Calpain regulates enterocyte brush border actin assembly and pathogenic Escherichia coli-mediated effacement.

This study identifies calpain as being instrumental for brush border (BB) microvillus assembly during differentiation and effacement during bacterial pathogenesis. Calpain activity is decreased by 25-80% in Caco 2 lines stably overexpressing calpastatin, the physiological inhibitor of calpain, and the effect is proportional to the calpastatin/calpain ratio. These lines exhibit a 2.5-fold reduction in the rate of microvillus extension. Apical microvillus assembly is reduced by up to 50%, as measured by quantitative fluorometric microscopy (QFM) of ezrin, indicating that calpain recruits ezrin to BB microvilli. Calpain inhibitors ZLLYCHN2, MDL 28170, and PD 150606 block BB assembly and ezrin recruitment to the BB. The HIV protease inhibitor ritonavir, which inhibits calpain at clinically relevant concentrations, also blocks BB assembly, whereas cathepsin and proteasome inhibitors do not. Microvillus effacement is inhibited after exposure of calpastatin-overexpressing cells to enteropathogenic Escherichia coli. These results suggest that calpain regulates BB assembly as well as pathological effacement, and indicate that it is an important regulator involved in HIV protease inhibitor toxicity and host-microbial pathogen interactions.

in intestinal microvilli, are Ca 2ϩ -sensitive, the role that Ca 2ϩ plays in the assembly and stability of microvilli is undetermined. Interest in Ca 2ϩ as a regulator of microvillus remodeling has been focused on its role in disrupting villin crosslinks of the microvillus actin filaments (reviewed in Ref. 1) and the activation of the actin filament severing activity of villin. It has been suggested that the Ca 2ϩ -dependent protease, calpain, cleaves the membrane linker protein ezrin, during cell motility-associated remodeling of cortical ␤-actincontaining structures (2)(3)(4). Because ␤-actin is the predominant actin isoform of microvilli (5) and because ezrin, which is abundant in microvilli, associates with ␤-actin in an isoform-specific and calpain-sensitive fashion (2), the question has arisen whether calpain regulates ␤-actin-ezrin interactions in microvilli, and microvillus assembly. Additionally, the finding that calpain levels exceed calpastatin levels in intestinal epithelial cells (6) suggests that calpain could play a role in intestinal differentiation.
Calpain has been implicated in cytoskeletal remodeling, including disruption of cell-matrix interactions at the rear of the cell during crawling (7) and lamellipodial and protrusion formation during spreading (4). These examples illustrate the role of calpain in remodeling dynamic actin filament structures at the periphery of the cell. Calpain has not been implicated in assembly of filamentous actin structures in the BB apical domain of enteric epithelial cells or membrane recruitment of ezrin, where it may play an important role in assembly of actin filaments (2,8).
A recently developed approach to study the role of calpain in cytoskeletal remodeling has been to create stable transfectants of established cell lines that overexpress calpastatin, calpain's physiological inhibitor (4). Calpastatin inhibits the ubiquitous calpain isoforms, m-and -calpain, so named for their respective millimolar and micromolar Ca 2ϩ ion requirements for activity in vitro. Calpastatin is specific for calpain, regulates no other protease (9 -11) and is the inhibitor of choice for implicating calpain in biological processes (12). Pharmacological inhibitors of calpain interact with the cysteine protease or the EF hand domains (13) of calpain and several are used to establish specific involvement of calpain proteases in cellular events. Fast-acting pharmacological inhibitors provide an advantage over antisense and siRNA approaches, which are limited by the 5-day half-life of calpain proteases (14). For the above reasons we chose to use calpastatin overexpression and combinations of pharmacologi-cal inhibitors to determine the role of calpain role in BB 1 assembly.
To determine the role of calpain in BB assembly, calpain activity levels, and corresponding BB microvillus elongation rates as well as ezrin and F-actin recruitment were measured for Caco 2 enterocyte cell lines stably overexpressing calpastatin. Down-regulation of calpain levels and activity were proportional to calpastatin overexpression, as was down-regulation of microvillus elongation. BB ezrin and F-actin recruitment were also blocked by calpastatin overexpression. The calpain inhibitors ZLLYCHN 2 , MDL28170, and PD 150606 also blocked BB assembly and ezrin recruitment. These results indicate that the calpain regulates assembly of the enterocyte BB, in part through regulation of ezrin recruitment. Calpastatin overexpression also conferred resistance to EPEC-initiated effacement. These findings indicate dynamic roles for calpain in epithelial morphogenesis and modulation of host-bacterial interactions during bacterial pathogenesis.
Quantitation of Calpastatin, m-Calpain, and Cytoskeletal Calpain Substrates in Caco 2 Clonal Cell Lines-Cytoplasmic and membrane/ cytoskeletal extracts were made by Dounce homogenization of cells and differential centrifugation of the lysate, in the presence of protease inhibitors, as described (4). The cytoplasmic extract contained ϳ95% of the extranuclear protein, and the membrane/cytoskeletal fraction, 5%. Immunoblot analysis of cytoplasmic extracts was performed after SDS-PAGE and transfer to 0.2 m nitrocellulose filters. Blots were blocked (5% nonfat dry milk, 50 mM Tris-HCl, 50 mM NaCl, 1 mM EDTA and 1 mM dithiothreitol) overnight and probed with primary antibodies, including monoclonal murine anti-human calpastatin PC-6 (17) (Research Diagnostics Inc., Flanders, NJ), rabbit anti-human m-calpain (D. E. Croall, University of Maine) (18), monoclonal murine anti-human ezrin-specific antibody 3C12 (Sigma), which reacted with a single 80-kDa protein in Caco 2 cell extracts, rabbit polyclonal ezrin-specific antibody 7-130 (Upstate Biotechnology, Lake Placid, NY), rabbit anti-human ␤-actin (5), and monoclonal murine anti-human villin (Chemicon International Inc., Temecula, CA). Secondary antisera were horseradish peroxidase-coupled goat anti-mouse IgG (Bio-Rad) or goat anti-rabbit IgG (Bio-Rad). Proteins were detected using the chemiluminescent peroxidase system (Amersham Biosciences) and exposed to film. The relative levels of antigens were quantified by densitometry and the Gelblot-Pro program (UV Products, San Gabriel, CA).
Calpain Activity in Clonal Calpastatin-overexpressing Caco 2-derived Cell Lines-Calpain activity in cell suspensions was determined by measuring Ca 2ϩ -ionophore-specific hydrolysis of the peptidyl 7-amino bond of the calpain substrate succinyl-leucyl-leucyl-valyl-tyrosyl-7-amino-4-methylcoumarin (suc-LLVY-AMC) (19) as previously described (4). Fluorimetry was performed with stirring at 37°C, with a PerkinElmer LS50B luminescence spectrometer. The initial rate of substrate cleavage in the absence of ionomycin was subtracted from the initial rate in the presence of ionomycin to determine the rate of Ca 2ϩ -dependent cleavage of the substrate.
Scanning Electron Microscopy-Caco 2 cells (5 ϫ 10 6 ) were plated on sterile glass coverslips in 35-mm plate wells. After 16 h, the medium was changed to remove non-adherent cells. Confluent monolayers were fixed in complete medium (CM) containing 3.7% formaldehyde, at 54 h of culture. Coverslips were dehydrated for 10 min in hexamethyldisilazane (100%), air-dried, and sputter-coated with gold/palladium. Scanning electron microscopy (SEM) was performed using an ISI-DS-130 scanning electron microscope. In time course experiments, cells were plated, and the medium was changed at 2 h to remove nonadherent cells. Slides were fixed at 2, 6, 12, and 24 h. SEM images were obtained of the apical microvilli, at ϫ18,400 magnification. Analysis of SEM images indicated that BB assembly rates could be assayed by measurement of apical non-border microvilli. Microvilli at the border areas of cells were not measured, because border microvilli were tightly clustered.
Microvillus Morphometry and Measurement of Microvillus Elongation Rates-SEM images of Caco 2 cell monolayers photographed at 18,400x magnification were scanned from Polaroid photographs and stored as Adobe Photoshop TIFF files (256 level gray scale). Each image contained a size bar generated by the scanning electron microscope. At least two separate micrographs, each containing several hundred apical microvilli, were measured for each condition. Automated morphometry of the apical microvilli was done using Metamorph 5.0r. Four filtering functions identified measurable apical microvilli. The first level of filtering involved thresholding to eliminate background surrounding the microvilli, which consisted primarily of pits and pockmarks in the apical membrane. Thresholding was done without significant loss of microvillus images, because of the high contrast of the microvilli. The second level of filtering involved gating objects to be measured on the basis of object width. Objects that were within Ϯ 1 S.D. of the mean fiber width were retained for length measurement. The third level of filtering involved rejection of objects lacking a centriod within the measured object, suggestive of overlapping microvilli. The fourth level of filtering involved removal of objects touching the edges of the image. Filtering left 150 -225 measurable microvilli per image and these were used to measure mean microvillus length. Calibration using the original size bar was used to convert all measurements to nanometers. The results obtained by automated microvillus morphometry were qualitatively comparable to curvilinear measurement of microvillus length of all measurable microvilli in the image, using NIH Image 1.61 (data not shown).
Ezrin Immunofluorescent Antibody Staining of Assembling BB-Caco 2 cells were plated on collagen type I (10 g/ml, 2 h, 37°C)-coated 8 well Lab-Tek II chamber slide system at 5 ϫ 10 5 cells per well (0.7 cm 2 ) and incubated in a tissue culture incubator under 5% CO 2 . Caco 2 cells plated under these conditions, in fresh complete medium (Dulbecco's modified Eagle's medium with 10% fetal calf serum), form polarized monolayers within 2 h, exhibiting polygonal borders and apical microvilli, as assayed by phalloidin or ezrin antibody staining. Two hours after plating, non-adherent cells were removed by gentle pipetting, and fresh complete medium was added. The 6 h time point was chosen for assay, because it exhibited the largest difference in BB assembly. At the assay end point slides were fixed in 3.7% formaldehyde in PiBS (PIPES buffer saline; 15 mM PIPES-OH, pH 7.0 and 140 mM NaCl) for 20 min at room temperature, rinsed three times with PiBS, and permeabilized in 0.5% Triton X-100 in PiBS for 20 s. After blocking for 30 min in PiBS with protease free 0.2% BSA (Sigma) (PiBS/BSA), all 8 cell patches on the slide were overlaid with a solution of rabbit polyclonal ezrin-specific peptide antibody (USB 7-130) (20, 21) (5 g/ml in PiBS/BSA) (Upstate Biotechnology) in humidified chamber at 37°C for 1 h. The ezrin antibody is a protein A-purified IgG rabbit antibody to the peptide C-EPVSYHVQESLQDEGAEPTG, amino acids 497-498 of human ezrin. The ezrin antibody used for these studies stained only apical microvilli and junctional microvilli at the junctions between cells and no other structures. Slides were washed three times with PiBS/BSA and incubated with FITC-coupled goat anti-rabbit IgG secondary antibody (Southern Biotechnology Associates, Inc., Birmingham, AL) for 1 h. To capture the maximum number of apical non-border microvilli per cell, the cell monolayers were photographed in the focal plane of the border microvilli using an epifluorescence microscope (Axioskop, Zeiss) equipped with a Spot Insight CCD camera, using the ϫ100 oil objective. Digital images were saved as TIFF files using Spot software 2.3 (Diagnostic Instruments, Inc, Sterling Heights, MI).
Quantitative Fluorometric Assay of BB Assembly-QFM assay measures BB assembly per cell by determining mean pixel intensity per cell of apical non-border microvillus staining with ezrin antibody 7-130/goat anti-rabbit-FITC. Cell patches (experimental and at least one control) in the wells of Lab-TEK II slides were stained under identical conditions for each slide (see ezrin immunofluorescent antibody staining, above). The image capture method used a Spot cooled CCD camera driven by the Spot 3.5 program. A 4-s exposure time was used for all images. The 4-s exposure time for the ezrin-stained monolayers resulted in less than 5% pixel saturation for the brightest image, based on the pixel intensity histogram (256 level grayscale). Captured images were focused in the plane of the intensely staining borders of the cells, which allowed the maximal number of apical non-border microvilli to be in focus. Previously unviewed areas of each cell patch were captured to avoid FITC bleaching. The TIFF image files were opened in NIH Image 1.63 and inverted. For each cell, the polygon tool was used to outline the area of apical non-border microvilli that was in focus. Out-of-focus microvilli over the perinuclear convexity near the center of the cell were excluded from analysis, as were the cell borders. Mean pixel intensity per cell was determined. The normalized mean pixel intensity (NMPI) per cell is reported, with a p value derived from a Student's t test for unpaired data with equal variance. The maximal reduction of NMPI per cell measured by this method was 50% (0.5-11 versus C9). This method can measure a 15% reduction in NMPI per cell (p Ͻ 0.05) with a sample size of 40 cells.
Treatment of Caco 2 Cells with Protease Inhibitors-Caco 2 cells grown to 50 -70% confluence and were then treated with vehicle (0.5% Me 2 SO), carbobenzyloxy-Leu-Leu-Tyr-diazomethyl ketone (ZLLY- The cells were replated on collagencoated Lab-Tek II 8 chamber slides in the presence of inhibitor. Microvillus assembly was measured by ezrin immunofluorescence staining by the QFM assay described above. Confocal Fluorescence Microscopy-Sterile glass coverslips were seeded with calpastatin-overexpressing Caco 2 line 2-1, which overexpresses calpastatin 2-fold or controls (C9). Cells were plated at 4-fold over confluence density. The medium was changed to remove nonadherent cells, at 16 h, and the monolayers were fixed in PBS containing 4% formaldehyde, at 54 h. The cells were permeabilized with Triton X-100 (0.1%) for 1 min, stained with Oregon-Green phalloidin (Molecular Probes, Eugene, OR), and photographed by fluorescence microscopy, as described (4). Confocal microscopy was performed with a Nikon inverted fluorescence microscope interfaced with a Noran laser illuminator, automated stage micrometer and digital CCD camera. Thirty images, at 500-nm spacing along the z-axis, were stored and reconstructed using Intervision 2D/3D software and a Silicon Graphics workstation.
EPEC-mediated BB Effacement-Caco 2 cells were plated on Permanox 4-well microscope slides. The medium was changed at 2 h and non-adherent cells were removed. The monolayers were cultured in CM in the presence of G418 for 2 weeks, with changes of medium every 3 days. The EPEC strain 2348/69 (EAFϩ/EAEϩ) was grown from fresh overnights in LB supplemented with mannose. At an OD of 0.8 -0.9, in log-phase growth, EPEC were harvested, washed in HBS, and resuspended in CM without antibiotics. Monolayers were infected for 30 min at 37°C or mock-infected with CM without bacteria. At the end of infection, monolayers were washed four times in warm phosphatebuffered saline (37°C) and fixed in warm 2.5% glutaraldehyde in 100 mM sodium cacodylate (37°C). Staining for transmission electron microscopy (TEM) was with 0.1% osmium tetroxide for 1 h at room temperature (22). TEM was performed with a Philips CM-10 electron microscope. Micrograph prints were digitized with an Agfa flat bed scanner and analyzed in Adobe Photoshop 4.0.

m-Calpain is Down-regulated in Calpastatin-overexpressing
Caco 2 Cells-To determine whether calpain plays a role in epithelial cell BB assembly, the ⌬3 isoform of calpastatin was over-expressed in Caco-2a enterocytes. 10 of 30 selected clonal lines were analyzed for calpastatin overexpression. Of the 10 lines screened by Western blotting, 5 lines demonstrated 2.0 -2.5-fold calpastatin overexpression and were further studied. Control cell lines, C9 and C12, were chosen from 10 identically derived cell lines transfected with the empty vector, pRC/CMV. Calpastatin is increased 2-fold in the cytoplasmic extracts of the lower overexpressing lines 2-1, 2-3, 2-6, and 2-7 and 2.4-fold in the higher overexpressing cell line 0.5-11. To determine whether limited (up to 2.5-fold) calpastatin overexpression observed in clonal Caco 2 lines is due to decreased clonogenicity related to increasing calpastatin overexpression, the C9, 2-3,    Calpastatin overexpression stably down-regulates the predominant calpain isoform, m-calpain (Table I), in proportion to the calpastatin/calpain ratio. The lower and higher over-expressor lines demonstrate calpastatin levels 2-fold or 2.5-fold background, respectively (Table I). Cytosolic m-calpain levels are reduced by 40% in the lower and higher calpastatin overexpressors (Table I). The lower and higher over-expressor lines, 2-3 and 0.5-11, respectively, demonstrate calpastatin/calpain ratios of 3:1 and 4:1 (Table I) and have calpain activities 73 and 22% of control, as determined by fluorometric measurement of cleavage of suc-LLVY-AMC (Fig. 1). These results indicate that a small change of the calpastatin/calpain ratio markedly affects calpain activity. This is consistent with the notion that the calpastatin/calpain ratio is tightly regulated in mammalian cells and directly modulates calpain activity (4,23).
Membrane/Cytoskeletal Ezrin and Villin Are Decreased by Calpastatin Overexpression-Membrane/cytoskeleton levels of calpastatin and the microvillus proteins villin and ezrin were measured by Western blot analysis of the calpastatin-overexpressing lines. This fraction represents about 5% of the total extranuclear protein. Membrane/cytoskeleton calpastatin was increased 1.6-fold in the highest over-expressor line, 0.5-11, and the calpastatin/calpain ratio was increased 1.4-fold, while the low over-expressor line, 2-3, exhibited a 1.2-fold increase of the calpastatin/calpain ratio (Table II). Membrane/cytoskeletal ezrin and villin were decreased up to 50% in calpastatin-overexpressing lines (Table II). These results indicate that calpastatin overexpression changes the calpastatin/calpain ratio in the cytosolic rather than in the membrane/cytoskeletal fraction and that trafficking of ezrin and villin to the membrane/ cytoskeletal fraction is impaired by inhibition of calpain. The  decrease of ezrin in the membrane/cytoskeletal fraction is consistent with impairment of BB assembly in calpastatin-overexpressing Caco 2 cell lines (Fig. 2, A and B).
Calpastatin Overexpression Decreases the Rate of Apical Microvillus Elongation During BB Assembly-Because calpain promotes assembly of ␤-actin and ezrin-rich cortical structures during cell motility (4, 24), we tested the hypothesis that calpain also regulates cortical actin remodeling involved in differentiation. Specifically, we tested whether BB assembly during enterocyte polarization is calpain regulated by measuring apical microvillus elongation in calpastatin-overexpressing cells. Calpastatin-overexpressing Caco 2 cells were plated under conditions that resulted in monolayer formation and BB assembly within 2 h (Fig. 2A). Qualitative assessment of SEM images indicated that the apical microvilli of the control cell line elongated continuously during the first 24 h of BB assembly ( Fig.  2A). The high calpastatin-overexpressing Caco 2 line, 0.5-11 displayed shorter microvilli at each time, while the microvilli of the low calpastatin over-expressor line 2-3 were of intermediate length. To determine apical microvillus length at each time by automated image analysis, a microvillus morphometry method was developed, using the program Metamorph 5.0r.
Images of the Caco 2 apical domain captured at ϫ18,400 magnification were analyzed as described (see "Experimental Procedures"). Calpastatin-overexpressing cells demonstrated a 30 -80% reduction in the length of apical microvilli during the first 24 h after plating (Fig. 2B) and remained decreased by 18 -29% even at 54 h (data not shown). The initial rate of microvillus elongation was 31 nm/h for the C9 control line, 12 nm/h for the 2-3 line and Ϫ38 nm/h for the 0.5-11 line, indicating microvillus shortening for the highest calpastatin-overexpressing line. The rates of microvillus elongation were similar for all three lines at 24 h, but there was a lag in length of 33% for the 2-3 line and 40% for the 0.5-11 line, relative to the C9 control line. All 5 clonal lines overexpressing calpastatin exhibited similar defects in microvillus assembly (data not shown). These results indicate that calpain plays a regulatory role in the initiation of apical microvillus elongation.
Calpastatin Overexpression Blocks BB Assembly and Ezrin Recruitment to the BB-To determine whether calpain regulates BB assembly by promoting ezrin recruitment to apical microvilli, as suggested by the membrane/cytoskeletal fraction (Table II), ezrin IF staining was used to develop a QFM assay for BB assembly. Ezrin IF staining has been used to visualize apical microvilli in epithelial cells (20,21,25). Ezrin IF staining of epithelial monolayers reveals a "belt-like" border structure at cell-cell junctions and a convex non-border apical domain, the highest region of which overlays the cell nucleus (20,21,25). IF staining of ezrin in the non-border apical domain correlates with BB (25)(26)(27) and was chosen to assay BB assembly. Caco 2 cell lines were plated at high density on a collagen IV for 6 h, fixed and stained for ezrin. Three images representative of at least 30 cells for each condition are shown for the C9 control line and the 2-3 and 0.5-11 calpastatin over-expressor lines (Fig. 3A). There were fewer apical microvilli in the apical domains of the calpastatin-overexpressing lines (Fig. 3A), confirmed by QFM. Both calpastatin-overexpressing lines exhibited a 50% reduction of the apical ezrin normalized mean pixel intensity per cell (NMPI) (line 2-3, p Ͻ 0.0023; line 0.5-11, p Ͻ 0.00010) suggesting that calpain regulates BB assembly and the recruitment of ezrin to the BB. These results suggest that reduced ezrin recruitment to apical microvillus structures leads to reduced ezrin in the cytoskeletal/membrane fraction.
Calpain Inhibitors Block BB Assembly and Ezrin Recruitment to the BB-To confirm that calpain regulates BB assembly and ezrin recruitment to apical microvilli, calpain inhibitors that specifically target the protease and EF-hand domains of calpain were tested for inhibition of BB assembly, by assaying incorporation of ezrin into apical microvilli. The selective calpain inhibitor, ZLLYCHN 2 , which binds irreversibly to the active site, does not inhibit the proteasome at concentrations less than 100 M (28) and has been used to demonstrate the role of calpain in lamellipodial protrusion formation (4). At concentrations selective for calpain inhibition, ZLLYCHN 2 blocks BB assembly and apical ezrin recruitment (Fig. 4, B and  C). The QFM assay demonstrates a 20% reduction of the NMPI (Fig. 4J, p Ͻ 0.0034). Another selective active site inhibitor of calpain, MDL 28,170 (29), also blocks BB assembly and apical ezrin recruitment (Fig. 4, E and F), resulting in a 25% reduction of the NMPI (Fig. 4J, p Ͻ 0.00020). The HIV protease inhibitor, ritonavir, which competitively inhibits m-calpain (K i ϭ 9 M) (30), blocks BB assembly (Fig. 4I), resulting in a 15% reduction of the NMPI (Fig. 4J, p Ͻ 0.042). Calpain activity was blocked by ritonavir under these conditions, by fluorometric assay of suc-LLVY-AMC cleavage in intact cells (data not shown). PD150606, which binds to the calcium-binding EF hand motif of calpain and inhibits its proteolytic activity (31) also blocks BB assembly (Fig. 4H), resulting in a 25% reduction of the NMPI (Fig. 4J, p Ͻ 0.00010). Inhibition of cathepsins by the lysosomotropic agent, NH 4 Cl (1 mM) had no effect on BB assembly (Fig. 4, D and J). The proteasome inhibitor, lactacystin, which binds specifically to the X(␤5) subunit of the proteasome, does not inhibit calpain (28), and has no inhibitory effect on BB assembly (Fig. 4, G and J). In summary, pharmacological calpain inhibitors that work through different mechanisms confirm that calpain regulates ezrin recruitment to the BB.
Apical Microvillus Actin Filament Assembly Is Inhibited by Calpastatin Overexpression-Because microvillus assembly is blocked by calpastatin, the ␤-actin content of the BB cytoskeleton was measured by Western blotting of membrane/cytoskeletal extracts of cells plated for 54 h (Table II). Although BB assembly is impaired in both calpastatin over-expressor lines, the BB ␤-actin content is reduced only 10%. The length of microvilli is decreased by 18 -29% in confluent monolayers plated for 54 h, suggesting that cytoskeletal/membrane ␤-actin of the BB may be in structures other than microvilli. To test this hypothesis, confocal microscopy of fluorescein-phalloidinstained monolayers was conducted. F-actin was localized in calpastatin over-expressing Caco 2 lines, to assess morphologic perturbation of the apical actin cytoskeleton induced by cal-pastatin overexpression. These images, assembled as a reconstruction looking down the z-axis, as well as a cross-sectional x-z plane, (Fig. 5) confirm that calpastatin over-expression decreases the density of apical microvilli and the apical microvillus F-actin content. These images also reveal abnormal deposits of actin filaments at the apical surface, localization of which is confirmed by an x-z plane cross-section, which provides a side view of the monolayer (Fig. 5). These findings suggest that calpain plays a role in localization and organization of ␤-actin in the microvilli of enterocytes, and that loss of calpain activity results in mislocalization of actin in abnormal filamentous structures at the enterocyte apical domain.
Calpastatin Overexpression Blocks EPEC-mediated Brush Border Effacement-To test whether calpain regulates BB effacement, which occurs during the first 30 min of EPEC infection and before intimate association between EPEC and Caco 2 cells (32), confluent monolayers of calpastatin over-expressing cells were infected with EPEC strain 2348/69 (EAFϩ/EAEϩ) in log-phase growth. Infection proceeded for 30 min and the monolayers were fixed, stained and evaluated by TEM. During this period of infection, pedestal formation does not occur (32). The control lines are sensitive to EPEC-mediated effacement (Fig.  6A), with few microvilli left 30 min after infection. The calpastatin over-expressing lines (Fig. 6, B and C) exhibit nearly intact microvilli, suggesting a role for calpain in EPEC-mediated effacement.

DISCUSSION
These studies demonstrate the Ca 2ϩ and calpain dependence of intestinal epithelial cell BB assembly and BB disassembly initiated by a common enteric pathogen, EPEC. Calpastatin overexpression blocks calpain activity up to 80%, in proportion to increase of the calpastatin/m-calpain ratio. Calpastatin overexpression inhibits the rate of apical microvillus elongation 2.5-fold, blocks ezrin recruitment to the BB and decreases apical microvillus density. The pharmacological calpain inhibitors ZLLYCHN 2 , MDL 28,170 and ritonavir, which selectively block the sulfhydryl protease domain, and the PD150606 inhibitor, which blocks the Ca 2ϩ -binding EF hand domain, sim- ilarly inhibit BB assembly and ezrin recruitment. The BB proteins ezrin and villin are decreased in the membrane/cytoskeletal fraction, while BB content of ␤-actin, the microvillus isoactin, is unchanged. The F-actin content of the apical domain is disorganized in calpastatin over-expressing cells, shifted from microvillus structures to irregular clumps. These results indicate a regulatory role for calpain in microvillus actin assembly and in the nucleation and extension of microvilli. Thus calpain is involved not only in motility-associated actin remodeling, but also actin remodeling associated with cell differentiation. We find also that calpain regulates adherens junction formation, 3 suggesting a general role for calpain in intestinal epithelial cell differentiation. The finding that EPEC-mediated BB effacement is regulated by calpain provides an alternative explanation for the Ca 2ϩ -sensitivity of microvilli (33), previously attributed to Ca 2ϩ activation of the actin severing activity of villin. These findings are consistent with the notion that the BB is not static, in agreement with studies of BB protein turnover (34).
There are at least three mechanisms by which calpain could facilitate actin filament, and thus microvillus, extension: (a) uncapping of actin filaments to foster microvillus extension, (b) disruption of ezrin and myosin I linkages between bundled microvillus actin filaments and the membrane needed for movement of the microvillus core relative to the membrane, thus creating space for addition of actin monomers at the barbed ends of actin filaments, or (c) remodeling of the terminal web or cortical membrane to facilitate microvillus actin filament nucleation. Supporting the first mechanism, ezrin is an important component of microvilli and associates specifically with the microvillus actin isotype, ␤-actin, through an indirect interaction at the actin filament barbed end (2,35). This complex of ezrin and ␤-actin is Ca 2ϩ -and calpain-sensitive, and involves the ubiquitous ␤-actin capping protein ␤cap73 (3). Because the binding of ezrin and ␤cap73 to the barbed end of actin filaments is calpain sensitive, it has been proposed that activation of calpain mediates Ca 2ϩ -regulated exposure of barbed ends (3).
Prior studies have argued against capping of actin filaments in microvilli (36), despite fact that microvilli tend to be uniform in length, arguing in favor of some form of capping complex. Isolated, permeabilized, microvilli have free barbed ends that can serve as sites of elongation of actin filaments (37). Nonetheless, ezrin is easily extractable under the membrane permeabilization conditions used in these studies, and capping proteins may be extractable as well (3). When isolated BB, not treated with detergent, is soaked in solutions of actin monomer, addition of monomer at the barbed ends indicates uncapped actin structure (37). Nonetheless, the added actin monomers have an altered filament structure, not found in control tips, suggesting that the tip structure of the microvillus is altered as a result of BB preparation. Our finding of calpain in the membrane/cytoskeletal fraction of Caco 2 enterocytes provides an alternative explanation. Calpain activation during BB isolation could disrupt ezrin-containing capping structures and promote addition of actin monomer to barbed ends.
A second mechanism by which calpain could facilitate mi- FIG. 6. Calpastatin overexpression blocks EPEC-mediated effacement of the BB. Calpastatin-overexpressing Caco 2 cell lines plated at confluence and cultured for 2 weeks on Permanox microscope slides were infected with log phase EPEC in complete medium (CM) for 30 min, without antibiotics. Infected monolayers were fixed and stained for TEM. A, control Caco 2 cell line C9 stably transfected with pRC/ crovillus actin polymerization is by disruption of linkages between the microvillus core and the membrane. This would allow the membrane to become untethered, permitting the addition of actin monomers to the barbed ends of actin filaments. The main linker proteins that connect the microvillus core to the membrane are the heavy chain of BB myosin I (BBMI), previously referred to as the p110 linker protein (38), and ezrin (39). Although BBMI is not known to be a calpain substrate, it is lost in microvilli following Ca 2ϩ -treatment (40), in contrast to other abundant core proteins, including villin (95 kDa), fimbrin (68 kDa), and actin (42 kDa). Furthermore, myosin and unconventional myosins, including myosin V, are calpain substrates (41,42). Ezrin, which binds the barbed end of microvillus actin filaments (2,3), is also found along the entire length of the microvillus (1). In addition to being a barbed end-associated molecule, ezrin oligomers bind to the sides of actin filaments (26). Since ezrin may tether the sides of the microvillus core to the membrane, we propose that calpain cleavage of side-binding ezrin may facilitate untethering of the membrane during microvillus protrusion formation. Elongation of the filaments could then result in increased ezrin binding capacity, consistent with our observation that calpain activity facilitates ezrin recruitment to the BB.
A third mechanism by which calpain could facilitate microvillus growth is by facilitating microvillus actin nucleation. Calpastatin overexpression results in a similar phenotype as cytochalasin B treatment, namely reduction in the density of microvilli (43). These results suggest that calpain and regulation of actin barbed end exposure play essential roles in nucleation of microvilli at the apical membrane or in the terminal web. The mechanism of nucleation of microvilli is unknown, but short ␤-actin oligomers, 7-10 monomers in length, present at the plasma membrane 4 could function as isoactin nucleation sites. Tethering of these oligomers to the plasma membrane through a calpain-sensitive ezrin linkage could provide a source of nuclei for elongation. Furthermore, these ␤-actin oligomers may be associated with spectrins of the terminal web, allowing an anchoring point for incipient actin filaments. Association of the spectrins TW 260/240 with microvillus actin rootlets (44) supports this notion. In this model, calpain could cleave spectrin filaments, allowing untethering of the membrane from the underlying terminal web, facilitating actin protrusion formation above the plane of the membrane.
The sensitivity of Caco 2 BB assembly to the HIV protease inhibitor, ritonavir, is of clinical importance. Diarrhea is one of the most common side effects of full dose ritonavir in adults (45) and children (46) and has been attributed to increased epithelial permeability (47) in HT29/B6. We present evidence that ritonavir inhibits differentiation-associated cytoskeletal remodeling and BB assembly in Caco 2 enterocytes, concomitant with calpain inhibition. Loss of BB in colonic epithelial cells could interfere with colonic H 2 O resorption, contributing to diarrhea. This work supports the notion that some of the toxicities of HIV protease inhibitors may not be due to host/ virus/drug interactions, but rather due to direct host/drug interactions.
BB effacement mediated by the EPEC virulence factor EspB (48) is Ca 2ϩ -dependent. Ca 2ϩ -sensitivity of microvilli has been described and attributed to Ca 2ϩ activation of the actin filament severing activity of villin (33,40,49). The rise in intracellular Ca 2ϩ known to correlate with EPEC binding led to the hypothesis that the effacement lesion may be due to Ca 2ϩ activation of villin (50,51). Baldwin hypothesized that the effacement lesion could be related to calpain proteases (50). We find evidence to support the latter hypothesis, since calpastatin overexpression blocks the EPEC effacement lesion. Villin, by itself, is unlikely to cause the dissolution of the microvillus actin core, since fimbrin would still function as the primary crosslinker (52). Furthermore, the finding that calpastatin overexpression blocks EPEC-mediated BB effacement, suggests that Ca 2ϩ -mediated activation of villin is insufficient for initiating actin filament disassembly. In addition, actin severing activity of villin induced by Ca 2ϩ cannot explain the vesiculation of the microvillus membrane induced by Ca 2ϩ (49,53), which may involve dissolution of the submembrane cytoskeleton. Calpain cleavage of ezrin and other linker proteins that connect the microvillus core to the membrane may necessary for microvillus dissolution and the finding that effacement of the enterocyte BB by enteropathogenic E. coli (EPEC) is Ca 2ϩand calpain-dependent, provides support for this hypothesis. Thus, calpain may play regulatory roles in both the physiological formation and pathological dissolution of the BB.