferring seedlings on plates containing various estradiol (in ethanol; light-tight tube aliquots) concentrations, which we determined experimentally in trial experiments (2-100 μM).

Evolutionary relationships of taxa and sequences analyses

The evolutionary history of SFH8 was inferred using the Neighbor-Joining method (Saitou and Nei, 1987). The optimal tree is shown. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) is shown next to the branches (Felsenstein, 1985). The evolutionary distances were computed using the Poisson correction method (Zuckerkandl and Pauling, 1965) 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 were a total of 611 positions in the final dataset. Evolutionary analyses were conducted in MEGA X (Kumar et al., 2018).

Transient N. benthamiana assays

For transient assays, Agrobacterium tumefaciens strain GV3101 was used. Infiltrations were done as described previously (Arora et al., 2020). During treatments, plants were kept in defined growth conditions in an Aralab growth chamber.

Root gravitropism assays

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

Cloning and Plasmids

Primer sequences used for amplicons are listed in Table S1. Clonings were done either by Gateway, restriction enzyme digestion, or In-fusion (Takara). The following constructs were produced in: i) pENTR vectors were generated via BP reaction with pDONR/Zeo (Invitrogen) and PCR products: SFH genes (amplicons from RT-PCR using cDNA from 1-week-old seedlings), truncations of SFH8 (amplicons from pDONR/Zeo-SFH8 was used as a template), mutations of SFH8 by mutagenesis PCR using pDONR/Zeo-SFH8. The Kin7.3 and truncations were previously described (Moschou et al., 2016); ii) In pGWB601: pSFH8:mNeon-gSFH8 (g, for genomic), pRPS5a::mNeon-gSFH8, pKin7.3:mNeon-gSFH8, pSFH8:gSFH8-mNeon, pRPS5a:gSFH8-mNeon, pKin7.3:gSFH8-mNeon, pRPS5a::HF-mScarlet-gSFH8-mNeon, pRPS5a:HF-gSFH8, pRPS5a:gSFH8-HF, pKin7.3:CFP-cKin7.3-mNeon. The pGWB601 empty vector was used as a backbone, cut open by XhoI and SacI, then the vectors assembled by four or five-fragments using In-fusion cloning (amplicons from Arabidopsis genomic DNA, template for mNeon, mScarlet, and proKin7.3-CFP-cKin7.3); iii) pGBKT7/pGADT7-gateway-compatible (Maier et al., 2008) for Y2H: LR reaction with the pENTR vectors of Kin7.3 and truncations, SFH8, SFH8 truncations, and mutations; iv) ratiometric bimolecular fluorescence complementation (rBiFC)-gateway-compatible system (Grefen and Blatt, 2012): Kin7.3 (pDONR/P3P2-Kin7.3) and SFH8 (pDONR/P1P4-SFH8); v) pRPS5a gateway compatible dual tagged vectors: modified pGWB517 and 560 empty vectors were used as a backbone, cut open by HindIII and XbaI, and then the vectors assembled by two fragments In-fusion cloning (amplicons from Arabidopsis genomic DNA-1.6kb promoter of RPS5a and template for FLAG and sGFP). The Kin7.3 (pDONR/Zeo-Kin7.3) and SFH8 (pDONR/Zeo-SFH8) were used to generate the FRET pair and the biosensor, respectively; vi) pGAT4 and pDEST15 gateway compatible vectors for His- and GST-protein expression in Escherichia coli: full-length or truncations of Kin7.3 and SFH8 pDONR/Zeo vectors were used; vii) inducible constructs under the promoter of Kin7.3 (XVE): vector was cut with PmeI and MluI and then the vectors assembled by two fragment In-fusion cloning (amplicons from Arabidopsis genomic DNA-1.7kb promoter of Kin7.3 and part of LexA); inserts from pENTR-GFP-ESP and pENTR-HA-Kin7.3 tail were introduced by LR reaction from the corresponding pDONR/Zeo vectors; viii) The pDR gateway compatible vector for yeast temperature-sensitive complementation: LR reaction with pDONR/Zeo-SFH8 and the indicated truncations (in pDONR/Zeo). The SYN4 cloning has been described in Miniana et al., 2017.

Yeast Two-Hybrid Screen

The genotype of the strain Y2HGold is MATa, trp1-901, leu2-3, 112, ura3-52, his3-200, gal4Δ, gal80Δ, LYS2::GAL1UAS-Gal1TATA-His3, GAL2UAS-Gal2TATA-Ade2, URA3::MEL1UAS-Mel1TATA-AUR1-C MEL1. The genotype of the strain Y187 is MATα, ura3-52, his3-200, ade2-101, trp1-901, leu2-3, 112, gal4Δ, gal80Δ, URA3::GAL1UAS-GAL1TATA-lacZ. Transformed yeast cells were incubated at 30°C until OD600=0.8 in a minimal medium (SD) lacking the amino acid tryptophan. To confirm the expression of the baits, total protein (10 µg) was extracted using alkaline lysis and resolved by SDS-PAGE. Fusion proteins were detected with α-Myc monoclonal antibodies (Roche, Stockholm, Sweden). The following constructs were used: pBKGT7-AtESP domain A (DomA; Moschou et al., 2016); pBKGT7-AtESP Domain C (DomC; Moschou et al., 2016); pBKGT7-AtMau2 (Minina et al., 2017); pBKGT7-cKin7.3; pBKGT7-cKin7.3-motor; pBKGT7-cKin7.3-tail (Moschou et al., 2016); The absence of self-activation was verified by a transformation of the baits alone to select on minimal medium (SD) lacking the amino acids leucine, histidine, and adenine. The baits were transformed into the strain Y2HGold and mated with the Universal Arabidopsis cDNA Library (Clontech) in Y187. For pairwise Y2H assays, the Gateway-compatible pGADT7 vector (Maier et al., 2008) and the yeast Y187 were used, including the following constructs: pGADKT7-SCC2 (Minina et al., 2017); pGADKT7-cKin7.3-tail; pGADKT7-cSFH8; pGADKT7-cSFH8-SEC14 domain (SD); pGADKT7-cSFH8-Nodulin; pGADKT7-cSFH8-Nodulin^2KtoA; pGADKT7-cSFH8-Nodulin^{6Kto A} (or other indicated variants).

Protein Expression and Purification

The pGAT4/PDEST15 constructs were transformed in BL21 (DE3) Rosetta or BL21 (DE3) Rosetta II Escherichia 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 expression was induced at OD600=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 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 (otherwise the SFH8 proteins precipitated).

Expression levels of proteins were estimated by CBB stainings in PAGE or by Western blots. The proteins were dialyzed overnight in assay buffers (2 L).

GST Pull-Down Assay

The GST pull-downs were performed by binding 10 μg of GST fusion proteins to glutathione Sepharose beads (GE Healthcare Life Sciences) and incubating them with 10 μg of his-tagged proteins in a binding buffer containing 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 0.5% Nonidet P-40. After an incubation of 1 h at 4°C, beads were washed three times with the wash buffer containing 25 mM Tris-HCl (pH 8.0), 150-300 mM NaCl, and 0.5% (v/v) Nonidet P-40. Proteins were then eluted from the washed beads with glutathione and analyzed by SDS-PAGE. Purified proteins were dialyzed against 75 mM PIPES (piperazine-N, N′-bis; pH 6.8), 1 mM EGTA (ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid), 1 mM MgCl2, and 10% (v/v) glycerol, 1 mM DTT (dithiothreitol), and 70 mM KCl. Expression levels of proteins were estimated by western blot. As a negative control, GST protein was used in all assays.

Protein Immunopurification

Agrobacterium strain GV3101 carrying constructs expressing various forms of Kin7.3, SYN4, SFH8, ESP or TAP-GFP were infiltrated into N. benthamiana leaves. Three-four days later, leaves were grounded in liquid nitrogen and resuspended in ten 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 14,000 x g for 20 min at 4 °C. The supernatant was filtered through four 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 300 g, washed three times with buffer A, and treated for 4 hr with PreScission protease (GE Healthcare). The TAP-ESP beads were used directly or the supernatant was incubated for 30 min with nickel beads and h-AtESP was eluted with 250 mM imidazole containing buffer A. Protein was dialyzed against 0.1M PIPES (pH 6.8), 5mM EGTA, 2mM MgCl2, and 20% (v/v) glycerol buffer. Protein levels were estimated by Western blot.

SFH8 in vitro cleavage assay

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 1-2 h, resolved by SDS-PAGE, and detected by Western blot.

Preparation of liposomes

1,2-dioleoyl-sn-glycerol-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycerol-3-phospho-L-serine (DOPS), 1,2-dioleoyl-sn-glycerol-3-phospho-(1’-myo-inositol-4’,5’-bisphosphate) (PI(4,5)P2) and 1,2-dioleoyl-sn-glycerol-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid) succinyl] (DGS-NTA(Ni)) lipids were purchased from Avanti Polar Lipids (Alabaster, Al, USA). Dipalmitoyl phosphatidylinositol 3-phosphate (PI3P) lipids were from Echelon Biosciences Incorporated (EBI). Lyophilized lipids were dissolved in 1:1 (v/v) chloroform:methanol mixed in glass flasks in the desired amount. The organic solvent was first evaporated under a gentle stream of nitrogen while gently turning the flask to form a thin lipid film onto the wall of the flask. The lipid film was further dried with nitrogen for at least 30 min. Lipid films were hydrated in a buffer containing 10mM Tris-Cl, pH 7.5 (Merck), 150mM NaCl at a final concentration of 2mg/ml. After vortexing for 30 minutes, the resulting multilamellar vesicle solutions were extruded 25 times through a polycarbonate membrane with 50 or 200 nm nominal pore diameter (LiposoFast, Avestin) leading to a stock solution of unilamellar vesicles which were stored at 4°C and used for 5 d at maximum.

Protein labeling for LLPS

The dyes used for labeling were Alexa Fluor™ 647 C2 Maleimide and Alexa Fluor™ 555 C2 Maleimide (ThermoFischer Scientific). The dyes were dissolved in water-free DMSO (dimethyl sulfoxide) at a concentration of 1mM. After the protein isolation with GST-affinity chromatography, the proteins were dialyzed against 20mM Tris-HCl pH 7.5, 150mM NaCl. Proteins were partially labeled in solution in a ratio of 1 mol protein to 0.1 mol (1:10 molar ratio labeled/unlabeled) of dye for 15 min on ice. The labeled 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 30min. The QCM-D experiments were performed using the Q-Sense E4 (Biolin, Q-Sense, Sweden) instrument and AT-cut quartz disks (5MHz). 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 (200μL, 0.2mg/mL protein solution). 5’-biotinylated, 3’-cholesterol modified DNA was then used (200μL, 0.075pmol/μL) as a binding anchor to liposomes of different lipid compositions as previously described (Milioni et al., 2020). All liposome solutions were used at a concentration of 0.2mg/ml and a final volume of 700μL. Finally, protein solutions of various concentrations were added at a volume of 500μL.

Supported Lipid bilayer with DGS-NTA(Ni)

Supported lipid bilayers containing 1% (v/v) DGS-NTA(Ni) were formed on SiO2 coated sensors (Q-Sense QSX 303) upon the addition of 0.05 mg/mL lipid solution. Final frequency and dissipation changes were Δf=-170 Hz and ΔD = 0.13 x 10^-6. These values are typical of the formation of a homogeneous supported lipid bilayer.

Liposome-protein binding assays

For the liposome-protein pull-down assays, we used 25 μl of Small Unilamellar Vesicles (SUVs, 2 mg/ml) combined with approximately 500ng of the protein of interest. Before binding we centrifuged the proteins at 16,000 × g, at 4°C, for 5 min, to avoid precipitates (Julkowska et al., 2013). Twenty-five μl of protein mix with 25 μL liposome suspension were combined and incubated on an orbital shaker platform for 30-45 min at room temperature. The binding buffer consisted of 150 mM KCl, 25 mM Tris–HCl pH 7.5, 1 mM DTT, 0.5 mM EDTA. The samples were centrifuged at 16,000g for 30 min at room temperature, we harvested the supernatant (we added 16.7 μL of 4× Laemmli sample buffer to a supernatant fraction) and resuspended the liposome pellet in 300-500 μl 1× binding buffer. The centrifugation was repeated in the same conditions and we resuspended the liposome-protein pellet in 33 μL 4x Laemmli buffer. All samples were incubated at 95°C for 5 min and loaded 10 μL on SDS-PAGE for Western blot.

Liposome DNA-mediated fusion assay and SUPER template setting

To establish DNA-zippers, we adjusted concentrations of oligomers-cholesterol and used various temperatures (4-25°C) to achieve DNA-zipper-driven fusion at a percentage lower than ∼10% (noise). Eventually, all liposomes irrespective of the conditions used, given enough time fused to a percentage of ∼50%. The liposomes used were 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)P2 0.5% (PS was included for SUPER template experiment to provide required membrane flexibility). We also used LUVs without PE-Texas Red lipid. Instead, they carried the fluorophore fluorescein (cf, 100 mM fluorescein in Tris-Cl pH 7.5). For the fusion of the two different liposomes populations, we incubated each with a DNA primer for 45 min at room temperature. The sequence of the single-stranded DNA molecules can be found in Table S1. For every liposome, we used 100 DNA molecules (Peruzzi et al., 2019). SUPER templates were prepared, as described previously (Pucadyil and Schmid, 2010). Silica beads (monodispersed 4.9 μm in diameter; Corpuscular Inc., 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 H2SO4/30% H2O2 4:1, v/v) for 1 h at room temperature and washed extensively with boiling water. The coverslips were allowed to attain room temperature, stored under water, 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 room temperature for 30 min with PEG-silane (2%, v/v; Gelest Inc., Morrisville, PA) in acetone. The coverslips were later washed extensively with water and air-dried before use. The samples were observed with confocal LIGHTNING SP8 module (high speed and resolution of 120 nm) using an observation chamber and an Apochromat 63x objective (N.A.=1.4). images were deconvoluted by Leica built-in software.

Phase separation assays

The slides and coverslips used for confocal microscopy were treated overnight with 1-2 mg/mL PLL-PEG, after cleaning with 2% (v/v) Hellmanex for approximately 2 h, at room temperature. 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 labeled protein was 1-2 μM for 3 h at room temperature. Some of our samples contained Large Unilamellar Vesicles (LUVs) consisting of DOPC 0.5%: DOPS 99%: DOPE-Texas Red 0.5%. Phase separation was imaged via confocal microscopy.

Whole-mount proximity ligation assay (PLA) and Immunocytochemistry

Four to five-day-old seedlings were fixed and permeabilized as described (Moschou et al., 2016). SFH8 antibody was raised against the unique peptide sequence 517-GNAIELGSNGEGVKEECRPPSPVPDLTET-545 in rabbits using standard immunogenic procedures. Primary antibody combination 1:200 for α-GFP mouse [Sigma-Aldrich, SAB2702197] and 1:100 for α-SFH8 rabbit, 1:100 for α-mNeon mouse [Chromotek, 32F6] and 1:100 for α-SFH8 rabbit or 1:200 for α-FLAG mouse [Sigma-Aldrich, F1804] and 1:200 for α-GFP rabbit [Millipore, AB10145]) were used for overnight incubation at 4°C. Roots were then washed with MT-stabilizing buffer (MTSB: 50 mM PIPES, 5 mM EGTA, 2 mM MgSO4, 0,1% (v/v) Triton X-100) and incubated at 37°C for 3 h either with α-mouse plus and α-rabbit minus for PLA assay (Sigma-Aldrich, Duolink). PLA samples were then washed with MTSB and incubated for 3 h at 37°C with ligase solution as described (Pasternak et al, 2018). Roots were then washed 2x with buffer A (Sigma-Aldrich, Duolink) and treated for 4 h at 37°C in a polymerase solution containing fluorescent nucleotides as described (Sigma-Aldrich, Duolink). Samples were then washed 2x with buffer B (Sigma-Aldrich, Duolink), with 1% (v/v) buffer B for another 5 min, and then the specimens were mounted in Vectashield (Vector Laboratories) medium.

Immunocytochemistry was done as described previously (Moschou et al., 2016). The primary antibodies used were rabbit α-SFH8 (diluted 1:500), rat α-tubulin YL1/2 (1:200; Santa Cruz Biotechnology), mouse α-FLAG (1:250), and sheep α-PIN2 (1:500). Specimens were washed three times for 90 min in PBST and incubated overnight with donkey α-sheep conjugated Alexa fluor 488, goat α-mouse tetramethylrhodamine isothiocyanate (TRITC), α-rat TRITC, and α-rabbit fluorescein isothiocyanate-conjugated (FITC) secondary antibodies diluted 1:200-250. After washing 3x in PBSΤ, specimens were mounted in Vectashield (Vector Laboratories) medium.

rBiFC and protoplast transformation

The rBiFC assay was done in Arabidopsis root protoplasts via PEG transformation as described previously (Jeong et al., 2021). After 16h transformation, protoplasts were observed with a Zeiss LSM780 laser scanning confocal microscope using 40×/1.2 W C-Apochromat in multi-track channel mode. Excitation wavelengths and emission filters were 514 nm/band-pass 530-550 nm for YFP, and 561 nm/band-pass 600-630 nm for RFP. Images were processed by Fiji.

Quantification of Fluorescent Intensity, FRAP and FRET

To create the most comparable lines to measure the fluorescence intensity of reporters in multiple mutant backgrounds, we crossed homozygous mutant bearing the marker with either a WT plant (outcross to yield progeny heterozygous for the recessive mutant alleles and the reporter) or crossed to a mutant only plant (backcross to yield progeny homozygous for the recessive mutant alleles and heterozygous for the reporter). 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, one bleach scan, and 96 post-bleach scans (or more). Bleaching was performed using 488, 514, and 561 nm laser lines at 100% transmittance and 20-40 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 2-10 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 x 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 20x with NA=0.8 M27 (Zeiss). GFP-AHA1, PIP2a-GFP, PIN2-GFP, mNeon-SFH8, and SFH8-mNeon fluorescence were detected using a water- or oil-corrected 40x objective. During analyses, FRAP mode of Zeiss Lsm780 ZEN software was set up for the acquisition of one pre bleach image, one bleach scan, and 40 or more post bleach scans. The following settings were used for photobleaching: 10 to 40 iterations; 10 to 60 s per frame; 100% transmittance with the 458- to 514-nm laser lines of an argon laser. Pre- and post bleach scans were at minimum possible laser power (1.4 to 20% transmittance) for the 488 nm and 0% for all other laser lines; 512x512 pixel format; and zoom factor of 5.1. The fluorescence intensity recovery values were determined, then the background values were subtracted from the fluorescence recovery values, and the resulting values were normalized against the first post bleach time point.

For FRET, either acceptor photobleaching (AB) or sensitized emission (SE) modules of the SP8 Leica confocal microscope were used, using standard modules and internal calibration. The acceptor photobleaching experiments were done as described previously (Moschou et al., 2016), using the aforementioned parameters for FRAP. Fluorescence was detected using a water- or oil-corrected 40x 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 (Lin et al., 2021).

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 with, 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 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 of FIJI or NanoTrackJ/TrackMate, 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.

LC-MS/MS Lipids and Proteins Analyses and Overlay Assays of Lipids

Lipid overlay assay PI strips were blocked in PBST (0.1% Tween 20 [vol/vol]) supplemented with 3% (w/v) fatty acid-free BSA (Sigma-Aldrich) overnight at 4°C and then incubated with 0.5 μg/ml GST fusion proteins for 1 h at room temperature. After washes with PBST-BSA, the membranes were incubated with rabbit α-GST for 1 h at room temperature, followed by IRDye 800 goat anti-rabbit secondary antibody for 1 h.

Lipids were extracted from roots after inactivating the tissue with boiling water, or from microsomes according to (Gasulla et al., 2013). 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 (Wewer et al., 2014).

For proteomic analyses of FLAG-tagged SFH8 immunoprecipitates, samples were analyzed by LC-MS using Nano LC-MS/MS (Dionex Ultimate 3000 578 RLSCnano System) interfaced with Eclipse Tribrid mass spectrometer (ThermoFisher 579 Scientific). Samples were loaded onto a fused silica trap column Acclaim PepMap 100, 580 75 μm x 2 cm (ThermoFisher 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 582 column (Nanoease MZ peptide BEH C18, 130 A, 1.7 μm, 75 μm x 250 mm, Waters) for 583 LC-MS/MS. Peptides were fractionated at 300 nL/min using a segmented linear gradient 584 4-15% B in 30min (where A: 0.2% formic acid, and B: 0.16% formic acid, 80% 585 acetonitrile), 15-25% B in 40min, 25-50%B in 44 min, and 50-90% 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 350–1600, automatic gain control (AGC) target 1E6, maximum injection time 100 ms). The top S (3 589 sec) and dynamic exclusion of 60 sec 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, automatic gain control (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 labeled with 5′-801 carboxytetramethylrhodamine at the N-terminus (TAMRA-PEP1) with an HPLC purity of 95.24% and molecular weight of 2905.24 (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 five 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×/1.2 W C-Apochromat M27 objective at a zoom factor of 3.5. TAMRA-PEP1 was excited at 559 nm, and fluorescence emission was captured between 570-670 nm, GFP was excited at 488 nm, and fluorescence emission was captured between 500-540 nm. Image quantification was done using Fiji software. To this end, the PM of individual cells was selected with the brush tool with a size of 10 pixels 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 plasma membrane fluorescence. Five epidermal cells from five plants were quantified.

The FM4-64 experiments were as described in Moschou et al., 2013. In brief pulse labeling with FM4-64 (2 μM FM464, Molecular Probes; made from a 2mM stock in DMSO) was done for 5 min (time 0) and then analyzed for 18 min at room temperature in Zeiss Lsm 780 with excitation at 488 nm and fluorescence emission captured between 540-670 nm. After the pulse with FM4-64, the roots were washed two times in an ice-cold MS medium. Quantifications were done as described for the TAMRA-PEP1 experiments above.

RNA Extraction, RNAseq, and Quantitative RT-PCR Analysis

Total RNA from the seedlings was extracted using the RNeasy Plant Mini Kit with DNaseI digestion (QIAGEN). Reverse transcription was carried out with 500 ng of total RNA using the iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer’s protocol. Quantitative PCR with gene-specific primers was performed with the SsoAdvanced SYBR Green Supermix (Bio-Rad) on a CFX96 Real-Time PCR detection system (BioRad). Signals were normalized to the reference genes ACTIN2 using the DCT method and relative expression of a target gene was calculated from the ratio of test samples to Col-0. For each genotype, three biological replicates were assayed in three qPCR replicates. qRT-PCR primers were designed using QuantPrime (www.quantprime.de)(Arvidsson et al., 2008). Primer sequences used for qRT-PCR can be found in Table S1.

Super-resolution Imaging, Transmission Electron Microscopy, and Negative Staining

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 63x water immersion objective with a theoretical x/y-axial 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 room temperature in an inverted microscope setting. Root samples were fixed in a mixture of 1.25% (v/v) glutaraldehyde and 1.25% (v/v) paraformaldehyde in 50 mM cacodylate buffer, pH 7.2, for 1 h. After washing in cacodylate buffer, samples were postfixed overnight in 1% osmium tetroxide (prepared in cacodylate buffer), dehydrated in a graded ethanol series, and embedded in London Resin White. Ultrathin sections were examined with a transmission electron microscope at 80 kV without post staining. Negative staining was essentially done as described for lipoproteins in (Rames et al., 2014).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was performed in R studio (R-project.org) or GraphPad Prism (version 9.2.0). Each data set was tested whether it followed a 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’s 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 (very low values are indicated as <0.0001). For tests involving multiple comparisons, 1-way ANOVA or the Kruskal-Wallis test (nonparametric analog of ANOVA) was used followed by Dunnet’s or Dunn’s multiple comparison test 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 truncated or untruncated for aesthetic reasons.

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 a region of interest. The dwell time rate of tagged proteins in FRAP experiments was calculated by the single exponential fit equations as described by (Chang et al., 2005). Co-localization was analyzed using Pearson statistics (Spearman or Manders analyses produced similar results)(French et al., 2008). 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 (Deli et al., 2022). 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).

AUTHOR CONTRIBUTIONS

P.N.M, conceptualization, project administration, supervision, formal analysis, writing-original draft, funding acquisition, complex identification; C.L., A.M., A.D., F.P., A.P., P.R., L.C., investigation, formal analysis, data curation; S.B., P.D., Y.J., E.R., E.G., and G.S., methodology; A.D., F.P., and E.G., methodology and establishment of liposome assays, curation; C.L., P.N.M, and A.M., established advanced imaging approaches, visualization. All authors reviewed, edited, and approved the paper.

DECLARATION OF INTERESTS

The authors declare no competing interests.

SUPPLEMENTAL MATERIALS

Supplemental Figures 1-7 (legends below figures) Supplemental Movies 1-4:

1. TIRFM of a diffusive cluster following a directional track or non-directional track (part a and b)

2. Non-diffusive clusters with low dwelling times

3. Immobile patches of SFH8 atop short filaments resembling the “beads on a string” localization

4. SFH8^IDR puncta showing dynamic morphology with frequent splitting, fusion, and interconnections resembling liquid-liquid phase-separated (LLPS) condensates

Supplemental Files 1-3

1. SFH-like proteins alignments

2. IDR region conservation in various SFH-like proteins

3. Overall steady-state lipid PM and total composition of various mutants used in this study Supplemental Table 1: oligos used in the study

Supplemental Information

See below

Supplemental Information

Localization of Kin7.3 interacting proteins in stable or transient expression

We expressed PP2C, Spc25, and POR in stable Arabidopsis Col-0 lines or N. benthamiana. The localization of these proteins is shown (Figure 1). Furthermore, we identified homozygous mutants of these proteins, showing a lethal phenotype for Spc25 and POR mutants confirming previous data (Shin et al., 2018; Steinborn et al., 2002).

The phosphatase PP2C (AP2C1), an Arabidopsis Ser/Thr phosphatase of type 2C, is a stress signal regulator that inactivates the stress-responsive MAPKs MPK4 and MPK6 (Bequette et al., 2018; Schweighofer et al., 2007). Mutant ap2c1 plants produce significantly higher amounts of jasmonate upon wounding and are more resistant to phytophagous mites (Tetranychus urticae). Plants with increased AP2C1 levels display lower wound activation of MAPKs, reduced ethylene production, and compromised innate immunity against the necrotrophic pathogen Botrytis cinerea. Phosphorylation by a MAPK module regulates Kin7 superfamily members Hinkel and Tetraspore, which impinge on cell plate formation (Takahashi et al., 2010).

The Spc25 is part of the centromere assembly complex. Kin7.3 is a centromeric protein-E homolog and thus this interaction as well might be biologically relevant especially during cell division. Spc25, is a component of the NDC80 complex, in mouse oocytes, an outer kinetochore complex comprising four subunits (ndc80/Hec1, nuf2, spc24, and spc25)(Xiong, 2011). The NDC80 constitutes one of the core MT-binding sites within the kinetochore. One hypothesis would be that KISC during cell division is involved in MT stabilization on the centromeres. Among the components of the outer kinetochore complex, the four proteins in the NDC80 complex, including NDC80 [nuclear division cycle 80, also known as Hec1 (highly expressed in cancer1) in humans], NUF2 (nuclear filament-containing protein 2), SPC24 (spindle pole body component 24) and SPC25, play critical roles in connecting spindle fibers to chromosomes. In vertebrates, the C-terminal ends of the NUF2-NDC80 heterodimer associate with the N-terminal coiled-coil domains of the SPC24-SPC25 heterodimer (Suzuki et al., 2016). Globular dimeric heads, containing the RWD (RING finger, WD repeat, DEAD-like helicases) domain of the SPC24-SPC25 dimer, bind to the inner kinetochore components, the KNL1-Mis12 complex. Therefore, the NDC80 complex serves as a central ‘hub’ connecting the kinetochore complexes, which provides the ‘bridges’ between the inner kinetochore complexes and the MTs of the spindle.

The PORCINO protein is a tubulin folding cofactor and its absence results in embryo lethality (Steinborn et al., 2002). The importance of Kin7.3 interaction with PORCINO (POR) remains to be established. The TFC C ortholog, PORCINO belongs to the four PILZ group genes encoding orthologs of mammalian tubulin-folding cofactors (TFCs) C, D, and E, and associated small G-protein Arl2 that mediate the formation of α/β-tubulin heterodimers in vitro. PORCINO was detected in cytosolic protein complexes and did not colocalize with MTs. Consistently, with previous reports, the por mutant was embryo lethal and the POR localized in the cytoplasm.

Figure 1. Localization of Kin7.3 interactors and putative interactions of Kin7.3.

(A) Localization of SPC25, SFH8, and PORCINO tagged N-terminally with tagRPF in N. benthamiana transient system. Micrographs are representative of an experiment replicated three times results. Scale bars, 20 μm.

Cleavage of Cohesin Subunit SYN4 by Separase

Sister-chromatid cohesion depends on cohesin, a 4-subunit protein complex that links the sisters as they are synthesized in the S phase. In anaphase when the protease separase cleaves the Scc1/ Mcd1 subunit of cohesin (or its Rec8 counterpart in meiotic cells), thereby allowing the sisters to be pulled apart by the mitotic spindle. The Arabidopsis genome encodes four kleisin subunits: the meiosis-specific ones SYN1 and SYN3, involved in gene expression of meiotic genes, but also expressed in somatic cells, and SYN2 and SYN4, which have been suggested to participate in mitotic cell division (Minina et al., 2017). As we have shown previously, plants expressing tagged SYN proteins, except the mitotic cohesin SYN4, were highly susceptible to even mild environmental changes and showed decreased fertility (Minina et al., 2017).

Principle of R2D2 Probe

R2D2 shows a diminishing green but not red fluorescence when auxin levels are high, as the conserved domain II (DII) marker (pRPS5A-driven DII fused to n3×Venus) is rapidly degraded in response to auxin. On the contrary, the pRPS5A-driven mutated DII-ntdTomato (red fluorescence) is not responding to auxin, thereby allowing the ratiometric fluorescence quantification of auxin response levels through the R2D2 (Liao et al., 2015).

ACKNOWLEDGMENTS

Funding for this work was through VR Research Council Grant 298264-2015, FORMAS Research Council Grant MOP-86675, HFRI-Always Strive for Excellence 1624, through the IMBB-FORTH Start-Up Funding (to P.N.M), and by the Deutsche Forschungsgemeinschaft (SCHA 1274/5-1, Germany’s Excellence Strategy EXC-2070-390732324 PhenoRob to G.S.).