Biomolecular condensates mediate bending and scission of endosome membranes

Plant materials and growth conditions

A. thaliana free1 mutant (a T-DNA insertion line) seeds and UBQ10::GFP–FREE1 transgenic seeds were provided by L. Jiang. All other transgenic A. thaliana lines were generated by transformation of corresponding constructs into the heterozygous free1 mutant background and subsequent genotyping of homozygous free1 from progeny. The vps2 mutant (a T-DNA insertion line) seeds were provided by Y. Cheng. The RHA1–mCherry marker line was gifted by L. Qu. Seeds were surface sterilized by ethanol and sown on half-strength Murashige and Skoog (MS) medium with 1.0% sucrose, 0.4% phytagel at pH 5.8. MS was supplemented with NaCl or sorbitol as indicated. Plate media were stratified at 4 °C for 2 days and transferred to a growth chamber under a long-day (16 h–8 h light (22 °C)–dark (18 °C)) photoperiod.

Chemical and osmotic seedling treatments

Wortmannin-treated seedlings were prepared by soaking for 45 min in liquid MS medium with wortmannin (Selleck, S2758) added at a final concentration of 33 μΜ from a 33 mM stock solution in dimethyl sulfoxide. For hexanediol treatment, 1,6-hexanediol was added at a final concentration of 4% (w/v) to MS medium with wortmannin (as above) and incubated for 5 min before imaging. For washout experiments, seedlings were incubated in half-strength MS medium with wortmannin (as above) without 1,6-hexanediol for 1 h before imaging.

For acute NaCl and sorbitol treatments, seedlings were soaked in liquid MS medium supplemented with 125 mM NaCl or 300 mM sorbitol for the indicated times before imaging. To measure germination and seedling survival rates, seeds were sown on MS plate medium supplemented with 125 mM NaCl or 300 mM sorbitol and grown at 22 °C under a long-day (16 h–8 h light (22 °C)–dark (18 °C)) photoperiod for 10 days.

Plasmid construction

To generate the constructs for in vitro protein expression, the coding sequences of FREE1 and FREE1 variants (FREE1(ΔIDR), FUS-IDR–FREE1, PTAP–FUS-IDR–FREE1, FUSm-IDR–FREE1 and FLOE1-IDR–FREE1) were amplified and inserted into the pET11-6×His-MBP⁵³ or pET11-6×His vector. The coding sequences of VPS23A, VPS28A, VPS37A, TOL6, TOL9, LIP5 and BRO1 were amplified and inserted into the pRSFduet-6×His-mCherry vector. All cloning was performed using the ClonExpress II One Step Cloning kit (Vazyme, C112).

To generate constructs for transient expression in A. thaliana protoplasts, the coding sequences of FREE1 and FREE1 variants (FREE1(ΔIDR), FUS-IDR–FREE1), and RHA1 were amplified and inserted into modified pBI221 vectors containing a GFP tag or an mCherry tag, respectively.

For complementation constructs, an approximately 1 kb FREE1 promoter region was amplified and inserted into the pCambia1300-N1-GFP vector to generate the pCambia1300-pFREE1-GFP construct. Moreover, the coding sequences of FREE1 and FREE1 variants (FREE1(ΔIDR), FUS-IDR–FREE1, PTAP–FUS-IDR–FREE1, FUSm-IDR–FREE1 and FLOE1-IDR–FREE1) were amplified and inserted into pCambia1300-pFREE1-GFP. All of the resulting constructs were introduced into Agrobacterium tumefaciens strain GV3101 and transformed into the free1 heterozygous mutant using the floral dip method. In brief, A. tumefaciens GV3101 cells were collected and resuspended in a 5% sucrose solution containing 0.02% (v/v) silwet L-77. A. thaliana flower buds were dipped into the A. tumefaciens suspension for 2 min and subsequently kept in the dark overnight before being transferred to the growth chamber.

For the b-isox precipitation assay, proteins were transiently expressed in Nicotiana benthamiana. The coding sequences of VPS23A, VPS28A, VPS37A, TOL6, TOL9, LIP5 and BRO1 were amplified and inserted into pCambia1300-N1-Flag vector. All of the constructs were introduced into the A. tumefaciens strain GV3101 and infiltrated into N. benthamiana leaves.

For yeast two-hybrid constructs, the coding sequences of FREE1 variants and VPS23A were amplified and inserted between the EcoRI and BamHI restriction sites of pGADT7 and pGBKT7 using the ClonExpress II One Step Cloning kit (Vazyme, C112). A list of all of the coding and primer sequences is provided in Supplementary Table 1.

In vitro protein expression and purification

All fusion proteins were expressed and purified from Escherichia coli (Rosetta) cell extracts on a Ni-NTA column. In brief, protein expression was induced by 0.4 mM isopropyl-β-d-1-thiogalactopyranoside at 18 °C overnight. Cells were collected by centrifugation and resuspended in lysis buffer (40 mM Tris-HCl pH 7.4, 500 mM NaCl, 10% glycerol). The suspension was sonicated for 30 min (2 s on, 4 s off, SCIENTZ) and centrifuged at 13,000g for 30 min at 4 °C. The supernatant was incubated with Ni-NTA agarose for 20 min, washed and eluted with 40 mM Tris-HCl pH 7.4, 500 mM NaCl and 500 mM imidazole (Sangon). Proteins were further purified by gel-filtration chromatography (Superdex-200; GE Healthcare) and stored in 40 mM Tris-HCl pH 7.4, 50 or 500 mM NaCl, 1 mM dithiothreitol on ice or at −80 °C.

In vitro phase-separation assay

The solubility tag MBP was cleaved using TEV protease for 30 min to induce LLPS. Protein concentrations were determined by measuring absorbance at 280 nm using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). Proteins were diluted to various salt (50–500 mM NaCl) and protein (0.05–20 μΜ) concentrations in 384-well plates (Greiner Bio One, 781090) and observed using the Zeiss LSM880 confocal laser-scanning microscope equipped with a ×63 objective.

To quantify the partition coefficient of VPS proteins by FREE1 condensates, regions of interest of equivalent size were analysed in the Fiji implementation of ImageJ to calculate the VPS signal intensity inside and outside FREE1 condensates. The partition coefficient was defined as intracondensate fluorescence intensity divided by the fluorescence intensity of the extracondensate solution.

FRAP analysis

FRAP analysis was performed on the Olympus Fluoview FV-1000 confocal laser-scanning microscope. Condensates were photobleached using a 488 nm Ar-laser pulse at maximum intensity. Time-lapse recordings of intensity changes were analysed using the Fiji implementation of ImageJ.

Transient expression in A. thaliana protoplasts

A. thaliana mesophyll protoplasts were isolated from 3-week-old Col-0 plants as previously described⁵⁴. In brief, leaf slices from the middle part of a leaf were incubated in enzyme solution (20 mM MES pH 5.7, 1.5% (w/v) cellulase R10, 0.4% (w/v) macerozyme R10, 0.4 M mannitol and 20 mM KCl) for 3 h. The suspension was centrifuged and washed twice with W5 solution (2 mM MES pH 5.7, 154 mM NaCl, 125 mM CaCl₂ and 5 mM KCl). Plasmids were transformed into protoplasts using the PEG-calcium-mediated method (40% PEG3350). The transfected protoplasts were incubated at 22 °C for 16 h in WI solution (0.5 M mannitol, 4 mM MES pH 5.7 and 20 mM KCl) and observed under the Zeiss LSM880 confocal laser-scanning microscope equipped with a ×63 objective. The colocalization coefficient was defined as the ratio of VPS23 fluorescence intensity relative to FREE1 fluorescence intensity in the same condensate. Data analysis was performed in the Fiji implementation of ImageJ.

b-isox precipitation

The precipitation of ESCRT proteins by b-isox was performed as described previously. In brief, 0.5 g of fine powder ground from tobacco leaves was lysed in extraction buffer (10 mM Tris-Cl pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 20 mM β-mercaptoethanol, 0.5% NP-40; 10% glycerol, 1× protease inhibitor cocktail). The cell lysate was then clarified by centrifugation. Then, 500 µl protein extract was incubated with 30 µM b-isox at 4 °C for 1 h and centrifuged. The precipitate was washed twice with ice-cold extraction buffer and boiled in SDS loading buffer (100 mM Tris-HCl pH 7.4, 4% SDS, 0.2% bromophenol blue, 20% glycerol, 5% β-mercaptoethanol). For precipitation of FREE1, ten-day-old A. thaliana Col-0 seedlings were used. For precipitation of other ESCRT proteins, each Flag-tagged protein was transiently expressed and resulting tobacco cells were used for precipitation.

Immunoblot analysis

To prepare total protein extracts, A. thaliana seedlings were ground in liquid nitrogen and lysed with extraction buffer (40 mM HEPES-KOH at pH 7.5, 1 mM EDTA, 10 mM KCl, 0.4 M sucrose with 1 × cOmplete Protease Inhibitor Cocktail). Debris was removed by centrifugation at 12,000g for 30 min. Protein samples were separated by 10% SDS–PAGE and transferred to a polyvinylidene difluoride membrane. Primary antibodies were used against FREE1, Flag (Merck, F1804, 1:2,000), GFP (Roche, 11814460001, 1:7,000), tubulin (Sigma-Aldrich, T5168, 1:2,000), ubiquitin (Santa Cruz Biotechnology, sc-8017; 1:1,000) or His-tag (Sangon, D110002, 1:1,000). The horseradish peroxidase (HRP)-conjugated secondary antibodies goat anti-Mouse (CWBIO, CW0102, dilute at 1:10,000) and goat anti-rabbit (CWBIO, CW0103, 1:10,000) were used for protein detection by chemiluminescence (ChemiDoc, LAS4000).

Confocal microscopy

Confocal imaging was performed using the Zeiss LSM880 or Olympus Fluoview FV-1000 microscope equipped with ×63/1.4 NA oil and ×60/1.2 NA water-immersion objectives. GFP was excited at 488 nm and detected between 490 nm and 530 nm or between 500 nm and 544 nm. mCherry and DiIC18 were excited at 561 nm and detected at 579–650 nm or 570–670 nm.

Cell culture and transfection

Maintenance of cell lines and transfection were performed as described previously⁵⁵. COS-7 and HEK293T cells were maintained in DMEM supplemented with 10% FBS at 37 °C in 5% CO₂. Transfection was performed using X-tremeGENE HP (Roche, 28088300) according to the manufacturer’s protocols.

Lentiviral transduction

For lentiviral transduction, HEK293T cells were transfected with pFUGW (together with VSVG and psPAX2 plasmids). Viruses were harvested at 60–72 h post transfection, and the viral supernatant was centrifuged at 600 g for 5 min to remove cell debris. The indicated cells were infected with the viral supernatant diluted with fresh medium (30% viral supernatant) containing 10 µg ml^–1 polybrene.

Immunofluorescence microscopy and quantification

After being infected with indicated lentiviruses for 72 h, the cells were fixed with 4% paraformaldehyde for 15 min at room temperature and used for imaging. For super-resolution microscopy, imaging experiments were performed using the Nikon combined confocal A1/SIM/STORM system with four excitation/imaging lasers (405, 488 and 561 nm from Coherent, 647 nm from MPBC) and a CFI Apo SR TIRF ×100/1.49 NA oil-immersion objective. Images were acquired using an Andor EMCCD camera (iXON 897). Data analyses were performed using the NIS-Elements AR (Nikon) software. All SIM images were acquired as z-stack images.

TEM and immuno-gold labelling in mammalian cells

COS-7 cells were fixed with 2.5% glutaraldehyde for 1 h at room temperature and washed three times (15 min each) with 0.1 M PB. Post-fixation staining was performed with 1% osmium tetroxide (SPI, 1250423) for 30 min on ice. Cells were washed three times (for 15 min each) with ultrapure water and placed in 1% aqueous uranyl acetate (EMS, 22400) at 4 °C overnight. The samples were then washed three times (15 min each) with ultrapure water, dehydrated in a cold-graded ethanol series (50%, 70%, 80%, 90%, 100%, 100%, 100%; 2 min each), and infiltrated with EPON 812 resin using 1:1 (v/v) resin and ethanol for 8 h, 2:1 (v/v) resin and ethanol for 8 h, 3:1 (v/v) resin and ethanol for 8 h, pure resin 2 × 8 h. After a final infiltration with fresh resin, the samples were polymerized at 60 °C for 48 h. Embedded samples were sliced into 75-nm-thick sections and stained with uranyl acetate and lead citrate (C1813156) before imaging on the HT-7800 120 kV transmission electron microscope (Hitachi High-Technologies).

Immuno-gold labelling was performed as described previously⁵⁶ with some modifications. In brief, COS-7 cells were incubated in 2% paraformaldehyde and 0.01% glutaraldehyde in PB buffer at 4 °C overnight and then washed with chilled PB/glycine. Cells were next scraped from the bottom of plastic dishes into 1% gelatine, centrifuged at 1,000 rpm for 2 min and resuspended in 12% gelatine at 37 °C for 10 min. The gelatine–cell mixture was then solidified on ice for 15 min. Small blocks (about 0.5 mm³) were cut and immersed in 2.3 M sucrose overnight at 4 °C. Then, 70-nm-thick cryosections were prepared at −120 °C with an ultramicrotome (Leica, EM FC7). After the sections were thawed at room temperature, immunolabelling was performed using rabbit anti-GFP antibodies followed by the immune-gold secondary antibody. The sections were then treated with methyl cellulose/uranyl acetate and subsequently imaged using the HT-7800 120 kV transmission electron microscope (Hitachi High-Technologies).

Expression of FREE1 in Saccharomyces cerevisiae

FREE1 ORFs were amplified and inserted into a pRS416 yeast expression vector containing the GAP1 promotor sequence (pAM199). Vectors were transformed into SEY6210 wild-type cells (yAM007) and grown in SDCA medium (0.17% yeast nitrogen base without amino acids and ammonium sulphate, 0.5% ammonium sulphate, 0.5% casamino acids, 2% glucose, 100 μM l-histidine, 100 μM l-tryptophan) overnight. Cells were then diluted into fresh SDCA medium and grown for 4 h. Before imaging, vacuolar membranes were stained for 15 min in medium containing 160 μM FM4-64 dye, washed twice and chased for a further 30 min in fresh medium. Imaging was performed on the SpinSR10 spinning-disc confocal microscope, and the recorded images were deconvoluted using the Cellsens software (Olympus).

Dot blot for PI3P-binding assay

PI3P lipids (Avanti Polar Lipids, 850187P) were spotted onto nitrocellulose membranes (Millipore) at the indicated concentrations (2%, 200 pmol; 1%, 100 pmol; 0.5%, 50 pmol; 0.1%, 10 pmol; 0.05%, 5 pmol). These membranes were then blocked in 5–10 ml blocking buffer (3% BSA in PBS-T buffer) for 1 h at room temperature, incubated with 10 nM of the indicated GFP-tagged protein (in blocking buffer) for 1 h at room temperature and washed three times (5 min each) with PBS-T buffer. GFP fluorescence was detected using the ChemiDoc imaging system (Bio-Rad).

Membrane flotation assay

SUVs were prepared as described previously⁵⁵. In brief, lipids were mixed as POPC:POPS:cholesterol:PtdIns(3)P:PE at a molar ratio of 62:10:25:1:1 in trichloromethane. The lipids were dried under a nitrogen stream and further dried for 1 h at 37 °C. The lipid film was then hydrated using Tris buffer (40 mM Tris HCl, pH 7.4, 150 mM NaCl) and subjected to 10 cycles of freezing in liquid nitrogen and thawing in a 42 °C water bath. Liposomes were extruded through a 400 nm pore size polycarbonate film to produce the SUVs. Proteins (100 nM) were then added to the SUV solution and incubated for 5 min at room temperature. To remove unbound protein, a membrane flotation procedure was then performed. For each 120 μl SUV–protein solution, 480 μl 50% OptiPrep was added. The mixture was overlaid with 480 μl 30% OptiPrep and 90 μl Tris buffer (40 mM Tris HCl, pH 7.4, 50 mM NaCl). After centrifugation at 100,000g for 2 h, 60 μl of the top fraction containing the protein-bound SUVs was collected. To ensure equal loading of SUVs, 1 μl of the top fraction was used to measure the PC concentration using the Phospholipid C Kit (Wako, 433-36201). The top fraction as well as the total fraction were boiled in SDS loading buffer and used for immunoblot analyses.

Yeast two-hybrid assay

The yeast two-hybrid assay was performed using the Matchmaker Gold Y2H system according to the manufacturer’s instructions (Clontech). In brief, the constructs were co-transformed pairwise into yeast strain AH109 and cultured on SD/−Trp−Leu medium for 3 days. The interaction was analysed by spreading transformed cells (10³â€“10⁷ cells per ml) onto selective SD/−Trp−Leu−His−Ade medium supplemented with 5 mM 3-amino-1,2,4-triazole (3-AT).

TEM analysis of plant samples

Observation of ILVs by TEM was performed as previously described. In brief, root tips from five-day-old A. thaliana seedlings or germinated seeds were subjected to high-pressure freezing (EM PACT2, Leica) and substituted by acetone containing 0.4% uranyl acetate at −85 °C overnight in an AFS freeze-substitution unit (Leica). Next, the samples were infiltrated with increasing concentrations of HM20 (33–66–100%), embedded and ultraviolet polymerized for 48–72 h. Ultrathin sections were cut on the Leica UC7 ultramicrotome. Ultrathin section on grids were observed using the 80 kV Hitachi H-7650 transmission electron microscope (Hitachi High-Technologies) equipped with a charge-coupled device camera.

GUV preparation

GUVs were generated by the electroformation method⁵⁷ with modifications⁵⁸. The lipid mixture contained 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC, Avanti Polar Lipids) and 1,2-dioleoyl-sn-glycero-3-phospho-(1′-myo-inositol-3′-phosphate) (PI(3)P, Avanti Polar Lipids,) at a 9:1 molar ratio. Vesicles were fluorescently labelled by addition of 0.4 mol% membrane dye (DiIC18, Sigma-Aldrich). Lipid films were deposited onto indium-tin-oxide-coated glass plates (PGO) using glass syringes and 4 mM lipid stocks in chloroform, maintained for 1 h under vacuum and subsequently assembled into a chamber with a 2 mm Teflon spacer. The chamber was held together by binder clips, filled with 170 mM sucrose solution (Osmomat 3000 basic, Gonotec) connected to a function generator and an alternating current of 3.5 V and 10 Hz frequency was applied for approximately 2 h at 30 °C. GUVs were carefully collected, concentrated by sedimentation following a 1:10 dilution in isosmotic glucose solution and stored at room temperature until use.

GUV wetting and ILV formation assays

FREE1 condensates were generated by TEV cleavage of MBP from FREE1-MBP in 96-well high-content imaging plates coated by applying and drying a 1% polyvinyl alcohol (146–186 kDa, Sigma-Aldrich) solution in water before use. To quantify FREE1 condensate–GUV contact angles, GUVs were added 1–2 min after micrometre-sized FREE1 condensates had formed in an iso-osmotic solution (40 mM Tris-HCl pH 8.0, 50 mM NaCl, 1 mM dithiothreitol). Confocal sections of condensate-GUV contacts were used to measure all three angles at the three-phase contact line and to compute the Young’s contact angle (Extended Data Fig. 7b). For membrane fission assays, GUVs and the isosmotic TEV solution were first mixed before addition of FREE1–MBP and immediate imaging by confocal microscopy (Olympus Fluoview FV-1000) as described above.

Computer simulations

The computational model incorporates three fluid phases: a fluid inside the vesicle and two immiscible fluids surrounding the vesicle. The vesicle was modelled as a fluidic inextensible membrane with bending stiffness. The model includes the fluid motion of each of the three fluid phases coupled to the motion of the membrane. Membrane mechanics are governed by out of plane bending stiffness κ and in-plane stretching elasticity. To solve the coupled physical system, we developed a numerical model based on the combination of the three-phase fluid model for wetting⁵⁹ and the elastic shell model⁶⁰ (further details are provided in Supplementary Note 1).

The stretching resistance of the membrane was assumed to be large enough that the membrane stretches locally less than 3%. An initially elongated membrane (Extended Data Fig. 7d) was put in contact with a half circular condensate with surface tension σ_βγ (Supplementary Note 1), where the excess membrane area was set to allow for the condensate to be fully enclosed. Owing to the capillary forces of the condensate, both the membrane and droplet mutually remodel. The simulation model captures the full interaction of membrane bending stiffness, fluid motions and wetting deformations, enabling it to characterize intermediate shapes and time scales of the process.

Theory of condensate-mediated membrane scission

To understand the conditions under which membrane necks close, the stability of such necks and the timing of their scission, the energy of the system E(R_ne), which depends on the radius of the neck R_ne, was considered. This energy includes contributions corresponding to the bending energy of the membrane (of which the properties are affected by the presence of the condensate), the wetting and surface tension energies of the condensate and the line tension of the three-phase contact line (Supplementary Note 2). The force on the closed neck was calculated as \(f={\left.\frac{{\rm{d}}E\left({R}_{{\rm{ne}}}\right)}{{\rm{d}}{R}_{{\rm{ne}}}}\right|}_{{R}_{{\rm{ne}}}=0}\). For f < 0, the force acts to expand the neck, giving rise to either lens-like morphologies or endocytosis-like morphologies with wide necks (Fig. 4a). For 0 ≤ f ≤ f∗, the force constricts the neck into a closed form, but is not large enough to induce scission. Finally, for f > f∗ ≈ 25 pN, strong constriction causes scission of the neck.

Statistics and reproducibility

Sample sizes are chosen as widely used in the field. Biological and technical replicates were performed as described in the Methods for each experiment and conform to standards in the field. Exact n numbers for each experiment are provided in each figure legend. No data were excluded from analysis. Randomization of samples was performed. Seedlings from different plates were collected. Blinding was not deemed necessary in our study because we made no a priori assumptions on the response of the different samples to the experimental treatment, samples were all treated in parallel and all samples treated were always measured. Cells used in this study have been tested negative for mycoplasma contamination.

Reporting summary

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