Membrane prewetting by condensates promotes tight-junction belt formation

Cell culture in two-dimensional monolayers

HEK293 and MDCK-II (00062107, Public Health England) cells were cultured as two-dimensional monolayers in minimal essential medium with 5% fetal bovine serum, 1% non-essential amino acids, 1% sodium pyruvate, and 1% GlutaMAX without addition of antibiotics at 37 °C with 5% CO₂. Two-dimensional monolayers on transgenic cell lines were selected in the presence of geneticin (400 μg ml^−1). Transient transfection of MDCK-II cells was carried out using Lipofectamine 2000 after cells reached a confluency of 70%. Transgenic cell lines were created from synthesized genes cloned using Not-I and Asc-I cutting sites into mammalian expression plasmids designed in-house (pOCC series) with N-terminal Dendra2, a selection marker against neomycin-geneticin and a CMV promotor. To obtain stable lines, single clone selection was carried out by fluorescence-activated cell sorting.

For seeding of two-dimensional monolayers, 400 µl of cell suspension (0.5 million cells per millilitre) was transferred on to the external side of a standing transwell filter (Corning 3460). Cells were left to adhere at 37 °C in a sterile environment for 30 min. Afterwards, each transwell filter was mounted in a one-well plate with 1 ml of culture medium on the bottom side and 500 µl of medium on the internal side.

Generation of three-dimensional cysts and transepithelial permeability assay

For the adherent three-dimensional cell culture cysts, MDCK-II cells were resuspended from a confluent monolayer into a single-cell suspension. The surface of a Mateck dish (35 mm glass bottom, P35G-0.170-14-C) was coated with a solution of laminin (0.5 mg ml^−1) for 1 h at 37 °C, 5% CO₂. Afterwards, a suspension of 20,000 cells were seeded on the coated surface in the respective culture medium complemented with 5% Matrigel on ice. Cells were cultured for 5–6 days until reaching 30–40 µm in diameter. To measure the permeability of dextran (Alexa647, 10k) cyst medium was supplemented with 10 μM dextran. After 15 min of incubation, cysts were imaged using a confocal microscope to evaluate the distribution of dextran. The permeability of dextran into the lumen was quantified by measuring the intensity in the lumen and dividing this by the average intensity outside the cysts.

CRISPR–Cas9 knock-in MDCK-II cells

CRISPR–Cas9 technology was used to generate several single and double knock-in MDCK-II cell lines. ZO-1–mNeon was generated by fusing a copy of mNeon at the N terminus of the initial exon of endogenous ZO-1. Using the cell line expressing this fusion construct as a background, we tagged several other scaffold proteins with mScarlet at the N terminus of the initial exon of endogenous PATJ or ZO-2 and at the carboxyl terminus of MAGI-3 to generate double knock-in lines. Briefly, specific CRISPR RNA (crRNA, Integrated DNA Technologies (IDT) lt-R CRISPR–Cas9 crRNA) for each gene of interest was designed using online tools Crispor (http://crispor.tefor.net) and ChopChop (https://chopchop.cbu.uib.no). Trans-activating crRNA (IDT catalogue no. 1072532) and crRNA were annealed at a ratio of 1:1 by incubation for 5 min at 95 °C and then 10 min at room temperature to generate guide RNA (gRNA). All gRNAs can be found in Supplementary Table 4. Next, the donor plasmid containing 5′ and 3′ homology arms was synthesized in a pUC57 Kan (Genescript) or a pUCIDT Kan (IDT) (Supplementary Table 4). A PTisy plasmid containing the fluorescence tag (mSc or mNn) was integrated into the donor plasmid through digestion using restriction enzymes. Afterwards, a ribonucleoprotein complex was assembled by mixing 1 μl (10 μg μl^−1) of the recombinant Alt-R S.p. HiFi Cas9 nuclease (IDT, catalogue no. 1081060) and 1 μl of gRNA (100 μM) with the reaction buffer, followed by incubation at room temperature for 20 min. The ribonucleoprotein complex and 1 μg of the donor plasmid were cotransfected through electroporation and overlapped with the exon sides. Electroporation of each complex was performed in 300,000 cells (Invitrogen NEON electroporation machine and kit, two pulses, 20 ms, 1,200 V). Cells were then plated in growth medium in a six-well plate. Medium was exchanged after 24 h. Cells were sorted 48–72 h after electroporation. Fluorescent cells (mNn or mSc) were enriched (one or two cycles), and single clones were plated in one well of a 96-well plate. To validate genetic modification, we first did a PCR amplification. The correct insertion of the fluorescent tag at the correct locus of the gene of interest was verified by sequencing genotyping. Sequencing confirmed homozygous insertion of the tags, and imaging confirmed that mS tagging of the endogenous proteins resulted in proper colocalization with mN-ZO-1 at the tight-junction belt in confluent monolayers.

CRISPR–Cas9 knockout in MDCK-II cells

For deletion of PATJ ∆exon 3, we used two gRNAs on each site of the targeting exon for each gene, selected on the basis of low off-target activity using http://crispor.tefor.net. The gRNAs were ordered as crRNAs from IDT. gRNAs were transfected in pairs, and cell pools were tested for deletion events by PCR using primers spanning the targeting exon. Best-performing gRNA pairs were used for cell cloning. Single cells were screened for the deletion by PCR using primers spanning the targeting exon. Flanking PCR with one primer outside and one primer inside the deletion was performed on the knockout candidate clones to verify the absence of the WT allele. Deletion alleles were verified by Sanger sequencing.

Sequencing of two clones confirmed deletion of the first exon, including a frame shift leading to an early stop codon. Western blots and quantitative PCR (qPCR) against the N-terminal L27 domain of PATJ, encoded by the first exon, confirmed deletion of this domain. Immunostaining of the more C-terminal PDZ4 confirmed that the remaining truncated protein was located in the cytosol and was not present at the tight junction. qPCR was performed using a Qiagen RNeasy (74104) mini kit after bioanalysis showed that the primers were of good quality (Supplementary Table 4).

Two-colour super-resolution STED microscopy

STED imaging was performed with a commercial confocal infinty-line STED microscope (Abberior Instruments) operated using the Imspector software (v.16.2.8415), with pulsed laser excitation (490 nm, 560 nm, 640 nm, 40 MHz), and ×60 water and ×100 oil objectives (Olympus). Star Orange was imaged with a pulsed laser at 560 nm, and excitation of Abberior Star Red was performed at 640 nm. The depletion laser for both colours was a 775 nm, 40 MHz, pulsed laser (Katana HP, 3 W, 1 ns pulse duration, NKT Photonics). The optimal combination of excitation intensity, STED power and pixel dwell time were established by minimizing the onset of strong photobleaching. To reduce high-frequency noise, STED images were filtered with a two- or three-dimensional Gaussian with a sigma of 0.8 pixels. Quantification and segmentation of STED data was done using custom software written in MATLAB⁵².

Live imaging microscopy

Live imaging kinetics were acquired using a Wide-Field Delta Vision Elite system equipped with an Olympus PlanApo N ×60/1.42 objective, SSI Lumencor illumination system and Roper Evolve EMCCD camera. Perturbation experiments were acquired with a wide-field General Electric Delta Vision system equipped with an Olympus PlanApo N ×60/1.42 objective, and mN and mCh were excited with an Applied Precision Xenon Arc Lamp (V300-Y18). High-resolution imaging was done with a Yokogawa Spinning Disk Field Scanning Confocal System (CSU-X1 Nikon, Andor iXOn Ultra) using a ×60 water objective.

MDCK-II calcium switch assay

The calcium depletion experiments were performed on a confluent monolayer of MDCK-II with different cell lines. Cells were grown for 18 h in calcium-free media until tight junctions were disrupted. For proteomics experiments, medium containing calcium was added to the cells at 37 °C, 5% CO₂, and each sample was prepared at 0 h, 0.5 h, 1 h, 3 h, and 18 h afterwards. For live imaging experiments, medium containing calcium was added to the cells directly on the microscope chamber at 37 °C, 5% CO₂, and tight-junction formation was imaged every 1 min for 3 h or every 30 min for 18 h.

Transepithelial resistance

MDCK-II cells were seeded into a Corning transwell plate with membrane inserts (12 mm, 0.4 mm pore), and transepithelial resistance was measured at various time points from 1 h after seeding until 6 days. Measurements were acquired using a two-electrode resistance system from Millipore (Millicell ERS-2) inserted between the apical and basal parts of the chamber. The electric resistance was multiplied by the growth area of the transwell filters (1.12 cm²), and the resistance of the medium without cells was subtracted to give a final result in Ω × cm².

Quantification of ZO-1 belt length and cell perimeter coverage

To determine the coverage of the cell perimeter by ZO-1 (%) or the ZO-1 belt length per cell (µm per cell), we performed the following steps. We segmented the cells in the monolayer with CellPose⁵³ using the bright-field channel as input. The outline of the CellPose segmentation was used as the cell perimeter length. We segmented the mN-ZO-1 channel through local intensity thresholding and skeletonized the segmented image to obtain the length of the ZO-1 belt using FIJI. The percentage of the cell perimeter covered by ZO-1 was calculated by the following formula: perimeter coverage = (TJ length/cell periphery) × 100%. The length of the ZO-1 belt per cell, L_ZO-1 (in μm per cell), was calculated using MATLAB as follows: L_ZO-1 = sum(skeletonized ZO-1 image)/cell number.

Determination of tight-junction protein recruitment kinetics

To determine the t_1/2 arrival kinetics of the mS-tagged tight-junction proteins from the two-colour time series, we segmented the condensed mN-ZO-1 signal and quantified the mScarlet intensity in the segmented ZO-1 condensates and the cytoplasm over time. Cell segmentation was done using CellPose (https://github.com/mouseland/cellpose), and ZO-1 condensate segmentation used the plugins Subtract Background and Make Binary in FIJI. To directly correct for photobleaching artefacts, we calculated the ratio between the junctional and the cytoplasmic mScarlet signal for each time point using custom MATLAB code⁵². Assuming that bleaching is spatially homogenous, this ratio is independent of bleaching and directly reports the enrichment of the protein in the condensed ZO-1 (tight junction). Kinetic data were fitted using a HiII slope model in Prism with a nonlinear fit variable slope (four parameters) to calculate the half-maximal effective concentration of each protein. To calculate the t_1/2 values for client protein arrival, we fitted the kinetic data to a Hill slope model and determined the difference in arrival time between ZO-1 and the client protein in living cells.

Determination of ZO-1 condensate extension and eccentricity

ZO-1 condensate eccentricity was quantified on segmented ZO-1 images using the function regionprops to measure the eccentricity of segmented condensates over time in MATLAB. The eccentricity is the ratio of the distance between the focus of the ellipse and its major axis length; it takes a value between 0 and 1. (Here, 0 and 1 are degenerate cases; an ellipse whose eccentricity is 0 is a circle, whereas an ellipse whose eccentricity is 1 is a line segment.)

Extension and intensity rates

To determine the extension of the condensates over time, we used the plugin JFilament in FIJI⁵⁴. The analysis directly provided the length of the condensates over time. To quantify the amount of ZO-1 material in the condensate over time, we used the segmentation of the JFilament tracks to measure the sum intensity of the condensate per time point in MATLAB. As the JFilament tracks were one-dimensional, we extended the width of the segmentation to fit the width of the condensate.

Fluorescence recovery after photobleaching

FRAP experiments in cells were carried out with the following settings on a confocal infinty-line STED microscope (Abberior Instruments). The region of interest was bleached using a 405 nm diode at 1.5 mW at the back focal plane of the objective, with 100 ms pixel dwell time. Prebleaching and postbleaching images were acquired using a 490 nm laser at 5 μW. Fluorescence recovery of mNeon was monitored for 1–20 min with a time resolution of 11 s. Cell movements during the recovery were corrected by registration of all frames to the first frame using the plugin StackReg in FIJI.

Client partitioning assays in HEK293

HEK293 cells were transfected with ZO-1–GFP, and clients were tagged with mCherry. We measured the fluorescence intensity of the client inside and outside the condensate of ZO-1, and the background outside the cell was subtracted from those values.

Immunoblotting and in-gel fluorescence

Aliquots of approximately 5 μg of input, elution and beads in lysate buffer (10% glycerol, 2% sodium dodecyl sulfate (SDS), 1 mM dithiothreitol (DTT), 1x Protease Inhibitor Cocktail, 50 mM Tris-HCl, pH 8) were subjected to SDS polyacrylamide gel electrophoresis (4–20% Tris-Glycine Novex gels) at 90 mV for 3 h. An iBlot2 gel transfer system (Thermo Fisher, IB21001) was used to transfer proteins on to a nitrocellulose membrane (10 min, 20 V). An iBind Flex system (Thermo Fisher, SLF2000) was used to detect protein levels by immunoblotting, with the following antibodies: IRDye-800CW streptavidin (1:4000, LI-COR, 92632230). Membranes were scanned with an LI-COR Odyssey system at 700 nm and 800 nm. In-gel fluorescence of the tagged proteins (Dendra2) was done in a Typhoon FLA 7000 scan system (GE). For western blot detection of PATJ, a wet system was used for transfer (260 mA for 2 h), and classical HRP detection of proteins was done using superSignal West Pico Plus chemiluminescent substrate (Thermo Fisher, catalogue no. 34580) on GE hyperfilms (Cytiva, 28-9068-50). Primary antibody incubation was performed overnight at 4 °C in 5% milk and 0.1% Tween, with PATJ-L27 (rabbit polyclonal, gifted by Le Bivic’s lab) 1:200, PATJ-PDZ4 (rabbit polyclonal, LSBio, LC-C410011), and 1:500 β-actin (rabbit polyclonal, Abcam, ab8227). The following secondary antibodies were used at 1:5000 dilution: goat anti-rabbit IgG-HRP (H+L) (Cell Signaling, catalogue no. 7074), and goat anti-mouse IgG-HRP (H+L) (Cell Signaling, catalogue no. 7076) (Supplementary Fig. 1).

Immunofluorescence

Fixation was performed in the same way for both two-dimensional monolayers and three-dimensional cysts. Cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 10 min at room temperature, followed by quenching in 300 mM glycine, and permeabilized with 0.5% Triton X-100 in PBS for 10 min. Cells were blocked with 2% bovine serum albumin and 0.1% Triton X-100 in PBS for 1 h at room temperature. Staining of all primary and secondary antibodies involved incubation at room temperature for 2 h or 30 min with a dilution of 1:50 or 1:200, respectively, in blocking buffer. Staining for neutravidin-647 (neutravidin, A-2666 and Alexa Fluor 647 succinimidyl ester A-20006, Invitrogen) was performed at 1:1000 dilution for 1 h at room temperature in blocking buffer. Primary antibodies were PATJ-L27 rabbit polyclonal (produced in-house), PATJ-PDZ4 rabbit polyclonal (LSBio, LC-C410011), PALS1 mouse monoclonal (Santa Cruz, sc-365411), ZO-1 mouse monoclonal IgG1 (Invitrogen, 33-9100), occludin rabbit polyclonal (Life Technologies, 71-1500), Lin7 rabbit polyclonal (Thermo Fisher, 51-5600), E-cadherin rabbit IgG, (Cell Signalling, 3195S). The secondary antibodies were goat anti-mouse (Abberior, star-red 2-0002-011-2) and goat anti-rabbit star-orange (Abberior, storange-1102)

APEX2 labelling during tight-junction formation

Cells were plated in a T75 flask after reaching confluency, and the full medium was substituted with medium without calcium for 18 h to enable the cells to reach a rounded, non-polarized state. Afterwards, the medium was replaced with full medium again, and APEX2 labelling was performed at various time points by adding 1 mM biotin phenol to the cells for 30 min before the addition of 1 mM H₂O₂ for 1 min. Immediately after, the reaction was quenched on ice by washing three times with 1× PBS supplemented with 10 mM sodium ascorbate (Sigma A4034), 10 mM sodium azide, and 5 mM Trolox ((+/−)-6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid, Sigma 238813). To validate biotinylation through fluorescence microscopy, cells were immediately fixed after this step. For proteomics and western blot analyses, cells were lysed by scraping them off the growth surface into ice-cold lysis buffer (2% SDS, 10% glycerol, 1 mM DTT, 50 mM Tris-HCl, pH 8, 1× Protease Inhibitor Cocktail Set III EDTA-Free (EMD Millipore, catalogue no. 539134), supplemented with 10 mM sodium ascorbate, 10 mM sodium azide and 5 mM Trolox. The lysate in was collected in a low-protein-binding reaction tube and 5 μl of benzonase was added, followed by incubation in a shaker for 15 min at 37 °C, 700 rpm. Finally, 1 mM EDTA and 1 mM EGTA were added, and the sample was spun down using a table centrifuge to remove debris. After lysis, the protein concentration was checked using a Pierce 660-nm (Thermo Fisher Scientific, catalogue no. 22660) protein assay test before pull-down to ensure the same starting protein concentration was used. Lysates were aliquoted into 1,500 μg of total protein, snap frozen and stored at −80 °C.

Streptavidin pull-down of APEX2 biotinylated proteins

All buffers used for pull-down proteomics experiments were freshly made and filtered with a 0.22 μm filter before use. Frozen lysates (1,500 g protein) were diluted 1:10 in 50 mM Tris, pH 8, and placed in an incubation buffer consisting of 0.2% SDS, 1% glycerol, 1 mM DTT, 1 mM EGTA, 1 mM EDTA in 50 mM Tris-HCl, pH 8, and 1× protease inhibitor. For affinity purification, approximately 100 l of streptavidin magnetic beads (Pierce, PI88817) were washed in incubation buffer two times before binding. Beads were added to the sample in a total volume of 2 ml and incubated for 2 h at room temperature in a rotating wheel. Next, beads were pelleted down using a magnetic rack, and the supernatant was collected and kept for western blot analysis. Each sample of beads containing the bound biotinylated proteins was washed with a series of ice-cold buffers (2 ml each) to remove unspecific binders. The beads were then washed twice with washing buffer (0.2% SDS, 1% glycerol, 1 mM DTT, 1 mM EGTA, 1 mM EDTA in 50 mM Tris-HCl pH 8, 1× protease inhibitor), once with 1 M KCl, once with 2 M urea in 50 mM Tris-HCl pH 8, once with 2 mM biotin 50 mM Tris-HCl pH 8, and finally three times with 50 mM Tris-HCl pH 8. Biotinylated proteins were eluted by boiling the beads in 50 μl of elution buffer (5% SDS, 10% glycerol, 20 mM DTT, in 50 mM Tris-HCl, pH 8, 1× protease inhibitor) at 95 °C for 10 min, followed by cooling on ice and a brief spin-down. Samples were placed on a magnetic rack, and the eluate was collected in a new tube for proteomics and SP3. Magnetic beads and 10 μl of eluate were together subjected to western blotting for validation of the biotinylation experiments before mass spectrometry.

To process all time points in the same quantitative mass spectrometry analysis, we used a multiplex proteomic approach based on tandem mass tag (TMT). The TMT isobaric tagging approach enables robust quantitative proteomics by measuring all samples in one run. This enabled statistical analysis of relative protein enrichment at different time points after calcium switch across proteins detected at all time points.

Mass spectrometry and TMT labelling

Reduction of cysteine-containing proteins was performed with dithiothreitol (56 °C, 30 min, 10 mM in 50 mM HEPES, pH 8.5). Reduced cysteines were alkylated with 2-chloroacetamide (room temperature, in the dark, 30 min, 20 mM in 50 mM HEPES, pH 8.5). Samples were prepared using the SP3 protocol^55,56, and 300 ng trypsin (sequencing grade, Promega) was added per sample for overnight digestion at 37 °C. Peptides were labelled with TMT10plex Isobaric Label Reagent (Thermo Fisher) according to the manufacturer’s instructions. In short, 0.8 mg reagent was dissolved in 42 μl acetonitrile (100%), and 8 μl of stock solution was added, followed by incubation for 1 h and quenching with 5% hydroxylamine for 15 min at room temperature. Samples were combined for the TMT10plex, and an OASIS HLB µElution Plate (Waters) was used for further sample clean-up⁵⁷. Offline high-pH reverse-phase fractionation was performed using an Agilent 1200 Infinity high-performance liquid chromatography system equipped with a quaternary pump, degasser, variable-wavelength ultraviolet detector (254 nm), and Peltier-cooled autosampler and fraction collector (both set at 10 °C for all samples).

The column was a Gemini C18 column (3 μm, 110 Å, 100 × 1.0 mm, Phenomenex) with a Gemini C18, 4 × 2.0 mm SecurityGuard (Phenomenex) cartridge as a guard column. The solvent system consisted of 20 mM ammonium formate (pH 10.0) (phase A) and 100% acetonitrile as the mobile phase (B). The separation was accomplished at a mobile phase flow rate of 0.1 ml min^−1 using the following linear gradient: 100% A for 2 min, from 100% A to 35% B in 59 min, to 85% B in a further 1 min, and held at 85% B for 15 min, before returning to 100% A and re-equilibration for 13 min. Thirty-two fractions were collected during liquid chromatography separation and subsequently pooled into six fractions. The first and the two last fractions of the 32 were discarded and not used at all. Pooled fractions were dried under vacuum centrifugation and reconstituted in 15 μl 1% formic acid, 4% acetonitrile, for liquid chromatography coupled with tandem mass spectrometry analysis.

Mass spectrometry data acquisition

Peptides were separated using an UltiMate 3000 RSLC nano liquid chromatography system (Dionex) fitted with a trapping cartridge (µ-Precolumn C18 PepMap 100, 5 µm, 300 µm i.d. × 5 mm, 100 Å) and an analytical column (nanoEase M/Z HSS T3 column 75 µm × 250 mm C18, 1.8 µm, 100 Å, Waters). Trapping was carried out with a constant flow of solvent A (3% dimethyl sulfoxide, 0.1% formic acid in water) at 30 µl min^−1 on to the trapping column for 6 min. Subsequently, peptides were eluted through the analytical column with a constant flow of 0.3 µl min^−1 with an percentage of solvent B (3% dimethyl sulfoxide, 0.1% formic acid in acetonitrile) increasing from 2% to 8% in 4 min, from 8% to 28% in 104 min, from 28% to 40% for a further 4 min and finally from 40% to 80% for 4 min before returning to the equilibration condition of 2%. The outlet of the analytical column was coupled directly to an Orbitrap Fusion Lumos Tribid mass spectrometer (Thermo Fisher) using the proxeon nanoflow source in positive ion mode.

The peptides were introduced into the Fusion Lumos through a Pico-Tip Emitter 360 µm OD × 20 µm ID 10 µm tip (New Objective) with an applied spray voltage of 2.2 kV. The capillary temperature was set to 275 °C. A full mass scan was acquired with a mass range of 375–1500 m/z in profile mode in the Orbitrap with a resolution of 120,000. The filling time was set to a maximum of 50 ms with a limitation of 4 × 10⁵ ions. Data-dependent acquisition was performed with the resolution of the Orbitrap set to 30,000, a fill time of 94 ms and a limitation of 1 × 10⁵ ions. A normalized collision energy of 36 was applied. MS2 data were acquired in profile mode.

Mass spectrometry analysis

IsobarQuant and Mascot (v.2.2.07) were used to process the acquired data. The data were then searched against the UniProt Canis lupus proteome database (UP000805418), which contains common contaminants and reversed sequences. The following modifications were included in the search parameters: carbamidomethyl (C) and TMT10 (K) (fixed modifications); and acetyl (N-term), oxidation (M), and TMT10 (N-term) (variable modifications). A mass error tolerance of 10 ppm and 0.02 Da was set for the full scan (MS1) and the MS/MS spectra, respectively. A maximum of two missed cleavages was allowed, and the minimum peptide length was set to seven amino acids. At least two unique peptides were required for protein identification. The false discovery rate (FDR) at the peptide and protein level was set to 0.01. The R programming language (ISBN 3-900051-07-0) was used to analyse the raw output data of IsobarQuant. Potential batch effects were removed using the limma package. Variance stabilization normalization was applied to the raw data using the vsn package. Individual normalization coefficients were estimated for different time points compared with the non-calcium condition (t₀). Normalized data were tested for differential expression using the limma package. The replicate factor was included in the linear model. For comparisons of different time points versus the t₀ condition, proteins with fold change greater than 2 were considered to be potential hits, and an FDR threshold of 0.05 was used to filter out noisy data. In addition, we excluded false positive hits due to non-junctional interactions (ribosomes, nucleus, mitochondria, endoplasmic reticulum (Supplementary Table 2).

In experiments comparing different time points, proteins were first tested for their enrichment compared with a −t₀ control. R package fdrtool35 was used to calculate FDRs using the t values from the limma output. Proteins with an FDR less than 5% and a consistent fold change of at least 10% in each replicate were defined as hits. The ggplot2 R package was used to generate graphics. Proteins matching a false positive list of non-biotinylated proteins were removed (Supplementary Table 3). RStudio code was adapted from previously reported code (https://github.com/fstein/EcoliTPP). R packages used were limma (https://bioconductor.org/packages/limma), MSnbase (https://bioconductor.org/packages/MSnbase), tidyverse (https://tidyverse.tidyverse.org), biobroom (https://bioconductor.org/packages/biobroom), ggrepel (https://cran.r-project.org/web/packages/ggrepel/vignettes/ggrepel.html) and ClusterProfiler (https://bioconductor.org/packages/clusterProfiler/). The interactome was created in Cytoscape (v.3.9.0) using the STRING database (v.11.5). Cell localization gene ontology annotations were from http://geneontology.org. The mass spectrometry proteomics data have been deposited at the ProteomeXchange Consortium through the PRIDE⁵⁸ partner repository with dataset identifier PXD052221.

Quantification and statistical analysis

Images were analysed with FIJI (https://fiji.sc/) and MATLAB (MathWorks). All data are expressed as the mean ± s.d., mean ± s.e.m. or mean ± 95% confidence interval, as stated in the figure legends and results. Values of n and what n represents (for instance, number of images, condensates or experimental replicates) are stated in figure legends and results. Two-tailed Student’s t-test or one-way ANOVA was used for normally distributed data. Statistical analyses used Kruskal–Wallis test with post hoc Dunn’s multiple-comparison test.

Numerical calculations for the thermodynamic model

Numerical calculations were done using programming language Python v.3.8.10; all code was run using IPython v.7.3.10. This software comes preinstalled in most of the Linux distributions. OS: Ubuntu 20.04.6 LTS, 64-bit, GNOME v.3.36.8. All code and a minimal dataset are available at Zenodo (https://zenodo.org/doi/10.5281/zenodo.11174400)⁵².

Reporting summary

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