Research ArticleBiochemistry

The Src family kinase Fgr is a transforming oncoprotein that functions independently of SH3-SH2 domain regulation

See allHide authors and affiliations

Science Signaling  23 Oct 2018:
Vol. 11, Issue 553, eaat5916
DOI: 10.1126/scisignal.aat5916

Fgr: The maverick in the Src family

Members of the family of kinases that includes Src, Hck, Lyn, and others (collectively referred to as Src family kinases) are understood to be regulated by intramolecular interactions involving SH2/SH3 domains and an inhibitory “tail” region. However, Shen et al. found that one member of this family, Fgr, is not regulated by these interactions because of a small and unique difference in its peptide sequence in its activation loop. Replacing this portion with that of other Src family members reduced the basal activity of Fgr. Simply expressing Fgr induced cancer-like growth in fibroblasts, and overexpressing it in human myeloid progenitor cells increased their sensitivity to a myeloproliferative cytokine. These findings may explain why Fgr expression is linked with the development of acute myeloid leukemia.


Fgr is a member of the Src family of nonreceptor tyrosine kinases, which are overexpressed and constitutively active in many human cancers. Fgr expression is restricted to myeloid hematopoietic cells and is markedly increased in a subset of bone marrow samples from patients with acute myeloid leukemia (AML). Here, we investigated the oncogenic potential of Fgr using Rat-2 fibroblasts that do not express the kinase. Expression of either wild-type or regulatory tail-mutant constructs of Fgr promoted cellular transformation (inferred from colony formation in soft agar), which was accompanied by phosphorylation of the Fgr activation loop, suggesting that the kinase domain of Fgr functions independently of regulation by its noncatalytic SH3-SH2 region. Unlike other family members, recombinant Fgr was not activated by SH3-SH2 domain ligands. However, hydrogen-deuterium exchange mass spectrometry data suggested that the regulatory SH3 and SH2 domains packed against the back of the kinase domain in a Src-like manner. Sequence alignment showed that the activation loop of Fgr was distinct from that of all other Src family members, with proline rather than alanine at the +2 position relative to the activation loop tyrosine. Substitution of the activation loop of Fgr with the sequence from Src partially inhibited kinase activity and suppressed colony formation. Last, Fgr expression enhanced the sensitivity of human myeloid progenitor cells to the cytokine GM-CSF. Because its kinase domain is not sensitive to SH3-SH2–mediated control, simple overexpression of Fgr without mutation may contribute to oncogenic transformation in AML and other blood cancers.


The human Src protein-tyrosine kinase family consists of eight members, which display ubiquitous (Src, Yes, and Fyn) as well as tissue-specific (Lck, Hck, Lyn, Fgr, and Blk) patterns of gene expression. Src family kinases are activated in response to a wide variety of extracellular stimuli, including growth factors, cytokines, hormones, antigens, and cell adhesion molecules, as well as proteins expressed by invading pathogens (13). These diverse signaling inputs illustrate the complex physiological roles for Src family members in the regulation of cellular growth, differentiation, migration, and survival, as well as more specialized functions in innate and adaptive immune responses.

Src family kinases are often overexpressed and active in several forms of cancer (4). For example, Src itself is commonly overexpressed and kinase active in colon and breast cancer, where it contributes to tumor cell growth, metastasis, and progression. Src family members normally expressed in hematopoietic cells, including Hck, Lyn, and Fgr, have been linked to myeloid leukemias (5). Hck cooperates with Bcr-Abl in the pathogenesis of chronic myeloid leukemia (CML) (6, 7), and direct phosphorylation of Bcr-Abl by Hck results in imatinib resistance in the absence of Bcr-Abl mutations (8). Hck has been implicated in the genesis of acute myeloid leukemia (AML) as well. Gene expression profiling has revealed that Hck transcripts are highly overrepresented in AML-derived leukemic stem cells in comparison to their normal counterparts (9), and RNA interference (RNAi)–mediated knockdown of Hck expression blocks AML cell growth (10). Furthermore, a pan-Src family kinase inhibitor (RK-20449; also reported under the name A-419259) (11, 12) markedly reduced primary AML bone marrow cell growth in engrafted immunocompromised mice, including patient AML cells resistant to conventional chemotherapy (10). Lyn has been shown to be active in clinical AML isolates and may cooperate with the AML-associated receptor tyrosine kinase mutant FLT3-ITD in the activation of Stat5 (signal transducer and activator of transcription 5) downstream (1315). Fgr is also highly expressed and active in AML bone marrow samples, and small interfering RNA knockdown of Fgr substantially reduced primary AML cell growth (16). Although knockdown studies suggest that Hck, Lyn, and Fgr are important for AML development in vitro, less is known about the mechanisms leading to unscheduled kinase activation in AML.

All members of the Src kinase family share a common domain organization, with a myristoylated N-terminal domain involved in membrane localization (Fig. 1) (17). This unique region, which is also palmitoylated in most cases, is followed by conserved SH3, SH2, and kinase domains as well as a negative regulatory tail. Crystal structures of near–full-length Src (18, 19) and Hck (20, 21) show that the SH3 domain engages a polyproline type II (PPII) helix formed by the linker connecting the SH2 and kinase domains, whereas the tyrosine-phosphorylated C-terminal tail engages the SH2 domain. Together, these noncatalytic domains lock the kinase domain in an inactive conformation. Mutations or peptide ligands that disrupt SH3- and SH2-mediated interactions stimulate kinase activity both in vitro and in cells (17, 22, 23). For example, the HIV-1 Nef protein, a selective ligand for the Hck SH3 domain, induces constitutive activation of Hck in vitro (23), whereas coexpression of Nef with Hck results in kinase activation and oncogenic transformation of rodent fibroblasts (24). These and many other studies have led to the general theory that all Src family kinases are negatively regulated by intramolecular interactions involving their SH3 and SH2 domains and that phosphorylation of the negative regulatory tail is required for suppression of kinase activity in vivo.

Fig. 1 Src family kinase domain organization and crystal structure of the inactive conformation.

The x-ray crystal structure of inactive, near–full-length Src is modeled at the top as a ribbon diagram (left) and space-filling model (right). The linear domain organization of full-length, wild-type Src is shown below the models, along with the near–full-length version with the modified tail sequence (Src-YEEI) used in this study. The N-terminal unique domains of all Src family members are modified by myristoylation (Myr) and palmitoylation (not shown) in most cases. The unique region is followed by the SH3 and SH2 domains, which are joined to the bi-lobed kinase domain through the SH2-kinase linker. The activation loop (A-loop), an important dynamic element involved in kinase domain regulation, has a single tyrosine residue (Tyr416). Autophosphorylation of Tyr416 stabilizes the active form of the kinase. The kinase domain is followed by the C-terminal tail, with a single tyrosine residue, Tyr527; phosphorylation of this site by Csk induces intramolecular SH2 engagement important for negative regulation.

In this study, we investigated the regulation and oncogenic potential of the myeloid Src family member, Fgr. Among the least-studied members of the Src family, Fgr normally attenuates Fcγ receptor signaling pathways linked to phagocytosis and other innate immune responses (2527). Although Fgr is overexpressed in a subset of primary AML blasts (16), little is known about the regulation of Fgr kinase activity in AML. No crystal structures of Fgr have been reported to date, and the mechanisms regulating its activity are unknown. Here, we report the unexpected finding that ectopic expression of wild-type Fgr induces oncogenic transformation of Rat-2 fibroblasts. Fgr was phosphorylated on both the activation loop and the negative regulatory tail, and tail tyrosine mutation did not further enhance transforming activity. A recombinant Fgr SH3-SH2–kinase core protein with a tyrosine-phosphorylated tail was insensitive to activation by SH3 and SH2 domain displacement, suggesting that Fgr kinase activity is independent of these regulatory domains. The amino acid sequence of the Fgr activation loop is unique, with a proline residue just C-terminal to the activation loop tyrosine. Substitution of the Fgr activation loop sequence with that of Src suppressed both kinase and transforming activities in fibroblasts. Last, wild-type Fgr was active after expression in human myeloid progenitor cells, where it enhanced sensitivity to the myelopoietic cytokine, granulocyte-macrophage colony-stimulating factor (GM-CSF). Together, these results indicate that, unlike other members of the Src family, the Fgr kinase domain is not under the allosteric control of the SH3-SH2 region. Simple overexpression of Fgr without mutation may therefore contribute to oncogenic transformation in AML, identifying this unique kinase as a target for therapeutic intervention.


Rodent fibroblasts are transformed by wild-type Fgr

Expression of the myeloid Src family member Fgr is often enhanced in primary AML blast cells. RNAi-mediated knockdown of Fgr expression or small-molecule inhibition of Fgr kinase activity impairs AML cell proliferation, implying that unregulated kinase signaling by Fgr contributes to AML pathogenesis (16, 28). To address the transforming potential of Fgr directly, we performed transformation assays with rodent fibroblasts (Rat-2 cells) that do not express Fgr or other hematopoietic Src family members, thereby providing a clean background for assessment of Fgr transforming potential. Along these lines, we have previously shown that active mutants of Hck readily transform Rat-2 cells in both focus-forming and soft-agar colony assays and that kinase activity correlates directly with transforming potential (2932). On the other hand, wild-type Hck is tail phosphorylated and nontransforming in this model system (32).

For these experiments, we created recombinant retroviral vectors for the efficient transfer of wild-type, full-length Fgr into Rat-2 fibroblasts. In addition, we created an Fgr mutant, in which the conserved negative regulatory tail tyrosine [Tyr527; all residue numbering as per the crystal structure of human Src; Protein Data Bank (PDB): 2SRC] (18) was replaced with phenylalanine (Y527F mutant). This substitution prevents phosphorylation by the regulatory kinase Csk (33) in the context of Hck and other Src family members, resulting in constitutive kinase activity and fibroblast transformation. However, the role of tail phosphorylation in the regulation of Fgr is not currently established. Rat-2 fibroblasts were infected with recombinant retroviruses carrying the wild-type or tail mutants of Fgr, or an empty control vector. Rat-2 cells expressing wild-type Fgr showed clear evidence of transformation, manifested as foci of cells growing on top of the otherwise contact-inhibited cell monolayer (Fig. 2A). Rat-2 cells expressing wild-type Fgr formed foci as efficiently as those expressing the tail mutant, indicating that overexpression of wild-type Fgr alone is sufficient to drive a transforming signal in this system. To quantify this result, the cells were plated in soft agar, and transformed colonies were counted 14 days later. Both wild-type and tail-mutant Fgr expression induced cells to form a similar number of colonies from each of three input amounts of cells (Fig. 2B). For comparison, we also transduced Rat-2 fibroblasts with wild-type and tail mutants of Hck. In this case, expression of tail-mutant Hck induced colony formation, whereas expression of wild-type Hck did not, consistent with previous results (2932).

Fig. 2 Expression of wild-type Fgr induces oncogenic transformation of Rat-2 fibroblasts.

(A) Micrographs of representative 14-day monolayer cultures of Rat-2 fibroblasts after infection with recombinant retroviruses carrying wild-type (WT) and tail mutants (Tyr527 to Phe; YF) of Fgr or Hck and selection with G418. Transformed foci appear as clumps of refractile cells growing on top of the contact-inhibited cell monolayer. Scale bar, 300 μm. (B) Colony-forming assays with the cells described in (A). Each cell population, as well as control cells transduced with an empty vector, was plated in semisolid medium and incubated for 2 weeks. Cultures were stained with Wright-Giemsa, and colonies were counted from scanned images of each plate using ImageJ software. Colony counts from each plate as a function of cellular input are shown, with the mean value represented by the horizontal bar (n = 3). This entire experiment was repeated twice using independently derived retroviral supernatants and produced comparable results. (C) Protein extracts from each of the Rat-2 cell populations shown in (A) were separated by SDS–polyacrylamide gel electrophoresis (PAGE), transferred to nitrocellulose, and probed with phosphospecific antibodies to the activation loop phosphotyrosine (pY416), the tail phosphotyrosine (pY527), and each kinase protein. Immunoreactive proteins were visualized using secondary antibodies conjugated to infrared dyes and imaged using the LI-COR Odyssey system and software. Representative blots of three independent determinations are shown.

To determine whether fibroblast transformation correlated with Fgr and Hck kinase activity, we performed immunoblots with antibodies specific for the activation loop phosphotyrosine, pTyr416. The Fgr activation loop was phosphorylated in transformed cells expressing either the wild-type or the tail mutant (Fig. 2C). In contrast, only the tail mutant of Hck was phosphorylated on the activation loop, consistent with the transformation data. Control blots with antibodies selective for the negative regulatory phosphotyrosine residue (pTyr527) confirmed that both Fgr and Hck were tail phosphorylated, whereas the Y527F tail mutants were not (Fig. 2C). These results suggest that tail phosphorylation is not sufficient for suppression of Fgr kinase activity in the Rat-2 fibroblast transformation assay.

Fgr kinase activity is independent of SH3 and SH2 domain control in vitro

As noted above, SH3 and SH2 domains play important roles in the regulation of Src family kinase activity as well as signal transduction. X-ray crystal structures of near–full-length Hck and Src showed that the SH3 and SH2 domains pack against the back of the kinase domain to hold the kinase in an inactive state (Fig. 1) (17). Previous studies have shown that peptide ligands for the SH3 and SH2 domains can displace these regulatory interactions in recombinant Hck and other Src family members, resulting in stimulation of kinase activity in vitro (22, 34). We therefore used a similar approach to investigate the roles of SH3 and SH2 in Fgr kinase domain regulation.

We expressed near–full-length Fgr in Sf9 insect cells and purified the protein to homogeneity using our established methods for Hck and Src (22). In this form of Fgr, the N-terminal unique domain is replaced with a His-tag, whereas the C-terminal tail sequence is modified to pTyr-Glu-Glu-Ile (“YEEI” tail). This modification promotes tail phosphorylation in insect cells and intramolecular engagement of the SH2 domain. Note that the crystal structure of an analogous form of near–full-length Hck-YEEI adopts an overall structure virtually identical to the wild-type kinase, with the tail tyrosine phosphorylated and SH2 bound in both cases (20, 21). Recombinant near–full-length Fgr-YEEI was purified using a combination of ion-exchange and immobilized metal affinity chromatography, followed by gel filtration. The identity of purified Fgr-YEEI was confirmed by electrospray mass spectrometry (MS), and peptic phosphopeptide mapping confirmed the presence of a single phosphate group on the C-terminal tail tyrosine, whereas the activation loop was not phosphorylated (fig. S1).

We next assessed the kinase activity of recombinant Fgr-YEEI, along with the analogous forms of Hck and Src, using the Z-LYTE in vitro kinase assay. This assay is based on phosphorylation of a fluorescence resonance energy transfer (FRET)–peptide substrate labeled with coumarin and fluorescein on its N and C termini, respectively, which form a FRET pair (35). After incubation with the kinase and adenosine 5′-triphosphate (ATP), the reaction was developed with a site-specific protease that selectively cleaves unphosphorylated peptides. Cleavage results in loss of the FRET signal, and the extent of phosphorylation is then calculated as the ratio of coumarin to FRET fluorescence relative to a control phosphopeptide. Recombinant Fgr was first titrated into the assay over a range of 0.5 to 500 ng per well, with the ATP concentration set to the Km (Michaelis constant) value. Fgr-YEEI displayed the highest specific activity in this assay, with a median effective concentration (EC50) value of 7.9 ng (fig. S2). Hck-YEEI was about half as active as Fgr-YEEI, with an EC50 value of 16.6 ng, whereas Src-YEEI was about 10-fold lower with an EC50 of 77.9 ng.

We then compared all three kinases for sensitivity to activation by peptides that displace SH3 (VSL12 peptide), SH2 (pYEEI peptide), or both (pFAK peptide) regulatory domain interactions (sequences in Fig. 3A). For these experiments, each kinase was added to the assay at a concentration corresponding to 20 to 25% of maximal activity (based on the obtained titration curve; fig. S1) and in the presence of increasing concentrations of each peptide ligand. Both Hck-YEEI and Src-YEEI were maximally activated by all three peptides (Fig. 3B). The pFAK peptide was the most potent activator, which likely reflects simultaneous engagement of both the SH3 and SH2 domains, as shown previously (34, 36). Unexpectedly, Fgr was virtually insensitive to activation by these peptides: VSL12 and pYEEI had no effect, whereas pFAK produced only a 10 to 15% increase in activity at the highest ratio tested (300-fold molar excess). In contrast, both Hck-YEEI and Src-YEEI were maximally stimulated by a threefold molar excess of the pFAK peptide.

Fig. 3 Activation of recombinant Src family kinases by SH3- and SH2-binding peptides.

(A and B) Near–full-length Fgr, Src, and Hck kinase activities were assayed in the presence of peptides that bind to the SH3 domain alone (VSL12), to the SH2 domain alone (pYEEI), or to both SH2 and SH3 (pFAK). The amino acid sequence of each peptide ligand is shown (A), in which amino acids involved in regulatory domain engagement are underlined. Each recombinant kinase was added to the assay at a concentration that yielded 20 to 25% of maximum activity, and the input of each peptide was varied over the range of peptide to kinase molar ratios shown (B). Each data point was measured in triplicate and is shown as the mean value ± SD; the error bars are smaller than the size of the data points.

The lack of activation of Fgr-YEEI by SH3 and SH2 domain peptide ligands may reflect low affinity for their target domains in Fgr. To control for this possibility, we performed surface plasmon resonance (SPR) studies using purified recombinant SH3 and SH2 domains from Fgr and Hck. For these studies, each of the peptides was immobilized on the SPR chip surface via an N-terminal biotin moiety. Recombinant SH3 and SH2 domains were then injected over a range of concentrations, and the interaction kinetics were monitored in real time. The resulting sensorgrams (Fig. 4) showed that the VSL12, pYEEI, and pFAK peptides all bound to the Hck and Fgr SH3 and SH2 domains with similar kinetics, indicating that the inability of these peptides to simulate Fgr activity is not due to reduced binding affinity for the SH3 and SH2 domains.

Fig. 4 SPR analysis of peptide ligand binding to recombinant Fgr and Hck SH2 and SH3 domains.

Biotinylated VSL12, pFAK, and pYEEI peptide ligands (sequences shown in Fig. 3) were immobilized on a streptavidin biosensor chip. Recombinant SH3 and SH2 domain proteins were then injected in triplicate over the range of concentrations shown until equilibrium was reached, followed by a 2-min dissociation phase. Reference-corrected sensorgrams were fit by a 1:1 Langmuir binding model, and equilibrium Kd values (M) were calculated from the resulting kinetic rate constants (koff/kon). The three replicate sensorgrams obtained at each domain protein concentration are shown (color code at top right), and a single representative fitted curve for each set is shown as a dark gray line. RU, resonance units.

We also performed SPR studies with these peptide ligands and near–full-length Fgr-YEEI, which includes the SH3, SH2, and kinase domains plus the tyrosine-phosphorylated tail. The experimental approach was the same as for the individual domains, with the three peptides bound to the chip and recombinant Fgr-YEEI protein as analyte. All three peptides bound to Fgr-YEEI, although the kinetics of binding differ from those of the isolated domains (fig. S3). With all three peptide ligands, the rate of association was slower with near–full-length Fgr-YEEI compared to the individual domains. This may reflect competition with intramolecular interactions of the SH2 and SH3 domains with the phosphorylated tail and linker, respectively. Once associated, the peptide ligands showed slower dissociation, resulting in calculated equilibrium dissociation constant (Kd) values in the 10−6 to 10−7 M range. Together, these SPR data suggested that intramolecular interactions present in the crystal structures of near–full-length Hck and Src are also found in Fgr.

Although the slower SPR binding kinetics observed with near–full-length Fgr support an assembled kinase structure, the Kd value derived from the data are not in complete agreement with previous experiments using Hck. Using a similar SH3 peptide ligand, Moarefi et al. (23) reported a Kd value of 0.25 μM for the isolated Hck SH3 domain versus 1.4 μM for tail-phosphorylated, near–full-length Hck. These results are consistent with an expected increase in the apparent Kd value for the external SH3 peptide ligand when there is competition with internal SH3 interaction and the SH2-kinase linker. For Fgr, however, we observed that the apparent Kd value for the isolated SH3 domain is about 8 μM (Fig. 4), whereas that of near–full-length Fgr is only 0.2 μM (fig. S3). Whether this difference reflects a different binding mode for the SH3 domain in the context of near–full-length Fgr, or just differences in the experimental approaches and ligands used, will require further investigation.

Fgr-YEEI adopts an assembled conformation in solution similar to that of Src-YEEI

Our results thus far indicated that wild-type Fgr is active in cells and is not activated by SH3 and SH2 domain ligands in vitro, unlike other members of the Src-kinase family. One possible explanation for these findings is that Fgr may not adopt the compact, assembled conformation observed in the x-ray crystal structures of inactive Hck and Src (Fig. 1). In this conformation, the SH3 domain binds the PPII helix formed by the SH2-kinase linker, whereas the SH2 domain engages the tyrosine-phosphorylated tail (discussed above). To explore the conformation of Fgr in solution, we used hydrogen-deuterium exchange MS (HX MS), which enables regional monitoring of deuterium uptake into folded proteins as a function of time. Regions of a protein that are exposed to solvent and are not hydrogen bonded take up deuterium more readily than solvent-protected, hydrogen-bonded regions as a function of protein-protein interactions or secondary structure. The application of HX MS to the study of Src, Hck, and other multidomain kinases is reviewed by Engen et al. (37).

The objective of the HX MS experiments was to determine whether the Fgr SH3-SH2 regulatory unit engages the kinase domain in recombinant Fgr-YEEI, as observed previously for Hck and Src. To approach this question, we expressed and purified recombinant Fgr and Src SH3-SH2 proteins without the kinase domains. We then compared deuterium incorporation into the SH3-SH2 proteins alone versus incorporation into the near–full-length proteins, which also contain the kinase domains. We reasoned that if the SH3-SH2 region was packed against the kinase domain in the near–full-length proteins, then the rate of deuterium uptake into peptides derived from SH3-SH2 would be reduced when compared to the isolated SH3-SH2 domains without the kinase domain.

Each of the four proteins was independently incubated in D2O-based buffer, and exchange was quenched at specific times ranging from 10 s to 4 hours. The quenched samples were digested with pepsin, and 10 peptides from the SH3 and SH2 domains were then followed for changes in mass over time, which reflects the exchange of backbone amide hydrogens with deuterium. The difference in deuterium uptake between the near–full-length kinases and the isolated SH3-SH2 proteins was plotted over time for each peptide (Fig. 5). The difference maps integrate all the data from the individual deuterium uptake plots for each peptide (fig. S4). Deeper shades of blue reflect increasing protection from exchange in the near–full-length proteins. Nine of the 10 peptides derived from the SH3 and SH2 domains of near–full-length Fgr and Src showed reduced uptake of deuterium when the kinase domain was present, compared to the corresponding peptides from the isolated SH3-SH2 proteins. The one exception was a peptide derived from each SH3 domain that is solvent exposed in the crystal structure of Src, providing a negative control (peptide sequence HILNNTEGDW in Fgr SH3; Fig. 5). Deuterium exchange in Fgr SH3-SH2 was nearly identical to deuterium exchange in Src SH3-SH2 in the presence of the kinase domain, suggesting that they adopt similar assembled conformations in solution. In addition, these HX MS results with near–full-length Fgr were very similar to previous HX MS studies of Src, the closely related kinases Hck and Lck, and the Abl core; in each case, reduced deuteration was found in SH3-SH2 when associated with the kinase domain (37). These data therefore indicate that Fgr adopts the assembled conformation previously associated with the crystal structures of inactive Src and Hck as well as the Abl core, which also consists of an SH3-SH2–kinase domain structure.

Fig. 5 HX MS shows that near–full-length Fgr adopts an assembled conformation similar to inactive Src.

Deuterium uptake by the Src and Fgr SH3-SH2 domains alone was compared to uptake in the near–full-length kinases by HX MS. Difference maps for both Fgr and Src are shown, where the level of deuteration in each SH3-SH2 peptide was subtracted from that in the near–full-length kinase (SH3-SH2-K) at each of the time points indicated, and the difference was colored according to the scale shown. Protection from deuterium exchange (increasingly deeper blue color) was observed in the SH3 and SH2 domains from both Src and Fgr when the kinase domain was also present. The sequences of 10 homologous peptides derived from the Src and Fgr SH3-SH2 region are indicated on the maps, from the N terminus (top) to the C terminus (bottom). The corresponding uptake plots for each peptide are shown in fig. S4. Conn, SH3-SH2 connector.

We also compared deuterium uptake between near–full-length Fgr-YEEI and Src-YEEI directly. Although the amino acid sequences of these kinases are not identical, which complicates head-to-head comparison of deuterium uptake at the peptide level, a qualitative comparison of the overall dynamics of each protein was performed using fractional deuterium uptake analysis of all peptides derived from Fgr-YEEI and Src-YEEI. The fractional deuterium uptake by each peptide was calculated by dividing the observed deuterium incorporation (in daltons) by the number of exchangeable backbone amide hydrogens in the peptide at each time point (detailed in Materials and methods). The results, presented as a heat map (fig. S5), demonstrate notable similarity in regional deuterium exchange rates across both proteins. This analysis is consistent with the idea that both proteins adopt similar conformations in solution.

Unique activation loop sequence of Fgr contributes to kinase activity

The surprising finding that near–full-length Fgr adopts a similar overall conformation as inactive Src in solution while exhibiting fibroblast-transforming activity equivalent to a “tail” mutant led us to explore sequence differences between the kinase domains of Fgr and other Src family members. Whereas the overall kinase domain sequences are conserved, the sequence of the Fgr activation loop in the region immediately adjacent to the autophosphorylation site (Tyr416) is unique among all eight human Src family members (Fig. 6A). In Src, Hck, and other family members, this sequence is Tyr416-Thr-Ala-Arg (“TAR” motif). In the crystal structures of inactive Src and Hck, this region forms a short α helix that orients the side chain of Tyr416 toward the catalytic cleft, where it forms a hydrogen bond with the catalytic aspartate (Asp386; modeled in Fig. 6A). In Fgr, however, the three amino acids adjacent to Tyr416 are replaced with Asn-Pro-Cys (“NPC” motif). The presence of proline at position 418 is particularly intriguing, because it may prevent α-helix formation and inward orientation of Tyr416. To evaluate the influence of this activation loop motif on kinase regulation, we replaced the Fgr NPC motif with the conserved Thr-Ala-Arg (TAR motif) found in Hck. The resulting Fgr-TAR clone (with a wild-type tail) was then expressed in Rat-2 fibroblasts, followed by colony-forming assays and assessment of Tyr416 phosphorylation. Fibroblasts expressing Fgr-TAR produced 16 to 20% fewer colonies than wild-type Fgr at each of three different inputs of cells (Fig. 6B). This decrease in transforming activity is mirrored in a 50% reduction in autophosphorylation of the activation loop compared to the wild-type kinase in these cells (Fig. 6C). These results show that the activation loop unique to the Fgr kinase domain may be responsible, at least in part, for its higher basal kinase activity both in vitro and in cell-based assays.

Fig. 6 Substitution of the Fgr activation loop with the conserved TAR motif suppresses kinase and transforming activity in Rat-2 cells.

(A) Left: Alignment of the activation loop sequences from the eight mammalian Src family members. Fgr is unique among the Src family in that the Thr-Ala-Arg (TAR) sequence adjacent to the activation loop tyrosine (YP) is substituted with Asn-Pro-Cys (NPC). Right: Src activation loop structure in the inactive conformation with the side chains of the TAR motif shown. The activation loop tyrosine (Y416) makes a hydrogen bond with the catalytic aspartate (D386) to stabilize the inactive structure (PDB: 2SRC). (B) Rat-2 cells expressing wild-type Fgr (WT), the tail mutant (YF), wild-type with the TAR substitution in the activation loop, and vector control cells were plated in triplicate in semisolid medium and incubated for 2 weeks. Cultures were stained with Wright-Giemsa, and colonies were counted from scanned images of each plate using ImageJ software. Average colony counts, ±SD, from each plate as a function of cellular input are shown from a representative of two experiments. *P < 0.05 by Student’s t test. (C) Protein extracts from the Rat-2 cell populations shown in (B) were immunoblotted for activation loop phosphorylation (pY416) and Fgr protein abundance. Data are the average ratio of pY416 to Fgr protein signals from three independent determinations ± SD; **P < 0.05 by Student’s t test.

Wild-type Fgr is active in myeloid leukemia cells and reduces the cytokine requirement for colony formation

Fgr is normally expressed in myeloid hematopoietic cells and has been implicated in AML. To determine whether Fgr is constitutively active in myeloid cells, we used the human TF-1 myeloid leukemia cell line, which requires GM-CSF for growth and survival in culture (38). These cells do not express detectable Fgr or Hck (28), making them a useful model system to study Fgr regulation in human cells of myeloid lineage. Using recombinant retroviruses, we created a population of TF-1 cells that express wild-type Fgr as well as matched vector control cells. Fgr was immunoprecipitated from the TF-1 + Fgr cell population, and activation loop tyrosine phosphorylation was assessed by immunoblotting (Fig. 7A). Wild-type Fgr was kinase active in TF-1 cells in the presence or absence of GM-CSF. Treatment of the cells with the pan-Src kinase inhibitor A-419259 resulted in near-complete suppression of Fgr activation loop phosphorylation at a concentration of 100 nM. The Fgr C-terminal tail tyrosine was also phosphorylated in TF-1 cells but was not affected by A-419259 treatment as anticipated. These results show that wild-type Fgr is active in human myeloid leukemia cells, consistent with the fibroblast transformation and in vitro kinase data.

Fig. 7 Wild-type Fgr is active in TF-1 myeloid cells and reduces the GM-CSF requirement for proliferation.

TF-1 myeloid cells were infected with recombinant retroviruses carrying the wild-type human Fgr coding sequence or an empty vector as a negative control, followed by selection with puromycin. (A) Cultures of TF-1/Fgr cells were expanded in the presence of GM-CSF. Cells were washed free of GM-CSF, split into two aliquots, and then incubated overnight in the presence or absence of GM-CSF (1000 pg/ml) or the pan-Src family kinase inhibitor A-419259 as indicated at the top. Fgr was immunoprecipitated from cell lysates and probed with phosphospecific antibodies to the activation loop phosphotyrosine (pY416), the tail phosphotyrosine (pY527), and total Fgr protein. Immunoreactive proteins were visualized using secondary antibodies conjugated to infrared dyes and imaged using the LI-COR Odyssey system and software. Images of representative blots are shown. Bar graphs below the images show the average intensity ratios for the phosphotyrosine signals divided by the Fgr protein abundance from three independent determinations. Ratios were normalized to the values for the +GM-CSF/no inhibitor conditions and are shown as mean values ± SE. *P < 0.05 compared to dimethyl sulfoxide control, all others not significant (by Student’s t test). (B) TF-1 cells expressing Fgr or vector (TF-1/vector and TF-1/Fgr, respectively) and vector control cells were plated in soft-agar colony-forming assays (1000 or 2500 cells per 35-mm plate) in the absence or presence of a suboptimal concentration of GM-CSF (100 pg/ml is 10% of the normal concentration used in routine cell culture). Colonies were visualized with Wright-Giemsa stain and quantified using ImageJ after 2 weeks. Each bar represents the average number of colonies observed ± SE (n = 3). *P < 0.05 by Student’s t test. (C) Fgr and Src gene expression data from 163 AML bone marrow samples were downloaded from TCGA database. Data are shown as the number of kinase complementary DNA fragments per kilobase of transcript per million mapped reads (FPKM). The dotted lines indicate the mean transcript values.

We also assessed the impact of Fgr expression on the TF-1 cell cytokine requirement for growth and survival. TF-1 + Fgr and vector control cell populations were plated in soft-agar colony-forming assays in the absence or presence of a suboptimal concentration of GM-CSF, and colony formation was recorded 2 weeks later. Whereas overexpression of Fgr alone did not result in cytokine-independent colony formation, it did significantly enhance colony counts in the presence of a low concentration of GM-CSF (100 pg/ml, which is 10% of the concentration used for routine culture of TF-1 cells; Fig. 7B). This observation suggests that Fgr may enhance the cytokine responsiveness of leukemic progenitors in the subset of AML cases that strongly express this oncogenic kinase. Analysis of Fgr transcripts in 163 cases of primary AML bone marrow samples from The Cancer Genome Atlas (TCGA) database revealed a remarkably broad range of Fgr expression levels, spanning nearly three orders of magnitude (Fig. 7C). In contrast, expression of c-Src varied by about 100-fold, with maximum values nearly 10-fold lower than those observed with Fgr.


Mutations that disrupt SH3-linker or SH2-tail interaction are sufficient to activate Hck and other Src family kinases and induce transformation after expression in rodent fibroblasts (2932). In contrast, overexpression of wild-type Hck does not transform Rat-2 cells, and its kinase activity remains suppressed. Here, we found that, unlike Hck, wild-type Fgr was active and promoted transformation in rodent fibroblasts. The transforming activity of wild-type Fgr was equivalent to that of Hck and Fgr tail mutants in which Tyr527 was replaced with phenylalanine. This finding challenges current dogma surrounding Src family kinase regulation. Major tenets of this regulatory theory include the following: (i) Wild-type Src kinases are inactive when expressed in rodent fibroblasts and in their physiological contexts in the absence of activating signals; (ii) Src family kinase activity is strictly regulated by intramolecular interactions involving their SH3 and SH2 domains, in a manner demonstrated by the x-ray crystal structures of inactive Src and Hck; and (iii) kinase activation and cellular transformation require disruption of SH3-linker interaction, SH2-phosphotail interaction, or both. Immunoblots from rodent fibroblasts transformed with wild-type Fgr showed that activation loop Tyr416 was phosphorylated in transformed cells, providing evidence that the kinase is active and therefore likely responsible for transformation. The negative regulatory tail tyrosine, Tyr527, was also phosphorylated in wild-type Fgr, demonstrating that Fgr was active in cells despite the presence of tail phosphorylation and presumptive SH2 domain engagement. In contrast to Fgr, wild-type Hck activity was strictly suppressed in Rat-2 cells, with no detectable Tyr416 phosphorylation.

Biochemical and structural studies support a model of Src family kinase negative regulation in which the SH3 domain engages a PPII helix formed by the SH2-kinase linker, whereas the SH2 domain interacts with the phosphotyrosine residue in the C-terminal tail (17). Four Src family members (Hck, Src, Fyn, and Lyn) can be activated by SH3 and SH2 domain displacement both in vitro and in cell-based systems (22, 23, 34, 36, 39). Here, we tested recombinant Fgr for sensitivity to activation in vitro using peptide ligands for the SH3 and SH2 domains, as well as a peptide that engages both domains simultaneously. Whereas recombinant Hck and Src were fully activated by all three peptides, Fgr kinase activity was unaffected by SH3- or SH2-binding peptides and only slightly activated by dual domain engagement. These results strongly suggest that the Fgr kinase domain is not under allosteric regulation by its SH3 and SH2 domains, unlike other members of the Src family.

The x-ray crystal structures of both Hck and c-Src show that SH2-kinase linker residues just C-terminal to the polyproline helix interact with the N-terminal lobe of the kinase domain (1821). Together with SH2 engagement of the tyrosine-phosphorylated tail, these interactions clamp the SH3-SH2 region against the back of kinase domain and are important for allosteric suppression of kinase activity. These observations led us to speculate that Fgr may not adopt this “assembled” conformation as a possible explanation for its transforming activity in Rat-2 cells. To test this possibility, HX MS was used to compare the rates of deuterium uptake by the Fgr and Src SH3-SH2 dual domains alone and in the context of near–full-length kinases that also include the linker, kinase domain, and phosphorylated tail. The SH3 and SH2 domains of both Fgr and Src were protected from deuterium uptake in the context of the near–full-length kinases, as reported previously for inactive Hck (40). This result suggests that the Fgr core region adopts an assembled conformation similar to inactive Src and Hck, although these intramolecular contacts do not appear to regulate Fgr kinase domain activity.

Kinase domain amino acid alignment revealed a unique sequence for the Fgr activation loop that may partially explain its divergent regulatory behavior. The three amino acids C-terminal to Tyr416 in Fgr (Asn-Pro-Cys; NPC motif) are distinct from all other Src family kinases, where this region is substituted with Thr-Ala-Arg (TAR motif). The substitution of proline for alanine is particularly interesting, because proline is likely to prevent the formation of the short α helix formed by this activation loop region in inactive Hck and Src (see Fig. 6A). Substitution of the Fgr NPC motif with the Src TAR sequence reduced both Fgr kinase activity and fibroblast transformation efficiency. This observation suggests that the unique NPC motif may be partially responsible for the persistent kinase activity of Fgr. Because substitution of the unique activation loop sequence of Fgr with that of Src did not completely suppress kinase activity, additional amino acid differences in the kinase domain may also contribute to uncoupling of kinase activity from SH3-SH2 regulation. Differences in activation loop sequences have also been reported to affect the intrinsic activity of members of the Tec kinase family (41). In that study, NMR (nuclear magnetic resonance) and HX MS analyses revealed that activation loop protein dynamics account for major differences in the catalytic efficiency between Itk and Btk, two prominent members of this family with roles in T and B cell receptor signaling, respectively.

Studies of primordial Src family kinase orthologs in unicellular organisms have revealed an arrangement of SH3, SH2, and kinase domains (42, 43) quite similar to those observed in metazoans. Although a C-terminal tail is also present, these Src-like kinases are not suppressed by Csk-mediated tail phosphorylation, suggesting that regulation of kinase activity by SH2-tail and other intramolecular interactions evolved later. One of these kinases (CoSrc1 from the unicellular protist, Capsaspora owczarzaki) also has a proline residue in +2 position C-terminal to the autophosphorylation site as observed for Fgr (42). Thus, Fgr may more closely resemble these primitive Src-like kinases, where intramolecular interactions involving the SH2 and SH3 domains do not allosterically regulate kinase activity but may control interactions with other proteins as well as subcellular distribution.

Wild-type Fgr was also active after expression in a human myeloid progenitor cell line, where it enhanced cellular survival in terms of colony-forming activity in response to low concentrations of GM-CSF. Given the remarkably high levels of Fgr expression in a subset of primary AML bone marrow samples (Fig. 7C), our data suggest that Fgr may provide a selective growth advantage in AML leukemic stem cells as well. Previous work has shown that knockdown of Fgr expression interferes with primary AML cell proliferation in vitro (16), supporting the development of Fgr inhibitors for targeted therapy in AML. A potent, Fgr-selective ATP-site inhibitor has been shown to induce growth arrest of AML patient samples that overexpress this kinase, providing additional evidence that unregulated Fgr kinase activity contributes to AML pathogenesis (28).


Rat-2 cell culture and transformation assay

Rat-2 cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (FBS) and antibiotic-antimycotic (Thermo Fisher). Rat-2 cells were transduced with recombinant Hck and Fgr retroviruses generated in 293T cells using the pSRαMSVtkneo retroviral vector and an ecotropic packaging plasmid as described elsewhere (24, 29, 32). Following retroviral infection, cells were selected with G418 at a dose of 800 μg/ml and maintained in the presence of G418 at a dose of 400 μg/ml. Oncogenic transformation was assayed as colony formation in culture medium containing 0.3% agarose as described (44). Transformed colonies were visualized with Wright-Giemsa stain 10 to 14 days later and were quantified using ImageJ.


Cultures of Rat-2 cells were washed twice with phosphate-buffered saline and lysed in 50 mM tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1.0% sodium deoxycholate, 0.1% SDS, and 1% Triton X-100 supplemented with 2.5 mM sodium orthovanadate, 25 mM sodium fluoride, benzonase (60 U/ml) (Novagen), and protease inhibitors (cOmplete protease inhibitor tablets; Sigma). Protein concentrations were determined with the Bradford protein assay (Bio-Rad). Cell lysates were then immunoprecipitated for Fgr (Cell Signaling Technology, 2755S) or Hck (Santa Cruz Biotechnology, sc-1428), and aliquots were separated by SDS-PAGE. Proteins were transferred to nitrocellulose membranes and probed with the following primary antibodies for 16 hours at 4°C: Fgr (1:1000; Santa Cruz Biotechnology, sc-514251), Hck (1:1000; Cell Signaling Technology, 14643S), phospho-Src pTyr416 (1:500; EMD Millipore, 05-677), phospho-Src pTyr527 (1:500; Cell Signaling Technology, 2105), and phospho-Hck pTyr521 (1:500; Thermo Fisher, PA5-37592). After washing with tris-buffered saline (three washes of 15 min each), blots were incubated for 1 hour at room temperature with secondary antibodies conjugated to LI-COR infrared dyes with the appropriate species specificity (1:10,000; LI-COR, 925-32210, 925-32211, 925-68020, or 925-68021). Blots were then developed and imaged using a LI-COR Odyssey instrument and quantified using the LI-COR Image Studio Lite software.

Expression and purification of near–full-length Src family kinases

Recombinant near–full-length human Fgr, Hck, and Src were expressed with an N-terminal 6-His purification tag replacing the unique domain, and the C-terminal tail sequence was modified to Tyr527-Glu-Glu-Ile (YEEI). This YEEI modification facilitates tail phosphorylation and SH2 engagement in Sf9 insect cells, allowing for purification of each kinase in the tail-phosphorylated form observed in previous crystal structures (20, 22, 45). Each kinase coding sequence was subcloned into the pFastBac1 vector and used to generate a recombinant bacmid by transforming DH10Bac Escherichia coli (Thermo Fisher). Recombinant baculoviruses were then produced by transfecting Sf9 insect cells with each recombinant bacmid. Both Hck-YEEI and Src-YEEI were coexpressed with Yersinia pestis YopH, a phosphatase that dephosphorylates the activation loop tyrosine to suppress kinase activity as described previously (22). All three kinases were purified using a combination of ion-exchange, immobilized metal affinity chromatography, and gel filtration as described elsewhere (22). MS confirmed the theoretical mass of each intact kinase, and MS analysis of peptic peptides from each kinase revealed a single phosphate group on C-terminal tail Tyr527; no Tyr416 phosphorylation was observed in peptides derived from the activation loop (fig. S1) (22).

In vitro kinase assays

The FRET-based Z-LYTE kinase assay (Thermo Fisher) was used to determine the effect of SH3 and SH2 domain displacement on the kinase activity of recombinant near–full-length Fgr-YEEI, Src-YEEI, and Hck-YEEI. This assay is based on a synthetic FRET-peptide substrate (Tyr-2) tagged with coumarin and fluorescein on its N and C termini, respectively. Reactions are developed with a protease that selectively cleaves the nonphosphorylated peptide, interrupting FRET. ATP concentrations were set to the Km values for each kinase, and the kinase inputs were adjusted to give 20% of maximal substrate phosphorylation. Kinases were preincubated with peptide ligands for the SH3 (VSL12), SH2 (pYEEI), or both domains (pFAK) over a range of peptide/kinase ratios before the assay. Data were normalized to basal kinase activity observed in the absence of each peptide, and EC50 values were calculated by nonlinear regression analysis of the resulting concentration-response curves using GraphPad Prism software.

Recombinant SH2 and SH3 regulatory domain expression and purification

The coding regions of the SH3, SH2, and dual SH3-SH2 domains of human Hck, Src, and Fgr were polymerase chain reaction–amplified and subcloned into the bacterial expression vector, pET21a. These constructs resulted in the expression of each domain with an additional LEHHHHHH sequence fused to the C terminus for purification. E. coli BL21 Star (DE3) cells (Thermo Fisher) were transformed with each plasmid, and cultures were induced with 1 mM isopropyl-β-d-thiogalactopyranoside for 18 hours at 25°C. Soluble regulatory domain proteins were purified from clarified cell extracts by ion-exchange and immobilized metal chromatography as described (22), and the molecular weight and purity of each recombinant protein were confirmed by MS.

Surface plasmon resonance

Biotinylated VSL12, pFAK, and pYEEI peptides were immobilized on a streptavidin biosensor chip (Reichert) in HBS-EP running buffer [10 mM Hepes (pH 7.4), 150 mM NaCl, 3 mM EDTA, and 0.05% (v/v) Surfactant P20]. Recombinant SH3 and SH2 domain proteins were then injected in triplicate over a range of concentrations (0.01 to 3.0 μM) at a rate of 30 μl/min for 1 min until equilibrium was reached, followed by dissociation in HBS-EP for 2 min. Kinetic rate constants were calculated from reference-corrected sensorgrams using the TraceDrawer software and were best fit by a 1:1 Langmuir binding model. Equilibrium Kd values were calculated from the resulting kinetic rate constants (koff/kon). All SPR data were collected using a Reichert four-channel SPR instrument.

Intact mass analysis of Fgr-YEEI

Near–full-length Fgr-YEEI (200 pmol) was injected onto a 2 mm × 20 mm analytical guard column (Alltech) packed with POROS 20-R2 reversed-phase medium. Desalting was performed by manually flushing 1 ml of 0.05% trifluoroacetic acid over the column. The protein was eluted from the column using a Shimadzu prominence high-performance liquid chromatography (HPLC) system (LC-20AD) with a 3-min 15 to 70% acetonitrile gradient at a flow rate of 50 μl/min. The eluent was directed into a Waters Xevo G2 mass spectrometer equipped with a standard electrospray source for mass analysis. Average protein mass was determined using the MassLynx v.4.1 software (Waters).

Hydrogen-deuterium exchange mass spectrometry

Recombinant purified proteins were diluted in 20 mM tris-HCl (pH 8.3), containing 100 mM NaCl and 3 mM dithiothreitol. Each protein (50 pmol) was diluted 15-fold in the identical buffer prepared with D2O. Labeling reactions were quenched with equal volumes of ice-cold acid phosphate buffer [100 mM potassium phosphate (pH 2.5)] at various times. Quenched samples were subjected to online pepsin digestion at 15°C and Ultra Performance Liquid Chromatography (UPLC) peptide separation at 0°C using a Waters nanoACQUITY system with HDX technology (46). A stainless steel column (2.1 mm × 50 mm) packed with POROS 20AL resin coupled to porcine pepsin (Sigma) was used for digestion. Peptic peptides were trapped and desalted on a VanGuard Pre-Column trap (2.1 mm × 5 mm; ACQUITY UPLC BEH C18, 1.7 μm) for 3 min and separated on an ACQUITY UPLC HSS T3 column (1.8 μm, 1.0 mm × 50 mm) with a 5 to 35% acetonitrile gradient over 6 min at a flow rate of 65 μl/min. Peptides were identified from triplicate undeuterated samples using Waters MSE and the Waters Protein Lynx Global Server. Deuterium incorporation was analyzed using Waters DynamX software. Fractional deuterium uptake in each peptide (fig. S5) is reported as a percentage and is calculated by dividing the uptake (in daltons) by the number of exchangeable backbone amide hydrogens in the peptide (peptide length minus 1 for the N-terminal residue and minus the number of prolines in the sequence) as shown in the following equation: Fractional uptake = uptake in Da/(n − 1 − #Pro), where n is the peptide length. Labeling reactions were performed in duplicate with a single preparation of each protein; then, a second biological replicate of Fgr (another labeling duplicate) was performed, with the same result. The error of measuring deuteration with this system was ±0.20 Da; therefore, deuterium differences larger than 1.0 Da were considered meaningful.

Retroviral transduction of myeloid cells and colony-forming assays

The human GM-CSF–dependent myeloid leukemia cell line TF-1 (38) was obtained from the American Type Culture Collection and maintained in RPMI 1640 supplemented with 10% FBS, antibiotic-antimycotic, and recombinant human GM-CSF (1000 pg/ml; Thermo Fisher, PHC2015). The full-length human Fgr coding sequence was subcloned into pMSCVpuro (Clontech), and retroviral stocks were produced by cotransfection of 293T cells with this retroviral vector and an amphotropic packaging plasmid as described (8, 47). TF-1 cells (106) were incubated with 5 ml of viral stock in the presence of polybrene (4 μg/ml; Sigma) and centrifuged at 3000g for 4 hours at room temperature to promote viral transduction. After infection, cells were washed, transferred to regular medium for 24 hours, and then put under puromycin selection (3 μg/ml) for 14 days. After selection, cells were maintained in puromycin (1 μg/ml). Assessment of Fgr activation loop and tail tyrosine phosphorylation in TF-1 cells was assessed in Fgr immunoprecipitates as described above for Rat-2 fibroblasts. Colony formation was assessed by suspending cells in growth medium supplemented with 0.33% agarose, followed by overlay on preset bottom layers of 0.5% agarose.

Statistical analysis

Data are presented as means ± SD for Rat-2 colony-forming assays and for quantitative immunoblotting experiments or means ± SEM for TF-1 myeloid cell colony-forming assays. Significant differences between control and experimental groups were determined by unpaired Student’s t test (GraphPad Prism 7.1, GraphPad Software Inc.) with statistical significance set at the level of P < 0.05.


Fig. S1. Analysis of intact Fgr-YEEI and phosphopeptides by MS.

Fig. S2. In vitro kinase assay for recombinant near–full-length Fgr, Hck, and Src activity.

Fig. S3. SPR analysis of Fgr-YEEI interaction with peptide ligands for the SH3 and SH2 domains.

Fig. S4. Deuterium uptake by peptic peptides derived from near–full-length Fgr and Src compared to their isolated SH3-SH2 domains.

Fig. S5. Fractional deuterium uptake by Src-YEEI and Fgr-YEEI.


Funding: This work was supported by grants from the NIH (CA185702 to T.E.S. and F32 GM113356 to H.R.D.) and a research collaboration with the Waters Corporation (to J.R.E.). R.K.P. was supported by the NIH Pharmacology and Chemical Biology Training Program Grant T32 GM08424. K.S. is supported by a China Scholarship Council award through the Tsinghua University School of Medicine. Author contributions: T.E.S., H.R.D., and J.R.E. designed the project. K.S., J.A.M., R.K.P., and H.S. performed experiments. K.S. and T.E.S. wrote the manuscript with editorial input from all other authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Navigate This Article