Membrane surfaces regulate assembly of ribonucleoprotein condensates

Abstract

Biomolecular condensates organize biochemistry, yet little is known about how cells control the position and scale of these structures. In cells, condensates often appear as relatively small assemblies that do not coarsen into a single droplet despite their propensity to fuse. Here, we report that ribonucleoprotein condensates of the glutamine-rich protein Whi3 interact with the endoplasmic reticulum, which prompted us to examine how membrane association controls condensate size. Reconstitution revealed that membrane recruitment promotes Whi3 condensation under physiological conditions. These assemblies rapidly arrest, resembling size distributions seen in cells. The temporal ordering of molecular interactions and the slow diffusion of membrane-bound complexes can limit condensate size. Our experiments reveal a trade-off between locally enhanced protein concentration at membranes, which favours condensation, and an accompanying reduction in diffusion, which restricts coarsening. Given that many condensates bind endomembranes, we predict that the biophysical properties of lipid bilayers are key for controlling condensate sizes throughout the cell.

Extended Data Fig. 1. Quantifying Whi3-ER co-localization.

(a-d) Images of an Ashbya hypha expressing Whi3-tdTomato (tagged endogenously) and ER marker Sec63-GFP (plasmid expression). Images show merged channels (a), detected Whi3 puncta after particle detection (b), the same detected puncta overlaid with the ER channel (c), and a random distribution of the same number of puncta overlaid with the ER channel (d). (e) Local intensity in the ER channel at the Whi3 puncta shown in the above images, expressed as a fraction of the median intensity of the ER channel throughout the hypha after masking and background subtraction. Blue points: observed Whi3 positions, orange points: randomized Whi3 positions corresponding to the puncta in image (d). n = 16 puncta. Dashed line indicates the threshold for co-localization, corresponding to the median intensity of the ER channel. Black horizontal bars represent mean, vertical bars represent first s.d. p-value from two-tailed, unpaired Student’s t-test. (f) Ratio of the average local intensity in the ER channel at detected Whi3 puncta within the above hypha relative to randomized Whi3 puncta positions. A value greater than one indicates that the local intensity within the ER channel is greater on average at detected Whi3 puncta compared to randomized puncta. Each data point represents the ratio to one of n = 50 random distributions. The mean, indicated by the black horizontal bar, corresponds to one of the 60 data points in Fig. 1c. Vertical bar represents first s.d.

Extended Data Fig. 2. Tracking Whi3 puncta and ER co-localization.

(a) Time-lapse montages of Whi3 puncta associated with ER, including ER tubules (yellow arrowheads) and nuclear-associated ER (blue arrowheads). White dashed lines in first frame indicate cell periphery. Similar to Fig. 1d. (b) First frame from time-lapse shown in Supplementary Video 1. Yellow arrows indicate puncta that appear co-localized with the ER but moved out of the imaging plane during the movie and were not included in the tracking. (c) ER channel from the image in (b) with overlaid Whi3 tracks, colored according to the fraction of the track lifetime spent co-localized with the ER. All tracks clearly co-localize with ER structures. Not all tracks begin in the indicated frame. (d) Relative, local intensity in the ER channel as a function of time for the Whi3 tracks shown in (c), expressed as a fraction of the median intensity in the ER channel throughout the cell. Values greater than one (red region) were defined as co-localized with the ER. In this example, all tracks spend 100% of the lifetime co-localized with the ER. (e) Histogram of the tracks in (c-d), binned according to the fraction of track lifetime co-localized with the ER (similar to Fig. 1f). n = 7 tracked Whi3 puncta in this representative example from 60 hyphae. (f) Average intensity of all tracked Whi3 puncta as a function of the fraction of the track lifetime spent associated with the ER. Data points show moving average of the raw data, with lifetime fraction increments of 0.2. Data are mean ± 95% c.i. n = 83, 37, 45, 69, and 535 tracks in bins centered at 0.1, 0.3, 0.5, 0.7 and 0.9, respectively.

Extended Data Fig. 3. Particle detection and single molecule calibration.

(a) TIRF image of 50 pM Whi3-Atto488 on a plasma-cleaned glass coverslip. Right image shows detected puncta from cmeAnalysis software. (b) Histogram of detected puncta intensities, obtained from fits to a two-dimensional Gaussian function with standard deviation equal to the microscope PSF. The indicated peak value (red arrow) was taken as the average intensity of a single Whi3 protein. n = 3,682 puncta from a representative example of 10 independent experiments. (c-d) This single molecule intensity estimate was validated using photobleaching measurements. (c) TIRF time-lapse of 50 pM Whi3-Atto488 puncta on glass at the indicated times. Images were acquired with the same TIRF angle, laser power, and camera exposure settings used for acquisition of single molecule calibration images in (a-b). (d) Average, background-subtracted peak intensities of the puncta in the indicated colored circles in (c). Each puncta bleaches to the level of the camera background in a single step, indicating that each puncta corresponds to a single Whi3-Atto488 protein. The average, pre-bleach intensity of each puncta was comparable to our estimate of the single molecule intensity from particle detection in (b).

Extended Data Fig. 4. Membrane binding drives Whi3 condensate assembly.

(a) Time-lapse images of SLBs composed of DOPC alone (top row) or 90 mol% DOPC + 10 mol% DOPS (lower row) at the indicated times after addition of 50 nM Whi3. Images show that Whi3 puncta non-specifically interact with SLBs, but macroscopic condensates do not form. (b) FRAP reveals rapid and complete recovery of the fluorescent lipid Texas Red (TR)-DHPE in SLBs. Plot shows normalized lipid intensity within bleached region as a function of time after bleaching. Black line shows fit to single-component exponential recovery model, with recovery time constant (τ) indicated. Data are mean ± first s.d., n = 4 bleached regions from 1 SLB. (c) Confocal section (top) and maximum intensity projection (bottom) of membrane-associated condensates assembled on a GUV after addition of 400 nM Whi3. (d) Time-lapse of membrane-associated Whi3 condensates on an SLB (top row, TIRF images) and on the top of a GUV (lower row, confocal section). Red arrows indicate contacting condensates which do not fuse or round. (e) Images of rapid and complete unbinding of membrane-bound 6his-GFP at the indicated times after addition of 10 mM EDTA. Images at 1 and 5 min are contrasted equally. Plot shows GFP intensity as a function of time after EDTA addition. Data are mean ± first s.d., n = 160,000 pixel intensity values from time-lapse images of 1 SLB. (f) Solution droplets formed with 41 μM Whi3 at the indicated times after assembly, induced by lowering the KCl concentration to 75 mM. Images are maximum intensity projections from confocal z-stacks. All images contrasted equally. (g) Average radius of solution droplets as a function of time after assembly, formed with 41 μM Whi3. Black line shows fitted power law function with indicated scaling exponent. Data are mean ± first s.d. n = 439 - 1,171 droplets per data point, from 3 biologically independent samples. Exact n per data point provided in Source Data Extended Data Fig. 4g. (h) Time-lapse of rapid droplet fusion, 4 h after assembly. GUV and SLB membrane composition: 96 mol% DOPC, 4 mol% DGS NTA-Ni; 0.03 mol% TR-DHPE included for FRAP.

Extended Data Fig. 5. RNA is clustered at the edges of pre-formed condensates.

(a) Time series of two example condensates formed in the presence of 50 nM Whi3 and 100 pM CLN3. Images show condensates formed in proximity to an RNA puncta (black dashed box) and with no associated RNA (gray dashed box). (b) Time-lapse of condensate assembly on SLB with 50 nM Whi3 and 100 pM CLN3. Frames span 2.5-8.5 min after addition of Whi3, 1 min between frames. White arrows indicate Whi3 condensates visible after 3.5 min, and yellow arrowheads indicate condensate edge-associated CLN3 puncta 1 min later. White asterisks in final frames indicate bright CLN3 clusters that continued to assemble at condensate edges. (c) Intensity histograms of CLN3 puncta adsorbed to condensate edges or in the dilute phase on the surrounding membrane, 20-30 min after addition of 50 nM Whi3. n = 467 and 24,954 edge-adsorbed and dilute phase puncta, respectively, from 4 biologically independent samples. p-value from two-sided Kolmogorov-Smirnov test. (d) Membrane-associated condensates do not attain measurable height. Maximum intensity projections from a spinning disc confocal z-stack of condensates formed in the presence of 500 nM Whi3, with z-spacing of 0.2 μm. SLB membrane composition: 96 mol% DOPC, 4 mol% DGS NTA-Ni.

Extended Data Fig. 6. Membrane-tethered RNA recruits Whi3 and forms condensates.

(a) Confocal slices of ER and CLN3 smFISH in Ashbya cells. Arrows indicate example CLN3 puncta showing ER co-localization. (b) Images of CLN3 puncta tethered to membrane (left) or PEG surface (right) prior to addition of Whi3. Images contrasted equally. (c) Initial densities of CLN3 puncta on membranes or PEG surfaces prior to addition of Whi3. Black bars show mean ± first s.d., n = 90 and 60 images from 18 and 12 independent membrane and PEG experiments, respectively. (d) Average CLN3 puncta intensity on membrane or PEG surfaces as a function of time after addition of 50 nM Whi3. PEG-tethered CLN3 did not recruit Whi3 or cluster to the same extent as membrane-tethered CLN3, despite the higher initial CLN3 density on PEG surfaces. Data are mean ± 95% c.i. n = 1,059 - 1,335 and 1,804 - 1,987 puncta per data point in membrane and PEG experiments, respectively, from 3 biologically independent samples in each experiment. Exact n per data point provided in Source Data Extended Data Fig. 6d. (e) Histograms of scaled Whi3 puncta intensity from Ashbya cells and membrane-tethered CLN3 experiments after clustering by 50 nM Whi3. Scaled distributions obtained by dividing by the distribution means. n = 13,198 and 867 Whi3 puncta in membrane-tethered CLN3 experiment and in live cells, respectively. (f) Membrane-tethered RNA condensates 15 min after exposure to Whi3 at the indicated concentrations. CLN3 and Whi3 channels contrasted equally in all images. (g-h) Intensity of CLN3 (g) and Whi3 (h) within membrane-tethered RNA condensates as a function of bulk Whi3 concentration. Data are mean ± 95% c.i. in (g-h). Each data point in (g-h) contains n = 7,484 - 35,945 and 9,238 - 37,992 puncta in membrane and PEG experiments, respectively, from 3 biologically independent samples in each experiment. Exact n per data point provided in Source Data Extended Data Fig. 6g,h. SLB membrane composition: 99 mol% DOPC, 1 mol% DOPE cap-biotin.

Extended Data Fig. 7. Average FCS traces of solution and membrane-bound GFP.

Black curves indicate fits to single-component diffusion model. n = 10 and 27 FCS traces in solution and membrane experiments, respectively. SLB membrane composition: 96 mol% DOPC, 4 mol% DGS NTA-Ni.

Figure 1.

Whi3 puncta persistently associate with the endoplasmic reticulum in Ashbya gossypii. (a) Ashbya hyphae expressing Whi3-tdTomato and Sec63-GFP (ER marker). White arrows indicate example Whi3 puncta showing ER co-localization. (b) Local intensity in the ER channel at detected Whi3 puncta, expressed as a fraction of the median intensity of the ER channel throughout the hypha, after masking and background subtraction (see methods). Blue points: observed Whi3 positions, orange points: randomized Whi3 positions. n = 642 puncta from 60 hyphae. Dashed line indicates the threshold for co-localization, corresponding to the median intensity of the ER channel. Horizontal bar represents mean, vertical bar represents first s.d. (c) Ratio of the average local intensity in the ER channel at detected Whi3 puncta within a hypha relative to randomized Whi3 positions. Each data point represents the average of 50 random distributions per hypha, n = 60 hyphae (see methods). Horizontal bar represents mean, vertical bar represents first s.d. (d) Montages of Whi3 puncta associated with ER tubule (left, yellow arrowhead), nuclear-associated ER (middle, blue arrowhead), and hyphal tip (right, red arrowhead). White dashed lines in first frame indicate cell periphery. Montages span 6 min, 30 s/frame. (e) First frame from a time-lapse showing ER (green) and Whi3 (magenta) with three overlaid Whi3 tracks. Plot shows the associated, local intensity in the ER channel at each time point of the three tracks, expressed as a fraction of the median intensity of the ER channel throughout the hypha. All three tracks spend the entire lifetime co-localized with the ER. (f) Histogram of Whi3 tracks, binned according to the fraction of track lifetime co-localized with the ER. n = 769 tracks from 60 hyphae.

Figure 2.

Membranes promote assembly of protein-only condensates with properties that deviate from LLPS predictions. (a) Schematic of SLB experiment. (b) Time series of SLB after addition of 50 nM Whi3-Atto488. Frames show different SLB regions at the indicated times. White arrows indicate condensates apparent after 5 min. (c) Top: Time-lapse of PEG surface after addition of 50 nM Whi3. Bottom: Histogram of the number of Whi3 proteins per puncta on PEG surfaces. n = 10,791 puncta from 3 biologically independent samples. (d) Top: Time-lapse of membrane-associated condensate, formed with 50 nM Whi3, after addition of 10 mM EDTA. Bottom: Average condensate intensity as a function of time after addition of 10 mM EDTA. Data are mean ± s.e.m. n = 10 condensates from 3 biologically independent samples. (e) Membrane-associated condensates 180 min after addition of Whi3 at the indicated bulk concentrations. Images are contrasted unequally to clearly show puncta or condensates. (f) Radius of membrane-associated condensates as a function bulk Whi3 concentration. Data are mean ± first s.d. (g-h) Protein density within the condensate dense phase (g) and dilute phase (h) as a function of bulk Whi3 concentration. Data in (g-h) are mean ± first s.d. Data in (f-g) from n = 717, 719, 592, and 687 condensates with 20, 50, 100, and 500 nM Whi3, respectively, from 3 biologically independent samples. Data in (h) from n = 30 images per Whi3 concentration, from 3 biologically independent samples. (i) Membrane-associated condensates, initially assembled with 50 nM Whi3-Atto488, approximately 20 min after addition of 50 nM Whi3-Atto594. SLB membrane composition: 96 mol% DOPC, 4 mol% DGS NTA-Ni.

Figure 3.

Delayed interactions with RNA reduce the sizes of pre-assembled Whi3 condensates. (a) Time series of SLB with CLN3-Cy3 in solution prior to Whi3 addition. Frames show different SLB regions at the indicated times after addition of 50 nM Whi3-Atto488. Final CLN3 concentration is 100 pM. White arrows indicate condensates apparent after 5 min. Yellow arrowheads indicate relatively bright CLN3 puncta at condensate edges. (b) Membrane-associated condensates 180 min after assembly with 50 nM Whi3 and 100 pM CLN3. (c) Histograms of condensate radii formed with 50 nM Whi3, in the presence and absence of CLN3, approximately 180 min after condensate assembly. n = 1,080 and 719 condensates in the presence and absence of CLN3, respectively, from 3 biologically independent samples. p-value from two-sided Kolmogorov-Smirnov test. (d) FRAP of condensates formed with 50 nM Whi3 in the presence (upper) and absence (lower) of 100 pM CLN3. Plot shows corresponding FRAP profiles. Red curves show fits to single-component exponential recovery model, with mobile fractions (m.f.) and recovery time constants (τ) indicated. Data are mean ± s.e.m. n = 12 and 8 condensates with and without RNA, respectively, from 2 biologically independent samples. (e) Intensity of “dilute” phase CLN3 puncta recruited by membrane-bound Whi3 as a function of time after addition of 50 nM Whi3. Black line shows fit to a power law function with indicated exponent. Data are mean ± 95% c.i. n = 44 - 2,264 puncta per data point, from 4 biologically independent samples. Exact n per data point provided in Source Data Fig. 3e. SLB membrane composition: 96 mol% DOPC, 4 mol% DGS NTA-Ni.

Figure 4.

Whi3 forms condensates with membrane-tethered RNA. (a) Representative confocal section of ER and CLN3 smFISH in Ashbya cell. Arrows indicate example CLN3 puncta showing ER co-localization. (b) Schematic of membrane-tethered RNA experiment. (c,d) SLB (c) and immobile, PEG-coated surface (d) with tethered CLN3 at indicated times after addition of 50 nM Whi3. Time lapse images show merged CLN3 and Whi3 channels, while the cyan (c) or yellow (d) boxed frames show separate channels at 10 min time point. CLN3 and Whi3 channels contrasted equally. (e) Intensity histograms of membrane or PEG surface-tethered CLN3 puncta before and after addition of 50 nM Whi3. SLB experiment: n = 35,945 and 13,198 puncta before and after Whi3, from 18 and 3 biologically independent samples, respectively. PEG experiment: n = 37,992 and 10,084 puncta before and after Whi3, from 12 and 3 biologically independent samples, respectively. p-values from two-sided Kolmogorov-Smirnov tests. SLB membrane composition: 99 mol% DOPC, 1 mol% DOPE cap-biotin.

Figure 5.

Condensates formed with membrane-tethered RNA coarsen by fusion but rapidly arrest. (a) Time series of two condensate fusion events at the indicated times after addition of 50 nM Whi3. Yellow arrowheads indicate condensate pairs one frame prior to fusion. (b) CLN3 intensity as a function of time after addition of 50 nM Whi3, plotted on log-log axes. Black line shows fit to a power law function over the first 5 data points, with indicated exponent. Data are mean ± 95% c.i. n = 1,059 - 1,335 puncta per data point, from 3 biologically independent samples. Exact n per data point provided in Source Data Fig. 5b. (c) Membrane-tethered RNA condensates 15 min after exposure to Whi3 at the indicated concentrations. CLN3 and Whi3 channels contrasted equally in all images. (d) CLN3 intensity as a function of time after addition of Whi3 at the indicated concentrations. 50 nM Whi3 data repeated from (b). Data are mean ± 95% c.i. n = 952 - 1,299 and 845 - 1,192 puncta per data point with 20 and 200 nM Whi3, respectively, from 3 biologically independent samples. Exact n per data point provided in Source Data Fig. 5d. (e) Proportion of CLN3 puncta with intensity below 50 brightness units as a function of Whi3 concentration. Data are mean ± first s.d. n = 18 biologically independent samples for 0 nM Whi3, n = 3 for all other data points. SLB membrane composition: 99 mol% DOPC, 1 mol% DOPE cap-biotin.

Figure 6.

Diffusion slows as particles gain mass. (a) Average, normalized FCS traces of Whi3 diffusion in solution or at membrane. Black curves indicate fits to single-component diffusion model. n = 15 and 20 FCS traces in solution and membrane experiments, respectively, from 3 biologically independent samples. (b) Diffusion coefficients of Whi3 and GFP in solution or at membranes. Data are mean ± first s.d., n = 15, 20, 10, and 27 FCS traces in Whi3 solution, Whi3 SLB, GFP solution, and GFP SLB experiments, respectively. GFP solution and SLB data from 2 and 4 biologically independent samples, respectively. (c) RNA puncta tracks before (left) and after (right) addition of 50 nM Whi3. Images contrasted equally. Each color represents a separate track. Tracks corresponding to low-mobility puncta appear as dots. (d) Histograms of mean CLN3 puncta velocity on SLBs before (black) and after (blue) addition of 50 nM Whi3. n = 20,485 and 1,668 tracks before and after Whi3, from 18 and 3 biologically independent samples, respectively. (e) Proportion of CLN3 puncta with average velocity below 0.5 μm/s as a function of Whi3 concentration. Data are mean ± first s.d. n = 18 biologically independent samples for 0 nM Whi3, n = 3 for all other data points. (f) Histograms of CLN3 puncta diffusion exponent, α, before and after addition of Whi3 at the indicated concentrations. n = 3,544, 384, and 510 tracks with 0, 20, and 50 nM Whi3, from 18, 3, and 3 biologically independent samples, respectively. (g,h) Mean CLN3 velocity (g) and total mass of Whi3 and CLN3 (h) plotted as a function of CLN3 intensity (measured at the start of the track), before and after addition of Whi3 at the indicated concentrations. Data in (g-h) are mean ± 95% c.i. and represent moving averages of the raw data, with CLN3 intensity bins of 35. Each bin in (g) contains n = 17 - 16,311, 123 - 623, 198 - 417, and 169 - 219 puncta with 0, 20, 50, and 400 nM Whi3, from 18, 3, 3, and 3 biologically independent samples, respectively. Each bin in (h) contains n = 46 - 27,372, 1,085 - 2,766, 1,741 - 3,084, and 737 - 1,432 puncta with 0, 20, 50, and 400 nM Whi3, from 18, 3, 3, and 3 biologically independent samples, respectively. Exact n per data point provided in Source Data Fig. 6g,h. (i) Mean CLN3 velocity as a function of the total mass of Whi3 and CLN3. Plot shows data points repeated from (g,h) and additional Whi3 concentrations. Plot also includes four additional CLN3 intensity bins per Whi3 concentration. Data points represent mean, horizontal and vertical error bars represent 95% c.i. Exact n per data point provided in Source Data Fig. 6i. Data are from 18 biologically independent samples for 0 nM Whi3 and 3 biologically independent samples for all other data points. (j) Summary schematic. Left: Whi3-CLN3 condensates assembled in solution nucleate, grow, and coarsen into micrometer-scale droplets over several hours^, (blue regime). Right: Addition of Whi3 to membrane-tethered CLN3 drives assembly of condensates that arrest within minutes (red arrow), resulting in condensates that more closely resemble size distributions in cells. While diffusion of Whi3 and CLN3 in solution remains unhindered throughout assembly (yellow boxes), the lateral diffusion of membrane-tethered CLN3 is reduced upon Whi3 binding and clustering (orange and red boxes). Sizes of gray arrows indicate relative magnitudes of diffusion. SLB membrane composition in (a-b): 96 mol% DOPC, 4 mol% DGS NTA-Ni; in (c-i): 99 mol% DOPC, 1 mol% DOPE cap-biotin.