In Vivo, Villin Is Required for Ca²⁺-Dependent F-Actin Disruption in Intestinal Brush Borders

A 6.5-kb BamHI fragment was isolated from a λDASHII phage containing 16 kb of the mouse villin gene (kindly furnished by G. Tremp, Rhône Poulenc RORER) and subcloned in pBS/KS+ (Stratagene). The neomycin resistance gene (pMC1neo; Stratagene) was introduced in a unique KpnI site in villin exon 2. This disrupts the open frame reading of the villin gene. The neo cassette was placed in the reverse transcriptional orientation compared with the villin gene. The construct contained 3.5 and 3 kb of homology regions, 5′ and 3′, respectively. In addition, the Herpes simplex virus (HSV) thymidine kinase expression cassette (Thomas and Capecchi 1987) was inserted in the unique ClaI site flanking the 5′ end of the construct. This resulted in the pvillin neo targeting construct.

The CK35 ES cell line (Kress et al. 1998) was cultured in DMEM (GIBCO BRL) supplemented by 1 mM Na-Pyruvate, 15% FCS (Techgen International), 1,000 U/ml LIF (ESGRO; GIBCO BRL), and 50 mM β-mercaptoethanol (GIBCO BRL) as described (Cohen-Tannoudji et al. 1998). 2 × 10⁷ CK35 ES cells were electroporated with 20 μg of the pvillin neo targeting construct linearized in the plasmid backbone (PvuI site). G418 (300 μg/ml) was added 36 h after plating for 12 d. Gancyclovir (2 μM) was added with G418 from day 2 to day 8 of selection. The G418-resistant clones were isolated and their genotype analyzed by Southern blot.

Chimeric mice were generated by microinjection of the targeted ES cells into C57Bl6 blastocysts as described (Bradley 1987). Crosses between chimeric male and C57Bl6/DBA2 females generated heterozygous animals which were then intercrossed to generate homozygous animals. The following experiments were performed in sibling villin-null mice and wild-type animals with the same mixed genetic background.

The mice were killed with an intraperitoneal injection of a lethal dose of pentobarbital, or anesthetized when necessary with a 50 μl/10 g body wt of a mixture of 750 μl xylazine (Rompun™ 2%; Bayer), 6 ml Imalgene™ (Rhône Merieux), and 300 μl Flunitrazepan (20 mg in 5 ml 100% ethanol; Sigma Chemical Co.) dissolved in 12 ml of PBS. The abdominal cavity was opened and the blood sample was obtained from aorta in heparinized tubes for further centrifugation and plasma-obtaining or EDTA-containing tubes (Hémo Kit 200; Melet Schoesing Laboratories) for blood cell counting. The intestine was then isolated and its length measured. Schematically, the intestine was divided in three identical parts in length corresponding to the duodenum, jejunum, and ileum. The proximal large intestine was isolated near the caecum and the distal part close to the rectum. The intestinal tube was washed with PBS containing 1 mM CaCl₂, 1 mM MgCl₂. Kidney was sampled and frozen in liquid nitrogen.

Brush border membranes were prepared from mucosa freshly scraped from the whole small intestine. The mucosa was diluted in 10 vol/mg wt with buffer A (10 mM imidazole, 5 mM EDTA, 1 mM EGTA, pH 7.4, 0.2 mM DTT, 200 μg/ml Pefabloc, 1 μl/ml of a mixture of protease inhibitors: 1 μg/ml antipapaine, 1 μg/ml pepstatine, 15 μg/ml benzamidine) (Sigma Chemical Co.) and stirred at 4°C for 1 h. A mechanical cellular disruption was then obtained with 5 strokes in a Dounce homogeneizer (model S852; Braun). After centrifugation (10 min, 4°C, 1,000 g), the pellet was washed 3 times with 10 ml buffer A followed by centrifugation. At this step, two procedures were carried out: either the crude membranes were directly used for studying the Ca²⁺ effect (see below) or brush border membranes were purified in a sucrose gradient. For the latter procedure, the pellet was resuspended in 10 ml buffer B (75 mM KCl, 5 mM MgCl₂, 1 mM EGTA, 10 mM imidazole, pH 7.4, and similar protease inhibitors to those in buffer A) and then mixed with a sucrose solution (40% sucrose final concentration in buffer B). This sample was collected in a tube containing the same volume of 65% sucrose (in buffer B), and centrifugation was performed at 4°C, 30 min, 15,000 g. The purified brush border membranes were obtained at the interface of the 40%:65% sucrose gradient. The protein content was determined and normalized with the initial weight of scraped mucosa.

In some experiments, brush border crude preparation was incubated for 10 min in a solution containing 10⁻⁵ to 10⁻³ M CaCl₂ in imidazole buffer (10 mM, pH 7.4, without EGTA and EDTA, plus protease inhibitors similar to those in buffer A). Microscopic observation was immediately performed using Nomarski optics (DMRD) or transmission electron microscopy after fixation (see below).

DNA from embryonic stem (ES) cell pellets or mouse tails was extracted in a lysis buffer (50 mM Tris, pH 7.5, 0.1 M NaCl, 0.5% SDS, 5 mM EDTA, 100 μg/ml proteinase K), and analyzed by Southern blots or PCR.

Southern blots were performed using DNA digested with HincII or Sca1 restriction enzymes and analyzed with a 0.4 kb BamHI-HincII fragment (pr A, Fig. 1 A), and a 0.5-kb BglII-StuI fragment (data not shown) as 3′ and 5′ external probes, respectively.

PCR analysis was performed using DNA in 50 μl for 30 cycles. Each cycle consisted of 60 s at 94°C, 60 s at 67°C, and 90 s at 72°C, using the 0.5 μl of Taq polymerase (Quantum Biotechnologies, Inc.) in a homemade buffer (67 mM Tris HCl, pH 8.8, 16.6 mM [NH₄]₂SO₄, 0.45% Triton X-100, 0.2 mg/ml gelatin, 2 mM MgCl₂). 2 pmol of 5′-GTCAAAGGCTCTCTCAACATCAC-3′ villin sense oligonucleotide (primer 1), 1 pmol of 5′-GACTACATAGCAGTCACCATC-3′ villin antisense oligonucleotide (primer 3), and 1 pmol of 5′-TCTGGATTCATCGACTGTGGC-3′ neo antisense oligonucleotide (primer 2) were used, generating a 0.9-kb fragment for the wild-type allele and a 1.2-kb fragment for the targeted allele (Fig. 1A and Fig. C).

Total RNA was extracted from different tissues using RNA Now kit (Ozyme).

For Northern blot analysis, RNA samples (10 μg) were separated on 4.2% formaldehyde-agarose gel and blotted onto nylon membrane (Amersham). The ³²P-labeled probe was synthesized with TransProbe T kit (Pharmacia Biotech). The RNA probe used corresponded to 530 bp of the 3′ end of the smaller human villin mRNA (Pringault et al. 1986).

For reverse transcription PCR analysis, 10 pmol of 5′-TCCAGCCAGCACATTCCTCTTCCC-3′ was hybridized with 2 μg of total RNA at 70°C for 10 min in distilled water. Reverse transcription with 200 U of Moloney murine leukemia virus reverse transcriptase (SuperScript II; Life Technologies, Inc.) was carried out at 37°C for 60 min in a 20 μl solution of 1× First Strand Buffer (Life Technologies, Inc.), 10 mM DTT, 0.5 mM deoxynucleoside triphosphates, and 0.4 U/ml RNasin. 3.5 μl of the resulting cDNA was amplified by PCR reaction in 50 μl for 35 cycles. Each cycle consisted of 60 s at 94°C, 60 s at 55°C, and 60 s at 72°C. 10 pmol of primers, 5′-ATGCCCAAGTCAAAGGCTCTCTCAACATCAC-3′ coding strand and 5′-TGCAACAGTCGCTGGACATCACAGG-3′ noncoding strand, was used, generating a 400-bp product.

Proteins were separated by standard gel electrophoresis and then transferred to nitrocellulose membrane (0.2 μm; Schleicher and Schuell). Immunostaining was performed using primary antibodies visualized by peroxidase-conjugated antiimmunoglobulin antibodies (Jackson ImmunoResearch Laboratories, Inc.) or by alkaline phosphatase–conjugated antiimmunoglobulin antibodies (Promega). Quantification analysis (Vilbert Lourmat) was performed by scanning the membrane comparing the signal obtained with that obtained with anti-tubulin antibody (Amersham) for tissue and scraped mucosa or with anti–β-actin antibody (Sigma Chemical Co.) for brush border preparation.

To perform histological analysis, organs from wild-type and mutant mice were dissected, fixed in formalin (10% vol/vol in PBS), ethanol dehydrated, and paraffin embedded. Sections of 5-μm width were prepared and stained with hematoxylin eosin–safranin according to standard histological procedures.

Immunofluorescence studies were performed in cryostat sections of OCT-embedded tissues or in paraffin-embedded sections for fimbrin isoform immunodetection. 5-μm sections were prepared and fixed in 3% paraformaldehyde. After washing in PBS, permeabilization was obtained with 0.2% Triton X-100. Sections were incubated for 40 min with primary antibodies and 2% BSA, and after washing for 30 min with secondary fluorescent antibody. Control sections were obtained in the absence of primary antibody. The different antibodies that have been tested are as follows: villin (rabbit polyclonals, 1–189 affinity purified, which recognizes mainly the head piece domain [Maunoury et al. 1992], and 1–149, which recognizes both the head piece and the core region, dilution 1:50 [our unpublished data]), actin (phalloidine, dilution 1:3,000; Sigma Chemical Co.), ezrin (rabbit polyclonal, dilution 1:100; Algrain et al. 1993), fimbrin (specific antibodies against the three isoforms of fimbrin, I- T- and L-fimbrin, have been produced in rabbit and then purified by affinity; our unpublished data), brush border myosin I (BBMI) (rabbit polyclonal, dilution 1:10; kindly provided by P.T. Matsudaira, Massachusetts Institute of Technology, Cambridge, MA), espin (rabbit polyclonal, affinity purified, dilution 1:200; kindly provided by J.R. Bartles, Northwestern University Medical School, Chicago, IL), E-cadherin (DECMA, rat monoclonal, dilution 1:100; Sigma Chemical Co.), sucrase isomaltase (goat polyclonal, dilution 1:50; Riby and Kretchmer 1984), dipeptidylpeptidase IV (CD26, rat monoclonal, dilution 1:4; Naquet et al. 1988), neutral aminopeptidase (CD13, rat ascite, dilution 1:300; kindly provided by P. Auberger, Contrat Jeune Formation, Nice, France), alkaline phosphatase (rabbit polyclonal, affinity purified, dilution 1:500; our unpublished data), carcinoembryonic antigen (mouse monoclonal, not diluted; kindly provided by J.-L. Teillaud, Institute Curie).

Small pieces of tissue (∼1–2 mm) were fixed for 2 h at room temperature in 2.5% glutaraldehyde and 2% paraformaldehyde in 80 mM cacodylate buffer, pH 7.2, 0.05% CaCl₂. After washing with the same buffer, the tissue was postfixed for 30 min at 4°C with 1% osmium and 1.5% potassium ferrocyanure in 80 mM cacodylate buffer, pH 7.2, and then at room temperature for 1 h with 2% uranyl acetate in 40% ethanol. The samples were dehydrated in a series of graded ethanol solutions and then embedded in Epon.

For brush border preparations, the procedure has been described previously (Burgess and Prum 1982). In brief, the brush border membranes were fixed for 30 min at room temperature in 0.1 M phosphate buffer, pH 7.0, with 3% glutaraldehyde and 0.2% tannic acid. Washing was performed for 10 min at room temperature with 0.1 M phosphate buffer, pH 7.0, containing 10% sucrose. Postfixation was done for 1 h at 4°C in 1% osmium in 0.1 M phosphate buffer, pH 6.0. The samples were then washed three times in water and processed as above. Silver sections were contrasted and then observed with CM120 Twin microscope (Philips).

Semithin sections (0.2 μm) were stained with toluidine blue and were examined with a Leitz (DMRD) microscope.

Images showing transverse sections through the microvilli and having a magnification of 22,000 were selected. Negatives were scanned on an Arcus II scanner (Agfa) to give a pixel size of 1.3 nm on the digitized image. All images were divided into boxes of 1,024 × 1,024 pixels to be processed using NIH image 1.60 (44 boxes for 8 wild-type animals, and 83 boxes for 12 villin-null mice). The Fourier transform and the power spectrum of each square area were calculated. Fourier analysis has allowed us to classify the specimens in three major categories (nonordered, slightly ordered, and well-ordered specimens) according to the characteristics of the power spectra of the transverse sections.

About 1 cm of intestine was fixed as described above for transmission electron microscopy. The intestine was then opened and fixed, luminal side up, on pieces of cork. Dehydration was then performed in a series of graded ethanol solution. The samples were dried by the critical point method using liquid CO₂ and coated with gold palladium. Observation was performed with a scanning electron microscope (Jeol JSM 840A).

Three series of experiments were conducted.

After mouse anesthesia, a jejunum loop was in situ isolated, taking care not to injure the local vasculature. Drugs were applied either from the apex or from the basolateral side of the intestine.

For the apical application of drugs, the intestinal loop was washed with PBS without Ca²⁺ and Mg²⁺. Then, ∼700 μl of a solution of ionophore A23187 (10 μM [Sigma Chemical Co.] diluted in PBS) was injected in the lumen of the intestine (∼3 cm) between two clamps, for 10 min. After washing with PBS, this loop was separated in two parts with a surgical clamp, and only the lower part was injected with ∼300 μl of PBS containing 2 × 10⁻⁴ M CaCl₂ for 30 min. This experiment was repeated in five wild-type and five villin-null mice. In separate animals, thapsigargin (1 mM; Sigma Chemical Co.) or ATP (1 mM; Sigma Chemical Co.) was injected in an isolated segment of the jejunal loop for 30 min. For each animal, a segment infused with only PBS served as control.

For the basolateral infusion of drugs, the isolated loop was carefully placed in a Petri dish containing 1 or 10 μM carbachol (Sigma Chemical Co.) in PBS plus Ca²⁺ for 20 min. This experiment was repeated in two wild-type and two villin-null mice.

At the end of the experiment, the animal was killed with a lethal dose of anesthetic. The intestinal loop treated with the drug or PBS was cut and prepared as described above for transmission electron microscopy and immunofluorescence study.

Experiments were performed in six awake animals (three wild-type and three mutants) after 24 h fasting with normal drinking. Then, a 60-min refeeding period was observed. The mice were then killed and the intestine was removed as described above for immunohistochemistry analysis.

DSS (2.5%, wt/vol, mol wt 40,000; ICN Biomedical) was administered in the drinking water of 27 wild-type mice and 26 mutant mice for 13 consecutive days (total numbers for 5 independent experiments performed in parallel, with ∼ 5 wild-type and 5 mutant mice in each series). This procedure is known to induce colic epithelial injury (Okayasu et al. 1990; Mashimo et al. 1996). Survival of the animals was surveyed during a period of 13 d. The remaining living animals were killed at day 13, and the small and large intestines were sampled as described above for histological analysis. The survival curves were analyzed by Kaplan Meier transform of probability versus days of DSS treatment. A P value <0.05 is considered as significant.

1 of 150 G418-gancyclovir–resistant ES cell clones underwent homologous recombination at the villin locus as determined by Southern blot analysis (Fig. 1 B). Hybridization with a neo probe failed to detect any additional sites of integration (data not shown). The targeted clone was injected into host C57Bl6 blastocysts, and the blastocysts were transferred to the uteri of pseudopregnant females. Germline transmission of the mutant allele was achieved with the targeted ES clone. The heterozygous mice did not display any obvious abnormalities in comparison with their wild-type littermates.

To examine whether animals homozygous for the villin mutation were viable, heterozygous animals were intercrossed and genotypes of the progeny were determined by Southern blot or by PCR analysis. Mice homozygous for the villin mutation were detected among the intercross progeny (Fig. 1 C). The genotypes of the progeny showed a good fit to Mendelian distribution (+/+: 84, 27%; +/−: 148, 48%; −/−: 74, 24%). Homozygous villin-deficient mice were indistinguishable from their heterozygous or wild-type littermates on the basis of size, activity, fertility, or aging. When interbred, villin-null females had litter sizes similar to their wild-type littermates. Animals were followed for as long as 2 yr and no obvious pathology developed.

To verify the absence of villin protein in homozygous mice, we examined the small intestine, colon, and kidney. By immunofluorescence, no staining was observed in the null mutant mice contrasting with an apical brush border staining in the wild-type animals (see Fig. 4A and Fig. B). Western blot analysis of the same tissues was performed using polyclonal antibodies which recognize either the head piece COOH-terminal domain or both this domain and the core of villin. No villin protein was detected in the mutant mice, whereas a significant amount was present in wild-type animals (Fig. 1 D). By Northern blot analysis and reverse transcription PCR, no villin mRNA was detected in the homozygous animals. No aberrant transcripts could be detected (data not shown).

Plasma values did not reveal any major differences between homozygous and wild-type animals (Na concentration: 147 mM ± 1.8, n = 17, and 145 mM ± 0.8, n = 10; glucose concentration: 14.9 mM ± 1.87, n = 7, and 14.3 mM ± 1.34, n = 5; creatinine concentration: 20.6 μM ± 3.79, n = 5, and 19.7 μM ± 3.14, in wild-type and villin-null mice, respectively, means ± SEM). This observation suggests, in a first approximation, that the transport functions of intestine and kidney were unaffected under basal conditions. This hypothesis could be indirectly correlated to the normal distribution in the intestinal epithelial cells of the major digestive enzymes. Indeed, the different enzymes—sucrase isomaltase, dipeptidylpeptidase IV, neutral aminopeptidase, and alkaline phosphatase—that have been studied by immunofluorescence have a normal localization in both wild-type and villin-null mice (data not shown).

Careful examination of histological sections and of scanning electron microscopy pictures of the different parts of intestine, duodenum, jejunum, ileum, and proximal and distal colon did not detect any differences in histological appearance between wild-type and mutant mice (data not shown).

At the cellular level, transmission electron microscopy revealed no difference on brush border microvilli structure in homozygous mice compared with wild-type (Fig. 2). Large heterogeneity in the length of the microvilli was observed in both wild-type and mutant mice. Fourier analysis was used as a quantitative method to describe the long range organization of the microvillus array. Representatives of all the defined categories, nonordered, slightly ordered, and well-ordered specimens (see Materials and Methods), indicating intra- and interindividual variations in the organization of the microvilli, were found in both wild-type and homozygote animals with a similar median. The percentage of the well-ordered crystalline lattices was similar for homozygote (34/83, 41%) and wild-type (14/44, 32%) mice (χ² test, P > 0.5). These crystalline lattices were characterized by the distance between two neighboring microvilli (a and b) and the lattice angle between these directions (γ) (see Fig. 2 C). The mean values are a = 119 ± 1.4 nm (mean ± SEM), b = 109 ± 1.8 nm, and γ = 58 ± 1.6° in wild-type animals (14 sections, n = 8), and a = 114 ± 2.1 nm, b = 107 ± 1.8 nm, and γ = 59 ± 0.92° in mutant animals (34 sections, n = 12), indicating no obvious difference (t test, P < 0.05) in microvilli organization between these animals.

To examine whether other actin-binding proteins take over the bundling activity of villin, protein extracts from purified brush border preparation and immunofluorescence staining of intestine sections were performed. Semiquantitative Western blot analysis of the three fimbrin isoforms (I, T, and L), espin, and ezrin as well as immunofluorescence staining (fimbrin isoforms, espin, ezrin, and BBMI) did not reveal any major difference between wild-type and homozygous animals. However, a slight increase in the labeling intensity of espin was systematically noticed in the mutant mice. This observation could be related to a better accessibility to espin epitopes in the absence of villin. Indeed, semiquantitative analysis of Western blot performed on brush border preparation failed to demonstrate any increase in espin concentration in villin-null mice. These results indicate that these known actin-binding proteins of the brush border microvilli are neither overexpressed nor downregulated in the absence of villin.

To examine Ca²⁺-dependent severing of actin in villin-null mice, both in vitro and in vivo approaches were undertaken. In vivo, pharmacological, physiological, and pathological conditions were explored.

In vitro, when brush borders were isolated in the absence of Ca²⁺ and purified on sucrose gradient, a marked difference was observed between wild-type and mutant mice. On average, a threefold increase in the yield of brush border recovery, calculated as the ratio of final protein content over initial mucosa weight, was obtained in the villin-null mice compared with wild-type. This observation suggests that the mutant mice brush borders were less fragile to mechanical cell disruption. Both wild-type and mutant mice presented a tightly packed array of microvilli (Fig. 3A and Fig. C) and a well-conserved ultrastructure with a normal actin filament bundle (Fig. 3E and Fig. F). The addition of Ca²⁺ in wild-type preparation (final concentration 10⁻⁵ to 10⁻³ M) resulted in a rapid alteration of brush borders (Fig. 3B and Fig. G) where the microvillar core disassembled with the concomitant vesiculation of the surrounding membrane. In contrast, the morphology of mutant mouse brush border remained unchanged in the presence (Fig. 3D and Fig. H) or absence of Ca²⁺ (Fig. 3C and Fig. F) as assessed by both optical Nomarski observation and transmission electron microscopy. These results demonstrated that, in the absence of villin, no fragmentation activity upon actin filaments occurred in response to increased Ca²⁺ concentrations.

In vivo, in pharmacological experiments, jejunum loops isolated in situ were used to test the effect of Ca²⁺. Different strategies known to increase intracellular Ca²⁺ concentration were monitored using ATP, carbachol, an acetylcholine agonist, Ca²⁺ ionophore A23187, or thapsigargin, a blocker of the endoplasmic reticulum Ca²⁺ ATPase. In the brush border from wild-type small intestine, the major effect observed with each of these conditions was a disruption of the apical F-actin phalloidin labeling, as illustrated in Fig. 4C and Fig. E, using 10 μM carbachol basolateral treatment in comparison to the intense and broad F-actin labeling observed in the untreated wild-type animal (see Fig. 5 A). Only thin phalloidin labeling remained at the apical pole, probably related to the terminal web region; the basolateral F-actin labeling was unaltered. Slight differences were noticed using the various conditions to increase intracellular Ca²⁺ concentration: carbachol basolateral treatment affected the apical labeling all along the villus, whereas the others drugs, administered in the lumen of the intestine, altered mostly the upper part of the villus, probably due to differences in accessibility of the drug. In these animals, the villin labeling was unaffected (Fig. 4 A). In contrast, in the villin-null mice, whatever conditions were used to increase Ca²⁺, the phalloidin F-actin labeling was unaffected: a large continuous F-actin ribbon (Fig. 4D and Fig. F) was observed along the villi similar to that observed in both control wild-type and villin-null mice (Fig. 5A and Fig. B). No significant alteration of the microvilli structure was observed by transmission electron microscopy on fixed tissues. This discrepancy between electron and light microscopy observations may be due to the different procedures and processing of the specimens (i.e., fixatives and freezing). Alternatively, rapidly reversible effects may have occurred, as it has been shown in primary culture of proximal tubule for the Ca²⁺-dependent effect of parathormone (Goligorsky et al. 1986). To assess this hypothesis in our in vivo conditions, extensive kinetics studies have to be performed in a large series of animals.

In vivo, in a physiological fasting/refeeding condition, most of the apical phalloidin labeling disappeared in wild-type animals, whereas in villin-null mice, it remained unaffected (Fig. 5C and Fig. D). This alteration affected the whole length of intestinal villi similarly in the three animals.

In pathological experiments where colitis was induced by DSS administration, although some variation was observed in the series of five experiments, villin-null mice appeared to be significantly more sensitive to the injuring effect of the DSS treatment. This is illustrated by the fact that 64% of mutant animals are dead at day 13 in contrast to only 30% of wild-type animals (Kaplan Meier transform, P = 0.008; Fig. 6 A). The median survival time was 10 d for the mutant mice; this parameter cannot be defined for the wild-type animals. Multiple and extended sites of epithelial injury were conspicuously found in the villin-null colons compared with wild-type animals (Fig. 6 B). These lesions included obvious large ulcerations, glandular atrophy, and inflammatory changes including neutrophil infiltrate, edema, and fibrosis (Fig. 6 B, panels b and d). In contrast, the wild-type colon lesions were less severe and limited to focal regions (Fig. 6 B, panels a and c). No alteration was observed along the small intestine.

Villin is one of the major structural actin-binding proteins associated with the actin cytoskeleton of the intestinal epithelial cell brush border microvilli. Among the actin-binding protein family, villin is the only member which has severing, capping, nucleating, and bundling in vitro activities. In vivo, the search for these properties needs to be performed at the level of the tissue, organ, or organism. In this study, we have addressed the biological role of villin by generating mutant mice lacking villin through homologous recombination in ES cells. In the villin-null mice, the ultrastructure of the intestinal brush border is not modified, suggesting normal bundling of the actin filaments. The other proteins expressed in the brush border (fimbrin, espin, BBM1, ezrin) are not overexpressed to compensate for the absence of villin. Interestingly, however, the severing properties of villin are not compensated in vivo, as demonstrated by the absence of actin fragmentation when the intracellular Ca²⁺ concentration was increased in mutant animals.

Villin contains at least three actin-binding sites, two of which are Ca²⁺ dependent and located in the core domain. The third is situated in the head piece domain and is Ca²⁺ independent. The in vitro activities of villin upon actin vary with the Ca²⁺ concentration (Fig. 7). At high Ca²⁺ concentration (>10⁻⁴ M), villin severs F-actin into short filaments. At lower concentration (10⁻⁷ to 10⁻⁶ M), villin caps the fast-growing end of actin microfilament and thereby prevents elongation. At the same Ca²⁺ concentration, villin nucleates microfilament growth when added to actin monomers. At very low Ca²⁺ concentration (<10⁻⁷ M), villin has no effect on actin polymerization but bundles actin filaments. In fibroblast cell culture, synthesis of large amounts of villin in cells which do not normally produce this protein induces both the growth of microvilli on the cell surface and the redistribution of F-actin. Villin lacking one actin-binding domain located at its COOH-terminal end did not induce growth of microvilli or stress fiber disruption (Friederich et al. 1989). Moreover, it has been demonstrated that a cluster of charged amino acid residues (KKEK) is crucial for the morphological activity of villin, indicating that this motif is part of an F-actin binding site that induces G-actin to polymerize (Friederich et al. 1992). Similar bundling properties of villin were demonstrated with a different approach using an antisense mRNA strategy in a colonic adenocarcinoma cell line. Indeed, stable expression of a cDNA encoding antisense villin RNA and thus downregulation of endogenous villin messenger induced dramatic brush border disassembly (Costa de Beauregard et al. 1995). The key role of villin in the morphogenesis of microvilli has not been verified in vivo (Pinson et al. 1998; this study). Indeed, both the length and the structural organization of the microvilli, in the null mouse intestine and colon are indistinguishable from those in wild-type animals. Even a sophisticated mathematical analysis of the microvilli organization such as Fourier analysis fails to demonstrate any significant difference. However, the heterogeneity observed in both types of animals may have prevented us from detecting significant differences. Nevertheless, because of the large number of animals studied in this paper and in one published previously (Pinson et al. 1998), we believe that had they existed, structural alterations should have been detected. The discrepancy between the in vivo studies and cell culture experiments is presumably due to the adaptation and plasticity of gene expression during organogenesis in vivo, properties that are lost or inefficient in in vitro systems.

Other actin-binding proteins are expressed in the brush border of epithelial cells (I isoform of fimbrin, BBMI, ezrin, and espin). Fimbrin is required, with villin, to bundle actin filaments in vitro (Coluccio and Bretscher 1989). Three fimbrin isoforms have been demonstrated with relatively specific cell expression: I in intestine and kidney epithelial cells, L in hemopoietic cells and many tumor cell lines, and T in various cells and tissues (Chafel et al. 1995). In villin-null mice, I fimbrin exhibited a cellular distribution and a semiquantitative expression similar to those observed in wild-type animals. No expression or labeling of the T and L isoforms was observed in the brush border. In addition, no modification was noticed for espin, another actin-binding protein localized in the intestinal epithelial cells (Bartles et al. 1996, Bartles et al. 1998) for which an actin bundling function has been demonstrated. The expression and distribution of the other actin-binding proteins present in the brush border (ezrin and BBMI) were also not modified. They do not bundle actin filaments but are involved in the link between the brush border cytoskeleton and the plasma membrane. Ezrin is known to play a structural and regulatory role in the assembly and stabilization of specialized plasma membrane domains (Bretscher et al. 1997). BBMI has been reported to connect laterally the microfilament core with the microvillus membrane (Matsudaira and Burgess 1979). The different members of the gelsolin/villin family of actin regulatory protein, if they are able to bundle actin in vivo as suggested by the presence of homologous COOH-terminal head piece sequences, would be good candidates to compensate for the absence of villin and might explain the normal organization of actin microfilament bundles in villin-null mice.

The Ca²⁺-regulated severing properties of villin have been established mainly in vitro. This study is the first report, to our knowledge, that supports a role for the villin-severing activity in vivo. In a normal situation, at the usual low intracellular Ca²⁺ concentration, the severing properties of villin should not be effective. However, a local increase of Ca²⁺ can occur in some physiological conditions, such as hormonal stimulation. In these conditions, in the cells expressing villin, severing of F-actin microfilaments would be induced, a phenomenon that can be part of the hormonal response. To test the role of villin in physiological or pathological situations, we have used different pharmacological (e.g., A23187, carbachol) or physiological (fasting/refeeding) strategies, to increase the intracellular Ca²⁺ concentration. For each of them, a net decrease of the F-actin microfilament labeling with phalloidin was observed only in wild-type animals, a result that suggests a severing effect of villin. In fasting/refeeding experiments, Ca²⁺ concentration should also increase. Indeed, in these conditions, catecholamine and digestive hormones increases are expected, and subsequently induce an increase in the intracellular Ca²⁺ concentration. Other reports also support this hypothesis. In the ileum, it has been demonstrated that F-actin fragmentation is necessary for carbachol to inhibit NaCl absorption (Khurana et al. 1997). The cholinergic agonist carbachol activates basolateral M3 muscarinic cholinergic receptors and then induces a biphasic increase in intracellular Ca²⁺ concentration involving a phosphoinositol phospholipase/phosphokinase (PLC/PKC) cascade (for review see Donowitz et al. 1998). The physiological effect of carbachol consists of an inhibition of the NaCl absorption and brush border Na⁺/H⁺ exchange (Cohen et al. 1991), presumably as a result of the internalization of the transport proteins (Donowitz, M., personal communication).

Another situation in which modifications of intracellular Ca²⁺ concentration occurs is during infection by microorganisms or during different conditions that induce cell injury (e.g., acid application, stress, wounding). During the intestinal infections by enteroaggregative or enteropathogenic Escherichia coli (EAEC, EPEC), vesiculation of the brush border has been observed (for review see Garcia and Le Bouguénec 1996). The increased Ca²⁺ concentration induced by the bacterial infection might rearrange the cytoskeleton and activate Ca²⁺-dependent kinase(s) resulting in morphological changes such as microvilli effacement and pedestal formation in the host cells. Damage of the intestinal mucosa has also been investigated by the use of DSS, a chemical agent that induces acute and chronic experimental ulcerative colitis in mice (Okayasu et al. 1990). DSS also causes changes in the intestinal microflora. This alteration, which can be compared with pathological intestinal infection, together with inappropriate macrophage function or toxic effects on colonic epithelium, is a possible mechanism by which enteral DSS induces ulcerative colitis in experimental animals. The observation that the death probability was two times higher in villin-null mice compared with wild-type suggests that villin might be involved in the cell injury processes and/or in the cell repairing processes. The exact role of villin in these processes remains to be further investigated. A working hypothesis is that the binding of villin to F-actin microfilaments plays a key role in the dynamics of the actin cytoskeleton, and therefore it might be a major factor in cell organization and motility. Interestingly, the present result may also have clinical implications. Indeed, the etiology of the human hemorrhagenic colorectitis is still unknown. Considering the large extension of the colon lesions in villin-null mice after DSS, an alteration of villin, such as a villin gene mutation, could be an interesting etiologic hypothesis for this pathology.

This study illustrates the complexity of the protein interactions at the cellular, tissue, organ, and animal levels. During mouse development, mechanisms might be used to compensate for the absence of villin and thus to organize well-structured microvilli in intestinal cells. In the adult mouse, this compensation persists and results in an apparently normal intestinal epithelium. We can postulate that fimbrin may play a role to compensate for the bundling activity of villin that leads to obtain normal morphogenesis, whereas no other actin-binding protein could efficiently provide the severing activity necessary for the dynamics of cortical actin microfilaments in enterocytes. This proposal can be evaluated in future experiments by creating villin- and fimbrin-null mice.

Moreover, cellular apoptosis that occurred at the upper part of the villi was similar in villin-null mice and in the wild-type animals (data not shown). Thus, the equilibrium between living and apoptotic cells in the intestine is maintained under physiological conditions in the mutant mice. As illustrated in this work, after inducing cell injury by the administration of DSS, the lack of epithelial cell plasticity might lead to dramatic effects. Starvation, unbalanced diet, cell injury, local infections, and carcinogenesis are stresses that might affect the dynamics of brush border F-actin microfilaments. Indeed, it is well established that actin microfilaments are essential to remodel cell shape and to drive cell motility. Hence, external stimuli leading, for instance, to epithelial plasticity and cell repair could be required in a variety of conditions in which villin-null mice might be found deficient in the course of future investigations.

We are grateful to Monique Arpin for her skillful advice in preparing fimbrin isoform antibodies, Julio Celis for his help in protein analysis, Hubert Reggio for his study of the Peyer's patches, Fabiola Terzi for performing transient renal ischemia, Daniel Meur and Dominique Morineau for the transmission electron microscopy photographic artwork, and Emmanuelle Fourme from the Biostatistics Department of Curie Hospital (Prof. Bernard Asselain) and Frédéric Fumeron (Laboratoire de Nutrition, Faculté Xavier Bichat) for the statistical analysis of the survival of the mice after DSS treatment. We also thank Manuel Buchwald, Evelyne Coudrier, Evelyne Friederich, and Thierry Galli for their helpful comments and valuable discussions. The scanning microscopy was performed at Centre Inter-universitaire de Microscopie Electronique (CIME) with the technical assistance of Michèle Grasset.

This work was supported in part by Association de Recherche contre le Cancer, Ligue Nationale contre le Cancer, Centre National de la Recherche Scientifique, Institut Pasteur, and Institut Curie. A. Lapillonne was a recipient of an Ipsen Laboratories fellowship.