The desmosome-intermediate filament system facilitates mechanotransduction at adherens junctions for epithelial homeostasis

Summary

Epithelial homeostasis can be critically influenced by how cells respond to mechanical forces, both local changes in force balance between cells and altered tissue-level forces ¹. Coupling of specialized cell-cell adhesions to their cytoskeletons provides epithelia with diverse strategies to respond to mechanical stresses ^2,3,4. Desmosomes confer tissue resilience when their associated intermediate filaments (IFs) ^2,3 stiffen in response to strain ^5–11, while mechanotransduction associated with the E-cadherin apparatus ^12,13 at adherens junctions (AJs) actively modulates actomyosin by RhoA signaling. Although desmosomes and AJs make complementary contributions to mechanical homeostasis in epithelia ^6,8, there is increasing evidence to suggest that these cytoskeletal-adhesion systems can interact functionally and biochemically ^{14–17 8,36,37, 18}. We now report that the desmosome-IF system integrated by desmoplakin (DP) facilitates active tension-sensing at AJ for epithelial homeostasis. DP function is necessary for mechanosensitive RhoA signaling at AJ to be activated when tension was applied to epithelial monolayers. This effect required DP to anchor IFs to desmosomes and recruit the dystonin cytolinker to apical junctions. DP RNAi reduced the mechanical load that was applied to the cadherin complex by increased monolayer tension. Consistent with reduced mechanical signal strength, DP RNAi compromised assembly of the Myosin VI-E-cadherin mechanosensor that activates RhoA. The integrated DP-IF system therefore supports AJ mechanotransduction by enhancing the mechanical load of tissue tension that is transmitted to E-cadherin. This cross-talk was necessary for efficient elimination of apoptotic epithelial cells by apical extrusion, demonstrating its contribution to epithelial homeostasis.

Graphical Abstract

Blurb

Desmosomes (Dsm)-bound intermediate filaments (IF) are important determinants of epithelial mechanics. Nanavati et al. show that the Dsm-IF system supports mechanosensitive RhoA signaling at adherens junctions by increasing the load that tissue tension applies to E-cadherin adhesions. This cross-talk facilitates apoptotic extrusion for homeostasis.

Results and Discussion

Desmoplakin supports tension-activated RhoA signaling at adherens junctions.

DP creates a mechanically integrated system by linking desmosome adhesions to IF ^2,3. Therefore, in order to uncouple the desmosome-IF system in Caco-2 intestinal epithelial cells, we targeted DP^19,20 with two independent siRNAs that significantly reduced total cellular DP (knock-down, KD, Figure S1A–D). Immunofluorescence revealed that IFs extended into cell-cell contacts in control cells (Figure 1A, 2A,G), but this was significantly reduced in DP KD cells, creating IF-sparse gaps between the cytosolic IF pool and the cell-cell borders (Figure 1A, 2G). Quantitative immunostaining showed that junctional levels of the desmosomal cadherins, Desmoglein 2 (Dsg2; Figure S1E) and desmocollin 2 (Dsc2; Figure S1F) were somewhat reduced by DP KD. Electron microscopy of control cells confirmed that IFs extended into electron-dense intercellular junctions (Figure S1J), consistent with desmosomes. Junction-associated IFs were largely absent in DP KD cells and the lateral intercellular spaces were more prominent, but intercellular densities remained evident (Figure S1J). Together, these findings indicate that DP KD uncoupled IFs from epithelial cell-cell junctions, but suggest that the cells remained capable of assembling desmosomal structures, albeit at somewhat reduced levels. Junctional levels of the AJ proteins E-cadherin (Figure S1G), β-catenin (Figure S1H) or α-catenin (Figure S1I) were not changed in DP KD cells. Then we stimulated cell contractility with the phosphatase inhibitor, calyculin A ^13,19,20, to test how DP KD monolayers responded to tensile stress. Cell-cell contacts fracture to create holes in the Caco-2 sheet when tension is increased (Figure S1K) ¹³. The onset of fracture was accelerated by DP KD (Figure S1K,L), consistent with the acknowledged role for the desmosome-IF system in supporting the mechanical resilience of epithelia ^3,6,21.

Figure 1. Desmoplakin supports tension-sensitive RhoA activation.

(A) Effect of DP KD on IF anchorage. Representative images of IFs and E-cad to visualise IF-anchorage in control and DP-depleted monolayers. E-cadherin was used to identify cell-cell contacts. White arrowheads indicate the gap between IF and the cell-cell contacts. Scale bar = 5μm.

(B,C) Effect of DP KD on response of active RhoA at cell-cell junctions to contractile stimulation with calyculin A (100 nM). GTP-RhoA was detected with the AHPH location sensor. (B) Representative images from movies; red arrowheads identify junctions where AHPH increases. (C) Quantification of junctional AHPH, normalized to junctional intensity at the beginning of each movie. DP was depleted by two separate siRNAs. Scale bar = 10μm.

(D,E) Response of junctional GTP-RhoA (AHPH) to monolayer stretch. Control and DP KD monolayers expressing AHPH were grown on flexible substrata before application of stretch. (D) Representative images and (E) quantification of change in junctional AHPH. Red arrowheads indicate junctions that showed an increase in AHPH. n>52 junctions from 3 independent experiments. Scale bar = 10μm.

Error bars represent standard error of mean calculated from mean of three independent experiments analysed using two-way ANOVA (C). In (E) error bars represent standard deviation with individual data points indicated from three independent experiments analysed using One-way ANOVA ****p<0.0001, ***p<0.0005, **p<0.005, *p<0.05. See also Figure S1.

Figure 2. DP recruits IF and dystonin to support tension-activated RhoA.

(A-C) Effect of DPNTP on intermediate filament (IF) coupling to cell-cell junctions. (A) Representative images of control and DPNTP-expressing cells immunostained for keratin 18, ZO1 or DPNTP. (B) Schematic of technique for quantifying IF anchorage to cell junctions, measured as the average distance between IF and cell-cell junctions (detailed description in Methods). (C) Quantification of IF anchorage to cell junctions in control and DPNTP-expressing cells. Scale bar = 10μm. n=10 junctions from 3 independent experiments. White arrowheads indicate the gap between IF and the cell-cell contacts.

(D) Effect of DPNTP on the junctional GTP-RhoA response to tensile stimulation. Control and DPNTP-expressing cells were transfected with AHPH and stimulated with calyculin (100 nM). Quantitation of junctional AHPH fluorescence. Representative images can be found in Supplementary Figure 2E.

(E,F) Development of the linker construct. Desmosome-binding domain is composed of 176 amino acids from the N-terminus of DP and the IF-binding domain is composed of 111 amino acids from the C-terminus of periplakin. (E) Schematic of the linker construct design (Linker-TagRFPT) and (F) its localization in DP KD cells; note that linker co-accumulates with cell-cell junctions marked by Dsg2 (Merged). Scale bar = 10μm. Representative image of linker construct localisation in control cells can be found in Supplementary Figure 2G.

(G,H) Effect of Linker-TagRFPT on IF anchorage to cell-cell junctions after DP KD. Control and DP KD cells expressing Linker-TagRFPT were immunostained for keratin 18 and ZO-1 or the Linker construct to mark cell-cell borders. (G) Representative images and (H) quantification of IF-junction anchorage in cells where Linker-TagRFPT was predominantly at the junctions. n=10 junctions from 3 independent experiments. Scale bar = 10μm. White arrowheads indicate the gap between IF and the cell-cell contacts.

(I) Quantification of junctional fluorescence intensity of AHPH after calyculin A treatment in control, DP KD and linker construct-expressing DP KD cells. Representative images can be found in Supplementary Figure 2H.

(J,K) Effect of DP KD on Dystonin. Control and DP KD cells immunostained with Dystonin and ZO-1 to mark apical cell junctions. (J) Representative image. Red arrowheads: Apical junctional accumulation of Dystonin. (K) Quantification of apical junctional intensity of Dystonin. Scale bar = 20μm. n = 70 junctions from 3 independent experiments.

(L) Quantification of junctional fluorescent intensity of AHPH after calyculin A treatment in control, DST KD and Plectin KD cells. Representative images can be found in Supplementary Figure 3I.

Error bars represent standard deviation with individual data points indicated from three independent experiments analysed using student t-test (C,K) or One-way ANOVA (H). (D,I,L) Error bars represent standard error of mean calculated from mean of three independent experiments analysed using two-way ANOVA. ****p<0.0001, ***p<0.0005, **p<0.005, *p<0.05. See also Figures S2 and S3.

Tension-sensitive RhoA signaling at AJs can also protect epithelial integrity against tensile stresses ¹³. Indeed, disabling this mechanoresponsive RhoA pathway accelerated the sensitivity of Caco-2 monolayers to calyculin A in a fashion that resembled the effect of DP KD ¹³. We therefore asked if DP KD might also affect tension-activated RhoA signaling at cell-cell junctions using AHPH, a location biosensor for active, GTP-RhoA ^22,23. We performed time-lapse imaging after calyculin A treatment to observe changes in GFP-AHPH (Figure 1B). Quantification of fluorescence intensity in control cells revealed that junctional AHPH increased continuously after calyculin A treatment (Figure 1C). In contrast, the increase in junctional intensity was significantly compromised in DP KD cells, an effect that was confirmed with the two siRNAs (Figure 1B,C). Therefore, DP depletion hindered the junctional activation of RhoA in response to calyculin A.

To reinforce this conclusion, we used an alternate method to apply tensile force by growing Caco-2 cell monolayers on a flexible substrate and then subjecting them to external stretch (10% sustained bi-axial, 10 min; Figure 1D,) ¹³. Junctional AHPH intensity increased significantly when control monolayers were stretched, indicating that external application of tension also activated RhoA in these cells (Figure 1 D,E). However, junctional AHPH did not respond to stretch in DP KD monolayers (Figure 1 D,E). Together, these experiments indicate that DP is required to support tension-sensitive RhoA signaling at AJ, providing an alternative mechanism for DP to influence how epithelia respond to tensile stress.

Intermediate filament coupling to desmosomes is necessary for monolayer tension to activate junctional RhoA.

DP is best-understood to mediate the physical linkage of IFs to desmosomes, although it can also interact with other cytoskeletal and signaling proteins ^24,25. Therefore, we asked if IF coupling was critical for DP to support tension-activated RhoA signaling at AJ.

First, we expressed DPNTP, a truncated mutant of DP, that has been reported to act as a dominant inhibitor that can uncouple IFs from desmosomes and affect cell-cell cohesion ^21,26. DPNTP contains the desmosome-binding N-terminus of DP but lacks the IF-binding domain. Transient expression of DPNTP in Caco-2 cells was confirmed by Western blot analysis (Figure S2A) and immunofluorescence demonstrated that the transgene localized to cell-cell junctions (Figure 2A, S2B). Spinning disc microscopy revealed that the subcellular localization of keratin IFs was disrupted by DPNTP, with a clear gap being apparent between the bulk of IFs in the bodies of the DPNTP cells and the cell-cell contacts (Figure 2A–C). Therefore, as previously reported ^21,26, expression of DPNTP could uncouple IFs from desmosomes. As we observed with DP KD, DPNTP-expressing cells showed somewhat reduced junctional levels of Dsg2 and Dsc2(Figure S2C,D).

Then we tested how DPNTP affected the RhoA response to tensile stress, measuring changes in junctional AHPH after stimulation with calyculin A. Whereas control cells showed a rapid and progressive increase in AHPH, this was abolished by DPNTP (Figure 2D, S2E). This suggested that linking desmosomes to IFs might be important for the contribution of DP to tension-activated RhoA signaling.

As an alternative test of this hypothesis, we then developed a gain-of-function strategy to restore desmosome-IF coupling in DP KD cells. For this, we engineered a fusion protein consisting of the desmosomal-binding domain of DP and the keratin 8-binding domain of periplakin ²⁷ (Linker-TagRFPT, Figure 2E, S2F,G), reasoning that this minimal construct could serve as an artificial linker to couple IFs to desmosomes. Linker-TagRFPT expressed efficiently, as measured by Western analysis (Figure S2F) and localized to cell-cell contacts both in WT (Figure S2G) and DP KD cell lines (Figure 2F), as identified with its TagRFPT epitope tag. Linker-TagRFPT effectively restored junctional coupling of IFs to DP KD cells. The aberrant gap between cell-cell contacts and IFs that was evident in DP KD cells was eliminated by expressing Linker-TagRFPT (Figure 2G, H). High-magnification views showed bundles of IFs extending into the cell-cell contacts in the presence of Linker-TagRFPT, as were evident in controls but not KD cells (Figure 2G). Therefore, Linker-TagRFPT provided a tool to more selectively restore junctional coupling of IFs to DP-depleted cells.

Importantly, Linker-TagRFPT, also restored the junctional RhoA response to tension when it was expressed in DP KD cells. Whereas DP KD abolished the junctional RhoA response to tension, KD cells expressing Linker-TagRFPT displayed a rapid-onset, progressive increase in AHPH (Figure 2I, S2H), similar to that seen in control cells. This indicated that coupling of IFs to cell-cell junctions was a critical factor in allowing DP to support tension-activated RhoA signaling.

Anchored IFs can also potentially recruit other proteins to the plasma membrane, such as cytolinkers which can connect IFs to other cytoskeletal components ^26–29. This prompted us to consider whether accessory cytoskeletal proteins might allow DP to support mechanosensitive RhoA signaling at AJ. To test this possibility, we examined whether DP KD affected the immunolocalization of the dystonin and plectin cytolinkers ^28–31, using antibodies whose specificity was demonstrated by reduced staining upon treatment with RNAi for the target proteins (Figure S3A–E). Interestingly, apical junctional staining was evident for dystonin (Figure 2J, S3A,C), but not for plectin (Figure S3D). Junctional dystonin staining depended on DP, being disrupted by DP KD (Figure 2J,K). However, neither Dsg2 nor DP staining were affected by dystonin KD (Figure S3F–H), suggesting that dystonin localization occurred as a response to junctional DP. Finally, we found that the RhoA response to tensile stimulation with calyculin was compromised in dystonin KD monolayers, but not by plectin KD (Figure 2L, S3I). Therefore, although dystonin has previously been identified at hemidesmosomes that link integrins to substrate ^24,32, our data identify a novel pool at cell-cell junctions that is recruited, directly or indirectly, by DP to support mechanosensitive RhoA signaling.

DP supports mechanical loading for mechanotransduction at AJ.

Mechanosensitive RhoA signaling at AJ is activated when tissue level stresses in epithelial monolayers increase molecular-level tension in the E-cadherin adhesion complex ¹³. Increased tension across the cadherin system promotes the association of Myosin VI with E-cadherin, the mechanosensor apparatus which engages a signaling cascade to ultimately stimulate RhoA ¹³. To test if DP KD affected this mechanical signaling process, we first examined molecular-level tension in the cadherin complex when tensile stress was applied to Caco-2 monolayers, reasoning that this was the upstream input signal for mechanotransduction. For this, we immunostained monolayers with the α18 mAb, which recognizes a cryptic epitope in α-catenin that is revealed upon application of tension ³³. The ratio of α18 to total α-catenin immunostaining intensity then represents an index of molecular-level tension at the cadherin complex ³⁴. In order to characterize the early-immediate input for mechanotransduction, these assays were performed shortly after stimulation, when junctions remained intact and before RhoA increased. Mechanical stretch of Caco-2 monolayers increased the α18/α-catenin ratio at AJ (Figure 3A,B), consistent with enhanced tension at the tissue-level of the monolayer being transmitted to cadherin complexes at AJ. In contrast, the α18/α-catenin ratio did not increase in DP KD cells (Figure 3A,B). This was confirmed with calyculin stimulation, where α18/α-catenin ratios increased in control but not in DP KD monolayers (Figure S4A). These results suggested that the integrated demosome-IF system mediated by DP facilitates the transmission of tissue-level tensile stress to mechanically load the cadherin complex.

Figure 3. Desmoplakin facilitates mechanical loading of the E-cadherin system for mechanotransduction of tissue tension.

(A,B) Effect of DP KD on tension in the cadherin complex upon application of sustained monolayer stretch for 8 min. Intensity of α18 immunofluorescence normalised to total α-catenin intensity at junctions was measured in stretched or unstretched Caco2 monolayers. (A) Representative images. (B) Quantification of ratiometric junctional intensity. Scale bar = 10μm. n>48 junctions from 3 independent experiments.

(C) Effect of DP RNAi on Myosin VI recruitment to AJ following contractile stimulation. Control and DP KD cells were stimulated with calyculin (100 nM) and immunostained for Myosin VI and E-cadherin. Junctional fluorescence intensity was quantified. Representative images can be found in Supplementary Figure 4B. n>138 junctions from 3 independent experiments.

(D,E) Effect of DP on the tension-induced biochemical interaction between Myosin VI and E-cadherin. Experiments were performed in control and DP KD Caco-2 cells bearing CRISPR/Cas9-engineered E-cadherin-GFP. (D) Representative E-cadherin-GFP precipitates probed for Myosin VI and E-cadherin, and (E) quantitation of Myosin VI recruitment to E-cadherin in response to calyculin. Western analysis of whole cell lysates of cells stimulated with calyculin or DMSO probed for myosin VI and DP can be found in Supplementary Figure 4C.

(F,G) A constitutively-active Gα12 transgene (Gα12^CA) rescues tension-activated RhoA signaling to DP KD cells. Junctional GTP-RhoA was identified with AHPH in control, DP KD and DP KD cells reconstituted with Gα12^CA (DP KD + Gα12^CA). (F) Representative images of AHPH before and after stimulation with calyculin (100 nM; red arrowheads show increased AHPH at junctions) and (G) quantitation of the response. Scale bar = 10μm. n>20 junctions from 3 independent experiments.

Error bars represent standard deviation with individual data points indicated from three independent experiments analysed using One-way ANOVA (B,C,G) or (E) Student t-test. ****p<0.0001, ***p<0.0005, **p<0.005, *p<0.05. See also Figure S4.

Then, we asked if DP KD affected the interaction of Myosin VI with E-cadherin, as an index of how the mechanotransduction apparatus responds to tension. Immunofluorescence showed that Myosin VI staining at cell-cell junctions increased when calyculin was applied to control monolayers (Figure 3C, S4B). However, DP KD monolayers showed reduced levels of Myosin VI at baseline, compared with controls, and these failed to increase further on stimulation with calyculin A (Figure 3C, S4B). As the total cellular levels of Myosin VI were not altered by DP KD (Figure S4C), this suggested that DP KD compromised the junctional recruitment of Myosin VI in response to tensile stress. We confirmed this with co-precipitation analyses, using a Caco-2 cell line whose E-cadherin was engineered by CRISPR/Cas9 genome editing to bear a C-terminal GFP tag ³⁵. E-cadherin-GFP was isolated with GFP-Trap and the protein complexes probed for Myosin VI (Figure 3D). As previously reported ¹³, calyculin A increased the amount of Myosin VI that associated with E-cadherin in control cultures, but this was substantially compromised by DP KD (Figure 3D,E). This effect was not confined to intestinal epithelial cells as DP KD also compromised the ability of calyculin to promote assembly of the Myosin VI-E-cadherin mechanosensor in MCF-7 mammary epithelial cells (Figure S4D). Together, these findings indicate that DP influences AJ mechanotransduction by supporting the strength of the mechanical signal that is applied to the cadherin complex when monolayer tension is increased. Such an action on the strength of the input signal can then account for engagement of AJ mechanotransduction, as reflected in assembly of the Myosin VI-E-cadherin mechanosensor apparatus and downstream activation of RhoA.

We further reasoned that the impact of depleting DP would be ameliorated if we could activate downstream elements of the E-cadherin-Myosin VI pathway. To do this, we focused on the Gα12 G protein, which is recruited to the E-cadherin complex when tension is applied, and then engages the p114 RhoGEF to activate RhoA ¹³. We expressed a constitutively active form of Gα12 (Gα12^CA) in DP KD cells and used the AHPH assay to measure the RhoA response to calyculin. Expression of Gα12^CA restored the tension-activated RhoA response in DP KD cells towards those of control levels (Figure 3F,G), indicating that activation downstream in the E-cadherin-Myosin VI pathway could overcome the effects of DP depletion. Altogether, these results indicate that DP supports tension-activated RhoA signaling at AJ by facilitating the assembly of the E-cadherin and Myosin VI tension-sensing apparatus.

DP supports apoptotic extrusion via AJ mechanotransduction.

Increasing evidence indicates that homeostatic disturbances are identified by epithelia because they cause local disturbances in the balance of forces between cells. For example, apoptotic cells become hyper-contractile ^36–38 and are eliminated by apical extrusion when that force imbalance is detected at AJ ¹⁷. As demosomes have recently been implicated in apoptotic extrusion ³³, we asked if this was mediated by an impact on AJ mechanotransduction.

First, we confirmed that DP was necessary for extrusion of sporadic apoptotic cells, induced by laser microirradiation of nuclei in E-cadherin-GFP Caco-2 monolayers. Time-lapse imaging confirmed that the apoptotic cell was apically expelled from control monolayers and the neighbours of the dying cell formed a rosette-like structure that sealed the monolayer (Figure 4A). However, as previously reported ³⁹, apical extrusion was compromised by DP KD, causing the apoptotic cells to be retained within the monolayers (Figure 4A,B). Effective apoptotic extrusion requires the engagement of AJ mechanotransduction in the neighbour cells. Therefore, we devised mixing experiments to test which cell population required DP for apoptotic extrusion to occur (Figure 4C). For these experiments, apoptosis was induced in Caco-2 cells that express the pro-apoptotic protein, p53 Upregulated Modulator of Apoptosis (PUMA-mCherry), under control of the Tet-ON promoter system ¹⁹. PUMA^TetON cells were then mixed with E-cadherin-GFP cells at a ratio (1:100) that created small groups of PUMA^TetON cells surrounded by PUMA^TetON-null cells. Induction of PUMA-mCherry expression with doxycycline (5 hr) caused the sporadic PUMA^TetON cells to undergo apoptosis and be apically extruded from the monolayer (Figure 4C). We then counted the proportion of apoptotic cells that underwent extrusion as an index of apoptotic extrusion.

Figure 4. Desmoplakin supports RhoA activation in neighbours of apoptotic cells to drive their apical extrusion.

(A,B) Effect of DP on the apical extrusion of apoptotic cells (apoptotic extrusion). Apoptosis was induced by laser micro-irradiation of selected cells in control and DP KD monolayers. (A) Stills from time-lapse movies and (B) quantitation of apoptotic extrusion. Red asterisk: cells undergoing apoptosis, Red arrowhead: formation of rosette-like structure by neighbours of the dying cell. Scale bar = 10μm.

(C-E) DP is selectively required in epithelium surrounding apoptotic cells to support apical extrusion. Apoptosis was induced with Tet-inducible PUMA (see main text) and mixing experiments were used to deplete DP selectively from the PUMA cells (DP KD in PUMA cells, mixed 1:100 with WT E-cad-GFP cells) or its surrounding epithelium (neighbours; PUMA cells with a WT-background mixed with DP KD in E-cadherin GFP cells, 1:100). (C) Stills from representative movies (imaged for 5 h after induction of PUMA with doxycycline and quantification of apical extrusion (% apoptotic cells that underwent extrusion) when DP was depleted in the PUMA (apoptotic cell; D) or in the surrounding epithelium (E). Red asterisk: cells undergoing apoptosis, Red arrowheads: formation of rosette-like structure by neighbours of the dying cell. Scale bar = 15μm.

(F,G) Effect of DP on the neighbour-cell RhoA response during apoptotic extrusion. Apoptosis was induced by laser microirradation of AHPH-negative cells (red asterisks) surrounded by AHPH-positive neighbours. DP siRNA were transfected to the whole cultures. (F) Stills from videos showing the RhoA response in neighbour cells, manifest as AHPH recruitment to the interface with the apoptotic cell (green arrowheads in controls) and (G) Quantitation of AHPH intensity in neighbour cells at their interface with the apoptotic cells. Scale bar = 15μm.

(H,I) Effect of rescuing IF anchorage on apoptotic extrusion in DP-depleted cells. Apoptosis was induced by 500nM/mL etoposide treatment for 6 hours in control, DP KD monolayers, and DP KD monolayers retransfected to express Linker-TagRFPT. Cleaved caspase 3 (magenta) was used to mark apoptotic cells and E-cad-GFP (cyan) was used to mark cell boundaries. (H) Representative images of cells undergoing apoptosis (I) Quantitation of apical extrusion (% apoptotic cells undergoing apical extrusion). Scale bar = 15μm.

Error bars represent standard error of mean from three independent experiments analysed using student t-test (B), One-way ANOVA (I) or analysed using two-way ANOVA (D,E,G). ****p<0.0001, ***p<0.0005, **p<0.005, *p<0.05.

To test if effective extrusion required DP in the apoptotic cell, we transfected PUMA^TetON cells separately with DP siRNA and mixed these with E-cadherin-GFP cells that had not been transfected with siRNA. Surprisingly, DP-depleted PUMA^TetON cells were extruded as efficiently as PUMA^TetON cells that retained DP (Figure 4C,D). This implied that DP was not required in the apoptotic cell for extrusion to occur. Then, to test a role for DP in the neighbour cell population, we mixed PUMA^TetON cells with E-cadherin-GFP cells that had been separately transfected with DP siRNA. Here there was a significant decrease in the proportion of apoptotic cells that were extruded (Figure 4C,E). This indicated that DP is selectively required in the epithelia around apoptotic cells for apical extrusion to be effectively achieved. This coincides with the cell population where AJ mechanotransduction engages RhoA.

Then, we used AHPH to test if DP siRNA affected the RhoA response in the neighbours of apoptotic cells (Figure 4F,G). To ensure that we only measured GTP-RhoA in the neighbours, we mixed WT cells with GFP-AHPH expressing cells (1:100 ratio), then induced apoptosis by laser microirradiation of cells not expressing GFP-AHPH. In control monolayers, AHPH in neighbours progressively accumulated at their immediate interface with the apoptotic cell as extrusion occurred (Figure 4F,G). However, this RhoA response was significantly compromised by DP siRNA, applied to all the cells in the culture, both those targeted by laser microirradiation as well as the surrounding cells. Therefore, the capacity for DP to support tension-activated RhoA signaling provided a potential mechanism for it to facilitate apoptotic extrusion. Finally, since coupling of IFs was necessary for DP to support AJ mechanotransduction, we tested if this also applied for apoptotic extrusion. For this, we induced apoptosis (etoposide, 6 hr) in DP KD cells that expressed Linker-TagRFPT and compared the proportion of apoptotic cells that were extruded with that in controls or DP KD cells (Figure 4H, I). Strikingly, Linker-TagRFPT restored the extrusion that was impaired by DP KD to control levels. This confirmed that the integrated desmosome-IF system was necessary for apoptotic extrusion, as it was for tension-activated RhoA signaling.

Conclusion

Local changes in the balance of forces between cells, associated with events such as apoptosis, are increasingly implicated in epithelial homeostasis ¹. Our current data suggest that the integrated desmosome-IF system contributes to this mechanobiology of homeostasis by modulating tension-sensing at AJ. Thus, DP supports mechanosensitive RhoA signaling at AJ by coupling IFs to desmosomes and recruiting the dystonin cytolinker ¹³. A key to understanding the underlying mechanism lies in finding that DP KD reduced the mechanical signal that is applied to the cadherin complex when monolayers come under increased tension. One explanation is that the desmosome-IF system which DP integrates ^6,8 amplifies the mechanical loading that cellular tension applies to E-cadherin. This implies some physical interaction between AJs and desmosomes, which could be mediated by interactions between the adhesion receptors themselves ⁴⁰ and/or coupling of their associated cytoskeletons by cross-linkers, such as dystonin. This effect on the upstream input signal would be predicted to alter the strength of the mechanotransduction response, exactly as we saw for assembly of the Myosin VI-E-cadherin mechanosensor apparatus and downstream RhoA signaling. Tension-activated RhoA signaling contributes to diverse processes that range from apical extrusion ¹⁹ to collective epithelial cell migration ⁴¹. Indeed, our experiments show that the connection with AJ mechanotransduction accounts for the ability of DP to support apoptotic extrusion ³⁹. How the desmosome-IF system affects mechanical loading of cadherins is an important question for future work. Nonetheless, our findings identify mechanical connectivity as a key avenue for these adhesion-cytoskeletal systems to collaborate for epithelial homeostasis.

Limitations of this study

Here we highlight two issues for future consideration. First, what is the relationship between desmosome-based cell-cell adhesion and the cross-talk pathway that we have identified? Manipulation of DP in our experiments was associated with a reduction in mature demosome assembly and more prominent lateral intercellular spaces, suggesting a change in adhesion. It is possible that these are secondary consequences of the defect in tension-sensitive RhoA signaling that we identified. Both desmosome assembly ⁴² and cell-cell adhesion ⁴³ are supported by RhoA. However, we cannot exclude the potential for these changes to also be involved in defective mechanotransduction. Of note, a reduction in desmosomal cell-cell adhesion would be predicted to reduce the mechanical load that this system bears when tissue tension is increased, and thereby the load that is transferred to AJ-based mechanotransduction. Second, proteins such as plakoglobin ², can associate with both desmosomes and classical cadherins, providing additional pathways for cross-talk at the mechanical and/or signaling levels. Elucidating these issues will be important goals for future research.

STAR Methods.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Alpha Yap: a.yap@uq.edu.au

Materials availability

All reagents generated in this study will be made available upon request to the lead contact

Data and code availability

• All microscopy data and original western blot images reported in this manuscript will be shared by the lead contact, Alpha Yap, upon request.

• This paper does not report any original code.

• Any additional information required to reanalyse the data reported in this work is available from the lead contact, Alpha Yap, upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell culture and transfection

Human colorectal adenocarcinoma (Caco-2) cells were obtained from ATCC (HTB-37) and cultured in Roswell Park Memorial Institute (RPMI) media supplemented with 10% Fetal Bovine Serum (FBS), 1% L-Glutamine, 1% MEM non-essential amino acids and 100 units/mL of penicillin/streptomycin. Human Embryonic Kidney 293 (HEK293T) cells and Michigan Cancer Foundation-7 (MCF-7) cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% FBS, 1% L-Glutamine, 1% MEM non-essential amino acids and 100 units/mL of penicillin/streptomycin. All cells were grown in a humidified chamber at 37°C in 5% CO₂ atmosphere.

Caco-2 cells were transfected with plasmids using Lipofectamine 3000 (Thermo Fisher Scientific #L3000015) according to the manufacturer’s protocol. For siRNA transfection, Caco-2 or MCF-7 cells were transfected with Lipofectamine RNAiMAX (Thermo Fisher Scientific #13778150) according to the manufacturer’s protocol. When transfected with plasmids, cells were analysed 24 hours post transfection. When transfected with siRNA, cells were analysed 48–72 hours post transfection.

Stable cell lines and transduction

Caco-2 E-cad-GFP knock-in cell line and MCF-7 E-cad-GFP knock-in cell line were established using CRISPR-Cas9 technology as described previously ³⁵. PUMA^TET-ON Caco-2 cells, and GFP-AHPH Caco-2 cells were generated as described previously ¹⁹. The Caco-2-Linker-TagRFPT cell line was generated using a lentiviral-based transduction technique. To generate lentiviruses, HEK-293T cells were transfected with a lentiviral expression vector, pTripZ and packaging constructs, pMDLg/pRRE, RSV-Rev and pMD.G using Lipofactamine 2000 according to the manufacturer’s protocol. The virus was concentrated from the supernatant of the HEK-293T cells after 48 hours of incubation, using concentrating spin columns. Subsequently, Caco-2 cells were infected with the concentrated virus. TagRFPT-positive Caco-2 cells were selected using FACS Aria Cell Sorter (Queensland Brain Institute, The University of Queensland).

METHOD DETAILS

DNA constructs

pC1 linker construct was cloned by Gene Universal, a gblock was synthesized containing 176 AA from N-terminus of DP and 111 AA from C-terminus of periplakin. The gblock was cloned into p-C1 vector by restriction digestion with Age1 and BamH1. Linker construct was subsequently cloned into pLVX-IRES-Puro lentiviral backbone by restriction digestion with Age1 and Hpa1. To transfer DPNTP-GFP into a lentiviral backbone, DPNTP-GFP containing Mlul and XbaI restriction sites was generated using PCR. DPNTP-GFP was subsequently cloned into pLVX-IRES-Puro lentiviral backbone by restriction digestion with Mlul and XbaI.

CalyculinA treatment

Intrinsic contractility was increased by treating 90–100% confluent Caco-2 cell monolayers with 100nM/mL calyculinA (Abcam #ab141784) for 10–15 min at 37°C.

Induction of apoptosis

a. Laser injury

E-cad-GFP Caco2 cells were cultured on 35 mm glass-bottom dish. When cells reached 90–100% confluency, apoptosis was induced by creating DNA damage using Mai Tai two-photon laser. To record cellular responses to apoptosis, time-lapse imaging was performed continuously for 30 min. AnnexinV (Life Technologies #A23204) was used as a marker to identify cells undergoing apoptosis.

b. Tet-inducible PUMA

PUMA^TetON cells were mixed with E-cad-GFP-Caco2 cells at a 1:100 ratio and cultured on 35 mm glass bottom dish, to create small clusters of PUMA cells surrounded by E-cad-GFP-Caco2 cells. When cells reached 90–100% confluency, apoptosis was induced in PUMA^TetON cells by treating the mixed monolayer with 1μg/mL doxycycline. To allow enough time for expression of PUMA to induce apoptosis and record cellular responses to apoptosis, time-lapse imaging was performed for 5 hours after the treatment with doxycycline.

c. Etoposide treatment

E-cad-GFP Caco-2 cells or E-cad-GFP Linker-TagRFPT Caco2 cells were cultured on coverslips in 6-well plates. When the cells reached 90–100% confluency, they were treated with 500nM/mL etoposide (Adooq Biosciences LLC #A10373) for 6 hours at 37°C. After the treatment, cells were fixed with 100% methanol and stained for Cleaved Caspase 3 to mark apoptotic cells and GFP to stain for E-cad to identify cell borders.

Immunofluorescence

Caco-2 cells were fixed using 4% Paraformaldehyde (PFA) in cytoskeleton stabilisation buffer (10mM PIPES at pH 6.8, 100mM KCl, 300mM sucrose, 2mM EGTA and 2mM MgCl₂) for 20 min at room temperature (RT) or were fixed using 100% methanol for 10 min at −20°C. Staining for MyosinVI, DP, keratin-8, keratin-18, Dsg2, tagRFPT and GFP required pre-permeabilisation of cells in pre-permeabilisation buffer (0.1% Triton in PBS) for 3 min on ice before fixation with PFA. Subsequently, the fixed cells were blocked in 3% Bovine Serum Albumin (BSA) in PBS at RT for 1 hour. Following this, cells were incubated with corresponding primary antibody for 1 hour at RT. For MyosinVI and Dsg2 primary antibodies, the cells were incubated overnight at 4°C. Then, the cells were washed with 0.05% Tween in PBS before incubating with the corresponding secondary antibodies at RT for 1 hour. Following this, the coverslips were mounted using Prolong gold (Genesearch #8916BC) with or without DAPI.

Immunoprecipitation

For GFP-trap experiments, Caco-2 E-cad-GFP cells or MCF-7 E-cad-GFP cells were cultured on 10 cm dishes. After 72–96 hours of plating the cells, they were treated with calyculinA or DMSO for 10 min at 37°C. Then, the cells were lysed for 30 min in lysis buffer (1% NP40, 10mM Tris-HCl pH7.4, 150mM NaCl, 2mM CaCl2, 1x protease inhibitor). Meanwhile, GFP beads (Chromotek #gta-20) were blocked in 3% BSA in Tris-Buffered Saline (TBS) for 4 hours at 4°C. The cell lysates were incubated with GFP beads overnight at 4°C. After washing the beads with lysis buffer several times, the protein complexes were resolved using SDS-PAGE.

Western blotting

Caco-2 cells were lysed in 1x lysis buffer (4x: 200mN Tris-HCl, 40% glycerol, 8% SDS and0.4% Bromophenol Blue) and incubated at 95°C for 9 min. The cell lysates were resolved in 8% or 12% SDS polyacrylamide gels. Then the samples were transferred to nitrocellulose membrane and blocked with 5% milk in TBS for 1 hour at RT. The membranes were incubated with primary antibodies overnight at 4°C. Later, the membranes were incubated with the corresponding horseradish (HRP) – conjugated secondary antibodies for 1 hour at RT. The blots were resolved using Supersignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher #34579) and imaged on Biorad Chemidoc. For myosinVI staining, the membranes were incubated with Alexa Fluor conjugated secondary antibodies. These blots were imaged using Odyssey CLX.

Microscopy

Fixed samples were imaged on Andor Dragonfly inverted confocal spinning disc microscope operated using Fusion and Imaris software. Images were acquired using 60x, 1.4 NA Plan Apo objective. Fixed samples were imaged on Upright Zeiss Axiolmager widefield microscope operated using Zeiss ZEN blue 3 software. Images were acquired using 40x, 0.95 NA Plan Apo objective or 63x, 1.40 NA Plan Apo objective. Time-lapse imaging after calyculinA treatment and for PUMA-induced apoptotic extrusion was performed on Nikon Ti-E deconvolution microscope operated using NIS elements software was used. Images were acquired using 20X, 0.75 NA Plan Apo objective.Time-lapse imaging for laser-induced apoptotic extrusion was performed on Zeiss LSM 710 Meta confocal microscope equipped with Mai Tai multi-photon laser operated using Zen Black software. Images were acquired using 43X, 1.3 NA Plan Apo objective. Subcellular distribution of E-cadherin and Desmoglein 2 was visualised on Leica SP8 STED 3X FLIM super resolution microscope operated using Leica LASX software. Images were acquired using HC Plan Apochromat 93×1.30 glycerol objective.

Electron Microscopy

Caco2 cells were fixed in 2.5% glutaraldehyde in PBS and embedded in Epon as described previously ⁴⁴. The sections were poststained with aqueous uranyl acetate (2%) and Reynold’s lead citrate ⁴⁵. Micrographs were acquired using a JEOL1011 transmission electron microscope at 80 kv fitted with a Morada CCD camera under the control of iTEM imaging software.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantitative Image analysis

All the quantitative image analysis was performed using ImageJ software.

a. Initiation of cell separation

The initiation of cell separation was quantified by noting the time point after calyculinA treatment when there was formation of gap in the cell monolayers.

b. Intermediate filament anchorage

The anchorage of IFs was measured as the average distance between IFs and cell-cell borders. We quantitated IF recruitment at entire bicellular junctions (i.e. from one multicellular vertex to another), as illustrated in the cartoon in Fig 2B. We measured the area of the gap (black space) between a bicellular junction (defined as between two vertices) and the body of IFs. Dividing this area by the length of the bicellular junction gave us the average distance between that junction and the IFs.

c. Apoptotic extrusion

Apoptotic extrusion was classified successful when: (1) the apoptotic-marker-positive cell expelled out of the monolayer, and (2) the neighbours of the apoptotic cell formed a rosette-like structure to avoid formation of gaps in the monolayer. Apoptotic extrusion was classified as a failure when (1) the apoptotic-marker-positive cell retained in the monolayer, and (2) the neighbours of the apoptotic cell failed to elongate to form a rosette-like structure leading to formation of gaps in the monolayer.

d. Junctional intensity

Junctional intensity of AHPH and myosinVI was measured by taking a ratio of the junctional intensity and cytoplasmic intensity. For measuring the changes in junctional intensity of AHPH after calyculinA treatment or induction of apoptosis, ratio of the junctional intensity of AHPH at multiple time points to the junctional intensity of AHPH before calyculinA treatment or the induction of apoptosis was taken.

e. Immunoprecipitation

To quantitate this, we normalized the Myosin VI signal to that of E-cadherin-GFP, then derived the ratio of the normalized Myosin VI signal in calyculin-treated cells compared with DMSO-treated controls.

Statistical analysis

All the statistical analysis was performed using GraphPad Prism 8 software. Variance of data is presented as SD or SEM of the mean, as specifically identified in the figure captions. All the experiments were performed at least 3 times independently. For 2 groups, student t test was used to compare the datasets. For 3 or more groups, one-way annova or two-way annova was used to compare the datasets depending on the number of variables. P<0.05 was considered significant. Statistical parameters for the individual experiments are included in the specific figure legends. The parameters include sample size (N), statistical test performed and statistical significance.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti cytokeratin 18	Abcam	Cat#ab668 RRID:AB_2533147
Rabbit monoclonal anti E-cadherin	Cell technologies	Cat#3195 RRID:AB_2291471
Mouse monoclonal anti E-cadherin (hECD1)	Gift from Dr. M Takeichi	N/A
Rabbit polyclonal anti Desmoplakin (NW6)	Made by Green lab	N/A
Rabbit polyclonal anti Desmoplakin (NW161)	Made by Green Lab	N/A
Mouse monoclonal anti Desmoplakin (DP2.15)	Abcam	Cat#ab16434 RRID:AB_443375
Rabbit polyclonal anti Zonula Occludens (ZO-1)	Thermo Fisher	Cat#61–7300 RRID:AB_2533938
Mouse monoclonal anti Zonula Occludens (ZO-1)	Thermo Fisher	Cat#33–9100 RRID:AB_2533147
Mouse monoclonal anti Desmoglein-2	Thermo Fisher	Cat#32–6100 RRID:AB_2533089
Rabbit polyclonal Desmoglein 2	Abcam	Cat#ab96761 RRID:AB_10679032
Mouse monoclonal Desmocollin 2/3	Thermo Fisher	Cat# 32–6200 RRID:AB_2533090
Mouse monoclonal anti β-catenin	BD Biosciences	Cat#610154 RRID:AB_397555
Rabbit polyclonal anti α-catenin	Thermo Fisher	Cat#71–1200 RRID:AB_2533974
Rabbit polyclonal anti Green Fluorescent Protein (GFP)	Thermo Fisher	Cat#A6455 RRID:AB_221570
Chicken monoclonal anti Green Fluorescent Protein (GFP)	Abcam	Cat#ab13970 RRID:AB_300798
Rabbit polyclonal anti Tag Red Fluorescent Protein (tRFP)	Evrogen	Cat#AB233 RRID:AB_2571743
Rabbit polyclonal anti Dystonin/BPA	Abcam	Cat#AB234644
Mouse monoclonal anti Dystonin/BPA (clone 279)	Cosmo Bio	Cat# CAC-NU-01-BP1 RRID:AB_1961833
Rabbit polyclonal anti GAPDH	R&D Systems	Cat#2275-PC-100 RRID:AB_2107456
Rabbit monoclonal anti Plectin	Thermo Fisher	Cat#MA5–32102 RRID:AB_2809395
Rat monoclonal anti α-catenin (α18)	Gift from Dr. A. Nagafuchi	N/A
Rabbit polyclonal anti Myosin VI	Sigma Aldrich	Cat#M5187 RRID:AB_260563
Rabbit polyclonal anti Myosin VI	Gift from Prof. F. Buss	N/A
Rabbit polyclonal anti cleaved caspase 3	Cell Signaling Technology	Cat#9661 RRID:AB_2341188
Goat anti mouse, rabbit, rat Alexa-Fluor-488	Thermo Fisher	Cat#A-11001 RRID:AB_2534069 Cat#A-11008 RRID:AB_143165 Cat#A-11006 RRID:AB_2534074
Goat anti mouse, rabbit, rat Alexa-Fluor-594	Thermo Fisher	Cat#A-11032 RRID:AB_2534091C at#A-11037 RRID:AB_2534095 Cat#A-48264 RRID:AB_2896333
Goat anti mouse, rabbit, rat Alexa-Fluor-647	Thermo Fisher	Cat#A-21236 RRID:AB_2535805 Cat#A-21245 RRID:AB_2535813 Cat#A-48265 RRID:AB_2896334
Goat anti Chicken Alexa-Fluor-488	Thermo Fisher	Cat# A-11039 RRID:AB_2534096
Goat anti mouse HRP	Bio-Rad	Cat#1706516 RRID:AB_2921252
Goat anti rabbit HRP	Bio-Rad	Cat#1706515 RRID:AB_11125142
Goat anti mouse IgG H&L (IRDye 680CW)	Abcam	Cat# ab216776 RRID:AB_2933974
Goat anti rabbit IgG H&L (IRDye 800RD)	Abcam	Cat# ab216773 RRID:AB_2925189
Bacterial and virus strains
NEB 10-beta Competent E.coli	New England Biolabs	Cat#C3019H
Stellar Competent E.coli	Clontech	Cat#636766
Chemicals, peptides, and recombinant proteins
Lipofectamin 3000	Thermo Fisher	Cat#L3000015
Puromycin	Sigma-Aldrich	Cat#P8833
Lenti-X concentrator	Clontech	Cat#631232
Calyculin A (phosphatase inhibitor)	Abcam	Cat#ab141784
Annexin V Alexa Fluor™ 647 conjugate	Thermo Fisher	Cat#A23204
DRAQ7™	Abcam	Cat#ab109202
Doxycycline	Sigma-Aldrich	Cat#D9891
Etoposide	Sigma-Aldrich	Cat#E1383
Chemiluminescent Substrate	Thermo Fisher	Cat#34579
ProLong Gold with DAPI	Cell Signaling	Cat#8961
ProLong Gold without DAPI	Cell Signaling	Cat#9071
Experimental models: Cell lines
Caco2	ATCC	HTB-37 RRID:CVCL_0025
HEK293T	ATCC	CRL-3216 RRID:CVCL_0063
MCF7	ATCC	HTB-22 RRID:CVCL_0031
Oligonucleotides
siRNA against Desmoplakin (DP siRNA 1) 5’ CGACAUGAAUCAGUAAGUA ’3	IDT	N/A
siRNA against Desmoplakin (DP siRNA 2) 5’ GAAGAGAGGUGCAGGCGUA ’3	IDT	N/A
siRNA smartpool against Dystonin containing the following sequences: 5’ GACCUAAGGACUCGAUAUA ’3 5’ CAGCAGAUCUCAUUAUUCA ’3 5’ GAAUUGAGCAACAGUAUCA ’3 5’ GCAGAUUGACAACAGGUUA ’3	Dharmacon	Cat#LQ-011596–00–0002
siRNA smartpool against Plectin containing the following sequences: 5’ GCACUCAUCUUGCGUGACA ’3 5’ UCGCAGGGCUGUUGCUGAA ’3 5’ GGCAAGACGGUGACCAUUU ’3 5’ GAAGAGACACAGAUCGACA ’3	Dharmacon	Cat#LQ-003945–00–0002
Recombinant DNA
GFP-AHPH	Gift from Dr. M. Glotzer	Piekny et al.22
Linker-Tag RFPT	This study	N/A
DPNTP-GFP	N/A	Broussard et al.9 Bornslaeger et al.26 Huen et al.21
GFP-G protein-alpha 12 FL (Q231L)	Constructed and validated in this lab.	Acharya et al.13
Tet-on-HA-PUMA-PGK-mCherry-CAAX	Constructed and validated in this lab.	Duszyc et al.19
Software and algorithms
Fiji	ImageJ	https://imagei.net/Fiii
Prism 9.0.0	Graphpad	https://www.graphpad.com/scientific-software/prism/
Adobe Illustrator	Adobe	https://www.adobe.com/

Highlights.

• Desmoplakin (DP) supports tension-sensitive RhoA signaling at adherens junctions.

• DP supports cross-talk by coupling intermediate filaments (IF) to desmosomes.

• The Desmosome-IF system supports mechanical loading at adherens junctions.

• This cross-talk pathway supports apoptotic extrusion for epithelial homeostasis.