Rac Activation by Lysophosphatidic Acid LPA 1 Receptors through the Guanine Nucleotide Exchange Factor Tiam1*

Lysophosphatidic acid (LPA) is a serum-borne phos-pholipid that activates its own G protein-coupled receptors present in numerous cell types. In addition to stim-ulating cell proliferation, LPA also induces cytoskeletal changes and promotes cell migration in a RhoA- and Rac-dependent manner. Whereas RhoA is activated via G (cid:1) 12/13 -linked Rho-specific guanine nucleotide ex- change factors, it is unknown how LPA receptors may signal to Rac. Here we report that the prototypic LPA 1 receptor (previously named Edg2), when expressed in B103 neuroblastoma cells, mediates transient activation of RhoA and robust, prolonged activation of Rac leading to cell spreading, lamellipodia formation, and stimulation of cell migration. LPA-induced Rac activation is inhibited by pertussis toxin and requires phosphoinositide 3-kinase activity. Strikingly, LPA fails to activate Rac in cell types that lack the Rac-specific exchange factor Tiam1; however, enforced expression of Tiam1 restores LPA-induced Rac activation in those cells. Tiam1-deficient cells show enhanced RhoA activation, stress fiber formation, and cell rounding in response to LPA, consistent with Tiam1/Rac counteracting RhoA. We conclude that LPA 1 receptors couple to a G i -phos-

Lysophosphatidic acid (LPA) 1 is a bioactive lysophospholipid that acts on its cognate G protein-coupled receptors (GPCRs) to induce a host of cellular responses, ranging from rapid morphological changes to stimulation of cell proliferation and survival (1)(2)(3). Extracellular LPA is produced following platelet activation and, hence, is an active constituent of serum (4,5). LPA is also found in conditioned media from cultured cells (6,7), and its levels are elevated in body fluids from cancer pa-tients (8,9). Three distinct G protein-coupled receptors for LPA have been identified to date, termed LPA 1 , LPA 2 , and LPA 3 (previously Edg2, Edg4, and Edg7, respectively), with LPA 1 / Edg2 being the first identified and most widely expressed subtype (3,10). Although much is known about LPA signaling and its cellular responses, the unique biological properties and specific signaling pathways of each individual LPA receptor remain to be fully characterized. LPA serves as the prototypic GPCR agonist that activates the mitogenic Ras-ERK1/2 cascade via G i (11) and evokes rapid contractile responses, such as cell rounding and neurite retraction, via G␣ 12/13 -mediated activation of the small GTPase RhoA (12,13); however, opposite morphological responses to LPA have also been observed, i.e. cell spreading and pseudopodia formation (14,15).
Although LPA action is most often associated with cell proliferation and morphological changes, less attention has been paid to the effects of LPA on cell motility and migration. Cell migration is fundamental to many normal and pathophysiological processes and plays a central role not only in embryonic development but also in the progression of tumors from a non-invasive to an invasive and metastatic phenotype. LPA stimulates the invasion of tumor cells across a monolayer of normal cells (14,16) and promotes wound healing both in vitro and in vivo (17,18). Interestingly, a migratory response to LPA is also observed in Dictyostelium discoideum amoebae (19), suggesting the possible existence of as-yet-unidentified LPA receptors in invertebrates. However, it remains unclear how LPA receptors signal cell migration. In general, cell migration is driven by signaling pathways controlled by the three Rho GTPases, RhoA, Rac1, and Cdc42, acting in a coordinate fashion (20). Rac1 regulates lamellipodia protrusion and forward movement; Cdc42 establishes cell polarity, and RhoA mediates actomyosin-driven cytoskeletal contraction and stress fiber formation but also detachment of the rear end of migrating cells (20,21).
The mechanisms by which LPA and other GPCR agonists activate RhoA have been studied in some detail and are reasonably well understood. These studies have shown that RhoA is activated via G␣ 12/13 subunits; furthermore, they have led to the identification of a family of Rho-specific guanine nucleotide exchange factors (RhoGEFs) that provide a direct link between G␣ 12/13 and RhoA-GTP accumulation (22)(23)(24). In contrast, very little is still known about how LPA receptors may modulate the activity of Rac and/or Cdc42. In particular, the identity of the Rac-GEF(s) (and/or Cdc42-GEFs) that may act downstream of LPA receptors remains unknown. In the present study, we set out to analyze the activation of the three Rho GTPases downstream of the prototypic LPA 1 /Edg2 receptor and to determine how their activity relates to LPA-induced cell migration. We show that the LPA 1 receptor, in addition to transiently activating RhoA, mediates prolonged activation of Rac via a G i -PI3K pathway leading to lamellipodia formation, cell spreading, and migration. Furthermore, we show that the Tiam1 exchange factor provides the link between LPA 1 /Edg2 receptors and Rac activation.
Expression Constructs, Retroviral Transduction, and Transfections-The LPA 1 /Edg2 cDNA was isolated from a human brain cDNA library (Stratagene) and its sequence submitted to GenBank TM (accession number U78192). C-terminally FLAG-or HA-tagged LPA 1 receptor cDNAs were cloned into the retroviral vector LZRS-IRES-Neo (26), and recombinant viruses were produced by transfection of retroviral cDNA constructs into Phoenix packaging cells (27). B103 cells were infected with virus-containing supernatants in the presence of 4 g/ml Polybrene, followed by G418 selection. Myc-tagged N17Rac, N17Cdc42, and N19RhoA cloned into LZRS-IRES-zeo (26) were introduced into B103-LPA 1 cells by retroviral transduction and selected on Zeocin (0.25 mg/ml). HA-tagged Tiam1 constructs have been described (28). COS7 cells were transfected using the DEAE-dextran method.
Cell Migration-Cell migration was measured in Transwell chambers (Costar Corp., pore size 8 m). Filters were coated with 10 g/ml laminin-1 (in PBS overnight at 4°C), rinsed once with PBS, and placed in the lower chamber containing serum-free DMEM supplemented with agonists. Cells suspended in serum-free DMEM containing 0.1% fatty acid-free bovine serum albumin were added to the upper chamber (1 ϫ 10 5 cells/well). Cells were allowed to migrate for 4 h at 37°C. Nonmigratory cells were removed from the top filter surface with a cotton swab. Migrated cells, attached to the bottom surface, were fixed in 3% formaldehyde/PBS, permeabilized in methanol, stained with crystal violet, and counted.
Fluorescence Microscopy-Cells were grown on laminin-1-coated glass coverslips and fixed in 3.7% formaldehyde/PBS for 10 min. Cells were permeabilized in 0.1% Triton X-100 and blocked with 2% bovine serum albumin in PBS. HA-tagged LPA 1 was detected by antibody 3F10 (Roche Molecular Biochemicals) and fluorescein isothiocyanate-labeled secondary antibodies (Zymed Laboratories Inc.). Cells were stained simultaneously with rhodamine-conjugated phalloidin (Molecular Probes). Images were collected by confocal microscopy (Leica).

LPA 1 /Edg2 Receptor Signaling and Motility Responses in
B103 Cells-Most cell types co-express at least two distinct LPA receptors, which hampers the dissection of receptor-specific signaling pathways. One exception is the B103 neuroblastoma cell line, which lacks detectable expression of LPA receptors (30). Through retroviral transduction, we obtained B103 cells that stably express the epitope-tagged LPA 1 /Edg2 receptor at relatively modest levels (B103-LPA 1 cells), as shown by immunoprecipitation and immunoblot analysis (results not shown). In serum/LPA-containing medium, the parental B103 cells have a rounded, refractile appearance and tend to grow in small islands. In contrast, B103-LPA 1 cells display a flattened morphology characterized by prominent lamellipodia and membrane ruffles, whereas they are more "scattered" throughout the dish (Fig. 1A), suggesting that LPA 1 receptor expression promotes random cell motility. Proper membrane localization of LPA 1 receptors was examined by confocal microscopy, which reveals that LPA 1 receptors localize preferentially to lamellipodia (Fig. 1B). We found that LPA 1 receptors undergo rapid internalization from the plasma membrane after LPA addition (Fig. 1B) and mediate pertussis toxin (PTX)-sensitive activation of ERK1,2; the latter response was not observed in the parental B103 cells (Fig. 1C). Thus, LPA 1 receptors expressed in B103 cells are functional and couple to G i/o -mediated activation of ERK1,2.
When maintained in serum-free medium, B103-LPA 1 cells underwent rapid cell rounding and process retraction following addition of LPA ( Fig. 2A), very similar to the contractile responses observed in LPA-treated N1E-115 neuroblastoma cells (12). LPA-induced cell contraction was complete after about 3 min and was observed in at least 80% of the cells in a randomly selected microscopic field ( Fig. 2A). In the continuous presence of LPA, rounded cells began to re-spread after 15-30 min. Cell rounding and re-spreading were fully prevented by retrovirally introduced dominant-negative N19RhoA and the Rho kinase inhibitor Y-27632 but not by PTX 2 (see also Ref. 30). LPA-induced cell rounding was associated with a rapid increase in RhoA-GTP levels, as measured by the GST-Rhotekin pull-down assay (Fig. 2B). In most experiments, however, RhoA activation was hard to detect and always very transient (peaking at 2-3 min and lasting Ͻ10 min). From these results, together with previous findings (13,(22)(23)(24)31), we conclude that LPA 1 receptors couple to a G␣ 12/13 -linked RhoGEF-RhoA-Rho kinase pathway to mediate rapid but transient actomyosin-driven cytoskeletal contraction.
Because LPA 1 receptor expression seems to confer a motile phenotype on B103 cells, we measured LPA-induced cell migration using a Transwell system, in which a laminin-1-coated membrane filter separated the upper cell-containing chamber 2 F. van Leeuwen, unpublished results. from the lower LPA-containing chambers. As shown in Fig. 3A, LPA strongly stimulates cell migration in B103-LPA 1 cells but not in the parental B103 cells. In the absence of added LPA, B103-LPA 1 cells are slightly more motile than the parental cells (Fig. 3A); one plausible explanation for this finding is that some "autocrine" LPA may accumulate in the cellular microenvironment during the course of the experiment (3-4 h). Very similar migratory responses were observed when LPA was present in both chambers, indicating that LPA 1 receptors mediate both random and directed cell migration (chemokinesis and chemotaxis, respectively). LPA-induced cell migration was markedly, although not completely, inhibited by PTX but not by the mitogen-activated protein kinase/ERK kinase inhibitor PD98059 at doses that block ERK1,2 activation (Figs. 3A and 1C). Migration in control cells was also reduced by PTX treatment, indicating that G i is required to sustain basal cell motil-ity. Thus, LPA 1 receptor-driven cell migration is mediated by G i but independent of the G i -ERK1,2 activation pathway.
Cell migration during wound healing of fibroblast monolayers (in serum-containing medium) depends on the activity of RhoA, Rac, and Cdc42, with active Rac inducing forward cell movement (20). We examined the requirement of Rac, RhoA, and Cdc42 for LPA-induced cell migration by expressing their dominant-negative versions in B103-LPA 1 cells via retroviral transduction. As shown in Fig. 3B, expression of dominantnegative N17Rac led to a significant inhibition of LPA-induced migration. Expression of dominant-negative N19RhoA and N17Cdc42 also inhibited cell migration, albeit to a lesser degree. These results indicate that LPA-induced cell migration requires the activity of all three Rho GTPases. We next measured the activation state of Rac and Cdc42 in LPA-stimulated B103-LPA 1 cells, using GST-PAK pull-down assays. As shown in Fig. 3C, LPA 1 receptor stimulation leads to a rapid increase in Rac-GTP levels. Unlike the relatively weak and transient RhoA activation response (Fig. 2B), LPA-induced Rac activation was robust and prolonged, decreasing to above basal levels after about 30 min (Fig. 3C). The GST-PAK fusion protein can also be used to detect Cdc42 activity. However, we did not detect increased Cdc42 activity above basal levels in LPA-stimulated B103-LPA 1 cells (n ϭ 6; results not shown).
Activation of Rac by cell-surface receptors in general, and GPCRs in particular, occurs through incompletely characterized effector routes. In many cases, however, Rac activation is critically dependent on PI3K activity. We found that LPAinduced Rac activation in B103-LPA 1 cells is inhibited by PTX and the PI3K inhibitor wortmannin (Fig. 3D), consistent with Rac being activated via G i -mediated stimulation of PI3K activity. In keeping with this, LPA activates the PI3K downstream target Akt/protein kinase B in a PTX-and wortmannin-sensitive manner (Fig. 3D). PI3K isozymes can be activated by binding to receptor protein tyrosine kinases, activated Ras or G protein ␤␥ subunits. Previous antibody-blocking experiments have implicated the PI3K␤ isoform in LPA signaling (32), whereas more recent findings indicate that this ubiquitously expressed ␤-isoform is activated by G␤␥ dimers both in vitro and in vivo (33)(34)(35). Therefore, LPA-induced Rac activation is most likely mediated by the G␤␥-regulated PI3K␤ isoform, although this remains to be established experimentally.

The GDP/GTP Exchange Factor Tiam1
Mediates LPA-induced Rac Activation-The available evidence indicates that activation of a given Rho GTPase occurs through stimulation of a GDP/GTP exchange factor (GEF), rather than by inhibition of a GTPase-activating protein (for review see Ref. 36). The above results therefore prompt the question of which GEF(s) may link LPA receptors to Rac activation in a PI3K-dependent manner. Little is still known about the identity of Rac-GEFs that act downstream of GPCRs. However, one attractive candidate is Tiam1, a Rac-specific GEF that was originally isolated as a lymphoma invasion-inducing gene product (37). Tiam1 is widely expressed and has been implicated not only in tumor cell invasion and metastasis but also in neurite outgrowth and cell-cell adhesion (38,39). Tiam1 function is PI3K-dependent by virtue of the presence of an N-terminal pleckstrin homology domain that binds preferentially to PtdIns(3,4,5)P 3 (40), but otherwise the upstream signaling pathways that lead to Tiam1 activation remain unknown.
Whereas Tiam1 is highly expressed in B103 cells, 2 interference approaches were not feasible since dominant-negative versions of Tiam1 are not available, and our experiments using RNAi-expressing vectors met with little success so far. To examine the possible role of Tiam1 in LPA 1 receptor signaling, we therefore took advantage of LPA-responsive cells that lack endogenous Tiam1, notably COS7 cells (38), Ras-transformed MDCK cells (41), and Tiam1-null fibroblasts (see below). These cell types predominantly express LPA 1 receptors, with little LPA 2 and/or LPA 3 mRNA detectable. 3 In COS7 cells, LPA activates the G i -mediated Ras-ERK1/2 pathway (42) yet fails to stimulate Rac activity (Fig. 4A). In Tiam1-transfected COS7 cells, however, a significant increase in Rac-GTP levels was observed in response to LPA.
We then turned to MDCK epithelial cells, which express endogenous Tiam1 (39). When transformed by activated V12Ras, MDCK cells (MDCK-f3) revert from an epithelial to a mesenchymal phenotype, which is associated with transcriptional down-regulation of Tiam1 and reduced Rac-GTP levels (41). As shown in Fig. 4B, LPA activates Rac in wild-type MDCK cells but not in the Tiam1-deficient MDCK-f3 cells. Re-introduction of Tiam1 restores Rac activation by LPA (Fig.  4B). Taken together, these results show that LPA-induced Rac activation requires Tiam1.
Next, we compared Rac activation in MEFs derived from Tiam1 homozygous knockout mice with that in wild-type MEFs. Tiam1-deficient mice are phenotypically normal, but they are resistant to skin carcinogenesis (25). Wild-type and Tiam1 knockout (Ϫ/Ϫ) MEFs express LPA 1 and little LPA 2 but not LPA 3 mRNA. 3 In serum-starved MEFs, LPA induced a significant increase in Rac-GTP levels (Fig. 4C). In contrast, no LPA-induced increase in Rac-GTP levels was detected in the Tiam1(Ϫ/Ϫ) cells, whereas ERK1,2 activation was not impaired (Fig. 4C). Thus, three independent lines of evidence obtained in different cell systems (COS7, MDCK, and Tiam1knockout MEFs) demonstrate that Tiam1 is necessary and sufficient for LPA to activate Rac. In other words, LPA-induced Rac activation is mediated by Tiam1. 4 Assessment of the migratory behavior of Tiam(Ϫ/Ϫ) MEFs was hindered by the fact that Tiam1 deficiency led to marked cell contraction, enhanced stress fiber formation, and reduced cell adhesion, when compared with wild-type MEFs (Fig. 4D and results not shown). LPA stimulation led to further cytoskeletal contraction in the Tiam1(Ϫ/Ϫ) cells but not in the wild-type MEFs. These observations strongly suggest that RhoA signaling is enhanced following Tiam1 deletion. Indeed, although no RhoA activation was detectable in wild-type MEFs, the Tiam(Ϫ/Ϫ) cells showed a significant RhoA activation response to LPA (Fig. 4D). It thus appears that the balance between Rac and RhoA activity depends on Tiam1 expression, a notion consistent with previous observations (29,38) showing that Tiam1/Rac activation inhibits RhoA. Whereas the mechanism underlying this negative cross-talk remains to be elucidated, these results show that the Rac-RhoA activity balance in a given cellular context is critical in determining whether LPA receptor stimulation leads to cell spreading and migration or to cell rounding and reduced adhesion.
In conclusion, our findings indicate that LPA 1 receptors couple to a G i -mediated PI3K-Rac activation pathway that is essential for the stimulation of cell motility. We have identified Tiam1 as the GDP/GTP exchange factor that is necessary and sufficient for LPA to activate Rac in three different cell types. This newly established G i -Tiam1-Rac pathway counteracts the G 12/13 -linked RhoGEF-RhoA activation pathway, as schematically depicted in Fig. 5. In the simplest signaling scheme that is compatible with the current evidence, Tiam1 is activated by direct binding of PI3K lipid products (particularly PtdIns(3,4,5)P 3 ; Ref. 40) to its N-terminal pleckstrin homology domain. However, it could well be that additional signals emanating from G i are required for full activation of Tiam1, perhaps similar to the situation in P-Rex1, a Rac-specific GEF that is synergistically activated by PtdIns(3,4,5)P 3 and G␤␥ subunits (44). In addition to the identification of Tiam1 as a key player in LPA-induced Rac activation, our findings highlight the importance of LPA 1 /Edg2 as a cell motility-stimulating GPCR. Whether the other LPA receptor members, LPA 2 /Edg4 and LPA 3 /Edg7, can play a similar physiological role is currently under investigation.
FIG. 5. LPA 1 receptor signaling pathways leading to activation of Rac and RhoA. In this scheme, Rac activation is mediated by G i (␤␥) and involves PI3K-dependent activation of the Tiam1 RacGEF; PI3K␤ is the most likely PI3K isoform involved (see text). RhoA activation occurs in parallel via one or more G␣ 12/13 -linked RhoGEF(s) (22)(23)(24). Through an inhibitory cross-talk mechanism that remains to be elucidated, Tiam1/Rac signaling suppresses both basal and LPA-induced RhoA activation. Coordinate regulation of Rac and RhoA activity thus controls LPA-induced cytoskeletal changes (cell spreading and rounding, respectively) and cell motility. See text for further details.