|Home | About | Journals | Submit | Contact Us | Français|
Vav proteins are guanine nucleotide exchange factors (GEFs) for Rho family GTPases. They control processes including T cell activation, phagocytosis and migration of normal and transformed cells. We report the structure and biophysical and cellular analyses of the five-domain autoinhibitory element of Vav1. The catalytic Dbl homology (DH) domain of Vav1 is controlled by two energetically coupled processes. The DH active site is directly, but weakly, inhibited by a helix from the adjacent Acidic domain. This core interaction is strengthened 10-fold by contacts of the calponin homology (CH) domain with the Acidic, pleckstrin homology and DH domains. This construction enables efficient, stepwise relief of autoinhibition: initial phosphorylation events disrupt the modulatory CH contacts, facilitating phosphorylation of the inhibitory helix and consequent GEF activation. Our findings illustrate how the opposing requirements of strong suppression of activity and rapid kinetics of activation can be achieved in multi-domain systems.
Vav proteins are guanine nucleotide exchange factors (GEFs) for small GTPases in the Rho family (Bustelo, 2001; Tybulewicz, 2005). They play important roles in actin regulatory pathways in numerous cell types and control diverse processes including immune cell development and activation, neuronal development and angiogenesis. N-terminal truncations of Vav proteins lead to cell transformation (Katzav et al., 1989), and the wild type proteins have been implicated in the development and severity of a variety of cancers, including those of the brain, pancreas and skin (Dong et al., 2006; Katzav, 2007). Vav1 is expressed in haematopoietic cells, where it activates the Rac GTPase in response to Src- and Syk-family kinase signals downstream of the T cell, B cell and Fcγ receptors. It controls functions such as the antigen response of T cells, phagocytosis by macrophages and superoxide production by neutrophils (Hall et al., 2006; Tybulewicz, 2005; Utomo et al., 2006). Vav2 and Vav3 are more widely expressed and play similarly important roles in development and cell function throughout the body (Cowan et al., 2005; Sauzeau et al., 2006).
Vav proteins are composed of eight domains: calponin homology (CH), Acidic (Ac), Dbl homology (DH), pleckstrin homology (PH), zinc finger (ZF), Src homology 3 (SH3), Src homology 2 (SH2) and a second SH3. The DH domain catalyzes guanine nucleotide exchange in Rho family GTPases (Abe et al., 1999), and many Vav functions have been ascribed to this activity. The five N-terminal domains (abbreviated CADPZ; other fragments are designated analogously) function together to regulate the biochemical activity of the DH domain; the C-terminal Src homology domains are thought to mediate cellular localization (Fig. 1A) (Bustelo, 2001; Tybulewicz, 2005). The GEF activity of Vav1 is autoinhibited by binding of a helix from the C-terminus of the Ac domain (residues 167–178) into the active site of the DH domain, blocking access to substrate (Abe et al., 1999; Aghazadeh et al., 2000; Lopez-Lago et al., 2000). This core inhibitory interaction is relieved by phosphorylation of Tyr174 in the inhibitory helix, which melts the helix and displaces it from the DH domain (Aghazadeh et al., 2000; Han et al., 1998). In the cell, Tyr174 is phosphorylated within seconds of receptor stimulation by Src- and Syk-family kinases, rapidly initiating downstream signaling. We recently demonstrated that an Ac-DH fragment of Vav1 rapidly fluctuates between a “closed” ground state, where the helix is bound to the active site, and an “open” excited state resembling the phosphorylated form, where the helix is dissociated and melted (Li et al., 2008). The closed state is favored by ~10:1, and correspondingly this equilibrium decreases both GEF activity and the rate of Tyr174 phosphorylation by ~10-fold, reflecting accessibility of the active site and the Tyr174 sidechain, respectively. Thus, the ratio of the closed and open states determines the magnitude of autoinhibition and rate of activation of the helix-DH core.
A variety of data indicate that interactions of the CH domain further suppress the activity of the DH domain, and that this suppression is critical for Vav1 function in the cell. For example, the original Vav oncogene encoded a protein lacking part of the CH domain (Katzav et al., 1989). Later studies showed that CH truncation increases the transforming activity of Vav1 and its homolog Vav3, concomitant with increased GEF activity in vitro and in vivo (Bustelo, 2001; Katzav et al., 1991; Llorca et al., 2005). A cryoelectron microscopy reconstruction of full length Vav3 has suggested that these effects arise from intramolecular binding of the CH domain to the ZF domain, and to a surface of the DH domain adjacent to the Ac helix interface (Llorca et al., 2005). But this mechanism has not been examined in atomic resolution structural analyses. Moreover, it is not clear how the CH interactions affect the helix-DH core to control GEF activity. Studies of other multi-domain systems have shown that the construction of Vav1, with a core autoinhibitory element whose biochemical activity is modulated by other domains, is widespread (DiNitto et al., 2007; Moarefi et al., 1997; Prehoda et al., 2000; Sondermann et al., 2004; Yohe et al., 2008). Thus, understanding how cooperativity is achieved in Vav1 should inform generally on regulation of multi-domain proteins.
Although phosphorylation of Tyr174 is clearly important to Vav1 activation, several pieces of data suggest the process is more complicated than modification of this single site. In addition to Tyr174, Vav1 is also phosphorylated at two additional sites in the N-terminus of the acidic element, Tyr142 and Tyr160 upon receptor stimulation (Lopez-Lago et al., 2000; Miletic et al., 2006). Mutation of these sites increases Vav1 signaling activity, suggesting that they may also contribute to autoinhibition in the wild type protein and that their phosphorylation could play a role in activation. The mechanisms by which Tyr142 and Tyr160 contribute to regulation of Vav1, and more generally the pathway(s) to Vav1 activation, are not understood.
Here we report the structure of the five-domain regulatory element of Vav1, CADPZ. The structure shows that the CH domain and the N-terminus of the Acidic element bind to each other and to a platform formed by the PH domain and a C-terminal extension of the DH domain. NMR, biochemical and cell biological analyses show that these interactions suppress GEF activity by modulating the core helix-DH equilibrium, shifting it toward the inhibited state by approximately 10-fold, likely by restraining the inhibitory helix to the DH domain. Phosphorylation of Tyr142 and Tyr160 relieves the modulatory interactions, making Tyr174 more accessible to kinases, suggesting a sequential activation mechanism for full length Vav1. The layered construction of Vav1 provides a means of achieving strong suppression of activity while still maintaining a route to rapid activation, features that are probably general among multi-domain systems in biology.
To understand how interdomain contacts act together to suppress GEF activity, and how phosphorylation drives activation, we determined the crystal structure of the CADPZ element of Vav1 (Table S1, Figs. 1B, C) using data to a resolution of 2.73 Å. The core module, consisting of the Tyr174-containing inhibitory helix and DH domain, closely resembles the previously reported solution structure of this fragment (Aghazadeh et al., 2000). The inhibitory helix lies perpendicular to the long axis of the DH domain in the bottom of the active site. The PH and ZF domains dock onto the C-terminus and center of an extended C-terminal helix of the DH domain, respectively, and make extensive contacts with each other as well. The organization of the DPZ portion of the structure closely resembles that seen in recently reported structures of the DPZ-Rac complex (Chrencik et al., 2008; Rapley et al., 2008). However, a small bend near the center of the C-terminal DH helix somewhat changes the relative orientations of the DH domain and PZ unit in the two structures (Fig. S1A, B). The CH domain binds the PH domain and the extended C-terminal helix of the DH domain, but does not contact the ZF domain, in contrast to previous inferences from EM and biochemical data (Llorca et al., 2005). The CH domain contacts the helix-DH element only through Arg58 (CH), whose sidechain is near that of Glu169 (helix).
The Ac element emerges from the side of the CH domain opposite the DH domain. A short N-terminal helix (residues 143–150) wraps around the CH domain and also contacts the PH domain. This is followed by a long extended element (residues 151–166) that connects to the inhibitory helix packed in the DH active site (Figs. 1B, C). Tyr142 is located immediately before the N-terminal Ac helix; Tyr160 is located in the extended element. Both sidechains lie on the surface of the CH domain, with their aromatic rings contacting primarily hydrophobic residues (Tyr142: Val30, Gly106, Ile109, Leu145; Tyr160: Val41, Leu159, Arg63) (Fig. S1C, D). The hydroxyl group of Tyr160 forms a hydrogen bond with the sidechain of Asp39 (Fig. S1D).
The overall domain organization of the structure here is quite different from the previously reported EM reconstruction of Vav3, which is 58 % identical to Vav1 in amino acid sequence (see Supplementary Materials). Most significantly, the contacts of the CH domain to the PH domain and extended C-terminal helix of the DH domain are not possible in the previous description. To validate the organization observed in the crystal, we mutated a series of residues at different contact surfaces of the CH domain, and compared the GEF activity of these proteins to several benchmarks. In the wild type CADPZ protein, GEF activity is inhibited by the effects of the Acidic helix on the DH active site and by the modulation of this core interaction by the interdomain contacts of the CH domain (referred to as modulatory interactions hereafter). Thus, as shown in Figure 2A, CADPZ has extremely low GEF activity toward Rac. Truncation of the CH domain (giving the ADPZ protein) results in partial activation, consistent with the absence of modulatory interactions of the CH domain, but the remaining presence of the core inhibitory interaction of the Acidic helix with the DH active site. Phosphorylation of CADPZ at all three tyrosine sites in the Acidic region (giving pCADPZ) disables both the modulatory and core elements of the autoinhibitory apparatus (Fig. S2A), leading to maximal GEF activity. GEF activity of fully active (mimicked by ADPZY174D), partially active (ADPZ) and maximally inhibited (CADPZ) proteins parallels affinity for Rac, as qualitatively assessed in pull-down assays (Fig. 2B), further supporting the idea that activity reflects accessibility of the DH active site (Li et al., 2008).
The activities of wild type CADPZ, ADPZ and pCADPZ provide reference points for analysis of a series of CADPZ mutants. These were designed based on the structure of the CH-Ac interface with the DPZ element (Fig. 3). The CH-Ac element contacts DPZ through two groups of residues. In the first group, the α3–α4 loop of the CH domain (residues 66–68) makes parallel β-sheet-like interactions with the first β-strand of the PH domain (residues 406–408); inter-strand hydrogen bonds are observed between the sidechain of Asp406 and mainchain NH of Phe69, and Ser67 is immediately adjacent to Asp406. (Fig. 3A). Additional contacts in this region occur within a hydrophobic cluster that contains Phe69 (CH), Phe425 (PH), Tyr441 (PH) and Phe386 (DH extended helix). The second interaction group is immediately adjacent, with the N-terminal helix of the Ac element contacting the basic β5–β6 and β7-α1 loops of the PH domain. Several oppositely charged sidechains are in close proximity across this interface, including Asp150 (Ac) and Lys487 (PH) (Fig. 3B). CADPZ mutations S67D, F69D and K487E were designed to strongly destabilize the β-strand, hydrophobic cluster and Ac-PH interactions, respectively. These CADPZ mutants had GEF activity ~10% of pCADPZ, very similar to that of ADPZ, which lacks the CH domain entirely. F69A and D406A mutations caused smaller increases in GEF activity, consistent with less severe perturbations of the CH interactions (Fig. 2A). Mutation of Met66, a residue near the CH-PH interface but making no interdomain contacts had no effect on GEF activity (not shown). Our crystal structure is also consistent with the activities of several previously reported CH domain mutants (Figs. S1E–G, Table S2).
The GEF activity of Vav1 is also manifest in signaling strength in cells. Previous reports have shown that overexpression of Vav1 proteins lacking the CH domain leads to increased Rac activation and cell transformation (Bustelo, 2001; Han et al., 1998; Llorca et al., 2005; Lopez-Lago et al., 2000), consistent with the increased GEF activity of the ADPZ protein. Further truncation to remove the Ac domain inhibitory helix increases cellular activity further (Abe et al., 1999), consistent with loss of the core inhibitory element. As shown in Figure 2C, destabilizing the modulatory interactions through the S67D mutation also increases Rac activation in cells. Rac activation is higher in cells expressing the Y174D helix mutant, consistent with higher GEF activity of the corresponding CADPZ protein (Fig. 2A), and in the S67D/Y174D double mutant. Similar behavior is observed in focus forming assays, where wild type Vav1 is much less active than the S67D, Y174D and double mutants (Fig. S2B–D). The similarity in this assay between the three mutants likely arises because the long timecourse of the assay flattens differences between partially and fully active proteins. Nevertheless these data suggest that the additional modulatory interactions are relevant for the control of Vav activity under non-stimulated conditions. The combined biochemical and biological data validate the crystal structure and illustrate the importance of the modulatory contacts in controlling GEF activity in vitro and in vivo.
The data above demonstrate that the modulatory interactions of the CH and Ac elements with the PH domain further suppress the GEF activity of the helix-DH core. However, they do not reveal how this suppression occurs. One possibility is that interactions of the CH domain shift the autoinhibitory helix-DH equilibrium to further favor the helix-bound state. That is, the helix-DH equilibrium may be thermodynamically coupled to the intramolecular binding equilibria of the CH domain such that the bound state of one favors the bound state of the other and vice versa. While analogous mechanisms have been proposed to explain interdomain cooperativity in other multi-domain systems (DiNitto et al., 2007; Prehoda et al., 2000; Sondermann et al., 2004; Young et al., 2001), these hypotheses have not to our knowledge been tested by direct measurement of the regulatory equilibria.
To test this hypothesis here, we used NMR to directly examine the autoinhibitory helix-DH equilibrium in the AsD protein (where As indicates a shortened Acidic region containing only the inhibitory helix) and CADPZ. We recently showed that in AsD the inhibitory helix fluctuates in and out of the DH active site on a μs-ms timescale, with the helix-bound and helix-free states in an ~10:1 ratio (Li et al., 2008). These fluctuations are reflected in a large chemical exchange contribution to transverse relaxation of NMR signals arising from methyl groups in the protein (large ΔR2, the difference between multiple quantum 1H/13C transverse relaxation measured at the lowest and highest transverse fields (Fig. S3A (Korzhnev et al., 2004)). In CADPZ, seven methyl signals that report on the helix-DH interactions are sufficiently free of overlap to be analyzed quantitatively: Leu177δ1 and Leu180δ1,2 from the helix, Val183γ1,2 from the helix-DH linker and Leu325δ1,2 from the DH domain. ΔR2 values for all these methyl groups are significantly reduced in CADPZ (Fig. S3A), indicating either a shift in populations toward the helix-bound state, a change in the fluctuation rate or both. This reduction in ΔR2 requires more than addition of the CH domain, as CAD and AD proteins have very similar values (Fig. S3A).
We used a chemical shift-based approach to quantify the populations of helix-bound and helix-dissociated states in AsD and CADPZ. The approach is based on the fact that if a system is fluctuating rapidly (fast exchange) between two states, then the chemical shift of any NMR signal will be a population weighted average of the chemical shifts of the pure states. Thus, when the chemical shifts of the pure states are known (or can be approximated), the chemical shift observed for any given protein reports on the relative populations of the two states in equilibrium. Using this approach we recently showed that a designed series of AsD mutants have different populations of helix-bound and helix-dissociated states across the autoinhibitory equilibrium (Li et al., 2008). These proteins ranged from ~5% to > 99% open. When their NMR spectra are overlaid, their signals lie on linear trajectories between those of pAsD and a nearly completely closed mutant. Three of the mutants, K208A, K208S and K208G populate predominantly the open state of AsD, with open populations of 74 ± 10 %, 83 ± 9 % and 93 ± 6 %, respectively (Li et al., 2008). To determine if the helix-DH equilibrium is different in CADPZ, here we examined methyl TROSY spectra of CADPZ proteins containing these same three mutations, as well as a Y174D mutant of CADPZ (CADPZY174D) that mimics the phosphorylated protein (not shown). As shown in Figure 4A, methyl chemical shifts from the helix and DH elements in pAsD and CADPZY174D are nearly identical, indicating that phosphorylation also shifts the helix-DH equilibrium in CADPZ to strongly favor the helix-dissociated state (Figs. 4A, S3B–D). Similarly, the helix and DH chemical shifts are in similar positions in AsD and CADPZ. Thus, the open and closed forms of the helix-DH element are very similar in AsD and CADPZ. Five of the interface methyl resonances in the CADPZ mutants are in fast exchange and free of overlap. As in AsD (Li et al., 2008), these resonances lie along the linear trajectory between the non-phosphorylated and phosphorylated chemical shifts (Figs. 4A, S3B-D). This linearity indicates that from the standpoint of the inhibitory interface, CADPZ exists in an equilibrium analogous to that in AsD, fluctuating between helix-bound and helix-dissociated states. However, for each mutant, the signals in CADPZ are shifted appreciably toward the non-phosphorylated position relative to those in AsD. Thus, the core inhibitory equilibrium is shifted toward the helix-bound state in CADPZ (Fig. 4A). These data are consistent with the reduction of ΔR2 in wild type CADPZ observed above. The observed chemical shifts indicate that on average the equilibrium is shifted ~10-fold for each mutant (range = 6–16-fold) (Fig. 4B). The shift in equilibrium toward the helix-bound state is also manifest in GEF activity. As in the AsD proteins (Li et al., 2008), GEF activity of CADPZ is appreciably lower than that of CADPZK208A, reflecting destabilization of autoinhibitory contacts in the mutant (black bars in Fig. 4C). Most importantly, the activities of the CADPZ proteins are significantly below those of their AsD counterparts due to the decreased accessibility of the DH active site (compare black and red bars in Fig. 4C).
These NMR data indicate a thermodynamic mechanism of interdomain cooperativity in which the 10-fold suppression of accessibility of the DH active site caused by the intrinsic thermodynamics of the AsD core is enhanced an additional 10-fold by modulatory interactions of the CH domain. How is coupling between these equilibria achieved structurally? In this regard, the paucity of contacts between the CH domain and the helix-DH core element is striking (Fig. 1C, right panel). Mutation of the lone direct CH-DH contact observed in the crystal (R58A) has no effect on GEF activity in CADPZ (not shown). This suggests that the CH domain acts on GEF activity in an indirect manner, perhaps by restraining the N-terminus of the Acidic domain against the DPZ element, and consequently entropically disfavoring the dissociated state of the inhibitory helix. One prediction of this structural mechanism is that cutting the Acidic element immediately before the helix should destroy the mechanical linkage between the modulatory and core elements, preventing the CH domain from affecting GEF activity. To test this hypothesis, we performed two titrations, monitoring GEF activity (Fig. 5). In the first, we added CH (residues, 1–134) to ADPZ (residues 135–584), an interaction with KD of 32 µM based on independent NMR analyses (not shown). As illustrated in Figure 5, the CH domain can act in trans to suppress GEF activity of ADPZ, with activity decreasing linearly with the fraction bound. Thus, a protein cleaved at the CH-Acidic junction can manifest coupling between the modulatory and core equilibria (see also Fig. S4A–C). This effect is specific, as titrations of CHS67D into ADPZ or CH into ADPZK487E produced much less inhibition (Fig. S4D). In contrast, addition of a CA construct lacking the inhibitory helix (residues 1–168) to AsDPZ (residues 169–584, KD = 290 µM) had no effect on GEF activity. Thus, cleaving the Ac element before the inhibitory helix prevents the modulatory equilibrium from communicating to the core equilibrium. In an analogous titration of CA into deuterated, methyl protonated AsDPZ, monitored by methyl TROSY spectra of AsDPZ, resonances at the CH-PH interface moved linearly from their positions in free AsDPZ toward their positions in CADPZ (Figs. S4E, F). Thus, the complex reproduced many of the packing interactions of the intact protein. In the same titration, resonances at the helix-DH interface were not affected, indicating that the helix-DH equilibrium is not affected by CH-PH interactions when the Ac element is broken (Fig. S4G, compare to Fig. S4C). Thus, the decrease in GEF activity tracks with modulation of the helix-DH equilibrium, not simply binding of the CH domain to the PH domain. The data are consistent with the idea that the modulatory interactions act on the helix-DH equilibrium by restraining the N-terminus of the Ac element, disfavoring dissociation of the helix.
To investigate the idea of coupled equilibria in greater detail, we examined the NMR properties of the proteins described above (Fig. 2A) which are impaired in one or both elements of the inhibitory apparatus (modulatory or core inhibitory interactions). For each protein we examined two sets of NMR signals. The first consists of methyl groups located at the CH-PH interface (Leu409δ1,2, Leu491δ, Val102γ; Figs. 2D, S2E–G). If the CH domain equilibrates rapidly between the state observed in the crystal structure (with contacts to Ac and PH intact) and a state where the contacts to Ac and PH are lost, then the chemical shifts of these methyl signals should reflect the populations of the two states. This is exactly analogous to our analysis of the helix-DH equilibrium in Figure 4, under the assumption that contacts of the CH domain to Ac and PH are highly cooperative (i.e. CH is either fully bound to both elements or fully dissociated). When interactions between the CH and PH domains in CADPZ are destabilized by mutation, these “CH-PH signals”, which report on stability of the modulatory interactions, will shift toward the dissociated extreme. The bound extreme is represented by the CADPZ chemical shifts, and the dissociated extreme by the chemical shifts of ADPZ or pCADPZ, where the modulatory contacts are absent or fully destabilized, respectively (Figs. 2D, S2A). For each protein we also examined NMR signals of methyl groups at the helix-DH interface (Figs. 2E, S2H, I). As in Figure 4, these signals report on populations across the core inhibitory equilibrium. By comparing CH-PH and helix-DH signals for each mutant, we can observe how changes to one process affect the other. We can also correlate both to GEF activity. The behavior of ADPZ is demonstrative. In this protein the CH-PH signals are strongly shifted from their CADPZ positions, reflecting the absence of modulatory interactions. In contrast, since the core inhibitory module is intact, the helix-DH signals in ADPZ are shifted only partially from the CADPZ positions toward the pCADPZ positions (Fig. 2E). The GEF activity of ADPZ, which is dependent on exposure of the DH active site, parallels the helix-DH signals (Fig. 2A). In the CADPZ mutants, S67D, F69D and K487E, which were designed to destabilize the modulatory interactions, the CH-PH signals shift appreciably toward their open positions (Fig. 2D). The helix-DH signals shift toward the open position but remain in the range between the CADPZ and ADPZ positions (Fig. 2E). These changes are consistent with the GEF activities of these mutants, which are at or below that of ADPZ (Fig. 2A). The less perturbing mutations F69A and D406A show smaller shifts in the CH-PH signals, much smaller shifts in the helix-DH signals (Fig. 2E) and lower GEF activity (Fig. 2A). Thus, shifting the modulatory equilibrium changes the helix-DH equilibrium and consequently GEF activity. Perturbations can also be observed in reciprocal fashion: introducing the Y174D (helix) mutation into the F69A background shifts the CH-PH signals toward their positions in ADPZ (Fig. 2D). Thus, destabilizing the helix-DH interactions also destabilizes the modulatory interactions. These data support our model that the core and modulatory equilibria are thermodynamically coupled, and that this coupling is a major contributor to the cooperative suppression of GEF activity by interactions of the CH domain.
Our model for autoinhibition creates a conundrum regarding activation of Vav1. That is, how is Tyr174 rapidly accessed by kinases, as is known to occur in vivo (Miletic et al., 2006), when its exposure is strongly suppressed by the combined actions of the core and modulatory equilibria? We asked whether the Tyr142 and Tyr160 phosphorylation sites could be important in providing a kinetic pathway to activation. As shown in Figure 6A, phosphorylation at Tyr160 or Tyr142 had little effect on GEF activity. However, the doubly phosphorylated pTyr142/pTyr160 protein has activity similar to ADPZ. To generate materials on larger scale we mimicked phosphorylation by mutation of Tyr142 and/or Tyr160 to aspartate. The Tyr to Asp mutants have GEF activity very similar to their pTyr counterparts, indicating that the mutations act similarly to phosphorylation. The change in activity of the mutants occurs concomitantly with a shift in the helix-DH equilibrium toward the ADPZ position as assessed by NMR (Figs. 6B, S5A, B). A Y142F/Y160F double mutant has similar GEF activity and NMR properties as the double Asp mutant (not shown), explaining previous data showing that Y142F and Y160F mutations increase the transforming activity of Vav1 (Lopez-Lago et al., 2000). These data suggest that initial phosphorylation at Tyr142 and Tyr160 could destabilize the modulatory interactions, increasing access of Tyr174 to kinases. In support of this model, Tyr174 is phosphorylated by the Lck kinase domain much more rapidly in the Y142D/Y160D mutant than in wild type CADPZ (Fig. 6C). Moreover, in wild type CADPZ Tyr142 and Tyr160 are phosphorylated more rapidly than Tyr174 (Fig. 6C). This is striking since in short peptides containing the different Vav1 tyrosine motifs, phosphorylation of Tyr174 is strongly favored over Tyr142 and Tyr160 (Amarasinghe and Rosen, 2005). These data suggest an ordered pathway dictated by the structural organization of CADPZ in which Tyr142 and Tyr160 are modified initially, followed by Tyr174. Interestingly, mutation of the highly conserved Asp39, which coordinates the Tyr160 hydroxyl group, to asparagine decreases both the initial rate of phosphorylation of Tyr160 and also the rate of subsequent phosphorylation of Tyr142 and Tyr174 (Fig. S5C). Thus, Asp39 may be selected to tune the system into an appropriate response regime.
Tyr142 and Tyr160 would not be accessible to kinases in the conformation observed in the crystal. However, both sidechains are on the surface of the protein, and NMR relaxation data indicate substantial µs-ms timescale fluctuations of methyl groups surrounding them in the Ac element (Fig. S5D). These dynamics likely represent transitions to kinase accessible conformations, explaining how Tyr142 and Tyr160 can be phosphorylated rapidly. We previously showed that pTyr142 in Ac peptides provides a docking site for the SH2 domain of the Lck kinase, and that this interaction allows Lck to phosphorylate Tyr174 in unimolecular fashion with increased efficiency (4-fold higher kcat/KM) (Amarasinghe and Rosen, 2005). Upon initial phosphorylation of Tyr142/Tyr160, destabilization of modulatory contacts and recruitment of kinase should act synergistically to increase phosphorylation of Tyr174 by Lck. Thus, the structural and energetic architecture of Vav1 provides a facile kinetic route to activation in the face of substantial suppression of basal activity. Since Tyr142 and Tyr160 are conserved in 31 of 33 available Vav sequences (Aedes aegypti and Tribolium castaneum sequences have only Tyr142), this mechanism is likely to be general across the family.
Our collective data show that the regulatory apparatus of Vav1 is composed of at least two thermodynamically coupled processes (Fig. 1A). A core process, in which the inhibitory helix blocks the DH active site, provides approximately 10-fold suppression of activity and phosphorylation rate (Li et al., 2008). Binding of the CH domain to the N-terminus of the Acidic element and to the DPZ fragment restricts the inhibitory helix, thermodynamically coupling these interactions to the core. This coupling shifts the helix-DH equilibrium ~10-fold further toward the inhibited (helix bound) state. Together, the two processes provide the high level of suppression (~100-fold) that is apparently needed for in vivo function (since CH truncation leads to cell transformation (Bustelo, 2001; Katzav et al., 1991; Llorca et al., 2005)). This 10-fold × 10-fold construction of Vav1 (rather than a single 100-fold inhibitory equilibrium) is likely functionally important, as we have shown that inhibition by the core alone can be increased to 50-fold simply by two point mutations (Li et al., 2008). Biochemical data on other multi-domain systems suggest similar (relatively small, 2–20-fold) magnitudes of core suppression and modulatory coupling (DiNitto et al., 2007; Moarefi et al., 1997).
This construction has important implications for the biology of Vav1. It provides a quantitative rationale for the increased activity of CH domain truncations of the protein. In such truncations, even though the core inhibitory module is intact, the inherent thermodynamics of the module produce high levels of the active state (9–33%, Fig. 2E, Fig. 4A, S3A) when coupling to the modulatory equilibrium is lost. This effect increases both the phosphorylation rate (and likely steady-state phosphorylation level) and basal GEF activity of Vav1, explaining increased Rac activation and cell transformation (Bustelo, 2001; Han et al., 1998; Llorca et al., 2005; Lopez-Lago et al., 2000). The ability of CH interactions to modulate the AD core also implies that molecules that disrupt these contacts could partially stimulate Vav1 activity, and act cooperatively with kinases to fully activate the protein in vivo.
The creation of suppression from multiple weak interactions also has important implications regarding the kinetics of Vav1 activation upon receptor stimulation. In many allosteric systems, it has been shown that activators gain access to their targets by recognizing the small population of active state present under basal conditions (Henzler-Wildman and Kern, 2007). This is the case in the AsD core of Vav1, where kinases can only recognize Tyr174 in the helix-dissociated state (Li et al., 2008). However, strong suppression of activity opposes this pathway, since it requires that the accessible population be very small, which would tend to make activation slow. The layered energetic and physical construction of modular proteins provides a likely general solution to this problem, which we have observed for Vav1. That is, the structure and energetics of multi-domain systems can be organized so that the modulatory elements are more accessible to activators (e.g. Tyr142 and Tyr160 in the CA element show significant µs-ms timescale dynamics and can be rapidly phosphorylated). This allows activation to occur in stepwise fashion, with initial events disabling the modulatory interactions, and consequently increasing access to the core. Thus, modularity enables evolution not only of strong inhibition, but also of ready routes to activation.
A variety of biochemical and structural data suggest that the energetic and kinetic mechanisms that we have found for Vav1 are likely to be widespread among multi-domain proteins. The Tim family of GEFs represents a structurally and functionally similar example (Yohe et al., 2008). The Tim proteins have a DH domain whose activity is inhibited by an adjacent helix that binds in the active site, and that can be displaced by phosphorylation, analogous to the Vav AD core. The Tim proteins also have a C-terminal SH3 domain that further suppresses the activity of the helix-DH core through binding an internal proline-rich motif (PRM). The mechanism of suppression by the SH3 domain has not been explored, but available data would be consistent with coupling of the helix-DH and SH3-PRM equilibria. Moreover, although the kinetic pathway to Tim activation has not been examined experimentally, it has been speculated that activation may occur stepwise, with initial displacement of SH3-PRM interactions facilitating phosphorylation of the inhibitory helix (Yohe et al., 2008). The unrelated Ras GEF, SOS, also appears to use the same regulatory logic, with the DH-PH element shifting an allosteric equilibrium in the Rem-Cdc25 element toward its inactive state (Sondermann et al., 2004).
The broadest class of, and best characterized, proteins that appear to behave in this manner are the protein kinases. A large body of data suggests that the isolated kinase domain fluctuates between an inactive state, where the active site is not properly organized for catalysis, and an active state, where the catalytic cluster and substrate binding site are catalytically competent (e.g. (Levinson et al., 2006; Masterson et al., 2008; Vogtherr et al., 2006)). Phosphorylation of residues in the so-called activation loop stabilizes the active conformation, increasing catalytic activity. While the active state is common among many kinases, the inactive states differ. Sampling of the inactive conformation has been capitalized upon in the Abl tyrosine kinase inhibitor imatinib, a widely used anti-cancer therapeutic (Schindler et al., 2000). Imatinib distinguishes Abl from other closely related kinases by selectively binding to its unique inactive conformation. NMR analysis of fluctuations between the active an inactive states in isolated kinase domains is an active area of research. However, to date such studies have not yielded a quantitative understanding of kinase thermodynamics (populations of the states) or a detailed view of the equilibrating structures in the absence of stabilizing inhibitors (Masterson et al., 2008; Vajpai et al., 2008; Vogtherr et al., 2006). Many kinases also contain additional domains that repress the activity of the catalytic domain (Huse and Kuriyan, 2002). Disabling such domains, either by truncation or through regulatory interactions, increases the activity of kinases toward various substrates. A detailed interpretation of the biochemical data in terms of allosteric equilibria are complicated by the fact that phosphorylation of the activation loop locks the kinase in an active state, and that the activity of non-phosphorylated kinases is very low and thus hard to quantify. However, molecular dynamics simulations of the Src kinase have suggested that, at least in that system, the additional SH3 and SH2 domains suppress activity through biasing the regulatory equilibrium in the kinase domain toward the inactive state (Faraldo-Gomez and Roux, 2007; Young et al., 2001). As in Vav, this construction also appears to dictate that kinases are activated in stepwise fashion, with initial events that disable modulatory interactions (e.g. in Src by ligand binding to the SH3 domain (Moarefi et al., 1997)) enabling phosphorylation of the activation loop and consequent maximum stabilization of the active state.
These various examples strongly suggest that although multi-domain proteins use a variety of structural mechanisms to achieve regulation, the underlying energetic mechanism that we have illustrated for Vav, and the requirement for an ordered activation process that arises from it, are likely to be widely observed. The NMR approach we developed here provides a means to directly measure the populations of different states across regulatory equilibria and has allowed us to quantitatively characterize the energetic landscape of Vav1. This in turn has established coupled equilibria as a major mechanism of interdomain cooperativity in this system. Application of these methods to other systems should reveal how widespread cooperative inhibition through coupled equilibria is in multi-domain proteins.
Detailed expression, modification and purification procedures for all materials are provided in the Supplementary Materials.
We created a murine CADPZ mutant (CADPZtryp) containing three additional tryptic sites: Arg135, Lys148 and Lys166. Trypsinolysis of CADPZtryp yields peptides containing Tyr142 (136–148), Tyr160 (149–166) and Tyr174 (167–179). Most NMR signals that report on the core and modulatory equilibria are in identical positions in CADPZ and CADPZtryp (not shown), suggesting that phosphorylation of CADPZtryp is representative of WT. Moreover, bulk phosphorylation kinetics of CADPZ and CADPZtryp are identical (not shown). Lck-mediated phosphorylation reactions were carried out with [γ32P]ATP, quenched with the kinase inhibitor LckI2 (Calbiochem #428206) and urea to 8M. Proteins were digested with trypsin. Peptides containing pTyr142, pTyr160 and pTyr174 were separated with reverse-phase HPLC and detected by an in-line scintillation counter (Bioscan Flow-Count). See Supplementary Materials for details.
Crystals of human Vav1 CADPZ (~50 × 100 × 200 µm) were grown at 4 °C using the hanging-drop vapor-diffusion method. Drops contained 2 µl of 3–10 mg/ml protein solution (20 mM Tris (pH 7.5), 50 mM NaCl, 5% (w/v) glycerol, 2 mM TCEP) and were equilibrated against 200 µl of water. Crystals of selenomethionine-labeled protein were obtained under similar conditions when seeded with native crystals. Crystals were cryoprotected by addition of ethylene glycol to 30% (v/v). Native and anomalous dispersion data were collected at 100 K using the Advanced Photon Source 19ID beamline. The native dataset was collected at the anomalous edge for zinc. Vav1 CADPZ crystallized in space group P21 with unit cell parameters of a = 85 Å, b = 59 Å, c = 161 Å, β= 97 degrees, containing two molecules in the asymmetric unit with ~55% solvent. Crystals diffracted x-rays to a minimum Bragg spacing of ~2.65 Å. Data were processed using the HKL3000 suite of programs (Minor et al., 2006). Data collection statistics are in Table S1.
Phases were obtained from a single-wavelength anomalous dispersion experiment using a seleno-methionine-Vav1 CADPZ crystal with data to a resolution of 2.80 Å. Thirty of 34 expected selenium sites were located using the program SHELXD (Schneider and Sheldrick, 2002). Phases were refined with the program MLPHARE (Otwinowski, 1991), resulting in a figure-of-merit of 0.27. Phases were further improved by density modification and two-fold non-crystallographic averaging with the program DM (1994) resulting in a figure-of-merit of 0.88. An initial model containing about 54% of all residues was automatically generated using the program ARP/warp (Morris et al., 2004). Additional residues were manually added using the program coot (Emsley and Cowtan, 2004). Refinement was carried out with native data to a resolution of 2.73 Å in the program PHENIX (Adams et al., 2002) and consisted of simulated annealing, conjugate-gradient minimization and refinement of individual B-factors and TLS parameters, interspersed with manual revisions of the model. The current model contains two Vav1 CADPZ monomers (labeled A and B, respectively) in the asymmetric unit with backbone r.m.s.d. of 0.13 Å for residues 2–129, 142–150, 156–180, 189–417, 419–456, 463–478, 482–564. Statistics of refinement and structural quality are listed in Table S1. The coordinates and structure factors have been deposited in the Protein Data Bank under accession number 3KY9.
NMR analyses were performed in buffer containing 20 mM Tris(hydroxymethyld3) amino-d2-methane, pH 7.5 (uncorrected), 100 mM NaCl, 2 mM TCEP, and 10% (w/v) d8-glycerol; 99.9% D2O. Data were acquired on 600 and 800 MHz Varian Inova spectrometers equipped with cold probes. CPMG experiments were performed using a constant relaxation period of 20 ms at 15 °C. Methyl-TROSY experiments were performed at 25 °C. Average open populations (po) for murine proteins were determined as previously described (Li et al., 2008). Note that po values for AsD reported here differ slightly from those reported previously due to small differences in buffer conditions. NMR reporters in CADPZ and ADPZ were assigned to residue number by mutagenesis and confirmed by comparison to spectra of AsD and CAD. Methyl groups at the helix-DH interface of CADPZ/ADPZ were stereospecifically assigned by comparison to signals in AsD (Aghazadeh et al., 2000).
Guanine nucleotide exchange assays were carried out and analyzed as previously described (Li et al., 2008). Assays contained 10 µM Vav1 CADPZ, AsDPZ, or ADPZ. In titrations of Figure 5, binding affinities used to calculate fraction bound (see text) were determined independently under identical solution conditions through NMR titrations of unlabeled ADPZ proteins into methyl-labeled CH-containing proteins.
In each assay 10 µM GST-Rac(GDP), 10 µM Vav protein, 0.5 µM GDP, 2 mM MgCl2 and 0.1 M Tris pH 7.5 were premixed to final volume of 200 µl and bound to 100 µl glutathione sepharose beads (GE Healthcare). After three washes, proteins were eluted with 30 mM glutathione. In control experiments, 10 µM GST was used instead of GST-Rac.
To measure Rac activation by various full length Vav1 proteins, Panc04.03 pancreatic cancer cells were transfected with 20 µg of pCDNA3 control vector or Vav1 expression vectors. The GTP-bound form of endogenous Rac1 was detected using the standard GST-PAK affinity precipitation assay 18 hours post transfection (see Supplemental Methods). The transformation of NIH3T3 cells was assayed 14 days after transfection with control or Vav1-expressing pCDNA3.1/HisA vector by staining cells with 0.4% crystal violet and counting foci (see Supplemental Methods).
We thank Carlos Amezcua and Chad Brautigam for technical assistance, Lewis Kay for providing NMR pulse sequences, Shae Padrick and Rama Ranganathan for discussion and Yuh Min Chook for critical reading of the manuscript. I.M. was supported by fellowships from the Foundation for Science and Technology (Portugal) and the Luso-American Development Foundation (Portugal). G.K.A. was supported by a postdoctoral fellowship from the Cancer Research Institute. This work was supported by NIH grants GM066930 to M.K.R., AI065474 and CA102701 to D.D. Billadeau and M.E.F. and by the Howard Hughes Medical Institute. D.D. Billadeau is a Leukemia and Lymphoma Society Scholar. Use of the Argonne National Laboratory Structural Biology Center beamlines at the Advanced Photon Source, was supported by the U. S. Department of Energy, Office of Energy Research, under Contract No. W-31-109-ENG-38.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.