Regulation of epithelial cell polarity by PAR-3 depends on Girdin transcription and Girdin–Gαi3 signaling

In multicellular organisms, epithelial cell layers form the boundaries between internal and external environments to regulate homeostasis and physiological functions (Yeaman et al., 1999). During epithelial morphogenesis, epithelial cells are arranged into cysts or tubular structures enclosing a central lumen (Bryant and Mostov, 2008; Lubarsky and Krasnow, 2003), which forms cell layers with apicobasal polarity, resulting in functionally distinct apical and basolateral membrane domains separated by junctional complexes such as tight junctions (Farquhar and Palade, 1963; Shin et al., 2006). Apicobasal polarity is crucial for a variety of biological processes, including asymmetric cell division and tissue morphogenesis, and its dysfunction contributes to various diseases including cancer (Feigin and Muthuswamy, 2009; Hirose et al., 2009; Knoblich, 2010). Thus, uncovering the regulatory mechanisms of apicobasal polarity should provide greater insight into fundamental aspects of cell and developmental biology and tumorigenesis.

Studies over the past decade have revealed that an evolutionarily conserved PAR–aPKC complex, comprising PAR-3 (also known as PARD3 in mammals), atypical protein kinase C (aPKC) and PAR-6 (note that both aPKC and PAR-6 have several isoforms in mammals), plays a pivotal role in regulating cell polarity (Ebnet et al., 2008; Goldstein and Macara, 2007; Suzuki and Ohno, 2006). In Drosophila melanogaster, cooperation of PAR-3 (which is known as Bazooka) with aPKC is indispensable for the establishment of apicobasal membrane identities (Bilder et al., 2003; Wodarz et al., 2000, 1999). Moreover, genetic manipulation in mammals, including the targeted disruption of PAR-3 in mice, has also revealed the important in vivo role of PAR-3 in tight junction formation, apical membrane development and lumen formation during epithelial morphogenesis (Chen and Macara, 2005; Hirose et al., 2006; Horikoshi et al., 2009; Hurd et al., 2003). PAR-3 interacts with the aPKC–PAR-6 complex to form the tripartite PAR-3–aPKC–PAR-6 complex (Izumi et al., 1998; Joberty et al., 2000; Lin et al., 2000; Suzuki et al., 2001), which is required for aPKC recruitment to cell–cell contacts and for apical domain development (Horikoshi et al., 2009). In addition, by binding to PAR-3, the LIMK2 and Tiam1 actin cytoskeleton regulatory proteins and the Sec 8 exocyst protein (also known as EXOC4) have been implicated in epithelial cell polarization (Bryant et al., 2010; Chen and Macara, 2005, 2006; Zuo et al., 2009). Moreover, the Lgl tumor suppressor protein (for which there are two isoforms in mammals, LLGL1 and LLGL2) negatively regulates the tripartite PAR-3–aPKC–PAR-6 complex by competing with PAR-3 for binding to the aPKC–PAR-6 complex (Chalmers et al., 2005; Hutterer et al., 2004; Yamanaka et al., 2006, 2003). These results show that the PAR–aPKC complex has a central role in establishing epithelial cell polarity through coordinating regulatory proteins of the cytoskeletal and vesicular transport systems. However, the molecular events downstream of the PAR–aPKC complex are not yet fully understood.

Heterotrimeric G proteins, composed of α, β and γ subunits, are molecular switches that control a wide array of signaling pathways (Simon et al., 1991). In the canonical G protein signaling cycle, G proteins bind to G-protein-coupled receptors (GPCRs), and receptor activation by extracellular ligands induces GDP-GTP exchange on the Gα, catalyzed by GEFs. The GTP-bound Gα then dissociates from Gβγ, and both function as signaling modules to regulate various effectors. This cycle is terminated by GTP hydrolysis of Gα, followed by the reassociation of GDP-bound Gα with Gβγ. Recent genetic studies in Caenorhabditis elegans embryos and Drosophila neuroblasts have shown that, in addition to canonical GPCR signaling, the Gi class of Gα subunits (Gαi and Gαo) plays an important role in regulating cell polarity in a receptor-independent manner (Bellaiche and Gotta, 2005; Siderovski and Willard, 2005). For example, C. elegans Gα subunits [GPA-16 (a Gαi) and GOA-1 (a Gαo)] are required for formation of the anterior–posterior axis downstream of PAR-3 polarity cues (Colombo et al., 2003; Gotta and Ahringer, 2001; Gotta et al., 2003). In mammals, pharmacological Gα activation and overexpression of constitutively active Gαi enhances tight junction biogenesis (Denker et al., 1996; Saha et al., 1998). These data suggest that GEF-mediated Gαi signaling is involved in establishing apicobasal polarity. However, the mechanism linking the PAR–aPKC complex to Gαi still remains to be elucidated.

Girdin (also designated GIV, CCDC88A, APE, HkRP1 and KIAA1212) was originally identified based on its ability to interact with the Gαi3 heterotrimeric G protein (also known as GNAI3), the Akt/PKB serine/threonine kinase and the dynamin GTPase (Anai et al., 2005; Enomoto et al., 2005; Le-Niculescu et al., 2005; Simpson et al., 2005). Subsequent studies have revealed that Girdin is a nonreceptor GEF for Gαi3 that plays a crucial role in cell migration through regulating actin remodeling (Garcia-Marcos et al., 2010, 2009; Ghosh et al., 2008). Moreover, recent studies have shown that Girdin overexpression promotes breast cancer metastasis (Jiang et al., 2008; Ling et al., 2011), and patients with colon tumors overexpressing Girdin tend to have shorter metastasis-free survival times (Garcia-Marcos et al., 2011). However, little is known about the mechanisms regulating Girdin expression and how Girdin–Gαi3 signaling determines the apicobasal polarity of epithelial cells.

In this study, we uncovered a new signaling pathway that functions downstream of PAR-3 – that PAR-3 cooperates with the AP-2 transcription factor to induce Girdin transcription. Furthermore, PAR-3 interacts with Girdin, and Girdin–Gαi3 signaling promotes apicobasal polarity by inducing tight junction formation and apical domain development.

Madin–Darby canine kidney (MDCK) epithelial cells have been widely used as a model to investigate apicobasal polarity in cell monolayers (Gonzalez-Mariscal et al., 1985) and in three-dimensional (3D) cultures of polarized cysts that model many processes involved in tube formation (O'Brien et al., 2002). Studies using this system have uncovered the molecular basis of the opposing actions of PAR-3 and Lgl in epithelial cell polarity regulation (Horikoshi et al., 2009; Yamanaka et al., 2006, 2003). To gain further insight into events downstream of the PAR–aPKC complex, we searched for genes that are reciprocally regulated by PAR-3 and Lgl. Through comparing microarray-based gene expression profiles of control and PAR-3-knockdown (kd) cells, or of control and Lgl kd cells, and subsequent confirmation by quantitative real-time PCR (qPCR), we identified Girdin as a gene reciprocally regulated by PAR-3 and Lgl (Fig. 1).

qPCR analysis confirmed that PAR-3 depletion dramatically reduced the level of Girdin mRNA, but not of LIMK2 mRNA (Fig. 1A). Immunoprecipitation using an anti-Girdin antibody showed that, compared with control, Girdin protein expression is much lower in two stable cell lines in which PAR-3 was knocked down by different short hairpin RNA (shRNA)-targeting sequences (Fig. 1B, supplementary material Fig. S1A). However, there were no effects on aPKCλ or LIMK2 protein expression, or on the expression of protein that are components of the tight and adherens junctions. Overexpression of EGFP-tagged shRNA-resistant wild-type (WT) PAR-3 (PAR-3–EGFP), but not of EGFP alone, restored Girdin mRNA and protein levels in PAR-3 kd cells (Fig. 1A,B), confirming the specific inhibitory effect of PAR-3 knockdown on Girdin mRNA expression. PAR-3–EGFP overexpression also rescued cell polarity defects in tight junction formation and apical domain development in PAR-3 kd cells (supplementary material Fig. S1B). These results suggest that PAR-3 specifically regulates Girdin mRNA expression.

A previous study has reported that the human PARD3 gene (encoding PAR-3) is mutationally inactivated in LNCaP prostate cancer cells (Kunnev et al., 2009). We therefore asked whether alterations in Girdin expression are common in prostate cancer cells. qPCR and immunoblot analysis showed that the levels of both Girdin mRNA and protein were drastically reduced in PAR-3-deficient LNCaP cells compared with PC-3 and DU145 cells, which express PAR-3 (supplementary material Fig. S1C,D). This result suggests that PAR-3 regulation of Girdin mRNA expression is not restricted to MDCK cells.

aPKC binding to PAR-3 is essential for PAR-3 rescue of apical membrane development in PAR-3 kd cells (Horikoshi et al., 2009). To examine whether aPKC also regulates Girdin mRNA expression, we investigated the effect of depleting aPKCλ, an aPKC predominantly expressed in MDCK cells, on Girdin expression. We confirmed that aPKCλ depletion causes aberrant tight junction formation and apical domain development using a Ca²⁺ switch assay (supplementary material Fig. S1E).

qPCR and immunoprecipitation analysis using an anti-Girdin antibody revealed that the levels of both Girdin mRNA and protein are reduced in aPKCλ kd cells (Fig. 1C,D). Consistent with a previous report, PAR-6β was also downregulated in aPKCλ kd cells (Imai et al., 2006). In contrast, aPKCλ depletion did not affect LIMK2 mRNA and protein levels (Fig. 1C,D). These results suggest that aPKCλ is also involved in regulating Girdin mRNA expression.

The activity of the PAR-3–aPKC–PAR-6 complex is suppressed by Lgl, which competes with PAR-3 for binding to the aPKC–PAR-6 complex (Yamanaka et al., 2006, 2003). We performed an immunoprecipitation analysis using anti-PAR-3 antibody to evaluate the association between PAR-3 and aPKC, and show that more aPKCλ could be coprecipitated with PAR-3 from MDCK cells depleted of the two mammalian Lgl homologs than from control cells (Fig. 1E, mLgl-1/-2 kd). Hence, we predicted that enhanced association of PAR-3 with aPKC would lead to Girdin upregulation. As expected, qPCR and immunoprecipitation analysis showed that Lgl knockdown caused a twofold elevation in both Girdin mRNA and protein levels, whereas levels of LIMK2 mRNA and of LIMK2 and PAR-1b proteins are unaltered (Fig. 1F,G). These results suggest that mammalian Lgl represses Girdin expression through inhibiting the PAR-3–aPKC complex.

To exclude the possibility that protein depletion nonspecifically changes Girdin expression, we examined the effect of knockdown of PAR-1b, another kinase in the PAR–aPKC system (Suzuki et al., 2004), on Girdin expression. Girdin immunoprecipitation from PAR-1b kd cells clearly showed that PAR-1b knockdown does not affect Girdin protein levels (supplementary material Fig. S1F).

An important question arising from these findings is how PAR-3 regulates Girdin expression. An mRNA degradation assay showed almost no difference in Girdin mRNA stability under PAR-3- and Lgl-depleted conditions (supplementary material Fig. S2A,B). Thus, our observation that PAR-3-depletion causes a drastic reduction in Girdin mRNA levels raises the possibility that PAR-3 regulates Girdin mRNA synthesis. We therefore investigated potential mechanisms of Girdin mRNA transcriptional regulation. The STAT3 transcription factor has recently been reported to activate Girdin transcription (Dunkel et al., 2012). However, PAR-3 silencing did not affect the levels of either STAT3 phosphorylated at Y705 or total STAT3 (supplementary material Fig. S2C), suggesting that other transcription factors are involved in regulating Girdin transcription. To obtain insight into the mechanism regulating Girdin transcription in our system, a series of luciferase reporter plasmids containing various lengths of the Girdin 5′-flanking region were transiently transfected into MDCK cells. As shown in Fig. 2A, the Girdin promoter region from −82 to +41 (−82/+41) activated the reporter gene and the Girdin −891/+41, −383/+41, −145/+41 and −112/+41 sequences had approximately the same capacity to promote luciferase activity. In contrast, the −47/+41 promoter region induced significantly lower levels of luciferase activity compared with the Girdin −82/+41 sequence. These results locate the regulatory elements responsible for Girdin transcription to the promoter region between −82 and −47. Analysis of the TRANSFAC database showed that this promoter region contains two types of putative binding sequences that are highly conserved in mammals: two AP-2-binding sites (positions −75 to −67 and −62 to −51) and one Sp1-binding site (position −71 to −63; Fig. 2B). Deletion of a sequence containing the distal AP-2- and Sp1-binding sites (Δ−82/−63 or Δ−75/−67) from the Girdin −891/+41 reporter construct had almost no effect on luciferase activity, whereas deletion of a region containing the proximal AP-2 site (Δ−66/−53) abrogated luciferase activity to the same level as deletion of the whole essential region (Δ−82/−48; Fig. 2B,C). These results suggest that the proximal putative AP-2-binding site is the cis-acting element essential for Girdin transcription.

AP-2 is a developmentally regulated transcription factor family that is crucial for the expression of genes such as tight junction regulators during vertebrate development (Choi et al., 2012; Eckert et al., 2005). To evaluate the involvement of AP-2 in Girdin transcription, we next co-transfected the Girdin promoter −891/+41 construct and an AP-2α expression vector into HepG2 cells, which express a low level of endogenous AP-2α and are therefore suitable for AP-2 overexpression studies (Braganca et al., 2003). As shown in Fig. 2D, AP-2α overexpression resulted in activation of the Girdin promoter in a dose-dependent manner, whereas Sp1 overexpression did not alter luciferase activity. AP-2α co-transfection with the reporter constructs revealed that AP-2α-dependent elevation in reporter activity requires the essential region (−82 to −47; Fig. 2E). To further validate AP-2α control of Girdin transcription, point mutations frequently observed in patients with branchio-oculo-facial syndrome (Milunsky et al., 2011), which are known to result in loss of AP-2α DNA binding (Garcia et al., 2000), were introduced into the basic region of human AP-2α expression plasmids. A reporter assay clearly indicated that these AP-2α point mutants failed to activate the Girdin promoter (Fig. 2F). Unlike the nuclear localization of WT AP-2α, we also observed that mutant proteins were abnormally localized to the cytoplasm, suggesting that the mutated amino acids are crucial for Girdin transcription (supplementary material Fig. S2D). Importantly, chromatin immunoprecipitation (ChIP) revealed that AP-2α binds to the Girdin promoter in vivo (Fig. 2G). These findings indicate that Girdin gene transcription is regulated by the AP-2α transcription factor.

Next, to determine whether AP-2α is involved in PAR-3-stimulated Girdin expression, we performed a gain-of-function study by introducing exogenous AP-2α into PAR-3-depleted cells. qPCR and immunoprecipitation analyses showed that AP-2α overexpression strikingly recovered the Girdin mRNA and protein expression in PAR-3 kd cells (Fig. 3A,B), suggesting that AP-2α functions downstream of PAR-3. In addition, a Girdin reporter assay showed that, compared with AP-2α transfection alone, PAR-3 co-transfection led to a fourfold stimulation of Girdin luciferase activity (Fig. 3C). aPKCλ overexpression also stimulated AP-2-regulated Girdin luciferase activity (supplementary material Fig. S2E). These findings indicate that PAR-3 and aPKCλ cooperate with AP-2α to induce Girdin expression.

In PAR-3 kd cells, there was no change in endogenous AP-2α protein expression and nuclear localization compared with control cells (supplementary material Fig. S2F,G). To test whether AP-2α is causally linked to cell polarity defects in PAR-3 kd cells, we performed a Ca²⁺ switch assay using control and PAR-3 kd cells with or without AP-2α. Control MDCK cells showed the correct distribution of the ZO-1 tight junction protein (also known as TJP1) at the cell–cell boundary and of the gp135 (also known as podocalyxin, PODXL) apical protein on the apical surface (Fig. 3D). In contrast, PAR-3 kd cells transfected with empty vector showed defects in tight junction formation and apical domain development. Importantly, we found that AP-2α overexpression rescued the defects in ZO-1 distribution at the cell–cell contacts and in apical domain development caused by PAR-3 deficiency (Fig. 3D). To confirm that Girdin is involved in regulating cell polarity downstream of PAR-3, we overexpressed WT Girdin in PAR-3 kd cells (supplementary material Fig. S2H). Consistent with the results of AP-2α overexpression, WT Girdin overexpression partially rescued ZO-1 distribution and actin organization in PAR-3 kd cells (Fig. 3E, supplementary material Fig. S2I). Taken together, these results support the hypothesis that AP-2-induced Girdin expression plays an important role in PAR-3 function during epithelial cell polarization.

When depolarized MDCK cells were subjected to a Ca²⁺ switch assay, Girdin mRNA and protein levels were transiently increased in the early-to-intermediate phases of cell polarization, whereas LIMK2 expression was almost unchanged (Fig. 4A, supplementary material Fig. S3A). This transient Girdin mRNA induction was not observed in PAR-3 kd cells (supplementary material Fig. S3B). These results are consistent with the involvement of Girdin in epithelial cell polarization. To directly test this hypothesis, we generated Girdin-depleted MDCK cells using shRNA targeting two different sequences to avoid off-target effects. qPCR and immunoprecipitation using an anti-Girdin antibody showed that Girdin expression was efficiently reduced in the two independent MDCK cell clones (Girdin kd #GA10, #GB9), with no effects on the expression of other proteins including PAR-3, aPKCλ and LIMK2 (Fig. 4B, supplementary material Fig. S3C). Immunostaining for tight junction markers such as occludin and ZO-1 after the Ca²⁺ switch revealed that tight junction formation was significantly delayed in Girdin kd cells compared with control cells. At 2 h after the Ca²⁺ switch, when normal occludin and ZO-1 localization to cell–cell contacts was seen in control cells, discontinuous occludin and ZO-1 staining at cell–cell contact zones was observed in Girdin kd cells (Fig. 4C,D). This was followed by the eventual formation of tight junctions up to 6 h after the Ca²⁺ switch at which time ZO-1 was localized to the apical junctional region in Girdin kd cells (Fig. 4C, supplementary material Fig. S3D). In contrast, E-cadherin and β-catenin staining showed that adherens junction assembly after the Ca²⁺ switch was almost unaffected by Girdin depletion (Fig. 4D, arrowheads; supplementary material Fig. S3E). Although the phenotypic changes seen in Girdin kd cells are not as severe as those of PAR-3 kd cells, the features of tight junction formation in Girdin kd cells strongly resemble those of PAR-3 kd cells (Chen and Macara, 2005), suggesting that Girdin is involved in tight junction formation. This was supported by a functional measurement of tight junction integrity, i.e. transepithelial electrical resistance (TER; Fig. 4E). During cell–cell junction maturation, correlating with the timing of transient Girdin expression, TER in control cells was elevated with a peak at around 8–14 h after the Ca²⁺ switch, and gradually decreased. TER development was strongly suppressed in Girdin kd cells and was abolished in PAR-3 kd cells (Fig. 4E), suggesting that Girdin contributes to the later processes that strengthen tight junction integrity, rather than the initial steps. Previous studies have demonstrated that tight junction permeability is controlled by the actin cytoskeleton (Madara et al., 1986; Stevenson and Begg, 1994). Phalloidin staining of F-actin showed that Girdin depletion caused aberrant organization of the actin cytoskeleton during polarization (Fig. 4F). In control cells, an actin ring rapidly localized to the cell periphery upon Ca²⁺ repletion, but F-actin-positive vesicle-like structures accumulated and were retained inside cells lacking Girdin even at 6 h after the Ca²⁺ switch. Taken together, these results suggest that Girdin contributes to tight junction integrity by promoting the correct organization of F-actin.

Immunostaining showed that the F-actin-positive intracellular vesicle-like structures that accumulate in Girdin kd cells show strong co-staining with the apical gp135 protein (Fig. 4F, supplementary material Fig. S3F,G), indicating that they correspond to vacuolar apical compartments (VACs). VACs are apical membrane structures formed by the fusion of short-lived apical carrier vesicles that contribute to the establishment of apical membrane domains (Vega-Salas et al., 1987, 1988). Manipulation of PAR-3 complex components affects VAC formation under low- Ca²⁺ conditions (Horikoshi et al., 2009; Yamanaka et al., 2006). In control cells, gp135 is retained at the cell periphery. However, Girdin depletion caused the accumulation of gp135-positive VACs (supplementary material Fig. S3H). The proportion of Girdin kd cells retaining gp135 at the cell periphery was significantly lower in Girdin-depleted cells than in control cells (supplementary material Fig. S3I), indicating that, similar to PAR-3 and aPKC, Girdin is involved in the maintenance of apical domains (Horikoshi et al., 2009). Taken together, these results support a role for Girdin in the regulation of apicobasal polarity.

We next investigated the molecular basis for the defects in tight junction formation and apical domain development caused by Girdin depletion. A recent study reported that Girdin physically interacts with PAR-3 (Brajenovic et al., 2004; Ohara et al., 2012). We confirmed that the Girdin C-terminus including the CT domain preferentially binds to PAR-3 and that the PAR-3 C-terminus is required for its interaction with the Girdin C-terminus (supplementary material Fig. S4A,B). To obtain further mechanistic insight, we next focused on Gαi3, a member of the heterotrimeric Gαi family and a Girdin-binding partner (Le-Niculescu et al., 2005). When we performed pulldown experiments to clarify whether Girdin interacts with Gαi3 in MDCK cells, GST-tagged C-terminal WT Girdin pulled down endogenous Gαi3 from MDCK cell lysates, whereas the F1658A point mutant (corresponding to F1685A in Girdin isoform 1) did not, as previously reported (Garcia-Marcos et al., 2009; supplementary material Fig. S4C). Furthermore, affinity purification experiments also revealed that streptavidin-binding peptide (SBP)-tagged WT Girdin, but not the F1658A mutant, coprecipitated with endogenous Gαi3 (Fig. 5A). These observations suggest that Girdin binding to Gαi3 contributes to the partial rescue of the PAR-3 kd phenotype achieved upon WT Girdin overexpression (Fig. 3E). Consistently, the F1658A mutant that is impaired in Gαi3 binding showed weaker activity than WT Girdin for restoring the defects in ZO-1 distribution caused by PAR-3 depletion (Fig. 3E). These results not only support the previous findings on the relationship between PAR-3 and Girdin, and Girdin and Gαi3 but also suggest a close physical and functional relationship between PAR-3 and the Girdin–Gαi3 complex.

To clarify this point, we investigated whether Gαi3 participates in regulating apicobasal polarity. Immunostaining of polarized MDCK cells with an anti-Gαi3 antibody revealed that Gαi3 is concentrated at the cell periphery in a very similar manner to F-actin (Fig. 5B, Ca²⁺ switch 24 h). Interestingly, in depolarized cells, Gαi3 localizes to VAC intracellular aggregates, which stain positively for both gp135 and F-actin (Fig. 5B, Ca²⁺ switch 0 h). Gαi3-positive vacuoles were redistributed to the cell peripheral cortex together with F-actin during cell polarization (Fig. 5B, Ca²⁺ switch 0.5 h, 2 h and 6 h). Transient siRNA-mediated Gαi3 knockdown confirmed the antibody specificity (supplementary material Fig. S4D). These results suggest that Gαi3 is involved in F-actin assembly, which has been implicated in regulating apicobasal polarity.

To evaluate the role of Gαi3 in this process, we generated Gαi3-depleted MDCK cells, in which Gβ, PAR-3 and aPKCλ expression was unaffected (Fig. 5C). We then performed a Ca²⁺ switch assay to assess apicobasal polarity. ZO-1 and gp135 immunostaining showed that tight junction formation and apical domain development were profoundly delayed in Gαi3 kd cells compared with control cells (Fig. 5D). F-actin organization was also altered in Gαi3 kd cells. The phenotype of Gαi3 kd cells therefore resembles those of PAR-3 kd and Girdin kd cells, strongly supporting a role for Gαi3 in regulating the formation of tight junctions and apical membrane domains.

These results indicating that, although there are differences in phenotypic severity, depletion of PAR-3, Girdin or Gαi3 leads to similar defects in cell polarity, prompted us to investigate the functional relationships among the three proteins. To do this, we monitored the behavior of Gαi3-positive vacuoles. As previously mentioned, in control cells Gαi3-positive vacuoles disappear after Ca²⁺ repletion and redistributed to the cell cortex up to 5 h after Ca²⁺ switch (Fig. 5E). In contrast, ∼40% of PAR-3 kd cells still contained Gαi3-positive vacuoles at this time point (Fig. 5E,F). When we next overexpressed WT Girdin into PAR-3 kd cells, the aberrant distribution of Gαi3-positive vacuoles was significantly rescued by the exogenous WT Girdin (Fig. 5E,F). Consistent with this, knockdown of Girdin or aPKCλ delayed Gαi3 redistribution (supplementary material Fig. S4E,F). These results provide mechanistic insight into the regulation of apicobasal polarity by Girdin: Girdin controls Gαi3 distribution downstream of PAR-3 during cell polarization.

To address whether PAR-3, AP-2, Girdin and Gαi3 are involved in epithelial morphogenesis, we embedded MDCK cells into type I collagen gel to form cysts. As previously reported (Horikoshi et al., 2009), control cells formed 3D cysts consisting of a single polarized monolayer surrounding a central lumen, whereas PAR-3 kd cells frequently developed abnormal cysts with multiple lumens (Fig. 6A,B). Importantly, overexpression of either AP-2α or WT Girdin partially rescued the multiple lumen phenotype of PAR-3 kd cysts (Fig. 6A,B). Consistent with this, Girdin knockdown resulted in the formation of cysts with multiple lumens (Fig. 6C,D). In Girdin-depleted cysts, E-cadherin was excluded from luminal surfaces containing apical gp135 (supplementary material Fig. S4G), indicating that the apical domain of each cell was not correctly integrated, similar to reports for PAR-3 and apical transport proteins (Bryant et al., 2010; Horikoshi et al., 2009; Torkko et al., 2008). Gαi3 colocalized with F-actin in the apical membrane of MDCK cysts (Fig. 6E), and Gαi3 depletion also resulted in the formation of cysts with multiple lumens (Fig. 6F,G). Taken together, these results suggest that Gαi3 and AP-2-induced Girdin transcription are involved, at least partly, in apical domain development downstream of PAR-3 during cystogenesis.

Finally, to confirm the importance of the interaction between Girdin and Gαi3 in cystogenesis, we examined the effect of overexpressing WT Gαi3 or a W258F mutant unable to bind Girdin (Garcia-Marcos et al., 2010; supplementary material Fig. S4H). We found that Gαi3 W258F had a dominant-negative effect on the formation of cysts with multiple lumens, whereas WT Gαi3 did not (Fig. 6H,I). These results support a role for the Girdin–Gαi3 interaction in apical domain development during epithelial cyst morphogenesis.

Establishing apicobasal polarity is a fundamental process in the development of epithelial tissues. In this process, PAR-3 provides a spatial cue for the development of tight junctions and apical domains. In the present study, we uncovered a new molecular network that functions downstream of PAR-3. On the one hand, PAR-3 controls Girdin protein expression at the transcription level by cooperating with the AP-2 transcription factor, and, on the other hand, PAR3 binds to Girdin and induces some aspects of apicobasal polarity such as tight junction formation, actin cytoskeletal organization and apical membrane development by regulating Gαi3 localization during epithelial cell morphogenesis (Fig. 6J). Thus, in addition to its established role of providing a spatial cue during cell polarization, PAR-3 regulates the expression of a downstream signaling mediator, Girdin, to enforce cell polarity.

During the initial phase of epithelial cell polarization, PAR-3 accumulates at primordial cell–cell contact sites and, together with various binding partners, including aPKC, promotes the subsequent assembly of tight junctions and apical membranes at the contact sites (Chen and Macara, 2005; Horikoshi et al., 2009). In this study, we confirmed that PAR-3 physically interacts with Girdin, which is consistent with previous studies (Brajenovic et al., 2004; Ohara et al., 2012). Furthermore, we found that PAR-3 upregulates Girdin expression through AP-2-induced Girdin transcription. Exogenous AP-2 or WT Girdin overexpression partially corrects the polarity defects caused by PAR-3 depletion. These data suggest that, besides their direct function at cell–cell contact sites, PAR-3 and aPKC transduce polarity cues to nuclear AP-2 to activate Girdin transcription in polarizing cells, which is suppressed by Lgl (Fig. 6J). In turn, Girdin might help to assemble the actin cytoskeleton with Gαi3 in the vicinity of PAR-3, which strengthens tight junction integrity, especially in the later phases of polarization. It is possible that, by coordinating with the tiny amount of residual PAR-3 in PAR-3 kd cells, overexpressed Girdin can overcome the polarity defects caused by PAR-3 knockdown. Several early studies have demonstrated that increased synthesis of regulatory proteins is required for the establishment of cell–cell junctions (Griepp et al., 1983; Sang et al., 1980; Yamada et al., 2005). Our findings provide additional insight into this mechanism. However, how PAR-3 regulates the transcriptional activity of AP-2 remains to be established. The activation status of STAT3, another transcription factor involved in Girdin expression (Dunkel et al., 2012), was unaffected in PAR-3 kd cells. We predicted that PAR-3 located at cell–cell contact sites might activate unknown signaling cascade(s), resulting in AP-2 activation. Another possibility is that a small nuclear PAR-3 population plays a role in AP-2-induced Girdin transcription. It has been previously reported that PAR-3 occasionally localizes to the nucleus, as well as to cell–cell contacts (Fang et al., 2007). Biochemical studies indicated that PAR-3 can form a complex with the Ku80, Ku70 and catalytic subunits of DNA-dependent protein kinase (DNA-PK), which are also AP-2-binding proteins (Fang et al., 2007; Nolens et al., 2009; Traweger et al., 2008). Thus, PAR-3 might activate Girdin transcription by binding to the DNA-PK–AP-2 complex, and aPKC might be involved in processes such as protein assembly. Further studies to identify AP-2-binding proteins might help to elucidate the mechanism.

During epithelial morphogenesis, lumen formation in MDCK cysts is mainly generated by a process known as ‘hollowing’, which involves exocytosis of small vesicles containing apical components into the central lumen (Andrew and Ewald, 2010; Bryant and Mostov, 2008). In addition to PAR-3, the results presented herein support a role for Girdin–Gαi3 signaling in the regulation of hollowing. First, we found that knockdown of Girdin or Gαi3 in MDCK cells causes defects in apical lumen formation by cysts. Second, Gαi3 localizes to apical-protein-containing vacuoles. Third, Girdin overexpression partially rescued the aberrant distribution of Gαi3-positive apical vacuoles caused by PAR-3 deficiency. Girdin has been previously shown to control actin cytoskeleton remodeling by associating with Gαi3 and acting as a GEF (Garcia-Marcos et al., 2010, 2009). We also confirmed the physical association between Girdin and Gαi3, which is an important downstream component of PAR-3. Taken together, these results suggest that Girdin controls assembly of the actin cytoskeleton through binding to and/or activating Gαi3 to promote apical domain development during epithelial morphogenesis (Fig. 6J). This notion is further supported by our finding that overexpression of inactive Gαi3 (a W258F mutant impaired in Girdin binding) causes apical lumen defects in cysts.

Similar hollowing processes have also been reported to occur during endothelial cell lumen formation (Iruela-Arispe and Davis, 2009). PAR-3, aPKC, AP-2 family members and Girdin regulate lumen and tube formation by endothelial cells (Kitamura et al., 2008; Koh et al., 2008; Park et al., 2007; Zovein et al., 2010). Thus, it will be interesting to investigate whether a conserved molecular network comprising PAR-3, AP-2, Girdin and Gαi3 also regulates cell polarization during tube formation in endothelial tissues.

Loss of cell polarity and subsequent tissue disorganization are frequently associated with cancer development (Feigin and Muthuswamy, 2009). PAR-3 gene mutations resulting in the loss of protein expression have been reported in prostate cancer LNCaP cells (Kunnev et al., 2009). AP-2α expression is also lost in LNCaP cells and in prostate adenocarcinoma tissues (Ruiz et al., 2001). We found that the Girdin mRNA and protein levels were markedly decreased in PAR-3- and AP-2α-deficient LNCaP cells. These findings raise the possibility that regulation of cell polarization by the PAR-3–AP-2α–Girdin pathway might be impaired in cells that lose expression of these proteins during prostate cancer progression.

Girdin has been reported to be highly expressed in invasive colon and breast carcinoma tissues, and high levels of Girdin expression correlate significantly with ErbB2 oncogene expression in breast cancer (Garcia-Marcos et al., 2011; Jiang et al., 2008; Ling et al., 2011). Interestingly, ErbB2 expression is also transcriptionally regulated by AP-2α and is associated with AP-2α in breast cancer (Allouche et al., 2008; Bosher et al., 1995; Pellikainen et al., 2004). Therefore, it is possible that accumulated genetic and epigenetic mutations during cancer progression might deregulate PAR-3, leading to aberrant transcriptional activity and/or increased expression of AP-2α, and resulting in Girdin and ErbB2 overexpression. Further studies, for example, immunohistochemical analysis of clinical cancer specimens, will help to determine whether dysfunction of the PAR-3–AP-2α–Girdin pathway contributes to cancer initiation and progression.

In conclusion, this study has revealed that the PAR-3–AP-2α–Girdin and PAR-3–Girdin–Gαi3 pathways contribute to epithelial cell polarity, and form a new regulatory mechanism for determining cell polarity in epithelial cells.

MDCK II, HepG2, HeLa and prostate cancer cells were grown as previously described (Horikoshi et al., 2009; Ishiguro et al., 2009). Using RNAiMAX, Lipofectamine 2000 (Invitrogen) or Hekfectin (Bio-Rad), cells were transfected with the appropriate siRNAs or plasmids. Ca²⁺ switch and 3D cyst formation assays were performed as previously described (Yamanaka et al., 2006).

Based on the sequence of GenBank ID NM018084.4 (amino acids 1–1843) and human EST clones (IMAGE ID 6176807, IMAGE ID 268857, DKFZp686D0630 and KIAA1212), cDNA encoding human Girdin isoform 2 was generated by standard PCR-based procedures and subcloned into SR-V5, pCAG-GS-HA-SBP (Yamashita et al., 2010) and pGEX expression vectors. The PAR-3–EGFP fusion construct driven by the CAG promoter was generated by the insertion of EGFP downstream of rat PAR-3 cDNA. The Sp1 expression plasmid was a generous gift from Yoshiaki Fujii (Tsukuba University, Tsukuba, Japan). cDNA encoding human AP-2α isoform b (amino acids 1–431; IMAGE clone ID 4432023, GenBank ID BC017754.1) was subcloned into the SR-HA, pCAG-GS-HA-SBP. cDNA encoding human Gαi3 (amino acids 1–354; IMAGE clone ID1687117) was purchased from Missouri S&T cDNA Resource Center and subcloned into pCAG-GS-HA-SBP.

Confluent MDCK cells were harvested and resuspended in RIPA lysis buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 10% glycerol, 1 mM dithiothreitol, 1 mM EGTA, 1% Triton X-100, protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktail (Roche)] for immunoprecipitation with an anti-PAR-3 antibody. SDS (0.1%) and deoxycholate (0.05%) were added to the lysis buffer for immunoprecipitation with an anti-Girdin or anti-AP-2α antibody. After incubation for 30 min on ice, cell lysates were clarified by centrifugation at 20,400 g for 30 min and preabsorbed onto protein-A–Sepharose (GE Healthcare) for 1 h at 4°C. Each antibody was incubated with cell lysates for 4 h or overnight at 4°C, followed by incubation with protein-A–Sepharose beads for 2 h at 4°C. After washing three times with lysis buffer, protein complexes were eluted with SDS sample buffer, separated by SDS-PAGE and immunoblotted.

Confluent MDCK cells were harvested and resuspended in lysis buffer [50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 10 mM MgCl₂, 2 mM dithiothreitol, 5 mM EDTA, 0.4% Nonidet P-40, protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktail (Roche)] (Garcia-Marcos et al., 2010). Cell lysates were clarified by two rounds of centrifugation at 20,400 g for 30 min. Streptavidin-Mag Sepharose (GE Healthcare) were added to cell lysates and incubated overnight at 4°C. After washing four times with washing buffer [4.3 mM Na₂HPO₄, 1.4 mM KH₂PO₄, pH 7.4, 137 mM NaCl, 2.7 mM KCl, 0.1% Tween 20, 10 mM MgCl₂, 5 mM EDTA] (Garcia-Marcos et al., 2010), affinity-purified protein complexes were eluted with 10 mM biotin, separated by SDS-PAGE and immunoblotted.

GST-fused Girdin C-terminal fragments (amino acids 1518–1630, 1665–1843 and 1518–1843) were expressed in Escherichia coli (strain DH-5α, BL21) and purified on glutathione–Sepharose-4B beads (GE Healthcare) by standard procedures.

TER was measured using an ERS electrical resistance system (Millipore), as previously described (Suzuki et al., 2001). TER values were calculated by subtracting the blank value (from an empty filter) and expressed in units of Ohm×cm².

PAR-3 kd (#13–32 and #25a), mLgl-2 kd (#2–10), and mLgl-1/-2 double kd (#24–15) stable MDCK clones and corresponding control clones (#11–10 and #1–5 for PAR-3 kd, #21–1 for mLgl kd) were characterized previously (Yamanaka et al., 2006). To establish aPKCλ kd (#B5–3), Girdin kd (#GA10 and #GB9) and Gαi3 kd MDCK clones, a pSUPERneo aPKCλ-specific shRNA expression plasmid, pSUPER.neo+gfp (OligoEngine) Girdin-specific shRNA expression plasmids or a pSUPERIOR.puro (Oligoengine) Gαi3-specific shRNA expression plasmid, respectively, was introduced into MDCK cells, followed by G418 or puromycin selection. The canine aPKCλ siRNA sequence was described previously (Suzuki et al., 2004). The two distinct Girdin shRNA target sequences were previously characterized (Enomoto et al., 2005), and are conserved in the Canis familiaris Girdin gene. The siRNA sequence 5′-GGAGTGATTAAACGGTTAT-3′ was used for canine Gαi3 knockdown.

Total RNA was extracted using an RNeasy Plus mini kit (QIAGEN). Samples were hybridized to a Canine Genome 2.0 array (Affymetrix) for microarray analysis. cDNA was reverse transcribed using a High Capacity cDNA Reverse Transcription kit (Applied Biosystems) and subjected to qPCR using an iCycler iQ system (Bio-Rad). cDNA produced from control cells was used to prepare standard curves. Relative mRNA expression levels were normalized to those of 18S rRNA. The following TaqMan primer sets (Applied Biosystems) were used in qPCR analysis: Canis CCDC88A/Girdin, Cf02688833_m1 or Cf02630272_m1; Canis LIMK2, Cf02637893_m1; Homo sapiens CCDC88A/Girdin, Hs_00250677_m1 or Hs_00214014_m1; H. sapiens GAPD, 4352934E; Eukaryotic 18S rRNA, #4352930E.

The human Girdin gene promoter was cloned into the pGL4.10 (luc2) vector (pGL4-Girdin; Promega). The pRL-CMV vector (Promega) was used as an internal control. MDCK or HepG2 cells were transfected with Girdin reporter plasmids and a pRL-CMV control reporter with or without SR-HA-AP-2α, pCMV-Sp1, SR-His-PAR-3 and SR-His-aPKCλ. Empty vector was added to each transfection mix such that the total amount of plasmid DNA was constant for each experiment. Luciferase activity was measured at 44–48 h after transfection using the dual-luciferase reporter assay system (Promega).

ChIP was performed according to the manufacturer's instructions (EZ ChIP kit; Millipore). Briefly, cells were fixed in 1% formaldehyde for 10 min, followed by a 5-min incubation after the glycine addition. Cells were then harvested, resuspended in lysis buffer and sonicated. After centrifugation, cell lysates were diluted in ChIP dilution buffer and incubated overnight with antibodies at 4°C. Immunoprecipitated samples were eluted and reverse cross-linked by overnight incubation at 65°C in elution buffer. Genomic DNA was then extracted and processed for PCR according to the manufacturer's instructions (LA Taq with GC buffer kit, TAKARA).

MDCK cells were fixed in 2% paraformaldehyde, permeabilized with 0.5% Triton X-100 in PBS and stained after blocking with 10% calf serum or 5% BSA in PBS. Secondary antibodies were Alexa-Fluor-488-, 555- or 647-conjugated goat antibodies against rabbit, mouse or rat immunoglobulin G (IgG; Life Technologies). Alexa-Fluor-488, 647- or Rhodamine-conjugated phalloidin and TOPRO (Life Technologies) were used to visualize filamentous actin and nuclei, respectively. Samples were examined using an AxioImager Z1 imaging microscope (Carl Zeiss) equipped with laser scanning confocal system (LSM510 or LSM700) or a CSU10 disk confocal system (YOKOKAWA) and a water-immersed 40× objective lens. According to each confocal driver system, images were analyzed by LSM Image Browser, Zen or MetaMorph software, and were assembled and adjusted using Photoshop software (Adobe Systems).

The rabbit anti-Girdin polyclonal antibody (pAb; #1AP) was raised against a mixture of GST-fused human Girdin peptide antigens (amino acids 1518–1630 and 1665–1843) and affinity purified on the antigen-conjugated column. The #1AP antibody cross-reacts with isoform 1 (NP_001129069.1, molecular mass 216 kDa), isoform 2 (NP_060554.3, molecular mass 213 kDa), isoform 3 (NP_001241872.1, molecular mass 208 kDa) and isoform X1 (XP005264475.1, molecular mass 233 kDa). Polyclonal rabbit anti-mLgl-1 (C-2AP), rabbit mLgl-2 (N13AP), rabbit anti-PAR-6β (BC31AP) and rabbit anti-PAR-1b (MAP1) antisera were described previously (Suzuki et al., 2004; Yamanaka et al., 2003). Anti-gp135 mouse monoclonal antibody (mAb; 3F2) was a generous gift from George K. Ojakian (SUNY Downstate Medical Center, NY). All other antibodies were purchased from commercial sources: rat anti-ZO-1 mAb, rabbit anti-aPKCζ pAb, rabbit anti-LIMK2 pAb, rabbit anti-Gαi3 pAb, rabbit anti-Gβ pAb, rabbit anti-AP-2α pAb, mouse anti-AP-2α mAb and mouse anti-GFP mAb (Santa Cruz Biotechnology); mouse anti-ZO-1 mAb, mouse anti-occludin mAb, rabbit anti-occludin pAb and rabbit anti-claudin-1 pAb (Zymed); mouse anti-E-cadherin mAb, mouse anti-aPKCι mAb, mouse anti-α-catenin mAb and mouse anti-β-catenin mAb (BD Biosciences); rabbit anti-PAR-3 pAb and rabbit anti-Gαi3 pAb (Millipore); mouse anti-β-actin mAb (Sigma); mouse anti-mLgl-2 mAb (Abnova); mouse anti-GAPDH mAb (Abcam); rabbit anti-GFP pAb (MBL); rat anti-HA mAb (Roche); rabbit anti-STAT3 pAb and rabbit anti-pY705-STAT3 pAb (Cell Signaling).

We thank Takahiro Nagase (Kazusa DNA Research Institute, Japan) for KIAA1212 cDNA, Yoshiaki Fujii for Sp1 cDNA, Sachiko Tsukita (Kyoto University, Kyoto, Japan) for HeLa cells and George K. Ojakian for the gp135 antibody. We also thank Hitoshi Ishiguro (Yokohama City University, Japan) for technical advice and suggestions. We are grateful to all members of the Ohno laboratory for their technical help and regular discussions.