Structural Analysis Identifies Imidazo[1,2-b]Pyridazines as PIM Kinase Inhibitors with In vitro Antileukemic Activity

Much attention has recently been focused on PIM kinases as potential targets for the treatment of hematopoietic malignancies and some solid cancers. Using protein stability shift assays, we identified a family of imidazo[1,2-b]pyridazines to specifically interact with and inhibit PIM kinases with low nanomolar potency. The high-resolution crystal structure of a PIM1 inhibitor complex revealed that imidazo[1,2-b]pyridazines surprisingly interact with the NH₂-terminal lobe helix αC rather than with the kinase hinge region. Thus, the identified inhibitors are ATP competitive but not ATP mimetic compounds, explaining their enhanced selectivity with respect to conventional type I kinase inhibitors. One of the identified imidazo[1,2-b]pyridazines (K00135) was further tested in several hematopoietic cellular systems. First, K00135 dose-dependently impaired survival of murine Ba/F3 cells that have been rendered cytokine independent by overexpression of human PIMs. Second, K00135 impaired survival and clonogenic growth of a panel of human acute leukemia cells. Third, exposure of K00135 significantly suppressed in vitro growth of leukemic blasts from five acute myelogenous leukemia patients but not of normal umbilical cord blood mononuclear cells. In vitro kinase assays and immunoblotting using lysates from human MV4;11 leukemic cells showed inhibition of phosphorylation of known PIM downstream targets, such as BAD and eukaryotic translation initiation factor 4E–binding protein 1, by K00135. Taken together, we report a family of small molecules that selectively interact and block PIM kinases and could serve as a lead to develop new targeted antileukemic therapeutics. [Cancer Res 2007;67(14):6916–24]

Genetic alterations leading to uncontrolled protein tyrosine kinase activity are a hallmark of myeloproliferative disorders. Fusion genes involving ABL or platelet-derived growth factor receptor are the molecular correlate of chronic myeloid leukemia, whereas activating mutations of FLT3 or KIT are recurrently found in human acute myelogenous leukemia (AML; refs. 1, 2). The success of small molecules that block oncogenic tyrosine kinase activity, like imatinib-mesylate (Gleevc, Novartis), provided a proof of principle for targeted antileukemic therapy (3). However, the successful clinical use of such specific inhibitors has been challenged by the development of drug resistance and a limited clinical efficacy in patients with acute leukemia (4). To overcome these limitations, identification of critical signaling mediators downstream of an oncogenic tyrosine kinase is essential to delineate new targets that would allow development of an efficient combined therapeutic approach. Strong evidence exists that most oncogenic tyrosine kinases mediate malignant transformation by parallel activation of several signaling pathways such as Janus-activated kinase/signal transducers and activators of transcription (STAT), phosphatidylinositol 3-kinase/Akt, Ras/Raf/mitogen-activated protein kinase, or nuclear factor κB (1, 2).

Retroviral gene tagging in c-myc–induced murine lymphomas led to identification of the PIM serine/threonine kinase family that, in humans, encompasses three members: PIM1, PIM2, and PIM3 (5, 6). PIM1 has been characterized as one of the first target genes of STAT5 (7, 8). Several groups have shown that constitutive STAT5 activation is a hallmark of a wide spectrum of hematologic malignancies and that STAT5 is an essential mediator for transformation by oncogenic tyrosine kinases (9, 10). Interestingly, PIM1 and PIM2 have been found to be overexpressed in leukemia and lymphoma (11–13). Further in vitro functional studies have provided evidence that activation of PIM1 and/or PIM2 is important for malignant transformation by oncogenic tyrosine kinases such as FLT3 internal tandem duplications (FLT3-ITD), expressed in ∼25% of human AML cases (12, 14–17). These studies suggested that small molecules inhibiting PIM kinases could provide a promising therapeutic avenue for hematologic malignancies.

To facilitate development of lead compounds, several groups have reported the crystal structure of PIM1 in complex with a number of ATP mimetic kinase inhibitors with broad kinase inhibition profile (18–22). The crystal structures of PIM1 and PIM2 (pdb code 2IWI)⁴

showed that PIM kinases have an atypical hinge region characterized by an insertion of one additional residue and the presence of a proline that allows formation of only one hydrogen bond with ATP or ATP mimetic inhibitors. Despite this unique structural feature, all inhibitors published to date interact with the kinase hinge region in a classic ATP mimetic way (18–22).

Here, we report imidazo[1,2-b]pyridazines as inhibitors of PIM kinases. Despite high sequence homology and conservation of the active site, the identified inhibitors were on average 100-fold selective for PIM1 over PIM2. The selected lead compounds were highly selective and cross-reacted only with one additional kinase (Cdc-like kinase 1) when screened against a panel of 50 kinase catalytic domains. Determination of the cocrystal structure with K00135 suggested that the unusual binding mode of the studied inhibitor is responsible for the enhanced selectivity of this scaffold. In addition, we show that this inhibitor has in vitro antileukemic potential as shown in various cell line models as well as primary AML patient samples.

Protein expression and purification. Full-length human PIM1 (gi|33304198) was subcloned and protein was expressed and purified as previously described (21). The masses of purified proteins were confirmed by liquid chromatography-mass spectrometry using an Agilent LC/MSD TOF system with reversed-phase high-performance liquid chromatography coupled to electrospray ionization and an orthogonal time-of-flight mass analyzer. The purified PIM1 protein was homogeneous and had an experimental mass of 35,546 Da as expected from its primary structure.

Kinase assay. Phosphorylation reactions were monitored using a coupled-enzyme assay in which ADP production was coupled to NADH oxidation by pyruvate kinase and lactate dehydrogenase (23). The assay was carried out in 100 μL of a buffer containing 50 mmol/L HEPES (pH 7.5), 100 mmol/L NaCl, 10 mmol/L MgCl₂, 1 mmol/L phosphoenolpyruvate, 0.1 mmol/L NADH, 30 μg/mL pyruvate kinase, 10 μg/mL lactate dehydrogenase, and either 20 nmol/L PIM1 or PIM2. The reaction was monitored at 340 nm at 25°C on a Spectramax spectrophotometer (Molecular Devices) and started by addition of 100 μmol/L ATP after a 10-min preincubation at 25°C. A recognition peptide of the PIM1 substrate p21 (RKRRQTSMTD) was used at 30 μmol/L. DMSO-dissolved inhibitors were added at the preincubation period resulting in a 2% final DMSO. Kinetic analysis was done by nonlinear regression fitting using the program KaleidaGraph (Synergy Software).

Protein stability shift assay. Thermal melting experiments were carried out using the Mx3005p real-time PCR machine (Stratagene). Proteins (2 μmol/L, final) were assayed in 10 μL of 10 mmol/L HEPES (pH 7.5), 150 mmol/L NaCl in a 96-well plate. Inhibitors were added at a final concentration of 10 μmol/L. SYPRO Orange (Molecular Probes; dilution, 1:1,000) was added as a fluorescence probe. Excitation and emission filters for the SYPRO-Orange dye were set at 465 and 590 nm, respectively. The temperature was raised at 1°C/min from 25°C to 96°C and fluorescence readings were taken at each interval. The temperature dependence of the fluorescence during the protein denaturation process was approximated by the equation

where ΔuG is the difference in unfolding free energy between the folded and unfolded state, R is the gas constant, and y_F and y_U are the fluorescence intensity of the probe in the presence of completely folded and unfolded protein, respectively (24). The baselines of the denatured and native state were approximated by a linear fit. The observed temperature shifts, ΔT_m^obs, for each inhibitor were recorded as the difference between the transition midpoints of sample and reference wells containing protein without inhibitor in the same plate and determined by nonlinear least squares fit. All compounds were purchased from BioFocus.

Crystallization. Nonphosphorylated PIM1 sample was concentrated (final, 10 mg/mL) in the presence of ligands 0.7 mmol/L “pimtide” (ARKRRRHPSGPPTA-amide) and 1 mmol/L imidazopyridazine inhibitor. Crystals were grown in a sitting drop setup at 4°C by mixing 100 nL of this sample with 50 nL of precipitant [40% polyethylene glycol 300, 0.1 mol/L Tris (pH 8.5), 0.2 mol/L Li₂SO₄]. The imidazopyridazine inhibitor [1-(3-{6-[(cyclopropylmethyl)amino]imidazo[1,2-b]pyridazin-3-yl}phenyl) ethanone; referred as K00135] was purchased from BioFocus and added to a final 1 mmol/L from a 50 mmol/L stock in DMSO.

Structure determination. PIM1 diffraction data were collected on a cryofrozen crystal (100K) at the Swiss Light Source synchrotron beam line SLS-X10. Images were indexed and integrated using MOSFLM (25) and scaled using SCALA (26) implemented in the CCP4 (CCP4, 1994) suite of programs. The structure was solved using molecular replacement and the program Phaser (27) using the PIM1 in a complex with the bisindolyl maleide inhibitor BIM1 as a search model. Both structures were refined with REFMAC5 (28) using iterative rounds of rigid-body and restrained refinement with translation-libration-screw against maximum likelihood targets, interspersed with manual rebuilding of the model using Xfit/XtalView (29).

Coordinates. Coordinates have been deposited in the Protein Data Bank (pdb accession code 2C3I).

Reagents. Anti–phospho-BAD (Ser¹¹²), anti–phospho-BAD (Ser¹³⁶), anti-BAD, anti–phospho-eukaryotic translation initiation factor 4E–binding protein 1 (4E-BP1; Thr^37/46), and anti–phospho-4E-BP1 (Thr⁷⁰) rabbit polyclonal antibodies were purchased from Cell Signaling Technology. Anti-FLAG (clone M2) mouse monoclonal antibody (mAb) was obtained from Sigma-Aldrich. Anti-actin antibody and anti-PIM1 mouse mAb were from Santa Cruz Biotechnology and anti-PIM2 rabbit polyclonal antibody was from Abgent. Z-Leu-Leu-Leu-al (MG132) was purchased from Sigma-Aldrich.

Generation of stable cell lines expressing human PIM1/PIM2 and inhibitor treatment. The murine interleukin-3 (IL-3)–dependent hematopoietic cell line Ba/F3 was maintained in RPMI 1640 containing 10% FCS and 1 ng/mL recombinant murine IL-3 (PeproTech EC). The cDNAs encoding human PIM1 and PIM2 were reverse transcription-PCR (RT-PCR) amplified (with or without a FLAG tag) from mRNA extracted from human leukemic cell line K562 and cloned into EcoRI/XhoI sites of pMSCV-IRES-EYFP (PIM1) and into BgllI/EcoRI sites of pMSCV-IRES-EGFP (PIM2). High-titer retroviral supernatants were produced as previously described (17). Transduced Ba/F3 cells were selected for IL-3–independent growth. Expression of FLAG-tagged PIM1/PIM2 was verified by real-time RT-PCR and Western blotting (on treatment with 10 μmol/L MG123 for various amounts of time). Cellular growth curves were established by daily counting of the cultures at a starting density of 10⁵ cells/mL. To investigate the effect on the cell proliferation, K00135 (dissolved in DMSO) was added to the culture medium (0.1, 1, and 10 μmol/L) and 0.1% DMSO was added to the control cultures. Cell viability was analyzed by trypan blue dye exclusion and normalized as percentage to control cultures after 24 h.

Human cell lines and treatment with PIM inhibitor. Human leukemia cell lines MV4;11, RS4;11, MOLM13, KOCL45, SEM, KOPN8, and K562 were purchased from DSMZ and maintained in RPMI 1640 with 10% FCS and 1% penicillin-streptomycin at 37°C in 5% CO₂. To investigate the effect on the cell proliferation, either K00135 (in DMSO, 0.1, 1, 10 μmol/L final) or 0.1% DMSO was added to the culture medium and the cultures were incubated for 48 h (1 × 10⁵ cells in 1 mL). Cell proliferation and viability were assayed using 200 μL of the initial culture and Cell Proliferation Reagent WST-1 from Roche Diagnostics according to the manufacturer's instructions. Cell survival was calculated as a percentage normalized to control cultures and IC₅₀ values were calculated using Forecast function of Excel.

Primary AML samples. Peripheral blood samples were collected with informed consent from patients with newly diagnosed or recurrent AML. Mononuclear cells were separated by Ficoll-Histopaque and frozen in 10% DMSO. Cells were then either used for colony formation assays, as described below, or cultured in AIM-V medium (Cambrex) supplemented with 10% FCS, 1 mmol/L l-glutamine, and 1% penicillin-streptomycin. Cord blood was isolated from healthy donors under informed consent, and mononuclear cells were separated by Ficoll-Histopaque gradient and used as a control for compound-related cytotoxicity. To investigate the effect on cell proliferation, either K00135 (in DMSO, 1, 5, 10 μmol/L final) or 0.1% DMSO was added to the culture (1 × 10⁵ living cells/mL in 2 mL) and cell viability analyzed by trypan blue dye exclusion after 24 h.

In vitro colony forming assay. Cells (10⁴) were plated in methylcellulose (2 mL; MethoCult SF^BIT H4236, StemCell Technologies) supplemented with 10% FCS and either K00135 (1 or 4 μmol/L) or DMSO (0.1%) alone, and grown in 35-mm Petri dishes in a humidified atmosphere with 5% CO₂ at 37°C. Light microscopy was done 8 days later to assess colony formation (a colony was defined as a cluster of ≥50 cells). Mononuclear blood cells (5 × 10⁵) from the AML patients were plated in “complete” methylcellulose (2 mL; MethoCult GF H4534, StemCell Technologies) supplemented with either K00135 (1, 5, and 10 μmol/L) or 1% DMSO alone and grown as above. All cultures were inspected 10 days later for leukemic cell growth and pictures were taken.

Analysis of apoptosis. MV4;11 cells were treated with 1 μmol/L K00135 for 48 h and analyzed at various time points by flow cytometry using both violet fluorescent dye PO-PRO-1 (a sensitive indicator of apoptotic cells) and 7-amino-actinomycin D (indicator of dead cells) from the Vybrant Apoptosis Assay Kit #13 (Molecular Probes) according to the manufacturer's instructions.

Protein extraction and Western blotting. MV4;11 cells were grown at 3 × 10⁵/mL to 5 × 10⁵/mL, treated with 1 to 3 μmol/L K00135 for various times, harvested, and rinsed with ice-cold PBS. Ice-cold lysis buffer [10 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 1% Triton X-100, 0.5 mmol/L EDTA, 10% glycerol, 10 mmol/L NaF, 1 mmol/L Na₃VO₄, protease inhibitor cocktail] was added to the cells and incubated on ice for 20 min followed by 5-min microcentrifugation. Proteins were precipitated from supernatant using methanol-chloroform, and then pellets were resuspended in Laemmli's sample buffer, separated by SDS-PAGE, and transferred to the membranes. The membranes were blotted with primary antibodies (diluted according to the manufacturer's recommendations), followed by horseradish peroxidase–conjugated secondary antibodies, and the proteins detected by SuperSignal West Femto Maximum Sensitivity Substrate (Pierce). The same blots were stripped and reprobed with desired antibodies to confirm equal loading.

Immunoprecipitation and in vitro kinase assay. Cell lysates from 2 μmol/L K00135–treated MV4;11 cells (60 min) were prepared as described above and incubated with either no antibodies or 2 μg of anti-PIM1/PIM2 antibody with gentle rocking at 4°C overnight. Protein A-agarose beads (Upstate; 50 μL of 50% beads slurry in 500 μL of the lysate) were added and incubated for additional 2 h, followed by 15-s microcentrifugation, three washes with 300 mmol/L lysis buffer, and two additional washes with kinase buffer [40 mmol/L Tris-HCl (pH 7.4), 20 mmol/L MgCl₂, 0.1 mg/mL bovine serum albumin]. Beads were then suspended in 50 μL of kinase buffer supplemented with 500 μmol/L ATP and either no inhibitor or 2 μmol/L K00135 and incubated at 30°C for 30 min. Four micrograms of soluble BAD protein (Upstate) were then added to the reaction and incubated for additional 30 min. The reactions were terminated by addition of the Laemmli sample buffer and phosphorylation of BAD was analyzed by Western blotting with anti–phospho-BAD (Ser^112/136) antibodies.

Overall structure and substrate binding. The crystal structure of the ternary complex of PIM1 with a consensus substrate peptide and the inhibitor K00135 was determined at 1.9-Å resolution (Fig. 1; Supplementary Table S1) revealing the kinase with the typical bilobal architecture. The overall structure of the protein was found to be similar to structures that have previously been described (Fig. 1A; refs. 18–21). The serine/threonine kinase PIM1 has been shown to have a strong preference for substrates with basic residues, particularly arginine, at positions −5 and −3 (21, 22). Peptide library screening has defined additional selectivity for histidine at −2, proline at −1, and glycine at +1, facilitating the design of a high-affinity consensus peptide, pimtide (ARKRRRHPS*GPPTA). As in the case of the PIM1-BIM1 peptide complex (22), the substrate peptide was well defined between residues Lys⁶ and Gly¹, whereas the termini were not visible in the electron density and assumed to be disordered.

Inhibitor binding. Although the imidazo[1,2-b]pyridazine inhibitor binds to the ATP binding site of PIM1, surprisingly, it does not mimic binding of ATP by forming hydrogen bonds with the kinase hinge region (shown in magenta in Fig. 1B), but rather binds to the opposite side of this pocket participating in a hydrogen bond network formed by the conserved active site lysine (Lys⁶⁷), a structural water molecule, the αC residue of Glu⁸⁹, and the activation loop residue Phe¹⁸⁷ (Fig. 1B). In addition, a number of hydrophobic contacts, particularly with PIM1 residues Leu⁴⁴, Phe⁴⁹, Ile¹⁰⁴, and Leu¹²⁰, stabilize this interaction. The unusual hinge architecture of PIM1, which has a proline at the hinge position 123, allows formation of only one hydrogen bond to ATP or ATP mimetic inhibitors. We believe that this active site architecture, together with the unexpected binding mode of the studied imidazo[1,2-b]pyridazine inhibitor, results in surprisingly PIM-specific inhibitors. Indeed, the imidazo[1,2-b]pyridazine that cocrystallized with PIM1 interacted with only one other kinase (Cdc-like kinase 1) in a panel of more than 50 diverse serine/threonine kinase catalytic domains purified for cocrystallization studies in our laboratory (Fig. 2).⁵

We generated a small library of inhibitors based on the imidazo[1,2-b]pyridazine scaffold targeting PIM1 kinase and used temperature shift assays to identify and rank-order ligands that interact with PIM kinases. This assay monitors the linear affinity–dependent stability increase of proteins and has been shown to correlate well with binding constants and IC₅₀ values (22). In the study done here, we also confirmed an excellent correlation of temperature shift data (T_m) with IC₅₀ values and identified several inhibitors (K00152 and K00486) with improved potency compared with K00135 (Fig. 2A). The structure-activity relationship of the library was established with two varying positions (R1 and R2; Fig. 2B). The limited variations in position R2 did not dramatically influence T_m shift values. This is supported by the cocrystal structure, which showed that chemical moieties at that position interact with the glycine-rich loop but larger moieties would point toward the solvent. Aliphatic or saturated six-member ring systems at that position resulted in slightly more potent inhibitors than the tricyclic ring present in K00135. Within the inhibitor series tested, a trifluoro-methoxy moiety in meta position of the aromatic substitution in R1 showed the highest affinity for PIM1 (Fig. 2). Our crystal structure suggests that the larger trifluoro-methoxy moiety might favorably interact with the side chain of Arg¹²². K00135 binding was also confirmed by isothermal titration calorimetry, revealing a K_D of 25 nmol/L and a strongly enthalpy-driven interaction (not shown). It was interesting to see that, in general, PIM2 inhibition was 10- to 100-fold weaker than IC₅₀ values measured for PIM1 (Fig. 2). This finding is very surprising considering the high homology between the PIM1 and PIM2 active sites. In fact, there is only one single conservative change within the ATP pocket: comparison with the PIM2 crystal structure (pdb code 1IWI) showed PIM1 Val¹²⁶ is an alanine residue in PIM2. The close contact of Val¹²⁶ side chain and the carbonyl oxygen of the substitution of the aromatic R2 moiety of the inhibitor is 4 Å. This interaction is not likely to contribute to the binding of the inhibitor, but an amino acid substitution at this position might influence the flexibility of the kinase hinge region. PIM3 interaction with the investigated inhibitors seems to be more similar to PIM1 (Fig. 2). Two hits identified in the screen were particularly interesting; K00486 had an IC₅₀ of 34 nmol/L against PIM1 and only an IC₅₀ of 2.5 μmol/L against PIM2 with an identical selectivity profile, whereas K00152 inhibited PIM1 with an IC₅₀ of 39 nmol/L (PIM2 7 μmol/L) but also inhibited polo-like kinase 1 in our profiling for specificity (data not shown).

Ba/F3 cells stably expressing human PIMs. To address inhibition of PIM kinases by K00135 in a cellular system, we established a series of murine IL-3–dependent Ba/F3 cells that stably expressed human hPIM1 or hPIM2 by retroviral gene transfer. Stable expression of hPIM1 or hPIM2 allows the cells to survive in the absence of IL-3 (Fig. 3A,, top; ref. 30). To increase the sensitivity of detection of hPIM proteins, we also generated Ba/F3 cells stably expressing NH₂-terminal FLAG-tagged human PIM kinases. No differences in cell growth were observed when tagged or untagged hPIM kinases were expressed (not shown). Interestingly, although we were able to detect high levels of hPIM mRNAs (not shown), only relatively low levels of steady-state hPIM proteins were observed (Fig. 3A,, bottom). This is probably due to the short half-life of PIM proteins in vivo and their rapid proteosomal degradation. Treatment of the cells with the proteosome inhibitor MG132 led to a substantial increase in hPIM protein levels, in particular for hPIM1 (Fig. 3A,, bottom; ref. 31). To address the effect of K00135 in Ba/F3 cells stably overexpressing hPIMs, we exposed the cells to different concentrations of K00135, ranging from nanomolar to low micromolar levels. Survival of the cells was determined after 24 h by trypan blue dye exclusion. As shown in Fig. 3B, a dose-dependent decrease in cell survival was observed after 24 h in Ba/F3-hPIM1 grown in the absence of IL-3. Cells expressing hPIM2 were less sensitive to K00135, as expected from the higher IC₅₀ values for PIM2 determined for this compound (Fig. 2). Despite better potency in in vitro kinase assays, effects on cell survival using the two inhibitors K00152 and K00486 were similar (data not shown). Importantly, no significant decrease in viability was observed in wild-type Ba/F3 cells growing in IL-3 and in the presence of all three inhibitors tested. These experiments show that imidazo[1,2-b]pyridazines interact with and inhibit hPIM1 and, to a lesser extent, hPIM2 in a cellular system. In addition, these compounds do not seem to interfere with molecules of murine origin that are critical for IL-3–mediated cell growth.

Effect of K00135 on human leukemic cell lines. Cytotoxic effects of K00135, as well as the effect on clonogenic cell growth, were tested in a series of human leukemia cell lines (MV4;11, RS4;11, MOLM13, KOCL45, SEM, KOPN8, and K562). As shown in Fig. 4, some cytotoxic effects of K00135 were observed in all cell lines tested. Interestingly, cell lines that harbor the FLT3-ITD mutation, such as MV4;11 or MOLM13, were up to 20 times more sensitive than K562 cells that originate from BCR/ABL–positive blast crisis patient (Fig. 4C). We also assessed the effect of K00135 on clonogenic growth of leukemic cells. Among the cell lines included in our study, only MV4;11, MOLM13, and K562 were able to form colonies (>50 cells) when plated in methylcellulose (not shown). A significant reduction in clonogenic growth was observed in MV4;11 and MOLM13 even at an inhibitor concentration of <1 μmol/L, whereas colony formation of K562 was not affected even at a higher dose as expected from the relative resistance of this cell line to the treatment with this inhibitor class (Supplementary Fig. S1A).

K00135 induces apoptosis in human leukemic cell line MV4;11. K00135 induced cytotoxicity in both hPIM1/hPIM2–overexpressing Ba/F3 cells and all leukemic cell lines tested. To determine the mechanism of K00135-mediated cell death, we measured by flow cytometry the induction of apoptosis in MV4;11 cells treated with 1 μmol/L K00135 at different time points after inhibitor exposure. We used both dye PO-PRO-1, which is a sensitive indicator of apoptotic cells, and 7-amino-actinomycin D, an indicator of dead cells. The percentage of apoptotic MV4;11 cells increased in a time-dependent manner, with 2.8% of the cells undergoing apoptosis and 15.3% dead after 12 h and 8.5% apoptotic and 40.5% dead cells after 36 h of treatment (Supplementary Fig. S1B).

K00135 impairs phosphorylation of PIM downstream targets. To show that PIM kinases are indeed targeted by K00135 in leukemic cells, we monitored the in vitro kinase activity of immunoprecipitated PIM on its downstream targets by Western blotting with phospho-specific antibodies. The antiapoptotic protein BAD has been shown to be directly serine phosphorylated by both PIM1 and PIM2 (32, 33). To show inhibition of PIM activity in MV4;11 cells, PIM protein was immunoprecipitated and analyzed for its ability to phosphorylate BAD (on Ser¹¹² and Ser¹³⁶) in vitro in the presence (1 h) or absence of K00135 (2 μmol/L). As shown in Fig. 5A, phosphorylation of BAD by PIM1 and PIM2 was abrogated on addition of the compound. To further analyze the effects of K00135 on PIM function, phosphorylation of two known direct PIM targets (BAD and 4E-BP1) was followed by immunoblotting. As shown in Fig. 5B, short-term exposure of the cells to 1 μmol/L K00135 led to a significant decrease of phosphorylation of BAD and 4E-BP1. In contrast, no changes in phosphorylation of FLT3-ITD or Akt/protein kinase B were observed (data not shown). These results suggest that K00135 exerts its effects in MV4;11 by interfering primarily with PIM kinases downstream of FLT3-ITD or Akt/protein kinase B.

K00135 impairs in vitro survival of primary leukemic blasts. Next, we analyzed the effects of K00135 on the growth of leukemic blasts from five patients with newly diagnosed AML. When cultured in methylcellulose, the cells from AML patients grew in small clusters (Fig. 6B) and were therefore not amendable to be analyzed by classic quantitative colony forming assays. However, as shown in Fig. 6A, exposure of leukemic blasts in a short-term liquid culture to K00135 resulted in a significant decrease in survival for all patient samples. Although K00135 showed an effect on colony formation of all primary cells analyzed, no significant difference in survival between the patients harboring FLT3-ITD (patients 1 and 2) or those without (patients 3–5) was observed. Finally, we have addressed the cytotoxic effects of K00135 on normal human cord blood cells at antileukemic doses. As shown in Fig. 6C, K00135, at concentrations of 1 to 10 μmol/L, had only a minor effect on survival of normal umbilical blood mononuclear cells from four healthy donors.

Recent reports showing that PIM kinases are involved in malignant transformation of hematopoietic cells by oncogenic tyrosine kinases provided a strong rationale for development of PIM kinase inhibitors as potential therapeutics (11–17). All PIM inhibitor structures that have been published to date interact with the kinase hinge region in a classic ATP mimetic way. Screening of a kinase-targeted library and structure-activity relationship analysis allowed us to identify several imidazo[1,2]pyridazines with high affinity for PIM1. Related inhibitors of the pyrazolpyrimidine class were first described as specific inhibitors of the Src family of tyrosine kinases, with several pyrazolo[3,4-d]pyrimidines having low nanomolar affinity for Lck and anticancer activity (34). One of these compounds, PP1, was later crystallized in complex with the autoinhibited form of Hck, providing further guidance for drug development (35). Subsequently, this scaffold yielded specific inhibitors for glycogen synthase kinase 3 and members of the cyclin-dependent kinase (CDK) family (36). Inhibitors based on the variant pyrazolo[1,5-a]pyrimidine scaffold have also been described with specificity for CDK2 and the KDR kinase domain (37–39). However, all published cocrystal structures of this inhibitor class show binding to the kinase hinge region in the expected type I binding mode. Here, we identified a binding mode for our compound in which the inhibitor interacts with the opposite side of the ATP binding pocket, without making significant polar contacts with the kinase hinge region. In addition, the DFG motif did not move into an “out” conformation as observed in cocrystal structures of kinases with type II inhibitor complexes. However, the described area of the binding pocket is much more diverse than the hinge region, bearing the promise that more specific inhibitors may be developed on the basis of the determined crystal structure. Indeed, we observed cross-reactivity with only one other kinase (Cdc-like kinase 1) within a panel of 50 serine-threonine kinases currently present in our laboratory.⁶

Several compounds that inhibit PIM1 in vitro with nanomolar potency were selected for in vivo profiling using murine Ba/F3 cells overexpressing human PIM1 or PIM2 (30). Three compounds showed a dose-dependent inhibition of cell survival in hPIM-expressing Ba/F3 cells, but not in wild-type cells growing in the presence of IL-3 (Fig. 3). This suggests that PIM activity may not be essential for growth and survival of murine Ba/F3 cells signaling through IL-3, or simply that the compounds can interact and inhibit only human but not mouse PIM kinases. Additional experiments carried out with IL-3–independent Ba/F3 cells overexpressing mouse PIMs strongly suggest the latter possibility (data not shown).

The nanomolar in vitro potencies of the selected inhibitors using purified PIM protein translated into low micromolar potencies in the Ba/F3 cell model. The compounds were further tested in several human leukemia cell lines. Some AML cell lines such as MV4;11 or MOLM13 were very sensitive to the selected PIM1 inhibitors with IC₅₀ values well below 1 μmol/L in cell survival studies; however, other cell lines, such as K562 (chronic myelogenous leukemia blast crisis), were less affected (Fig. 4C). The molecular correlates that determine the sensitivity of PIM inhibitors are currently unknown, but preliminary results suggest that effects on cell survival may not fully correlate with PIM kinase expression levels (data not shown). In addition to cytotoxic effects, the tested inhibitors also significantly reduced the self-renewal capacity of leukemic cells as shown by inhibition of their clonogenic activity. Interestingly, there are several lines of evidence suggesting that PIM1 may functionally cooperate with Gfi1, a known regulator of self-renewal of hematopoietic stem cells (40). In addition, the homeobox transcription factor HOXa9, another regulator of self-renewal capacity and frequently deregulated in human acute leukemia, seems to regulate expression of PIM1 in hematopoietic stem cells, further supporting PIM1 as a therapeutic target for this disease (41).

Indeed, exposure of blasts from five AML patients to our PIM1 kinase inhibitors leads to a significant reduction in cell survival after 24 h (Fig. 6A). Interestingly, exposure of mononuclear cells from human cord blood from four donors to the compounds was associated with minimal cytotoxic effects (Fig. 6C). Mice lacking all known PIMs (PIM1, PIM2, and PIM3) are fertile and have a normal life span, suggesting that PIMs may not be essential for steady-state homeostasis, although deregulated on malignant transformation (42). Biochemical analysis strongly suggests that our compounds indeed block PIM1 function and its downstream targets such as BAD or 4E-BP1. However, a drawback of the identified compounds may be that significant reduction of growth and survival of human leukemia cell lines or primary samples could only be achieved at relatively high (micromolar) compound concentrations. Further experiments are necessary to determine whether micromolar serum level can be achieved in a living organism. Additional modification of the molecule core may further increase the affinity of the compounds. Strategies to modify PIM expression and/or stabilization may also be explored for the treatment of hematopoietic malignancies, and possible additive effects of combinations of PIM kinase inhibitors with heat shock protein-90 inhibitors such as 17-N-allylamino-17-demethoxygeldanamycin derivatives are currently investigated (31, 43). One should also keep in mind that PIM kinases have been intensively characterized as proteins of relatively weak inherent oncogenicity but potent in cooperation with other oncogenes such as c-myc (44–47). Therefore, it is likely that synergistic effects may be achieved in combination with either traditional chemotherapeutics and/or other small molecules that target signaling cascades linked to survival of the malignant cells.

K00135 shows selectivity for PIM1. Although previous studies have shown that an optimal PIM-targeted tumor therapy would include inhibition of PIM2 (12, 15, 17), thus far, there is no potent inhibitor for PIM2 available. Interestingly, all reported inhibitors show selectivity for PIM1, suggesting that PIM2 is intrinsically more difficult to target. This finding is surprising considering the high sequence conservation of the PIM1 and PIM2 active sites. It is therefore likely that dynamic variables that are currently not understood may also play a role in the recognition process. Strong research efforts are currently put into a search for PIM1 and PIM2 inhibitors.

This is the first report of a class of compounds that inhibit PIM kinases with clear anticancer activity as shown in leukemic cell lines as well as primary blasts from AML patients. Deregulated expression of PIM1 in other hematologic tumors (high grade lymphoma) and solid tumors, such as prostate carcinoma or oral squamous cell cancer, suggests that efficient inhibition of these kinases by small molecules may lead to potent targeted cancer therapy (48–50).

Grant support: Gertrude von Meissner Foundation, the Swiss National Science Foundation and ONCOSUISSE (J. Schwaller). The Structural Genomics Consortium is a registered charity (no. 1097737) funded by the Wellcome Trust, GlaxoSmithKline, Genome Canada, the Canadian Institutes of Health Research, the Ontario Innovation Trust, the Ontario Research and Development Challenge Fund, the Canadian Foundation for Innovation, VINNOVA, The Knut and Alice Wallenberg Foundation, The Swedish Foundation for Strategic Research, and Karolinska Institutet.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Sabine Ehret, Michael Sundström, and other members of our laboratories for their input in this work.