Rescue of atypical protein kinase C in epithelia by the cytoskeleton and Hsp70 family chaperones.

Atypical PKC (PKCι) is a key organizer of cellular asymmetry. Sequential extractions of intestinal cells showed a pool of enzymatically active PKCι and the chaperone Hsp70.1 attached to the apical cytoskeleton. Pull-down experiments using purified and recombinant proteins showed a complex of Hsp70 and atypical PKC on filamentous keratins. Transgenic animals overexpressing keratin 8 displayed delocalization of Hsp70 and atypical PKC. Two different keratin-null mouse models, as well as keratin-8 knockdown cells in tissue culture, also showed redistribution of Hsp70 and a sharp decrease in the active form of atypical PKC, which was also reduced by Hsp70 knockdown. An in-vitro turn motif rephosphorylation assay indicated that PKCι is dephosphorylated by prolonged activity. The Triton-soluble fraction could rephosphorylate PKCι only when supplemented with the cytoskeletal pellet or filamentous highly purified keratins, a function abolished by immunodepletion of Hsp70 but rescued by recombinant Hsp70. We conclude that both filamentous keratins and Hsp70 are required for the rescue rephosphorylation of mature atypical PKC, regulating the subcellular distribution and steady-state levels of active PKCι.


Introduction
The protein kinase C (PKC) family in humans comprises three subclasses: conventional, novel and atypical. They control essential cellular functions ranging from cell proliferation to differentiation, share a highly conserved kinase domain, and differ in their regulatory domains. PKCι (iota, ι/λ; one of the two atypical isoforms, which are collectively referred to as aPKC) is an oncogene product causative of lung non-small cell carcinoma and is a predisposing factor for colon cancer when overexpressed (Fields and Regala, 2007). It is also a key regulator of asymmetry in various tissues, and is conserved in evolution. It was originally identified as such in C. elegans along with its partners (partition deficient, PAR) PAR3 and PAR6 (Rose and Kemphues, 1998). In neurons, PKCι controls axonal polarity under extrinsic Wnt signaling (Zhang et al., 2007). In epithelial cells, the PKCι-PAR3-PAR6 complex is associated with tight junctions and generates a signaling gradient that specifies the apical domain and the localization of the junctions. The tight-junctional scaffold for this complex includes Crumbs, Pals1 and PatJ (Suzuki and Ohno, 2006). The functionality of the PKC kinase domain depends on its conformation, controlled by phosphorylations in the activation loop (T403 in PKCι) by PDK-1 and in the turn motif (T555 in PKCι) (Newton, 2003).
Unlike other kinases, PKC isozymes lose their activation phosphorylation and conformation in a substrate-dependent manner (Dutil et al., 1994), i.e. they become inactivated as a consequence of their own function. This characteristic has been used to inactivate conventional and novel PKC by prolonged treatment with phorbol esters (Savart et al., 1992;Hansra et al., 1999;Lang et al., 2004). The loss of PKC conformation is controlled by phosphatases and followed by attachment to still-unidentified cytoskeletal elements in the Triton-insoluble fraction (Gao et al., 2008). Once dephosphorylated, the PKC molecules are ubiquitinylated and degraded . Hsp70 chaperone proteins (Hsp70.1 hereafter referred to as Hsp70, and the constitutively expressed Hsc70) (Daugaard et al., 2007) bind dephosphorylated PKC (including α, β and the atypical ζ isoforms) (Gao and Newton, 2002) and rescue them from degradation to regain function (Gao et al., 2006). The activation loop and turn motif are rephosphorylated, although it is unclear whether PDK-1 or other kinases are involved in the process. This mechanism effectively extends the active life of PKC (Newton, 2003) and, conceivably, can regulate the overall steady-state cellular levels of the kinase.
A role of Hsp70 in rescuing dephosphorylated atypical PKCι isoform has not yet been reported. However, based on the structural conservation of the PKC catalytic domain and the presence of the invariant Leu554 motif (NFDSQFTNEPVQL*TPDDDDI) necessary for Hsp70 binding (Gao and Newton, 2006), it is reasonable to hypothesize that Hsp70 interaction with atypical PKCι is also involved in maintaining the functional pool of this kinase. Phosphorylation in the turn motif (T555) abrogates Hsp70 interaction (Gao and Newton, 2002), thus making this phosphorylation an ideal marker for the active-conformation PKCι pool that cannot interact with Hsp70. To our knowledge, no attempts have been made to identify the specific cytoskeletal proteins that bind PKCι in the Atypical PKC (PKCι) is a key organizer of cellular asymmetry. Sequential extractions of intestinal cells showed a pool of enzymatically active PKCι and the chaperone Hsp70.1 attached to the apical cytoskeleton. Pull-down experiments using purified and recombinant proteins showed a complex of Hsp70 and atypical PKC on filamentous keratins. Transgenic animals overexpressing keratin 8 displayed delocalization of Hsp70 and atypical PKC. Two different keratin-null mouse models, as well as keratin-8 knockdown cells in tissue culture, also showed redistribution of Hsp70 and a sharp decrease in the active form of atypical PKC, which was also reduced by Hsp70 knockdown. An in-vitro turn motif rephosphorylation assay indicated that PKCι is dephosphorylated by prolonged activity. The Tritonsoluble fraction could rephosphorylate PKCι only when supplemented with the cytoskeletal pellet or filamentous highly purified keratins, a function abolished by immunodepletion of Hsp70 but rescued by recombinant Hsp70. We conclude that both filamentous keratins and Hsp70 are required for the rescue rephosphorylation of mature atypical PKC, regulating the subcellular distribution and steady-state levels of active PKCι.
inactive conformation. However, various chaperones, including Hsp70 involved in PKC rescue, are bound to keratin intermediate filaments (IFs) (Liao et al., 1995;van den Ijssel et al., 1999;Planko et al., 2007). IFs are formed by family of cytoskeletal proteins, encoded by approximately 60 different genes in mammals. Mutations identified in IF proteins responsible for human disease outnumber the pathogenic mutations in both actins and tubulins. In some cases, these mutations do not abrogate filament polymerization or structural integrity , and, therefore it is unlikely that the phenotype results from mechanical weakness of the cells. An emerging body of evidence suggests that IFs have non-mechanical functions (Oriolo et al., 2007a), including antiapoptotic activity (Marceau et al., 2007) related to intracellular signaling, but mechanistic explanations for those functions are still missing Toivola et al., 2005).
In addition, we reported a broad loss of the apical domain phenotype in the villus enterocytes of keratin 8 (K8)-null mice. Yet, a molecular model for that phenotype has been elusive (Ameen et al., 2001). Together, these facts led us to hypothesize that a pool of PKCι might be associated with a keratin scaffold in epithelial cells. The results not only corroborated the existence of a pool of keratin-bound PKCι, but also showed a novel function of the cytoskeleton in the post-translational regulation of PKCι expression and localization.

A functional pool of PKCι co-distributes and co-purifies with the keratin cytoskeleton
The CACO-2 (human colon carcinoma) cell line is an extensively used model of intestinal epithelium, slowly polarizing after 5 days in culture (Pinto et al., 1983). Like in other epithelial cells (Suzuki and Ohno, 2006), XZ confocal sections showed PKCι distributed only in the apical domain of CACO-2 cells, colocalizing with PAR3 only in the apical-most part of the cell-cell contacts (Fig. 1A,B). In the rest of the apical region, away from tight junctions, it colocalized with PAR6 (not shown), along with keratin IFs (Fig. 1B,C), but not with PAR3. Triton-100-insoluble PKCι (Fig. 1C, green) colocalized with the apical keratin signal (Fig. 1C, red). PAR3 was not detected in the Triton X-100 extracted monolayers (not shown). To characterize the cytoskeletal association of PKCι and its partners, we used an extraction method modified from a well-established method to isolate keratins that dissects actin. The Triton X-100 soluble supernatant (S1) contains cytosolic and membrane proteins. The Triton-insoluble (cytoskeletal) pellet was further extracted in 1.5 M KCl, and separated into supernatant (S2) and pellet (P). Together, the three fractions contained the entire set of cellular proteins. Actin was present in S1 and S2, but not in P (Fig. 1D). Keratins were found almost exclusively in P (Fig. 1D). PKCι and pT555 PKCι were present in all three fractions. It was surprising to find pT555 PKCι in the P fraction because previous publications indicated that only the nonphosphorylated inactive forms of PKC bind to the cytoskeleton (Gao et al., 2008). Well-known PKCι partners and components of the conventional tight-junction scaffold that binds PKCι, including PAR3, PAR6, 14-3-3 (PAR5 homolog), ZO-1 and PALS1 (Suzuki and Ohno, 2006), were not present in the keratin pellet P, but were found in S1 or S2. Hsp70 proteins, however, were present in all three fractions (Fig. 1D). Protein partitioning in lipid rafts represents an exception to the Triton insolubility criterion of cytoskeletal association. We ruled out raft association of the Triton-insoluble PKCι by lack of co-purification with the ganglioside GM1, found only in The cells were extracted with 0.5% TX-100, fixed, and processed with anti-K8 antibody (red channel) and anti-PKCι antibody (green channel). High magnification of XY confocal section through the apical plane is shown in the upper panels. Lower panels represent XZ sections, apical side up (DAPI light blue). Scale bars: 10 μm. (D) PKCι co-purifies with intermediate filaments in a highly insoluble fraction of CACO-2 cells. Equally cultured cells were extracted to obtain the three fractions: room temperature Triton-X-100-soluble fraction S1 (first supernatant); Triton-X-100-insoluble, 1.5 M KClsoluble fraction S2 (second supernatant); and pellet P (Triton-X-100-and 1.5 M KCl-insoluble). Samples of 50 μg/lane of each fraction were seeded, separated by PAGE and blotted. Ponceau-S staining is shown as a loading control. Parallel blots were processed with the antibodies indicated at the top of each blot. Arrowheads on the right hand side of each blot indicate the expected relative molecular mass (M r ). Standards on the far left are 203, 126, 80 and 39ϫ10 3 . S1, and by lack of effect of cholesterol sequestering agents filipin III and β-cyclodextrin (CD) (supplementary material Fig. S1).
To independently determine whether PKCι in the three different fractions was active kinase we used a standard PKC activity assay that is not isoform-specific. Thus, although PKCι is expressed as the prevalent isoform in CACO-2 cells (Wald et al., 2008), it was expected that other PKC isoforms would contribute to the activity as well. To dissect PKCι activity from conventional and novel PKC activities, we used inhibitors that preferentially block atypical PKC (RO-31-8220) or conventional and novel PKCs (GF10923X). A PKA inhibitor (H-89) was used to control for possible nonspecificity of the assay. Substantial PKC activity was found in S1, S2 and P fractions ( Fig. 2A). Importantly, PKC activity in the P fraction was only blocked by RO-31-8220, indicating that it is due to atypical PKC ( Fig. 2A). More direct evidence was obtained by shRNA knockdown of PKCι by lentiviral transduction. It was very effective (>90%), with only a small amount of PKCι remaining in the S1 fraction ( Fig. 2C). When the same fractions were assayed for PKC activity, PKCι knockdown abrogated PKC activity in the keratin pellet P. It is of note that all the results in Figs 1 and 2 were normalized by protein amount, but the size of each fraction, in terms of total cellular protein was very different. S1 comprised 75% of the total cellular protein and P only 7% (Table 1). Accordingly, the relative amounts of pT555-PKCι signal and PKC activities were proportional to the size of the corresponding fraction (Table 1).

PKCι and Hsp70 form a ternary complex with intermediate filaments in vitro
To establish whether PKCι binds directly to keratins, we expressed V5-6ϫHis-tagged PKCι in HEK293T cells and purified it on Ni 2+ columns under non-denaturing conditions. The desalted eluates containing PKCι but not endogenous Hsc70 (Fig. 3A) were used to overlay blots of highly purified recombinant keratins (K8, K18 and K19). We found direct binding of PKCι (V5 signal) to K8. However, no pT555 signal was detected despite the fact that a fraction of the PKCι in the input was phosphorylated in T555 ( Fig.  3A,B), suggesting that only the dephosphorylated-T555 (inactive conformation) PKCι binds to K8.
Because of PKC regulation by Hsp70 (Gao and Newton, 2006), next we tested the interaction of PKCι with Hsp70. Surprisingly, we did not find any Hsp70 co-immunoprecipitation with PKCι in the soluble fraction (S1) of CACO-2 cells (Fig. 3C). Separately, both Hsp70 (not shown) and PKCι (Fig. 3D, pull-down, IF) bound directly to highly purified intermediate filaments from CACO-2 cells in a pull-down assay. The direct interaction of PKCι with filamentous keratins confirmed the results from overlay assays described above. However, the addition of Hsp70 to the PKCι- Fig. 2. PKCι in the intermediate filament fraction is capable of substrate phosphorylation. CACO-2 cells were grown to confluency in 75-mm plates for 10 days, serum starved for 24 hours, and subjected to the three-fraction (S1, S2 and P) preparation described in Fig. 1D. PKC activity was measured using a PKC kinase assay kit. Samples of 20 μg protein were diluted into the kinase assay dilution buffer and loaded on 96-well plates coated with a PKC substrate peptide. (A) Inhibitors were added to the appropriate wells in the following concentrations: 1 μM PKA inhibitor H-89, 1 μM GF10923X (at this concentration inhibiting only conventional and novel PKCs), 1 μM RO-31-8220 (inhibits more effectively atypical PKCs). Relative kinase activity was normalized per μg of protein. The data are means ± s.d. from three independent experiments (Student's t-test significance, *P<0.001). (B) A similar assay was performed from the three fractions obtained from cells expressing a specific anti-PKCι shRNA or infected with empty vector lentiviral particles (contrl). After 10 days in culture and puromycin selection, the cells were fractionated in S1, S2 and P as described in Fig. 1D, and relative kinase activity was measured in each fraction as described above. The data are means ± s.d. from three independent experiments (t-test significance, **P<0.001). (C) In parallel experiments, S1, S2 and P fractions from cells expressing anti-PKCι shRNA and controls were analyzed by immunoblot. Ponceau-S staining of the same membrane is shown as loading control. M r standards: 194, 127, 87 and 39ϫ10 3 . Percentage of total PKC and pT555-PKCι were measured by densitometry on immunoblots using the anti-PKCι (total protein) or anti-pT555-PKCι antibodies, respectively. For all measurements, non-saturated images were used. ‡ Percentage of total PKC activity in the cell was calculated from the data in Fig. 2A. § Percentage of total PKC activity in cells expressing shRNA was calculated from the data in Fig. 2B. The data are means ± s.d. from at least three independent experiments. keratin pull-down mix increased the amount of PKCι bound to intermediate filaments threefold (Fig. 3D,E), showing the existence of a ternary complex of filamentous keratins, Hsp70 and PKCι. Similarly to the results from overlay assays, we did not detect any pT555 signal in the PKCι bound directly to intermediate filaments or in the additional PKCι bound to the filaments in the presence of Hsp70 (Fig. 3D, pull-down, pT555), albeit the input did contain pT555 PKCι. This result indicates that only the pT555dephosphorylated form of PKCι interacts with Hsp70 in a ternary complex with IF. This is consistent with the binding of another atypical isoform, PKCζ, to Hsp70 that occurs only when the turn motif is dephosphorylated (Gao and Newton, 2002).
Keratins and Hsp70 post-translationally regulate steady state levels of PKCι Hsp70 and Hsc70 are products of two different genes that display 86% identity at the amino acid level. Hsc70 is constitutively expressed whereas Hsp70 is expressed only under stress in normal tissues (Daugaard et al., 2007) such as intestinal epithelium (Baumler et al., 2007). However, in some tissue culture cells, including CACO-2 cells, both Hsp70 and Hsc70 are constitutively expressed (Hirasaka et al., 2008;Broquet et al., 2003), albeit Hsp70 expression can be further upregulated by stress (Pierzchalski et al., 2008). Importantly, whereas it is unclear to what extent the functions of both proteins are redundant, Hsp70 and Hsc70 share several clients and a substantial level of functional overlap has been demonstrated (Singh et al., 2008;Freeman and Morimoto, 1996;Haag Breese et al., 2006).
To understand the possible biological implications of a PKCι-Hsp70 interaction based on filamentous keratins, we started by knocking down IFs using lentiviral-mediated expression of anti-K8 shRNA. This is possible because differentiated CACO-2 cells express only one type-II keratin (K8; see two-dimensional gel analysis in supplementary material Fig. S2A), which is necessary to form the obligate heterodimers of a type-II and a type-I keratin. As in previous similar experiments (Wald et al., 2008), the knockdown was efficient, although some cells still expressed filaments ( Fig. 4A, dark blue). Cells in which IF knockdown was very effective showed no PKCι, except for some remnant signal in the tight-junction region (Fig. 4A, green). Both Hsp70 and Hsc70 were modestly decreased and delocalized (Fig. 4A, red). Immunoblot analysis of total extracts confirmed the K8 knockdown, and showed a modest reduction in Hsp and/or Hsc70, an almost 70% reduction in total PKCι, and extensive loss (>90%) of the pT555 PKCι signal (Fig. 4B,C). After testing several sequences, we have found only one shRNA sequence that effectively knocks down K8 in CACO-2 cells. The lack of a second independent shRNA sequence carried a small risk of observing 'off-target' effects. However, having observed a similar decrease of PKCι in K8-null mouse enterocytes (see Fig. 7L), we can reasonably rule out that possibility for the PKCι readout. The reduction in total PKCι was clearly much larger than the size of the PKCι pool that co-purifies with IFs (Table 1), suggesting that the interaction of PKCι with IF-Hsp70 is necessary to maintain a population of PKCι molecules that is normally not attached to IFs.
To determine whether changes in PKCι steady state levels were transcriptionally regulated, we measured PKCι mRNA levels by qPCR. Triplicate determinations (normalized to GAPDH) yielded 1.25±0.01 (control) and 1.28±0.02 (K8 knockdown), showing that K8 knockdown did not significantly affect steady-state levels of PKCι mRNA. By contrast, incubation of CACO-2 cells with the proteasome inhibitor MG-132 for only 6 hours resulted in a threefold increase in the steady-state amount of PKCι (along with a tenfold increase in total protein ubiquitinylation) (supplementary material Fig. S3), indicating that, like in other cells (Gao and Newton, 2006), PKC turnover is very high. Along with the lack of transcriptional effects shown before, this result highlighted the Journal of Cell Science 122 (14) The 6ϫHis-and V5-tagged human PKCι was expressed in HEK293T cells and purified under nondenaturing conditions on Ni 2+ columns. Run-throughs and 350 mM imidazole eluates were analyzed by immunoblot with anti-V5, anti-Hsc70 or anti-pT555-PKCι antibodies. Of note, HEK293T cells do not express Hsp70; blots for this protein were negative (not shown). M r of standards is ϫ10 3 . (B) For blot overlay, recombinant bacterially expressed keratins (5 μg/lane K8, K18 and K19) were separated in a 12% acrylamide gel and blotted onto nitrocellulose. The membrane was transiently stained with Ponceau-S red to show protein load (bottom panels). His-V5-tagged PKCι, eluted from the Ni 2+ column (A), desalted and concentrated by ultrafiltration, was added onto the caseinsaturated keratin blots and incubated overnight at 4°C. The overlay was extensively washed and the membranes were processed as for a regular immunoblot with anti-V5 or anti-pT555-PKCι antibodies. (C) Hsp70 and PKCι do not coimmunoprecipitate in the soluble fraction of CACO-2 cells. Soluble fraction (S1) of 10-day-old CACO-2 cells was prepared as described in Fig. 1 and immunoprecipitated (IP) with anti-PKCι (+) or with non-immune rabbit IgG (-). Samples of the extracts (input) and the eluates were probed with anti-PKCι antibody (raised in mouse). The membrane was later re-probed with anti-Hsc70 and/or anti-Hsp70 antibody (raised in rabbit). (D-E) The amount of PKCι pulled down with IF increases in the presence of Hsp70. Highly purified (>99.9%, purity control; see supplementary material Fig. S2A) native polymerized IFs were covalently coupled to CNBr-activated Sepharose beads and incubated overnight with gentle shaking at 4°C in PBS containing 1% TX-100 with recombinant PKCι together with Hsp70 (input 1) or with the recombinant purified PKCι alone (input 2). After extensive washes, the beads were eluted in sample buffer and analyzed by immunoblot (pull-down, IF). Incubation with Sepharose beads coupled to non-immune rabbit serum served as a negative control (pull-down, control). Notice that the K8 blot is shown at a much shorter exposure to avoid saturation. (E) Quantification of the result shown in D (pull-down). The bars represent the means ± s.d. of the ratio of densitometric values of the PKCι bands relative to K8 bands in the same lane, in the pull-down with IFs without (left-hand bar) or in the presence of recombinant Hsp70 (right-hand bar) from three independent experiments. For all measurements, non-saturated images were used (Student's t-test significance, *P<0.01).
likelihood of post-translational mechanisms effectively maintaining PKCι steady-state levels.
Because of the protective role of Hsp70 in PKC regulation, it was important to test the prediction that Hsp70-mediated rescue is involved in the maintenance of normal steady-state levels of PKCι. In CACO-2 cells, two members of the Hsp70 family, Hsp70 and Hsc70, are expressed in similar amounts because Hsp70 expression is leaky (Broquet et al., 2003;Liu et al., 2003). Accordingly, two different shRNA sequences were designed to knockdown Hsc70 (shRNA1) or Hsp70 (shRNA2) and were delivered by lentiviral transduction followed by puromycin selection. Both shRNAs knocked down Hsp70 and Hsc70 (Fig. 5A), possibly because of the very high level of sequence homology between the two mRNAs (>90%). Both shRNAs had no effect on other proteins, such as actin or K8, which were used as load controls (Fig. 5A). However, there was a 60-70% decrease in total PKCι levels. More importantly, knockdown of Hsp and/or Hsc70 proteins with both shRNAs abrogated pT555 PKCι signal (Fig. 5A,B). Because knockdown of Hsp70 proteins results in apoptosis (Aghdassi et al., 2007), all these experiments were performed within 48 hours of the lentiviral infection. Within this time window there was no increase in annexin V binding, although longer times of Hsp70 knockdown resulted in massive cell death (not shown). Additional internal controls for lack of caspase-3 activation and cleavage were also performed (Fig. 5C). Similar Hsc and/or Hsp70 knockdown experiments were analyzed by immunofluorescence and confocal microscopy. They confirmed that the knockdown of Hsp and/or Hsc70 proteins was extensive in the majority of the cells (Fig. 5D,E), and indicated that Hsp70 knockdown did not affect the expression or localization of IFs (Fig.  5F,G, also see Fig. 5A), but abrogated the pT555 PKCι signal (Fig.  5H,I).
Hsp70 is involved in facilitating the correct conformation of many proteins, so it is likely that several client kinases might be affected. Likewise, the effect of K8 knockdown could be a general phenomenon affecting several kinases. To test this idea we run a screen of 12 different kinases in a commercial facility that uses a panel of validated, normalized antibodies for immunoblot. The At 4 days after seeding, the cells were transduced with lentiviral particles carrying shRNA against K8 (K8 shRNA) or empty vector particles (control). After 10 days in culture and puromycin selection, the cells were fixed and processed with anti-K8 antibody (blue channel), anti-PKCι antibody (green channel), anti-Hsp70 (total, Hsp70 and/or Hsc70) antibody (red channel) and DAPI (light blue), and analyzed by confocal microscopy. XZ sections are shown with the apical side up. DAPI staining is shown in light blue. Scale bars: 10 μm. (B) CACO-2 cells were transduced as described above, but total SDS extracts were analyzed by immunoblot. M r of standards is ϫ10 3 . (C) The reduction in band intensity was obtained from the ratios of band intensity in knockdown and control cells, normalized by the intensity of the actin band re-probed in the same lane. For all measurements, non-saturated images were used. The means ± s.d. from three independent experiments are shown. pattern of relative kinase expression and activation in control cells (Table 2) was identical to those published before from various pools of cell extracts [see table 1, 8 days, in Wald et al. (Wald et al., 2008)], showing the consistency of the method from one set of experiments to another analyzed at the same facility. Comparing extracts from CACO-2 cell cultures transduced with either anti-K8 shRNA lentivirus or mock viral particles, we independently confirmed the decrease in PKCι (Table 2; Fig. 4B,C). Apart from this kinase, of five known Hsp70 clients, one did not change (Raf-1), whereas three showed a steep decrease (PKCβII, PKCζ and Aktα) and another was only mildly affected (Chk-1). Interestingly, most PKC isoforms decreased, with the exception of PKCα. By contrast, the PKA catalytic subunit α actually showed an almost twofold increase (Table 2). Importantly, PDK-1, which is known to phosphorylate newly synthesized PKC was not affected by K8 knockdown. Because of the consistency of these measurements, it seems safe to rely on the results that show no change. However, the screen does not provide any measurement of variability, such as standard deviation. Bearing this limitation in mind, it seems safe to conclude that the phenomenon described here might affect other kinases, but is not generalized to all kinases, or even to all Hsp70 clients. To statistically confirm the significance and extend the observation to cells in vivo, we compared the levels of two Hsp70 clients in small intestine villous enterocytes of K8-null mice with control littermate animals. Thus, we could confirm that the differences shown by the kinase screen in CACO-2 cells were reproducible in the animal model and statistically significant for the other atypical PKC isoform PKCζ (pT410) and for pT308 Aktα (Table 3), two known Hsp70 clients (Gao and Newton, 2002).
To confirm our results in vivo, we used two mouse knockout models. The K19-null, K18 +/and hK18 R89C (knock-in) mice (K18R89C animals) have been shown to lack IFs in the intestinal crypts, where keratins form small, non-filamentous aggregates (Hesse et al., 2007). In normal littermates of K18R89C mice, the small intestinal crypts showed apical layers of Hsc70 and pT555 PKCι (Fig. 6A,G). By contrast, in K18R89C transgenic mice the apical localization of Hsc70 and pT555 PKCι was abrogated along with the expression of IFs (Fig. 6C,D,I,J), with the signals close to those in antibody negative controls (Fig. 6E,F,K,L). In the same animals, the expression and distribution of Hsc70 and pT555 PKCι were normal in the villi (not shown). The second knockout model we used, K8-null mice, lack intermediate filaments in the villus, but display normal filaments in the crypts, because of the expression of redundant K7 (Baribault et al., 1994;Ameen et al., 2001). Accordingly, in K8-null mice the apical layer of Hsc70 was not observable in villus enterocytes (Fig. 7C). The pT555 PKCι signal in villus enterocytes from control littermates was present only in the tight-junctional compartment (Fig. 7G, arrows). In K8-null mice, the IFs disappear after enterocytes move over the crypt-villus boundary (Ameen et al., 2001;Wald et al., 2008). Unlike in control Journal of Cell Science 122 (14) Identical amounts of protein from total SDS extracts from 1.5ϫ10 7 CACO-2 cells infected with empty lentiviral particles (Control) or particles expressing anti-K8 shRNA and selected in puromycin were analyzed in a commercial facility (Kinexus Kinetworks KCPS 1.0, Vancouver, Canada) by immunoblot using a panel of validated standardized antibodies. The antibody mixtures were validated by the provider to prevent cross-reactivity among target proteins. Normalization was based on the intensity of fluorescence detection of immunoreactive standard proteins (Pelech, 2004). All the normalization and statistics were done by Kinexus. The values of normalized units, are, therefore, comparable throughout the assay. Probe indicates the antibody used, either against total protein or a phosphoepitope in a consensus regulatory domain. Chk-1, Checkpoint kinase 1; ND, not detected.
*-signifies an increase. † Yes, confirmed direct clients; Ind, kinases known to indirectly depend on Hsp70. Cells were isolated from 15-cm loops including the duodenum and part of jejunum from three control and four knockout animals as described in Materials and Methods and subjected to SDS-PAGE and immunoblot. The band intensity was measured and normalized to the corresponding actin band reprobed in the same lane as described in Fig. 4C (relative units). Control animals were were K8 +/littermates. Rescue of aPKC and the cytoskeleton animals, K8-null mice displayed the pT555 PKCι signal associated with tight junctions only until the mid-villus region (Fig. 7I, arrows), not reaching the tip of the villus (Fig. 7I, arrows) as in controls. This suggests that even the tight-junction pool of PKCι needs IFs to subsist.
To validate these observations at the biochemical level, villus enterocytes were isolated from the mucosa of control and K8-null mice and analyzed by immunoblot. The intensity of Hsc70 signal was very variable among all the animals, but it was not significantly decreased in K8-null mice. The average actually showed a modest, non-significant increase (negative values, Fig.  7L). This effect was different from CACO-2 cells under K8 knockdown (Fig. 4) and might indicate some degree of cellular stress in the epithelium in some animals. Conversely, pT555 signal was greatly and significantly decreased in all knockout animals as expected (Fig. 7K,L, where n=3 and n=4, respectively). Because the best-known function of atypical PKC is to maintain the structure and localization of tight junctions (Hurd et al., 2003) and paracellular permeability (Hirose et al., 2002), we hypothesized that a decrease in pT555 PKCι might affect tight junctions in the upper part of villi. We did not observe any morphological changes by electron microscopy (not shown). However, transepithelial uptake of fluorescent Dextran 3000, a bona fide probe for paracellular permeability, was significantly increased in small intestine loops with their normal blood supply (Fig. 7M). No evidence of inflammation was observed microscopically, or by myeloperoxidase assay of the intestinal mucosa in the same animals (control, 0.124±0.05 OD units/50 μg protein; K8-null, 0.05±0.08 OD units/50 μg protein). Therefore, the decrease in PKCι correlates with a modest but significant increase in permeability of the tight junction in the small intestine, as predicted by previous studies on aPKC functions.
Hsp70 and filamentous keratins are essential for the rephosphorylation of the turn motif in mature PKCι The results described above indicate that Hsp70 and the keratin cytoskeleton are necessary to maintain the steady-state levels of PKCι but do not distinguish between direct participation of those molecules in the rescue mechanism and indirect effects mediated by other molecules, such as kinases. To independently corroborate the role of Hsp70 and the cytoskeleton in the maintenance of PKCι levels we designed an in-vitro rescue of PKCι turn motif phosphorylation assay, using different cellular fractions. Because PKCι lacks the phorbol-ester binding domain, it was impossible to dephosphorylate PKCι by phorbol-ester overstimulation, as with conventional PKCs. Therefore, we incubated S1, S2 and P fractions of CACO-2 cells (described in Figs 1 and 2) in the presence of ATP and an excess of the consensus PKCι substrate peptide. The experiments were performed in the presence of proteases and proteasome inhibitor cocktails, but in the absence of phosphatase inhibitors.
After 5-hour incubations with the PKC substrate peptide, we succeeded in substantially decreasing the pT555 PKCι signal in all three fractions, without noticeable changes in the PKCι protein levels ( Fig. 8A for S1, not shown for S2 and P; the fractions containing dephosphorylated PKCι are hereafter referred to as S1* and P*). Then, the peptide was removed by ultrafiltration, and ATP was replenished. Separately, none of the fractions were able to rephosphorylate PKCι (S1*+ATP, Fig. 8A; P*+ATP, Fig. 8B). Mixing S2 with either S1 or P in the presence of ATP also failed to achieve any PKCι rephosphorylation (not shown). However, when fractions S1* and P* were mixed together in the presence of ATP, pT555 signal was restored, in most cases to levels similar to those at the beginning of the experiment (S1*+P*+ATP, Fig. 8C). Because the P fraction is enriched in keratins but still contains keratin-associated proteins, we also used a highly purified preparation of keratins to supplement S1*. When analyzed by twodimensional gels and silver stain, these keratins were found to be >99.9% pure [no contaminants were detected within the sensitivity of silver stain; supplementary material Fig. S2A (Oakley et al., 1980)]. Such a preparation was found to be as equally active as the P fraction in supplementing S1* for PKCι rephosphorylation (S1*+IF+ATP, Fig. 8D). By contrast, single isolated non-filamentous K8, K18, or K19 failed as supplements for S1* to induce PKCι rephosphorylation (Fig. 8E). The specificity of Hsp70 and/or Hsc70 for the PKCι rephosphorylation was tested by immunodepleting S1* with three anti-Hsp70 and anti-Hsc70 antibodies. As expected, after immunodepletion of Hsp70 and/or Hsc70, S1* supplemented with purified filamentous keratins also failed to rephosphorylate PKCι (Hsp70-depleted, S1*+IF+ATP, Fig. 8F). However, if the same immunodepleted S1* fraction was supplemented with recombinant purified Hsp70 (single-band by silver stain, supplementary material Fig. S2B), it regained its ability to rephosphorylate pT555 PKCι (Fig. 8G).
Finally, we wanted to check whether Hsp70 that is bound to IF in the pellet fraction (P) of CACO-2 cells is functionally competent to enable PKCι rephosphorylation. To that end, we added the intermediate filament pellet P* fraction to the PKCιdephosphorylated Hsp70-immunodepleted S1* fraction (Fig. 8H). Indeed, we found that addition of the P* fraction and ATP (+P+ATP, Fig. 8H) resulted in PKCι rephosphorylation, whereas purified IFs, lacking Hsp70, did not rescue pT555 signal (+IF+ATP, Fig. 8H) in the presence of Hsp70-immunodepleted S1* (Fig. 8H). We conclude that Hsp70 and/or Hsc70 and filamentous keratins, present together in an insoluble cytoskeletal pellet compatible with the complexes shown in vitro in Fig. 3, are equally necessary to rescue dephosphorylated PKCι in the inactive conformation and enable rephosphorylation of the turn motif.

Overexpression of keratin filaments affects the subcellular distribution of PKCι
Bearing in mind that there is an excess of soluble Hsp70 in intestinal cells (Fig. 1), and that filamentous keratins are essential for the rescue of PKCι phosphorylation (Fig. 8), we hypothesized that overexpression of keratin filaments might affect the distribution of PKCι within an epithelial cell, because overexpressed keratins result in redistributed and mislocalized excessive IFs (Wald et al., 2005). The rationale was that the assembly of the ternary complex (PKCι, Hsp70, filamentous keratins) might be the rate-liming step for PKCι rescue, considering that the keratin-Hsp70 compartment is much smaller than the cytosolic compartment of soluble Hsp70 (Table  1). The prediction was that additional IFs would increase the ability of the cell to rescue more PKCι. Furthermore, the additional newly refolded and rephosphorylated PKCι would be expected to interact with nearby membranes through its C-terminal domain and become mislocalized. To test these predictions of the model, mild K8 overexpressors (approximately twofold increase in K8 protein) were used first. Unlike 'severe' HK8-4 animals with an intestinal phenotype (approximately fourfold increase in K8) (Casanova et al., 1999;Wald et al., 2005), these animals display only a chronic pancreatic phenotype (Toivola et al., 2008). The typical distributions of pT555 PKCι (Fig. 9B, arrows) and Hsc70 (Fig. 9J) were not Journal of Cell Science 122 (14) Frozen sections were processed with anti-Hsp70 antibody (green) or (G-J) with anti-pT555 PKCι antibody (green, arrowheads point at tight-junction compartment of PKCι), and anti-K8 antibody (K8, red), and counter-stained with DAPI (light blue). Control sections were processed with non-immune IgG at the same dilutions (E,F). The images are single confocal optical sections. Scale bars: A-F, 20 μm; G-J, 10 μm. (K) Villus enterocytes from control mice (+/-; 1-3), and K8-null mice (-/-; 4-6) were isolated, extracted in SDS, run in SDS-PAGE and blotted. Immunoblots of the same membranes are shown for keratins, actin, Hsc70 and pT555. All the blots are shown in non-saturated exposures, except for Hsc70. In this case, the variability in protein expression was required to saturate the more intense bands to make the weaker bands visible. (L) The reduction in band intensity was calculated as in Fig. 4C (negative values indicate increase).
Results are means ± s.d. from four knockout and four control animals normalized as percentage of reduction with respect to the average of bands for the same protein from control animals (+/-) run in the same gel. For all measurements, non-saturated images were used. Student's t-test statistics were performed directly on means ± s.d. of actin-normalized values (*P<0.025). (M) Texas-red Dextran 3000 (10 mM in 1:1 PBS:H 2 O) was perfused inside 10cm loops of duodenum and the first part of jejunum in anesthetized K8-null (-/-) or control littermates (+/+) for 50 minutes. Blood samples were collected at time 0 (and pooled together because fluorescence values were indistinguishable) or at 50 minutes when animals were sacrificed. Samples of serum were analyzed by spectrofluorometry. The results are means ± s.d. from five animals in each group (Student's t-test significance, *P<0.05, **P<0.02). changed by K8 overexpression, but, for both of them, a signal above background was visualized in the cytoplasm, especially in regions where abundant supernumerary IFs were observed (arrows, Fig. 9E-F; Fig. 9M-N). Second, we also analyzed the distribution of Hsc70 and pT555 PKCι in villous enterocytes of an independent cohort of five hemizygous HK8-4 animals ('severe' overexpressors). A result identical to the mild overexpressors was observed in flurorescence (not shown). SDS extracts from purified enterocytes of the HK8-4 animals showed the expected increase in keratin, and also a significant fivefold increase in pT555 signal (Fig. 9Q,R). We conclude that IF overexpression does not interfere with the normal localization mechanisms of PKCι through its PAR partners, but does increase steady-state levels of phosphorylated aPKC turn motif, possibly through an increased rescue, and results in delocalization of pT555 PKCι signal.

Discussion
This work highlights the importance of chaperone-mediated rescue in maintaining functional levels of atypical PKC and also indicates that filamentous keratins are essential for that function. Because overexpression of active PKCι is causative of neoplasia, understanding PKCι rescue might be important for interventions in abnormal cell growth. From a functional perspective, the apical polarity phenotype of keratin knockdown (Salas et al., 1997) or knockout (Ameen et al., 2001;Toivola et al., 2004;Satoh et al., 1999) has been documented, and includes loss or mistargeting of apical membrane proteins, depolarization of γ-tubulin, and disorganization of microtubules. It is conceivable that some of these anomalies might be due to the steep decrease in active PKCι activity. In agreement with this notion, here we report an increase in paracellular permeability in the intestine (Fig. 7M), as expected from the functions of aPKC on tight junction assembly and integrity reported by others (Hirose et al., 2002).
The data in this work, along with previous publications (Satohisa et al., 2005), support a model of two different apical PKCι pools, one associated with PAR3 at the tight-junction, and the other located in the apical membrane, away from junctions (Fig. 10A). In vivo, the latter is observed in mouse intestinal crypts, which CACO-2 cells mimic well, whereas the former is observed in the more differentiated villus enterocytes. Upon continuous activity, PKCι from both pools becomes dephosphorylated and acquires an inactive conformation (Fig. 10B). This continuous and rapid turnover was Fig. 8. Keratins and Hsp70 are required for rephosphorylation of mature dephosphorylated PKCι. S1, S2 and P fractions were prepared from CACO-2 cells as described in Fig. 1 except that anti-phosphatase cocktails were omitted. The phosphorylation state of PKCι turn motif was examined using anti-pT555-PKCι antibodies (pT555). (A) To induce the activity-dependent dephosphorylation of PKCι, S1 was incubated in the presence of 150 μM of PKC substrate peptide and 1 mM ATP at 30°C with gentle shaking for 5 hours, resulting in S1 with dephosphorylated PKCι (S1*). After treatment, the peptide was removed from the solution by ultrafiltration. Subsequent incubation of that fraction (S1*) with 1 mM ATP for 4 hours (S1*+ATP) did not lead to rephosphorylation of PKCι. (B) Incubation of the dephosphorylated pellet fraction (P*) alone with 1 mM ATP for 4 hours (P*+ATP) failed to show PKCι rephosphorylation. (C) 50 μg of protein in S1* was then incubated with 20 μg of protein in the equally dephosphorylated pellet (P*) fraction or (D) with 15 μg of purified native filamentous keratins (IF) in the presence or absence of 1 mM ATP at 30°C for 4 hours. Pan-keratin antibodies (pan-keratin) were used to show keratin load. (E) Only filamentous keratins are able to rephosphorylate PKCι. S1* fraction was supplemented with purified native filamentous keratins (IF) in the presence (S1*+IF+ATP) or absence of 1 mM ATP (S1*+IF) as in D, or with the recombinant K8 (S1*+K8+ATP), K18 (S1*+K18+ATP) or K19 (S1*+K19+ATP) in the presence of 1 mM ATP for 4 hours. (F-H) PKCι rephosphorylation in the presence of purified IFs is Hsp70-dependent. (F) Immunodepletion by incubation with anti-Hsc70 and/or anti-Hsp70 antibodies resulted in substantial decrease in the amount of Hsp70 protein in S1 fraction (Hsp70-depleted). Incubation of S1 fraction with normal rabbit serum served as a control (control). The activity-dependent dephosphorylation of PKCι in the presence of the substrate peptide was unaffected by either Hsp70 immunodepletion or incubation with the normal rabbit serum (S1*). However, reduction of Hsp70 protein in S1 fraction resulted in the abrogation of mature PKCι rephosphorylation in the presence of intermediate filaments and 1 mM ATP (S1*+IF+ATP; Hsp70-depleted versus control *). (G) Addition of the recombinant Hsp70 protein to the immunodepleted S1* fraction restores PKCι rephosphorylation. 50 μg of protein from the dephosphorylated Hsp70-immunodepleted S1* fraction was incubated with 15 μg of purified native filamentous keratins (IF) and supplemented (+) or not (-) with 2 μg of the recombinant Hsp70 protein in the presence of 1 mM ATP at 30°C for 4 hours. (H) Hsp70 bound to IFs is sufficient for PKCι rephosphorylation. 50 μg of protein from the dephosphorylated Hsp70immunodepleted S1* fraction was incubated with 15 μg of purified native filamentous keratins (+IF, +IF+ATP) or with 20 μg of the equally dephosphorylated pellet (+P*, +P*+ATP) fraction in the presence or absence of 1 mM ATP at 30°C for 4 hours. The blots are typical of very consistent experiments repeated at least three times.
documented here by blocking proteasomal activity (supplementary material Fig. S3). Part of the inactive PKCι is ubiquitinylated and degraded (Fig. 10C), but a fraction is rescued by Hsc70 in association with keratins (Fig. 10E), to be rephosphorylated and relocalized back to its normal apical compartments. We do not know what determines the fate of an inactive PKCι molecule, although the data in Fig. 3 encourage the speculation that binding to IFs might be the first step of the rescue process. The pull-down data suggest that there are at least two subpopulations of PKCι in the inactive conformation bound to IFs: one bound directly and another one attached via a ternary complex that includes Hsp70 and/or Hsc70. We cannot dissect the role of each population, but the functional data (Fig. 8) strongly suggest that the latter represents a PKCι in the process of being refolded.
To further test the notion that the ternary complex of inactive PKCι, Hsp70 and IFs might represent the first step of the rescue process, we verified that Hsp70 and/or Hsc70 bound to IFs (Fig.  1D) are functional, because the P* fraction is sufficient to rescue Hsp70-immunodepleted S1 activity to rephosphorylate the turn motif (Fig. 8H). However, the existence of pT555 and active PKCι in the IF pellet (Figs 1, 2) suggests that the release step might occur after rephosphorylation. In the absence of a positive identification of the kinase responsible for rephosphorylating the turn motif, we are currently unable to analyze the steps that follow the interaction of PKCι with Hsp 70 and/or Hsc70. However, it was surprising that PKCι in the IF pellet can be functional in the absence of all its known PAR partners and membranes, just associated with the keratin cytoskeleton. Whereas the size of PKCι cytoskeletal compartment is modest (Table 1), the presence of an active PKCι pool on the cytoskeleton, away from the plasma membrane, might have important implications for other functions (e.g. membrane traffic) involving PKCι-dependent phosphorylations within 1 μm or less from the plasma membrane.
The reconstitution experiments enabled us to independently conclude that there is a direct role for Hsp70 and filamentous keratin in PKCι conformational rescue behind the phenotype of keratin knockdown and knockouts. However, there remains the possibility that the keratin knockdown effect might be mediated by the observed decrease in steady-state Hsp70 levels. This possibility seems to be dispelled by the high levels of Hsc70 in K8-null enterocytes. Conversely, the Hsp70 knockdown did not affect the steady-state amount of keratins. Together, these data allow us to conclude that filamentous keratins and Hsp70 are independent but concurrent requirements for the first step of the PKCι rescue mechanism, acting in a complex attached to the cytoskeleton.
The apical distribution of keratin IF in intestinal cells is conserved from worms to vertebrates (Oriolo et al., 2007a). From the results in Fig. 8, it seems safe to conclude that the rescue of PKCι would occur only at locations where IFs and Hsp70 are present together. Because there is an excess of soluble Hsp70, the availability of IFs in the cell is expected to be a rate-limiting step in the PKCι rescue mechanism. We tested this prediction of the model in two different K8-overexpression animal models. In both cases, there was grossly mislocalized pT555 PKCι signal in the regions of the cell where abnormal, excessive IFs were observed. Yet, the typical localization to the tight junctions was not affected (Fig. 9), indicating that the Journal of Cell Science 122 (14) Fig. 9. Subcellular distribution of PKCι and Hsc70 modified by overexpression and redistribution of keratins. Frozen sections of duodenum from genetically unmodified FVB/n mice (A-D,I-L) or transgenic K8 overexpressors (K8 oe; E-H,M-P) were processed with anti-K8 antibody (K8; red channel), and anti-pT555 PKCι antibody (A,B,E,F), anti-Hsc70 antibody (I,J,M,N) or non-immune rabbit IgG (control; C,D,G,H,K,L,O,P) (green channel). Notice that intestinal epithelium does not express Hsp70 when not stressed. In each case, controls were performed on parallel sections from the same block from which the images with specific antibody were obtained (shown above the corresponding control). DNA was counterstained with DAPI (light blue). The images are representative of sections from six overexpressor animals and five FVB/n littermate controls and are confocal optical sections obtained at 0.9 Airy units maintaining the same gains in the red and green channels. Arrows: B, pT555 signal consistent with tight junction distribution; E, mislocalized K8 signal; F, pT555 signal both in the normal expected distribution (apical junctions) and in the region of abnormally localized IFs. Scale bars: 10 μm. (Q) Duodenum and jejunum enterocytes were isolated from five HK8-4 K8 transgenic overexpressors (K8 oe) or four control littermates (FVBn), and analyzed by immunoblot (30 μg/lane total protein), reprobing the same membranes with anti-atypical PKC phospho-turn motif antibody (Epitomics) (pT555), and anti-keratin and anti-actin (loading control) antibodies. For the pT555 signal, bands in lanes 5, 6, 7 and 8 are saturated to show visible bands in lanes 1 and 4 at the same exposure. (R) The relative intensity of pT555 signal respect to the actin loading control signal was measured in non-saturated exposures of the same blots. Significance was calculated by the Student's t-test (*P<0.001).
canonical localization mechanisms were not disturbed by keratin overexpression.
In summary, we conclude that in the intestinal epithelium IFs display a non-mechanical function, compartmentalizing the Hsp70dependent rescue of PKCι. Results in Table 2 and Table 3 suggest that a similar mechanism might apply to other members of the ABC kinase family with the notable exception of PKCα. Interestingly, Akt1, a key signaling molecule for cell survival, might be within the set of Hsp70 and/or Hsc70 clients that depend on filamentous keratins (Tables 2,3). This unsuspected possibility might be important to understand how keratin knockouts sensitize cells for apoptosis. Conversely, the dependence of IF for kinase rescue is clearly not universal for all Hsp70 clients. Such a compartmentalization provides an additional, novel mechanism for sustaining the normal steadystate levels of active atypical PKC along with its functions, also aiding to preserve its normal asymmetric distribution.
The data presented here adds to a growing body of evidence indicating that IFs are integral parts of signaling networks (Pallari and Eriksson, 2006;Kim and Coulombe, 2007). This might be especially important to at the time of understanding phenotypes resulting from IF protein mutations. Such phenotypes might not be the consequence of 'mechanical' weakness of the filaments, but rather emerging from dysfunction of signaling pathways. It will be important to incorporate this notion into the testable hypotheses of future research efforts in this field.

CACO-2 cell extraction and fractionation
The procedure was modified from Steinert et al. (Steinert et al., 1982). At 10 days after seeding, cells were washed in PBS and then extracted in PBS containing 1% Triton X-100, 2 mM EDTA (extraction buffer, EB) supplemented with cocktails of protease and phosphatase inhibitors (Calbiochem) at room temperature. After 15 seconds (three 5-second intervals) of sonication the cell extract was spun for 10 minutes at 16,000 g). This first supernatant was referred to as the S1 fraction. The pellet was resuspended in 1.5 M KCl, sonicated for 15 seconds (three intervals), incubated for 10 minutes on ice, and spun for 10 minutes at 16,000 g). The resulting supernatant was referred to as the S2 fraction, and the pellet was referred to as the P fraction. For functional assays, S1 was used directly, S2 desalted by ultrafiltration, and P resuspended as a coarse suspension by a 3-second sonication. For immunoblots, S1 and S2 were acetone-precipitated, washed in double-distilled H 2 O, and resuspended in sample buffer. In all cases protein was determined by Lowry assay.
PKC activity assay PKC activity was measured using a PKC Kinase Non-Radioactive Assay kit (Assay Designs, Stressgen, Ann Arbor, MI) as follows: CACO-2 cells were grown to confluency for 10 days in 75-mm plates, serum starved for 24 hours, washed twice with PBS and subjected to three fractions preparation (S1, S2 and P) as described above. About 18-20 μg of protein was diluted into the kinase assay dilution buffer (Stressgen) and loaded per well in 96-well plates coated with a PKC substrate peptide. Inhibitors were added to appropriate wells in the following concentrations: 1 μM PKA inhibitor H-89, 1 μM GF10923X (at this concentration inhibiting only conventional and novel PKCs), 1 μM Ro-31-8220 (inhibits atypical PKCs more effectively). The PKC assay was performed according to manufacturer's specifications.
Non-denaturing PKCι purification, blot overlay, pull-down assays and immunoprecipitation PKCι was transfected into HEK293TN cells, harvested, purified on Ni 2+ columns under non-denaturing conditions and used in blot overlays as described (Wald et al., 2008). Pull-down experiments were performed as described elsewhere (Oriolo et al., 2007b). Briefly, highly purified intermediate filaments from CACO-2 cells were covalently bound to CNBr-activated Sepharose beads (Amersham) according to manufacturer's protocol (25 μg of keratin/10 mg beads per reaction). The beads were blocked with 1% casein, incubated with 2 μg/ml recombinant PKCι (Invitrogen) in the presence (2 μg/ml) or absence of recombinant Hsp70 (Stressgen) in PBS containing 1% Triton X-100 overnight at 4°C with gentle shaking and then extensively washed. Sepharose beads coupled to normal rabbit serum (Jackson ImmunoResearch Laboratories) served as a negative control. Immunoprecipitation of PKCι was performed as described before (Wald et al., 2005), except that the protein-A-coupled beads (Santa Cruz Biotechnology) were used instead of Sepharose beads. A rabbit Fig. 10. PKCι acquires an inactive conformation after normal activity and can be rescued from ubiquitinylation and degradation by a Hsp-70-and keratindependent mechanism. The model is based on data originally published by Newton and coworkers for conventional PKC (Gao et al., 2006) and on the results in this work. TJ, tight junction; AJ, adherens junction. polyclonal anti-PKCι antibody (Santa Cruz Biotechnology) was used for immunoprecipitation.

Transgenic mice, analyses of intestinal epithelium, and immunofluorescence
The K8-null transgenic mice in the FVB/n background were originally obtained by Baribault and coworkers (Baribault et al., 1994), and the intestinal phenotype is described elsewhere (Ameen et al., 2001). Mice deficient in intermediate filaments in intestinal crypts [K18 +/-, K19 -/-, hK18 R89C (K18R89C animals) (in a mixed FVB/n ϫ C57BL/6 ϫ 129P2 genetic background)] and K8 overexpressors have been described (Hesse et al., 2007;Toivola et al., 2008;Casanova et al., 1999). Isolation of villus enterocytes was described by McNicholas et al. (McNicholas et al., 1994). For paracellular permeability assays, animals were deeply anesthetized, and a 10-cm loop comprising duodenum and the first part of the jejunum was cut, leaving the normal blood supply. A solution of 10 mM Texas-red-coupled Dextran 3000 (Invitrogen) in 1:1 H 2 O:PBS supplemented with 5 mg/ml glucose was instilled into the lumen and then the open ends of the loop were clamped. Blood was collected at time 0 from the tail, and by cardiac puncture at 50 minutes at the time of euthanizing the animal. Texas-red fluorescence was analyzed in 70-μl samples of serum in a spectrofluorometer. Myeloperoxidase assay was performed on mucosa scraped from 2 cm of jejunum just distal to the Dextran-perfused loop, as described elsewhere (Bradley et al., 1982) Procedures for immunofluorescence, frozen sectioning, and confocal microscopy have been described before (Ameen et al., 2001). Triton X-100 extraction before fixation is described elsewhere (Salas et al., 1988) and was performed with the following modification in the extraction buffer: PBS was supplemented with 0.5% Triton X-100, 1 mM MgCl 2 , 1 mM EGTA and the Sigma cocktail of antiproteases described above. Trichloroacetic acid (TCA) fixations were performed in 10% TCA in water for 10 minutes. Confocal microscopy was performed in a Leica TCS SP5 confocal microscope. Confocal stacks were normally collected at 0.4-μm intervals under a 63ϫ immersion objective. For XZ three-dimensional reconstructions, the entire stack was cropped in 7-voxel thick volumes, reconstructed, and rotated 90°u sing SlideBook 4.2 software (Intelligent Imaging Innovations).

Analysis of intermediate filament-dependent rephosphorylation of PKCι in the soluble fraction of CACO-2 cells
CACO-2 cells were grown and fractionated (S1, S2 and P) as described above, with the exception that the extraction buffer was not supplemented with phosphatase inhibitors. To induce the activity-dependent dephosphorylation of PKCι, the S1, S2 and P fractions were incubated in the presence of 150 μM PKC-substrate peptide (Upstate, cat. no. 12-536) and 1 mM ATP at 30°C with gentle shaking for 5 hours. After treatment, the peptide was removed by ultrafiltration in Centricon YM-3 (Millipore). To assess PKCι rephosphorylation, 50 μg of S1 fraction protein was then incubated with 20 μg of the P fraction protein or with 15 μg of purified IFs from CACO-2 cells [obtained as described by Steinert et al. (Steinert et al., 1982)] in the presence or absence of 1 mM ATP at 30°C for 4 hours. For isolated, non-filamentous recombinant keratins, 5 μg of K8, K18 or K19 were used instead of native IFs under the same conditions. Following incubation, the phosphorylation state of PKCι was examined by western blot using anti-pT555-PKCι antibodies.
To immunodeplete Hsc70 and/or Hsp70 protein, the S1 fraction was incubated with rabbit anti-Hsc70 and two different rabbit anti-Hsp70 antibodies together at 4°C overnight (non-immune rabbit serum served as a control). Then, protein-A beads (Santa Cruz Biotechnology) were added and incubated for 2 hours at 4°C, spun, and the supernatant used in subsequent in vitro PKCι rephosphorylation experiments.
We are deeply thankful to Yolanda Figueroa-Menendez for the excellent technical assistance, to Nevis Fregien for help with lentivirus preparations, and to Richard Rotundo for critically reviewing the manuscript. Supported by NIDDK (NIH) grants RO1DK057805 and R01DK076652 (to P.J.S.) and RO1DK47918 (to M.B.O.). The TROMA I monoclonal antibody developed by Philippe Brulet and Rolf Kemler was obtained from the Developmental Studies Hybridoma Bank, University of Iowa, Department of Biological Sciences. Andrea S. Oriolo was a recipient of a scholarship from Department of Defense training grant 4-49497-LS-HSI, and Flavia A. Wald was a recipient of a Crohn's and Colitis Foundation of America post-doctoral award. Deposited in PMC for release after 12 months.