Research ArticleAcute Myeloid Leukemia

A Pyrrolo-Pyrimidine Derivative Targets Human Primary AML Stem Cells in Vivo

See allHide authors and affiliations

Science Translational Medicine  17 Apr 2013:
Vol. 5, Issue 181, pp. 181ra52
DOI: 10.1126/scitranslmed.3004387


Leukemia stem cells (LSCs) that survive conventional chemotherapy are thought to contribute to disease relapse, leading to poor long-term outcomes for patients with acute myeloid leukemia (AML). We previously identified a Src-family kinase (SFK) member, hematopoietic cell kinase (HCK), as a molecular target that is highly differentially expressed in human primary LSCs compared with human normal hematopoietic stem cells (HSCs). We performed a large-scale chemical library screen that integrated a high-throughput enzyme inhibition assay, in silico binding prediction, and crystal structure determination and found a candidate HCK inhibitor, RK-20449, a pyrrolo-pyrimidine derivative with an enzymatic IC50 (half maximal inhibitory concentration) in the subnanomolar range. A crystal structure revealed that RK-20449 bound the activation pocket of HCK. In vivo administration of RK-20449 to nonobese diabetic (NOD)/severe combined immunodeficient (SCID)/IL2rgnull mice engrafted with highly aggressive therapy-resistant AML significantly reduced human LSC and non-stem AML burden. By eliminating chemotherapy-resistant LSCs, RK-20449 may help to prevent relapse and lead to improved patient outcomes in AML.


Acute myeloid leukemia (AML) is the most common type of acute leukemia in adults, with ~28,000 new cases reported in the United States, Europe, and Japan in 2010 (1). The prognoses for AML other than acute promyelocytic leukemia remain poor. Although 70 to 80% of patients younger than 60 years and 40 to 50% of patients 60 years and older initially achieve complete remission after a combination chemotherapy regimen that includes cytarabine (AraC) and an anthracycline, the majority of patients ultimately experience disease relapse, with 5-year overall survival of 40 to 45% for patients below 60 years of age and <10% in patients 60 years and older (25). Therefore, prevention of relapse is crucial for improving long-term outcomes in AML.

Over the last decade, leukemia stem cells (LSCs) have become recognized as key players in human AML pathogenesis as well as chemotherapy resistance and relapse (6, 7). In two recent studies, the expression of a LSC gene signature was independently associated with poor clinical outcome (8, 9). We have previously demonstrated chemotherapy resistance of human LSCs in an in vivo mouse model for human primary AML (10). Forced cell cycle entry increases the chemotherapy responsiveness of human primary LSCs and leads to decreased capacity for in vivo leukemia initiation (11). Recently, we determined that mRNA for hematopoietic cell kinase (HCK), a member of the Src family of nonreceptor tyrosine-protein kinases (SFKs), is overrepresented in primary human AML LSCs in comparison to normal hematopoietic stem cells (HSCs) (12).

SFKs have been implicated as intermediates in signaling pathways that regulate cellular functions such as proliferation, differentiation, survival, and migration in various cell lineages (13). Of these, Lyn, Fgr, and HCK are expressed in myeloid lineages and are involved in normal myelopoiesis (14). The importance of HCK, Lyn, and Fgr in myeloid proliferation and differentiation has been demonstrated in knockout mice (15, 16). The myeloid-specific SFKs participate in wild-type and mutant KIT and FLT3 signaling and in the activation of the signal transducer and activator of transcription 5 (STAT5) and extracellular signal–regulated kinase (ERK) pathways downstream of BCR-ABL, and are involved in leukemogenesis in mouse model of BCR-ABL+ B-ALL (1723).

Here, we have used a screening strategy that integrates a high-throughput enzyme inhibition assay, in silico structure prediction, and crystal structure determination to identify potential small-molecule therapeutic agents, which we tested in a murine in vivo engraftment and treatment model for human primary AML.


HCK expression and function in chemotherapy-resistant human AML LSCs

Nonobese diabetic (NOD)/severe combined immunodeficient (SCID)/IL2rgnull (NSG) mice lack multiple immune subsets, making them highly useful recipients for xenotransplantation models of human normal and malignant hematopoiesis (24). Consistent with our previous report that HCK transcript is overrepresented in human AML LSCs, HCK protein was expressed in human (h)CD45+ AML cells in the endosteal region of the bone marrow (BM) of human AML–engrafted NSG recipients, where Ki-67–negative, cell cycle–quiescent, chemotherapy-resistant human LSCs reside (Fig. 1A, left panel) (1012). HCK expression persisted in hCD45+ AML cells that remained in the BM endosteal region after in vivo treatment with the chemotherapy agent cytarabine (AraC) (Fig. 1A, right panel). We therefore hypothesized that HCK is a promising molecular target for anti–human AML LSC therapy development.

Fig. 1 HCK expression in human AML–engrafted BM and inhibition of HCK expression by shRNA.

(A) Femurs from NSG recipients engrafted with human primary AML at baseline and after in vivo AraC administration were labeled with anti-CD45 (green) and anti-hHCK (red) antibodies. Nuclei were labeled by 4′,6-diamidino-2-phenylindole (DAPI; blue). (B) After lentiviral transduction, GFP+ hCD34+CD38 TF1a and GFP+ hCD45+ K562 cells were cultured in triplicate (1 × 104 cells per well). Photomicrographs of TF1a (upper left panels) and K562 (lower left panels) cells transduced with either control or HCK shRNA at day 7. Cells were harvested, and the numbers of viable cells were determined by flow cytometric beads enumeration (upper right and upper left panels). Dotted lines indicate starting number of sorted cells. *P = 0.0048 by two-tailed t test.

To assess the functional role of HCK, we examined the effect of HCK down-regulation by short hairpin RNA (shRNA). We transduced the human AML cell line TF1a and the human chronic myeloid leukemia (CML) cell line K562 with HCK shRNA and control shRNA. Green fluorescent protein (GFP)–positive hCD34+CD38 cells were fluorescence-activated cell sorting (FACS)–purified and cultured in vitro for 7 days to determine the effect of HCK down-regulation on proliferation and survival. Reduced HCK expression led to significant inhibition of proliferation and reduction in the survival of TF1a cells (which expressed HCK before transduction), whereas the cell growth and viability of K562 cells [which do not express HCK (25)] were not affected (Fig. 1B and fig. S1, A and B). HCK shRNA did not affect the transcript levels of the SFKs Lyn and Src, reportedly involved in human AML pathogenesis (fig. S1C). These findings suggested that HCK expression is required for proliferation and survival of human AML cells and that pharmacologic inhibition of HCK may eliminate human AML LSCs.

Identification of small-molecule HCK inhibitors

To identify inhibitors for HCK, we performed a combination of a high-throughput enzyme inhibition assay, in silico screening, and x-ray crystallography (Fig. 2A). First, using a kinase mobility shift assay, which rapidly and directly determined the ability of a compound to inhibit HCK, we screened 47,574 commercially available small molecules with a mean molecular weight of 329.6 and an ALogP of 3.32. As a primary screen, each compound (final concentration of 2 μM) was tested for its effect on HCK activity by the mobility shift assay (26) with 0.5 mM adenosine 5′-triphosphate (ATP). The mean z′ value (27) was 0.61 across all plates, indicating a stable assay performance. Among 47,574 compounds, 146 hit compounds inhibited HCK activity by more than 50%, and among these, we identified 12 compounds with IC50 (half maximal inhibitory concentration) less than 1 μM.

In parallel with the enzyme inhibition assay, we performed in silico screening that allowed rapid screening of a much larger pool of compounds. We used a computational approach combining ligand-based two-dimensional (2D) and 3D similarity search against known inhibitors of HCK and other SFKs, docking simulation with published HCK crystal structures, and a machine-based learning algorithm for structural fingerprints of known HCK inhibitors. From 8,927,055 commercially available small molecules, we selected 3088 candidate compounds. HCK inhibition in enzymatic assays was determined for these compounds, yielding three compounds with IC50 less than 1 μM.

Fig. 2 Identification of small-molecule HCK inhibitors.

(A) Screening strategy integrating high-throughput enzyme assay, in silico prediction, and x-ray crystallography. (B) Upper left: Chemical structures of PP2 and RK-24466. Upper right: Superimposed model of RK-24466 with PP2. Lower left and right: Docking models of RK-24466 and PP2 with HCK (PDB ID: 1QCF). The phenoxy moiety of HCK where hydrophobic interaction with RK-24466 may occur is indicated by squares. (C) RK-24466 bound to the ATP binding site of HCK. RK-24466 and AMPPNP [PDB ID: 1AD5 (61)] are represented in ball and stick model, and carbon atoms are colored yellow and cyan, respectively. The nitrogen, oxygen, and sulfur atoms are colored blue, red, and green, respectively. Hydrogen bonds are shown as broken lines. (D) RK-24466 binding with HCK based on crystal structure determination, highlighting strategies for designing compounds with higher affinity with HCK. Broken lines indicate hydrogen bonds. Meshed regions show contact preferences of ligand atoms by Molecular Operating Environment (MOE; Chemical Computing Group). Green regions favor hydrophobic ligand atoms, and magenta regions favor hydrophilic ligand atoms. Blue and pink shading indicate positively and negatively charged regions of HCK, respectively. The pyrrolo-pyrimidine nucleus of RK-24466 is circled, with the seventh position indicated. The areas targeted for modification strategies are indicated by a rectangle.

Of the 15 compounds, we chose RK-24466 for further evaluation because it was the only compound with IC50 <100 nM against HCK (IC50 = 7.7 nM) and had the most favorable chemical structure (without reactive moieties or structures associated with nonspecific protein reactivity) (28). A 3D structural model of RK-24466 showed similarity (ROCS ComboScore) to a known HCK inhibitor, PP2 (Fig. 2B). An in silico docking model of RK-24466 with HCK [Protein Data Bank (PDB) ID: 2C0I] showed that RK-24466 inserts into the ATP-binding pocket of HCK and forms hydrogen bonds with Met341 and Glu339. This is predicted to stabilize RK-24466’s interaction with the hinge region of HCK, resulting in competitive interference with ATP binding (29) (fig. S2). The Glide docking score of RK-24466 binding with HCK predicted that it has a higher binding affinity than the known SFK inhibitor PP2 (IC50 = 630 nM) (Fig. 2B) (30). The higher potency of HCK inhibition by RK-24466 than by PP2 may also be a result of additional protein–ligand surface contacts, for instance, at the phenoxy moiety where there is a hydrophobic interaction. The crystal structure of HCK complexed with RK-24466 (Fig. 2C and table S1) confirmed that the compound binds within the ATP-binding pocket and forms hydrogen bonds with the hinge region. The phenoxy moiety is buried in a hydrophobic pocket formed by Met314, Val323, Leu325, Ile336, Ala403, Phe405, and Leu407, which may contribute to the potency of this compound against HCK.

We next proceeded to evaluate the activity of RK-24466 against LSCs derived from AML patients (patient information summarized in Table 1). In an in vitro assay in which FACS-purified CD34+CD38 AML cells are cultured in the presence of human stem cell factor (SCF), thrombopoietin (TPO), and FLT3 ligand, the phenotype and in vivo function of CD34+CD38 LSCs are maintained, as demonstrated by leukemia initiation and generation of CD34+CD38+ cells and CD34 cells in NSG recipients (fig. S3). After exposure of the cells to the test compounds for 3 days, we harvested the cells for flow cytometric determination of viability and proliferation. In vitro, RK-24466 potently inhibited CD34+CD38 human primary AML cell proliferation to a greater extent than did the known SFK inhibitor PP2 (fig. S4). However, pharmaceutical use of RK-24466 was expected to be difficult because of its poor aqueous solubility. Therefore, we went on to explore modifications of RK-24466 to obtain HCK inhibitors with superior chemical profile.

Table 1 Patient information.

FAB, French-American-British classification; NOS, not otherwise specified; MRC, myelodysplasia-related changes; Ida, idarubicin; AraC, cytarabine; MIT, mitoxantrone; HiDAC, high-dose cytarabine; MEC, mitoxantrone, etoposide, and cytarabine; CAG, cytarabine, aclarubicin, and granulocyte colony-stimulating factor; BHAC, behenoyl cytarabine; DMP, daunorubicin, 6-mercaptopurine, and prednisolone; DNR, daunorubicin; 6-MP, 6-mercaptopurine; CA, cytarabine and aclarubicin; FLAG, fludarabine, cytarabine, and granulocyte-colony stimulating factor; SCT, stem cell transplantation; GO, gemtuzumab ozogamicin; na, not available.

View this table:

There remained open spaces in the crystal structure of RK-24466–bound HCK, which may accommodate new substituents to improve the aqueous solubility as well as the potency of RK-24466 (Fig. 2D). One of these spaces was between the 7-substituent of the pyrrolo-pyrimidine nucleus and Asp348, which hydrogen-bonds with the ribose moiety of ATP (Fig. 2C). The corresponding space has been used to increase the potency and specificity of inhibitors against Lck, another SFK (31). Another available space extends toward Asp404 in the universally conserved Asp-Phe-Gly (DFG) motif in the activation segment of HCK. To use these two spaces, we synthesized seven compounds that are potentially better ligands for HCK than RK-24466 (table S2 and fig. S5). Of these, RK-20449 [compound 13 (31); Fig. 3A] demonstrated the most potent inhibition of HCK, with an IC50 of 0.43 nM.

Fig. 3 Crystal structure of HCK in complex with RK-20449, the leading candidate inhibitor.

(A) Chemical structure of RK-20449. (B to D) Crystal structures of RK-24466 and RK-20449 bound to HCK, demonstrating an additional interaction that leads to the improved affinity of RK-20449 for HCK. Broken lines (yellow) indicate favorable interactions between each compound and HCK, with the interaction distances in angstrom units. (B) RK-24466 (green) forms hydrogen bonds with Met341, Glu339, and Thr338 in the hinge region of HCK (blue), enhancing its binding within the ATP-binding pocket, and the phenoxy moiety interacts with the hydrophobic pocket featuring Phe405. (C) In addition to the interactions with the hinge region and the hydrophobic pocket, the methylpiperazine group of RK-20449 (green) interacts with Asp348 of HCK (blue), leading to increased binding affinity. (D) An alternative view [45° rotation from (C)] showing how the acral tertiary amine nitrogen in the methylpiperazine group of RK-20449 interacts with Asp348 of HCK. (E) RK-20449 was screened against a panel of 106 human kinases, which are represented by circles on the kinase dendrogram. Kinases with IC50 less than 1 μM are shown by red circles, with circle size indicating potency of inhibition. Kinases with IC50 greater than 1 μM are indicated by small green circles. The complete data set is shown in table S3. The kinase dendrogram was adapted and is reproduced with permission from Cell Signaling Technology Inc. (

To examine the mechanism of highly potent HCK inhibition by RK-20449, we determined the crystal structures and analyzed the binding modes of these seven compounds with HCK. RK-20444, RK-20445, and RK-20446 have phenyl groups connected through N-methylurea, amide, and urea linkers, respectively, which extend toward the DFG motif (fig. S6, A and B). In the structures of HCK complexed with these three compounds, the conformations of the Phe405 side chain vary. Correspondingly, the phenyl groups of the three compounds assume different orientations, but none of them reaches the Phe405 side chain. In addition, the side-chain conformation of Asp404 in the RK-20446–bound HCK structure is different from those in the RK-20444– and RK-20445–bound HCK structures, presumably to avoid a steric clash with the urea linker. These findings explain why these three compounds exhibit increased IC50 values, as compared with the compounds with diphenylether, such as RK-24466 (table S2).

Next, we determined the crystal structures of HCK bound with RK-20448, RK-20449, RK-20450, and RK-20451, four compounds with 7-substituents extending toward Asp348. Although all four compounds maintain hydrogen-bond interactions with the hinge region of HCK similar to RK-24466 (Fig. 3B), RK-20450 and RK-20451, with more hydrophilic oxane and methylpiperazine groups, respectively, do not interact with Asp348 (fig. S6C). In RK-20448, with the attachment of methylpiperazine to cyclohexane in the cis configuration, the side chain moves away from the negatively charged surface of HCK, precluding its interaction with Asp348 (fig. S6, C and D). In contrast, the trans configuration of methylpiperazine attached to cyclohexane in RK-20449 allows electrostatic interaction of the basic acral tertiary amine nitrogen atom with the acidic side chain of Asp348 (3.2 Å) (Fig. 3, C and D, and fig. S6, C to E). This interaction is expected to be stronger than the hydrogen bond between Asp348 and the ATP ribose moiety (3.4 Å) (fig. S6D), resulting in higher binding affinity and superior potency of inhibition of HCK by RK-20449.

To examine the spectrum of inhibition by RK-20449, we determined the IC50 against a panel of 106 human kinases. The findings are summarized in Fig. 3E and table S3. We found that RK-20449 showed an IC50 <100 nM against seven kinases, all of which were SFKs (FYN, HCK, LCK, LYNa, LYNb, SRC, and YES), consistent with the high level of amino acid sequence homology among the SFKs. Four non-SFK kinases (EGFR, RET, PDGFRα, and FLT3) were inhibited at IC50 of 100 nM to 1 μM. RK-20449 did not highly inhibit the other kinases tested (IC50 >1 μM).

Because constitutive activation of FLT3, most frequently in the form of FLT3ITD+ mutations, is commonly associated with normal karyotype AML, we examined the significance of HCK expression in FLT3ITD+ human AML by HCK shRNA. In two FLT3ITD+ human AML cell lines, MV-4-11 and Molm13, we found that inhibition of HCK by shRNA reduced both viable cell number and viability in the remaining cells (fig. S7), suggesting that HCK inhibition by RK-20449 may be effective against AML with activating mutations of FLT3.

In vitro inhibition of human AML LSC by RK-20449

We then evaluated the activity of RK-20449 in the in vitro LSC culture assay. Compared with SFK inhibitors PP2, PP20, and RK-24466; multikinase inhibitors PP121 and PI-103; and a mammalian target of rapamycin (mTOR) inhibitor, RK-20449 demonstrated the highest activity against human AML CD34+CD38 cells (Fig. 4, A and B) (32). RK-20449 showed dose-dependent inhibition of LSC proliferation with IC50 <1 μM in 17 of 25 AML cases that we examined, exhibiting higher in vitro activity against human AML CD34+CD38 cells than PP2 (IC50 <1 μM in 0 of 18 cases) and RK-24466 (IC50 <1 μM in 7 of 21 cases) (Fig. 4C, table S4, and Table 1).

Fig. 4 Potent in vitro activity of RK-20449 against human AML LSCs.

(A and B) Potency of various SFK inhibitors, RK-24466, and RK-20449 against hCD34+CD38 cells in vitro. Representative (A) photomicrographs and (B) flow cytometry plots of sorted hCD34+CD38 cells exposed to 300 nM PP2, RK-24466, RK-20449, PP121, PP20, PI-103, or rapamycin. (C) Inhibition of cell proliferation by (left) PP2 (n = 18), (middle) RK-24466 (n = 21), and (right) RK-20449 (n = 25). For each dose, the numbers of viable cells, determined by flow cytometry bead enumeration, are represented as fraction of viable cells in the DMSO alone–treated group. Broken line indicates 50% reduction in the number of viable cells compared with the DMSO-alone group. (D) RK-20449 inhibition of phosphorylation of SFK-activating tyrosine and downstream intermediates. P-flow scores are obtained by log2 transformation of the fold change of the mean fluorescent intensity (MFI). Fold change was calculated by dividing the MFI of RK-20449– or rapamycin-treated cells by the MFI of DMSO only–treated cells. Case #1 RK-20449 n = 3, rapamycin n = 2; case #2 RK-20449 n = 1, rapamycin n = 1; case #3 RK-20449 n = 6, rapamycin n = 2; case #4 RK-20449 n = 3, rapamycin n = 3. Red, green, purple, and blue bars show P-flow scores (mean ± SEM) after 1 μM RK-20449 for cases #1, 2, 3, and 4, respectively. Gray bars, P-flow scores (mean ± SEM) after 1 μM rapamycin for cases #1 to 4 combined.

Next, we examined the effect of RK-20449 binding to HCK on the phosphorylation of the activating tyrosine residue in the homologous SFK protein kinase domain (pY418) and further downstream mediators in four AML patient cases (representative flow cytometry plots are shown in fig. S8). P-flow score, representing the change in phosphorylation status compared with dimethyl sulfoxide (DMSO) alone–treated cells, showed consistent suppression of SFK, S6, p4E-BP1, and STAT5 phosphorylation in hCD45+ total AML cells and in hCD34+CD38 LSCs with in vitro exposure to RK-20449 (Fig. 4D). In contrast, rapamycin suppressed phosphorylation of the TORC1 substrate S6 but not of pSRC and pSTAT5, consistent with previous reports (3335).

We then compared the effect of 100 and 300 nM RK-20449 at 24 and 48 hours on human AML LSCs and human cord blood (CB) CD34+CD38 HSCs in vitro. Although there was some reduction in CB HSC cluster size with higher concentration of RK-20449, the effect was more pronounced in AML LSCs, suggesting relative sparing of normal HSCs (fig. S9A). To determine the effect of RK-20449 treatment on LSC function, we evaluated the in vivo leukemia initiation capacity of in vitro RK-20449–treated LSCs through NSG transplantation. After 3 days in culture with RK-20449, viable hCD45+ cells were FACS-purified and transplanted into irradiated newborn NSG mice. We found that recipients of LSCs that had been treated in vitro with RK-20449 showed attenuated AML engraftment [representative flow cytometry plots of peripheral blood (PB) engraftment are shown in fig. S9B]. In addition, recipients of RK-20449–treated LSCs showed significantly improved survival compared with recipients of DMSO only–treated LSCs, consistent with reduced PB engraftment of hCD45+ AML cells (fig. S9, C and D). These data indicated reduced leukemia initiation capacity with in vitro exposure to RK-20449. When CB HSCs were similarly exposed to RK-20449 and transplanted into newborn NSG mice, we observed long-term, multilineage human hematopoietic reconstitution in the recipient BM, showing that HSCs retained in vivo hematopoietic repopulation capacity after exposure to RK-20449 (fig. S9, E to H).

In vivo elimination of primary human AML LSCs by RK-20449

Next, we evaluated the effect of RK-20449 in vivo in NSG recipients engrafted with primary human LSCs (from seven patients) that showed in vitro RK-20449 IC50 values of less than 1 μM. The recipients were treated twice daily with RK-20449 or vehicle alone and were analyzed after 14 to 21 days of treatment. After in vivo RK-20449 treatment, BM and spleen of recipient mice showed significant reduction of the absolute numbers of viable hCD45+ cells (total AML cells) and hCD34+CD38 cells (enriched in AML-initiating cells) compared with controls, indicating that RK-20449 as a single agent can effect reduction of both total AML cells and LSCs in vivo (Fig. 5, A and B). To assess the effect of RK-20449 treatment on normal human hematopoiesis in vivo, we treated CB HSC–engrafted NSG recipients for 30 days. We found nonsignificant reduction in total hCD45+ hematopoietic chimerism with maintenance of frequencies of total T cells and B cells (fig. S10). However, frequency of CD33+ myeloid cells was reduced in the BM of all three treated recipients.

Fig. 5 Elimination of chemotherapy-resistant human AML LSCs by RK-20449 in vivo.

(A and B) Effect of RK-20449 as a single agent on the absolute numbers of viable hCD45+ total AML cells and hCD34+CD38 LSCs in (A) BM and (B) spleen of AML-engrafted recipients treated for 14 to 21 days. Recipients were engrafted with LSCs obtained from AML cases #1 to 7 (Table 1). Vehicle only control recipients: n = 22 (case #1 n = 5, case #2 n = 5, case #3 n = 3, case #4 n = 2, case #5 n = 1, case #6 n = 2, case #7 n = 4); RK-20449–treated recipients: n = 20 (case #1 n = 5, case #2 n = 3, case #3 n = 4, case #4 n = 1, case #5 n = 1, case #6 n = 4, case #7 n = 2). *P < 0.0001 by two-tailed t test. (C to G) In vivo effect of RK-20449 on AML cells and LSCs in PB, BM, and spleen of the recipients in two cases of clinically chemotherapy-resistant AML (cases #1 and 2). (C and D) Representative flow cytometry plots showing percentage hCD45+ cells in PB from (C) case #1– and (D) case #2–engrafted recipients treated with RK-20449. (E) Human AML chimerism in PB over time. (Left) Case #1: control, n = 7; AraC alone, n = 7; RK-20449 alone, n = 7; RK-20449 after AraC, n = 3. (Right) Case #2: control, n = 7; AraC alone, n = 4; RK-20449 alone, n = 5; RK-20449 after AraC, n = 3. (F and G) Absolute numbers of viable hCD45+ total AML cells and hCD34+CD38 LSCs in (F) BM and (G) spleen of recipients treated for 21 days or longer. Vehicle only control recipients: n = 10 (case #1 n = 5, case #2 n = 5); RK-20449–treated recipients: n = 10 (case #1 n = 5, case #2 n = 5); AraC alone–treated recipients: n = 12 (case #1 n = 9, case #2 n = 3); recipients treated with RK-20449 after AraC: n = 6 (case #1 n = 4, case #2 n = 3). *P < 0.001 by two-tailed t test.

Because chemotherapy-resistant LSCs likely contribute to AML relapse, we next investigated the effect of in vivo RK-20449 administration in recipients engrafted with primary human AML with known chemotherapy resistance. To do so, we chose NSG recipients engrafted with LSCs obtained from patient #1 who failed the induction chemotherapy regimen containing AraC. Compatible with this clinical history, AraC was not highly effective against these cells in vitro, whereas they showed substantial response to RK-20449 (fig. S11). NSG recipients engrafted with AML case #1 were treated with either vehicle alone, RK-20449 alone, AraC alone, or AraC followed by RK-20449. Compared with mice receiving vehicle alone or AraC alone, mice treated with RK-20449 alone and RK-20449 after AraC exhibited a reduction in PB hCD45+ AML chimerism. As expected, in vehicle-alone recipients, the percentage of hCD45+ cells in the recipient PB increased over time, and in vivo treatment with AraC alone did not reduce the frequency of hCD45+ AML cells (representative flow cytometry plots are shown in Fig. 5C). The mice receiving vehicle alone or AraC alone became moribund at 7 to 40 days of treatment as PB hCD45+ AML cell frequency increased and approached 100% (Fig. 5E, left panel). In contrast, RK-20449 either alone or after AraC reduced hCD45+ cells in the recipient PB to nearly undetectable levels (representative flow cytometry plots are shown in Fig. 5C). In PB of all engrafted recipients treated with RK-20449 as a single agent or after AraC, the percentage of hCD45+ rapidly decreased, approaching zero by 2 to 3 weeks of treatment (Fig. 5E, left panel). In recipients engrafted with AML case #2, another AML with primary resistance to AraC-containing induction chemotherapy, in vivo treatment with RK-20449, either alone or after AraC, resulted in nearly complete elimination of hCD45+ AML cells in the recipient PB [Fig. 5, D and E (right panel)]. For both cases #1 and 2, RK-20449–treated recipients remained healthy and showed no weight loss during 15 weeks of RK-20449 treatment. The BM and spleen of recipients treated with RK-20449 alone or RK-20449 after AraC showed a significant reduction in absolute numbers of viable hCD45+ and hCD34+CD38 cells compared with controls (Fig. 5, F and G).

Macroscopic examination of control recipient mice revealed tibiae that were white, indicating a lack of erythroid cells, and enlarged spleens (Fig. 6A). In contrast, RK-20449–treated recipients demonstrated tibiae with red marrow, indicating recovery of murine erythropoiesis, and spleens reduced in size compared with control recipients (Fig. 6A). These changes were present relatively early in the course of RK-20449 treatment, as seen in recipient tibia and spleen at day 6 of treatment (Fig. 6A). We confirmed the recovery of murine hematopoiesis and the elimination of human AML cells in RK-20449–treated recipient BM by flow cytometry (Fig. 6B). In control (no treatment) recipients, BM was nearly completely replaced by human AML cells. On day 6 after a single dose of AraC, there was a modest reduction in the frequency of CD45+ human AML cells in the recipient BM. This was accompanied by enrichment of hCD34+CD38 AML stem cell fraction, associated with elimination of non-stem hCD34+CD38+ and hCD34 cells. With RK-20449 treatment in vivo, BM hCD45+ chimerism declined rapidly, with a corresponding increase in mouse CD45+ hematopoietic cells and TER119+ erythroid cells. In addition, the frequency of hCD34+CD38 LSC fraction declined, indicating elimination of LSCs along with hCD34+CD38+ AML cells. Recovery of murine hematopoiesis was also demonstrated by expansion of lineagecKit+ mouse hematopoietic progenitors with 6 to 8 weeks of RK-20449 administration. Similar effects were observed in recipients treated with AraC followed by RK-20449.

Fig. 6 RK-20449 treatment associated with recovery of murine hematopoiesis.

(A) Top three rows: Macroscopic observation of tibia and spleen obtained from NSG recipients engrafted with case #1. Top row: Control recipient (no treatment). Second row: AraC- and RK-20449–treated recipients on treatment day 6. Third row: RK-20449–treated recipient at treatment day 52. Bottom row: Tibia and spleen from NSG recipients engrafted with normal human CB HSCs for comparison. (B) BM of NSG recipients engrafted with case #1 showing mouse TER119+ erythroid cells, mouse CD45+ leukocytes, and mouse lineageSca-1+cKit+ primitive hematopoietic stem/progenitor cells with RK-20449 administration. (From left) First column: Untreated case #1–engrafted recipient; second column, recipient analyzed at day 6 after AraC treatment; third and fourth columns, recipients analyzed at days 6 and 52 of RK-20449 treatment; fifth column, recipient analyzed at day 45 of AraC followed by RK-20449 treatment.

To demonstrate the in vivo effect of RK-20449 on human primary AML cells in situ, we performed histopathological examination of organs from recipients engrafted with case #1. In a control recipient, femur, sternum, and spleen were packed with hCD45+ AML cells (Fig. 7, A and B, and fig. S12, A and B, left panels). In contrast, on day 6 of in vivo RK-20449 treatment, clearance of hCD45+ AML cells as well as recovery of murine erythroid cells was observed (Fig. 7, A and B, and fig. S12, A and B, middle panels). On day 16 of treatment, human AML cells were nearly completely eliminated in the recipient femur, sternum, and spleen, with the appearance of murine erythroid cells, megakaryocytes, and granulocytes (Fig. 7, A and B, and fig. S12, A and B, right panels; the cells of interest are indicated in high-magnification images in fig. S12A). In the BM of multiple bones and in the spleen of a recipient maintained 52 days on RK-20449, hCD45+ AML cells remained infrequent, whereas murine hematopoiesis continued to recover (Fig. 7C, middle panels). In contrast, AraC alone–treated recipient BM was obliterated with human AML cells, as expected from flow cytometric data showing little effect of in vivo AraC in case #1–engrafted recipients as well as from the clinical course of the patient (Fig. 7C, left panels). However, AraC treatment followed by in vivo administration of RK-20449 resulted in nearly complete elimination of human AML cells associated with recovery of normal murine hematopoiesis (Fig. 7C, right panels). Similar findings were observed in case #2–engrafted recipients treated with RK-20449 as a single agent (fig. S13). These findings demonstrate that reduction of whole-body AML burden is achieved and is sustained long-term by in vivo RK-20449 treatment.

Fig. 7 Eradication of hCD45+ AML cells from hematopoietic organs of engrafted recipients and elimination of human AML stem cells with in vivo RK-20049.

(A and B) hCD45+ AML cells in situ in femur, sternum, and spleen of AraC-resistant AML case #1–engrafted recipients treated in vivo with RK-20449 as shown by (A) hematoxylin and eosin (H&E) and (B) hCD45 staining. In both (A) and (B): left column, untreated recipient; middle column, day 6 of in vivo RK-20449; right column, day 16 of in vivo RK-20449. (C) Low-magnification (×20) and high-magnification (×40) photomicrographs showing hCD45+ AML cells in situ in BM of AraC-resistant AML case #1–engrafted recipients after long-term in vivo treatment with RK-20449 alone or after AraC. hCD45 staining of bones and spleen from case #1 recipients treated with (left) AraC alone, (middle) in vivo RK-20449 alone for 52 days, and (right) in vivo RK-20449 after AraC for 50 days. (D) Frequency of hCD45+ AML cells in the BM of secondary transplantation recipients observed for up to 18 weeks after transplantation. Untreated, AraC-treated (sacrificed 7 days after 960 mg/kg ip × 1), or RK-20449–treated (30 mg/kg ip twice daily for 20 to 60 days) NSG mice engrafted with AML case #1, 2, or 17 were used as primary recipients. All secondary recipients from untreated or AraC-treated primary recipients showed human AML engraftment (total n = 22), whereas none of the secondary recipients of RK-20449–treated primary recipients showed engraftment (total n = 28).

Finally, to establish that RK-20449 inhibits human primary AML LSCs in vivo, we performed secondary transplantation of hCD45+ AML cells isolated from RK-20449–treated, human AML–engrafted recipient BM. Viable hCD45+ BM cells were FACS-purified from untreated, AraC-treated [sacrificed 7 days after 960 mg/kg intraperitoneally (ip) × 1], or RK-20449–treated (30 mg/kg ip twice daily for 20 to 60 days) primary recipients engrafted with cases #1, 2, and 17 and transplanted into secondary recipients. Among secondary recipients from untreated and AraC-treated primary recipients, 9 of 9 (case #1 primary), 10 of 10 (case #2 primary), and 3 of 3 (case #17 primary) showed human AML engraftment in PB by 10 weeks after transplantation, became moribund, and were sacrificed by 18 weeks after transplantation. Secondary recipients of RK-20449–treated, case #1, 2, or 17–engrafted BM hCD45+ cells did not show PB engraftment and remained well until they were sacrificed at 18 weeks after transplantation. Although secondary recipients of untreated or AraC-treated primary recipients showed BM engraftment by hCD45+ AML cells, secondary recipients of RK-20449–treated primary recipients showed no human hematopoietic engraftment, demonstrating elimination of LSCs by in vivo RK-20449 treatment (Fig. 7D).


Currently, the most important prognostic determinant for AML at diagnosis is clonal chromosomal abnormalities, which are found in 40 to 50% of cases and form the basis for cytogenetic risk groups. In AML where chromosomal aberrations are undetected, molecular defects with significant prognostic implications have been identified, including mutations in nucleophosmin 1 (NPM1), CCAAT enhancer–binding protein α (CEBPA), mixed-lineage leukemia (MLL), KIT, and FLT3 genes (3638). Of these, about 50% are activating mutations in receptor tyrosine kinases such as FLT3 and KIT (39). Moreover, the involvement of nonreceptor protein tyrosine kinases in the pathogenesis of AML is suggested by the deregulation of intracellular signaling pathways associated with cell proliferation and survival, such as mitogen-activated protein kinase (MAPK), phosphatidylinositol 3-kinase (PI3K)/Akt, mTOR, and nuclear factor κB (NF-κB) (4046). In particular, SFKs have been implicated as intermediates in signaling pathways that regulate cellular functions such as proliferation, differentiation, survival, and migration in various cell lineages (13), and HCK and other SFKs are expressed in AML cells obtained from patients (39).

RK-20449 is a multikinase inhibitor with activity against a relatively restricted range of human kinases. For instance, RK-20449 did not significantly inhibit the PI3K pathway, whose correct temporal regulation is required for normal hematopoiesis including B lymphocyte development, myeloid differentiation, and HSC maintenance (47). In addition, KIT, the receptor for SCF, involved in normal HSC development, myeloid differentiation, as well as mast cell development and function, was not significantly inhibited by RK-20449 (48, 49). RK-20449 also does not highly inhibit ABL, consistent with a previous report that RK-20449 inhibits the proliferation and survival of BCR-ABL+ CML cells through HCK (25). Reflecting the structural similarity among SFKs, RK-20449 inhibits other SFKs at nanomolar concentrations. Among them, Lyn is frequently expressed in AML, and inhibition of Lyn by RK-20449 may inhibit AML (39). In addition, RK-20449 inhibits FLT3 with IC50 in the range of 500 nM to 1 μM, suggesting a role for this agent in FLT3-mutated AML. At the same time, HCK expression appears to be important in FLT3-mutated AML because HCK knockdown significantly inhibits proliferation of FLT3ITD+ AML cell lines. The elucidation of the role of HCK-FLT3 interaction in AML awaits further investigation.

In vivo activity of RK-20449 was demonstrated in the newborn NSG xenotransplantation system, in which recipient mice engrafted with primary human LSCs develop AML with high whole-body leukemia burden (10). As a single agent, RK-20449 treatment resulted in significant reduction of total human AML cells as well as human LSC burden in BM and spleen of recipients. Recovery of murine leukocytes and erythroid and megakaryocyte lineage along with cKit+ stem/progenitors in recipients maintained on RK-20449 long-term suggests the presence of a therapeutic window. In addition, RK-20449 was effective in vivo in mice engrafted with two cases (#1 and 2) of FLT3ITD+ AML. HCK, LYN, and Src associate with FLT3 and FLT3-ITD through their SH2 (Src homology 2) domains, and inhibition of SFKs in FLT3ITD+ AML cell lines impairs growth and survival (19, 21, 50). The in vivo activity of RK-20449 against FLT3ITD+ AML cases may be related to interactions of HCK and other SFKs with FLT3 affecting its activation and signaling.

Although we first identified HCK as a target molecule for anti-LSC therapy because of its selective expression in LSCs compared with HSCs, whether small-molecule inhibitors such as RK-20449 has significant selectivity against AML cells over normal hematopoietic elements remains to be seen. To begin to address this question, we examined the effect of RK-20449 on human HSCs in vitro (by the assessment of long-term, multilineage hematopoietic repopulation capacity in vivo in NSG recipients) and on mature human hematopoietic lineages in vivo (by the administration of RK-20449 to NSG recipients repopulated with normal human hematopoiesis). The in vivo hematopoietic repopulation capacity of normal human HSCs appears not to be impaired by in vitro exposure to RK-20449. With in vivo administration of RK-20449, there is some decline in the frequency of myeloid lineage cells, whereas lymphoid lineage cells appear less affected. Further testing is required to clarify the effect of in vivo RK-20449 administration on normal hematopoiesis.

In CML, the discovery of the BCR-ABL kinase inhibitor imatinib has led to substantial improvement in patient survival, establishing kinase inhibition as first-line therapy. However, recent studies have indicated that disease-initiating cells in CML survive BCR-ABL inhibition, leading to relapse upon withdrawal of kinase inhibitor therapy. RK-20449 offers a potential avenue for preventing and overcoming AML relapse through elimination of disease-initiating human LSCs.

Materials and Methods

Human samples

All experiments were performed with authorization from the Institutional Review Board for Human Research at RIKEN RCAI. All patient samples were collected with written informed consent. CB samples were collected by Tokyo Cord Blood Bank with written informed consent. AML BM mononuclear cells (MNCs) and CB MNCs were isolated by density-gradient centrifugation.


NOD.Cg-PrkdcscidIl2rgtmlWjl/Sz (NSG) mice were developed at The Jackson Laboratory by backcrossing a complete null mutation at the Il2rg locus onto the NOD.Cg-Prkdcscid (NOD/SCID) strain. Mice were bred and maintained under defined flora at the animal facility of RIKEN and at The Jackson Laboratory according to guidelines established by the Institutional Animal Committees at each institution.

Flow cytometry and FACS

The following fluorochrome-conjugated monoclonal antibodies (mAbs) were used for flow cytometry: mouse mAbs to hCD45, hCD34, hCD38, hCD33, hCD3, hCD4, hCD8, and hCD19; rat mAbs to mCD45, mCD117, mSca-1, mGr1, mCD11b, and mTER119 (BD Biosciences). Analyses were performed with FACSAria and FACSCanto II (BD). To obtain cells for xenogeneic transplantation, MNCs obtained from AML patient BM or PB were labeled with fluorochrome-conjugated mouse mAbs to hCD3, hCD4, hCD8, hCD34, and hCD38, and recipient BM MNCs were labeled with mouse mAbs to hCD45, hCD34, and hCD38 followed by cell sorting with FACSAria (BD). The purity of sorted cells was >98%.

Xenogeneic transplantation

Newborn NSG recipients received 150 of cGy total body irradiation followed by intravenous injection of sorted cells. To generate AML-engrafted recipients, 103 to 105 sorted 7AADlineage (hCD3/hCD4/hCD8)hCD34+hCD38 AML patient BM cells were injected per recipient as described (10). PB human cell engraftment was assessed by retro-orbital phlebotomy.

Immunofluorescence labeling and imaging

Paraformaldehyde-fixed, decalcified, paraffin-embedded femoral sections were labeled with mouse mAbs to hCD45 (DAKO) and hHCK (Novus Biologicals) for confocal imaging. Laser-scanning confocal imaging was obtained with Zeiss LSM 710 (Carl Zeiss). H&E staining was performed by standard methods. Mouse anti-hCD45 mAb (DAKO) was used for immunohistochemical labeling of hCD45.

Lentiviral transduction and HCK knockdown

Cell lines TF1a, K562, and SV-4-11 (American Type Culture Collection) and Molm13 (German Collection of Microorganisms and Cell Cultures) were maintained under conditions specified by the provider. Human primary AML CD34+CD38 cells were FACS-purified either directly from patient BM or PB or from human AML–engrafted recipient BM. Lentivirally packaged HCK shRNA and control GFP shRNA were obtained from Sigma. Cells were infected at a multiplicity of infection (MOI) of 100 for 4 days; harvested; labeled with mouse mAbs to hCD45, hCD34, and hCD38; and sorted to obtain HCK shRNA– or control GFP shRNA–transduced (GFP-positive) hCD34+hCD38 (TF1a) or hCD45+ (K562, SV-4-11, and Molm13) cells. Sorted cells were then cultured in cell type–appropriate media in the case of cell lines or in Hematopoietic Progenitor Growth Medium (HPGM; Lonza) supplemented with recombinant human SCF (50 ng/ml; Wako), recombinant human TPO (50 ng/ml; PeproTech), and recombinant human FLT3 ligand (50 ng/ml; PeproTech) in the case of primary human AML cells. After photomicrographs were taken with Zeiss Axiovert 200 (Carl Zeiss Microimaging), cells were harvested and viable cell numbers were determined by flow cytometry with AccuCount Beads (BD). To demonstrate the level of gene expression suppression by shRNA, total RNA was extracted from HCK shRNA–transduced (GFP positive) hCD34+hCD38 TF1a cells. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed on complementary DNA (cDNA), amplified from total RNA with WT-Ovation RNA Amplification System (Nugen), with Platinum Quantitative PCR SuperMix (Invitrogen) on LightCycler 480 (Roche). The sequences of dual-labeled fluorogenic probes and gene-specific primers (Sigma-Aldrich) are listed in table S5. The absolute copy number of each transcript (HCK, LYN, SRC) was calculated by the standard curve method, using quantified PCR product–containing pCR 2.1-TOPO vectors (Invitrogen) as standard samples. Hypoxanthine phosphoribosyltransferase 1 (HPRT1) was used as an internal control. For initial normalization to the internal control, the relative copy number of each transcript was calculated as the ratio to the absolute copy number of HPRT1. The relative expression level of each transcript after knockdown experiment with shRNA was calculated as the percent of the relative copy number of non–target shRNA–transfected control sample.


We obtained 47,574 commercially available compounds, with a mean molecular weight of 329.6 and an ALogP of 3.32, from ASINEX Ltd., ChemBridge Corp., and Maybridge Chemical Co. Ltd. In addition, for in silico screening, we obtained 2D structures and supplier information of commercially available compounds distributed by Namiki Shoji Co. Ltd. and Kishida Chemical Co. Ltd. From this information, an integrated database of 8,927,055 commercially available compounds with distinct structures was prepared and used for in silico screening. From the in silico screening, 3088 compounds were then purchased from the companies. The compounds were dissolved in DMSO and stored at −20°C as 10 mM stock solutions. Absorption, distribution, metabolism, elimination, and toxicity (ADMET) assay was performed by Cerep.

In silico screening and design

Published HCK structures and small-molecule inhibitors of HCK and SFKs were collected from the PDB, Thomson Reuters Integrity patented compound database, and ChEMBL published small ligand database (51). As of 2010, seven HCK x-ray crystal structures and 36 HCK inhibitors and 141 inhibitors for other SFKs with IC50 values of less than 10 μM were deposited, and these were used for in silico screening. On the basis of published structure information, we performed in silico screening for HCK inhibitors from 8,927,055 commercially available compounds (as described in the “Compounds” section) by ligand-based prediction (similarity searches and machine-based learning) and structure-based docking simulation (5254). We used similarity searches to select compounds with molecular similarity to known inhibitors from the viewpoints of 2D structure, 2D pharmacophore locations, and 3D shape and pharmacophore locations. To do so, we calculated the Tanimoto coefficient between 36 known inhibitors and each of the 8,927,055 compounds on the basis of 2D fingerprints (MACCS public keys, ECFP4, and GpiDAPH3) and 3D shape metrics (ComboScore) with the software Pipeline Pilot version 7.5 (Accelrys, 2008), MOE 2009.10 (Chemical Computing Group), and OpenEye ROCS version 3.0.0 (OpenEye Scientific Software, 2010). To construct machine-learning models, we first prepared a training data set containing 36 known HCK inhibitors (positive cases) and 5000 randomly selected decoy compounds (negative cases) calculated by MOE. Using this training data set, we constructed two machine-learning models, one based on 186 MOE 2D descriptors and another based on GpiDAPH3 fingerprints, using Random Forest on Pipeline Pilot R component (Accelrys, 2008). We then screened 8,927,055 compounds using these machine-learning models and selected positively predicted compounds. For docking simulation, we first validated the efficiency of virtual screening and the reproducibility of the x-ray conformation of 36 known HCK inhibitors and seven published HCK crystal structures using the Schrodinger Glide software (30). Among the seven HCK structures, we selected PDB entries 2C0I and 2HCK as protein structures optimized for detecting diverse inhibitors for all docking conditions. We then performed docking simulation of 8,927,055 commercially available compounds with the two HCK protein structures and selected compounds according to the docking score and important protein-ligand interactions of known HCK inhibitors. Using these in silico screening methods, we selected 3088 compounds for kinase inhibition assay.

Kinase mobility shift assay

Kinase mobility shift assay was performed as reported (26). Briefly, reaction buffer contained 50 mM Hepes (pH 7.4), 10 mM MgCl2, 0.01% Brij-35, 1 mM dithiothreitol (DTT), 1% Protease Inhibitor Cocktail Set V (Calbiochem), 1% Phosphatase Inhibitor Cocktail Set III (Calbiochem), and 0.6% DMSO. Purified HCK sample [75 to 526 amino acids of NCBI (National Center for Biotechnology Information) Reference Sequence Database accession number NP_002110.3] was prepared and purified as described below. HCK in the reaction buffer was first incubated with various concentrations of the compounds (0 to 10 μM) for 30 min. After this preincubation step, FAM-labeled substrate peptide [FL-Peptide 4 (5-FAM-EGIYGVLFKKK-CONH2) (Caliper Life Sciences)] and ATP were added to final concentrations of 1.5 μM and 0.5 mM, respectively. The kinase reactions were performed at room temperature for 30 min and stopped by the addition of termination buffer [50 mM Hepes (pH 7.4), 140 mM EDTA, and 0.01% Brij-35]. The phosphorylated and unphosphorylated peptides obtained were separated and quantified with a Lab-Chip EZ Reader II (PerkinElmer). To determine IC50, we measured kinase activity in triplicate with the assay mixture containing 1 mM ATP.

Protein expression for structural study

The human HCK gene fragment, encoding residues 81 to 526, was PCR-amplified and cloned into the modified pDEST 10 vector (Invitrogen), in which sequences around the tobacco etch virus (TEV) cleavage site were changed. To make the mutant vector HckQQQ523EEI that has the same amino acid sequence as reported [PDB ID: 1QCF (29)], we changed residues 523 to 525 (QQQ) to EEI by PCR and Dpn I digestion. The DH10Bac Escherichia coli cells (Invitrogen) were transformed with the HckQQQ523EEI vector and were grown to isolate the recombinant bacmid DNA. The baculovirus was generated by transfecting the bacmid DNA into Spodoptera frugiperda Sf9 cells and was amplified to increase in titers. Sf9 cells in suspension culture, grown to a cell density of 2 × 106 cells/ml, were infected with the baculovirus at an MOI of 1 and were incubated at 27°C for 48 hours. The Sf9 cells were collected, frozen, and stored at −80°C before use.

HCK purification

The Sf9 cells were sonicated in 20 mM tris-HCl buffer (pH 8.0) containing 20 mM imidazole, 500 mM NaCl, 10% glycerol, and 0.1 mM DTT. The insoluble cell debris was removed by centrifugation at 30,000g at 4°C for 30 min, and the supernatant was applied on a HisTrap column (GE Healthcare) with an AKTA prime system. The protein was eluted by a linear gradient of 20 to 500 mM imidazole in 20 mM tris-HCl buffer (pH 8.0) containing 500 mM NaCl, 10% glycerol, and 0.1 mM DTT. The protein was collected and mixed with TEV protease and was dialyzed against 20 mM tris-HCl buffer (pH 8.0) containing 100 mM NaCl, 20 mM imidazole, and 0.1 mM DTT for overnight. The sample was centrifuged at 30,000g at 4°C for 30 min to remove precipitated contaminating proteins and then applied to a HisTrap column equilibrated with 20 mM tris-HCl buffer (pH 8.0) containing 100 mM NaCl, 20 mM imidazole, 10% glycerol, and 0.1 mM DTT. The flow-through was collected and applied to the MonoQ column (GE Healthcare) using an AKTA 10S system and eluted by a linear gradient of 100 to 800 mM NaCl in 20 mM tris-HCl buffer (pH 8.5) containing 2 mM DTT. The eluted fractions were collected and concentrated up to 3 ml, using an Amicon Ultra Centrifugal Filter [molecular weight cutoff (MWCO) 10,000; Millipore], and were subjected to gel-filtration chromatography (HiLoad 16/60 Superdex 75, GE Healthcare) using an AKTA system equilibrated with 20 mM tris-HCl buffer (pH 8.0) containing 200 mM NaCl, 10% glycerol, and 3 mM DTT. The purity of the protein was more than 98%, as judged by SDS–polyacrylamide gel electrophoresis. The protein yield was 1.2 mg per liter of culture. The purified protein was concentrated to 10 mg/ml, using an Amicon Ultra Centrifugal Filter (MWCO 10,000). Before crystallization, the protein sample was mixed with 30% (w/v) 1,5-diaminopentane dihydrochloride (Hampton) so that its final concentration becomes 3% (w/v). The compounds were dissolved in DMSO and stored at −80°C.

Crystallization, data collection, and data processing of HCK complexed with compounds

Crystals were grown by the sitting drop vapor diffusion method. Cocrystals of HCK with RK-24466, RK-20445, RK-20450, and RK-20451 were grown at 4°C and with RK-20444 and RK-20446 at 15°C. One microliter of the HCK protein was mixed with an equal volume of the reservoir solution consisting of 100 mM tris-HCl buffer (pH 7.4 to 8.0), containing 90 to 100 mM calcium acetate, 18 to 20% glycerol, and 20 to 22% polyethylene glycol 8000 (PEG 8000), and equilibrated against 70 μl of reservoir solution. Crystals were grown to the maximal size in 2 weeks. RK-20448 and RK-20449 were soaked into the cocrystal with RK-20446, in 100 mM tris-HCl buffer (pH 7.0 to 8.0), containing 100 mM calcium acetate, 20% glycerol, 0.5 mM compound, and 22% PEG 8000, for 1 or 2 days. For data collection, crystals were transferred to 100 mM tris-HCl buffer (pH 8.0), containing 100 mM calcium acetate, 30% glycerol, 24% PEG 8000, 0.5 to 1 mM compound, and 2% DMSO, and flash-cooled in liquid nitrogen. The statistics of the x-ray data set collection are summarized in table S1. The data were processed with the program XDS (RK-24466, RK-20444, RK-20445, RK-20448, and RK-20450) (55) or with the program HKL2000 (RK-20446, RK-20449, and RK-20451) (56). The phases were determined by molecular replacement by the program PHASER (57), using the coordinates of HCK [PDB ID: 2C0T (58)] as the search model. The asymmetric unit contains two HCK molecules. The model was built and corrected with the program COOT (57, 59). The statistics of model building and refinement are summarized in table S1. The Ramachandran plot was calculated with the program Procheck (57).

Panel kinase inhibition assay

Inhibition of a panel of 106 human kinases by RK-20449 was assayed by Carna Biosciences. Reaction buffer contained 20 mM Hepes (pH 7.4), 0.01% Triton X-100, 2 mM DTT (pH 7.5), and 5 mM MgCl2 (additional 1 mM Mn for CSK and EGFR, 25 μM sodium orthovanadate for KIT). The assay mixture contained various concentrations of RK-20449 (0 to 3 μM) with 1 μM substrate and ATP (at Km of 5 to 400 μM for each kinase). The kinase reactions were performed at room temperature for 1 hour (5 hours for CSK) and stopped by the addition of termination buffer (QuickScout Screening Assist MSA, Carna BioSciences). The phosphorylated and unphosphorylated peptides obtained were separated and quantified with the LabChip3000 system (Caliper Life Sciences). The kinase activities were measured in triplicate, and the IC50 values were calculated.

In vitro culture assay for primary human AML LSCs and CB HSCs

Human primary AML CD34+CD38 cells were FACS-purified either directly from patient BM/PB or from human AML–engrafted recipient BM. 7AADlineage (hCD3/hCD4/hCD8)hCD34+hCD38 HSCs were FACS-purified from healthy donor CB. The sorted cells were cultured in HPGM (Lonza) supplemented with recombinant human SCF (50 ng/ml; Wako), recombinant human TPO (50 ng/ml; PeproTech), and recombinant human FLT3 ligand (50 ng/ml; PeproTech) in the presence of kinase inhibitors at the concentrations indicated or DMSO alone in wells of 96-well tissue culture plates. After 3 days, the cells were harvested from the wells, and cell number, surface phenotype, and viability were analyzed by flow cytometry. To examine the leukemia-initiating capacity of treated AML CD34+CD38 cells and the hematopoietic repopulation capacity of treated CB CD34+CD38 cells, 7AADCD45+ cells were sorted from cells harvested from each well at the indicated concentrations of test compound and transplanted into three NSG newborns as described above. Transplant recipients underwent retro-orbital phlebotomy every 3 weeks starting at 6 weeks after transplantation to assess AML engraftment. Survival was estimated with the Kaplan-Meier method.

Phospho-specific flow cytometry

BM MNCs obtained from primary human AML–engrafted NSG recipients were cultured as above in HPGM (Lonza) supplemented with cytokines in the presence of DMSO alone, 1 μM RK-20449, or 1 μM rapamycin and incubated for 90 min at 37°C. Cells were fixed with Fix/Lyse buffer, permeabilized with Perm Buffer III, and resuspended in Stain Buffer with mouse Fc Block (clone 2.4G2) according to the manufacturer’s instructions (BD). Cells were then labeled with phospho-specific antibodies and analyzed with FACSCanto II (BD). The following phospho-specific antibodies were used: Alexa Fluor 488–conjugated mouse IgG1k isotype control (clone MOPC-21), mouse anti-STAT5 (pY694) (clone 47), mouse anti-Src pY418 (clone K98-37), mouse anti-S6 (pS235/pS236) (clone N7-548), and mouse anti–4E-BP1 (pT36/pT45) (clone M31-16) (all from BD). Because of sequence similarities in the protein kinase domain, anti-Src pY418 recognizes homologous phospho-tyrosine residues in HCK (pY410). P-flow scores were calculated as described (60).

In vivo treatment of human primary AML–engrafted recipients and secondary transplantation

For all in vivo treatment experiments, the numbers of recipients engrafted with AML from each patient and for each treatment group are detailed in the figure legends. The in vivo treatment experiments were performed in sets of AML-engrafted littermates transplanted on the same day with the same number of hCD34+CD38 cells. Each set contained two to four recipients. The littermates were selected for in vivo treatment study when their PB chimerism reached 20% or above. Pretreatment PB hCD45+ chimerism was similar, and treatment assignments were made randomly within each set of littermates. For short-term in vivo treatment study, NSG littermates engrafted with primary AML from seven patients (cases #1 to 7) received vehicle alone or RK-20449 (30 mg/kg ip twice daily) for 14 to 21 days. For in vivo treatment study on recipients engrafted with AML cells obtained from patients with chemotherapy-resistant disease (cases #1 and 2), littermates received vehicle alone, RK-20449 alone (30 mg/kg ip twice daily), AraC alone (960 mg/kg ip), or AraC (960 mg/kg ip) followed by RK-20449 (30 mg/kg ip twice daily). The mice were sacrificed when they became moribund or after 8 weeks of treatment, and human AML chimerism in BM, spleen, and PB; total numbers of cells in BM and spleen; and PB complete blood count were obtained. Secondary transplantation of viable hCD45+ FACS-purified from case #1, 2, or 17–engrafted primary recipient BM into 150 cGy–irradiated newborn NSG secondary recipients was performed. Untreated, AraC-treated (sacrificed 7 days after 960 mg/kg ip × 1), or RK-20449–treated (30 mg/kg ip twice daily for 20 to 60 days) NSG mice engrafted with AML case #1, 2, or 17 were used as primary recipients. To compare the effect of each treatment on AML stem cell burden, each recipient was transplanted with viable hCD45+ cells derived from the same number of primary recipient BM cells. Secondary recipients were sacrificed when moribund or at 18 weeks after transplantation.

Statistical analysis

Numerical data are presented as means ± SEM. The differences were examined with two-tailed t test (GraphPad Prism, GraphPad). Survival was estimated by the Kaplan-Meier method, and the curves were compared by log-rank (Mantel-Cox) test.

Supplementary Materials

Fig. S1. Inhibition of human AML cell line TF1a and human CML cell line K562 by HCK shRNA.

Fig. S2. Crystal structure of HCK showing target regions of small-molecule inhibitors.

Fig. S3. In vitro assay for human primary AML LSCs.

Fig. S4. RK-24466 suppresses human LSC proliferation in vitro.

Fig. S5. Chemical structures of compounds complexed with HCK for crystal structure determination.

Fig. S6. Crystal structures of candidate HCK inhibitors complexed with HCK.

Fig. S7. Inhibition of HCK expression by shRNA in FLT3ITD+ human AML cell lines MV-4-11 and Molm13.

Fig. S8. Representative plots showing changes in phosphorylation after in vitro exposure to RK-20449.

Fig. S9. In vitro treatment by RK-20449 reduces the in vivo leukemia initiation capacity of human primary AML LSCs, whereas RK-20449–treated CB CD34+CD38 cells maintain the capacity for long-term multilineage hematopoietic reconstitution in vivo.

Fig. S10. In vivo treatment of CB-derived HSC–engrafted NSG recipients with RK-20449.

Fig. S11. In vitro assay of PP1, PP2, RK-24466, RK-20449, and AraC against CD34+CD38 cells derived from case #1.

Fig. S12. RK-20449 treatment eliminates chemotherapy-resistant human AML cells (case #1) in vivo (high-resolution images of Fig. 7, A and B).

Fig. S13. RK-20449 treatment eliminates chemotherapy-resistant human AML cells (case #2) in vivo.

Table S1. Data collection and refinement statistics.

Table S2. Seven derivatives of RK-24466 based on in silico structure prediction.

Table S3. Enzymatic IC50 of RK-20449 against a panel of 106 human kinases.

Table S4. IC50 of PP2, RK-24466, and RK-20449 against primary human AML LSC proliferation.

Table S5. qRT-PCR probe and gene-specific primer information.

References and Notes

  1. Acknowledgments: We thank R. Takeuchi and M. Yasuda for administrative assistance. We thank N. Ohsawa for plasmid construction, M. Ikeda for protein expression using the baculovirus/Sf9 system, K. Honda for protein purification, and M. Toyama, H. Niwa, and L. Parker for help in x-ray data collection; the Beamline staff, K. Hirata and Y. Kawano at BL32XU (SPring-8), K. Hasegawa and H. Okumura at BL41XU (SPring-8), N. Matsugaki at BL1A (Photon Factory), and T. Caradoc-Davies and N. Cowieson at MX2 (Australian Synchrotron) for help in x-ray data collection; and T. Kanabayashi for preparation of paraffin-embedded sections and CD45 immunostaining. Funding: This work was supported by the Japan Society for the Promotion of Science (JSPS) through the “Funding Program for Next Generation World-Leading Researchers (NEXT Program),” initiated by the Council for Science and Technology Policy (CSTP) (to F.I.), the Targeted Proteins Research Program (TPRP), the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan (to S.Y.), and NIH Cancer Core grant CA34196 (The Jackson Laboratory/L.D.S.). This research is granted by the JSPS through the “Funding Program for World-Leading Innovative R&D on Science and Technology (First Program).” Author contributions: Y.S. and F.I. designed and performed the research, analyzed the data, and wrote the manuscript. H.Y. and T.H. designed and performed the in silico screening and docking simulations, analyzed the data, and wrote the manuscript. M.K. designed and performed the crystallographic experiments and wrote the manuscript. Y.H. designed and synthesized small molecules. S. Takagi, Y.N., A.S., I.O., M.T.-M., Y.A., A.K., S. Tanaka, N.S., and H.K. performed the experiments. A.T. and J.M. designed and performed the high-throughput screening, analyzed the data, and wrote the manuscript. M.S. designed and performed the crystallographic experiments and reviewed the manuscript. N.H. and M.W. designed and performed the crystallographic experiments. Y.T. performed the crystallographic experiments. N.U. and S. Taniguchi were involved in study design, provided patient information and samples, and reviewed the manuscript. O.O., T.F., and T.G. were involved in study design and reviewed the manuscript. L.D.S. created the NSG xenograft model and reviewed the manuscript. S.Y. designed and supervised the crystallographic part of the project and reviewed the manuscript. Competing interests: Japanese patent application no. 2012-167553 (Agent for treating and preventing relapse in AML) has been filed relating to the designed compounds that target human AML stem cells. The authors declare that they have no competing interests. Data and materials availability: Coordinates and structure factors have been deposited in PBD under accession codes 3VRY (RK-24466), 3VRZ (RK-20444), 3VS0 (RK-20445), 3VS1 (RK-20446), 3VS2 (RK-20448), 3VS3 (RK-20449), 3VS4 (RK-20450), and 3VS5 (RK-20451).
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article