Atypical protein kinase iota (PKCι) is a key organizer of the apical domain in epithelial cells. Ezrin is a cytosolic protein that, upon activation by phosphorylation of T567, is localized under the apical membrane where it connects actin filaments to membrane proteins and recruits protein kinase A (PKA). To identify the kinase that phosphorylates ezrin T567 in simple epithelia, we analyzed the expression of active PKC and the appearance of T567-P during enterocyte differentiation in vivo. PKCι phosphorylated ezrin on T567 in vitro, and in Sf9 cells that do not activate human ezrin. In CACO-2 human intestinal cells in culture, PKCι co-immunoprecipitated with ezrin and was knocked down by shRNA expression. The resulting phenotype showed a modest decrease in total ezrin, but a steep decrease in T567 phosphorylation. The PKCι-depleted cells showed fewer and shorter microvilli and redistribution of the PKA regulatory subunit. Expression of a dominant-negative form of PKCι also decreased T567-P signal, and expression of a constitutively active PKCι mutant showed depolarized distribution of T567-P. We conclude that, although other molecular mechanisms contribute to ezrin activation, apically localized phosphorylation by PKCι is essential for the activation and normal distribution of ezrin at the early stages of intestinal epithelial cell differentiation.
The sorting mechanisms responsible for the polarization of membrane proteins in epithelial cells, which result in the segregation of apical and basolateral domains, have been extensively studied (Nelson, 2003; Zegers et al., 2003). However, our understanding of how cytosolic proteins interface the cytoskeleton with the membrane in the form of submembrane scaffolds is still incomplete. Ezrin is a polarized extrinsic membrane protein, and although it is also expressed in other tissues, ezrin is apical in simple epithelial cells (Bretscher et al., 2002). It connects actin filaments either directly to membrane proteins, such as CD44 (Martin et al., 2003), or indirectly via NHERF (EBP50) proteins, which in turn bind membrane proteins by means of two PDZ domains (Bretscher et al., 2000). Among others, cystic fibrosis transmembrane regulator and NHE-3 are apical transmembrane proteins that interact with these PDZ domains. More importantly, ezrin recruits PKA and thus enables cAMP-mediated regulation of these membrane proteins (Dransfield et al., 1997; Kurashima et al., 1999; Sun et al., 2000), a key pathophysiological event in secretory diarrheas, such as cholera (Kunzelman and Moll, 2002).
Ezrin is synthesized in a `dormant' (inactive) configuration in which the N-terminal domain (N-ERMAND) binds the C-terminal domain (C-ERMAND) mutually blocking their binding capacities to other molecules (Bretscher, 1999). Phosphorylation of T567, as well as interactions with phosphatidylinositol-4,5-bisphosphate (PIP2), open the dormant form, freeing the C-ERMAND to bind actin and the N-ERMAND to bind either NHERF or certain membrane proteins. This configuration is known as the `active' form of ezrin (Fievet et al., 2004). In normal intestinal epithelial cells, ribosomes are concentrated in the perinuclear region, away from the subapical cytoplasm (Kasai et al., 2003). More importantly, in situ hybridization data on the distribution of ezrin mRNA in enterocytes suggests a diffuse, perhaps supranuclear distribution rather than apical localization of ezrin mRNA (Barilá et al., 1995). In tall, slender cells such as enterocytes, it means that newly synthesized ezrin must travel several micrometers to reach its final destination under the apical membrane. It has been generally assumed that apical localization of ezrin is due to binding of the active form to its cognate partners. However, many of those partners exist in the basolateral domain: CD44, for example, is basolateral in intestinal cells (Gouyer et al., 2001). Actin is also extensively present in the basolateral submembrane region (Shigeta et al., 2003; Oriolo et al., 2007). Therefore, it is reasonable to speculate that ezrin must be translocated from the point of synthesis to the subapical region in the dormant configuration and become activated locally, in the immediate vicinity of its normal apical partners. We presented evidence that the dormant form can be recruited by apical intermediate filaments (Wald et al., 2005), but in this model, a second step of local apical activation is necessary, because activation of dormant ezrin before reaching the apical cytoskeleton would prevent the recruitment step. Therefore, we investigated which kinase, if any, might specifically activate ezrin under the apical domain of epithelial intestinal cells.
Several kinases have been shown to phosphorylate ezrin T567, including Rho kinase (Matsui et al., 1998; Haas et al., 2007), PKCθ (theta) in vitro (Simons et al., 1998) and Akt2 (Shiue et al., 2005). In human skin squamous carcinoma cells (A431), PKCθ has been shown to phosphorylate and participate in the localization of ezrin to microvilli (Stapleton et al., 2002). In addition, phosphoinositides are considered essential to activate ezrin (Yonemura et al., 2002). Rho and ROCK are active at the basolateral domain in epithelial cells (Van Aelst et al., 2002; Walsh et al., 2001; Takaishi et al., 1997). If Rho activated ezrin in simple epithelia, one would predict a basolateral activation. In fact, Yonemura and co-workers (Yonemura et al., 2002) demonstrated that ezrin activation is independent of Rho in these cells. Likewise, only small amounts of PKCθ may be localized to the tight junctions (Banan et al., 2004). Finally, Akt2 is enriched in the apical membrane of intestinal cells in vivo and in CACO-2 (human colon carcinoma) cells in culture. However, its expression in the intestine is circumscribed to the distal part of the villi, where cells are terminally differentiated (Li et al., 2004). In the intestine, stem cells are localized at the bottom of the crypts, where cell division occurs, and cells rapidly move toward the villi, where they desquamate. The crypt-villus axis, thus, shows the life cycle and differentiation processes (Ahuja et al., 2006): the deeper that cells express a protein within the crypt, the earlier the expression. Ezrin is expressed and activated in the crypts, albeit at a lower rate than in the villus. Expression of Akt2 near the tip of the villus rules out its possible involvement in activating ezrin, at least in the earliest stages of enterocyte differentiation. Thus, we looked at other possible kinases as candidates for ezrin activation. Notably, T567 is localized within a highly conserved PKC consensus phosphorylation site: RDKYKT*LRQIRQ, which contains two of the most common PKC phosphorylation motifs together: (S/T*)x(K/R) and (K/R)xx(S/T*) (Pearson and Kemp, 1991). In addition, at least two PKC isoforms are known to phosphorylate ezrin T567 under different circumstances: PKCα in migrating cells (Ng et al., 2001) and, as mentioned above, PKCθ. PKC also is known to phosphorylate NHERF, one of the apical partners of ezrin (Fouassier et al., 2005).
Among other PKC isoforms, atypical PKC (PKCι/λ, hereafter referred to as PKCι) has been extensively shown to be a key determinant of apical polarity, which is conserved in worms, arthropods and mammalians. It is localized to the apical domain of simple epithelial cells (reviewed by Suzuki and Ohno, 2006). However, little is known about its targets. Therefore, this work was undertaken to test the hypothesis that PKCι locally (i.e. apically) phosphorylates ezrin at T567 and that this activation is essential for the apical localization of ezrin in intestinal epithelial cells.
Expression of active PKCι and phosphorylation of T567 in ezrin appear together during the differentiation of intestinal epithelial cells
In the intestine, cells replicate at the bottom of the crypts and move toward the tip of villus. Therefore, the crypt-to-villus axis can be equated to time of differentiation (Ahuja et al., 2006). To assess the timing of ezrin phosphorylation, we analyzed correlation in the expression of active (T555-P) PKCι and the phosphorylation of ezrin in T567 in 36 longitudinal sections of crypts extending to the bottom of the gland. In all cases, PKCι was apical (Fig. 1A) and T555-P PKCι signal codistributed with T567-P ezrin signal within the crypt. In some cases, PKCι expression extended slightly deeper (∼2-3 rows of cells) in the crypt than the T567-P ezrin signal (Fig. 1C). In other words, the expression of active PKCι is synchronous (or shortly precedes) with the onset of ezrin phosphorylation (Fig. 1E, E', arrowhead). Conversely, Akt2, another possible candidate involved in phosphorylation of ezrin T567 (Shiue et al., 2005), was found expressed in the villus, especially around the tips, but the signal was undetectable in the crypts in the same sections (supplementary material Fig. S1). This result does not demonstrate that PKCι is the kinase involved in the phosphorylation of ezrin T567, but it indicates that PKCι is expressed at the right time and place in vivo to fulfil this function under physiological conditions in the early stages of differentiation of intestinal cells.
Atypical PKCι can phosphorylate ezrin at T567 in vitro and in cells that lack endogenous ezrin phosphorylation
For in vitro studies, 6×His-tagged human ezrin was expressed in Sf9 insect cells by means of baculovirus infection. It has been shown that Sf9 lack the machinery to phosphorylate ezrin at T567, and that human (h)-ezrin expressed in these cells is in the dormant configuration (Martin et al., 1997; Wald et al., 2005). H-ezrin was then purified in Ni2+ columns and challenged with recombinant PKCι and [γ-32P]ATP (Fig. 2A) or cold ATP (Fig. 2B). In both cases, ezrin was found to be a target of PKCι in vitro. More importantly, immunoblotting with anti-T567-P antibody indicated that the T567 site specifically is phosphorylated by PKCι in vitro (Fig. 2B). A similar experiment was performed, but the phosphorylated ezrin and the corresponding nonphosphorylated control were pulled down with phalloidin-stabilized F-actin. PKCι-mediated phosphorylation increased the ezrin pulled down by actin (Fig. 2C), indicating that this phosphorylation resulted in the exposure of at least a fraction of the C-ERMAND domains.
To study ezrin phosphorylation by PKCι within eukaryotic cells, we expressed these proteins in Sf9 cells by means of baculovirus infection. When expressed separately, h-ezrin or h-PKCι showed a diffuse distribution in the cytoplasm, along with the endogenous F-actin (Fig. 2D, top two rows). This broad cytoplasmic F-actin distribution was observed even in cells infected with a wild-type baculovirus. In addition, empty virus infections showed that under the current infection protocol, the overall size and shape of the cells were not substantially changed (supplementary material Fig. S2). When Sf9 cells were co-infected with h-ezrin and h-PKCι, F-actin and h-ezrin redistributed to the cortex (Fig. 2D, arrows). Moreover, the free surface of Sf9 cells was relatively smooth under SEM when separately infected with virus expressing either protein separately, but 40% of the cells doubly infected with both viruses showed abundant protrusions and blebs (Fig. 2E), consistent with cortical redistribution of F-actin (Fig. 2D). These protrusions resembled the membrane ruffles induced by expression of T567D (active) ezrin (Gautreau et al., 2000). In similar experiments, cell extracts were purified in Ni2+ columns, and analyzed by immunoblot. Co-expression of PKCι and ezrin resulted in a substantial increase of T567-P signal (Fig. 2F). A faint band observed with the T567-P antibody, in the absence of PKCι expression, is consistent with a low-affinity recognition of the nonphosphorylated epitope by this antibody, as demonstrated by dephosphorylation of active ezrin with λ-phosphatase (not shown). It was also noted that the morphology of the cell surface in doubly infected Sf9 cells was also remarkably reminiscent of the appearance of apoptotic cells. To rule out this possibility, we analyzed the genomic DNA of Sf9 cells noninfected, or infected with viruses expressing PKCι or ezrin separately, or doubly infected. A positive control for apoptosis was performed by a long incubation in hydrogen peroxide, which caused the DNA smearing typical of apoptosis. Evidence of apoptotic degradation of DNA, however, was not found in infected cells (Fig. 2G). Likewise, staining for apoptotic cells was positive for cells incubated in hydrogen peroxide, but not for cells infected with baculovirus expressing the proteins under study (Fig. 2H). Therefore, we concluded that the redistribution of F-actin and ezrin in doubly infected Sf9 cells and the remodeling of the surface are probably the consequence of PKCι-mediated phosphorylation of ezrin in T567.
Atypical PKCι knockdown abrogates phosphorylation of endogenous ezrin T567 in intestinal CACO-2 cells
The normal endogenous expression of PKC isoforms in CACO-2 cells was studied in a specialized facility by immunoblot using a standardized panel of anti-phosphoepitope antibodies that recognize the active form of each kinase. The advantage of this study is that the normalized results (Table 1) enable semiquantitative comparisons of the relative amounts of different kinases. Because CACO-2 cells start differentiating and polarizing at around day 5 after seeding (Pinto et al., 1983), the study was performed in CACO-2 cell extracts from confluent monolayers 4 and 8 days after seeding. PKCι was found to be the most abundant of all PKC isoforms in these cells, both in the undifferentiated and the early differentiation (8 days) stage. It was nearly tenfold more abundant than any of the other two PKC isoforms known to be localized to the apical domain (PKCα and ζ). Furthermore, its level of expression increased from the nondifferentiated stage to early differentiation, even though the rate of increase was much more pronounced for PKCα (Table 1). To determine whether PKCι binds to ezrin in vivo, we first analyzed whether they coimmunoprecipitated. Most of endogenous ezrin is incorporated in insoluble scaffolds, and, therefore, not available for coimmunoprecipitation. To increase the available soluble ezrin we transfected CACO-2 cells with the full-length ezrin ORF in a pcDNA vector. Nontransfected monolayers were used as controls (Fig. 3A, ezrin transfection –), and the immunoprecipitation was controlled with nonimmune mouse IgG (Fig. 3A, ip –). The cells were extracted in 0.5% Triton X-100, and sample of the extracts was analyzed by immunoblot with anti-PKCι antibody (Fig. 3A, input). The eluates of the ezrin immunoprecipitation were analyzed by immunoblot with anti-PKCι antibody, and the same membrane was then reprobed with anti-ezrin antibody. In cells overexpressing ezrin, PKCι co-immunoprecipitated with ezrin (Fig. 3A, *). Considering its relative abundance and direct interaction with ezrin, PKCι was a good candidate for knockdown studies.
To knockdown PKCι, five sequences expressing shRNA against human PKCι were obtained from a commercial source in lentivirus vectors, which also confer puromycin resistance when transduced into mammalian cells. In preliminary experiments, only one was found to be highly effective in knocking down PKCι in CACO-2 cells, although all the lentiviruses displayed very high levels of infection (typically >85% of cells). A subclone of CACO-2 cells, CACO-2 C2BBe was used because of its homogeneity, early and robust polarization and differentiation (Peterson and Mooseker, 1992). The cells were transduced with lentivirus carrying a control empty vector or the same vector with the shRNA sequence under a pol III promoter. All the experiments were performed within the first three passages after transduction and puromycin selection, as the knockdown effect of shRNA was found to fade thereafter, even when the cells were kept in puromycin selection. The knockdown was >95% effective 3-7 days after plating the cells, and typically ∼80% effective at day 10 (Fig. 3B; load controls, Fig. 3C). Transduced cells had only negligible rates of apoptosis (<0.1%, supplementary material Fig. S3), although knockdown cells grew at a slower rate than cells transduced with empty virus (approximately half the doubling time, not shown). For this reason, they were routinely passaged at twice the seeding density to match the confluence time of the cells transduced with empty virus.
When observed by immunofluorescence in cells transduced with empty vector virus and cultured for 8 days, PKCι (total protein) was found in a continuous apical layer, slightly more concentrated in the region of the tight junctions (Fig. 3D,F). Using an antibody against T555-P PKCι (active form), a similar pattern was observed, but the difference between the signal in the vicinity of tight junctions and the rest of the apical domain was more evident (Fig. 3E,G), suggesting that active PKCι is more concentrated near cell-cell contacts. Expression of anti-PKCι shRNA resulted in a drastic reduction of the fluorescence signal, compatible with the immunoblot results. The signal away from the tight junctions fell to background levels for both total PKCι protein and T555-P epitope. A small amount of remnant PKCι after knockdown was observed in the vicinity of the tight junction region (Fig. 3H-K).
In parallel monolayers, we also analyzed the expression of ezrin (total protein) and its phosphorylation on T567 upon knockdown of PKCι. The extent of the reduction in T567-P signal was similar to the extent of decrease in PKCι (Fig. 4A). Interestingly, there was also a decrease in the steady-state level of total ezrin, although not as great as the decrease in T567-P signal (35-55%) (Fig. 4B). The ratios of p-T567 signal/total ezrin signal were used to compare the samples from both sets of cells. In all cases, the ratios were lower in the samples from knockdown cells, although they slowly increased with time (supplementary material Fig. S4A). Ezrin mRNA levels were assessed semiquantitatively by RT quantitative PCR, and did not show significant differences among cultures transduced with empty virus and cultures expressing anti-PKCι shRNA (supplementary material Fig. S4B), suggesting that the differences at the protein level are not transcriptional. The same results were observed at the immunofluorescence level under the same conditions described in Fig. 3. The remnant ezrin protein and the T567-P signals were found localized around the tight junction region (Fig. 4D,F,G-I), precisely where remnant PKCι was found (Fig. 3). Indeed, in X-Z confocal reconstructions of PKCι-knockdown cells, the apical layer of T567-P was found to be discontinuous and the remnant signal often clustered in the vicinity of ZO-1 (tight junction) signal (Fig. 4G-I).
Functional effects of the decrease of active ezrin upon knockdown of PKCι
Ezrin is a scaffolding protein, and few functions can be directly associated with its knockdown. From the phenotype of ezrin knockout mice, we know that intestinal cells lacking ezrin display shorter and fewer microvilli (Saotome et al., 2004). Also, ezrin is known to recruit the regulatory subunit of PKA to the apical domain (Sun et al., 2000). We tested both functions. When the surface of the cells was analyzed by SEM, CACO-2 C2BBe cells expressing anti-PKCι shRNA showed a decrease in number and size of microvilli (Fig. 4K, shRNA) compared with control cells under identical culture conditions and infected with mock lentiviral particles (Fig. 4J), consistent with the microvillus phenotype of ezrin-null enterocytes. This result was also confirmed by phalloidin fluorescence, which showed disorganization of the apical actin cytoskeleton (supplementary material Fig. S5A,B,E,F) with little or no effect on the basal actin organization in stress fibers (supplementary material Fig. S5C,D). In addition, the regulatory subunit II of PKA was localized in 6-day CACO-2 BBE cultures infected with `empty' lentiviral particles. It was found to be concentrated under the apical domain in most cells (Fig. 4L, arrows). By contrast, in cells depleted in PKCι, it was depolarized in all the cells (Fig. 4M, shRNA), confirming the need for active ezrin to recruit PKA.
Expression of a dominant-negative PKCι mutant inhibits phosphorylation and a constitutively active PKCι mutant results in depolarized phosphorylation of ezrin T567
Dominant-negative K274W PKCι and constitutively active A120E PKCι have been described before (Spitaler et al., 2000; Lim et al., 1999). Both mutants were 6×His and V5-tagged and expressed by transient transfection in cells before day 5 of culture because the cells become resistant to transfection when they polarize. In transfected cells, V5 signal for both mutants filled the cytoplasm and did not show the peculiar apical localization of the endogenous protein (Figs 5 and 6, red). In immunoblot experiments (not shown), we assessed the stability of the transfected protein by purifying the protein extracted from monolayers transfected for 1, 2 and 3 days through Ni2+ columns. Immunoblots using the V5-tag antibody showed that the K274W mutant was present for 3 days after transfection, whereas the A120E mutant was only present for 2 days after transfection. When reprobing the same blots with anti-T555 PKCι antibody, we found that both mutants were phosphorylated in T555, and are therefore likely to have the appropriate conformation 1 day after transfection. Later, the ratio of phosphorylated to total protein decreased, and thus, the kinase was unlikely to display the correct active-loop conformation (Newton, 2003). Accordingly, all the transient transfection experiments were designed for readouts 24 hours after transfection.
CACO-2 C2BBe cells at day 5 of culture showed variable levels of T567-P ezrin signal (Fig. 5A,C). Cells transfected with V5-K274W PKCι also showed variable levels of V5 expression. Examples of low (cell 1, Fig. 5B), medium (cell 4, Fig. 5D) or high levels of expression (cells 2 and 3, Fig. 5B) are shown in Fig. 5. In all cases, transfected cells showed levels of T567-P ezrin signal comparable with those of the lowest T567-P signal in nontransfected cells (Fig. 5A,C). Because of the variability in the T567-P ezrin signal in nontransfected cells, a qualitative scoring of 121 transfected cells was conducted. In 81% of transfected cells, T567-P signal was scored as lower than the nontransfected neighbors, whereas the rest displayed similar levels of signal, because nontransfected cells also had low levels of expression. Not a single case was observed of a transfected cell expressing more T567-P than a neighboring nontransfected cell (Table 2). The difference between the score distribution in control and transfected cells was highly significant (P<0.01, Table 2), reflecting the fact that all the transfected cells had the lowest levels of T567-P signal in the sample. Therefore, expression of K274W PKCι effectively abrogated phosphorylation of ezrin T567.
Wild-type CACO-2 cells express less ezrin at day 5 than the C2BBe clone, yet most cells displayed some apical T567-P ezrin signal (Fig. 6A). Therefore, they were used to test the effect of the constitutively active A120E PKCι mutant. Nearly all the cells transfected at day 4 displayed a similar phenotype. The cells appeared as flat disks, disconnected from the filter, and forming a second layer on top of nontransfected cells (Fig. 6C,D). These transfected cells were negative for apopercentage staining (supplementary material Fig. S6A-C) and excluded an aldehyde-fixable extracellular marker (Lucifer yellow CH, supplementary material Fig. S6D-G). More importantly, when falling off the monolayer in the supernatant, they were viable for reseeding (supplementary material Fig. S6H,I), indicating that they are neither apoptotic nor necrotic, but viable cells that possibly display different adhesive or motility properties than their nontransfected neighbors. Surprisingly, confocal images at the apical surface did not show an increase in T567-P signal (Fig. 6A,C, arrows), although in nonconfocal microscopy the cells showed clearly more T567-P signal than their neighbors (not shown, see example in Fig. 6I,J).
The T567-P signal was disperse in the cytoplasm and excluded from the nucleus (Fig. 6C). The effect was milder in the same cells transfected at day 2 and fixed on day 3, presumably because the total expression of ezrin is lower. In fact, many cells did not show any T567-P ezrin signal at all (Fig. 6E). At this early stage, A120E PKCι-transfected cells did not detach from the filter, but approximately 30% of them showed T567-P signal in the basal sections (Fig. 6F, arrows). In these cells, X-Z sections showed also a nonpolarized distribution of the mutant PKCι and T567-P signal (Fig. 6G,H, arrowheads) contrasting with the faint but well-polarized T567-P signal in some of their neighbors. As described before (Fig. 4), CACO-2 C2BBe cells expressing anti-PKCι shRNA showed little T567-P signal, mostly localized near the tight junction region (Fig. 6J, nontransfected cells). Transfection with A120E PKCι (Fig. 6I, red) for 24 hours rescued T567-P signal (Fig. 6J, transfected cells). However, as in previous examples of CACO-2 cells, the distribution of this signal was not polarized, largely localized below the nucleus (Fig. 6M, arrowhead) in transfected cells, once again in steep contrast with the highly polarized signal in their nontransfected neighbors. The phenotype of cells detaching from the filter and forming a second layer as described in Fig. 6C,D, was also very frequent in these cells (60-70%) (results not shown). In brief, expression of the constitutively active PKCι mutant increased the phosphorylation of T567-P ezrin, but in nonpolarized fashion, consistent with the nonpolarized distribution of the overexpressed PKCι.
Various kinases can phosphorylate ezrin at T567, a site important in ezrin activation, in different cells and even under different stages of differentiation. Our results indicate that atypical PKCι is essential to phosphorylate T567 in the early stages of intestinal cell differentiation. Not only it is expressed in crypts in vivo immediately before the onset of ezrin T567 phosphorylation, but it also phosphorylates this site in vitro, in cells that normally do not activate ezrin by overexpression of a PKCι constitutively active form. Furthermore, knockdown of endogenous PKCι abrogated most ezrin phosphorylation of T567. More importantly, the apical localization of the kinase is essential for the apical localization of T567-P ezrin, because overexpressed delocalized kinase results in diffuse unpolarized phosphorylation of T567 ezrin.
The conformational switch of ezrin from the dormant to the active form is a complex phenomenon, in which phosphorylation of T567 is necessary, although not sufficient (Chambers and Brestcher, 2005). The interaction of ezrin with PIP2 is also important for ezrin activation (Yonemura et al., 2002). An apical gradient of PIP2 versus basolateral PIP3 seems to play an essential role in epithelial polarization (Martin-Belmonte et al., 2007). Therefore, apical PIP2 is likely to play a synergistic role with T567 phosphorylation for the apical localization of ezrin. However, the PKCι knockdown data presented here suggest that phosphorylation of T567 is still necessary, because in the absence of the kinase there is an overall decrease in total ezrin apical localization. Moreover, most of the remnant ezrin distributes in the same region where the remnant PKCι is still active (Fig. 4). An alternative possibility that cannot be ruled is that PKCι knockdown might affect the apical PIP2 gradient itself. Although, to our knowledge, there is no published data supporting this possibility, it is worth noticing that the phosphatase and tensin homolog (PTEN), that maintains the apical PIP2 gradient (Martin-Belmonte et al., 2007) is physically associated with the Bazooka (PAR-3) complex, at least in arthropods (von Stein et al., 2005). Such a possibility will certainly deserve further investigation. Finally, it is also important to consider that although the PIP2 gradient seems to be important in general in epithelial cells, it is uncertain that it is active in CACO-2 cells in particular, since these cells show very small amounts of PTEN associated with the membrane (Li et al., 2004).
Although the apical localization of ezrin seems to obey to a combination of synergistic mechanisms during the assembly of newly synthesized ezrin molecules, the final apical concentration may also respond to the stability of active ezrin. In fact, the results in this work must be also understood from the perspective of how stable forms of ezrin accumulate. Little is known about the relative stability of ezrin in its various possible stages (e.g. dormant vs active). However, published data indicate that once assembled in its normal scaffold, active ezrin is very stable (half life »1 day) whereas conditions that destabilize the apical scaffold, such as oxidative stress, cause a steep increase in ezrin degradation mediated by proteasomes (Grune et al., 2002). Therefore, it is likely that the fully activated, membrane and actin-bound ezrin can accumulate, whereas the intermediaries cannot. Such a scenario would explain the lack of ezrin in the EBP50-null mouse (Morales et al., 2004). In the absence of a conditional PKCι-knockout mouse, we can only speculate that activation at the apical domain stabilizes ezrin and enables the apical ezrin accumulation characteristic of the differentiated brush border. In fact, it is possible that further phosphorylation and accumulation may be mediated by Akt2 in the villi (Shiue et al., 2005). Similar considerations may apply when attempting to explain the distribution of the remnant PKCι after knockdown (Fig. 3H,K). Although no data is available on turnover of PKCι pools in the apical domain, it is known that the Par-3–Par-6–PKCι complex interacts with the Crumbs–PatJ–Pals–ZO-3 complex at the tight junction (reviewed by Suzuki and Ohno, 2006). Whether a different scaffold is used away from the tight junction is unknown, but it is conceivable that the turnover of PKCι is different when the kinase is associated with the tight-junctional complexes.
The phenotype of CACO-2 cells overexpressing A120E (constitutively active) PKCι is fully consistent with previous reports showing `apicalization' of the basolateral domain with overexpressed PKCι. This effect is probably due to the phosphorylation of other targets such as PAR-1b (Suzuki et al., 2004) and not to the phosphorylation of ezrin. This interpretation is also consistent with our observation that the effect of A120E-PKCι expression changed dramatically with the differentiation of CACO-2 cells. Key to the understanding of these experiments was the fact that overexpressed PKCι appeared homogeneously distributed in the cytoplasm, rather than accumulated under the apical domain. Many otherwise well-localized proteins become delocalized upon overexpression, possibly because the compartmentalization mechanisms become overwhelmed, as we have shown before for ezrin overexpression (Wald et al., 2005). For the purpose of this study, there are two features of the constitutively active PKCι overexpression phenotype that must be highlighted: (1) transfected cells did not show an increase in apical T567-P ezrin signal; and (2) such an increase was observed in the cytoplasm, or at the basal domain. Because of the considerations mentioned above, we cannot assert that the cytoplasmic T567-P signal that characterized A120E-PKCι expressing cells represents fully active ezrin. Yet, our interpretation is that molecules phosphorylated on T567 outside the apical microenvironment cannot become incorporated into the apical scaffold. Conversely, it could be argued that other mechanisms (e.g. apical PIP2) may be the rate-limiting step for apical localization of ezrin. However, neighboring nontransfected cells, which should be at the same stage of differentiation, were capable of polarizing their endogenous ezrin (Fig. 6). The fact that transfected cells showed equal or less T567-P at the apical domain than their nontransfected neighbors is more reminiscent of situations where ezrin is sequestered before reaching the apical scaffold, such as keratin overexpression (Wald et al., 2005). Moreover, localization of ezrin at the basal domain, especially in nondifferentiated cells (day 3), highlights the possibility of mistargeting of ezrin if the phosphorylation does not occur in the appropriate apical environment. In this regard, Poullet et al. (Poullet et al., 2001), for example, demonstrated that ezrin N-ERMAND in the open (active) configuration can interact with the focal adhesion kinase, typically a basal protein. These results are also consistent with published data on the distribution of T567D-ezrin-GFP. The ezrin phosphomimetic mutant showed a relatively modest incorporation into the apical domain (compared with wild-type ezrin) and a very robust association with the basolateral domain (i.e. mistargeting) (Coscoy et al., 2002). However, these authors obtained their data in the context of overexpression of ezrin or its mutants, whereas we show the behavior of endogenous ezrin in early stages of differentiation, when the overall levels of ezrin expression are low. However, the distribution of the phosphomimetic ezrin mutant also suggests that phosphorylation of ezrin T567 before it arrives to the apical domain results in heavy mistargeting. In summary, although it is likely that other PKC isoforms (e.g. PKCα) or other kinases (e.g. Akt2 at the tip of intestinal villi) may also play a role in ezrin activation, apically localized PKCι is a major player in that mechanism during the early stages of intestinal cell differentiation and polarization. In addition, it seems safe to conclude that any mechanism phosphorylating ezrin T567 must operate exclusively at the apical domain to achieve full apical localization of ezrin.
Materials and Methods
Vectors and reagents
Human ezrin and PKCι complete coding sequences were obtained from Open Biosystems, AL and Origene, MD, respectively, and cloned in the pAcHLT-C Baculovirus Transfer Vector (BD Pharmingen). For expression in CACO-2 cells, the same coding sequences and the PKCi mutants (see below) were cloned in the pcDNA3.1/V5-His TOPO vector (Invitrogen). The anti-PKCι shRNA (CCGGGCCTGG ATACAATTAA CCATTCTCGA GAATGGTTA ATTGTATCCA GGCTTTTT), in the pLKO.1 lentivirus vector was obtained from Open Biosystems (cat. no. TRCN0000006037). Purified active PKCι was obtained from Upstate USA. λ-phosphatase was purchased from New England Biolabs. Actin filaments polymerization was carried out using a nonmuscle actin Biochem Kit (Cytoskeleton) according to the manufacturer's instruction. The resultant F-actin was stabilized by adding 70 nM phalloidin to the filament suspension. CNBr-activated Sepharose beads for pull down were obtained from American Biosciences. [γ-32P]ATP was obtained from Perkin Elmer Life Sciences (Boston, MA). Lucifer yellow CH was from Molecular Probes. The antibodies used in this study were as follows: mouse monoclonal anti-ezrin (clone 3C12, Abcam); mouse monoclonal anti-phospho-T567 (T567-P) ezrin (Becton Dickinson), rabbit monoclonal anti-T567-P ezrin and rabbit monoclonal type-specific anti-Akt2 (Cell Signaling); mouse monoclonal anti-PKCι (Becton Dickinson); rabbit anti-PKCι (Santa Cruz Biotechnology); rabbit anti-phospho-T555 (T555-P) PKCι (Biosource Invitrogen); mouse monoclonal anti-baculovirus gp64 (Novagen); rabbit polyclonal antibody anti-PKA IIβ regulatory subunit (Santa Cruz Biotechnology); and mouse monoclonal anti-ZO-1 antibody (Zymed). Secondary antibodies were obtained from Jackson Immunoresearch Laboratories and were all affinity purified and with no crossreactivity with IgG of other species.
Cells and baculovirus
CACO-2 cells and the CACO-2 C2BBe clone were obtained from American Type Culture Collection. The cells were cultured as described (Salas et al., 1997) in the absence of antibiotics. Sf9 cells were obtained from Becton and Dickinson as part of the Baculogold kit used for baculovirus production.
Lentivirus production and infection
Lentivirus production was done by cotransfecting the shRNA pLKO.1 vector and packaging DNA plasmids, pMDLg/pRRE#54′, pRSV-Rev and pMDLgVSVG using calcium phosphate c-precipitation into HEK293T cells. All DNAs were isolated using Qiagen endotoxin-free kits. Virus-containing medium was collected every 24 hours for 3 days, centrifuged at 2500 g for 5 minutes and passed through a 0.45 μm filter. The filtrate was used directly for infection, or virus was concentrated by ultracentrifugation at 50,000 g for 90 minutes. The virus titers were estimated by p24 ELISA (PerkinElmer, Wellesley, MA). CACO-2 cells were infected in the presence of polybrene. Transduced cells were selected for at least two passages in 10 μg/ml puromycin and used for experiments during the first three passages after transduction. The effect of shRNA expression vanished thereafter, although the cells remained puromycin resistant.
Cell extracts and ezrin purification and phosphorylation in vitro
To obtain genomic DNA extract from Sf9 cells, 4-day-old cells were harvested in 50 mM Tris-HCl, pH 8, 10 mM EDTA, 0.5% SDS and 0.5 mg/ml proteinase K, and incubated at 55°C for 1 hour. After a 10-minute centrifugation, the supernatant was incubated at 55°C for 1 hour in the presence of 25 μg RNase A. After another round of centrifugation, the supernatant containing the genomic DNA was incubated at 70°C for 5 minutes and analyzed by 1% agarose gel in Tris-acetate buffer.
Ezrin extracts and Ni2+ column purifications were carried out as follows. Sf9 cells were seeded at 70% confluency and infected at ∼10 MOI with h-ezrin-producing baculovirus. After 3 days, the cells were harvested in RIPA buffer (150 mM NaCl, 50 mM Tris-HCl, 1% NP-40, 0.5% DOC, 0.1% SDS, pH 8.0) supplemented with an antiprotease cocktail (Sigma, cat. no. P-8340) and two antiphosphatase cocktails (Calbiochem, cat. no. 524624 and 52625). The ezrin protein was purified on Ni2+ columns (ProBond resin; Invitrogen) using the nondenaturing protocol as described before (Wald et al., 2005), with washes in 50 mM imidazole and elution in 350 mM imidazole. The eluates were desalted and concentrated by ultrafiltration in Centricon YM-10 (Millipore). A single band was detected by western blot using anti-ezrin and anti-His tag antibodies.
For in vitro phosphorylation assay 6×His-tagged ezrin purified from the Ni+2 column was resuspended in 20 mM HEPES, pH 7.4, 0.03% Triton X-100, 5:l PKC lipid activator (20-133, Upstate), 14 mM MgCl2 and, either 250:Ci [γ-32P]ATP (DuPont-NEN) or 1 mM cold ATP. It was then incubated for 30 minutes at 30°C with 100 ng active PKCι (Upstate). The samples containing the [γ-32P]ATP were precipitated with 10:l of StrataClean resin beads (Stratagene) and analyzed by SDS-PAGE and PhosphorImager. The samples containing the cold ATP were blotted and analyzed by western blot using anti-ezrin and anti-T567 antibodies. For immunoblotting, CACO-2 cells were extracted in 75 mM Tris buffer, pH 7, supplemented with 1% SDS. The extract was boiled for 2 minutes, sonicated and protein was measured by Micro Lowry, Peterson's modification assay (Sigma) to seed equal amounts for PAGE and immunoblot.
Immunoprecipitation of ezrin was performed as described before (Wald et al., 2005), except that an affinity-purified goat anti-mouse IgG (Jackson Immunoresearch Laboratories) was coupled to the Sepharose beads, instead of the primary antibody. A mouse monoclonal anti-ezrin antibody was used to immunoprecipitate.
Site-directed mutagenesis and transient transfection
Mutations of alanine to glutamate, A120E, and lysine to tryptophan, K274W, in PKCι were carried out using the QuikChange mutagenesis kit (Stratagene) according to the manufacturer's specifications. Both of the mutated full-length cDNAs were cloned in pcDNATM3.1/V5-His TOPO vector (Invitrogen) and confirmed to be correct by PCR sequencing of the full-length open reading frame. Transient transfection was performed with ExGene (Fermentas) according to manufacturer's specifications.
Reverse-transcriptase quantitative PCR
Total RNA was extracted and purified from 2.8×106 cells/culture using the RNAqueous kit (Ambion) in triplicate. Reverse-transcription and quantitative PCR were done using the TaqMan® EZ RT-PCR Kit (Applied Biosystems). Ezrin-specific primers spanning more than one exon were used from the Taqman Gene Expression Assay Villin 2/Ezrin (Applied Biosystems). The reactions were run in a Roche LightCycler® 480 at the Oncogenomics Core Facility, Sylvester Cancer Center, University of Miami, and relative quantification was performed using the calibrator-normalized method taking a sixth culture as calibrator.
Staining for apoptotic cells was done with Apopercentage kit (Biocolor) according to the manufacturer's specifications.
FVB/n mice were kept according to a protocol approved by the Internal Animal Care Committee and the guidelines of the Public Health Service Policy on Humane Care and Use of Laboratory Animals. Small intestine (jejunum) was removed under deep anesthesia, immediately rinsed with ice-cold phosphate-buffered saline (PBS) containing a cocktail of anti-proteases described above. The intestinal segments were then fixed (Hayashi et al., 1999) (see below) perfusing the fixative through the lumen with the yellow tip of a pipette, and kept in fixative for 20 minutes. The intestinal segments were cut in 3-mm-long pieces, embedded in OCT (Sakura Finetek) and frozen in isopentane at melting point. Frozen sections from mouse small intestine were sectioned at –24°C, mounted on glass slides and kept at the same temperature until the immunofluorescence procedure was performed, usually within 24 hours of sectioning.
Immunofluorescence, scanning electron microscopy and confocal microscopy
For immunofluorescence, Sf9 cells were seeded directly on glass coverslips, whereas CACO-2 cells were grown on Transwell filters or glass engraved coverslips (EM Biosciences). Immunofluorescence was performed as described (Wald et al., 2005), with the following exception: when phospho-epitopes were to be localized, especially T567-P, ezrin, tissues and tissue culture cells were fixed in 10% tricholoroacetic acid (TCA), as described by Hayashi et al. (Hayashi et al., 1999). For immunolocalization of antigens in mouse tissues using an antibody raised in mice or obtained from mouse hybridomas, the following technique was used: (1) the tissues were preincubated with Fab fragments of an anti-mouse IgG antibody raised in goats, to quench endogenous mouse IgGs, before all other antibody incubations; (2) the primary monoclonal antibody was preincubated with an approximately equal molar amount of CY3-conjugated Fab fragments of an anti-mouse IgG antibody raised in goats for 30 minutes; (3) this mixture was immunoadsorbed onto nonimmune mouse IgG covalently bound to Sepharose, and the supernatant was taken; (4) the adsorbed Fab-CY3 anti-mouse bound to the primary antibody was incubated with the tissue along with any other secondary antibody. For this procedure, negative controls were done with Cy3-conjugated Fab fragments of the anti-mouse antibody preincubated with an excess of nonimmune mouse IgG. Finally, all tissue culture cells and tissues were routinely counterstained for DNA with DAPI. Frozen sections were thawed in PBS and processed as described above for tissue culture cells. Images were obtained in a Leica DMRB microscope (40×, 0.75 NA objective), through a Hamamtsu Orca camera using SlideBook 4.2 software (Intelligent Imaging Innovations), or in a Leica TCP SP5 spectral confocal microscope (63× water-immersion, 1.2 NA objective) from preparations in 30% glycerol, 8.3% polyvinyl alcohol under no. 1.5 coverslips. Confocal images were further analyzed and 3-dimensional (3D) reconstructions performed using SlideBook high-quality 3D rendering on 7-voxel thick cropped image stacks (unless indicated otherwise). Statistics of the effects of K274W PKCι transfection on the distribution of T567-P expression were done by ordinal analysis (http://home.clara.net/sisa/binomial.htm) and the significance analyzed by χ2 distribution.
For scanning electron microscopy, the cells were grown on glass coverslips, fixed in 3% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.8, and counterstained in 2% OsO4. The cells were then dehydrated in ethanol of increasing graduation, critical-point dried and coated with gold. The samples were photographed with a Jeol 35 CF scanning electron microscope at 15 kV.
We are grateful to Yolanda Figueroa for providing excellent technical support. Supported by NIDDK grants RO1DK057805 and R01DK076652. A.S.O. is a recipient of a scholarship from DOD training grant 4-49497-LS-HSI and F.A.W. is a recipient of a Crohn's and Colitis Foundation of America postdoctoral award.