14-3-3γ-mediated transport of plakoglobin to the cell border is required for the initiation of desmosome assembly in vitro and in vivo

Desmosomes are adherens-like junctions that anchor intermediate filaments, leading to the generation of a tissue-wide intermediate filament network. Three different protein families contribute to desmosome structure and function – the desmosomal cadherins desmocollins (DSCs) and desmogleins (DSGs), the armadillo (ARM) proteins and the plakins (Green and Gaudry, 2000). Desmosome composition varies with respect to tissue type and differentiation status, as the cadherins and the associated ARM family members show tissue- and cell-type-specific expression (Bass-Zubek et al., 2009; Dusek et al., 2007), leading to changes in the organization and function of desmosomes in different tissues.

The ARM proteins participate in the regulation of desmosome assembly and cell–cell adhesion (Marcozzi et al., 1998; Palka and Green, 1997). Plakoglobin (encoded by JUP) localizes to both desmosomes and adherens junctions and is required for the initiation of desmosome formation by adherens junctions (Acehan et al., 2008; Knudsen and Wheelock, 1992; Lewis et al., 1997). Decreases in DSG3, the density of the plaque and the levels of plakophilin 1 (PKP1) at the cell border have been observed in plakoglobin-null keratinocytes (Acehan et al., 2008; Caldelari et al., 2001), suggesting that plakoglobin is required for desmosome formation and function in cultured keratinocytes. Plakoglobin-knockout mice die during embryogenesis owing to defects in desmosome formation in cardiac tissue (Ruiz et al., 1996). Some discrepancies exist in the literature regarding the effects of plakoglobin loss on desmosome formation in the epidermis. Ruiz and colleagues have reported that the epidermis of embryos at 11.5 days post coitum are normal upon loss of plakoglobin (Ruiz et al., 1996), whereas others have reported that defects in epidermal organization and desmosome function are observed in mice lacking plakoglobin at 17.5 days post coitum (Bierkamp et al., 1996). Plakoglobin has been reported to form a complex with both P-cadherin and E-cadherin, and the total levels of the classical cadherins dictate desmosome formation and organization (Lewis et al., 1997; Michels et al., 2009; Tinkle et al., 2008). These results suggest that plakoglobin and other ARM proteins might serve as a link between adherens junction formation and desmosome formation. Consistent with this hypothesis, plakoglobin and E-cadherin are independently required for the recruitment of plakophilin 3 (PKP3) to the cell border in order to initiate desmosome formation in HCT116 cells (Gosavi et al., 2011), whereas plakoglobin and the plakophilin family members collaborate with the desmoplakin N-terminus to regulate the clustering of the desmosomal cadherins at the cell surface (Chen et al., 2002). These data are indicative of plakoglobin being required for the initiation of desmosome formation and maintenance. However, the mechanisms by which the ARM proteins are transported to the cell border to initiate the process of desmosome formation remain unclear.

Spermatogenesis occurs in seminiferous tubules in the testis and is intrinsically dependent upon cell–cell adhesion between spermatocytes and Sertoli cells, and the formation of the blood–testis barrier between two Sertoli cells (Russell et al., 1990). These adhesive interactions are crucial for the progression of spermatogenesis. The blood–testis barrier comprises tight junctions, basal ectoplasmic specializations, gap junctions and desmosome-like junctions (Cheng and Mruk, 2002; Lie et al., 2011). Disruption of cell–cell adhesion perturbs the normal progression of spermatogenesis (Wong et al., 2004). Importantly, decreasing the expression of PKP2, DSG2 and DSC2 affects cell–cell adhesion, indicating that the development of spermatozoa is regulated by the formation of desmosome-like junctions in the testis (Li et al., 2009; Lie et al., 2010).

The 14-3-3 protein family is a family of small acidic proteins (Yaffe, 2002) that bind to proteins that contain a phosphorylated serine residue in a consensus motif (Muslin et al., 1996; Yaffe et al., 1997). Loss of 14-3-3ε and 14-3-3γ leads to the overriding of checkpoint function and premature entry into mitosis (Hosing et al., 2008; Telles et al., 2009). Therefore, we wanted to determine whether loss of 14-3-3γ in the mouse led to defects in checkpoint function. When we attempted to generate 14-3-3γ-knockdown mice by using a novel transgenic protocol that had been developed in our laboratory (Sehgal et al., 2011), we observed that loss of 14-3-3γ led to sterility in male mice due to a decrease in cell–cell adhesion and a defect in the transport of plakoglobin and other desmosomal proteins to the cell border in the seminiferous tubules of mice. Similar results were obtained in the human HCT116 colorectal cancer cell line. Furthermore, our results demonstrate that 14-3-3γ might load plakoglobin onto the KIF5B–KLC1 complex in order to transport plakoglobin to the cell border to initiate desmosome formation, both in HCT116 cells in culture and in the mouse testis, thus demonstrating that 14-3-3γ is required for desmosome formation.

To determine whether loss of 14-3-3γ leads to a loss of checkpoint regulation in vivo, we attempted to generate knockdown mice for 14-3-3γ by using a sperm-mediated gene transfer protocol that was developed in our laboratory (Sehgal et al., 2011). However, when mice that had been injected with viruses expressing the shRNA construct against 14-3-3γ were mated with female mice, no pups were obtained. The levels of 14-3-3γ were substantially decreased in the testis of mice that had been injected with viruses expressing the shRNA construct against 14-3-3γ (sh14-3-3γ) in comparison with the mice that had been injected with the vector control (Vec, Fig. 1A). Loss of 14-3-3γ led to an almost complete absence of mature spermatozoa in the epididymis in comparison with the control mice (Fig. 1B,C). In addition, the organization of the seminiferous tubule was severely disrupted upon 14-3-3γ knockdown, as evidenced by individual sections of the seminiferous tubule being dissociated from one another in comparison with those of control mice (Fig. 1B). Furthermore, primary germ cells and Sertoli cells were detached from the basal lamina. This did not lead to a large increase in transferase dUTP nick end labeling (TUNEL)-positive cells (supplementary material Fig. S1A). Finally, histological examination revealed an abrogation of cell–cell adhesion between Sertoli cells, and between Sertoli cells and germ cells, in the testis, upon knockdown of 14-3-3γ (Fig. 1D), which was confirmed by electron microscopy (Fig. 1E). Thus, these results suggest that loss of 14-3-3γ leads to a decrease in cell–cell adhesion in vivo.

To identify the mechanisms that lead to a decrease in cell–cell adhesion, we used a HCT116 cell line model in which 14-3-3γ had been knocked down (sh14-3-3γ) (described previously by Hosing et al., 2008). 14-3-3γ mRNA and protein levels were lower in the sh14-3-3γ cells in comparison with the control cells (Fig. 2A,B). The protein levels of 14-3-3ε and 14-3-3σ, or the mRNA levels of 14-3-3ε, 14-3-3β, 14-3-3τ and 14-3-3ζ (Fig. 2A,B) were not altered in the sh14-3-3γ cells in comparison with vector control cells (Fig. 2A). In comparison with the control cells, the sh14-3-3γ cells showed a decrease in cell–cell adhesion in hanging-drop assays (Fig. 2C,D), and cell adhesion to fibronectin and collagen IV was also diminished in these cells (supplementary material Fig. S1B), which is consistent with the detachment of cells in the testis from the basal lamina.

To determine whether the defect in cell–cell adhesion was induced specifically by the loss of 14-3-3γ, cells that had been subjected to 14-3-3ε knockdown were generated (sh14-3-3ε). Western blot analysis demonstrated that, although14-3-3ε protein levels were decreased in sh14-3-3ε cells, the levels of 14-3-3γ were unaltered. No substantial difference was observed in cell–cell adhesion in the sh14-3-3ε-knockdown cells in comparison with the control cells (Fig. 2F,G). These results suggest that the differences in cell–cell adhesion that are observed upon knockdown of 14-3-3γ are specific to the 14-3-3γ isoform.

To determine the causes of the decrease in cell–cell adhesion, the levels of adhesion proteins in the control and sh14-3-3γ cells were determined by western blot analysis or reverse transcription PCR (RT-PCR) (Fig. 3A–C). The levels of these proteins or mRNAs were not decreased upon 14-3-3γ knockdown. However, in the case of DSG2, and DSC2 and DSC3, increased protein levels were observed in the 14-3-3γ knockdown cells (Fig. 3B), although no substantial increase was observed in mRNA levels (Fig. 3C). PKP2 mRNA levels were not altered substantially in the sh14-3-3γ cells in comparison with the control cells (Fig. 3C), and our previous reports have suggested that HCT116 cells do not express PKP1 (Kundu et al., 2008). Notably, the levels of the desmosomal proteins plakoglobin, PKP3, desmoplakin, DSC2 and DSC3, DSG2 and PKP2 were significantly lower at the cell borders in sh14-3-3γ cells than in control cells (Fig. 3E,F), even though the total levels of these proteins were unaltered in the sh14-3-3γ cells. Intensity profiles for the staining are shown in supplementary material Fig. S2. Importantly, no change in the detergent solubility of the desmosomal proteins was observed in the 14-3-3γ-knockdown cells when compared with the control cells (supplementary material Fig. S4D).

By contrast, the levels of adherens junction components [E-cadherin (also known as CDH1), P-cadherin (also known as CDH3), β-catenin, p120-catenin (also known as CTNND1) and α-E-catenin], tight junction components (ZO-1) and polarity proteins (Par-3, also known as PARD3) were not reduced at the cell border in sh14-3-3γ cells (supplementary material Fig. S1D,E). Loss of 14-3-3ε did not result in a decrease in the levels of plakoglobin at the cell border (Fig. 3D). HCT116 cells that lacked both copies of 14-3-3σ (Chan et al., 1999) showed a decrease in plakoglobin levels (supplementary material Fig. S4A) and a decrease in cell–cell adhesion (supplementary material Fig. S4B). However, there was no defect in localization of plakoglobin to the border in these cells (supplementary material Fig. S4C), suggesting that 14-3-3σ is required to maintain plakoglobin protein levels, but not plakoglobin localization to the border.

Expression of a green fluorescent protein (GFP)-tagged shRNA-resistant 14-3-3γ cDNA (GFP–14-3-3γR) resulted in the recruitment of plakoglobin to the cell border in sh14-3-3γ cells in contrast with cells transfected with GFP alone (Fig. 4A). To determine whether the desmosomal proteins formed a complex with 14-3-3γ, protein extracts from HCT116 cells were incubated with either glutathione S-transferase (GST) alone or GST–14-3-3γ. 14-3-3γ formed a complex with plakoglobin, PKP3 and desmoplakin but not with DSG2 or adherens junctions proteins, such as E-cadherin (Fig. 4B). Our previous results have suggested that plakoglobin is present at the cell border in low-calcium medium, and that plakoglobin is required for the recruitment of other desmosomal proteins to the cell border in HCT116 cells and for the initiation of desmosome formation upon the addition of calcium (Gosavi et al., 2011). Therefore, to determine whether 14-3-3γ is required for the initiation of desmosome formation, calcium-switch assays were performed. Plakoglobin was present at the cell border in low-calcium medium and the levels at the border increased upon the addition of calcium in the vector control cells. DSC2 and DSC3 were not detectable at the border in the control cells in low-calcium medium. However, DSC2 and DSC3 localized to the border 60 minutes after the addition of calcium in the control cells. By contrast, in the sh14-3-3γ cells, plakoglobin, DSC2 and DSC3 were not present at the border in low-calcium medium and accumulated at significantly lower levels at the border upon the addition of calcium in comparison with the control cells (Fig. 4C). E-cadherin levels at the border were unaffected in the sh14-3-3γ cells in comparison with the control cells (Fig. 4C). These results suggest that 14-3-3γ specifically is required for the localization of plakoglobin to cell borders and is required for the initiation of desmosome formation.

We observed that, although the levels of DSC2 and DSC3 at the cell borders were lower in 14-3-3γ-knockdown cells than in the control cells, a substantial fraction of DSC2 and DSC3, nevertheless still localized to the borders in the knockdown cells. These results suggested that, to some extent, localization of DSC2 and DSC3 to the border might be independent of the presence of 14-3-3γ and plakoglobin at the cell border. We have previously shown that, in this cell system, plakoglobin is required to recruit PKP3 and desmoplakin to the cell border (Gosavi et al., 2011). To determine whether plakoglobin is required for the recruitment of desmosomal cadherins to the border, the localization of these cadherins was studied in HCT116-derived plakoglobin-knockdown cells, which have been described previously (Gosavi et al., 2011). The levels of DSC2 and DSC3, and DSG2, were not reduced at the border in the plakoglobin-knockdown cells (supplementary material Fig. S1C), suggesting that 14-3-3γ, in addition to being required for plakoglobin localization to the border, might have other functions in desmosome formation. These data are consistent with the observation that 14-3-3γ forms a complex with PKP3 and desmoplakin (Fig. 4B).

Analysis of the plakoglobin amino acid sequence led to the identification of a potential 14-3-3 binding site at serine residue 236 (S236) (supplementary material Fig. S3A) (Obenauer et al., 2003). To determine whether S236 is required for an interaction between plakoglobin and 14-3-3γ and to mediate the targeting of plakoglobin to the surface, residue S236 was replaced with alanine (S236A), and the ability of this mutant to bind to 14-3-3γ and to localize to the border was investigated. GST–14-3-3γ formed a complex with wild-type plakoglobin but not with that of the S236A mutant (Fig. 5A). In contrast with wild-type plakoglobin, which localized to the cell border, we observed reduced levels of the S236A protein at the border and an increased pan-cellular localization (Fig. 5B). Although some of the S236A mutant protein still localized to the border, the localization to the border was attenuated in comparison with that of the wild-type protein, suggesting that binding of plakoglobin to 14-3-3γ is required for the efficient localization of plakoglobin to the border.

The S236 residue is a potential site for phosphorylation by PKCμ (supplementary material Fig. S3A), suggesting that phosphorylation of plakoglobin by PKCμ is required for plakoglobin localization to the border. HCT116 cells were treated with the vehicle control dimethyl sulfoxide (DMSO), a pan PKC inhibitor that doesn't inhibit PKCμ (bisindolylmaleimide I, BisI) or an inhibitor that is specific for PKCμ and PKCα (Go6976). A GST pulldown assay was then performed using 14-3-3γ. Importantly, GST–14-3-3γ was unable to pull down plakoglobin from protein extracts that had been prepared from cells treated with Go6976, in contrast to cells that had been treated with DMSO (Fig. 5C), suggesting that the activity of either PKCα or PKCμ is required for complex formation with 14-3-3γ and localization of plakoglobin to the cell border. However, 14-3-3γ formed a complex with desmoplakin and PKP3 in extracts that had been derived from cells treated with the inhibitor (Fig. 5C), suggesting that the activity of PKCμ or PKCα is not required for the association of desmoplakin or PKP3 with 14-3-3γ. In agreement with these results, immunofluorescence analysis using antibodies against plakoglobin demonstrated that the treatment of cells with Go6976 decreased the localization of plakoglobin at the cell border in comparison with that of cells that had been treated with either DMSO or BisI (Fig. 5D).

Because the motif scan software identified S236 as a potential site for phosphorylation by PKCμ, we inhibited the expression of PKCμ using vector-driven RNA interference (RNAi) in HCT116 cells using a previously described sequence that inhibits PKCμ expression but not PKCα expression (Park et al., 2009). HCT116 cells were transfected with either the vector plasmid or a plasmid that expressed shRNA sequences that target PKCμ (shPKCμ). Forty-eight hours post transfection, the cells were transferred to medium containing puromycin in order to enrich for transfected cells. Western blot analysis demonstrated that PKCμ levels were reduced, as expected; however, plakoglobin and desmoplakin protein levels were reduced in cells that had been transfected with the shRNA against PKCμ, suggesting that PKCμ might also regulate the stability of these proteins (Fig. 5E). No change in the levels of PKP3 was observed, and western blots for actin were performed as loading controls. The levels of plakoglobin, desmoplakin and PKP3 at the cell borders were decreased in cells that had been transfected with the PKCμ shRNA in comparison with cells that had been transfected with the vector control, and the remaining protein in these cells was not localized to the border (Fig. 5F). These results suggest that, in addition to the decreased protein levels, there is a decrease in the localization of the desmosomal components to the cell borders in the absence of PKCμ. This is in contrast with the results obtained for cells that lacked 14-3-3σ, in which a decrease in the levels of plakoglobin was observed but there was no defect in plakoglobin localization to the border (supplementary material Fig. S4C). To determine whether PKCμ phosphorylated plakoglobin directly, the first 300 amino acids of plakoglobin, which comprise the putative 14-3-3-binding site, were purified from bacteria as a GST fusion protein and used as a substrate in an in vitro kinase assay using purified PKCμ. A peptide derived from CREB (catalog number C50-58, Signal Chem) was used as a positive control in these assays. As shown in Fig. 5G, the GST–PG1-300 fusion protein was phosphorylated in vitro by PKCμ, and the level of phosphorylation increased markedly with an increase in substrate concentration, in contrast with the results observed with GST alone. These results suggest that PKCμ regulates localization of plakoglobin to the cell border, and plakoglobin expression or stability.

A proteomic screen conducted in our laboratory identified the kinesin 1 family member KIF5B (Cross and Carter, 2000) as a potential ligand for 14-3-3γ (data not shown). As kinesin motor proteins are required for the transport of proteins to the cell border (Cross and Carter, 2000) and have been shown to be required for the transport of the desmosomal cadherins to the cell border (Nekrasova et al., 2011), we hypothesized that plakoglobin was transported to the border by KIF5B. To determine whether 14-3-3γ forms a complex with KIF5B, HCT116 cells were transfected with either the vector control (pcDNA3) or hemagglutinin (HA)-tagged 14-3-3γ, and immunoprecipitation reactions were performed using antibodies to the HA epitope. 14-3-3γ formed a complex with KIF5B, suggesting that 14-3-3γ might load plakoglobin onto KIF5B and lead to the transport of plakoglobin to the cell border (Fig. 6A). To test this hypothesis, KIF5B expression was stably downregulated using vector-driven RNAi. Five clones (K1–K5) were isolated that showed a decrease in KIF5B protein expression in comparison with the vector control cells (Fig. 6B). Loss of KIF5B led to a perinuclear accumulation of mitochondria, as reported previously (Tanaka et al., 1998) (Fig. 6C). Furthermore, hanging-drop assays demonstrated that loss of KIF5B led to a decrease in cell–cell adhesion in comparison with the control cells (Fig. 6D; supplementary material Fig. S3B).

To determine whether the decrease in cell–cell adhesion that was observed upon KIF5B knockdown was accompanied by a decrease in the localization of the desmosomal proteins to the cell borders, the control cells and the KIF5B knockdown clones K3 and K5 were stained with antibodies against components of desmosomes or adherens junctions. Similar to that observed in the sh14-3-3γ cells, the levels of plakoglobin, PKP3, desmoplakin, DSC2 and DSC3, and DSG2 substantially decreased at the border in K3 and K5 KIF5B-knockdown cells – a phenotype similar to that of the control cells (Fig. 6E,F). Intensity profiles for the staining are shown in supplementary material Fig. S2. By contrast, knockdown of KIF5B did not lead to a decrease in the levels of E-cadherin, p120 catenin, α-E-catenin and β-catenin at the cell border (supplementary material Fig. S3C,F,G). The decrease in the levels of desmosomal proteins at the border was not due to a decrease in protein expression levels, as indicated by western blot analysis. In contrast with the results that were observed for plakoglobin, PKP3 and desmoplakin, a large increase in the levels of DSC2 and DSC3, and DSG2 were observed in the kinesin-knockdown cells (Fig. 6F), although this was not observed at the mRNA level (supplementary material Fig. S3E). In cells that had been fixed with methanol, we observed low levels of DSC2 and DSC3, and DSG2, in the cytoplasm (Fig. 6E); however, fixation with paraformaldehyde revealed high levels of DSC2 and DSC3, and DSG2, in the cytoplasm (supplementary material Fig. S4E), suggesting that these proteins accumulate in the cytoplasm upon loss of KIF5B. PKP2 mRNA levels were not altered upon knockdown of kinesin (supplementary material Fig. S3E).

Because kinesin motor proteins transport their cargo on microtubules, we investigated whether disruption of the microtubule network would lead to a decrease in the concentration of plakoglobin at the cell border. Treatment with nocodazole, but not the vehicle control, did indeed lead to a decrease in the levels of plakoglobin at the border (supplementary material Fig. S3D). To then determine whether complex formation between KIF5B and plakoglobin is dependent upon PKCμ, protein extracts from Go6976- or DMSO-treated HCT116 cells were incubated with GST–14-3-3γ. GST–14-3-3γ formed a complex with KIF5B under both conditions but did not interact with plakoglobin in protein extracts from cells that had been treated with Go6976 (Fig. 5C), indicating a requirement for active PKCμ for the association between plakoglobin and 14-3-3γ but not for the association between KIF5B and 14-3-3γ.

To further confirm the role of KIF5B in transporting plakoglobin to the border, we determined whether a dominant-negative KIF5B construct could inhibit transport of plakoglobin to the cell border. The dominant-negative kinesin heavy chain constructs used here have a point mutation in the ATPase domain (see Materials and Methods), which results in an inability of the proteins to move along the microtubules but preserves the ability of kinesins to bind to cargo (Cross and Carter, 2000). Yellow fluorescent protein (YFP)-tagged versions of wild-type and dominant-negative (mutation T92N) KIF5B, or the GFP-tagged wild-type and dominant-negative (mutation T107N) kinesin 2 family member KIF3A (Cross and Carter, 2000) constructs were transfected into HCT116 cells. After transfection, the cells were stained with antibodies against plakoglobin and visualized by using confocal microscopy. Cells that expressed wild-type YFP–KIF5B showed border staining for plakoglobin. By contrast, plakoglobin was not localized at the cell border in cells that expressed dominant-negative YFP–KIF5B (Fig. 7A). Cells that had been transfected with either of the KIF3A constructs showed border staining for plakoglobin. Cells that had been transfected with either of the dominant-negative constructs had round edges and showed a morphology that was different from that of cells that had been transfected with either of the wild-type constructs, presumably because overexpression of the dominant-negative constructs affects other cellular processes, such as microtubule organization (Silver and Harrison, 2011). However, despite the change in morphology, only the cells that had been transfected with the dominant-negative KIF5B construct, and not those transfected with the dominant-negative KIF3A construct, showed a disruption of plakoglobin localization (Fig. 7A). These results suggest that KIF5B specifically is required for transport of plakoglobin to the border.

Kinesin motor proteins are heterodimers that consist of two heavy chains and two light chains (Cross and Carter, 2000). KIF5B associates with two different kinesin light chains (KLCs) – KLC1 and KLC2 (Verhey and Hammond, 2009). To determine which of these is required for plakoglobin transport to the cell border, bacterially produced GST-tagged KLC1 and KLC2 were incubated with protein extracts that had been prepared from HCT116 cells, and the reactions were resolved on SDS-PAGE gels followed by western blotting with antibodies against plakoglobin, KIF5B and 14-3-3γ. Both KLC1 and KLC2 formed a complex with KIF5B; however, only KLC1 could form a complex with both plakoglobin and 14-3-3γ (Fig. 7B). KLCs contain a coiled-coiled domain, which is required for complex formation with the kinesin heavy chain, and a cargo-binding domain [which comprises tetratricopeptide repeats (TPRs)] (Verhey and Hammond, 2009) (Fig. 7C). To determine whether the coiled-coiled domain of KLC1 could form a complex with plakoglobin, protein extracts from HCT116 cells were incubated with either GST–KLC1 or the GST-tagged coiled-coiled domain of KLC1. GST–KLC1 could form a complex with KIF5B, plakoglobin and 14-3-3γ, whereas the coiled-coiled domain bound to KIF5B but not to plakoglobin and 14-3-3γ (Fig. 7D). Therefore, the coiled-coiled domain of KLC1 might be acting as a dominant-negative mutant because it should bind to the heavy chain but fail to form a complex with plakoglobin and 14-3-3γ. To test this hypothesis, HCT116 cells were transfected with either GFP–KLC1 or the GFP-tagged KLC1 coiled-coiled domain only and stained with antibodies against plakoglobin. Overexpression of the KLC1 coiled-coiled domain disrupted the transport of plakoglobin to the cell border, whereas cells that expressed the wild-type protein did not show any alteration in plakoglobin localization (Fig. 7E). These results suggest that the KIF5B–KLC1 complex is required for the transport of plakoglobin to the cell border.

The results described above suggest that 14-3-3γ and KIF5B are required for desmosome formation in epithelial cells. To determine whether loss of KIF5B leads to a decrease in cell–cell adhesion in the seminiferous epithelium, KIF5B expression was inhibited in the testis, as described previously (Sehgal et al., 2011). Loss of KIF5B in the testis led to a phenotype similar to that observed for the loss of 14-3-3γ (Fig. 8A–C), suggesting that both KIF5B and 14-3-3γ are independently required to regulate cell–cell adhesion in the testis. Importantly, an immunohistochemical analysis demonstrated that loss of either KIF5B or 14-3-3γ did not lead to a decrease in the levels of the other protein (Fig. 8B). Loss of 14-3-3ε did not lead to a decrease in cell–cell adhesion and spermatogenesis, suggesting that the effects observed upon knockdown of 14-3-3γ are specific to 14-3-3γ (Fig. 8A,B), a result that is consistent with those observed in HCT116 cells (Fig. 2). In addition, loss of either 14-3-3γ or KIF5B led to a detachment of the cells from the basal lamina. To determine whether the detachment of the primary germ cells and the Sertoli cells from the basal lamina led to an increase in cell death, the testis sections were stained using a TUNEL staining kit. As shown in supplementary material Fig. S1A, treatment of the positive-control sections with DNase resulted in a strong positive signal in the TUNEL assay. Testis sections from mice that had been injected with the 14-3-3γ-knockdown construct showed low levels of TUNEL positivity in comparison with testis sections from the control or KIF5B-knockdown mice. Given that the testis morphology of the KIF5B-knockdown mice and the 14-3-3γ-knockdown mice was very similar, it is likely that the loss of cell–matrix adhesion does not lead to cell death, which is consistent with our results in the cell line model. These results suggest that cell–cell adhesion in the testis requires both 14-3-3γ and KIF5B and that loss of either protein leads to defects in cell–cell adhesion.

To determine whether desmosome organization was altered in the 14-3-3γ- and KIF5B-knockdown testis, testis sections were stained with antibodies against plakoglobin, PKP3, DSC2 and DSC3, N-cadherin and E-cadherin. Plakoglobin, PKP3 and DSC2 and DSC3 localized to the border in testis that had been injected with the control virus; however, the levels of these proteins at the cell border were greatly diminished in the 14-3-3γ- and KIF5B-knockdown testis (Fig. 8D). By contrast, there was no change in E-cadherin localization in the 14-3-3γ- and KIF5B-knockdown testis in comparison with testis sections that had been injected with the vector control, a phenotype similar to that observed in HCT116 cells in culture. In contrast with the results obtained for E-cadherin, it was observed that N-cadherin localized to the border in the vector control and 14-3-3γ-knockdown testis but not in the KIF5B-knockdown testis (Fig. 8D). These results are consistent with our observations that the KIF5B-knockdown testis showed a more severe adhesion phenotype than the 14-3-3γ-knockdown testis and that KIF5B might be required for the transport of other cell–cell adhesion molecules to the border, in addition to desmosomal proteins. Overall, our results suggest that 14-3-3γ and KIF5B are required for the formation of desmosomes in vivo.

Our results suggest that 14-3-3γ and the KIF5B–KLC1 complex are required for the transport of plakoglobin to cell borders in human cell lines and in the mouse testis, and disruption of this interaction with 14-3-3γ leads to a defect in the localization of plakoglobin. In addition, loss of 14-3-3γ leads to a disruption of cell–matrix adhesion, which might affect cell–cell adhesion in vitro and in vivo. The association of plakoglobin with 14-3-3γ, and the transport of plakoglobin to the border is dependent on PKCμ activity; thus, loss of either 14-3-3γ or KIF5B inhibits plakoglobin transport, resulting in a decrease in desmosome formation and cell–cell adhesion in HCT116 cells and the testis, leading to male sterility. The decrease in the levels of N-cadherin at cell borders in KIF5B-knockdown testis resulted in a more drastic phenotype in these animals than in those that had been subjected to knockdown of 14-3-3γ, suggesting that N-cadherin is required for cell–cell adhesion in the testis as previously reported (Andersson et al., 1994; Lee et al., 2003).

Previous results have shown that loss of plakoglobin leads to the depletion of desmosomal proteins from the cell border and to defects in desmosome formation (Acehan et al., 2008; Gosavi et al., 2011; Knudsen and Wheelock, 1992; Lewis et al., 1997). Although the presence of a classical cadherin seems to be required for plakoglobin recruitment to the cell border (Michels et al., 2009; Tinkle et al., 2008), the mechanisms by which plakoglobin is transported have not been identified. The results presented here suggest that phosphorylation of plakoglobin by PKCμ at residue S236 leads to the generation of a binding site for 14-3-3γ, and that 14-3-3γ is required for the transport of plakoglobin to the cell border in order to initiate desmosome formation, presumably in a complex with a classical cadherin (Michels et al., 2009; Tinkle et al., 2008). 14-3-3γ might be required to load plakoglobin onto the KIF5B–KLC1 complex, which precedes the transport of plakoglobin on microtubules (Fig. 8E). This is consistent with previous observations that showed that PKCμ localizes to the Golgi complex (Hausser et al., 2002; Prestle et al., 1996) and is required for the fission of vesicles carrying cargo to the cell border (Liljedahl et al., 2001). A 14-3-3γ dimer is required for carrier formation at the Golgi complex along with PKCμ (Valente et al., 2012), suggesting that the loss of 14-3-3γ could disrupt desmosome formation owing to defects in plakoglobin transport.

The treatment of cells with an inhibitor that inactivates both PKCα and PKCμ led to a decrease in the localization of plakoglobin at cell borders and abolished complex formation between plakoglobin and 14-3-3γ. However, treatment with the inhibitor does not inhibit the interaction between 14-3-3γ and other desmosomal proteins or KIF5B. These results suggest that phosphorylation of plakoglobin by PKCμ is required for complex formation between plakoglobin and 14-3-3γ and the loading of plakoglobin onto KIF5B, which is essential for the transport of plakoglobin. Alternatively, it is possible that plakoglobin, desmoplakin and PKP3 are transported independently to the border and that the absence of plakoglobin at the border prevents the formation of a functional desmosome. Furthermore, it was observed that loss of plakoglobin did not affect the localization of DSC2 and DSC3, and DSG2 to cell borders, unlike the loss of 14-3-3γ or KIF5B. This might be because 14-3-3γ is required for the localization of PKP3, desmoplakin and plakoglobin; therefore, the defects in the localization of the desmosomal cadherins that is observed upon 14-3-3γ loss are not observed upon plakoglobin loss. Plakoglobin is probably required to stabilize and maintain the organization of the desmosome, which is why loss of plakoglobin leads to a decrease in cell–cell adhesion in these cells as previously reported (Gosavi et al., 2011). PKCμ was also shown to directly phosphorylate a peptide comprising the first 300 amino acids in plakoglobin, which contains the S236 residue, consistent with the notion that S236 serves as site for phosphorylation by PKCμ. This does not exclude the possibility that PKCμ phosphorylates other residues in plakoglobin and might provide an explanation for the observation that plakoglobin protein levels decreased when PKCμ expression was inhibited. It is possible that PKCμ regulates other aspects of plakoglobin function that are not limited to the transport of plakoglobin to the cell border, such as the retention of plakoglobin at the border or the stability of the plakoglobin protein.

In the absence of KIF5B, or upon expression of a dominant-negative KIF5B construct, plakoglobin does not localize to the cell border, leading to a decrease in the recruitment of other components of the desmosome, as previously reported (Gosavi et al., 2011; Lewis et al., 1997). Previous experiments have suggested that inhibiting KIF5B expression does not affect the localization of plakoglobin to the border in SCC9 cells (Nekrasova et al., 2011). Our results, however, suggest that loss of KIF5B in HCT116 cells and the testis results in an inhibition of plakoglobin transport to the border. Consistent with the results reported by Nekrasova and colleagues (Nekrasova et al., 2011), we did not observe any defects in plakoglobin localization upon expression of a dominant-negative KIF3A mutant. The discrepancy in these two reports might be due to the use of different cell types; other kinesin-family members could regulate plakoglobin transport in SCC9 cells in the absence of KIF5B. Similarly, Pasdar and colleagues have reported that a microtubule network is not essential for desmosome formation (Pasdar et al., 1991); however, the results from this report and others suggest that the transport of desmosomal components to the border is dependent on an intact microtubule network (Gloushankova et al., 2003; Nekrasova et al., 2011). This is consistent with reports that suggest that desmosome organization, function and composition vary in different cell types (reviewed in Cross and Carter, 2000; Garrod and Chidgey, 2007; Getsios et al., 2004; Green and Gaudry, 2000; Hatzfeld, 2007) and with the observations that loss of the different desmosomal components in the mouse leads to a vast variety of phenotypes (Chidgey et al., 2001; Gallicano et al., 1998; Grossmann et al., 2004; Koch et al., 1997; Lechler and Fuchs, 2007; Ruiz et al., 1996; Sklyarova et al., 2008; Vasioukhin et al., 2001).

Although 14-3-3γ seems to be required for the transport of plakoglobin, it might also be required for the transport of PKP3 and desmoplakin to the cell border in a manner that is independent of plakoglobin transport. This is consistent with our observation that inhibition of both PKCα and PKCμ did not lead to the disruption of the interaction between 14-3-3γ and desmoplakin or PKP3. The transport of DSG2 to the border is dependent on KIF5B, whereas the transport of DSC2 requires KIF3A and PKP2 (Nekrasova et al., 2011). Our results suggest that plakoglobin transport to the border is dependent on KIF5B, but not KIF3A, suggesting that the transport of desmosomal cadherins and plaque proteins to the border occurs through independent microtubule-dependent pathways. Alternatively, because 14-3-3 proteins bind to their ligands as dimers, with each member of the dimer forming a complex with a phospho-peptide (Brunet et al., 2002; Yaffe et al., 1997), it is possible that 14-3-3γ bridges interactions between the different desmosomal plaque proteins, thereby allowing the formation of an intact desmosome, as previously postulated by Bonné and colleagues (Bonné et al., 2003). The only argument against this hypothesis is that, according to our study, 14-3-3γ does not localize to the desmosome (Fig. 4A). In fact, the lack of any colocalization between 14-3-3γ and plakoglobin suggests that any function that 14-3-3γ performs, with respect to plakoglobin localization, is transient in nature; therefore, we favor the hypothesis that 14-3-3γ is required for loading plakoglobin onto the KIF5B–KLC1 complex.

In contrast with the results reported here, previous work has suggested that 14-3-3γ-knockout mice are viable, and no adhesion or sterility defects have been reported in these mice (Steinacker et al., 2005). Plakoglobin-knockout mice die during embryogenesis owing to cardiac defects arising as a result of decreased desmosome formation, and, as these mice die before the testis is formed, no information is available on the effects of loss of plakoglobin in the testis (Ruiz et al., 1996). Taken together, these previously published reports suggest that the functions of plakoglobin that are required for desmosome formation are not altered in the 14-3-3γ^−/− mice. Our results indicate that 14-3-3γ regulates desmosome formation in multiple cell types, as loss of 14-3-3γ in HCT116 cells, which are derived from the colon, and in the seminiferous epithelium leads to a decrease in cell–cell adhesion and in desmosome formation. Our results are also consistent with the previously reported role of 14-3-3γ in the transport of proteins from the Golgi complex to the cell border (Valente et al., 2012). One reason for the differences in these results and those reported by Steinacker and colleagues (Steinacker et al., 2005) could be that another 14-3-3 family member binds to plakoglobin and stimulates desmosome formation in the 14-3-3γ^−/− mice, and that this compensation is not observed upon shRNA-mediated knockdown in the testis. Another possibility is that these are strain-specific variations that are due to differences in the genetic background of the mice used in the two studies. The generation of an inducible knockdown of 14-3-3γ, in either the Swiss mice used in this study or in other mouse strains, could help determine whether loss of 14-3-3γ leads to defects in desmosome formation and cell–cell adhesion in other tissues and not just in the testis.

The results described above point to the following model. 14-3-3γ binds to plakoglobin that has been phosphorylated at residue S236 by PKCμ and loads plakoglobin onto the KIF5B–KLC1 complex for transport to the cell border (Fig. 8E). Loss of either 14-3-3γ or KIF5B, or the dominant-negative versions of KIF5B and KLC1, inhibit the transport of plakoglobin to the border. Because 14-3-3γ forms a complex with PKP3 and desmoplakin, loss of 14-3-3γ might also lead to defects in transport of these proteins to the border. As loss of plakoglobin does not substantially affect the localization of the cadherins to the border, it is possible that loss of plakoglobin leads to a defect in cadherin retention or the formation of an intact desmosome at the border in the absence of 14-3-3γ. Furthermore, loss of KIF5B in the testis led to sterility and a decrease in desmosome formation. This is consistent with previously reported results that showed that disruption of cell–cell adhesion (Cheng and Mruk, 2002) and desmosome-like junction formation in the testis leads to an increase in sterility (Li et al., 2009; Lie et al., 2010).

To conclude, 14-3-3γ and the KIF5B–KLC1 complex are required for regulating the transport of plakoglobin to the cell border. A decrease in 14-3-3γ levels leads to a decrease in desmosome formation and the recruitment of other desmosomal proteins to the border. 14-3-3γ might be required to maintain cell–cell adhesion in multiple tissues; however, this can only be confirmed by additional experiments in vivo.

Swiss mice Crl:CFW(SW) were bred and maintained in the laboratory animal facility of the Advanced Centre for Treatment Research and Education in Cancer (ACTREC). Protocols for the experiments were approved by the Institutional Animal Ethics Committee of ACTREC. The animal study proposal number is 11/2008 dated August 19, 2008. The testis injections were performed as previously described (Sehgal et al., 2011).

Details of the oligonucleotides used in this study are available in supplementary material Table S1. The GST–14-3-3γ, HA-epitope-tagged 14-3-3γ and shRNA-resistant GFP–14-3-3γ constructs have been previously described (Hosing et al., 2008). Site-directed mutagenesis (Stratagene) was used to generate the MYC–PG-S236A construct. The full-length KLC1 and KLC2 cDNA (Rahman et al., 1998), KLC-1 deletion (GST-KLC1 WT, GST-KLC1-CC and GST-KLC1-TPR) (Aoyama et al., 2009), GFP–KIF3A (Haraguchi et al., 2006), wild-type YFP–KIF5B (Gu et al., 2006), dominant-negative kinesin (GFP–KIF3A-T107N and YFP–KIF5B-T92N) (Wiesner et al., 2010), wild-type GFP–KLC1 and the GFP-tagged KLC1 coiled-coiled domain (Araki et al., 2007) constructs have been described previously. To generate the shRNA constructs against KIF5B and PKCμ, oligonucleotide pairs (supplementary material Table S1) were ligated into the pLKO.1 Puro or pLKO.1 EGFP-f Puro vectors as described previously (Sehgal et al., 2011). GST–PG1-300 was generated by amplifying the first 300 amino acids of plakoglobin (supplementary material Table S1) and cloning the fragment into the pGEX4T1 vector (Amersham).

The HCT116 (American Type Culture Collection) and the HCT116-derived stable cell lines were cultured as described previously (Hosing et al., 2008). To generate the KIF5B-knockdown clones in the HCT116 cell line, the cells were transfected with 1 µg of the shRNA constructs that targeted human KIF5B. Sixty hours post transfection, the cells were transferred to medium that contained 0.5 µg/ml of puromycin (Sigma) to generate single-cell clones. The clones K3 and K5 were used for the indicated experiments. The HCT116-derived plakoglobin-knockdown clones have been described previously (Gosavi et al., 2011).

The antibodies against PKP3, both DSC2 and DSC3, DSG2, plakoglobin, desmoplakin, K8 (keratin 8), actin, E-cadherin, β-catenin and α-E-catenin were used in western blots as previously described (Gosavi et al., 2011; Khapare et al., 2012; Kundu et al., 2008). Tissue culture supernatants of the antibodies against HA (12CA5), 14-3-3γ (CG31) and 14-3-3σ (CS112) were used at a dilution of 1∶50. The antibodies against 14-3-3ε (T-16, Santa Cruz, dilution of 1∶2000), p120 catenin (mouse monoclonal from BD Transductions, catalog number 610134, dilution of 1∶1000) and PKCμ (rabbit monoclonal obtained from Abcam, catalog number 3146-1, dilution of 1∶500) were used for western blot analysis. The secondary antibodies against mouse and rabbit IgGs were conjugated to horseradish peroxidase and used at a dilution of 1∶1000 (Invitrogen) and 1∶5000 (Pierce), respectively.

The cells were cultured on chromic-acid-treated, poly-L-lysine-coated glass coverslips at a confluence of 70–80%. Before fixation, the cells were carefully washed twice with 1× PBS. HCT116-derived clones were fixed in absolute methanol for 10 minutes at −20°C to detect α-tubulin, KIF5B, PKP2, Par3, ZO1, P-cadherin, desmoplakin, plakoglobin, DSC2 and DSC3, DSG2, E-cadherin and PKP3. In some experiments, cells were fixed in 4% paraformaldehyde and permeabilized with Triton X-100 as described previously (Gosavi et al., 2011). The antibodies against PKP3, both DSC2 and DSC3, DSG2, plakoglobin, desmoplakin, K8, actin, E-cadherin, β-catenin, ZO-1 and α-E-catenin were used in immunofluorescence analysis as described previously (Gosavi et al., 2011; Khapare et al., 2012; Kundu et al., 2008). Antibodies against PKP2 (BD Clontech, dilution 1∶25), KIF5B (Abcam, dilution 1∶100), α-tubulin (Abcam, dilution 1∶150), Par3 (Millipore, dilution 1∶50), ZO-1 (Abcam, dilution 1∶100), P-cadherin (BD Transduction Laboratories, dilution 1∶100), HA (12CA5, supernatant), p120 catenin (BD Transductions, dilution 1∶100), N-cadherin (Life Technologies, catalog number 33-3900, dilution 1∶50), α-E-catenin (Santa Cruz Biotechnology, dilution 1∶25) and E-cadherin (clone 36/E-cadherin, mouse monoclonal, BD Transduction Laboratories, dilution 1∶100) were incubated with the cells for 1 hour at room temperature at the indicated dilutions as described previously (Gosavi et al., 2011). To stain mitochondria, Mitotracker Green FM (Invitrogen) was used at a concentration of 100 nM to stain live cells. Confocal images were obtained by using a LSM 510 Meta Carl Zeiss confocal system with argon 488-nm and HeNe 543-nm lasers. All images were obtained by using an Axio Observer Z.1 microscope (numerical aperture 1.4) at a magnification of ×630 (×63 objective and ×10 eyepiece) with a ×2 or ×4 optical zoom. The surface intensity of staining was measured for the different proteins in a minimum of 24 cells using the Axiovision software, and the mean and standard deviation were plotted.

Hanging-drop assays were used to measure cell adhesion as described previously (Kundu et al., 2008).

These assays were performed as described previously (Dalal et al., 1999).

GST alone or GST–PG1-300 was purified from bacteria as described previously (Dalal et al., 2004). The purified proteins were used in kinase assays with recombinant PKCμ and a peptide that had been derived from CREB as a positive control (Signal Chem). Kinase activity was determined by using the ADP glow assay kit (Promega) according to the manufacturer's instructions.

Mouse testes were fixed in 10% formaldehyde overnight and processed for histology as described previously (Kundu et al., 2008). TUNEL assays were performed as per the manufacturer's instructions (Promega).

Wild-type and 14-3-3γ-knockdown testes were fixed with 3% glutaraldehyde and post-fixed with 1% osmium tetraoxide (Tedpella). Grids were contrasted with alcoholic uranyl acetate for 1 minute and lead citrate for 30 seconds. The grids were observed under a Carl Zeiss LIBRA120 EFTEM transmission electron microscope at an accelerating voltage of 120 kV and at ×325,000 magnification. Images were captured using a Slow Scan CCD camera (TRS, Germany).

All P-values were generated using a Student's t-test.

We thank Young-hoon Lee (Korea Advanced Institute of Science and Technology, Daejeon, Korea), Stefan Linder (Institute of Medical Microbiology, Hamburg, Germany), Tetsu Akiyoma (University of Tokyo, Tokyo, Japan), Chen Gu (Ohio State University, Columbus, OH) and Lawrence Goldstein (University of California San Diego, San Diego, CA) for supplying us with constructs that were used during the course of this study. We would also like to thank the ACTREC imaging facility for help with confocal microscopy, the ACTREC animal facility, and Sharda Sawant for helping with the preparation of the grids for electron microscopy.