The ability of proteins to regulate their own enzymatic activity can be facilitated by changes in structure or protein dynamics in response to external regulators. Because many proteins contain SH2 and SH3 domains, transmission of information between the domains is a potential method of allosteric regulation. To determine if ligand binding to one modular domain may alter structural dynamics in an adjacent domain, allowing potential transmission of information through the protein, we used hydrogen exchange and mass spectrometry to measure changes in protein dynamics in the SH3 and SH2 domains of hematopoietic cell kinase (Hck). Ligand binding to either domain had little or no effect on hydrogen exchange in the adjacent domain, suggesting that changes in protein structure or dynamics are not a means of SH2/SH3 crosstalk. Furthermore, ligands of varying affinity covalently attached to SH3/SH2 altered dynamics only in the domain to which they bind. Such results demonstrate that ligand binding may not structurally alter adjacent SH3/SH2 domains and implies that other aspects of protein architecture contribute to the multiple levels of regulation in proteins containing SH3 and SH2 domains.
Modular protein domains are used as the building blocks of many proteins (Pawson et al. 2002; Gavin and Superti-Furga 2003; Bru et al. 2005). Unique combinations of domains within a particular protein can dictate the overall function of that protein. Proteins with multiple domains may have multiple functions, determined by the spatial and/or temporal binding or interaction(s) of a particular domain(s). Integration of information as a result of domains interacting with and influencing one another, either in an intraprotein or interprotein fashion, may play a significant role in multidomain protein function.
There are several possible methods of communicating information within and between multidomain proteins. One potential method involves communication of structural changes between domains. Such allosteric regulation has been demonstrated for many proteins by comparing crystal structures of different forms of the same protein. Hydrogen exchange (HX) analysis has also been used to gain insight into the role of structural changes within proteins (for recent reviews, see Hoofnagle et al. 2003; Redfield 2004; Yan et al. 2004; Busenlehner and Armstrong 2005). By monitoring the incorporation of deuterium into the domains of the protein in different states (i.e., active, inactive, bound, unbound), information about the location and magnitude of structural changes, and the associated allosteric alterations, can be obtained.
To determine if dynamics changes play a significant role in interdomain communication, we have chosen the Src homology 2 (SH2) and 3 (SH3) domains as a model system. SH2 and SH3 are small, modular protein domains that participate in a wide variety of cellular functions (Musacchio et al. 1994; Schaffhausen 1995; Kuriyan and Cowburn 1997). SH3 and SH2 were first discovered in the Src family of protein tyrosine kinases, where an SH3 domain and an SH2 domain precede a catalytic domain in sequence (Fig. 1A). While SH3 and SH2 are most often found in signal transduction proteins, they have also been discovered in other proteins, where they are primarily involved in protein–protein interactions. SH2 domains, with ∼100 amino acids, bind with sub-micromolar affinity to phosphotyrosine-containing sequences (Ladbury et al. 1995). The SH3 domain, in contrast, contains 60–80 amino acids and binds with somewhat lower affinity to proline-rich sequences that form class II polyproline (PPII) helices (Pawson 1992).
At first, SH2 and SH3 were primarily seen as binding domains that recruited or allowed recognition of target proteins. It is now evident that they also participate in regulation of enzymatic activity in some proteins. For example, within the Src family of kinases, the domains participate in protein regulation. SH2 binding to the C-terminal tail, a regulatory sequence found at the C-terminal end of the same molecule (Fig. 1A), is critical for maintaining the inactive state (Brown and Cooper 1996). Src-family proteins lacking SH3 are not regulated (Superti-Furga et al. 1993; Erpel et al. 1995), and SH3 displacement from its regulatory position without concomitant SH2 displacement is sufficient for kinase activation (Lerner and Smithgall 2002).
To determine the role of dynamics crosstalk between domains, HX mass spectrometry (MS) was used to directly test whether ligand binding to SH3 alters the structure and dynamics of the adjacent SH2 domain and vice versa. Further, we tested whether a natural ligand for SH3 found in full-length Src-family kinases (the SH2-kinase linker, see Fig.1A) associates with the SH3 domain in the absence of the rest of the protein, and if there were any changes in dynamics within the SH2 domain as a result of the tethered linker binding to SH3. No evidence for communication of binding between the domains via changes in protein dynamics was found, thereby supporting the hypothesis that organization of the domains within the protein and potential intraprotein interactions among the domains may be more significant in protein regulation than communication of dynamics between domains.
MS was used to compare binding-induced changes in HX in 12 small peptide fragments of Hck SH32 (a construct of SH3+SH2) that were produced after incubating Hck SH32 in D2O (Fig. 1B). MS with peptic fragmentation allows changes in deuterium levels to be localized and permits analysis of events that occur on a broader range of time scales than can be easily observed with HX NMR (Zhang and Smith 1993; Engen et al. 1999b). Changes in HX associated with peptide ligand binding to the isolated Hck SH3 (Engen et al. 1997) and Hck SH2 (Engen et al. 1999a) domains have previously been determined with this approach. HX in isolated SH3/SH2 domains has also been compared with exchange in a combined SH3+SH2 construct to probe dynamics changes as a result of the domains being covalently attached to one another (Engen et al. 1999b).
Free Hck SH32 and SH32 separately bound to either an SH3 ligand or to an SH2 ligand were independently incubated in D2O under identical conditions. The deuterium exchange-in reaction was quenched at various times by lowering the pH to 2.5 and the temperature to 0°C. Following quenching of isotopic exchange, the labeled proteins were digested into fragments with the acid protease pepsin, and the fragments were quickly separated with perfusion HPLC and directed into a mass spectrometer, where the mass of each fragment was determined. In this way, the deuterium levels in short segments of free and bound SH32 (see Fig. 1B) were measured at various exchange-in times. Deuterium incorporation into free SH32 versus bound SH32 was compared to determine which regions experienced changes in HX, and hence protein dynamics, in response to binding.
When SH32 was incubated with a high-affinity binding peptide from the HIV Nef protein (Lee et al. 1995), significant changes in HX were observed throughout the SH3 domain, but no significant changes in HX were found in the SH2 domain (Figs. 2, 3A). A significant change was defined as a change in which the difference in the relative deuterium level was greater than the experimental uncertainty (±0.2–0.3 Da) of each data point. No distinction has been made about where in the exchange time course the difference was found (i.e., fast-exchange amide hydrogens, slow-exchange amide hydrogens, etc.). When the SH32 construct was bound to a high-affinity SH2 phosphopeptide ligand from hamster polyomavirus middle T antigen (pYEEI peptide) (Songyang et al. 1993), HX was altered for the SH2 domain, but not for the SH3 domain (Figs. 3B, 4). These results show that changes in protein dynamics as a result of binding are not communicated from one domain to the other in the SH32 construct. In other words, SH3 is unaware of binding to the SH2 domain, and vice versa. Further, these results demonstrate that the binding status of each domain can be determined by monitoring changes in deuterium incorporation.
In addition to monitoring deuterium levels upon binding, partial, cooperative unfolding that occurs in the Hck SH3 domain under physiological conditions (hereafter referred to as “SH3 unfolding”) can be used as an SH3 binding assay (Engen et al. 1997). This unfolding event is revealed by the appearance of a bimodal isotope pattern in mass spectra after ∼10–15 min of D2O labeling. An example is shown in Figure 5A for the Hck SH3 domain. Partial unfolding still occurs when SH3 binds to a ligand but is significantly slower than when SH3 is unbound (Engen et al. 1997; Gmeiner et al. 2001). A slowdown factor can be calculated (see Materials and Methods) and represents the degree to which SH3 unfolding is inhibited by binding, with higher numbers correlating to a greater extent of SH3 occupancy by ligand. The HIV Nef protein and Nef peptide, which have 0.2 μM and 90 μM affinity for Hck SH3, respectively (Lee et al. 1995), inhibited SH3 unfolding by a factor of 3.6–4.4 when incubated at concentrations such that ∼60% of the SH3 molecules were bound (Fig. 5B). Because these binding experiments were done with a free peptide, the maximal amount of slowdown that could be obtained was influenced by the dissociation constant for the complex. We reasoned that by covalently attaching a peptide to the SH3 domain, we could increase the effective local concentration of the peptide and substantially increase the slowdown factor. We tested this idea by preparing a construct in which a proline-rich peptide with moderate Hck SH3 affinity was covalently attached to SH3 by means of a short linker, a construct termed SH3-Pro (Gmeiner et al. 2001). SH3-Pro had a slowdown factor of ∼39, consistent with reduced SH3 domain dynamics as a result of an increase in the local concentration of the peptide. Measuring a decrease in the SH3 unfolding rate in various bound states illustrates that slowdown factors can be used to determine the binding status of the SH3 domain, both for peptides in trans and peptides that are covalently attached.
When the slowdown factor for Hck SH3 was measured for the SH32 construct, no difference was seen versus SH3 alone (Fig. 5B). From this it was apparent that the physical presence of the SH2 domain does not influence SH3 unfolding. No increase in SH3 slowdown factor was seen when SH32 was bound to the high-affinity SH2 peptide YEEI, meaning that changes in the dynamics of the SH2 domain were not sensed by the SH3 domain (Fig. 5B). These results, in addition to the data presented in Figures 2 and 3A, rule out changes in domain dynamics as a means of interdomain communication within the SH32 construct.
The crystal structures of Hck (Sicheri et al. 1997; Schindler et al. 1999) and Src (Xu et al. 1997, 1999) revealed that the short sequence that connects the SH2 domain to the kinase domain, termed the SH2-kinase linker, is an intramolecular ligand for SH3 in the down-regulated form of the enzyme (see Fig. 1A). To determine whether the natural SH2-kinase linker was a ligand for the SH3 domain in the absence of the kinase domain and possibly a bridging molecule that could help communicate binding status from the SH3 domain to the SH2 domain, the SH32 construct was incubated with a molar excess of the SH2-kinase linker in trans (as a free peptide). No alteration of the unfolding rate of the SH3 domain was observed, consistent with a lack of interaction (Fig. 5B). In contrast, the positive control Nef peptide elicited a substantial slowdown factor when incubated with SH32. These results suggest that the SH2-kinase linker peptide does not adopt the same binding-capable structure in solution as it does in downregulated Src-family kinases (discussed below).
We next investigated whether covalent attachment of the SH2-kinase linker sequence to the SH32 construct (construct termed SH32L) led to SH3 binding and caused structural changes within the SH2 domain. This covalent ligand-tethering approach was effective for ligand binding in the SH3-Pro construct described above. Surprisingly, tethering the SH2-kinase linker peptide onto SH32 was not sufficient to promote SH3 binding (no change in the slowdown factor; Fig. 5B), indicating that simply increasing the local concentration of the natural SH2-kinase linker by tethering is not sufficient to convert this sequence into a ligand capable of binding Hck SH3. Consistent with a lack of binding, detailed measurements and comparisons of HX in the SH32L form versus the SH32 form showed no changes in the SH3 domain as a result of tethering the linker to SH32 (Figs. 3C, 6). Only a few changes were observed in the SH2 domain, which likely result from conformational stabilization by the linker peptide.
To test whether the linker peptide was able to elicit changes in SH2 as a result of binding to SH3, a linker peptide was needed that bound tightly to the SH3 domain. A mutant version of SH32L was created (termed SH32HAL, for SH32 with a high affinity linker) that contained an SH2-kinase linker sequence with much higher affinity for the SH3 domain, as demonstrated by surface plasmon resonance measurements (Lerner et al. 2005). Two lysines within the linker were changed to proline, thereby changing the propensity of the sequence to spontaneously form a PPII helix. The slowdown factor in the SH32HAL construct was so great that it could not be measured within the 8-h time course of the experiment (Fig. 5B). Such results imply that the SH3 domain in the SH32HAL construct was bound very tightly to the high-affinity linker. Detailed comparisons of HX for the SH32HAL form versus the SH32 form showed significant changes in HX in the SH3 domain and only minor changes to the SH2 domain (Figs. 3D, 7), again likely caused by the restricted movements of the C-terminal part of the SH2 domain. Taken together, these data indicate that although the SH3 domain was very aware of a bound ligand, no significant changes in structural dynamics were transmitted to the SH2 domain as a result of ligand binding to SH3.
SH3 and SH2 domains somehow coordinate with each other to regulate the activity of Src-family kinases and other proteins in which they are found. We had hypothesized that they communicate through classic allosteric-like mechanisms; that is, by changes in structural dynamics. We have used hydrogen exchange to probe protein unfolding and dynamics in the tandem SH3, SH2 construct (SH32) to determine if crosstalk occurs via changes in structural dynamics. Further, we tested whether a covalently attached ligand for one of the domains could elicit a change in dynamics in the other domain.
Our data indicate that the binding status of SH2 and SH3 is not communicated directly between the two domains by changes in their structure or dynamics, both of which can be detected by differences in amide HX rates. When the SH32 domain was incubated with an SH2 ligand, there were significant changes in the HX, and therefore protein unfolding and dynamics in the SH2 domain. The changes were comparable to those observed when isolated SH2 was bound to the same peptide (Engen et al. 1999a). However, no changes were observed in the SH3 domain. These results indicate that occupancy of the SH2 binding site is not communicated to the SH3 domain. Similarly, when the SH3 domain was incubated with a high-affinity ligand, HX was altered in SH3 but not in SH2, suggesting that changes in structural dynamics within the SH3 domain are not a means of communicating the occupancy of the SH3 binding site to the SH2 domain.
Tethering either a binding or non-binding SH3 ligand onto the SH32 construct did not change dynamics in the SH2 domain to a significant degree. Only moderate changes were observed in the region where the peptide ligand was attached to the end of the construct. The native SH2-kinase linker sequence that was tethered to the SH32 construct bears little resemblance to a high-affinity SH3 ligand, leading to the hypothesis that additional interactions are required to structure it into a PPII helix capable of binding SH3 (Gonfloni et al. 1997, 1999; Barila and Superti-Furga 1998). According to our results, the presence of the SH2 domain was not sufficient to structure the linker into a PPII helix competent for binding to the SH3 domain. A lack of SH2-kinase linker binding in the absence of the rest of the protein has been demonstrated for other Src-family kinase SH3/SH2 constructs. A Src SH32L construct showed a lack of SH2-kinase linker affinity for SH3 by NMR (Tessari et al. 1997), and the Abl SH3 domain also seems incapable of binding to the Abl SH2-linker peptide (Pisabarro et al. 1998; L. Serrano, pers. comm.). A recent report (Cobos et al. 2004) indicated that additional interactions provided by protein scaffolding can stabilize formation of polyproline helix conformation and convert low-affinity proline-containing sequences into high-affinity SH3 ligands. Based on our results, a lack of additional interactions from the rest of the protein (presumably from the kinase domain in the case of Src-family kinases, but perhaps from other domains in other proteins) alters the affinity of the linker for SH3, a circumstance that may be involved in diverse modes of activation.
Having ruled out dynamics crosstalk as a means of communication for SH3 and SH2, an alternative hypothesis is that interactions with these domains must be communicated to their parent proteins primarily through means that do not involve changes in the dynamics within the domains themselves. Proper positioning of the domains with respect to each other (e.g., Young et al. 2001) or with respect to the rest of the protein may be the means of domain communication and information integration within multidomain proteins. In the case of Src-family kinases, this likely involves SH3/SH2 interactions with the rest of the protein, as demonstrated in the crystal structures of the inactive forms of the kinases (Schindler et al. 1999; Xu et al. 1999). Interactions in Csk, another SH3/SH2-containing protein involved in Src regulation, are likewise facilitated by domain interactions with other parts of the protein (Wong et al. 2005). It appears that the SH3 and SH2 domains are compact enough to participate only in ligand binding and can influence their parent proteins by means of interactions (hydrophobic, electrostatic, steric) that are unaltered by dynamics changes with the domains. Further investigation of other multidomain proteins will be required before this hypothesis can be generally applied to all proteins containing these domains.
Materials and methods
Preparation of proteins and peptides
Recombinant human Hck SH3, SH3 covalently attached to a binding sequence from human Ras-GAP (SH3-Pro), SH32, SH32 plus the natural SH2-linker sequence (SH32L), and SH3–SH2 plus a high-affinity linker sequence (SH32HAL) were prepared in Escherichia coli as previously reported (Engen et al. 1997, 1999b; Gmeiner et al. 2001). The 72-residue SH3 domain and the 107-residue SH2 domain encompass amino acids 72–143 and 140–245 of Hck, respectively, while SH32L encompassed amino acids 72–256 (all c-Src numbering). SH3-Pro has been described previously (Gmeiner et al. 2001). The SH32HAL mutation is described in detail elsewhere (Lerner et al. 2005). The sequence SKPQKP in the Hck SH2-kinase linker was changed to SPPQPP by site-directed mutagenesis. The HIV Nef peptide and SH2 high-affinity peptide are described elsewhere (Engen et al. 1997, 1999a). The SH2-kinase linker peptide, [Ac]-KPQKPWEKDAWE-[NH2], had the sequence of Hck residues 245–256 and was synthesized by conventional solid-phase methods at the Alberta Peptide Institute.
Continuous-labeling deuterium exchange experiments were carried out following methods previously described (Engen et al. 1997, 1999b) with the various concentrations of the different ligands as noted in the legend of Figure 5B. The calculation of the percent of SH3 bound was based on KD,SH2 of 0.5 μM (Ladbury et al. 1995) and KD,SH3 of 90 μM (Lee et al. 1995) and followed essentially the equations described by Mandell et al. (2001). The protein concentration was estimated with the Bradford assay. As a negative control, in which no SH3 or SH2 ligand was present, SH32 was incubated with 1500 μM of the non-binding peptide, angiotensin I.
Analysis of deuterium incorporation by mass spectrometry
After D2O labeling of the intact protein, but prior to MS analysis, 200–350 pmol of each sample was incubated with pepsin at a ratio of 1 : 1 (weight : weight) for 5 min at 0°C. The resulting peptides were separated in 7 min by a 5%–60% acetonitrile : water gradient using a 100 mm × 0.25 mm (ID) reversed-phase capillary perfusion HPLC column (POROS 10 R2 media, PerSeptive Biosystems) or a C18 reversed phase column (Michrom Bioresources). Both components of the mobile phase contained 0.05% trifluoroacetic acid, and the flow rate was 40 μL/min. The injector and column were cooled to 0°C to minimize deuterium back-exchange. Under these conditions, the average amount of deuterium lost during analysis was 12%–13%. Although deuterium loss during HX MS experiments can span a range of 10%–25%, as described elsewhere (Zhang and Smith 1993), adjustment for back-exchange was not performed because all experiments were done at nearly the same time under identical experimental conditions. Hence, all uptake curves are noted as relative deuterium level. The HPLC step was performed with protiated solvents, thereby removing deuterium from side chains and amino/carboxy termini that exchange much faster than amide linkages (Bai et al. 1993). Therefore, an increase in molecular mass was a direct measure of deuteration at peptide amide linkages. Identification of the peptic fragments of all constructs was as described previously (Engen et al. 1999b). Analyses of deuterium incorporation were performed with a Waters-Micromass QTOF2 in ESI mode. Data were processed by centroiding an isotopic distribution corresponding to the +2, +3, or +4 charge state of each peptide. The relative amount of deuterium in each peptide was plotted as deuterium level versus the exchange-in time, and the experimental data were fitted with a series of first-order rate terms as described previously (Engen et al. 1999b). To calculate the slowdown factor (SF) from the rate constant for unfolding, the natural log of the percent of folded molecules was determined from the area ([A]) of the peak representing the folded form (lower mass peak in bimodal pattern, Fig. 5A) and the total area ([A]o) of the bimodal distribution; the slope of Ln% folded plotted against D2O labeling time provided the rate constant (simple first-order kinetics) and was used to calculate the t½ for unfolding (see also Gmeiner et al. 2001). The SF calculation was SF = (t½ of SH3 unfolding for test sample) / (t½ of SH3 unfolding for unbound SH3).
We thank G. Superti-Furga for initial help with the manuscript and discussions about SH3/SH2, and L. Serrano for advice and critical reading of parts of the manuscript. We acknowledge the Alberta Peptide Institute for synthesizing the peptides. This work was supported by grants from the National Cancer Institute (R01-CA81398, T.E.S. and R24-CA088339, J.R.E.), the National Institute of Allergy and Infectious Diseases (R01-AI57083, T.E.S.), the National Institute of General Medical Sciences (R01-GM070590, J.R.E.), and the National Institute for Research Resources (P20-RR016480, J.R.E.).