Tmod3 regulates polarized epithelial cell morphology

Tropomodulins (Tmods) are a conserved family of actin filament pointed-end capping proteins. Previous studies have identified four Tmod isoforms in vertebrates (Fischer and Fowler, 2003). The first Tmod to be identified, Tmod1, was discovered as part of the membrane skeleton of erythrocytes, where it caps the short tropomyosin-coated actin filaments in the spectrin-actin network (Fowler, 1987; Fowler, 1996). In addition to capping the pointed-end of actin filaments, Tmods also bind to tropomyosin, which enhances the ability of Tmods to cap pointed ends and stabilize actin filaments (Weber et al., 1994). Thus in erythrocytes, the role of Tmod has long been proposed to stabilize the short tropomyosin-coated actin filaments that form the junctions of the spectrin membrane skeleton (Fowler, 1996; Ursitti and Fowler, 1994), the assembly of which is required for erythroid morphogenesis (Lazarides and Moon, 1984; Woods and Lazarides, 1988). The more recently discovered Tmod3 is an ubiquitous Tmod isoform present in non-erythroid cells, where it regulates dynamic actin processes, such as lamellipodial protrusion and cell motility in endothelial cells, suggesting that Tmods have much broader biological functions than initially thought (Fischer and Fowler, 2003).

Individual polarized epithelial cells collaborate to form a sheet of tall cuboidal cells held together by cell-cell interactions at their lateral membranes. Since reorganization of the actin cytoskeleton is intrinsic to attaining a specific cell shape as well as to assemble and maintain cell-cell contacts, a great deal of research has focused on the roles actin dynamics play at cell-cell junctions that contain E-cadherin (Gates and Peifer, 2005; Weis and Nelson, 2006). However, the spectrin-based membrane skeleton also plays important roles in epithelial cell morphogenesis as well as in biogenesis of membrane domains in these and other cell types (Bennett and Baines, 2001; Kizhatil et al., 2007; Thomas, 2001). Whereas cortical actin filaments clearly appear to be important in epithelial cell morphogenesis (Lecuit and Lenne, 2007) and comprise integral linker elements in the spectrin membrane skeleton (Fowler, 1996), exactly how actin filament regulation contributes to achievement of the tall, cuboidal morphology characteristic of cells in such polarized epithelia remains unclear.

Here, we have used cultured immortal epithelial cells to further study the role of Tmod3 and pointed-end capping in the establishment of cell shape in a polarized epithelium. We show that Tmod3 localizes to lateral membranes in a number of cultured epithelial cell lines and is the sole Tmod isoform present in Caco-2 intestinal epithelial cells. Reduction of Tmod3 levels by short hairpin RNA (shRNA) results in reduced fluorescence intensity of F-actin and tropomyosin on lateral membranes, and significantly decreases cell height and increases cross-sectional area. These changes occur in the absence of altered localizations of E-cadherin or ZO-1, markers for adherens and tight junctions, suggesting a specialized role for actin pointed end dynamics in regulating the tall, cuboidal morphology of the Caco-2 epithelial cells, rather than in regulating the polarized distribution of the E-cadherin and ZO-1 apical markers. Instead, reduction of Tmod3 leads to disruption and disorganization of αII-spectrin staining on lateral membranes, suggesting that disassembly of tropomyosin-coated actin filaments in the absence of Tmod3 leads to perturbation of the spectrin membrane skeleton. We propose that Tmod3-capping of tropomyosin-actin filaments on lateral membranes plays a key role in the assembly or stability of the spectrin membrane skeleton, which has been shown previously to control the height of polarized epithelial cells in monolayers (Kizhatil et al., 2007).

Tmod3 was initially identified in endothelial cells where it caps pointed-ends of actin filaments and negatively regulates cell motility by controlling turnover of F-actin (Fischer et al., 2003). We investigated the role of Tmod3 in epithelial cells by growing several immortal epithelial cell lines to confluence, and fixing and staining cells for F-actin and Tmod3 (Fig. 1). In all epithelial cell types that we have studied to date, including porcine kidney epithelial cells (LLC-PK1 CL4, Fig. 1A-C), human bronchial epithelial cells (HBE, Fig. 1D-F) and human intestinal epithelial cells (Caco-2, Fig. 1G-I), Tmod3 is associated with lateral membranes present at intracellular junctions. There is also substantial cytoplasmic Tmod3 staining in all of these cell types, which appears somewhat punctate, possibly suggesting association with vesicular structures (Fig. 1C,F,I). This is similar to the distribution of Tmod3 in endothelial cells where the majority of Tmod3 is cytoplasmic and only 30-40% of endogenous Tmod3 is associated with the cytoskeleton (Fischer et al., 2003). By contrast, Tmod3 does not localize to stress fibers at the basal surface of epithelial cells (data not shown).

We used Caco-2 cells, which faithfully represent differentiated polarized epithelial cells when cultured in vitro (Alvarez-Hernandez et al., 1991), to investigate Tmod3 function in the establishment of epithelial cell polarity and morphology. Immunoblots confirm that our Tmod3 antibody recognizes a specific band of the appropriate size in Caco-2 cell lysates (Fig. 1J). Since this antibody crossreacts weakly with purified recombinant Tmod1, we also probed Caco-2 lysates with a monoclonal anti-Tmod1 antibody. Immunoblotting with anti-Tmod1 antibody showed no Tmod1 in Caco-2 cells (Fig. 1J). Given that Tmod2 and Tmod4 have very narrow tissue distributions in neurons and skeletal muscle, respectively (Conley et al., 2001), we concluded that Tmod3 is the only isoform known so far that is expressed in Caco-2 cells.

XZ sections from 3D reconstructions of confocal stacks reveal that Tmod3 localizes along the length of the lateral membranes in confluent monolayers (Fig. 1K-M, arrowheads), as well as to the sub-apical domain of Caco-2 cells underlying the F-actin in the microvilli (Fig. 1K-M, arrows). Careful examination of the brush-border region reveals that Tmod3 colocalizes with the lower portion of the F-actin staining at the base of the brush border, in what is presumably the terminal web region of the cells where the pointed ends of the microvillar actin bundle rootlets are located (Fig. 1K-M and data not shown). The abundant punctate cytoplasmic staining noted above is also readily apparent in these XZ sections. It is also notable that the distribution of Tmod3 at the lateral membrane is somewhat less defined than that of F-actin, and tends to be more concentrated towards the apical portion of the lateral membrane (Fig. 1L,M). To confirm the localization of Tmod3 observed by immunofluorescence, we infected Caco-2 cells with adenovirus expressing GFP-Tmod3, and fixed and stained the cells for F-actin. Similar to antibody staining, GFP-Tmod3 was localized to the lateral membranes in the subapical domain and in the cytoplasm of Caco-2 cells (data not shown).

To determine the function of Tmod3 in epithelial cells, a silencing vector coexpressing GFP and shRNA (pEGFP-H1RNAi) was designed against human Tmod3. Typical initial transfection rates were ∼75%. Transfection of Caco-2 cells with the silencing vector resulted in a significant decrease of total cellular Tmod3 levels when compared with cells transfected with the silencing vector that encoded a mismatch control sequence with three nucleotides altered from the target sequence (Fig. 2A). Total cellular GFP, actin and tubulin levels were also examined to compare relative transfection efficiency and to control for equal loading (Fig. 2A, bottom panels), and showed that reduction in Tmod3 levels was not accompanied by changes in actin or tubulin levels.

Caco-2 cells were examined by confocal microcopy to determine morphological effects of decreasing cellular Tmod3 levels by shRNA (Fig. 2B-I). After the relatively long culture times that are required for cells to become polarized and achieve their tall, cuboidal shape (2-3 weeks) (Peterson and Mooseker, 1992), we observed that many cells no longer expressed GFP and presumably had ejected the silencing vector. Nonetheless, patches of cells that did express the silencing vector were located and examined for phenotypic effects of Tmod3 knockdown. Expression of the silencing vector results in decreased fluorescence intensity of Tmod3 (Fig. 2B-I). As assessed by fluorescence intensity, F-actin levels in these cells are also diminished overall, including F-actin associated with lateral membranes and the subapical domain of the cells (Fig. 2B-I). In addition to decreased F-actin, cells transfected with the silencing vector appear to be noticeably shorter than their untransfected neighbors (evident from the XZ perspective, Fig. 2B-E), and to have increased cross-sectional area (evident from the XY perspective, Fig. 2F-I; note dim F-actin staining outlining the lateral membranes in panel G).

To quantify the effects of decreased cellular Tmod3 levels on cell morphology, the mean cell height and cross-sectional area were measured randomly throughout the sample while viewing only the channel showing F-actin; each measurement was later assigned to `transfected' or `untransfected' while viewing all channels. Expression of the silencing vector resulted in a highly significant decrease in cell height of approximately 30%, and a highly significant increase in cross-sectional area of approximately 40% (Fig. 2J,K; P<0.005). There was also a significant decrease (∼20%) in total F-actin fluorescence intensity in cells expressing the silencing vector, as determined by comparing the average fluorescence intensity in XY projections of image stacks of GFP-expressing transfected cells with untransfected cells in the same fields of view (Fig. 2L; P<0.05). There was no difference in height, cross-sectional area or total F-actin intensity between the untransfected control cells and cells transfected with the mismatch control vector (Fig. 2J-L).

To examine changes in F-actin intensity specifically on lateral membranes, we performed line scans in cells transfected with the silencing vector compared with control cells (Fig. 2M,N). Line scans were taken from XY confocal sections, and untransfected cells were compared with cells transfected with the silencing vector (Fig. 2M) or to the mismatch control vector (Fig. 2N). Line scans shown for comparison were taken from the same sample within the same field of view. The increase in GFP fluorescence indicative of transfection correlates with a decrease in fluorescence intensity and increased spacing between peaks of F-actin signal, which marks the lateral membranes in the cells transfected with the silencing vector when compared with untransfected cells in the same field of view (Fig. 2M). Quantification of the area under the F-actin peaks corresponding to the lateral membranes reveals that there is a highly significant ∼30% decrease in intensity at lateral membranes in cells transfected with the Tmod3-silencing vector (Fig. 2O; P<0.005), considerably greater than the ∼20% decrease in total F-actin intensity in these cells (Fig. 2L). By contrast, there is no change in either the fluorescence intensity or distribution of F-actin on lateral membranes in the cells transfected with the mismatch control vector when compared with untransfected cells in the same field of view (Fig. 2N,O). These data demonstrate that the F-actin at lateral cell membranes is reduced, cell height is decreased and average cell diameter is increased as a function of reduced Tmod3 levels. Since no changes in total cellular F-actin levels were observed by western blotting (Fig. 2A), it is probable that decreased phalloidin staining for F-actin reflects depolymerization of F-actin from the lateral membranes.

Because Tmods enhance tropomyosin association with F-actin (Mudry et al., 2003; Weber et al., 1994), and because tropomyosin is associated with F-actin along lateral cell membranes in epithelial cells (Temm-Grove et al., 1998), we examined the localization of tropomyosin in Caco-2 cells with reduced levels of Tmod3. Caco-2 cells were transfected with the silencing vector that coexpresses GFP, and polarized monolayers were fixed and stained for F-actin and tropomyosin. In addition to the morphological changes and decrease in F-actin intensity described above, cells transfected with the silencing vector (coexpressing GFP, shown in blue; Fig. 3D,H) have markedly reduced fluorescence intensity of tropomyosin when compared with untransfected cells within the same field of view (Fig. 3A-H). To quantify the changes in total F-actin and tropomyosin over the entire cell, the average fluorescence intensities of F-actin or tropomyosin for cells transfected with the silencing vector were determined and compared with untransfected cells in the same field of view (Fig. 4E-H). The results showed a ∼20% decrease in total F-actin fluorescence intensity, as described above, together with a ∼30% decrease in total tropomyosin staining intensity in cells expressing the silencing vector (Fig. 3I; P<0.05). Comparison of the F-actin and tropomyosin fluorescence intensities between untransfected cells and cells transfected with the mismatch control vector were quantified as an additional control, and showed no changes (Fig. 4I).

To examine the changes in F-actin and tropomyosin fluorescence intensity at lateral membranes in cells transfected with the silencing vector, line scans were taken from XY confocal sections. For this, a single line was drawn within the field of view from a region transfected with the silencing vector into a region of untransfected cells (Fig. 4J). Regions transfected with the silencing vector, evident by increased GFP fluorescence on the right side of panel (green line), had decreased peaks of F-actin at the lateral membranes (red line) and an even more dramatic decrease in tropomyosin signal (blue line) when compared to their untransfected neighbors on the left side of the panel (Fig. 4J). These data show that the decreases in total and lateral-membrane-associated F-actin that take place upon reduction of Tmod3 levels are accompanied by similar decreases in total and lateral-membrane-associated tropomyosin levels, suggesting that absence of Tmod3 leads to tropomyosin dissociation from F-actin followed by F-actin disassembly.

To discern whether the morphological changes in cells with decreased Tmod3 levels described above were accompanied by alterations in the distribution of cell polarity markers, cells were transfected with the silencing vector that coexpresses GFP, and polarized monolayers were fixed and stained for several proteins with well-established localizations in polarized epithelial cells (Fig. 4A). Relative localization of the adherens junction component E-cadherin (Fig. 4B-E) and the tight junction marker ZO-1 (Fig. 4F-I) were unchanged in cells transfected with the silencing vector (expressing GFP, shown in blue; Fig. 4E,I) when compared with untransfected cells in the same field of view. These results suggest that the changes in F-actin and tropomyosin intensity, and in cell height and cross-sectional area induced by reducing cellular levels of Tmod3 are not a result of abnormal targeting of apical polarity markers for adherens and tight junctions, E-cadherin and ZO-1, respectively.

Tmod1-capped short actin filaments are key linker elements in the spectrin-actin network in the membrane skeleton underlying the erythrocyte membrane (Fowler, 1996). To investigate whether reduction of Tmod3 levels leads to effects on the spectrin membrane skeleton in Caco-2 cells, cells transfected with the Tmod3 silencing vector were fixed and stained for αII-spectrin. In untransfected Caco-2 cells, αII-spectrin staining is localized to apical and lateral membranes (Fig. 4J-M), as expected (Thomas, 2001). By contrast, αII-spectrin exhibits a considerably broader pattern of staining in the vicinity of the lateral membranes in cells transfected with the Tmod3 silencing vector (GFP shown in blue, Fig. 4M), when compared with untransfected cells in the same field of view (Fig. 4J-M). Quantification of the distribution of the αII-spectrin near lateral membranes from XY sections (data not shown) demonstrated an approximately twofold increase in breadth of αII-spectrin staining upon reduction of Tmod3 as compared with untransfected cells, which was highly significant (P<0.005) (Fig. 4N). No changes in the distribution of αII-spectrin were observed in cells transfected with the mismatch vector. In contrast to F-actin and tropomyosin, the intensity of αII-spectrin staining at lateral membranes was not decreased upon reduction of Tmod3 (Fig. 4L, and data not shown). As a control, we also quantified the breadth of the E-cadherin staining on lateral membranes, which showed no changes in cells transfected with the silencing vector as compared with untransfected cells or cells transfected with the mismatch vector (Fig. 4O). These results suggest that reduction of Tmod3 levels, and decreases in F-actin and tropomyosin are accompanied by disorganization of αII-spectrin at the lateral membranes of Caco-2 epithelial cells. Thus, Tmod3 might play a role in assembly or organization of the spectrin-based membrane skeleton in polarized epithelial cells.

Our data strongly suggest a cooperative role for tropomyosin and Tmod3 in stabilizing a subset of F-actin on lateral membranes, thus controlling the ability of epithelial cells in polarized monolayers to acquire or maintain their full height. Previous investigations have established that tropomyosin is likely to have a specialized function at lateral membranes of epithelial cells, based on isoform-specific localization and isoform switching after cell-cell junctions become established (Temm-Grove et al., 1998). The presence of tropomyosin on actin filaments enhances capping affinity of Tmod for pointed-end actin more than 1000-fold (Weber et al., 1999; Gregorio and Fowler, 1995) and, conversely, Tmod stabilizes the association of tropomyosin at pointed ends of actin filaments (Kostyukova and Hitchcock-DeGregori, 2004; Mudry et al., 2003; Weber et al., 1994). Thus, both assembly and disassembly of Tmod3-tropomyosin-bound actin filaments are inhibited, making this population of actin filaments more stable. Using FRAP analysis of cells microinjected with fluorescently labeled actin, Yamada and colleagues observed two actin filament populations on lateral membranes of polarized epithelial cells, a highly dynamic and a more stable population (Yamada et al., 2005). We hypothesize that Tmod3-capped tropomyosin-actin filaments may comprise this latter, more stable population. In agreement with Yamada's results, our data also imply that the stable Tmod3-capped tropomyosin-actin filaments on lateral membranes do not play a direct role in controlling apical-basal cell polarity, based on normal polarized distributions of E-cadherin and ZO-1 in cells with decreased Tmod3.

In erythrocytes, the spectrin-actin network has been established to control membrane stability, cell shape and deformability (Mohandas et al., 1983). Given the existence of components of this membrane skeleton in other cell types, such as epithelial cells and neurons, it has long been proposed that a similar network is in place in these cells, presumably with similar architecture, where it functions to generate and maintain membrane domains and confer mechanical stability (Thomas, 2001; Yeaman et al., 1999). Our observation that Tmod3 stabilizes αII-spectrin organization with lateral membranes suggests that Tmod3-capped tropomyosin-actin filaments form linkers in a spectrin-actin membrane skeleton network on the lateral membranes of epithelial cells. To our knowledge, these are the first data to indicate that a specific population of tropomyosin-actin filaments participates in the organization of the spectrin membrane skeleton in epithelial cells. Taking these data in light of previous studies, such a linkage between Tmods, tropomyosin and the spectrin-actin membrane skeleton is likely to occur in many cell types, from lens fiber cells to erythrocytes (Fischer and Fowler, 2003).

How the actin filaments regulated by Tmod3 contribute to the epithelial cell morphological phenotype is less clear. Caco-2 cells with decreased Tmod3 were flatter, with significantly decreased depth of lateral interactions. Thus, the Tmod3-regulated actin filaments appear to determine the extent of cell-cell interactions at lateral membranes, rather than to stabilize adherens junctions or control epithelial cell polarity. Similarly, perturbations in spectrin or ankyrin functions lead to defects in epithelial cell morphogenesis, exhibited by markedly flatter epithelial cells without defects in apical-basal cell polarity (Kizhatil and Bennett, 2004; Kizhatil et al., 2007; McKeown et al., 1998; Praitis et al., 2005; Zarnescu and Thomas, 1999). However, our data do not allow us to distinguish whether Tmod3 is required for assembly and biogenesis of lateral membranes or to stabilize and maintain them once formed. Epithelial cells with decreased ankyrin-G or β2-spectrin have been demonstrated to have impaired lateral membrane biogenesis, resulting in a fivefold decrease in cell height while retaining apical-basal polarity (Kizhatil and Bennett, 2004; Kizhatil et al., 2007). Since this phenotype is much more pronounced than that we observe with Tmod3 depletion, we favor the interpretation that Tmod3-tropomyosin stabilized actin filaments are required for maintenance of established epithelial cell morphology, rather than assembly of lateral membranes de novo.

Alternatively, a recent study indicates that the changes in cell shape necessary for an epithelial monolayer to attain its characteristic tall, cuboidal morphology require actomyosin-dependent contraction of the circumferential belt of actin that is established during the initiation of cell-cell contacts (Zhang et al., 2005). Zhang and colleagues found a ∼35% reduction in epithelial cell height when cells were treated with the myosin 2 inhibitor blebbistatin, similar to the ∼30% reduction we observe by decreasing cellular levels of Tmod3 (this study). It is possible that Tmod3 is somehow acting to stabilize F-actin such that, when Tmod3 protein levels are reduced via shRNA, the tropomyosin-actin filaments upon which myosin pulls are unstable and disassembled. Thus, full contraction of the actin belt is not possible and, therefore, the cells do not achieve their maximum height. Indeed, effects of spectrin mutations on morphogenesis in epithelial sheets in development have been postulated to be owing to indirect effects on apical contraction (Praitis et al., 2005; Thomas, 2001). Thus, it is possible that Tmod3 contributes to acquisition or maintenance of cell height in polarized epithelial cells by stabilizing the spectrin-actin membrane skeleton on lateral membranes, which is in turn required for maximal cortical actomyosin contractility. In either case, what has become clear is that a specific population of tropomyosin-actin filaments capped by Tmod3 at the lateral cell membranes of epithelial cells is crucial to acquisition or maintenance of the tall cuboidal shapes of polarized epithelial cells. Further investigations into the kinetics and temporal relationship between actin dynamics, spectrin membrane skeleton assembly and lateral membrane biogenesis are necessary to illuminate how these processes control epithelial cell shape.

LLC-PK1 CL4 cells were a gift from J. Bartles (Northwestern University, Chicago, IL) and were cultured in MEMα supplemented with 10% FBS on 0.4 μm Corning transwell filters (Fisher). HBE cells were a gift from V. Bennett (Duke University, Durham, NC) and were cultured in DMEM supplemented with 10% FBS on glass coverslips. Caco-2 cells (ATCC) were cultured on transwell filters in MEMα supplemented with 20% FBS and penicillin-streptavidin. Cells at ∼50% confluency were transfected using Lipofectamine 2000 (Invitrogen). For microscopy, cells were trypsinized 1-3 days after transfection, replated onto 0.4 μm Corning transwell filters, and cultured for 2-3 weeks until polarity was well established (Peterson and Mooseker, 1992).

Full-length human Tmod3 (accession number AF177172) and human Tmod1 (accession number AF131836) cDNAs were inserted in-frame in the pGEX-KG vector (Amersham Biosciences). GST-Tmods were expressed in Escherichia coli and affinity-purified on a glutathione column; then, Tmods were released from the GST moiety by thrombin proteolysis and purified to homogeneity as previously described (Babcock and Fowler, 1994). Adenoviral vectors expressing a fusion protein between GFP and human Tmod3 were made as previously described (Fischer et al., 2003). A silencing vector coexpressing GFP and shRNA (pEGFP-H1RNAi) was a gift from T. Wittmann (University of California San Francisco, San Francisco, CA). Briefly, the CMV promoter and EGFP were cloned into pSuper vector, which independently drives transcription of the RNA from the H1 promoter (described in Kojima et al., 2004). For the experiments described here the Tmod3 target sequence used was 5′-TTGTGTGACCTCGCAGCAA-3′ (Fischer et al., 2003). The mismatch sequence was 5′-TTGAGTGAGCTCGCACCAA-3′.

For immunoblots of whole-cell lysates, confluent cells were rinsed with PBS, scraped directly into 2× SDS sample buffer and analyzed by SDS-PAGE 4-5 days after transfection. The following antibodies were used for immunoblot analyses: anti-tubulin (DM1A, Sigma), anti-GFP (JL-8, Clontech), anti-Tmod3 [R5168 rabbit serum (see Fischer et al., 2003)], anti-Tmod1 [mAb9, monoclonal (see Gregorio et al., 1995)], anti-mouse-HRP (Promega) and anti-protein-A–HRP (Sigma). Monoclonal antibody against actin (C4) was a gift from J. Lessard (University of Cincinnati, Cincinnati, OH).

For visualization of GFP-Tmod3, Caco-2 cells were infected with adenoviral vectors 3 weeks after plating on transwell filters (after polarization), allowed to express GFP-Tmod3 for 3 days, then fixed and stained as follows. Polarized Caco-2 and LLC-PK1 CL4 cells on transwell filters were rinsed with PBS supplemented with 0.05% Triton X-100 (PBST) and fixed for 15 minutes with PBST supplemented with 3.7% paraformaldehyde. After fixation, cells on filters were rinsed and permeabilized with PBST for 20 minutes. Autofluorescence was quenched for 10 minutes with PBST supplemented with 100 mM sodium borohydride. Cells on filters were then rinsed with PBS and blocked for at least 1 hour at room temperature in PBST supplemented with 50 mM EGTA, 4% BSA and 1% goat serum. Cells were incubated with primary antibodies and fluorescent phalloidin overnight at 4°C, washed in PBS, incubated with secondary antibodies for 2 hours at 37°C, and again washed in PBS. HBE cells on glass coverslips were fixed for 15 minutes with PBS supplemented with 1% paraformaldehyde. After fixation, cells were permeabilized for 30 minutes with PBST supplemented with 0.25% paraformaldehyde. Subsequent processing was performed as described above for Caco-2 and LLC-PK1 CL4 cells.

For fluorescence staining, Rhodamine or Alexa Fluor-568-labelled phalloidin (Molecular Probes) were used to label F-actin. Primary antibodies used for immunofluorescence staining were: rabbit polyclonal antibody serum (R5168) against recombinant human Tmod3 (Fischer et al., 2003), affinity-purified rabbit polyclonal antibody (R14) against human erythrocyte tropomyosin (Ursitti and Fowler, 1994), mouse monoclonal antibodies against E-cadherin (G-10, Santa Cruz), α fodrin (nonerythroid αII-spectrin; AA6, ICN), ZO-1 (1A12, Zymed). Secondary antibodies were Alexa Fluor-488 and Alexa Fluor-647 anti-rabbit IgG (Molecular Probes), Cy5 anti-rabbit IgG (Jackson Labs), and Alexa Fluor-647 anti-mouse IgG (Molecular Probes).

Images were acquired using Bio-Rad 1024 (63×/1.4 n.a.) or 2100 (40×/1.3 n.a. or 60×/1.4 n.a. objectives) laser-scanning confocal fluorescence microscopes. The Z-section step size was in the range of 0.3-0.5 μm, and Caco-2 monolayers were 10-25 μm thick. Images were processed using Volocity and Adobe Photoshop and image figures were constructed in Adobe Illustrator. Quantifications were performed with ImageJ and Metamorph. Microsoft Excel was used for statistical analyses. Linescans were generated using Metamorph and Microsoft Excel. For Fig. 2, cell height and cross-sectional area were measured randomly throughout the samples while viewing only the channel showing F-actin, and each measurement was later assigned to `transfected' or `untransfected' while viewing all channels. Cell height was measured by reconstructing the monolayer as a Z-stack, and drawing a line from the basal border of the monolayer (as defined by stress fibers) to the bottom of the microvilli. Cell area was measured in XY section by drawing a line around the circumference of each cell at its midsection. 70 cells each were measured for shRNA-transfected and untransfected cells in the same fields of view, and 20-30 cells were measured for mismatched control transfected and untransfected cells in the same fields of view.

Total cellular F-actin and tropomyosin fluorescence intensity were determined from an XY projection of Z stacks by encircling the entire patch of GFP-expressing transfected cells in a given field of view, calculating average fluorescence intensity of the encircled cells in the appropriate channel, and comparing it directly to the average fluorescence intensity of the surrounding untransfected cells in the same field of view. For F-actin in Fig. 2, n=7 fields of view encompassing 1457 cells (524 transfected) for shRNA; n=3 fields of view encompassing 1123 cells (256 transfected) for mismatched control. For tropomyosin and F-actin in Fig. 3, n=5 fields of view encompassing 781 cells (286 transfected) for shRNA; n=3 fields of view encompassing 264 cells (60 transfected) for mismatched control.

F-actin intensity at lateral membranes in Fig. 2O was measured by performing line scans across XY projections of Z stacks and calculating the area under the peaks using Image J. n=7 fields of view containing 141 lateral membranes each from shRNA transfected and untransfected cells, and n=3 fields of view containing 61 lateral membranes each from mismatched control and untransfected cells. The breadth of E-cadherin or αII-spectrin staining on lateral membranes in Fig. 4N,O was measured by drawing a line across arbitrary positions on lateral membranes in an XY projection, counting fluorescent pixels and converting to μm. n=2 fields of view with 40 lateral membranes each for untransfected and shRNA transfected cells, and one field of view with 20 lateral membranes each for untransfected and mismatched transfected cells.

We thank Jeannette Moyer for assistance in purifying recombinant Tmods and Tmod antibodies, and members of the Fowler lab for helpful discussions. This research was supported by NIH grants R01-HL083464 (formerly GM034225) to V.M.F., and R03-EY014972 to R.S.F. K.L.W. was supported by an NRSA fellowship from the NIH (F32-GM072448). We also acknowledge the assistance of the Imaging Module of the NEI Core Grant for Vision Research (P30-EY012598) for confocal image microscopy, processing and analysis with Metamorph and Volocity software.