Anomalous apical plasma membrane phenotype in CK8-deficient mice indicates a novel role for intermediate filaments in the polarization of simple epithelia

Over 50 genes encode intermediate filament (IF) proteins, in higher eukaryotes, which are expressed in a tissue specific and often redundant manner (Magin, 1998). In general, a family of proteins with these features would be expected to have key tissue-specific functions. However, the only known functions of IF proteins are mechanical. Unfortunately, in contrast with microtubules and microfilaments, the lack of specific ‘anti-IF’ drugs has prevented pharmacological approaches to study IF function. While there is no doubt that IF protein are essential for the mechanical properties of the epidermis and multi-layered mucosae (Fuchs, 1996), an emerging body of evidence points to other possible functions (Chou et al., 1997).

The epithelial IF proteins, the cytokeratins (CKs), are obligate heterodimers composed of a type I and a type II keratin. In simple epithelia, CK8 has been deemed to be the major type II CK available (Moll et al., 1982). Therefore, a transgenic knockout of the gene for CK8 was expected to render these tissues IF negative. Such a knockout was first created by Oshima and co-workers (Baribault et al., 1993) in the C57Bl/6 strain and was mostly lethal in utero. The majority of the knockout embryos died at 12 days of gestation with no apparent histological phenotype, except bleeding inside the liver. A second attempt in a different mouse strain (FVB/n), however, resulted in a fivefold increase in the percentage of knockouts born alive, with animals reaching adulthood. Thus, in the FVB/n mice, the CK8 knockout escapers display a relatively mild phenotype, comprising colorectal hyperplasia, anal prolapse, female infertility (Baribault et al., 1994) and a defect in hepatocyte integrity (Loranger et al., 1997). The latter correlates well with the liver phenotype of dominant-negative CK18 transgenic mice (Ku et al., 1996) and other CK18 mutations (Ku et al., 1997; Ku et al., 1998), as expected from the fact that adult hepatocytes express only CKs 8 and 18 (Moll et al., 1982). The molecular mechanisms responsible for this phenotype, however, remain unknown.

The interest of our laboratory in IF proteins in simple epithelia arose when we first identified a subpopulation of apical plasma membrane proteins directly attached to apical cortical IF protein that contained CK19 (Rodriguez et al., 1994). This finding supported the hypothesis that, in single layered polarized epithelia, cortical IF proteins may provide apical anchoring sites that facilitate the organization of the apical domain. In other words, IF proteins may be the apical counter-part of the well known fodrin-actin cortical cytoskeleton on the basolateral domain of the same cells (Mays et al., 1994). This potential role in the organization of epithelial polarity was further tested in two different human epithelial cell lines by partially down-regulating CK19 with antisense oligonucleotides, which delayed the acquisition of polarity, decreased the number of microvilli and mildly disorganized the apical population of microtubules (Salas et al., 1997). The last result suggested the possibility that IF proteins might be involved in the positioning of apical microtubule-organizing centers (MTOC) (Meads and Schroer, 1995; Buendia et al., 1990), a hypothesis that was recently confirmed in our laboratory in CACO-2 cells (Salas, 1999). All these results in tissue culture, strongly suggesting the participation of IF proteins in the organization of the apical domain, seemed contradictory with the mild phenotype of CK8-deficient FVB/n mice mentioned above. The important functions of apical membrane proteins in the intestine and kidney epithelia, among others, would predict such a phenotype to be severe or possibly lethal.

In the current studies we sought to resolve that apparent contradiction between the observations in tissue culture cells and the CK8-minus phenotype in vivo. We examined the extent of the lack of IF proteins in various epithelia using antibodies that recognize several CKs. The results indicate that, when IF proteins are effectively downregulated in the absence of the promiscuous CK7, alterations of the apical domain can be observed. Furthermore, these results support an as yet unsuspected role of IF proteins in the organization of the cytoskeleton.

CK8-deficient transgenic mice were a generous gift from Dr Hélène Baribault and 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. The animals were tagged by toe clipping under pentobarbital anesthesia (Kumar, 1979). The cut toes were used at the same time to obtain genomic DNA by proteinase K and phenol/chloroform extraction, and to PCR-genotype the animals, as described elsewhere (Baribault et al., 1994). Because of sterility of homozygous female CK8 knockouts (Baribault et al., 1994), FVB/n CK8 null mice were obtained by breeding heterozygous females and analyzing the genotype of the offspring. To harvest tissues, the animals were anaesthetized with pentobarbital (40 mg/kg).

The jejunum was rinsed in phosphate-buffered saline (PBS) supplemented with 2 mM EDTA and cut open. The pellet from the scraped mucosa was then extracted in PBS supplemented with 1% Triton X-100, 1.5 M KCl and a complete cocktail of antiproteases (Sigma). CKs were isolated by cycles of solubilization in 8M urea and repolymerization by dialysis in PBS (Steinert et al., 1982). The last pellet was subjected to 2D electrophoresis (IEF and SDS-PAGE; O’Farrell, 1975) and analyzed by Coomassie Blue staining.

Different procedures were used according to the organ and the antigen to be localized. For jejunum, two different procedures were followed: (1) some samples were rinsed in PBS, cut in short (2 mm) segments, embedded in OCT and frozen in isopentane at melting point; and (2) other samples were rinsed and cut in 10 mm loops, which were loaded with 1 mg/ml Lucifer Yellow CH (Molecular Probes, Eugene, OR) in PBS at 4°C, clamped and incubated in ice-cold PBS for 10 minutes. Then, the clamps at each end of the loop were released and the luminal solution rapidly changed to 3% formaldehyde in PBS without washes. The loops were then incubated in the same fixative for 2 hours, infused overnight in 27% (w/v) sucrose in PBS, and frozen in the same solution in melting isopentane. To localize microtubules, the same procedure was followed, omitting the Lucifer Yellow, washing and incubating always at 37°C, and supplementing the formaldehyde fixative with 0.2% (v/v) glutaraldehyde. In all cases, the blocks were stored at −60°C and cryosectioned at −25°C. Immunofluorescence procedures were performed as described before (Salas et al., 1997) with the following exceptions. The sections were thawed in fixative: for alkaline phosphatase, sucrase isomaltase and syntaxin 3 in 3% formaldehyde; for CKs, CFTR, Na⁺-K⁺ATPase and γ-tubulin, in 100% methanol at −20°C. The fixed sections that had been frozen in sucrose were thawed in PBS and not subjected to further fixation.

Kidneys were obtained from anaesthetized animals perfused by intracardiac injection of 3% formaldehyde. The kidneys were immediately removed and cut into 1 mm sections with a razor blade, incubated in the same fixative for 2 hours and infused overnight in 27% sucrose in PBS. Pieces containing cortex and medulla were frozen in the same sucrose solution in melting isopentane. Samples of liver and uterus were rinsed in ice-cold PBS, rapidly cut into 1 mm cubes (or sections perpendicular to the lumen in the case of uterus) and frozen in OCT as described above. Frozen sections were thawed in PBS (kidney) or methanol (liver and uterus) and processed for immunofluorescence.

The primary antibodies were polyclonal anti-intestinal calf alkaline phosphatase (iAP, Dako); mAb anti-sucrase isomaltase was a generous gift from Dr H. P. Hauri (Department of Pharmacology, Biozentrum, University of Basel, Switzerland; Hauri et al., 1985); anti-PAN cytokeratin mAb clone C-11 (Sigma, St Louis, MO); TROMA (rat monoclonal) antibodies (Developmental Studies Hybridoma Bank); R3195 polyclonal anti-CFTR antibody was a gift from Dr C. R. Marino (Veterans Affairs Medical Center, University of Tennessee); polyclonal antibody against a synthetic polypeptide comprising amino acids 38-53 (EEFATEGTDRKDVFFY) of the N-terminal region of human γ-tubulin (Sigma); polyclonal antibody anti-Na⁺-K⁺ATPase was a generous gift from Dr W. James Nelson (Stanford University, Stanford, CA); mAb against α-tubulin (DM 1A) (Sigma); mAb anti-CK7 (RCK105, Euro-Diagnostica/Accurate Chemical, Westbury, NY); polyclonal Ab anti-bile canaliculus HA4, ECTO-ATPase (Margolis et al., 1988, Margolis et al., 1990); polyclonal Ab against synthetic peptide (KDRLEQLKAKQLTQD) of syntaxin 3 (Alomone Labs, Israel); polyclonal Ab anti-carbonic anhydrase IV (a generous gift from Dr William S. Sly, St Louis University School of Medicine, St Louis, MO; Schwartz et al., 1999); and polyclonal Ab anti-ASGP-2/Muc 4 (a generous gift from Dr Kermit Carraway; Idris and Carraway, 1999).

To ensure no crossreactivity with endogenous mouse immunoglobulins, the secondary antibodies (anti-rabbit and anti-rat) were grown in mouse and affinity purified (Jackson ImmunoResearch Laboratories). The primary mouse monoclonal antibodies were biotinylated in the carbohydrate residues (O’Shannessy, 1987) with biotin-LC-hydrazide (Pierce) and localized in the tissue with Cy3-conjugated streptavidin (Jackson ImmunoResearch Laboratories). Biotinylated antibodies were used freshly after quenching and dialysis of free biotin (storage of these antibodies after biotinylation resulted in a marked loss of reactivity). Alternatively, in the case of RCK105 and DM 1A, the tissues were pretreated with an excess of anti-mouse immunoglobulin Fab fragments to quench the endogenous signal for the secondary antibody. Negative controls were performed with non-immune IgG of the same species or biotinylated non-immune mouse IgG.

The preparations were observed and photographed with a Leica DM RB epifluorescence microscope. For localization of microtubules, the sections were observed with an Olympus Fluoview laser confocal microscope using a 60 μm pinhole with a 1.4 NA lens.

The relative distribution of a membrane protein in the basolateral versus apical domains of hepatocytes was determined by digital image analysis, a method adapted from previous publications in MDCK cells (Rodriguez-Boulan et al., 1989). Briefly, 5 μm liver frozen sections were digitized with a Leica LEI-470 video camera using a Leica Quantimet 500+ video analysis system. The camera/digitizer gain was set so that none of the significant pixels (either bile canaliculus or vascular pole of the plasma membrane) were out of the linear range, and kept constant throughout the measurements. Lines crossing both bile canaliculus and the vascular pole of the same cell were marked at random and the peak pixel values of the intensity profile on each membrane were taken. One ratio of fluorescence intensity values (basolateral/apical) per cell was calculated after subtracting the average background value measured on nuclei. The significance of the difference between means was estimated by non-paired Student’s t test.

To assay for apoptosis, a DNA fragmentation detection kit (Oncogene Research Products, Cambridge, MA), based on the end labeling of fragmented DNA with biotin-labeled deoxynucleotides by Klenow DNA polymerase I, was used. The biotinylated nucleotides were then labeled with CY3-coupled streptavidin (Jackson ImmunoResearch Laboratories) and visualized by epifluorescence microscopy.

Mouse jejunum was fixed in 3% glutaraldehyde as described previously (Salas et al., 1997) and embedded in Spurr. Thick sections for light microscopy were counterstained with 1:1 Toluidine Blue and Malloy’s Azur II (1% azur II, 1% methylene blue), and observed under phase-contrast microscopy. Thin sections were counterstained with lead citrate and photographed with a Jeol JEM-100 CX II electron microscope.

The mild phenotype of FVB/n CK8 null mice escapers prompted us to hypothesize that IF proteins (IFs) may be still present in a variety of epithelia in these animals. In light of Baribault et al.’s observations that CK18 and 7 are present in intestinal crypts, our first task was to determine if the intestinal epithelium displays any IFs at all in the CK8-deficient mice (Baribault et al., 1994). IFs were prepared from jejunal intestinal mucosa by detergent-high salt extraction, and the corresponding cytokeratins purified by cycles of urea solubilization and re-polymerization in phosphate buffered saline (Steinert et al., 1982). When such a preparation from a heterozygous animal was analyzed by 2D electrophoresis (IEF and SDS-PAGE), five cytokeratins were found. The two spots of lower MW and lower pI (type I CKs) were identified as CKs 19 and 18 (from lower to higher MW, Fig. 1A; parallel immunoblot, not shown), the third type I CK was tentatively identified as CK 20, in accordance with previous publications (Chandler et al., 1991; Flint et al., 1994). Likewise, CK8 was identified (Fig. 1A, arrow, blots not shown). An additional type II CK of higher MW was also found, and identified as CK7 in parallel immunoblots with RCK105 antibody (data not shown), a result that confirms previous observations in the mouse intestine (Baribault et al., 1994). As expected, CK8 disappeared in the CK8 knockout, along with two of the type I spots (Fig. 1B). Because some degradation is possible during CK purification, these experiments do not rule out the possibility that other CKs may also be present but clearly point at the persistence of CKs 18-7 pairs in the knockouts.

To analyze the distribution of these persistent IF, frozen sections of intestinal tissues from Ck8^+/− and Ck8^−/− were stained with antibodies that recognize a broad range of cytokeratins (anti-PAN cytokeratin): biotinylated C11 mAb (Bartek et al., 1991) and TROMA 1-3 cocktail (Brulet et al., 1980). For both antibodies the images were identical. The Ck8^+/− animals showed a typical CK distribution in all epithelial cells. In the enterocytes the signal was typically concentrated in the cortical cytoskeleton with a clear higher density in the apical region (terminal web) (Fig. 2A,C). As expected from the data in Fig. 1, some CK IFs were observed in the knockouts. The CK signal was restricted to the apical submembrane region of the crypt cells (Fig. 2B, arrows) and to the goblet cells. However, IFs were clearly absent in villus enterocytes (Fig. 2D).

To analyze the expression of apical membrane proteins in small intestinal enterocytes, we used three markers. Intestinal alkaline phosphatase (iAP) is a glycosyl-phosphatidyl inositol anchored protein, while sucrase isomaltase and the cystic fibrosis transmembrane conductance regulator (CFTR) are integral membrane proteins. As expected, in all three cases, these proteins were apically distributed in the heterozygous mice (Fig. 3A,C,E). However, they were not present in the apical domain of villus enterocytes of the upper two thirds of the villus in the knockouts (Fig. 3B,D,F). Moreover, none of the other subcellular localizations of these markers (e.g. a faint basolateral distribution of sucrase isomaltase, Fig. 3D, or a supranuclear localization of iAP, Fig. 3B) was affected by CK8 knockout.

The effect of the knockout was dependent on the position of the enterocytes along the crypt-villus axis. Apical membrane markers were normally expressed in crypt cells in the knockout (Fig. 3H, shown only for CFTR). Likewise, the surface expression of all markers examined was indistinguishable between knockouts and littermate heterozygous in the proximal third (closer to the crypts) of the villi (not shown). The analysis of the basolateral marker Na⁺-K⁺ATPase showed no changes in its normal distribution in the intestinal epithelium of Ck8^−/− animals (not shown).

The ultrastructure of the brush border of villus enterocytes was analyzed by phase-contrast microscopy using plastic embedded sections (Fig. 4A,B) and by transmission electron microscopy (Fig. 4C,D). At the light microscopic level, the knockout mice showed patches of enterocytes (ranging from one to several cells) randomly distributed on the villus, with a clearly thinner brush border (Fig. 4B, small arrowheads, compare with the brush border to the left of the arrows). These patches represented a modest percentage of the total surface (5-9%, as determined by the ratio of section lengths: Si/St=4/π Li/Lt, where Li is the length of section of brush border with shorter microvilli and Lt the total length of brush border in that section, and Si/St represent the fraction of surfaces covered; see Weibel, 1979). Examination by electron microscopy revealed that the microvilli corresponding to the ‘thin’ brush border areas (Fig. 4D) were present, about 20% of the normal microvillus length (Fig. 4C), and less packed, but otherwise showed a normal ultrastructure. Some of the cells with thin brush borders also displayed dilated ER cisternae in the apical cytoplasm (Fig. 4D), in contrast with normal enterocytes (Fig. 4C). These results, however, indicate that the lack of apical membrane proteins is not due to a gross loss of brush border. Because these ultrastructural changes may suggest some mechanical damage in CK8 knockout enterocytes, we decided to analyze whether the lack of apical membrane markers in the apical domain of villus enterocytes may be due to mechanical injury of cells.

Because CK8 null hepatocytes are more susceptible to tumor necrosis factor-induced apoptosis (Caulin et al., 2000) and Ifs are generally considered essential for the mechanical strength of cells, the possibility arose that the lack of apical membrane markers described above might be due to either increased numbers of apoptotic cells in the distal segment of the villus, or simple mechanical damage of the enterocytes. The latter would cause membrane permeabilization, loss of the brush border and, ultimately, necrosis. To test these possibilities, we analyzed membrane permeabilization and apoptosis in enterocytes of the distal segment of the villus, and correlated them with the lack of expression of apical membrane markers in these cells.

Traditionally, damaged cells have been identified in tissue culture by the permeability of their plasma membrane to Trypan Blue. This technique, however, cannot be used in tissues because the dye is soluble even after fixation and would diffuse away from the permeabilized (damaged/necrotic) cells during fixation and sectioning/processing. Therefore, we used Lucifer Yellow CH, a dye of similar MW to Trypan Blue (457 versus 960 Daltons, respectively) which is fixed in situ by aldehydes without any loss of fluorescence. Lucifer Yellow is not permeable through intact plasma membranes and has been extensively used as an intracellular marker. Loops of jejunum were incubated in the cold with Lucifer yellow in PBS in the lumen, and then fixed with formaldehyde without washes. As expected, desquamating epithelial cells at the tip of the villi were brightly positive both in control and knock-outs (Fig. 5A,B, arrows). Lucifer Yellow was also trapped/fixed in the space among microvilli (possibly within the glycocalyx), and became a reliable – although extracellular – marker of the brush border, a by-product of this technique. In addition to the desquamating cells at the tips of the villi, a very small population of enterocytes was marked by luminal Lucifer Yellow (e.g. Fig. 5E, arrow). These cells (possibly necrotic) represented less than 1% of the villus enterocytes and were randomly distributed in the epithelium. The fraction of Lucifer Yellow-positive enterocytes was identical in the knockouts and their heterozygous littermates. When Lucifer Yellow was co-localized with iAP signal, both co-localized to the brush border, thus demonstrating the extracellular trapping of the former described above (Fig. 5C,E). The very rare enterocytes permeable to Lucifer Yellow lacked iAP signal in the apical domain (Fig. 5C,E, arrow). In the knockouts, the iAP signal disappeared from the apical domain in the middle third of the villus (Fig. 5D), as described in Fig. 3. However, co-localization with Lucifer Yellow in the same field showed that the brush border was intact, and that none of the enterocytes lacking iAP were permeable to Lucifer Yellow (Fig. 5D,F).

To analyze the possibility that the lack of apical membrane markers may be due to an increased number of apoptotic enterocytes in the distal half of the villus, the presence of apoptotic cells was determined by DNA fragment end labeling. As described before (Westcarr et al., 1999), a cap of enterocytes at the very tip of the villus were apoptotic. Normally, three to eight cells per section of the villus tip are positive for DNA fragmentation (Fig. 5G, arrow). Interestingly, co-localization with iAP signal in the same field indicated that apoptotic cells display clearly polarized apical iAP signal, although with lower intensity than in the non-apoptotic neighbors (compare Fig. 5I with 5G, arrow). The extent of the apoptotic caps at the villus tips in knockout animals was identical to that in the heterozygous littermates (Fig. 5H, arrow). Co-localization of iAP signal in the same field indicated that the downregulation of apical membrane markers in CK8-deficient enterocytes occurs distant to the area of apoptosis at the villus tip. Because of the displacement of enterocytes in the crypt-villus tip axis, this indicates that apical iAP disappears from the apical domain at a much earlier time than the occurrence of apoptotic DNA fragmentation (Fig. 5J). Considering that truly apoptotic cells still display moderate levels of apical iAP in the control (Fig. 5I), it seems safe to conclude that the lack of apical markers is not due to an increased number of apoptotic cells in the knockouts.

Our recent observation that apical IFs attach and localize MTOCs (Salas, 1999) suggested the need to analyze MTOC distribution in CK8-deficient villus enterocytes that lack terminal web IF (Fig. 2D). As expected from previous work in tissue culture cells (Meads and Schroer, 1995; Salas, 1999), a layer of non-centrosomal γ-tubulin was found underlying the terminal web (Fig. 6A, arrows) in villus enterocytes from heterozygous siblings. In CK8-deficient cells, this layer was replaced by a homogeneous cytoplasmic distribution. In some animals (∼30%), γ-tubulin was also found in supranuclear accumulations (Fig. 6B). Such accumulations were observed in Ck8^+/− cells limited to the crypt-villus boundary, but never extended to the villus, as was evident in a fraction of the knockouts (Fig. 6B). The layer of non-centrosomal MTOC γ-tubulin was apical and identical in knockouts and controls in the crypts (Fig. 6C,D), where IFs persist in the knockout.

To determine whether the redistribution of γ-tubulin has an effect on the architecture of microtubules, jejunal mucosa was fixed to preserve microtubules (at 37°C and in the presence of glutaraldehyde) before cryosectioning. The distribution of polymerized α-tubulin was observed by indirect immunofluorescence on frozen sections using a confocal laser microscope with the smallest pinhole setting to achieve the highest possible degree of confocality. In this way, single MT and the general microtubular architecture could be observed. In heterozygous littermates, MT were always oriented in the apico-basal axis and highly concentrated in the supranuclear cytoplasm. Few MT extended around and below the nuclei, and were always parallel to the same axis (Fig. 6E,G). This type of distribution is consistent with observations in other epithelial cells (Bacallao et al., 1989; Salas, 1999) and with the distribution of γ-tubulin described above. The architecture of MT was identical in the knockouts, except in the enterocytes in the upper third of the villus. In these cells, contrasting with the images observed in controls, MT appeared disorganized, often perpendicular to the apico-basal axis or radiating from perinuclear nodes (Fig. 6F,H, arrows). Fewer MT were observed attaching to the apical domain than in controls (Fig. 6F,H, versus 6E,G). Additionally, in these images it is easier to show that CK8 null enterocytes in the distal third of the villus were often shorter (in the apico-basal axis) than normal cells.

To understand the mechanism underlying the lack of apical membrane markers in the upper half of villus enterocytes, we analyzed the expression of syntaxin 3, considered a key element of the machinery that mediates membrane fusion of apically bound exocytic vesicles in enterocytes (Delgrossi et al., 1997; Riento et al., 1998). As reported before, syntaxin 3 was localized to the brush border of all crypt and villus enterocytes (Fig. 7A), including the apoptotic ones at the tip of the villus (not shown) in heterozygous littermates. In the CK8 null mice, syntaxin 3 was present in the crypts and in the epithelium at the boundary between the opening of the crypts and the base of the villi (Fig. 7B, large arrow), but conspicuously absent in the villus enterocytes from the very base of the villi. In Fig. 7B, the small arrows point at the position of the negative brush border of the lower third of two adjacent villi. That position was verified by co-localization in the same field with luminal Lucifer Yellow trapped and fixed in the space among the microvilli (Fig. 7C). The loss of syntaxin 3 that appeared at the lower (proximal) third of villi, therefore, is an event that occurs earlier than the loss of iAP, sucrase isomaltase or CFTR, which occurred at the middle third of the villi (Fig. 3).

The observations in the small intestine described above prompted us to study other simple epithelia. In every case we analyzed the expression of IFs using an anti-pan-cytokeratin Ab and the expression of CK7, another type II CK expressed in several simple epithelia that is potentially promiscuous, and may replace CK8 in the knockouts. Two organs arose as interesting candidates to analyze apical polarity: the kidney and the liver. In the former, CK7 expression has been reported to be restricted to the thin limbs of Henle and the medullary segment of collecting ducts in the human (Moll et al., 1991). So we hypothesized that IFs would be missing in CK8 null mice in all other segments of the nephron. In liver, however, it has been reported that CK8 is the only type II CK in hepatocytes (Moll et al., 1982).

The analysis of IFs in kidney led to an immediate surprise. Against the prediction based on previous publications, the immunofluorescence analysis of formaldehyde perfused kidneys from Ck8^−/− animals showed CK7 expression in the majority of the tubule sections. In proximal tubules (PT), for example, CK7 signal appeared highly concentrated to a terminal web-like structure in the apical domain (Fig. 8A, arrows). Accordingly, two different apical membrane markers, carbonic anhydrase IV and iAP, that label segments I and II, and segment III of the PT, respectively (Brown et al., 1990; Nouwen and De Broe, 1994), appeared in their expected distribution and were well polarized (not shown).

In the liver we verified previous reports indicating the lack of CK7 in hepatocytes and the ablation of IF in the hepatocytes of CK8 null mice (not shown). Thus, hepatocytes became the second example in which the CK8 knockout results in true downregulation of IFs. Immunofluorescence label using a polyclonal antibody (HA4) against the bile canaliculus Ecto-ATPase, showed apical (bile canaliculus) staining in the hepatocytes of Ck8^+/− littermates (Fig. 8B). A very faint basolateral signal, very difficult to photograph, was picked up by digital image analysis of these preparations. When digitalization was carried out within the linear range of response of the camera and digitizer, ratios of fluorescence were calculated for the signal at the vascular pole/bile canaliculus of hepatocytes. In control animals (Ck8^+/−) this ratio was 0.2, indicating that the signal in the bile canaliculus was, on the average, fivefold stronger than the signal in the vascular pole (Table 1). In the CK8 knockouts, HA4 signal was still detected in the bile canaliculi, but a distinctively higher level of fluorescence was observed in the basolateral domain of hepatocytes (Fig. 8C). In digitized images, the ratio of HA4 fluorescence basolateral/apical became 0.8 (Table 1), indicating that, although bile canaliculi were still brighter than the vascular pole, the polarization of HA4 in the knockout was poorer than in their siblings expressing IFs. Apoptotic cells were not observed by DNA fragmentation detection in CK8 null mouse livers (not shown).

Finally, because Ck8^−/− females are sterile, we also analyzed the epithelia in the uterus. However, CK7 and IFs were observed in both glandular and luminal epithelia. Likewise, an apical membrane marker in rodent uterus, ASGP-2 (Idris and Carraway, 1999) was found to be well polarized in these cells (not shown). The results for the expression of CK7, IFs and the polarization of apical membrane markers in various organs are summarized in Table 2. It can be concluded that only in the two tissues where IFs were truly absent in CK8 null mice, problems in the organization of apical polarity were observed. It is puzzling, however, that the effect of the lack of IFs in enterocytes and hepatocytes is different. In the first case apical membrane proteins disappeared from the apical domain. In the second case, HA4 levels at the bile canaliculus seemed quite normal and an increase in the basolateral signal appeared to be responsible for a decrease in the degree of polarization.

The results of the current study indicate that CK8 null mice express IFs in a number of simple epithelia, presumably because of the almost ubiquitous expression of CK7. Interestingly, in mouse kidney, the expression of CK 7 (Fig. 8A) is much more widespread than previously reported in humans (Moll et al., 1991). Whether this is a species-or strain-specific feature, or whether an increased transcription of the CK7 gene enables survival of CK8 null fetuses is beyond the scope of this work. In ten different simple epithelia analyzed in five different organs (jejunum, colon, liver, kidney, and uterus), only in two cases IFs were manifestly missing: villus enterocytes and hepatocytes. In all other cases CK7 expression was verified and apical membrane markers displayed normal polarization (Table 2). The changes in apical polarization in enterocytes and hepatocytes, therefore, correlated well with the absence of IF. The phenotype, however, was different in each case and will be discussed separately.

To better interpret the data in villus enterocytes, it is important to remember that intestinal epithelial cells originate from a small set of stem cells located near the deep part of crypts, the only site of mitosis. From there, differentiating enterocytes and goblet cells slowly migrate first to the opening of the crypts and then up the villus, desquamating from the tip of the villus just a few (3-4) days later (Schmidt et al., 1988). Therefore, enterocytes found at the base of the villus have been localized at the crypts only a few hours before. In general, the position in the crypt-to-villus axis is equivalent to the point in the life time of the enterocyte and its differentiation. It is also accepted that important changes in enterocyte differentiation occur when the cells exit the crypts and move up the villi. For instance, changes in the expression of CK18 (Flint et al., 1994) and CK20 (Chandler et al., 1991) have been reported. Baribault and co-workers found CK7 expression limited to the crypts and, therefore, also switched off as the cells migrate onto the villus (Baribault et al., 1994). The expression of CK7-CK18 heterodimers (Fig. 1) in the crypts of CK8 knockouts (Fig. 2), therefore, is fully consistent with the expression of a redundant type II keratin (CK7), promiscuous for the function of the missing CK8.

The defects that we found in the Ck8^−/− phenotype can be ordered along the crypt-villus axis as follows: disappearance of IFs (opening of crypts) → lack of apical syntaxin 3 (proximal third of villi) → lack of apical membrane markers (middle third of villi) → changes in the localization of MTOCs and disorganization of MTs (distal third of villi).

Because of the migration of enterocytes, these changes also reflect a temporal order in which these defects appear. While the order of precedence does not imply a cause- and-effect relationship, it helps to rule out some possibilities. Specifically, the disorganization of MTs occurs after the disappearance of apical membrane proteins, so it cannot be its direct cause. However, it is generally accepted that syntaxin 3 participates in the insertion of newly synthesized apical proteins. Thus, it is quite likely that, with blocked exocytosis, enterocytes will slowly remove the apical proteins already in the brush border by normal turnover. In other words, depending on the rate of removal, the steady-state levels of a given apical protein will be lower hours or days after downregulation of syntaxin 3. Therefore, we speculate that the effect on apical polarity may well depend on the turnover of each specific component. Some apical features may be stable enough to persist until the enterocyte reaches the tip of the villus and desquamates. That may be the case for microvilli, that appeared normal in more than 90% of the enterocytes, including those at the tip of the villus. In fact, the patches of short microvilli (Fig. 4B), which represent less than 10% of the surface and had no obvious correlation with the crypt-to-villus axis, may represent the inability of the cells to fully recover from an injury to the apical domain in the absence of IFs (or in the absence of syntaxin 3). Paradoxically, no gross damage was observed in terminally differentiated enterocytes (located near the tip of villi) that lack IFs in the Ck8^−/− mice. These cells, therefore, seemed mechanically sound, indicating that IFs may be necessary in internal epithelia for non-mechanical reasons. Obviously, our knowledge of the molecular mechanisms underlying the role of IF in the organization of the apical domain is still very limited, and further investigations will be necessary to fully understand the features of this phenotype. The results in this work seem to point to an unexpected role of IFs in the subcellular localization of syntaxin 3, a possibility that is currently under investigation in our laboratory.

Because of the results in this work, it is intriguing that CK8 knockout mice do not show signs of a malabsorption syndrome, although Baribault et al. reported watery stools (Baribault et al., 1994). The lack of malabsorption may be explained in two different, non-excluding ways.

First, it is well known that the surface of the small intestine mucosa represents an excess with respect to the minimum needed for normal digestive and absorptive functions. In rats, up to 45% of the small intestine can be removed without any significant absorptive compromise (Menge and Robinson, 1978). Since only a fraction of the enterocytes seem to be affected in the knockout (those in the distal half of villi), it is conceivable that the surface of villi covered by still normal enterocytes is sufficient to carry out the necessary absorptive and digestive functions. In other words, assuming that the enterocytes in the distal half of villi are unable to carry out absorptive/digestive functions, the CK8 knockout should be functionally equivalent to a partial resection of the small intestine. In this regard, it is known that a decrease in the absorptive surface of the intestine (either after surgical resection or in short bowel syndrome), results in hyperplasia and inflammation of the distal remnant of intestine (Chang and Rao, 1994). In rats, that response has been attributed to specific nutrients, such as pectin, that become overloaded in the distal remnant segment of ileum and the colon (Booth, 1994). Thus, the macroscopic intestinal phenotype of CK8 null mice described before, colorectal hyperplasia, is perfectly consistent with the lack of apical membrane proteins in a fraction of small intestine enterocytes shown here.

Second, the lack of three proteins does not imply that all the apical components are missing. In fact, because microvilli are normal in the majority of the enterocytes, it can be speculated that other apical features may be normal as well. In other words, digestive enzymes and membrane channels (e.g. membrane proteins with very slow turnover) other than iAP, sucrase isomaltase or CFTR may be still apically localized in the terminally differentiated enterocytes.

Finally, it has been shown that apical membrane proteins in enterocytes participate in the control of the intestinal microflora (Hooper et al., 2000). The absence of those proteins may, therefore, cause changes in the normal bacterial content of the intestine. We speculate that the colon hyperplasia is not an intrinsic problem of the colon, where, after all, the epithelium does have IFs (Table 2). Rather, it may rather be secondary to slight functional defects of the small intestine resulting from the lack of apical membrane proteins in the distal half of villi.

Our data indicate that the number and distribution of apoptotic cells in the intestinal mucosa is similar in knockouts and controls (Fig. 5). Furthermore, in heterozygotes, apoptotic cells do display apical iAP, albeit at a reduced level. The data do not indicate any relationship between necrotic (Lucifer Yellow-positive) cells, which represent less than 1% of the population and are never seen in patches, and the patches of cells with shorter microvilli (Fig. 4), about 5-10% of the population. Neither does there seem to be a relationship between the enterocytes lacking apical membrane markers, which are restricted to the upper half of the villi, and the patches of cells with short microvilli or the apoptotic cells. In short, the data indicate that the downregulation of apical membrane markers is not due to cell damage or apoptosis.

The CK8 null mice display an interesting liver phenotype. In general, it has been shown that the lack of IFs (either by knockout of CK8 or dominant negative mutation of CK18) renders hepatocytes more labile to hepatotoxic drugs (Toivola et al., 1998; Ku et al., 1996), prolongs the effect of barbiturates, drugs cleared solely by the liver (Loranger et al., 1997) and sensitizes the cells to the apoptotic effects of tumor necrosis factor (Caulin et al., 2000). In the absence of injury, however, livers of Ck8^−/− mice show little or no morphological anomalies (Loranger et al., 1997). Unlike in enterocytes, HA4, an apical marker, was well expressed in the bile canaliculus surface, but its levels were increased in the basolateral surface. Three reasons can be argued to explain this difference. First, hepatocytes are a quiescent, long lived population of cells, while enterocytes originate, differentiate and desquamate in 3-4 days. If a defect in the exocytosis pathway exists in IF-deficient mice, it is conceivable that it may be by-passed by parallel or compensatory mechanisms in long-lived cells. Second, syntaxin 3, although expressed in hepatocytes, is not entirely apical, but significantly present in transport vesicles, while syntaxin 2 is also apical (Fujita et al., 1998). As our current knowledge about the relationship between IFs and subcellular localization of syntaxins is embryonic, it remains a matter of speculation that only syntaxin 3 or only a group of syntaxins may be affected by the lack of IFs. Finally, it has been known for years that the mechanisms of polarization in hepatocytes and enterocytes are different, at least for some proteins. In the former, all apical membrane proteins make an intermediate stop at the basolateral domain in route to their apical destination (Ihrke and Hubbard, 1995), while in the latter, only some proteins do so, and others are vectorially inserted in the apical domain (Nelson, 1992; Mostov, 1995). The participation of IFs in the organization of each pathway is currently unknown.

The results in this work suggest that IF may participate in the tissue-specific determination of subcellular domains in multicellular organisms, a set of functions that would justify the great redundancy and tissue-specific expression of the IF family of proteins. For example, IFs have been implicated in the localization of mRNAs in Xenopus oocytes (Forristall et al., 1995), of proteasomes (Palmer et al., 1994), of mitochondria in muscle (Reipert, 1999), in the positioning of MTOCs in various cell types (Salas, 1999; Trevor et al., 1995; Pockwinse et al., 1997) and in connecting microfilaments (Yang et al., 1996). Understanding possible non-mechanical functions of the IF family may shed light on as yet unsuspected ways in which the cytoskeleton participates in the generation of cellular and tissue architecture.

The authors thank Dr Hélène Baribault, not only for providing the CK8 knockout mice, but also for her key suggestions and helpful criticisms during the performance of this work. It is a pleasure to thank Dr Ann L. Hubbard (Johns Hopkins University) for providing us with HA4 polyclonal antibody and Dr Richard Rotundo for critically reading the manuscript. We are grateful to Ms Susan Decker for her skillful technical assistance. This work was supported by a grant from National Heart, Lung and Blood Institute to P.S.