Enhanced SH3/Linker Interaction Overcomes Abl Kinase Activation by Gatekeeper and Myristic Acid Binding Pocket Mutations and Increases Sensitivity to Small Molecule Inhibitors

Background: Abl kinases are regulated by noncatalytic domains that allosterically impact kinase domain structure and inhibitor sensitivity.

Results: Enhanced SH3/linker interaction suppresses c-Abl core protein dynamics and sensitizes Bcr-Abl to kinase domain inhibitors.

Conclusion: SH3/linker interaction influences kinase dynamics in the context of Bcr-Abl.

Significance: Stabilizers of SH3/linker interaction may sensitize Bcr-Abl to kinase domain inhibitors, providing a new route to allosteric kinase control.

Abstract

Multidomain kinases such as c-Src and c-Abl are regulated by complex allosteric interactions involving their noncatalytic SH3 and SH2 domains. Here we show that enhancing natural allosteric control of kinase activity by SH3/linker engagement has long-range suppressive effects on the kinase activity of the c-Abl core. Surprisingly, enhanced SH3/linker interaction also dramatically sensitized the Bcr-Abl tyrosine kinase associated with chronic myelogenous leukemia to small molecule inhibitors that target either the active site or the myristic acid binding pocket in the kinase domain C-lobe. Dynamics analyses using hydrogen exchange mass spectrometry revealed a remarkable allosteric network linking the SH3 domain, the myristic acid binding pocket, and the active site of the c-Abl core, providing a structural basis for the biological observations. These results suggest a rational strategy for enhanced drug targeting of Bcr-Abl and other multidomain kinase systems that use multiple small molecules to exploit natural mechanisms of kinase control.

Introduction

Chronic myelogenous leukemia (CML)³ is characterized by the Philadelphia chromosome translocation, which fuses the BCR (break point cluster) region locus on chromosome 22 with the c-ABL (Abelson tyrosine kinase) proto-oncogene on chromosome 9. This translocation results in the expression of Bcr-Abl, a constitutively active protein-tyrosine kinase that drives CML pathogenesis through downstream pathways that promote cell growth and survival. Expression of Bcr-Abl in bone marrow cells induces a CML-like syndrome in mouse models, demonstrating that Bcr-Abl kinase activity alone is sufficient to cause the disease (1, 2).

Clinical management of CML has been revolutionized by imatinib mesylate, a selective ATP-competitive inhibitor of Bcr-Abl kinase activity (3). Despite this clinical success, imatinib is less effective in advanced CML due to selection of drug-resistant mutants of Bcr-Abl (4). Resistance mutations often arise in the drug binding site, and include the T315I gatekeeper mutation that also enhances Bcr-Abl kinase and transforming activities (5). Other mutations can occur outside of the active site and allosterically reduce drug binding by promoting an active kinase domain conformation incompatible with imatinib binding (6). Second generation ATP-competitive inhibitors, including nilotinib and dasatinib, have been approved for the clinical management of imatinib-resistant CML (7). Although these newer inhibitors are more potent, they do not inhibit the Bcr-Abl T315I mutant.

In contrast to Bcr-Abl, c-Abl kinase activity is tightly regulated in cells. Structural and functional studies attribute intramolecular interactions to down-regulation of the c-Abl kinase core, which consists of a myristoylated N-terminal region (Ncap), followed by regulatory SH3 and SH2 domains, the SH2-kinase linker and the kinase domain (8). The kinase domain is comprised of a smaller N-lobe connected to a larger C-lobe through a flexible hinge, allowing for articulation of the two lobes during kinase activation. The Ncap, SH3, and SH2 domains work in concert to keep the kinase in the autoinhibited state (8–10). By binding to the SH2-kinase linker, which adopts a polyproline type II helical structure, the SH3 domain stabilizes the N-lobe of the kinase domain in the inactive state. Mutations within the SH3 domain, as well as the linker, switch on the kinase and transforming activities of c-Abl (11). Moreover, phosphorylation of residues in the linker (Tyr²⁴⁵) or in the SH3 domain (Tyr⁸⁹) disrupt SH3/linker engagement and also enhance Abl kinase activity (12–14). The SH2 domain docks onto the back of the kinase domain C-lobe through a network of hydrogen bonds to further stabilize the down-regulated conformation of the core. Mutation of SH2 Tyr¹⁵⁸ disturbs this interaction and leads to kinase activation (10). Finally, the N-terminal myristate group penetrates into a deep pocket in the C-lobe, inducing a kink in helix αI that is critical for SH2/C-lobe interaction (10). Mutations in the hydrophobic pocket of the C-lobe that prevent myristic acid insertion (e.g. A356N) activate the kinase (10). Recently, a new class of allosteric Bcr-Abl inhibitors has been described that target the C-lobe myristic acid binding site (15, 16). These compounds, of which GNF-2 is the prototype, stabilize the inactive conformation of the Abl core and work in concert with ATP-competitive inhibitors to overcome imatinib-resistant mutants of Bcr-Abl, including T315I (16, 17).

In the context of Bcr-Abl, Bcr fusion prevents N-terminal myristoylation of c-Abl and deletes most of the Ncap, thereby removing one important element of kinase down-regulation. In addition, Bcr adds an N-terminal coiled-coil oligomerization domain (18) that induces clustering of Bcr-Abl and promotes kinase activation through trans-phosphorylation of the activation loop. Unlike the c-Abl core, the contribution of the SH3 and SH2 domains to kinase domain control within Bcr-Abl is less clear. Despite the constitutive activation of Abl that results from Bcr fusion, mounting evidence suggests that the SH3 and SH2 domains are not necessarily displaced from their regulatory positions on the back of the kinase domain. Mutations in the Bcr-Abl SH3 and SH2 domains as well as the linker promote imatinib resistance, consistent with the kinase domain adopting the active conformation incompatible with drug binding (6). Similarly, phosphorylation of the Bcr-Abl linker and SH3 domain by the Src-family kinase Hck also results in imatinib resistance, most likely by perturbing SH3/linker interaction (19). A logical conclusion from these findings is that the SH3 and SH2 domains maintain their regulatory influence on the kinase domain even in the context of Bcr-Abl. This raises the exciting possibility that strengthening the regulatory SH3/linker interaction present in Bcr-Abl may stabilize a down-regulated kinase domain conformation, thus sensitizing the kinase domain to both imatinib and allosteric inhibitor action.

In this study, we explored this possibility by creating a series of modified c-Abl and Bcr-Abl proteins with enhanced SH3/linker interactions. By systematically increasing the proline content of the linker, we identified high-affinity linkers that stabilized intramolecular SH3 binding without disturbing the overall regulation of the kinase core. Enhanced SH3/linker interaction completely reversed c-Abl core activation by a C-lobe myristate binding pocket mutation, and substantially reduced activation by mutations in the ATP binding site (T315I) as well as the SH2/C-lobe interface (Y158D). Remarkably, enhanced SH3/linker interaction sensitized Bcr-Abl not only to imatinib but also to the allosteric inhibitor, GNF-2. Hydrogen exchange mass spectrometry of recombinant Abl core proteins with high-affinity linkers revealed dynamic coupling between the SH3/linker interface and the GNF-2 binding site in the C-lobe myristate binding pocket. Taken together, these studies provide strong evidence that the regulatory SH3/linker interaction is retained in the context of Bcr-Abl, and that overall kinase regulation is controlled by an “allosteric triangle” linking the SH3 domain, the C-lobe myristate binding pocket, and the active site. Small molecules enhancing natural regulatory interaction at the SH3/linker interface may have utility as chemical sensitizers of existing Bcr-Abl drug action.

EXPERIMENTAL PROCEDURES

Cell Culture

The human GM-CSF-dependent myeloid leukemia cell line TF-1 was obtained from the American Type Culture Collection (ATCC) and maintained in RPMI 1640 supplemented with 10% fetal bovine serum (FBS), 100 units/ml of penicillin G, 100 μg/ml of streptomycin sulfate, 0.25 μg/ml of amphotericin (antibiotic-antimycotic, Invitrogen), and 1 ng/ml of human recombinant GM-CSF. Sf9 insect cells were maintained in Grace's medium (Invitrogen) supplemented with 10% FBS. 293T cells were obtained from the ATCC and cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS and antibiotic-antimycotic.

Expression and Purification of Abl SH3-SH2-Linker Proteins

The high affinity linker mutants HAL1 (K241P), HAL2 (V244P), HAL4 (T243P), and HAL5 (G246P) were created by site-directed mutagenesis using the QuikChange method (Stratagene) and the pET-14b/SH32L plasmid as a template. This plasmid was previously used to express wild-type Abl Ncap-SH3-SH2-linker protein for HX MS analysis (12, 20). The pET-14b-HAL1 and -HAL4 vectors were then used in a subsequent round of mutagenesis to generate HAL3 (K241P, V244P) and HAL6 (T243P, G246P), respectively. The pET-14b-HAL3 vector was used to generate HAL7 (K241P, T243P, and V244P), whereas the pET-14b-HAL7 vector was used to create HAL8 (K241P, T243P, V244P, and G246P). The HAL8 construct was then used to create HAL9 (K241P, T243P, V244P, G246P, and V247P) and subsequently HAL10 (K241P, T243P, V244P, G246P, V247P, and S248P). All Abl SH32L proteins were expressed in Escherichia coli Rosetta2 (Novagen) and purified using affinity chromatography with nickel-nitrilotriacetic acid-agarose beads (Qiagen). Following cleavage of the hexahistidine tag at the N terminus by human thrombin protease, the proteins were further purified by size exclusion chromatography. The theoretical mass for each protein matched the measured mass to within 0.5 Da by electrospray mass spectrometry (data not shown).

Expression and Purification of Ncap-Abl Core Proteins

The Ncap-Abl core encompasses residues 1–531 of human c-Abl-1b with an internal deletion of residues 15–56 and a C-terminal cleavage site for the tobacco etch virus protease followed by a hexa-histidine tag. All 10 HAL sequences were introduced into the Ncap-Abl core coding region using a two-step PCR-based strategy and the corresponding Abl SH3-SH2-HAL constructs as templates. The Ncap-Abl core coding sequences with modified linkers were assembled in the cloning vector pSP72 (Promega) and subsequently subcloned into pCDNA3.1 (Invitrogen) for transient expression in 293T cells and pVL1392 (BD Biosciences) for expression in Sf9 insect cells. The pVL1392/Ncap-Abl constructs were used to create high-titer recombinant baculoviruses in Sf9 insect cells using linearized Baculogold DNA and the manufacturer's protocol (BD Biosciences). For protein production, Sf9 cells were grown in monolayers on large plates and co-infected with Ncap-Abl and YopH phosphatase baculoviruses. YopH phosphatase promotes the down-regulated conformation of Ncap-Abl and facilitates high-yield purification (21). Sf9 cells were grown for an additional 72 h post-infection, harvested by centrifugation, and resuspended in 20 mm Tris-HCl (pH 8.3), 10% glycerol, and 5 mm β-mercaptoethanol. Pellets were lysed by sonication and the lysates were clarified by centrifugation at 16,000 × g for 30 min. The proteins were purified from the supernatant using a combination of ion exchange and affinity chromatography as described previously (21). Purified proteins were dialyzed against 20 mm Tris-HCl (pH 8.3) containing 100 mm NaCl and 3 mm DTT.

Transient Expression of c-Abl Core Proteins in 293T Cells

Human 293T cells (10⁶) were plated in 60-mm dishes and incubated at 37 °C overnight, followed by transfection with 2.5 μg of plasmid DNA and X-tremeGENE9 DNA transfection reagent (Roche Applied Science). Cells were lysed by sonication 24 h later in ice-cold RIPA buffer supplemented with protease and phosphatase inhibitors as described (19). Cell lysates were clarified by centrifugation at 16,000 × g for 10 min at 4 °C and the protein concentrations were determined using the Bradford assay reagent (Pierce). Aliquots of total protein were heated directly in SDS sample buffer and separated by SDS-PAGE. For immunoprecipitation, protein concentrations were first normalized in lysis buffer, followed by addition of 1 μg of anti-His antibody (Abcam) and 20 μl of protein G-Sepharose (50% slurry; GE Healthcare). Following incubation at 4 °C overnight, immunoprecipitates were washed three times in RIPA buffer and heated in SDS sample buffer. Following SDS-PAGE, proteins were transferred to nitrocellulose membranes (Bio-Rad) for immunoblot analysis. Immunoreactive proteins were visualized and quantitated with appropriate infrared (IR) dye-labeled secondary antibodies using the Odyssey imaging system (LI-COR Biosciences). Antibodies used in this part of the study include Abl polyclonal (sc-131; Santa Cruz Biotechnology), Abl monoclonal (sc-23; Santa Cruz), phospho-Abl (Tyr⁴¹², Tyr²⁴⁵, Tyr⁸⁹; Cell Signaling Technology), and anti-phosphotyrosine (pY99; Santa Cruz).

Construction of Bcr-Abl HAL Vectors and Retroviral Transduction of the Human Myeloid Leukemia Cell Line, TF-1

Oligonucleotides (648 bp) spanning the HAL9 and Bcr-Abl kinase domain plus flanking restriction sites were commercially synthesized (DNA 2.0). Each of these DNA fragments was then swapped for the corresponding region of wild-type Bcr-Abl and subcloned into the retroviral vector pMSCVneo (Clontech). The Bcr-Abl kinase domain mutation T315I was generated by site-directed mutagenesis as described above. Following DNA sequence verification of all constructs, retroviral stocks were produced by co-transfection of 293T cells with each pMSCV/Bcr-Abl plasmid and an amphotropic packaging vector as described previously (19, 22). TF-1 cells (10⁶) were incubated with 5 ml of viral stock in the presence of 4 μg/ml of Polybrene (Sigma) and centrifuged at 3000 × g for 3 h at room temperature. Following infection, cells were washed, transferred to regular medium for 24 h, and then put under G418 selection (800 μg/ml) for 14 days. Following selection, cells were maintained in medium supplemented with 400 μg/ml of G418.

Proliferation and Apoptosis Assays

Proliferation of Bcr-Abl-transformed TF-1 cells was assayed in the absence of GM-CSF using the CellTiter-Blue Cell Viability assay (Promega) according to the manufacturer's protocol. Fluorescence intensity was measured using a Gemini XS microplate spectrofluorimeter (Molecular Devices) at 544/590 nm excitation and emission wavelengths, respectively. Caspase activation was measured using the ApoOne Caspase 3/7 assay (Promega) according to the manufacturer's protocol. Fluorescence intensity was measured at 485/520 nm excitation and emission wavelengths, respectively.

Hydrogen Exchange Experiments

Hydrogen exchange experiments were performed essentially as described previously (21). Stock solutions of each Abl core protein (WT, HAL9, A356N, and HAL9-A356N) were prepared at 20 pmol/μl in 20 mm Tris-HCl (pH 7.96), 100 mm NaCl, 3 mm DTT in H₂O. Deuterium exchange was initiated by dilution of each protein 15-fold with an identical buffer prepared in D₂O at room temperature. At each deuterium exchange time point (ranging from 10 s to 4 h) an aliquot from the exchange reaction was removed and labeling was quenched by adjusting the pH to 2.5 with an equal volume of quench buffer (0.8 m guanidinium HCl, 0.8% formic acid) followed by freezing on dry ice. Samples were stored at −80 °C prior to pepsin digestion and mass analysis.

Mass Analysis

Each frozen sample was thawed rapidly and injected into a custom Waters nanoACQUITY UPLC HDX Manager^TM (Waters Corp., Milford, MA) and analyzed as previously described (23). The protein samples were digested using a Poroszyme immobilized pepsin cartridge (Applied Biosystems), which was accommodated within the UPLC system. The cooling chamber of the UPLC system that housed all the chromatographic elements was held at 0.1 °C for the entire analysis. The injected peptides were trapped and desalted for 3 min at 100 μl/min and then separated over 6 min with an 8–40% acetonitrile:water gradient at 40 μl/min. Chromatography was performed on a 1 × 100-mm Waters ACQUITY UPLC C18 BEH column containing 1.7-μm particles and the back pressure averaged 8800 p.s.i. at 0.1 °C. The average amount of back-exchange using this experimental setup was 18 to 25%, based on analysis of highly deuterated peptide standards. Deuterium levels were not corrected for back-exchange and are therefore reported as relative (24). However, all comparisons were done under identical experimental conditions thus negating the need for back-exchange correction (24). The UPLC step was performed with protonated solvents, thereby allowing deuterium to be replaced with hydrogen from side chains and the amino/carboxyl termini, which exchange much faster than amide linkages (25). All experiments were performed in duplicate and the error of determining the deuterium levels was ±0.20 Da in this experimental setup consistent with previously obtained values (26).

Mass spectra were obtained with a Waters XEVO G2 TOF equipped with a standard ESI source. The instrument was configured as follows: capillary at 3.2 kV, trap collision energy at 6 V, sampling cone at 35 V, source temperature of 80 °C, and desolvation temperature of 175 °C. Mass spectra were acquired over an m/z range of 100 to 2000. Mass accuracy was ensured by calibration with 500 fmol/μl of GFP, and was less than 10 ppm throughout all experiments. The mass spectra were processed with the software package DynamX^TM (Waters) by centroiding an isotopic distribution corresponding to the +2, +3, or +4 charge state of each peptide. Deuteration levels were calculated by subtracting the centroid of the isotopic distribution for peptide ions of undeuterated protein from the centroid of the isotopic distribution for peptide ions from the deuterium-labeled sample. The resulting relative deuterium levels were automatically plotted versus the exchange, in time. Identification of the peptic fragments was accomplished through a combination of exact mass analysis and MS^E using Identity Software (Waters) (27). MS^E was performed by a series of low-high collision energies ramping from 5 to 30 V, therefore ensuring proper fragmentation of all the peptic peptides eluting from the LC system. Peptic maps were obtained using the DynamX software package. Local changes in isotope exchange were mapped to the crystal structure of c-Abl (Protein Data Bank code 2FO0) using PyMOL (28).

RESULTS

Design and Characterization of High Affinity Linker (HAL) Variants of Abl

The first objective of this study was to modify the Abl SH2-kinase linker to enhance intramolecular interaction with the SH3 domain. In the structure of the c-Abl core, this linker forms a polyproline type II helix that serves as an internal docking site for the SH3 domain (see Introduction). This interaction is essential to down-regulation of c-Abl activity. Compared with high affinity peptide ligands for the Abl SH3 domain (29), however, the linker represents a suboptimal SH3 docking site, with a charged residue (Lys²⁴¹) facing the hydrophobic binding surface of SH3 (Fig. 1A). To strengthen the internal SH3/linker interaction, we systematically substituted linker residues in close proximity with the SH3 surface with prolines, resulting in a set of 10 modified linkers (see Fig. 1B for sequences). These engineered sequences are referred to hereafter as HALs.

FIGURE 1.

Abl linker proline substitutions enhance SH3 engagement. A, model of the down-regulated wild-type c-Abl core is shown at the top left, and includes the Ncap (light blue), SH3 domain (red), SH2 domain (green), SH2-kinase linker (orange), and kinase domain (purple). The SH3/linker interface (boxed) is enlarged in the center to highlight the side chains of the linker residues that were modified in the HAL9 mutant. A model of the SH3/linker interface from HAL9 shows the positions of the five proline substitutions (left). Models and residue numbering are based on the x-ray crystal structure of the myristoylated c-Abl core (Protein Data Bank 2FO0). B, the sequence of the wild-type c-Abl linker is aligned with each of the 10 HAL sequences as well as the linker of the Src family kinase, Hck. Proline substitutions introduced into each of the HAL sequences are highlighted in red. C, analysis of intramolecular SH3/linker interaction by HX MS. Recombinant c-Abl proteins consisting of the SH3 and SH2 domains plus each of the linkers shown were expressed in bacteria, purified, and subjected to HX MS analysis as described in the text. Data are expressed as the unfolding half-life of the SH3 domain, which is directly proportional to the strength of the internal SH3/linker interaction. No change in the unfolding of SH3 could be observed with HAL10 over the time scale of this experiment (∞). The error associated with the slow-down factor is approximately ±1 min based on triplicate determinations done for HAL1, HAL2, and HAL3. The wild-type protein was analyzed five times, and the remaining HAL proteins were done twice.

To determine whether enhanced proline content stabilized intramolecular SH3 binding, the modified linkers were first incorporated into bacterial expression vectors for the Abl SH3-SH2-linker region. The resulting proteins were expressed, purified, and examined for enhanced SH3/linker interaction using hydrogen exchange mass spectrometry (HX MS), a method previously developed to explore the impact of linker engagement on Abl SH3 dynamics (12, 20). With HX MS, protein dynamics are determined by measuring the rate of backbone amide hydrogen exchange following transfer to a deuterated solvent (24). Under physiological conditions, HX MS has shown that the c-Abl SH3 domain undergoes a cooperative partial unfolding event that is stabilized by ligand binding either in cis (linker) or in trans (high affinity peptide). The extent of SH3 unfolding is directly proportional to the affinity of the ligand, and complete inhibition of SH3 unfolding can be achieved by binding to the BP1 peptide, a sequence with relatively high affinity toward the Abl SH3 domain (12, 20). HX MS measurements provide a direct assay for the impact of the HAL substitutions on intramolecular SH3 engagement.

HX MS was performed on the wild-type Abl SH3-SH2-linker protein as well as all 10 recombinant HAL mutants, and the results are summarized in Fig. 1C. Introduction of single proline residues in the P₊₃ (HAL1) or P₀ (HAL2) positions (30, 31), which directly oppose the SH3 domain in the down-regulated structure of the c-Abl core, did not alter SH3 unfolding relative to the wild-type protein. However, double proline substitution at both of these positions (HAL3) showed a moderate increase in the SH3 unfolding half-life (i.e. reduced SH3 dynamics due to enhanced linker binding) compared with the wild-type and singly substituted linkers, although this effect was subtle (about 2-fold). Substitution of a single proline at the P₊₁ position (HAL4) also resulted in a doubling of the unfolding half-life relative to the wild-type linker, whereas a single substitution at the P₋₂ position (HAL5) had no effect. HAL6, which combines these substitutions, showed the same unfolding half-life as HAL4, and addition of a third proline at position P₊₁ did not further enhance SH3 engagement (HAL7). However, sequential addition of prolines to HAL7 at the P₋₂ (HAL8) and P₋₃ (HAL9) positions resulted in dramatic increases to the unfolding half-life of about 5- and 7-fold for HAL8 and HAL9, respectively. Addition of one final proline to HAL9 at position P₋₄ (HAL10) completely suppressed cooperative unfolding, suggesting that the SH3 domain is essentially locked to this modified linker. Binding to the BP1 peptide (20) in trans produced results similar to those obtained for HAL10, thereby acting as a positive control for very tight binding and total inhibition of Abl SH3 unfolding (data not shown).

The next goal was to identify HAL sequences that enhanced SH3 interaction without disturbing overall down-regulation of the c-Abl core. To accomplish this, we introduced all 10 HAL sequences into a c-Abl kinase core protein composed of the Ncap, the SH3 and SH2 domains, the SH2-kinase linker, and the kinase domain. The x-ray crystal structure of this core protein has been determined previously and it contains all of the structural elements necessary for down-regulation (8). To determine the impact of the HAL sequences on Abl kinase activity, the wild-type and HAL core proteins were expressed in 293T cells, immunoprecipitated, and immunoblotted with antibodies against key phosphotyrosine residues in the activation loop (Tyr(P)⁴¹²) and the SH3 domain (Tyr(P)⁸⁹). As shown in Fig. 2A, Abl core proteins with HAL sequences 1, 4, 6, 8, and 9 were phosphorylated at both of these sites to the same or reduced levels relative to wild-type Abl, suggesting that these modified linkers adopt a polyproline type II helical conformation compatible with down-regulation of the wild-type Abl kinase core. As an independent measure of kinase activity, we also assessed levels of tyrosine phosphoproteins in lysates from the transfected cells. As shown in Fig. 2B, the extent of protein-tyrosine phosphorylation in the cell lysates parallels the level of phosphorylation at each of the regulatory tyrosines in each c-Abl core protein.

FIGURE 2.

Abl-HAL core protein expression and relative kinase activity. A, each of the HAL sequences shown in Fig. 1B were introduced into the c-Abl kinase core, consisting of the Ncap, the SH3 and SH2 domains, the SH2-kinase linker, and the kinase domain. Each Abl core protein was expressed in 293T cells, immunoprecipitated, and immunoblotted with phosphospecific antibodies against phosphotyrosine residues in the activation loop (pY412) and the SH3 domain (pY89) as well as for Abl protein recovery. B, overall protein-tyrosine phosphorylation was assessed in the cell lysates by immunoblotting with Abl blots performed as a control. Two independent experiments resulted in the same rank order of activities; a representative example is shown.

In contrast, Abl core proteins with HAL substitutions 2, 3, 5, and 7 showed much higher phosphorylation of autoregulatory tyrosine sites and a concomitant increase in cellular phosphotyrosine content (Fig. 2). The HAL10 core protein also appeared to be up-regulated, albeit to a lesser extent. HAL sequences that enhance SH3 engagement but also increase kinase activity are likely to produce additional changes to the linker structure that interfere with effective down-regulation of kinase activity.

Enhanced SH3/Linker Interaction Overcomes Abl Core Activation by Gatekeeper and Myristate Binding Pocket Mutations

Previous studies have shown that the Abl “gatekeeper” mutant associated with imatinib resistance (T315I) is more active and dynamic in vitro than the corresponding wild-type kinase (21). Furthermore, mutations of the N-terminal c-Abl myristoylation site or the complementary myristate binding pocket in the kinase domain C-lobe (e.g. A356N) strongly up-regulate Abl kinase activity (10). These observations led us to explore whether enhanced SH3/linker interaction could reverse these activating influences through an allosteric mechanism. To test this possibility, we first confirmed the activating effects of these Abl mutations using the 293T cell expression system. Three Abl core protein mutants were expressed: the T315I gatekeeper mutant, a myristoylation-defective mutant in which the essential N-terminal glycine is replaced with alanine (G2A), and the myristic acid binding pocket mutant (A356N) described above. All three mutations strongly enhanced phosphorylation of the activation loop (Tyr(P)⁴¹²), the linker (Tyr(P)²⁴⁵), and the SH3 domain (Tyr(P)⁸⁹) and resulted in a parallel increase in overall protein-tyrosine phosphorylation in cell lysates (Fig. 3).

FIGURE 3.

Activating mutations of the Abl-core protein. A, model of the down-regulated, myristoylated c-Abl core showing the positions of key tyrosine autophosphorylation sites in the SH3 domain (Tyr⁸⁹), SH2-kinase linker (Tyr²⁴⁵), and kinase domain activation loop (Tyr⁴¹²). The positions of three activating mutations are also shown, which involve Ile substitution for the gatekeeper residue (Thr³¹⁵), Asn substitution for Ala³⁵⁶ in the myristate binding pocket (A356N), and substitution of Gly² with Ala (G2A), which prevents myristoylation. (Numbering is based on the crystal structure of the c-Abl core (Protein Data Bank code 2FO0) with the exception of the gatekeeper residue, which is numbered as Thr³¹⁵ by convention; this position corresponds to Thr³³⁴ in this structure.) B, the wild-type (WT) c-Abl core and active mutants described in A (T315I, A356N, and G2A) were expressed in 293T cells, immunoprecipitated, and immunoblotted with phosphospecific antibodies against Tyr(P)⁴¹², Tyr(P)²⁴⁵, and Tyr(P)⁸⁹ as well as for Abl protein recovery. Untransfected cells served as the negative control. C, overall protein-tyrosine phosphorylation was assessed in the cell lysates by immunoblotting; Abl blots were also performed as a control.

To determine whether enhanced SH3/linker interaction influences the effect of activating mutations in the Abl core, we combined all 10 HAL sequences with the myristate binding pocket mutation (A356N). We then expressed these proteins in 293T cells and looked for changes in autophosphorylation of the two regulatory tyrosines. As shown in Fig. 4A, HAL9 completely reversed the potent activating effects of the A356N mutation as judged by the reduced phosphorylation of regulatory activation loop and SH3 domain tyrosine sites. In addition, the overall phosphotyrosine content of lysates from cells expressing Abl-HAL9-A356N was dramatically reduced. These data support the idea that the SH3/linker interaction allosterically overrides the activating effect of the myristic acid binding pocket mutation.

FIGURE 4.

Effect of HAL9 substitution on c-Abl activation caused by myristic acid binding pocket mutation (A356N), loss of myristoylation (G2A), and gatekeeper mutation (T315I). Each of the 10 HAL sequences (Fig. 1B) was combined with: A, the A356N mutation in the myristic acid binding pocket; B, the G2A mutation in the myristoylation signal sequence; and C, the T315I gatekeeper mutation in the kinase domain. The resulting compound mutants were expressed in 293T cells, and Abl proteins were immunoprecipitated and immunoblotted with phosphospecific antibodies against the activation loop (pY412) and the SH3 domain (pY89) as well as for Abl protein recovery. Overall protein-tyrosine phosphorylation was assessed in the cell lysates by immunoblotting (pTyr) with Abl blots performed as a control. Three independent experiments resulted in the same rank order of activities; representative examples are shown.

We next combined each of HAL sequences with the myristoylation-defective Abl core mutant, G2A. In this case, HAL9 substitution only partially reversed the activating effects of this mutation (Fig. 4B). This result suggests that disruption of Myr-Ncap interaction with the kinase domain C-lobe by mutating the myristoylation site results in a different active state of the Abl core compared with mutation of the myristic acid binding pocket (see “Discussion”).

We have previously observed that the activating gatekeeper mutation in the c-Abl core (T315I) causes local conformational changes in the kinase domain and at a distance in the SH3 domain (21). To investigate whether enhanced SH3/linker interaction impacted the activating effect of the gatekeeper mutation, we combined all 10 HAL sequences with T315I and determined their relative activity in 293T cells. As shown in Fig. 4C, HAL9 reversed the Abl core T315I activation loop of tyrosine phosphorylation (Tyr(P)⁴¹²) by more than 60%, whereas phosphorylation of the SH3 domain (Tyr(P)⁸⁹) was almost completely suppressed. Consistent with these observations, HAL9 substitution also substantially reduced the cellular phosphotyrosine content. These results show that the destabilizing impact of the Abl gatekeeper mutation on the Abl kinase domain remains under the allosteric control of the SH3/linker interaction.

Enhanced SH3/Linker Interaction Overcomes Abl Kinase Activation by SH2-Kinase Interface Mutation

The interface of the SH2 and kinase domain C-lobe is also important for autoinhibition of the c-Abl kinase. In the x-ray crystal structure of the down-regulated core, Tyr¹⁵⁸ in the SH2 domain makes a π-stacking interaction with Tyr³⁶¹ of kinase domain helix αE and is also hydrogen bonded to Asn³⁹³ (8, 9) (modeled in Fig. 5A). Substitution of SH2 Tyr¹⁵⁸ with aspartate (Y158D) has been shown to increase Abl kinase activity, presumably by disturbing this interaction (10). To determine whether enhanced SH3/linker interaction influences the activating effect of the Y158D mutation, the HAL9 sequence was combined with this SH2 mutation in the c-Abl core. As shown in Fig. 5B, the Y158D mutation alone strongly enhanced autophosphorylation of the activation loop (Tyr⁴¹²), the SH3 domain (Tyr⁸⁹), as well as tyrosine phosphorylation of cellular proteins. Remarkably, introduction of the HAL9 sequence almost completely reversed kinase activity resulting from this SH2 domain mutation, both at specific regulatory sites and overall protein-tyrosine phosphorylation in cell lysates (Fig. 5C). These data show that the enhanced SH3/linker interaction overcomes the activating effect of disturbing the SH2/kinase domain interaction, and support an allosteric connection between the SH3/linker and SH2/C-lobe (see “Discussion”).

FIGURE 5.

HAL9 substitution suppresses c-Abl core activation by SH2-kinase interface mutation (Y158D). A, model of the c-Abl core highlighting tyrosine residues in the SH2 domain (Tyr¹⁵⁸) and kinase domain C-lobe (Tyr³⁶¹) that interact as part of the down-regulated conformation (boxed and enlarged on the right). Abl domains are colored as per Fig. 1A. B, the wild-type c-Abl core protein (WT), HAL9, the SH2 Tyr¹⁵⁸ to aspartate mutant (Y158D), as well as the compound mutant (HAL9-YD) were expressed in 293T cells. Untransfected cells were included as a negative control. Abl proteins were immunoprecipitated and immunoblotted with phosphospecific antibodies against the activation loop (pY412) and the SH3 domain (pY89) as well as for Abl protein recovery. C, overall protein-tyrosine phosphorylation was assessed in cell lysates by immunoblotting; Abl blots were also performed as a control. This experiment was repeated twice and produced comparable results; a representative example is shown.

High Affinity Linkers Sensitize Bcr-Abl-transformed Cells to Imatinib-induced Apoptosis

In the context of Bcr-Abl, fusion to Bcr removes the N-terminal myristoylation signal and provides an oligomerization motif, resulting in constitutive kinase activation. Although Bcr fusion is sufficient to activate Abl and drive CML, the SH3 and SH2 domains may remain in their regulatory positions on the back of the kinase domain in the context of Bcr-Abl (see Introduction). Results presented so far show that the enhanced SH3/linker interaction suppresses the activity of the c-Abl core, suggesting that the HAL sequence may stabilize the inactive Abl kinase domain conformation and enhance imatinib sensitivity. To test this idea, we transformed human TF-1 myeloid cells with wild-type and HAL9 forms of p210 Bcr-Abl. Each transformed cell population was then treated over a range of imatinib concentrations followed by assays for apoptosis, measured as effector caspase activity. As shown in Fig. 6A, TF-1 cells transformed with wild-type Bcr-Abl showed a dose-dependent increase in apoptosis following imatinib treatment. When this experiment was repeated with cells expressing Bcr-Abl-HAL9, the apparent potency of imatinib was increased, supporting the idea that enhanced SH3/linker interaction stabilizes the kinase domain conformation required for imatinib binding (32, 33). Immunoblots for Bcr-Abl activity were consistent with the apoptosis data, showing enhanced sensitivity to imatinib at the activation loop (Tyr(P)⁴¹²), the SH3 domain (Tyr(P)⁸⁹), and in overall tyrosine phosphorylation of proteins in cell lysates (Fig. 6B). In contrast, HAL9 substitution did not sensitize TF-1 cells transformed with Bcr-Abl T315I to imatinib, even at concentrations as high as 10 μm (data not shown). This is consistent with the steric clash and loss of the Thr³¹⁵ H-bond to imatinib that results from the T315I mutation (34).

FIGURE 6.

HAL9 sensitizes Bcr-Abl-transformed cells to imatinib and GNF-2. Human TF-1 myeloid cells were transduced with Bcr-Abl retroviruses with either wild-type (WT) or high affinity (HAL9) SH2-kinase linkers. Transformed cells were plated in triplicate in the presence of the indicated concentrations of imatinib (A) or GNF-2 (C) for 48 h, and apoptosis (assayed as Caspase3/7 activity) was measured as described under “Experimental Procedures.” Caspase activity is presented in the bar graphs as the average fold-change relative to the untreated controls ± S.D. (n = 3). B and D, TF-1 cells transformed with wild-type or HAL9 Bcr-Abl were treated with the indicated concentrations of imatinib or GNF-2 for 16 h. Cell lysates were immunoblotted with antibodies to the Bcr-Abl protein, phosphospecific antibodies for the Bcr-Abl activation loop (pY412) and SH3 domain (pY89), and for overall levels of protein-tyrosine phosphorylation (pTyr). Immunoblot experiments were repeated twice with comparable results; a representative example is shown.

High Affinity Linkers Sensitize Bcr-Abl-transformed Cells to Allosteric Kinase Inhibitors

Recent studies have identified a novel class of Bcr-Abl inhibitors that interact with the myristic acid binding pocket in the C-lobe of the kinase domain (15, 16). Structural and dynamics studies show that these inhibitors, of which the phenylamino-pyrimidine compound GNF-2 is the prototype, stabilize the active site of the kinase through an allosteric mechanism (16, 17). These observations led us to investigate whether HAL substitution also enhances the sensitivity of Bcr-Abl-transformed TF-1 cells to allosteric inhibitors such as GNF-2. We first evaluated the sensitivity of TF-1 cells transformed with the wild-type and HAL9 forms of Bcr-Abl to GNF-2 treatment (Fig. 6C). In this case, the presence of the high affinity linker dramatically enhanced the apoptotic response to GNF-2, with half-maximal induction of apoptosis observed at about 200 nm. In contrast, cells transformed with wild-type Bcr-Abl were much less sensitive to GNF-2, with an IC₅₀ value greater than 3 μm. Immunoblots for Bcr-Abl kinase activity closely parallel the apoptosis results, with nearly complete inhibition of Bcr-Abl-HAL9 kinase activity with as little as 100 nm GNF-2 (Fig. 6D). These results support an allosteric connection between the SH3 domain and the myristic acid binding pocket, and show that inhibitors targeting this C-lobe binding site are sensitive to the stabilizing influence of the SH3/linker interaction.

Although the binding site for GNF-2 is localized to the C-lobe of the kinase domain, the imatinib-resistant Bcr-Abl mutant T315I is also much less sensitive to this allosteric inhibitor (15). Given the strong sensitizing effect of the HAL substitution on the apoptotic response to GNF-2 in wild-type Bcr-Abl, we repeated these experiments using TF-1 cells transformed with Bcr-Abl T315I bearing either a wild-type linker or the HAL9 substitution. As shown in Fig. 7A, HAL9 substitution enhanced the sensitivity of TF-1 cells transformed with Bcr-Abl T315I to the apoptotic effects of GNF-2, although the impact was not as strong as that observed for wild-type Bcr-Abl. Very similar results were obtained with the second generation analog GNF-5, for which in vivo efficacy has recently been demonstrated (16) (data not shown). GNF-2 also partially inhibited Bcr-Abl T315I-HAL kinase activity by immunoblot analysis, whereas no effect was observed with T315I (Fig. 7B). These results show that the enhanced SH3/linker interaction can sensitize the T315I kinase domain to these allosteric inhibitors in the context of Bcr-Abl.

FIGURE 7.

HAL9 sensitizes myeloid cells transformed with the Bcr-Abl gatekeeper mutant T315I to GNF-2. Human TF-1 myeloid cells were transduced with Bcr-Abl T315I retroviruses with either wild-type or high affinity (HAL9) SH2-kinase linkers. A, transformed cells were plated in triplicate in the presence of the indicated concentrations of GNF-2 for 48 h, and apoptosis (assayed as Caspase3/7 activity) was measured as described under “Experimental Procedures.” Caspase activity is presented in the bar graphs as the average fold-change relative to the untreated controls ± S.D. (n = 3). B, TF-1 cells transformed with wild-type (WT) or HAL9 Bcr-Abl were treated with the indicated concentrations of GNF-2 for 16 h. Cell lysates were immunoblotted with antibodies to the Bcr-Abl protein, phosphospecific antibodies for the Bcr-Abl activation loop (pY412) and SH3 domain (pY89), and for overall levels of protein-tyrosine phosphorylation (pTyr). Immunoblot experiments were repeated twice with comparable results; a representative example is shown.

We also investigated whether high affinity linker substitution affected the transforming activity of Bcr-Abl in the absence of inhibitor treatment. For these experiments, we compared the proliferation rate of the wild-type and T315I forms of Bcr-Abl with and without HAL9 substitutions using the Cell Titer Blue assay. HAL9 substitution did not affect the rate of cytokine-independent proliferation of the Bcr-Abl-transformed cells (data not shown). These results suggest that unlike c-Abl, enhanced SH3/linker interaction alone is not sufficient to inhibit Bcr-Abl kinase and transforming activities (see “Discussion”).

HX MS Supports Allosteric Interplay between the SH3 Domain, the GNF-2 Binding Pocket, and the Active Site of the c-Abl Core

Results presented above demonstrate that enhanced SH3/linker interaction completely suppresses the activity of a c-Abl core domain myristate binding pocket mutant (A356N; Fig. 4A), and dramatically enhances the sensitivity of Bcr-Abl to allosteric inhibitors that bind to this C-lobe site (Figs. 6 and 7). These findings suggest that the SH3/linker interface exerts allosteric control over the regulatory influence of the C-lobe on the active site. To test this idea from a structural perspective, we used HX MS to investigate the effects of the A356N mutation and HAL9 substitution on global conformational changes in recombinant c-Abl kinase core proteins.

The c-Abl core proteins chosen for comparison were wild-type (WT), the A356N mutant, HAL9, and the combined HAL9-A356N protein. All four proteins were expressed in their myristoylated forms using Sf9 cells as the host, and purified to homogeneity. The intact mass of each protein was determined, and showed that the WT and HAL9 proteins were largely unphosphorylated, whereas the A356N mutant was heavily phosphorylated, consistent with the activating effect of this mutation (Fig. 8). Interestingly, HAL9 substitution nearly reversed autophosphorylation of A356N, supporting the dominant allosteric control of the enhanced SH3/linker interaction on overall kinase activity in vitro.

FIGURE 8.

Mass analysis of recombinant c-Abl core proteins. Wild-type c-Abl core (WT), A356N, HAL9, and A356N-HAL9 proteins were overexpressed in Sf9 insect cells together with the Yersinia phosphatase, YopH, and purified as described under “Experimental Procedures.” Intact protein mass spectral analysis was performed by injecting each protein onto a POROS 20 R2 protein trap and desalting with 0.05% TFA at a flow rate of 100 μl/min. The proteins were eluted into the mass spectrometer using a linear 15–75% acetonitrile gradient over 4 min at 50 μl/min and a Shimadzu HPLC system (LC-10ADvp). Mass analyses were performed on an LCT-Premier instrument (Waters) equipped with a standard electrospray source. The capillary voltage was 3.2 kV and the cone voltage was 35 V. Nitrogen was used as desolvation gas. A source temperature of 175 °C and a desolvation temperature of 80 °C were applied. The instrument was calibrated by infusing a solution of 500 fmol/μl of myoglobin and the mass accuracy was 10 ppm. The raw m/z data are shown on the left and the transformed, mass-only spectra are shown on the right. The measured and theoretical molecular masses are indicated, and the additional +80 Da peaks correspond to the tyrosine-phosphorylated species. All masses are consistent with stoichiometrically myristoylated proteins.

For HX MS experiments, each recombinant Abl core protein was diluted in D₂O buffer, aliquots were removed at various time points, and the exchange reaction was rapidly quenched. To refine the location of deuterium incorporation, each protein was digested with pepsin prior to mass spectrometry. The peptic peptide coverage maps are shown in supplemental Fig. S1. Deuterium uptake curves were then generated for each peptic peptide and analyzed pairwise for the WT versus A356N mutant, WT versus HAL9, and A356N versus A356N-HAL9. All of the uptake curves are presented in supplemental Fig. S2. Overall, the rate of deuterium incorporation, and therefore protein dynamics, was the same for most of the peptides across all three comparisons during the time frame of the experiment (4 h). However, several important differences were observed that provide remarkable insight into the impact of the opposing effects of the A356N and HAL9 substitutions on c-Abl core dynamics.

Comparison of exchange in peptic peptides derived from WT versus A356N Abl proteins revealed nine peptides with more rapid deuterium uptake in the mutant, supporting the idea that disruption of the myristate binding pocket results in a more open, mobile structure. Most of the increases in deuterium incorporation observed in the A356N mutant localize to the N-lobe of the kinase domain. The N-lobe peptide corresponding to Abl residues 325–336 not only showed more deuterium uptake in the A356N mutant, but the mass spectra indicate the presence of two populations of peptides, one of which is labeled more rapidly than the other (Fig. 9A). The half-life of conversion between the two populations in the mutant started around 10 min, whereas the conversion took much longer for the WT protein. These results indicate that the mutant was much more dynamic in this region, displaying dynamics of the EX1 type (35). Analysis of exchange into four overlapping peptides in this region shows that just four amino acids (Ile-Ile-Thr-Glu) are responsible for the altered dynamics of the A356N mutant (Fig. 9, B and C). These residues are located in the tip of the two-strand β-hairpin pointing into the ATP-binding site, and include the gatekeeper threonine, Thr³¹⁵ (gatekeeper designation is based Abl-1a numbering (34); all other peptide numbering is based on the crystal structure of the Abl-1b core; Protein Data Bank code 2FO0 (8)). These results support a long-range allosteric connection between the myristic acid binding site in the C-lobe and part of the active site derived from the N-lobe, consistent with previous results (17). In addition to peptides derived from the N-lobe, four peptides encompassing C-lobe region 471–490 also displayed more deuterium uptake in A356N compared with WT (supplemental Fig. S2). This region is adjacent to the site of the A356N substitution and indicates that this activating mutation also causes a local increase in solvent accessibility. Peptide 475–489 from this region is highlighted on the c-Abl core structure in Fig. 10.

FIGURE 9.

A kinase domain N-lobe peptide encompassing the gatekeeper residue (Thr³¹⁵) is sensitive to A356N mutation in the myristate binding pocket. A, mass spectra of deuterium incorporation into the c-Abl N-lobe peptide 325–336 were recorded over the time intervals shown. The start of the unfolding in this peptide from the A356N mutant is readily observed after 10 min, but is not observed in the wild-type peptide over this time frame. B and C, comparison of the deuterium uptake curves corresponding to four overlapping peptides across the 324–336 region identifies four amino acids (Ile-Ile-Thr-Glu) including the gatekeeper position as responsible for the dynamic and conformational changes in the A356N mutant (IITE). The position of this peptide in the crystal structure of the c-Abl core (Protein Data Bank code 2FO0) is shown at the lower right; the locations of the mutation (A356N) and the myristate group are highlighted.

FIGURE 10.

High-affinity linker substitution reduces hydrogen exchange in the c-Abl myristic acid binding pocket, SH3 domain, and kinase domain N-lobe. The locations of the peptides that exhibited altered hydrogen exchange kinetics are modeled on the structure of the myristoylated, down-regulated c-Abl core (Protein Data Bank code 2FO0). Changes are indicated in pairwise fashion for the wild-type c-Abl core (WT) versus the A356N myristic acid binding pocket mutant (left), for WT versus HAL9 (middle), and for A356N versus HAL9-A356N (right). Differences in deuterium uptake for WT versus HAL9 and A356N versus HAL9-A356N are color-coded with blue indicating peptides with a significant reduction in uptake between 0.7 and 1.0 Da and yellow indicating peptides that displayed subtle changes (less than 0.7 Da difference between the uptake curves). Magenta indicates peptides with more deuterium uptake (0.7–1.0 Da) in the A356N sample when compared with WT. The cumulative error of measuring deuterium uptake in these assays is ±0.20 Da. Proline residues that were introduced in HAL9 and HAL9-A356N are colored orange and rendered as sticks. The A356N mutation is represented in blue sticks, and the position of the myristate group is shown in green. All deuterium uptake curves are presented in supplemental Fig. S2.

Comparison of exchange between the WT and HAL9 forms of the c-Abl core revealed several regions with less deuterium uptake in HAL9, strengthening the idea that the enhanced SH3 interaction with the linker stabilizes the down-regulated conformation of the kinase domain (Fig. 10). The regions that showed reduced deuterium uptake in HAL9 included peptide 100–107, which includes the N-terminal portion of the RT loop of the SH3 domain, SH2 domain peptide 158–168, which is part of the α-helix that contacts the kinase domain C-lobe, and peptide 325–331 from the N-lobe of the kinase domain. The latter peptide overlaps with the same one that is more mobile in the A356N mutant as described above. Reduction of deuterium incorporation in these HAL9 regions supports the hypothesis that increasing the SH3/linker interaction results in global stabilization of the inactive Abl core conformation.

A final comparison of HX MS data from the active A356N mutant with HAL9-A356N showed that the HAL9 substitution completely reverses the dynamic changes induced by A356N (Fig. 10). Overall, the HAL9-A356N protein had a very similar HX MS profile to that of HAL9 alone, with reduced deuterium uptake in SH3 domain peptides 100–107 and 119–133 as well as SH2 domain peptide 158–168, which contacts the kinase domain C-lobe. This SH2 peptide contains Tyr¹⁵⁸, which when mutated to aspartate increases Abl kinase activity (Fig. 5). As discussed above, HAL9 substitution completely reversed kinase activation resulting from this SH2 domain mutation, consistent with global control of the Abl kinase core conformation by SH3 interaction with the linker.

DISCUSSION

The discovery that the tyrosine kinase activity of Bcr-Abl is the driving force behind CML led to the development of imatinib, a type II ATP-competitive inhibitor of Abl, as the first-line treatment for this rare form of leukemia (3). Imatinib revolutionized CML therapy and laid the foundation for targeting kinase activity as a therapeutic approach to other forms of cancer. However, despite the clinical success of imatinib, long term therapy often leads to the emergence of drug resistance and disease relapse (4). Imatinib resistance results from kinase domain mutations that impact drug binding (e.g. gatekeeper T315I (34)) or from mutations outside of the kinase domain, which induce an active site conformation incompatible with drug binding (6). The second generation ATP-competitive Abl inhibitors nilotinib and dasatinib have been approved for therapy of imatinib-resistant CML (7). However, these drugs are also associated with resistance and do not overcome the T315I imatinib-resistance mutation.

The need to overcome the gatekeeper mutation and to target Abl kinase activity more specifically led to the discovery of allosteric kinase inhibitors. GNF-2, the prototype of this inhibitor family, targets the myristate binding pocket in the C-lobe to stabilize the inactive conformation of the kinase domain (15–17). GNF-2 inhibits some imatinib-resistant forms of Bcr-Abl, although the T315I gatekeeper mutant is a notable exception. This observation suggests that the T315I mutation uncouples the active site from allosteric control by the myristate binding pocket. However, when combined with nilotinib or dasatinib, GNF-2 and related compounds overcome T315I mutations both in vitro and in vivo (16). Combinations of GNF-2-type allosteric inhibitors and ATP competitive drugs also dramatically decrease the rate of experimental drug resistance, strongly suggesting that other therapeutic approaches targeting natural mechanisms of kinase regulation may be of value in combating drug resistance. In this study, we focused on the potential of the SH3/linker interaction as an alternative allosteric regulator of kinase activity. Our approach was inspired by growing evidence that the Abl SH3 and SH2 domains maintain regulatory control over the Abl kinase domain even in the context of the active Bcr-Abl kinase. Indeed, imatinib resistance can arise from mutations in these regulatory domains as well as phosphorylation of the SH3 domain by Src-family kinases (6, 19).

Previous HX MS studies have shown that the Abl linker remains bound to the SH3 domain even in the absence of the kinase domain, supporting a prominent role for this interaction in Abl kinase regulation (20). Our goal was to enhance SH3 interaction through systematic substitution of linker amino acid residues with proline without disturbing the overall structure of the down-regulated c-Abl core. Using HX MS, we first showed enhanced SH3:linker engagement as a function of increased linker proline content in small proteins consisting of the SH3, SH2, and linker regions. Within the context of the larger Abl core protein, most of the HAL modifications maintained or reduced phosphorylation of key regulatory tyrosines relative to wild-type protein. Among these, the HAL9 protein, with five additional linker prolines, showed the greatest enhancement of SH3/linker interaction without disturbing overall kinase regulation.

The long-range allosteric influence of the SH3/linker interaction on Abl kinase regulation became readily apparent when HAL9 was combined with mutations known to activate the c-Abl core. Particularly striking was the observation that HAL9 substitution completely reversed the activating effect of the A356N mutation in the myristic acid binding pocket. In contrast, HAL9 only partially reversed the activity of a nonmyristoylated (G2A) Abl core protein, suggesting that G2A and A356N produce distinct active conformations of the c-Abl core despite their functional relationship. In the case of the nonmyristoylated Abl-G2A mutant, small-angle x-ray scattering supports a dramatic reorientation of the SH2 domain onto the N-lobe of the kinase (8), where it stabilizes an active kinase domain conformation (36). This active “top hat” conformation is also supported by previous HX MS studies of nonmyristoylated Abl core proteins (21). On the other hand, HX MS of the Abl-A356N core protein presented here revealed no changes in deuterium incorporation at the SH2:C-lobe interface, the Ncap or the SH2-kinase linker, indicating that Abl-A356N does not adopt the top hat conformation. These differences in active conformations may help to explain why HAL9 completely reverses the activity of the A356N mutation but is less effective in the context of the G2A mutant. In the case of A356N, the overall structure of the down-regulated c-Abl core is likely to be maintained, with the SH3 domain bound to the linker and the myristoylated Ncap bound to its C-lobe binding pocket. Enhanced SH3/linker interaction via HAL9 substitution may compensate for the allosteric uncoupling that results from the A356N mutation. Indeed, HX MS studies showed that introduction of HAL9 restabilized the C-lobe peptide in the Abl-A356N protein adjacent to the Myr-binding pocket (475–489; Fig. 10), and decreased deuterium uptake in the N-lobe adjacent to the active site as well as the SH3 and SH2 domains. These HX MS data also support an allosteric connection between the SH3 domain, the myristic acid binding pocket, and the active site.

In addition to HAL9, we also observed a partial suppressive effect of HAL4 on Abl-A356N activity. Like HAL9, HAL4 shows enhanced SH3/linker interaction by HX MS (Fig. 1C), whereas maintaining proper down-regulation in the context of the wild-type c-Abl core (Fig. 2). These observations provide additional support for the idea that enhanced SH3/linker interaction can allosterically suppress kinase activation, provided that proper down-regulation can be achieved. The partial effect of HAL4 fits well with its lower impact on SH3 unfolding relative to HAL9.

The allosteric impact of HAL9 was observed not only with the Abl kinase core protein but also within the context of full-length Bcr-Abl. Using human TF-1 myeloid cells transformed with Bcr-Abl as a model system, we found that HAL9 substitution enhanced sensitivity to imatinib both in terms of Bcr-Abl kinase activity and apoptosis. These data provide direct evidence that the SH3/linker interaction maintains control over Bcr-Abl kinase activity. The enhancement in the apparent potency of imatinib against the HAL9 form of Bcr-Abl most likely results from restructuring of the kinase domain in the down-regulated conformation required for drug binding (32, 33).

In addition to imatinib, enhanced SH3/linker interaction dramatically sensitized Bcr-Abl-transformed TF-1 cells to the allosteric inhibitor GNF-2 both in terms of kinase inhibition and the apoptotic response. These results support a three-way allosteric connection between the SH3 domain, the myristic acid binding pocket, and the active site in Bcr-Abl. This point is underscored by the observation that HAL9 substitution also sensitized the T315I mutant of Bcr-Abl to GNF-2.

Although HAL9 substitution enhanced Bcr-Abl inhibitor sensitivity, it did not affect the rate of transformed cell proliferation, suggesting that enhanced SH3/linker interaction alone cannot suppress Bcr-Abl kinase and transforming activities. This observation may relate to the clustering and constitutive activation of Abl kinase domains that result from Bcr fusion, which cannot be overcome by SH3/linker engagement. This result is also consistent with the inability of the HAL9 substitution to completely overcome the activating effect of the “G2A” mutant of the Abl core, which, like Bcr-Abl, cannot be myristoylated. Nevertheless, our data strongly suggest that enhanced SH3/linker engagement can push the Bcr-Abl kinase domain toward a conformation more susceptible to imatinib and especially GNF-2 action.

One notable difference between our data and previous work relates to the sensitivity of Bcr-Abl-transformed TF-1 cells to GNF-2. The original study describing the discovery of GNF-2 reported IC₅₀ values for inhibition of cell proliferation in the triple-digit nanomolar range (15), whereas our data is in the micromolar range for Bcr-Abl with a wild-type linker. Previous studies of GNF-2 were performed with the IL-3-dependent cell line BaF3, which was transformed to cytokine independence by Bcr-Abl. BaF3 is a pro-B cell line of mouse origin, whereas TF-1 is a myeloid cell line of human origin (37). TF-1 cells require GM-CSF-dependent for growth, but can be readily transformed with Bcr-Abl to a cytokine-independent phenotype (13, 22). These species and lineage differences most likely account for the differential sensitivity of the two cell lines to GNF-2. Nevertheless, this difference does not affect the interpretation of the result that enhanced SH3/linker interaction strongly enhances Bcr-Abl sensitivity to GNF-2.

In summary, our data show that the SH3/linker interface is a key node controlling the overall dynamics and regulation of the c-Abl kinase core, and that this regulatory influence is retained in the context of the Bcr-Abl oncoprotein. Furthermore, the combination of HX MS, biochemical, and biological studies used here strongly support allosteric interplay between the SH3/linker interface, the myristic acid binding pocket in the C-lobe, and the active site. Enhanced SH3/linker interaction sensitizes Bcr-Abl to both conformationally sensitive ATP competitive inhibitors such as imatinib as well as compounds that bind to the myristic acid binding pocket in the C-lobe. Our work strongly supports future drug discovery campaigns to identify small molecules that stabilize SH3/linker interaction as allosteric inhibitors of both wild-type and drug-resistant forms of Bcr-Abl, and in other kinase systems where SH3 domains play a role in regulation such as Src-family kinases (38).