Xenopus Kazrin interacts with ARVCF-catenin, spectrin and p190B RhoGAP, and modulates RhoA activity and epithelial integrity

In common with other p120-catenin subfamily members, Xenopus ARVCF (xARVCF) binds cadherin cytoplasmic domains to enhance cadherin metabolic stability or, when dissociated, modulates Rho-family GTPases. We report here that xARVCF binds and is stabilized by Xenopus KazrinA (xKazrinA), a widely expressed conserved protein that bears little homology to established protein families, and which is known to influence keratinocyte proliferation and differentiation and cytoskeletal activity. Although we found that xKazrinA binds directly to xARVCF, we did not resolve xKazrinA within a larger ternary complex with cadherin, nor did it co-precipitate with core desmosomal components. Instead, screening revealed that xKazrinA binds spectrin, suggesting a potential means by which xKazrinA localizes to cell–cell borders. This was supported by the resolution of a ternary biochemical complex of xARVCF–xKazrinA–xβ2-spectrin and, in vivo, by the finding that ectodermal shedding followed depletion of xKazrin in Xenopus embryos, a phenotype partially rescued with exogenous xARVCF. Cell shedding appeared to be the consequence of RhoA activation, and thereby altered actin organization and cadherin function. Indeed, we also revealed that xKazrinA binds p190B RhoGAP, which was likewise capable of rescuing Kazrin depletion. Finally, xKazrinA was found to associate with δ-catenins and p0071-catenins but not with p120-catenin, suggesting that Kazrin interacts selectively with additional members of the p120-catenin subfamily. Taken together, our study supports the essential role of Kazrin in development, and reveals the biochemical and functional association of KazrinA with ARVCF-catenin, spectrin and p190B RhoGAP.


Introduction
Catenins were initially defined as molecules that bind cadherins (Ozawa et al., 1989) and, with the exception of -catenin, that possess a central Armadillo (Arm) domain that mediates cadherin and other protein-protein interactions (Choi and Weis, 2005;Huber et al., 1997;Peifer et al., 1994). Catenins have numerous roles in varied cellular compartments. Members of the p120-catenin subfamily, such as p120-catenin, Armadillo-repeat protein deleted in velo-cardio-facial syndrome (ARVCF)-catenin and -catenin, modulate cadherin stability at cell-cell junctions (Davis et al., 2003;Fang et al., 2004;Gu et al., 2009;Ireton et al., 2002;Xiao et al., 2003). They also directly or indirectly associate with and regulate small GTPases, enabling intracellular signaling and cytoskeletal control (for a review, see Anastasiadis, 2007). Most catenins additionally enter the nucleus (for a review, see McCrea et al., 2009). -catenin, for example, is known to relieve TCF/LEFmediated transcriptional repression in response to canonical Wntpathway stimulation, thereby activating target genes important in development or in pathologies such as cancer (for a review, see Cadigan and Peifer, 2009). Although further study is required to address unresolved issues (Ruzov et al., 2009a;Ruzov et al., 2009b), we and others find that p120 likewise contributes to Wnt signaling (Hong et al., 2010;Iioka et al., 2009;Kim et al., 2004;Park et al., 2006;Park et al., 2005;Spring et al., 2005). In comparison to -catenin and p120-catenin, relatively little is known concerning the interactions and functional roles of ARVCF-catenin or other catenins.
ARVCF is one of several gene products deleted in velo-cardio facial syndrome (Sirotkin et al., 1997). In the context of the cadherin complex, p120 subfamily members including ARVCF bind in a mutually exclusive manner to the cadherin juxtamembrane domain (Mariner et al., 2000). In the nucleus, p120-but not ARVCF-catenin binds the transcriptional repressor Kaiso (Daniel and Reynolds, 1999;Kim et al., 2002) (for a review, see Daniel, 2007). The C-terminal PDZ binding motifs within ARVCF, catenin and p0071 (absent in p120-catenin) bind the ERBIN scaffolding protein (Laura et al., 2002), the ZO-1 and ZO-2 tight junction proteins (Kausalya et al., 2004), and the less-characterized FRMPD2 (Stenzel et al., 2009). In brief, the p120-subclass proteins exhibit both shared and non-overlapping interactions.
To reveal novel ARVCF functions, we screened a Xenopus neurula (stage 18) cDNA library for proteins that interact with Xenopus laevis ARVCF (xARVCF) and identified Xenopus laevis Kazrin (xKazrinA). Biochemically, human KazrinA was previously shown to associate with the peripheral desmosomal proteins periplakin and envoplakin in human keratinocytes (Groot et al., 2004), with microtubules (isoformE) (Nachat et al., 2009), and to modulate RhoA (Sevilla et al., 2008a). We found that xKazrinA interacts directly with xARVCF but not with Xenopus laevis p120 (Xp120) or -catenin, and as reported earlier is present at cell-cell junctions (Groot et al., 2004). Surprisingly, we found that the xARVCF-xKazrinA complex associates and colocalizes with the spectrin cytoskeleton, rather than with cadherins at adherens junctions (Kaufmann et al., 2000;Mariner et al., 2000;Paulson et al., 2000), or with desmosomal core proteins (Groot et al., 2004). Our depletion of xKazrinA resulted in lessened embryonic tissue integrity (Sevilla et al., 2008b). In parallel, xARVCF protein levels were reduced and, supporting their functional interaction, exogenous xARVCF significantly rescued xKazrinA depletion phenotypes. xKazrinA depletion additionally led to RhoA activation, microfilament alterations, and lowered cadherin and cell adhesion levels, which are probably relevant to ectodermal fragility. An additional screen for novel xKazrinA partners resolved Xp190B RhoGAP. In common with xARVCF, p190B partially rescued xKazrinA depletion effects, consistent with functional links existing between components of the xARVCF-xKazrinA-Xp190B complex. Finally, two additional Xenopus p120 subfamily catenins, x-catenin and Xp0071 directly bound xKazrinA. Taken together, we propose that xKazrinA enables xARVCF association with the spectrin-actin network, and that the xARVCF-xKazrinA-Xp190B complex modulates RhoA activity and thereby cytoskeletal organization, cell adhesion and ectodermal integrity.
Prior findings demonstrated that seven human Kazrin splice isoforms (KazrinA-KazrinF and KIAA1026) contain variable N-and C-termini (Groot et al., 2004;Kikuno et al., 1999;Nachat et al., 2009;Wang et al., 2009). As discussed below, KazrinA-KazrinF each encode a putative nuclear localization sequence (NLS), as do the Xenopus isoforms when assessed using PredictNLS server (http://cubic.bioc.columbia.edu.predictNLS) (Cokol et al., 2000) ( Fig. 1A,C). KIAA1026, on the other hand, lacks this cluster of multiple lysine residues; we have designated KIAA1026 as KazrinK. As a preliminary test of the putative NLS functionality, we compared xKazrinA intracellular localization with that of human KazrinK (KIAA1026) in MCF7 cells. Human KazrinK appeared fully excluded, although xKazrinA resided in the nuclei of multiple cells (supplementary material Fig. S1A). This pattern was likewise observed in HEK293 and Xenopus A6 cells (data not shown). The putative NLS lysine residues of xKazrin (365-368; supplementary material Fig. S1B), were then mutated to glutamines to alter their charge but maintain the steric properties of the region (Cokol et al., 2000). Whereas full-length xKazrinA exhibited an even pattern within the cytoplasm and nucleus, or showed predominantly nuclear staining (supplementary material Fig. S1C, left panel), the NLS mutant was excluded from the nucleus. Collectively, these data suggest that the C-terminal polylysine tract of xKazrin functions as an NLS. Although KazrinK and KazrinE isoforms are represented in other mammals, we have not yet identified their homologs in Xenopus. In all cases, the existence of variably spliced and localized Kazrin isoforms is suggestive of functional distinctions.
Using predictive software, the central region of xKazrin and a limited N-terminal region (amino acids 82-255 and 16-41, respectively) were strongly predicted to be coiled coils (both scoring >90% using COILS) (http://www.ch.embnet.org/software/ COILS_form.html) (Lupas et al., 1991). This region bears limited sequence similarity with known coiled-coil domains (Table 2). Using the structural property of 3.5 amino acid residues per helical turn, we modeled five coiled coils within the central region of xKazrin (leucine zipper), and an additional coil in the N-terminal region (supplementary material Fig. S1B and Fig. S2).

xKazrinA interacts with xARVCF-catenin but not with Xp120-catenin
To determine whether the xARVCF-xKazrinA complex could be isolated from in vivo extracts, Myc-tagged xKazrinA was coexpressed in Xenopus embryos with HA-tagged xARVCF or HA-tagged Xp120. In agreement with our yeast two-hybrid results, coimmunoprecipitation assays showed that xKazrinA associates with xARVCF but not with Xp120-catenin ( Fig. 2A).
To test whether xKazrin and xARVCF interact directly, each protein was purified as glutathione-S-transferase (GST)-or maltosebinding protein (MBP)-fusion constructs. Then, GST and MBP pull-down assays were performed (Fig. 2B, left versus right panels). GST-xKazrinA coimmunoprecipitated with MBP-xARVCF but not (or considerably more weakly) with MBP-Xp120, indicating specificity and a direct xARVCF-xKazrinA interaction. Using the same approach, Xenopus -catenin (x-catenin) did not interact with xKazrinA (data not shown).
We further assayed for direct xARVCF-xKazrinA association using blot overlays (Fig. 2C) (Hall, 2004). Purified and immobilized GST-xKazrinA or GST were probed with purified MBP-xARVCF, 4130 Journal of Cell Science 123 (23) MBP-Xp120 or MBP. A clear interaction was observed between membrane-bound GST-xKazrinA with MBP-xARVCF, although none was observed using MBP-Xp120 or MBP probes. Although cofactors might modulate their interaction in vivo, this data indicates that xARVCF and xKazrinA can selectively bind each other in vitro.
To outline the region interacting with xARVCF, we used GST-xKazrinA polypeptides representing the N-terminal, C-terminal, and coiled-coil regions of xKazrinA (GST-xKazrin NT, GST-xKazrin CT, and GST-xKazrin CC, respectively). Using a blot-overlay approach, these fusion proteins were assayed along with full-length GST-xKazrinA for association with MBP-xARVCF (Fig. 2D). xARVCF interacted with both full-length and coiled-coil GST-xKazrinA, but not with N-terminal or C-terminal fusions, suggesting that the central coiled-coil domain of xKazrin is required for xARVCF-xKazrin association. (A)Sequence alignment of X. laevis Kazrin (Xl Kazrin), X. tropicalis Kazrin (Xt Kazrin) and two isoforms of human Kazrin (hKazrinA and hKazrinK). Identical and similar residues are highlighted in black or grey, respectively. The central coiledcoil region (brackets), conserved potential leucine zipper and the predicted nuclear localization sequence (NLS) are indicated. (B)Genomic structure of X. tropicalis Kazrin. Genomic information was obtained from Ensembl database (Ensembl ID #ENSXETG00000013469 and scaffold 257). Asterisk indicates an alternatively spliced exon in X. laevis. (C)Comparison of Xenopus and human Kazrin isoforms. The coiled coil and polylysine tract (NLS) are respectively depicted as grey or black shaded areas. LZ indicates the putative leucine zipper region. Alternative splicing at the Cterminus of the human KazrinA isoform results in additional isoform human KazrinK without the polylysine tract. The total number of amino acids of each protein is indicated. (D)Identification of xKazrin isoforms using primers (F1 and R1) binding to the 5Ј-and 3Ј-flanking regions of exon 6. cDNA was synthesized from embryo stage 18 genomic RNA. Open arrowhead indicates PCR product of xKazrinB cDNA and closed arrowhead indicates that of xKazrinA. MW signifies molecular weight markers.

xKazrinA associates with xARVCF within cells
To investigate the xARVCF-xKazrinA interaction within the physiological conditions provided by vertebrate cells, xARVCFcatenin or Xp120-catenin constructs were targeted to the mitochondrial outer membrane (MOM) via C-terminal fusion to the MOM localization domain of Bcl-X L (Kaufmann et al., 2003;Waibler et al., 2001). Such xARVCF or Xp120 fusion proteins were localized to the MOM when expressed in Xenopus A6 kidney epithelial cells, resulting in a characteristic punctuate pattern (Fig. 3A). xKazrinA relocalized to the MOM when co-transfected with xARVCF-MOM but not with Xp120-MOM (Fig. 3B, top versus bottom panels). Similar patterns appeared when using human embryonic kidney (HEK-293) or human breast cancer cells (MCF-7) (data not shown). These experiments support the specificity and authenticity of the xARVCF-xKazrinA interaction within a cellular context.

Interaction domain mapping of the xARVCF-xKazrinA complex
To resolve xARVCF-xKazrinA interaction domains, deletion constructs were employed. Seven xKazrinA constructs were tested for interaction with xARVCF-MOM in A6 cells ( These data indicate that the region of xKazrinA necessary for interaction with xARVCF resides between amino acid residues 70-354, containing the coiled-coil domain of xKazrin up to the potential A/K isoform splice site at amino acid residue 354. This region is present in all identified isoforms of Kazrin (frog, human, mouse, rat and Xenopus), suggesting that ARVCF-Kazrin interactions might be shared. Indeed, as shown in supplementary material Fig. S9B, human KazrinE and KazrinK isoforms coimmunoprecipitate with xARVCF.
To map the interaction region(s) of xARVCF, we used MOMtargeted deletion constructs comprised of one or two of the three xARVCF domains (N-terminus, Arm domain and C-terminus;  (A)Xenopus embryos expressing HA-xARVCF or HA-Xp120 with (+) or without (-) coexpressed Myc-xKazrinA were harvested at gastrula stage (stage 10.5). Lysates were immunoprecipitated using pre-immune, xARVCF or Xp120 polyclonal antisera, and interactions with xKazrinA detected by immunoblotting using anti-Myc antibody. Total lysates (5%) were also blotted. (B)In vitro coprecipitation of GST-xKazrinA with MBP-xARVCF. GST-xKazrinA was incubated with MBP-xARVCF or MBP-Xp120, precipitated using glutathione-agarose (left panels) or amyloseagarose (right panels) beads and the samples subjected to SDS-PAGE and immunoblotting using anti-MBP (top panels) or anti-Kazrin (bottom panels) antibodies. (C)Direct interaction of purified xARVCF with xKazrinA using blot overlay. Top panel: Duplicate lanes of purified GST-xKazrinA were electrophoresed and immobilized on nitrocellulose then probed using purified MBP-xARVCF, MBP-Xp120 or MBP. Protein complexes were then identified using anti-MBP antibodies (top-left panel). Presence of GST-xKazrinA was detected using anti-GST antibodies (top-right panel). As a negative control, purified GST was immobilized on nitrocellulose and processed in an identical manner (bottom panels). (D)Blot overlay of GST-xKazrinA deletion constructs. Purified GST-xKazrinA, GST-xKazrinA N-terminal (NT), GST-xKazrinA C-terminal (CT), GST-xKazrinA coiled-coil (CC), and GST proteins were electrophoresed and immobilized on nitrocellulose. Total loading was detected using anti-GST antibodies (left panel). The right panel was probed with MBP-xARVCF and the resulting protein complexes detected using anti-MBP antibodies. (Kausalya et al., 2004;Laura et al., 2002;Stenzel et al., 2009)]. Relocalization also failed when using the xARVCF Arm domain (Fig. 3C, HA xARVCF-Amom), or upon removal of the xARVCF C-terminus (Fig. 3C, HA xARVCF-NAmom). These data collectively indicate that a region(s) comprised of the Arm and Cterminal domains of xARVCF is required for interaction with the coiled-coil domain of xKazrin (Fig. 3D).

xKazrinA and xARVCF colocalize at cell-cell contacts in Xenopus blastula ectoderm
To assess xKazrin intracellular distribution in Xenopus tissue, immunofluorescence was performed of Xenopus blastula ectoderm (animal caps) expressing Myc-xKazrinA. Most xKazrinA protein was present at cell-cell borders, with some staining evident in the nucleus (data not shown). The nuclear presence of Kazrinin in vivo is in accordance with its nuclear localization in cell lines and human skin (supplementary material Fig. S1) (Nachat et al., 2009;Sevilla et al., 2008a).

Journal of Cell Science 123 (23)
To assess HA-xARVCF and Myc-xKazrinA colocalization following coexpression in vivo, immunofluorescent images were acquired from the outer ectodermal cell layer at blastula stages. Colocalization occurred along cell-cell borders in the X-Y plane ( Fig. 4A-C), and to a significant if not complete extent along the Z-axis (Fig. 4D). In general, these findings are consistent with the biochemical association of xARVCF and xKazrinA.

The xARVCF-xKazrinA complex is not a core component of adherens junctions or desmosomes
Because xARVCF directly binds cadherin juxtamembrane regions (Kaufmann et al., 2000;Mariner et al., 2000;Paulson et al., 2000), we tested whether xKazrinA and xARVCF colocalize with Ccadherin, a major cadherin essential for Xenopus early development (Heasman et al., 1994;Lee and Gumbiner, 1995). Xenopus Ccadherin (xC-cadherin) colocalized at cell-cell borders with xKazrinA as well as with xARVCF ( Fig. 4E-L), consistent with a possible association of xKazrinA with adherens junction components. . Once fused to a peptide sequence tag from human Bcl-Xl directing their ectopic localization to the MOM, xARVCF and Xp120 displayed a punctate pattern characteristic of mitochondrial localization (bottom panels). (B)Coexpression of Myc-xKazrinA with MOM-targeted HA-xARVCF or HA-Xp120 (top and bottom panels, respectively). Cells were coimmunostained for localization of the catenin (HA-epitope, left panels) and for xKazrinA (Myc-epitope, center panels). Co-localization of xKazrinA was observed with xARVCF but not with Xp120, as shown in the merged images (right panels). (C)List of Myc-tagged (MT) xKazrinA and HA-tagged xARVCF constructs (see text for details) and their interaction status. Asterisk indicates mutation of lysine to glutamine; LZ, leucine zipper; mom, mitochondrial outer membrane; ARM, Armadillo domain. The co-relocalization of xKazrinA deletion mutants was tested in the presence of full-length MOM-targeted xARVCF (upper set). Conversely, full-length xKazrinA was tested for co-relocalization with deletion constructs of MOM-targeted xARVCF (lower set). A positive score (+) indicates that numerous cells coexpressing the indicated xKazrinA or xARVCF deletion constructs showed colocalization of xKazrinA with the MOM-targeted xARVCF construct. (D)Representation of the xARVCF-xKazrinA interaction, indicating the putative regions of association. Scale bars: 20m.
To test whether xKazrinA and C-cadherin form a larger complex, presumably bridged by xARVCF, we overexpressed Myc-xKazrinA followed by C-cadherin immunoprecipitation and western blotting (Fig. 4M). As anticipated, endogenous C-cadherin co-precipitated endogenous xARVCF, as well as Xenopus -catenin and plakoglobin. Importantly, however, it did not co-precipitate Myc-xKazrinA. In further tests, we simultaneously overexpressed three components: HA-xKazrinA, HA-xARVCF and the cytoplasmic tail of C-cadherin (Myc-tagged). Whereas xARVCF coimmunoprecipitated with the cytoplasmic domain of C-cadherin as expected, HA-xKazrinA was not present in conjunction with cadherin (Fig. 4N). Furthermore, we could not detect an interaction between xKazrinA and xC-cadherin following chemical crosslinking [using DTSSP; 3,3Ј-dithiobis(sulfosuccinimidyl propionate)] (data not shown). These results indicate xC-cadherin-xARVCF association as anticipated, but given our coimmunoprecipitation results, not cadherin association with xKazrinA.
Because Kazrin associates with envoplakin and periplakin at desmosomal or interdesmosomal regions in human keratinocytes (Groot et al., 2004), and ARVCF-and p120-catenins have been reported at desmosomes (Borrmann et al., 2006;Johnson and Boekelheide, 2002;Kanno et al., 2008a;Kanno et al., 2008b), we wanted to know whether ARVCF is resident at desmosomes in human keratinocytes. As shown in Fig We then employed transmission EM, but in our hands this did not further clarify ARVCF or Kazrin presence at desmosomes and/or adherens junctions (data not shown).
To further probe desmosomal associations, we turned to biochemical precipitations from Xenopus embryo ( Fig. 5J) and human epithelial carcinoma cell extracts (A431) (Fig. 5K). The cytoplasmic tail of human desmoglein1 ( Fig. 5J) or full-length desmoglein1 ( Fig. 5K) failed to co-precipitate Xenopus or human ARVCF from either system. Plakoglobin and C-cadherin were positive controls for the respective binding functionality of desmoglein and ARVCF in Xenopus embyos (Fig. 5J). As evaluated by these criteria, our data do not support xARVCF as being a core component of the desmosome.

xKazrinA and xARVCF form a complex with the x2-spectrin cytoskeletal protein
To further address how the xARVCF-xKazrinA complex localizes to junctional areas, we performed yeast two-hybrid screening of xKazrinA to identify novel binding partners (Hybrigenics). From an adult mouse brain cDNA library, we obtained 64 positive clones (partial list in supplementary material Table S1). Notably, nine clones encoded plasma-membrane-associated proteins. These included two p120-catenin subfamily members (-catenin and p0071), the tight junction protein symplekin (Keon et al., 1996), the spectrin binding protein Camsap1 (Baines et al., 2009;Yamamoto et al., 2009), and three spectrin family members (1-, 2and 2-spectrins) (for a review, see Bennett and Healy, 2009). This screen was not saturating because we did not isolate ARVCF- (M)Co-immunoprecipitation of Myc-xKazrinA (MT-xKazrinA) with adherens junction components. Myc-xKazrinA was expressed in early embryos lysed at stage 9-10 and endogenous xC-cadherin was immunoprecipitated. Co-precipitated proteins were detected by immunoblotting using the indicated antibodies. (N)Co-immunoprecipitation tests following the coexpression of Myc-xC-cadherin tail, HA-xARVCF and HA-xKazrinA. Myc-xC-cadherin tail was precipitated using anti-Myc antibodies. Co-associated HA-xARVCF and HA-xKazrinA were detected (versus not) using anti-HA antibody. Asterisk indicates migration of IgG heavy chains (50 kDa). Scale bars: 50m. catenin, envoplakin or periplakin. Contributing factors could be lower ARVCF expression in mammals than in amphibians (Mariner et al., 2000), and little envoplakin or periplakin presence in neural tissues (Aho et al., 1998).
Spectrin as a potential Kazrin-binding partner was intriguing given that our earlier structural analysis of xKazrinA predicted its coiled-coil domain to be similar to the spectrin-repeat region of spectrin (GenThreader program in PSIPRED Protein Structure Prediction Server; http://bioinf.cs.ucl.ac.uk/psipred/; P-value  4e-04) (Jones et al., 1999). In mammals, there are two -spectrin genes and five -spectrin genes. Their gene products form heterodimers that orient end-to-end to form  2  2 -tetramers. Tetramers in turn interact with microfilaments, neighboring spectrin tetramers, and other proteins such as band 4.1, ankyrin, adducin and calmodulin, forming a cytoskeletal web that functions in many contexts (for a review, see Bennett and Healy, 2009). Anchoring points, such as provided by band 4.1 and ankyrin, facilitate the 4134 Journal of Cell Science 123 (23) plasma membrane association of the spectrin cytoskeleton, as do junctional cadherins (see later) (Bennett and Healy, 2009). In Xenopus, two spectrins are known, -spectrin (also known as fodrin) and -spectrin (Giebelhaus et al., 1987;Klein et al., 2002). Xenopus-spectrin (x-spectrin), although highly expressed in oocytes, becomes reduced following fertilization (Giebelhaus et al., 1987), whereas x-spectrin-encoding mRNA is expressed throughout development, as assessed via semiquantitative RT-PCR (supplementary material Fig. S4B).
We wondered whether the spectrin network might be relevant to xARVCF-xKazrinA localization to cell-cell borders. Confocal microscopy and biochemical approaches were employed using a partial x2-spectrin cDNA (GenBank #BC046267.1). This construct is the longest available x-spectrin, and is homologous to human 2-spectrin isolated from our yeast two-hybrid screen (74.6% identical). It encodes the N-terminal actin-binding domain of 2spectrin (ABD; which binds band 4.1 and adducin), and five spectrin repeats ( Fig. 6A and supplementary material Fig. S4A). Isolation of shorter yeast clones from our two-hybrid screen suggested that the 1or 2-spectrin regions extending from the end of the second spectrin repeat to the middle of the third repeat are sufficient to bind Kazrin (data not shown).
The xKazrinA-x2-spectrin interaction was supported using coprecipitation tests of Xenopus embryo lysates in which full-length xKazrinA and partial x2-spectrin proteins were coexpressed (Fig.  6B). Furthermore, using confocal microscopy, xKazrinA and x2spectrin colocalized as expected to cell-cell borders in blastula ectoderm (Fig. 6C). To test whether x2-spectrin binds to the xARVCF-xKazrinA complex in vivo, varying amounts of spectrin were expressed in Xenopus embryos with a constant amount of xKazrinA, and the endogenous xARVCF complexes isolated and assayed (Fig. 6D). x2-spectin as well as xKazrinA co-precipitated with xARVCF. Because x2-spectrin expression did not alter xARVCF-xKazrinA co-precipitation, we expect that a ternary xARVCF-xKazrinA-x2-spectrin complex was formed. However, because a prior study reported association of 2-spectrin with Ecadherin (through ankyrin) (Kizhatil et al., 2007), it was conceivable that xARVCF association with x2-spectrin occurred via their mutual association with cadherin. We tested this possibility using co-precipitation assays. As shown in Fig. 6E, xARVCF but not xCcadherin associated with x2-spectrin and xKazrinA. Thus, the ARVCF-Kazrin complex probably associates with spectrin through Kazrin binding to spectrin.

xARVCF partially rescues embryonic phenotypes resulting from xKazrin depletion
Morpholino-directed Kazrin depletion in early X. tropicalis embryos was previously reported to result in ectodermal blistering, as well as a shortened body axis, head reductions and somitic defects (Sevilla et al., 2008b). Because the mRNA sequence targeted by the X. tropicalis morpholino is completely conserved with that in X. laevis, we used X. tropicalis morpholino (KMO1) in our study, and demonstrated its efficacy in blocking in vitro translation of Kazrin-encoding mRNA, and in knocking-down X. laevis Kazrin to 60% of endogenous levels by neurulation stages (significant maternal protein loading precluded obtaining earlier effects) (supplementary material Fig. S6A and Fig. S6B, respectively). The phenotypes appearing in X. laevis were similar to those previously indicated in X. tropicalis (Sevilla et al., 2008b). However, Kazrin depletion in X. laevis appeared to have a more severe impact upon the ectoderm, as evident in abnormal In addition, Kazrin localization was compared to that of endogenous desmoglein3 (H), a desmosomal cadherin. Merged images (F,I) show partial colocalization of Kazrin with desmoplakin and extensive colocalization with desmoglein3. Colocalization between desmoplakin and ARVCF is shown in C. Photos represent projection images of deconvolved Z-stacks. Scale bars: 10m. (J)Interaction between desmosomal cadherin desmoglein1 (Dsg1) and xARVCF in Xenopus embryo lysates. Indicated amounts of in vitro transcribed RNAs were microinjected into one-cell stage embryos. Then, early gastrula embryos were lysed and human desmoglein1 tail or xARVCF were immunoprecipitated, followed by SDS-PAGE and western blotting. (K)Endogenous desmoglein1 or E-cadherin were immunoprecipitated from confluent A431 human epithelial carcinoma cells. Bound endogenous ARVCF was detected with human ARVCF-specific antibody. neural tube closure during neurulation (supplementary material Fig. S6C, upper panels), and in regional ectoderm shedding in addition to blistering (noted previously) at early tailbud stages (supplementary material Fig. S6C, bottom panels and Fig. 7A, middle panel). The fraction of embryos exhibiting such phenotypes increased in a dose-dependent manner (supplementary material Fig. S6D). Because we obtained the same results using an independent non-overlapping Kazrin morpholino in X. laevis, the specificity of our phenotype was supported (data not shown). To test for an in vivo functional interaction between xARVCF and xKazrinA, we attempted to rescue ectodermal shedding through ectopic expression of xARVCF (Fig. 7, Table 3). As shown in Fig. 7B, Kazrin depletion was partially rescued by an appropriately titrated dose of ectopic xARVCF (dose subphenotypic in isolation), consistent with xKazrin-xARVCF functional interactions in maintaining ectodermal integrity.

Kazrin depletion decreases cadherin protein levels and cell-cell attachment
Our data suggested that xKazrinA and xARVCF form a complex that is capable of associating with spectrin, and furthermore that the xKazrinA-xARVCF complex participates in maintaining ectodermal integrity. However, it remained unclear at the mechanistic level how xKazrinA depletion alters ectodermal integrity. Given their 4135 ARVCF-catenin interacts with Kazrin established contributions to cell adhesion, we first considered the effects upon C-cadherin and E-cadherin. Interestingly, the levels of both cadherins were decreased after xKazrin depletion (Fig. 8A), as was that of desmoglein1, a desmosomal cadherin (supplementary material Fig. S5). A previous report employing cell lines used immunofluorescence to indicate that Kazrin overexpression lowers desmoplakin and perhaps E-cadherin levels (Sevilla et al., 2008a). Intriguingly, both in our and others' hands (Sevilla et al., 2008b) (data not shown), Kazrin overexpression in Xenopus does not generate an obvious phenotype(s). Although speculative, it is possible that effects occur but are more effectively compensated for in vivo than in vitro. To evaluate xKazrin depletion effects on cell-cell adhesion in a more controlled setting, we performed cell dissociation and cell re-aggregation assays using blastula ectoderm tissue (animal caps). xKazrin-depleted tissue exhibited accelerated dissociation compared to controls (Fig. 8B) and, correspondingly, they experienced delayed reaggregation (data not shown). Considering the key roles of cadherins in determining the outcome of such dissociation-reaggregation assays (Turner et al., 1992), together with our biochemical evidence, the observed reduction of ectoderm integrity is probably due to reduced cadherin levels following xKazrin depletion. In support of this proposition, xKazrin depletion phenotypes were partially rescued upon ectopic expression of an appropriately titrated dose of C-cadherin (subphenotypic in isolation) Enzymatically synthesized mRNAs encoding Myc-xKazrinA and HA-x2-spectrin were co-injected into one-cell stage embryos, and the blastula embryo lysates were HAimmunoprecipitated followed by Myc immunoblotting. (C)Colocalization of xKazrinA and x2-spectrin in blastula ectoderm. Animal caps were isolated from blastula embryos (stage 9-10) expressing Myc-xKazrinA and HA-x2-spectrin, followed by their respective (Myc and HA) immunofluorescent detection. (D)Formation of xARVCF, x2-spectrin and xKazrinA ternary complex. Blastula lysates from embryos injected with indicated amounts of each mRNA were immunoprecipitated with xARVCF antibody, followed by immunoblotting for Myc-xKazrinA and HA-x2-spectrin. (E)ARVCF-specific ternary complex. Lysates from embryos expressing x2spectrin and xKazrinA were immunoprecipitated using the indicated antibodies. Co-precipitating x2-spectrin or xKazrinA were detected by immunoblotting. (Fig. 8C). Together, our in vivo data suggests that Kazrin positively modulates cell-cell contacts, with the xARVCF-xKazrinA complex probably being pertinent to effects upon cadherin levels.

Kazrin depletion leads to increased RhoA activity and cortical actin disorganization
A remaining issue was how xKazrinA, or the xARVCF-xKazrinA complex, modulates cadherin stability. p120 subfamily catenins including ARVCF are thought to stabilize cadherins by binding cadherin membrane-proximal domains and blocking certain protein associations (such as with Hakai or presenilin-1) (Baki et al., 2001;Fujita et al., 2002;Marambaud et al., 2002), that would otherwise promote cadherin degradation and/or endocytosis (Davis et al., 2003;Ireton et al., 2002;Xiao et al., 2003). Regulation of Rho-GTPases by p120 subfamily catenins is further relevant to cadherin 4136 Journal of Cell Science 123 (23) stabilization and/or endocytosis (Akhtar and Hotchin, 2001;Charrasse et al., 2006;Izumi et al., 2004;Lamaze et al., 1996;Leung et al., 1999) (for a review, see Anastasiadis, 2007;Xiao et al., 2007). Interestingly, a non-biased screen earlier correlated Kazrin overexpression with inhibition of clathrin-mediated endocytsosis of transferrin receptor (Schmelzl and Geli, 2002). Thus, Kazrin together with p120 subfamily catenins might be relevant to cadherin endocytosis. Indeed, cadherin levels after Kazrin depletion were restored when embryos were incubated with a specific dynamin inhibitor (Dynasore; Fig. 8D) (Macia et al., 2006), and ectodermal defects were partially rescued in parallel (data not shown). This suggests a potential role of Kazrin in cadherin endocytosis, although we never resolved Kazrin in complex with cadherin using standard immunoprecipitation (Fig.  4M,N) or chemical crosslinking (using DTSSP, data not shown). Furthermore, although less compelling, we never observed that association of xARVCF with cadherin was altered by xKazrinA expression (data not shown).
In common with cadherin, the level of xARVCF, Xp120 and xcatenin became decreased in response to xKazrin depletion (data not shown). One possible explanation for the modulation of cadherin protein levels by Kazrin is its above-shown relationship with ARVCF. On the basis of published reports, lowered ARVCF (as for p120) levels are expected to result in lowered cadherin levels (Davis et al., 2003;Fang et al., 2004;Ireton et al., 2002;Xiao et al., 2003). Symmetrically, lowered cadherin presence can result in reduced catenin levels (e.g. of -catenin) (Papkoff, 1997;Thoreson et al., 2000). At this time, we do not know the temporal hierarchy of events following Kazrin perturbation that result in altered levels and/or functions of cadherin-catenin complex components. This will require further study.
Nonetheless, given the known effects of ARVCF-catenin (and p120-catenin) on small GTPases (Anastasiadis et al., 2000;Fang et al., 2004;Grosheva et al., 2001;Noren et al., 2000), we turned our attention to xKazrinA effects on RhoA and Rac, and on cortical actin organization in plasma-membrane regions. We initially evaluated RhoA activity in embryos following the depletion of xKazrin and/or xARVCF (Fig. 8E). Interestingly, xKazrin depletion significantly increased RhoA activity, an effect likewise observed (and expected) following xARVCF knockdown (Fang et al., 2004). Coordinate xKazrin and xARVCF depletions had an additive effect on RhoA activity, as seen by comparing 10 ng deliveries of each morpholino (Fig. 8E, lane 6 versus 2 and 4). Because Rac1 was not affected by xKazrin depletion (data not shown), Kazrin might be a RhoAspecific modulator. Our results are in line with a report indicating that Kazrin overexpression inhibits RhoA in human keratinocytes, with no apparent effect on Rac1 (Sevilla et al., 2008a). Thus, although Kazrin and xARVCF appear to have shared inhibitory effects on   Table 3. RhoA, one distinction (relative to ARVCF and related p120 subfamily members) is that Kazrin is not likely to modulate Rac1 in this context (Anastasiadis et al., 2000;Fang et al., 2004;Grosheva et al., 2001;Noren et al., 2000). Because RhoA activation might have contributed to the ectoderm fragility we observed following Kazrin depletion, we tested the rescuing capacity of carefully titrated dominant-negative RhoA (versus dominant-active RhoA), upon coinjection with KMO1. In line with RhoA being functionally relevant and possibly downstream of Kazrin, co-injection of dominantnegative RhoA mildly rescued observed phenotypes, whereas titrated dominant-active RhoA doses only worsened developmental defects (data not shown).
Because ectopic RhoA activation modulates actin and cell structure and/or cell morphology (Paterson et al., 1990), we employed phallodoin staining to grossly resolve Kazrin-depletion effects on actin organization in embryos (Fig. 8F). Whereas control embryos exhibited a defined enrichment of cortical actin at cell-cell borders, Kazrin-depleted embryos displayed a disorganized pattern and an actin signal of reduced thickness along the Z-axis (Fig. 8F). The cytoskeletal disorganization we resolved in vivo appears to be in keeping with a previous report employing keratinocytes (Sevilla et al., 2008a). In summary, Kazrin appears to be a negative regulator of RhoA in vivo, suggesting that increased RhoA activity upon

xKazrinA interacts with Xp190B RhoGAP
We wondered whether Kazrin might modulate RhoA via another protein(s), because predictive analysis suggested that Kazrin was unlikely to directly bind RhoA. Common modulators of RhoA include Rho GDIs (Rho GDP dissociation inhibitors), RhoGAPs (Rho GTPase activation proteins) and Rho GEFs (Rho guanine nucleotide exchange factors). Intriguingly, from our xKazrinA yeast two-hybrid screen, we identified multiple independent clones of mouse p190B RhoGAP (supplementary material Table S1). This novel Kazrin interaction suggested one means by which it might influence RhoA activity and function.
There are two p190 RhoGAPs, p190A and p190B, and both are widely functioning negative regulators of Rho GTPases (Vincent and Settleman, 1999). Recently, p120-catenin was shown to mediate RhoA inhibition via the recruitment of the p190A RhoGAP (Niessen and Yap, 2006;Wildenberg et al., 2006). This suggested that xARVCF-catenin might analogously associate with p190B, bridged or facilitated by Kazrin. Upon analysis of our yeast two-hybrid findings (alignment of limiting sequences from 22 independent  Table S2. (D)Effect of inhibiting clathrin-mediated endocytosis on the reduced xE-cadherin levels observed following xKazrin depletion. One-cell stage embryos were injected with the indicated morpholinos. Embryos were subsequently (32-64 cell stages) incubated with 50 mM Dynasore (versus DMSO solvent). Lysates from neurula (stage 18) embryos were immunoblotted for xE-cadherin and GAPDH. (E)Rho activity in response to xARVCF and/or xKazrin depletion. The indicated morpholinos were injected into one-cell stage embryos, and lysates from late blastula (stage 9-10) embryos were tested for Rho activity in vitro. (F)Actin staining of blastula ectoderm after Kazrin depletion. Animal cap were isolated as mentioned in B and filamentous actin in deep (inner) cells was visualized using phalloidin-Alexa 488 fluorescence. Scale bar: 50m. clones), we deduce that xKazrinA binds to a region within or spanning the third to fourth FF domains of p190B (   Table S3. (E)Ectopic p190B rescues xE-cadherin levels, reduced following xKazrin depletion. The indicated morpholinos and mRNAs were injected into one-cell embryos. Neurulation (stage 18) extracts were immunoblotted as indicated. (F)In vivo ternary complex formation of xARVCF-xKazrinA-Xp190B. Extracts from gastrula-stage embryos expressing the indicated proteins and endogenous xARVCF were precipitated for xARVCF. Co-precipitated xKazrinA and Xp190B were detected using antibodies directed against Myc and HA, respectively. (G)In vitro ternary protein complex of xARVCF-xKazrinA-Xp190B. In vitro transcribed and translated (IVT) Xp190B was incubated with bacterially purified GST-xKazrinA and MBP-xARVCF. Pull-down of xARVCF occurred using amylose-agarose beads, followed by immunoblotting as indicated.
To further test the validity of the interaction of Kazrin with p190B, we undertook precipitations from embryo extracts and GST purified-component systems, using cDNA encoding human p190B, as well as the longest available (partial) cDNA encoding Xenopus p190B (79% sequence identity with human) (Fig. 9A; supplementary material Fig. S7). Following expression in vivo, both full-length human p190B and our partial Xp190B construct robustly coimmunoprecipitated with xKazrinA (Fig. 9B), and they colocalized to cell-cell borders (Fig. 9C). Finally, in a reductionist context suggesting their direct interaction, GST-xKazrin was co-precipitated with in vitro translated p190B (supplementary material Fig. S8).
Given their physical association, we asked whether the xKazrinA-p190B interaction might be functionally evident in vivo. For a read-out, we evaluated phenotypic rescue of ectodermal shedding, as described above for embryos depleted of xKazrin. Indeed, shedding phenotypes were partially but significantly rescued by titrated amounts of ectopic p190B (human full-length), and this occurred in a dose-dependent manner ( Fig. 9D and supplementary material Table S3). These results support the concept that p190B and Kazrin are functionally coupled. Furthermore, the decrease in E-cadherin levels upon Kazrin depletion was restored by ectopic expression of human p190B RhoGAP (Fig. 9E), consistent with the possibility that lowered cadherin levels in Kazrin knockdown embryos might result from abnormal activation of RhoA.
To investigate whether xKazrin might facilitate p190B association with xARVCF, we asked whether Kazrin expression (versus not) had an impact on p190B association with xARVCF. As indicated in Fig. 9F (lane 3), a trace but reproducible amount of p190B was detected in complex with ARVCF only upon Kazrin coexpression. We next turned to in vitro pull-down assays to evaluate how p190B might associate with ARVCF (Fig. 9G). As expected on the basis of the precipitations from Xenopus extracts, in vitro translated Xp190B was pulled down with MBP-xARVCF in the presence of GST-xKazrinA (Fig. 9G, lane 3). This suggests that these three proteins form a ternary complex. Even in the absence of xKazrinA, Xp190B appeared to weakly bind xARVCF (Fig. 9G, lane 1). Thus, although requiring further study, it appears that Kazrin might recruit and/or enhance p190B association with ARVCF. In all cases, our data support the direct association and functional interaction of xKazrin with Xp190B RhoGAP.

In addition to xARVCF-catein, xKazrinA binds x-catenin and Xp0071-catenin
As mentioned, our xKazrin yeast two-hybrid screen suggested that, in addition to xARVCF-catenin, two related members of the p120 subfamily interact with xKazrinA, namely x-catenin and Xp0071catenin (supplementary material Table S1). Pull-down validation tests were performed using purified GST-xKazrinA protein and in vitro translated x-catenin or Xp0071-catenin. Positive and negative controls, respectively, included xARVCF-and Xp120-catenin (Fig.  10). In agreement with our yeast two-hybrid assays, xARVCFcatenin, x-catenin and Xp0071-catenin bound xKazrinA, whereas Xp120-catenin did not. This suggests that the biochemical and presumably functional interactions of Kazrin involve some but not all p120 subfamily catenins.

Discussion
In this study we describe the identification of xKazrinA as a direct binding partner of xARVCF-catenin, a member of the p120-catenin subfamily. We additionally resolved the association of of xKazrinA with spectrin and p190B RhoGAP. Distinct from reported protein interactions of ARVCF or KazrinA at the adherens junction or desmosome, we find that KazrinA associates with spectrin. Specrin forms a cytoskeletal network that includes interactions with junctional constituents (Bennett and Healy, 2009;Kizhatil et al., 2007). Although requiring further study, Kazrin appears to enhance the association of ARVCF with p190B RhoGAP, with the reduction of RhoA activity by Kazrin being likely to occur via these same interactions (at least in part). Borrowing from independent studies (Gliem et al., 2010;Lamaze et al., 1996;Schmalzing et al., 1995;Sullivan et al., 1999), we postulate that such RhoA modulation is relevant to the organization of cortical actin and cadherin stability. On the basis of our rescue analyses, these biochemical and functional relationships appear to be relevant to effects observed following Kazrin depletion in Xenopus embryos, which included reduced ectodermal integrity (see Fig. 11). Finally, we find that some, but not all, p120 subfamily members (not p120-catenin) bind xKazrinA. Our results thus constitute an initial outline of the role of Kazrin in the context of catenins, RhoA and p190B RhoGAP.

Kazrin isoforms and structure
Kazrin was originally identified as a cDNA (KIAA1026) expressed in human brain (Kikuno et al., 1999). In human keratinocytes, five Kazrin splice isoforms (KazrinA-Kazrin E) were then reported, along with their biochemical interactions and functional effects on keratinocyte differentiation, small GTPases and actin or microtubule networks (Groot et al., 2004;Nachat et al., 2009;Sevilla et al., 2008a). Adding to the functional complexity of Kazrin is KazrinF, recently reported to inhibit apoptosis through binding BAX (Bcl-2associated X protein) and ARC (apoptosis repressor with caspase recruitment domain) in a human glioma cell line (Wang et al., 2009).
Our evaluation of KIAA1026 in particular showed that it lacks the NLS we characterized in other Kazrin isoforms, resulting in its exclusion from the nucleus. We propose that KIAA1026 be renamed KazrinK. Although the nuclear roles of all NLS-containing Kazrin isoforms are uncertain, both human KazrinE and human KazrinK associate with xARVCF (supplementary material Fig. S9B), possibly suggesting they have shared cytoplasmic and cytoskeletal functions. Structurally, xKazrin and its well-conserved vertebrate homologs are likely to represent a new protein group. All Kazrins examined share limited homology with the known coiled-coil regions as represented in Table 2. Secondary structural prediction of the coiled-coil region of Kazrin further indicated resemblances to the folding of spectrin repeats within -spectrin (Pascual et al., 1997). Each of these proteins possesses extensive -helical or coiled-coil structure engaging in homotypic or heterotypic protein-protein interactions (Burkhard et al., 2001). Indeed, we find that the heterotypic interactions between Kazrin and ARVCF requires the predicted coiled-coil domain of Kazrin (Fig. 3C,D). Although possible homotypic coiled-coil associations will require further study, we observed that xKazrinA constructs bearing different epitope tags co-precipitate (supplementary material Fig. S9A), and that human Kazrin isoforms co-associate in a human cell line (Nachat et al., 2009).

Protein partners of ARVCF-catenin
A number of proteins have been shown to directly bind ARVCF. These include cadherins, RhoA, Vav2, ERBIN, FRMPD2, ZO-1 and ZO-2 (Fang et al., 2004;Izawa et al., 2002;Kaufmann et al., 2000;Kausalya et al., 2004;Laura et al., 2002;Magie et al., 2002;Mariner et al., 2000;Noren et al., 2000;Paulson et al., 2000;Stenzel et al., 2009). The latter four each contain PDZ domains, which are thought to associate with PDZ binding motifs within the C-termini of ARVCF, -catenin and p0071. Our present study adds Kazrin as the newest binding partner of p120 subfamily members, excluding p120 itself. Kazrin seems to bind a relatively broad region of ARVCF (and probably other catenins), encompassing portions of the central Arm and C-terminal domains. In turn, the extended coiled-coil domain of Kazrin associates with ARVCF, such that the coiled-coil regions of each protein presumably interact. Because the Arm domain of ARVCF exhibits some homology with p120-catenin (55% identity and 74.4% similarity) (Sirotkin et al., 1997), it may prove interesting to resolve the basis of specificity among different p120 subfamily proteins. It will furthermore be relevant to test whether the larger ARVCF-Kazrin complex includes proteins known to bind via the PDZ binding motif within ARVCF, such as ERBIN (Laura et al., 2002), as this would provide further insight to the funcions of the complex.

Subcellular localization of the xARVCF-xKazrinA complex
In cornified epithelia and other primary and cultured cell types, Kazrin was visualized at plasma-membrane areas containing 4140 Journal of Cell Science 123 (23) desmosomal and intermediate filament proteins, extradesmosomal areas enriched in cortical actin, and partially within cytoplasmic and nuclear compartments (Groot et al., 2004). In accordance with these results, and in common with ARVCF-, p120-and -catenin, we find that xKazrinA localizes to three subcellular regions: cellcell borders, cytosol and the nucleus. Consistent with its nuclear localization, we resolved an NLS in the C-terminal region of xKazrinA (supplementary material Fig. S1). Interesting questions to address in future work include the nuclear function of Kazrin, possibly in a nuclear complex with ARVCF.
Initially, we addressed xARVCF and xKazrin localization to cellcell borders, because although Kazrin is reported to bind periplakin and envoplakin at peripheral desmosomal or interdesmosomal regions in human keratinocytes (Groot et al., 2004), ARVCF is present at adherens junctions via binding cadherins (Kaufmann et al., 2000;Mariner et al., 2000;Paulson et al., 2000). Our present study indicates that the xARVCF-xKazrinA complex is not likely to be a core component of either adherens or desmosomal junctions. Instead, screening of xKazrinA resolved three (2-, 1and 2-) spectrin proteins as possible binding partners. Spectrins form a cytoskeletal network with actin filaments at the plasma membrane cortex, assisting in functions including plasma membrane and junctional strengthening, restriction of select proteins to plasma membrane subdomains, transduction of nerve action potentials, and contraction of heart cells (for a review, see Baines, 2009;Bennett and Healy, 2009). In this study, we found that the xARVCF-xKazrinA complex associates with x2-spectrin (Fig. 6), suggesting a novel interaction and localization of Kazrin and ARVCF-catenin (perhaps also catenin and p0071). Previously, it was reported that the spectrin network associates with E-cadherin via ankyrin, contributing to lateral cell membrane integrity (Kizhatil et al., 2007). We, however, found neither C-cadherin nor p120 co-precipitating with 2-spectrin, despite significant homology within the cyto-domains of Xenopus E-and C-cadherin (80.5% identity). Possibly, cadherins such as Ccadherin form lesser interactions with the spectrin cytoskeleton in developmentally transitioning embryonic tissues.

Roles of the ARVCF-Kazrin complex
Earlier work predicted the involvement of Kazrin in clathrinmediated endocytosis (Schmelzl and Geli, 2002), which is known to contribute to cadherin internalization and thereby to the modulation of junction function (Ivanov et al., 2004;Le et al., 1999;Xiao et al., 2005) (for reviews, see Bryant and Stow, 2004;Fig. 11. Working model of xARVCF-xKazrinA complex localization and function at the plasma membrane. Kazrin (Kaz) facilitates formation of the trimeric ARVCF-p190B RhoGAP complex, which localizes to the plasma membrane via association of Kazrin with spectrin. This novel RhoGAP complex locally reduces RhoA activity at the plasma membrane and stabilizes the actin cytostructure, supporting cell-cell contacts. An interaction of p120-catenin with p190A RhoGAP at adherens junctions was previously reported (Wildenberg et al., 2006). D' Souza-Schorey, 2005;Delva and Kowalczyk, 2009;Yap et al., 2007). Perhaps related are reports that p120 subfamily members protect cadherins from endocytosis and lysosomal degradation, and might also chaperone cadherins to the plasma membrane (Chen et al., 2003;Davis et al., 2003;Fang et al., 2004;Gu et al., 2009;Ireton et al., 2002;Xiao et al., 2003;Xiao et al., 2005) (for reviews, see Kowalczyk and Reynolds, 2004;Reynolds and Carnahan, 2004;Xiao et al., 2007). In desmosomal regions, Kazrin colocalizes with desmoglein3 ( Fig. 5G-I), a desmosomal cadherin prone to accelerated endocytosis and degradation in the autoimmune disease pemphigus vulgaris (Calkins et al., 2006). We speculate that the role of Kazrin in endocytosis, or the role of the ARVCF-Kazrin complex, could be linked to reduced levels of cadherin and adhesion following Kazrin depletion.
If this were the case, one question would be how Kazrin, or the ARVCF-Kazrin complex, modulates cadherin endocytosis at the molecular level. Endocytosis requires cortical actin to be locally reorganized, with such changes involving the actions of small GTPases (for reviews, see Ellis and Mellor, 2000;Symons and Rusk, 2003). Multiple groups have reported Rho and Rac (and Cdc42) modulation by p120 subfamily catenins including ARVCF (Anastasiadis et al., 2000;Fang et al., 2004;Grosheva et al., 2001;Noren et al., 2000) (for a review, see Anastasiadis, 2007). Kazrin is similarly implicated in RhoA modulation (Sevilla et al., 2008a). Our results here point to negative xARVCF and xKazrin effects on RhoA activity (Fig. 8E), with the actin network becoming reorganized upon Kazrin depletion in vivo (Fig. 8F). Requiring future study, we speculate that the ARVCF-Kazrin complex modulates RhoA activity at cell contacts, having an impact on actin organization and thereby on cadherin stability and endocytosis. Interestingly, Charrasse and co-workers found that constitutively active RhoA induces M-cadherin endocytosis and degradation (following p120-catenin dissociation from M-cadherin) (Charrasse et al., 2006). Although there is a discrepancy between our results and theirs regarding p120 (Kazrin knockdown in Xenopus embryos decreased catenin protein levels, whereas RhoA activation in mouse myoblasts does not change endogenous p120 protein levels), both groups find that altered RhoA activity affects cadherin levels.
We identified p190B RhoGAP, ubiquitously expressed in mammals, as a novel binding partner of xKazrinA ( Fig. 9 and supplementary material Table S1). Because p190A RhoGAP interacts with p120-catenin (Niessen and Yap, 2006;Wildenberg et al., 2006), and Xp190B with xARVCF ( Fig. 9), p190 RhoGAP members might be core RhoA effectors of the p120 subfamily proteins (or vice versa). This view might also relate to an earlier report implicating p190 RhoGAP in RhoA inhibition upon cadherin engagement (Noren et al., 2003). Although we detected only weak endogenous association of Xp190B RhoGAP with xARVCF in vivo (Fig. 9F), their interaction might be regulated, conceivably in response to Rac1, Rnd proteins or tyrosine phophorylation by Src or Arg (Abl related gene) kinase (Brouns et al., 2001;Bustos et al., 2008;Hernandez et al., 2004;Wennerberg et al., 2003). Of note, overexpressed xARVCF partially rescues defective ectoderm integrity resulting from Kazrin depletion. Via the weak association of ARVCF with p190B, we speculate that properly titrated ARVCF overexpression might assist in recruiting p190B RhoGAP to cellcell borders, thereby restoring a more physiological range of local RhoA activity.
Although Kazrin resides in both the nucleus and cytoplasm, as well as at cell-cell borders, the nuclear functions of Kazrin are uncertain. This and a prior study indicate that Kazrin affects cortical actin function via RhoA modulation (Sevilla et al., 2008a). Speculatively, Kazrin might also modulate nuclear actin, and thereby chromosome positioning and gene expression (Carmo-Fonseca, 2007;Chuang et al., 2006;Dundr et al., 2007). Additional nuclear functions might take place through association of Kazrin with coilin, a component of cajal bodies engaged in snRNP assembly and RNA splicing (Lim et al., 2006). Kazrin is further reported to associate with the meiotic apparatus of mouse unfertilized eggs (Gallicano et al., 2005), and with intermediate filaments (Groot et al., 2004), whereas a liprin-like isoform (KazrinE) associates with stable (acetylated) microtubules (Nachat et al., 2009). Given that ARVCF might bind Kazrin in the context of varied cell structures and compartments (Kaufmann et al., 2000;Mariner et al., 2000;Waibler et al., 2001), it will be especially interesting to discern whether either protein is relevant to the potential role of the other (or that of -catenin or p0071-catenin) in gene activity.
In all cases, our work here establishes xKazrin as a binding partner of xARVCF-catenin, with further related interactions occurring with p190B RhoGAP and spectrin. In keeping with a prior study (Sevilla et al., 2008b), xKazrin was shown to be essential in vivo, with our rescue analysis extending the functional interactions of Kazrin with xARVCF, cadherin and RhoA.
Screening using full-length xKazrinA employed an adult mouse brain cDNA library in collaboration with Hybrigenics ULTImate Y2H (Rain et al., 2009). We obtained 281 positive clones that were then sequenced and analyzed.

GST pull-downs and blot overlay
For in vitro pull-down between xARVCF and xKazrinA, 1 mg of bacterially purified MBP-xARVCF or MBP-Xp120 were incubated with 1 mg of bacterially purified GST-xKazrinA in 1 ml of buffer (PBS, 1 mM EDTA, 0.1% NP-40, 100 ng/ml BSA) for 30 minutes with rocking at 4°C. Proteins were then precipitated using either glutathione-resin or amylose-resin, rinsed, and analyzed using SDS-PAGE and western blotting.
For in vitro MBP pull-down assays involving Xp190B, xARVCF and xKazrinA, 4 g of DNA encoding pCS2/HA-Xp190B was in vitro transcribed and translated using the TNT SP6 High-Yield Wheat Germ Protein Expression System (Promega), following the manufacturer's protocol. Synthesized Xp190B protein was then incubated with 1 mg of GST-xKazrinA and/or MBP-xARVCF and pulled-down using amylose-resin.
For blot overlays, purified GST-xKazrinA or GST (250 ng each) was subject to SDS-PAGE and transferred to nitrocellulose. Resulting blots were incubated in PBS with 0.1% Tween-20 and 5% non-fat dry milk overnight at 4°C to facilitate refolding of the purified proteins, and then incubated for 1 hour at room temperature with 1 mg/ml purified MBP-xARVCF, MBP-Xp120 or MBP diluted in the same buffer. Associated MBP fusion proteins were detected using anti-MBP antibody.

Xenopus embryo manipulations, coimmunoprecipitations and cell-cell dissociation assays
Induction of female X. laevis, in vitro fertilization of eggs, and embryo injections were carried out using standard methods (Sive et al., 2000). Capped mRNAs for injection were generated using mMessage mMachine kits (Ambion). Embryos were incubated at 18°C to stage10 (early gastrula), lysed in embryo IP buffer (50 mM Tris-HCl pH 7.4, 1% Triton X-100, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, 1 mM Na 3 VO 4 , 1 mM NaF, and 1 mg/ml each of Aprotinin, leupeptin and pepstatin), centrifuged, and the supernatants divided for immunoprecipitation. Immune complexes were precipitated using Protein-A/G agarose beads (Santa Cruz Biotechnology), and subjected to western blotting.
For assay of cell-cell dissociation, naive ectoderm tissue fragments (animal caps) were isolated from late blastula embryos (stages 9-10), placed in 0.6ϫ MMR solution with 2 mM EGTA and rotated at room temperature for 2-3 hours until control morpholino-injected animal caps were completely dissociated.

Cell culture
X. laevis A6 kidney epithelial cells were maintained in L15 media (Invitrogen) with 5% fetal bovine serum (Invitrogen), and kept at 22°C in a humidified incubator of normal atmosphere. Human MCF-7 breast cancer and HeLa cervical carcinoma cells were cultured in DMEM (Invitrogen), using standard mammalian cell culture conditions. A6 and MCF-7 cells were transfected using FuGene6 (Roche). Coimmunoprecipitations employed TX100 lysis buffer.
Normal human keratinocytes were isolated from neonatal foreskins and cultured in keratinocyte growth medium (KGM, Cambrex, East Rutherford, NJ), as described (Calkins et al., 2006). Keratinocytes at passage 4 or less were seeded into tissue culture vessels. Cells were switched into KGM medium containing 0.5 mM calcium 16-18 hours prior to fixation and immunostaining. Keratinocyte transient transfections were performed as described (DiColandrea et al., 2000).

Immunofluorescence
X. laevis A6 or human MCF-7 cells on coverslips were fixed 24-48 hours posttransfection in PBS with 3.7% formaldehyde for 5 minutes at room temperature. Samples were then permeabilized in PBS with 0.2% Triton X-100 for 10 minutes at room temperature, blocked for over 1 hour in antibody dilution buffer (PBS with 0.02% Triton X-100, 2% BSA and 2% normal donkey serum), and incubated for 1 hour each at room temperature with primary followed by secondary antibodies (donkey anti-rabbit-Cy3 or anti-mouse-FITC, Jackson Immunolabs). Samples were washed in blocking buffer three times and mounted on slides.
Keratinocytes on coverslips were fixed with 4% paraformaldehyde (PFA) and incubated with the indicated antibodies. Various Alexa-Fluors (Invitrogen, Carlsbad, CA) conjugated to species-specific secondary antibodies were used for dual-label immunofluorescence. Images were acquired using a Leica DMR-E fluorescence microscope, captured with a Hamamatsu Orca camera, and processed using Simple PCI software (Compix, Cranberry Township, PA). In limited cases, image brightness was adjusted using Adobe Photoshop software.
For Xenopus ectoderm confocal microscopy, in vitro transcribed mRNAs encoding Myc-xKazrinA and/or HA-xARVCF were injected into the animal pole of one-cell stage embryos. Vitelline membranes of late blastula embryos (stage 9) were manually removed, followed by a 2-hour fixation in MEMFA and methanol dehydration. Embryos were then serially re-hydrated with PBS-diluted methanol, incubated for 1 hour in blocking buffer (20% goat serum in PBST), incubated for 3 hours with primary antibody, washed for 1 hour with PBST, incubated for 3 hours with secondary antibodies and washed (3-4 times) with PBST. Blastula ectoderm tissue (animal caps) were isolated as described (Fang et al., 2004), and mounted on glass slides. Images obtained from an Olympus IX-70 inverted microscope were digitally captured using Olympus Fluoview FV500 software, and adjusted for brightness in select cases using Adobe Photoshop software.
For actin staining, naive ectoderm (animal caps) isolated at stage10 was fixed with 3.7% PFA and 0.25% glutaraldehyde in PBST for 30 minutes at room temperature. After rinsing with PBST, samples were incubated overnight at 4°C with Alexa Fluor 488-phalloidin (Molecular Probes/Invitrogen). Stained ectodermal tissues were washed with PBST, mounted and viewed under a confocal microscope.

In vitro Rho activity assay
In vitro Rho activity assays were performed as described (Gu et al., 2009). Briefly, Xenopus embryos injected with xARVCF and/or xKazrinA morpholinos were subject to lysis buffer (25 mM HEPES pH 7.5, 150 mM NaCl, 10 mg MgCl 2 , 1% NP-40, 1 mM EDTA and 2% glycerol), and lysates incubated with recombinant GST-Rhotekin RBD (Rho binding domain) immobilized to glutathione-Sepharose 4B (GE Healthcare Life Science), followed by pull-down and subsequent SDS-PAGE and western blotting detection of active Rho.