Effective Concentration of a Multikinase Inhibitor within Bone Marrow Correlates with In Vitro Cell Killing in Therapy-Resistant Chronic Myeloid Leukemia

Philadelphia chromosome–positive chronic myeloid leukemia (CML) is caused by constitutive activation of the oncogenic p210^BCR-ABL tyrosine kinase as result of a reciprocal translocation between chromosomes 9 and 22 (1). Therefore, CML treatment in clinic has been focused on blocking kinase activity of the fusion protein with tyrosine kinase inhibitors (TKI), such as imatinib, dasatinib, nilotinib, and recently, ponatinib (2–5). However, patients develop resistance to these targeted therapy drugs (6,7). Among the potential mechanisms for therapy resistance include development of new mutations in the fusion gene such as the T315I mutation (7), unfavorable pharmacokinetics and biodistribution of TKIs (8,9), and a protective bone marrow microenvironment (10,11). Strategies to address these critical issues will be effective in treating CML.

To identify a dose range for effective cell killing, CML cells have traditionally been treated with TKIs for 24 to 72 hours in cell culture (12–14). However, it is impossible to maintain the therapeutic magnitude and duration of the drugs in vivo due to their rapid drug metabolism and clearance, as the plasma half-lives for dasatinib and nilotinib are 2 hours and 1 hour in mice, respectively (15,16). Although the IC₅₀ value for nilotinib on inhibition of Ba/F3 cells overexpressing the BCR-ABL fusion protein in cell culture is less than 10 nmol/L (14), a daily dosage of 75 to 100 mg/kg is needed to treat animals in order to achieve a desirable therapeutic efficacy in murine CML models (17,18). Because the peak plasma drug concentration could already reach 14 μmol/L at a 25 mg/kg treatment dosage (16), which is over 1,000 folds of the IC₅₀ value in cell culture, the peak plasma concentration at these therapeutic dosages will be even higher. Thus, the in vitro cell growth inhibition study provided little guidance on the design of in vivo efficacy studies. Therefore, more reliable approaches are needed to predict therapeutic outcome based on the cell killing data.

In the current study, we applied TG101209 to treat CML cells that are resistant to BCR-ABL–targeted therapy in order to (i) establish an in vitro–in vivo correlation on treatment dosage, and (ii) develop an effective treatment for therapy-resistant CML. The TG compounds (TG101209 and TG101348) were originally developed as inhibitors of the JAK2/STAT5 signaling (19,20). STAT5 is one of the critical mediators for CML initiation, maintenance, and TKI resistance (21). Upon BCR-ABL inhibition, CML progenitor cells depend on high levels of cytokine-mediated JAK2/STAT5 activation for continued viability inside the bone marrow (22). So targeting the JAK2/STAT5 signaling with inhibitors, such as TG101209, is an ideal approach to prevent CML cell escape from BCR-ABL–targeted therapy. Interestingly, a recent study indicated that TG101209 could also inhibit the p210^BCR-ABL tyrosine kinase activity (21). Thus, TG101209 might serve as a multikinase inhibitor to block p210^BCR-ABL tyrosine kinase–dependent and independent pathways. Because CML cells carrying a T315I mutation in the BCR-ABL gene (p210^T315I) are resistant to imatinib and dasatinib (6), and clinical cases of resistance to ponatinib have also been identified (7), we applied cells with overexpressed p210^T315I to test drug efficacy in this study. We performed transient treatments of murine myeloid 32D cells overexpressing p210^T315I with TG101209 in cell culture, and identified the concentration range where CML cells were sensitive to TG101209 treatment. We then developed a bone-targeted formulation to achieve bone marrow TG101209 concentration at or above the effective concentration range in a sustained manner in vivo so as to effectively kill leukemia cells. Subsequently, we applied two murine leukemia models to demonstrate therapeutic efficacy.

32D cells were grown in RPMI 1640 medium supplemented with 10% FBS, 2 mmol/L glutamine, penicillin/streptomycin, and 10% WEHI-3 conditioned medium as previously described (23). 32D cells overexpressing wide-type BCR-ABL (32Dp210^WT) or T315I-mutant BCR-ABL (32Dp210^T315I) were generated by infecting cells with retroviruses carrying a wild-type or mutant p210^BCR-ABL gene, respectively. All cell lines were routinely tested for cellular morphology and microbial presence by microscopic observation. No authentication was performed by the authors. 32Dp210^WT and 32Dp210^T315I cells were cultured in the RPMI 1640 medium supplemented with 10% FBS, 2 mmol/L l-glutamine, and penicillin/streptomycin.

32Dp210^WT and 32Dp210^T315I cells were seeded in 6-well plates and incubated with increasing concentrations of TG101209 (Selleck Chemicals). Drug was washed out by centrifugation after 2 hours of treatment, and cells were resuspended in the growth medium and seeded in a 96-well plate at a seeding density of 4,000 cells/well. Cell proliferation was measured 48 hours later using a cell counting kit-8 (CCK8) viability assay (Dojindo Molecular Technologies Inc.). Half maximal effective concentration (EC₅₀) was calculated as the mean of three independent experiments. All values were normalized to the negative control.

32Dp210^WT and 32Dp210^T315I cells were seeded in 6-well plates and treated with increasing concentrations of TG101209 or vehicle for 2 hours. Cells were then washed with PBS and lysed in the RIPA buffer (Pierce, Thermo Fisher). Equal amount of proteins was applied for Western blot analysis. Primary antibodies specific for the following proteins and phosphoproteins were used for detection: JAK2, phospho-JAK2 (Y1007/1008), STAT5, phospho-STAT5 (Y694), histone H3, phospho-histone H3 (ser10), cleaved caspase-3, and cleaved PARP from Cell Signaling Technologies; and c-ABL, phospho-c-ABL (Y412), and Aurora B from Abcam.

32Dp210^T315I cells were plated in triplicates in 6-well plates supplied with 2 mL growth medium. Cells were treated with 5, 10, or 25 μmol/L of TG101209 for 2 hours before switching back to growth medium. Cell growth was maintained for 24 to 48 hours thereafter. Cells were washed with PBS and fixed in 70% ethanol overnight at −20°C. For flow cytometric analysis, cells were washed twice with PBS, resuspended in the PI/RNase staining buffer (BD Pharmingen), and incubated for 30 minutes at room temperature before they were applied for analysis with a BD LSRII flow cytometer (BD Sciences).

32Dp210^T315I cells were treated with TG101209 at 10 μmol/L for 2 hours followed by incubation in growth medium for 48 hours. They were fixed with 4% paraformaldehyde in phosphate buffer for 20 minutes, transferred to a cover slip by centrifugation, and permeabilized with 0.2% Triton X-100 in PBS for 5 minutes. Cytoskeleton was stained with Alexa Fluor 488 Phalloidin (Invitrogen) for 20 minutes, and nuclei were stained with DAPI for 5 minutes. Images of the stained cells were captured with a Nikon A1 spectral confocal microscope.

Cell lysates from posttreatment samples were incubated with a primary anti-Aurora B antibody (Abcam) overnight at 4°C. Twenty microliters of protein A/G agarose magnetic beads (Thermal Scientific) were added into 500 μL cell lysate and incubated for 2 hours at 4°C. The pellet was washed three times with a pH 7.5 kinase assay buffer (20 mmol/L Tris, 10 mmol/L MgCl₂, 5 mmol/L glycerophosphate, 0.1 mg/mL BSA, and 2 mmol/L dithiothreitol), and resuspended in 30 μL kinase buffer supplemented with 200 μmol/L ATP and 2 μg histone H3.1 protein (Biolabs Inc.). The mixture was incubated for 30 minutes at 30°C to allow protein phosphorylation and terminated with a SDS sample buffer before applying for Western blot analysis for histone H3.

Bone-targeted TG101209 micelles were prepared by encapsulating TG101209 in alendronate-polyethyleneglycol-polylactic acid (PEG-PLA) using a co-solvent evaporation method (24). Briefly, 8 mg TG101209 was mixed with 50 mg methoxy PEG₂₀₀₀-PLA₂₀₀₀ and alendronate-conjugated PEG-PLA (molar ratio: 3:2), and dissolved in 0.5 mL acetone. This solution was added drop-wise into 10 mL distilled water under stirring. The mixture was then vigorously stirred at 20°C for 2 hours, and residual acetone was removed under vacuum. The micelle solution was filtered through a 0.45 mm syringe filter and subsequently lyophilized.

Pharmacokinetics of TG101209 in solution or micelles was evaluated in C3H mice. For free TG101209, the compound was dissolved in acetate solution (pH 4.0) containing 20% Tween 80 and administrated by oral gavage (100 mg/kg). Blood samples were collected at the 0.5, 1, 2, 3, 4, and 6 hour time points. For TG101209 micelles, the CML mice were treated with untargeted (40 mg/kg) or targeted (20 or 40 mg/kg) TG101209 by intravenous administration, and blood samples were collected at 0.25, 0.5, 1, 2, 3, 4, and 6 hour time points. TG101209 was extracted from plasma by mixing methanol with plasma (9:1, v/v) followed by centrifugation at 21,000 × g for 10 minutes.

Biodistribution analysis was carried out after mice were treated i.v. or p.o. with TG101209. Mice dosed with the drug intravenously were euthanized 0.5, 2, or 6 hours later, and those with oral gavage were sacrificed 2, 4, or 6 hours later. Major organs, including lung, liver, heart, kidney, spleen, femur, and tibia, were collected for drug content analysis. The samples were homogenized and centrifuged at 21,000 × g for 10 minutes. Methanol was added to the supernatant (9:1, v/v) to remove proteins by centrifugation at 21,000 × g for 10 minutes. Femur and tibia were decalcified in 0.5 mol/L ethylenediaminetetraacetic acid at pH 7.2 for 48 hours before homogenization. TG101209 concentration in plasma and tissue samples was measured with high-performance liquid chromatography.

One million GFP-positive 32Dp210^T315I cells were inoculated into each C3H mouse by tail vein injection. Three days after inoculation, mice were treated intravenously every other day with 40 mg/kg untargeted or targeted TG101209 for 6 days. In the control group, imatinib was administrated by daily oral gavage (150 mg/kg) for 6 days. Spleen and bone marrow samples were collected 24 hours after the last dosing for Western blot analysis, flow cytometry, and immunohistochemical staining. To measure phosphorylated proteins in cells isolated from bone marrow, cells were fixed and permeabilized with the BD Perm/Wash buffer by following the manufacture's protocol. To analyze histone H3 phosphorylation, cells were stained with a phycoerythrin (PE)-conjugated anti–p-Histone H3 (ser10) antibody. To detect pSTAT5, cells were incubated with an anti-pSTAT5 antibody (Cell Signaling Technology) followed by incubating with a secondary PE-conjugated goat anti-rabbit immunoglobulin.

Two murine leukemia models were applied in this study. The C3H syngeneic leukemia model was developed by intravenous injection of 1 × 10⁶ 32Dp210^T315I cells. The bone marrow transplantation models were developed by inoculating p210^WT or p210^T315I retrovirus–infected bone marrow cells into sublethally irradiated NOD/SCID mice. Briefly, C3H donor mice were treated daily with 5-fluorouracil (200 mg/kg) for 4 days. Mice were euthanized by CO₂, and bone marrow cells were harvested from the tibia and femur by flushing with PBS. Bone marrow cells were resuspended at 5 million/mL in DMEM containing 10% FBS, 5% WEHI medium, 10 ng/mL recombinant murine IL3, 20 ng/mL recombinant murine IL6, 100 ng/mL recombinant murine stem cell factor, 100 U/mL penicillin/streptomycin, and 2 mmol/L l-glutamine. The cells were infected with two rounds of Mig1-BCR-ABL-GFP or Mig1-BCR-ABL-T315I-GFP retroviruses in the presence of 8 μg/mL of polybrene, and infection efficiency was determined by flow cytometry 24 hours after the second round of retrovirus infection. The infected bone marrow cells were then injected into the sublethally irradiated (600 cGy) recipient NOD/SCID mice at 1 × 10⁶ cells/mouse.

To evaluate therapeutic efficacy in the C3H leukemia model, mice were treated beginning the second day after cell inoculation with (i) imatinib (150 mg/kg, p.o. daily), (ii) vehicle (empty targeted micelles, i.v., every other day), (iii) untargeted TG101209 micelles (40 mg/kg, i.v., every other day), (iv) targeted TG101209 micelles (40 mg/kg, i.v., every other day), or (v) free TG101209 in DMSO (40 mg/kg, i.p., every other day). To examine therapeutic efficacy in the bone marrow transplantation models, leukemia development was confirmed based on peripheral blood cell counts and GFP⁺ cells in peripheral blood 1 week after transplantation. Mice were then treated with (i) imatinib (150 mg/kg, p.o., daily), (ii) untargeted TG101209 micelles (40 mg/kg, i.v., every other day), (iii) targeted TG101209 micelles (20 mg/kg, i.v., every other day), or (iv) targeted TG101209 micelles (40 mg/kg, i.v., every other day). Percentage of circulating GFP⁺ leukemia cells in peripheral blood was examined by flow cytometry on days 15 and 25. Mice were euthanized on signs of morbidity, and Kaplan–Meier plots were generated based on animal survival.

C3H mice were treated with targeted TG101209 micelles (20 mg/kg or 40 mg/kg, i.v., every other day) for 2 weeks. The untreated mice were as control. The animals were observed for signs of toxicity throughout the experiment. Water and food consumption were recorded weekly. Body weight of the mice was measured every 3 days. On day 7 and day 14, blood samples were collected for hematology analysis. At the end of the 2-week treatment, the major organs were collected, and histologic analyses were performed.

Statistical analyses were performed with the GraphPad Prism 5 software (GraphPad Software, Inc.). For all in vivo analysis, sample sizes were chosen to ensure adequate power to detect a prespecified effect size. P values of less than 0.05 and 0.01 were considered statistically significant and very significant, respectively. Data are presented as mean ± SD.

Resistance to the second-generation p210^BCR-ABL TKI dasatinib was confirmed in 32Dp210^T315I cells by Western blot analysis. Although low nanomolar concentrations of dasatinib were potent enough to inhibit phosphorylation of BCR-ABL, JAK2, and STAT5 in 32Dp210^WT cells after 2 hours treatment, this drug was not effective in blocking phosphorylation of these proteins in the mutant line even at micromolar concentrations (Supplementary Fig. S1). In contrast, TG101209 treatment suppressed phosphorylation of these proteins in both 32Dp210^WT and 32Dp210^T315I cells (Fig. 1A). Interestingly, phosphorylation of all three proteins was inhibited within the same drug concentration range in the same cell line, although the IC₅₀ value was 7-fold as high in 32Dp210^T315I (IC₅₀ 17.5 μmol/L) as in 32Dp210^WT (IC₅₀ 2.5 μmol/L) cells (Fig. 1B). This observation indicates that activation of BCR-ABL, JAK2, and STAT5 might be mediated through a common signal transduction pathway. It has been previously reported that STAT5 phosphorylation is mediated by the upstream p210^BCR-ABL in CML cells, an event independent of JAK2 activity (21). Our result indicates that JAK2 phosphorylation in 32Dp210^WT and 32Dp210^T315I cells is most likely also controlled by the BCR-ABL kinase activity. It is important to note that low micromolar plasma drug concentrations are commonly detected in animal studies treated with the more potent analogue TG101348 and in the clinic (25,26).

We treated 32Dp210^WT and 32Dp210^T315I cells with TG101209 at their respective IC₅₀ values for 2 hours before switching cells to a drug-free growth medium, and tracked JAK2 and STAT5 phosphorylation status in the next 24 hours. Phosphorylation of both proteins was inhibited in the first 3 hours after drug treatment, but gradually recovered thereafter (Fig. 1C). We also measured cell viability 48 hours later. Interestingly, the EC₅₀ value matched with the IC₅₀ value for inhibition of phosphorylation in 32Dp210^WT cells (EC₅₀ 3.15 μmol/L), but was 37% less than the IC₅₀ value in 32Dp210^T315I cells (EC₅₀ 10.9 μmol/L; Fig. 1D). The discrepancy between IC₅₀ and EC₅₀ in 32Dp210^T315I cells indicates that TG101209 might have inhibited BCR-ABL–independent signaling and triggered cell death before reaching an effective concentration for inhibition of BCR-ABL, JAK2, and STAT5 activities.

We treated 32Dp210^T315I cells with increasing concentrations of TG101209 in order to identify factors critical for 32Dp210^T315I cell growth and viability. Surprisingly, a 2-hour incubation with 10 μmol/L TG101209 was already very effective in cell-cycle blockage as indicated by increased percentage of cells in G₂–M and accumulation of polyploids (Fig. 2A). Cytostaining of cells treated with 10 μmol/L TG101209 for 2 hours revealed enlarged overall cell size and enrichment of multinucleated cells (Fig. 2B). Drug treatment also triggered cell apoptosis as indicated by increased levels of the G₀ population (Fig. 2A) and cleaved PARP and caspase 3 (Fig. 2C). As expected, phosphorylation of p210^T315I, JAK2, or STAT5 was not blocked at 10 μmol/L TG101209 (Fig. 1A).

It has been previously reported that TG101209 can inhibit the activities of multiple protein kinases, including an Aurora kinase (19). Aurora kinase inhibitors achieved clinical response in CML patients bearing T315I-mutated BCR-ABL (27). Interestingly, blockage of Aurora kinases is an effective approach for cell-cycle arrest (28). Selective Aurora B inhibition can block histone H3 phosphorylation at serine 10, and lead to G₂–M arrest, polyploidy, and apoptosis (28,29), a result similar to treatment with 10 μmol/L TG101209. We applied a selective aurora B inhibitor, barasertib, to treat CML cells to confirm the cell killing effect. When 32Dp210^WT and 32Dp210^T315I cells were treated with barasertib for 2 hours, p-histone H3 (ser10) was inhibited at 1 μmol/L in both cell lines (Supplementary Fig. S2A). 32Dp210^T315I cells were blocked in G₂–M phase, and the ratio of polyploidy and apoptotic cells increased after 2-hour barasertib treatment followed by 24 and 48 hours incubation (Supplementary Fig. S2B and S2C). We next examined whether Aurora B kinase was involved in TG101209-mediated cell killing. 32Dp210^WT and 32Dp210^T315I cells were treated with increasing concentrations of TG101209 for 2 hours, and Aurora B kinase activity was analyzed based on phosphorylation of the kinase substrate histone H3 at the Ser10 residue (30). Potent inhibition of histone H3 phosphorylation was achieved at 5 μmol/L concentration in both cell lines (Fig. 2D), indicating that Aurora B kinase activity was effectively inhibited by TG101209. Because phosphorylation of p210^T315I, JAK2, or STAT5 was not affected by TG101209 below 10 μmol/L (Fig. 1A), this result also demonstrated that Aurora B kinase activity was independent of the BCR-ABL signaling in 32Dp210^T315I cells. To confirm inhibition of kinase activity directly, we treated 32Dp210^T315I cells with increasing concentrations of TG101209 for 2 hours, immunoprecipitated Aurora B kinase from cell lysates, and then measured the protein level and kinase activity of Aurora B. Although the expression level of Aurora B kinase was unchanged after drug treatment, its ability to phosphorylate histone H3 was dramatically reduced (Fig. 2E). Thus, Aurora B kinase mediates cell-cycle blockage in 32Dp210^T315I cells treated with TG101209.

In order to reach a high tissue drug concentration so as to effectively block activities of key kinases in CML cells inside the bone marrow, we developed an intravenously injectable bone-targeted formulation of TG101209 for in vivo studies (Fig. 3A). Briefly, the hydrophobic TG101209 was encapsulated in PEG-PLA micelles, and the surface of the micelles was conjugated with alendronate molecules as the bone-targeting moieties. Binding between bisphosphonate drugs including alendronate and the mineral hydroxyapatite in the bone has been well studied (31), and application of bisphosphonates as bone-targeting moieties has been previously documented (32). The untargeted micelles had an average size of 29.9 nanometers and a surface charge of −4.37 mV (Supplementary Table S1). Alendronate conjugation did not dramatically alter the size of micelles; however, the surface charge of micelles decreased to −38.2 mV owing to the negatively charged phosphate groups in the targeting moiety.

Mice bearing 32Dp210^T315I CML were dosed i.v. with TG101209 in untargeted or targeted formulation (40 mg/kg untargeted TG101209 and 20 to 40 mg/kg targeted TG101209), and pharmacokinetics and tissue biodistribution were compared between the treatment groups. These parameters were also compared with mice treated p.o. with 100 mg/kg of the free drug, as oral administration of TG101209 at this dosage has been explored in preclinical studies (11,19,33,34). Overall, mice treated intravenously with targeted or untargeted micelles showed similar pattern on changes in plasma drug concentration (Fig. 3B), and had similar pharmacokinetic parameters (Supplementary Table S2). Mice in the 20 mg/kg targeted TG101209 treatment group had a comparable AUC with those in the 100 mg/kg oral gavage treatment group, in line with a previous report that the orally administrated TG101209 has a bioavailability of 23% in mice (19).

As expected, majority of the drug accumulated in the liver, lung, and spleen in all treatment groups regardless of the route of administration, and the highest drug concentrations in the liver and spleen were detected in the oral gavage group (Supplementary Fig. S3). However, bone marrow biodistribution analysis revealed striking difference among the different treatment groups (Fig. 3C). As with the oral gavage group (100 mg/kg), intravenous administration with untargeted micelles at the 40 mg/kg dosage only briefly raised bone marrow TG101209 concentration above 10 μmol/L, but not the 17.5 μmol/L threshold for inhibition of p210^T315I/JAK2/STAT5. It is thus unlikely that phosphorylation status of these proteins is altered in the bone marrow CML cells in the untargeted formulation or oral administration treatment groups. In contrast, bone marrow drug concentration could already reach 17.5 μmol/L in mice treated i.v. with the low-dosage (20 mg/kg) targeted micelles, and a high bone marrow drug concentration was maintained for over 2 hours in mice dosed with the high-dosage (40 mg/kg) targeted micelles. Based on the high bone marrow drug concentration level, it is anticipated that treatment with 40 mg/kg targeted micelles can potently and sustainably block signaling from both Aurora B kinase and p210^T315I/JAK2/STAT5 so as to effectively kill CML cells inside the bone marrow.

A murine model of CML was generated by inoculating 32Dp210^T315I cells i.v. into C3H mice. Mice were treated daily with imatinib (150 mg/kg, p.o.) or every other day with untargeted or bone-targeted TG101209 micelles (40 mg/kg, i.v.) for 6 days, and the spleen and bone marrow samples were collected 24 hours after the last treatment for expression and histologic analyses. Mice in the imatinib treatment group served as controls, and twice a day treatment with imatinib at the indicated dosage is a common practice for preclinical studies on CML (35,36). Western blot analysis revealed no alteration on phosphorylation status of p210^T315I/JAK2/STAT5 in the spleen from imatinib treatment (Fig. 4A). However, phosphorylation of these proteins was effectively blocked in the spleen by untargeted or bone-targeted TG101209 (Fig. 4A and C), correlating with the high tissue drug concentration in this organ (Supplementary Fig. S3). In comparison, phosphorylation of p210^T315I/JAK2/STAT5 and histone H3 proteins was effectively blocked in bone marrow by bone-targeted, but not untargeted TG101209 (Fig. 4B). In addition, flow cytometry was performed to examine inhibition of STAT5 and histone H3 phosphorylation in leukemia cells isolated from the bone marrow in different treatment groups. Although treatment with imatinib or untargeted TG101209 had limited effect on the phosphorylation status of STAT5 and histone H3, complete inhibition of phosphorylation in these two proteins was observed in leukemia cells from the 40 mg/kg targeted TG101209 treatment group (Fig. 4D), correlating with drug concentration analysis in bone marrow (Fig. 3C).

Inhibition of STAT5 phosphorylation might provide additional benefit for CML treatment by suppressing stimuli from the bone marrow microenvironment. In a recent study, we have shown that, in response to TKI treatment, bone marrow mesenchymal stromal cells secrete cytokines which activate signal transduction pathways leading to therapy resistance by the leukemia cells (37). We incubated 32Dp210^T315I cells in growth medium with or without supplemented IL3, and treated them with ponatinib or TG101209. Ponatinib treatment inhibited phosphorylation of p210^BCR-ABL and the downstream JAK2 and STAT5 in the absence of IL3, but was not effective in blocking IL3-mediated JAK2/STAT5 phosphorylation; in contrast, TG101209 treatment abolished both p210^BCR-ABL-mediated and IL3-mediated STAT5 phosphorylation (Supplementary Fig. S4A). As a result, TG101209 treatment effectively killed leukemia cells regardless of the presence of IL3, whereas cells showed resistance to ponatinib treatment in the presence of IL3 (Supplementary Fig. S4B and S4C).

Survival benefit from targeted delivery of TG101209 was evaluated in two murine models of p210^T315I CML. In the C3H CML model, the syngeneic 32Dp210^T315I CML cells grew aggressively inside the body, and all untreated mice succumbed to leukemia growth 12 to 20 days after cell inoculation, with a median survival of 16 days (Fig. 5A). No survival benefit was observed from daily treatment with 150 mg/kg imatinib or every other day treatment with 40 mg/kg free TG101209 or untargeted TG101209 micelles. In contrast, significant therapeutic benefit was achieved in mice treated with 40 mg/kg targeted TG101209 (Fig. 5A).

The bone marrow transplantation CML model was more sensitive to TG101209 treatment. This CML animal model was generated by inoculating p210^T315I retrovirus–infected C3H bone marrow cells into sublethally irradiated NOD/SCID mice. Peripheral blood samples were applied to monitor leukemia progression during treatment, and significantly reduced number of CML cells was observed from the targeted TG101209 micelle treatment groups (Supplementary Fig. S5). All mice in the nontreatment group died from leukemia growth 25 to 32 days after bone marrow transplantation (Fig. 5B). Imatinib treatment did not offer any survival benefit on the CML-bearing NOD/SCID mice, with a median survival time of 27.5 days (Fig. 5B). Treatment with untargeted or the low-dosage targeted micelles provided a moderate but statistically significant survival benefit over the nontreatment group. Their median survival times were 33 and 34.5 days, respectively. Extension of survival was most likely from cell-cycle blockage as a result of inhibition of Aurora B kinase activity, because bone marrow drug levels in these two groups would not be high enough to exert sustained inhibition of p210^T315I/JAK2/STAT5 (Fig. 3C). The most significant survival benefit was obtained in mice treated with the high-dose targeted TG101209 micelles, with a median survival time of 42 days (Fig. 5B), demonstrating the power of blocking both p210^T315I/JAK2/STAT5 activities and cell cycle in suppressing CML. We also applied the same treatments on p210^WT CML in the bone marrow transplant model. As expected, imatinib (150 mg/kg, daily) treatment significantly extended animal survival, and the best therapeutic outcome was from treatment with the targeted TG101209 (40 mg/kg, every other day; Supplementary Fig. S6).

Because daily treatments with high oral doses (100 and 200 mg/kg) of TG101209 could cause severe toxicity such as anemia (11,33), we carried out in vitro and in vivo studies to evaluate potential toxicity from the bone-targeted formulations. In the in vitro study, cell viability and colony-formation ability were monitored in bone marrow cells from C3H mice upon transient treatment with increasing concentrations of TG101209. No significant impact from therapeutic dosages of TG101209 was observed (Supplementary Fig. S7). In the in vivo study, C3H mice were dosed every other day with bone-targeted TG101209, and no changes on blood chemistry or histology were detected (Fig. 6). Thus, bone marrow enrichment of the drug resulted in dramatically decreased overall treatment dosage and significantly reduced drug exposure to normal tissues, which caused no detectable toxicity in our study.

In order to determine whether enhanced bone marrow drug accumulation could be translated into improved therapeutic efficacy in clinic, we isolated leukemia cells from peripheral blood in CML patients at the blast crisis stage and treated them with increasing concentrations of TG101209. All patients had gone through multiple rounds of treatment and had become resistant to the first- or second-generation CML TKIs in clinic. Two-hour TG101209 treatment was effective in blocking phosphorylation of p210^BCR-ABL, JAK2, STAT5, and histone H3 at 2.5 to 25 μmol/L in all samples (Supplementary Fig. S8), indicating that the patients would benefit from the bone marrow–targeted drug formulation in clinic.

Multiple TKIs have been developed for the treatment of Philadelphia chromosome–positive leukemia so far; however, most patients eventually develop therapy resistance and hence disease relapse (38). The underlying mechanism for therapy resistance includes frequent mutations in the BCR-ABL gene (7,39), activation of BCR-ABL–independent signal transductions pathways (40–42), and protection of leukemia cells by the bone marrow microenvironment (11,37,43,44). In addition, poor pharmacokinetic profile and biodistribution of the therapeutic agent contribute to development of therapy resistance (8). It is thus important to take into consideration of the multiple factors in developing an effective therapeutic strategy for CML. In this study, we have demonstrated that the effective dosage to overcome CML therapy resistance can be identified in an in vitro setting, and the information can serve as a guidance to develop the proper drug formulation for in vivo efficacy studies. Our results have revealed there exist two groups of proteins that are vital for CML cell survival, one is regulated by p210^BCR-ABL tyrosine kinase activity (e.g., STAT5) and the other is independent of the p210^BCR-ABL signaling (e.g., Aurora B kinase). Because CML cells bearing a wild-type BCR-ABL gene are sensitive to low-dose TG101209 treatments, there is no need to overload the patients with the drug during treatment. Consequently, drug effect on the second group of proteins is minimal. In patients with a mutant BCR-ABL gene, however, the p210^T315I tyrosine kinase is more resistant to treatment, and activities of the second group proteins are modulated by the small-molecule inhibitor at concentrations below the threshold for inhibition of p210^T315I/JAK2/STAT5. Thus, TG101209 treatment results in two stages of responses in the mutant cells: cell-cycle arrest and apoptosis due to blockage of p210^T315I-independent proteins at a relatively low drug concentration, and enhanced cell death caused by inhibition of p210^T315I/JAK2/STAT5 once the threshold concentration is reached. Our results have shown that potent inhibition of both groups of proteins with a high bone marrow concentration of TG101209 is essential for effective treatment of therapy-resistant CML.

We have previously demonstrated JAK2 activation by the p210^BCR-ABL tyrosine kinase activity in CML cells (45), and have proposed JAK2 as a potential target for drug intervention in CML (46). We have also shown that one of the JAK2 TKIs, ON044580, inhibits both JAK2 and p210^BCR-ABL kinase activities in in vitro assays using their respective peptide substrates (47). Multiple JAK2 TKIs have since then been applied in combination with BCR-ABL TKI for CML treatment (22,48). Interestingly, the cell killing effect from TG101209 and TG101348 seems to be predominantly from direct inhibition of the p210^BCR-ABL kinase rather than JAK2 (21). In the current study, we have observed synchronized blockage of phosphorylation of BCR-ABL, JAK2, and STAT5 kinases in both wild-type and T315I-mutant cells by TG101209. The increased drug concentration needed for inhibition of p210^T315I/JAK2/STAT5 in the mutant cells supports the notion of direct inhibition of the p210^T315I tyrosine kinase by the TKIs. It is unknown whether other JAK2 TKIs in various stages of drug development have a similar spectrum of target inhibition and consequently mechanism of action.

Small-molecule JAK TKIs have commonly been dosed through oral administration for the treatment of myelofibrosis and hematopoietic neoplasia in preclinical (19,33,49) and clinical studies (50,51). Our analysis indicates that only a very small portion of the drug can reach the bone marrow, and majority of the drug is accumulated in the liver, lung, and spleen in mice treated with 100 mg/kg TG101209 by oral gavage. Thus, the body needs to be overloaded with a high dosage of the drug in order to achieve a therapeutic level inside the target organ (i.e., the bone marrow). However, sustained inhibition of the JAK/STAT pathway in the major organs may cause severe side effects (11). In comparison, the gap between in vitro activity and in vivo efficacy of TG101209 was bridged by identifying the effective concentration in vitro and achieving this concentration in bone marrow in vivo. The bone-targeted formulation is very efficient in delivering drug to the bone marrow. With this new formulation, targeted therapy is achieved through targeted delivery of the TKI molecules. Bone enrichment of the drug with significantly decreased drug exposure to normal tissues resulted in remarkable therapeutic efficacy and minimal toxicity in our study.

In summary, we have demonstrated that incubation of the therapy-resistant CML cells with TG101209 in the 10 to 25 μmol/L range blocks BCR-ABL–dependent and –independent signaling and triggers cell death in vitro. The information has guided us to develop a bone marrow–targeted approach to enrich drug in the tissue and effectively treat CML. This in vitro–in vivo correlation strategy for development of effective cancer treatment might be applicable to diseases beyond CML in clinic.

No potential conflicts of interest were disclosed.

Conception and design: X. Xia, R.B. Arlinghaus, H. Shen

Development of methodology: C. Mu, X. Wu, H. Ma, G. Zhang, J. Shen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Mu, X. Wu, J. Mai

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Mu, X. Wu, X. Xia, T. Sun, H. Shen

Writing, review, and/or revision of the manuscript: C. Mu, X. Wu, T. Sun, H. Shen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Tao, X. Sun

Study supervision: H. Shen

The authors thank Drs. Yu Mi and Guohui Wang for their assistance.

This work was supported by grants from Golfers Against Cancer (to H. Shen) and NCI 1R01CA193880-01A1 (to H. Shen).

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.