Dependence of fibroblast migration on actin severing activity of gelsolin.

Gelsolin nucleates actin filament assembly, blocks the fast-exchanging ends of actin filaments, and severs filaments, processes that may play an important role in cell motility. We studied the relationship between cell migration, gelsolin content, and actin severing activity in human gingival fibroblasts. These cells were keratin negative and desmin negative but expressed vimentin and myosin II. Cells were separated by their ability to migrate in response to a chemoattractant stimulus. Northern analysis of mRNA, [35S]methionine incorporation into immunoprecipitated gelsolin, immunoblots of cell lysates, and quantitative confocal microscopy showed 1.4-2-fold higher levels of gelsolin in nonmigrant compared with migrant cells. Because the concentration of intracellular gelsolin did not appear to be a central determinant of cell migration, we assessed its requirement for motility. Cells that were electroinjected with a gelsolin antibody that inhibits actin severing by gelsolin in vitro showed a 72% reduction of the number of migrant cells compared with controls treated with an irrelevant antibody. Cells that were electroinjected with free gelsolin exhibited a 33% increase in migration compared with cells electroinjected with bovine serum albumin. Compared with nonmigrant cells, migrant cells contained abundant free gelsolin and exhibited gelsolin-dependent F-actin severing activity, which required Ca2+. Serum stimulation of cell migration required increases in [Ca2+]i because incubation with 3 μM 1,2-bis-(o-aminophenoxy)-ethane-N,N,N′,N′-tetraacetic acid tetra(acetoxymethyl)-ester blocked calcium flux and cell migration. Serum also stimulated the recruitment of gelsolin into the supernatants of lysates from migrant but not from nonmigrant cells. In fibroblasts, gelsolin concentration alone does not apparently determine migratory capacity. Instead, the Ca2+-dependent actin severing activity of free gelsolin appears to be a major determinant of cell migration.

Gelsolin nucleates actin filament assembly, blocks the fast-exchanging ends of actin filaments, and severs filaments, processes that may play an important role in cell motility. We studied the relationship between cell migration, gelsolin content, and actin severing activity in human gingival fibroblasts. These cells were keratin negative and desmin negative but expressed vimentin and myosin II. Cells were separated by their ability to migrate in response to a chemoattractant stimulus. 2؉ ] i because incubation with 3 M 1,2-bis-(o-aminophenoxy)-ethane-N,N,N,N-tetraacetic acid tetra(acetoxymethyl)-ester blocked calcium flux and cell migration. Serum also stimulated the recruitment of gelsolin into the supernatants of lysates from migrant but not from nonmigrant cells. In fibroblasts, gelsolin concentration alone does not apparently determine migratory capacity. Instead, the Ca 2؉ -dependent actin severing activity of free gelsolin appears to be a major determinant of cell migration.

Northern analysis of mRNA, [ 35 S]methionine incorporation into immunoprecipitated gelsolin, immunoblots of cell lysates, and quantitative confocal microscopy showed 1.4 -2-fold higher levels of gelsolin in nonmigrant compared with migrant cells. Because the concentration of intracellular gelsolin did not appear to be a central determinant of cell migration, we assessed its requirement for motility. Cells that were electroinjected with a gelsolin antibody that inhibits actin severing by gelsolin in vitro showed a 72% reduction of the number of migrant cells compared with controls treated with an irrelevant antibody. Cells that were electroinjected with free gelsolin exhibited a 33% increase in migration compared with cells electroinjected with bovine serum albumin. Compared with nonmigrant cells, migrant cells contained abundant free gelsolin and exhibited gelsolindependent F-actin severing activity, which required Ca 2؉ . Serum stimulation of cell migration required increases in [Ca
Studies from a wide variety of organisms have shown that a large number of proteins interact with actin to augment the versatility of actin filament structures and contribute to the dynamics of actin filament rearrangements (1,2). Consequently these actin-binding proteins are likely to play important roles in regulating cell motility, cell differentiation, and cell shape. One such protein that has major effects on actin assembly/disassembly cycles and on actin filament architecture is gelsolin (3), a Ca 2ϩ -and polyphosphoinositide-regulated multifunctional actin-binding protein (4) that severs actin filaments, nucleates actin filament assembly, and blocks the fast exchanging end of actin filaments (5). Gelsolin-actin complexes form transiently during the activation of neutrophils (6), macrophages (7), and platelets (8,9), and the activation process is associated with the shuttling of gelsolin between membranebound and cytoplasmic sites (10). Among the actin-modulating functions exhibited by gelsolin, severing of filaments has the most stringent Ca 2ϩ requirement (11,12). Functional studies indicate that gelsolin has two actin-binding sites (13). In the presence of Ca 2ϩ , a 2:1 actin-gelsolin complex is formed. Chelation of free Ca 2ϩ with EGTA releases actin from one site only, resulting in the formation of an EGTA-resistant 1:1 actingelsolin complex (13), which can be dissociated by phosphatidylinositol 4,5-bisphosphate (4).
There are two types of gelsolin. Both the secreted form (plasma gelsolin) and the cytoplasmic form are derived from a single gene by alternate initiation sites and differential splicing (14). Although the cytoplasmic form almost certainly contributes to the dynamics of actin structure, the plasma form may be important in clearing actin complexes and cell debris from blood vessels (15). The abundant distribution of gelsolin in Xenopus laevis oocytes suggests a role in embryogenesis (16) and gelsolin content increases during differentiation of murine carcinoma cells (17). However, gelsolin expression is also downregulated during chicken erythroid differentiation (18).
Several studies have shown a relationship between gelsolin content and cell migration. For example increasing gelsolin content in cultured mouse fibroblasts by transfection proportionally enhances the rate of cell migration toward a chemoattractant serum gradient (19). Large increases in gelsolin content are also associated with the enhanced motility exhibited during differentiation of myeloid cell lines into macrophagelike cells (20). However, other investigators have shown that high gelsolin content may be associated with reduced cell migration (21,22). Further, the motility of Dictyostellium mutants that lack this severing protein is normal (23), and transgenic mice lacking gelsolin expression exhibit normal motility of embryonic fibroblasts (24). Collectively, these data indicate considerable cell-type dependence of gelsolin in migratory behavior, and the potential functional significance of how gelsolin expression is related to cell migration may be strongly affected by the differentiation state of the cell (17). Thus, the relative abundance of gelsolin may not be the sole factor in determining its relationship to cell motility. Because the functional activities of gelsolin are highly regulated, conceivably it is not the total cellular content but the relative actin severing, nucleating, and binding activities of gelsolin that are responsible for the gelsolin-dependent variations in migration between motile and nonmotile cells.
One alternative approach to study the role of gelsolin in motility is to correlate gelsolin expression in separate subpopulations of motile and nonmotile cells that arise from common precursors. In this context fibroblasts provide a useful model because they constitutively express largely cytoplasmic gelsolin (14), and it is possible to select for differentiated subtypes in culture (25)(26)(27) including nonmotile cells (28). Notably, gingival fibroblasts spontaneously differentiate in culture as measured by filamentous actin content (29) and ␣-smooth muscle actin expression (30). In this study we have separated and characterized motile and nonmotile gingival fibroblasts and have examined the dependence of serum-induced migration on gelsolin expression, severing activity.

EXPERIMENTAL PROCEDURES
Cell Culture and Assay of Chemotaxis-Experiments were conducted using fourth to tenth passage human gingival fibroblasts grown from explants and maintained in cell culture as described (30). To obtain sufficient cell numbers for the various experiments performed in this study, migrant and nonmigrant cells were separated and assayed in a large, customized chemotactic chamber (16 ϫ 16 cm; Mount Sinai Hospital Biomedical Engineering, Toronto, ON). In some experiments migration assays were also performed in conventional 48-well Boyden chambers (Neuroprobe, Cabin John, MD) as described (31,32). Prior to incubation with cells the polycarbonate membranes (8-m pore size; Poretics Corporation, Livermore, CA) were treated with 0.01% acetic acid, washed, and coated with a 0.1% (w/v) collagen solution. Before incubation in chambers, cells were counted electronically (ZM Coulter counter, Hialeah, FL), and 3 ϫ 10 6 cells were added to the large chamber and 5 ϫ 10 4 cells/well for the smaller Boyden chamber. Cells were allowed to attach to filters in normal growth medium for 2 h. Filter-attached fibroblasts were thoroughly washed in phosphate-buffered saline (PBS) 1 at 37°C, and the chamber was assembled. The chemoattractant solution consisted of ␣minimal essential medium containing antibiotics and 20% fetal bovine serum in the bottom chamber, whereas the top chamber contained ␣ minimal essential medium and antibiotics only. The chambers were incubated at 37°C in a humidified atmosphere of 95% air/5% CO 2 . At the end of the assay filters were thoroughly washed with PBS, and cells were collected by scraping into appropriate buffers for each experiment.
To establish the time at which all potentially migrant cells had migrated, timed chemotactic assays were performed in 48-well Boyden chambers. At 2, 6, 8, 16, and 24 h, the number of migrant cells was counted in a fluorescence microscope after staining with 4,6-diamidino-2-phenylindole (1 g/ml; Sigma). Control wells were established without the chemoattractant gradient and contained normal growth medium without serum. To circumvent the dissipation of the chemotactic gradient, which occurs over longer incubation periods (15 h), the chemoattractant and medium in the chambers were replaced every 6 h. Following separation into migrant and nonmigrant cells, we characterized the separate populations by detaching cells from the upper and lower sides of the filters with mild trypsinization and replating in culture dishes containing normal growth medium.
To determine the requirement for increases in [Ca 2ϩ ] i for cell migration in the chemotactic assay, cells were loaded with or without (control) the intracellular calcium chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,NЈ,NЈ-tetraacetic acid tetra(acetoxymethyl)-ester (BAPTA/AM; 3 M) in nominally calcium-free medium. The loaded cells were continuously exposed to BAPTA/AM during the assay, which was performed in a Boyden chamber at various time points as described above. Cell viability was assessed by staining with 0.4% trypan blue.
Gelsolin Fractionation and Immunoblotting-To assess total gelsolin, cells were rinsed twice with PBS and were then lysed in an extraction buffer (19) plus protease inhibitors (1 mM benzamidine, 1 mM leupeptin, and 1 mM aprotonin). Triton-soluble supernatants and triton-insoluble cytoskeletons (pellets) were fractionated by centrifugation at 14,000 ϫ g for 2 min at 4°C. Lysates were boiled in solubilizing buffer (10% SDS, 20% glycerol, 2% mercaptoethanol, Tris-HCl, pH 6.8) for 10 min, and equal amounts of each fraction were then separated by electrophoresis on 7.5% SDS-polyacrylamide gels (Phast system; Pharmacia Biotech Inc.). Proteins on the gel were transferred to nitrocellulose paper (Schleicher & Schull), and gelsolin was identified by immunoblotting with an anti-gelsolin monoclonal antibody (clone 2C4, Sigma) and with horseradish peroxidase-labeled rabbit anti-mouse IgG as a second antibody. To adjust for variations in gel loading and to provide a standard for quantification, blots were also probed for ␤-actin (Clone AC-15, Sigma). The blots were developed by a chemiluminescence method (ECL, Amersham Corp.), autoradiographed, and scanned for quantification of intensity.
We examined whether the motility difference between the subpopulations was due to differences in the availability of free gelsolin and the complexing of gelsolin with actin. We quantified gelsolin that was associated with actin in the ultracentrifuged pellet and free gelsolin in the supernatant. Migratory and nonmigratory cells were collected from the polycarbonate filters and suspended in a nondetergent buffer (1 mM EGTA, 2 mM Tris, pH 7.4, 0.2 mM ATP, 0.2 mM MgCl 2 , 0.2 mM dithiothreitol, 2 mM phenylmethylsulfonyl fluoride) containing 1 M phalloidin. Phalloidin at this concentration stabilizes F-actin. The freshly prepared cell lysate was pipetted several times with a narrow gauge pipette tip and ultracentrifuged at 100,000 ϫ g for 1 h at 4°C (33). The supernatant and pellet were collected, separated by SDS-polyacrylamide gel electrophoresis, and immunoblotted for detection of actin and gelsolin. The pellet contained gelsolin complexed with F-actin and was suspended in 0.1 M KCl, 1 mM MgCl 2 , 0.1 mM CaCl 2 , 0.2 mM ATP, 1 mM NaHCO 3 , and 1 mM NaNa 3 . Protein concentrations in the fractions were equalized. In separate experiments designed to assess the effect of serum stimulation on availability of free gelsolin, migrant and nonmigrant cells were separated, plated (2 each), and serum-deprived for 24 h. One plate from each group was incubated with 10% FBS for 20 min. Fresh lysates were collected from all four plates in a Triton buffer (19) containing 1 mM EGTA and ultracentrifuged at 100,000 ϫ g for 1 h at 4°C (33). The supernatants were adjusted for equal protein loading, separated by SDS-polyacrylamide gel electrophoresis, blotted, and probed for actin and gelsolin.
Severing Assays-Gelsolin severing activity was performed essentially as described (35,36). Affinity purified rabbit muscle actin (1.0 mg/ml; Cytoskeleton, Denver, CO) was resuspended in polymerization buffer (50 mM KCl, 2 mM MgCl 2 , 0.5 mM ATP, 2 mM Tris, pH 8.0, 0.5 mM dithiothreitol) and labeled with N-(1-pyrenyl)iodoacetamide (Molecular Probes, Eugene, OR) dissolved in N,N-dimethylformamide. The molar ratio of the dye to actin was 1:1. The reaction was allowed to proceed in the dark overnight. Unreacted dye was removed by incubation with SM-2 bio-beads (Bio-Rad). Because not all of the labeled actin could participate in polymerization/depolymerization cycles, it was removed during repeated cycles of polymerization and depolymerization (37). Severing assays were performed on the dialyzed cell lysates obtained from the migratory and nonmigratory samples by fluorescence spectrophotometry (excitation ϭ 365 nm; emission ϭ 386 nm; slit width ϭ 3 nm). Cell lysates prepared with detergent as described above (Ref. 19 plus protease inhibitors) were dialyzed with several changes of buffer containing 2 mM MgCl 2 , 50 mM KCl, 2 mM Tris-HCl, and 1 mM EGTA, 0.5 mM ␤-mercaptoethanol, with 1 mM NaN 3 . The volume of the dialyzed cell lysate was adjusted to 200 l in dialysis buffer; 100 l of labeled F-actin in polymerizing buffer (50 mM KCl, 2 mM MgCl 2 ) was added to a final concentration of 200 nM. Severing assay were performed in the presence of calcium (final concentration, 2 mM CaCl 2 /1 mM EGTA).
In separate experiments designed to assess the in vitro motility block by electroinjected gelsolin (2C4) antibody, severing assays were performed. Increasing concentrations of antibody were titrated to assess the concentration of antibody that would block severing. Gelsolin (1 M; purified as described in Ref. 34) was incubated with antibody prior to incubation with pyrene-labeled actin (2 M) before the severing assays were conducted. Severing activity was also assessed with an irrelevant antibody (pan-cytokeratin; clone PCK-26; Sigma).
Single cell suspensions were prepared for flow cytometry analysis by trypsinization of cells from the filters after the termination of the migration assay. Cells were fixed in 3% formaldehyde with PBS and 0.5% dimethyl sulfoxide for 15 min and permeabilized in 0.1% Triton X-100. To obtain accurate estimates of fluorescent intensity that were not confounded by inclusion of gelsolin-containing serum, cells were blocked with 3% BSA to avoid spurious introduction of gelsolin into the cells that is associated with a serum block. Cells were stained for gelsolin (2C4), washed, and resuspended in Ca 2ϩ -and Mg 2ϩ -free PBS. Samples were analyzed on a FACSTAR Plus flow cytometer (Becton Dickinson, Mississauga, ON) at 200 cells/s with 488 nm laser excitation.
Electroinjection-To optimize the method by which the anti-gelsolin antibody inhibits the intracellular activity of gelsolin, in vitro studies were performed first to determine the amount of antibody that would complex known amounts of gelsolin. From estimates of cell number and the amount of protein in the lysates, we determined the optimal concentration of purified gelsolin for experiments (1 M; i.e. 83 g/ml). Gelsolin was incubated with various dilutions of gelsolin antibody (mouse monoclonal anti-gelsolin antibody; 2C4; in electroporation buffer). The gelsolin-antibody complexes were removed with protein A-Sepharose. The residual, free gelsolin was detected by immunoblotting, and the optimal concentration of antibody for complete complexing was calculated (100 g/ml). For electroporation and severing assays an equivalent molar amount of antibody was used.
To directly assess the requirement for gelsolin in cell migration in the chemotactic assay, the gelsolin antibody or an irrelevant antibody (pancytokeratin antibody, clone PCK-26; diluted in electroporation buffer) was electroinjected into cells by electroporation as described (41). Electroporation was performed with a Bio-Rad gene pulser, a capacitance extender, and sterile cuvettes. An 800-l aliquot of 2 ϫ 10 6 cells/ml was placed into an electroporation cuvette (0.4-cm interelectrode distance), and the cells were electroporated in a buffer containing 5.37 mM KCl, 0.52 mM KH 2 PO 4 , 0.64 mM MgCl 2 , 0.63 mM MgSO 4 , 85.5 mM NaCl, 5.8 mM NaHCO 3 , 0.50 mM NaH 2 PO 4 , and 12.5 mM HEPES at a field strength of 250 V/cm and capacitance of 500 microfarads. The cells were collected by centrifugation and suspended in growth medium without serum, and 5 ϫ 10 4 cells/well were assayed in the Boyden chamber with a chemotactic gradient for various time periods. The filter-attached cells were fixed and stained with 4,6-diamidino-2-phenylindole before counting. Previous detailed assessments of this protocol have demonstrated that 10 8 antibody molecules/cell can be introduced into viable cells (41) and that electroinjected antibodies can specifically bind and inactivate specific functional proteins (40,42).
[Ca 2ϩ ] i Measurement-To examine the role of intracellular calcium fluxes in chemotactically induced cell migration, attached, serumstarved cells were loaded with fura2/acetoxymethyl ester (3 M) and were stimulated with 20% FBS. [Ca 2ϩ ] i was estimated as described (43). In separate experiments, cells were co-loaded with fura2/AM and BAPTA/AM (3 M) in calcium-free buffer, and the [Ca 2ϩ ] i was determined as described. Cell viability was assessed by adding Ca 2ϩ and 1 M ionomycin.
Northern Analyses-Total cellular RNA was prepared from the migratory and nonmigratory cells and quantified by UV spectroscopy. Prehybridization and hybridization were at 60°C, and washing precedures were at 55°C as described (44). Three different customized 27mer oligonucleotide probes for gelsolin based on the published human gelsolin cDNA sequence (45) and one 29-mer oligo from the human ␤-actin cDNA sequence (46) were designed with the OLIGO program to minimize self-hybridization, synthesized (General Diagnostics Systems, Toronto, ON), and purified by high pressure liquid chromatography. Oligos were 5Ј end-labeled with [␥-32 P]ATP (Ͼ3000 Ci/mmol, Du-Pont NEN) by T4 Polynucleotide Kinase (Pharmacia). The labeled probes were purified with NICK columns (Pharmacia), and 1 l of labeled probe was used in the scintillation counter to determine the total 32 P incorporation before hybridization. The blots were sequentially probed for gelsolin and actin as described (46). After hybridization, the blots were exposed to Kodak X-Omat film at Ϫ70°C using intensifying screens. The radioactive bands were quantified by densitometry and normalized to the hybridization obtained for the ␤-actin.
Biosynthetic Labeling and Immunoprecipitation-After the chemotactic assay was performed and cells had been separated on the filter membrane, the medium was replaced with serum-free and methioninedeficient medium containing 100 Ci/ml of [ 35 S]methionine (specific activity, 1000 Ci/mmol; DuPont NEN) for 1 h. Cells were washed three times with ice-cold PBS and solubilized with Triton-containing lysis buffer (6,7), and protein contents were equalized for immunoprecipitation by gelsolin antibody bound to Sepharose beads as described (7). After separation, proteins were detected by fluorography. Biosynthetic activity and protein degradation were determined by scanning densitometry of the autoradiographs of immunoprecipitated gelsolin. Total [ 35 S]methionine incorporation into cytoplasmic proteins was determined by trichloroacetic acid precipitation. The amount of radioactivity was measured by scintillation counting.
Statistical Analysis-For all assays, three or more separate experiments were performed; the means Ϯ S.E. were calculated for continuous variables, and comparisons were made by unpaired t tests.

RESULTS
Chemotactic Response-Boyden chambers were used to determine the kinetics of the chemotactic migration of the fibroblasts to a concentration gradient of FBS. Migration of cells continued for up to 16 h (Fig. 1, top panel) when there was 0% FBS in the upper and 20% FBS in the lower chamber, indicating that migration occurs in response to a concentration gradient of FBS. The maximum migratory response to FBS was 202 Ϯ 25 cells/microscope field on the lower surface of the filter and the background migration in the absence of attractant was 20 Ϯ 5 cells/field. After 16 h there was no further migration of the cells onto the lower filter in spite of re-establishment of the chemotactic gradient every 6 h. On the basis of these findings the two cell types collected from the lower and upper sides of the assay filters were designated as migratory and nonmigratory, respectively. We also counted the number of cells on the upper surface of the filter (i.e. nonmigrant cells) and found that the number of cells decreased between 0 -24 h, indicating that cells had indeed migrated on to the lower surface of the filter. Total cell number on the upper and lower sides of the filters was not detectably different, indicating that cell proliferation over this time interval was minimal. Estimation of cell size by flow cytometry showed that the forward scatter of migrant and nonmigrant cells was not significantly different, indicating that the lack of migration of the nonmigrant cells was not simply due to their potentially greater size blocking the crawling of cells through the pores of the filter (nonmigrant, 322 Ϯ 6 fluorescence units; migrant, 313 Ϯ 7 units; p Ͼ 0.2).
We assessed whether the migratory cells required increases of intracellular calcium. Cells were serum-starved and loaded with BAPTA/AM or not. Unloaded control cells that were stimulated with 20% serum showed whole cell [Ca 2ϩ ] i transients, whereas [Ca 2ϩ ] i reduction with BAPTA abrogated calcium transients (Fig. 1, middle panel). We determined the dependence of [Ca 2ϩ ] i increases for motility by measuring the migration of BAPTA/AM loaded cells in the migration assay. There was a 76 Ϯ 8% reduction of the number of migrant cells compared with controls. The viability of the cells in the Boyden chamber after BAPTA/AM treatment as determined by trypan blue was 98 Ϯ 0.5%. After incubation of cells in calcium-containing buffer, increases of [Ca 2ϩ ] i in response to ionomycin confirmed that the cells were indeed viable (Fig. 1, bottom  panel).
Subpopulations-Immunoblots of lysates prepared from cells that had been freshly scraped from filters and adjusted for cell number showed 1.8-fold more gelsolin in nonmigrant cells ( Fig. 2A). Northern analyses were performed in which the same amount of total RNA from the two cellular subpopulations was probed for gelsolin and actin (Fig. 2B). Gelsolin mRNA, adjusted to equal actin mRNA content, was 1.4 times more abundant in nonmigratory than in migratory cells. Actin expression was not detectably different between the migrant and nonmigrant cells (Ͻ5% variation by densitometry). [ 35 S]Methionine incorporation into gelsolin was determined by densitometry of autoradiographs of immunoprecipitated gelsolin at the end of a 1-h pulse. There was approximately 1.4 times more radioactivity in the cell-associated gelsolin of the nonmigratory cells compared with the migratory cells, suggesting that there was more synthesis of gelsolin in the nonmigrant cells (Fig. 2C). Because the reduced gelsolin content and synthesis may have been a reflection of a global reduction of protein synthesis in migrant cells, we examined [ 35 S]methionine incorporation into cellular proteins by trichloroacetic acid precipitation. These data showed ϳ50% higher incorporation in migratory compared with nonmigratory cells (nonmigratory, 4.04 ϫ 10 6 Ϯ 448 cpm/mg; migratory, 5.95 ϫ 10 6 Ϯ 783 cpm/mg; n ϭ 3 for each cell type), indicating that there was more synthesis of cellular proteins in the migratory cells. We localized gelsolin, G-actin, and F-actin in the two subpopulations separated by the chemotactic assay. Filters were cut into small pieces and stained so that the migratory and nonmigratory fibroblasts on either side of the same filter could be analyzed. This procedure reduced variations in staining intensity between different filters. Quantitative analysis of fluorescence intensity by confocal microscopy showed that gelsolin staining was more intense in nonmigratory cells (90 Ϯ 4 fluorescence units; n ϭ 5) than in migratory cells (44 Ϯ 8; p Ͻ 0.01; n ϭ 5). G-actin (by rhodamine DNase staining) was distributed throughout the cytoplasm, in the cell periphery, and around the nucleus, but the intensity of the staining in the two cell types was not significantly different (nonmigratory cells, 67 Ϯ 6; migratory cells, 77 Ϯ 7; p Ͼ 0.2). In preparations of double labeled nonmigrant cells, staining for F-actin by tetramethylrhodamine isothiocyanate-phalloidin colocalized with gelsolin in stress fibers (Fig. 3, A and B), whereas migrant cells showed no obvious colocalization. F-actin in nonmigrant cells was more intense than the migrant cells (nonmigrant cells, 81 Ϯ 4; migrant cells, 65 Ϯ 5; p Ͻ 0.05).
We assessed whether the migrant and nonmigrant subpopulations may exhibit variations of cytoskeletal protein expression by staining with a panel of antibodies (Fig. 4, A-H). Both the migrant and nonmigrant cells stained for vimentin as shown earlier (data not shown; Ref. 43) but did not stain for desmin or cytokeratin, confirming that neither phenotype were epithelial or smooth muscle cells (Fig. 4, A-D). We measured the ␣-smooth muscle actin (SMA) content because this protein is a differentiation marker for contractile fibroblast subtypes (30,38). ␣-SMA was localized in the cytoplasm and stress fibers. Nonmigrant cells were more intensely stained for ␣-SMA than migrant cells (nonmigrant, 66 Ϯ 1; migrant, 39 Ϯ 4; p Ͻ 0.01), indicating that the nonmigrant cells were indeed more differentiated than the migrant cells (Fig. 4, E and F). Staining with nonmuscle myosin was localized to the leading edge of cells in both subpopulations and appeared identical (Fig. 4, G and H).
In separate experiments, the two cell types collected from either side of the filters were isolated and grown in cell culture. The cultured, migrant, and nonmigrant cell populations exhibited essentially the same characteristics as they exhibited on the filters. The relative gelsolin content per cell in whole population cell suspensions was estimated by flow cytometry (n ϭ 10,000 cells/sample). The nonmigrant cells exhibited higher mean fluorescent channel numbers (110 Ϯ 9) than the migrant cells (45 Ϯ 4; p Ͻ 0.01). Similar patterns of increased gelsolin content in nonmigratory cells were observed in lysates of cells passaged once in culture and immunoblotted. The lysates exhibited ϳ1.8-fold more gelsolin in the nonmigrant cells. Collectively these data on protein content and immunofluorescence staining intensity of the separated subpopulations revealed similar patterns between cells on filters and cells grown from filters, indicating the stability of the separated subpopulations. The functional stability of the separated subpopulations was also confirmed by subjecting them to chemotactic assays, which showed that only 15% of the passaged nonmigrant cells responded to a second application of the chemotactic serum gradient over a 24-h assay, whereas virtually all of the migrant cells (Ͼ95%) exhibited migration. The stability of these subpopulations was preserved for two or three passages, after

FIG. 4. Fluorescence micrographs of nonmigrant (A and B) and migrant (C and D) cells showing negative staining for keratin (A and C) and desmin (B and D). ␣-SMA staining is stronger in nonmigrant cells (E) compared with migrant cells (F). Staining for nonmuscle myosin in the submembrane region of nonmigrant (G) and migrant cells (H) is indistinguishable.
which the differences of actin, ␣-smooth muscle actin, and gelsolin contents between migrant and nonmigrant cells dissipated and were not detectably different.
Requirement for Gelsolin in Cell Migration-We directly assessed the requirement of gelsolin for cell migration 6 h after the initiation of the Boyden chamber assay. Cells electroporated in the presence of gelsolin antibody showed a 72 Ϯ 7% reduction of the number of migrant cells compared with control cells electroporated with an irrelevant antibody (controls, 86 Ϯ 10 cells/well; gelsolin electroporation, 24 Ϯ 13 cells/well; 6-h chemotactic gradient), indicating that gelsolin is required for cell migration. We confirmed that the electroporated cells contained antibody as shown in cells that were stained with an fluorescein isothiocyanate-conjugated anti-mouse antibody. Because these experiments indicated that gelsolin was required for migration, we determined if electroinjection of excess free gelsolin would induce increased migration. Cells were electroinjected with either 1 M gelsolin or with 1 M BSA, incubated in the Boyden chamber for 6 h, and assayed as described above. Electroinjected gelsolin significantly increased cell migration compared with the BSA (gelsolin, 72 Ϯ 8 cells/field; BSA, 54 Ϯ 6 cells/field; p Ͻ 0.05).
The distribution of gelsolin antibody in nonmigrant cells showed a variety of staining patterns (Fig. 5) that was in sharp contrast to nonmigrant cells stained for endogenous gelsolin content without electroporation (Fig. 3B). Gelsolin antibody was distributed either throughout the cytoplasm (Fig. 5A), at the leading edge (Fig. 5B), or rarely in fibrillar structures (Fig.  5C). The staining of the few migrant cells that had migrated in spite of electroporated antibody showed a barely detectable reaction (Fig. 5D), indicating that these cells were poorly injected with antibody. Cells that had been electroinjected with an irrelevant control antibody (cytokeratin) showed diffuse staining similar to Fig. 5A.
We determined whether the basis of the motility block by the electoporated gelsolin antibody was due to complexing and immobilization of the gelsolin and consequent inhibition of severing activity within the cell. In vitro severing assays were performed in which previously determined optimal concentrations of gelsolin antibody (100 g/ml) were used to complex gelsolin. As expected, purified gelsolin showed robust severing activity (Fig. 5E), whereas the gelsolin antibody by itself showed no severing activity (Fig. 5F). The gelsolin-antibody complex exhibited slight but detectable severing activity, presumably due to the inability of the antibody to completely prevent interaction of the severing domain of gelsolin with actin. Incubation with an irrelevant antibody (pan cytokeratin) showed no inhibition of severing (not shown).
Comparison of Gelsolin Fractions-We quantified the total gelsolin in the two subpopulations and assessed the partitioning of gelsolin into Triton-soluble and Triton-insoluble extracts by immunoblotting and densitometry (47) of cells lysed in 1 mM EGTA. Using low speed centrifugation of Triton lysates, a large proportion (97 Ϯ 4%; n ϭ 3) of the total cellular gelsolin was solubilized by the detergent, whereas only a very small portion was found in the insoluble fraction. In contrast, a large proportion (73 Ϯ 4%) of the cellular actin was in the Triton-insoluble fraction and 25 Ϯ 7% was in the soluble fraction. Immunoblots of cells freshly prepared from filters and adjusted for equal cell number showed 1.9-fold more gelsolin in the supernatants from nonmigrant cells.
Because immunostaining showed that a large proportion of gelsolin in nonmigrant cells was localized to stress fibers, we determined whether gelsolin co-sedimented with F-actin in both migrant and nonmigrant cells. Buffer conditions (i.e. with phalloidin and EGTA but no Triton) were used to maintain the stability of the actin filaments and the gelsolin-actin complexes. Ultracentrifugation was used to obtain an approximate separation of free gelsolin and gelsolin that was bound to filaments. There was more sedimentable gelsolin in nonmigrant cells compared with migrant cells, whereas free gelsolin was more abundant in the migrant cells compared with the nonmigrant cells (Fig. 6).
Surface ruffling of motile cells is thought to reflect an underlying mechanism that is also involved in locomotion (48). We induced membrane ruffling by incubating serum-starved (24 h) migrant and nonmigrant cells with 10% FBS. Ultracentrifuged supernatants prepared from lysates were probed for gelsolin and actin. Actin was barely detectable without serum, but only migrant cells showed increased amounts of actin in response to serum stimulation (Fig. 7).
Severing Activity: Roles of Calcium and Gelsolin-Fluorimetric assays were used to evaluate the actin filament-severing activity of cell lysates that were obtained in a Triton buffer and then dialyzed. The proportions of gelsolin-actin complex and of free gelsolin were similar to those shown in Fig. 6. Migratory cells showed significant severing activity, whereas severing activity in nonmigratory cells was barely detectable (Fig. 8A), a finding that was consistent with the low levels of free gelsolin shown in Fig. 6. Severing assays performed in the absence of Ca 2ϩ showed no detectable severing activity in either cell type (Fig. 8B). Because the severing activity in the cell lysates could be attributed to proteins other than gelsolin, we immunodepleted the lysates from migrant cells by incubation with antigelsolin antibody coupled to Sepharose beads. Treatment of lysates with this protocol completely eliminated severing activity compared with controls (Fig. 8C), demonstrating that the severing activity recovered from cell lysates might be due to gelsolin activity. DISCUSSION In this study we assessed the dependence of fibroblast migration on gelsolin expression and severity activity. Cell migration was clearly dependent on gelsolin as shown by the 3-fold reduction of migrant cell number after electroinjection of gelsolin antibody. In vitro assays showed that this antibody directly inhibited actin severing. Our immunofluorescence localization data indicated that the electroporated antibody was efficiently injected into cells at high concentrations. Compared with control cells containing endogenous gelsolin (i.e. no electroporation), the gelsolin antibody disrupted the normal spatial location of gelsolin and presumably prevented gelsolin from moving into cellular sites required for its function (10). Taken together with the BAPTA experiments showing inhibition of cell migration, it appears that the calcium-dependent actin severing activity of gelsolin was required for cell motility in response to a chemotactic gradient.
Previous studies have used cell transfection to generate cell lines expressing different levels of gelsolin (19). Notably, cell transfection experiments in which forced expression of gelsolin is used to study the role of cytoskeletal and actin-binding proteins are complicated by the instability of expression levels (49) and the disrupted growth potential of transfectants. In contrast, we have used a functional assay to separate migrant and nonmigrant cells from a common, mixed population of fibroblasts. The detection of cells with widely different migratory behavior is not surprising in view of earlier data demonstrating distinctive migratory subpopulations through collagen gels in skin fibroblasts (28) and in subpopulations of gingival fibroblasts (31,50). We determined whether constitutive differences of cell migration may be reflected in gelsolin expression and functional activities.
Gelsolin Expression-Comparison of gelsolin expression in the migrant and nonmigrant cells showed higher mRNA levels and higher [ 35 S]methionine incorporation into immunoprecipitated gelsolin in the nonmigrant cells. Thus there was abundant gelsolin mRNA and translated product in nonmigrant cells. Similar observations of high gelsolin content at the protein and mRNA levels in quiescent differentiated arterial smooth muscle cells compared with proliferative cells have been reported (21) and in revertant cells with an ordered growth pattern (51). Notably, previous reports have indicated FIG. 7. Migrant and nonmigrant cells were plated, serumstarved, treated with 10% serum, and extensively washed, and lysates were freshly prepared. Lysates were ultracentrifuged. Equal amounts of protein from the supernatants were run on SDS-polyacrylamide gel electrophoresis gels and immunoblotted for actin and gelsolin. Actin is barely detectable in cells without serum despite the relatively large amount of cell protein loaded onto each lane. Serum-treated migrant cells show increased amounts of actin in the supernatants. FIG. 6. Fresh cell lysates stabilized with phalloidin in nondetergent buffer were prepared from migrant and nonmigrant cells. Ultracentrifugation was used to prepare pellets (P) and supernatants (S). In nonmigrant cells there was more gelsolin in the pellets, presumably complexed to F-actin. With the amount of cell protein loaded on the gel, we were unable to detect actin in the supernatants. that migratory fibroblasts are relatively undifferentiated and exhibit characteristics of fetal cells (28,50). Taken together with previous data on gelsolin content and differentiation in embryonal carcinoma cells (17), it is likely that high levels of gelsolin and ␣-SMA expression in gingival fibroblasts are indicative of a more differentiated, nonmigrant cell type. Conceivably, these cells may be involved in remodelling of matrix by traction (27,43) and by collagen phagocytosis.
Gelsolin Localization-Staining of gelsolin and F-actin showed their co-distribution in stress fibers of nonmigrant cells, whereas migrant cells exhibited diffuse staining of gelsolin in the cytoplasm. Similar techniques have shown that gelsolin was distributed along the stress fibers in human lung fibroblasts using phalloidin and anti-gelsolin antibody (52) and that gelsolin molecules visualized by colloidal gold are associated with actin filaments in platelets and macrophages (10). However, gelsolin localization by immunofluorescence probably does not detect all intracellular gelsolin because free gelsolin may be inadequately fixed and thereby lost on permeabilization (54,55). Because the two subpopulations in our system were treated identically for immunostaining, we believe that the immunohistochemical methods provided comparable results in terms of actin-bound gelsolin. Notably, quantification of gelsolin staining by confocal microscopy was more intense in nonmigrant than migrant cells, particularly in the vicinity of stress fibers. This morphological finding is consistent with the higher density of gelsolin bands in immunoblots of ultracentrifuged pellets of nonmigrant cells.
Gelsolin Activity-We have shown that cell migration and actin severing are dependent on Ca 2ϩ fluxes in vivo, consistent with the accepted functional regulation of gelsolin in vitro (56). In this study we have used a combination of powerful in vivo techniques to provide strong evidence that despite abundant gelsolin in the stress fibers of nonmigrant cells, the gelsolin cannot sever actin. The apparently preferential localization of gelsolin to actin stress fibers, its inability to sever actin, and its high levels in nonmigrant cells suggest diverse functions of gelsolin. In this context previous studies have shown that gelsolin associates with actin-myosin complexes and may participate in the regulation of cell contraction (57)(58)(59) and myofibrillar assembly (60). It has also been suggested that the presence of nebulin in sarcomeric thin filaments may confer resistance to actin severing by gelsolin (61). Because the nonmigrant gingival fibroblasts examined here have prominent stress fibers that are enriched in ␣-smooth muscle actin and actively remodel extracellular matrices in vivo (43), we speculate that the gelsolin in these cells is sequestered in the stress fibers specifically for local remodelling of these structures.
In addition to calcium regulation of gelsolin activity, polyphosphoinositides are also important because they can dissociate gelsolin from actin (62). We found that electroinjection of exogenous gelsolin increased cell migration by increasing the availability of free gelsolin in nonmigrant cells. Conceivably, the blocked gelsolin in stress fibers does not have access to phosphatidylinositol 4,5-bisphosphate for dissociation. Thus the local regulation of gelsolin by polyphosphoinositide and phospholipase C binding (53) may be another factor in determining actin severing in stress fibers.