An in vitro model for the analysis of intestinal brush border assembly II. Changes in expression and localization of brush border proteins during cell contact-induced brush border assembly in Caco-2_BBe cells

We have developed an in vitro cell culture model for analysis of intestinal brush border (BB) cytoskeletal assembly, as described in the preceding companion paper. We demonstrated that Caco-2_BBe (C2_BBe) cells follow an ultrastructural pathway comparable to that observed for intestinal epithelial cells during embryonic differentiation. In the present study we have examined, by immunoblotting and immunocytochemical methods, the patterns of BB cytoskeletal protein expression during this cell contact-induced differentiation, and have related these patterns to the observed ultrastructural events. The aim of this analysis was to evaluate further the utility and validity of the C2_BBe model by comparing BB protein expression and localization in the model to the molecular events previously characterized in a number of studies in vivo (reviewed by Heintzelman and Mooseker, 1992; Mamajiwalla et al., 1992).

Briefly, the BB consists of tightly packed, uniform, apical microvilli (MV) rooted in the filamentous meshwork of the subjacent terminal web (TW) domain from which organelles are excluded (reviewed by Mooseker, 1985; Bretscher, 1991). Most of the cytoskeletal components of these two domains have been identified, localized within the BB, and characterized in vitro with respect to their interactions with actin. The polarized actin cores of the MV are bundled by villin, a BB-specific protein, and the more generally distributed fimbrin (Mooseker, 1985; Bretscher, 1991). The actin bundles are linked to the overlying membrane by spirally arranged crossbridges that are comprised, at least in part, of BB myosin I (reviewed by Mooseker et al., 1991). The actin cores extend into the inter-rootlet domain of the TW and are crosslinked by a network of myosin II and non-erythroid spectrin (fodrin; reviewed by Coleman et al., 1989). A recent study demonstrated a possible role for myosin II in maintaining MV in an upright position (Temm-Grove et al., 1992). Fodrin is also a key component of the basolateral membrane skeleton. The other subdomain of the TW is the junctional complex at the lateral margin of the cell. A circumferential band of actin filaments of mixed polarity is associated with the adherens junction, along with myosin II; microfilaments are also associated with the tight junction (Madara, 1992).

In vivo, the patterns of expression and localization of actin, villin, and fimbrin during embryogenesis have been described in several species (for review and references, see Heintzelman and Mooseker, 1992; Louvard et al., 1993). During chick embryogenesis, F-actin and villin are colocalized cortically in the enterocyte at 3-4 days by immuno-fluorescence, a time at which very few MV are present. Fimbrin is present by 7-8 days as a diffuse cytoplasmic stain. However, it does not localize apically until day 10, coinciding with an increase in microvillar density. These same patterns of expression and localization have also been observed during mammalian intestinal embryogenesis, as well as differentiation of the visceral endoderm. Again, while both proteins are expressed early in development, villin localizes apically ahead of fimbrin by two to three days in intestinal epithelial cells. As in avian differentiation, apical villin localization precedes the appearance of MV in the fetal mouse gut. Finally, this sequence of expression and localization of BB proteins is also conserved in mouse F9 teratocarcinoma cells, an in vitro model of the visceral endoderm (Ezzell et al., 1992).

Expression of unconventional myosins, including the differentiation-specific BB myosin I, is more complex in mammalian than in avian intestinal differentiation. Unconventional myosins lack α-helical coiled-coil tails and/or do not form filaments; some, including BB myosin I, are able to bind to membranes (reviewed in Cheney and Mooseker, 1992). In chicken, BB myosin I is not detectable cytoplasmically until day 7-8, where it remains throughout BB assembly (for references, see Heintzelman and Mooseker, 1992). At day 18 it is found basolaterally and does not move to its final apical localization until just before hatching. In both human and rat intestinal epithelia, on the other hand, initially a 130-135 kDa immunogen is detected by immunoblot analysis with a polyclonal antiserum to avian BB myosin I (Rochette-Egly and Haffen, 1987; Rochette-Egly et al., 1988). The 110 kDa immunogen, presumably BB myosin I, appears later in development, after initial assembly of the BB is already underway in both species. Apical fluorescence is observed in the rat intestinal epithelial cells concomitant with the appearance of the 110 kDa protein on immunoblots. The 135 kDa immunogen disappears in the rat intestine at birth, but the 130 kDa immunogen is present in both adult colon and small intestine in humans (Peterson and Mooseker, 1992).

Studies of terminal web formation during embryogenesis of the intestine have been less extensive. Three TW proteins interdigitate and presumably crosslink the actin rootlets extending down from the MV: myosin II, fodrin and TW 260/240, a BB-specific spectrin in chicken (for review and references, see Mooseker, 1985). In the chicken embryo, fodrin and myosin II are apically located at day 13 of embryogenesis, when MV with short rootlets are present (for review and references, see Heintzelman and Mooseker, 1992). TW 260/240 is expressed later in development (day 15-16, Glenney and Glenney, 1983; day 19, Takemura et al., 1988) and shows an immediate apical distribution. It does not replace fodrin in the TW. At this point the MV rootlets are elongating, and the MV become increasingly uniform overall.

Fodrin expression and localization patterns have also been examined during human and rat intestinal differentiation. By immunofluorescence, fodrin is evenly distributed along the plasma membrane of unpolarized rat intestinal cells at 16 days of gestation. At this point, remodeling of the intestinal epithelium from a stratified layer of cells to a simple columnar epithelium has not yet started (Rochette-Egly and Haffen, 1987; Amerongen et al., 1989). Fodrin is restricted, however, from the basal membrane of cells in contact with the basement membrane and the apical poles of cells bordering the lumen of the intestine. At 19 days fodrin is detected at the apical domain of the cells bordering on the lumen, concurrent with BB assembly (Amerongen et al., 1989). In human fetal intestine, on the other hand, fodrin is evenly distributed about the cell periphery, including the apical domain, of all cells in the stratified epithelium at 8 weeks of gestation (Rochette-Egly et al., 1988). MV are present at this point, but the TW has not assembled (Kelley, 1973; Moxey and Trier, 1979). Fodrin expression increases apically at 12 weeks in the small intestine and at 14 weeks in the colon (Rochette-Egly et al., 1988). This appears to occur before or just as the TW assembles (Moxey and Trier, 1979), although studies specifically correlating BB ultrastructural and immunocytochemical data have not been done.

Along the crypt-villus axis in adult chicken intestine, most of the BB cytoskeletal components, villin, fimbrin, the α subunit of non-erythroid spectrin, myosin II, and tropomyosin, are already apically localized in the polarized crypt cells (for review and references, see Heintzelman and Mooseker, 1992; Mamajiwalla et al., 1992); many of these cells also have rudimentary MV starting to protrude from the apical surface. BB myosin I is the exception, exhibiting cytoplasmic and some basolateral staining. It assumes an apical localization at the crypt-villus transition zone. The fodrin α subunit has also been localized along the human colonic crypt villus axis. While a small amount is detected apically in the crypt cells, α-fodrin is primarily localized along the basolateral membrane. Apical localization increases substantially when the cells leave the crypts. In addition to localization of BB proteins, studies of populations of epithelial cells isolated from different points along the crypt-villus axis in adult avian intestine have examined both protein and mRNA levels as the cells differentiate (Fath et al., 1990). While a small increase in the mRNA levels for villin, calmodulin, and tropomyosin were detected by northern blotting, protein levels for all components examined (villin, actin, myosin II, α-spectrin, tropomyosin, myosin I, and α-actinin) did not change. Villin mRNA has also been localized along the crypt-villus axis by in situ hybridization in mouse. A villin cRNA clone revealed the greatest accumulation of villin mRNA in cells at the base of the villus.

In this report we followed the expression of a subset of BB proteins (actin, villin, fodrin, unconventional myosin immunogens, and sucrase-isomaltase) in C2_BBe cells throughout cell contact-induced BB assembly, as described ultrastructurally in the previous paper. All proteins showed a shift from a relatively even, cortical distribution to an enhanced apical localization and, in most cases, a less intense but distinct basolateral staining. Villin protein levels showed an average 10-fold increase from 0 d to 19 d, while sucrase-isomaltase was not expressed until 9 d, well after initial BB assembly. C2_BBe cells also express multiple unconventional myosin immunogens of 140 kDa, 130 kDa, and 110 kDa; they are the only human intestinal cell line that expresses substantial levels of the 110 kDa immunogen, presumed to be the BB myosin I heavy chain (Peterson and Mooseker, 1992). The 110 kDa BB myosin I is barely detectable in the depolarized cells. During BB assembly, however, the level of BB myosin I starts to increase at 9 d, coincident with sucrase-isomaltase expression and after initial BB assembly. Ribonuclease protection assays revealed a very low level of BB myosin I mRNA in the depolarized cells that increased approximately 9-fold in confluent cells. This shift in unconventional myosin expression mirrors that observed in mammalian fetal development, suggesting that this will be a powerful system for dissecting unconventional myosin function.

Caco-2 clones C2_BBe-1 and -2 were maintained as described in Peterson and Mooseker (1992). Cells for the time course studies were set up as described in the previous paper. Two filters per time point were necessary for the protein level studies; depolarized cells were also set up in T75 tissue culture flasks (Corning, NY) for the cell fractionation studies and for RNA isolation. The experiments reported here were carried out primarily with C2_BBe-1, although C2_BBe-2 was also used; no significant differences in the results were noted. Cells were used between passages 42-58.

Antibodies to human brain fodrin, Caco-2 sucrase-isomaltase, human platelet myosin II, chicken BB myosin I (CX-1), and chicken villin were those used in an earlier study and are described there (Peterson and Mooseker, 1992). A different anti-villin mon-oclonal (mAb), clone ID2C3, was obtained from AMAC, Inc., (Westbrook, ME) and used for immunofluorescence.

Immunofluorescence staining was performed as in Peterson and Mooseker (1992) with the following modifications. For optimal staining with some of the antibodies, we used a 15 min fixation in 3% formaldehyde in 80 mM K-Pipes, 5 mM EGTA, 2 mM MgCl₂, pH 6.5, rather than the pH-shift fixation. This change in fixation resulted in a greater likelihood of specimen compression. Two 10 min incubations, one each in 50 mM ammonium chloride and 100 mM Tris-Cl, pH 7.5, were substituted for the sodium borohydride quenching steps used by Bacallao et al. (1990), as the bubbles generated by that method lifted the newly formed monolayers from the filters. The cells were permeabilized with increasing concentrations of Nonidet P-40 (Particle Data Laboratories, Elmhurst, IL), depending on confluency: 0.1% was used at the zero time point on the depolarized cells, 0.2% at 1 d and 2 d after return to regular medium, and 0.5% for the remainder of the time course. Difco gelatin (Detroit, MI) was substituted for fish skin gelatin. Fluorescein-conjugated goat anti-rabbit secondary antibody was obtained from Cappel/Organon Teknika (West Chester, PA), while fluorescein-conjugated goat anti-mouse secondary was purchased from Jackson Immuno Research Laboratories (West Grove, PA). Rhodamine-phalloidin was obtained from Molecular Probes (Eugene, OR). Data were collected with a Bio-Rad MRC 500 mounted on a Zeiss Axiovert 10 using a ×63 apochromat objective (1.4 NA). Lateral views were obtained by integration of images gathered at a z axis increment of 0.15 μm using the accompanying software. En face images at selected intervals were obtained by optical sectioning. Images were photographed from the monitor with Kodak T-Max 400 and developed according to the manufacturer’s instructions. Data shown are compiled from four experiments.

Total cell extracts were collected by two different methods in an attempt to minimize proteolysis; each yielded similar results. Initially, cells were collected by trypsinization. Usually, C2_BBe cells on plastic undergo an initial 5 min incubation in Ca²⁺,Mg²⁺-free phosphate-buffered saline (PBS), followed by 5 min in trypsin-EDTA/PBS. In order to remove the cells from the filter, more time was added to the initial incubation in PBS as the cells became more confluent. Increasing the time in trypsin led to damaged and dead cells. Time in PBS did not exceed 12 min, and time in trypsin did not exceed 5 min. The cell suspension trypsinized from the filter was transferred immediately to a tube containing low-Ca²⁺ complete medium and 1.3 μM diisopropyl fluorophosphate (DFP). A total cell extract was obtained as described in Peterson and Mooseker (1992). The trichloroacetic acid-precipitated cell protein was collected by sedimentation, resuspended in 400 μl of ice cold Tris-buffered saline with protease inhibitors, and 100 μl of 10× SDS sample buffer containing bromphenol blue was added. Ammonia vapors were used to neutralize the pH of the sample, which was then heated in a boiling water bath.

Alternatively, a filter was washed in Ca²⁺,Mg²⁺-free PBS, and cell protein was collected from the filter by trituration with hot 3× SDS sample buffer supplemented with 1.3 μM DFP, 2 μg/ml aprotinin, 200 μM dithiothreitol, and 200 μM phenylmethylsulfonyl fluoride. Care was taken not to puncture the filter and to triturate thoroughly, as the DNA present caused the sample to become quite viscous. The sample was transferred to a tube and heated in a boiling water bath for several minutes.

Cells were trypsinized from a second, identical filter for each time point as described above. The cell pellet was resuspended in 1 ml of low-Ca²⁺ complete medium, and a 50 μl aliquot was reserved for a cell count. The remaining 950 μl was pelleted in a microfuge and resuspended in 950 μl of Reagent A from the Pierce BCA assay kit (Rockville, IL); the sample was frozen at −80°C until all samples were collected. The 50 μl aliquot was diluted appropriately in complete medium, and a total cell count was performed.

Total protein in the sample was determined using the BCA assay and following the manufacturer’s instructions for the micro procedure (Pierce, Rockville, IL). As a control for freezing in Reagent A, a highly concentrated stock of bovine serum albumin (BSA; Sigma, St. Louis, MO) was prepared and diluted to 2 mg/ml in Reagent A. This stock was also frozen. Standard curves using BSA treated in this fashion or the BSA standard supplied by the manufacturer did not vary significantly.

Confluent cells or depolarized cells were scraped from T75 tissue culture flasks and fractionated exactly as described in Peterson and Mooseker (1992). The data shown are from one of two experiments.

Gel samples for scanning densitometry were diluted in 2× SDS sample buffer using a Hamilton syringe. Each time point was diluted such that cell number would be equivalent across the time course. These samples in turn were diluted 1:2 and 1:5. Equal amounts of each of the three were loaded per time point. SDS-PAGE, transfer of proteins to nitrocellulose, and immunoblotting were performed as described by Carboni et al. (1988) and Peterson and Mooseker (1992). The immunoblots were photographed, air dried, and quantified by scanning densitometry on a Visage 2000 Scanner (BioImage, a division of Millipore, Bedford, MA) in reflectance mode. Data obtained were processed with the accompanying whole band analysis software. Three time courses were quantitated.

Standard protocols were followed (Ausubel et al., 1989). RNA from depolarized and from confluent cells was isolated by guanidium thiocyanate lysis, followed by purification by CsCl gradient (Ausubel et al., 1989). The BB myosin I probe was derived by PCR from reverse transcribed C2_BBe-1 total RNA (Tung et al., 1992). Degenerate primers based on conserved sequences flanking the ATP binding site of the myosin head domain were used (W. M. Bement, J. A. Wirth, T. B. Hasson, R. E. Cheney and M. S. Mooseker, unpublished results); the resultant PCR products were isolated and ligated into pBluescript (Stratagene, La Jolla, CA). Plasmid minipreps were prepared from individual transformants (Sambrook et al., 1989), and their inserts were sequenced according to the method of Sanger et al. (1977). A putative human homologue of BB myosin I was identified based on its high amino acid sequence identity with bovine and avian BB myosins I (see Results). The plasmid with the putative BB myosin I homologue was linearized with BamHI. Transcription from the T7 promoter yielded a 260 nucleotide probe that should yield a 141 nucleotide maximum protected product upon ribonuclease treatment. Several exposures of autoradiograms from each of three assays were scanned with a Bio-Rad Model 1650 scanning densitometer in transmittance mode and analyzed with the Hoefer Scientific Instruments GS360 Eletrophoresis Data Reduction System.

To establish a further basis of comparison to in vivo observations, we first undertook a series of immunofluorescence localization studies using laser scanning confocal microscopy (LSCM). Each of the proteins examined showed a similar pattern of peripheral distribution in depolarized C2_BBe cells that shifts to an intensification of the signal apically during the BB assembly time course. In lateral views of cells depolarized as described in the previous paper, F-actin and villin are each distributed relatively evenly at the periphery of the C2_BBe cells (Figs 1A and 2A). The staining underlying the membrane appeared frequently as large punctae (data not shown), consistent with the large, bleb-like protrusions seen by SEM (Figs 1B and 2A, previous paper). Both actin and villin also exhibited a concentrated signal at the base of the cells in a suction cup-like ring (data not shown). The ring-like distribution of these proteins appeared to correspond to the dense mesh of filaments seen at the base of these cells by TEM (Fig. 10C, Peterson and Mooseker, 1993). By 1 d after return to regular maintenance medium, the cells had spread and flattened and both proteins were still localized peripherally, underlying the plasma membrane in a generally even fashion (Figs 1B, 2B). A faint, diffuse cytoplasmic staining was occasionally detected in cells immunostained for villin (Fig. 2B).

The first obvious change in localization was noted at 2 d; actin and villin frequently appeared to be more concentrated at the apical domain (Figs 1B and 2B versus 1C and 2C). This transition led to the differentiated or enterocyte-like distribution at 2-3 d for actin and villin (Figs 1C,D and 2C,D). Subsequent time points revealed few other changes. Intense apical localization and fainter but distinct lateral staining was observed for actin (Fig. 1E-H), while villin was primarily concentrated apically (Fig. 2E-H). Finally, at later time points, a small percentage of the cells contained what appeared to be vacuolar apical compartments (VACs; Gilbert and Rodriguez-Boulan, 1991), manifest as small, brightly staining circles near the base of the cell (Fig. 1G). Optical sectioning showed some, but not all, of them to be entirely within a cell (data not shown), unlike a deep invagination of the apical surface (e.g. see Fig. 5D). Others may be small intercellular cysts, which we have detected occasionally by TEM (M. D. Peterson and M. S. Mooseker, unpublished observations). They stain most brightly for actin, although we have detected them with sucrase-isomaltase and unconventional myosin antibodies as well.

In contrast to villin and actin, sucrase-isomaltase was not expressed in the depolarized cells (Fig. 3A), consistent with previous studies (e.g. see Pinto et al., 1983). It was reliably detected at 9 d (Fig. 5B), although rare positive cells were spotted at 5 d (data not shown). The variable expression from cell to cell (Fig. 5C,D) is characteristic of sucrase-isomaltase in the C2_BBe cells (Peterson and Mooseker, 1992) and Caco-2 cells in general (Pinto et al., 1983; Beaulieu and Quaroni, 1991).

The distribution of the unconventional myosin immunogens was similar to those of actin and villin throughout the time course. Like actin and villin, the unconventional myosins were localized at the periphery of the depolarized cells (Fig. 4A), as well as concentrated in a basal ring (data not shown). The unconventional myosin immunogens frequently localized as large puncta on the ring (data not shown). It should be noted that CX-1 (Carboni et al., 1988), the anti-chicken BB myosin I mAb used in this study, detects multiple putative unconventional myosin immunogens of 140 kDa, 130 kDa, and 110 kDa on immunoblots of both human intestine and the C2_BBe cells (Peterson and Mooseker, 1992; we do not yet have specific antibody probes to investigate any possible differential distribution of the immunogens). The unconventional myosin immunogens remained distributed evenly about the cell periphery through 2 d (Fig. 4B,C). Unlike the other MV cytoskeletal proteins examined, the unconventional myosin immunogens did not exhibit an apical redistribution until 3d (Fig. 4D). This ‘adult’ localization did not change throughout the rest of the time course (Fig. 4E-H).

The TW protein fodrin displayed an almost identical localization pattern to that of the unconventional myosin immunogens, with an apical shift in the relative intensity of the peripheral staining occurring at the 3d time point (compare Figs 4A-D to 5A-D). The apparent apical shift in localization of fodrin was clearer in en face optical sections. At 2 d, the C2_BBe cells displayed a faint, fine, dot-like staining at the apical membrane in some of the cells (Fig. 6A). By 3 d, however, a more intense, velveteen apical staining, consistent with an end-on view of MV, was evident in a larger number of cells (Fig. 6B), confirmed by a complete series of sections along the z axis at both time points (data not shown). While villin and actin redistribute earlier than fodrin or the unconventional myosins, the en face images flanking the transition are similar for each protein (data not shown). The differentiated distribution of fodrin was achieved by ∼5 d (Fig. 5E); the pattern was similar to that of actin. No further change in localization was noted in subsequent time points (Fig. 5F-H).

For all of the proteins examined, staining patterns were indistinguishable between 19 d cells that had undergone suspension culture in the presence of cytochalasin D (Figs 1G, 2G, 3G, 4G and 5G) and those that had not (Figs 1F, 2F, 3F, 4F and 5F). The cells also continued to increase in height, although variability in monolayer heights at the same time point are readily evident (compare Figs 1F and 2F). C2 _BBe monolayers generally tend to be heterogeneous in height (Peterson and Mooseker, 1992), and this has been exacerbated by variable degrees of compression by the cover slip (see Materials and Methods).

Large, brightly staining intracellular patches were observed within a number of depolarized C2_BBe cells stained for villin (Fig. 2A), fodrin (Fig. 5A), actin (Fig. 1A), and unconventional myosin immunogens (Fig. 4A). It was not clear what this staining represented; obvious BB remnants were not observed within the depolarized cells by TEM (previous paper, Fig. 4). In order to determine whether or not this staining corresponded to a soluble cytoplasmic pool of BB proteins, depolarized and confluent cells were fractionated by differential centrifugation as described previously (Peterson and Mooseker, 1992). Under these conditions, only proteins that are strongly associated with the BB in confluent cells and other sedimentable membrane fractions will pellet during a low speed spin. The high speed pellet contains mitochondria and BB fragments, while the ultra speed supernatant comprises the soluble cytoplasmic components (Peterson and Mooseker, 1992).

Immunoblot analysis of stoichiometrically loaded fractions revealed virtually identical fractionation patterns between the two cell types (Fig. 7). In both depolarized and confluent cells, villin was retained almost entirely in the low speed pellet (Fig. 8, lane 2, V), with a small amount released in the low speed supernatant of both cell types (lane 3). The unconventional myosin immunogens also pre-dominantly partitioned to the low speed pellet (Fig. 8, MI, lane 2), although trace amounts are detected in the other fractions. Interestingly, the presumed 110 kDa BB myosin

I is present at a very low level in the depolarized cells; on some blots it was undetectable (see also below). Sucrase-isomaltase was not expressed at all in depolarized cells (S/I). Sucrase-isomaltase in the confluent cells was also retained in the low speed pellet (S/I, lane 2). Fodrin, too, remained associated with the low speed pellet (F, lane 2); a small amount was detected in other fractions, particularly visible in the confluent cells. Finally, myosin II partitioned relatively evenly between sedimentable and cytoplasmic fractions (MII, lanes 2, 3 and 6), as demonstrated previously (Peterson and Mooseker, 1992). Identical fractionation profiles for both depolarized and confluent cells were obtained in the same buffer containing 1% Triton X-100 (data not shown).

We next wished to determine if, in addition to the observed redistribution of cytoskeletal components, a change in the level of expression of any of these components occurred in the C2_BBe cells during cell contact-induced BB assembly. Immunoblot analysis of total cell extracts revealed a change in the pattern of expression of some of the BB proteins (data not shown). These changes were quantified by scanning densitometry of identical immunoblots loaded for equal cell number across the time course. The total protein per cell ranged from 0.25 ng/cell to 0.5 ng/cell with no obvious upward or downward trend throughout the time course (data not shown). Villin levels started to rise between 5 d and 9 d, increasing on average 10-fold (Fig. 8A). Fodrin, in contrast, appeared to remain unchanged throughout the time course; results shown are from one experiment (Fig. 8B). Absolute quantitation was difficult, given the extensive proteolysis that occurred despite the presence of inhibitors that controlled the degradation of other proteins (data not shown). Some processing of fodrin may possibly be occurring, as the interaction of fodrin and actin in vitro has been shown to be regulated by calmodulin and calcium-dependent protease I (Harris and Morrow, 1990). Sucrase-isomaltase was first detected at 9 d, and its level did not change after 14 d (data not shown).

Interestingly, the unconventional myosin immunogens exhibited differential expression patterns (Fig. 8C) reminiscent of those observed during differentiation of the intestinal epithelia in mammalian embryogenesis (Rochette-Egly and Haffen, 1987; Rochette-Egly et al., 1988). As assayed by immunoblot, the 140 kDa species gradually declined to half of its initial level (Fig. 8C, triangles). The 130 kDa myosin immunogen, on the other hand, exhibited variable levels of expression not consistent with any sort of change over time (Fig. 8C, squares). The 110 kDa immunogen, presumed to be BB myosin I, was present at almost negligible levels in depolarized cells and rose 6- to 10-fold by 19 d (Fig. 8C, circles; see also Fig. 7). BB myosin I expression does not begin to increase until 5 d to 9 d. This is after apical fluorescence is observed to increase (Fig. 4D) and initial BB assembly has occurred (see Figs 12B, 13A, previous paper).

Because the CX-1 mAb recognizes multiple unconventional myosin immunogens, we confirmed these intriguing results by examining BB myosin I expression at the mRNA level. We have recently cloned a short segment of BB myosin I from the C2_BBe cells. At the amino acid level, this portion of human BB myosin I is 97% identical to bovine BB myosin I and 85% identical to avian BB myosin I (Fig. 9A). Using this presumed BB myosin I-specific probe, a ribonuclease (RNase) protection assay was performed on equal amounts of total RNA isolated from depolarized cells and from confluent cells. These times were chosen based on the protein expression results described above. Paralleling the protein level results, protected BB myosin I mRNA was barely detectable in the depolarized cells, but it had increased approximately 9-fold in confluent cells (Fig. 9B). Equal amounts of total RNA from human small intestine and colon were included as a positive control. A substantial amount of a protected product of the predicted size range was detected in the small intestine sample (Fig. 9B); this is likely to be an underestimate of the actual amount present in the epithelial cells themselves, as other mucosal and submucosal cells, which do not express BB myosin I mRNA, are present in the sample. The amount of protected product in the colonic sample was roughly equal to that of the confluent C2_BBe cells (data not shown). Because degenerate primers were used to obtain the human BB myosin I PCR product, the ends of the probe are heterogeneous. Consequently, multiple bands were detected at the predicted size (110-140 nucleotides). As a comparison, RNase protection was also performed in parallel on the same samples with a probe for the α subunit of fodrin, a protein whose level does not change substantially over the time course. The amount of protected fodrin message, on average, did not vary much. Levels of protected product increased or decreased no more than 33% from depolarized to confluent cells, depending on the experiment (data not shown), whereas the levels of BB myosin I mRNA exhibited a large increase, as mentioned above.

The previous ultrastructural analysis revealed the similarity of the C2_BBe in vitro model of BB assembly to the events that occur in vivo during embryonic differentiation of the intestinal epithelium (Peterson and Mooseker, 1993). The data presented here further support that conclusion. Fetal expression and localization patterns of BB proteins were generally conserved during C2_BBe cell BB assembly. The BB proteins redistributed to the apical domain in a discrete sequence, most characteristic of embryonic intestinal differentiation. These changes are summarized in Fig. 10.

The TW component fodrin was distributed relatively evenly about the cortex of both depolarized C2 _BBe cells and in cells of the newly formed 1 d and 2 d monolayers. It did not begin to concentrate at the apical domain until 3 d (Fig. 5A-D). At this time the MV were becoming more upright and just starting to extend rootlets, although an organized TW was not yet apparent (Fig. 12A, previous paper). This pattern is similar to that observed during human fetal intestinal differentiation. Fodrin is also localized at the periphery of both small intestinal and colonic epithelial cells by 8 weeks of gestation, as detected by immunofluorescence (Rochette-Egly et al., 1988). It, too, redistributes and intensifies apically just as the TW assembles, or slightly prior to that time (Rochette-Egly et al., 1988; Moxey and Trier, 1979). These results suggest that some post-translational modification or other signal is necessary for the redistribution of fodrin to the BB. In contrast, during avian intestinal differentiation, the BB-specific isoform of spectrin, TW260/240, is not expressed until the actin rootlets begin to elongate and TW assembly starts in earnest (Takemura et al., 1988).

The localization of fodrin in depolarized C2_BBe cells differed strikingly from that observed in ‘contact-naive’ MDCK cells (Nelson and Veshnock, 1987). In addition to cortical immunofluorescence localization, fodrin is also present in intracellular aggregates. These aggregates were not part of a soluble pool of cytoskeletal and junctional components, as indicated by the fractionation studies. Contact-naive MDCK cells in low-Ca²⁺ medium, on the other hand, have a substantial soluble, cytoplasmic pool of fodrin, as demonstrated by both extraction in isotonic buffers with Triton X-100 and by immunofluorescence (Nelson and Veshnock, 1987). Upon addition of Ca²⁺, fodrin assembles into a stable, insoluble membrane cytoskeleton (Nelson and Veshnock, 1987). However, direct comparisons between the two systems may not be valid, as they differ in several notable respects. First, tissue-specific differences in other cellular functions have been described, such as protein sorting and targeting. As reviewed by Louvard et al. (1993), Caco-2 cells not only sort proteins at the trans-Golgi network like MDCK cells, but they also sort from the basolateral membrane, transcytosing apical proteins to their proper domain as in hepatocytes. Second, differences among species may also account for the dissimilar assembly pattern of fodrin. As described earlier, fodrin is absent from the apical domain in rat but is present in human during comparable stages of BB assembly (Amerongen et al., 1989; Rochette-Egly et al., 1988).

Actin and villin redistributed to the apical domain earlier than fodrin, appearing at 2 d in the C2_BBe cells (Figs 1A-C and 2A-C). This shift in immunofluorescence localization was concomitant with an increase in MV density and possibly length (Figs 3 and 11, previous paper). The overall increase in villin protein levels (Fig. 8A) has also been noted in other in vitro models (Ezzell et al., 1992; Dudouet et al., 1987), and to a lesser extent in vivo during chicken intestinal embryogenesis and mouse crypt cell differentiation (Shibayama et al., 1987; Boller et al., 1988). For example, villin expression and localization has also been examined in HT29-18 cells, a clone derived from the HT-29 cell line. This clone is undifferentiated in glucose-containing medium (HT29-18 glu), while cells adapted to medium lacking glucose (HT29-18 gal) express a differentiated phenotype, including formation of a BB and apical villin localization (Dudouet et al., 1987). Both villin mRNA and protein levels are 10-fold higher in the differentiated HT29-18 gal clone as compared to the clone maintained in glucose medium (Dudouet et al., 1987; Pringault et al., 1986). In contrast, the MV membrane hydrolase sucrase-isomaltase was not detected by immunoblot until 9 d, although rare positive cells were spotted by immunofluorescence. At this time the BB is basically assembled; the only significant change after this point is a continued increase in MV density (see Fig. 3 in preceding paper).

As discussed above for fodrin during TW assembly, villin and actin are also expressed in the C2_BBe cells before they are actually assembled into MV. This is similar to the expression patterns in both chicken and mouse intestinal differentiation (Heintzelman and Mooseker, 1992). MV cores are capable of self-assembly in vitro from actin, villin, fimbrin, and BB myosin I purified from avian BBs (Coluccio and Bretscher, 1989), yet this does not occur in either the C2_BBe cells or in vivo. Expression of these MV proteins alone then, does not appear to be sufficient to trigger at least some aspects of BB cytoskeletal assembly. Rather, these results, taken together, suggest that MV assembly is mediated and/or regulated by protein(s) or post-translational modifications yet to be identified. Lending credence to the latter notion is a study in which levels of phosphotyrosine-containing proteins were compared between crypt cells and differentiated enterocytes; two major cytoskeletally associated peptides of 36 and 17 kDa were found to have 15 times more phosphotyrosine in crypt cells than those in differentiated cells (Burgess et al., 1989).

Vacuolar apical compartments (VACs) have also been hypothesized to be a mechanism of apical domain assembly (Vega-Salas et al., 1988). Gilbert and Rodriguez-Boulan (1991) described the presence of these structures in confluent Caco-2 cells after treatment with either nocodazole or colchicine and hypothesized that they were the result of misdirected targeting. They noted that VACs were seen in only about 5% of the cells; these structures are also commonly seen in cancerous cell lines (Louvard et al., 1992). We noted the presence of VACs in a minority of cells at later time points. These structures do not appear until after initial BB assembly is completed and most likely do not play a significant role during BB assembly in C2_BBe cells. Instead, they may just be a reflection of the neoplastic origin of these cells.

Of the changes in protein expression during BB assembly described in this report, one of the most exciting is the differential expression of unconventional myosin immunogens. The immunoblot analysis revealed a similar pattern of unconventional myosin immunogen expression during differentiation of C2_BBe cells and human fetal intestine: expression of a 130-135 kDa species first, followed by expression of a 110 kDa protein after the BB has assembled. While the in vivo localization study was performed using a polyclonal antiserum to avian BB myosin I (Rochette-Egly et al., 1988), the same results were later obtained with the CX-1 mAb used in this study (Rochette-Egly and Mooseker, unpublished observations). The levels of the 130 kDa species varied in an inconsistent fashion during the C2_BBe time course. Close examination of some immunoblots showed a tight doublet at 130 kDa that was not resolved by the densitometer. Differential behavior of these immunogens may have contributed to such a result.

As described earlier, the CX-1 mAb recognizes several unconventional myosin immunogens in the C2_BBe cells, as well as human intestinal epithelial cells. Immunofluorescence localization performed with CX-1 showed an intensification of the apical signal at 3 d, at which time only the higher molecular mass species were detected by immunoblot. The level of the presumed BB myosin I began to rise between 5 d and 9 d. However, we are unable to tell if the presumed BB myosin I immediately distributes to the apical domain. We are presently generating specific probes to determine more precisely the localization of these myosins within the cell, ultimately at the immunoEM level. Previous cell fractionation of confluent monolayers does not show differential partitioning of these immunogens; they all appear to be present in the BB (Peterson and Mooseker, 1992).

The presence of these multiple unconventional myosins in the BB raises several questions. How are these molecules related to one another? Why does the cell needs three (or more) unconventional myosins in the BB? We initially assumed that all of the immunogens detected by the CX-1 mAb in the C2_BBe cells were other myosin I isoforms. Proteins similar, but not identical, to BB myosin I have been isolated from bovine adrenal medulla and cortex, and bovine brain (Barylko et al., 1992) and rat kidney (Coluccio, 1991). It is possible that the immunogens we observe in the C2_BBe cells might be myosin I isoforms generated by alternative mRNA splicing or from separate genes. Alternatively, these immunogens may be representatives of other classes of unconventional myosins. Amino acid sequence analysis of myosin head domains has revealed seven distinct classes of myosins (Espreafico et al., 1992; Cheney et al., 1993). The CX-1 mAb recognizes a conserved epitope in the myosin head near the ATP binding site (Carboni et al., 1988). CX-1 has been shown to crossreact with one member each of two other classes of myosins, skeletal muscle myosin II (Carboni et al., 1988) and chicken myosin V (Espindola et al., 1992). It is not unreasonable to expect that some of these other classes may be expressed in the C2_BBe cells and that they may also be recognized with this antibody. This is consistent with preliminary data indicating the presence of at least 8 unconventional myosins in the C2_BBe cells (Bement et al., 1992). The C2_BBe cell contact-induced differentiation will provide an invaluable model for analysis of the functions of these and other unconventional myosins, as well as for dissecting the events of BB assembly.

The authors would like to thank Peter Lewis for assistance with the confocal microscopy, Philippe Male for the scanning densitometry, Dan Skovronsky for early versions of the graphs and Tama Hasson for troubleshooting the RNase protections. We also thank Carol Cianci and Jon Morrow for the fodrin RNA probe and antibody, and Dr Andrea Quaroni for the monoclonal anti-body to sucrase-isomaltase. Finally, the support and encouragement from all of the Mooseker lab members has been invaluable. This work was supported by NIH grants DK 25387 and GM 37756 to M.S.M., M.D.P. was supported in part by an NSF predoctoral fellowship. W.M.B. is a postdoctoral fellow of the American Cancer Society.