Mutant and Wild-Type Isocitrate Dehydrogenase 1 Share Enhancing Mechanisms Involving Distinct Tyrosine Kinase Cascades in Cancer

The terms metabolic “reprogramming” and “rewiring” have emerged to describe the increasingly better understood metabolic changes observed in cancer cells (1, 2). From a definitional perspective, “metabolic reprogramming” represents “software changes” in cancer cells and describes metabolic alterations that are normally induced by growth factors in proliferating cells but are hijacked by oncogenic signals, whereas “metabolic rewiring” represents “hardware changes” and describes metabolic alterations due to neofunctions of oncogenic mutants, which are not found in normal cells (3). For example, oncogenic signals reprogram cancer cells in an acute manner involving diverse post-translational modifications of metabolic enzymes that also exist in proliferating normal cells (4). The identification of mutations in isocitrate dehydrogenase (IDH) 1 and 2 in glioma and acute myeloid leukemia (AML) represents a rewiring because the mutations confer a neofunction to IDH1/2 to produce the oncometabolite 2-hydroxyglutamate (2-HG) to regulate cancer epigenetics, which is not found in normal cells harboring wild-type (WT) IDH1/2 (5–8). We previously reported that oncogenic BRAF^V600E rewires the ketogenic pathway to allow cancer cells to benefit from ketone body acetoacetate-promoted BRAF^V600E–MEK1 binding, which is not found in cells expressing BRAF WT (3). Thus, clearly distinguishing and characterizing metabolic “reprogramming” and “rewiring” in cancer cells offers apparent advantages to inform therapy development because targeting rewiring (e.g., IDH mutant inhibitors) in cancer cells will have minimal toxicity to normal cells.

IDH1 and IDH2 are two highly homologous members of the IDH family of metabolic enzymes and are located in the cytoplasm and mitochondria, respectively. IDH1/2 form homodimers and convert isocitrate to α-ketoglutarate (αKG) with the reduction of NADP⁺ to NADPH (9). αKG is a key intermediate in the Krebs cycle and glutaminolysis, an important nitrogen transporter, and a ligand for αKG-dependent enzymes, including histone demethylases such as JHD1 and methylcytosine dioxygenase enzyme TET2 (10). NADPH not only fuels macromolecular biosynthesis, such as lipogenesis, but also functions as a crucial antioxidant to quench the reactive oxygen species (ROS) produced during rapid proliferation of cancer cells, which is important for the maintenance of cellular redox homeostasis to protect against toxicity of ROS and oxidative DNA damage (11). Thus, IDH1/2 are important for many metabolic processes in cells including bioenergetics, biosynthesis, and redox homeostasis. Moreover, recent evidence demonstrates that IDH1/2 play an important role in reductive carboxylation that is enhanced in cells under hypoxia, allowing the generation of isocitrate/citrate from αKG and glutamine, which is in particular important in cancer cells for producing citrate and acetyl-CoA that are essential for lipid synthesis during tumorigenesis, as well as reducing mitochondrial ROS to sustain redox homeostasis during anchorage-independent growth (12, 13).

Missense mutations of R132 in the enzyme active site of IDH1 were identified in patients with glioblastoma (GBM) and in AML cases (5–7, 14, 15), and corresponding IDH2 R172 mutations as well as a novel R140Q mutant repeatedly occur in patients with AML (14, 16, 17). Overall, IDH1/2 mutations are identified in >75% of grade 2/3 glioma and secondary GBM cases and in >20% of AML cases. IDH mutations were also identified in other cancer types such as chondrosarcoma and cholangiocarcinoma (9). IDH mutations are heterozygous events, resulting in loss of function of WT IDH1 enzyme activity but a gain of function to mutant IDH1, allowing NADPH-dependent reduction of αKG to produce the oncometabolite 2-HG. 2-HG competitively inhibits the function of αKG-dependent enzymes, such as TET2, which in turn causes epigenetic dysregulation including DNA hypermethylation in both GBM and AML, and consequent block of differentiation (18). In AML, IDH1 R132, IDH2 R172, and IDH2 R140 mutations are mutually exclusive, and although IDH1 R132 mutations were detected in patients with AML with mutations in NPM1, FLT3, CEBPA, or NRAS, co-occurrence of IDH1 mutations and FLT3 internal tandem duplication (FLT3-ITD) is less common (19).

The pathogenic role of IDH mutations suggested mutant IDH proteins as promising therapeutic targets. Small-molecule inhibitors, including AG-120, enasidenib (AG-221/CC-90007), AG-881, IDH305, and FT-2102, that selectively target mutant IDH1 or IDH2 have been developed and evaluated in preclinical and clinical studies as single agents and in combination with other anticancer agents. These inhibitors were designed to bind to the catalytic site of mutant IDH proteins and inhibit the enzyme activity of mutant IDH to convert αKG to 2-HG (20). In addition, recent reports demonstrate the ability of single-agent IDH2 R140 mutant inhibitor AG-221 to reverse the DNA hypermethylation and differentiation block induced by the IDH2 mutant in leukemia cells, and improve efficacy when given in combination with a FLT3 tyrosine kinase (TK) inhibitor in IDH2 mutant–expressing leukemia cells (21, 22). It remains unknown how oncogenic signals such as oncogenic TKs, including FLT3, regulate WT IDH1 and/or mutant IDH1, what the regulatory mechanism is, and whether WT and mutant IDH1 share the same regulatory mechanism in the corresponding related cancers.

To understand the pathogenic link between mutant FLT3 and IDH1, we treated primary leukemia cells from representative patients with AML with the FLT3 inhibitor quizartinib and/or the IDH1 R132 mutant inhibitor AG-120. We found that single-agent treatment for 24 hours resulted in decreased cell viability of IDH1 R132H mutant–expressing primary leukemia cells harboring concurrent FLT3-ITD mutation (left, Fig. 1A; Supplementary Fig. S1A), or expressing FLT3 WT (right, Fig. 1A; Supplementary Fig. S1A), whereas combined treatment resulted in further decreased cell viability of these cells. Surprisingly, treatment with either AG-120 or quizartinib resulted in decreased mutant IDH1 activity, which was further reduced upon combined treatment in all of the tested primary AML cells, accompanied by decreased tyrosine phosphorylation levels of IDH1 proteins (Fig. 1A; Supplementary Fig. S1A). Similar results were obtained using primary leukemia cells from patients with AML expressing IDH1 WT and FLT3-ITD or FLT3 WT, where treatment with quizartinib resulted in decreased IDH1 enzyme activity (Fig. 1B). These data together suggest an intrinsic link between FLT3 and IDH1 regardless of the mutational state of both proteins, likely involving FLT3-dependent tyrosine phosphorylation and activation of IDH1.

We next performed in vitro kinase assays coupled with mutant IDH1 activity assays using purified, recombinant IDH1 R132H mutant protein (rIDH1 R132H) incubated with recombinant active forms of diverse TKs, including rFLT3, rFGFR1, rSrc, rPDGFR, rEGFR, rMET, rKIT, or rJAK2. We found that purified rIDH1 R132H mutant protein demonstrated basal-level enzyme activity, whereas all of the tested TKs were able to directly phosphorylate purified rIDH1 R132H mutant, leading to increased mutant IDH1 activity (Fig. 1C; Supplementary Fig. S1B). We next performed mutational analysis and generated diverse phospho-deficient Y→F mutants of IDH1 R132H (top, Fig. 1D) based on publicly available data (https://www.phosphosite.org/proteinAction.action?id=10630&showAllSites=true), which identified phosphorylated tyrosine residues of IDH1 in human cancer cells. We found that mutations at Y139 and Y219 intrinsically abolished mutant IDH1 activity, whereas substitution of Y391 with phenylalanine resulted in the abolishment of FLT3-enhanced activation of IDH1 R132H mutant (Fig. 1D).

However, we found that a group of TKs (which we termed group I), including EGFR, JAK2, ABL1, PDGFR, and FGFR1, were still able to phosphorylate and enhance activation of IDH1 R132H mutant with Y391F mutation (Fig. 1E, left and 1F), whereas FLT3 or Src (termed group II)–enhanced activation of IDH1 R132H/Y391F mutant was abolished (Fig. 1E, right). Only the Y42F mutation abolished FGFR1-enhanced activation of mutant IDH1 in the coupled kinase and mutant IDH1 activity assays (Fig. 1F). Similar results were obtained using purified rIDH1 WT proteins incubated with rFLT3 (group II) or rFGFR1 (group I), where Y391F or Y42F mutations exclusively abolished FLT3 or FGFR1-enhanced activation of IDH1 WT, respectively (Supplementary Fig. S1C). Further studies confirmed that group I TKs, including EGFR and JAK2 (Fig. 1G, left) as well as ABL1, FGFR3, c-MET, and PDGFR (Supplementary Fig. S2A), enhanced activation of IDH1 R132H mutant by phosphorylating Y42, whereas group II TK Src enhanced activation of mutant IDH1 through Y391 phosphorylation (Fig. 1G, right). Similar results were obtained using IDH1 WT proteins incubated with diverse TKs (Supplementary Fig. S2B–S2C). These data demonstrate that two groups of TKs enhance activation of IDH1 WT or R132H mutant in the same manner through preferential phosphorylation of Y42 or Y391 of IDH1 despite mutational status, suggesting that WT and mutant IDH1 might share the same regulatory mechanisms involving tyrosine phosphorylation.

We next sought to elucidate the distinct mechanisms underlying Y42 and Y391 phosphorylation–dependent enhancement of IDH1 activation. Further studies confirmed that purified dimeric and monomeric IDH1 R132H mutant or WT proteins represented active and inactive forms of IDH1, respectively (Fig. 2A, left and right, respectively), and IDH1 R132H mutant or WT enzyme activity correlated with increasing levels of dimer in different compositions of purified monomeric and dimeric IDH1 proteins (Fig. 2B, left and right, respectively). In addition, incubation with the group I TKs rFGFR1, rMET, or rEGFR but not group II TKs rFLT3 or rSrc resulted in increased dimer formation of IDH1 R132H mutant or WT proteins assessed by Western blot after cross-linking (Fig. 2C, left and right, respectively), suggesting that Y42 phosphorylation by group I TKs may contribute to IDH1 dimer formation. This was further confirmed by incubation of diverse group I or II TKs with IDH1 R132H mutant proteins with or without Y42F or Y391F mutation, where FGFR1 (group I) but not FLT3 (group II; Fig. 2D, top), as well as EGFR and MET (group I) but not Src (group II; Supplementary Fig. S3A) were able to promote the homerdimerization of IDH1 R132H mutant and R132H/Y391F but not R132H/Y42F proteins through tyrosine phosphorylation. Similar results were obtained using IDH1 WT, Y42F, and Y391F proteins incubated with diverse group I and II TKs (Fig. 2D, bottom; Supplementary Fig. S3B).

We purified monomeric and dimeric IDH1 proteins using a sucrose density ultracentrifugation approach (Fig. 2E, top). We found that the majority of recombinant IDH1 R132H proteins were monomers, whereas phosphorylation by FGFR1 (group I) resulted in a shift to dimer formation (Fig. 2E, middle). This shift was abolished when using rIDH1 R132H/Y42F mutant but retained using rIDH1 R132H/Y391F mutant (Fig. 2E, bottom two plots). Similar results were obtained using rIDH1 WT, Y42F, and Y391F proteins incubated with FGFR1 (Supplementary Fig. S3C).

We next determined the underlying mechanism by which Y42 phosphorylation–enhanced dimer formation contributes to IDH1 activation. We separated monomeric and dimeric IDH1 R132H mutant or WT proteins on native gels (Fig. 3A, bottom). The gels loaded with IDH1 R132H mutant or WT proteins were incubated with ¹⁴C-labeled αKG or ³H-labeled isocitrate, respectively. The binding ability of ¹⁴C-αKG or ³H-isocitrate to monomeric/dimeric IDH1 R132H mutant or IDH1 WT proteins, respectively, was assessed by scintillation counting of excised gel bands containing radiolabeled IDH1 monomers or dimers (Fig. 3A). The results of this “overlay” assay suggested that αKG or isocitrate as substrates demonstrated higher binding ability to dimeric IDH1 R132H mutant or IDH1 WT proteins, respectively, compared with corresponding monomers of IDH1 proteins (Fig. 3A, left and right, respectively), suggesting IDH1 dimer formation promotes substrate binding.

Moreover, incubation with group I TKs, including rMET and rEGFR, but not group II TKs, including rFLT3 and rSrc, promoted ¹⁴C-αKG binding to IDH1 R132H mutant, whereas control IDH1 WT showed minimal binding ability to ¹⁴C-αKG, which was not altered when phosphorylated by group I TK rMET or group II TK rFLT3 (Fig. 3B, left). In contrast, incubation with group I TKs, including rFGFR1, rEGFR, rMET, and rPDGFR, but not group II TKs, including rFLT3 and rSrc, promoted ³H-isocitrate binding to IDH1 WT, whereas control IDH1 R132H mutant showed minimal binding ability to ³H-isocitrate, which was not altered when phosphorylated by group I TK rFGFR1 or group II TK rFLT3 (Fig. 3B, right).

This was further confirmed by incubation of diverse group I or II TKs with IDH1 R132H mutant proteins with or without Y42F or Y391F mutation, where FGFR1 (Fig. 3C, left), EGFR, MET, and PDGFR (Supplementary Fig. S4A, top plots; group I TKs) but not FLT3 (Fig. 3C, right) or Src (Supplementary Fig. S4A, lower; group II TKs) were able to promote ¹⁴C-αKG binding to IDH1 R132H mutant and R132H/Y391F but not R132H/Y42F proteins through tyrosine phosphorylation, whereas IDH1 WT as a negative control showed minimal binding ability to ¹⁴C-αKG, which was not altered when phosphorylated by group I or II TKs (Fig. 3C; Supplementary Fig. S4A). Similar results were obtained using IDH1 WT, Y42F, and Y391F proteins incubated with diverse group I and II TKs in the presence of ³H-isocitate (Fig. 3D; Supplementary Fig. S4B).

In multiple structures of IDH1, the location of Y391 is proximal to the NADP⁺ binding site (approximately 10Å); we thus hypothesized that Y391 phosphorylation may enhance IDH1 activation by affecting cofactor NADP⁺ binding. To test this hypothesis, we performed an NADP⁺ binding experiment using Blue Sepharose CL-6B (Sigma-Aldrich). CL-6B mimics NADP⁺ and is a pseudo-affinity ligand of many dehydrogenases using NADP⁺ as a substrate. We found that incubation of group II TKs, including rFLT3 or rSrc, but not group I TKs, including rFGFR1, rMET, or rEGFR, resulted in a significant increase in the amount of IDH1 protein bound to the CL-6B agarose beads, indicating increased binding between IDH1 and NADP⁺ (Fig. 3E), suggesting the Y391 phosphorylation by group II TKs may contribute to cofactor NADP⁺ binding to IDH1. This was further confirmed by incubation of diverse group II or I kinases with IDH1 WT, Y42F, or Y391F mutant proteins, where Src and FLT3 (group II TKs; Fig. 3F), but not FGFR1, EGFR, or MET (group I; Fig. 3G), were able to promote binding of IDH1 WT and Y42F mutant but not Y391F mutant proteins to CL-6B agarose beads.

Further kinetics studies confirmed that, in the presence of increasing concentrations of isocitrate as a substrate, phosphorylation of IDH1 WT and Y391F mutant but not Y42F mutant by FGFR1 (group I TK) resulted in increased Vmax values (Fig. 4A, top two plots) with decreased Km values for isocitrate (Fig. 4A, bottom two plots), suggesting that Y42 phosphorylation by group I TKs enhances IDH1 WT activation by promoting substrate binding. In contrast, in the presence of increasing concentrations of NADP⁺ as a cofactor, although phosphorylation of IDH1 WT and Y391F mutant but not Y42F mutant by FGFR1 resulted in increased Vmax values (Fig. 4B, top two plots), Km values for NADP⁺ as a cofactor were not altered among IDH1 variants (Fig. 4B, bottom two plots), suggesting that Y42 phosphorylation does not affect cofactor binding. Furthermore, in the presence of increasing concentrations of isocitrate as a substrate, despite the increased Vmax values due to phosphorylation of IDH1 WT and Y42F mutant but not Y391F mutant by FLT3 (group II; Fig. 4C, top two plots), the Km values for isocitrate remained unaltered (Fig. 4C, bottom two plots), whereas in the presence of increasing concentrations of NADP⁺ as a cofactor, phosphorylation of IDH1 WT and Y42F mutant but not Y391F mutant by FLT3 resulted in increased Vmax values (Fig. 4D, top two plots) with decreased Km values for NADP⁺ as a cofactor (Fig. 4D, bottom two plots). These data suggest that Y391 phosphorylation by group II TKs enhances IDH1 WT activation by promoting cofactor but not substrate binding.

Similar results were obtained using IDH1 R132H mutant proteins phosphorylated by FGFR1 or FLT3 in the presence of increasing concentrations of αKG or NADPH (Fig. 4E–H). These results together also demonstrate that the Y42F and Y391F mutations do not affect intrinsic enzyme properties of IDH1 WT or IDH1 R132H mutant.

In order to confirm whether TKs enhance IDH1 activation predominantly through tyrosine phosphorylation, we performed an in vitro kinase assay followed by protein Tyr phosphatase (PTP) treatment. We found that rIDH1 WT but not Y42F phosphorylated by FGFR1 (group I) showed enhanced enzyme activity, which was abolished by PTP treatment (Supplementary Fig. S5A, left). In addition, rIDH1 WT but not Y391F phosphorylated by FLT3 (group II) also showed enhanced activation, which was eliminated by PTP (Supplementary Fig. S5A, right). Consistent with these findings, treatment of IDH1 immunoprecipitates from MOLM-14 cells with PTP resulted in decreased IDH1 enzyme activity with reduced tyrosine phosphorylation (Supplementary Fig. S5B, top left), and PTP treatment of MOLM-14 cell lysates resulted in a decrease in the amount of dimeric IDH1 proteins (Supplementary Fig. S5B, bottom left). Similar results were obtained using IDH1 immunoprecipitates from MOLM-14 cells treated with the FLT3 inhibitor TKI258 (Supplementary Fig. S5B, right).

We sought to determine whether Y42 and Y391 can be phosphorylated in monomeric and/or dimeric IDH1. Interestingly, we found that Y42 phosphorylation by FGFR1 (group I) occurred only in purified monomeric IDH1 WT protein, leading to increased dimer formation from IDH1 monomers accompanied with enhanced enzyme activity, whereas purified dimeric IDH1 WT protein treated with FGFR1 was not phosphorylated at Y42 with unaltered dimer amount and enzyme activity (Fig. 5A). In contrast, Y391 phosphorylation by FLT3 (group II) occurred in both purified monomeric and dimeric IDH1 WT proteins (Fig. 5B). However, Y391 phosphorylation of IDH1 monomers did not promote dimer formation or enhance enzyme activity, whereas Y391 phosphorylation of IDH1 dimers enhanced enzyme activity despite unaltered dimer amount, which is likely due to increased cofactor NADP⁺ binding. Indeed, incubation of rFLT3 resulted in increased Y391 phosphorylation of both monomeric and dimeric IDH1 WT proteins, with a significant increase in the amount of both monomeric and dimeric IDH1 proteins bound to the CL-6B agarose beads, indicating increased binding between Y391-phosphorylated monomeric and dimeric IDH1 and NADP⁺ (Fig. 5C). Similar results were obtained using purified monomeric and dimeric IDH1 R132H proteins incubated with rFGFR1 or rFLT3 (Fig. 5D and E).

We next determined whether Y42 phosphorylation affects dimer formation from monomers and/or dimer disruption to monomers. We found that purified, recombinant monomeric IDH1 WT or R132H mutant proteins spontaneously form dimers in a time-dependent manner (Fig. 5F, right halves of left and right, respectively), which explains the basal levels of enzyme activity of nonphosphorylated IDH1 WT or R132H mutant proteins, whereas Y42 phosphorylation of monomeric IDH1 WT or R132H mutant proteins by FGFR1 resulted in enhanced dimer formation (Fig. 5F, left halves of left and right, respectively). In addition, purified, recombinant dimeric IDH1 WT or R132H mutant proteins spontaneously disrupt to form inactive monomers in a time-dependent manner (Fig. 5G, right halves of left and right, respectively), whereas dimeric IDH1 WT or R132H mutant proteins purified from IDH1 proteins incubated with rFGFR1, which contained Y42-phosphorylated dimers that were likely formed from Y42-phosphorylated monomers, showed attenuated dimer disruption to form monomeric IDH1 proteins (Fig. 5G, left halves of left and right, respectively). These data together suggest that Y42 phosphorylation occurs in IDH1 monomers, which promotes dimer formation, and once Y42-phosphorylated monomers form dimers, Y42 phosphorylation attenuates dimer disruption.

IDH1 mutations exist as heterozygous mutations in AML (5–7, 14, 15), and R132H/WT heterodimers were suggested to be more efficient than R132H homodimers in 2-HG (23). To determine the effect of Y42 phosphorylation on IDH1 heterodimers, we purified monomers of FLAG-tagged IDH1 and HA-tagged IDH1 R132H. The monomers were mixed in the presence and absence of recombinant FGFR1. Theoretically, the mixture contains FLAG-IDH1 and HA-IDH1 R132H monomers, FLAG-IDH1 and HA-IDH1 R132H homodimers, and heterodimers of FLAG-IDH1/HA-IDH1 R132H (Fig. 5H; top left). FLAG pulldown enriched FLAG-IDH1 monomers, FLAG-IDH1 homodimers, and FLAG-IDH1/HA-IDH1 R132H heterodimers (Fig. 5H; middle of left). The FLAG pulldown samples were applied either to native gel followed by Western blotting (Fig. 5H, top right), or to HA pulldown to further purify the heterodimers (Fig. 5H; bottom left). As shown in Fig. 5H, top right, HA blotting detected the FLAG-IDH1/HA-IDH1 R132H heterodimers, and the results suggested that phosphorylation by rFGFR1 resulted in increased heterodimer formation of IDH1/R132H (left) and IDH1 Y391F/R132H Y391F (right), but not IDH1 Y42F/R132H Y42F (middle). Similar results were obtained using FLAG-IDH1/HA-IDH1 R132H heterodimers purified by sequential FLAG-HA pulldowns detected by FLAG blotting (Fig. 5H, bottom right). Moreover, as shown in Fig. 5I, purified FLAG-IDH1/HA-IDH1 R132H heterodimers spontaneously disrupt to form monomers in a time-dependent manner, whereas heterodimers formed by FGFR1-phosphorylated monomers of FLAG-IDH1 and HA-IDH1 R132H (left) or monomers of FLAG-IDH1 Y391F and HA-IDH1 R132H Y391F (right), but not monomers of FLAG-IDH1 Y42F and HA-IDH1 R132H Y42F (middle), showed attenuated dimer disruption to monomers.

We next sought to determine whether IDH1 R132H mutant and IDH1 WT may be phosphorylated at both or either of the Y42 and Y391 sites in diverse cancer cells expressing different oncogenic TKs. We found that both Y42 and Y391 were phosphorylated in lung cancer H1299 cells expressing FGFR1 and leukemia cell lines including K562 cells expressing BCR–ABL, HEL cells expressing JAK2 V617F mutant, and MOLM-14 cells expressing FLT3-ITD (Supplementary Fig. S5C and S5D). Interestingly, treatment with distinct TK inhibitors to target group I TKs, including FGFR1 (TKI258), BCR–ABL (imatinib), and JAK2 V617F (AG490; Supplementary Fig. S5C), or group II TK FLT3-ITD (quizartinib; Supplementary Fig. S5D) in corresponding cancer/leukemia cells resulted in decreased phosphorylation levels of both Y42 and Y391, suggesting these oncoge nic/leukemogenic group I or II TKs activate partner kinases in group II or I, respectively, to achieve both Y42 and Y391 phosphorylation to further activate IDH1. Indeed, we found that treatment with group II TK inhibitors PP2 (Src) but not quizartinib (FLT3) reduced Y391 phosphorylation of IDH1 in H1299 and K562 cells expressing FGFR1 and BCR–ABL, respectively, which are group I TKs (Fig. 6A, left), suggesting that group I TKs may activate Src, but not FLT3, to achieve Y391 phosphorylation. This was further confirmed by results showing that treatment with group I TK inhibitors TKI258 (FGFR1) or imatinib (BCR–ABL) resulted in decreased Src phosphorylation levels in H1299 and K562 cells, respectively (Fig. 6A, bottom right), and reduced activity of Src kinases in an in vitro kinase assay using immunoprecipitated Src incubated with myelin basic protein (MBP) as an exogenous substrate, where Src activity was assessed by the amount of phospho-MBP (Fig. 6A, top right).

Similar results were obtained using two FLT3-ITD–expressing leukemia cell lines, including MOLM-14 and MV4-11, where treatment with the FLT3 inhibitor quizartinib resulted in decreased phosphorylation levels of both Y42 and Y391 of IDH1, whereas JAK2 inhibitors, including AG490 and ruxolitinib, reduced only Y42 phosphorylation (Fig. 6B, left; Supplementary Fig. S5E). In contrast, control inhibitors, including imatinib and PP2, did not affect Y42 or Y391 phosphorylation of IDH1 (Fig. 6B, left). These data suggest that group II TK FLT3 may activate group I TK JAK2 to achieve Y42 phosphorylation of IDH1. This is consistent with results showing that treatment with the FLT3 inhibitor quizartinib resulted in decreased JAK2 phosphorylation levels in MOLM-14 cells (Fig. 6B, bottom right), and reduced JAK2 kinase activity in an in vitro kinase assay using immunoprecipitated JAK2 incubated with MBP as an exogenous substrate (Fig. 6B, top right). Similar results were obtained using siRNAs to specifically knock down different TKs in MOLM-14 cells, where FLT3 siRNA resulted in reduced phosphorylation levels of both Y42 and Y391 of IDH1, whereas JAK2 siRNA decreased only Y42 phosphorylation, but both siRNAs targeting ABL and Src did not alter IDH1 phosphorylation (Fig. 6C). Further in vitro kinase assays using immunoprecipitated group I TK JAK2 or group II TK Src (Supplementary Fig. S5F, left and right, respectively) pretreated with protein tyrosine phosphatase PTP1B to remove tyrosine phosphorylation prior to incubation with group II TK rFLT3 or group I TKs rABL, rFGFR1, or rJAK2, respectively, showed that FLT3 directly phosphorylated JAK2, whereas ABL, FGFR1, and JAK2 phosphorylated Src. These data supported the proposed distinct group II→I and I→II TK cascades, respectively, which achieve phosphorylation of both Y42 and Y391 of IDH1.

Proliferation of TF-1 erythroleukemia cells depends on the cytokine GM-CSF, and when transformed by expression of IDH1 R132H, TF-1 cells can grow in the absence of GM-CSF and are resistant to erythropoietin (EPO)-induced cell differentiation (24, 25). We generated TF-1 cell lines with stable expression of FLAG-tagged IDH1 R132H or R132H/Y42F, R132H/Y391F, and R132H/Y42F/Y391F mutants by lentiviral transduction. We found that expression of IDH1 R132H resulted in increased GM-CSF–independent cell proliferation, whereas cells expressing R132H/Y42F, R132H/Y391F, and R132H/Y42F/Y391F mutants showed attenuated potential for GM-CSF–independent cell proliferation (Fig. 6D, left), which was partially “rescued” by adding cell-permeable TFMB-(R)-2-HG (24) in the culture media (Fig. 6D, right). Consistent with these findings, compared with TF-1 cells expressing IDH1 R132H, cells expressing diverse phospho-deficient R132H mutants showed attenuated IDH1 mutant enzyme activity (Fig. 6E), 2-HG production (Fig. 6F), histone hypermethylation (Fig. 6G), and resistance to EPO-induced differentiation of TF-1 cells that was assessed using staining with glycophorin A (ref. 24; Fig. 6H; Supplementary Fig. S6A).

We next generated “rescue” H1299 cells with stable knockdown of endogenous IDH1 and stable rescue expression of FLAG-IDH1 WT, Y42F, Y391F, or Y42F/Y391F mutants (Fig. 6I, left). Expression of IDH1 WT reversed the decreased cell proliferation and IDH1 activity of H1299 cells under both normoxia (Supplementary Fig. S6B, left and middle, respectively) and hypoxia (Fig. 6I, left and middle of right plot, respectively), whereas Y→F mutants demonstrated attenuated “rescue” of reduced cell proliferation due to decreased IDH1 activity compared with IDH1 WT. In addition, knockdown of endogenous IDH1 or “rescue” expression of IDH1 variants had no effect on lipogenesis under normoxia (Supplementary Fig. S6B, right), whereas IDH1 knockdown resulted in decreased lipogenesis under hypoxia, which was completely or partially rescued by IDH1 WT or IDH1 Y→F mutants, respectively (Fig. 6I, right of right plot). Consistently, in a xenograft experiment, rescue expression of IDH1 WT reversed the decreased tumor growth rate and masses (Supplementary Fig. S6C, left, and 6J, left, respectively) of injected H1299 cells with stable knockdown of endogenous IDH1, whereas IDH1 Y→F mutants demonstrated attenuated rescue ability with reduced IDH1 activity due to abolishment of Y42 and/or Y391 phosphorylation (Fig. 6J, middle two plots, respectively). In tumor cells, IDH1 or “rescue” expression of IDH1 variants had no effect on lipogenesis under normoxia (Supplementary Fig. S6C, middle), whereas IDH1 knockdown resulted in decreased lipogenesis under hypoxia, which was completely or partially rescued by IDH1 WT or IDH1 Y→F mutants, respectively (Fig. 6J, right). In addition, these tumor growth results correlate with tumor cell proliferation potential assessed by IHC staining of Ki67 (Supplementary Fig. S6C, right two plots). Similar results were obtained using “rescue” MOLM-14 cells expressing Y→F mutants of IDH1 with decreased cell proliferation under both normoxia and hypoxia, with reduced lipogenesis only under hypoxia (Supplementary Fig. S6D–S6F).

Phosphorylation of Y42 and Y391 was detected in human primary leukemia cells from patients with AML harboring IDH1 R132H mutation despite different FLT3 mutational state (Fig. 7A), whereas upregulated phosphorylation levels of Y42 and Y391 of IDH1 WT were detected in primary leukemia cells from representative patients with AML compared with control peripheral blood cells from healthy donors (Fig. 7B). We next confirmed that PTP treatment resulted in decreased enzyme activities of immunoprecipitated IDH1 R132H (Fig. 7C, top left) or IDH1 WT (Fig. 7C; Supplementary Fig. S7A, top right), with reduced phosphorylation levels of Y42 and Y391 (Fig. 7C; Supplementary Fig. S7A, middle), and that PTP treatment of primary AML cell lysates reduced the amount of dimeric IDH1 proteins (Fig. 7C, bottom). Moreover, treatment with the FLT3 inhibitor quizartinib resulted in decreased amounts of IDH1 R132H dimers but increased monomers (Fig. 7D), with reduced 2-HG production (Fig. 7E) in primary leukemia cells from patients with AML expressing mutant IDH1 with either FLT3-ITD or WT. In addition, treatment with quizartinib resulted in decreased Y42 and Y391 phosphorylation of IDH1 R132H mutant (Fig. 7F; Supplementary Fig. S7B) or IDH1 WT (Fig. 7G) in primary leukemia cells from patients with AML expressing either FLT3-ITD or WT, whereas treatment with the JAK2 inhibitor ruxolitinib decreased only Y42 phosphorylation of IDH1 R132H or WT (Fig. 7F and G, right, respectively). Furthermore, IHC staining results showed that relative Y42-phosphorylated IDH1 levels compared with overall IDH1 protein levels were commonly upregulated in diverse primary tumor tissue samples from human patients with GBM, lung, breast, head and neck, prostate, and colon cancers, but not pancreatic cancer, compared with corresponding normal control tissue samples (Fig. 7H; Supplementary Fig. S7C).

Our findings reveal a molecular mechanism shared by WT and mutant IDH1 in related human cancers, which involves distinct TK cascades containing group I and II TKs that preferentially phosphorylate Y42 and Y391, respectively, to fulfill activation of WT and mutant IDH1. Moreover, our finding of tyrosine phosphorylation–dependent IDH1 activation represents a novel isoform-specific activating mechanism for IDH1 in AML, because the functional phosphorylation sites of IDH1 reported in our article, Y42 and Y391, are replaced with F82 and F431, respectively, in IDH2. Our results suggest that spontaneous formation of active IDH1 dimers from monomers and disruption of dimers to form monomers coexist and eventually reach an equilibrium, such that IDH1 dimers but not monomers are able to recruit substrate (isocitrate for WT and αKG for R132H mutant) and thus represent the active portion of total IDH1 proteins (Fig. 7I, top left). Group I TKs preferentially phosphorylate Y42 on monomeric but not dimeric IDH1, and Y42-phosphorylated IDH1 monomers exhibit enhanced ability to form active dimers, leading to increased substrate binding and enzyme activity (Fig. 7I, top left). Moreover, active dimers formed by Y42-phosphorylated monomers showed attenuated disruption to form monomers, suggesting that Y42 phosphorylation has a dual role in enhancing IDH1 activation by promoting dimer formation from Y42-phosphorylated monomers and preventing disruption of Y42-phosphorylated dimers, which represents a mechanism to sustain IDH1 activation in cells (Fig. 7I, top left). In contrast, Y391 phosphorylation by group II TKs occurs in both monomeric and dimeric IDH1 proteins, which is able to promote cofactor (NADP⁺ for WT and NADPH for R132H mutant) binding to IDH1 monomers and dimers, respectively (Fig. 7I, bottom left). Future structural studies of IDH1 monomers and dimers are warranted to decipher the underlying mechanisms by which Y42 phosphorylation contributes to monomer–dimer equilibrium and subsequent substrate binding, whereas Y391 phosphorylation promotes cofactor binding to IDH1 despite monomeric or dimeric status. Moreover, it will be interesting to determine whether one or both molecules of an active dimer need to be phosphorylated at Y42 and/or Y391, or both Y42 and Y391 need to be phosphorylated in one molecule in an active dimer to functionally contribute to IDH1 activation.

In human cancers expressing IDH1 WT, diverse group I TKs activate Src (group II), or FLT3 (group II) activates JAK2 (group I) to achieve phosphorylation of both Y42 and Y391, leading to sustained levels of active dimers and enhanced substrate and cofactor NADP⁺. Phosphorylation-enhanced IDH1 activation is important to provide metabolic advantages to cancer cell proliferation and tumor growth, including reductive carboxylation for lipogenesis under hypoxia (Fig. 7I, middle). In human cancers harboring IDH1 mutations, mutant IDH1 retains such an enhancing mechanism. For example, in AML, FLT3 (group II) activates JAK2 (group I) to achieve phosphorylation at both sites and subsequently full activation of mutant IDH1, leading to 2-HG production that provides an epigenetic advantage to tumorigenesis and tumor growth (Fig. 7I, right). However, because IDH1 mutations exist as heterozygous mutations in AML and glioma cases (5–7, 14, 15), it is believed that homodimers of IDH1 WT and mutant, as well as WT/mutant heterodimers, coexist. It has been suggested that, although WT:R132H heterodimers exhibit almost abolished IDH1 WT enzyme activity (26), IDH1 mutants require IDH1 WT activity to produce 2-HG (27). Moreover, R132H/WT heterodimers demonstrate a significantly lower Km for αKG as substrate compared with R132H homodimers (23). These studies suggest an overall advantage of R132H/WT heterodimers that favors the R132H mutant–driven conversion of αKG to 2-HG. Our results demonstrate that phosphorylation of Y42 similarly enhances the IDH1 mutant/WT heterodimer formation from monomers and attenuates disruption of heterodimers to monomers as Y42 phosphorylation does to homodimers of IDH1 WT and mutant, which further strengthens the importance of IDH1 tyrosine phosphorylation in mutant IDH1-expressing human cancers.

Our studies revealed that a spectrum of tyrosine residues of IDH1 WT and R132H can be phosphorylated. These include the previously reported Y139, which is believed to play a critical role in IDH1 R132H–mediated catalysis of αKG by compensating the increased negative charge on the C2 atom of αKG during reduction of αKG to 2-HG (28). IDH1 R132H/Y139A mutant demonstrates <1% activity (28), which is consistent with our finding that Y139F mutant is catalytically deficient. Future structural and mechanistic studies are warranted to determine whether phosphorylation at Y139 also contributes to IDH1-mediated catalysis. In contrast, unlike Y139F or Y219F mutations that almost abolish IDH1 enzyme activity, Y42F and Y391F mutations do not alter the intrinsic enzymatic properties of IDH1 WT or R132H mutant, which was supported by the diverse kinetic studies.

These results also provide an additional rationale supporting the combination of TK and mutant IDH1 inhibitors for the clinical treatment of IDH1 mutant–expressing human cancers. It is plausible to consider combined therapy using inhibitors of mutant IDH1 and FLT3 to treat patients with AML harboring IDH1 mutations regardless of the mutational status of FLT3, given both FLT3 WT and ITD mutant are able to phosphorylate IDH1 mutant directly and indirectly through activation of JAK2. It was reported that FLT3-ITD is able to induce STAT5 phosphorylation independent of JAK2, where FLT3 may directly phosphorylate STAT5 (29). Thus, it is likely that FLT3 WT or ITD mutant is able to phosphorylate and activate JAK2, which might be dispensable for STAT5 phosphorylation and activation but is critical for the phosphorylation and subsequently enhanced activation of WT and mutant IDH1 in AML leukemia cells. Taken together, our studies elucidate a novel regulatory mechanism of IDH1, which not only provides insight into how signaling affects metabolic function but also informs mechanism-based combination therapeutic studies.

The approval for use of human specimens was given by the Institutional Review Board of Emory University School of Medicine. The clinical samples numbered with UPN were obtained with informed consent with approval by the Emory University Institutional Review Board. The clinical samples from the Memorial Sloan Kettering Cancer Center were approved by the Institutional Review Board of Memorial Sloan Kettering Cancer Center. Patients provided written informed consent in all cases at the time of enrollment. Only samples from patients with leukemia who were not previously treated with chemotherapy or radiotherapy were used. Mononuclear cells of peripheral blood and bone marrow samples from patients with leukemia or healthy donors were isolated using lymphocyte separation medium (Cellgro). Cells were then counted and cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin for further indicated treatments.

TF-1 cells (#CRL-2003, ATCC; purchased in 2018; not authenticated) were cultured in RPMI-1640 medium with 10% FBS and 2 ng/mL recombinant human GM-CSF (R&D Systems). K-562 (#CCL-243, ATCC; purchased in 2015; not authenticated; Mycoplasma tested in November 2018), H1299 (#CRL-5803, ATCC; purchased in 2015; not authenticated; Mycoplasma tested in November 2018), MOLM-14, MV-4-11, and HEL cells were cultured in RPMI-1640 medium with 10% FBS. HEK293T cells were cultured in Dulbecco's Modified Eagle Medium with 10% FBS. Cells were cultured at 37°C with 5% CO₂. MOLM-14, MV-4-11, HEL, and HEK293T cells (not authenticated) were obtained from Dr. Gary Gilliland's laboratory at Brigham and Women's Hospital in 2004.

The shRNA sequence for WT IDH1 is CGAATCATTTGGGAATTGATT. The WT IDH1 and diverse YF mutant plasmids were mutagenized by PCR to generate IDH1 plasmids that contain three silent mutations introduced in the 21 bp IDH1 shRNA target sequence (CGTATCATATGGGAGTTGATT). The mutations were confirmed by sequence analysis.

Stable knockdown of endogenous IDH1 was achieved by using lentiviral vectors harboring shRNA constructs. To be specific, each shRNA construct was cotransfected with psPAX2 packaging plasmid and pMD2.G envelope plasmid (Addgene) into HEK293T cells using TransIT-LT1 Transfection Reagent (Mirus Bio) according to the manufacturer's instructions. The supernatant of culture medium containing lentivirus was collected 36 to 48 hours after transfection and filtered by 0.2-μm filter, followed by addition to host cell lines. Twenty-four hours after infection, target cells were subjected to puromycin (2 μg/mL) selection for 24 to 48 hours. The knockdown efficiency of the endogenous IDH1 protein was confirmed by Western blotting. Stable overexpression of IDH1 WT and mutants in MOLM14, H1299, and TF-1 cells was conducted using retroviral vectors harboring RNAi-resistant FLAG-tagged IDH1 WT, FLAG-tagged IDH1 R132H, and RNAi-resistant FLAG-tagged IDH1 WT or R132H plus diverse YF mutants. Briefly, to produce retrovirus, each construct was cotransfected with VSVG, EcoPak packaging plasmid, and envelope plasmid (Addgene) into HEK293T cells using FuGENE Transfection Reagent (Promega) according to the manufacturer's instructions. Retrovirus-containing supernatant medium was collected 48 hours after transfection and filtered before addition to the indicated host cell lines (1 mol/L HEPES was used to adjust the pH to 7.4 in culture medium). Twenty-four hours after infection, target cells were subjected to hygromycin selection (Invitrogen). The overexpression of proteins was confirmed by Western blotting using antibodies against IDH1.

Antibodies against DYKDDDDK (FLAG) tag, p-Tyrosine (pTyr-100), JAK2, pJAK2 (Tyr1007/1008), dimethyl-histone H3 (Lys9), histone H3, FLT3, and Src were from Cell Signaling Technology (CST). Antibodies against β-actin and p-Src (Tyr418) were from Sigma-Aldrich. Antibody against HA was from Santa Cruz Biotechnology. Antibody against trimethyl (Lys9) was from Abcam. Antibodies against IDH1 R132H and Trimethyl-Histone H3 (Lys4) were from EMD Millipore. Antibody against IDH1 was from R&D Systems. Antibody against PE-Cy5 Mouse Anti-Human CD235a was from BD Pharmingen. Goat anti-Mouse IgG (H + L) secondary antibody and goat anti-rabbit IgG (H + L) secondary antibody were from Thermo Fisher Scientific. Antibodies against pIDH1 Y42 and pIDH1 Y391 were custom-made by Shanghai Genomics, Inc. Antibodies against Ki67 and IDH1 for IHC staining studies were from Abcam.

MBP from bovine, DL-isocitric acid trisodium salt hydrate, β-nicotinamide adenine dinucleotide phosphate hydrate (NADP⁺), β-nicotinamide adenine dinucleotide 2′-phosphate reduced tetrasodium salt hydrate (NADPH), α-ketoglutaric acid sodium salt, diaphorase from Clostridium kluyveri, resazurin sodium salt, and ATP disodium salt hydrate were purchased from Sigma-Aldrich. Ketoglutaric acid sodium salt, α-[1-¹⁴C], 50μCi (1.85MBq) and glucose, D-[U-¹⁴C]- were from PerkinElmer. Glutamine L-[5-¹⁴C] and isocitric acid[³H(G)] were from ARC. Inhibitors including dovitinib (TKI-258, CHIR-258), imatinib, AG-490, quizartinib, PP2, erlotinib, ruxolitinib, dasatinib, and AG-120 were purchased from Selleckchem. Recombinant proteins including FLT3, SRC, PDGFRα, PDGFRβ, EGFR, MET, KIT, JAK2, FGFR3, and ABL1 were purchased from Thermo Fisher.

Cell proliferation assays were performed by seeding 5 × 10⁴ cells in a 6-well plate in normoxia (5% CO₂ and 95% air) or hypoxia (5% CO₂, 1% O₂, and 94% N₂) with daily counting of cell numbers. Cell proliferation was determined by cell numbers recorded after being seeded and normalized to that of each cell line at the starting time (T = 0 hour) by trypan blue exclusion using TC20 Automated Cell Counter (Bio-Rad). For cell viability assay, 5 × 10⁴ cells were seeded in each well of a 96-well plate and incubated with ivosidenib (AG-120) or quizartinib for 48 hours. Relative cell viability was determined using the CellTiter 96 Aqueous One solution proliferation kit (Promega).

An in vitro or in vivo IDH1 R132H activity assay was performed as previously described (30). Specifically, 20 ng purified IDH1 R132H and variant proteins treated with or without recombinant active TKs in an in vitro kinase assay were added to 50 μL assay buffer [150 mmol/L NaCl, 20 mmol/L Tris-HCl (pH7.5), 10 mmol/L MgCl₂, 0.05% (w/v) bovine serum albumin (BSA)] containing 15 μmol/L NADPH (Sigma-Aldrich) and 1.5 mmol/L α-KG (Sigma-Aldrich). After 1-hour incubation at room temperature, 20 μmol/L resazurin and 10 μg/mL diaphorase were added to the reaction mixture and incubated for another 10 minutes at room temperature. To determine the IDH1 R132H activity in cells, 1 × 10⁶ cells were harvested and directly lysed using 100 μL 1% NP-40 cell lysis buffer. Whole-cell lysates (25 μL) were then added with 25 μL 2× assay buffer [300 mmol/L NaCl, 40 mmol/L Tris-HCl (pH7.5), 20 mmol/L MgCl₂, 0.05% (w/v) BSA] containing 30 μmol/L NADPH and 3 mmol/L α-KG. After 1-hour incubation at room temperature, 20 μmol/L resazurin and 10 μg/mL diaphorase were added to the primary reaction and incubated for another 10 minutes at room temperature. Fluorescence was recorded on a Spectramax GEMINIEM plate reader (Molecular Devices) at Ex544/Em590.

An in vitro enzyme activity assay for WT IDH1 was performed as previously described (8). In brief, 20 ng purified IDH1 and YF variant proteins that were pretreated with or without recombinant active TKs in an in vitro kinase assay were added to 100 μL enzyme activity assay buffer [25 mmol/L Tris-HCl (pH7.5), 10 mmol/L MgCl₂, 5 mmol/L DTT] containing 0.5 mmol/L NADP⁺ (Sigma-Aldrich) and 1 mmol/L isocitric acid (Sigma-Aldrich). To determine the WT IDH1 activity in cells, endogenous IDH1 from 1 × 10⁷ cells was bound to protein G-Sepharose 4 Fast Flow beads (Sigma-Aldrich) by immunoprecipitation using IDH1 antibody (CST). To test IDH1 activity in xenograft tumor tissues, around 2 mg of total tumor lysates were used. After immunoprecipitation, the IDH1-bound beads were washed 3 times and eluted using IDH1 peptide (CST). Then, 10 μL of the supernatant was added to 100 μL assay buffer (25 mmol/L Tris-HCl, pH7.5, 10 mmol/L MgCl₂, 5 mmol/L DTT) containing 0.5 mmol/L NADP⁺ and 1 mmol/L isocitric acid. IDH1 activity was measured by recording absorbance of 340 nm in kinetic mode every 20 seconds for 15 minutes using a SpectraMax Plus spectrophotometer (Molecular Devices).

Cells or tumor lysates (1–2 mg) were incubated with anti-IDH1 antibody (R&D Systems) or anti-IDH1 antibody (CST) overnight at 4°C. After incubation, protein G-Sepharose was used for precipitation for 2 hours. The beads were then washed 3 times with 1× TBS and eluted by boiling in SDS sample buffer for Western blotting analysis.

6×His-FLAG-IDH1 or 6×His-HA-IDH1 and variant proteins were purified by sonication of high expression BL21(DE3) pLysS cells obtained from a 250 mL culture subjected to IPTG induction for 16 hours at 30°C. Bacteria cell lysates were obtained by centrifugations and loaded onto a Ni-NTA column within 20 mmol/L imidazole. The bound proteins were eluted with 250 mmol/L imidazole, followed by desalting using a PD-10 column. The purified recombinant IDH1 proteins were examined by Coomassie Brilliant Blue staining and Western blotting.

In vitro kinase assays were performed as previously described (31, 32). Specifically, 1 μg recombinant IDH1, IDH1 R132H, and variant proteins were incubated with diverse recombinant active form of TKs for 90 minutes in the presence of 800 μmol/L ATP (Sigma) at 30°C in the following assay buffers, respectively: for FGFR1 assay buffer, 10 mmol/L HEPES (pH 7.5), 10 mmol/L MnCl₂, 150 mmol/L NaCl, 5 mmol/L DTT, and 0.01% Triton X-100 were used; for FLT3 assay buffer, 60 mmol/L HEPES (pH 7.5), 3 mmol/L MgCl₂, 3 mmol/L MnCl₂, 3 μmol/L Na₃VO₄, and 1.2 mmol/L DTT were used; for FGFR3 assay buffer, 60 mmol/L HEPES (pH 7.5), 3 mmol/L MgCl₂, 3 mmol/L MnCl₂, 3 μmol/L Na₃VO₄, and 1.2 mmol/L DTT were used; for JAK2 assay buffer, 25 mmol/L HEPES (pH 7.5), 10 mmol/L MgCl₂, 0.5 mmol/L EGTA, 0.5 mmol/L Na₃VO₄, 5 mmol/L β-glycerophosphate, 2.5 mmol/L DTT, and 0.01% Triton X100 were used; for ABL1 assay buffer, 60 mmol/L HEPES (pH 7.5), 3 mmol/L MgCl₂, 3 mmol/L MnCl₂, 3 μmol/L Na₃VO₄, and 1.2 mmol/L DTT were used; for SRC buffer, 50 mmol/L HEPES (pH 7.5), 10 mmol/L MgCl₂, 10% glycerol, 2.5 mmol/L DTT, and 0.01% Triton X-100 were used; for EGFR buffer, 20 mmol/L Tris (pH 7.5), 10 mmol/L MgCl₂, 1 mmol/L EGTA, 1 mmol/L Na₃VO₄, 5 mmol/L β-glycerophosphate, 2 mmol/L DTT, and 0.02% Triton X-100 were used; for MET assay buffer, 25 mmol/L Tris (pH 7.5), 10 mmol/L MgCl₂, 0.5 mmol/L EGTA, 0.5 mmol/L Na₃VO₄, 5 mmol/L β-glycerophosphate, 2.5 mmol/L DTT, and 0.01% Triton X-100 were used. The amount of each recombinant active kinase used for the reaction was as follows: 100 ng rFGFR1 (0.1 U), 160 ng rMET (0.06 U), 800 ng rEGFR (0.1 U), 200 ng rJAK2 (0.06 U), 120 ng rPDGFRα (0.066 U), 250 ng rPDGFRβ (0.066 U), 25 ng rFLT3 (0.06 U), 100 ng rSRC (0.2 U), and 120 ng rABL1 (0.06 U). For in vitro kinase assay of IDH1 dimer or monomer, 1 mg IDH1 WT or R132H protein was used for sucrose gradient centrifugation to separate monomer or dimer, then 1 μg monomer or dimer protein was used for in vitro kinase assay with or without recombinant active form of rFGFR1 or rFLT3. Then, samples were injected for the Western blotting analysis with native gel.

Total protein (200 μg) from whole-cell lysates was incubated with 30 μL of ANTI-FLAG M2 Affinity Gel (Sigma-Aldrich) for 4 hours at 4°C, followed by washing with phosphate-buffered saline (PBS) 3 times to remove unbound materials. The bound proteins were then eluted from beads by boiling in SDS buffer [50 mmol/L Tris-Cl (pH 6.8), 2% (w/v) SDS, 0.1% (w/v) bromophenol blue, 100 mmol/L DTT]for 10 minutes and visualized via immunoblotting analysis.

Native gels were prepared as described (31) using 0.375 mol/L Tris-HCl, pH 8.8, 10% acrylamide, without SDS. Cell lysates were mixed with 5× native gel sample buffer (0.05% bromophenol blue, 10% glycerol, 312.5 mmol/L Tris-HCl, pH 6.8) before loading. The samples were applied to native PAGE that ran in buffer (25 mmol/L Tris-HCl, 192 mmol/L glycine), followed by Western blotting.

Sucrose gradient centrifugation was performed as previously described (31). In brief, purified recombinant 6×His-FLAG-IDH1 or R132H and variant protein was laid on a 13.75% to 36% sucrose gradient and spun at 50,000 rpm for 12 hours using a Beckman MLS-50 rotor. Fractions in each section were collected and analyzed by Coomassie blue staining.

Purified recombinant 6×His-FLAG-IDH1 and variants that are immobilized on anti-FLAG beads were treated with or without diverse recombinant active TKs in an in vitro kinase assay. The beads were then incubated with 0.1 mmol/L isocitric acid [³H(G)] (ARC) at room temperature for 1 hour and then washed twice with TBS to remove the unbound isocitric acid [³H(G)] molecule. IDH1 and variant proteins were then eluted with 3× FLAG peptide. Bound isocitric acid molecules to proteins were measured using a scintillation counter. Purified recombinant 6×His-FLAG-IDH1 R132H and variants that are immobilized on anti-FLAG beads were treated with or without diverse recombinant active TKs in an in vitro kinase assay. The beads were incubated with 0.1 mmol/L α-ketoglutaric acid [1-¹⁴C] (PerkinElmer) at room temperature for 1 hour and then washed twice with TBS to remove the unbound α-ketoglutaric acid [1-¹⁴C]. The IDH1 R132H and variant proteins were eluted with 3× FLAG peptide. Bound α-ketoglutaric acid molecules to proteins were measured using a scintillation counter.

Purified recombinant IDH1 monomers and dimers were separated on a native gel. The gel was incubated with 0.1 mmol/L isocitric acid [³H(G)] (ARC) for 2 hours at 4°C and then washed twice with TBS to remove the unbound isocitric acid [³H(G)]. The IDH1 proteins in monomer and dimer form were cut from the gel and the retained isocitric acid [³H(G)] on both forms of IDH1 proteins were measured using a scintillation counter. Purified recombinant IDH1 R132H monomers and dimers were separated on a native gel, then incubated with 0.1 mmol/L α-ketoglutaric acid [1-¹⁴C] (PerkinElmer) for 2 hours at 4°C and then washed twice with TBS to remove the unbound α-ketoglutaric acid [1-¹⁴C]. The IDH1 R132H proteins in monomer and dimer form were cut from the gel, and the retained α-ketoglutaric acid [1-¹⁴C] on both forms of IDH1 R132H proteins was measured using a scintillation counter.

NADP⁺ binding assay was performed as previously described (33). Briefly, 200 ng of 6×His-FLAG-IDH1 and variant proteins was incubated with 30 μL Blue Sepharose CL-6B beads (Amersham Biosciences), which mimics NADP⁺ at 4°C for 2 hours, then washed with 20 mmol/L Tris-HCL (pH 8.6) 3 times. The beads were then subjected to SDS-PAGE, followed by Western blotting analysis. For NADP⁺ binding assay with IDH1 dimer or monomer, 1 mg IDH1 WT protein was used for sucrose gradient centrifugation to separate monomer or dimer, and then 1 μg monomer or dimer protein was used for in vitro kinase assay with or without FGFR1. After kinase assay, 200 ng protein were incubated with 30 μL Blue Sepharose CL-6B beads (Amersham Biosciences) at 4°C for 2 hours, then washed with 20 mmol/L Tris-HCL (pH 8.6) 3 times. The beads were subjected to SDS-PAGE, followed by Western blotting. The same amount of protein was loaded in parallel as loading control.

Twenty nanograms purified IDH1 and YF variant proteins pretreated with or without recombinant active TKs in an in vitro kinase assay were incubated with various concentrations of NADP⁺ (0–200 μmol/L; Sigma-Aldrich) or isocitrate acid (0–200 μmol/L; Sigma-Aldrich) and 1 mmol/L isocitric acid (Sigma-Aldrich) or 0.5 mmol/L NADP⁺ (Sigma-Aldrich), respectively, at 100 μL enzyme activity assay buffer (25 mmol/L Tris-HCl pH7.5), 10 mmol/L MgCl₂, 5 mmol/L DTT). Absorbance at 340 nm was measured every 20 seconds for 5 minutes using a SpectraMax Plus spectrophotometer (Molecular Devices).

Twenty nanograms purified IDH1 R132H and YF variant proteins pretreated with or without recombinant active TKs in an in vitro kinase assay were incubated with various concentrations of NADPH (0–10 μmol/L; Sigma-Aldrich) and 1.5 mmol/L α-KG (Sigma-Aldrich) or various concentrations of α-KG (0–1500 μmol/L; Sigma-Aldrich) and 15 μmol/L NADPH (Sigma-Aldrich) at 100 μL enzyme activity assay buffer [25 mmol/L Tris-HCl (pH7.5), 10 mmol/L MgCl₂, 5 mmol/L DTT]. Absorbance at 340 nm was measured every 20 seconds for 5 minutes using a SpectraMax Plus spectrophotometer (Molecular Devices). Nonlinear regression analysis (Michaelis-Menten) was performed in GraphPad Prism 6.0.

One milligram IDH1 WT or R132H protein was used for sucrose gradient centrifugation to separate monomer, and then 1 μg monomer protein was used for in vitro kinase assay with FGFR1 at indicated time and stopped in a minus 80 freezer. Then samples were injected for the Western blotting analysis with native gel.

One microgram of 6×His-FLAG-IDH1 protein was incubated with 100 μL Blue Sepharose CL-6B beads (Amersham Biosciences) at 4°C for 2 hours, then washed with 20 mmol/L Tris-HCL (pH 8.6) 3 times and separated to 3 tubes equally. The beads were incubated with indicated concentrations of NADP⁺ at 4°C for 1 hour. Then the samples were centrifuged to separate the supernatant and beads. The supernatant or beads were subjected to SDS-PAGE, followed by Western blotting analysis.

One milligram FLAG-IDH1 WT and HA-IDH1 R132H protein was used for sucrose gradient centrifugation to separate monomer, then 100 μg monomer protein was mixed and used for in vitro kinase assay with FGFR1 for 2 hours. The protein was incubated with 50 μL of ANTI-FLAG M2 Affinity Gel (Sigma-Aldrich) for 4 hours at 4°C, followed by washing with PBS 3 times to remove unbound materials. Then bound protein was eluted with 100 μg FLAG peptide (Sigma-Aldrich). Then half of the eluted protein was injected for the Western blotting analysis with native gel. The rest of the eluted protein was incubated with 30 μL of monoclonal anti-HA agarose (Sigma-Aldrich) for 4 hours at 4°C, followed by washing with PBS 3 times to remove unbound materials. The bound protein was eluted with 10 μg HA peptide (Sigma-Aldrich), and then the samples were applied to native gels for the Western blotting analysis.

One milligram IDH1 WT or R132H protein was used for in vitro kinase assay with FGFR1, and then the dimer was separated by sucrose gradient centrifugation. One microgram dimer was collected at indicated time, and the monomer conversion was stopped in a minus 80 freezer; then the samples were loaded on native gels for the Western blotting analysis.

For the purification of IDH1 heterodimer, 1 mg FLAG-IDH1 WT and HA-IDH1 R132H protein was used for in vitro kinase assay with FGFR1 for 2 hours. The protein was incubated with 100 μL of ANTI-FLAG M2 Affinity Gel (Sigma-Aldrich) for 4 hours at 4°C, followed by washing with PBS 3 times to remove unbound materials. Then bound protein was eluted with 200 μg FLAG peptide (Sigma-Aldrich). The eluted protein was incubated with 1000 μL of monoclonal anti-HA agarose (Sigma-Aldrich) for 4 hours at 4°C, followed by washing with PBS 3 times to remove unbound materials. Then bound protein was eluted with 50 μg HA peptide (Sigma-Aldrich). Then the heterodimer was collected at indicated time and the monomer conversion was stopped by a minus 80 freezer. Then the samples were loaded on a native gel for the Western blotting analysis.

Treatments with TK inhibitors were performed by incubating cells with 0.1 μmol/L TKI-258, 1 μmol/L imatinib, 50 μmol/L AG-490, 100 nmol/L erlotinib, 0.1 μmol/L ruxolitinib, 200 nmol/L quizartinib, 50 nmol/L dasatinib, or 1 μmol/L AG-120 for 4 hours. PP2 treatment was performed by incubating cells with 10 μmol/L PP2 for 4 hours.

The transfection of siRNA into MOLM-14 cells was carried out using Lipofectamine 2000 transfection reagent (Thermo Fisher), according to the manufacturer's instructions. Briefly, siRNA and Lipofectamine 2000 reagent were mixed in Opti-MEM medium (Thermo Fisher) and incubated for 30 minutes at room temperature to allow the complex formation. Then the cells were washed with Opti-MEM medium (Thermo Fisher), and the mixtures were added. Twelve hours after transfection, the culture medium was replaced by a fresh complete medium. The cells were harvested 72 hours after transfection, followed by further analysis. The following siRNA sequences were used for knockdown: negative control siRNA (nonsilencing; QIAGEN; SI03650325); Hs_ABL1_6 Flexi Tube siRNA (QIAGEN; SI00299089); Hs_SRC_7 Flexi Tube siRNA (QIAGEN; SI02223928); Hs_JAK2_7 Flexi Tube siRNA (QIAGEN; SI02659657); and Hs_FLT3_6 Flexi Tube siRNA (QIAGEN; SI02659608).

TF-1 stable cell lines were washed four times with plain RPMI medium and cultured for 24 hours without GM-CSF. Then 2 U/mLEPO (recombinant human erythropoietin) was added for differentiation. Eight days later, glycophorin A was used for erythroid differentiation assay by flow cytometry. Cell-surface staining for glycophorin A was performed as follows: Cells were washed once in PBS supplemented with 2% FBS (PBS–FBS) and then resuspended in a 1:200 dilution of PE-Cy5–conjugated mouse anti-human CD235a antibody (BD Pharmingen) in PBS–FBS. The cells were incubated for 20 minutes at 4°C in the dark, then washed once with PBS-FBS. The cells were analyzed on a FACSCanto II cytometer (Becton Dickinson). All flow cytometry data were analyzed using FlowJo software (TreeStar, Inc.).

The sample preparation for ex vivo nuclear magnetic resonance (NMR) analysis followed a previously described method (34). Briefly, 1 × 10⁷ cells were thawed in 99.996% saline deuterium oxide (D2O; Sigma-Adrich) in the sample holder/rotor (4 mm ZrO2). A 50 μL insert was subsequently placed in the sample holder to stabilize the sample and to provide the balance for the rotor. For NMR experiments, D2O (99.996%) containing 0.75% 3-(trimethylsilyl) propionic acid (TSP) was added to get a frequency-lock signal, as well as to serve as an internal reference for chemical shift and concentration measurements. Each sample added with TSP containing D2O was reweighed for metabolite quantification. All NMR samples were treated rapidly on ice to avoid possible degradation. HRMAS NMR experiments were conducted with a dedicated 4-mm HRMAS probe at 4°C using a Bruker AVANCE 600 WB solid state NMR spectrometer (Bruker Instruments, Inc.). The entire process throughout the experiment was maintained at 4°C (±0.1°C) via a variable temperature control unit, and the probe-head was precooled to 4°C before loading the sample. To ensure the spin sidebands did not affect the spectrum, the spinning rates of the sample were controlled in the range of 2,800 kHz (±2 Hz) or at the lower spin rate of 800 Hz if the rotor-synchronized delay alternating with nutation for tailored excitation sequence was used. The presaturation of water was achieved with a zqpr sequence before acquisition pulses. A rotor-synchronized Carr–Purcell–Meibom–Gill pulse sequence was used to suppress broad signals from macromolecules. The number of transients was 256. In all experiments, the repetition time was 5 seconds, and the spectral width was 10 kHz. A 95% pure 2HG compound (Santa Cruz Biotechnology) in solution was used as 2HG resonance identification and assignment. One-dimensional NMR spectra of a pure 2HG compound, and a mixture of 2HG, and a combination of glutamate (Glu) and glutamine (Gln) compounds (together termed Glx) were obtained at ∼10 mmol/L 2HG in D2O, pH 7.0 and collected at 300 MHz at 25°C. J-coupling correlations and patterns of protons in the pure 2HG were then analyzed by a two-dimensional (2-D) J-coupled correlated spectroscopy (COSY) method. For the tumor sample analysis, data of 2-D COSY were collected at 4°C with 6,000 Hz spectral width and 1.5-second relaxation delay. Thirty-two transients in the time domain t2 were averaged for each of the 512 increments in time domain of t1 with a total acquisition time of ∼3 hours.

For ¹⁴C-lipid biosynthesis assay, cells were preincubated with 4 μL of glucose, D-[¹⁴C(U)] (PerkinElmer) under normoxia (5% CO₂ and 95% air) for 2 hours or 4 mmol/L glutamine L-[5-¹⁴C] (ARC) under hypoxia (5% CO₂, 1% O₂, and 94% N₂) for 24 hours. Lipids were then extracted by the addition of 500 μL of hexane: isopropanol (3:2 v/v), dried, resuspended in 50 μL of chloroform, and subjected to scintillation counting.

Approval for the use of mice and designed experiments was given by the Institutional Animal Care and Use Committee of Emory University. Briefly, NSG mice (NOD/SCID gamma, female 6-week-old; The Jackson Lab) were subcutaneously injected with 1 × 10⁶ H1299 cells stably expressing IDH1 WT, IDH1 Y42F, IDH1 Y391F, and IDH1 Y42F/Y391F with stable knockdown of endogenous IDH1 on the left and right flanks, respectively. Tumor growth was recorded every 2 days from 7 to 12 days after inoculation by measurement of two perpendicular diameters using the formula 4π/3 × (width/2)² × (length/2). The masses of tumors (mg) derived from treatments were analyzed. For single-cell isolation from tumor tissues, fresh tumors were removed from xenograft mice and were placed in PBS. The tumors were then minced into small pieces using scissors and digested with 5 mL digestion buffer (Accumax@ Stemcell) for 1 hour at room temperature on a horizontal shaker. The digestion buffer was then neutralized with an equal amount of cell culture media after digestion, and xenograft tumor cells were strained through a 70-μm cell strainer, followed by washing twice with PBS. Cells were then counted and cultured in RPMI-1640 medium supplemented with 10% FBS and penicillin/streptomycin (100 unit/mL) for further treatment.

For in vivo assay, 1 × 10⁷ cells were harvested and lysed using 900 μL NP-40 cell lysis buffer with 100 μL 10 × PTP1B buffer (500 mmol/L HEPES, pH 7.2, 10 mmol/L EDTA, 10 mmol/L DTT, 0.5% NP-40) containing 0.5 μg BSA, followed by incubation with 2 μg PTP1B protein for 90 minutes at 30°C. Reaction mixture (200 μL) was then stored as input samples, whereas the remaining 800 μL reaction mixture was subjected to IDH1 immunoprecipitation using anti-IDH1 antibody (R&D Systems) for 6 hours at 4°C. After immunoprecipitation, protein G-Sepharose was added for precipitation for 2 hours. The beads were then washed 3 times with TBS, followed by analysis using PAGE and Western blotting. Input samples were applied to native gel followed by Western blotting. For the in vitro assay, recombinant FLAG-tagged IDH1 and mutant proteins (1 μg each) were incubated with FLAG beads (30 μL) for 3 hours at 4°C. After incubation, the beads were washed 3 times with TBS, followed by in vitro TK assay in the presence or absence of recombinant active form of FGFR1 or FLT3 for 90 minutes at 30°C. The beads were then washed 3 times with TBS to remove kinases, followed by addition of 1 μg PTP1B in 100 μL reaction buffer (50 mmol/L HEPES, pH 7.2, 1 mmol/L EDTA, 1 mmol/L DTT, 0.05% NP-40 containing 0.1 μg of BSA) and incubated for 90 minutes at 30°C. At the end of the reaction, the beads were washed with TBS 3 times, then IDH1 proteins were eluted by adding 2 μg of FLAG peptide, followed by Western blotting analysis and enzyme activity assay.

IHC staining for Ki67, phospho-IDH1 Y42, and IDH1 was performed as previously described (35–37). In brief, tumor tissues from xenograft mice were fixed in 10% buffered formalin, embedded in paraffin, and mounted on slides. Slides of xenograft tumor tissues or tumor tissue arrays were deparaffinized and rehydrated, followed by incubation in 3% hydrogen peroxide to suppress endogenous peroxidase activity. High-pressure antigen retrieval was achieved in 10 mmol/L sodium citrate (pH6.0). Sections were then blocked by incubation in 10% goat serum. Human Ki67 antibody (Abcam; 1:500), human phospho-IDH1 Y42 antibody (Shanghai Genomics, Inc.; 1:200), or human IDH1 antibody (Abcam; 1:500) was applied overnight at 4°C. A Dako IHC kit (Agilent Technologies) was used for detection. A secondary antibody was applied at room temperature for 1 hour. Slides were then stained with 3,3′-diaminobenzidine, washed, counterstained with hematoxylin, dehydrated, and mounted.

A two-tailed Student t test was used in studies in which statistical analyses were performed to generate P values, except a two-way ANOVA was used for cell proliferation assay and tumor growth. P values less than or equal to 0.05 were considered significant. Data with error bars represent mean ± SD, except for tumor growth curves which represent mean ± SEM. There is no estimate of variation in each group of data, and the variance is similar between the groups. No statistical method was used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. All data are expected to have normal distribution. Statistical analysis and graphical presentation were performed using Prism 6.0 (GraphPad) and Microsoft Office Excel 2016.

S. Lonial reports receiving commercial research grants from Takeda, Celgene, and Janssen and is a consultant/advisory board member for Takeda, Celgene, BMS, AbbVie, Merck, Amgen, and Janssen. R.L. Levine is on the supervisory board of Qiagen, reports receiving commercial research grants from Roche, Prelude, and Celgene, has ownership interest (including stock, patents, etc.) in Loxo, Isoplexis, and C4, and is a consultant/advisory board member for Celgene, Roche, Morphosys, Janssen, Incyte, C4, and Astellas. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Chen, S. Xia, M. Wang, T.J. Boggon, R.L. Levine, J. Chen

Development of methodology: D. Chen, S. Xia, M. Wang, J. Fan, L. Jin, L. Song, J. Peng, J. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Chen, S. Xia, M. Wang, Y. Li, M. Aguiar, C.A. Famulare, A.H. Shih, Y. Pan, S. Liu, J. Fan, L. Jin, L. Song, A. Zhou, R.O. Pieper, A. Mishra, J. Peng, M. Arellano, W.G. Blum, R.L. Levine

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Chen, S. Xia, M. Wang, Y. Li, H. Mao, M. Aguiar, J. Fan, L. Jin, L. Song, A. Mishra, J. Peng, M. Arellano, S. Lonial, R.L. Levine, J. Chen

Writing, review, and/or revision of the manuscript: D. Chen, S. Xia, M. Wang, R. Lin, Y. Li, H. Mao, C.W. Brennan, M. Arellano, W.G. Blum, S. Lonial, R.L. Levine, J. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Wang, R. Lin, C.A. Famulare, A.H. Shih, C.W. Brennan, X. Gao, J. Chen

Study supervision: R. Lin, J. Peng, J. Chen

Other (provided cell lines developed in the study): J. Mukherjee

We thank Dr. Anthea Hammond for critical reading and editing of the manuscript. This work was supported in part by NIH grants, including CA140515, CA183594, CA174786 (J. Chen), R35197594, CA173636, MSKCC Support Grant/Core Grant P30 CA008748 (R.L. Levine), CA169937 (H. Mao), and K08CA181507-01A1 (A.H. Shih), and Joel A. Katz Music Medicine Fund supported by the T.J. Martell Foundation/Winship Cancer Institute (J. Chen and R. Lin). R. Lin and A.H. Shih are Special Fellows of the Leukemia and Lymphoma Society. J. Chen is the Winship 5K Scholar and the R. Randall Rollins Chair in Oncology.