The RhoGAP ARHGAP32 interacts with desmoplakin, and is required for desmosomal organization and assembly Free

Cell–cell junctions play a crucial role in the formation and maintenance of tissue barrier integrity (Adil et al., 2021; Garcia et al., 2018; Wu and Sun, 2023). Desmosomes, acting as ‘rivets’ between keratinocytes, are abundant in tissues that experience considerable mechanical stress, such as the skin (Bharathan et al., 2023; Kowalczyk and Green, 2013). Desmogleins (DSGs) and desmocollins (DSCs), facilitate cell-to-cell connections via extracellular adhesive interactions, whereas the armadillo proteins including plakoglobin (PKG; also known as JUP) and plakophilins (PKPs) play a role in forming the intracellular plaque. Desmoplakin (DSP) is proposed to specifically link PKG and PKP proteins to the intermediate filament network (Kowalczyk and Green, 2013). Genetic disorders in desmosomal components cause diseases, such as pemphigus, cardiomyopathies and cancers (Broussard et al., 2015; Kottke et al., 2006; Schmidt et al., 2019; Spindler et al., 2023).

The actin cytoskeleton is involved in the assembly and turnover of desmosomes (Fuchs et al., 2023; Godsel et al., 2010; Hiermaier et al., 2021; Moch et al., 2022). The relocation of DSP to desmosomes during desmosome assembly can be accelerated by active RhoA, but the maturation of desmosomes is hindered by persistent activation of RhoA (Godsel et al., 2010). Overexpressing a catalytic DHPH domain of the RhoA activator ARHGEF11, which enhances actomyosin contractility, also leads to downregulation of DSP and DSG1 (Ning et al., 2021). Additionally, induction of a microtubule-severing protein, spastin, in differentiated epidermal cells, which also leads to increased actomyosin contractility, decreases levels of cortical DSG1, DSP, and DSC2 (also known as DSC3) (Muroyama and Lechler, 2017). Therefore, an overload of actomyosin contractility caused by increased RhoA needs to be controlled to facilitate desmosome assembly. However, little is known how F-actin is controlled during desmosomal turnover and assembly.

Desmosomes also function as a signaling hub to influence actin organization (Muller et al., 2021). PKP1 colocalizes with F-actin and is essential for cortical actin organization. Expression of PKP1 lacking its desmosomal-binding domain induces formation of filopodia and long protrusions (Godsel et al., 2010; Hatzfeld et al., 2000; Keil et al., 2016). PKP2 recruits active RhoA to the cell–cell interface, driving actomyosin filament reorganization. Similarly, loss of DSP results in increased cellular migration and filopodia length and aberrant Rac1 signaling (Bendrick et al., 2019; Godsel et al., 2010). Moreover, Tctex-1 (also known as DYNLT1), a part of the dynein complex, regulates localization of Dsg1 which recruits cortactin–Arp2/3 actin nucleation complexes to promote actin polymerization (Nekrasova et al., 2018). Desmosome-depleted junctions have aberrant actomyosin networks and weakened junctional tension, suggesting a mechanical coupling between desmosomes and actomyosin contractility (Thomas et al., 2020). In pemphigus, autoantibodies against DSG1 and DSG3 disrupt desmosomes, and the downregulation of RhoA activity contributes to the pathogenesis of the disease (Jin et al., 2021; Spindler and Waschke, 2011; Waschke et al., 2006). However, how desmosomes influence F-actin, and whether this is direct or occurs indirectly through players such as Rho guanine exchange factors (GEFs) or RhoGTPase-activating proteins (GAPs), is still an open question.

Here, we demonstrated that ARHGAP32 is localized in desmosomes by interacting with DSP via its GAB2-interacting domain (GAB2-ID). KO of ARHGAP32 impaired desmosomal organization and length, as well as their assembly and maturation. ARHGAP32 inhibition also increased stress fiber formation across the nucleus and led to T18/S19 phosphorylation of the regulatory myosin light chain Myl9 (hereafter simply denoted myosin). Inhibition of ROCK (herein referring to ROCK1 and ROCK2) activity rescued the desmosomal localization as well as the epithelial cell sheet integrity caused by ARHGAP32 KO. In summary, our work highlights a RhoGAP associated with desmosomes, linking desmosome dynamics with stress fiber formation, which is crucial for maintaining epithelial cell sheet integrity.

We previously generated a mouse line, TRE-Arhgef11^CA, which allows inducible expression of a constitutively active version of a GEF for RhoA to promote actomyosin contractility. Induced actomyosin contractility in suprabasal differentiated epidermis in K10rtTA;TRE-Arhgef11^CA mice caused basal cell hyperproliferation and impaired desmosomal assembly in the epidermis (Ning et al., 2021). Consistent with this, we observed increasing actomyosin contractility impaired cortical localization of desmosomal components, including DSP and DSG1 (Fig. 1A–D). These results suggest that stronger actomyosin contractility is detrimental for desmosomal assembly in vivo (Fig. S4J). Therefore, we speculate that there might be F-actin inhibitors associated with desmosomes to control actomyosin contractility during the rapid turnover of desmosome.

To figure out the desmosomes-associated F-actin inhibitors, we screened the proteomic datasets of DSP interactome (Badu-Nkansah and Lechler, 2020). Four candidates including ARHGAP32, NDRG1, CFL1 and DSTN were selected for further validation (Fig. 1E). ARHGAP32 is a GAP promoting GTP hydrolysis on RhoA, CDC42 and RAC1 small GTPases (Diring et al., 2019; Okabe et al., 2003). NDRG1 inhibits the ROCK1-pMLC2 pathway (Sun et al., 2013). CFL1 and DSTN are actin-severing proteins (Kanellos and Frame, 2016; Yeoh et al., 2002). The unique peptide counts detected for these proteins are listed, highlighting ARHGAP32 as a promising candidate for association with desmosomes. We failed to clone an NDRG1-expressing plasmid, and neither DSTN1–GFP nor CFL1–GFP colocalized with desmosomal components (Fig. 1F,G). Notably, ARHGAP32 colocalized with desmosomal components, including PKG, DSP and DSG1, in HaCaT keratinocytes (Fig. 1H–J). ARHGAP32 also showed punctate colocalization with DSP in the cytoplasm (Fig. S1A). An effective antibody recognizing endogenous ARHGAP32 further confirmed that it colocalized with DSP at the cell cortex (Fig. S1B–E). Association of ARHGAP32 with desmosomes in other epithelial cell lines, including DLD1 and Caco-2, was also confirmed (Figs S2A–C and S3I–K).

ARHGAP32 consists of a PX domain, SH3 domain, RHO-GAP domain, GAB2-ID and FYN domain (Fig. 2A). To determine how ARHGAP32 localizes at desmosomes, we created truncations that removed the indicated domains of ARHGAP32 and assessed their desmosomal localization. Notably, truncations lacking the GAB2-ID showed no desmosomal localization, whereas those containing GAB2-ID still localized at desmosomes (Fig. 2B–F). To further confirm that GAB2-ID mediates the desmosomal localization of ARHGAP32, we constructed ΔGAB2-ID–RFP, which lacks the GAB2-ID, and GAB2-ID–RFP, which contains only the GAB2-ID. ΔGAB2-ID–RFP did not exhibit desmosomal localization (Fig. 2G), whereas the GAB2-ID alone was sufficient for the desmosomal localization of ARHGAP32 (Fig. 2H,I). The GAB2-ID-mediated desmosomal localization was further confirmed in DLD1 and Caco-2 cells (Fig. S2). These findings suggest that GAB2-ID is necessary and sufficient for the desmosomal localization of ARHGAP32.

To investigate how ARHGAP32 localizes at desmosomes, we first explored its interaction with DSP, based on proteomic data suggesting their proximity. We employed the BiFC technique to validate their interaction. Compared to control transfection of ARHGAP32–VC155 with VN173, ARHGAP32–VC155 co-expressed with DSP–VN173 showed robust fluorescence at the desmosomes (Fig. 3A), indicating a potential interaction between ARHGAP32 and DSP. Furthermore, immunoprecipitation experiments confirmed the interaction between ARHGAP32 and DSP, specifically through GAB2-ID (Fig. 3B,C; Fig. S3F,G). Additionally, a proximity ligation assay (PLA) was performed between DSP and endogenous ARHGAP32, as well as with exogenous ARHGAP32–Flag, ΔGAB2-ID–Flag and GAB2-ID–Flag (Fig. S2K–M). This assay further confirmed the interaction between DSP and ARHGAP32 at the cell cortex. Moreover, the loss of DSP, but not α-catenin, impaired the desmosomal localization of ARHGAP32 (Fig. S3I–M). These data indicate that ARHGAP32 associates with desmosomes via interacting with DSP through its GAB2-ID (Fig. 2J).

To investigate whether ARHGAP32 plays a role in regulating desmosomes, we generated stable ARHGAP32-KO HaCaT cells by CRISPR-Cas9 gene editing (Fig. S3A). Interestingly, the DSP ARHGAP32-KO HaCaT cells was found in elongated fibers and showed reduced border intensity compared to that seen in control (Fig. 3D–F; Fig. S3B,C). Additionally, in an air–liquid HaCaT 3D equivalent assay, we demonstrated impaired desmosomal localization of DSP and PKG, indicating that ARHGAP32 is required for desmosomal assembly in ex vivo skin equivalent models (Fig. S3D,E). Next, we performed Ca²⁺ switch assays using control and ARHGAP32-KO cells (Fig. 3G,H). ARHGAP32 loss delayed desmosomal assembly compared to control, as observed at 12 and 24 h post-Ca²⁺ switch (Fig. 3G,H). This delay indicates that ARHGAP32 is required for efficient desmosomal assembly. Furthermore, impaired association between DSP and keratin filaments at cell borders was also observed after KO of ARHGAP32 (Fig. 3I,J). Transmission electron microscopy (TEM) analysis revealed that desmosomes in ARHGAP32-KO cells were significantly smaller, with the majority ranging from 100–150 nm, compared to the 150–200 nm range observed in control cells (Fig. 3K–M). In addition, ΔGAB2-ID–RFP, which lacks the desmosomal localization domain but retains its GAP activity, as shown by unchanged F-actin formation (Fig. S4D), failed to rescue desmosomal organization compared to full-length ARHGAP32 (Fig. S4B,C). Additionally, forced linkage of α-catenin to ΔGAB2-ID–RFP also failed to rescue the DSP organization in ARHGAP32-KO cells (Fig. S4E–G). Collectively, these results indicate that desmosomal association of ARHGAP32 is crucial for its regulation in desmosome organization and assembly.

Next, we tested whether the known ARHGAP32-interacting proteins colocalized with ARHGAP32 at desmosomes. Our results showed that exogenous GAB1, GAB2, RASA1 as well as RhoA did not colocalize with ARHGAP32 significantly at desmosomes in HaCaT cells (Fig. 4A–E; Fig. S4A). These data further support the idea that the desmosomal localization of ARHGAP32 is independent of GAB2 protein despite GAB2-ID mediating this process. Although CRK has been reported to localize at desmosomes (Badu-Nkansah and Lechler, 2020), we observed very weak colocalization between CRK and ARHGAP32 at cell cortex (Fig. 4D; Fig. S4A). Interestingly, the ΔGAB2-ID–RFP truncation showed strong colocalization with RhoA in the cytoplasm compared to full-length ARHGAP32 (Fig. 4E,F). Furthermore, depletion of DSP resulted in the disassociation of ARHGAP32 from the cell cortex and decreased levels of myosin phosphorylated at T18/S19 (Fig. S3L–O). This implies that desmosomes might function to sequester ARHGAP32 though DSP interaction, thereby inhibiting RhoA activation and controlling F-actin around desmosomes.

F-actin stress fibers were also increased in ARHGAP32-KO cells compared with control (Fig. 4G), extending across the nucleus (Fig. 4H), indicating increased actomyosin contractility. Immunofluorescence staining and immunoblotting confirmed elevated levels of T18/S19-phosphorylated myosin, consistent with the increased stress fibers (Fig. 4I,J). Treatment with the ROCK inhibitor Y27632 rescued the desmosomal organization in two ARHGAP32-KO cell lines (Fig. 4K–N), suggesting that ARHGAP32 is required for desmosomal localization by regulating actomyosin contractility. Dispase disassociation assays revealed increased cell debris after ARHGAP32 KO compared to control, which could be mitigated by Y27632 treatment (Fig. 4O–Q), highlighting the importance of ARHGAP32 for cell sheet integrity. We further investigated whether the regulation of desmosomal organization is specific to ARHGAP32. Knocking down ARHGAP21, another RhoGAP involved in cell–cell adhesion formation (Barcellos et al., 2013; Zilberman et al., 2017), surprisingly led to a significant increase in T18/S19-phosphorylated myosin, but only caused a slight lengthening of DSP puncta at cell borders without affecting DSP border intensity (Fig. S4H–K). This observation suggests that the presence of ARHGAP32 at desmosomes might help reduce local actomyosin contractility around desmosomes when ARHGAP21 is knocked down (Fig. S4L).

In conclusion, we show that ARHGAP32, a RhoGAP, interacts with DSP through its GAB2-ID, playing a crucial role in desmosomal organization and assembly by regulating actomyosin contractility, which is necessary for preserving cell sheet integrity (Fig. S4M). In our conditions, deletion of ARHGAP32 increases F-actin stress fiber formation and levels of T18/S19-phosphorylated myosin, implying elevated RhoA activity. Active RhoA accelerates DSP relocation but the persistent activation hinders desmosome plaque maturation (Godsel et al., 2010). It is notable that depletion of DSP results in decreased cortical ARHGAP32 and T18/S19-phosphorylated myosin. Deletion of the ARHGAP32 GAB2-ID leads to its cytoplasmic release, enhancing its association with RhoA, suggesting that desmosomes might sequester ARHGAP32 to fine-tune RhoA activity. In pemphigus, where desmosome assembly is disrupted by auto-antibodies against DSG, downstream pathways, including altered RhoA activity, p38 MAPK signaling and protein kinase C signaling also affected (Bektas et al., 2013; Schmidt et al., 2019). Therefore, ARHGAP32 located at desmosomes might participate in controlling RhoA activity and F-actin contractility during desmosome assembly and maturation. However, the exact mechanism remains unclear and warrants further investigation.

In fibroblasts, ARHGAP32 has been reported to mediate transport of the N-cadherin–β-catenin complex from the endoplasmic reticulum to the Golgi (Nakamura et al., 2008). This implies that ARHGAP32 might facilitate the transport of desmosomal component to cell adhesions (Fig. S4M). Furthermore, we noted a slightly reduction in border intensity of α-catenin (Fig. S3H). Although we cannot rule out the role of adherence junctions (AJs) in regulating desmosome assembly, the failure of ΔGAB2-ID truncation and the AJ-linked ΔGAB2-ID to rescue desmosome organization underscores the importance of the desmosomal localization of ARHGAP32 in its regulation of desmosomes. The lack of effect on DSP border intensity by another AJ-regulating RhoGAP, ARHGAP21, further implies the unique and essential role of desmosomal ARHGAP32.

HaCaT cells (CL-0090, Procell, China), DLD1 and HEK293T cells (provided by Professor Jianwei Sun from Yunnan University), and Caco-2 cells (provided by Professor Qun Lu from Yunnan University) were cultured in DMEM (Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Vivacell, 24765, China), 100 units/ml penicillin and 100 mg/ml streptomycin (Beyotime, C0222, China). The cells were grown in a humidified incubator at 37°C and 5% CO₂ and passaged at ∼80–90% confluence. HEK293 T cells were transfected with PEI (Polysicences, 24765, USA) reagent for virus generation. Stable ARHGAP32-KO HaCaT cell lines were generated following the protocol by Shalem et al. (Shalem et al., 2014) followed by puromycin selection and confirmation of ARHGAP32 mutation through single-colony PCR and sequencing. HEK293 T cells for the immunoprecipitation assay were transfected using Highgene plus transfection reagent (ABclonal, RM09014P, China). Transient transfections of HaCaT or DLD1 cells were performed using Lipofectamine 3000 reagent (Thermo Fisher Scientific, L3000015, USA). In the desmosome Ca²⁺ switch experiment, HaCaT keratinocytes were cultured in low Ca²⁺ medium then switched to high Ca²⁺ (1.5 mM) medium (DMEM/F-12, 3:1; D9811-14D; US Biological Life Sciences, USA). For the actin drug treatment experiment, cells were treated with 10 μM Y27632 (Glpbio, GC15712, USA) or DMSO for 2 h at room temperature before fixation for immunofluorescence staining and imaging.

All animal work was approved by Yunnan University. Mice were maintained in a barrier facility with 12-h-light–12-h-dark cycles. Mouse strains used in this study were obtained and generated as described previously (Ning et al., 2021).

Full-length coding sequences (CDS) for CFL1 and DSTN were PCR amplified from human cDNA and inserted into pEGFP-N1 plasmid (Addgene, Cat. No. 6085-1) by infusion cloning (Vazyme, C112, China). The full-length ARHGAP32 CDS and its truncations were also amplified and inserted into pRFP-N1. The pRFP-N1 plasmid was acquired by removing GFP fragment and replaced with an RFP fragment of pEGFP-N1 plasmid. For the bimolecular fluorescence complementation assay, ARHGAP32 and DSP CDS were inserted into pBiFC-VC155 and pBiFC-VN173, as abbreviated described VC155 and VN173 in the main text and figures. pBiFC-VN173, pBiFC-VC155 were Addgene plasmid #22010 and #22011 (deposited by Chang-Deng Hu; Shyu et al., 2006). For immunoprecipitation assay, the full-length ARHGAP32 CDS and its truncations were PCR amplified and cloned into pCMV-HA plasmid (self-made by inserting an HA tag into pCMV-Tag2B plasmid between restriction sites NotI and PstI), the PCR amplified full-length CDS for DSP was cloned into pCMV-Tag 2B plasmid (provided by Professor Jianwei Sun) with a flag tag. For CRISPR-Cas9 mediated gene KO mutant construct, the following sgRNA for ARHGAP32 were used: sgRNA1, 5′-CGAACTCAACATTCTCATAA-3′; sgRNA2, 5′-CCTGACAAGCAATCTGCACG-3′. These were inserted into lentiCRISPR-V2 (Addgene plasmid #52961) by T4 ligation (NEB, M0202S). For shRNA gene knockdown plasmid constructs, shRNAs were inserted into pLKO.1 vector (Addgene plasmid #8453) by T4 ligation (NEB, M0202S). The following shRNAs were used: 5′-CCTCACAACTTCAGCTCCATT-3′ for ARHGAP21shRNA1, 5′-GCATCTCAAAGCACGACAGAT-3′ for ARHGAP21 shRNA2, 5′-CCTGATGTCGCAGCCTATAAG-3′ for α-catenin shRNA, 5′-GCAGAATGATTCACAAGCAAT-3′ for DSP shRNA, and 5′-CAACAAGATGAAGAGCACCAA-3′ for a scramble control target.

Tissues or cells were immobilized with 4% paraformaldehyde at 37°C for 10 min or with methanol at −20°C for 3 min. After fixation, cells were washed with 1× PBS containing 0.1% Triton X-100 (PBST) and then blocked with 1% BSA for 30 min. Then, the samples were incubated with the specified primary antibody for 1 h at room temperature (RT), washed three times by PBST and then incubated with secondary antibodies for 30 min. After washing with PBST, coverslips or slides were sealed with anti-fluorescence quenching agent (UElandy, China). Frozen blocks were sectioned into 10 µm sections on a cryostat, stained as described above, and mounted on glass slides. The following primary antibodies were used: rabbit anti-DSP (Proteintech, 25318-1-AP, 1:100), anti-DSG1 (Proteintech, 24587-1-AP, 1:100), anti-DSG1 (Santa Cruz Biotechnology, SC-137164, 1:100), anti-PKG (Santa Cruz Biotechnology, SC-8415, 1:100), anti-Flag (Sigma, F1804, 1:500), anti-HA (Invitrogen, 26183, 1:250), anti-phospho-MYL9-T18/S19 Rabbit (Abclonal, AP0955, 1:100) and anti-ARHGAP32 (CUSABIO, CSB-PA415667LA01HU, 1:50). The secondary antibodies were used were: Coralite594-conjugated goat anti-rabbit-IgG(H+L) (Proteintech, SA00013-4, 1:500), Coralite594-conjugated goat anti-mouse-IgG(H+L) (Proteintech, SA00013-3, 1:500), Coralite488-conjugated goat anti-rabbit-IgG(H+L) (Proteintech, SA00013-2, 1:500) and Coralite488-conjugated goat anti-mouse-IgG(H+L), (Proteintech, SA00013-1, 1:500). Phalloidin–Coralite 488 (PF00001, 1:500) and phalloidin–Coralite 594 (PF00003, 1:500) were from Proteintech, and DAPI was from Meilunbio (MA0127, 1:500).

PLA reagents were purchased from Sigma-Aldrich, and PLA was performed following its protocol. Cells on coverslip were rinsed and fixed using cold methanol at −20°C for 5 min, then washed using PBS containing 0.1% Triton X-100 three times for 5 min each. Cells were then blocked using 1% BSA for 30 min, and then incubated with primary antibody for 60 min at room temperature. Cells were then incubated with PLA secondary antibodies conjugated to DNA oligonucleotides for 60 min at 37°C in a humid chamber, and then washed using wash buffer A (0.01 M Tris-HCl pH 7.4, 0.15 M NaCl and 0.05% Tween 20) twice for 5 min each. Cells were then subjected to 30-min incubation at 37°C in a humid chamber for ligation of nucleotides, washed using wash buffer A twice for 5 min each, followed by a 100-min incubation at 37°C for rolling circle polymerization. Cells were then washed using wash buffer B (0.2 M Tris-HCl pH 7.5 and 0.1 M NaCl) twice for 10 min each. DAPI was used to stain the nucleus. ImageJ software was used to quantify the number of PLA and DAPI signals per image field.

The dispase-based dissociation assay (DDA) experiment was modified from a previously described protocol (Calautti et al., 1998). Briefly, human keratinocytes in 24-well plates were grown to confluence in 10% FBS medium. Then the cells were washed twice with PBS and incubated with 2.4 U/ml dispase II (Sigma, D4693, USA) at 37°C for 1 h. The floating monolayer was rotated in a shaker (300 rpm) for 5–10 min. For Y27632 treatment in the DDA assay, ARHGAP32-KO cells were first treated with DMSO or Y27632 at 4 μM for 2 h, then the DDA was performed with additional with DMSO or Y27632. A photograph of the debris was taken with the camera, and the cell counter function was used to determine the numbers of debris fragments using the ImageJ software.

The 3D skin equivalent experiment was modified from a previously described protocol (Wang et al., 2022). Briefly, ∼2×10⁵ HaCaT cells were seeded onto Transwell inserts (Corning, cat. 3493, USA) in DMEM high-glucose supplemented with 10% FBS in a 24-well plate. The culture medium was replenished daily. After 3 days, the cultures were transitioned to CnT-PR-3D medium (CELLnTEC, cat. Cnt-PR-3D, Switzerland) for 24 h and then maintained at the air–liquid interface for 6 days. The culture medium was refreshed every other day. Following 6 days of air exposure, the 3D skin cultures were subjected to immunofluorescence staining analyses.

HEK293 T cells were transfected with indicated Flag-tagged plasmids for 36 h in 6 cm plates and the cell lysate were extracted using RIPA buffer. Then the immunoprecipitation procedure was carried out using an anti-Flag affinity gel kit (Beyotime, P2271-10 ml, China). The eluted proteins were then analyzed by western blotting (full images of blots in this paper are shown in Fig. S5). The primary and secondary antibodies used were as below: anti-Flag (Sigma, F1804, USA, 1:4000), anti-HA (Invitrogen, 26183, USA 1:4000), anti-phospho-MYL9-T18/S19 rabbit (Abclonal, AP0955, China, 1:1000), and anti-DSP (Proteintech, 25318-1-AP, China, 1:2000), horse anti-mouse-IgG, HRP-linked antibody (Cell Signaling Technology, 7076S, USA, 1:4000), goat anti-rabbit IgG, HRP-linked antibody (Cell Signaling Technology, 7074S, USA, 1:4000).

TEM sample preparation was performed following previous protocols (Badu-Nkansah and Lechler, 2020). Cell samples were washed with PBS and fixed with 2.5% glutaraldehyde on ice for 2 h, then washed in PBS three times. Cells were then fixed with 1% osmium tetroxide for 1 h and washed with distilled water three times, then stained with 2% uranyl acetate for 30 min, and washed three times with distilled water. Samples were then dehydrated in 50%, 70%, 80%, 90% and 100% ethanol for 2 min respectively, then were infiltrated in 50%, 66% and 75% epoxy resin for 30 min, finally were infiltrated in 100% resin overnight at room temperature. On the second day, samples were embedded in fresh resin for 1 h and labeled, then polymerized at 60°C for 48 h. After polymerization, the samples were section using an EM UC7 ultramicrotome, then stained with 2% uranyl acetate for 10 min, and washed with distilled water for 10 min, then stained with 0.2% lead citrate for 5 min, and finally washed with distilled water for 10 min. After the samples were naturally dried, imaging was performed with a transmission electron microscope (HITACHI, HT7800).

Images were analyzed and quantified using FIJI software. For the quantification of the border intensity and length, the fluorescence intensity along the defined border was measured and divided by the border length, and then normalized to background noise. The fluorescence intensities of cortical DSP and DSG1 were measured by drawing lines across cell junctions and their profiles were analyzed and plotted for significance and relative P-value based on the highest fluorescence intensity at cell junctions. The relative fluorescence intensity of DSP and PKG at desmosomes was calculated by dividing the highest fluorescence intensity at desmosomes by their average fluorescence intensity in the cytoplasm. Desmosomes length was analyzed by measuring the length of dense desmosomes in TEM images. The number of stress fiber above the nucleus was quantified. The DDA quantified the amount of debris until control cell sheets were shaken off from plates after treatment with dispase. All experiments were performed at least three times independently. For two group comparisons, P-values were calculated using two-tailed unpaired Student's t-tests. For multiple group comparisons, statistical differences between means were determined using one-way ANOVA followed by a Tukey's multiple comparisons test. P<0.05 was considered statistically significant. All calculations were performed using ImageJ software. Pearson's correlation coefficients were calculated using ImageJ software to assess the degree of colocalization between DSP and keratin or ARHGAP32 as indicated in the figure legends. All statistical analysis was performed using Microsoft Excel.

Images were acquired using Zeiss LSM 800 confocal with Airyscan using a 63× oil objective at ∼5% laser intensity and master gain at 800 V. When comparing immunofluorescence intensity in different conditions, images were all acquired with same parameters including the laser power, master gain and pin hole size.

We are grateful to Professor Jianwei Sun from Yunnan University for sources of HEK293T and DLD1 cell lines, Professor Qun Lu for Caco-2 cells, and Professor Maorong Chen for the kindly provision of the Duolink PLA kit. We thank the core facilities and animal center of Yunnan University for imaging and mice care.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.261901.reviewer-comments.pdf