The Abl SH2-kinase linker naturally adopts a conformation competent for SH3 domain binding

Abstract

The core of the Abelson tyrosine kinase (c-Abl) is structurally similar to Src-family kinases where SH3 and SH2 domains pack against the backside of the kinase domain in the down-regulated conformation. Both kinase families depend upon intramolecular association of SH3 with the linker joining the SH2 and kinase domains for suppression of kinase activity. Hydrogen deuterium exchange (HX) and mass spectrometry (MS) were used to probe intramolecular interaction of the c-Abl SH3 domain with the linker in recombinant constructs lacking the kinase domain. Under physiological conditions, the c-Abl SH3 domain undergoes partial unfolding, which is stabilized by ligand binding, providing a unique assay for SH3:linker interaction in solution. Using this approach, we observed dynamic association of the SH3 domain with the linker in the absence of the kinase domain. Truncation of the linker before W254 completely prevented cis-interaction with SH3, while constructs containing amino acids past this point showed SH3:linker interactions. The observation that the Abl linker sequence exhibits SH3-binding activity in the absence of the kinase domain is unique to Abl and was not observed with Src-family kinases. These results suggest that SH3:linker interactions may have a more prominent role in Abl regulation than in Src kinases, where the down-regulated conformation is further stabilized by a second intramolecular interaction between the C-terminal tail and the SH2 domain.

The Abelson tyrosine kinase (c-Abl) is a nonreceptor protein tyrosine kinase involved in a wide variety of cellular functions including regulation of cell growth and differentiation, oxidative stress and DNA-damage responses, actin dynamics, and cell migration (for reviews, see Pendergast 2002; Hantschel and Superti-Furga 2004). c-Abl is also a frequent translocation partner in human leukemias, of which chronic myelogenous leukemia (CML) is arguably the best characterized. The cytogenetic hallmark of CML is the Philadelphia (Ph) chromosome, which arises from a reciprocal translocation between the c-abl locus on chromosome 9 and the breakpoint cluster region (bcr) gene on chromosome 22. CML and other Ph-positive leukemias such as acute lymphocytic leukemia (ALL) express a chimeric Bcr-Abl fusion protein with constitutive protein-tyrosine kinase activity (de Klein et al. 1982; Heisterkamp et al. 1983). Bcr-Abl tyrosine kinases cause cytokine-independent proliferation of hematopoietic progenitor cells and are sufficient to induce CML- and ALL-like syndromes in mice, providing strong evidence that they are the driving force behind the disease. Gleevec (imatinib mesylate) and other small molecule inhibitors of Bcr-Abl catalytic activity are dramatically effective in chronic-phase CML, providing further evidence that Bcr-Abl is essential for initiation and maintenance of the transformed phenotype (for reviews, see Giles et al. 2005; Ren 2005; Jabbour et al. 2006).

The tyrosine kinase core of c-Abl makes up the N-terminal half of the molecule and is followed by a large unique domain (the last exon region or C-terminal domain) important for subcellular localization and protein/DNA binding (Pendergast 2002). The kinase core is similar in many regards to the Src-family of tyrosine kinases (Courtneidge 2003; Harrison 2003; Hantschel and Superti-Furga 2004). In fact, the X-ray crystal structures of the down-regulated forms of the Src-family kinases (SFKs), c-Src (Williams et al. 1997; Xu et al. 1997, 1999) and Hck (Sicheri et al. 1997; Schindler et al. 1999), are nearly superimposable on the structures of the kinase core of c-Abl (Nagar et al. 2003, 2006). In both SFKs and c-Abl, modular SH3 and SH2 domains interact with the backside of the kinase domain (see Fig. 1A,B), resulting in downregulation of kinase activity. In both c-Abl and SFKs, the SH3 domain contributes to downregulation by binding an internal ligand formed by the sequence connecting the SH2 domain to the small lobe of the kinase domain (referred to hereafter as the linker). In contrast, the contribution of the SH2 domain to downregulation is different in c-Abl from that in SFKs. In SFKs, the SH2 domain binds to a phosphotyrosine-containing sequence at the C-terminal tail of the kinase and helps hold SH2 in position to inhibit kinase activity (Brown and Cooper 1996; Harrison 2003). In c-Abl, however, there is no C-terminal tail because the last exon region of c-Abl begins just after its kinase domain. Prior to the crystal structure of the c-Abl kinase core, it was not understood how c-Abl could accomplish efficient downregulation without an SFK-like SH2:C-tail interaction. The crystal structure and related mutagenesis experiments revealed intramolecular interactions between the SH2 domain and the large lobe of the kinase domain that hold SH2 in a position that contributes toward downregulation of kinase activity (Hantschel et al. 2003; Nagar et al. 2006). Mutations in the SH2-kinase linker similarly cause release of c-Abl kinase activity, supporting a regulatory role for SH3:linker engagement in kinase regulation (more below; Hantschel and Superti-Furga 2004).

Figure 1.

Abl tyrosine kinase and the constructs used in this study. (A,B) Crystal structure of c-Abl including the N-cap (PDB code 2FO0) (Nagar et al. 2006). Coloring is according to the convention (Nagar et al. 2003). Myristic acid and an inhibitor (PD166326) in the active site are shown in stick form. (C) Summary of the amino acids used in each construct. The sequence of c-Abl is shown in the middle with numbering according to the human sequence of Abl 1b. The length of each linker mutant is shown with a distinctive color and labeled above the sequence. An alignment with several other Src-family members is shown immediately below. The residues mutated in the HAL mutants are indicated. The sequence of the BP1 peptide (Rickles et al. 1994) is also shown and aligned with the sequences according to the P₀ and P₊₃ positions (indicated in gray). (D,E) Abl as in A and B but with the linker colored according to the constructs used in C. A small black line marks the end of each linker construct. The P₀ and P₊₃ residues that were mutated are shown as sticks and labeled.

Other parts of c-Abl also contribute to regulation and partially compensate for the lack of a C-tail. c-Abl has an N-terminal “cap” that is important for downregulation (Pluk et al. 2002). A second crystal structure of the c-Abl kinase core (Nagar et al. 2006) showed that the residues of the cap immediately N-terminal to the SH3 domain, including a phosphorylated serine at position 69, loop back to interact with the SH3–SH2 linker (Fig. 1A). Cap interactions with the SH3 and SH2 domains are thought to stabilize the regulatory domain interactions with the back of the kinase domain and mutations of select cap residues (i.e., K70), making contact with SH3/SH2 result in kinase activation (Hantschel et al. 2003). Finally, a deep pocket in the C lobe of the kinase domain was shown to be a binding site for myristic acid (see Fig. 1A). The crystal structures strongly suggest that the myristoylated N terminus of c-Abl binds within this pocket and thereby positions the cap and associated sequence to help hold SH2 and SH3 in the down-regulatory position (for reviews, see Courtneidge 2003; Harrison 2003; Hantschel and Superti-Furga 2004). Interestingly, small molecules that mimic the interaction of the cap myristoyl group with the C lobe are also effective c-Abl kinase inhibitors, suggesting that the myristic-acid-binding pocket is a unique feature that allosterically controls the kinase active site (Adrian et al. 2006).

While the SH2 domain interactions in SFKs and c-Abl are unique to each protein, the SH3:linker interaction is a feature shared by both the SFKs and c-Abl. As such, it is a critical element in the regulation of these enzymes. Disruption of the SH3:linker interaction leads to kinase activation and transforming capability in both Abl (Barila and Superti-Furga 1998; Hantschel et al. 2003; Meyn et al. 2006) and SFKs (Gonfloni et al. 1997; Briggs and Smithgall 1999). In previous analyses of the SFK Hck (Lerner et al. 2005; Hochrein et al. 2006), there was no evidence for an intrinsic ability of the linker sequence to bind to SH3 in the absence of stabilizing interactions from the kinase domain. Because the SFK and c-Abl linker structures are inherently different, we speculated that the c-Abl linker might display intrinsic SH3-binding activity and thereby confer unique properties to the regulatory nature of the c-Abl linker:SH3 interaction. Using hydrogen-deuterium exchange mass spectrometry, we provide direct biophysical evidence that the c-Abl linker sequence does, indeed, interact with c-Abl SH3 in solution, and that unlike Src family kinases, this interaction occurs in the absence of the kinase domain. This analysis reveals a unique feature of the c-Abl regulatory mechanism key to the conformational control of both c-Abl and Bcr-Abl.

Results

Sequence comparison of SFK and Abl linkers

To begin to evaluate possible differences in linker function between SFKs and c-Abl, we first aligned the linker sequences of c-Abl and several closely related SFKs (Fig. 1C). The c-Abl linker has two more residues than the Src linker and three more than Hck, Lyn, and Lck. As the positions of the SH2 domain and small lobe of the kinase domain are highly similar in both SFKs and c-Abl, the additional residues in the c-Abl linker must either fold in a different manner to occupy the same space or loop away from the path they take in SFKs. A number of constructs were prepared to identify the linker sequences required for intramolecular engagement of SH3 by hydrogen-deuterium exchange mass spectrometry as described below. Figure 1C shows the names and boundaries of each c-Abl linker construct. The locations of the amino acids in the down-regulated crystal structure of c-Abl (Nagar et al. 2006) are shown in Figure 1, D and E. Mutations of the residues in positions P₊₃ and P₀ (nomenclature according to Lim et al. 1994) of the putative polyproline type II helix (PPII) that contacts the SH3 domain were also undertaken as described below.

SH3 unfolding and ligand binding

Much of what is known about the interactions in the down-regulated state of c-Abl comes from crystal structures. The use of other biophysical methods (i.e., SAXS as in Nagar et al. 2006) to probe the motions and conformational movements of c-Abl constructs in solution is helping to uncover the conformational dynamics. We have previously used hydrogen-deuterium exchange (HX) and mass spectrometry (MS) (for reviews, see Hoofnagle et al. 2003; Wales and Engen 2006a) to examine intramolecular regulation in the SFKs Hck (Engen et al. 1997; Hochrein et al. 2006) and Lck (Weis et al. 2006b). With HX MS, we have shown that the SH3 domain of c-Abl, like other SH3 domains, has a unique characteristic in that it partially unfolds slowly in solution under physiological conditions (Wales and Engen 2006b). Briefly, unfolding in SH3 domains is an event that exposes a certain number of backbone amide hydrogens to solvent in a coordinated fashion. The resulting cooperative deuterium exchange that occurs causes the shape of the mass spectrum to change from a binomial isotopic distribution to a bimodal isotopic pattern (explained in detail in Weis et al. 2006c). By monitoring the appearance and changes to the isotopic pattern in various constructs over time, binding can be ascertained. We first described the use of this assay to probe protein:ligand binding in the Hck SH3 domain (Engen et al. 1997) and later illustrated its use in larger constructs of Hck (Hochrein et al. 2006). Binding of peptide ligands to the Lck SH3 domain also has been investigated with this method (Weis et al. 2006b). In this assay, the interaction with a partner protein or peptide (inter- or intramolecularly) is gauged by the ability of the partner to inhibit unfolding. The correlation between slowed protein unfolding (or reduced protein dynamics) as a result of binding has been observed previously with other assays, including FTIR (Li et al. 1997), NMR (Seeliger et al. 2005), and differential scanning calorimetry (Martin-Sierra et al. 2003). In the HX MS assay, strong binders stabilize the protein so much that evidence for unfolding disappears and the isotopic distribution becomes completely binomial over the time course of the entire deuterium exchange experiment. Weaker binders interact with the protein less, and the bimodal distribution persists, although the unfolding half-life is shifted to longer times than free protein. In order to determine K _d values with HX MS, a titration experiment must be performed (Engen et al. 1997; Engen 2003). We have used this HX MS unfolding assay to probe intramolecular association of the c-Abl SH3 domain with the linker in the various constructs of c-Abl shown in Figure 1C.

SH3 binding in trans

To validate the use of the unfolding assay with the Abl SH3 domain, interactions were first tested with a known peptide ligand (BP1) (Rickles et al. 1994). BP1 exhibits relatively high affinity for the Abl SH3 domain (K _d = 2 μM) (Rickles et al. 1994) and is useful to illustrate both the appearance of the mass spectra and the resulting data processing steps. Purified recombinant Abl SH3 was labeled with deuterium either alone or in the presence of BP1 such that >90% of the SH3 molecules were bound and mass spectra were obtained after various amounts of labeling time. The spectra of the free SH3 domain (Fig. 2A, left) showed that the peak broadening characteristic of EX1 unfolding (Weis et al. 2006c) occurred with a half-life of nearly 5 min. At later time points, the peak returned to the width it had before the unfolding event occurred. The amount of deuterium that was exchanged-in after each time point was determined (Fig. 2B) and the peak width at each time point measured (Fig. 2C). The change in peak width with time is most obvious in the peak-width plot (Fig. 2C), where an increase in width is created as a result of the unfolding event. The centroid of the peak in the width plot (Fig. 2C) is the approximate unfolding half-life (see Materials and methods). Ligand-bound Abl SH3 (Fig. 2A, right), in contrast, did not exhibit peak broadening like free SH3, was not able to incorporate as much deuterium (Fig. 2B), and had a peak-width plot that was almost flat (Fig. 2C). Such a dramatic difference between the free and ligated form indicates the binding of c-Abl SH3 with BP1 and illustrates that HX MS is a sensitive method for detecting SH3:ligand interactions.

Figure 2.

Binding to the Abl SH3 domain. (A) Transformed electrospray mass spectra of deuterium labeled Abl SH3 when free (left) and when bound (right) to the BP1 peptide (Rickles et al. 1994). Representative labeling times (there were 11 time points between 10 sec and 8 h) are shown and indicated on the left of each spectrum. Binding was carried out according to K _d,BP1 = 2 μM such that >90% of SH3 molecules were bound. The dotted line is provided for optical guidance. (B) Relative deuterium uptake with time for intact Abl SH3 when alone (▴) or when incubated with BP1 (○). BP1 binding was done as in A. The results have not been adjusted for back-exchange but were obtained under identical conditions (see Wales and Engen 2006a). (C) Peak-width change during the time course of deuterium labeling for Abl SH3 alone (▴) or when incubated with BP1 (○). The mass spectral peak width was measured at FWHM for each time point (see Materials and Methods).

We next applied this technique to investigate the binding of the natural linker sequence with the Abl SH3 domain. A peptide with the sequence of the natural Abl SH2-kinase linker was synthesized and independently incubated (in 50-fold molar excess) with the SH3 domain alongside the analysis of Abl SH3 binding to BP1. The mass spectra for Abl SH3 in the presence of the linker peptide were nearly identical to those of free SH3, and the uptake curves and peak-width plots were highly similar (data not shown). These results suggest that the natural linker sequence, when untethered to the rest of the Abl core, is unable to bind to the SH3 domain.

A construct of c-Abl that included both the SH3 and SH2 domains (SH32) was then investigated with similar methodology (Fig. 3). The shape of the mass spectra, the uptake, and the peak-width properties were similar to that observed for the SH3 domain alone. The SH3 unfolding in the Abl SH32 domain occurred at a similar time to that of the isolated SH3 domain (centroid of the peak in the peak-width plot in Fig. 3C). These results suggest that the presence of the SH2 domain did not affect the SH3 unfolding reaction. Again, HX MS analysis of the SH32 construct in the presence of the high-affinity peptide ligand BP1 abolished the unfolding, indicative of SH3 binding. As was the case for the SH3 domain alone, the natural linker did not bind to SH3 of the SH32 construct (Fig. 3B,C), suggesting that there was no affinity for the linker sequence when added in trans.

Figure 3.

Binding to the Abl SH32 construct. (A) Transformed electrospray mass spectra of Abl SH32 when free (left) and when bound (right) to the BP1 peptide. As in Figure 2, not all spectra are shown. The deuterium labeling time is indicated on the left of each spectrum. The binding conditions are the same as those described in Figure 2. The dotted line is provided for optical guidance. (B) Relative deuterium uptake with time for intact Abl SH32 when alone (▴), when incubated with the natural linker peptide (•), or when incubated with BP1 (○). The binding conditions for BP1 are the same as those described in Figure 2 for Abl SH3. For incubations with the natural linker, a molar ratio of 1:50 (SH32:linker peptide) was used. The results have not been adjusted for back-exchange. (C) Peak-width change during the time course of deuterium labeling for Abl SH32 alone (▴), when incubated with the natural linker peptide (•), or when incubated with BP1 (○). The mass spectral peak width was measured at FWHM for each time point (see Materials and Methods).

Covalent linker attachment

We speculated that while the c-Abl linker did not associate with the SH3 domain in the SH32 construct in trans, perhaps covalent attachment would result in binding due to an increased local concentration. HX MS analyses of Abl SH32L, a version of SH32 in which the linker was covalently attached to the SH2 domain, showed a detectable and reproducible slowing of the unfolding (Fig. 4A,C; Fig. 4E, cf. open and closed triangles). The shift in unfolding was not as dramatic as that afforded by binding to BP1, but the unfolding half-life was still nearly doubled in SH32L versus SH32. The alteration was highly reproducible and therefore considered significant. Modest alteration in the unfolding rate of the SH3 domain implies that the SH3 domain interaction with the linker in SH32L was relatively weak.

Figure 4.

Peak-width changes during the time course of deuterium labeling for various lengths of natural linker covalently attached to Abl SH32 (see Fig. 1C). (A–D) Transformed electrospray mass spectra of (A) Abl SH32, (B) SH32L_1/3, (C) SH32L, and (D) SH32L₊. As in Figure 2, not all spectra are shown. The deuterium labeling time is indicated on the left of each spectrum. A gray bar of constant width is shown in each spectrum at FWHM. (E) Peak-width plot for Abl SH32 (▴), Abl SH32L_1/3 (○), Abl SH32L (△), and Abl SH32L₊ (•). Mass spectral peak width was measured at FWHM for each time point (see Materials and Methods).

We next defined the linker length in Abl SH32L that was needed to inhibit the unfolding. For initial work (SH32L), the length was the same as the linker peptide that was added in trans. A shorter construct with half the length (SH32L_1/3) (see Fig. 1C) and a longer construct with extra residues (SH32L₊) were engineered and investigated in the same HX MS assay. While the SH32L and SH32L₊ constructs had similar behavior and reduced the SH3 unfolding half-life by approximately twofold, SH3 domain unfolding in the SH32L_1/3 construct was similar to that observed for the SH32 construct that did not contain any linker sequence at all. These results indicate that the L_1/3 was unable to bind to Abl SH3 and that the residues in the latter half of the linker sequence were responsible for imparting the ability of the linker to associate with the SH3 domain.

The results of the analyses of the unfolding of the Abl SH3 domain in the various versions of Abl SH32L are summarized in Figure 5. Figure 5 plots the half-life of the unfolding for each construct. Where the unfolding rate was slower than the labeling time range (8 h), the unfolding half-life was designated as infinity. As discussed above, SH3 did not bind to the SH2-kinase linker peptide but bound strongly to the BP1 peptide. This effect is clearly shown in the top two panels of Figure 5 for Abl SH3 and Abl SH32. The effects of covalent attachment of the linker illustrated in Figure 4 are shown again in the plot in Figure 5. Five independent observations showed that the half-life of SH3 unfolding in SH32L was significantly shifted to a longer half-life versus SH3 unfolding in SH32. The addition of more residues (SH32L₊) or subtraction of nearly half the linker (SH32L_1/3) indicated that the sequence NYDKWE is critical to SH3 domain recognition. To localize residues in this sequence important for SH3 interaction, two more constructs were made: SH32L_3/4 and SH32L_4/5 (see Fig. 1C). The linkers of both SH32L_3/4 and SH32L_4/5 did not slow the unfolding of their SH3 domains (Fig. 5, third panel from top). These results implicate Trp 254 and Glu 255 as playing an essential role in stabilizing the linker for interactions with the SH3 domain, even though these residues do not appear to contact the SH3 domain directly in the crystal structure (discussed below).

Figure 5.

Summary of the c-Abl SH3 domain unfolding half-life in various constructs. The top box summarizes the data in Figure 2, the second box from the top summarizes Figure 3, and the third box from the top summarizes parts of Figure 4. Each independent measurement of unfolding half-life is shown as a diamond in the plot, and the number of measurements that were obtained for each construct is shown on the right-hand side of the figure. Replicate measurements were independent determinations from independent protein purifications and labeling experiments. When four or more measurements were made, the 95% confidence interval is shown surrounding the average half-life (large black dot). For three or less replicates, no confidence interval is shown, and the large black dot is the average of the determinations. The average half-life value is indicated under each black dot. In cases in which binding was assessed, the free protein (as a negative control) was always analyzed alongside the bound form under identical experimental conditions (see Materials and Methods). The various constructs are described in Figure 1C.

High-affinity linker mutants

To determine if the propensity to form a polyproline helix (PPII) was a prerequisite for linker:SH3 interaction in c-Abl, mutations were engineered into the c-Abl SH32L construct. Residues corresponding to the P₀ and P₊₃ positions (described by Lim et al. 1994) of the putative polyproline helix in the c-Abl linker were mutated to proline. In previous work, similar substitutions in the Hck linker dramatically stabilized intramolecular association with SH3, presumably by stabilizing the PPII helix and providing a more favorable interface for SH3 binding (Hochrein et al. 2006). This previous Hck linker mutant was termed Hck-HAL, for high affinity linker (Lerner et al. 2005; Hochrein et al. 2006). The corresponding mutants of the Abl SH32L protein were therefore named HAL1, HAL2, and HAL3 (Fig. 1C); and SH3 unfolding in each of these forms of Abl SH32L was assayed in the same way as described above. The results are summarized in the bottom panel of Figure 5. To our surprise, the single mutations (HAL1, HAL2) actually reduced linker binding to the SH3 domain as indicated by a shift in the unfolding half-life toward the value of constructs where there was no binding (i.e., SH32). Such results imply that the single mutants prevented polyproline helix formation or disrupted the conformation of the linker enough to reduce the association it underwent in the wild-type sequence. The double mutant (HAL3) showed a moderate increase in the unfolding half-life, implying that both proline mutations together raised the affinity of the linker sequence for the SH3 domain. Based on these data, it appears that increasing the propensity for polyproline helix formation by the introduction of proline residues in the P₀ and P₊₃ positions in the c-Abl linker does not increase the affinity of the linker for the SH3 domain to the level observed for a strong binder like BP1.

Discussion

While the overall structures of the c-Abl kinase core and SFKs are highly similar, important differences are present. The lack of a C-terminal tail in Abl and the additional interactions involving the N-terminal cap region are all suggestive of other unique properties within Abl that participate in kinase downregulation. We have shown in this study that in constructs of Abl that include the regulatory SH3/SH2 domains and the linker, there are obvious differences in the ability of the SH3 domain to associate with its own linker when compared to the results obtained previously for the Src-family kinase, Hck. In the case of c-Abl, incubation of the natural linker in trans with SH32 elicited no binding to the SH3 domain, whereas tethering the linker onto SH32 was sufficient for binding. This result is in striking contrast to that obtained with an analogous construct from the SFK Hck in which tethering of the natural linker sequence had no effect whatsoever on the SH3 unfolding half-time (Hochrein et al. 2006). These results imply that the Abl linker has evolved for enhanced interaction with SH3, perhaps to overcome the lack of a negative regulatory tail. Enhanced interactions with the linker may also help explain why Abl-1a, the splice variant that is nonmyristoylated, can still be down-regulated (Hantschel and Superti-Furga 2004).

Figure 6 shows a structural alignment of Hck and Abl, optimized for the best fit between the SH2 domains, the linker, and the small lobe of the catalytic domain. Unlike the alignment between Src and Abl (Nagar et al. 2003), there are significant differences between the linker positions in Hck versus Abl due to the three additional residues in the c-Abl linker (see also Fig. 1C). These extra residues lead to a bulge in the C-terminal end of the linker in the crystal structure. Our HX MS data suggest that these residues assist the SH3:linker interaction.

Figure 6.

Comparison of the linker conformation for Abl (PDB code 2FO0) (Nagar et al. 2006) and Hck (PDB code 1QCF) (Schindler et al. 1999). General structural similarity (A) from the front and (B) from above. Hck is shown in green and Abl in gray; the Hck and Abl linkers are colored blue and red, respectively. The two crystal structures were overlaid such that the backbone carbons of the SH2 domains and the small lobes of the catalytic domains were structurally aligned as well as possible (the RMSD for this alignment was 2.35 Å). The amino acids at positions P₀ and P₊₃ in Abl are shown in stick form and labeled. Tryptophan residues W254 (Abl) and W260 (Hck) are shown in stick form. (C) Close-up of the linkers with the same alignment as in A and B. The backbone of the c-Abl linker is colored according to the length of linker described in the constructs in Figure 1C. The Hck linker is colored blue. (Inset) Model of hypothetical π-stacking interactions between the side chains of Y245 and W254.

In the SH32L construct, the SH3 unfolding rate was slowed, indicative of intramolecular SH3:linker binding. However, when this construct was shortened by just two residues (SH32L_4/5), SH3 unfolding was completely restored, indicating that linker binding was completely lost. Figure 6 shows that these two additional residues form the distal part of the “bulge” in the Abl linker (green) near the top of the structure. This bulging loop projects backward (in the classical view; see Fig. 6B) to fill the space between the SH3 domain and the small lobe of the kinase domain. Perhaps these residues are able to pack onto the rest of the linker and stabilize the polyproline helix required for SH3 domain engagement. Such stabilization has been reported for other polyproline constructs (Cobos et al. 2004; Zellefrow et al. 2006) and for the SH3:linker interaction in intact Hck (J.M. Hochrein, T.E. Wales, and J.R. Engen, unpubl.). An alternative proposal for stabilization is that in the absence of interactions with the kinase domain, there might be π-stacking interactions between W254 and Y245 in the Abl linker (see Fig. 6C), which mold the linker into a conformation that has more interactions with the SH3 domain. Simple rotation of Y245 and minor backbone adjustment brings these two π systems into good alignment for such a stabilizing interaction (Fig. 6C, inset). Interestingly, Y245 has been mapped as an important Abl phosphorylation site that has a positive effect on kinase activity; phosphorylation may disturb the π-stacking interaction proposed here; π stacking has also been established as a stabilizing feature of SH2:kinase domain interaction (Hantschel et al. 2003).

When mutations in the P₀ and P₊₃ residues in the linker region of the Hck SH32L construct were created previously (Lerner et al. 2005; Hochrein et al. 2006), the Hck SH3 half-life for unfolding approached infinity, and the ability of the SH3 domain in this construct to bind a high-affinity peptide ligand was inhibited. Comparable proline substitutions in the c-Abl SH32L construct, on the other hand, were much more subdued and did not push the unfolding half-life anywhere near infinity. Single mutations, in fact, shifted the SH3 unfolding half-life to shorter times, suggesting that the introduction of single proline residues at the P₀ and P₊₃ positions decreased rather than increased the affinity of the linker for SH3. In the double P₀, P₊₃ mutant, there was a modest increase in the unfolding half-life, interpreted as an increase in the binding affinity of the linker and SH3. These results may mean that the comparable P₀, P₊₃ proline mutations in the c-Abl linker either do not force the Abl linker into a polyproline helix as they do for Hck, that a polyproline helix is not a requirement for tight binding of the linker to the SH3 domain, or that other factors make the Abl SH3:linker interaction more complex than just the simple formation of a polyproline helix.

We conclude that a unique conformation of the Abl SH2-kinase linker is created by the insertion of two residues (W254, E255) not present in SFK linkers, and that these additional residues stabilize the linker and promote SH3:linker interaction. These residues may be instrumental in keeping the SH3 domain in the proximity of the catalytic domain for downregulation of catalytic activity and for assisting kinase domain:SH2 domain interactions (Hantschel et al. 2003). In the Bcr-Abl protein, fusion to Bcr may alter linker interactions and contribute to disruption of the down-regulatory conformation. Future studies will address this possibility.

Materials and Methods

DNA constructs

The coding regions of the human c-Abl SH3 domain (G76-S145), SH32 (G76-K238), SH32L_1/3 (G76-T243), SH32L (G76-E255), and SH32L₊ (G76-T259) were amplified by PCR and subcloned into the bacterial expression vector pET-14b (Novagen; see Fig. 1C for construct boundaries). pET-14b encodes a 6xHis tag at the N terminus of each construct for purification. The pET-14b-SH32L vector was used as a template to generate the truncation mutants SH32L_3/4 (G76-Y251) and SH32L_4/5 (G76-K253) by mutation of the appropriate codon to a stop codon using the QuickChange site-directed mutagenesis kit (Stratagene). The high-affinity linker mutants HAL1 (K241P) and HAL2 (V244P) were also created with QuickChange using the pET-14b-SH32L plasmid as a template. The pET-14b-HAL1 vector was subsequently used to generate mutant HAL3 (K241P, V244P). All mutations were verified by DNA sequencing and by confirming the mass of each protein (see below).

Protein production and peptides

Proteins were overexpressed in Escherichia coli Rosetta2 (DE3) strains (Novagen). Cells were grown at 37°C in LB medium supplemented with ampicillin (100 μg/mL) and chloramphenicol (20 μg/mL). When the optical density (600 nm) reached 0.8, protein expression was induced by addition of 0.2 mM IPTG. After 4 h of induction at 37°C, cells were collected by centrifugation, resuspended in binding buffer (50 mM NaH₂PO₄ at pH 8.0, 300 mM NaCl, 10 mM imidazole, 0.5 mM PMSF), and lysed by sonication at 4°C. The cell lysate was clarified by centrifugation, and the supernatant was mixed end-over-end with Ni-NTA agarose beds (QIAGEN) for 1 h at 4°C. The beads were washed with the binding buffer, and the bound recombinant proteins were eluted with 250 mM imidazole. Fractions containing the recombinant proteins were determined by UV absorption and SDS-PAGE. The relevant fractions were pooled and incubated with 80 units of human thrombin protease (Amersham Biosciences) for 12 h with gentle agitation at room temperature to cleave the 6×His tag. The cleaved proteins were eluted and separated from the 6×His tag by reincubation with Ni-NTA agarose beads. The proteins were further purified by size exclusion chromatography with a G-75 column (100 cm × 1.5 cm) that had been equilibrated with 25 mM NaH₂PO₄, 25 mM Na₂HPO₄ (pH 7.0), 0.1 mM NaCl, and 0.005 mM NaN₃. The final purity of each protein was estimated to be >98% by SDS-PAGE and electrospray mass spectrometry. Mass spectrometry also verified that the mutations were correct, as each theoretical mass matched the measured mass to within 0.5 Da.

The BP1 peptide ([Ac]-AEPPPYPPPPIPGGK-[NH₂]) (Rickles et al. 1994) and the natural SH3 linker peptide ([Ac]-RNKPTVYGVSPNYDKWE-[NH₂]; c-Abl residues 239–255) were both synthesized by the Sigma-Genosys Company and purified (>95%) by reversed phase HPLC. The mass of each peptide was verified by electrospray mass spectrometry. The peptides were reconstituted to a final concentration of 20 mM in 50 mM phosphate buffer (pH 7.2).

Deuterium exchange reactions

Protein stock solutions were prepared by diluting the recombinant proteins to a final concentration of 100 μM (by Bradford assay) with 50 mM sodium phosphate buffer (pH 7.0, H₂O). Deuterium exchange was initiated by dilution of the stock solution 15-fold with 50 mM sodium phosphate (pD 8.3, D₂O) at 25°C. At each time point (ranging from 10 sec to 8 h), ∼300 pmol of protein were removed from the exchange reaction, and the labeling was quenched by adjusting the pH to 2.5 with 0.5 M HCl. Quenched samples were immediately frozen on dry ice and stored at −80°C until analysis.

For incubations including peptides in trans, the percent SH3 or SH32 bound was estimated using a K _d of 2 μM for BP1 (Rickles et al. 1994). BP1 was added such that 92.5% of SH3 or SH32 molecules were bound to peptide ligand in the labeling solution. For the natural SH2-linker peptide where the K _d was unknown, the peptide was added in a 50-fold molar excess. All mixtures were incubated for 30 min at room temperature before labeling began. As a negative control, SH3 or SH32 was incubated with 20 mM of the nonbinding peptide angiotensin I (Sigma); therefore, all data listed as free are actually the constructs in the presence of angiotensin I.

Analysis of deuterium incorporation by mass spectrometry

Labeled proteins were rapidly thawed at 0°C, and ∼300 pmol of protein were injected onto a Shimadzu HPLC (LC-10ADvp) fitted with a C-18 protein trap and desalted for 3 min with 2% ACN, 98% H₂O, and 0.05% TFA at a flow rate of 100 μL/min. To minimize deuterium back-exchange, the trap, the injector, and the associated tubing were placed in an ice bath. Proteins were eluted into a Waters QTOF2 mass spectrometer with a single step to 98% ACN at a flow rate of 50 μL/min. The relative deuterium level was determined by subtracting the mass of the labeled protein from the mass of unlabeled protein at each exchange time point. The deuterium levels were not corrected for back-exchange and are therefore reported as relative (Wales and Engen 2006a). To determine the peak width, the mass/charge data were transformed to a mass-only scale using MassLynx software. The peak width was measured at full-width at half-maximum (FWHM) either by hand or with HX-Express (Weis et al. 2006a) for each time point and plotted against the deuterium labeling time. The SH3 domain unfolding half-life (t _1/2) of each c-Abl construct was determined by an intersection method. A linear regression was performed on each side of the peak in each peak-width plot (i.e., Figs. 2C, 3C, 4E). The intersection point of the two linear equations was determined and used as the half-life of unfolding. This method was found to be simpler and just as accurate as fitting a Gaussian equation to the peak-width peak. When more than one measurement was made for each construct, the t _1/2 value reported corresponds to the average of the replicates (as described in the Fig. 5 legend). The 95% confidence interval was also calculated when more than four independent experiments were made.