PALS1 is a key regulator of the lateral distribution of tight junction proteins in renal epithelial cells

Apicobasal cell polarization establishes an asymmetric distribution of lipids and proteins in the plasma membrane and forms the basis for the physiological role of numerous epithelia in mammalian organs. Studies from the past two decades in Drosophila melanogaster, in the zebrafish Danio rerio and in mammalian cell cultures have demonstrated the essential role of the evolutionarily conserved Crumbs (CRB) complex in epithelial cell polarity and epithelial tissue morphogenesis (Martin et al., 2021; Thompson et al., 2013). The CRB core complex consists of the name-giving type I transmembrane protein Crumbs (in mammals mainly CRB3a; an isoform of CRB3), lin-7 homolog C (LIN7c), protein associated with LIN7c (PALS1) and PALS1-associated tight junction protein (PATJ) (Martin et al., 2021; Pieczynski and Margolis, 2011). Recently, it has been shown that the CRB complex components CRB3a, PATJ and PALS1 localize directly apical to the tight junction (TJ) in a novel compartment called the vertebrate marginal zone (VMZ) (Mangeol et al., 2022; Tan et al., 2020). Several in vitro studies have identified an essential role for mammalian CRB3, PALS1 and PATJ in TJ formation (Fogg et al., 2005; Lemmers et al., 2004; Michel et al., 2005; Straight et al., 2004; Tan et al., 2020; van Rossum et al., 2006; Wang et al., 2007). In addition, PALS1, PATJ and CRB3 are required for lumen formation in 3D mammalian epithelial cell cultures (Schlüter et al., 2009; Shin et al., 2005; Straight et al., 2004). Taken together, this shows that the CRB complex is a critical regulator of epithelial tissue morphogenesis.

Cell polarization and the formation of cell contacts are of outstanding importance for mammalian kidneys, particularly for renal nephron epithelia. The physiological function of various tubular nephron epithelia depends on the asymmetric distribution of numerous channels and transporters to regulate the resorption and recycling of nutrients, as well as on the formation of TJs to regulate the salt and ion homeostasis and to concentrate the primary ultra-filtrate into excretable urine.

Several studies indicate an important role for the CRB complex in kidney function and imply that dysfunction of CRB core components might act as a causative or aggravating factor for renal diseases. For instance, mice lacking CRB3 or LIN7c develop cysts in the renal tubules of the kidneys (Charrier et al., 2015; Olsen et al., 2007; Whiteman et al., 2014). In addition, heterozygous PALS1-knockout (KO) mice with a reduced PALS1 expression in all nephron epithelia (Weide et al., 2017) develop proteinuria and form numerous cysts leading to death within the first 2 months after birth (Berghaus et al., 2022; Weide et al., 2017). Remarkably, in the case of PALS1, the loss of only one allele is sufficient to cause this phenotype, indicating a crucial and dose-dependent function of PALS1 in renal epithelial cells (Weide et al., 2017). However, it remains unclear whether reduced PALS1 levels lead to defects in cell polarization or rather affect the formation of cell–cell contacts between the epithelial cells of the nephron.

In this study, we analyzed the role of PALS1 in the formation of AJs and TJs in Madin Darby canine kidney (MDCKII, hereafter just called MDCK) cells in which both PALS1 alleles were inactivated using CRISPR/Cas9. These cells show moderate defects in cell polarization but exhibit dramatically increased paracellular permeability. Strikingly, several TJ components such as ZO-1 and ZO-3 (also known as TJP1 and TJP3, respectively) and occludin as well as the PALS1-binding partner PATJ are lost from bicellular TJs in PALS1 KO cells. Moreover, reestablishment of the tight junctional belt after transient F-actin depolymerization by Latrunculin A was significantly delayed in the absence of PALS1. Thus, our data provide novel evidence that PALS1 regulates the barrier function in MDCK cells by controlling the lateral distribution of TJ components and its coupling to the actin cytoskeleton.

To ensure the complete absence of PALS1, we established MDCK KO cells by CRISPR/Cas9 and selected a target sequence within the exon 7 of the canine PALS1 gene, encoding the PALS1 PDZ domain (Fig. S1A). The two selected knockout clones KO1 and KO2 showed no expression of the two PALS1 splice variants (Fig. 1A). Immunofluorescence (IF) analyses confirmed the complete absence of PALS1 at cell–cell junctions (Fig. 1B).

To test whether loss of PALS1 affects the barrier properties of MDCK cells, we measured the trans-epithelial electrical resistance (TEER) of MDCK wild-type (WT) and PALS1 KO cells. TEER measurements are suitable to analyze the tightness of cell–cell junctions and the ion transport properties (pore pathway) of epithelial cells. We first compared TEER of WT and PALS1 KO cells during the establishment of epithelial monolayers on Transwell filters. Whereas WT cells established peak TEER values around 3 days post seeding, the typical TEER peak was not observed in KO cells, in line with previous studies (Straight et al., 2004; Fig. S1B). Next, we tested TEER development using the well-established Ca²⁺ switch assay. The withdrawal of Ca²⁺ ions from the medium led to a complete disruption of cell–cell contacts in MDCK monolayers and collapse of the TEER. The effect was reversible, as the subsequent addition of Ca²⁺ to the medium results in the re-formation of cell–cell contacts and TEER. TEER measurements during the Ca²⁺ switch assay revealed strongly reduced TEER values upon withdrawal of Ca²⁺ in both MDCK WT and PALS1 KO cell monolayers. WT cells completely re-established the high TEER values within 24 h after Ca²⁺ re-addition, whereas PALS1 KO cells showed no TEER increase during this time period (Fig. S1C). This demonstrates loss of TJ resistance in PALS1 KO cells.

The altered TEER in PALS1 KO cells grown in 2D monolayers suggested differences in the formation or functionality of cell–cell junctions. To investigate whether PALS1 KO caused the permeability to be increased for ions only or for larger molecules as well, we analyzed MDCK WT and PALS1 KO cells using a 3D cyst permeability assay. MDCK cells embedded in extracellular matrix (ECM) form 3D spherical cysts with a single apical lumen. These cysts consist of a layer of apicobasally polarized cells. 3D MDCK cysts share many properties with in vivo epithelia and are an excellent in vitro model for studying polarity and cell–cell contact formation and integrity (Elia and Lippincott-Schwartz, 2009). To address the function of PALS1 in this system, the accumulation of fluorescently labeled dextran (Alexa Fluor 647, 10 kDa; FITC, 3–5 kDa) in the cyst lumens was analyzed over time. Although neither single-lumen nor multiple-lumen WT cysts showed any accumulation of labeled dextran in their lumen over the time period of 15 min, dextran intraluminal accumulation in PALS1 KO cyst was evident within the first two minutes of dextran administration and increased over time (Fig. 1C,D; Fig. S1D,E). These data indicate that PALS1 regulates the establishment of the paracellular ion barrier (pore pathway) as well as the barrier for small solutes (leak pathway). Given that WT1 and WT2 as well as KO1 and KO2 showed identical phenotypes, from now on only data from WT1 (denoted WT) and KO1 (denoted KO) are presented as representatives of each phenotype.

Next, we analyzed the cell–cell junctions in 3D cysts for potential changes in their structure by transmission electron microscopy (TEM). The apical regions of cell–cell contacts, which contain the TJs, appeared more diffuse and less electron-dense in PALS1 KO cells compared to in the WT cells. In contrast, desmosomal structures showed no obvious differences between WT and PALS1 KO cells in 3D cysts, indicating that PALS1 regulates the structure of cell–cell junctions in the apical region of epithelial cells (Fig. 1E).

The increased permeability of cell–cell contacts in PALS1 KO cysts and the ultrastructural changes in the apical region of cell–cell contacts of KO cells found by TEM argue for changes in TJ organization and function. To test this, we first investigated the distribution of ZO-1 as a marker for TJs, and E-cadherin as a marker for adherens junctions (AJs) in WT and KO cysts.

In the cells lining the cysts, ZO-1 showed an apical distribution facing the lumen. E-cadherin, on the other hand, was found basolaterally between the individual cells of the cysts. This was the case both for WT and PALS1 KO cells, and was observed in single lumen cysts as well as in cysts with multiple lumens (Fig. S2A). Of note, both WT and PALS1 KO cells formed cysts with single, multiple and no lumen. However, multiple- and no-lumen cysts were more prevalent in PALS1 KO cell lines (Fig. S2B), in agreement with previous work using PALS1 KD MDCK cell lines (Straight et al., 2004).

Strikingly, in cysts, PALS1 deficiency resulted in ZO-1 positive ‘spike-like’ elongations at tricellular junctions, which was accompanied by a discontinuous ZO-1 pattern in bicellular junctions (Fig. 2A, middle row). These spikes were absent in the MDCK WT cysts (Fig. 2A, upper row).

To verify the specificity of the CRISPR/Cas9-mediated PALS1 KO we re-expressed HALO-tagged PALS1 in PALS1 KO cell lines. PALS1 KO cysts that expressed PALS1–HALO fusion proteins re-established the continuous ZO-1 distribution along bi- and tri-cellular junctions in 3D. Spike-like elongations at tricellular junctions were not observed (Fig. 2A, lower panel).

Next, we analyzed the distribution of TJ and AJ proteins in 2D monolayers. MDCK WT cells showed a ‘chicken-wire-like’ distribution of E-cadherin and ZO-1 (Fig. 2B). In PALS1 KO cells, ZO-1 disappeared from bicellular junctions, leading to a ‘dot-like’ pattern of ZO-1 that was caused by strong ZO-1 signals at the tricellular junctions (Fig. 2B). The TJ protein occludin also exhibited a reduced intensity along bicellular junctions in PALS1 KO cells (Fig. S3A). By contrast, E-cadherin (Fig. 2B) and the lateral membrane protein JAM-A (also known as F11R) (Iden et al., 2012; Martìn-Padura et al., 1998) (Fig. S3B) showed a normal chicken-wire-like distribution both in WT and PALS1 KO cells.

The strong decrease of TJ proteins at bicellular junctions prompted us to analyze the vertical distribution of ZO-1 and E-cadherin in polarized 2D monolayers (z-axis). To this end, we determined the ratio of the lateral E-cadherin and ZO-1 signals over the total junctional length (length of ZO-1 plus E-cadherin signals). The analysis revealed no overt changes in the lateral length of E-cadherin in polarized WT and PALS1 KO monolayers. However, PALS1 KO cells showed a small but significant elongation of the ZO-1 signal (Fig. S4).

We also analyzed PALS1 KO cells expressing the PALS1–HALO fusion proteins in 2D monolayers. Indeed, cells expressing the PALS1–HALO fusion protein re-established the continuous distribution of endogenous ZO-1 in bi- and tri-cellular junctions (Fig. 2C). Bicellular contacts between PALS1–HALO-positive cells and cells without a detectable PALS1–HALO expression maintained a disrupted ZO-1 distribution. Thus, PALS1 expression is required for the continuous distribution of ZO-1 in 2D and 3D cultures of epithelial cells.

Data shown in the Fig. 2C indicate that the expression of PALS1 only in one cell of a cell–cell contact is unable to trigger full TJ formation in the adjoining cell. To investigate the impact of PALS1 on TJ proteins of neighboring cells in more detail, we used co-culture monolayers (1:1) of MDCK WT and PALS1 KO cells and performed IF studies with antibodies against PALS1, ZO-1 (Fig. 3A), E-cadherin, occludin and claudin 1 (Fig. S5).

At bicellular contacts, the highest signal intensity of ZO-1 was observed between junctions of two WT cells (WT–WT). Cell junctions between one WT and one KO (WT–KO) cell showed reduced signal intensities and the contacts of two KO cells (KO–KO) showed the lowest levels (Fig. 3A,B). In comparison to the expected stepwise decrease of the PALS1 signal (WT–WT versus WT–KO and KO–KO, respectively), the reduction in the ZO-1 signal in WT–KO and KO–KO contacts was moderate, with average values not below 80% of those of the WT–WT contacts (Fig. 3C). IF studies of co-cultures grown on glass coverslips confirmed our observations on Transwell filters and again showed a reduction in TJ proteins (ZO-1, occludin and claudin 1) at bicellular junctions (Fig. S5). For the AJ protein E-cadherin, we could not observe an alteration at the bicellular junctions (Fig. S5). In contrast, tricellular junctions shared by KO cells (KO–KO–KO) showed the strongest ZO-1 signal intensity. The intensity in mixed junctions (WT–WT–KO and WT–KO–KO, respectively) was in between the values of the homogenous tricellular junctions composed only of WT or KO cells (Fig. 3D). Together, these data suggest that the continuous distribution of ZO-1 in bicellular pools is directly linked to the PALS1 expression level within a cell–cell contact.

Membrane binding of PALS1 is mediated by its interaction with PATJ (Roh et al., 2002a). We therefore wondered whether the interaction of PALS1 with PATJ is needed for and linked to the proper distribution of TJ proteins. To address this, we expressed two PALS1 mutants with a C-terminal GFP tag in PALS1 KO cells (Fig. S6A). One mutant lacked the N-terminal evolutionarily conserved region (ECR), which mediates binding to PAR6 [ΔECR:, amino acids (aa) 118–675; Hurd et al., 2003; Penkert et al., 2004; Wang et al., 2004]. The other mutant additionally lacked the N-terminal L27 domain (L27N), which binds to PATJ (ΔECR+L27N, aa 178–675) (Roh et al., 2002a). Full-length PALS1 fused to GFP (PALS1–GFP full length) was used as a positive control. Transient transfection of constructs encoding these PALS1–GFP fusion proteins in PALS1 KO cells showed that only GFP-tagged full-length PALS1 and the ΔECR deletion mutant localized to cell junctions and were able to re-establish a continuous ZO-1 distribution. The ΔECR+L27N deletion mutant, lacking the first 177 aa, failed to localize to TJs and showed a diffuse cytoplasmic localization (Fig. S6B). These data indicate that the interaction between PALS1 and PATJ is essential for proper TJ distribution.

IF studies and western blot analyses were used to investigate to what extent PALS1 determines the intracellular localization and expression of PATJ. We observed a strong colocalization of the TJ protein ZO-1 with PALS1, PATJ and the PATJ-binding partner ZO-3 in bi- and tri-cellular junctions of WT cells (Fig. 4). For PALS1 KO cells, the IF studies revealed a strong reduction of junctional PATJ indicating that PALS1 regulates the junctional level of PATJ. Remaining PATJ and ZO-3 showed a dot-like pattern, that strongly overlapped with that of ZO-1 (Fig. 4A). Western blot analyses of MDCK WT and PALS1 KO cells revealed a strong reduction of PATJ in PALS1 KO cells, whereas LIN7c and actin expression levels remain unchanged. TJ proteins ZO-1 and ZO-3 showed slightly reduced expression levels in PALS1 KO lysates, in contrast to what was seen for E-cadherin, which showed a similar expression in WT and KO cells (Fig. 4B). Strikingly, expression of the PALS1–HALO fusion protein in PALS1 KO cells resulted in the re-establishment of the endogenous PATJ expression levels as well as of its junctional localization (Fig. 4C).

To clarify to what extent these effects are dose dependent, we analyzed PALS1 knockdown (KD) cell lines stably expressing a PALS1 shRNA (Fig. S7). PALS1 KD cells showed a strong reduction in PALS1 and PATJ expression. Residual endogenous PALS1 in PALS1 KD cells showed a weak punctate pattern that was different to the continuous belt-like staining seen in WT cells (Fig. S7A). TJ proteins ZO-1 and ZO-3 were strongly reduced along bicellular junctions and showed a dot-like pattern in tricellular junctions, similar or identical to PALS1 KO cells. Moreover, co-cultures of PALS1 KD cells and MDCK cells expressing a control shRNA showed a strong decrease in PATJ and occludin levels at bicellular TJs. PALS1 KD cells stably expressing a shRNA-resistant PALS1–GFP fusion protein rescued the expression and the normal bicellular TJ distribution of endogenous PATJ (Fig. S7). Together, these data indicate that the expression of PALS1 at WT levels is crucial for the continuous lateral distribution of PATJ and a range of bona fide TJ proteins at bicellular TJs.

The data summarized in Figs 2–4 indicate that PALS1 acts upstream of TJ proteins. However, it remains unclear how PALS1, perhaps in concert with PATJ, orchestrates the lateral distribution of TJ proteins. Previous studies have identified a connection between F-actin and the ZO proteins (Belardi et al., 2020; Fanning and Anderson, 2009; Fanning et al., 2002; Odenwald et al., 2017; Rouaud et al., 2020). Moreover, it is well established that F-actin polymerization and actomyosin-mediated contractility regulate TJ formation and the permeability properties of epithelia (Belardi et al., 2020; Ivanov et al., 2005; Van Itallie et al., 2009, 2015).

We therefore treated WT and PALS1 KO cells with Latrunculin A (LatA). The LatA toxin binds to monomeric G-actin and inhibits its polymerization into F-actin oligomers, thereby affecting the self-assembly of G-actin into F-actin. In fully polarized MDCK WT cells, LatA treatment for 4 h resulted in the depolymerization of F-actin, leading to irregular (phalloidin-positive) F-actin accumulation inside the cytosol of cells and a dot-like pattern along the bicellular TJs. (Fig. 5). Junctional F-actin-positive dots were also positive for the polarity proteins PALS1 and PATJ (Fig. 5A,B), as well as the TJ proteins ZO-1 and ZO-3 (Fig. 6A,B). In almost every case, the dots were also found at tricellular regions of the TJs. In PALS1 KO cells, LatA had a similar effect – LatA-resistant F-actin pools colocalized with ZO-1 and ZO-3 dots along bicellular TJs (Fig. 6). Thus, the dot-like pattern of TJ proteins in LatA-treated MDCK WT cells strongly resembled the distribution of TJ proteins in untreated PALS1 KO cells.

To investigate whether F-actin re-polymerization-induced elongation of junctional TJ protein requires PALS1, we performed LatA-washout experiments with WT and KO cells. For this, we treated cells for 4 h with LatA (Fig. 7), removed the toxin and analyzed the cells 10 min or 60 min after the LatA washout. After the LatA washout in WT cells, the dot-like pattern disappeared, and F-actin as well as ZO-1 re-assembled stepwise within 60 min in the typical chicken-wire-like pattern (Fig. 7A). In PALS1 KO cells, LatA washout also led to the reestablishment of the circumferential F-actin belt, but – unlike in WT cells – without the corresponding formation of a continuous lateral distribution of TJ proteins along bicellular contacts (Fig. 7B).

To evaluate the effects of the LatA-washout assays, we analyzed the distribution of fluorescence intensities (FI) of F-actin and ZO-1 along the cell–cell junctions (detailed description in the Materials and Methods). Positive values represent with FI higher than the mean for the whole junction, and negative values indicate pixels with the FI lower than the average for the junction. The results of the junctional FI distributions in WT and KO cells with and without LatA treatments are visualized in violin plots (Fig. 7C,D). Untreated WT cells showed a ‘drop-like’ distribution, with most values around the mean FI (around zero), indicating equal protein distribution along the junction. LatA treatment of cells resulted in a spread of FI values, due to an increase of both strong positive and negative values. The number of negative values is significantly higher, as the localization of the proteins is limited to a few very intense spots. After the LatA washout (10 min and 60 min), FI values for F-actin and ZO-1 showed a stepwise recovery towards the FI distribution of untreated WT cells. The F-actin FI distributions of PALS1 KO cells resembled that of the WT cells, resulting in similar violin plot patterns. However, actin reassembly might be slightly delayed as after 10 min of washout as there was no stepwise rearrangement in the KO cells as seen in the WT (Fig. 7C,D). In contrast to F-actin recovery, LatA washout in PALS1 KO cells did not lead to recovery of the ZO-1 distribution (Fig. 7B,D). Together, these results demonstrate that PALS1 KO cells exhibit a strong defect in the reformation of TJs upon release from actin depolymerization.

In addition to the LatA washout experiments, we also conducted Ca²⁺ switch assays and analyzed the intracellular localization of ZO-1 and F-actin by IF analyses. The withdrawal of Ca²⁺ for 18 h resulted in a complete disassembly of TJs in WT and KO cells (Fig. 8). Both WT and KO cells showed a disrupted ZO-1 pattern accompanied by a ring-like F-actin distribution inside the cells (Fig. 8A,B). In WT cells, the re-establishment of F-actin belts and the uniform lateral distribution of TJ proteins along the bicellular and tricellular TJs occurred within 6 h of Ca²⁺ re-addition. By contrast, in PALS1 KO cells, the re-addition of Ca²⁺ only re-established the F-actin belt, whereas the distribution of ZO-1 remained incomplete, even after 24 h (Fig. 8B). In both junction reformation assays, we had the impression that the re-polymerization of F-actin in PALS1-depleted cells showed a slight delay. Since local actin polymerization can be regulated by the GTPase RhoA (de Seze et al., 2023; Terry et al., 2011), we analyzed the localization of RhoA in MDCK WT and PALS1 KO cells. Indeed, IF analyses of co-cultures demonstrated elevated levels of RhoA at the junctions formed between WT cells, with minimal RhoA signal detected in the cytosol. In contrast, KO cells exhibited reduced RhoA pools at cell–cell contacts (Fig. 8C).

In this study, we analyzed the functions of PALS1 in MDCK cells. Genetic inactivation of the PALS1 gene using CRISPR/Cas9 resulted in an increased formation of multiple- and no-lumen cysts in 3D cell culture, and a strong increase in paracellular permeability in cysts derived from PALS1 KO cells. PALS1 KO cells that formed single-lumen cysts also showed an increase in paracellular permeability, whereas single- and multiple-lumen cysts of MDCK WT cells maintained a tight paracellular diffusion barrier for solutes at least up to 10 kDa. This is in line with earlier studies (Straight et al., 2004) and suggests that PALS1 is involved in the regulation of the paracellular transport of ions and solutes in epithelial cells and that the formation and function of TJs is not directly coupled to the establishment of a single apical luminal domain.

Proteins of the ZO family are essential for the formation of TJs and the regulation of paracellular transport in epithelia (Haas et al., 2022; Otani and Furuse, 2020; Otani et al., 2019; Umeda et al., 2006; Van Itallie et al., 2009; Zihni et al., 2016). Indeed, the loss of ZO-1 and ZO-2 in 3D MDCK cysts is associated with a dramatic increase in paracellular permeability to water, solutes and ions (Otani et al., 2019), and ZO-1/ZO-2 double KO cell lines show a phenotype similar to MDCK cells lacking PALS1 or the PALS1-binding site in PATJ (ΔL27-PATJ; (Pombo-García et al., 2022 preprint). PALS1-deficient cells revealed a significant reduction in TJ proteins along bicellular junctions, accompanied by an accumulation of these proteins at tricellular junctions, resulting in a dot-like (or spike-like) distribution pattern. Remarkably, this discontinuous distribution was preserved in bi- and tri-cellular junctions that were formed between WT and PALS1 KO cells. In other words, PALS1 KO cells that formed cell–cell contacts with cells that expressed PALS1 (MDCK WT cells) maintained a disrupted TJ distribution, indicating that PALS1 on one side of cell–cell contacts is unable to trigger full TJ formation in adjoining cells. These data collectively suggest two things: first, PALS1 governs the correct lateral distribution of TJ proteins along bicellular contacts and, second, the continuous lateral distribution of ZO-1 at intercellular TJs is directly associated with the cell autonomous expression of PALS1.

PALS1 binds to PATJ, which in turn binds to the TJ proteins ZO-3 and claudin 1, via the sixth and eighth PDZ domain, respectively (Roh et al., 2002b). Moreover, previous studies identified the L27N domain of PALS1 as the PATJ-binding module that targets PALS1 to TJs, demonstrating that membrane targeting of PALS1 requires the presence of PATJ (Roh et al., 2002a) and suggesting that PALS1 localization requires PATJ.

We show here that PALS1, in turn, controls the protein level and the localization of PATJ to bicellular TJs, because in PALS1 KO and KD cells PATJ expression was strongly reduced and residual PATJ was no longer efficiently targeted to bicellular TJs. Several other TJ proteins were similarly affected.

Moreover, PALS1 deletion mutants lacking the L27N domain did not rescue the ZO-1 distribution phenotype of PALS1 KO cell lines. These PALS1 KO phenotypes, together with previous work showing that PATJ regulates the localization of occludin and ZO-3 to TJs (Michel et al., 2005), argue for a tight functional symbiosis between PALS1 and PATJ, in which PALS1 likely acts as a stabilizing factor upstream of PATJ to trigger the continuous distribution of TJ proteins. Recent super-resolution imaging studies demonstrating a high degree of co-localization of these proteins at the apical-lateral border support the strong interconnection between PALS1 and PATJ (Mangeol et al., 2022).

Recently, it has been demonstrated that phase separation of ZO proteins drives the formation of TJs (Beutel et al., 2019; Schwayer et al., 2019). Strikingly, cells lacking PATJ, or expressing a PATJ variant lacking the PALS1-binding module (ΔL27-PATJ) failed to form a continuous TJ belt due to reduced spreading dynamics of ZO-1 condensates (Fiedler et al., 2023; Pombo-García et al., 2022 preprint). Intriguingly, the disrupted ZO-1 distribution along bicellular junctions in ΔL27-PATJ MDCK cells strongly resembles the dot-like pattern that we observed in MDCK PALS1 KO and KD cell lines (Pombo-García et al., 2022 preprint). These observations strengthen the hypothesis that the close functional coupling between PALS1 and PATJ is required for the formation of a continuous and functional TJ belt in epithelial cells. Therefore, our observations along with those by Pombo-Garcia et al., and earlier studies by the groups of Margolis and Le Bivic indicate that the lateral continuous belt formation of ZO-1 (and other TJ proteins) might be predominantly organized via a PALS1–PATJ axis (Michel et al., 2005; Pombo-García et al., 2022 preprint; Roh et al., 2002a; Straight et al., 2004).

Moreover, F-actin polymerization and actomyosin-mediated contractility are known to control TJ formation and epithelial permeability (Belardi et al., 2020; Ivanov et al., 2005; Van Itallie et al., 2009, 2015). Depolymerization of the F-actin cytoskeleton results in disrupted TJs in which the colocalization of F-actin and TJ proteins is maintained (Stevenson and Begg, 1994; Sun et al., 2023 preprint). Recently, the groups led by Honigmann and Weber suggested that receptor-mediated clustering of ZO-1 is coupled to actin polymerization and bundling, which drives elongation of ZO-1 along the TJ (Sun et al., 2023 preprint). According to the proposed model, actin polymerization is the rate-limiting process for TJ belt elongation (Sun et al., 2023 preprint). Our current data indicate that the situation might be more complicated, as polymerization of actin in the absence of PALS1 is not able efficiently drive TJ elongation. In contrast to the biochemical in vitro experiments by Sun and colleagues, where weak binding of ZO-1 to actin via the actin binding regions (ARBs) was sufficient to accelerate actin polymerization and couple ZO-1 elongation along F-actin fibers, in MDCK cells this ZO-1–actin coupling requires PALS1 (this study) and most likely other members of the CRB complex (Pombo-García et al., 2022 preprint). Taken together, this suggests that TJ formation relies on two interconnected self-assembly processes: on the one hand, localized actin polarization along TJ (Sun et al., 2023 preprint), and, on the other hand, a phase separation-dependent condensation of cytoplasmic TJ proteins (such as ZO proteins) followed by elongation along bicellular TJs (Beutel et al., 2019; Schwayer et al., 2019). Our findings underscore a role for PALS1 – and likely PATJ – in the spatiotemporal coordination of these processes, which is essential for the lateral elongation of TJ protein condensates.

Interestingly, we observed decreased RhoA pools at TJs in PALS1 KO cells. Although it remains unknown to what extent these RhoA pools are activated, it is tempting to speculate that reduced RhoA pools in PALS1-depleted cells might correlate with diminished local RhoA-dependent F-actin polymerizations. It has been demonstrated previously that PATJ binds to and recruits p114RhoGEF (also known as ARHGEF18) to apical cell junctions of epithelial cells (Nakajima and Tanoue, 2011, 2012). Interestingly, this interaction appears to be conserved, as the RhoGEF Cysts, a Drosophila ortholog of p114RhoGEF, is also recruited to junctional membranes through an interaction with the Drosophila CRB complex (Silver et al., 2019). Moreover, p114RhoGEF activates local RhoA pools at TJs and thereby plays a crucial role in the formation of the circumferential F-actin belt (Terry et al., 2011). Strikingly, depletion of p114RhoGEF in MDCK cells caused a similar phenotype to that seen in MDCK cells lacking PALS1, with disrupted TJs and multiple lumen formation in 3D cultures (Terry et al., 2011). This together with our data provides evidence that a PALS1–PATJ–p114RhoGEF–RhoA axis regulates the F-actin-dependent formation of TJ in epithelia.

However, the situation is probably far more complex, as most TJ and polarity proteins are large multi-domain scaffolding proteins with numerous protein–protein interaction domains (Fanning and Anderson, 2009; Martin et al., 2021; Otani and Furuse, 2020; Rouaud et al., 2020; Thompson et al., 2013). PATJ, for example, contains up to 10 PDZ domains that mediate interactions with other proteins, including TJ proteins and several components of the Hippo signaling pathway (Duning et al., 2008, 2010; Ernkvist et al., 2009; Kohli et al., 2014; Morthorst et al., 2022; Roh et al., 2002b; Varelas et al., 2010; Wells et al., 2006). Of note, recent proximity labeling studies have identified additional binding partners of PATJ (e.g. HOMER proteins) that might provide links to TJ proteins, regulate F-actin formation or dynamics at TJ or control epithelial functions beyond the formation of cell–cell contacts (Tan et al., 2020). Finally, there is evidence that not only F-actin but also other components of the cytoskeleton, such as microtubules or members of the septin protein family control TJ formation in epithelial cells (Bowen et al., 2011; Yano et al., 2021). Given that p114RhoGEF has also been identified as a septin-associated GEF (Nagata and Inagaki, 2005) it is possible that mechanisms other than RhoA-mediated F-actin dynamics contribute to the PALS1 KO phenotypes we present here.

The epithelia of the nephron segments in the kidney ensure the excretion of noxious substances and the concentration of the primary ultra-filtrate into the final urine, as well as the recycling and resorption of numerous nutrients, organic and inorganic ions. In other words, minor changes in TJ formation dynamics, maintenance or in the distribution of central TJ proteins might have very severe consequences in vivo. The dysfunctional lateral distribution of TJ proteins in PALS1 KO and KD cells and the strong phenotype of Pals1 haplo-insufficient mice in which even a partial Pals1 depletion causes severe defects – although the well-defined apicobasal distribution of important proteins is in principle preserved – supports this assumption (Berghaus et al., 2022; Weide et al., 2017). Further research will be necessary to clarify to what extent even minor modifications of the ‘tightness of TJs’ contribute to pathomechanisms and how the complex interplay between polarity, junction assembly and cytoskeleton modulation is regulated.

We generated pENTR constructs with human PALS1 cDNA inserts (Gene ID: 64398) encoding for PALS1 full length (aa 1–675) and two PALS1 deletion mutants (ΔECR, aa 118–674; ΔL27N, aa 178–675) using the pENTR^TM Directional TOPO® Cloning Kit (Invitrogen). The pENTR constructs were shuttled into modified pQCXIH-GW-GFP or pQCXIH-GW-Halo plasmids using the L/R reaction of the Gateway^TM Cloning system (Invitrogen), according to the manufacturer's instructions. All primers and oligonucleotides are listed in Table S1.

To generate a PALS1 knockout in MDCKII cells (Sigma, #ECACC 00062107) frameshift mutations were introduced via CRISPR/Cas9. CRISPOR (Concordet and Haeussler, 2018) was used to design the following guide RNA/DNA-sequences 5′-CACCGATCATTAGTCGGATAGTAAA-3′ and 5′-AAACTTTACTATCCGACTAATGATC-3′. The gRNAs (100 µM, 8 µl each) were phosphorylated and annealed via adding 2 µl 10× PNK buffer, 1 µl ATP (5 mM) and 1 µl polynucleotide kinase (Promega, M4101) and incubation for 1 h at 37°C followed by incubation at 95°C for 5 min to stop the reaction. Resulting oligonucleotides were inserted into the px459 plasmid via digestion using BbsI and ligation by T4 ligase (Thermo Fisher Scientific, EL0011). pSpCas9(BB)-2A-Puro (PX459) V2.0 was Addgene plasmid #62988 (deposited by Feng Zhang; Ran et al., 2013). MDCK cells with a confluency of 60% were transfected using Lipofectamine™ 2000 following the manufacturer's protocol. Selection was done using puromycin (4 µg/ml, Invivogen, ant-pr) for 48 h. Potential knockout cells were separated to analyze single cell clones. Potential clones were sequenced and only clones with frameshift mutations were kept. Heterozygous clones were further analyzed after cloning with Zero Blunt^® TOPO^® PCR cloning kit to separate individual alleles.

MDCK WT (Merck/Sigma, ECACC 00062107) and PALS1 KO (MDCK WT cells with CRISPR/Cas9-mediated knockout of the PALS1 gene) cell lines were grown in modified Eagle's medium (MEM, Lonza) containing 5% fetal calf serum (FCS; Bio&Sell, FBS.S 0615) and 1% 100× penicillin/streptomycin/l-glutamine (P/S/G; Sigma, G6784). HEK293T cells were grown in DMEM medium containing 10% FCS and 1% P/S/G. The cells were cultivated at 37°C and 5% CO₂. For transient transfection, cells were transfected using Lipofectamine™ 2000 according to the manufacturer's manual.

MDCK cells were seeded on coverslips, on ThinCert™ cell culture inserts (0.4 mm-pore transwell filters, Greiner Bio-one) to reach full polarization, or in a matrix hydrogel (basement membrane extract, BME, from R&D Systems) in three dimensions (3D) to form cysts. Medium was exchanged every other day. On coverslips cells were grown up to a confluency of 100%, in transwell filters for 7–10 days and in BME for 5-7 days before using them for immunofluorescence staining. In case of mixed monolayers, wildtype and KO cells were mixed 1:1 and seeded on coverslips or ThinCert™ cell culture inserts, respectively.

Stable cell lines were generated via retroviral gene transfer, as described previously (Schulze et al., 2014). In brief, HEK293 T cells were transiently transfected with the helper plasmids pCMV-VSV-G and pUMVC and pQCXIH plasmids encoding for full-length PALS1 or deletion mutants fused to HALO or EGFP fluorescence tags. Supernatant containing the virus particles were added to MDCK WT and PALS1 KO cells after being filtered through a sterile 0.45 µm filter (Millipore). Here, one volume of MEM including supplements was mixed with one volume of virus-containing medium supplemented with polybrene (8 µg/ml). After 24 h, virus-containing medium was collected and stored at 4°C. Cells were left in medium without virus to regenerate for 24 h. After regeneration period, the second round of transduction using the same virus-containing medium was performed for another 24 h. After another 24 h for regeneration, transduced cells were selected using hygromycin (250 µg/ml). All established cell lines were tested for a constitutive overexpression of PALS1 proteins by live-cell imaging or western blotting.

Short hairpin RNAs (shRNAs) were purchased as complementary oligonucleotides that when annealed contained the HindIII (5′) and BamHI (3′) sticky ends for cloning into the pSuper vector (Oligoengine). The PALS1 shRNA B oligonucleotides had the following sequence: 5′-CCGGAGATGAGGTTCTGGAAATTCAAGAGATTTCCAGAACCTCATCTCCTTTTT-3′ (underlining highlights the targeting sequence). All plasmids were verified by Sanger sequencing. The shRNA-transfected MDCK cells were grown in 200 µg/ml hygromycin to select for stable clonal cell lines. The PALS1 rescue line was generated by transfecting shRNA-resistant PALS1–GFP into the clonal PALS1 shRNA line B1, followed by clonal selection in 400 µg/ml geneticin (Tan et al., 2020). PALS1 shRNA and PALS1-GFP rescue IF and western blotting experiments were performed as described previously (Tan et al., 2020).

For analyses of junction reformation in MDCK WT and PALS1 KO cells, Ca²⁺ detraction or LatA treatment were used. Prior to treatment, cells were grown on ThinCert™ cell culture inserts (12-mm, 0.4 mm-pore transwell filters, Greiner Bio-one) for 7 days. For Ca²⁺ switch experiments, cells were washed three times with 1× PBS and cultivated for 18 h in Ca²⁺ reduced medium (Sigma, M8167). For refeed of the cells, this Ca²⁺ depletion was followed by re-cultivation in normal cultivation medium (containing Ca²⁺) for 1–24 h before cells were fixed and used for IF staining. For LatA experiments, cells were treated with 1 µM LatA (Sigma, L5163) for 4 h. Afterwards, washout was performed by replacing the medium with normal cultivation medium for 10 min, 60 min before cells were fixed and stained.

MDCK cells were seeded on ThinCert™ cell culture inserts (12-mm, 0.4 mm-pore transwell filters, Greiner Bio-one) to measure the transepithelial electrical resistance (TEER). The electrical resistance between the apical and the basolateral chamber was measured using a two-electrode resistance system. To calculate the unit area resistance (Ω×cm²), the electric resistance was multiplied by the growth area of the Transwell filters (1.13 cm²). The TEER measurement was performed with an interval of 5 min from the moment of seeding the cells to 7 days after seeding. Medium was changed once after 3–4 days. Values were normalized to the WT control values.

For TEER measurement after Ca²⁺ depletion (Ca²⁺ switch assay) to break down the established junctions, cells were grown for 7 days on ThinCert™ cell culture inserts in cultivation medium until confluency. For Ca²⁺ reduction, cells were washed three times with 1× PBS and cultivated for 18 h in Ca²⁺ reduced medium (Sigma, M8167). Subsequently, Ca²⁺ reduced medium was replaced by normal cultivation medium (containing Ca²⁺). TEER measurement was started immediately and performed every 5 min for 24 h. Values were normalized to a medium control (well without cells).

To grow 3D MDCK cysts, ibidi µ-Slide 8-well chambers were coated with 8 µl BME [(Bio-Techne (R&D Systems), Cultrex PathClear ReducedGrowth Factor BME, Type 2; cat. no. 3533-005-02]. For solidification of the BME-coated, wells were incubated at 37°C with 5% CO₂ for 15–30 min. Cells were detached from a two-dimensional (2D) monolayer using incubation at 37°C with 5% CO₂ in 10× trypsin/EDTA solution until reaching a single-cell suspension. Cells were seeded (5000 cells/cm²), resuspended in 250 µl in medium containing 10% FCS, 1% P/S/G and 2.5% BME. Cells were cultured for 5–7 days at 37°C with an atmospheric CO₂ concentration of 5%. Medium was exchanged every second day. Cysts were either used for immunofluorescence staining (7 days) or for 3D permeability assays (5 days).

To measure the permeability of cysts, fluorescently labeled dextran of different sizes was used (Alexa Fluor 647, 10 kDa; FITC, 3-5 kDa). Medium was exchanged with 200 µl live-cell imaging medium (15 mM HEPES in Hanks’ balanced salt solution). Cysts were imaged live using the HC PL APO 40×/1.10 W motCORR CS2 objective of a confocal microscope (TCS SP8, Leica Biosystems). Cysts were imaged every minute for 15–20 min in total. The permeability assays were performed with simultaneous detection of both dextran samples: FITC-labeled 3–5 kDa dextran and Alexa Fluor 647-labeled 10 kDa dextran. After imaging time point t₀, medium was supplemented with 50 µl dextran solution (10 µM Alexa Fluor 647, 12.5 µM FITC). Cyst permeability was quantified by determining the signal intensity, comparing the lumen intensity against the average intensity outside of the cyst.

MDCK cysts cultivated in cell culture plates were washed with PBS and then fixed in 2% (v/v) formaldehyde and 2.5% (v/v) glutaraldehyde in 100 mM cacodylate buffer, pH 7.4, at 4°C overnight. After washing in PBS, specimens were postfixed in 0.5% (v/v) osmium tetroxide and 1% (w/v) potassium hexacyanoferrate (III) in 0.1 M cacodylate buffer for 2 h at 4°C followed by washing with distilled water. After dehydration in an ascending ethanol series from 30% to 100% ethanol, specimens were two times incubated in propylene oxide each for 15 min. The cells were detached from the surface of the cell culture plate during the propylene oxide treatment. Finally, detached cells were embedded in Epon using Beem capsules. Ultrathin sections were cut on an ultramicrotome and collected on copper grids. Electron micrographs were taken at a Phillips EM-410 electron microscope using imaging plates (Ditabis, Pforzheim, Germany).

Cells were grown on six-well plates up to a confluency of 95%, washed twice with 1× PBS and lysed with 300 µl 1× SDS sample buffer [20% (v/v) glycerin, 125 mM Tris-HCl pH 6.8, 10% (w/v) SDS, 0.2% (w/v) Bromophenol Blue, 5% mercaptoethanol and 1 l H₂O]. For homogenization, samples were passed through a blunt 20 G needle (0.9 mm diameter) and heated up to 95°C. Lysates were centrifuged for 1 min at 14,000 g and stored at −20°C until usage.

Western blot analysis was performed as previously described (Berghaus et al., 2022; Weide et al., 2017). First, cell lysates were boiled at 95°C for 5 min. Proteins were separated via SDS-PAGE using 10% gels (Bio-Rad). Then, proteins were transferred to a PVDF membrane (Millipore) following by incubation in blocking powder (5% milk powder in TBST [0.05% Tween 20]) for 1 h at room temperature (RT). Probes were incubated with primary antibodies (Table S2) diluted in BSA (2.5 g BSA powder in 50 ml TBST) o/n at 4°C. Afterwards, the membrane was washed three times with TBST followed by incubation with horseradish peroxidase coupled secondary antibodies (Table S3) for 1 h at RT. The membrane was washed again for three times with TBST before a chemiluminescence detection reagent (Clarity; Bio-Rad) was added and the membrane was imaged using an Azure Biosystems imager (c600; Azure Biosystems). Uncropped images of western blots are shown in Fig. S8.

Cells on coverslips and Transwell filter insets were grown either as pure cultures or as co-cultures of two different cell lines mixed in a ratio of 1:1. 3D cysts were grown as described above. All cells were washed twice with cold 1× PBS and fixed in 4% paraformaldehyde (PFA) for 20 min at RT. After cells were washed three times with 1× PBS, they were quenched for 10 min in 50 mM NH₄Cl followed by another three rounds of washing in 1× PBS. Then, cells were permeabilized in 0.2% Triton X-100 for 5 min. After washing the cells three times in IF wash buffer [0.2% (w/v) gelatin, 10% (v/v) 10× PBS, 0.2% (v/v) Triton X-100 and H₂O)] and blocked [90% (v/v) IF wash buffer, 10% (v/v) normal goat serum (NGS) for 20 min at RT. Cells were incubated o/n in antibody dilutions buffer (98% (v/v) IF wash buffer 2% (v/v) NGS, containing specific antibodies (Table S2). Next, cells were washed three times with IF wash buffer and transferred into fluorophore-conjugated secondary antibodies (Table S3) solved in antibody solution containing DAPI (1:5000). After an incubation of 30 min at RT, cells were washed the last time with 1× PBS and were carefully rinsed in distilled H₂O before either being mounted in 6–12 µl mowiol on object slides (coverslips, ThinCert™ cell culture inserts) or in IMM mounting medium (ibidi GmbH, Cysts). Samples were imaged on a confocal microscope (TCS SP8, Leica Biosystems) using a 40× water objective or a spinning disk microscope (CorrSight, Thermofisher) using a 40× or 63× oil objective (NA 1.3 or 1.4; Zeiss Plan Apochromat). Confocal stacks were processed, and analyses were performed with Fiji. Final figures were mounted using CorelDRAW.

Images were analyzed with Fiji (https://fiji.sc/). Movies were created using LAS X (Leica Biosystems). For quantification of the RhoA level, co-cultures of MDCK WT and PALS1 KO cells grown on ThinCert™ cell culture inserts were stained for RhoA, ZO-1 and PALS1 as described before and imaged via confocal microscopy. Fluorescence intensity (FI) measurements were performed in Fiji using maximum intensity projection images of the confocal stacks. Localization of RhoA was measured as RhoA FI at the TJ region (apicobasal plane defined by ZO-1) of a cell–cell contact compared to the cytoplasmic FI within the two neighboring cells building the contact. RhoA FI was analyzed in cell–cell contacts in corresponding WT–WT or KO–KO cells and, for quantification, the intensities at the junction were normalized to the values of the corresponding cytosols.

For analyzing of the distribution of F-actin and ZO-1 along the TJ region, violin plot profiles of the bicellular junction were created. Resulting FIs for each pixel were normalized to the mean FI of each junction. To determine the deviation from an equal protein distribution along the junction (relative fluorescence=1) each normalized value had 1 subtracted. Results of this analysis were visualized as violin plots. Hence, values >0 are equal to increased FI per pixel, whereas values <0 visualize pixels with reduced FI values compared to an equal fluorescence signal at the junction (junctional ΔFI Index). The effect of LatA treatment was visualized by examining the xy distribution of proteins along the junction using the maximum projection images. The apical-basal distribution of junction proteins was shown by examining the xz images resulting from z-stacks.

Each experiment was performed at least n=3 times. Graphs were created and statistical analysis was performed with the GraphPad PRISM software. Data show the mean±s.e.m. The value of n and the definition is indicated in the corresponding figure legends and results. For comparison of two groups, the Mann–Whitney U test was used, for comparison of more than two groups the Kruskal–Wallis-test was performed corrected for multiple testing with the Dunn's correction. *P<0.05, **P<0.01, ***P<0.001. If not indicated differently, all data represent at least three independent repeats.

We would like to thank Dr Alf Honigmann (MPI Dresden) for providing us details of the permeability protocol. We would like to thank Karin Gäher for excellent technical support.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261303.reviewer-comments.pdf.