Phase separation of Hippo signalling complexes

Abstract

The Hippo pathway was originally discovered to control tissue growth in Drosophila and includes the Hippo kinase (Hpo; MST1/2 in mammals), scaffold protein Salvador (Sav; SAV1 in mammals) and the Warts kinase (Wts; LATS1/2 in mammals). The Hpo kinase is activated by binding to Crumbs-Expanded (Crb-Ex) and/or Merlin-Kibra (Mer-Kib) proteins at the apical domain of epithelial cells. Here we show that activation of Hpo also involves the formation of supramolecular complexes with properties of a biomolecular condensate, including concentration dependence and sensitivity to starvation, macromolecular crowding, or 1,6-hexanediol treatment. Overexpressing Ex or Kib induces formation of micron-scale Hpo condensates in the cytoplasm, rather than at the apical membrane. Several Hippo pathway components contain unstructured low-complexity domains and purified Hpo-Sav complexes undergo phase separation in vitro. Formation of Hpo condensates is conserved in human cells. We propose that apical Hpo kinase activation occurs in phase separated "signalosomes" induced by clustering of upstream pathway components.

Keywords: Hippo signalling; condensates; epithelia; mechanobiology.

Figure 1. Hpo‐Venus can form cytoplasmic condensates in a mechanically‐ and nutritionally dependent manner

1. Drosophila egg chambers are composed of a large oocyte surrounded by a columnar follicle cell epithelium that activates a Hpo kinase bimolecular fluorescence complementation (BiFC) dimerization sensor (HpoKD‐Venus; green) apically in small punctae. Note that the Hpo sensor is activated in columnar cells but not in mechanically stretched cells.

2. Nutrient restriction (starvation) of adult females for 24 h causes formation of large micron‐scale Hpo sensor punctae.

3. Quantification of the size of Hpo sensor punctae from (A) and (B). Biological replicates are plotted as individual data points (n = 13 fed; n = 13 starved), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

4. Inhibition of PI3K by overexpression of a dominant‐negative version (the p60 subunit alone) is sufficient to induce formation of large micron‐scale Hpo sensor punctae, while activation of Akt by overexpression of a myristylated form of the protein is sufficient to reduce formation of the small apical Hpo sensor punctae that are normally observed in control animals.

5. Schematic diagram of pathways regulating formation of supramolecular Hpo kinase punctae in vivo.

Source data are available online for this figure.

Figure EV1. Phase separation of Hippo signalling complexes at the apical domain is enhanced by starvation or overexpression

1. Crb‐GFP localises apically in cuboidal and columnar epithelial cells of the ovarian follicular epithelium.

2. Kibra‐GFP localises apically in cuboidal and columnar epithelial cells of the ovarian follicular epithelium.

3. Wts‐GFP localises apically in cuboidal and columnar epithelial cells of the ovarian follicular epithelium.

4. Hpo‐YFP protein is mostly cytoplasmic, although weak apical signal is detectable in columnar epithelial cells.

5. A HpoKD‐Venus dimerization sensor can be detected at the apical domain of densely packed columnar follicle cells, similar to Crb, Kib and Wts.

6. Hpo‐YFP localises to the apical junctions in the follicular epithelium of control (fed) egg chambers and becomes strongly enriched into apical junction puncta under conditions of nutrient restriction.

7. HpoKD‐Venus is apically enriched in columnar cells but not in stretch cells (high mag view of right‐hand panel in E).

8. Hpo‐Sav Venus bimolecular fluorescence complementation (BiFC) sensor based on split‐Venus proteins fused to Hpo and Sav. Puncta formation in the cytoplasm is evident. Elevated signal at the apical domain is evident.

9. Hpo‐Sav Venus BiFC sensor levels are elevated upon starvation for 24 h, producing an increase in puncta formation.

10. Endogenously tagged Hpo‐YFP localises to apical punctae, which are more easily detected when the tagged allele is homozygous, rather than heterozygous.

11. Ectopically expressed HpoKD‐Venus BiFC sensor forms larger puncta when expressed at higher levels. Quantification of puncta number and size is shown below. Biological replicates are plotted as individual data points.

Source data are available online for this figure.

Figure EV2. Hpo‐Venus dimers are phosphorylated at Thr183 and condensate formation is inhibited by ectopic expression of active Akt but not Rheb

1. Clonal expression of HpoKD‐Venus dimers stain positively for phospho‐Thr183, indicating phosphorylation of the kinase activation loop. Clonal expression of HpoKD‐Venus dimers fails to enrich apically in the follicular epithelium when co‐expressed with constitutively active Akt. Clonal expression of active Hpo‐Venus dimers localises to large puncta within the cytoplasm which stain positive for phospho Thr183.

2. Hpo punctae induced by nutrient restriction are diminished by constitutively active Akt signalling in the follicular epithelium, but not by expression of Rheb (which activates TOR).

3. Quantification of Hpo punctae size in the clonal conditions described in (A). Biological replicates are plotted as individual data points (n > 8 per sample), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

4. Quantification of Hpo punctae size under nutrient restriction in the presence or absence of constitutively active Akt signalling. Biological replicates are plotted as individual data points (n > 15 per sample), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

5. Minimal autophagic activity is observed in the follicular epithelium of egg chambers from well‐fed females. Nutrient restriction promotes a strong autophagic response in the follicular epithelium. The autophagosome compartment, marked my mCherry‐Atg8a, does not colocalise with Hpo puncta.

Source data are available online for this figure.

Figure 2. Hpo‐Venus punctae are sensitive to 1,6‐hexanediol, enhanced by macromolecular crowding, and undergo fusion and dynamic exchange with the cytoplasm

1. Drosophila egg chambers at stages 9/10 of oogenesis expressing a Hpo kinase dimerization sensor that forms small punctae at the apical domain of the follicle cell epithelium that surrounds the oocyte.

2. Treatment of egg chambers with 1,6‐hexanediol (10%) for 5 min abolishes formation of Hpo sensor punctae.

3. Treatment of egg chambers with 1 M Sorbitol to osmotically withdraw water, and thereby induce macromolecular crowding, is sufficient to induce formation of large micron‐scale Hpo sensor punctae.

4. Nutrient restriction (starvation) of adult females for 24 h induces large micron‐scale Hpo sensor punctae that are similar in size to those in (C).

5. Treatment of egg chambers from adult females starved for 24 h with 1,6‐hexanediol (10%) for 5 min abolishes formation of Hpo sensor punctae.

6. Quantification of the size of Hpo sensor punctae from (A–E). Biological replicates are plotted as individual data points (n = 15 control; n = 8 hex; n = 15 sorb; n = 16 starv; n = 5 starv + hex), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

7. Live‐imaging of HpoKD‐Venus (green) condensates within follicle cells at high magnification reveals fusion events, occurring over the timescale of seconds. White arrowheads point to individual punctae undergoing fusion.

8. Fluorescence Recovery After Photobleaching (FRAP) experiments reveal that photobleaching of individual condensates is followed by rapid recovery after 150 s.

9. Quantification of FRAP recovery as a fraction of the initial fluorescence intensity. The rapid initial slope of recovery suggests dynamic exchange between the condensate and the cytoplasmic pool of the proteins. Temporal averages are plotted as individual data points (n > 3 biological replicates), error bars represent one standard deviation from the mean.

Source data are available online for this figure.

Figure EV3. Hpo kinase condensates do not co‐localise with endosomal markers and are reversibly sensitive to 1,6‐hexanediol treatment

1. A

   Hpo puncta formed at the apical domain of the follicular epithelium under wild‐type conditions or strongly recruited into puncta by Ex overexpression, do not co‐localise with markers for early (Rab5) or late (Rab7) endosomal compartments.

2. B

   Hpo puncta formed at the apical domain of the follicular epithelium under wild‐type conditions or Ex overexpression, do not co‐localise with Rab11, a marker of recycling endosomes.

3. C

   1,6‐hexanediol treatment reversibly disperses HpoKD‐Venus apical clusters.

4. D, E

   1,6‐hexanediol treatment also disperses endogenous Hpo‐YFP clusters that appear under starvation.

Source data are available online for this figure.

Figure 3. Endogenous Hpo‐YFP punctae are regulated by nutritional and mechanical stimuli

1. Apical cross‐sections of Drosophila ovarian follicle cell epithelia at high magnification, showing accumulation of Hpo‐YFP punctae at the apical domain, which is strongly enhanced upon 24 h starvation to reduce Insulin/IGF‐1 signalling. Hpo puncta are lost following a short treatment with 1,6‐hexanediol. Samples were fixed before imaging.

2. Quantification of Hpo‐YFP puncta number and size in (A). Biological replicates are plotted as individual data points (n > 13 for all samples), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

3. Apical cross‐sections of Drosophila ovarian follicle cell epithelia at high magnification, showing loss of Hpo‐YFP punctae upon mechanical strain in “stretch cells”. The Drosophila follicle cell epithelium at stage 10 is composed of anterior stretch cells (flat) and columnar cells. Endogenously tagged Hpo punctae are strongly enriched at the apical domain of columnar cells but not in flattened stretch cells. White asterisks demarcate nuclei of individual stretch cells.

4. Quantification of Hpo‐YFP total intensity and size in (C). Biological replicates are plotted as individual data points (n > 10 for all samples), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

Source data are available online for this figure.

Figure 4. Expanded & Kibra overexpression induces formation of large Hpo punctae in the cytoplasm

1. Expanded (Ex; endogenously tagged with YFP) localises to the apical domain of columnar follicle cells at stages 9/10 of oogenesis.

2. The Hpo dimerization sensor localises in a similar pattern to Ex.

3. Overexpression of Ex is sufficient to induce formation of large micron‐scale condensates in the cytoplasm.

4. Overexpression of Ex is sufficient to recruit the Hpo sensor into large micron‐scale condensates in the cytoplasm.

5. Condensates induced by high levels of Ex in the cytoplasm co‐localise with the Hpo sensor but do not co‐localise with Crb, which remains apical.

6. Condensates induced by high levels of Ex in the cytoplasm immunostain positively for phosphorylated Wts.

7. Moderate overexpression of untagged Kib is sufficient to induce the formation of large micron‐scale Hpo^KD‐Venus condensates that primarily localise apically. DAPI marks nuclei (blue).

8. Overexpression of Kib‐GFP with the Gal4/UAS system is sufficient to induce formation of large micron‐scale cytoplasmic condensates.

Source data are available online for this figure.

Figure 5. Human Kibra overexpression induces formation of large pLATS‐containing punctae in the cytoplasm and induces signalling

1. Human U2OS cells in culture immunostained for endogenous Hippo pathway components reveals punctate subcellular localisation within the cytoplasm.

2. Ectopic expression of Flag‐tagged human Kibra (KIB‐Flag) generates large spherical punctae in the cytoplasm of human Caco2 epithelial cells (top). Co‐localisation of KIB‐GFP with active phosphorylated LATS (pLATS) in the cytoplasm of human Caco2 epithelial cells (middle). Note that cells expressing KIB‐GFP have increased Hippo signalling as indicated by reduced levels of nuclear YAP localisation (bottom). Dashed lines demarcate transfected cells.

Source data are available online for this figure.

Figure EV4. The spectrin cytoskeleton promotes assembly of Hippo signalling condensates via local clustering at the apical plasma membrane

1. α‐Spectrin is required for active Hpo signalling at the apical domain of follicular epithelial cells.

2. Loss of α‐Spectrin significantly reduces Hpo puncta size. Quantification of Hpo punctae size in the presence or absence of α‐Spectrin‐RNAi. Biological replicates are plotted as individual data points (n > 6 per sample), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

3. Endogenously tagged β_H‐Spectrin forms a dense mesh at the apical domain of follicular epithelial cells in stage 9/10 egg chambers.

4. Kibra localises predominantly to the medial‐apical region of follicular epithelial cells, and this localisation is dependent on α‐Spectrin.

5. Mislocalisation of Kibra in α‐Spectrin clones is not a consequence of reduced Kibra levels at the apical domain.

6. Kibra is required for active Hpo signalling at the apical domain of follicular epithelial cells.

7. Disorder prediction of Drosophila Hpo pathway proteins, using two independent algorithms. For meta‐predictor PONDR‐FIT, residues with a score above 0.5 are predicted disordered. The per‐residue confidence score (pLDDT) generated by AlphaFold predicts regions below 50 pLDDT to be unstructured in isolation. AlphaFold structure prediction models for Hpo pathway proteins, colour‐coded according to model confidence. Orange regions correspond to very low model confidence (pLDDT < 50), predicted to represent intrinsically disordered protein sequence.

8. Molecular modelling of Drosophila Hpo components using the protein structure prediction server Phyre2. Where protein structures were incomplete or could not be predicted (e.g. Sav, Kib), protein sequences were submitted to I‐TASSER. The PDB outputs from both servers were uploaded to Illustrate (https://ccsb.scripps.edu/illustrate/) to generate the graphics. Predictive reconstructions of the Crbs‐Ex complexes at the apical junctions and Kib‐Mer complexes at the medial‐apical domain, highlighting the contribution of intrinsically disordered protein structures to both networks. Ex and Kib nucleate Hpo condensate formation in vivo, a process which may be driven by their intrinsically disordered domains.

Source data are available online for this figure.

Figure 6. Intrinsically disordered regions (IDRs) are required for KIB to form large spherical punctae and to activate pLATS

1. Expression of KIB‐GFP, KIB∆C‐GFP or KIB∆M + C‐GFP in human HEK293T cells and co‐staining for pLATS. DAPI marks nuclei in blue.

2. Schematic diagram of KIB intrinsically disordered regions (M) and (C) and the ∆C and ∆M + C deletion constructs.

3. Quantification of KIB‐GFP puncta number and size showing dependence upon the IDRs. Biological replicates are plotted as individual data points (n > 12 for all samples), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

4. Quantification of KIB‐GFP and pLATS intensity showing dependence of both condensate formation and signalling upon the IDRs. Biological replicates are plotted as individual data points (n > 40 for all samples), error bars represent one standard deviation from the mean, and statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

Source data are available online for this figure.

Figure EV5. Phase separation of mammalian Hippo signalling condensates in vitro and in vivo

1. GFP‐LATS1 forms cytoplasmic punctae when expressed in Caco2 epithelial cells. KIB‐Flag colocalises with GFP‐LATS1 in cytoplasmic punctae when co‐expressed in Caco2 epithelial cells.

2. Ectopic KIB‐GFP expression in human Caco2 epithelial cells causes formation of punctae in the cytoplasm. Quantification of KIB puncta size and number reveals a linear correlation with the total expression level of KIB in the cell. Ectopic SAV1‐GFP expression in human Caco2 epithelial cells causes the formation of punctae in the cytoplasm.

3. Purified mYFP‐SAV1:MST2 complex undergoes phase separation to form distinct puncta in a concentration‐dependent manner. Additionally, fewer condensates are observed for complexes containing a variant of SAV1 that lacks the predicted disordered regions, or upon treatment with 1,6‐hexanediol. Biological replicates are plotted as individual data points (n > 4 per sample), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

4. In U‐2 OS cells, expression of monomeric YFP‐tagged SAV1 together with mCherry‐tagged MST2 was sufficient to form condensates in which both proteins co‐localise.

5. In Caco‐2 epithelial cells, expression of monomeric YFP‐tagged Sav1 together with mCherry‐tagged MST2 was sufficient to form condensates in which both proteins co‐localise.

Source data are available online for this figure.

Figure 7. Mechanical strain inhibits the formation of pLATS‐containing punctae and drives YAP nuclear localisation

1. Human Caco2 intestinal epithelial cells are relatively flat when cultured as cell lines in 2D (Caco‐2 cells, top, middle) and cuboidal when cultured in 3D (spheroid, bottom). Punctae of pLATS (green) are strongly enriched at the apical domain of cuboidal cells grown in 3D culture but not in flattened cells grown in 2D culture. Note the increased apical enrichment in medium‐density culture in 2D, where cells become more cuboidal than at low density. Phalloidin staining marks the actin cytoskeleton in red. DAPI marks nuclei in blue.

2. Mouse small intestinal organoids exhibit highly columnar cells which feature a striking concentration of pLATS‐containing punctae into an apical ring and high levels of signalling (YAP becomes largely cytoplasmic).

3. Mouse small intestinal organoids can become stretched upon growth as inflated spheres, which leads to mechanical strain in epithelial cells and reduced pLATS localisation apically, which correlates with reduced Hippo signalling (leading to increased YAP nuclear localisation). White arrows highlight the apical domain of epithelial cells.

4. Quantification of pLATS punctae total intensity in mouse small intestinal organoids from (C). Biological replicates are plotted as individual data points (n > 17 for all samples), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

5. Schematic diagram of YAP nuclear localisation in mechanically stretched cells and YAP cytoplasmic localisation in columnar epithelial cells.

Source data are available online for this figure.

Figure EV6. Apical localisation of pLATS and other Hippo pathway components in intestinal organoids and colorectal epithelial cells in vivo

1. LATS1 and p‐LATS localise apically in intestinal organoids.

2. Loss of apical p‐LATS enrichment (white arrow) in mouse intestinal organoids treated with 10% 1,6‐hexanediol for 5 min.

3. Hippo pathway components localise apically in colorectal cancer epithelial cells from patient biopsies. Data were mined from the Human Protein Atlas (proteinatlas.org).

4. Schematic diagram of Hippo condensate/signalosome formation at the apical domain and its regulation by mechanical strain and growth factor signals.

Source data are available online for this figure.

Figure 8. Growth Factor signalling via Akt inhibits pLATS punctae and increases nuclear YAP

1. Murine small intestinal organoids cultured in control medium or medium supplemented with the NRG1 growth factor, which stimulates organoid growth (see graph, right). Biological replicates are plotted as individual data points (n > 9 per sample), error bars represent one standard deviation from the mean, statistical significance ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05.

2. High‐resolution imaging of intestinal organoids cultured in control medium or medium supplemented with the NRG1 growth factor for 1 h which strongly stimulates pAkt, reduces pLATS staining, and drives YAP to the nucleus.

3. Quantification of pAKT, pLATS and YAP immunostaining from (B).

4. Schematic diagram of columnar intestinal epithelial cells with and without growth factor stimulation.

Biological replicates are plotted as individual data points (n > 40 per sample), error bars represent one standard deviation from the mean, statistical significance was determined using a t‐test ****P < 0.0001 ***P < 0.001 **P < 0.01 *P < 0.05. Source data are available online for this figure.