Discovery of a Highly Selective JAK2 Inhibitor, BMS-911543, for the Treatment of Myeloproliferative Neoplasms

Abstract

JAK2 kinase inhibitors are a promising new class of agents for the treatment of myeloproliferative neoplasms and have potential for the treatment of other diseases possessing a deregulated JAK2-STAT pathway. X-ray structure and ADME guided refinement of C-4 heterocycles to address metabolic liability present in dialkylthiazole 1 led to the discovery of a clinical candidate, BMS-911543 (11), with excellent kinome selectivity, in vivo PD activity, and safety profile.

Myeloproliferative neoplasms (MPNs) are a subset of myeloid malignancies that are characterized by the expansion of a hematopoietic progenitor stem cell. MPNs encompass polycythemia vera (PV), essential thrombocytopenia (ET), and primary myelofibrosis (PMF).¹ In the majority of cases, this cluster of diseases has been shown to be associated with the somatic mutation JAK2-V617F that constitutively activates the Janus kinase 2 (JAK2) enzyme, a member of the JAK family of nonreceptor tyrosine kinases.² In MPNs the acquisition of the JAK2-V617F and other JAK2-STAT pathway mutations result in cytokine-independent activation of the pathway and the uncontrolled growth of hematopoietic cells with erythrocytes, platelets, and granulocyte/monocytes being the predominant lineages expanded in ET, PV, and PMF, respectively.³ The uncontrolled growth of these cell lineages in MPNs results in severe patient complications including splenomegaly, hemorrhage, thrombosis, bone marrow fibrosis, and transformation to acute myeloid leukemia. The overall survival rate for patients afflicted with advanced myelofibrosis is estimated to be 3–5 years.⁴

The causal role of JAK2 in MPNs is supported by significant genetic and pharmacological data. Transgenic reconstitution of JAK2 mutations into rodent bone marrow stem cells results in a phenotype mirroring the main features of human MPNs including splenomegaly, bone marrow fibrosis, and elevated levels of certain hematopoietic lineages (e.g., erythrocytes, leukocytes).⁵ Administration of small molecule JAK2 kinase inhibitors reverses the pathophysiological features of the transgenic phenotype.⁶ Moreover, clinical testing of two small molecule JAK2 inhibitors (e.g., ruxolitinib, fedratinib, pacritinib) showed effects on splenomegaly and normalization of blood counts (Figure 1).⁷⁻⁹ Based upon the late stage clinical efficacy and tolerability profile, ruxolitinib was approved to treat myelofibrosis (MF). Several other JAK2 inhibitors with varying degrees of JAK family as well as overall kinome selectivity profiles are in mid to late-stage clinical trials for MF.¹⁰ However, it is important to note that most of these compounds also inhibit other JAK family members that could be associated with immunosuppression, an undesired side effect for this indication. Additionally, other off-target kinome activities (e.g., FLT3) could further compromise the anticipated high safety window needed for long-term treatment in MPNs.

Figure 1.

JAK2 inhibitors.

We have recently disclosed the identification of a potent and highly selective JAK2 inhibitor 1.¹¹ X-ray crystallographic studies of 1 bound to the JAK2 kinase domain indicated that it engages the Tyr931 residue through hydrogen bonding with the thiazole nitrogen atom. In addition, unfavorable interactions of the 4,5-dimethylthiazole fragment with nonconserved residues in the extended hinge region of other JAK family members provided high selectivity. Further ADME profiling indicated that 1 was rapidly metabolized across species and was susceptible to generation of reactive metabolites (vide infra), which prevented its further progression. Herein we report ADMET and structure-guided optimization of heterocycles at the C-4 position of the imidazopyrrolopyridine core leading to the discovery of a highly selective JAK2 inhibitor, BMS-911543 (11), as a clinical candidate for the treatment of MPNs. Earlier we have reported biological characterization of BMS-911543.¹²

In vitro biotransformation studies with compound 1 using human liver microsomes revealed the formation of a thiourea metabolite in significant amounts (15%). This metabolic process presumably involves cytochrome P450-mediated oxidation of the thiazole ring to an epoxide and subsequent opening to form a diol intermediate. Decomposition of the diol intermediate leads directly to a thiourea and diketo species (Figure 2). Thioureas have the potential to form an active metabolite in vivo, which could disrupt thyroid function and could also have adverse effects in lung, liver, and bone marrow.¹³

Figure 2.

Metabolism of 1.

Literature reports indicated that the introduction of either electron withdrawing groups on the thiazole or increasing sterics may reduce initial complexation with CYP enzymes, which could suppress oxidative ring opening.^13,14 Accordingly, analogues 4–8 were prepared using the approach outlined in Scheme 1, starting from aminopyridine 2.^15,16

Scheme 1.

Reagents and conditions: (a) benzoylisothiocyanate, acetone; then 1 N NaOH, ethanol, 60 °C, 72%; (b) 3-bromopentan-2-one, 60 °C, 44%; (c) methyl 2-bromo-3-oxobutanoate, ethanol, 65 °C, 84%; (d) 1 N NaOH, methanol, 65 °C, 93%; (e) methylamine, HATU, 2,6-lutidine, DMF, 75%; (f) methyl 3-bromo-2-oxobutanoate, ethanol, 65 °C, 70%; (g) 1-bromo-1-(methylsulfonyl)propan-2-one ethanol, 65 °C, 62%; (h), 1,1-dioxo-1-thiomorpholine, HATU, 2,6-lutidine, DMF, 62%.

As anticipated, adding substitution on the 4-methyl group (4) maintained the JAK family selectivity profile of the parent compound (Table 1). However, this change still resulted in formation of thiourea in significant amounts in in vitro biotransformation studies. Introduction of electron withdrawing groups led to identification of potent JAK2 inhibitors (compounds 5–7) , which were significantly less susceptible to generation of the thiourea metabolite. However, such substitutions caused considerable erosion of selectivity versus other JAK family kinases. With respect to the selectivity of the amide-containing thiazoles, differences in the JAK family members around JAK2 Gln 853 may provide a rationale for the loss in selectivity observed. The smaller serine present in JAK3 would likely be able to accommodate the larger polar amide-containing analogues. However, the arginine present in JAK1 could be positioned such that a favorable hydrogen bond to the amide at the 4- or 5-position on the thiazole may also contribute to loss of selectivity (Figure 3 for JAK1 model with compound 6). Further steric bulk on the amide (8) failed to improve JAK family selectivity (compared to 1) and reduced cellular potency.

Table 1. Biotransformation and Thiazole Substitution SAR^a,b,c.

^aAssay protocols are provided in the Supporting Information.

^bAssay results are the average of at least two replicates.

^cPercent thiourea determined in human liver microsomes.

^dOnly trace levels detected by mass spectrometry.

^eNo data was generated.

Figure 3.

Model of 6 bound to the kinase catalytic domain of JAK1. The carbons of 6 are colored in pink, and the carbons for JAK1 are colored in green except for the residues near the C-4 group, which differ in the JAK family (carbons are colored cyan). Oxygen atoms are colored red, nitrogens blue, and sulfurs yellow. Hydrogen bonds are indicated with dashed lines.

We next turned our attention to explore other closely related five-membered isosteric dialkylthiazole ring isosteres that would dispose alkyl groups in the extended hinge region similar to 1 (Table 2). Accordingly, triazole analogue 10 was prepared (Scheme 2).¹⁷

Table 2. C-4 Heterocycle SAR^a,b.

^aAssay protocols are provided in the Supporting Information.

^bAssay results are the average of at least two replicates.

Scheme 2.

Reagents and conditions: (a) acetyl isocyanate, acetone, 35 °C, 44%; (b) methyl hydrazine, AcOH, 80 °C, 31%.

Triazole 10 displayed modest JAK2 potency and high JAK family selectivity. We postulated the loss of potency may be due to a disfavored interaction of the triazole ring nitrogen with the pyridyl nitrogen forcing the rings to adopt a less planar conformation than compound 1. Removal of a nitrogen from the triazole ring re-established the planarity (vide infra) and led to the discovery of 1,5-dimethyl pyrazole analogue 11. Compound 11, henceforth referred to as BMS-911543, displayed an IC₅₀ of 1.1 nM against JAK2 and was approximately 350-, 75-, and 65-fold selective vs JAK1, JAK3, and TYK2, respectively. Assessment of dissociation constants of BMS-911543 for JAK1, JAK2, and JAK3 indicated greater selectivity with K_i values of 110, 0.48, and 360 nM, respectively. BMS-911543 was also evaluated in the KinomeScan (formerly Ambit) panel (consisting of 451 kinases) as well as the internal kinase panel to assess overall kinome selectivity. It displayed a high level of selectivity across the kinome (see Supporting Information for complete data set).¹⁸

The X-ray structure of BMS-911543 bound to the JAK2 kinase domain displayed a similar binding mode as 1. One of the nitrogens of the pyrazole ring formed a hydrogen bond with Tyr931 while maintaining coplanarity with the pyrrolopyridine scaffold. The 1,5-dimethyl pyrazole occupied the extended hinge region where key residue differences such as JAK2-Gln853, JAK3-Ser826, JAK1-Arg868, and TYK2-Arg901 resulted in high selectivity within the JAK family (see Figure 4 for location of other nonconserved residues that may affect selectivity).¹⁹

Figure 4.

Crystal structure of BMS-911543 bound to the kinase catalytic domain of JAK2. The carbons of BMS-911543 are colored in pink and the carbons for JAK2 are colored in green except for the residues near the C-4 group, which differ in the JAK family (carbons are colored cyan). Oxygens are colored red, nitrogens blue, and sulfurs yellow. Hydrogen bonds are indicated with dashed lines.

As expected the introduction of bulkier substitution on the pyrazole nitrogen resulted in further enhancement of selectivity within the JAK family (>140-fold) as observed for the analogue 12 while retaining the positive attributes found in BMS-911543. In contrast, 1,3-dimethylpyrazole substitution (13) was found to be detrimental to JAK2 potency and selectivity, probably due to lack of the hydrogen bonding interaction with Tyr931 and suboptimal hydrophobic interaction with the extended hinge region. Consistent with our model, removal of the methyl group from the pyrazole nitrogen (14) led to significant loss of JAK family selectivity.

Although compound 12 displayed superior JAK family selectivity, it demonstrated higher potential for QT prolongation in the patch clamp hERG channel assay (75% and 20% inhibition for 12 and BMS-911543 at 30 μM, respectively). In addition, 12 also showed an inferior pharmacokinetic profile compared to BMS-911543 and hence was not progressed further for additional studies.

BMS-911543 and related pyrazoles were synthesized using the reaction sequence depicted in Scheme 3. Condensation of 2 with 4,4-bis(methylthio)but-3-en-2-one gave intermediate 15. Initially, 14 was directly combined with methyl hydrazine furnishing a 1:9 mixture of desired 1,5- and undesired 1,3-dimethylpyrazole regioisomers (BMS-911543 and 13, respectively).²⁰ By subjecting 15 to condensation with tert-butyl 1-methylhydrazinecarboxylate followed by subsequent treatment with formic acid, BMS-911543 was formed exclusively.²¹ The reaction sequence proceeds through the kinetically formed intermediate 16, which undergoes cyclization after Boc-deprotection to yield the desired regioisomer. Pyrazoles 12 and 14 were prepared from 15 in analogous manner as BMS-911543.

Scheme 3.

Reagents and conditions: (a) NaH, 4,4-bis(methylthio)but-3-en-2-one, DMF, RT, 75%; (b) methyl hydrazine, EtOH, 85 °C, 30%; (c) tert-butyl 1-methylhydrazinecarboxylate, AcOH, 50 °C; (d) formic acid, 60 °C, 65%.

In cellular assays, BMS-911543 showed potent antiproliferative activity in the SET-2 as well as BaF3-V617F engineered cell lines (both dependent upon JAK2 pathway), with IC₅₀ values of 60 and 70 nM, respectively. The antiproliferative activity of BMS-911543 in SET-2 and BaF3-V617F cells correlated with similar activity on constitutively active pSTAT5 (IC₅₀ 80 and 65 nM, respectively). In contrast, non-JAK2-dependent cell lines (A549, MDA-MB-231, MiaPaCa-2) were significantly less sensitive to the inhibitor treatment. The excellent biochemical selectivity versus JAK1/3 translated to good cellular and functional selectivity in an IL-2 mediated T-cell proliferation assay (IC₅₀ 990 nM).¹² Also, cell lines that rely on other JAK family members, including CTLL2 and parental BaF3 cells stimulated with IL-3, showed weak antiproliferative activity for BMS-911543 (IC₅₀ 2.9 and 3.5 μM, respectively) (Figure 5).¹²

Figure 5.

Biochemical and cellular data summary of BMS-911543.

BMS-911543 suppressed the pSTAT5 levels (mediated by wild type JAK2) relative to vehicle control when stimulated with thrombopoetin (TPO) in a mouse pharmacodynamic model.¹² The responses were dose dependent and resulted in nearly complete normalization of pSTAT5 levels for 18 h at the highest oral dose of 30 mg/kg. At an intermediate 10 mg/kg oral dose, ∼65% reduction was observed up to 18 h, whereas at the 5 mg/kg dose, approximately 50% reduction in pSTAT5 for 8 h was achieved. Observed pSTAT5 reductions correlated with exposures of BMS-911543, with AUC_0–8h values of 23, 41, and 109 μM·h, respectively, for dose levels of 5, 10, and 30 mg/kg. In addition, BMS-911543 demonstrated a potent and sustained (2 mg/kg up to 7 h) PD effect in blocking pSTAT5 formation in mice grafted with human SET-2 cells harboring JAK2-V617F mutation.

In in vitro ADMET profiling assays, BMS-911543 showed good metabolic stability, excellent intrinsic permeability, and moderate drug–drug interaction potential based upon CYP inhibition of the CYP3A4 and CYP1A2 isoforms. In an in vitro safety panel consisting of 45 targets, BMS-911543 showed IC₅₀ > 25 μM for all targets except PDE4 (IC₅₀ 5.6 μM). BMS-911543 was not mutagenic or clastogenic in exploratory Ames and in vitro micronucleus assays, respectively. In addition to weak activity in the patch clamp hERG assay, BMS-911543 also showed similar trends in in vitro Na⁺ and Ca²⁺ binding assays, indicating a low potential to cause cardiovascular effects (Table 3). In in vitro biotransformation studies with human liver microsomes, BMS-911543 formed small amounts (∼4%) of 1-demethylated metabolite (compound 14) as the major metabolite.

Table 3. In Vitro ADMET Profile of BMS-911543.

metabolic stability (T1/2 min)	49 (H), 108 (M), 69 (R), 70 (D)
PAMPA (pH 7.4)	698 nm/sec
HLM CYP IC50	all >40 μM; except CYP3A4 (3 μM) and 1A2 (1 μM)
hERG patch clamp	20% inh at 30 μM
Na+ patch clamp	<10% inh at 10 μM (1 Hz and 4 Hz)
Ca2+ patch clamp IC50	>80 μM
PXR EC50	>50 μM
HEPG2 IC50	>50 μM

The pharmacokinetics of BMS-911543 was investigated in mice, rats, dogs, and monkeys (Table 4 and Supporting Informatoin). The absolute oral bioavailability in solution was >50% in all the species tested. In addition, the absorption of BMS-911543 was not significantly impacted by particle dissolution (suspension formulation), with a relative bioavailability (vs solution) of ∼60% in rats and ∼100% in dogs.

Table 4. Pharmacokinetic Profile of BMS-911543.

PK	mousea	ratb	dogc	cynod
CL (mL/min/kg)	0.55	0.7	6.5	5.3
Vss (L/kg)	0.26	0.3	1.6	1.1
T1/2 (h)	5.1	5	2.2	2.4
F (%)	100	100	82	53

^aMice were dosed 2.0 mg/kg IV and 10.0 mg/kg PO.

^bRats were dosed 1.0 mg/kg IV and 1.0 mg/kg PO.

^cDogs were dosed 0.2 mg/kg IV and 0.2 mg/kg PO.

^dCyno were dosed 1.0 mg/kg IV and 1.0 mg/kg PO.

In single-dose toxicological studies, BMS-911543 was well tolerated up to 100 mg/kg in rats (mean AUC_0–72h 11300 μM·h) and dogs (AUC_0–24 610 μM·h). In two-week repeat dose studies in rats, a 15 mg/kg/day dose (Day 14 AUC_0–24 3200 μM·h) was well tolerated. The most sensitive effects observed were decreases in reticulocytes and subsequent reductions in red blood cell mass. These effects, and observed decreases in platelets, are consistent with JAK2 inhibition.

In summary, ADMET and X-ray structure-guided refinement of the C-4 heterocycle to address metabolic liability present in 4,5-dimethylthiazole 1 led to the discovery of BMS-911543 (11), with excellent kinome selectivity, in vivo pharmacodynamic activity, and safety profile. BMS-911543 is currently in clinical trials for the treatment of MF.²²

Supporting Information Available

Full experimental details for key compounds and biological protocols. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.5b00226.

Author Present Address

^† Immunomedics, 300 The American Road, Morris Plains, New Jersey 07950, United States.

Author Present Address

^‡ Xenex, 121 Interpark, Suite 104, San Antonio, Texas 78216, United States.

Author Present Address

^§ Janssen Pharmaceutical Companies of Johnson and Johnson, 1400 McKean Rdoa, Spring House, Pennsylvania 19002, United States.

Author Present Address

^∥ BioMotiv, 20600 Chagrin Boulevard, Suite 210, Cleveland, Ohio 44122, United States.

Author Present Address

^⊥ ARMGO Pharma Inc., 777 Old Saw Mill River Road, Tarrytown, New York 10591, United States.

Author Present Address

^# Biocon BMS R&D Center (BBRC), Bangalore, Syngene International Ltd., Plot No. 2 and 3, Bommasandra IV Phase, Jigani Link Road, Bangalore 560 099, India.

Author Contributions

All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.