Myosin-1a Is Critical for Normal Brush Border Structure and Composition

Generation of the Myo1a Null Mutation

Mouse genomic DNA containing the Myo1a gene was obtained by screening a mouse 129SV genomic library. An ∼9-kb BamHI restriction fragment was cloned into a yeast shuttle vector containing the TK gene to obtain pTK Myo1a. DNA primers with 50 base pairs of Myo1a sequence at the ends were synthesized and used to amplify a neo/URA3 fragment. The PCR product was purified and cotransformed with pTK Myo1a into yeast strain H1515. Yeast clones positive for URA3 were selected. The neo/URA3 fragment replaced the first three Myo1a exons via yeast-mediated homologous recombination. Embryonic stem (ES) cell transfection and selection were performed as described previously (Wilson et al., 2000). Proper positioning of the KO cassette was confirmed by genomic PCR (Expand; Roche Diagnostics, Indianapolis, IN) by using primers that flanked the first four exons of Myo1a (Figure 1A, F1 and B1) and southern blots with probes positioned 5′ and 3′ to the targeted region. After germline transmission of the KO allele, the resulting heterozygotes were bred to generate animals homozygous for the Myo1a deletion. Although our initial phenotypic studies were performed on mice with a mixed genetic background, KO mice were eventually backcrossed to a pure 129X1/SvJ background, enabling us to use commercially available 129X1/SvJ mice as wild-type (WT) controls. All procedures involving mice were performed in accordance with Yale-IACUC protocol #10214.

Histology

For histological staining with hematoxylin and eosin (H&E), paraffin sections were prepared from the small intestine. Briefly, duodenal segments were excised from WT and KO mice, washed with room temperature (RT) phosphate-buffered saline, and fixed overnight in 4% paraformaldehyde (PFA)/phosphate-buffered saline. Tissue fragments were then embedded in paraffin, sectioned, and stained by Research Histology at Yale University School of Medicine. For TUNEL assays (Roche), unstained sections were deparaffinized in xylene, rehydrated in a graded ethanol series, and stained using protocols provided by the manufacturer. For the quantification of terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) intensities in MetaMorph (version 5.0; Universal Imaging, Downingtown, PA), confocal images of TUNEL signal from whole sections were thresholded to eliminate background and then binarized so that all TUNEL-positive pixels were equivalent to 1. The total integrated intensity from each section (proportional to the number of TUNEL-positive cells) was then measured and normalized to the highest WT value for plotting in SigmaPlot (version 7.0; Systat Software, Point Richmond, CA).

Immunocytochemistry

Immunostaining was performed as described previously (Heintzelman and Mooseker, 1990b). The following dilutions were used for primary antibodies: anti-Myo1a (1:250), anti-SI (1:2000), anti-alkaline phosphatase (AP) (1:2000), anti-Myo6 (10 μg/ml), anti-wide-spectrum cytokeratin (WSCK) (1:200), and anti-Myo1c (10 μg/ml). Secondary antibodies and phalloidin conjugated to Alexa-488 or Alexa-568 (Molecular Probes, Eugene, OR) were used at 1:500.

Cell Fractionation, BB Isolation, and Detergent-resistant Membrane (DRM) Microdomain Preparation

Cell fractionation and BB isolation were performed as described previously (Mooseker and Tilney, 1975). For the membrane shedding assay (Figure 8), phalloidin stabilized BBs in A′ (1 mM EGTA, 75 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol, 1 mM Pefabloc, and 10 mM imidazole, pH 7.2) were exposed to 1 mM MgATP for 15 min at RT. Samples were then fixed with 4% PFA and labeled with concanavalin A (ConA)-Alexa-488 (Molecular Probes) as described previously (Tyska and Mooseker, 2004). For the fractionation shown in Figure 10, whole cell lysates were spun at 1500 × g for 10 min to generate low-speed samples (low-speed pellet [LSP] and low-speed supernatant [LSS]). The LSS was then spun at 20,000 × g for 10 min to generate high-speed samples (high-speed pellet [HSP] and high-speed supernatant [HSS]), and the HSS was spun at 100,000 × g for 60 min to create ultrahigh-speed samples (ultraspeed pellet [USP] and ultraspeed supernatant). All pellet fractions were resuspended in volumes equal to the initial lysate and equal volumes of all fractions were subject to SDS-PAGE as described below. Preparation of DRMs and subsequent Optiprep density gradient analysis was performed as described previously (Tyska and Mooseker, 2004). However, for the work presented here, DRMs were derived from purified intact BBs and not MV.

Light Microscopy

All of the images presented here were acquired on one of three microscopes: 1) Zeiss Axiovert with 4×/0.1 and 63×/1.4 PlanApo objectives coupled to a laser scanning confocal unit (MRC-1024; Bio-Rad, Hercules, CA) (Figures 1D; 2E, inset; 8A, D–G, H, K–N; 9D; and 10B); 2) Nikon Diaphot 300 with 20×/0.75 and 40×/1.2 PlanApo objectives coupled to a Photometrics Cool-SNAP HQ cooled charge-coupled device camera (Figures 2E and 8, B and I); and 3) Nikon Eclipse E800 with a 20×/0.75 PlanApo objective coupled to SPOT imaging hardware and software (SPOT Image, Chantilly, VI; Figure 2C). Images were LUT stretched using MetaMorph (version 5.0) and assembled into figures by using PowerPoint (version X; Microsoft, Redmond, WA). All images comparing WT and KO preparations were acquired with identical parameters (e.g., pinhole diameter, detector gain) and postprocessed in an identical manner.

Electron Microscopy

Segments of WT and KO intestine were fixed and dehydrated for transmission electron microscopy (TEM) and scanning electron microscopy (SEM) as described previously (Heintzelman and Mooseker, 1990b). Isolated BBs were processed according to established methods (Mooseker and Tilney, 1975). TEM negatives were digitized at 1200 dpi, and morphometry (measurements of angles, lengths, and diameters as in Figure 7) was performed using MetaMorph (version 5.0). Numerical data were exported to Excel (version X) and plotted using SigmaPlot (version 7.0). For SEM, tissue segments were dried in hexamethyldisilazane, sputter coated with 100-Å gold-palladium, and then imaged with an ISI SS40 at 10 kV (Yale University) or a JEOL JSM-840 at 15 kV (Marine Biological Laboratory, Woods Hole, MA). Prints from the ISI SS40 were digitized at 600 dpi; images from the JEOL JSM-840 were digitally recorded at 150 dpi (3360 × 2600).

SDS-PAGE and Immunoblotting

Protein fractions were analyzed with SDS-PAGE by using 5–20% gradient gels. Protein concentrations were assessed with the bicinchoninic acid assay (Pierce Chemical, Rockford, IL). For immunoblotting, gels were transferred to nitrocellulose (85 V, 3.5 h) at 4°C. Immunogens were visualized using the enhanced chemiluminescence method (Amersham Biosciences, Piscataway, NJ). Primary antibodies used for immunoblots in these studies include (used at 1:1000 unless otherwise noted) the following: anti-Myo1a (Skowron et al., 1998); anti-CaM (monoclonal antibody [mAb]; Upstate Biotechnology, Lake Placid, NY); anti-Myo1e (Skowron et al., 1998), anti-Myo2a (Biomedical Technologies, Cambridge, MA); anti-Myo5 (Espindola et al., 1996); anti-Myo6 (Hasson and Mooseker, 1994); anti-Myo10 (gift from R. Cheney, University of North Carolina, Chapel Hill, NC); CX-1 (1:500) (Carboni et al., 1988); anti-villin (mAb; AMAC, Westbrook, Maine); anti-fimbrin (Heintzelman and Mooseker, 1990a); anti-spectrin (Pearl et al., 1984); anti-WSCK, anti-cytokeratin (CK)8, and anti-CK18 (mAb; Abcam, Cambridge, MA); anti-adaptin β and anti-clathrin (BD Transduction Laboratories, Lexington, KY); anti-rat SI (gift from A. Quaroni, Cornell University, Ithaca, NY); anti-AP (1:2000; Sigma-Aldrich, St. Louis, MO); anti-galectin-4 (Gal4) (R&D Systems, Minneapolis, MN); and anti-Myo1c (gift from P. Gillespie, OHSU).

Development of the Myo1a KO Mouse

Conventional ES cell/recombination protocols were used to replace the first three exons of the Myo1a gene with a cassette conferring resistance to neomycin (Figure 1A). Proper positioning of the KO cassette was confirmed with genomic PCR by using primers that flanked the targeted region (Figure 1A, F1 and B1). The resulting KO product demonstrated the expected 2311-base pair insertion and the introduction of two new EcoRI sites (Figure 1B). Ablation of the Myo1a gene product in homozygote animals was confirmed by immunoblot analysis of whole gut, mucosal homogenate, and crude BB fractions by using a polyclonal directed against the TH1 domain of Myo1a (Figure 1C). Immunostaining of KO duodenal sections with the same antibody revealed the expected lack of signal in the BB region (Figure 1D, arrowheads).

Myo1a KO Mice Exhibit No Overt Phenotype

Homozygous KO mice demonstrated no overt phenotype at the whole animal level (Figure 2A). With respect to growth rate, weight gain, rate of stool production, and consistency (Figure 2B), KO mice seemed no different from WT. Moreover, levels of physical activity, reproductive rate, and litter size also were normal. Although a recent report suggests that mutations in Myo1a may result in deafness in humans (Donaudy et al., 2003), KO animals responded well in a startle test and performed normally in balance beam trials, suggesting that hearing and vestibular function were unperturbed.

Intestinal Histology

To assess irregularities in the morphology of KO small intestine, duodenal sections were embedded in paraffin and stained with H&E. KO sections revealed no obvious perturbations in villus or crypt organization (Figure 2C). Both mucosa and submucosa seemed normal; no defects in enterocyte structure were apparent at this level of resolution. However, measurements of villus length in the KO revealed a small but statistically significant decrease relative to WT (KO, 440 ± 57 μm, n = 144 villi vs. WT, 468 ± 46 μm, n = 144 villi; p < 0.0005). Villus shortening is a hallmark of intestinal inflammation (Kuisma et al., 2001) and can be linked to higher levels of apoptosis at villus tips. Consistent with this, TUNEL assays demonstrated that KO sections possessed significantly higher numbers of apoptotic cells relative to WT (Figure 2D). High-magnification views of KO villus structure also revealed that TUNEL signal was not restricted to villus tips where continuous cell shedding normally occurs (Figure 2E). Instead, the lamina propria, which is not normally apoptotic, also demonstrated bright TUNEL staining (Figure 2E, insets). To assess duodenal morphology with higher resolution, we used SEM. Most villi in these micrographs demonstrated normal morphology; yet occasional villi with obvious thickening and fractured tips were also observed (Figure 2F). Thus, although KO mice do not exhibit significant phenotypes at the whole animal level, this evidence indicates that the KO small intestine may persist in a state of elevated stress.

Ultrastructural Analyses

Apical Membrane Defects in KO Enterocytes. Our initial ultrastructural analyses focused on assessing the topology and integrity of the apical membrane. Low-magnification SEM revealed that the surfaces of villi from KO duodenum were much “rougher” than WT (Figure 3, A and B). The perceived roughness at this magnification is due to the nearly ubiquitous presence of membrane extrusions at the tips of MV, as well as greater disorder in MV length and packing (see below). Although absent in WT, these irregularities were present on nearly every cell in a given field and observed along the full length of the villus. Apical surface defects similar to those observed in the duodenum also were found in the ileum (Figure 4, D and E), but no obvious flaws were observed in the colon (Figure 4F). The confinement of these phenotypes to the small intestine is consistent with the expression profile of Myo1a along the gut; immunoblots showed that Myo1a expression levels in colon are much lower than duodenum or ileum (Figure 4A, inset).

Closer examination of the apical surface in KO duodenum also revealed very large, pleomorphic extrusions of membrane emanating from the tips of individual MV (see asterisks in Figure 5A [for mild] and B [for severe]). We speculate that these extrusions may represent precursors to the larger, more dramatic herniations that were observed at a lower frequency but encompassed multiple MV (Figure 5C, brackets). TEM showed that these larger herniations contained cytoplasm (Figure 5, D–M) and often retained MV actin bundles associated with the plasma membrane at the periphery (Figure 5E, arrowheads).

KO MV Are Susceptible to Core Solation. TEM and SEM of Myo1A KO enterocytes also revealed extensive vesiculation, similar in appearance to isolated MV (Mooseker et al., 1983) and BBs (Burgess and Prum, 1982) treated with elevated Ca²⁺ (Figure 6). The “sausage link” appearance of MV in these BBs (Figure 6B, arrowheads) is due to the lack of an intact actin core, likely the result of Ca²⁺-induced, villin-dependent severing (Mooseker, 1985). The presence of extensive vesiculation, which was absent in WT preparations, suggests that KO BBs may have difficulties buffering the Ca²⁺ fluxes normally encountered during ad libitum feeding conditions. Compositional changes in KO BBs (discussed below) provide an explanation for this lost buffering capacity.

Structural Disarray in KO BBs. KO BBs exhibited a pronounced lack of overall organization. Low-magnification TEM on whole gut sections showed that KO enterocytes lacked the highly ordered, tightly packed arrays of MV (Figure 7, D–F) that were obviously present in WT BBs (Figure 7, A–C). Although WT BBs are packed so tightly that free space between MV is eliminated, KO BBs exhibit a striking looseness and clearly discernible gaps between individual MV. These differences were obvious in longitudinal sections (compare Figure 7, B and E), but cross-sectional views of the BB (compare Figure 7, C and F) allowed us to quantify this defect by assessing the packing angle (γ) between adjacent MV. WT MV were packed at a mean γ of 59.0 ± 8.0° (Figure 7G, WT), very close to the ideal 60° that would be expected for the hexagonal pattern created by the tight packing of perfect cylinders. However, the mean γ between KO MV was significantly higher and exhibited a much broader range (67.5 ± 16.8°), reflecting a loss of hexagonal packing (Figure 7G, KO). Cross-sectional views also allowed us to make measurements of MV and actin core cross-sectional area. These observations revealed that the distance between the actin core and apical membrane (calculated from the difference between these two areas) is significantly reduced in the KO MV (Figure 7H), presumably due to the absence of Myo1a lateral bridges (only visible in isolated BBs; see below). Although we observed no significant differences in actin core cross sectional area (which reflects the number of bundled actin filaments), an accurate assessment of this parameter was precluded by the fact that it varies as a function of MV length (our unpublished observations), which cannot be determined in cross-sectional views. However, in longitudinal sections, KO MV did seem more variable in length, even within a single BB (compare Figure 7, B and E). WT MV length was relatively constant within a single cell with a SD <5% of the mean length (Figure 7I, WT), whereas KO variability was significantly higher (∼10% of mean length; Figure 7I, KO). In addition to overall MV disarray, the KO terminal web (TW) region seemed highly disorganized with rootlets of variable length and a poorly defined connective meshwork that failed to exclude cytosolic material (compare Figure 7, B and E). Although we did attempt to quantify the apparent disarray, in general KO rootlets were much more difficult to distinguish, a fact that confounded angle and length measurements. Together, these ultrastructural observations suggest that Myo1a plays a critical role in maintaining the exquisite organization and tight packing of MV; hallmark features of the enterocyte BB.

Compositional Analyses

Lateral Bridges Are Missing in KO MV. Although we were able to isolate KO BBs with apical membranes and actin cytoskeleton intact (Figure 8, H and I), they appeared more disheveled and less refractile compared with WT BBs (Figure 8, A and B). This appearance is most likely an in vitro manifestation of the structural disarray described in our ultrastructural analyses of whole gut sections (see above). Isolated BBs also were incubated in buffer containing high levels of Mg²⁺ to enhance the visualization of lateral bridges (Mooseker and Tilney, 1975). The prominent bridges that were clearly visible in WT MV were absent in the KO (compare Figure 8, C and J). This absence resulted in apical membranes that were collapsed onto the actin core surface and extruded from MV tips. In combination with our ultrastructural analyses (see above), these results clearly indicate that Myo1a lateral bridges are required to maintain proper spacing between the MV actin core and the apical membrane.

Based on the results of in vitro motility studies (Mooseker and Cheney, 1995), it has been assumed that Myo1a bridges exert an outward (plus-end–directed) force on the MV membrane, but direct evidence for this has been lacking. We observed that the addition of 1 mM ATP to WT BBs induced both myosin-IIa (Myo2)-dependent TW contraction (Mooseker, 1985) and extrusion of apical membranes from the underlying actin cytoskeleton (Figures 8, D–G). In WT BBs, any membrane shedding or damage that occurs in the presence of ATP could result from dissociation of Myo1a tethers or active Myo1a-dependent membrane movements. However, only TW contraction was observed in KO (Figures 8, K–N), providing the first direct evidence that Myo1a can exert concerted outward force on the BB membrane.

KO BBs Are Deficient in Membrane and Cytoskeletal Components. SDS-PAGE analysis of isolated BB fractions from KO indicated that the majority of proteins (visualized with Coomassie Blue) are present at levels comparable with WT (Figure 9A). However, prominent bands were missing at ∼110 and ∼20 kDa; immunoblots confirmed that these correspond to Myo1a heavy chain and its light chain, CaM, respectively (Figure 9A). The latter finding provides a biochemical basis for the vesiculation defects that were apparent in our ultrastructural analyses, because CaM serves as the primary Ca²⁺ buffer in this region of the enterocyte.

Isolated BBs from KO mice also were subject to an extensive immunoblot screen with probes directed against a number of BB components (Figure 9 and Table 1). It should be noted that BBs from KO enterocytes exhibiting the most extreme perturbations (e.g., herniations and vesiculation) are probably lost during the initial mucosal homogenization (i.e., purification selects for the most robust BBs). Thus, differences between WT and KO BB preparations reported here may actually underrepresent differences that are present in vivo. Immunoblots revealed that many BB cytoskeletal proteins exhibited normal levels of association in KO preparations: villin and fimbrin (MV core components), myosin-5a (Myo5), Myo2, and spectrin (TW components) all exhibited WT levels (Figure 9B). In contrast, the lipid raft-associated proteins sucrase-isomaltase (SI) and Gal4 were both reduced in KO, whereas AP was unchanged (Figure 9B and Table 1). Other cytoskeletal components that were deficient in KO BB preparations include another class I myosin, myosin-1e (Myo1e), and myosin-VI (Myo6), a motor implicated in endocytosis. A slight reduction in the clathrin adaptor protein, adaptin β, was also observed and may be related to the loss of Myo6 from these fractions (Figure 9B and Table 1). Immunoblots of whole cell homogenates showed that the reductions in CaM, Myo1e, and Myo6 were not the result of large changes in cellular expression levels, but instead due to the redistribution of these components in KO enterocytes (Figure 9C).

Consistent with our immunoblot analyses, immunostaining of duodenal tissue sections for SI revealed reduced staining intensity in the BB and low, but detectable, levels of basolateral signal not seen in WT (Figure 9D). Moreover, the TW localization of Myo6 normally observed in WT was absent in the KO, confirming that the reduction observed in immunoblots was not due to a differential loss during KO BB isolation (Figure 9D). Other BB components that were screened in KO duodenal tissue sections but that demonstrated no difference in localization relative to WT include actin, villin, fimbrin, Ca²⁺ transporter-1, and myosin-10 (our unpublished data).

In contrast to the numerous deficiencies in KO BBs discussed above, immunoblots of KO BBs demonstrated a striking increase in the intermediate filament protein pair CK8 and CK18 (Figure 9B and Table 1). Immunostaining of KO duodenal tissue sections with a WSCK polyclonal (recognizes both CK8 and CK18) showed much higher levels of signal in the BB region of KO enterocytes (Figure 9D). Because immunoblots of whole cell homogenates with the same antibody revealed no change in total cellular cytokeratin (Figure 9C), these elevated levels are likely the result of redistribution within KO enterocytes.

Myosin-1c Is Ectopically Recruited to the BB in KO Enterocytes

One of the most striking revelations from our immunoblot screen was the ectopic recruitment of Myo1c into KO BBs (Figure 10A). Fractionation of KO enterocytes revealed that recruitment to the BB was fueled by a loss of Myo1c from other membrane compartments (most obvious when comparing low-speed pellet and supernatant fractions; Figure 10A). KO duodenal sections stained for Myo1c revealed the same profound redistribution. In WT enterocytes, this motor preferentially associated with basolateral membranes and exhibited very little staining in the apical domain (Figure 10B, arrowheads). In KO sections, the pronounced basolateral signal was lost and Myo1c was instead found in the BB (Figure 10B, arrowheads).

Given our previous studies establishing Myo1a as a lipid raft component, required for the proper localization of SI (Tyska and Mooseker, 2004), we sought to determine whether Myo1c also was recruited to raft fractions in KO BBs. To this end, a detergent-insoluble raft fraction was produced from equal quantities of purified WT and KO BBs (see Materials and Methods). Density gradient analysis of this fraction revealed that the lowest density (raft-containing) fraction from KO BBs was in fact enriched with Myo1c relative to the WT fraction (Figure 10E). This first fraction also contained SI, but densitometry of the blot film images revealed that levels were reduced ∼20% compared with WT (Figure 10F). The total amount of SI (the sum of SI signal across the entire gradient) also was reduced by ∼30%. Interestingly, much less flocculent material coalesced at the top of KO gradients, indicating a reduced amount of detergent-insoluble lipid (our unpublished data). Consistent with this, the total amount of detergent-insoluble protein liberated from KO BBs (as visualized with Coomassie Blue) also was significantly reduced relative to the WT preparation (Figure 10C), suggesting an overall perturbation in lipid raft stability.

These observations raise the possibility that the loss of Myo1a function in KO mice may be rescued by the ectopic recruitment of Myo1c to the BB. However, the ultrastructural and biochemical perturbations described above suggest that this rescue is only partial, not complete. Reasons for this are discussed in more detail below.