Integrin β4 Regulates Migratory Behavior of Keratinocytes by Determining Laminin-332 Organization*

Whether α6β4 integrin regulates migration remains controversial. β4 integrin-deficient (JEB) keratinocytes display aberrant migration in that they move in circles, a behavior that mirrors the circular arrays of laminin (LM)-332 in their matrix. In contrast, wild-type keratinocytes and JEB keratinocytes, induced to express β4 integrin, assemble laminin-332 in linear tracks over which they migrate. Moreover, laminin-332-dependent migration of JEB keratinocytes along linear tracks is restored when cells are plated on wild-type keratinocyte matrix, whereas wild-type keratinocytes show rotation over circular arrays of laminn-332 in JEB keratinocyte matrix. The activities of Rac1 and the actin cytoskeleton-severing protein cofilin are low in JEB keratinocytes compared with wild-type cells but are rescued following expression of wild-type β4 integrin in JEB cells. Additionally, in wild-type keratinocytes Rac1 is complexed with α6β4 integrin. Moreover, Rac1 or cofilin inactivation induces wild-type keratinocytes to move in circles over rings of laminin-332 in their matrix. Together these data indicate that laminin-332 matrix organization is determined by the α6β4 integrin/actin cytoskeleton via Rac1/cofilin signaling. Furthermore, our results imply that the organizational state of laminin-332 is a key determinant of the motility behavior of keratinocytes, an essential element of skin wound healing and the successful invasion of epidermal-derived tumor cells.


EXPERIMENTAL PROCEDURES
Cell Culture, Drug Treatments, Matrix Preparation, SDS-PAGE, Immunoblotting, and Immunoprecipitation-Human epidermal (WT) keratinocytes were a gift from Dr. Lou Laimins of Northwestern University. The ␤4-deficient keratinocytes, derived from a JEB patient, were described previously (Russell et al. (36)). Both lines were immortalized in identical fashion using human papilloma virus E6 and E7 genes (39). They were maintained at 37°C in defined keratinocyte serum-free medium (Invitrogen). The Rac1 inhibitor NSC23766 was obtained from the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, NCI, National Institutes of Health. It was added to keratinocyte growth medium at 50 M 24 h prior to cell analysis. Brefeldin A (Sigma-Aldrich Co.) was added to growth medium at a concentration of 50 M.
Keratinocyte matrix was prepared as described previously (40). For immunoprecipitations, keratinocytes were extracted in 1% Brij 97, 25 mM Hepes, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 supplemented with a protease inhibitor mixture (Sigma-Aldrich Co.). The extract was clarified by centrifugation, and either primary antibody or control IgG was added. After shaking for 4 h at 4°C, rProtein G-agarose beads (Invitrogen) were added to the mix, which was incubated for 90 min at 4°C. The beads were washed, collected by centrifugation, and then incubated in SDS-PAGE sample buffer.
Protein preparations were processed for SDS-PAGE and immunoblotting as detailed elsewhere (30). Immunoblots were scanned and quantified using a MetaMorph Imaging System (Universal Imaging Corp., Molecular Devices, Downingtown, PA).
Retroviral and Adenoviral Constructs and Infection of Keratinocytes-The wild-type ␤4 integrin cDNA or a mutant, laminin-332 binding-deficient ␤4 integrin bearing a mutation in the specificity determining loop (K150A) were described previously (41). Both were fused with sequences encoding GFP at their 3Ј ends, cloned into the LNCX2 vector (Clontech, Palo Alto, CA), and sequenced using ABI Big Dye chemistry (Applied Biosystems, Foster City, CA). The subsequent constructs were introduced into PT67 packaging cells using calcium phosphate-mediated transfection. Briefly, 10 g of DNA was diluted in 440 l of sterile water, mixed with 500 l of 2ϫ HBS (280 mM NaCl, 10 mM KCl, 1.5 mM Na 2 HPO 4 , 12 mM dextrose, 50 mM HEPES, pH 7.05) and 60 l 2 M CaCl 2 , and added dropwise to cells in medium containing 25 M chloroquine. Following overnight incubation, the medium was removed, cells were rinsed in phosphate-buffered saline, and fresh medium was added. The following day, cells were trypsinized and replated into medium (without penicillin/ streptomycin) supplemented with G418 (Invitrogen). Surviving cells were subsequently cloned to produce stable cell lines. Four days prior to infecting keratinocytes, PT67 cell lines expressing the ␤4 integrin construct were transfected with a plasmid encoding the vesicular stomatitis virus glycoprotein using the protocol described above. The following day, the medium was replaced, and the cells were switched to 32°C for 2 days. The supernatant was harvested and filtered, and 8 g/ml Polybrene was added. Keratinocytes were incubated in the viral supernatant in plates which were centrifuged at 2000 rpm for 2 h at room temperature (42). After removal of viral supernatant, cells were subcultured for subsequent analyses.
Keratinocytes were infected with adenoviruses containing the cytomegalovirus promoter (courtesy of Dr. Kathy Rundell, Northwestern University), the N19 DN mutant of RhoA, the N17 DN mutant of Rac1, the N17 dominant-negative mutant of Cdc42, the V12 CA Rac1 mutant (all c-Myc epitope-tagged, courtesy of Dr. Anne Ridley of University College, London, UK), XAC1 GFP S3E (inactive, phospho-mimetic) cofilin, XAC1 GFP S3A (active, non-phosphorylatable) cofilin (generously provided by Dr. James Bamburg of the University of Colorado at Boulder, CO), and the YFP-tagged ␤3 laminin subunit at a multiplicity of infection of 1:50 in keratinocyte medium (43). To generate virus encoding YFP-tagged full-length ␤3 laminin, ␤3 laminin cDNA was cloned into the pEYFP-N vector (BD Biosciences Clontech, Palo Alto, CA) at the BamHI/XhoI sites in the polylinker 5Ј to the start of sequences encoding the YFP tag. This produces a fusion protein consisting of the ␤3 laminin subunit, a spacer of 21 amino acid residues (RAQASNSAVDGTAGPGSPVAT), and YFP. This entire cassette was subcloned into the pENTR-4 vector (Invitrogen). This entry vector was used in a site-directed recombination reaction to place the cassette into the pAD/CMV.V-5-DEST vector following the manufacturer's directions (Invitrogen). The adenoviral vector was linearized with PacI and transfected into 293A cells using Lipofectamine 2000 (Invitrogen). Approximately 12 days post transfection, the adenovirus-containing 293A cells were harvested and lysed to prepare a crude viral stock. The resultant viral stock was amplified and titered. Keratinocytes were incubated in virus at 37°C overnight, and the virus-containing medium was removed. The cells were then fed with fresh medium, followed by incubation at 37°C. After 18 h, cells were trypsinized and processed for motility assays. To confirm expression of the above constructs, immunoblotting of cell extracts was performed using GFP antibodies or antibody 9E10 against the c-Myc epitope, and/or cultures were viewed by fluorescence microscopy to identify cells expressing GFP-or YFPtagged protein (not shown). Greater than 80% infection rates were obtained in each of our studies.
For flow cytometry, freshly trypsinized cells were resuspended in phosphate-buffered saline containing a 50% dilution of normal goat serum, incubated with monoclonal antibodies against ␣3, ␤4, or ␣6 integrin for 45 min at room temperature, washed with phosphate-buffered saline, incubated with Cy5conjugated goat anti-mouse IgG for 45 min, washed, resuspended in phosphate-buffered saline, and analyzed using a Beckman Coulter Elite PCS sorter (Beckman Coulter). For negative controls, primary antibody was omitted from the above procedure.
Immunofluorescence Microscopy-Cells on glass coverslips were processed as detailed elsewhere (30). All preparations were viewed with a confocal laserscanning microscope (LSM 510, Zeiss Inc., Thornwood, NY). Images were exported as TIFF files, and figures were generated using Adobe Photoshop software.
Observations of Live Cells, Motility Assays, and TIRF-Scratch assays of confluent cell monolayers were performed in triplicate as described elsewhere (3). For studies on the motility of individual and groups of cells, cells were plated onto uncoated or matrixcoated 35 mm glass-bottom culture dishes (MatTek Corp., Ashland, MA). For function-blocking assays, fresh medium containing 20 g/ml antibody or immunoglobulin was added 15 min prior to visualization. Cells were viewed on a Nikon TE2000 inverted microscope (Nikon Inc., Melville, NY) or the Zeiss LSM 510 confocal microscope. Images were taken at 2-to 5-min intervals over a 2-h period, and cell motility was analyzed by MetaMorph (Molecular Devices Corp.). Motility assays were performed a minimum of three times with at least 50 cells being counted in each trial. Wide-field and TIRF observations of live cells were performed on a TIRFM microscope (Olympus America Corp., Melville, NY).

Expression of ␤4 Integrin in JEB and WT Keratinocytes-We
have used keratinocytes derived from a patient with junctional epidermolysis bullosa-pyloric atresia (JEB), a blistering skin disease whose pathogenesis results from an absence of ␤4 integrin expression (45). Fluorescence-activated cell sorting analyses confirmed that these JEB cells are deficient in ␤4 integrin (Fig.  1A), but express ␣3 and ␣6 integrin at levels comparable to those of wild-type (WT) cells (Fig. 1A). The JEB keratinocytes were infected with retrovirus encoding either GFP-tagged wildtype ␤4 integrin or ␤4 integrin bearing a mutation (K150A) that inhibits ␣6␤4 integrin interaction with LM-332 ligand (41). Fluorescence-activated cell sorting analyses reveal that surface expression of ␤4 integrin protein in these cells (JEB ␤4 and JEB ␤4Mut ) is at levels slightly higher than their WT counterparts; surface expression of ␣3 and ␣6 integrin in the cells was comparable to JEB and WT cells (Fig. 1A). Furthermore, when viewed either by confocal or total internal reflection fluorescence (TIRF) microscopy, tagged ␤4 integrin protein localized along the substratum-attached surface of JEB ␤4 and JEB ␤4Mut cells in patterns identical to the distribution of endogenous ␤4 integrin in WT cells (Fig. 1B). JEB cells showed no staining FIGURE 1. Analyses of integrin expression in WT, JEB, JEB ␤4 , and JEB ␤4Mut keratinocytes. A, cell-surface expression of ␤4 integrin, ␣3 integrin, and ␣6 integrin was compared in WT, JEB, JEB ␤4 , or JEB ␤4Mut keratinocytes (black, secondary antibody alone). In B: i and ii, expression of GFP-tagged ␤4 integrin was viewed in the same live JEB ␤4 cell using wide-field fluorescence microscopy and TIRF optics, respectively. iii, a WT keratinocyte, prepared for indirect immunofluorescence using 3E1 antibodies against ␤4 integrin, was viewed by confocal laser scanning microscopy. iv, a JEB ␤4Mut keratinocyte was viewed by TIRF optics. v and vi, a JEB cell, prepared for immunofluorescence using 3E1 antibody, was viewed by phase contrast (v) and by confocal microscopy (vi ). iii and vi, the focal plane was close to the substratum-attached surface of the cell. Bar, 10 m.
JEB Cells Exhibit Aberrant Motility Behavior-We next assayed the migration of WT, JEB, JEB ␤4 , and JEB ␤4Mut keratinocytes in scratch wound motility assays (Fig. 2). Consistent with others (36), both JEB and JEB ␤4Mut cell populations displayed delayed wound closure compared with WT and JEB ␤4 keratinocytes (Fig. 2). We have also monitored migration of individual and groups of cells following plating onto an uncoated substrate. These assays were performed at 6 -8 h after plating to allow the cells to adhere and spread. JEB and JEB ␤4Mut cells showed significantly lower overall migration and displacement over a 2-h period than their WT counterparts (Fig. 3A). These deficiencies were not observed in JEB ␤4 cells, which exhibit migration similar to WT cells (Fig. 3A). We also calculated the directed migration index (net displacement divided by total distance traveled) for the four cell types. None of the indices approached one indicating that the four keratinocytes failed to exhibit directed/processive migration under these conditions (Fig. 3A). Nevertheless, upon careful examination of the images we derived, we noticed that our cells show two distinct motile behaviors. WT cells move along linear tracks (Fig. 3, B (panel i) and C, and supplemental video 1). In addition, a number of WT cells change trajectory and move back over the same regions of the substrate that they have already traversed (backtracking is exhibited by 34 (Ϯ3) % of WT cells that we have examined; n ϭ 163, in three trials), which is the clear reason why they do not exhibit directed migration. In contrast, JEB cells move primarily in circles and exhibit little or no migration along linear tracks; this is true of individual cells as well as groups of adherent cells (Fig. 3, B (panels ii and iii) and C, and supplemental video 2). The migratory patterns displayed by the JEB cells explain why these cells failed to close a wound in a scratch assay (Fig. 2) (36). The motility of JEB ␤4 keratinocytes is similar to WT cells (Fig. 3, B (panel iv) and C, and supplemental video 3), whereas JEB ␤4Mut cells showed migration patterns comparable to JEB cells (Fig. 3, B (panel v) and C, and supplemental video 3).
The motility of WT, JEB, JEB ␤4 , and JEB ␤4Mut keratinocytes is inhibited by removal of growth factors (not shown) or by ␣3 and ␤1 integrin antagonists (P1B5 and 6S6, respectively, only results using P1B5 are shown; Fig. 3B (panel vi)). The antagonists were added to the cells at 6 h after plating, at a time when the cells have adhered to substrate and assembled an LM-332 matrix (not shown). Our analyses were begun 30 min later. The treated cells appeared to "wobble" but showed no obvious motility (Fig. 3B (panel vi)). Control immunoglobulins had no apparent effect on migration (not shown). These results indicate that the motility of all of our assayed keratinocytes is likely growth factor-and ␣3␤1 integrin-regulated, consistent with previous studies (16, 36).
Laminin-332 Matrix Organization Regulates Keratinocyte Motility Behavior-The aberrant motility of JEB cells detailed above might reflect an inherent migration machinery defect in the cells. To assess this possibility, we compared the motility of JEB and WT cells after plating on WT cell matrix. We also compared the migration of JEB and WT cells on JEB cell matrix. Our analyses were performed at 2 h after plating, because the cells rapidly adhere to and spread on the preformed matrices, and we wished to minimize the possibility that the cells would modify the matrix upon which they were plated. JEB keratinocytes and WT cells on the matrix of WT cells exhibited comparable levels of overall distance moved and directed migration; however, JEB cells exhibited somewhat lower displacement values than WT cells, following plating on WT matrix (Fig. 4A). In addition, WT cells and JEB cells both displayed comparable levels of overall migration and displacement on the matrix of JEB cells, albeit at levels lower than that when they were plated onto WT cell matrix (Fig. 4A). Moreover, even on preformed matrices, both WT and JEB cells failed to exhibit directed migration. More important, however, the majority of JEB and WT cells showed movement along linear tracks when plated onto the matrix of WT cells (Fig. 4, B (panel i) and C), whereas WT and JEB keratinocytes moved in a relatively circular fashion on the matrix of JEB cells (Fig. 4, B (panel ii) and C). To rule out the possibility that the cells secrete matrix components and/or proteases that change the functional properties of the matrix, keratinocytes were treated with Brefeldin A (supplemental Fig. S1A). Treatment had no impact on the motile behaviors of the cells, regardless of the matrix on which they were plated. In summary, these data imply: 1) that the intrinsic motile machinery of JEB cells is functional and 2) that the matrix of keratinocytes regulates the precise motile behavior of adherent keratinocytes.
The matrices of WT and JEB cells were both enriched in LM-332, whose subunit composition was comparable (supplemental Fig. S2). To determine whether LM-332 is the functionally active component of the matrix in supporting migration behavior, keratinocytes were allowed to adhere and spread on matrix for 2 h, and then the ␣3 LM subunit   NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 35491
antagonist antibody RG13 was added to the cells (44). After a 30-min incubation, analyses of migration of the cells were begun. RG13 antibody inhibited migration of JEB cells on WT cell matrix and vice versa (only results for JEB cells on WT cell matrix are shown; Fig.  4 (B (panel iii) and C) (control immunoglobulin had no impact; not shown). Similarly, the ␣3 integrin antagonist P1B5 inhibited the motility of JEB and WT cells plated onto each other's matrix (Fig. 4, B (panel iv) and C). These results indicate that motile behavior of the cells, irrespective of matrix type, its processing, or components, is LM-332-and ␣3 integrin-dependent.
If motility of keratinocytes on both JEB and WT cell matrix is LM-332-dependent, does LM-322 organization determine the differences in motile behaviors exhibited by JEB versus WT keratinocytes? WT keratinocytes secreted LM-332 primarily as linear tracks that appeared for some distance beyond the cells (Fig. 5i). Moreover, live cell imaging of WT keratinocytes, which have been induced to express a YFP-tagged ␤3 LM subunit via adenoviral delivery (the YFP is at the C-terminal end of the molecule), revealed that the cells move along these labeled, linear LM-332-rich tracks ( Fig. 6i; in the example shown the cells undergo the backtracking behavior mentioned earlier; supplemental video 4). In these studies, there was a 42:58 ratio of tagged to untagged ␤3 laminin subunits in the matrix of the cells (supplemental Fig. S3). Moreover, the YFP tag provided an accurate representation of the secreted and assembled laminin-332 matrix, because it showed precise colocalization with the ␥2 subunit when the matrix was stained by GB3 antibody (not shown).
JEB cells assembled LM-332 in intersecting arcs/circles with which the ␣3 integrin subunit colocalized (Fig. 5, ii and iii). The ␣6 integrin-blocking antibody, GoH3, when added to the medium of the cells prior to plating onto uncoated coverslips, did not inhibit assembly of the LM-332 circles, indicating that their formation is not ␣6 integrin-dependent (not shown). Furthermore, JEB cells, expressing a YFP-tagged ␤3 LM subunit, move over the labeled circular arrays of LM-332 that they secrete ( Fig. 6ii and supplemental video 4). The distinct organizational states of the proteins in the matrix of both WT and JEB cells appear specific for LM-332; we have observed no obvious LM-10/11 staining, and only diffuse fibronectin staining, in WT and JEB cell matrix (not shown).
The above results imply that LM-332 organization in the matrix of keratinocytes not only mirrors the motile behaviors of adherent cells but also specifies keratinocyte motile behavior patterns. To   i and ii and red in iii-v). The WT cells in i and JEB cells in ii were double-labeled with rhodamine phalloidin (red); the JEB cells in iii were processed with ␣3 integrin antibody (green); in iv and v, green represents GFP-tagged wild-type and mutant ␤4 integrin, respectively. Bars: in iii, 12.5 m; in v, 25 m. provide further support for this, we assayed migration of WT cells at 2 h after plating onto YFP-labeled LM-332-rich arrays of JEB cell matrix and vice versa. In both instances, cells migrated precisely along the tracks or circles defined by the preformed arrays of LM-332 (Fig. 6 (panels iii and iv) and supplemental video 5).
The morphologically distinct arrangement of LM-332 in the matrix of JEB versus WT cells suggested the possibility that ␤4 integrin may be involved in determining LM-332 organization. To assess if this is the case, we assayed LM-332 organization in JEB ␤4 cells. JEB ␤4 keratinocytes assembled a matrix in which LM-332 was organized into linear patterns rather than circles (Fig. 5iv). JEB ␤4Mut cells, expressing a ligand-binding defective ␤4 integrin, assembled a LM-332 matrix organized into circles (Fig. 5v).
Motile Behavior of Keratinocytes and Their Matrix Organization Are Dependent on Active Rac1-The small GTPases of the Rho family regulate assembly of fibronectin and LM matrices (46 -48). In this regard, previous analyses indicate that the activity of the small GTPase Rac1 is significantly lower in JEB cells than in WT cells and that Rac1 activity is regulated by ␣6␤4 integrin (35)(36)(37)(38). Based on these results, we evaluated whether Rac1 is involved in determining motile behavior of keratinocytes via effects on LM-332 matrix organization. To do so, we expressed dominant-negative (DN) Rac1 as well as DN versions of two other small GTPases (cdc42 or RhoA) in WT cells. Expression of the mutant proteins was confirmed by immunoblotting (Fig. 7A), and motility assays were performed on the cells. WT cells expressing either DN cdc42 or DN Rac1 exhibited motility defects (Fig. 7, B (panels i-iii) and C, and supplemental video 6). Most of the cells in the former instance appeared to wobble in place, although an occasional cell showed circular motion (Fig. 7, B (panel i) and C). In contrast, cells expressing DN Rac1 exhibited the same circular migration patterns that are shown by JEB cells (Fig. 7, B (panels ii and iii) and C, and supplemental video 5). The same was true when WT cells were treated with the Rac1 inhibitor NSC23766 (49, 50) (Fig. 7, B (panel iv) and C). DN RhoA had little or no effect on the motility behavior of WT cells (Fig. 7C). WT keratinocytes expressing DN Rac1 or WT cells treated with an Rac1 inhibitor also assembled circles of LM-332 (Fig. 7D, panels i and ii). In addition, untreated WT cells move in a circular fashion on the matrix assembled by WT cells in which the activity of Rac1 has been inhibited (supplemental Fig. S1B). Because the converse is also true (WT cells expressing DN Rac1 move in linear tracks on the matrix of untreated WT cells (supplemental Fig. S1B)), this implies that Rac1 regulates keratinocyte motility behavior by determining LM-332 matrix organization rather than negatively impacting the internal motile machinery of the cells. However, expression of constitutively active (CA) Rac1 is incapable of rescuing the motility defects of JEB cells, and it does not induce the cells to assemble linear tracks of LM-332 ( Fig. 7A; not shown).
Cofilin Regulates the Motile Behavior of Keratinocytes and Their Matrix Organization-The actin-severing protein cofilin is involved in regulating actin organization and, as a downstream target of Rho GTPases, is a key regulator of migration (51)(52)(53)(54). The relative level of Ser3 phosphorylated (inactive) cofilin in JEB cells was greater than in uninfected WT cells, JEB ␤4 cells, or WT cells expressing DN RhoA or DN cdc42 (Fig.  9A). However, the level of inactive cofilin in JEB cells was com-parable to that in WT cells expressing DN Rac1 and in WT cells treated with the Rac1 inhibitor NSC23766 (Fig. 9A), indicating that cofilin activity is linked to the distinct motile behavior of our keratinocytes lines. We have confirmed this, because WT cells induced to express a dominant-inactive, phospho-mimetic cofilin mutant via adenoviral infection show the same motility phenotype as JEB cells (Fig. 9, B and C, and supplemental video 6) (43). The infected cells assemble circular arrays of LM-332 in their matrix (Fig. 9D). In addition, WT cells move in a circular fashion on the matrix laid down by WT cells in which the activity of cofilin has been inhibited (supplemental Fig.  S1C). The relative activity level of cofilin in JEB ␤4Mut cells is comparable to that of WT cells even though the former showed an aberrant circular migration behavior (Figs. 3B (panel v) and 9A). This suggests that the mutant ␤4 integrin is capable of triggering certain signaling pathways, but its inability to bind ligand precludes assembly of LM-332 into linear tracks. Moreover, expression of CA Rac1 in JEB cells also results in an increase in active cofilin without rescuing the motility phenotype of the cells (Fig. 9). Similarly, JEB cells induced to express a dominant-active, non-phosphorylatable cofilin mutant exhibit the same circular migration behavior and matrix organization of their non-infected counterparts (43) (not shown). These data imply that the activity of cofilin is positively regulated by Rac1 and that active cofilin is necessary but not sufficient to induce LM-332 assembly into an organization that supports migration of keratinocytes along linear tracks (55).

DISCUSSION
Most studies on LM-332 have focused on its receptor-binding motifs, its role in disease, and the functions of its fragments (3, 4, 20, 24 -26, 56). Many of these studies have assayed cell behavior on LM-332 randomly spread onto surfaces. Our data reveal that there is an added complexity. The organizational state of LM-332 specifies cellular migration patterns. JEB cell migration is restricted to circular arrays of LM-332. Likewise, WT keratinocytes show LM-332-dependent migration and move along linear tracks of LM-332. Our data provide evidence that this is independent of LM-332 subunit processing and whatever accessory proteins reside in the matrix, because an LM-332 antagonist blocks migration of keratinocytes on both WT and JEB cell matrix. FIGURE 8. Rac1 co-precipitates with ␣6␤4 integrin. Extracts of WT cells were processed for immunoprecipitation using either a ␤4 integrin monoclonal antibody, control IgG (Ϫ), or an ␣3 integrin monoclonal antibody. The precipitated proteins were subjected to immunoblotting using Rac1, ␤1 integrin, or ␣6 integrin antibodies as indicated.
Although there is circumstantial evidence that ␤4 integrin regulates cell motility (31)(32)(33), a recent report has presented data that spontaneously immortalized mouse keratinocytes lacking ␤4 integrin show enhanced, rather than defective, migration (57). The latter result has led Raymond and coworkers (57) to conclude that ␣6␤4 integrin functions to "impede" cell migration. In contrast, our data indicate that immortalized, ␤4 integrin-deficient, human keratinocytes lack the ability to move in a linear fashion and migrate primarily in circles. It remains unclear why our results diverge from those of Raymond et al. (57), although the divergent results may reflect differences between spontaneous versus viral-immortalized cells. Regardless of any discrepancies, the mutant motility phenotype of JEB cells is corrected following expression of wild-type ␤4 integrin. This finding directly pinpoints ␣6␤4 integrin as an active participant in keratinocyte motility. Furthermore, our data are consistent with reports linking ␣6␤4 integrin to regulation of motility (32)(33)(34).
There is only one report that ␣6␤4 integrin plays a role in matrix assembly (58); most studies have focused on how ␤1 integrin is required for LM matrix assembly (see, for example, Refs. 59 -61). Indeed, our own previous studies have pointed to a crucial role for the ␣3 integrin subunit in LM-332 matrix assembly, because cells lacking ␣3(␤1) integrin assemble aberrant arrays of LM-332 (28). Thus, one possible explanation for our results is that ␣6␤4 integrin indirectly determines LM-332 matrix organization via its influence on ␣3␤1 integrin. Although we cannot rule out this scenario, we believe it is unlikely. It implies that ␣6␤4 integrin positively regulates the functions of ␣3␤1 integrin, whereas it is widely accepted that the reverse is the case (35)(36)(37)(38). Moreover, ␤4 integrin must be capable of binding LM-332 ligand in order for LM-332 matrix to be properly organized implying that ␣6␤4 integrin is actively organizing LM-332 matrix. Indeed, based on our previous data and the results we have presented here, we propose that the LM-332 matrix of keratinocytes assembles in "two phases." Specifically, LM-332 is first organized into circles in the extracellular matrix in an ␣3 integrin-dependent manner (28). We hypothesize that ␣6␤4 integrin may remodel circular arrays of LM-332 into linear tracks. Cells deficient in ␤4 integrin retain LM-332 in circular arrays and show ␣3␤1 integrin-dependent circular motility. Consistent with this hypothesis, ␣3 integrin co-localizes with LM-332 in JEB cells. Moreover, our previous data (28) suggest that, in the absence of ␣3 integrin, and therefore without the initial, ␣3 integrin-dependent phase of LM-332 organization, LM-332 is aberrantly arranged into arrowhead arrays. We propose that an initial, ␣3 integrin-dependent phase of LM-332 organization is a necessary prerequisite for the ␤4 integrin-mediated arrangement of LM-332 into linear tracks along which keratinocytes move in a ␣3 integrindependent manner.
We also report that assembly of LM-332 matrix into linear arrays depends on Rac1. Rac1 co-precipitates robustly with ␣6␤4 integrin from extracts of migrating keratinocytes, suggesting that ␣6␤4 integrin/Rac1 complexes are central to the mechanism via which ␣6␤4 integrin regulates LM-332 matrix organization. One might explain this by proposing that ␤4 integrin signals through Rac1 and indirectly regulates matrix assembly. However, this would be inconsistent with our finding that constitutively active Rac1 fails to rescue the motility defect of JEB cells and that a ligand-binding defective ␤4 integrin is capable of activating Rac1-regulated pathways but does not induce LM-332 assembly into linear tracks. These data indicate that Rac1 activation is necessary but not sufficient for LM-332 to be assembled into linear arrays.
How does Rac1 regulate matrix assembly and, in so doing, keratinocyte motility behavior? Rac1 does so via its impact on the actin-severing protein cofilin. We suggest that ␤4 integrin interacts with and activates Rac1, which then induces activa-tion of cofilin. This would parallel Rac1-mediated cofilin activation by epidermal growth factor-related growth factors in MCF-7 epithelial cells (55) (Fig. 10). Our data support this in that cofilin activity is down-regulated in WT keratinocytes where Rac activity is inhibited but not when the cells are expressing DN RhoA or cdc42 or in cells expressing CA Rac1. The mechanism we propose is distinct from a pathway via which active Rac1 induces an inactivation of cofilin through LIM kinase 1 (62)(63)(64).
The involvement of Rac1 and cofilin in regulating LM-332 matrix assembly/keratinocyte motility leads us to put forward a model (Fig. 10). ␣6␤4 integrin "structures" LM-332 in keratinocyte matrix by shuttling over the surface of keratinocytes. We propose that this mechanism is actin-dependent. There is support for this, because actin microfilaments act as a brake to the mobility of ␣6␤4 integrin complexes in the plane of the membrane (65). This supposes that ␤4 integrin links to the actin cytoskeleton for which there is well documented evidence; ␤4 integrin binds actin via plectin (66,67). That plectin, through its interaction with ␤4 integrin, is involved in regulating motility is also supported by other studies (34). Thus, we hypothesize that plectin regulates the ability of ␣6␤4 integrin to organize LM-332 matrices (34,(65)(66)(67)(68)(69) (Fig. 10). This is consistent with data showing that the interaction of ␤4 with plectin stabilizes ␣6␤4 integrin and LM-332 interaction (69). Furthermore, we propose that cofilin functions to release the brake imposed by plectin/actin, allowing ␣6␤4 integrin to become dynamic, thereby mediating LM-332 track assembly. This mechanism has a parallel. Integrin receptors move along actin cables and remodel fibronectin matrices leading to fibronectin fibril assembly (46).
In summary, we have provided new insight into the molecular mechanisms governing how LM-332 is organized in the matrix of keratinocytes. Moreover, we have shown that the organizational state of LM-332 dictates keratinocyte motile behavior. How the assembly of the exquisite arrays of LM-332 in the extracellular matrix is regulated is key to understanding the migration patterns of keratinocytes during wound healing in the skin and skin tumorigenesis.