To investigate how extracellular adhesive interactions contribute to PCP asymmetry, we examined polarity establishment in Celsr1^Crsh/Crsh mutant embryos (Figure 1A–D). In basal epithelial cells of the epidermis, PCP proteins accumulate preferentially at anterior–posterior junctions and are depleted from mediolateral junctions (Devenport and Fuchs, 2008; Aw et al., 2016). For simplicity, we refer to these as ‘vertical’ and ‘horizontal’ junctions, respectively (Figure 1A). Celsr1 asymmetry arises progressively between embryonic day E12.5 and E15.5 of development where both the magnitude and collective alignment of polarity increases (Devenport and Fuchs, 2008; Aw et al., 2016). To quantify Celsr1 asymmetry in Celsr1^Crsh/Crsh mutant epidermis, we used automated segmentation and calculated the nematic order of Celsr1 fluorescence intensity. In agreement with our previous qualitative analysis, Celsr1 enrichment at vertical junctions was completely lost in E15.5 Celsr1^Crsh/Crsh embryos with Celsr1 localizing uniformly to both vertical and horizontal cell borders (Figure 1C and D; Devenport and Fuchs, 2008). Heterozygous Celsr1^Crsh/+ embryos also displayed a polarity defect, where both the magnitude and collective alignment of Celsr1 asymmetry were reduced compared to Celsr1^+/+ littermates (Figure 1C and D). Thus, a D1040G substitution in the cadherin repeats of Celsr1 abolishes its ability to localize asymmetrically and partially interferes with Celsr1 function in a dominant manner.

Interestingly, the loss of Celsr1 asymmetry observed in Celsr1^Crsh/Crsh mutants is comparable to that caused by mutations that eliminate Fz6 or Vangl2 protein or prevent Vangl2 trafficking (Guo et al., 2004; Torban et al., 2007; Devenport and Fuchs, 2008; Merte et al., 2010; Yin et al., 2012; Cetera et al., 2017). By contrast, Celsr1^Crsh strongly impairs PCP establishment without significantly altering Celsr1 trafficking or cell surface levels, or affecting expression of other core PCP proteins (Devenport and Fuchs, 2008; Murdoch et al., 2014). Thus, Celsr1^Crsh is a unique PCP mutation that allows us to specifically address the function of Celsr1 adhesion in PCP.

The position of the Celsr1^Crsh mutation between EC7 and EC8 suggested that it may inhibit homotypic adhesion of Celsr1. Consistent with previous work, transfected Celsr1^Crsh-GFP failed to enrich at cell–cell interfaces between neighboring keratinocytes in contrast to the strong border enrichment of wild-type (WT) Celsr1 (Celsr1^WT-GFP) (Figure 1E and F; Devenport and Fuchs, 2008). When Celsr1^WT-GFP and Celsr1^Crsh-GFP were co-expressed in keratinocytes, enrichment of Celsr1^WT at junctions was reduced (Figure 1—figure supplement 1A–C), supporting the idea that the Celsr1^Crsh mutation interferes with Celsr1 adhesion in a dominant-negative manner (Qu et al., 2010; Murdoch et al., 2014).

To directly test whether Celsr1^Crsh impairs Celsr1 adhesion, we established a cell aggregation assay to determine if Celsr1 alone is sufficient to mediate cell–cell adhesion. K-562 cells, which are cadherin-free, non-adhesive, and grow in suspension (Ozawa and Kemler, 1998; Schreiner and Weiner, 2010; Baykal-Köse et al., 2020), were stably transfected with Celsr1 variants and assayed for their ability to induce cell aggregation, an indicator of cell adhesion activity. Celsr1^WT-GFP expressing K-562 cells formed aggregates similar to positive control cells expressing a chimeric protein containing the extracellular domain of E-cadherin. In contrast, negative control cells expressing GFP alone remained as a single cell suspension (Figure 2A–C). When mixed with cells expressing the E-cadherin chimera, Celsr1^WT expressing cells did not co-aggregate, but rather sorted into distinct clusters, demonstrating Celsr1 mediates homophilic binding (Figure 2D). Homophilic adhesion was dependent on the Celsr1 extracellular domain as cells expressing a Celsr1 variant lacking the extracellular domain (Celsr1^ΔN) failed to aggregate and also failed to co-aggregate with cells expressing full-length Celsr1 (Figure 2E and F). Like classical cadherins, Celsr1-mediated aggregation was calcium-dependent (Figure 2G and H). These results confirm that Celsr1 mediates homophilic, calcium-dependent cell adhesion in trans.

Surprisingly, K-562 cells expressing Celsr1^Crsh-GFP formed aggregates comparable to Celsr1^WT-GFP expressing cells, demonstrating the Celsr1^Crsh mutation does not impair the ability of Celsr1^Crsh to mediate adhesion in trans (Figure 2I and J). To test whether the Celsr1^Crsh variant could mediate homophilic adhesion with the WT protein, we performed mixing assays between cells expressing either Celsr1^WT-GFP or Celsr1^Crsh-mCherry. Unlike controls where Celsr1^WT-GFP cells co-aggregated with Celsr1^WT-mCherry cells, Celsr1^Crsh-mCherry expressing cells failed to co-aggregate with Celsr1^WT-GFP and sorted out into distinct cell clusters (Figure 2K and L, Figure 2—figure supplement 1A). This aggregate sorting behavior is likely specific to the effect of the Celsr1^Crsh mutation and not due to differences in expression or cell surface localization as cell surface expression of Celsr1^Crsh was similar to Celsr1^WT (Figure 2—figure supplement 1C and D). Interestingly, at early timepoints post mixing, Celsr1^Crsh-mCherry and Celsr1^WT-GFP cells did adhere, but later separated into distinct, unmixed aggregates (Figure 2M and N, Figure 2—figure supplement 1B). In support of these results, Celsr1^Crsh-mCherry and Celsr1^WT-GFP also failed to interact in trans at cell borders in mixed keratinocytes (Figure 2—figure supplement 2). These findings indicate that Celsr1^Crsh is a novel Celsr1 variant with altered homophilic binding preferences that fails to form stable trans-adhesive interactions with the WT protein.

Cadherin family members typically engage in two types of interactions to mediate cell–cell adhesion, intercellular trans-interactions between cadherins on opposing cells and lateral cis-interactions within the same cell (Brasch et al., 2012; Priest et al., 2017; Honig and Shapiro, 2020). Trans- and cis-interactions are thought to cooperate, functioning together to contribute to junction assembly, stability, and overall adhesive strength (Zhang et al., 2009; Wu et al., 2010; Harrison et al., 2011; Niessen et al., 2011; Brasch et al., 2012; Priest et al., 2017). Given that the Celsr1^Crsh mutant was still able to mediate adhesion in trans, we hypothesized that the Celsr1^Crsh mutation may impair Celsr1 interactions in cis. Consistently, high resolution confocal imaging of K-562 cell aggregates revealed stark differences in the subcellular localization of Celsr1^WT versus Celsr1^Crsh. Whereas Celsr1^WT-GFP strongly enriched at interfaces connecting neighboring cells (Figure 3A, open arrows) and was depleted from non-contacting edges (closed arrows), Celsr1^Crsh-GFP was distributed diffusely and broadly across the cell surface, localizing to both contacting (open arrows) and non-contacting cell edges (closed arrows). Quantifying the ratio of Celsr1-GFP fluorescence intensity at contacting versus non-contacting surfaces revealed a significant reduction in border enrichment (Figure 3B), suggesting that Celsr1^Crsh disrupts the ability of Celsr1 to concentrate at intercellular junctions.

In Drosophila, core PCP proteins organize into punctate assemblies along cell junctions (Strutt and Strutt, 2009; Strutt et al., 2011). By both confocal and super-resolution structured illumination microscopy (SIM), we observe that Celsr1 is similarly organized into puncta at cell borders in mouse epidermis and keratinocytes (Devenport and Fuchs, 2008; Aw et al., 2016; Figure 4A, Figure 4—figure supplement 1). To test how Celsr1^Crsh affects the nanoscale organization of Celsr1-containing puncta, we performed super-resolution direct stochastic optical reconstruction microscopy (dSTORM), which provides ~20 nm resolution (Rust et al., 2006). In keratinocytes, Celsr1^WT-GFP localized into discrete puncta that were relatively uniform in size and evenly spaced along junctions. Despite appearing largely diffuse at cell–cell borders by widefield microscopy, Celsr1^Crsh-GFP localized into discernable puncta by dSTORM, but these puncta were distributed across a much broader junctional interface and appeared highly disorganized and less compact compared to Celsr1^WT (Figure 4B–D, Figure 4—figure supplement 2A and B).

To characterize the molecular organization of Celsr1 puncta, individual puncta were segmented and those meeting a molecule threshold (≥10) were analyzed further (Figure 4—figure supplement 2C). The number of pixels that failed to meet this qualifying puncta criteria was substantially higher for Celsr1^Crsh-GFP borders (Figure 4—figure supplement 2D), consistent with its diffuse and less clustered localization. A spatial cluster analysis, or Ripley’s K-function (Ripley, 1977), confirmed that Celsr1^Crsh does cluster beyond that expected for a random distribution, but to a lesser degree than Celsr1^WT (Figure 4—figure supplement 2E). Though similar in area, we found that Celsr1^Crsh puncta contain fewer molecules than Celsr1^WT, resulting in puncta with a dramatic reduction in molecular density (Figure 4E–H, Figure 4—figure supplement 2F–H). Together, these findings indicate that the Celsr1^Crsh mutation disrupts the ability of Celsr1 to laterally organize into dense, clustered assemblies.

Cis-interactions stabilize classical cadherins at cell junctions and limit their mobility within the membrane (Harrison et al., 2011; Erami et al., 2015). The diffuse cell border distribution and inability to cluster into dense membrane domains suggested that Celsr1^Crsh may reduce Celsr1 stability. To measure the mobility of Celsr1 at intercellular junctions, we performed fluorescence recovery after photobleaching (FRAP) in keratinocytes expressing three Celsr1 variants – WT, Crsh, and deltaN (ΔN), a deletion that removes most of the extracellular domain. Regions along cell borders between neighboring Celsr1-expressing cells were photobleached and imaged to monitor fluorescence recovery. Celsr1^WT-GFP showed little recovery into the bleached region and only a small fraction of the junctional protein was mobile, indicating Celsr1 is highly stable at junctions (Figure 5, Figure 5—figure supplement 1). Removing the extracellular domain (Celsr1^ΔN), which prevents adhesion in trans, resulted in a dramatic increase (5.5-fold) in mobility at cell borders, demonstrating junctional stability requires intercellular interactions. The rate and total amount of fluorescence recovery for Celsr1^Crsh-GFP were intermediate compared to Celsr1^WT and Celsr1^ΔN variants (3.7-fold increase relative to WT). Additionally, we observed that Celsr1^Crsh mobility at cell borders was similar to both its mobility and that of Celsr1^WT at non-junctional regions or ‘free edges’ (Figure 5—figure supplement 2), with only junctional Celsr1^WT displaying substantial stability. These results suggest that the Celsr1^Crsh mutation lies within a region of Celsr1 that further stabilizes Celsr1 once Celsr1 is already engaged in adhesive interactions between neighboring cells. Overall, given Celsr1^Crsh still mediates adhesion in trans, its higher mobility and diffuse distribution at cell junctions relative to Celsr1^WT suggest that the mutation may impair lateral cis-interactions that function to stabilize Celsr1 adhesion.

Our data thus far support the hypothesis that the Celsr1^Crsh mutation lies within a putative cis-interaction domain. If correct, forcing cis-dimerization of the Celsr1^Crsh mutant protein should rescue its junctional localization. To model lateral cadherin clustering, we fused the FKBP dimerization domain to the cytoplasmic tail of Celsr1^Crsh (Figure 6A). Addition of a cell-permeable, small molecule dimerizer agent rapidly induces FKBP dimerization, driving lateral clustering of the tagged-cadherin (Yap et al., 1997; Chen et al., 2012; Cadwell et al., 2016). Keratinocytes were transfected with Celsr1^Crsh-FKBP-HA and treated with dimerizer or ethanol as a control. Dimerized Celsr1^Crsh-FKBP-HA became enriched at cell–cell borders (twofold increase) and clustered into discrete puncta, whereas the protein remained diffusely localized in ethanol-treated controls (Figure 6B and C, Figure 6—figure supplement 1A). Puncta were not readily observed outside of junctional contacts upon dimerization (refer to non-junctional regions; Figure 6, Figure 6—figure supplement 1). In agreement with non-junctional Celsr1 displaying increased mobility (Figure 5—figure supplement 2), these results suggest that trans-interactions between adjacent cells may be a prerequisite for dimerization to stabilize Celsr1 into discernable puncta.

To test whether induced dimerization could also rescue Celsr1^Crsh interactions with the WT protein in trans, we performed a cell mixing assay (Figure 6D). Keratinocytes were transfected with either Celsr1^WT-GFP or Celsr1^Crsh-FKBP-HA prior to mixing and junction formation. In ethanol-treated controls, neither the WT nor the Celsr1^Crsh variant became enriched at the junctional interfaces between Celsr1^Crsh (bottom cell of pair) and Celsr1^WT (top cell of pair) expressing cells, consistent with the inability of the two variants to form mixed cell aggregates in K562 cells (Figure 2). Strikingly, in the presence of dimerizer, Celsr1^WT-GFP and Celsr1^Crsh-FKBP-HA became concentrated at cell borders, colocalizing in puncta that were similar in size and organization to cells expressing WT Celsr1 alone (Figure 4—figure supplement 1, Figure 6D, Figure 6—figure supplement 1C). Together, these data demonstrate that forced cis-dimerization rescues the adhesion defects caused by Celsr1^Crsh in keratinocytes, supportive of the hypothesis that the Celsr1^Crsh mutation may selectively interfere with the ability of Celsr1 to mediate lateral, cis-interactions.

We next sought to investigate the mechanism by which Celsr1 cis-interactions promote PCP asymmetry. Celsr1, and its Drosophila homologue Fmi, are known to physically associate with core PCP components Fz6 and Vangl2, and are required for their recruitment to cell junctions (Bastock et al., 2003; Chen et al., 2008; Devenport and Fuchs, 2008; Strutt and Strutt, 2008; Struhl et al., 2012). One possible explanation for the loss of asymmetry in Celsr1^Crsh/Crsh mutants is that the Celsr1^Crsh mutation simply abolishes the physical interaction between Celsr1 and Fz6 and/or Vangl2. However, three key pieces of evidence argue against this possibility. First, examination of Fz6 and Vangl2 localization in the epidermis in vivo showed that although their asymmetric localization was severely disrupted in Celsr1^Crsh/Crsh embryos, Fz6 and Vangl2 still localized to cell junctions (Figure 7A). Second, similar to Celsr1^Crsh co-expression with Vangl2 (Devenport and Fuchs, 2008), when co-expressed with Fz6-tdTomato in cultured keratinocytes, Celsr1^Crsh-GFP was still able to drive the redistribution of Fz6 from the cytoplasm to cell contacts (Figure 7—figure supplement 1). Third, Celsr1^Crsh-GFP was able to co-immunoprecipitate both Fz6-tdTomato and tdTomato-Vangl2 from keratinocyte extracts (Figure 7B and C), indicating that the Celsr1^Crsh variant retains the ability to physically interact with both core PCP proteins. Thus, we find that the loss of asymmetry in Celsr1^Crsh/Crsh mutants cannot be attributed to a lack of physical interactions between Celsr1 and Fz6 or Vangl2.

To determine the nanoscale organization of intercellular PCP complexes in vivo, we turned to super-resolution SIM. Given the ~150–200 nm size of Celsr1 puncta observed by dSTORM (Figure 4), we reasoned the ~100 nm resolution of SIM (Galbraith and Galbraith, 2011) would be capable of resolving individual PCP puncta, and perhaps even asymmetric Fz6 and Vangl2 localization. SIM was performed on whole mount preparations of E14.5 dorsal skin labeled with Fz6, Vangl2, and Celsr1 antibodies. This stage of embryonic development was carefully selected because the epidermis is still thin and mostly unstratified, allowing for optimal proximity to the coverslip for imaging, yet far enough along developmentally that PCP proteins are adequately polarized (Aw et al., 2016).

We began by analyzing Fz6 and Vangl2 organization along vertical borders (anterior–posterior junctions), where PCP proteins are enriched, and found that similar to Celsr1, Fz6 and Vangl2 colocalized in punctate assemblies. We often observed asymmetry, or separation of the two proteins across individual junctions with Fz6 toward the left, or anterior of the junction, and Vangl2 to the right, or posterior of the junction (Figure 8A–C). We refer to this distribution as cross-junctional asymmetry. Further, borders often contained multiple asymmetric puncta with Fz6 and Vangl2 partitioned in the same orientation. Although SIM was able to resolve Fz6-Vangl2 asymmetry across junctions, the separation was subtle and a large fraction of their red and green fluorescent signals overlapped, appearing as yellow, colocalized signal. Nevertheless, we reasoned the increased resolution of SIM would allow us to address how Celsr1^Crsh alters the spatial organization of Fz6 and Vangl2.

We predicted that Fz6 and Vangl2 distributions might overlap completely in Celsr1^Crsh/Crsh epidermis, indicating a failure of Fz6 and Vangl2 to partition to opposite sides of the junction. Alternatively, Fz6 and Vangl2 might organize into asymmetric puncta, displaying cross-junctional separation, but their asymmetry would fail to align in a common orientation along cell borders. Surprisingly, we observed neither. Rather, Fz6 and Vangl2 segregated laterally along borders, into distinct, non-overlapping, and often alternating domains (Figure 8A–C). We refer to this distribution as lateral-junctional separation, which is further illustrated by plotting fluorescence intensity measurements along cell borders. Fz6 and Vangl2 signal were out of phase with each other along Celsr1^Crsh/Crsh borders in contrast to the cross-junctional separation and co-aligned distribution in WT (Figure 8D). Occasionally, cross-junctional separation was observed in Celsr1^Crsh/Crsh epidermis within asymmetric puncta of both correct (Fz6–Vangl2) and incorrect (Vangl2–Fz6) orientations (Figure 8B, images 8 and 9). The predominant phenotype, however, was lateral-junctional separation of Fz6 and Vangl2 into distinct, alternating domains. Colocalization and co-alignment of Celsr1 with either Fz6 or Vangl2 along vertical borders was similarly reduced in Celsr1^Crsh/Crsh epidermis (Figure 8—figure supplement 1; Figure 8—figure supplement 2). Thus, despite the ability to physically interact with both Fz6 and Vangl2 biochemically, Celsr1^Crsh fails to properly organize asymmetric assemblies of Fz6 and Vangl2 between cells in vivo.

Though PCP proteins predominately enrich along vertical borders, they do localize to horizontal borders in WT epidermis, albeit at lower levels (Aw et al., 2016). Intriguingly, we also observed lateral-junctional separation of Fz6 and Vangl2 along horizontal borders similar to their distribution in Celsr1^Crsh/Crsh mutants (Figure 8E). To quantify the cross- and lateral-junctional separation phenotypes, we used Pearson’s correlation to measure the co-occurrence of Fz6 and Vangl2 at cell borders. At vertical borders, Fz6 and Vangl2 co-occurrence was greater than 0.7 (±0.04) in WT, Celsr1^+/+ epidermis, and was significantly reduced in Celsr1^Crsh/Crsh (0.38 ± 0.03) (Figure 8F). Along horizontal borders, Fz6 and Vangl2 co-occurrence was significantly lower (0.39 ± 0.03) compared to vertical borders in Celsr1^+/+, and was more in line with their co-localization in Celsr1^Crsh/Crsh junctions (Figure 8F). Thus, at horizontal borders, Fz6 and Vangl2 do not organize into asymmetric co-assemblies between cells, as they do at vertical borders. Similarly, we observed lateral separation of Celsr1 from both Fz6 and Vangl2 at horizontal borders in Celsr1^Crsh/Crsh epidermis (Figure 8—figure supplement 1). This separation of Fz6 from Celsr1 at horizontal borders occurs to a similar extent as observed in Celsr1^+/+ epidermis. Interestingly, Vangl2 separation from Celsr1 at horizontal borders is enhanced in Celsr1^Crsh/Crsh epidermis relative to Celsr1^+/+ (Figure 8—figure supplement 1), suggesting that Celsr1^Crsh may impart a greater deleterious effect on Vangl2 interactions with Celsr1. Even though Vangl2 colocalization with Celsr1 is slightly reduced at Celsr1^Crsh/Crsh horizontal junctions, we collectively observe that the nanoscale organization of PCP proteins in Celsr1^Crsh/Crsh mutants more closely mirrors their organization at WT horizontal borders.

Together, these data demonstrate that SIM resolves the asymmetry of endogenous Fz6 and Vangl2 across individual junctions, consistent with their predicted organization into ‘asymmetric bridges’ from mosaic overexpression studies. Further, these findings provide evidence to support a model (Figure 8G) in which cis-interactions that dimerize and cluster Celsr1 are required for organizing asymmetric Fz6-Vangl2 junctional assemblies.

Mice were housed in accordance with the Guide for the Care and Use of Laboratory Animals in an AAALAC-accredited facility. Animal maintenance and husbandry were in full compliance with the laboratory Animal Welfare Act. All procedures involving animals were approved by Princeton University’s Institutional Animal Care and Use Committee (IACUC). Mixed background cryopreserved Celsr1^Crsh/+ were a kind gift from Elaine Fuchs (Rockefeller University) with permission from Jennifer Murdoch (Royal Holloway University of London, UK), and with rederivation done through the Rutgers Cancer Institute of New Jersey. Experiments were performed on E14.5–15.5 embryos (male and female) from matings of heterozygous mice.

Primary mouse keratinocytes, passage 17–21, derived from Mus musculus CD-1 (Charles River, Strain #022) P0 dorsal epidermis in the Devenport Lab were cultured in E-medium supplemented with 15% fetal bovine serum (FBS) and 0.05 mM Ca²⁺, at 37°C with 5% CO₂, and were free of mycoplasma. For immunofluorescence experiments, cells were cultured on #1.5 glass coverslips coated with fibronectin. Keratinocytes were switch to 1.5 mM Ca²⁺ for the durations indicated to induce junction assembly. Effectene transfection reagent (QIAGEN) was used for transfection prior to processing for immunofluorescence as described below. K-562 cells (CCL-243, ATCC, Manassas, VA) were cultured in Iscove’s Modified Dulbecco’s Medium (IMDM; ATCC, Cat #30–2005 or Fisher, Cat #SH30228) supplemented with 10% FBS at 37°C with 5% CO₂. Stable K-562 cell lines expressing various Celsr1 variants and controls as indicated were generated by FACS sorting post-transfection with SuperFect transfection reagent (QIAGEN) and selection with G418.

Celsr1^Crsh-FKBP-HA was generated by replacing EGFP in Celsr1^Crsh-GFP (Devenport and Fuchs, 2008) with HA-tagged FKBP (pC₄-Fv₁E; ARGENT Regulated Homodimerization Kit, Ariad) digested with HindIII and NotI sites. K14-tdTomato-Vangl2 was constructed by replacing the mCherry cassette (K14-mCherry-Vangl2 Devenport and Fuchs, 2008) with tdTomato via InFusion cloning. For stable K-562 cell lines, Celsr1^ΔN-GFP (Devenport et al., 2011) was subcloned into the pEGFPN1 vector (Clontech) using EcoRI and Not1 sites, WT Celsr1-mCherry was generated by replacing the EGFP cassette of Celsr1^WT-GFP (pEGFPN1-Celsr1 Devenport and Fuchs, 2008) with mCherry using HindIII and NotI sites, and mCherry was subcloned in place of GFP for Celsr1^Crsh-GFP (Devenport and Fuchs, 2008) and E-Cad-Celsr1^CT-GFP (Devenport et al., 2011) to generate mCherry-tagged variants.

Keratinocytes were co-transfected with Celsr1^WT-GFP, Celsr1^Crsh-GFP, or GFP and either tdTomato-Vangl2 or Fz6-tdTomato. Cells were lysed for 1 hr on ice in cadherin extraction buffer (50 mM Tris pH 7.4, 150 mM NaCl, 5 mM CaCl₂, 56 mM MgCl₂, 1% NP40, 1% Triton X) following 48 hr post-transfection with an 8 hr calcium shift. Celsr1^WT-GFP, Celsr1^Crsh-GFP, or GFP immunoaffinity purification was performed using GFP-Trap_MA GFP antibody-coupled magnetic beads (Chromotek) overnight at 4°C. Proteins were resolved on a 7.5% SDS gel and transferred to a nitrocellulose membrane (Bio-Rad). Standard protocols were performed for western blotting using primary antibodies against RFP (rabbit, Rockland Inc, 1:1000) and GFP (chicken, Abcam, 1:2500), and IRDye680 and IRDye800 secondary antibodies (LI-COR, 1:10,000). Immunofluorescence detection of the bands was performed using the LI-COR Odyssey CLx imaging system.

K-562 cells expressing either Celsr1^WT-mCherry or Celsr1^Crsh-GFP were seeded in a 6-well tissue plate and biotinylated with PBS⁺ (containing calcium and magnesium) containing 0.5 mg/mL EX-Link Sulfo-NHS-SS-Biotin at 37°C for 30 min. The following was carried out at 4°C: unbound biotin was quenched with cold PBS+ containing 50 mM Tris (pH 8.0); following cold PBS⁺ washes, cells were lysed in RIPA buffer containing 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, 10 mM Tris-HCl (pH 8.0), 140 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, and protease inhibitors for 10 min; samples were subjected to centrifugation at 13,200 RPM for 10 min and biotinylated protein from the supernatant were incubated with streptavidin beads for 1 hr with rotation; beads were captured by centrifugation at 2500 × g for 1 min and washed with cold PBS⁺. Proteins were released from beads with the addition of sample buffer. Samples were processed for western blot analysis as described above.

K-562 cell lines were cultured for a minimum of 24 hr post thawing prior to aggregation experiments. Cells were washed with PBS without calcium twice and resuspended in fresh growth medium. Cells were then placed on a Gyro mini shaker in an incubator at 37°C and 5% CO₂ for 1–24 hr as indicated to allow for aggregation. Cells were either imaged immediately following aggregation or fixed in 4% PFA for 10 min. Cells were rinsed with PBS and cleared in 50% glycerol for imaging. Cell aggregates were imaged by widefield or confocal microscopy.

Cultured keratinocytes on #1.5 coverslips were fixed in 4% paraformaldehyde (PFA) in PBS for 10 min at room temperature. Cells were washed three times in PBS and permeabilized in 0.1% Triton X-100 (in PBS, PBT) with subsequent PBT washes. Cells were incubated with primary antibodies (in PBT) for 1 hr and secondary antibodies for 30 min with PBT washes in between. Coverslips were washed and mounted in ProLong Gold antifade reagent (Molecular Probes). For whole mount dorsal skin preparations, E14.5 or E15.5 embryos were dissected in PBS+ to maintain calcium-dependent epidermal cell–cell adhesions and fixed in 4% PFA (in PBS+) for 1 hr at room temperature with rocking. Dissected skins were blocked in 0.2% Triton X-100 (PBT2) containing 2.5% normal donkey serum, 1% BSA, and 1% fish gelatin for 2 hr at room temperature or overnight at 4°C. Skins were incubated with primary antibodies overnight at 4°C followed by three 30 min PBT2 washes at room temperature and secondary antibodies plus Hoechst (1:2000) labeling for 2–3 hr at room temperature or overnight at 4°C. Following additional washes in PBT2 and PBS, skins were mounted in either ProLong Gold or Glass mounting reagent (Invitrogen).

Unless specified, widefield images were acquired on a Nikon Ti inverted epifluorescent microscope equipped with Plan Apo objectives (4×/0.13NA, 10×/0.45NA, 20×/0.75NA, 40×/1.3NA oil, 60×/1.4NA oil, 100×/1.4NA oil), four channel epifluorescent filters, CCD camera, and motorized stage.

ImageJ Fiji was used for image processing unless noted otherwise. Confocal and SIM images were background subtracted in ImageJ with a rolling ball radius of 100 and 50 px respectively. Figures were assembled in Adobe Illustrator CS5.1.

Statistical analysis was performed in Microsoft Excel or Prism (GraphPad Software, San Diego CA). See figure legends for additional experimental and statistical details. t-tests between WT and Crsh groups, or as indicated, were performed. p-value significance was defined as follows: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.