Restoration of β1A Integrins is Required for Lysophosphatidic Acid-induced Migration of β1-null Mouse Fibroblastic Cells*

Cells lacking the β1 integrin subunit or expressing β1A with certain cytoplasmic mutations have poor directed cell migration to platelet-derived growth factor or epidermal growth factor, ligands of receptor tyrosine kinases (Sakai, T., Zhang, Q., Fässler, R., and Mosher, D. F. (1998)J. Cell Biol. 141, 527–538). We investigated the effect of expression of β1A integrins on lysophosphatidic acid (LPA)-induced migration of fibroblastic cells derived from β1-null mouse embryonic stem cells. These cells expressededg-2, a G-protein-linked receptor for LPA, as well as the related edg-1 receptor. Cells expressing wild type β1A demonstrated enhanced cell migration across filters coated with gelatin or adhesive proteins in response to LPA, whereas β1-deficient cells lacked LPA-induced cell migratory ability. Checkerboard analyses indicated that LPA causes both chemotaxis and chemokinesis of β1-replete cells. Cells expressing β1A with mutations of prolines or tyrosines in conserved cytoplasmic NPXY motifs, threonine in the inter-motif sequence, or a critical aspartic acid in the extracellular domain had low migratory responses to LPA. These findings indicate that active β1A integrin is required for cell migration induced by LPA and that the cytoplasmic domain of ligated β1A interacts with pathways that are common to both receptor tyrosine kinase and G-protein-linked receptor signaling.

Directed cell migration in a concentration gradient (chemotaxis) is important for a large number of physiological and pathological processes, including development, immunity, wound healing, and cancer metastasis (1)(2)(3)(4). Chemotaxis involves the sensing of the gradient of chemoattractant, reorganization of the actin cytoskeleton, and subsequent movement toward the chemoattractant. Cell movement requires the fine control of cellular association with and release from the extracellular matrix (5,6). Integrins are transmembrane heterodimeric cell adhesion receptors that mediate organization of focal contacts, actin-containing cytoskeleton, and extracellular matrix and may also contribute to cell migration by participating in signal transduction cascades (7)(8)(9)(10)(11)(12)(13)(14). It has been suggested that integrin function involves interaction with adhesive ligands ("outside-in" signaling) and cellular control of binding avidity ("inside-out" signaling) (9,15,16). In addition, cell migration may be regulated in part by the cycling of integrins between cytoplasmic compartments and the cell surface (17,18).
Lysophosphatidic acid (LPA) 1 is a product of activated platelets and cells and has diverse actions on cells (19). LPA is the serum enhancement factor of fibronectin matrix assembly; enhancement of assembly closely correlates with LPA-induced actin stress fiber formation and cell contraction (20 -22). LPA is a mitogen for a number of cells, including endothelial cells (23,24). LPA induces in vitro invasion across host cell monolayers by several types of tumor cells (25,26). LPA also stimulates random, nondirectional migration (chemokinesis) of Rat1 fibroblasts (27). Recently, the specific LPA receptor, ventricular zone gene-1 (vzg-1) (28), also known as endothelial differentiation gene-2 (edg-2) (29), has been identified in mouse and human. Related proteins, Edg-1 (30,31) and Edg-3 (32,33), have been identified as receptors for sphingosine 1-phosphate. LPA signaling is mediated by the Raf-ERK pathway via Ras activation and through the small GTP-binding protein Rho, leading to activation of mitogen-activated protein kinase (19,29,34). Expression of a dominant negative Ras mutant inhibits migration of NIH(M17) cells in response to LPA as well as to other chemoattractants such as platelet-derived growth factor (PDGF) (35).
We recently expressed a set of ␤ 1 A integrin subunits with point mutations of the cytoplasmic domain in fibroblasts derived from ␤ 1 -null stem cells and showed that cells lacking the ␤ 1 integrin subunit or expressing ␤ 1 A with certain cytoplasmic mutations had impaired ability to migrate toward PDGF or epidermal growth factor (EGF), ligands of receptor tyrosine kinases (36). This system allows the effects of wild type and mutated ␤ 1 on cell migration to be analyzed without the confounding presence of endogenous ␤ 1 subunits. In the present study, we demonstrate that restoration of ␤ 1 integrins is stringently required for LPA-induced cell migration of ␤ 1 -null fibroblasts.

EXPERIMENTAL PROCEDURES
Cells-The GD25 and GD10 fibroblast lines were established after differentiation from ␤ 1 -null stem cells and immortalization with SV40 large T antigen (37,38). ␤ 1 As with mutations of the cytoplasmic domain were constructed from pBS␤ 1 A encoding full-length mouse ␤ 1 A integrin subunit and expressed in GD25 cells as described (36). A point mutation in the extracellular domain of ␤ 1 A (D130A) was introduced as follows. cDNA for ␤ 1 A was excised with XbaI and Acc65I from pBS␤ 1 A (38) and cloned into pGEM-7Zf. The mutant was generated by oligonucleotideprimed DNA synthesis using a mutagenesis kit (Pharmacia Biotech, Uppsala, Sweden). The region spanning the HindIII and ClaI sites was analyzed by DNA sequence analysis, and the correctly mutagenized HindIII-ClaI fragment was isolated and ligated into HindIII-ClaI-digested pGEM-7Zf containing the original XbaI-BglII fragment of ␤ 1 A. The XbaI-NcoI fragment was then excised and ligated into XbaI-NcoIdigested pBS␤ 1 A, yielding cDNA encoding full-length ␤ 1 A polypeptide containing the point mutation. The plasmid was linearized with XbaI and transfected into GD25 cells by electroporation. Clones stably expressing mutant ␤ 1 A were obtained with the selection by puromycin and analyzed for expression of ␤ 1 A by flow cytometry (36). The population of higher expressing cells in several clones was selected by fluorescence-detected cell sorting and expanded. By flow cytometry, the cells selected for study expressed mutant cell-surface ␤ 1 A at comparable levels with cells expressing wild type ␤ 1 A.
Wild type ␤ 1 A was expressed in GD10 cells by transfecting with pBS␤ 1 A using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions.
Cell Migration-Cell migration assays were performed in modified Boyden chambers containing Nucleopore polycarbonate membranes (5-m pore size; Costar Corp., Cambridge, MA) as described previously (36). The filters were soaked overnight in a 10 g/ml solution of vitronectin, fibronectin, or laminin-1 or 100 g/ml gelatin, briefly rinsed with phosphate-buffered saline containing 0.2% (w/v) bovine serum albumin (BSA), air-dried, and then placed in the chamber. PDGF from porcine platelets (R&D systems, Minneapolis, MN) or 1-oleoyl-LPA (Avanti Polar Lipids, Birmingham, AL) in Dulbecco's modified Eagle's medium containing 0.2% fatty acid-free BSA was added to the lower (and/or upper) compartment of the chambers. Cells suspended in Dulbecco's modified Eagle's medium containing 0.2% fatty acid-free BSA were introduced into the upper compartment. The chambers were then incubated for 6 h at 37°C. The filters were fixed and stained, and the cells that had migrated to the lower surface were counted at ϫ 400 magnification. In each experiment, two areas from each of two wells were counted. Values are the mean Ϯ S.D. of cells per 0.16-mm 2 field.
Immunoprecipitation and Immunoblotting-Immunoprecipitation analysis was performed as described previously elsewhere (39), with a slight modification. Briefly, cells grown to 80 -90% confluence were starved in medium without serum for 16 h and then stimulated with agonists for 5 min or left unstimulated. The cells were then lysed on ice in buffer containing 1% (v/v) Triton X-100, 150 mM NaCl, 5 mM EDTA, 100 mM sodium fluoride,1 mM sodium orthovanadate, 0.5 mM sodium molybdate, 2 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 0.1 g/ml pepstatin A, 0.4 mM pefabloc SC, and 20 mM Tris-HCl, pH 7.4 and centrifuged for 15 min at 12,000 r.p.m. The same amounts of protein from different experimental samples were used for analyses, as determined using a BCA protein assay (Pierce). The supernatants were precleaned with protein A-Sepharose 4 fast-flow (Pharmacia LKB Biotech, Uppsala, Sweden) and subsequently incubated with antibody. The complexes were precipitated with protein A-Sepharose 4 fast-flow, and the proteins were eluted from the resins by incubation with SDSsample buffer. Samples were then subjected to SDS-polyacrylamide gel electrophoresis and transferred onto polyvinylidene difluoride membranes.
For immunoblotting, the blots were probed with the primary antibody, then with a horseradish peroxidase-conjugated secondary antibody (Organon Teknika Corp., Westchester, PA). Immunoreactive bands were developed using the enhanced chemiluminescence (ECL) substrate system (NEN Life Science Products). Rabbit polyclonal antibodies that recognized mouse EGF and PDGF receptors were from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse monoclonal antibody against phosphotyrosine (mAb 4G10) was from Upstate Biotechnology Inc. (Lake Placid, NY). Human recombinant EGF was from Upstate Biotechnology Inc.
Migration through Gelatin-coated Filters-The effects of ␤ 1 A integrin on LPA-induced cell migration through gelatin-coated filters were analyzed. GD25 cells lacking ␤ 1 A migrated very little when LPA (500 nM) was added to the lower compartment ( Fig. 2A and Table I). GD25 cells expressing wild type ␤ 1 A migrated 10-fold more than nontransfected GD25 cells in response to LPA. GD25 cells and GD25 cells expressing ␤ 1 A migrated equally well in response to PDGF (Table I). GD10 cells and GD10 cells expressing wild type ␤ 1 A behaved similarly to the GD25 counterparts (Table I). Thus, ␤ 1 -null fibroblasts demonstrate a profound defect in migration through gelatin-coated filters in response to LPA that is not because of an intrinsic inability to migrate through the gelatin-coated pores.
Cells expressing mutant ␤ 1 As were tested for an ability to migrate in response to LPA. The mutations in the cytoplasmic domain fell into two groups, active (D759A; Y783F; Y795F; Y783,795F) and inactive (T788P; P781,793A), based upon the reactivity of expressing cells with anti-␤ 1 (9EG7) antibody, ␣6␤ 1 -dependent adhesion to laminin-1, and ability to support fibronectin assembly (36). T788P or P781,793A cells, which express inactive ␤ 1 As, migrated much less in response to LPA through gelatin-coated filters than cells expressing wild type ␤ 1 A (Fig. 2B). D130A cells, carrying a point mutation of extracellular domain of ␤ 1 A known to impair ligand binding (40), also showed less ability to migrate. In contrast, cells expressing ␤ 1 A with an activating D759A mutation of conserved aspartate in the membrane-proximal region of the cytoplasmic domain migrated 9-fold more than GD25 cells lacking ␤ 1 A and equivalently to the cells expressing wild type ␤ 1 A. Although mutation of one or both of the tyrosines in the two NPXY motifs (Tyr 783 , Tyr 795 ) to phenylalanines results in active ␤ 1 A as assessed by adhesion to laminin-1 and reactivity with the 9EG7 antibody (36), Y795F or Y783,785F cells migrated much less than cells expressing wild type ␤ 1 A and no more than GD25 cells lacking ␤ 1 A (Fig. 2B) 1. Expression of mRNAs for the vzg-1(edg-2) LPA receptor and the related sphingosine 1-phosphate receptor edg-1 detected by RT-PCR. Amplification products were resolved on 2% agarose gel. laminin-1 as well as migration through such filters in response to PDGF or EGF (36). Migration of cells expressing wild type ␤ 1 A through vitronectin-, fibronectin-, or laminin-1-coated filters in response to LPA was greater than migration through gelatin-coated filters (compare Table I and Fig. 3A). In response to LPA, GD25 cells expressing ␤ 1 A migrated 5-fold (on vitronectin) or 6-fold (on fibronectin) more than cells lacking ␤ 1 A and 10-fold (on vitronectin) or 3.5-fold (on fibronectin) more than the haptotactic responses in the absence of LPA (Fig. 3A). These results suggest that the participation of ␤ 1 A integrins in cell-matrix interactions synergizes with the response to LPA. LPA-induced migration through laminin-1coated filters required expression of ␤ 1 A (Fig. 3A), as is also the case for EGF-or PDGF-induced migration (36). Migration of cells expressing wild type ␤ 1 A was half-maximal in response to approximately 50 nM LPA through all the adhesive substrates (Fig. 3A). Migration of cells expressing the cytoplasmic D759A mutation was somewhat less than that of cells expressing wild type ␤ 1 A (Fig. 3B). Y783,795F cells migrated much less, nearly equivalent to GD25 cells lacking ␤ 1 A. Y783F and Y795F cells had migratory behavior that was intermediate between Y783,795F cells and cells expressing wild type ␤ 1 A (Fig. 3B). D130A cells also had an impaired migratory response through filters coated with fibronectin, vitronectin, or laminin-1 (not shown).
Checkerboard Analysis-Previous studies suggested that LPA causes random, nondirectional migration (chemokinesis) rather than chemotaxis of Rat1 fibroblasts (27). Checkerboard analysis in which different concentrations of chemoattractant were added to the upper and lower chamber of the apparatus was therefore carried out to characterize the effect of LPA on cell migration by derivatives of GD25 cells (Fig. 4). Such an analysis differentiates directed migration across the filter in response to the gradient of chemoattractant (chemotaxis) from increased random motility because of the presence of the chemoattractant per se (chemokinesis). Cells expressing wild type ␤ 1 A, the D759A mutant, or the Y783F mutant displayed directional motility in a LPA gradient, although there were substantial increases in random motility when LPA was present at equal concentrations in both chambers: the chemotactic responses were 1.2-1.5-fold greater than the chemokinetic responses. Migration of Y795F cells was greater when LPA was present in both chambers than when there was a gradient of LPA, indicating that the main effect was a chemokinetic response. Migration of Y783,795F cells or nontransfected GD25 cells was too low to classify responses as chemotaxis or chemokinesis.
Relation to EGF and PDGF Signaling-Signaling pathways initiated by high concentration of LPA (10 -25 M) are known to "cross-talk" with EGF signaling pathways to cause phosphorylation of the EGF receptors (41,42). Because GD25 cells are also deficient in migration in response to EGF or PDGF (36), we checked the presence of EGF and PDGF receptors in GD25 cells

TABLE I Effect of LPA or PDGF on cell migration through gelatin-coated filters
Cells and chemotactic agents were added to the upper and lower compartments, respectively. Numbers represent the mean Ϯ S.D. of at least four determinations.  and GD25 cells expressing ␤ 1 A. Lysates of both cells contained comparable amounts of both receptors when analyzed by SDSpolyacrylamide gel electrophoresis and immunoblotting (not shown). As assessed by anti-phosphotyrosine immunoblotting of immunoprecipitated receptors, both receptors in both cells were activated appropriately by concentrations of 3 ng/ml EGF or PDGF but not by 500 nM LPA that caused migration (not shown). DISCUSSION We previously found that ␤ 1 A with an intact cytoplasmic tail, including intact NPXY motifs, is important for optimal chemotaxis of GD25 fibroblasts through filters coated with vitronectin, fibronectin, or laminin-1 in response to PDGF or EGF (36). The purposes of the present studies were to learn if GD25 cells or its derivatives migrate in response to LPA and then to compare and contrast LPA-induced migration to that induced by PDGF or EGF. The fact that the extracellular D130A mutation ablated ability of ␤ 1 A to support migration through filters coated with fibronectin or vitronectin indicates that ␤ 1 integrins must be able to interact with extracellular ligands. However, which integrins are responsible for migration on the various coated filters is open to question, with the exception of laminin-1, to which adhesion of ␤ 1 A-expressing GD25 cells is blocked by anti-␣6␤ 1 antibody (36).
GD25 and GD10 cells lacking ␤ 1 A expressed the vzg-1/edg-2 LPA receptor and also the related edg-1 receptor. The generality of the findings for better defined cell types remains to be elucidated. Bovine heart or aortic endothelial cells expressing edg-1 respond to LPA with enhanced cell migration. 3 In contrast, human MG63 osteosarcoma cells expressing vzg-1/edg-2 respond to LPA with inhibited migration. 2,4 Thus, the germ cell-derived lines should be considered as models of cellular behavior rather than of any one cell type.
At least two functions of ␤ 1 integrins may account for the need for ␤ 1 integrins in migration of the germ cell-derived cells: binding of ␤ 1 integrins by insolubilized ligands to allow haptotaxis and synergism between integrin-initiated and LPA-initiated signaling pathways to drive chemotaxis and chemokinesis. Although high concentrations of LPA (10 -25 M) have been shown to induce tyrosine phosphorylation of EGF receptors in a variety of cell types (41,42), we could not show such crosstalk for EGF or PDGF receptors of GD25 cells with a concentration of LPA (500 nM) that induced optimal chemotaxis, suggesting that the initial steps in LPA-induced signaling are independent of tyrosine kinase receptors. LPA (19) and ligated ␤ 1 integrins (43) both signal by multiple pathways. Our results indicate that the ␤ 1 A integrin must not only be active but have intact NPXY motifs in the cytoplasmic domain to support LPAinduced migration. The latter requirement differentiates the effect of LPA on migration from the effect of LPA on fibronectin matrix assembly, which is up-regulated by LPA in GD25 cells expressing ␤ 1 A with conservative Tyr to Phe substitutions (36). We previously hypothesized that conversion of both tyrosines between nonphosphorylated and phosphorylated states was critical for directed movement, based upon the fact that the Y783F and Y795F mutations caused loss of migration ability to PDGF or EGF (36). One scenario is that the conversion between phosphorylated and dephosphorylated states is important for cycling of integrins to facilitate migration in response to a variety of agonist-receptor systems. Phosphorylated integrins may initiate a pathway leading to changes in F-actin containing cytoskeleton and thus to generation of the cellular polarity required for directional movement. In addition, NPXY motifs may regulate cycling of ␤ 1 integrins. Upon phosphorylation of the NPXY motifs, the integrin may lose its affinity for both extracellular ligand and cytoplasmic components of the focal contacts and exit the focal contact. Dephosphorylation of the motifs would allow the integrin to participate in a new round of ligation and focal contact formation. The phosphorylation-dephosphorylation cycle may also allow polarization of receptors for chemotactic agents. An alternative scenario is that the enzymatic cascade initiated by ligation of integrins and phosphorylation of ␤ 1 A is linked with a common pathway initiated by ligated LPA receptor or tyrosine kinase receptors. Expression of a dominant negative Ras inhibits migration in response to both LPA and PDGF but not to soluble fibronectin (35). Ligated LPA receptors (19), ligated ␤ 1 integrins (43), and ligated EGF or PDGF receptors (41,42) all can work through Ras. Thus, downstream targets of Ras are the most likely common pathway.