Lgl mediates apical domain disassembly by suppressing the PAR-3-aPKC-PAR-6 complex to orient apical membrane polarity

Lethal giant larva (Lgl), one of the basolateral tumor suppressor proteins in Drosophila (Gateff, 1978), has been shown to be an indispensable protein for the establishment of apical-basal polarity of Drosophila epidermis and epithelial-derived neural progenitor cells (Bilder et al., 2000; Ohshiro et al., 2000; Peng et al., 2000). In mammals, mammalian homologues of Lgl (mLgl), mLgl-1 and mLgl-2, have also been shown to localize to the basolateral membrane domain of epithelial cells (Musch et al., 2002; Yamanaka et al., 2001). In addition, mLgl-1 knockout mice have been reported to show hyperproliferation of neuroepithelial cells in the developing brain (Klezovitch et al., 2004), suggesting that mLgl is also important for the regulation of epithelial tissue morphology and brain development in mammals.

An apical protein complex containing atypical PKC (aPKC), PAR-6 and PAR-3, the PAR-3-aPKC-PAR-6 complex, has also been shown to be involved in the establishment of apical-basal polarity of epithelial cells (Bilder, 2004; Knust and Bossinger, 2002; Macara, 2004; Ohno, 2001). In Drosophila, this protein complex localizes to the subapical regions, and is indispensable for the establishment of apical junctions and apical-basal membrane polarity (Muller and Wieschaus, 1996; Petronczki and Knoblich, 2001; Wodarz et al., 2000). In mammalian epithelial cells, this protein complex localizes to the apical junctions (Hirose et al., 2002; Izumi et al., 1998; Joberty et al., 2000; Lin et al., 2000; Suzuki et al., 2001) and has an important role in the establishment of apical junctions during cell-cell adhesion-mediated epithelial cell polarization (Chen and Macara, 2005; Gao et al., 2002; Hirose et al., 2002; Mizuno et al., 2003; Suzuki et al., 2002; Suzuki et al., 2001; Yamanaka et al., 2001).

Recently, a direct interaction between basolateral Lgl and apical aPKC-PAR-6 complex has been shown in Drosophila and mammalian epithelial cells (Betschinger et al., 2003; Hutterer et al., 2004; Yamanaka et al., 2003). In addition, Lgl has been shown to be phosphorylated by aPKC, which is required for the process to exclude Lgl from apical region in epithelial cells (Betschinger et al., 2003; Hutterer et al., 2004; Musch et al., 2002; Yamanaka et al., 2003). On the contrary, in mammalian cells, overexpressed mLgl competes with PAR-3 for its interaction with the aPKC-PAR-6 complex and suppresses the cell-cell adhesion-induced formation of apical junctions in mammalian epithelial cells (Yamanaka et al., 2003). A genetic study using Drosophila has shown that a reduction in aPKC levels suppresses the development of Lgl mutant phenotypes, such as cell polarity defects and tumorigenesis (Rolls et al., 2003). Further, studies using Drosophila or Xenopus embryo indicate that mutual inhibition between apical aPKC and basolateral Lgl is important for the maintenance of epithelial membrane polarity (Chalmers et al., 2005; Hutterer et al., 2004). Although these observations suggest the importance of an antagonistic interaction of Lgl with apical PAR-3-aPKC-PAR-6 complex on the development of polarized membrane domains in epithelial cells, the precise mechanism of Lgl-mediated regulation of membrane domain polarization remains obscure.

Using small interfering RNA (siRNA), we provide evidence that endogenous mLgl facilitates the disassembly of apical membrane domains in MDCK cells depolarized by Ca²⁺-depletion. Importantly, this effect is mediated by the suppression of apical PAR-3-aPKC-PAR-6 complex activity. Furthermore, analyses using 2D- or 3D-cultured MDCK cells in collagen gel suggest that this mLgl-mediated suppression of apical protein complex activity is important for the collagen-mediated orientation of apical membrane polarity.

To examine the mechanism of Lgl-mediated regulation of epithelial cell polarity, we initially established MDCK cell clones (24-15 and 24-16) in which both mLgl-1 and mLgl-2 were stably knocked down (mLgl kd), by using pSUPER-based siRNA vectors. Control clones (21-1 and 21-7) were similarly established, by using a negative control sequence. Western blot analysis revealed that the expression of mLgl-1 or mLgl-2 was reduced to less than 10% in both (24-15 and 24-16) clones compared with the 21-1 clone (Fig. 1A). The expression of other proteins including PAR-3, aPKCλ, PAR-6β and PAR-1b was not affected (Fig. 1A).

Since we have shown previously that the overexpression of mLgl-2 in MDCK cells inhibits the formation of apical junctions induced by cell-cell adhesion (Yamanaka et al., 2001), we first performed a Ca²⁺-switch assay to assess the role of mLgl-1 and/or mLgl-2 in the formation of epithelial cell polarity initiated by cell-cell adhesion. However, ZO-1 (tight junction), ezrin (apical) and E-cadherin (basolateral) localized almost normally in the mLgl kd cells 3 and 24 hours after the Ca²⁺ switch (supplementary material Fig. S1), suggesting that mLgl is dispensable for the repolarization process of 2D-cultured MDCK cells that was initiated by cell-cell adhesion. Alternatively, the remaining mLgl-1 and mLgl-2 levels might be sufficient for the repolarization process of MDCK cells induced by the Ca²⁺ switch in mLgl kd cells.

By contrast, the most prominent effects of mLgl knockdown were detected in depolarized MDCK cells. As reported previously (Vega-Salas et al., 1987; Vega-Salas et al., 1988), incubation of control cells (21-1) with Ca²⁺-depleted culture medium induced reductions in levels of the peripheral apical proteins gp135 and ezrin, as well as F-actin, and then redistributed them into intracellular compartments, called vacuolar apical compartments (VACs) (Fig. 1B,C). Importantly, in both mLgl kd cells (24-15 and 24-16) the redistribution of these proteins into VACs was markedly suppressed. Most of the mLgl kd cells retained the apical proteins and F-actin at the cell periphery and intracellular apical proteins were hardly detectable (Fig. 1B,C). Quantification analysis revealed that approxymately 100% of the mLgl kd cells retained gp135 at the cell periphery, compared with only 33% of the control cells (Fig. 1D). In addition, staining of aggregated ezrin was decreased in both mLgl kd cells (Table 1). These results indicate that endogenous mLgl is required for the disassembly of apical membrane proteins and their subsequent redistribution into VACs in response to depolarization of MDCK cells. By contrast, overexpression of human Lgl-2, which has two mismatched bases in the siRNA sequence used for knockdown of canine Lgl-2, decreased the peripheral apical proteins in depolarized mLgl kd cells as well as in control cells (supplementary material Fig. S2A,B), supporting the positive role of mLgl in the disassembly of apical proteins in MDCK cells.

To obtain further insights into the formation of abnormal phenotypes in mLgl kd cells, we analyzed the initial process of depolarization induced by Ca²⁺-depletion. In this experiment, the localization of apical-surface gp135 was monitored by prelabeling it with an antibody against the extracellular region of gp135. In the control cells, surface gp135 gradually disappeared from the cell cortex and localized to the intracellular regions 5 hours after Ca²⁺-depletion. F-actin gradually coaggregated with gp135 in the intracellular regions (arrowheads in Fig. 2A,B). Notably, in mLgl-kd cells, gp135 and F-actin remained at the cell cortex even 5 hours after Ca²⁺-depletion. These results suggest that mLgl is required for the internalization of apical membrane proteins during the depolarization of MDCK cells. Interestingly, the disappearance of ZO-1 and occludin from the cell-cell contact region was partially suppressed in the mLgl-kd cells after Ca²⁺-depletion (supplementary material Fig. S3). This implies the involvement of mLgl in the disassembly of apical junctions in addition to apical membrane proteins during the depolarization of MDCK cells.

We have previously shown that overexpressed mLgl-2 antagonizes the interaction of PAR-3 with the aPKC-PAR-6 complex in epithelial cells (Yamanaka et al., 2001). Thus, we next examined whether the knockdown of endogenous mLgl affects the amount of the PAR-3-aPKC-PAR-6 ternary complex in MDCK cells. The amount of PAR-3 co-precipitated with anti-PAR6β antibody from polarized MDCK cells increased 1.5-fold in both mLgl kd cells (24-15 and 24-16) (Fig. 3A-D), supporting the hypothesis that endogenous mLgl competes with PAR-3 for aPKC-PAR-6 binding in MDCK cells. The amount of aPKC in PAR-6β immunoprecipitates was not affected by mLgl knockdown (Fig. 3A-D), which is consistent with the idea that aPKC and PAR-6 form a stable complex independently of mLgl. Thus, endogenous mLgl has a suppressive effect on the formation of the PAR-3-aPKC-PAR-6 ternary complex in epithelial cells.

Interestingly, mLgl knockdown also caused a marked increase in the co-precipitation of endogenous Cdc42 with PAR-6β, which was hardly detected in the control cells (Fig. 3A). These results suggest that endogenous mLgl not only prevents the aPKC-PAR-6 complex from associating with PAR-3, but also suppresses the interaction of Cdc42 with PAR-6β. It should be noticed that no co-precipitation of Rac1 with PAR-6β was detected (Fig. 3A), suggesting that the Lgl knockdown specifically stabilized the interaction of PAR-6β with Cdc42. Taken together with our previous observation that Cdc42 activates aPKC through PAR-6 in vitro (Yamanaka et al., 2001), these results imply that mLgl also regulates aPKC activity in the aPKC-PAR-6 complex, in addition to its interaction with PAR-3 in epithelial cells.

To examine whether the mLgl-mediated disassembly of apical membrane proteins results from the suppression of PAR-3-aPKC-PAR-6 complex activity in MDCK cells, we examined the effects of PAR-3 knockdown on the distribution of apical proteins in depolarized MDCK cells. For this purpose, we established the PAR-3-knockdown (PAR-3 kd) clones (13-32, 13-33). Control clones (11-10, 25-2) were similarly established by using a negative control sequence. Western blot analysis revealed that the expression level of PAR-3 in the PAR-3 kd cells decreased to 10% (13-32) or 30% (13-33) compared with the control cells (11-10), whereas the expression levels of other proteins including mLgl-1, mLgl-2, aPKCλ, PAR-6β and PAR-1b were unchanged (Fig. 4A).

Interestingly, PAR-3 kd cells showed opposite phenotypes compared with mLgl kd cells: gp135, ezrin and F-actin levels were decreased in the cell periphery but were increased in the intracellular regions in PAR-3 kd cells (Fig. 4B,C). Quantification analysis revealed that only approximately 10% of cells retained gp135 at the cell periphery in both PAR-3 kd clones (Fig. 4D), whereas staining of aggregated ezrin was increased in these cells (Table 1). Similar results were also obtained from another PAR-3 kd clone (25a) which was established by using other siRNA sequences (supplementary material Fig. S2C,D). By analyzing the initial depolarizing process induced by Ca²⁺-depletion, we showed that, in contrast to mLgl kd cells, intracellular localization of surface gp135 from the cell periphery was enhanced in PAR-3 kd cells (Fig. 5A,B). Quantification analysis revealed that aggregated gp135 staining was increased by around sixfold in PAR-3 kd cells after incubation with LC medium for 5 hour (Fig. 5C). These results suggest that the PAR-3 knockdown facilitates the disassembly of apical proteins and redistribution of these proteins into the intracellular regions possibly through the enhanced internalization of apical membrane proteins. Thus, endogenous PAR-3 has suppressive effects on the disassembly of apical proteins. By contrast, the overexpression of PAR-3 induced the retention of the peripheral apical proteins in depolarized MDCK cells (supplementary material, Fig. S2C,D), supporting the negative role of PAR-3 in the disassembly of apical proteins in MDCK cells.

In addition, the overexpression of a dominant-negative mutant of aPKC showed a similar phenotype to PAR-3 kd cells in depolarized MDCK cells: only 20% of cells expressing dominant-negative aPKCλ had peripheral gp135 staining, whereas approximately 40% of cells expressing LacZ or wild-type aPKCλ showed peripheral gp135 staining (Fig. 6A,B). Importantly, the overexpression of dominant-negative aPKCλ antagonized the effect of the mLgl knockdown on the redistribution of apical proteins. Overexpression of dominant-negative but not wild-type aPKCλ in mLgl kd cells reduced the number of cells with peripheral gp135 staining to 47% (Fig. 6A,B). These results indicate that aPKC activity also plays a role in suppressing the disassembly of apical proteins during the depolarization process, and mLgl counterbalances this aPKC activity to facilitate apical protein disassembly. Taken together, these data support the notion that mLgl suppresses PAR-3-aPKC-PAR-6 complex activity to facilitate the disassembly of apical membrane proteins during depolarization of MDCK cells.

It should be noticed that VACs have characteristics of cortical apical membrane domains, that is, they display microvilli and apical protein markers and exclude basolateral proteins (Vega-Salas et al., 1987; Vega-Salas et al., 1988). In addition, VACs do not contain incorporated apical proteins but newly synthesized proteins (Gilbert and Rodriguez-Boulan, 1991). Thus, the above data raise the possibility that mLgl has an important role not only in the disassembly of but also on the re-establishment of de novo apical membrane domains in MDCK cells. To test this, we performed a collagen-gel overlay assay. When MDCK cells cultured sparsely on collagen-coated glass were overlaid with collagen gel, apical domains and apical junctions disappeared, after which de novo apical domains surrounded by apical junctions, called intercellular lumens, were re-established at the intercellular regions (Ojakian et al., 1997; Ojakian et al., 2001; Schwimmer and Ojakian, 1995).

After 24-48 hours of incubating the control cells with collagen gel, the gp135 staining was reduced at the free surface and was detected at the intercellular regions with ZO-1 ring-like staining as previously described (Fig. 7A,B, control cells) (Ojakian et al., 1997). The degree of E-cadherin staining at cell-cell contact regions were also reduced but not diminished completely, as described previously (Fig. 7A) (Ojakian et al., 2001). Interestingly, in mLgl kd cells, the disappearance of gp135 and ZO-1 staining from the free surface and their relocalization at the intercellular regions was markedly suppressed (Fig. 7A). Thus, endogenous mLgl is required for the re-establishment of apical domains in response to collagen-gel overlay.

By contrast, PAR-3 kd cells overlaid with collagen gel showed opposite phenotypes to mLgl-kd cells. Although PAR-3 kd cells induced the formation of intercellular lumens similarly to the control cells at 24 hours after collagen-gel overlay (Fig. 7B,C), at an early stage, the population of colonies with lumens was increased in PAR-3 kd cells. At 10.5 hours after collagen-gel overlay, more than 80% of the colonies of PAR-3 kd cells were positive for intercellular gp135 staining, whereas only 20% of the colonies of the control cells were positive for the staining (Fig. 7B,C). Thus, PAR-3 knockdown accelerated the collagen-gel-mediated re-establishment of apical domains, suggesting a negative role of endogenous PAR-3 in the re-establishment of apical domains. This is in contrast to the role of mLgl, supporting the notion that mLgl facilitates the collagen-mediated re-establishment of intercellular apical domains by counterbalancing PAR-3 function.

In vivo, apical-basal polarity of epithelial cells is subjected to dynamic regulation during tissue development and morphogenesis. The above results imply that mLgl exerts a similar role in the various physiological processes that reorganize apical-basal polarity. To evaluate this, we examined the effect of mLgl knockdown on MDCK cell cystogenesis. When MDCK cells were cultured in a collagen-I gel, the cells proliferated, acquired an apical-basal polarity and finally created cysts with a single lumen at a central position (O'Brien et al., 2002). It has been suggested that collagen-mediated interaction of cell with extracellular matrix (ECM) is important for the orientation of apical membrane polarity; perturbation of this step resulted in inversion of apical membrane polarity and induction of cell aggregate formation without an obvious lumen (O'Brien et al., 2001; Yu et al., 2005).

We first examined the effects of mLgl knockdown on the establishment of apical membrane domains during the early phases of cystogenesis (days 2-3). In this experiment, two independent mLgl kd cells (#1 and #2) and control cells were established by using an autonomous replication siRNA vector, pEB6-SUPER (see Materials and Methods). As described previously (O'Brien et al., 2002), in control cysts, gp135 staining was detected only at the inner surfaces of the cysts (Fig. 8), and concentrated F-actin staining was detected at these inner surfaces (Fig. 9A, days 2-3). Notably, in cysts of mLgl kd cells, gp135 staining was detected at the outer surfaces of the cysts (Fig. 8), and concentrated F-actin staining at inner surfaces was not detected (Fig. 9A, days 2-3). These data suggest that mLgl knockdown suppresses the distribution of apical proteins in the inner surfaces of cysts, resulting in the formation of cysts with inverted apical membrane polarity. It should be noticed that the basolateral proteins, β-1 integrin and E-cadherin, were detected in the entire cell periphery including the cell-cell contact regions and outer surfaces of cysts of mLgl kd cells (Fig. 8, Fig. 9A), suggesting that basolateral membrane polarity was not inverted in the cysts of mLgl kd cells. These data suggest that mLgl knockdown affects the orientation of apical membrane polarity but does not induce an inverted apical-basal polarity phenotype during the early phase of cystogenesis.

Moreover, subsequent cystogenesis and lumen formation were also perturbed in cysts of mLgl kd cells, resulting in the formation of cell aggregates without obvious lumens (Fig. 9A, days 5-7). Quantification analysis of day 7 cysts revealed that approximately 80% of cysts of both mLgl kd cells (#1 and #2) establish cell aggregates without obvious lumens, whereas only 20% of control cysts established cell aggregates (Fig. 9B). Concentrated F-actin staining and distribution of PAR-6β in the inner surfaces of the cysts were affected by mLgl knockdown (Fig. 9A,C), suggesting that the establishment of apical lumens was disorganized in cysts of mLgl kd cells. These data imply that the mLgl-mediated orientation of apical membrane polarity seems to be one of the important steps for lumen formation during cystogenesis. We further revealed that the overexpression of PAR-3 showed a similar phenotype to that of mLgl kd knockdown; approximately 80% of PAR-3-expressing cysts formed cell aggregates without lumens (supplementary material Fig. S4). These data further suggest that the regulated counterbalance between mLgl and PAR-3-aPKC-PAR-6 is also important for cystogenesis.

This study suggests that in some specific conditions, such as Ca²⁺-depletion and collagen-gel overlay, mLgl actively suppresses apical PAR-3-aPKC-PAR-6 complex to facilitate the redistribution of apical membrane domains in epithelial cells. Further, analysis of 3D-cultured cells suggests the importance of this mLgl-function for the orientation of apical membrane polarity to create cysts with lumen (Fig. 10).

Analysis of depolarization of MDCK cells suggests that mLgl-mediated disassembly of apical membrane proteins is mediated by the enhancement of internalization of apical proteins. A similar phenomenon has also been observed when the apical membrane polarity in cysts is reversed; that is, when MDCK cysts formed in a suspension culture are placed in a collagen gel, the pre-existing apical membrane domains at outer surface of the cysts are disassembled by internalization, after which new apical domains are formed at the central position of cysts (Wang et al., 1990). Thus, this study suggests that suppressive regulation of the apical PAR-3-aPKC-PAR-6 complex by mLgl is required for collagen-induced disassembly of peripheral apical membrane domains by internalization of apical proteins, through which cells can orient their apical membrane polarity. Interestingly, we found that only an mLgl-2 knockdown showed a similar effect to mLgl-1-mLgl-2 double-knockdown on apical protein localization in depolarized cells (data not shown). This is consistent with our previous observation that mLgl-2 preferentially coimmunoprecipitates with PAR-6 from MDCK cell lysates (Yamanaka et al., 2003), and suggests that the mLgl-2 isoform mainly regulates the apical domain in these cells.

In addition to apical domain disassembly, we showed here that the mLgl knockdown also partially suppressed the disassembly of tight junction proteins induced by Ca²⁺-depletion. Interestingly, we also found that the localization of one of the adherens junction proteins, E-cadherin, was also affected in mLgl kd cells; that is, E-cadherin accumulation in the interacellular region in depolarizing cells was suppressed in mLgl kd cells (supplementary material Fig. S5A). Thus, it seems that mLgl is involved in the regulation of localization of tight junction and adherens junction proteins in addition to that of apical membrane proteins (Fig. 10). Therefore, does mLgl regulate junction proteins and apical membrane proteins independently or interdependently? The following data suggests that the localization of apical proteins seems to be regulated by mLgl independently of tight junctions and adherens junctions. First, apical proteins were retained at the cell periphery even in highly depolarized cells, in which peripheral ZO-1 staining almost disappeared (Fig. 1). Second, PAR-3 overexpression - showing similar effects to mLgl knockdown on apical protein localization, or PAR-3 knockdown - showing opposite effects to mLgl knockdown, did not show clear effects on E-cadherin localization (supplementary material Fig. S5B,C). The mechanism of mLgl on the regulation of tight junctions and adherens junctions remains obscure and requires further investigations.

This study suggests that the suppression of apical PAR-3-aPKC-PAR-6 complex activity by mLgl is mediated through the suppression of the interaction of the aPKC-PAR-6 complex with PAR-3 or Cdc42. Thus, what are the roles of PAR-3 and Cdc42 in this complex? PAR-3 seems to serve as a scaffold of the aPKC-PAR-6 complex to the cortical region of epithelial cells because, in addition to the apical cortical proteins, a PAR-3 knockdown also accelerates the disappearance of aPKC and PAR-6 from the cell periphery in response to Ca²⁺-depletion (T.Y. and S.O., unpublished data). By contrast, Cdc42 seems to be involved in the regulation of aPKC activity because Cdc42 activates aPKC through PAR-6 in vitro (Yamanaka et al., 2001). Another possible role of Cdc42 is that it regulates the interaction of the aPKC-PAR-6 complex with another apical protein complex, the CRB3-Pals1 complex (Hurd et al., 2003; Lemmers et al., 2004; Wang et al., 2004). This complex is important for the establishment of apical membrane domains in epithelial cells (Cao et al., 2005; Roh et al., 2003), and Cdc42 has been shown to be required for the interaction between PAR-6 and Pals1 (Hurd et al., 2003). Thus, PAR-3 and Cdc42 seem to have important roles in localization, activity and protein-protein interaction of the aPKC-PAR-6 complex, which are affected by mLgl during the redistribution of apical proteins in epithelial cells (Fig. 10). The second important question is how mLgl regulates the interaction of PAR-6 with Cdc42 in addition to PAR-3. One possible explanation is that mLgl physically competes with Cdc42 for its interaction with PAR-6, because the CRIB region of PAR-6 is located just before the PDZ domain that interacts with mLgl (Yamanaka et al., 2003). Alternatively, mLgl might cause a conformational change in the CRIB-PDZ region of PAR-6 and, consequently, the dissociation of Cdc42 from PAR-6.

Previous studies have suggested that Rac1, which is activated by interaction of integrin with collagen, is also important for the orientation of apical membrane polarity by regulating laminin-ECM organization (O'Brien et al., 2001; Yu et al., 2005). Here, we showed that Cdc42 but not Rac1 was detected in PAR-6-IP from mLgl kd cells, suggesting that aPKC-PAR-6 interacts with Cdc42 rather than Rac1 under the control of mLgl to regulate apical membrane polarity. Consistent with this, recent observation suggests that, in addition to Rac1, Cdc42 is also involved in the regulation of apical membrane polarity in cysts, that is, the overexpression of a dominant-negative mutant of Cdc42 induces a polarity reversal phenotype in cysts (Rogers et al., 2003). Thus, in addition to the integrin-Rac1 pathway, mLgl-mediated inhibition of the interaction of the apical protein complex with Cdc42 seems to be one of several important pathways influencing the orientation of apical membrane polarity.

In epithelial tissues, the axis of each polarized epithelial cell must be aligned with the overall structure of the tissue to achieve a specific architecture. This study suggests the intriguing possibility that mLgl functions as a mediator of cell adhesion stimuli to create a corresponding apical membrane domain and lumen during epithelial tissue morphogenesis. This can be one of several mechanisms that integrate one cell polarization to epithelial tissue morphogenesis. Further studies are needed to obtain deeper insights into the mechanisms underlying the regulation of epithelial tissue morphology.

Adenoviral expression vectors encoding mouse wild-type aPKCλ (aPKCλ wt), a kinase-deficient dominant-negative mutant (aPKCλ dn) and LacZ generated with pAxCAwt were described previously (Suzuki et al., 2001). Adenoviral expression vectors encoding hemagglutinin (HA)-tagged mLgl-2 (HA-mLgl-2) and an empty vector generated with the Adeno-X expression system (Clontech, Palo Alto, CA) were described previously (Yamanaka et al., 2003). cDNA for T7-tagged full-length PAR-3 (1-1337 aa) was subcloned into vector pEB6-CAG, an Epstein-Barr virus (EBV)-based extrachoromosomal vector carrying a replication origin (oriP) and a replication initiation factor (EBNA-1) that are sufficient for autonomous replication in human and canine cells (Tanaka et al., 1999). pEB6-SUPER vector was established by replacing the CAG promoter region of pEB6-CAG vector with the H1 promoter of pSUPER vector as described previously (Suzuki et al., 2004).

Rabbit anti-PAR-6β polyclonal antibody (pAb) (BC31AP), rat anti-PAR-6β pAb (BCR12AP), rabbit anti-mLgl-2 antibodies (N13AP and 100-278 2-3AP), rabbit anti-mLgl-1 pAb (C-2 AP), rabbit anti-PAR-3 pAb (C2-3AP), and rabbit anti-EMK (PAR-1b) pAb (EM1) have been described previously (Yamanaka et al., 2003; Akimoto et al., 1994; Izumi et al., 1998; Suzuki et al., 2004). Mouse anti-gp135 monoclonal antibody (mAb) (3F2) was generously provided by J. D. Nelson (Stanford University, CA). The following commercially available antibodies were also used: mouse anti-aPKCλ (ι) mAb, mouse anti-E-cadherin mAb, mouse anti-Cdc42 (C70820) mAb (Transduction Laboratories, KY); mouse or rabbit anti-ZO-1 mAb (ZYMED Laboratories Inc., CA); rabbit anti-T7 (Omni) pAb, rabbit anti-ezrin pAb, rabbit anti-aPKC (C20) pAb (Santa Cruz, CA); rabbit anti-HA pAb (Bethyl); rat anti-ZO-1 mAb (Chemicon); mouse anti-β-actin mAb, rat anti-uvomorulin (E-cadherin) mAb (Sigma); rabbit anti-PAR-3 pAb, mouse anti-Rac mAb (05-389: Upstate); rat anti-HA mAb (Roche); and rat β1-integrin mAb (A2B2: Hybridoma bank).

Madin-Darby canine kidney (MDCK) II cells were grown in Dulbecco's modified essential medium (DMEM) containing 10% fetal calf serum (FCS), penicillin and streptomycin, at 5% CO₂ atmosphere and constant humidity. For the Ca²⁺-depletion assay, cells were seeded on Transwell™ filters (12 mm diameter; Corning Coaster Corp., Cambridge, MA) with a pore size of 0.4 μm (3×10⁵ cells/cm²) and grown for 1 or 2 days to produce a confluent monolayer. Cells were washed with PBS twice and incubated in a low-Ca²⁺ (LC) medium containing 5% FCS, 3 μM Ca²⁺, penicillin and streptomycin for 20 hours. The Ca²⁺-switch assay was performed by replacing the LC medium with a normal growth medium for these cells. Adenovirus infection was performed as described previously (Suzuki et al., 2001). Briefly, confluent cells were pre-incubated with LC medium for 2 hours; cells were then incubated with 200 μl of the appropriate virus solution, diluted to 3×10⁸ pfu/ml in LC medium, for 2 hours. LC medium was prepared as follows: 25 ml of FCS was incubated with 1.25 g of analytical-grade Chelex 100 resin (BIA-RAD) at 4°C for 1 hour to chelate Ca²⁺ ions. Then, FCS was added to 500 ml of Ca²⁺-free DMEM (GIBCO), after which CaCl₂was added to a final concentration 3 μM, together with penicillin-streptomycin and glutamine.

pSUPER vector was generously provided by R. Agami (The Netherlands Cancer Institute, Amsterdam, The Netherlands) (Brummelkamp et al., 2002). pSUPER based vectors containing different antibiotics resistant genes were established: pSUPER-neo, which was established using the pEGFP-N3 vector in which the sequence for CMV-EGFP was replaced by the sequence for the H1 promoter of the pSUPER vector; and pSUPER-puro, which was established by insertion of the sequence for PGK and PAC into XhoI site of the pSUPER vector. Partial cDNA sequences of canine homologues of mLgl-1, mLgl-2 and PAR-3 were obtained by reverse transcriptase (RT)-PCR of the total RNA of MDCK II cells using a one-step RT-PCR kit (Qiagen) and cloned into the TOPO vector (Invitrogen). To generate mLgl-1-mLgl-2 double knockdown (kd) MDCK cell clones (24-15 and 24-16), mLgl-2 kd cell lines (2-10 and 2-13) were first established by transfection with pSUPER-puro encoding the mLgl-2 shRNA sequence (5′-GATGAGAGTTTCACACTGC-3′) and selection with puromycin. Then, clone 2-10 was further transfected with pSUPER-neo encoding the mLgl-1 shRNA sequence (5′-GGTGGCACTCTGCAAGTAC-3′) and selected with G418. Control clones (21-1 and 21-7) were established similarly by using pSUPER-puro and pSUPER-neo vectors with a negative control sequence (5′-CAGTCGCGTTTGCGACTGG-3′). Neomycin-resistant PAR-3 kd clones (13-32 and 13-33) were established by transfection with pSUPER-neo encoding the PAR-3 shRNA sequence (5′-GACAGACTGGTAGCAGTGT-3′) and selection with G418. Control clones (11-10 and 25-2) were established similarly by using pSUPER-neo with the negative control sequence. The puromycin-resistant PAR-3 kd clone (25a) was established by the cotransfection with pSUPER-puro encoding the PAR-3 shRNA sequence (5′-CATGGAGATGGAGGAATAC-33′) and pSUPER-puro encoding another PAR-3 shRNA sequence (5′-GAACAGGATGAGGATGGGA-33′). Control clone (1-5) was established similarly using pSUPER-puro vector with the negative control sequence.

The collagen-gel overlay assay was performed as previously described (Ojakian et al., 1997). Briefly, MDCK II cells were seeded on coverslips coated with collagen-I (2×10⁴ cells/cm²). After 1 day, cells were overlaid with collagen-I gel containing 10% FCS in DMEM.

MDCK cell cyst formation was performed as described previously (O'Brien et al., 2001). Briefly, MDCK cells were trypsinized and suspended with ice-cold water containing 2 mg/ml calf skin type-I collagen (Koken, Tokyo, Japan), DMEM, 20 mM HEPES (pH 7.4), and 5% FBS. The resulting suspension was placed on transwell membrane filters (10 mm in diameter, Corning Costar). After the formation of a gel at 37°C, growth medium was added to the wells. The cells were incubated for several days to form cysts in the gel. Two independent mLgl kd cells, #1 and #2, were generated by introducing pEB6-SUPER with mLgl-1 shRNA into the mLgl-2 kd clone 2-10 and 2-13. Control cells were generated by introducing pEB6-SUPER with the negative-control sequence into control clone 1-5. PAR-3-expressing cells were generated with pEB6-CAG-PAR-3 vector as described previously (Suzuki et al., 2004).

MDCK II cells were cultured on 10-cm dishes. Cells were suspended in lysis buffer containing 50 mM Tris-HCl at pH 7.4, 150 mM sodium chloride, 10% glycerol, 1% Triton X-100, 1.5 mM magnesium chloride, 1 mM EGTA, 10 mM sodium fluoride, 10 mM Na⁴P²O⁷, 1 mM Na³VO⁵, 1 mM phenyl methylsulphonyl fluoride, 10 μg/ml leupeptin and 16 μg/ml aprotinin. After centrifugation at 17,610 g for 30 minutes, the supernatants were subjected to immunoprecipitation with 2 μg of antibody, followed by SDS-PAGE and western blotting as described previously (Suzuki et al., 2001).

MDCK cells were fixed with 2% paraformaldehyde and stained as described previously (Suzuki et al., 2001). Secondary antibodies were Alexa Fluor 488-conjugated goat anti-rabbit or goat anti-mouse antibodies, Alexa Fluor 647-conjugated goat anti-mouse antibody (Molecular Probes Inc., Eugene, OR), and Cy3-conjugated goat anti-rabbit or goat anti-rat antibodies (Amersham Bioscience). Rhodamine-phalloidin was used to visualize filamentous actin (Molecular Probes). Samples were mounted with Vectashield (Vector Laboratories, Inc.) and viewed using a confocal microscope system (LSM 510; Carl Zeiss, Germany) or a fluorescence microscope (BX40, Olympus) equipped with a CCD camera (Princeton Instruments, Trenton, NJ).

We thank Y. Miwa (Tsukuba University, Japan) for the EBV-based vector, R. Agami (The Netherlands Cancer Institute, Amsterdam, TheNetherlands) for pSUPER, and J. D. Nelson (Stanford University, CA) for the gp135 antibody. This work was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (S.O.), the Japanese Society for the Promotion of Science (S.O.) and the Japanese Science and Technology Agency (S.O.).