In Vitro Evolution Reveals a Single Mutation as Sole Source of Src-family Kinase C-helix-out Inhibitor Resistance

Abstract

Understanding cancer cell drug resistance to protein-tyrosine kinase inhibitors, which often arises from acquired mutations in the target kinase, is central to the development of more durable therapies. Experimental systems that reveal potential paths to resistance for a given inhibitor and kinase target have an important role in preclinical development of kinase inhibitor drugs. Here we employed a codon mutagenesis strategy to define the mutational landscape of acquired resistance in HCK, a member of the SRC tyrosine kinase family and therapeutic target in acute myeloid leukemia (AML). Using PCR-based saturation mutagenesis, we created a cDNA library designed to replace each codon in the HCK open reading frame with all possible codons. This HCK mutant library was used to transform Rat-2 fibroblasts, followed by selection for resistant colonies with A-419259, a pyrrolopyrimidine HCK inhibitor and drug lead for AML. X-ray crystallography has shown that A-419259 binding induces outward rotation of the kinase domain αC-helix, a conformation incompatible with phosphotransfer. Remarkably, only a single resistance mutation evolved during A-419259 selection: histidine substitution for threonine at the gatekeeper position in the kinase domain. Deep sequencing confirmed representation of nearly all other missense mutations across the entire HCK open reading frame. This observation suggests that A-419259 and other C-helix-out Src-family kinase inhibitors may have a narrow path to acquired resistance in the context of AML cases where Hck is an oncogenic driver.

Graphical Abstract

Introduction

SRC-family kinases have been implicated in many forms of cancer, including myeloid leukemias and other hematologic malignancies. In chronic myeloid leukemia (CML), the myeloid SRC-family member HCK cooperates with BCR-ABL in signaling pathways related to CML cell survival and also influences sensitivity to imatinib and other tyrosine kinase inhibitors.^1–3 More recently, HCK has also attracted attention as a therapeutic target in the context of acute myeloid leukemia (AML). Comparison of gene expression signatures in normal hematopoietic stem cells vs. leukemic stem cells from AML patients identified HCK as a leukemia-specific transcript.⁴ Selective knockdown of HCK expression with siRNAs resulted in decreased cell proliferation and increased apoptosis in primary AML cells.⁵ A recent analysis of SRC-family kinase expression in AML identified HCK as one of the most highly expressed SRC-family members and showed that HCK expression levels correlate with worse outcomes in terms of AML patient survival.⁶ These data support the development of HCK inhibitors as precision therapy for AML patients that express high levels of this kinase.

One promising inhibitor candidate for HCK is the pyrrolopyrimidine A-419259, which was originally developed as part of a medicinal chemistry campaign to identify selective, orally active inhibitors of the closely related T cell kinase, LCK.⁷ Initial studies showed that A-419259 induced growth arrest and apoptosis of Philadelphia chromosome-positive CML cells at concentrations that did not directly inhibit BCR-ABL.⁸ An HCK gatekeeper mutant with engineered resistance to A-419259 reversed its anti-CML effects, demonstrating that HCK is a primary target for A-419259 in CML.² More recently, A-419259 was rediscovered in a high-throughput screening campaign for HCK inhibitors and shown to have potent activity against in mice bearing AML patient-derived xenografts following oral administration.⁵ As observed for CML, gatekeeper mutants of HCK with engineered resistance to A-419259 rescued its anti-leukemic activity in FLT3-ITD⁺ AML cell lines, supporting HCK as a target for this compound in AML.⁶

A-419259 inhibits HCK by binding to a pocket in the ATP-site of the kinase domain (Figure 1A). X-ray crystallography of near-full-length HCK bound to A-419259⁹ shows contacts between the pyrrolopyrimidine inhibitor core and the side chain of the gatekeeper residue (Thr338) as well as the main chains of hinge residues Glu339 and Met341 (all residue numbering based on a crystal structure of SRC¹⁰). In addition, the methylpiperizine moiety of A-419259 makes a polar contact with the side chain of Asp348, which extends from an α-helix in the C-lobe adjacent to the catalytic cleft. KINOMEscan analysis shows that A-419249 exhibits a remarkably narrow specificity profile, with the most likely targets restricted to HCK, other SRC-family members, as well as FLT3 and other Class III receptor tyrosine kinases relevant to AML.⁶ This specificity profile makes A-419259 a promising candidate for further development against AML cases that over-express of HCK and other myeloid SRC-family kinases, as well as mutants of FLT3.⁶

Figure 1. X-ray crystal structure of downregulated HCK bound to A-419259.

A) The overall structure of HCK is shown on the left, with the SH3 and SH2 domains packed against the back of the kinase domain. In this inactive conformation, the SH3 domain engages the SH2-kinase linker, while the SH2 domain binds to the tail when phosphorylated on Tyr527 (pTyr⁵²⁷). A close-up view of the kinase domain active site is shown at right, with the carbon skeleton of A-412959 modeled in cyan. The 4-amino group of the pyrrolopyrimdine makes a hydrogen bond with the gatekeeper residue (Thr338), and together with a ring nitrogen contacts main chain residues Glu339 and Met341 in the hinge region. One of the A-419259 methylpiperizine nitrogens also contacts the side chain of Asp348 in the C-lobe of the kinase domain. The overall domain organization of Hck is shown at the bottom, illustrating the N-terminal unique domain (U; not present in the crystal structure) and its post-translational modifications with myristate (Myr) and palmitate (Palm). Also shown are the relative positions of the gatekeeper residue (Thr338) as well as the autophosphorylation site in the activation loop (pTyr416; not resolved in this crystal structure). B) A-419259 binding induces an outward rotation of the αC-helix (αC_out) such that Glu310 cannot ion-pair with Lys295, an interaction essential for kinase activity. In addition, the Glu-Phe-Gly motif is rotated inward (DFG_in), with Asp404 contacting Lys295. A-419259 contacts with HCK as per part A. Models created with PyMol (Schrödinger, Inc.) and PDB files 1QCF (PP1) and 4LUE (A-419259).

Development of precision therapies targeting tyrosine kinases in myeloid leukemias has exploded over the past two decades. This effort was spawned by the success of imatinib, a relatively selective inhibitor of the BCR-ABL tyrosine kinase associated with CML.¹¹ Imatinib owes its specificity in part to its ability to trap a unique inactive conformation of the ABL kinase domain, in which the conserved Asp-Phe-Gly (DFG) motif is rotated outward while the A; helix is rotated inward.^{12, 13} This binding mode is sensitive to the allosteric effects of mutations that impact the conformation of the drug binding pocket. Random mutagenesis studies with BCR-ABL followed by imatinib selection in vitro revealed a diverse array of mutations that can produce inhibitor resistance, including many outside of the inhibitor binding site.¹⁴ Some of these mutations have also been observed in the clinic,¹⁵ illustrating the power of this approach in the identification of potential pathways to resistance during drug development. A similar approach has also been used to find clinically relevant resistance mutations against the Flt3-ITD inhibitors quizartinib¹⁶ and crenolanib,¹⁷ which also show promise for FLT3-ITD⁺ AML.

Further assessment of the X-ray crystal structure of the HCK•A-419259 complex also reveals a conformation-dependent binding mode, albeit one quite distinct from that observed with Bcr-Abl and imatinib. In this case, the DFG motif is rotated inward, with Asp404 from this motif forming a hydrogen bond with Lys295 (Figure 1B). The αC helix is rotated outward, moving Glu310 away from the ATP binding site such that it is unable to pair with Lys295, an interaction essential for HCK activity.¹⁸ Recent work has shown that other SRC-family kinase inhibitors displaying a similar ‘C-helix-out’ mechanism induce allosteric changes in the overall structure of the kinase, enhancing negative regulatory interactions of the SH2 and SH3 domains with the kinase domain.¹⁹ These observations are consistent with the overall crystal structure of the HCK•A-419259 complex, in which the SH3-SH2 unit is assembled on the back of the kinase domain (Figure 1A). Taken together, these observations suggest that A-419259 inhibitory activity may be affected by mutations that destabilize the assembled form of the kinase, providing alternative pathways to clinical resistance.

Given the therapeutic potential of A-419259 in AML as described above, herein we report a forward genetic screen designed to identify all possible HCK mutations that give rise to resistance. Using PCR-based codon mutagenesis, we generated an HCK cDNA library in which all possible codons are represented at each position across the entire HCK open reading frame, except for some essential to function. The library was used to transform Rat-2 fibroblasts, followed by selection of cell colonies resistant to A-419259 and nucleotide sequencing of HCK from these resistant clones. Despite the great diversity of this library as confirmed by deep sequencing, the only resistance mutation identified mapped to the gatekeeper residue in the inhibitor binding pocket (substitution of Thr338 with histidine). This unexpected finding suggests that A-419259, and possibly other SRC-family kinase inhibitors that induce a C-helix-out conformation, may be less susceptible to allosteric mutations that drive inhibitor resistance.

RESULTS AND DISCUSSION

A forward genetic screen to identify potential A-419259 resistance mutations in HCK.

HCK is one of eight mammalian SRC-family members that share a similar architecture consisting of four domains (Figure 1A). These include an N-terminal unique domain, which is myristoylated and palmitoylated for membrane anchoring, followed by the SH3, SH2, and kinase domains as well as a C-terminal tail with a conserved tyrosine (Tyr527) essential for kinase regulation. Phosphorylation of Tyr527 by the regulatory kinase CSK induces intramolecular engagement of the SH2 domain.²⁰ Together with intramolecular interaction of the SH3 domain with the SH2-kinase linker, SH2-tail interaction holds HCK and other SRC-family kinases in an inactive, assembled conformation. Mutation of HCK Tyr527 to phenylalanine (HCK-YF mutant) prevents tail phosphorylation, leading to constitutive kinase activity and oncogenic transformation following ectopic expression in Rat-2 fibroblasts.^{18, 21, 22} This Rat-2 transformation assay, driven by Hck-YF, forms the basis of the codon mutagenesis screen for A-419259 resistance described below.

To generate a fully diverse library of HCK mutants, we used a PCR-based saturation mutagenesis procedure covering the entire open reading frame of the p59 form of human HCK. This method involves a large set of overlapping PCR primers, each with a central degenerate codon flanked by adjacent coding sequences. The primers are then combined, and subsequent annealing and low-cycle PCR results in a single product in which every possible codon is represented theoretically at each position along the sequence. In the case of HCK, we excluded codons for residues critical to HCK activity and subcellular localization, while substituting Tyr527 with phenylalanine to ensure transformation in Rat-2 fibroblasts (see Materials and Methods). In total, 488 of the 505 HCK codons underwent mutagenesis; with 64 codons possible at each position, the maximum theoretical diversity of the HCK mutagenesis library is 488 codons × 64 possible codons = 31,232 individual clones. The final PCR product was subjected to large-scale ligation into a retroviral vector for gene transfer. The resulting ligation reaction was used to transform E. coli, and subsequent plating produced approximately 100,000 individual bacterial colonies or roughly 3-fold coverage of the calculated number of unique mutants.

To assess the diversity of the HCK codon mutagenesis library, twenty individual clones were picked at random and analyzed by Sanger sequencing in their entirety. Nucleotide changes were observed in all three positions within each mutated codon, with a slight bias towards single nucleotide substitutions when compared to the predicted outcome (Figure 2A). This bias may reflect the enhanced PCR efficiency of primers with single nucleotide substitutions. Next, we compared the number of mutations in each clone to the expected Poisson distribution (Figure 2B). The expected distribution was observed for clones with zero or one mutation, while the fraction observed for 2 mutations per clone was somewhat lower than expected while that for 3 mutations was somewhat higher. This result likely reflects the relatively small sample size. On average, we observed about two mutations per clone, with 72% of the clones exhibiting at least one mutation. We also examined the cumulative distribution of observed mutations across the HCK-YF coding sequence (Figure 2C). Perfectly distributed mutations would yield a cumulative distribution plot that follows a diagonal line. However, PCR-based codon mutagenesis tends to bias mutations at the beginning and end of the gene, an effect that is more pronounced as a function of target sequence length. This effect is observed in our cumulative distribution plot, although mutations are present across the entire length of the coding sequence as desired.

Figure 2. Frequency and distribution of mutations in the HCK-YF codon mutagenesis library.

Sanger sequence analysis was performed on 20 random clones from the codon mutagenesis library. A) Number of nucleotide substitutions in each codon position found across the 20 clones. B) Distribution of the number of mutations per clone. C) Cumulative distribution of mutations across the entire Hck-YF coding sequence. In each panel, the predicted outcomes are represented in black, while the observed result is plotted in gray.

To evaluate codon substitution in a more comprehensive fashion, we performed deep sequencing of the HCK mutant library. The frequency of codons at each amino acid position (as well as stop codons) were calculated for each of the 488 coding positions that were targeted across the HCK cDNA. Individual examples of substitution frequencies are presented in Figure 3A. The first example is Thr338, which is the gatekeeper residue in the HCK kinase domain; the side chain of Thr338 makes an essential hydrogen bond with A-419259 in the X-ray crystal structure (Figure 1). Here substitutions were detected that correspond to replacement of Thr338 with every other amino acid (except for isoleucine) along with nonsense mutations (stop codons). The proportional representation of each amino acid reflects the redundancy of the genetic code. For example, leucine, which is represented by six codons, was the most highly represented substitution at this position, while methionine and tryptophan, which are encoded by single codons, were the represented at the lowest level. A second example is presented for Asp348, the side chain of which makes a hydrogen bond with the methylpiperizine ring of A-419259 (Figure 1). In this case, arginine is most highly represented, consistent with its six codons in the genetic code. In contrast, the single codon for tryptophan is represented at the lowest frequency while the codon for methionine was not observed. A color-coded plot of the deep sequencing result for every mutated codon shows remarkably similar substitution rates across the entire coding sequence (Figure 3B). An expanded version of this plot is provided in the Supporting Information as Figure S1.

Figure 3. Deep sequencing reveals a wide and nearly complete distribution of codon substitutions in the Hck-YF library.

The HCK-YF library inserts in the retroviral vector were amplified by PCR, and subamplicons were then PCR-amplified using primers to add barcode sequences and Illumina adapter fragments as described under Materials and Methods. Sequencing was performed on the Illumina MiSeq platform and generated 300 base-pair paired-end reads. A) Proportion of substitutions present at individual codons 338 (left) and 348 (right). Amino acids are represented as the single letter code, with X = nonsense mutations (stop codons). The relative representation of each amino acid is represented by each different colored bar. B) Proportion of amino acids present at each codon position across the entire Hck-YF coding region. Color scheme as in part A. Gaps represent positions of essential residues that were not targeted for mutation. A higher resolution version of this plot is presented in the Supporting Information (Figure S1).

Transformation of Rat-2 cells with the HCK-YF mutant library and selection of A-419259-resistant clones.

Recombinant retroviruses carrying the HCK-YF codon mutagenesis library were produced in 293T cells and used to infect Rat-2 fibroblasts. Following G-418 selection, the transduced Rat-2 cell population was plated in soft-agar colony-forming assays in the presence of A-419259. Rat-2 cells expressing HCK-YF grow in an anchorage-independent fashion that is dependent on HCK kinase activity, forming tight colonies amenable to subsequent isolation and subculture. To ensure complete coverage of the library, 40 plates were prepared with 5,000 cells per plate, for a total of 200,000 individual cells that underwent selection which represents more than six times the calculated number of individual mutant clones present. Selection was performed at a final concentration of 1 μM A-419259, which completely blocked colony formation by control cells expressing HCK-YF (Figure 4). Following two weeks of selection, 13 colonies appeared over 5 of the selection plates. These colonies were picked and expanded in the absence of agarose and inhibitor under regular culture conditions. Eight of the thirteen original colonies regrew and were subsequently retested for colony formation in the presence of A-419259. Of these, only a single clone retained the ability to form equal numbers of colonies in the presence or absence of A-419259. This verified resistant clone (5–1) was analyzed in detail as described below.

Figure 4. Isolation of A-419259-resistant Rat-2 fibroblasts transformed with the HCK-YF mutant library.

Rat-2 cells expressing the HCK-YF codon mutagenesis library were plated in colony-forming assays in semi-solid medium in the presence of A-419259 (1.0 μM). Resistant colonies were picked and expanded in 2D culture, followed by re-assay for colony-forming activity in the presence of A-419259 or the DMSO carrier solvent as control. A) Of eight colonies assayed, only one (colony 5–1) produced the same number of colonies in the presence and absence of A-419259. Each cell population was assayed in triplicate, with the mean colony numbers represented by the bars ± SE. B) Images of representative 60 mm soft-agar culture plates from part A. Colonies were visualized with Wright-Giemsa stain prior to imaging. A-419259 (1 μM) completely suppressed colony formation by control cells expressing HCK-YF (left). Colonies 3-1 and 5-1 both formed colonies in the absence of A-419259, but only 5-1 formed equal numbers of colonies in the presence and absence of the inhibitor.

Genomic DNA was isolated from resistant clone 5–1, and the integrated HCK-YF coding sequence was amplified by PCR and analyzed by Sanger sequencing. Missense mutations were observed at the codons for Pro32 in the unique domain, Asp158 in the SH2 domain, and Met302, Ile334, and Thr338 in the kinase domain (Table 1). To determine whether these mutations were linked in individual clones, the full-length HCK-YF PCR product from colony 5–1 was subcloned into a plasmid. Following transformation, individual bacterial colonies were picked, and plasmid DNA isolated for nucleotide sequencing. This analysis revealed two variants of the HCK-YF coding sequence within the inhibitor-resistant Rat-2 cell population. The first clone contained three mutations, P32H-D158I-T338H, while the second encoded a stop codon at position 34, along with the two other missense mutations (M302W and I334S).

Table 1. Mutations associated with A-419259 resistance in Rat-2 cells transformed with the HCK-YF codon mutagenesis library.

Nucleotide sequence analysis was performed on human HCK present in A-419259-resistant colony 5-1 (see Figure 4 and main text). Six codons were modified in two independent HCK clones isolated from these cells, resulting in the amino acid changes shown. Note that the nucleotide and amino acid numbering for the HCK unique domain mutation (P32 position) is based on the human p59 HCK coding sequence due to lack of homology with SRC in this region; all other numbering is based on homology to SRC as per convention.

Mutation	Amino Acid Change	Domain
C95G to AT	P32H	Unique
G394AC to ATA	D158I	SH2
A826T to TG	M302W	Kinase
A922TC to TCC	I334S	Kinase
A934CG to CAC	T338H	Kinase

The HCK gatekeeper mutant T338H alone confers resistance to A-419259.

Analysis of the HCK-YF coding sequence from A-419259-resistant clone 5–1 identified seven missense mutations that were distributed across two independent clones. The M302W and I334S mutants were linked to a stop codon upstream, making this mutant very unlikely to be expressed or to contribute to the resistant phenotype. Nevertheless, each of the five missense mutations observed was reintroduced individually into a wild-type HCK cDNA clone for subsequent analysis.

To evaluate the effect of the mutations on HCK protein stability, we first expressed each single mutant in 293T cells and blotted for HCK protein expression relative to actin as a control (Supporting Information, Figure S2). This analysis revealed that several of the mutations resulted in reduced levels of full-length HCK relative to the wild-type kinase, including P32H, and D158I. The HCK-I334S mutant protein underwent a significant shift in mobility on the western blot, consistent with proteolytic cleavage. Cultures expressing HCK were also treated with A-419259, which enhanced the levels of the P32H, D158I mutants relative to the untreated cells. This observation suggests that the presence of the inhibitor may alter the conformation of HCK in such a way as to make the protein less susceptible to proteolytic degradation. Indeed, some ATP-site inhibitors have long-range effects on overall Src-family kinase conformations, with C-helix-out inhibitors stabilizing the overall assembled state.¹⁹ In contrast, the M302W and T338H mutants were at least as stable as wild-type HCK, and the presence of A-419259 did not alter their protein expression levels. We also compared the stability of the T338H mutant to that of a previously characterized gatekeeper mutant, T338M⁹. Expression of HCK T338M was reduced in comparison to T338H, and the presence of A-419259 appeared to accentuate this effect. These differences in HCK mutant protein stability may influence the overall sensitivity of cells to A-419259 as described in the next section.

To determine the effect each mutation on HCK sensitivity to A-419259, transfected 293T cells were treated in the presence or absence of A-419259 at final concentrations of 100 and 1,000 nM. HCK proteins were immunoprecipitated followed by immunoblotting for activation loop autophosphorylation as a measure of kinase activity (pY416 antibody) as well as HCK protein recovery (Figure 5). Of the five mutants tested, only the gatekeeper mutant (T338H) was resistant to A-419259, with no change in activation loop phosphorylation at either inhibitor concentration. In comparison, the HCK-T338M gatekeeper mutant showed complete resistance to A-419259 at 100 nM but only partial resistance at 1,000 nM. In addition, the HCK-T338M protein is less stable in the presence of the higher concentration of the inhibitor, while HCK-T338H remains stable as described above. Taken together, this analysis suggests that the T338H gatekeeper mutant alone is sufficient to generate A-419259 resistance without affecting protein stability, which may explain the selection of this mutant from the codon mutagenesis screen.

Figure 5. HCK-T338H is resistant to A-419259.

Human 293T cells were transfected with wild-type HCK or the six mutants associated with acquired A-419259 resistance from the Rat-2 cell codon mutagenesis screen (P32H, D158I, M302W, I334S and T338H). An engineered gatekeeper mutant of HCK (T338M) was also included for comparison. Cells were then treated with A-419259 at the concentrations shown (0, 100, 1000 nM), followed by immunoprecipitation of HCK and immunoblotting to assess activation loop phosphorylation (pY416) and HCK protein recovery. Immunoreactive proteins were visualized using secondary antibodies conjugated to infrared dyes and imaged using the LI-COR Odyssey system. Bar graphs below each set of representative images show the mean ratios of the signal intensities for pY416 to HCK protein from two independent experiments, normalized to the DMSO (no inhibitor) control ± SE. Full-length HCK-reactive bands were not observed with the I334S mutant (far right).

Human FLT3-ITD⁺ AML cells expressing HCK-T338H are resistant to A-419259.

Expression of HCK is primarily restricted to normal cells of myeloid lineage and has been implicated in the etiology of AML (see Introduction). To determine whether the HCK-T338H gatekeeper mutation has the potential to cause A-419259 resistance in the context of myeloid cells, we used human TF-1 myeloid cells as a model system. TF-1 cells require GM-CSF to support their proliferation and survival in culture and can be transformed to a cytokine-independent phenotype by retroviral transduction of the AML-associated receptor tyrosine kinase mutant, FLT3-ITD.^{6, 23, 24} TF-1 cells were first transformed to cytokine independence with FLT3-ITD. The TF-1/Flt3-ITD cell population was then transduced with retroviruses carrying wild-type HCK or each of the mutants recovered from the Rat-2 cell codon mutagenesis screen. Each population of cells was then tested for sensitivity to growth inhibition by A-419259 over a wide range of concentrations (Figure 6). Of all of the mutants tested, only HCK-T338H expression resulted in reduced sensitivity of TF-1/FLT3-ITD cells to the inhibitor, and the effect was greater than that observed with the previously described gatekeeper mutant, HCK-T338M⁹.

Figure 6. The HCK-T338H gatekeeper mutant confers A-419259 resistance in TF-1 myeloid cells transformed with FLT3-ITD.

TF-1 cells were transformed to cytokine-independent growth by expression of the FLT3-ITD receptor tyrosine kinase mutant associated with AML. Wild-type HCK (WT), as well as the six HCK mutants associated with acquired resistance in the Rat-2 cell codon mutagenesis screen (P32H, D158I, M302W, I334S and T338H) were then expressed in the TF-1/FLT3-ITD cells. Viability of each cell population over the range of A-419259 concentrations shown was assessed by Cell Titer-Blue assay, and the resulting concentration-response curves were best-fit by nonlinear regression analysis (GraphPad Prism v8) to estimate the IC₅₀ values shown. The upper left panel compares responses of TF-1 cells expressing FLT3-ITD with wild-type HCK (black curve) vs. FLT3-ITD alone (blue curve). All other panels compare responses of TF-1 cells expressing FLT3-ITD and each HCK mutant (red curves); the FLT3-ITD + wild-type HCK curve is plotted on each panel for reference (dotted curves). TF-1/FLT3-ITD cells expressing a previously described engineered resistance mutant, HCK-T338M, were also included for comparison (lower right panel). Each data point was assayed in triplicate, normalized to the DMSO (no inhibitor) control, and plotted as the mean value ± SE.

To correlate growth suppression with effects on kinase activity in the presence of A-419259, HCK was immunoprecipitated from each cell population and immunoblotted for activation loop phosphorylation and HCK protein recovery as before (Figure 7A). HCK was expressed and constitutively active in TF-1/Flt3-ITD cell populations expressing wild-type HCK as well as five of the six HCK mutants; the one exception was the I334S mutant which underwent proteolytic cleavage as previously observed in 293T cells (data not shown). In cells expressing wild-type HCK, A-419259 potently inhibited autophosphorylation with an IC₅₀ value of less than 30 nM, consistent with previous observations in this system as well as established FLT3-ITD⁺ AML cell lines.⁶ The P32H, D158I, and M302W mutants of HCK all remained sensitive to A-419259, with IC₅₀ values similar to wild-type. However, HCK-T338H remained phosphorylated on Tyr416 in the presence of A-419259 at concentrations up to 100 nM, consistent with the reduced sensitivity of cells expressing this gatekeeper mutant to growth suppression by the inhibitor. Cells expressing T338M were also assayed for comparison and showed somewhat higher resistance than T338H in terms of activation loop phosphorylation. However, HCK-T338M was expressed at much lower levels than T338H, suggesting that mutant protein stability as well as inhibitor resistance may contribute to the overall sensitivity of the cells to the compound.

Figure 7. HCK-T338H is resistant to A-419259 in TF-1 cells transformed by FLT3-ITD.

TF-1 cells were transformed with FLT3-ITD, followed by expression of wild-type HCK or the six mutants associated with acquired A-419259 resistance from the Rat-2 cell codon mutagenesis screen (P32H, D158I, M302W, I334S and T338H). An engineered gatekeeper mutant of HCK (T338M) was also included for comparison. Cells were treated with A-419259 at the concentrations shown (0, 30, 100, 300 nM), followed by immunoprecipitation of HCK and immunoblotting to assess activation loop phosphorylation (pY416) and HCK protein recovery. Immunoreactive proteins were visualized using secondary antibodies conjugated to infrared dyes and imaged using the LI-COR Odyssey system. Bar graphs below each set of representative images show the ratio of pY416 to HCK signal intensities from three independent experiments. Ratios from each experiment were normalized to the DMSO controls, and the average values ± S.E. are shown. B) Analysis of wild-type and mutant HCK transcript levels in TF-1/FLT3-ITD cell populations. Total RNA was isolated from parental TF-1 cells (Control) and TF-1/FLT3-ITD cells co-expressing wild-type and mutant forms of HCK as in part A. HCK transcript levels were analyzed by quantitative real-time RT-PCR, and relative expression levels are presented as the base 2 antilog of the ΔC_t values relative to GAPDH. Each cell population was analyzed in duplicate, with triplicate determinations per sample, and all values are presented. The mean value in each group is represented by the vertical line ± S.E.

To determine whether the differences in HCK protein expression observed with the different mutants were due to protein stability rather than mRNA expression levels, we performed quantitative real-time RT-PCR analysis of HCK transcript levels in each TF-1 cell population (Figure 7B). HCK transcript levels were very similar across all TF-1 cell populations, except for the I334S mutant which was somewhat reduced. Thus, the differences observed in HCK protein expression levels are most likely related to effects of the mutations on protein stability.

The degree of A-419269 resistance observed in TF-1/FLT3-ITD cells over-expressing the HCK-T338H mutant is relatively modest, an observation that may reflect the nature of the TF-1 cell model system for AML. Over-expression of HCK alone is not sufficient to cause cytokine-independent growth of this cell line, requiring prior transformation with FLT3-ITD, a common AML-associated mutation as described above. Under these conditions, expression of the HCK-T338H gatekeeper mutant in TF-1/FLT3-ITD cells causes a 2-fold shift in the IC₅₀ for A-419259, supporting the idea that HCK contributes to the inhibitor response. Importantly, the shift in potency for growth suppression correlates with the degree of kinase inhibition (IC₅₀ of about 100 nM in each case), suggesting that an HCK mutant with higher intrinsic resistance to A-419259 may produce a greater reduction in the inhibitor response. Future work will assess this possibility in the context of A-419259 selection in a myeloid cell background.

Structural basis of HCK-T338H resistance to A-419259.

The HCK gatekeeper residue, Thr338, forms a hydrogen bond with the C4 amino group of the A-419259 pyrrolopyrimdine core (Figure 8A). Substitution of T338 with histidine, the sole resistance mutation isolated from the codon mutagenesis screen, is anticipated to result in steric and electrostatic clash with the ligand (modeled in Figure 8B). Importantly, substitution with histidine at the gatekeeper position did not alter HCK protein stability or kinase activity, demonstrating that this missense mutation does not impart a fitness cost in terms of kinase function. X-ray crystallography of near-full-length HCK bound to A-419259 shows that this compound induces a DFG-in/C-helix-out conformation (Figure 1B).⁹ Work with other SRC-family kinase inhibitors that induce this active site conformation has shown that they also enhance the overall stability of the SH3-SH2-kinase core;¹⁹ mutations that destabilize assembly of this inactive conformation may therefore be predicted to cause resistance through an allosteric mechanism. However, no mutations of this type were observed in our unbiased screen, suggesting that A-419259 and possibly other C-helix-out inhibitors may be less prone to allosteric resistance mutations. This observation is in marked contrast to inhibitors that stabilize a complementary DFG-out/C-helix-in ‘Type II’ conformation, like imatinib for BCR-ABL in CML or quizartinib for FLT3-ITD in AML. In these cases, resistance can result from a variety of mutations both within and outside of the drug binding site. Allosteric loss of inhibitor action relates to conformational changes that bias the kinase active site toward the DFG-in conformation, which is no longer capable of high-affinity inhibitor recognition.^{15, 16} In the case of Bcr-Abl, which shares a similar SH3-SH2-kinase core structure with HCK, mutations that disturb SH3-linker interaction promote imatinib resistance, while experimental enhancement of this interaction increases inhibitor potency.²⁵ Whether our findings with A-419259 and HCK are more generally predictive of narrow resistance profiles for other DFG-in/C-helix out inhibitors will require unbiased codon mutagenesis screens with additional inhibitors in this class.

Figure 8. Molecular model of the HCK-T338H gatekeeper mutant in complex with A-419259.

A) Close-up view of the active site of HCK bound to A-419259. The side chain of the wild-type gatekeeper residue (T338) is modeled as spheres and makes a hydrogen bond with the C4 amino group of the A-419259 pyrrolopyrimidine scaffold. The methylpiperizinyl group of A-419259 forms a hydrogen bond with the side chain of Asp348. B) In silico mutagenesis of T338 to histidine removes this critical H-bond, creating electrostatic and steric clash predicted to disrupt inhibitor action.

METHODS

Sources of cell lines, antibodies, inhibitors, and other reagents used in this study, as well as methodological details, are provided in the Supporting Information.