Research ArticleBiochemistry

Mechanisms of autoregulation of C3G, activator of the GTPase Rap1, and its catalytic deregulation in lymphomas

See allHide authors and affiliations

Science Signaling  01 Sep 2020:
Vol. 13, Issue 647, eabb7075
DOI: 10.1126/scisignal.abb7075

Activating Rap1 with mutant C3G in lymphoma

The GTPase Rap1 promotes integrin-mediated cell adhesion and is associated with the development of various cancers, including B cell malignancies. Carabias et al. found two missense mutations in the Rap1 activator C3G in non-Hodgkin’s lymphomas (NHLs) from a database of cancer-associated mutations. Examination of the structure of C3G revealed that the basal autoinhibitory structure of the protein is disrupted by these mutations, enabling constitutive activity that promoted Rap1-integrin signaling independently of the normal activation mechanisms for C3G. Developing ways to target C3G in NHL cells may lead to targeted therapeutics for patients with NHL and possibly other cancers.


C3G is a guanine nucleotide exchange factor (GEF) that regulates cell adhesion and migration by activating the GTPase Rap1. The GEF activity of C3G is stimulated by the adaptor proteins Crk and CrkL and by tyrosine phosphorylation. Here, we uncovered mechanisms of C3G autoinhibition and activation. Specifically, we found that two intramolecular interactions regulate the activity of C3G. First, an autoinhibitory region (AIR) within the central domain of C3G binds to and blocks the catalytic Cdc25H domain. Second, the binding of the protein’s N-terminal domain to its Ras exchanger motif (REM) is required for its GEF activity. CrkL activated C3G by displacing the AIR/Cdc25HD interaction. Two missense mutations in the AIR found in non-Hodgkin’s lymphomas, Y554H and M555K, disrupted the autoinhibitory mechanism. Expression of C3G-Y554H or C3G-M555K in Ba/F3 pro–B cells caused constitutive activation of Rap1 and, consequently, the integrin LFA-1. Our findings suggest that sustained Rap1 activation by deregulated C3G might promote progression of lymphomas and that designing therapeutics to target C3G might treat these malignancies.


Small guanosine triphosphatases (GTPases) act as molecular switches that play key roles in signal transduction. They cycle between an inactive conformation, bound to guanosine diphosphate (GDP), and an active guanosine 5′-triphosphate (GTP)–bound state that binds to downstream effectors. Spontaneous exchange of the bound GDP by GTP is very slow. This conversion is accelerated by guanine nucleotide exchange factors (GEFs) that induce the dissociation of GDP from the GTPase and the binding of GTP, which is more abundant than GDP in the cells.

C3G [Crk Src homology 3 domain (SH3 domain)–binding GEF; also known as RapGEF1] is a GEF that directly activates the small GTPases of the Ras family Rap1a and Rap1b (1, 2) and the GTPase of the Rho family TC10 (3). C3G also moderately activates R-Ras and TC21 (4). C3G is ubiquitously expressed and is required for mouse embryonic development (5). C3G promotes integrin-mediated adhesion through the activation of Rap1 (6). C3G also regulates other cellular functions including migration (7), actin remodeling (8), apoptosis (911), proliferation (12), differentiation (13, 14), and exocytosis (15).

The main splice variant of C3G (isoform a, 1077 residues) has a three-segment structure (16) (Fig. 1A). The N-terminal domain (NTD; residues 1-245) mediates the interaction with E-cadherin (17). The central region (residues 246-670) contains five proline-rich motifs (P0 to P4) that are binding sites for proteins containing SH3 domains; hence, this region is named SH3-binding or SH3b. Crk adaptor proteins bind to motifs P1 to P4 (18, 19); p130Cas binds to the P0 (20); SH3-containing tyrosine kinases Hck (10), c-Abl (8, 21), and the oncoprotein Bcr-Abl (22) also bind to the SH3b. The C-terminal catalytic region consists of a Ras exchanger motif (REM) and a Cdc25 homology domain (Cdc25HD); the latter harbors the GEF activity.

Fig. 1 The NTD of C3G interacts with the REM domain.

(A) Schematic representation of the domain structure of C3G. The constructs used to study the interaction mediated by the NTD are shown aligned underneath. Molecular weights (MW) calculated from the sequence are indicated. (B) Analysis by coimmunoprecipitation (coIP) of the interaction of FLAG-NTD with other regions of C3G (HA-tagged) coexpressed in COS-1 cells. Cell lysates were immunoblotted for HA, FLAG, and β-actin. Proteins were immunoprecipitated with agarose-immobilized antibody against FLAG and were immunoblotted for HA. The asterisk indicates a nonspecific band. Blots are representative of at least two independent experiments.

In resting C3G, noncatalytic regions block the activity of the Cdc25HD, preventing uncontrolled signaling (2325). Other GEFs of the Cdc25 family are also repressed by autoinhibitory mechanisms (26); yet, there is no similarity between the noncatalytic regions of C3G and other GEFs.

C3G-Rap1 signaling is activated by several stimuli, including ligation of B and T cell receptors (27, 28), growth factors (29), cytokines (30), and mechanical cues (31). In unstimulated cells, C3G is associated to Crk proteins in the cytoplasm (27, 32); these adaptor proteins (CrkII and CrkL) contain an Src homology 2 domain (SH2) and two SH3 (SH3N and SH3C) domains; only the SH3N binds to Pro-rich motifs, whereas the noncanonical SH3C does not (33). During activation, the Crk-C3G complex is recruited to the plasma membrane through the interaction of the SH2 domain of Crk with tyrosine-phosphorylated proteins (25). C3G is then phosphorylated in Tyr504 of the SH3b domain and possibly other tyrosine residues (23). C3G is a substrate of kinases Src (34), Hck (10), Fyn (28), c-Abl (35), and Bcr-Abl (22). In addition, Crk also activates C3G directly in cells (23) and in vitro (24), although the mechanisms of activation remain unknown.

C3G has a dual role in cancer, acting as a tumor suppressor or promoter depending on the cellular context. C3G is down-regulated in cervical squamous cell carcinoma (36). On the other hand, C3G is overexpressed in non–small cell lung cancer (37), head and neck squamous cell carcinoma (38), and metastatic epithelial ovarian tumors, in which C3G-Rap1 signaling promotes cell invasion (39). The Crk-C3G-Rap1 pathway is involved in the transformation of papillary thyroid carcinoma (40), and alterations in C3G associate with reduced survival of patients with hepatocellular carcinoma (41). The oncoprotein Bcr-Abl, which causes chronic myeloid leukemia (CML), induces proliferation, survival, and migration of CML cells through the activation of Crk-C3G-Rap1 (42). CML cells abundantly express a truncated variant of C3G named p87C3G, which lacks the NTD, and the levels of p87C3G protein diminish upon treatment with the Bcr-Abl inhibitor imatinib (22). CrkL and C3G are phosphorylated and recruited to signaling complexes in a Bcr-Abl–dependent manner, suggesting that Rap1 activation in CML cells occurs through CrkL-C3G (43). C3G is constitutively phosphorylated in human lymphoblastoid cell lines derived from B lymphocytes transformed with Epstein-Barr virus (EBV) (44), and the RAPGEF1 gene coding for C3G is highly expressed in EBV-transformed lymphocytes [publicly available Genotype-Tissue Expression (GTEx) portal data]. A high-throughput screening in which the random insertion of the Sleeping Beauty transposon was used to test the ability of murine Ba/F3 pro–B cells to grow independently of interleukin-3 (IL-3) found recurrent transposon insertions in the Rapgef1 gene, suggesting that increased expression of C3G contributes to oncogenic signaling (45).

Despite the widely held view that the GEF activity of C3G is autoregulated, the details of this mechanism remain unknown. Here, we report the identification of two regulatory intramolecular interactions between noncatalytic and catalytic regions: (i) A contact between the NTD and REM domains stimulates the GEF activity, and (ii) an autoinhibitory region within the SH3b domain interacts with the Cdc25HD and blocks its activity. We also show that CrkL activates C3G by displacing this autoinhibitory interaction. We have mapped residues in the autoinhibitory region that are essential for blocking the activity of C3G. Noteworthy, two point mutations detected in hematological tumors target the autoinhibitory interaction and cause constitutive activation of C3G-Rap1 in cells.


The NTD of C3G interacts with the REM domain

Because the noncatalytic regions of C3G might regulate the GEF activity through interactions with the catalytic region, we searched systematically for potential intramolecular contacts. Initially, we analyzed whether the NTD interacts with other regions of C3G. The NTD with a FLAG-tag (FLAG-NTD) was transiently coexpressed in COS-1 cells with hemagglutinin (HA)–tagged C3G, full-length, or deletion mutants (Fig. 1A), and the interaction was assessed by coimmunoprecipitation (coIP) (Fig. 1B). Full-length C3G did not interact with the NTD, whereas a construct that lacks the NTD (C3G-ΔNTD, 246-1077) coimmunoprecipitated with FLAG-NTD. This suggests that in full-length C3G, the NTD-binding site is internally masked by an intramolecular interaction, and the site is exposed in the absence of the NTD. Four constructs carrying increasing deletions of the central SH3b region (277-1077, 292-1077, 465-1077, and 550-1077) and the catalytic region alone (REM-Cdc25HD, 670-1077) also interacted with FLAG-NTD. On the other hand, the isolated SH3b region did not interact with FLAG-NTD. Within the catalytic region, the REM domain associated with the NTD, whereas the Cdc25HD did not. Two constructs that contain the REM domain and segments of the SH3b domain, P3-P4-REM (465-814) and P4-REM (550-814), also interacted with FLAG-NTD. Last, none of the HA-tagged fragments coimmunoprecipitated with when they were cotransfected with empty pCEFLAG vector, indicating that the interaction was specifically mediated by the NTD.

To analyze whether the NTD/REM interaction occurs within the same molecule, we created C3G constructs carrying the fluorescent proteins CFP (cyan fluorescent protein) and Venus at the N and C termini, respectively (Fig. 2A). CFP and Venus form a donor-acceptor pair for Förster resonance energy transfer (FRET), and the FRET signal yields information about their proximity. Upon excitation of CFP in COS-1 cells expressing CFP-C3G-Venus (abbreviated C-C3G-V), a peak appeared at ~527 nm, which corresponds to the emission of Venus due to FRET (Fig. 2, B and C). This suggested proximity between the fluorophores at both ends of C3G. The Venus peak was not detected in cells coexpressing the individually labeled constructs C-C3G and C3G-V. Thus, the FRET signal of C-C3G-V is related to an intramolecular event.

Fig. 2 The NTD/REM interaction is intramolecular and dispensable for autoinhibition of C3G.

(A to C) Schematic representation of C3G fusion proteins used as FRET-based conformational sensors (A); a representative, normalized fluorescence emission spectra (excitation, 435 nm) in COS-1 cells expressing C3G constructs with CFP and Venus (B); and the fluorescence intensity ratios at Venus and CFP emission maxima used as a measurement of the relative FRET efficiency (C). Data are means ± SD; n = 3 to 6 independent experiments as indicated. *P < 0.05 , **P < 0.01, and ****P < 0.0001, pairwise comparisons by Welch’s t test. (D) Schematic representation of the conformations of the C3G sensors based on the FRET data above. (E) Structural model of the REM domain of C3G. The side chains of the mutated residues are shown as sticks. (F) coIP between the FLAG-NTD and WT or point mutant HA-P3-P4-REM in lysates from transfected COS-1 cells, assayed by FLAG antibody followed by Western blot (WB). Cotransfection with empty pCEFLAG vector was used as a negative control, and β-actin was used as a loading control. Blots are representative of two independent experiments. (G and H) Representative emission spectra (excitation, 435 nm) (G) and ratios of the fluorescence intensity (H) in COS-1 cells expressing C-C3G-ΔCdc25HD-V WT or point mutants. Data are means ± SD; n = 3 to 10 independent experiments as indicated. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by ANOVA followed by Dunnett’s test. ns, not significant. (I) Representative nucleotide exchange reactions of Rap1:mant-dGDP alone and in the presence of C3G proteins (1 μM). Single exponential decays were fitted (dashed lines) to determine the apparent nucleotide exchange rates (kobs). Image is a representative of at least three experiments. (J) Exchange rates (kobs) produced by C3G full-length (FL), C3G-ΔNTD, C3G-E731R/E784R, and the isolated Cdc25HD. Means ± SD. n = 3 to 5 independent experiments, as indicated, one shown in (I). *P < 0.05 and ****P < 0.0001 by ANOVA followed by Dunnett’s test. The linear dependence of the kobs with the concentration of C3G FL and Cdc25HD is shown in fig. S1. a.u., arbitrary units.

We also analyzed deletion mutants of C-C3G-V in a similar manner. The ratios of the fluorescence intensity at the CFP (475 nm) and Venus (527 nm) maxima were used to assess the FRET signal (Fig. 2, B and C). Mutants that lack either the NTD (C-C3G-ΔNTD-V) or the complete catalytic region (C-C3G-ΔREM-Cdc25HD-V) showed lower FRET ratios than the full-length construct, suggesting that they adopt open conformations (Fig. 2D). In contrast, deletion of only the Cdc25HD (C-C3G-ΔCdc25HD-V) resulted in the highest FRET ratio. This suggests that CFP at the N terminus is apparently closer to the REM domain than to the C terminus of the Cdc25HD. Collectively, coIP and FRET data support the notion of a direct intramolecular interaction between the NTD and REM domains, which does not require the Cdc25HD.

To map the NTD-binding site, we generated point mutants in the REM domain and analyzed their interaction with the NTD. To assist in the design and interpretation of the mutagenesis data, we created a structural homology model of the REM domain of C3G (Fig. 2E) using the structure of the REM domain of SOS as template (46). REM domains consist of five α helices (α2 to α6) that form a helical bundle and an additional short helix (α1) that packs on one side of the bundle. To sample multiple areas of the REM domain, we generated the single substitutions L714R, E731R, K756E, R764E, L774R, E784R, and K803E, which are spread along the solvent-exposed surface of the REM domain and are not predicted to be occluded by the Cdc25HD (Fig. 2E). Mutations were introduced in the construct HA-P3-P4-REM, which is expressed in higher levels than the isolated REM, and their interaction with FLAG-NTD was analyzed by coIP (Fig. 2F). Only the mutations E731R and E784R prevented the interaction with the NTD. The double mutant E731R/E784R also showed marginal interaction with FLAG-NTD. We also analyzed the effect of these point mutations on the conformation of C3G by FRET (Fig. 2, G and H). We used the construct C-C3G-ΔCdc25HD-V, because it had the highest FRET signal (see above) and because the Cdc25HD is dispensable for the NTD/REM interaction. Mutations L714R, K756E, R764E, and K803E did not alter the FRET signal. The mutant L774R had slightly, but not significantly, lower FRET ratio. Mutants E731R and E784R had significantly lower FRET ratios and, when combined (E731R/E784R), caused a further reduction of the FRET signal, suggesting an additive effect of these two substitutions. In summary, Glu731 and Glu784, which are adjacent on the surface of the REM domain, define a putative binding site for the NTD around helices α3 and α5 (Fig. 2E).

We then analyzed the role of the NTD/REM interaction on the regulation of the GEF activity. We measured the release of the fluorescent nucleotide mant-dGDP from Rap1b in the presence of purified C3G fragments (Fig. 2, I and J, and fig. S1). The spontaneous nucleotide exchange of Rap1b:mant-dGDP is extremely slow under these conditions; the apparent dissociation constant (kobs) was ~0.02 × 10−3 s−1. Full-length C3G induced a moderate increase in the exchange rate (kobs = 0.55 × 10−3 s−1), whereas the activity of the isolated Cdc25HD (kobs = 2.8 × 10−3 s−1) was five times faster. The truncated construct C3G-ΔNTD had slightly lower activity (kobs = 0.41 × 10−3 s−1), and the full-length mutant E731R/E784R was significantly less active (kobs = 0.15 × 10−3 s−1) than wild-type (WT) C3G. The role of the NTD/REM interaction on the regulation of the activity of C3G was analyzed later in detail (presented below). The data thus far suggest that the GEF activity of C3G is self-repressed and that the NTD is dispensable for this autoinhibition.

The SH3b region binds directly to the Cdc25HD and blocks the GEF activity

Because C3G-ΔNTD is autoinhibited, we next explored whether the SH3b domain might regulate the GEF activity. HA-tagged constructs of the catalytic region (REM-Cdc25HD) or the individual REM and Cdc25H domains were expressed in human embryonic kidney (HEK) 293T cells, and their interaction with purified GST-SH3b (274-646), or glutathione S-transferase (GST) alone as a control, was analyzed using a pull-down (PD) assay (Fig. 3A). HA-REM-Cdc25HD and HA-Cdc25HD interacted with the SH3b, but no binding of HA-REM was observed. Because HA-REM was expressed in very low levels, we also created constructs of these three fragments fused to monomeric enhanced green fluorescent protein (mEGFP) (Fig. 3B). Both REM-Cdc25HD-mEGFP and Cdc25HD-mEGFP bound specifically to GST-SH3b, whereas no specific interaction was observed between REM-mEGFP and GST-SH3b. Last, purified Cdc25HD produced in bacteria was also pulled down with GST-SH3b (Fig. 3C). In summary, the SH3b domain binds directly to the Cdc25HD.

Fig. 3 The final segment of the SH3b domain binds to the Cdc25HD and blocks its GEF activity.

(A and B) Analysis by pull-down (PD) of the interaction of the SH3b domain (274-646) with the catalytic region. The region REM-Cdc25HD and its individual domains were expressed in HEK293T cells with HA (A) or mEGFP (B) tags and were subjected to PD with GST-SH3b. Tagged C3G fragments and β-actin in the cell lysates and C3G fragments in the PD were detected by Western blot. (C) Binding of the Cdc25HD to GST-SH3b fragments analyzed by PD. The Cdc25HD produced in E. coli was detected in the PD by Western blot with an antibody against C3G. GST-SH3b fragments were visualized by Coomassie staining. PDs are representative of two independent experiments. (D) Schematic representation of the fragments of the SH3b domain analyzed and a summary of their binding to the Cdc25HD and inhibition of the GEF activity (n.a., not analyzed). (E) Exchange reactions of Rap1:mant-dGDP catalyzed by full-length C3G (1 μM; top left), the Cdc25HD alone (1 μM; top left), and the Cdc25HD in the presence of fragments of the SH3b domain (1 and 40 μM, respectively, displayed over all three graphs). The kobs were determined by fitting a single exponential decay model (dashed lines). Data are from a representative of three independent experiments. (F) GEF activity (kobs) of C3G full-length (FL) and Cdc25HD alone and in the presence of fragments of the SH3b domain. Data are means ± SD, n = 3 independent experiments, one shown in (E). Dashed lines mark the exchange rates of autoinhibited full-length C3G and the uninhibited isolated Cdc25HD. (G) Dose-dependent effect of untagged AIR on the activity of the Cdc25HD (1 μM). A sigmoidal function (solid line) was fitted to estimate the IC50. (H) Schematic representation of the two segments of the AIR and their contribution to the repression of the GEF activity of the Cdc25HD. The model depicts the behavior observed using independently purified AIR and Cdc25HD.

We used deletion mapping to identify the segment within the SH3b that binds to the Cdc25HD. GST-SH3b fragments were produced in bacteria, and their binding to purified Cdc25HD was assessed by PD (Fig. 3, C and D). The Cdc25HD did not bind to fragments of the first half of the SH3b (274-371 and 274-500), but it bound to the second half of the SH3b (501-646), which contains the P3 (539-547) and P4 (607-616) motifs. The segment 501-536, which contains Tyr504, was neither required nor sufficient for the interaction, because the Cdc25HD bound to 537-646 but not to 501-536. The final part of the SH3b (579-646), which contains the P4, did not bind to the Cdc25HD, and this region was not required for binding, as observed by the binding of the Cdc25HD to the fragments 274-578, 501-578, 537-588, and 537-578. A construct further shortened at the C terminus, 537-569, also bound to Cdc25HD. Additional trimming, as in the fragment 537-560, reduced the interaction. Similarly, the segment 545-560 also bound weakly. Other two fragments that lack the P3 motif, 545-646 and 545-569, bound to the Cdc25HD, revealing that the P3 is not required for binding. In summary, the minimal segment of the SH3b that binds to the Cdc25HD corresponds to a 25-residue segment (545-569) located immediately downstream of the P3 motif.

Next, we analyzed whether the SH3b/Cdc25HD interaction affects the GEF activity. We measured the exchange activity over Rap1b in vitro of the isolated Cdc25HD, alone and in the presence of the GST-SH3b fragments (40 μM) that bind to the Cdc25HD, and compared it to the activity of the autoinhibited full-length C3G (Fig. 3, D to F). The complete SH3b (274-646) reduced the activity of the Cdc25HD to the levels of full-length C3G, revealing that the SH3b/Cdc25HD interaction alone is sufficient for the autoinhibition, given that it did not require the presence of the NTD or REM domains. The SH3b fragments 372-646, 501-646, 537-646, and 545-646 also efficiently blocked the activity of the Cdc25HD (80 to 100% inhibition). On the other hand, fragments that lack the P4 and the C-terminal segment of the SH3b (274-578, 501-578, 537-588, 537-578, 537-569, and 545-569) did not block the GEF activity, or they did only partially (20 to 45% inhibition). When the fragment 537-646 was produced without any tag, it blocked the Cdc25HD as efficiently as the equivalent GST-tagged construct, excluding any contribution from the GST moiety. A dose-dependent analysis of the effect of the fragment 537-646 on the exchange activity of the Cdc25HD yielded a half-maximal inhibitory concentration (IC50) of ~13 μM (Fig. 3G).

Collectively, our results revealed an autoinhibitory region within the SH3b domain, hereafter AIR (residues 545-646), that contains two functionally distinct segments (Fig. 3, D and H). A first short segment of the AIR (residues 545-569), which is adjacent to the P3 motif, drives the binding to the Cdc25HD; hence, we name it Cdc25HD-binding region or AIR-CBR. A second segment located in the final part of the AIR (residues 570-646) is required to repress the GEF activity of the Cdc25HD; therefore, we name it AIR-inhibitory tail or AIR-IT.

The AIR/Cdc25HD interaction is the main autoinhibitory mechanism of C3G

Analysis of the sequences of C3G from 209 species (table S1) revealed that the catalytic region is the most conserved segment (Fig. 4A). The SH3b is globally the less conserved domain; yet, the Pro-rich motifs are highly conserved. The segment 545-583, which is adjacent to the P3 and includes the AIR-CBR, is almost as conserved as the catalytic region, suggesting that the regulatory interaction is preserved in C3G from multiple species.

Fig. 4 Mutations that disrupt the AIR/Cdc25HD interaction activate C3G constitutively.

(A) Conservation scores of C3G per residue (colored bars) and average values per domain (boxes). Alignment of the sequences of the P3 motif and part of the AIR from representative species; positions are colored according to their conservation. Secondary structure predictions by three methods are shown under the sequences. Secondary structure prediction for the complete SH3b domain is shown in fig. S2. (B and C) Binding of the Cdc25HD to GST-AIR, WT, and mutants, analyzed by PD assays. Cdc25HD in the PD was detected by Western blot. Two types of substitutions were assayed: reverse-charge replacements (B) and changes to alanine and mutations described in lymphomas (C). PDs are representative of two independent experiments. (D) Helical wheel representation of the predicted helix in the AIR-CBR. Met551, Tyr554, and Met555 define the putative binding site for the Cdc25HD. (E and F) Representative nucleotide exchange reactions of Rap1:mant-dGDP catalyzed by full-length C3G WT and the indicated point mutants (1 μM) (E) and the exchange rates (kobs) (F). n = 3 to 5 independent experiments, as indicated, one shown in (E). (G) Schematic representation of autoinhibited C3G and the uninhibited conformation induced by mutations that destabilize the AIR/Cdc25HD interaction.

To identify residues of the AIR that contribute to the interaction with the Cdc25HD, we generated point mutants within and near the AIR-CBR, focusing on conserved positions. Initially, we created six single and eight double mutants that collectively probed 22 residues in the 546-581 segment; acidic and noncharged residues were replaced by Arg, and basic residues were changed to Glu. Mutations were introduced in GST-AIR 537-646, and binding to the Cdc25HD was analyzed by PD (Fig. 4B). Only the individual change Y554R and the double mutations H550E/M551R and M555R/Q556R strongly reduced binding to the Cdc25HD. The single substitutions M551R and M555R compromised the interaction, whereas H550E and Q556R did not affect the binding. Replacing Tyr554 by Ala also severely reduced the interaction, whereas mutants M551A and M555A showed partial reduction of the binding (Fig. 4C). The AIR is predicted to be mostly disordered (fig. S2), yet an α helix is predicted in the AIR-CBR, with Met551, Tyr554, and Met555 being on the same side of that helix (Fig. 4D). Collectively, Tyr554 makes essential contributions to the binding to the Cdc25HD, and it is flanked by Met551 and Met555, suggesting that these three residues form a continuous site that contacts the Cdc25HD.

To assess the contribution of the AIR/Cdc25HD interaction to the regulation of C3G, the point mutations that disrupt this contact were introduced in the full-length protein, and the exchange activities of C3G mutants were analyzed (Fig. 4, E and F). C3G mutants M551R, Y554R, and M555R were constitutively in a high-activity state. C3G-Y554R showed the highest exchange activity (kobs = 20.5 × 10−3 s−1) that was ~50-fold higher than that of C3G-WT. C3G-M551R and M555R showed slightly slower exchange rates (kobs = 12.9 × 10−3 s−1 and kobs = 10.3 × 10−3 s−1) than Y554R, in concordance with their moderate reduction of the AIR/Cdc25HD binding. In summary, disruption of the AIR/Cdc25HD interaction is sufficient to unleash the GEF activity of C3G and suggests that this contact is the main autoinhibitory mechanism (Fig. 4G).

Cancer-related missense mutations in the AIR activate C3G constitutively

The Catalogue of Somatic Mutations in Cancer database (COSMIC; version 90) describes ~170 missense single-nucleotide variants (SNVs) in the RAPGEF1 gene; 12 of those are in or near the AIR-CBR (table S2). Two mutations found in patients with non-Hodgkin’s lymphomas, Y554H and M555K (47, 48), affect residues that are critical for the autoinhibition. This prompted us to analyze their impact on the regulation of C3G. Mutation Y554H abrogated the binding of GST-AIR to the Cdc25HD, and M555K partially reduced the interaction (Fig. 4C). When introduced in the full-length C3G, Y554H induced constitutive high GEF activity (Fig. 4, E and F, and fig. S1). M555K also activated C3G, although to a lesser extent.

Next, we analyzed the effect of C3G-activating mutations on the activation of Rap1 in cells. First, WT and point mutants of C3G-mEGFP, or isolated mEGFP, were transiently expressed in HEK293T cells, and the levels of Rap1-GTP were measured (Fig. 5, A and B). As a positive control, we expressed C3G-WT-mEGFP carrying the CAAX sequence of K-Ras, which is constitutively located on the plasma membrane, causing the activation of Rap1 (23). Expression of C3G-WT, the mutant C3G-E731R/E784R (with reduced NTD/REM interaction), or mEGFP alone did not result in detectable levels of Rap1-GTP. On the other hand, C3G-M551R and the lymphoma mutants C3G-Y554H and C3G-M555K caused a constitutive increase of Rap1-GTP. Rap1 activation induced by C3G and its mutants in HEK293T cells correlated with their activity in vitro (Fig. 5C).

Fig. 5 Lymphoma-related mutations in C3G activate Rap1 and LFA-1 in cells.

(A) Analysis of Rap1-GTP in HEK293T cells expressing mEGFP; C3G-mEGFP WT; the mutants M551R, Y554H, or M555K; or the membrane-targeted WT with a CAAX sequence. Expression of mEGFP and endogenous Rap1 and β-actin in cell lysates were analyzed by Western blot. In the PD, Rap1-GTP was detected by Western blot, and GST-RalGDS-RBD was detected by Ponceau S staining. PD is representative of three independent experiments. (B) Scatterplot and bar chart of three independent measurements of Rap1 activation in HEK293T cells, as described and represented in (A). Rap1-GTP levels in cells expressing C3G-mEGFP-CAAX were used to normalize the data from different experiments. (C) Correlation between the exchange activity in vitro (kobs) of C3G WT and mutants (shown in Fig. 4F) and the Rap1-GTP levels that they induced in HEK293T cells [shown in (B)]. (D) Rap1-GTP levels in Ba/F3 cells expressing C3G-mEGFP, WT or the indicated mutants, or mEGFP alone and in uninfected cells. Proteins were detected in the cell lysates and in the GST-RalGDS-RBD PDs as described in (A). In addition, C3G (endogenous and mEGFP-tagged) was also detected using an antibody against C3G. PD is representative of two independent experiments. (E) Time course activation of Rap1 after stimulation with IL-3 of Ba/F3 cells expressing mEGFP, C3G-mEGFP WT, or Y554H. Rap1-GTP was detected as described in (A) and (D). PD is representative of two independent experiments. (F) Activation of the integrin LFA-1 in Ba/F3 cells expressing mEGFP, C3G-mEGFP WT, or mutants, determined by flow cytometry (n = 3, biological replicates, means ± SD). Statistical comparison to cells expressing mEGFP was analyzed using ANOVA followed by Dunnett’s test. ***P < 0.001; ****P < 0.0001.

Because mutations Y554H and M555K were described in lymphomas, we also analyzed their impact in the pro–B cell line Ba/F3, in which C3G-mEGFP, WT and mutants, or mEGFP alone was stably expressed by lentiviral infection. Expression of C3G-Y554H or C3G-M555K resulted in constitutively higher levels of Rap1-GTP than when Ba/F3 cells expressed C3G-WT, C3G-E731R/E784R, and isolated mEGFP, or when cells did not express heterologous proteins (Fig. 5D). Upon stimulation with IL-3, which is a known activator of Rap1 in hematopoietic cells (30), Rap1-GTP increased in Ba/F3 cells expressing C3G-WT or C3G-Y554H, with Rap1-GTP levels being higher at all time points in the presence of C3G-Y554H (Fig. 5E). Rap1 was also activated by IL-3 in cells expressing mEGFP, although the levels of Rap1-GTP were lower than in the presence of heterologous expression of C3G; in this case, Rap1 may be activated by endogenous C3G or other GEFs. Last, we analyzed whether C3G mutants affected the activation of integrins. Because Rap1 is a potent activator of the leukocyte function-associated antigen 1 (LFA-1) integrin in Ba/F3 cells (49), we analyzed the impact of mutant C3G on its activation. Active LFA-1 was detected in Ba/F3 cells by flow cell cytometry using the antibody mAb24 that recognizes an active conformation of this integrin (Fig. 5F). Cells expressing C3G-Y554H or C3G-M555K had higher basal levels of active LFA-1 than cells expressing C3G-WT, C3G-E731R/E784R, or mEGFP. In summary, the missense SNVs, Y554H and M555K, found in patients with non-Hodgkin’s lymphoma cause catalytic deregulation of C3G, resulting in increased basal stimulation of Rap1 and integrin LFA-1 and strong activation of Rap1 upon stimulation with IL-3.

CrkL activates C3G by releasing the autoinhibitory contact

Binding of Crk proteins to the SH3b and tyrosine phosphorylation of C3G directly stimulate the GEF activity (23, 24). Given that disruption of the AIR/Cdc25HD interaction by point mutations activates C3G, we analyzed whether Crk or Src-mediated phosphorylation may act on the autoinhibitory interaction. First, we assessed the contribution of these two stimuli to the activation of C3G in vitro (Fig. 6, A to C). CrkL induced an ~8-fold increase of the exchange activity of full-length C3G. Phosphorylation of C3G by the kinase domain of Src (SrcKD) caused a ~4-fold increase of the exchange rates with respect to unphosphorylated C3G. Adding CrkL to phospho-C3G (p-C3G) induced the highest activity, ~30-fold increase, suggesting that these two stimuli are independent and additive. Neither CrkL nor phosphorylation altered the activity of the isolated Cdc25HD, supporting the notion that these stimuli act on the autoinhibitory mechanism and not directly on the catalytic domain.

Fig. 6 The AIR/Cdc25HD interaction is the main autoinhibitory mechanism and is disrupted by CrkL for activation.

(A) Phosphorylation of purified C3G-WT with SrcKD, analyzed by Western blot. Similar analysis of Cdc25HD and the C3G mutants M551R, Y554R, and Y554H are shown in fig. S3 (A to D). (B) Representative dissociation reactions of Rap1:mant-dGDP catalyzed by C3G (1 μM), unmodified or phosphorylated with SrcKD (p-C3G), alone and in the presence of CrkL (10 μM). (C) Nucleotide exchange rates (kobs) catalyzed by full-length C3G WT and point mutants and the isolated Cdc25HD. Unphosphorylated and SrcKD-phosphorylated samples were analyzed alone and in the presence of CrkL, n = 3 to 10 independent experiments as indicated. Representative dissociation reactions are shown in (B) (WT) and in fig. S3 (A to D) (Cdc25HD and mutants). (D) Analysis by PD of the competition of CrkL with the Cdc25HD for binding to four constructs of the AIR that contain the P3 and P4 (GST-537-646-WT), only the P3 (GST-537-646-P4A), only the P4 (GST-545-646-WT), or none of these proline-rich motifs (GST-545-646-P4A). PDs are representative of two independent experiments. (E) Time course of the in vitro phosphorylation of the AIR (537-646) by the SrcKD, analyzed by Western blot; representative of two independent experiments. (F) Binding of Cdc25HD to GST-AIR (537-646) phosphorylated with SrcKD in PD assays. PD is representative of two independent experiments. (G) Exchange activity (kobs) of the Cdc25HD alone and in the presence of AIR (untagged or GST-fusion) or the AIR mutants in (D). n = 3 to 4 independent experiments as indicated. Representative dissociation experiments are shown in fig. S3E. (H) Schematic representation of the effect of phosphorylation and CrkL binding to the AIR on the release of the inhibitory interaction when the AIR and the Cdc25HD are assayed as individual proteins.

The activity of p-C3G with CrkL was similar to that of the constitutively active mutants C3G-Y554R and Y554H. The activity of C3G-Y554R was not increased by CrkL, phosphorylation, or their combination, and those stimuli only produced a minor increase in the activity of C3G-Y554H (~1.2- to ~1.6-fold) (Fig. 6C and fig. S3, A to D). On the other hand, the partially active mutant C3G-M551R was stimulated by CrkL or phosphorylation to similar levels to C3G-Y554R/H, and the combination of CrkL and phosphorylation did not activate C3G-M551R any further than the individual stimuli. Collectively, our data suggest that CrkL and phosphorylation by Src stimulate C3G by acting on the autoinhibitory mechanism.

We analyzed the impact of CrkL on the AIR/Cdc25HD interaction using independently purified AIR and Cdc25HD. CrkL competed in a dose-dependent manner with the Cdc25HD for binding to the AIR construct 537-646 that contains the P3 and P4 motifs (Fig. 6D). To analyze the independent contribution of CrkL binding to the P3 and P4, we used two constructs of the AIR, each containing only one of these motifs. The P4 was inactivated by introducing simultaneously five point mutations P608A, P609A, L611A, P612A, and K614A (hereafter P4A). CrkL bound to GST-537-646-P4A and displaced the binding to the Cdc25HD with a similar efficiency as for GST-537-646-WT. CrkL also bound to GST-545-646-WT, which only contains the P4, but CrkL binding only caused a minor displacement of the Cdc25HD. Last, CrkL did not bind to GST-545-646-P4A and consequently did not displace Cdc25HD from binding to this construct.

We also analyzed the effect of the phosphorylation on the AIR/Cdc25HD interaction. The untagged AIR (537-646) was directly phosphorylated by SrcKD (Fig. 6E). The Cdc25HD bound to phospho-GST-AIR (537-646) similarly to the unphosphorylated protein (Fig. 6F).

Next, we analyzed the effect of CrkL and SrcKD phosphorylation on the ability of the isolated AIR to block the Cdc25HD activity (Fig. 6, G and H, and fig. S3E). Binding of CrkL to the AIR 537-646 prevented its inhibition of the Cdc25HD. CrkL also precluded the inhibition of Cdc25HD by the P3-only mutant AIR 537-646-P4A. On the other hand, when CrkL bound only to the P4 in the fragment AIR 545-646, this complex still blocked the activity of the Cdc25HD, although slightly less efficiently than 537-646-P4A alone. Last, the phosphorylated AIR blocked the activity of the Cdc25HD, albeit slightly less efficiently than unphosphorylated AIR (Fig. 6, G and H). In summary, our results suggest that CrkL activates C3G by releasing the AIR/Cdc25HD autoinhibitory interaction and that, although the P3 is not required for the autoinhibition, binding of CrkL to the P3 apparently disrupts the interaction between the adjacent AIR-CBR and the Cdc25HD.

The NTD/REM interaction is required for the activation of C3G

The isolated Cdc25HD has lower GEF activity than the full-length C3G activated by point mutations (for example, Y554H) or by the combined stimuli of CrkL and phosphorylation, suggesting that the activity is favored by other regions outside the Cdc25HD. In other GEFs, the REM domain makes close contact with the Cdc25HD, suggesting that in C3G, the REM domain could stabilize the Cdc25HD and favor its activity. Unfortunately, the activity of the region REM-Cdc25HD of C3G could not be analyzed directly because this construct is unstable. Yet, we had observed that autoinhibited full-length C3G has higher residual activity than the C3G-E731R/E784R mutant (see Fig. 2J). This prompted us to analyze in detail the role of the NTD/REM interaction on the regulation of C3G.

We compared the activity of full-length C3G and four constructs in which the NTD/REM interaction was absent or disturbed: C3G-ΔNTD, C3G-454-1077, C3G-530-1077, and full-length C3G-E731R/E784R (Fig. 7A). We also analyzed C3G-454-1077-E731R/E784R to assess the direct effect of these mutations in the absence of the NTD. Before stimulation, all constructs had low GEF activity; yet, full-length C3G was slightly but significantly more active than the other proteins (Fig. 7B and fig. S4A). Three independent purifications of full-length C3G were used, discarding the idea that the differences could be due to variability between protein batches. CrkL (10 μM) activated all these constructs, but those with the disrupted NTD/REM interaction reached slower dissociation rates (kobs = between 1.1 × 10−3 s−1 and 2.2 × 10−3 s−1) than full-length C3G-WT (kobs = 4.3 × 10−3 s−1) (Fig. 7C and fig. S4B). All the constructs were phosphorylated by the SrcKD to apparently similar levels (fig. S4C); yet, full-length p-C3G-WT showed higher activity than the other proteins (Fig. 7D and fig. S4D). Similarly, when CrkL was added to the phosphorylated proteins, C3G-WT was more active than all the other proteins (Fig. 7E and fig. S4E).

Fig. 7 The NTD/REM interaction stimulates the GEF activity of the Cdc25HD.

(A) Schematic representation of full-length C3G-WT, truncation fragments, and the mutant E731R/E784R (red crosses) in which the NTD/REM interaction is destabilized. (B to E) Nucleotide exchange activity (kobs) of the proteins depicted in (A), in the absence of stimuli (B), in the presence of CrkL (C), phosphorylated with SrcKD (D), and when CrkL and phosphorylation were combined (E). The number of independent experiments (n) is indicated. Lines are means ± SD. Representative nucleotide dissociation reactions are shown in fig. S4 (A, B, D, and E). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by one-way ANOVA followed by Dunnett’s test (B and C) or unequal variance Brown-Forsythe ANOVA followed by Dunnett’s test (D and E). (F) Analysis of Rap1 activation in HEK293T cells expressing C3G-mEGFP WT, the mutant E731R/E784R, or C3G with the membrane targeting CAAX-tag. Results of two independent experiments are shown. (G) Schematic illustration of the contribution of the NTD/REM interaction to the activation by CrkL and phosphorylation of C3G WT and mutants.

We also analyzed whether the NTD/REM interaction favors the GEF activity of C3G in cells. mEGFP-tagged C3G-WT, C3G-E731R/E784R, or C3G-WT-CAAX was expressed in HEK293T cells either alone or in combination with heterologous expression of CrkL. The latter induces activation of C3G in cells (25). Expression of C3G-WT or C3G-E731R/E784R did not increase the basal levels of Rap1-GTP in HEK293T (Figs. 5A and 7F) or in Ba/F3 cells (Fig. 5D). CrkL activated Rap1 in HEK293T cells expressing C3G-WT but not in those with C3G-E731R/E784R. In summary, the NTD/REM interaction indirectly enhances the activity of the Cdc25HD and favors the activation of C3G by CrkL and Src phosphorylation (Fig. 7G).


The direct interaction of the AIR with the Cdc25HD is the main mechanism that represses the GEF activity of unstimulated C3G. This interaction is driven by the AIR-CBR (residues 545-569). Residues Met551, Tyr554, and Met555 are predicted to form a single continuous site that interacts with the Cdc25HD. Because the AIR-CBR alone does not block the GEF activity, most likely it contacts the Cdc25HD outside the catalytic GTPase-binding site. Inhibition requires the AIR-IT (570-646), which is predicted to be disordered. Once the AIR-CBR binds to the Cdc25HD, the AIR-IT is likely to interact, albeit weakly, with the catalytic binding site, blocking the access of the GTPase. This bipartite interaction of the AIR with the Cdc25HD resembles a lock-lid system in which the AIR-CBR acts as the lock and the AIR-IT acts as a lid that blocks the catalytic site. The mechanism of autoinhibition of C3G has similarities with that of RasGRP1, a GEF of Ras. In RasGRP1, the EF domain binds to a surface on the Cdc25HD that is adjacent to the catalytic Ras-binding site, and the linker that connects the Cdc25HD and EF domain blocks the Ras-binding site (50).

The AIR-CBR is a key regulatory site for the activation of C3G. Mutations that disrupt the AIR-CBR/Cdc25HD interaction, such as those described in non-Hodgkin’s lymphomas, are sufficient to activate C3G to a similar extent as when C3G-WT is stimulated by CrkL and Src-mediated phosphorylation. The deregulated active mutants (Y554H and Y554R) are not further stimulated by CrkL and phosphorylation, suggesting that these stimuli operate through the release of the AIR-CBR/Cdc25HD interaction. Binding of CrkL to the P3, which is juxtaposed to the AIR-CBR, displaces the interaction of the AIR with the Cdc25HD. On the other hand, Src phosphorylation of the isolated AIR did not noticeably reduce its binding to Cdc25HD in PD experiments. Yet, phospho-AIR blocked the GEF activity slightly less efficiently than unphosphorylated AIR (Fig. 6G), and phosphorylation of full-length C3G by Src increased the GEF activity, albeit to a lesser extent than the stimulation by CrkL, suggesting that phosphorylation releases at least partially the autoinhibition. It is possible that phosphorylation of Tyr504 or other tyrosine residues might disrupt, directly or indirectly, the blockage of the Cdc25HD activity by the AIR-IT while having a minor effect on the binding of the AIR-CBR to the Cdc25HD. This mechanism is compatible with the apparently independent effects of CrkL binding and Src phosphorylation on the activation. In addition, phosphorylation of C3G could stabilize the interaction with CrkL.

The key role of Tyr554 in the autoinhibition apparently contrasts with the moderate activation caused by tyrosine phosphorylation of C3G. Tyr554 is within a sequence (551-MLAYMQL-557) that does not match the favored motifs of Src substrates, which have strong preference for Ile/Leu/Val preceding (position −1) the target Tyr residue and moderate preference for Glu in positions −3 and −2, Gly/Glu in position +1, and Phe/Ile/Leu in position +3 (5153). In addition, phosphorylation of Tyr561, Tyr570, and Tyr579, but not of Tyr554, has been reported (publicly available data at portal). Collectively, these data suggest that Tyr554 is not a substrate for Src, which is in agreement with the negligible impact of Src phosphorylation on the AIR/Cdc25HD interaction.

The isolated Cdc25HD has lower GEF activity than activated full-length C3G (Fig. 6C), revealing that activation requires the release of the autoinhibition and the concurrence of positive signals. The presence of the REM domain is not sufficient to fully activate the Cdc25HD. Instead, the REM domain acts as a mediator that channels the activation boost caused by the NTD/REM interaction. The NTD interacts with helices α3 and α5 of the REM domain, which are not part of the predicted Cdc25HD interaction surface, and the NTD/REM interaction does not require the Cdc25HD. Thus, we propose that the NTD enhances the GEF activity by inducing a high-activity conformation of the REM-Cdc25HD tandem. Such allosteric stimulation of C3G is reminiscent of the positive feedback activation of SOS, where binding of a Ras-GTP molecule to a regulatory site, at the rim of the REM/Cdc25HD interaction surface, induces a high-activity conformation of the Cdc25HD (54). Yet, the NTD interaction site in the REM of C3G is different from the Ras-GTP allosteric site of SOS. Despite the fact that no other interactors of the REM domain of C3G have been described so far, binding of other molecules to the REM domain could also modulate the GEF activity, suggesting that C3G might be regulated by feedback mechanisms as observed in other GEFs (26). Collectively, C3G works as a “dimmer switch” that can adopt multiple activity levels, which allows fine-tuning of its signaling.

Rap signaling contributes to the progression of several types of lymphomas and leukemias. Rap activation in the murine B lymphoma cell line A20 increases their dissemination in vivo, favoring their adhesion to endothelial cells and their transendothelial migration (55). Constitutive activation of Rap proteins due to high expression of CD38 in human chronic lymphocytic leukemia (CLL) B cells stimulates cell adhesion and migration (56). Mice deficient in the Rap1 GTPase-activating protein SIPA1 (signal-induced proliferation associated protein 1, also known as SPA-1), which have increased levels of Rap1-GTP, develop disorders that resemble human CML (57) and CLL (58). Transplantation in mice of hematopoietic progenitor cells with constitutively activated Rap1, by the expression of a membrane-targeted engineered form of C3G, causes T cell acute lymphoblastic leukemia (59, 60). CML cell lines and primary cells from patients with CML overexpress the variant p87C3G that lacks the NTD, and the CML-causing oncoprotein Bcr-Abl signals through p87C3G (22). Because activation of p87C3G is weakened by the lack of the NTD/REM interaction, CML cells might overexpress p87C3G to compensate for its attenuated signaling. We have also shown that two missense mutations in C3G detected in non-Hodgkin’s lymphomas disrupt the autoinhibitory mechanism and cause constitutive activation of C3G and Rap1, suggesting that mutant C3G may favor the progression of those lymphomas. Expression of deregulated C3G mutants also caused increased activation of the integrin LFA-1 in Ba/F3. Given the crucial role of LFA-1 in trafficking and homing of leukocytes and in lymphoma dissemination (61, 62), abnormal C3G-induced activation of LFA-1 might contribute to the dissemination of B cell lymphomas. This is the first description of a catalytic deregulation of C3G linked to a disease. Further work is required to fully understand the impact of C3G alterations in lymphomas and other types of cancer. The detailed characterization of the regulatory mechanisms presented here will pave the way to identify other potentially activating mutations in C3G and to understand their contribution to diseases. This opens to a potential use of C3G mutations as prognostic markers, as well as to the consideration of C3G as a potential therapeutic target.


DNA constructs and mutagenesis

All constructs of human C3G refer to the isoform a (UniProt Q13905-1). Constructs for expression in mammalian cells are listed in table S3. Constructs with an N-terminal HA-tag or FLAG-tag were created in the pCEFLHA and pCEFLAG vectors (63), respectively. Complementary DNA (cDNA) fragments were amplified by the polymerase chain reaction (PCR) using primers that added Bgl II and Not I sites at the 5′ and 3′ ends, respectively, and were cloned into those sites in the vectors. Primers used to create constructs of C3G are shown in table S4. Correctness of these and all other constructs was confirmed by DNA sequencing.

Constructs of C3G FRET sensors were created in the vector pcDNA3-CFP-Venus that codes for N-terminal CFP (between Hind III and Bam HI sites) and C-terminal Venus (between Not I and Xho I sites) tags. The cDNA fragments of C3G were amplified by PCR using primers that add Bgl II and Not I sites at the 5′ and 3′ ends, respectively, and do not contain a stop codon. The PCR products were cloned in between the Bam HI and Not I sites of the vector. C3G-Venus was created in the vector pcDNA3-Venus-N in a similar manner; in this case, the C3G cDNA was amplified using a direct primer that added a Bgl II site and a Kozak sequence. The construct CFP-C3G was created in the vector pcDNA3-CFP-C; in this case, the reverse primer used to amplify C3G included a stop codon.

Constructs of C3G with a C-terminal mEGFP tag for transient expression were created in the vector pEF1-mEGFP, which is a modification of pEF1/V5-HisA (Invitrogen), in a similar manner to the pcDNA3 constructs. C3G cDNA fragments were amplified by PCR that added Bgl II and Not I sites at the 5′ and 3′ ends, respectively, and were cloned into pEF1-mEGFP digested with Bam HI and Not I.

Lentiviral constructs with C3G cDNA were created in a modified version of the pLenti-C-HA-IRES-BSD vector (OriGene), which codes for a C-terminal mEGFP tag. To create this vector, hereafter pLenti-C-mEGFP-IRES-BSD, the region coding for the Kozak sequence and mEGFP tag of pEF1-mEGFP was PCR-amplified using primers that added Bgl II and Psp OMI sites at the 5′ and 3′ ends, respectively. The PCR product was digested with Bgl II and Psp OMI and was ligated into the compatible cohesive ends produced by digestion of pLenti-C-HA-IRES-BSD with Bam HI and Not I, which destroyed these restriction sites upon ligation. The resulting pLenti-C-mEGFP-IRES-BSD vector contains newly introduced Bam HI and Not I sites upstream the mEGFP cDNA. Next, the cDNAs of C3G WT and mutants were amplified by PCR, adding Bam HI and Not I sites at the 5′ and 3′ ends, respectively, and were ligated into the pLenti-C-mEGFP-IRES-BSD vector using the same sites.

C3G constructs for expression in bacteria are listed in table S5. The cDNAs of full-length C3G, C3G-ΔNTD, C3G-454-1077, and C3G-530-1077 were PCR-amplified using primers that added 5′–Nco I and 3′–Xho I sites. The PCR products were cloned using Nco I/Xho I in a modified pGEX-4T3 vector (pGEX-4T3-2xTEV-cHis) that codes for an N-terminal GST followed by a sequence recognized by the tobacco etch virus protease (TEV site) and a C-terminal TEV site and a 6×His tag. The cDNA of the Cdc25HD was amplified by PCR adding Nco I and Bam HI sites and was inserted in a variant of the pETEV15b vector (64) with unique Nco I and Bam HI sites for cloning.

The GST-fusion constructs of the SH3b and its fragments were created in the vector pGEX-4T3-TEV that codes for a TEV site downstream the GST and has Nde I, Nco I, and Bam HI sites for cloning. The cDNAs of the SH3b fragments 274-646, 372-646, 501-646, and 579-646 were PCR-amplified using primers that added Nco I and Bam HI sites, which were used for ligation in pGEX-4T3-TEV. cDNAs coding for fragments 537-588, 537-646, and 545-646 were PCR-amplified with Nde I and Bam HI sites at the 5′ and 3′ ends and were cloned in pGEX-4T3-TEV using these sites. The cDNA of the fragment 537-646 was also cloned in pETEV15b using the same restriction sties. Constructs coding for fragments 274-371, 274-500, 274-578, 501-536, 501-578, 537-560, 537-569, 537-578, 545-560, and 545-569 in pGEX-4T3-TEV were created by introducing a stop codon after the desired positions of constructs coding longer fragments. Stop codons and other point mutations in C3G constructs were introduced by PCR using the QuikChange method. Oligonucleotides for site-directed mutagenesis are shown in table S6.

The cDNA coding for residues 1-167 of human Rap1b was amplified from the IMAGE clone 2900837 by PCR. Primers added Nde I and Bam HI sites at the 5′ and 3′ ends, respectively. The amplified DNA was cloned in between the Nde I and Bam HI sites of the pETEV15b vector. The cDNA of human CrkL was amplified by PCR, adding Nde I and Bgl II sites at the 5′ and 3′ ends, respectively, and was cloned into the Nde I and Bam HI sites of pETEV15b. The cDNAs of the SrcKD (residues 254-536) and the phosphatase YopH were cloned respectively in the first and second multicloning regions of a modified pETDuet-1 vector that codes for an N-terminal His-tag and TEV site in the first expression region. The cDNA of YopH was PCR-amplified using as template the plasmid pEF5HA-YopH (a gift from A. Alonso) (65); the primers added Nde I and Xho I sites at the 5′ and 3′ ends, respectively, which were used for cloning to create pETDuet-1-YopH. The cDNA of the SrcKD was amplified by PCR from the plasmid pcDNA3-c-SRC (Addgene no. 42202, a gift from R. Lefkowitz) (66); the primers added Eco RI and Hind III sites at the 5′ and 3′ ends, respectively, and the DNA was ligated into pETDuet-1-YopH using these sites to create pETDuet-1-SrcKD-YopH. Oligonucleotides for Rap1b, CrkL, Src, and YopH are shown in table S7.

Expression and purification of recombinant proteins

All proteins were produced in Escherichia coli strain BL21(DE3) grown in Terrific Broth medium, and expression was induced with 0.2 mM isopropyl-β-d-thiogalactopyranoside at 15°C overnight. Proteins full-length C3G, C3G-ΔNTD, C3G-454-1077, and C3G-530-1077, which carry GST- and His-tags, were purified by two-step affinity chromatography. In the first step, they were purified by immobilized metal affinity chromatography (IMAC) using 5 ml of His-Trap Ni2+-chelating columns (GE Healthcare) as previously described (67). Subsequently, the sample was loaded in a 6-ml glutathione-agarose column [Agarose Bead Technologies (ABT)] and was washed with buffer A [20 mM tris-HCl and 300 mM NaCl (pH 7.5)] first supplemented with 0.1% Triton X-100 and second without detergent. The GST-C3G-His proteins were eluted with 20 mM reduced glutathione in buffer A. The GST- and His-tags were removed by digestion with recombinant TEV protease (rTEV) for 3 hours at room temperature, and the sample was dialyzed against buffer A. Free GST was removed by a reverse glutathione-affinity chromatography. C3G proteins eluted in the flow-through, were concentrated, and further purified by size exclusion chromatography (SEC) in a Superdex 200 10/300 GL column (GE Healthcare) equilibrated in buffer A. Proteins were concentrated by ultrafiltration, flash-frozen in liquid nitrogen, and stored at −80°C until they were used.

His-tagged Cdc25HD was purified by IMAC in a His-Trap column as above with minor changes. Bacteria were lysed in 20 mM Hepes, 500 mM NaCl, and 10 mM imidazole (pH 7.0) (buffer H). In the IMAC, after loading the sample, the column was washed with buffer H, and the Cdc25HD was eluted with 500 mM imidazole in the same buffer. Immediately after the IMAC, the buffer was changed to 20 mM citrate and 250 mM NaCl (pH 6.0) (buffer C) using a Sephadex G25 desalting column. The His-tag was removed by digestion with rTEV, and the Cdc25HD was further purified by SEC in the Superdex 200 10/300 GL column equilibrated in buffer C.

GST-SH3b fragments and GST-RalGDS Ras-binding domain (RBD; residues 701-851) were purified by affinity chromatography using glutathione-agarose columns (ABT). Proteins were eluted with phosphate-buffered saline (PBS) supplemented with 20 mM reduced glutathione and were extensively dialyzed against buffer B [20 mM tris-HCl and 150 mM NaCl (pH 7.5)].

His-tagged C3G-537-646 and CrkL were purified by IMAC, and the His-tag was removed by rTEV digestion as previously described (67). Rap1b (1-167) was produced by coexpression with the GroEL/GroES chaperones (plasmid pBB541, a gift from B. Bukau; Addgene no. 27394) (68). Rap1b was purified by IMAC, the His-tag was cleaved with rTEV, and the protein was subjected to SEC in the Superdex 200 10/300 GL column equilibrated in buffer B with 10 mM MgCl2. SrcKD was produced in E. coli by coexpression with the phosphatase YopH as previously described (69).

Western blotting

Proteins were separated by tris-glycine SDS–polyacrylamide gel electrophoresis (PAGE) and transferred onto Immobilon-P membranes (Millipore). The following primary antibodies were used at 1/1000 dilution unless otherwise indicated: mouse monoclonal antibody (mAb) against HA (HA11, Covance), rabbit polyclonal antibody (pAb) against HA Y11 (sc-805, Santa Cruz Biotechnology), mouse mAb M2 against FLAG (F1804, Sigma-Aldrich), mouse mAb against polyHis (H1029, Sigma-Aldrich), mouse mAb G9 against C3G (sc-393836, Santa Cruz Biotechnology), mouse mAb B-2 against green fluorescent protein (GFP) (sc-9996, Santa Cruz Biotechnology), mouse mAb 4G10 against pTyr (05-321, Millipore), rabbit pAb 121 against Rap1 (sc-65, Santa Cruz Biotechnology), and mouse mAb against β-actin (dilution 1/2000; AC-15, Sigma-Aldrich).

The following secondary antibodies were used (1/5000 dilution): horseradish peroxidase (HRP)–conjugated goat antibody against rabbit immunoglobulin G (IgG) (sc-2004, Santa Cruz Biotechnology), HRP-linked sheep antibody against mouse IgG (NXA931, GE Healthcare), goat pAb DyLight 680 against mouse IgG, goat pAb DyLight 680 against rabbit IgG, goat pAb DyLight 800 against mouse IgG, and goat pAb DyLight 800 against rabbit IgG (Thermo Fisher Scientific). Signals of DyLight secondary antibodies were detected by infrared fluorescence using an Odyssey Imaging System (LI-COR Biosciences) and were quantified using the Odyssey Application Software. HRP-conjugated secondary antibodies were detected by enhanced chemiluminescence (ECL) using Clarity Western ECL Substrate (Bio-Rad) and Super RX-N films (Fujifilm); films were scanned using an Epson V700-PHOTO scanner, and the intensities of the bands were quantified using the program ImageJ (70).

Cell lines

HEK293T cells and COS-1 cells were grown in Dulbecco’s modified Eagle’s medium (Sigma-Aldrich) with 10% fetal bovine serum (FBS; Life Technologies), penicillin (100 U/ml), and streptomycin (100 μg/ml) (Life Technologies) at 37°C and 5% CO2. Cells were transfected with plasmids using polyethylenimine (PEI; Polysciences Inc.) (PEI:DNA ratio 2.5:1). IL-3–dependent mouse Ba/F3 cells were grown in RPMI 1640 medium (Sigma-Aldrich) with 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml), supplemented with 20% conditioned medium of the WEHI-3B cell line as a source of IL-3.

coIP assays

COS-1 cells were transfected with HA-C3G constructs and FLAG-NTD or the empty pGEFLAG vector. After 36 to 48 hours, cells were lysed in lysis buffer consisting of 20 mM tris-HCl (pH 7.5), 150 mM NaCl, 0.5% Triton X-100, 1 mM sodium orthovanadate, 25 mM NaF, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1× protease inhibitor cocktail (cOmplete, Roche). Debris was removed by centrifugation (16,000g, 15 min, 4°C), and the supernatants were incubated with monoclonal anti-FLAG M2 agarose (Sigma-Aldrich) at 4°C for 1 hour. The beads were washed three times with lysis buffer. Bound proteins were extracted in sample buffer for electrophoresis and boiled. Samples of the cell lysates and the IPs were analyzed by Western blot.

Affinity PD assays

Purified GST-fusion proteins produced in E. coli were used as baits in PD assays against proteins either expressed in HEK293T cells or produced in bacteria. In the first case, HEK293T cells were transfected with plasmids expressing HA- or mEGFP-tagged C3G constructs and, 24 to 48 hours later, were lysed as above. Samples were clarified by centrifugation (16,000g, 15 min, 4°C), and the total protein content was estimated by Bradford assay. Typically, lysates containing 1 to 2 mg of total protein were incubated with 30 μg of GST-SH3b fragments, 0.01% bovine serum albumin (BSA), and 20 μl of glutathione-agarose resin at 4°C for 20 min. The resin was washed four times with buffer B with 0.1% Triton X-100, and proteins were released by adding Laemmli sample buffer. HA- and GFP-tagged proteins were detected in cell lysates, and PD fractions were detected by Western blot with specific antibodies; the GST-SH3b proteins in the PD fraction were detected by Coomassie staining of the gels. PDs of proteins produced in bacteria were done in a similar manner. In this case, ~30 μg of recombinant proteins was incubated in the presence of 0.01% BSA with 30 μg of the GST-fusion proteins prebound to 20 μl of glutathione-agarose resin. In competition experiments, increasing amounts of CrkL were added to the reaction samples. Incubation, washing, and release of bound proteins were done as above. The PD fractions were analyzed by SDS-PAGE, and the proteins were detected either by Coomassie staining (for example, CrkL) or by Western blot.

Fluorescence FRET measurements

Fluorescence of C3G conformational FRET biosensors was measured in COS-1 cells transiently expressing fluorescently tagged C3G constructs. Cells were trypsinized 48 hours after transfection, were collected by centrifugation, and were suspended in PBS prewarmed at 37°C. Fluorescence was measured in a FluoroMax-3 fluorometer (HORIBA Jobin Yvon) at 37°C using a 5-mm by 5-mm quartz cuvette. Samples were excited at 435 nm, and emission was acquired from 450 to 600 nm. The FRET signal was quantified as the ratio between the fluorescence intensities at 527 and 475 nm, which are the emission maxima of Venus and CFP, respectively. For representation, the spectra were normalized to the intensity at 475 nm.

Protein phosphorylation with Src

For in vitro phosphorylation, C3G proteins were prepared at 20 to 100 μM in kinase buffer [50 mM tris-HCl (pH 7.5), 300 mM NaCl, 10 mM MgCl2, 1 mM dithiothreitol (DTT), and 1 mM adenosine 5′-triphosphate]. Recombinant SrcKD was added at 1 μM, and the reaction was carried out at 30°C for 20 min unless otherwise specified. Phosphorylation was analyzed by Western blot against pTyr.

Nucleotide exchange activity assays

The nucleotide exchange activity of C3G toward Rap1b was determined in vitro using the GDP fluorescent analog 2′-deoxy-3′-O-(N′-methylanthraniloyl)guanosine-5′-O-diphoshpate (mant-dGDP) (Biolog), as previously described (71). The fluorescence intensity of mant-dGDP bound to the GTPase is about twice the intensity of the free form in aqueous solution. Rap1b (200 μM) was loaded with mant-dGDP by incubation in 20 mM tris-HCl (pH 7.5), 50 mM NaCl, and 4 mM ethylenediaminetetraacetic acid (EDTA) with a 10-fold excess (2 mM) of mant-dGDP for 90 min at 4°C. Afterward, 10 mM MgCl2 was added, and the protein was separated from the excess of mant-dGDP by SEC using the Superdex 200 10/300 GL column equilibrated in buffer G [50 mM tris-HCl (pH 7.5), 150 mM NaCl, and 10 mM MgCl2]. Exchange reactions were done at 25°C using 200 nM Rap1b:mant-dGDP in buffer G and were initiated by adding 40 μM GDP. Unless otherwise indicated, experiments were done in the presence of 1 μM C3G proteins, which is within a range where the kobs are directly proportional to the concentration of C3G (fig. S1). Fluorescence was measured in a FluoroMax-3 fluorometer (HORIBA Jobin Yvon) using 3-mm by 3-mm quartz cuvettes. The dissociation of mant-dGDP was monitored exciting at 370 nm (1 nm bandwidth) and registering the emission at 430 nm (10 nm bandwidth). The sample was illuminated only during data acquisition (typically 600 data points per time course) to prevent photobleaching. Data were recorded until the signal reached a plateau. The time-dependent decay of the fluorescence intensity could be described by the following single exponential decay modelIt=A0 ekobst+Bwhere It is the fluorescence intensity at each time point, B is the fluorescence of free mant-dGDP, A0 + B is the initial intensity before the reaction is started, and kobs is the apparent dissociation rate constant. Fitting was done using the program SigmaPlot. Data were normalized to A0 + B = 1 and B = 0 for representation.

Analysis of dose-response experiments

Data of the inhibition of the Cdc25HD by multiple concentrations of the AIR fragment were analyzed fitting the following logistic Hill equationk[AIR]=kmin+(kmaxkmin)1+(IC50[AIR])HillSlopewhere k[AIR] is the kobs determined at a given concentration of the AIR, kmax and kmin are the maximum and minimum values of kobs, IC50 is the concentration of AIR that produces the half-maximal inhibition, and HillSlope is the parameter that describes the steepness of the curve. Fitting was done with the program SigmaPlot.

Lentiviral production and cell transduction

Lentiviral particles were produced by transient transfection of HEK293T cells. WT and mutant C3G constructs in pLenti-C-mEGFP-IRES-BSD lentiviral transfer vector, or the empty vector, were cotransfected with plasmids pMD2.G (provided by D. Trono; Addgene no. 12259) and pCMV-deltaR8.91 (Lifescience Market). Lentiviral particles were collected from the medium at 24 and 48 hours by centrifugation (2 hours, 30,000g). Ba/F3 cells were infected with the recombinant viruses in the presence of polybrene (8 μg/ml) (hexadimethrine bromide, Sigma-Aldrich). After incubation at 37°C for 15 hours, infected cells were selected with blasticidin (20 μg/ml; InvivoGen) at least 3 weeks before analysis, which resulted in more than 90% of the cells expressing GFP.

Rap1 activation assay in cells

Rap1-GTP was detected in HEK293T and Ba/F3 cells by PD using GST-RalGDS-RBD, which binds selectively to the active conformation of Rap1. HEK293T cells in a 9-cm plate were lysed in 0.5 ml of magnesium lysis buffer (MLB) [25 mM Hepes (pH 7.5), 150 mM NaCl, 1% Igepal CA-630, 10 mM MgCl2, 1 mM EDTA, 2% (v/v) glycerol, 1 mM Na3VO4, 1 mM PMSF, 1× protease inhibitor cocktail (Roche), and 1 mM DTT] supplemented with 30 μg of purified GST-RalGDS-RBD protein. Samples were clarified by centrifugation (16,000g, 15 min, 4°C). Twenty microliters of glutathione-agarose resin was added to each sample and was incubated 10 min at 4°C. The resin was washed three times with 1 ml of MLB. Bound proteins were extracted and denatured with Laemmli buffer supplemented with 10 mM DTT for 12 to 20 hours at room temperature. Proteins were analyzed by SDS-PAGE and Western blot using anti-Rap1 antibody and HRP-conjugated secondary antibody. Total Rap1 was detected in cell lysates, and Rap1-GTP was detected in the PD. GST-RalGDS-RBD was detected in the PD samples by staining with Ponceau S (membranes) or Coomassie (gels). At least two independent experiments were performed. The levels of Rap1-GTP in cells expressing C3G-mEGFP-CAAX, which is constitutively attached to the membrane and produced the highest activation of Rap1, were used to normalize data from multiple experiments. The levels of Rap1-GTP in Ba/F3 cells were analyzed in a similar manner using 5 × 106 cells. For time course experiments of IL-3 stimulation, Ba/F3 cells were starved for 12 hours in media lacking FBS and IL-3, and then cells were stimulated by adding 20% WEHI-3B–conditioned media, with IL-3 but without FBS.

Flow cytometry analysis of integrin LFA-1 activation

The percentage of Ba/F3 cells expressing C3G-mEGFP, WT or mutants, or mEGFP alone with activated integrin LFA-1 was determined by flow cytometry using the antibody mAb24 (ab13219, Abcam), which recognizes the activated conformation of this integrin. Ba/F3 cells were washed in ice-cold PBS containing 3% BSA and were stained with mAb24 (1/100 dilution; 60 min, 4°C). Cells were washed and stained with Cy5-conjugated goat anti-mouse IgG secondary antibody (1/200 dilution; 30 min, 4°C) (Jackson ImmunoResearch). Cells were washed; resuspended in cold PBS, 3% BSA, and 1% sodium azide; and analyzed using an Accuri C6 flow cytometer (BD Biosciences). Cells processed in a similar manner but without the primary antibody were used as a control of unspecific labeling by the secondary antibody. Data were analyzed using the BD Accuri C6 software.

Protein sequence analysis

The search for C3G orthologs was done using the HMMER server (72). Initially, a search using as bait a segment of the NTD (residues 89-250) of human C3G identified sequences from 119 species, which exhibit a C3G-like domain architecture that contains NTD, REM, and Cdc25H domains. These sequences were used to create a hidden Markov model profile that was used in a second search, which identified sequences from 90 additional species. The 209 sequences were aligned with Clustal Omega (73), and conservation scores were calculated using the ConSurf server (74). Sequence-based secondary structure predictions were done with the methods Porter (75), PsiPred (76), and JPred4 (77).

Molecular modeling

The structure of the REM domain of C3G was modeled with Modeller (78) using the ProtMod server ( The structure of the REM of SOS (Protein Data Bank entry 3KSY) (46), which shares 18% sequence identity with the REM of C3G, was used as template. Molecular figures were created with PyMOL (79).

Quantification and statistical analysis

Data are presented as scatterplots with lines marking the means ± SD of independent experiments. Statistical significances of two-group comparisons were analyzed with two-tailed unpaired Welch’s t test. Comparisons of several groups with respect to a control group were done with one-way analysis of variance (ANOVA) with Dunnett’s test when variances were equivalent or with Brown-Forsythe ANOVA with Dunnett’s T3 test when variances were not equivalent. Differences in mean values were considered statistically significant at P < 0.05. Significance levels are as follows: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Statistical analyses were done with the program GraphPad Prism 8.


Fig. S1. Concentration dependence of the nucleotide exchange activity of C3G proteins.

Fig. S2. Secondary structure prediction for the SH3b domain of C3G.

Fig. S3. Effect of phosphorylation and CrkL on the GEF activity of C3G proteins.

Fig. S4. GEF activity of WT and mutant C3G in which the NTD/REM interaction was absent or weakened.

Table S1. List of C3G orthologs identified with HMMER.

Table S2. Human cancer-associated somatic missense mutations in the AIR-CBR of C3G.

Table S3. Constructs of human C3G for expression in mammalian cells.

Table S4. Oligonucleotides used to create constructs of C3G.

Table S5. Constructs of human C3G for expression in E. coli.

Table S6. Oligonucleotides used for site-directed mutagenesis of C3G.

Table S7. Oligonucleotides used to create constructs of Rap1b, CrkL, Src, and YopH for expression in E. coli.

References (8085)


Acknowledgments: We thank A. Alonso (Instituto de Biología y Genética Molecular, Valladolid, Spain), B. Bukau (EMBL, Heidelberg, Germany), R. Lefkowitz (Duke University Medical Center, Durham, NC), and D. Trono (Ecole Polytechnique Fédérale de Lausanne, Switzerland) for providing plasmids and J. M. Sánchez Santos (University of Salamanca, Spain) for assessing the statistical analysis. Funding: This work was supported by the Spanish Ministry of Economy, Industry and Competitiveness (MINECO), Agencia Estatal de Investigación (AEI), and the European Regional Development Fund (ERDF) (grants BFU2015-69499-P and PID2019-105763GB-I00 to J.M.d.P. and SAF2016-76588-C2-2-R and PID2019-104143RB-C21 to C.G.) and by the Consejería de Educación, Junta de Castilla y León (grant SA017U16 to C.G. and J.M.d.P.). A.C. received funding from MINECO (FPU14/06259). M.G.-H. and A.R.-B. were funded by Banco Santander and the University of Salamanca. The authors’ institution is supported by the Programa de Apoyo a Planes Estratégicos de Investigación de Estructuras de Investigación de Excelencia cofunded by Castilla y León autonomous government and ERDF (CLC-2017-01). Author contributions: A.C., M.G.-H., and J.M.d.P. conceived the study and designed experiments. C.G. designed and contributed to the interpretation of assays in cells. M.G.-H. characterized the NTD/REM interaction. A.C. performed most of the other experiments. S.d.C. purified functional C3G-Cdc25HD. A.R.-B. performed in vitro activity assays. A.M.-V. performed assays in cells. P.G.-S. performed part of the analysis of the NTD/REM interaction. A.C., C.G., and J.M.d.P. wrote the paper with input from all authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article