RhoGEF Specificity Mutants Implicate RhoA as a Target for Dbs Transforming Activity

Abstract

Dbs is a Rho-specific guanine nucleotide exchange factor (RhoGEF) that exhibits transforming activity when overexpressed in NIH 3T3 mouse fibroblasts. Like many RhoGEFs, the in vitro catalytic activity of Dbs is not limited to a single substrate. It can catalyze the exchange of GDP for GTP on RhoA and Cdc42, both of which are expressed in most cell types. This lack of substrate specificity, which is relatively common among members of the RhoGEF family, complicates efforts to determine the molecular basis of their transforming activity. We have recently determined crystal structures of several RhoGEFs bound to their cognate GTPases and have used these complexes to predict structural determinants dictating the specificities of coupling between RhoGEFs and GTPases. Guided by this information, we mutated Dbs to alter significantly its relative exchange activity for RhoA versus Cdc42 and show that the transformation potential of Dbs correlates with exchange on RhoA but not Cdc42. Supporting this conclusion, oncogenic Dbs activates endogenous RhoA but not endogenous Cdc42 in NIH 3T3 cells. Similarly, a competitive inhibitor that blocks RhoA activation also blocks Dbs-mediated transformation. In conclusion, this study highlights the usefulness of specificity mutants of RhoGEFs as tools to genetically dissect the multiple signaling pathways potentially activated by overexpressed or oncogenic RhoGEFs. These ideas are exemplified for Dbs, which is strongly implicated in the transformation of NIH 3T3 cells via RhoA and not Cdc42.

Dbs is a Rho-specific guanine nucleotide exchange factor (RhoGEF) that was identified in a screen for cDNAs whose expression causes deregulated growth in NIH 3T3 mouse fibroblasts (37). The RhoGEFs are a large family of structurally related proteins that share a common ability to catalyze the exchange of GDP for GTP on members of the Rho family of small GTPases (39). Like all GTP-binding proteins, Rho proteins function as binary switches cycling between a biologically inactive GDP-bound state and a biologically active GTP-bound state. Conversion to the GDP-bound form is achieved through intrinsic GTP hydrolysis, which can be further stimulated by the actions of GTPase-activating proteins (18). Conversion to the GTP-bound form is mediated by the actions of RhoGEFs, which stimulate the dissociation of bound GDP, thus providing an opportunity for GTP to bind (39). Since most Rho proteins exhibit biological activity only when in the GTP-bound state, RhoGEFs, such as Dbs, are primarily thought to be Rho activators.

The region of structural similarity that defines members of the RhoGEF family consists of an approximately 200-amino-acid Dbl homology (DH) domain that is always arranged in tandem with a pleckstrin homology (PH) domain (39). The DH domain contains most of the residues that are required for RhoGEF catalytic activity, while the PH domain is generally perceived to be noncatalytic in nature. Evidence from a variety of sources suggests that the PH domain may perform several functions in the context of a DH domain, which could include structuring the DH domain (41), enhancing catalytic activity (20), and promoting translocation to the plasma membrane (38, 40). The sites within the plasma membrane that interact with PH domains probably contain phosphoinositides, since these are known ligands for the PH domains of a variety of RhoGEFs (10, 20, 26, 34).

Although the RhoGEFs will utilize only members of the Rho subfamily of small GTPases as substrates, they can exhibit varying degrees of target specificity within the family. For example, FGD1 and Cdc24 preferentially use Cdc42 as a substrate (42, 43) and Trio and Tiam1 interact only with Rac1 (5, 13, 42, 43), while Lfc, Lsc, Lbc, and Net1 are able to exchange only on RhoA (2, 8). In contrast, Dbs and Dbl can activate both RhoA and Cdc42 (but not Rac1) (40, 44), while Vav1 and Vav2 are able to activate RhoA, Rac1, and Cdc42 (1, 9, 33). Whereas most of these specificity studies have been performed in vitro with purified exchanger and substrate, it has not yet been determined whether these RhoGEFs exhibit equivalent specificities in vivo.

The deregulated expression of many members of the RhoGEF family, including Dbs, causes tumorigenic growth and promotes the invasive potential of a variety of cell types (39). This is particularly evident in NIH 3T3 mouse fibroblasts, in which deregulated expression of RhoGEFs is often associated with loss of contact inhibition, growth factor independence, anchorage-independent growth, and tumorigenicity in nude mice. Since guanine nucleotide exchange is the only activity that has been demonstrated by many of the oncogenic RhoGEFs and structure-function analyses of RhoGEFs generally indicate a precise convergence between catalytic and transforming activities (11, 40), it is assumed that Rho proteins mediate RhoGEF transformation. Consistent with this, dominant active mutants of several Rho family members (including RhoA, Rac1, and Cdc42) have been shown previously to harbor substantial oncogenic potential when expressed in a variety of cell types including NIH 3T3 cells (3, 17, 22-24, 36). However, activated derivatives of Rho GTPases often have much-diminished transforming activity relative to the oncogenic RhoGEFs, suggesting that the combined activities of multiple GTPase substrates may be required to account for the full transforming activity of a RhoGEF (19).

Like many members of the RhoGEF family, Dbs is transforming when expressed in NIH 3T3 cells as measured by loss of contact inhibition and growth factor independence (37, 40). We have shown previously that Dbs can catalyze the exchange of GDP for GTP on both Cdc42 and RhoA, but not Rac1, in an in vitro system (40). Since activated derivatives of both Cdc42 and RhoA are known to be transforming in NIH 3T3 cells, it is likely that one or both of these GTPases contribute to Dbs transforming activity. However, the relative contribution of RhoA and Cdc42 to any cellular activities associated with Dbs, including transformation, is unclear.

We have recently reported the structures of Dbs in complex with Cdc42 (Dbs·Cdc42) (29) and RhoA (Dbs·RhoA) (35). In this study, we have used these structures as a molecular framework to begin to probe the rules governing DH domain specificity. Our analysis has allowed us to introduce point mutations into the catalytic domain of Dbs that selectively narrow its specificity of exchange. Expression of these mutants in NIH 3T3 cells reveals a precise correlation between Dbs transforming activity and its ability to utilize RhoA (but not Cdc42) as a substrate. This is consistent with observed increases in RhoA-GTP but not Cdc42-GTP in Dbs-transformed cells and is further confirmed by our ability to block Dbs transformation with a competitive inhibitor of RhoA. These studies provide valuable insights into the rules governing substrate recognition by RhoGEFs and important genetic tools that can be used in the analysis of RhoGEF function.

MATERIALS AND METHODS

Molecular constructs.

The pAX142 and pCTV3H mammalian expression vectors have been described previously (37). pAX142-dbs-HA6, pCTV3H-dbs-HA6, and pCTV3H-dbl-HA1 contain cDNAs that encode transforming derivatives of the respective GEFs fused to a hemagglutinin (HA) epitope tag (40). A pET28a (Novagen) bacterial expression construct that contains a carboxyl-terminal His₆-tagged, Dbs DH-PH domain (murine, residues 623 to 967) has been described previously (40). cDNAs encoding human placental Cdc42 (residues 1 to 188; C188S) and Rac1 (residues 1 to 189; C189S) were prepared by PCR and inserted into pET21a (Novagen) for bacterial expression. A cDNA encoding human RhoA (residues 1 to 190; C190S) was subcloned into the pProEX HT vector (Life Technologies) for bacterial expression. Site-directed substitutions in the Dbs DH-PH domain and Cdc42 constructs were prepared with the Quickchange site-directed mutagenesis kit (Stratagene) per the manufacturer's instructions. Site-directed mutagenesis was also used to introduce equivalent point mutations into the DH domain of the pAX142 and pCTV3H versions of Dbs-HA6. cDNA sequences of all expression constructs were verified by automated sequencing. pAX142-rac1, pAX142-rac1(61L), pAX142-rhoA, pAX142-rhoA(63L), pAX142-cdc42, and pAX142-cdc42(12V) have been described previously (40, 45). GST-PBD and GST-C21 contain the Rho binding domains from the Cdc42/Rac1 effector protein PAK3 and the RhoA effector protein Rhotekin, respectively (7, 28). pAX142-GST-C21 and pCTV3H-GST-C21 were made by transferring the GST-C21 sequences (27, 28) into the pAX142 and pCTV3H expression vectors, respectively. The (SREm)₂-luc construct utilized in the transcriptional assays has been described previously (36).

Cell culture, transfection, and transformation assays.

NIH 3T3 cells were maintained in Dulbecco's modified Eagle's medium (high glucose) supplemented with 10% bovine calf serum (BCS; JRH, Lenexa, Kans.). Primary focus formation assays were performed in NIH 3T3 cells exactly as described previously (4). Briefly, NIH 3T3 cells were transfected by calcium phosphate coprecipitation in conjunction with a glycerol shock. Focus formation was scored at 14 days.

NIH 3T3 cells that stably express pCTV3H, pCTV3H-dbs-HA6, pCTV3H-dbl-HA1, or the pCTV3H versions of the Dbs specificity mutants were generated by calcium phosphate coprecipitation followed by selection for 6 days in growth medium (10% BCS) supplemented with hygromycin B (200 μg/ml). Cell lines were then split 1:4, grown in the absence of selection until they reached 60% confluence (approximately 4 days), and then frozen. All affinity precipitation and transformation assays were performed with freshly thawed, single-passage cell lines.

Secondary focus formation assays and assays for anchorage-independent growth were performed with single-passage stable cell lines that had been maintained under subconfluent conditions. For secondary focus assays, 10³ stably selected cells were mixed with 10⁶ untransfected NIH 3T3 cells and then plated on 60-mm-diameter dishes. Foci were scored at 7 days. Anchorage-independent growth was measured as described previously (4). Briefly, cell lines were seeded at 10⁴ cells per 60-mm-diameter dish in growth medium containing 0.3% agar over a base layer of 0.6%. Colonies were counted after 21 days. All assays for transformation were performed in triplicate.

Transient-expression reporter gene assays.

For transient-expression reporter assays, NIH 3T3 cells were transfected by calcium phosphate coprecipitation as described previously (4). Cells were allowed to recover for 30 h and were then starved in Dulbecco's modified Eagle's medium that was supplemented with 0.5% serum for 14 h before lysate preparation. Analysis of luciferase expression was as described previously with enhanced chemiluminescent reagents and a Monolight 3010 luminometer (Analytical Luminescence, San Diego, Calif.) (12, 25). β-Galactosidase activity was determined with Lumi-Gal substrate (Lumigen, Southfield, Mich.) according to the manufacturer's instructions. All assays were performed in triplicate.

Protein expression and purification.

Protein expression in stably transfected NIH 3T3 cells was determined by Western blot analysis as described previously (38). Protein was visualized with enhanced chemiluminescence reagents (Amersham Pharmacia, Piscataway, N.J.). Protein expression and purification of Dbs DH-PH domain proteins were performed as described previously (29). Human placental Cdc42 (residues 1 to 188; C188S) and Rac1 (residues 1 to 189; C189S) were expressed and purified similarly as described previously (29). Human RhoA (residues 1 to 190; C190S) was expressed in Escherichia coli strain BL21(DE3) with a pProEX HT vector (Life Technologies) which encodes an amino-terminal His₆ tag. The His₆ tag was removed by digestion with tobacco etch virus protease, and RhoA was further purified as described previously (29).

Guanine nucleotide exchange assays.

Fluorescence spectroscopic analysis of N-methylanthraniloyl(mant)-GTP incorporation into bacterially purified Rho GTPases was carried out with a Perkin-Elmer LS 50B spectrometer at 20°C. Exchange reaction assay mixtures containing 20 mM Tris (pH 7.5), 50 mM NaCl, 10 mM MgCl₂, 1 mM dithiothreitol, 50 μg of bovine serum albumin/μl, 10% (vol/vol) glycerol, 400 nM mant-GTP (Biomol), and 2 μM relevant GTPase were prepared and allowed to equilibrate with continuous stirring. After equilibration, Dbs DH-PH domain proteins were added at 100 nM, and the relative fluorescence (λ_ex = 360 nm, λ_em = 440 nm) was monitored.

Cdc42, Rac1, and RhoA activation assays.

All affinity purification assays were performed on cell lysates derived from single-passage stable cell lines that had been established and maintained in subconfluent conditions. The p21 binding domains of Pak3 (GST-PBD) (7) or Rhotekin (GST-C21) (28) were expressed as glutathione S-transferase (GST) fusions in BL21(DE3) cells and immobilized by binding to glutathione-coupled Sepharose 4B beads (Amersham Pharmacia). The immobilized Rho binding domains were then used to precipitate activated GTP-bound Rac1, Cdc42 (GST-PBD), or RhoA (GST-C21) from NIH 3T3 cell lysates. Cells were washed in cold phosphate-buffered saline and then lysed in 50 mM Tris-HCl (pH 7.4)-2 mM MgCl₂-100 mM NaCl-10% glycerol-1% NP-40-1 μg of leupeptin/ml-1 μg of pepstatin/ml-1 μg of aprotinin/ml-1 μg of phenylmethylsulfonyl fluoride/ml. Cell lysates were then cleared by centrifugation at 10,000 × g for 10 min at 4°C. The expression of proteins was confirmed by Western blotting prior to affinity purification. Lysates used for affinity purification were normalized for endogenous Rho levels. Affinity purifications were carried out at 4°C for 1 h, and the mixtures were washed three times in an excess of lysis buffer and then analyzed by Western blotting. Cdc42, RhoA, and Rac1 were detected by monoclonal antibodies (Cdc42 and RhoA [Santa Cruz Biotechnology, Santa Cruz, Calif.] and Rac1 [BD Transduction Laboratories, Los Angeles, Calif.]).

RESULTS

Dbs-HA6 activates RhoA, but not Cdc42, in stably transformed NIH 3T3 cells.

We have shown previously that Dbs is highly transforming when stably expressed in NIH 3T3 cells as measured by loss of contact inhibition and growth in low serum (40). Since activated derivatives of both RhoA and Cdc42 are also known to be transforming when stably expressed in NIH 3T3 cells, both are potential mediators of Dbs transforming activity. To determine which (if either) of these two substrates is contributing to Dbs-mediated transformation, we wished to measure levels of endogenous RhoA-GTP, Cdc42-GTP, and Rac1-GTP in NIH 3T3 cells that stably express an activated derivative of Dbs. Dbs-HA6 is an oncogenic version of the Dbs protein that includes intact DH and PH domains along with some flanking sequences (40). Because Dbs and Dbl share identical in vitro exchange profiles, we also assayed an oncogenic derivative of Dbl (Dbl-HA1) for comparison (40). To avoid the risk of characterizing clonal variants, four independent matched pairs of stable cell lines were established for vector and Dbs-HA6 (designated Dbs-HA6-S1, -S2, -S3, and -S4), while an additional fifth matched pair was established for Dbl-HA1. Surprisingly, all the cell lines that express either Dbs-HA6 or Dbl-HA1 showed elevated levels of RhoA-GTP relative to vector controls but not Cdc42-GTP or Rac1-GTP (Fig. 1A and Table 1). To determine whether lack of Cdc42 and Rac1 activation could be attributed to saturation of the pull-down assay, we serum starved the cells for 18 h prior to lysate collection and then repeated the assays. Even though the endogenous levels of Rac1-GTP and Cdc42-GTP were reduced in response to the serum deprivation, we were still unable to observe any elevation in Rac1-GTP or Cdc42-GTP levels in the Dbl-HA1- or Dbs-HA6-transformed cell lines (Fig. 1A and Table 1). As an additional verification that our assay was sufficiently sensitive to detect changes in the endogenous pool of activated Cdc42 and Rac1, we transiently transfected NIH 3T3 cells with Cdc42(12V) (Fig. 1B) or Rac1(61L) (Fig. 1C). Although the level of expression of both the Cdc42(12V) and Rac1(61L) mutants was undetectably low in whole-cell lysates (lower panels in Fig. 1B and C), we were still able to affinity precipitate both and demonstrate their presence in the Cdc42-GTP and Rac1-GTP pools (upper panels in Fig. 1B and C). Thus, the lack of Cdc42 and Rac1 activation in our stable assays cannot be attributed to a lack of sensitivity of the affinity precipitations.

FIG. 1.

Elevated levels of RhoA-GTP in NIH 3T3 cells stably transfected with a transforming derivative of Dbs. (A) Lysates were collected from NIH 3T3 cell lines that stably express either Dbs-HA6, Dbl-HA1, or the cognate pAX142 vector. Lysates were examined by Western blotting for expression of RhoA (Total RhoA), Cdc42 (Total Cdc42), Rac1 (Total Rac1), and the transforming HA-tagged derivatives (anti-HA). Lysates were then split into three parts, each of which was normalized for expression of either RhoA, Rac1, or Cdc42. Each lysate was then subjected to affinity purification with immobilized GST-PBD (Rac1 and Cdc42) or GST-C21 (RhoA). GTP-bound Cdc42 (GTP-Cdc42) and Rac1 (GTP-Rac1) that were precipitated with GST-PBD or GTP-bound RhoA (GTP-RhoA) that was precipitated with GST-C21 were visualized by Western blotting. For the Rac1 and Cdc42 assays affinity purifications were performed on cells that had been maintained under normal serum levels (10%) or serum starved (0.5%) for 18 h prior to lysate collection. (B and C) NIH 3T3 cells were transiently transfected with plasmids that encoded activated derivatives of Cdc42 (B) or Rac1 (C). Lysates were collected at 48 h, and affinity purifications were performed as described above. All experiments were performed a minimum of three times, and data shown constitute a representative data set.

TABLE 1.

Rho-GTP loading in NIH 3T3 cell lines that stably express Dbs-HA6 or Dbl-HA1^a

Cell line	RhoA	Rac1b		Cdc42
		10% BCS	0.5% BCS	10% BCS	0.5% BCS
Dbs-HA6-S1	5.1 ± 0.8	1.1 ± 0.1	1.1 ± 0.1	1.0 ± 0.1	0.9 ± 0.1
Dbs-HA6-S2	8.2	1.1	0.9	1.3	0.8
Dbs-HA6-S3	3.8	1.0	1.3	1.2	1.2
Dbs-HA6-S4	5.6	1.2	1.1	1.2	1.3
Dbl-HA1	5.3 ± 0.7	1.3 ± 0.2	1.2 ± 0.1	1.1 ± 0.1	1.0 ± 0.1

^aRho-GTP levels were determined by phosphorimager analysis and are expressed as fold activation relative to matched vector cell lines. For the Dbs-HA6-S1 and Dbl-HA1 cell lines assays were repeated a minimum of four times and data are presented ± standard deviations. For the remaining cell lines assays were performed twice for each condition and a representative data set is shown.

^bRac1-GTP and Cdc42-GTP levels were determined under normal serum conditions (10% BCS) or after 18 h of serum starvation (0.5% BCS).

To determine whether Dbs and Dbl transformation is associated with a transient activation of Cdc42 or Rac1, affinity precipitation assays were also performed on NIH 3T3 cells that were transiently transfected with Dbl-HA1 or Dbs-HA6. Although both proteins failed to generate any detectable increases in Cdc42-GTP or Rac1-GTP levels when expressed transiently in NIH 3T3 cells, we were unable to detect expression of the proteins in the whole-cell lysates (data not shown). Thus, it may not be possible to achieve sufficiently high transient expression of Dbl-HA1 and Dbs-HA6 in NIH 3T3 cells to have any measurable effects on Cdc42 or Rac1 GTP levels. To summarize, although we cannot formally exclude the possibility that transient activation of Cdc42 or Rac1 may contribute to the initiation of Dbs- and Dbl-mediated transformation, neither GTPase seems to make a contribution to the maintenance of the transformed phenotype.

Altering substrate specificity of the Dbs DH domain.

Although we did not observe any increases in Cdc42-GTP in Dbs-transformed NIH 3T3 cells, it is possible that Cdc42 is cycling more rapidly in these cells and that this accelerated cycling is not being reflected in the Cdc42-GTP steady-state levels. Alternatively, Dbs transformation may be associated with a transient activation of Cdc42 that we are unable to measure with our affinity precipitations. To address these concerns, it was necessary to develop an alternative and complementary approach to determine the relative contributions of Cdc42 and RhoA to the transforming activity of Dbs. Towards this end, we wished to introduce mutations into the Dbs DH domain that would act to uncouple activation of RhoA and Cdc42. Such mutants would be valuable genetic tools that could be used to determine the relative contribution of each GTPase to the various in vivo functions of Dbs, including transformation. We have recently reported structures of several RhoGEFs in complex with their cognate GTPases (29, 35, 41), and based on these analyses, we are able to test specific predictions regarding the molecular bases for discrimination of Rho GTPases by DH domains.

As proposed for Tiam1·Rac1 (41), the α4-α5a region of the Dbs DH domain likely recognizes a specificity patch on β-strands 1 to 3 of Rho GTPases (Thr 3, Ala 41, Thr 43, Thr 52, and Phe 56 in Cdc42) to discriminate between Rho GTPases (Fig. 2A and B). Importantly, Trp 56 of Rac1, analogous to Phe 56 in Cdc42 and Trp 58 in RhoA, has recently been reported to be a primary determinant mediating DH domain specificity toward Rho GTPases (6, 15). For example, the Rac1-to-Cdc42 alteration (W56F) abolishes Rac1 activation by Tiam1 and allows activation by Intersectin, a Dbl family member specific for Cdc42 (15). Similarly, the structure of Dbs·Cdc42 highlights the insertion of Phe 56 of Cdc42 into a cleft of Dbs formed by leucines 759 and 766 to generate favorable hydrophobic packing interactions that promote binding (Fig. 2A). Docking a model of Rac1 upon the Dbs DH domain suggests that Rac1 residues Ser 41, Asn 43, and Asn 52 would likely be tolerated in a Dbs·Rac1 complex, although the bulkier side chain of Trp 56 would sterically impinge upon L766 of Dbs to hamper binding and exchange and may explain why Dbs cannot support exchange upon Rac1 (data not shown).

FIG. 2.

Specificity determination within the structures of Dbs·Cdc42 and Dbs·RhoA. (A) A ribbon diagram of the Dbs DH domain (yellow) bound to a GTPase (green) highlighting the specificity patch on β-strands 1 to 3 (magenta) of Rho GTPases and the complementary interface on DH domains (α4-α5a region; cyan). (B) Close-up of the interactions within the specificity patch for Dbs·Cdc42 (Protein Data Bank accession no. 1KZ7) highlighting the hydrophobic pocket formed by leucines 759 and 766 of Dbs and encompassing Phe 56 of Cdc42. (C) Within the equivalent region of Dbs·RhoA (Protein Data Bank accession no. 1LB1), the hydrophobic pocket adjusts to accommodate Trp 58 of RhoA. Val 43 and Ala 56 of RhoA also participate favorably with the hydrophobic pocket. Also shown is the electrostatic interaction of Lys 758 of Dbs with Asp 45 and Glu 54 of RhoA.

RhoA also possesses a tryptophan (Trp 58) equivalent to position 56 of Rac1, raising the question of how Dbs catalyzes nucleotide exchange on RhoA but not Rac1. Within the structure of Dbs·RhoA (Fig. 2C), Lys 758 of Dbs is located between Asp 45 and Glu 54 of RhoA (Asn 43 and Asn 52 in Rac1 and Thr 43 and Thr 52 in Cdc42) to form significant electrostatic interactions not possible with either Rac1 or Cdc42. In addition, Leu 759 of Dbs is located between Val 43 and Ala 56 in RhoA (Ser 41 and Gly 54 in Rac1 and Ala 41 and Gly 54 in Cdc42), resulting in a more extensive hydrophobic interface than possible with Rac1 or Cdc42. The additional binding energy afforded by these sets of interactions could compensate for a seemingly poor fit of Trp 58 of RhoA into the Dbs DH domain cleft, allowing Dbs binding and exchange of RhoA.

Based on our structures of Dbs·Cdc42 and Dbs·RhoA, we have mutated Dbs to alter its specificity for Cdc42 relative to RhoA (Fig. 3 and Table 2). Mutations were placed primarily at positions 759 and 766 of Dbs (Fig. 2) to remodel the binding cleft normally sequestering Phe 56 of Cdc42 or Trp 58 of RhoA. Substitution of either methionine or isoleucine for Leu 759 of Dbs indiscriminately crippled nucleotide exchange on both Cdc42 and RhoA. However, substitution of methionine or isoleucine for Leu 766 (analogous to Ile 1187 in Tiam1) conferred on Dbs an enhanced ability to exchange RhoA while showing wild-type ability (L766I) or a diminished ability (L766M) to exchange Cdc42. Furthermore, both mutations allowed Dbs to support exchange upon Rac1, with Dbs (L766I) exhibiting robust activity toward Rac1 (Fig. 3C). Therefore, L766M and L766I in Dbs likely enlarge the binding cleft to accommodate Trp 56 of Rac1 and Trp 58 of RhoA. In support of the idea that Dbs (L766I) relieves steric crowding within the binding pocket formed by leucines 759 and 766, Cdc42(F56W) is a poor substrate for wild-type Dbs but an exceedingly robust substrate for Dbs(L766I) (Fig. 3J). We caution, however, that the equivalent of position 56 cannot be the sole determinant of specificity utilized by DH domains, since Cdc42(F56W) is not exchanged by the DH-PH domains of Tiam1 (data not shown and reference 15).

FIG. 3.

Mutations within the Dbs DH domain that alter specificity toward Rho GTPases. Substitutions within the DH domain of Dbs differentially affect its ability to activate Cdc42 (A, D, and G), RhoA (B, E, and H), and Rac1 (C, F, and I). L766I within the DH domain of Dbs effectively compensates for the substitutions for F56W within Cdc42 (J). For each reaction, 2 μM GTPase was preincubated with 400 nM mant-GDP for 300 s prior to the addition of catalytic amounts (100 μM) of wild-type (wt; gray traces) or mutant Dbs as indicated. Spontaneous exchange rates are also shown (black traces) and indicated with the names of the relevant GTPases. RhoGEF addition is defined as the experimental start (time zero).

TABLE 2.

Initial rates of guanine nucleotide exchange reactions catalyzed by wild-type and mutant Dbs proteins on Cdc42, RhoA, and Rac1^a

Protein designation	Cdc42		RhoA		Rac1		Cdc42 (F56W)
	(RF s−1)	Fold activation	(RF s−1)	Fold activation	(RF s−1)	Fold activation	(RF s−1)	Fold activation
WT	3.95	49	1.44	48	0.02	1	0.65	8
L759M	1.94	24	0.44	15	0.02	1
L759I	0.24	3	0.17	6	0.02	1
L766M	0.50	6	3.02	101	0.08	4
L766I	4.39	55	8.00	267	1.13	57	6.45	81
Q752E	2.05	26	1.62	54	0.02	1
L759M/L766M	0.25	3	1.35	45	0.05	3
L759I/L766M	0.07	1	0.38	13	0.02	1
Q752E/L766M	0.31	4	2.61	87	0.08	4

^aThe initial rates of guanine nucleotide exchange for 2 μM wild-type (WT) GTPase stimulated by 0.1 μM concentrations of various Dbs proteins were determined by linear regression analysis of approximately the first 60 s of the Dbs-catalyzed reactions from Fig. 3. The fold activation for each Dbs protein reflects the ratio of the initial exchange rate measured for the GEF-containing reaction to the intrinsic rate of exchange for each GTPase. Intrinsic value of exchange: Cdc42, 0.08; RhoA, 0.03; Rac1, 0.02; and Cdc42 (F56W), 0.08 relative fluorescence (RF) units per second.

Substitution of glutamate (Q752E) for Gln 752 within the Dbs DH domain was hypothesized to bolster Dbs activity toward RhoA by encouraging electrostatic interactions with Arg 5 of RhoA (Fig. 2C). Indeed, a small increase in RhoA activation by Dbs (Q752E) is observed along with reduced exchange of Cdc42, indicating that a glutamate at position 752 confers increased RhoA specificity on Dbs (Fig. 3D to F).

The functional effects of double mutations within a protein are, in general, equivalent to the sums of the effects of the constituent single substitutions (30). To generate mutants that further impair the ability of Dbs to activate Cdc42, we separately combined the single substitutions Q752E, L759M, and L759I with L766M and assessed their effects upon Dbs-catalyzed exchange for Cdc42, RhoA, and Rac1 (Fig. 3G to I; Table 2). All three doubly substituted Dbs proteins showed further decreases in their ability to activate Cdc42, with Dbs (L759I/L766M) being completely inactive toward Cdc42. Dbs (L759M/L766M) and Dbs (Q752E/L766M) retained wild-type and greater-than-wild-type activity toward RhoA, respectively, while Dbs (L759I/L766M) showed a marked decrease in the ability to activate RhoA. None of the three doubly substituted Dbs proteins exhibited significant activity toward Rac1.

Determination of substrate usage by Dbs specificity mutants in NIH 3T3 cells.

Having generated Dbs mutants that are selectively impaired in their ability to catalyze exchange on Cdc42, we wished to use these mutants to determine the contribution of Cdc42 to Dbs transformation in NIH 3T3 cells. For this analysis, a subset of six Dbs mutants were selected to reflect the various ratios of catalyzed activation possible between Cdc42 and RhoA: Q752E/L766M, L759M/L766M, and L766M engender Dbs with elevated exchange on RhoA while significantly reducing exchange on Cdc42; L766I increases RhoA exchange while not perturbing exchange on Cdc42; and L759I and L759I/L766M significantly reduce exchange on both GTPases. Initially we wanted to determine whether the in vitro exchange activity of our panel members could be recapitulated in NIH 3T3 cells. Towards this end, all of the mutations were placed into the background of Dbs-HA6 and stable NIH 3T3 cell lines were constructed that expressed each of the mutants (Fig. 4). To avoid the risk of analyzing a clonal variant, two independent polyclonal cell lines were established for each cell line. Although Fig. 4 shows the results of a single matched set of cell lines, we observed identical results with the second matched set (data not shown). Expression of all mutants was verified by Western blotting with an anti-HA antibody and was found to be equivalent for all members of the panel (Fig. 4, bottom panel). Lysates were then collected and subjected to affinity precipitation to measure levels of activated RhoA and Cdc42 (Fig. 4). The levels of RhoA-GTP that we observed were in general accordance with what we had observed in our in vitro assays. The two cells lines that express mutants that are defective in RhoA exchange (L759I and L759I/L766M) contained very low levels of RhoA-GTP compared to Dbs-HA6, while the remainder of the cell lines contained RhoA-GTP levels that were equivalent to that of the Dbs-HA6 cell line. The enhanced RhoA exchange that we observed in vitro with the L766I, L766M, and Q752E/L766M mutants was not reflected in measurable increases in steady-state levels of RhoA-GTP in vivo. This may reflect an inability to detect the modest increases of steady-state levels of RhoA-GTP that are associated with these mutants (Fig. 3B and H). As expected, no changes were detected in endogenous Cdc42-GTP levels for any of the mutants, which is consistent with our previous observation that Dbs-HA6 does not activate Cdc42 in stable NIH 3T3 cell lines. This also suggests that the decreased levels of RhoA-GTP observed in the L759I and L759I/L766M cell lines cannot be attributed to a nonspecific overall reduction in Rho-GTP levels in these cell lines. Since the L766I mutant had acquired the ability to exchange on Rac1 in vitro, we also examined this cell line for Rac1-GTP levels and observed no difference relative to the wild-type control (data not shown). To address the possibility that Cdc42-GTP and Rac1-GTP levels are saturated under 10% serum conditions, assays were repeated following serum starvation (0.5% BCS) for 18 h with no significant differences observed (data not shown).

FIG. 4.

Substrate specificities of Dbs mutants are confirmed in vivo. GTP-RhoA and GTP-Cdc42 levels were measured in NIH 3T3 cells stably transfected with Dbs mutants. Lysates were collected from NIH 3T3 cell lines that stably express either Dbs-HA6, Dbs-HA6 mutants, or the cognate pAX142 vector. Lysates were examined by Western blotting for expression of RhoA (Total RhoA), Cdc42 (Total Cdc42), and the HA-tagged Dbs mutants (anti-HA). Lysates were then split into two parts, each of which was normalized for expression of either RhoA or Cdc42. Each lysate was then subjected to affinity purification with immobilized GST-PAK (Cdc42) or GST-C21 (RhoA). GTP-bound Cdc42 (GTP-Cdc42) that was precipitated with GST-PAK or GTP-bound RhoA (GTP-RhoA) that was precipitated with GST-C21 were visualized by Western blotting. All experiments were performed a minimum of three times on two independent sets of polyclonal cell lines. Data shown constitute a representative data set.

Dbs transforming activity in NIH 3T3 cells correlates with endogenous levels of RhoA-GTP.

The altered specificity of the Dbs mutants for their substrates, both in vitro and in vivo, suggested that they may be useful tools to examine the relative contributions of RhoA and Cdc42 to Dbs transforming activity. It has been shown previously that Dbs has transforming activity in NIH 3T3 cells as measured by primary focus formation assays (37, 40). To determine whether this parameter of transformation was impaired in our mutants, a primary focus formation assay was performed with the complete panel (Fig. 5A). Focus formation is observed in NIH 3T3 cells at approximately 10 to 14 days posttransfection, which is the point at which we are able to detect elevated RhoA-GTP in our stable cell lines. An examination of the transforming potential of the three Cdc42-defective mutants (Q752E/L766M, L759M/L766M, and L766M) indicates that there is no meaningful relationship between Cdc42 activation and transforming activity. Whereas all are substantially impaired in Cdc42 exchange, all retain wild-type levels of transforming activity. In contrast, there appears to be a more precise correlation between transformation and RhoA activation. The L766I mutant is selectively enhanced in RhoA exchange activity and exhibits increased transformation, while the two mutants that are impaired in RhoA exchange (L759I and L759I/L766M) are impaired in transformation. Our observation that the L766I mutant exhibits increased transforming activity in the absence of any discernible increase in steady-state RhoA-GTP levels may reflect increased cycling associated with endogenous RhoA. Alternatively, it may reflect the ability of this mutant to exchange on Rac1, although we also did not observe any increase in steady-state levels of Rac1-GTP in this cell line. To summarize, Dbs transforming activity in NIH 3T3 cells correlates more closely with its ability to activate RhoA than with its ability to activate Cdc42.

FIG. 5.

Dbs transforming activity correlates with in vivo activation of RhoA. The transforming activity of the Dbs specificity mutants was compared in primary focus assays (A), secondary focus assays (B), and soft agar assays (C). Assays were performed as described in Materials and Methods. For primary focus assays NIH 3T3 cells were transfected with 3 μg of plasmid DNA for each condition. Data shown are representative of three assays performed on triplicate plates.

It has been argued previously for the Dbl protein that different parameters of transformation in NIH 3T3 cells are mediated by different Rho GTPase targets (19). For example, it has been suggested elsewhere that loss of contact inhibition (which is measured by primary and secondary focus formation assays) may be mediated by RhoA, while anchorage-independent growth (as measured by growth in soft agar) may be mediated by Cdc42 (19). To determine whether this also applies to Dbs, we performed secondary focus formation assays (Fig. 5B) and soft agar assays (Fig. 5C) with the NIH 3T3 cell lines that stably express the Dbs DH domain mutants. In both assays the transformation profile of the panel was similar to what we observed with the primary focus formation assay. This strongly suggests that the loss of contact inhibition and the anchorage-independent growth associated with Dbs are unrelated to its ability to exchange on Cdc42 but rather correlate with its activity toward RhoA. In contrast with the primary focus assay, the L759I/L766M mutant exhibited some transforming activity in the secondary focus and soft agar assays. This is consistent with the modest RhoA activation that is observed with this mutant in the in vitro exchange assays (Fig. 3H) and may reflect greater sensitivity associated with these two transformation assays. Surprisingly, we did not observe any significant increases in transforming activity (relative to the Dbs-HA6 control) in the soft agar assay with any of the mutants. This may indicate that there is an upper threshold of activity that can be achieved in this assay with activated endogenous RhoA. Collectively, these results further support the hypothesis that Dbs transformation in NIH 3T3 cells is unrelated to Cdc42 activation and further implicate RhoA as the more meaningful in vivo substrate in this cell type.

A competitive inhibitor of RhoA blocks Dbs transformation.

Although we have established a good correlation between RhoA activation and Dbs-mediated transformation of NIH 3T3 cells, we wanted to make a more direct demonstration that RhoA activation is necessary for the transformed phenotype. Toward this end, we wished to determine whether a competitive inhibitor of RhoA function could block Dbs transformation. Although dominant-inhibitory mutants of RhoA have been described which could be used for such an analysis, such mutants have relatively low utility since they can bind to multiple RhoGEFs in any given cell type, thus potentially blocking additional signaling pathways (21, 22). In contrast, the GST-C21 fusion protein that we utilize for our GST pull-down assays interacts specifically with the GTP-bound forms of RhoA and RhoC (32). Since RhoC is not expressed in NIH 3T3 cells (unpublished observations), this reagent should function as a highly specific inhibitor of RhoA in this cell type. This high degree of target specificity selectively associated with Rho binding domains has been exploited previously to generate inhibitors of Rho proteins that have a much higher degree of target specificity (21, 22). With this strategy in mind we transferred GST-C21 into the pAX142 mammalian expression vector and confirmed its stable expression by transient transfection and Western blotting in COS-7 cells (Fig. 6A). To confirm that GST-C21 can block RhoA-mediated signaling events in NIH 3T3 cells, we performed a transient reporter assay with the (SREm)₂-luc reporter and either RhoA(63L), Cdc42(12V), or Dbs-HA6 in the presence or absence of the inhibitor (Fig. 6B). RhoA(63L) is a GTPase-defective, constitutively activated derivative of RhoA that should form nonproductive interactions with the GST-C21 inhibitor. Cdc42(12V) is a constitutively activated derivative of Cdc42 that should not interact with the inhibitor and is included as a control for specificity. As predicted, reporter activation by RhoA(63L) was substantially reduced in the presence of the inhibitor (>60%) while activation by Cdc42(12V) was unaffected. Activation of the reporter by Dbs-HA6 was also substantially impaired in the presence of the inhibitor (>80%), suggesting that Dbs activation of the (SREm)₂-luc reporter is occurring in a RhoA-dependent manner. We then determined whether the inhibitor could block transformation by Dbs-HA6. Dbs-HA6 exhibited substantially reduced transforming activity (>60%) in the presence of the inhibitor in an NIH 3T3 cell primary focus formation assay (Fig. 6C). This reduction could not be attributed to nonspecific growth inhibition, since the inhibitor, when expressed alone, did not exhibit any growth inhibition relative to vector controls (Fig. 6D). These observations support the model that the activated RhoA that we are able to detect in stably transfected NIH 3T3 cells is contributing to Dbs-HA6 transforming activity.

FIG. 6.

Suppression of Dbs-mediated transformation by a genetic inhibitor of RhoA function. (A) Expression of GST-C21 in transiently transfected COS-7 cells was determined with an anti-GST mouse monoclonal antibody (SC138; Santa Cruz). (B) GST-C21 blocks activation of serum response factor by RhoA. NIH 3T3 cells were cotransfected with 3 μg of either pAX142 (vector), pAX142-rhoA(63L), pAX142-cdc42(12V), or pAX142-dbs-HA6 and 3 μg of pAX142-GST-RBD, along with 2.5 μg of (SRE_m)₂-luc, and 500 ng of pCMVnlac as an internal control for transfection efficiency and/or growth inhibition. Luciferase and β-galactosidase levels were measured and expressed as fold activation relative to the level of activation seen with empty vector control. Luciferase activity was then standardized relative to β-galactosidase activity. Data shown are representative of three independent experiments performed on triplicate plates. Error bars indicate standard deviations. (C) GST-RBD blocks focus formation by Dbs-HA6. NIH 3T3 cells were cotransfected with 3 μg of pAX142-dbs-HA6 and 3 μg of either pAX142 or pAX142-GST-RBD. Foci were counted at 14 days. The data presented are representative of three independent experiments performed on triplicate plates. Error bars indicate standard deviations. (D) Clonogenic analyses showed that expression of the GST-C21 plasmid did not cause nonspecific inhibition of cell growth. NIH 3T3 cells were transfected with 3 μg of pCTV3H-GST-C21 or pCTV3H (vector) by calcium phosphate precipitation. At 24 h posttransfection, cells were split 1:4 and selected for 10 days in hygromycin B (200 μg/ml). Plates were then stained, and cell clones were counted.

DISCUSSION

Expression of RhoGEFs in many cell types is often associated with multiple cellular events including transcriptional regulation, entry into the cell cycle, cytoskeletal rearrangements, and oncogenic transformation. Historically, however, the relevant GTPase substrates that mediate these events have been difficult to identify. This difficulty can be generally attributed to a lack of substrate specificity in vitro, coupled with a lack of information about substrate utilization in vivo. Although recently described assays which allow for accurate measurements of Rho-GTP levels in cell lysates have been extremely helpful in addressing the latter deficiency (27, 28), the promiscuous targeting of GTPases by RhoGEFs remains problematic for their analysis. In this study we have described a genetic approach that could be used to determine the relative contributions of GTPase substrates to the various cellular events that are regulated by a single RhoGEF. Using the structures of several RhoGEF·GTPases as a molecular framework, we have been able to accurately predict amino acid substitutions that uncouple the specificity of Dbs for its two in vitro targets, RhoA and Cdc42. In our first application of this approach, we have used these mutations to identify RhoA, but not Cdc42, as an important mediator of Dbs transformation in NIH 3T3 cells. As we refine our understanding of the structural determinants necessary to dictate productive coupling between RhoGEFs and GTPases (35), we anticipate being able to more fully dissect and understand the myriad biological outputs initiated by RhoGEFs operating on the large family of Rho GTPases.

Since detailed structure-function analyses of Dbs have revealed a precise convergence between its catalytic and transforming activities (37, 40), it is presumed to cause deregulated growth in NIH 3T3 cells through the aberrant stimulation of Rho GTPase substrates. Since constitutively activated mutants of RhoA and Cdc42 also exhibit transforming potential when expressed in NIH 3T3 cells (22, 24, 40), they would both appear to be attractive candidates for mediators of Dbs transformation. However, despite the fact that chronic activation of Cdc42 is associated with deregulated growth in NIH 3T3 cells, we and others have made observations that suggest that RhoGEF-mediated transformation of NIH 3T3 cells can occur in a Cdc42-independent manner. For example, several RhoGEFs have been identified which exhibit potent transforming activity in NIH 3T3 cells without exhibiting any measurable exchange activity for Cdc42 in vitro (i.e., Lfc, Lsc, and Net1) (2, 8). In addition, although many RhoGEFs (including Dbs) have been isolated based on their activity in an NIH 3T3 cell primary focus formation assay, Cdc42 does not exhibit this particular parameter of transformation. This suggests that Cdc42 activation alone cannot account for the full transforming activity associated with many RhoGEFs in NIH 3T3 cells. Our present observation that a highly transforming derivative of Dbs does not influence steady-state levels of Cdc42-GTP in stably transfected NIH 3T3 cells further supports the contention that Cdc42 does not contribute to the transforming activity of this GEF. Although it is possible that Cdc42 activation may be required for the initiation but not the maintenance of transformation, our observation that Dbs specificity mutants which lack any appreciable Cdc42 exchange activity can retain full transforming activity argues against such a possibility.

What then is the in vivo target for Dbs? Since RhoA is the only GTPase that has been shown elsewhere to be capable of inducing the formation of foci in an NIH 3T3 primary focus formation assay (16), and RhoA is an in vitro substrate for Dbs, it is a strong candidate for such a target. In this study we observed elevated levels of RhoA-GTP in Dbs-transformed NIH 3T3 cells. Similarly, Dbs mutants that were impaired in RhoA exchange were consistently impaired in their transforming activity. Importantly, a mutant that is selectively enhanced in its RhoA activity exhibited enhanced transformation in both primary and secondary focus formation assays. Finally, a competitive inhibitor that targets RhoA in NIH 3T3 cells can effectively block Dbs transforming activity.

It has been suggested previously that Dbl transformation of NIH 3T3 cells may be accounted for by the simultaneous activation of RhoA, Rac1, and Cdc42 and that each GTPase may regulate a particular aspect of Dbl transformation (19). Since activated RhoA can cause loss of contact inhibition in NIH 3T3 cells, as measured by focus formation assays, while activated Cdc42 can regulate changes in anchorage-independent growth, as measured by growth in soft agar, both may cooperate to account for the Dbs-transformed phenotype. However, NIH 3T3 cells that stably express Dbs exhibit both anchorage-independent growth and loss of contact inhibition, even though they do not contain elevated levels of Cdc42, and Dbs specificity mutants that are deficient in Cdc42 exchange are not deficient in either of these two parameters of transformation. Thus, although it is formally possible that additional GTPases may contribute to Dbs transformation in NIH 3T3 cells, it seems unlikely that Cdc42 is making a contribution. These results are consistent with a recent demonstration that RhoA is downregulated in NIH 3T3 cells that stably express either activated Rac1 or Cdc42, suggesting that the simultaneous activation of RhoA with either Rac1 or Cdc42 in NIH 3T3 cells may not be achievable (31).

Although the bulk of our evidence at present implicates only RhoA as a mediator of Dbs transformation, we cannot exclude the possibility that other GTPases (other than Rac1 or Cdc42) may also make contributions. Dbs can activate transcriptional reporters in COS-7 cells in the absence of any detectable influence on RhoA-GTP and Cdc42-GTP levels, suggesting that it is capable of forming productive interactions with additional GTPases (unpublished observations). One additional GTPase that may contribute to Dbs transformation is RhoC. RhoC is highly related to RhoA structurally and will also form nonproductive complexes with the GST-Rhotekin inhibitor. However, Ost, the rat ortholog of Dbs, is unable to catalyze exchange on RhoC in vitro (14), and we are unable to detect expression of RhoC in NIH 3T3 cells (unpublished observations). Thus, at this time, we have no evidence that any GTPase other than RhoA is contributing to NIH 3T3 transformation by Dbs.