Small Espin: A Third Actin-bundling Protein and Potential Forked Protein Ortholog in Brush Border Microvilli

Homogenates (∼4%, wt/vol) were prepared from the testis, kidney, and small intestinal mucosal scrapings of adult rats in 0.25 M sucrose, 3 mM imidazole-HCl, pH 7.4, containing protease inhibitors (1 mM PMSF, 1 μg/ ml each of antipain and leupeptin, and 10–20 TIU/liter of aprotinin) using eight up-and-down strokes of a motor-driven Teflon-glass homogenizer (3,000 rpm). Brush borders were isolated from the small intestines of adult rats in the presence of the protease inhibitors using the procedure of Keller and Mooseker (1982). In some experiments, isolated brush borders were extracted for 45 min at 4°C with solution A (75 mM KCl, 5 mM MgCl₂, 1 mM EGTA, 10 mM imidazole-HCl, pH 7.2; Keller and Mooseker, 1982) containing the protease inhibitors in the presence or absence of 1% (vol/vol) Triton X-100 or 0.6 M KI before preparing a supernate and pellet fraction by centrifugation at 100,000 g for 60 min. Samples for PAGE and Western blotting were prepared from these various fractions by adding concentrated SDS gel sample buffer containing dithiothreitol and boiling for 3 min. Espin was detected by Western blotting using affinity-purified rabbit polyclonal antibodies directed against the COOH-terminal 379–amino acid peptide of the ∼110-kD rat Sertoli cell espin isoform followed by ¹²⁵I–protein A (Bartles et al., 1996). The amount of espin present in isolated small intestinal brush borders was determined by scanning laser densitometric analysis of Western blot autoradiograms and comparison to a standard curve generated from internal standards containing known amounts of recombinant small espin (see below). The amount of actin present in the isolated small intestinal brush borders was determined by scanning laser densitometric analysis of Coomassie blue– stained SDS gels and comparison to a standard curve generated from internal standards containing known amounts of rabbit skeletal muscle actin (Cytoskeleton, Inc., Denver, CO).

A cDNA (1,088-bp) that included the full coding sequence of the small isoform of espin was obtained using the affinity-purified rabbit polyclonal antibodies to screen a 5′-STRETCH rat kidney λgt11 cDNA library (CLONTECH Laboratories, Palo Alto, CA). cDNAs were introduced into the pBluescript SK phagemid vector (Stratagene, La Jolla, CA), and the coding and noncoding strands were sequenced in their entirety using [³⁵S]dATPαS and the Sequenase Version 2.0 kit (Amersham Corp., Arlington Heights, IL). Sequence analysis and database searching were carried out using the programs available through the University of Wisconsin Genetics Computer Group (GCG) Sequence Analysis Software Package (Devereux et al., 1984) and the BLAST Search (Altschul et al., 1990) available through the National Center for Biotechnology Information. Peptides with a high probability to form a coiled coil were identified using the programs of Lupas (1997) and Berger et al. (1995).

For immunoperoxidase labeling, rat intestine and kidney were fixed by perfusion through the abdominal aorta with Bouin's fluid and embedded in paraffin (Bartles et al., 1996). 5-μm sections were deparaffinized with xylenes, labeled with affinity-purified rabbit polyclonal espin antibody or preimmune IgG followed by horseradish peroxidase–conjugated donkey anti–rabbit F(ab′)₂, reacted with H₂O₂ and 3,3′-diaminobenzidine, and counterstained with hematoxylin (Bartles et al., 1996). In preparation for immunofluorescence, isolated small intestinal brush borders were fixed to poly-l-lysine–coated slides or coverslips with 2% paraformaldehyde in solution A containing 20 mM Hepes, pH 7.2, and 20 mM n-octyl-β-d-glucopyranoside, quenched with 0.25% (wt/vol) NH₄Cl and 0.5% (wt/vol) BSA and labeled with affinity-purified polyclonal espin antibody or preimmune IgG and fluorescein-labeled phalloidin (Molecular Probes, Eugene, OR) followed by rhodamine-labeled goat anti–rabbit IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) in PBS containing 0.02% (wt/vol) saponin (Bartles et al., 1996). In preparation for postembedment immunogold labeling, freshly isolated rat small intestinal brush borders were washed twice by low-speed centrifugation in 60 mM Pipes, 25 mM Hepes, 10 mM EGTA, 2 mM MgCl₂, pH 6.9, containing the protease inhibitors, fixed for 30 min at 4°C in the same buffer containing 4% paraformaldehyde, pelleted by centrifugation at 13,000 g for 20 min, post-fixed as a pellet for an additional 1.5 h at 4°C, washed overnight with two changes of the same buffer minus paraformaldehyde, dehydrated through 30, 50, and 70% ethanol at −20°C, infiltrated overnight with LR white (Electron Microscopy Sciences, Ft. Washington, PA) at 4°C with eight changes, placed in gelatin capsules, and then polymerized for 3–5 d at 38– 40°C. Ultrathin sections were quenched with 0.5% (wt/vol) BSA in Tris-buffered saline, pH 7.4, and labeled for 15 h with affinity-purified espin antibody or preimmune IgG followed by goat anti–rabbit IgG 12- or 6-nm-diameter colloidal gold conjugate (Jackson ImmunoResearch Laboratories) and counterstained with 2–3% (wt/vol) uranyl acetate. To minimize aggregation of the gold conjugate, the tracer was blended thoroughly on a vortex mixer and then precleared by centrifugation at 13,000 g for 5 min immediately before use. For preembedment immunogold labeling, isolated rat small intestinal brush borders were fixed in 4% paraformaldehyde in solution A containing 20 mM Hepes, pH 7.2, and 0.5% (vol/vol) Triton X-100, quenched with 0.25% (wt/vol) NH₄Cl and 0.5% (wt/vol) BSA, and labeled in suspension with affinity-purified espin antibody or preimmune IgG followed by colloidal gold–labeled secondary antibody (see above). In between steps, the brush borders were washed by low-speed centrifugation. After the final wash, the labeled brush borders were postfixed in 2% glutaraldehyde in 0.1 M sodium cacodylate, pH 7.4, postfixed in 1% OsO₄, stained en bloc with uranyl acetate, dehydrated in ethanol and propylene oxide, and embedded in Poly/Bed 812 (Polysciences, Warrington, PA). Ultrathin sections were stained with uranyl acetate and lead citrate.

For transient transfection, the 1,088-bp cDNA encoding small espin was introduced into the EcoRI site of the pEGFP-C2 vector (CLONTECH Laboratories) and used to express a green fluorescent protein (GFP)¹–small espin fusion protein in cells of the BHK fibroblastic line (American Type Culture Collection, Rockville, MD) by transient transfection with lipofectamine (Life Technologies, Gaithersburg, MD). The cells were cultured on coverslips in DME containing 10% (vol/vol) calf serum and penicillin/streptomycin. The GFP–small espin fusion protein was localized in these cells by conventional fluorescence microscopy, either without fixation or after fixation with 2% paraformaldehyde in PBS, pH 7.4, extraction for 5 min with ice-cold 0.1% (vol/vol) Triton X-100 in PBS, and labeling with rhodamine-phalloidin (Molecular Probes). In some transfection experiments, small espin was expressed without GFP, using the pcDNA3 expression vector (Invitrogen, Carlsbad, CA), and detected by immunofluorescence (Bartles et al., 1996).

cDNAs encoding the full-length small espin protein and NH₂- or COOH-terminally truncated versions were prepared from the library cDNA clone by PCR using Deep Vent DNA polymerase (New England Biolabs, Beverly, MA) with limiting amounts of template and enzyme. Primers were designed to introduce a BamHI site at the 5′ end and an EcoRI and/or a HindIII site at the 3′ end. The cDNAs were introduced into the BamHI and EcoRI or HindIII sites of the pProEX HT prokaryotic expression vector (Life Technologies) that gave the proper reading frame. The resulting constructs were checked by automated DNA sequencing (using a BigDye terminator and model 377 sequencer; Applied Biosystems, Inc., Foster City, CA) and used to transform Escherichia coli DH5α (Life Technologies). These resulting recombinant proteins included an additional 28 amino acids at their NH₂ termini: MSYY, followed by HHHHHH (the 6xHis tag), DYDIPTT (a spacer region), ENLYFQ (Tobacco Etch Virus protease cleavage site), and GAMGS. The 6xHis-tagged small espin constructs were isolated from 50 mM Tris-HCl, 10 mM 2-mercaptoethanol, pH 8.5, extracts of frozen-thawed, sonicated bacteria by batch affinity chromatography on Ni-NTA resin (Qiagen, Chatsworth, CA) using the 0.1 M KCl, 10% (vol/vol) glycerol, 10 mM 2-mercaptoethanol, 20 mM Tris-HCl, 20 mM imidazole-HCl, pH 8.5, buffer system recommended by Life Technologies. The 0.2-M imidazole eluate was dialyzed against 0.1 M KCl, 10 mM imidazole-HCl, 1 mM NaN₃, pH 7.4; treated at a concentration of 0.02–0.05 mg/ml with 15 U/ml of recombinant 6xHis-tagged Tobacco Etch Virus protease (Life Technologies) for 3 h at 37°C; incubated for 30 min at 4°C with Ni-NTA resin to remove cleaved 6xHis-tag, any uncleaved 6xHis-tagged small espin, and the residual 6xHis-tagged viral protease; adjusted to 2 mM MgCl₂; and freed of any insoluble protein by centrifugation at 150,000 g for 90 min at 4°C in preparation for use in F-actin binding and bundling assays. After removal of the 6xHis tag with Tobacco Etch Virus protease, the constructs retained only an extra five amino acids, GAMGS, at their NH₂ terminus.

F-actin was prepared by dilution of purified rabbit skeletal muscle actin or human platelet nonmuscle actin (85% β-actin and 15% γ-actin) (Cytoskeleton, Inc.) into 0.1 M KCl, 2 mM MgCl₂, 1 mM ATP, 1 mM NaN₃, 10 mM imidazole-HCl, pH 7.4, and incubation for 60 min at 37°C. To assay for F-actin binding or bundling, an equal volume of solution containing different amounts of recombinant small espin protein in the same buffer minus ATP was added to preformed actin filaments (at a final actin concentration of 0.1–0.5 mg/ml) and incubated for 60 min at 37°C. Samples were either taken for negative staining with 0.5–1% (wt/vol) uranyl acetate on 300-mesh Formvar and carbon-coated copper grids (Cooper and Pollard, 1982) or were centrifuged at 4°C for either 15 min at 22,000 g (bundling assay; Edwards et al., 1995) or for 90 min at 150,000 g (binding assay; Bartles et al., 1996). The levels of small espin protein and actin present in the supernate and pellet fractions were determined by scanning laser densitometric analysis of Coomassie blue–stained SDS gels using the rabbit skeletal muscle actin as the protein standard.

For sedimentation equilibrium, the purified recombinant espin was dialyzed against 0.1 M KCl, 0.02 M Tris-HCl, 2 mM MgCl₂, 1 mM NaN₃, pH 7.4, and centrifuged for 22–24 h at 15,000 rpm in an analytical ultracentrifuge (model XLA-1; Beckman Instruments, Fullerton, CA). The molecular mass was estimated from plots of log A₂₈₀ versus radius squared. The modeling of molecular dimensions (semimajor [a] and semiminor [b] axes multiplied by 2) and sedimentation coefficient using subunit molecular mass and frictional coefficient ratio for three basic asymmetrical shapes (prolate ellipsoid, oblate ellipsoid, and long rod) were carried out using the XL-A-UltraScan-Origin Software Package Version 2.93 produced by Dr. Borries Demeler (Missoula, MT) and supplied with the instrument. To measure sedimentation coefficient, the purified espin was centrifuged for 3 d at 25,000 rpm under isokinetic conditions in a linear 3–15% (wt/ vol) sucrose gradient in 0.1 M KCl, 2 mM MgCl₂, 10 mM imidazole-HCl, 1 mM NaN₃, pH 7.4. To measure Stokes' radius, the purified espin was chromatographed on a 93 × 1.5-cm column of Sephadex G-100 in the same buffer minus sucrose. The sedimentation coefficient and Stokes' radius were determined by SDS gel analysis of the gradient and column fractions, respectively, with comparison to the protein standards in the MW-GF-1000 gel filtration standards kit plus horse heart cytochrome c (Sigma Chemical Co., St. Louis, MO). The small espin was examined in SDS gels before and after centrifugation or gel filtration to ensure that it had not undergone proteolysis during the procedure. Circular dichroism spectroscopy was carried out at room temperature at rate of 50 nm/min using a freshly calibrated spectropolarimeter (model J-715; Jasco, Inc., Easton, MD) on samples containing ∼50 μg/ml of purified small espin in 50 mM NaP_i, pH 7.4, in a 0.1-cm path length cell. Molecular cross-linkers were purchased from Pierce Chemical Co. (Rockford, IL).

To determine whether the ∼1.7-kb crosshybridizing RNAs detected previously in our Northern blot analysis of total RNA isolated from rat kidney and small intestine (Bartles et al., 1996) encoded espin-related proteins, homogenates of rat kidney and small intestinal mucosa were compared with homogenate of rat testis on Western blots using affinity-purified rabbit polyclonal antibody directed against the COOH-terminal 379–amino acid peptide of rat Sertoli cell espin. In addition to reacting with the ∼110-kD Sertoli cell espin band present in homogenate of testis, the espin antibody also specifically labeled a closely spaced doublet of ∼30-kD present in homogenate of rat kidney and small intestinal mucosa (Fig. 1 A).

To elucidate the sequence of the ∼30-kD protein detected by Western blotting, the espin antibody was used to screen a commercial rat kidney cDNA library. A single 1,088-bp cDNA clone was obtained, and its sequence was determined. The single large open-reading frame present in this cDNA began at nucleotide 63, in the context of a Kozak consensus sequence (Kozak, 1987), and was predicted to encode a small version of the espin protein that was ∼1/3 the size of Sertoli cell espin, 253 amino acids versus 837 amino acids,² with a molecular mass of 28,240 D and an isoelectric point of 10.5. Attempts at 5′ rapid amplification of cDNA ends failed to identify any additional in-frame translation start sites upstream (data not shown), and the recombinant protein encoded by nucleotides 63– 821 of this cDNA was found to exhibit the expected apparent molecular mass (∼30 kD) in SDS gels (see below). The nucleotide sequence, the predicted amino acid sequence, and a comparison between the small and large espin isoforms are shown in Fig. 2. Along much of its length, small espin was identical in sequence to the COOH-terminal portion of the larger Sertoli cell isoform. As a result, it contained the potential P-loop (A32-S39), the 66–amino acid peptide that showed 39% identity to a peptide encoded by exon A5 and is present in both the large and small forked proteins of Drosophila (I140-K205), and the two peptides that contained clusters of multiple glutamates bracketed by arginines, lysines, and glutamines in a pattern reminiscent of the repetitive motif found in the protein trichohyalin (R160-R182 and R195-R219). Small espin did not, however, contain the ankyrin-like repeats and proline-rich peptides present in the large isoform. In their place, it contained a considerably shorter NH₂ terminus with two unique small peptides (M1-P13 and S61-R86) interrupted by another peptide (S14-A60) that was present in the large isoform. When analyzed using the secondary structure algorithms of Chou and Fasman (1978) or those of Garnier et al. (1978), a large fraction of the COOH-terminal half of small espin (R160-A238) was predicted to adopt an α-helical secondary structure. In addition, when analyzed using the algorithms or Lupas (1997) or Berger et al. (1995), two COOH-terminal peptides (Q156-L186 and R190-G231 or L194-G231) were predicted to form a coiled coil with high probability.

When localized at the light-microscopic level using the immunoperoxidase technique, small espin was found to be concentrated over the F-actin–rich brush borders of absorptive epithelial cells in the small intestine and kidney (arrowheads in Fig. 3, A and B, respectively). On sections of rat small intestine, some weaker labeling was also detected more basally within enterocytes. Compared with the brush borders of the mature enterocytes on intestinal villi, the apical margins of the enterocyte precursors found within the crypts of Lieberkühn (e.g., Fig. 3,A, arrow) had considerably lower levels of labeling. This suggested that small espin was expressed or accumulated in a developmental fashion during the assembly of the brush border (Ezzell et al., 1989; Heintzelman and Mooseker, 1992; Fath and Burgess, 1995). On sections of rat kidney, only the epithelial cells lining the distal straight portion of the proximal tubule near the cortico-medullary junction, presumably corresponding to segment S3 (Kriz and Kaissling, 1992), showed intense labeling along their brush borders (Fig. 3 B). Proximal convoluted tubules near glomeruli in the renal cortex showed considerably less labeling (data not shown). The proximal tubular epithelial cells with intensely labeled brush borders appeared to have an even higher level of cytoplasmic labeling than the enterocytes.

The majority (83–86%) of small espin present in homogenates of rat kidney and small intestinal mucosa resisted extraction with 1% (vol/vol) Triton X-100 (data not shown). Small espin was recovered in isolated rat small intestinal brush borders, where it was colocalized with F-actin at the light-microscopic level (Fig. 4, A–C). On the basis of quantitative Western blotting, we could estimate that ∼1 small espin was recovered for every 130 actin monomers in small intestinal brush borders isolated from adult rats. Fig. 1,B shows a Western blot in which dilutions of isolated rat small intestinal brush borders containing known amounts of actin are compared with internal standards containing known amounts of recombinant small espin (see below). Consistent with an association between small espin and the actin cytoskeleton, the small espin present in the isolated rat small intestinal brush borders resisted extraction with 1% (vol/vol) Triton X-100 but could be partially extracted with 0.6 M KI (Fig. 1 C).

Post- and preembedment immunogold electron microscopy were used to examine the localization of the small espin recovered on the isolated rat small intestinal brush borders at a higher level of resolution. When LR white sections of paraformaldehyde-fixed brush borders were labeled with affinity-purified polyclonal espin antibody followed by secondary antibody–colloidal gold conjugate, the bound gold particles were observed over the microvilli proper and over the actin bundles of the microvillar rootlets, but not over the terminal web (Fig. 4,D). Neither the sections labeled with preimmune IgG instead of the espin antibody nor the occasional nuclear fragment that can contaminate this subcellular fraction showed bound gold particles (data not shown). After examining more than 250 brush border profiles, we concluded that the specifically bound gold particles could be found anywhere along the length of the microvillar actin bundle. Preembedment immunogold labeling (Fig. 4 E) indicated that a significant pool of small espin was accessible to primary antibody and secondary antibody–colloidal gold conjugate in isolated brush borders that were fixed with paraformaldehyde in the presence of Triton X-100. Specimens labeled with preimmune IgG instead of espin antibody showed no bound gold particles (not shown). Compared with postembedment labeling, the specifically bound gold particles appeared to be more evenly distributed along microvilli, but fewer gold particles were detected over rootlets, suggesting that there was reduced accessibility of the tracer using the preembedment technique. Comparable results were obtained when the diameter of the secondary antibody– colloidal gold was reduced from 12 to 6 nm (data not shown).

To obtain additional evidence in support of an interaction between small espin and the actin cytoskeleton in vivo, we examined the localization of a GFP–small espin fusion protein in transiently transfected cells of the BHK fibroblastic line. On the basis of immunofluorescence and Western blotting (not shown), these cells were judged not to contain significant amounts of espin in the absence of transfection. When living cells were examined by conventional fluorescence microscopy 1 d after transfection, the GFP–small espin fusion protein was observed to decorate fine (Fig. 5,A) or more coarse (Fig. 5, B and C) stress fiber–like structures, whereas GFP without the small espin attached was found to be distributed more diffusely throughout the cytoplasm with evidence of concentration within the nucleus (Fig. 5,B, inset). To further examine the identity of the GFP–small espin–labeled fibers observed in living cells, the cells were fixed with paraformaldehyde, permeabilized with Triton X-100 and double labeled with rhodamine-phalloidin as a marker for F-actin. The patterns of GFP–small espin (Fig. 5,D) and rhodamine-phalloidin (Fig. 5,E) were found to be coincident, suggesting that exogenous small espin was associating with F-actin– containing structures in the transfected cells. In contrast, GFP without the small espin attached was detected in perinuclear aggregates in the fixed and permeabilized cells (Fig. 5,D, inset). Interestingly, the level of F-actin, as reflected by the intensity of rhodamine-phalloidin labeling, was consistently higher in cells expressing GFP–small espin than in the corresponding untransfected control cells (Fig. 5 E) or in cells expressing GFP alone (not shown). This latter result suggested that small espin could elicit the accumulation and/or bundling of F-actin–containing fibers or cables in the transfected cells. Neither the apparent accumulation of these F-actin–containing structures nor their decoration with small espin required the presence of the GFP because the same results were obtained when small espin was expressed without the GFP using the pcDNA3 vector and was detected by immunofluorescence (data not shown).

To examine the properties of small espin and its interactions with F-actin under defined conditions in vitro, it was necessary to obtain relatively large amounts of the purified protein. Since small espin was difficult to extract from organ/tissue homogenates and isolated brush borders (see above), we elected to use recombinant small espin for these studies. A full-length version of small espin containing a 28–amino acid NH₂-terminal extension that included a 6xHis tag was expressed in E. coli and purified by affinity chromatography on Ni-NTA agarose under nondenaturing conditions. The 6xHis tag was released from the protein by cleavage with recombinant 6xHis-tagged Tobacco Etch Virus protease, and the protein, which now contained only five extra amino acids (GAMGS) at its NH₂ terminus, was quickly repurified by a second challenge with Ni-NTA agarose before use.

The purified recombinant full-length small espin was found to bind to and bundle F-actin at physiological ionic strength and pH. This is illustrated in the left part of Fig. 6,A, where the supernate and pellet fractions resulting from an F-actin–bundling assay, which uses low-speed centrifugation to separate bundles from preformed actin filaments, were analyzed in an SDS gel. In the absence of small espin, the vast majority of the rabbit skeletal muscle F-actin remained in the supernate fraction upon low-speed centrifugation. However, when the small espin was added to the F-actin (in this case, at a molar ratio of approximately one espin/seven actin monomers), the solution became noticeably turbid, and the vast majority of the F-actin and the small espin were obtained as higher molecular weight cross-linked structures (bundles) in the pellet. In contrast, in the absence of F-actin, the small espin remained in the supernate. When examined by negative staining electron microscopy, the vast majority of the actin filaments present in the small espin/F-actin mixtures were found to be assembled into bundles (Fig. 7). In contrast, the F-actin controls contained randomly oriented actin filaments but no discernible bundles (data not shown). The small espin–induced bundles were partially ordered, frequently showing extended segments with transverse striations, aligned imperfectly across the width of the bundle, at integral multiples of ∼12 nm, which is characteristic of maximally cross-linked, hexagonally packed filaments (e.g., see Stokes and DeRosier, 1991; Tilney et al., 1995). Unlike the situation for many other actin-bundling proteins, which display actin-bundling activities that are inhibited by Ca²⁺(Glenney et al., 1981; Alicea and Mooseker, 1988; Namba et al., 1992; Lin et al., 1994), concentrations of CaCl₂ up to 1 mM had no effect on the bundling of F-actin by small espin in the low-speed centrifugation assay (data not shown). And, unlike the bundling of F-actin that results from counterion condensation by positively charged ions and small peptides (Tang and Janmey, 1996), relatively high concentrations of negatively charged small molecules, e.g., 10 mM ATP, also had no effect on the bundling caused by small espin (not shown).

To determine the affinity of the interaction between recombinant small espin and F-actin, high-speed centrifugation was substituted for the low-speed centrifugation to cause all of the F-actin to be recovered in the pellet. The data resulting from the comparison of high-speed pellet and supernate fractions at different inputs of small espin were used to construct Scatchard plots. As is illustrated in Fig. 6 B, the recombinant small espin was found to bind to human platelet nonmuscle F-actin with a threefold higher affinity than to rabbit skeletal muscle F-actin. The K_d for nonmuscle F-actin was determined to 50 nM, whereas the K_d for skeletal muscle F-actin was 150 nM. From the intercepts of the resulting linear regression lines on the abscissa, it could be calculated that the binding reached saturation at approximately one espin for every three to four actin monomers.

To further elucidate the mechanism of actin bundling by small espin, we used sedimentation equilibrium to determine whether the purified recombinant small espin was a monomer, dimer or higher oligomer. From the slope of plots of log A₂₈₀ versus radius squared at equilibrium (e.g., Fig. 8 A), the molecular mass was estimated to be 25,500 D, only ∼12% below the value of ∼29,000 D (28,240 D + NH₂-terminal GAMGS) calculated on the basis of sequence, suggesting that recombinant small espin was a monomer. In further support of a monomeric structure for recombinant small espin, we failed to observe covalently cross-linked dimers or higher oligomers when the purified protein was reacted with molecular cross-linkers, such as 3,3′-dithiobis(sulfo-succinimidyl propionate) and 1,5-difluoro-2,4-dinitrobenzene, and analyzed in SDS gels (data not shown).

Given that recombinant small espin appeared to be a monomer, it was found to have an unexpectedly large Stokes' radius by gel filtration. When chromatographed in columns of Sephadex G-100, the recombinant small espin eluted only one fraction after bovine serum albumin, a standard protein with more than twice the subunit molecular mass (66 kD), and seven fractions before carbonic anhydrase, a standard protein with approximately the same subunit molecular mass (29 kD) (Fig. 8,B). From the plot of logarithm of the Stokes' radius versus peak elution fraction for the standard proteins (Fig. 8,B, inset), the Stokes' radius of recombinant small espin was estimated to be 3.4 nm, suggesting that it had a frictional coefficient ratio (f/f_o) of 1.7 and was, therefore, an asymmetrical molecule. For example, it is possible to calculate that a frictional coefficient ratio of 1.7 for a protein of ∼29,000 D (28,240 D + NH₂-terminal GAMGS) is compatible with an oblate ellipsoid with dimensions of 10 × 0.6 nm, a prolate ellipsoid with dimensions of 23 × 1.8 nm, or a long rod with dimensions of 17 × 1.2 nm. Using the molecular mass of the small espin monomer and the frictional coefficient ratio of 1.7, one can estimate that the protein should have a sedimentation coefficient of ∼2 S. The sedimentation coefficient of the recombinant small espin was measured directly by centrifugation under isokinetic conditions in sucrose density gradients (Fig. 8,C). From a plot of sedimentation coefficient versus peak fraction for the standard proteins (Fig. 8 C, inset), the sedimentation coefficient of small espin was determined to be 2.1 S, just slightly above the predicted value. The observation that small espin sedimented significantly slower than the 29-kD protein carbonic anhydrase, the standard protein with approximately the same molecular mass, also supported the conclusion that small espin was an asymmetrical monomer.

When examined by far-UV circular dichroism spectroscopy, the spectrum of the purified full-length recombinant small espin had large negative peaks at 220 and 207 nm and a large positive peak at 192 nm (Fig. 8 D), corresponding to the major n-π* and π-π* transitions displayed by α-helices (Chen et al., 1974). These results suggested that the recombinant small espin was a highly folded protein and that, as predicted on the basis of sequence (see above), it contained a high percentage of α-helix. Unlike the spectra of many other α-helical proteins, however, the negative peak at 207 nm was significantly larger than the one at 220 nm, suggesting that the helical region was relatively long (>50 amino acids; Chen et al., 1974) and/or that, despite the prediction of a coiled-coil structure (see above), its various helical segments did not interact with one another (e.g., Zhou et al., 1992; Su et al., 1994).

PCR was used to generate a series of small espin NH₂- or COOH-terminal deletion constructs with NH₂-terminal 6xHis tags to allow for ready purification. The sequences of these constructs were verified by automated DNA sequencing, and the constructs were compared in the low-speed centrifugation actin-bundling assay after removal of the 6xHis tag and repurification. Deletion of the NH₂ terminus up to the 167–amino acid COOH-terminal peptide shared by the Sertoli cell and brush border isoforms (ΔN86), or up to the beginning of the forked homology domain (ΔN137), produced recombinant proteins that displayed actin-bundling activities equal to, or perhaps slightly greater than, that of full-length small espin (Fig. 9). An example of an SDS gel depicting the results of an actin-bundling assay for construct ΔN137 is shown on the right side of Fig. 6,A. Further deletion from the NH₂ terminus, up to the middle of the forked homology peptide (ΔN170), caused a dramatic (80-fold) decrease in actin-bundling activity to near background levels (Fig. 9). Deleting the last 5 amino acids from the COOH terminus of small espin (ΔC5) had no effect on actin-bundling activity, but deletion of an additional 13 amino acids (ΔC18), or deletion of a total of 73 amino acids from the COOH terminus (ΔC73), also caused a dramatic decrease in actin-bundling activity to near background levels (Fig. 9). These results suggested that the COOH-terminal 116–amino acid peptide common to the two known isoforms of espin (A138-K253 in small espin) was sufficient to cause maximum actin bundling in vitro. Since small espin was found to be a monomer in solution (see above), the COOH-terminal 116–amino acid peptide must presumably contain two or more actin-binding sites. The dramatic losses in actin-bundling activity observed in going from ΔN137 to ΔN170 and from ΔC5 to ΔC18 suggested further that the 33–amino acid peptide at the NH₂ terminus of the forked homology peptide (A138-E170) and the 13–amino acid peptide K236-K248 were necessary for bundling activity and, therefore, may contribute to the formation of small espin's actin-binding sites.

Taken together, our data strongly support the hypothesis that small espin is a newly identified actin-bundling protein of the brush border. Small espin is localized to the parallel actin bundles of brush border microvilli, and it accumulates there coincident with the migration/differentiation of enterocytes along the crypt–villus axis in the adult. The protein is recovered in the Triton X-100–insoluble cytoskeletal fraction prepared from isolated brush borders. Recombinant small espin binds to F-actin with high affinity, and it cross-links the actin filaments into partially ordered bundles. The recombinant small espin protein is a monomer in solution and appears to contain at least two relatively closely spaced actin-binding sites. And finally, when expressed exogenously in transfected fibroblastic cells, small espin decorates F-actin fibers and cables and appears to cause their accumulation, suggesting that small espin can bind to, stabilize, and/or bundle F-actin in vivo.

Small espin showed no significant sequence similarity to other proteins known or suspected to bind/bundle F-actin, with the exception of the large isoform of espin and the forked proteins of Drosophila. With the identification of the small isoform of espin, it is clear that the espins constitute a new family of actin-binding/bundling proteins. The two members of the espin family we have described thus far, the 837–amino acid ectoplasmic specialization isoform (Bartles et al., 1996) and the 253–amino acid brush border isoform, are expressed in different cell types and are markedly different in size. Nevertheless, both isoforms are localized to specializations of the cell surface, an invagination or an evagination, that contain hexagonally packed parallel actin bundles, and both contain an identical COOH-terminal 167–amino acid peptide that is sufficient to bundle F-actin in vitro. The observation that the peptides specific to the small isoform (M1-P13 and S61-R86) are found interspersed among peptides identical in sequence to those present in the large isoform suggest that the two espin isoforms arise through differential splicing. We are currently investigating the organization of the espin gene(s) and comparing the properties and activities of the small and large isoforms to ascertain how their different NH₂ termini influence interactions with F-actin.

Although small espin is a basic protein and contains many highly charged peptide segments, the fact that the bundling took place in 0.1 M KCl at pH 7.4 and was not inhibited by a small, highly charged anion, in this case up to 10 mM ATP, suggests that the bundling is due to the presence of specific actin-binding sites in small espin and is not the result of counterion condensation (Tang and Janmey, 1996). Through deletion mutagenesis, we established that the COOH-terminal 116–amino acid peptide shared by the two espin isoforms was sufficient to cause maximal actin bundling in the low-speed centrifugation assay. Since small espin proved to be a monomer in solution, this suggests that the COOH-terminal 116–amino acid peptide contains two, or possibly more, actin-binding sites. The simplest model that can account for the dramatic losses in actin-bundling activity noted in going from ΔN137 to ΔN170 and from ΔC5 to ΔC18 is one in which small espin contains two actin-binding sites disposed roughly at opposite ends of the COOH-terminal 116–amino acid peptide. Our sedimentation and gel filtration analyses suggest that small espin is a relatively small, asymmetrical molecule with dimensions in the range of 10–23 × 0.6–1.8 nm. Further experiments are underway to elucidate the three-dimensional structure of small espin and to map its actin-binding sites and its footprint on F-actin at high resolution.

The dramatic loss in the actin-bundling activity of small espin noted in going from ΔN137 to ΔN170 suggests that the NH₂-terminal half of its 66–amino acid forked homology peptide is necessary for actin-bundling activity and may therefore be important in forming one of small espin's actin-binding sites. This lends support to the hypothesis (Bartles et al., 1996) that the espins and the forked proteins are orthologous families of proteins. Although, to our knowledge, no forked protein has yet been observed to bundle F-actin directly, genetic studies have demonstrated that the forked proteins are necessary to form normal actin bundles in the developing bristles of Drosophila pupae (Hoover et al., 1993; Petersen et al., 1994; Tilney et al., 1995; Wulfkuhle et al., 1998). Forked proteins have been localized to the developing bristles of Drosophila pupae by immunofluorescence (Petersen et al., 1994), suggesting that they, too, may be components of plasma membrane–associated parallel actin bundles. The forked proteins also appear to come in multiple isoforms: six forked transcripts have been detected during mid to late pupal development, and five of the six isoforms are predicted to share a COOH-terminal peptide, which includes the 66–amino acid peptide with 39% identity to the forked homology peptide present in the shared COOH-terminal peptide of the espins (Hoover et al., 1993; Bartles et al., 1996). On the basis of the ability of one of the smaller forked proteins to restore normal actin bundle and bristle morphology in transgenic forked mutant pupae, it is presumably the shared COOH-terminal peptide that contains the actin-bundling module postulated to exist within the forked proteins (Petersen et al., 1994). A peptide present within the shared COOH-terminal peptide of the forked proteins—the one encoded by exon A5 of the forked gene that includes the 66–amino acid peptide with homology to the espins—has, in fact, been observed to bind to F-actin (Petersen, N.S., personal communication). Although other primary structural features of the forked proteins (e.g., the alternating histidine and proline residues of the PRD-like domain, the polyglutamine tracts encoded by the CAX trinucleotide repeats, and the peptides composed of large numbers of predominantly hydrophobic amino acids) are notably absent from the espins, the larger isoforms of the forked proteins do contain two proline-rich peptides and NH₂-terminal ankyrin-like repeats with an overall spacing that approximates that observed in the large isoform of espin (Hoover et al., 1993; Bartles et al., 1996). It is presently unclear whether forked proteins will be found in other tissues of Drosophila. The forked phenotype appears to be limited to the neurosensory bristles of pupae and to the hairs and denticles of larvae and adults, and to our knowledge, these are the only structures in Drosophila found, or postulated, to contain a forked protein to date (Hoover et al., 1993; Petersen et al., 1994; Tilney et al., 1995; Dickinson and Thatcher, 1997; Wulfkuhle et al., 1998).

At a minimum, our results suggest that microvillar actin bundles are even more complex and heterogeneous than thought previously. On the basis of our findings, we propose that in some cells it may take not two, but three actin-bundling proteins to construct and/or maintain a fully functional microvillar actin bundle. An extension of the prevailing two-bundling protein model (Tilney et al., 1995), the three-bundling protein model would appear to be consistent with the results of recent experiments examining pupal bristle formation in singed/forked double mutants of Drosophila that imply the involvement of at least one additional actin-bundling protein (Wulfkuhle et al., 1998). It would also appear to follow logically from our observation that small espin is present at different levels in the brush borders of cells that line different segments of the proximal tubule in rat kidney. In this regard, it may be relevant that the cells in the kidney we found to contain the largest amounts of brush border small espin, those lining the distal straight segment (S3) of the proximal tubule, are also known to display microvilli that are longer and more densely packed than those in other segments of the proximal tubule in the rat (Kriz and Kaissling, 1992).

Small espin appeared to be 10- to 20-fold less abundant than the two known major actin-bundling proteins, fimbrin/plastin and villin, in small intestinal brush borders isolated from adult rats. Barring losses due to degradation or extraction during brush border isolation, we can estimate that there is only ∼1 espin for every 130 actin monomers present in the intestinal brush border. Whether calculated assuming the theoretical value for maximally cross-linked, hexagonally packed filaments of ∼1 cross-link for every 4.3 actin monomers (Tilney et al., 1995), or calculated using the ∼10:0.65:1.3 molar ratio of actin monomer/fimbrin/ villin in isolated intestinal brush borders (Coluccio and Bretscher, 1989), this stoichiometry suggests that only ∼1 in every 30 crossbridges between actin filaments would be due to small espin in the microvillar actin bundles of the adult small intestine. Although its lower abundance may help to explain why small espin escaped the attention of earlier investigators, it makes it more challenging for us to identify the role(s) of small espin in microvillar actin bundle organization and/or function.

In the context of a three-bundling protein hypothesis, it is possible to envision at least a few different scenarios that would be compatible with the available evidence. One especially intriguing possibility is that small espin might confer a special property upon the bundle and that it is sufficient to have the cross-links involving this third bundling protein spaced relatively far apart in the completed structure. One obvious special property that small espin could confer upon a bundle is high-affinity cross-linking that is insensitive to changes in the concentration of Ca²⁺. Small espin exhibits a binding affinity for F-actin that is at least an order of magnitude higher than those displayed by many of the other known actin-bundling proteins, including fimbrin/plastin and villin (Glenney et al., 1981; Burgess et al., 1987; Pollard, 1993). In addition, both villin and fimbrin/plastin have binding sites for Ca²⁺, and the bundling activities of villin and fimbrin/plastin are known to be disrupted by physiological concentrations of Ca²⁺ (Glenney et al., 1981; Alicea and Mooseker, 1988; Namba et al., 1992; Lin et al., 1994). However, we found that the actin-bundling activity of small espin was insensitive to changes in the concentration of Ca²⁺ over a wide range. We are currently testing the hypothesis that strategic placement of small espin at intervals along the filaments might reinforce the bundle by protecting it from changes in Ca²⁺ concentration.

A second special property that small espin might be able to confer upon actin bundles is stabilization to depolymerization. This would be one possible explanation for the apparent accumulation of F-actin we noted in the small espin-expressing transfected cells. A similar stabilization effect, apparently independent of cross-linking, has already been noted in vitro for one actin-bundling protein, the ∼34-kD actin-bundling protein of Dictyostelium (Zigmond et al., 1992). In the case of the small espin-expressing transfected cells, it is difficult to sort out the effects of stabilization versus bundling. For example, we do not yet know the extent to which the F-actin fibers and cables observed in the transfected cells contain parallel actin bundles in addition to the antiparallel (contractile) bundles typical of stress fibers. We are currently carrying out experiments to determine whether the espin isoforms can stabilize F-actin independently of cross-linking. We are also investigating the possibility that the apparent accumulation of F-actin observed in the transfected cells reflects an increase in actin synthesis to maintain a sizable pool of monomer. Nevertheless, it is striking that the results we obtained with small espin are quite different from those obtained by others examining the localization and effects of villin and fimbrin/plastin in transfected fibroblastic cells: villin causes a loss of stress fibers and the redistribution of F-actin into long spike-like microvilli at the dorsal surface (Friederich et al., 1989), whereas T- and L-plastins cause fibroblastic cells to round up and reorganize their stress fibers into geodesic structures (Arpin et al., 1994). We are currently comparing the localizations and effects of the espin isoforms in transfected epithelial cells.

Another special property that could be conferred upon the bundles by small espin might be to link the bundles to other proteins, possibly even to those of signaling cascades. The ability to act as a linker or adaptor protein would appear to be especially likely for the large isoform of espin, with its multiple motifs for protein–protein interaction—the ankyrin repeats and the proline-rich peptides (for additional discussion, see Bartles et al., 1996)—but could also readily apply in the case of the small isoform. We are currently using a combination of affinity approaches and two-hybrid screens to identify espin-binding proteins other than actin.

One possible explanation for why small espin is relatively minor in the intestinal brush borders of adult rats is that it might be required at an earlier step in brush border assembly. In such a case, it might be expressed at lower levels or be more unstable in adults. The results of our immunoperoxidase localization of small espin in the small intestines of adult rats suggest, however, that this is probably not the case. In many respects, the assembly of the intestinal brush border along the crypt–villus axis in the adult approximates an accelerated version of intestinal brush border assembly during late embryonic development (Heintzelman and Mooseker, 1992; Fath and Burgess, 1995). We noted a large differential in the intensity of apical immunoperoxidase staining between crypt cells and villus cells in adult intestine. Although there are some limitations inherent in trying to use immunoperoxidase histochemistry as a quantitative measure, our results suggest that maximal small espin accumulation occurs not early, but relatively late during enterocyte differentiation and brush border assembly. In contrast, both villin and fimbrin/plastin have been found to be nearly as prevalent in the apical cytoplasm of crypt cells in the adult as they are in the brush borders of the mature enterocytes (Fath et al., 1990; Heintzelman and Mooseker, 1990; Landry et al., 1994). The increase in immunoperoxidase staining for small espin was first noted for enterocytes nearing the crypt–villus junction, roughly coinciding with the timing of the terminal elongation of microvilli and the arrival of brush border myosin I (Heintzelman and Mooseker, 1992; Fath and Burgess, 1995). To address this issue more fully, we are carrying out an analysis of small espin expression and localization in relation to other brush border proteins during the development of intestine and kidney and in cell culture models of enterocyte differentiation.

Even though small espin appears not to be as major as fimbrin/plastin or villin in the intestinal brush borders of adults, it is still possible that it is much more abundant in other F-actin–containing structures in the body. This situation would be somewhat analogous to that of ezrin, which is a minor protein relative to brush border myosin I in the intestinal brush border, but appears to supplant brush border myosin I as the major F-actin–membrane linker of placental microvilli (Bretscher et al., 1997). In such a case, small espin could be partially redundant in function in the intestinal brush border of normal adult rats and of primary importance only when one of the other actin-bundling proteins is impaired or missing. In this regard, it has been reported that mice that are missing villin due to targeted gene disruption have surprisingly normal looking intestinal brush borders (Pinson et al., 1998). In preliminary experiments, increased levels of brush border espin have been detected in villin gene knock-out mice by immunofluorescence and Western blotting (Ferrary, E., and S. Robine, personal communication). In addition, we have obtained preliminary evidence that espin isoforms are indeed more plentiful in other structures that contain parallel actin bundles (unpublished data). The further characterization of the espin isoforms found in these other locations is in progress.

We thank Dr. Mike Shaw (University of Pennsylvania, Philadelphia, PA) for advice about immunogold labeling of brush borders; Dr. Dan Kiehart (Duke University, Chapel Hill, NC) for advice about PHEM buffers and coiled coils; Maya Moody for help with sectioning, microscopy, and digital imaging; Dr. Rob Moir and Miri Yoon for advice about bacterial expression and GFP transfection systems; Dr. Kate Spiegel for help with the analytical ultracentrifuge; Dr. Brian Schoichet for help with the circular dichroism spectroscopy; Jodi Irwin for help with automated sequencing; Dr. Nancy Petersen (University of Wyoming, Laramie, WY) and Dr. Evelyne Ferrary and Dr. Sylvie Robine (Institut Curie, Paris, France) for granting us permission to discuss their results before publication; and Dr. Rex Chisholm for valuable discussions.

This work was supported by American Cancer Society research grant #CB-203, National Institutes of Health (NIH) grant R01 HD35280, and NIH Independent Scientist Award K02 HD01210 awarded to J.R. Bartles.