|Home | About | Journals | Submit | Contact Us | Français|
The identification of Pim-1/2 kinase overexpression in B-cell malignancies suggests that Pim kinase inhibitors will have utility in the treatment of lymphoma, leukemia, and multiple myeloma. Starting from a moderately potent quinoxaline-dihydropyrrolopiperidinone lead, we recognized the potential for macrocyclization and developed a series of 13-membered macrocycles. The structure–activity relationships of the macrocyclic linker were systematically explored, leading to the identification of 9c as a potent, subnanomolar inhibitor of Pim-1 and -2. This molecule also potently inhibited Pim kinase activity in KMS-12-BM, a multiple myeloma cell line with relatively high endogenous levels of Pim-1/2, both in vitro (pBAD IC50 = 25 nM) and in vivo (pBAD EC50 = 30 nM, unbound), and a 100 mg/kg daily dose was found to completely arrest the growth of KMS-12-BM xenografts in mice.
Pim-1, -2, and -3 are highly homologous and constitutively active serine/threonine kinases belonging to the CAMK (calmodulin dependent kinase) family.1 The three Pim isoforms phosphorylate a diverse group of proteins with known roles in proliferation, survival, apoptosis, and differentiation. The identification of oncogene-driven aberrant Pim kinase overexpression in subsets of B-cell malignancies2 including lymphomas, leukemias, and multiple myeloma, as well as in subsets of solid tumors,3 has led to intense efforts to identify small molecule Pim kinase inhibitors.4 Because Pim-1 and Pim-2 overexpression are commonly found in B-cell malignancies and because of the overlapping and compensatory nature of Pim-1 and Pim-2, we believe dual Pim-1/2 kinase inhibitors are likely to provide the most broad and durable clinical benefit in the treatment of B-cell malignancies. It is now widely recognized that cellular inhibition of Pim-2 kinase is a significant challenge, due in part to the low Km of Pim-2 kinase for adenosine triphosphate (ATP), and very few Pim inhibitors have been reported to achieve submicromolar IC50 values in Pim-2 overexpressing cell lines.5 Reported herein is the discovery and optimization of a class of macrocyclic Pim-1/2 kinase inhibitors with double-digit nanomolar activity in KMS-12-BM, a multiple myeloma cell line with relatively high endogenous levels of Pim-1/2.6
Screening of kinase inhibitors from prior kinase programs identified a naphthyridine pyrrolodihydropiperidinone as a potent Pim-1/2 kinase inhibitor that exhibited poor selectivity against many kinase off-targets (Figure Figure11). This screening hit is related to 2-(pyridin-4-yl)-dihydro-pyrrolopyridinones,7−9 which are known to interact with many kinases in part due to favorable interaction of the 4-pyridyl nitrogen with the backbone hydrogen bond donor presented by most kinase hinge residues. We found that replacement of the naphthyridine with a methylquinoxaline (1) resulted in similar Pim-1/2 potency and improved selectivity against kinase off-targets.10 Pim-1/2 are able to accommodate both the naphthyridine and quinoxaline cores because the hinge region of the Pim kinases contains a proline residue and therefore lacks the canonical hydrogen-bond donor of all other kinases. In an effort to diversify the aniline substituent, molecules containing N-alkyl groups were found to be similarly potent (Figure Figure11, 2–3). Consideration of the overall shape of these N-alkyl molecules led to the hypothesis that the linkage of the N-alkyl group and the pyrrolodihydropiperidinone to form a macrocycle would be feasible, and molecular modeling established that a total of five atoms exhibited the minimum necessary length to bridge the quinoxaline and dihydropiperidinone without introducing significant conformational strain (Figure Figure11, model). We speculated that macrocyclization could result in improved potency due to stabilization of the bound conformation11−13 and recognized that the macrocyclic linker itself, or substituents projecting from it, may interact with the protein, specifically P-loop residues Leu-44, Gly-45, and the side chain of Phe-49, more favorably than nonmacrocycles such as 2 and 3.
Ring-closing olefin metathesis14 proved to be a reliable strategy for installing the 13-membered macrocycle although in some cases an SNAr macrocyclization was necessary. The olefin resulting from ring-closing metathesis could be reduced via catalytic hydrogenation to provide additional molecules with a saturated bridge. The biological data for a simple set of macrocyclic compounds differing in the macrocyclic bridge and configuration at C9 (8–10, a–b) is presented in Table 1. For the evaluation of compounds, our program relied on the Pim substrate Bcl-2-associated death promoter (BAD), a proapoptotic BH3 family member that is inactivated by phosphorylation at Ser-112. A cell-free kinase assay with recombinant Pim-1 and Pim-2 was used to assess the ability of test compounds to inhibit Pim phosphorylation of a short sequence of BAD, and a cellular assay constructed with KMS-12-BM, a multiple myeloma cell line with relatively high endogenous levels of Pim-1/2, was used to assess the ability of test compounds to inhibit Pim phosphorylation of Ser-112 of the full-length BAD protein. Relative to the initial lead molecules 1–3 (Figure Figure11), the macrocycles show comparable or improved potency for Pim-1 and Pim-2 enzyme inhibition, with the 9-(S) stereochemistry associated with improved potency relative to the 9-(R) stereochemistry.
Encouraged by the favorable data for the simple macrocycles, we synthesized additional analogues with methyl substitution on the atoms linking the quinoxaline to the dihydropiperidinone, focusing on the 9-(S) stereochemistry (Table 1, 8–10, c–f). Substitution with a methyl group was chosen for two reasons. First, due to the presence of accessible lipophilic space in the Pim protein, especially in the region of the macrocyclic linker proximal to the amino group, we believed additional potency could be gained by addition of a small hydrophobe. Second, the initial macrocycles exhibited poor oral bioavailability (%F in rat for 8a, 9a, and 10a was 2, 11, and 16, respectively) believed to be due largely to poor aqueous solubility. The poor aqueous solubility, in turn, was likely due to high crystallinity afforded by the planar nature of the quinoxaline-pyrrole biaryl and the lack of rotatable bonds, and it was hypothesized that substitution might serve to reduce crystallinity and improve aqueous solubility. On the (E)-alkene framework, the vinyl-methyl macrocycle 8e showed considerable improvement in both potency and solubility, relative to the parent 8a. On the (Z)-alkene framework, both allylic substitution (9c) and vinyl substitution (9d) provided molecules with excellent potency and improved solubility. It was not possible to prepare the analogue of 9c with an epimeric allylic stereocenter via the olefin metathesis route, likely due to A1,3-strain and the unique constraints of the macrocycle. While methyl-substituted analogues (10c–f) of the saturated macrocycle 10a showed good enzymatic and cellular potency, the solubility was uniformly poor.
The X-ray structure of 9c bound to Pim-1 was obtained and is presented in Figure Figure22. The presence of a proline (Pro-123) residue in the ATP-binding “hinge” of the Pim kinases results in the absence of the canonical backbone NH found in other kinases, and this uniquely accommodates the C–H at C19 of the quinoxaline. The amino group at C15 of the quinoxaline is 6.0 Å from the NH of Asp-128, and two water molecules are resolved in this space, consistent with a bridged hydrogen-bonding interaction. Another feature of this series is that C20 of the quinoxaline packs closely against the C=O of Glu-121 (3.1 Å). The carbonyl of the dihydropiperidinone engages in a direct hydrogen bonding interaction with the Lys-67 side chain (C=O to N distance 2.7 Å) and also forms a water-bridged hydrogen bond to the backbone NH of Asp-186. The dihydropiperidinone NH is closely packed against the Asp-186 side chain (N to O distance 3.0 Å). The (S)-configuration at C9 is compatible with the Phe-49 side chain position, which is often observed in Pim-1 structures to be folded under the glycine-rich loop. The macrocyclic linkage connecting the quinoxaline and dihydropyrrolopiperidinone provides a continuous hydrophobic surface, which forms extensive contacts with the P-loop residues Leu-44, Gly-45, and the side chain of Phe-49 (Figure Figure22b). These interactions may contribute to the improved potency of the macrocyclic series over the related nonmacrocyclic compounds 2 and 3.
The kinase selectivity profile of 9c was determined in a panel containing 392 nonmutant kinases (scanMAX KINOMEscan, DiscoveRx) at a compound concentration of 1 μM, and the S(10) and S(1) selectivity scores, representing the fraction of tested kinases with percent-of-control (POC) values <10 and <1% were determined to be 0.048 and 0.003, respectively.15 Nineteen kinases including Pim-1/2/3 exhibited POC values of less than 10%, with possible off-targets being six members of the Casein kinase family, two members each of the CLK, DYRK, and HIPK families, as well as MYLK4, MAPK15, haspin, and YSK4. These kinases are, to the best of our knowledge, not known to contribute to BAD phosphorylation. The alpha isoforms of Casein kinase I (CK1, CSNK1)16 and II (CK2, CSNK2)17 have been implicated in contributing to multiple myeloma pathology, although the relevance of CK1 and CK2 in the setting of Pim overexpression is not known. Therefore, we believe off-target kinase inhibition to be a limited risk in the interpretation of the impact of 9c on BAD phosphorylation, but activity on CK1α/CK2α may impact the viability of KMS-12-BM.
Pharmacokinetic profiles and calculated physicochemical properties of macrocycles possessing good potency and reasonable aqueous solubility (8e, 9c, and 9d) are presented in Table 2. In general, these molecules exhibited moderate microsomal stability, high plasma protein binding, and good permeability. The in vivo profiles in male Sprague–Dawley rats revealed significant differences in clearance (0.47 to 2.4 L/h/kg, representing 14–72% of liver blood flow) and oral bioavailability (17–48%). Microsomal turnover was not predictive of in vivo clearance, suggesting that conjugation, transporter-mediated processes, or extrahepatic metabolism may have a role in the clearance of these compounds in rat. Compound 9c was judged to have the best overall profile (lowest clearance and highest oral bioavailability), and this molecule was advanced to in vivo xenograft studies.
As 9c exhibited potent Pim-1/2 kinase activity in the KMS-12-BM cell line (Table 1, IC50 = 25 nM), potent inhibition of KMS-12-BM cell viability in a 72 h assay (Figure SI-2, IC50 = 151 nM), and an acceptable pharmacokinetic profile in rodents, further studies were conducted in a mouse xenograft model. The ability of 9c to impact Pim-dependent phosphorylation of BAD (Ser-112) in vivo was determined in 350 mm3 KMS-12-BM tumor xenografts in female SCID-beige mice (Figure Figure33a). In this model, we typically observe a maximum inhibition of 75% BAD phosphorylation at Ser-112 for selective Pim kinase inhibitors; the remainder of BAD phosphorylation is attributed to other kinases, possibly PKB (AKT).18 Compound 9c dosed orally at 25, 50, or 100 mg/kg provided a dose- and concentration-dependent decrease in KMS-12-BM xenograft pBAD levels, with 100 mg/kg achieving greater than 50% inhibition of pBAD through 16 h postdose. The EC50 for pBAD inhibition was determined to be 3.6 μM total plasma 9c by fitting all individual concentration–response data presented in Figure Figure33a. As the fraction unbound of 9c in mouse plasma is 0.0084,19 this corresponds to an unbound plasma concentration of 30 nM, in good agreement with the measured in vitro IC50 of 25 nM (Table 1). The same KMS-12-BM xenograft model was employed to determine the impact of 9c on KMS-12-BM tumor growth in vivo (Figure Figure33b). Compound 9c dosed orally once daily at 10, 25, 50, or 100 mg/kg provided a dose-dependent decrease in KMS-12-BM tumor volume compared to vehicle treatment, with 100 mg/kg achieving complete tumor growth inhibition. All doses were well tolerated, with minimal impact on body weight over the course of the study. These data suggest that in vivo exposures above the in vitro IC50 for at least 16 h with greater than 50% pBAD inhibition result in significant tumor growth inhibition.
In conclusion, a series of macrocyclic Pim-1/2 kinase inhibitors was developed from nonmacrocyclic precursors. Some macrocycles exhibited large improvements in potency in both Pim-1 and Pim-2 cell-free enzyme assays as well as in a cellular assay measuring Pim-1/2-dependent phosphorylation of BAD. The improved potency may result from stabilization of the bound conformation as well as additional favorable van der Waals interactions between the macrocyclic linker and Pim protein. The (Z)-configured, allylic-methyl substituted macrocycle 9c was judged to exhibit the best combination of enzyme and cellular potency, solubility, and pharmacokinetics. In a KMS-12-BM tumor xenograft model, 9c dosed orally was capable of inhibiting Pim-dependent phosphorylation of BAD with an EC50 of 3.6 μM (30 nM free unbound 9c), and the highest dose studied (100 mg/kg) provided greater than 50% inhibition of BAD phosphorylation for at least 16 h. A daily oral dose of 100 mg/kg 9c was capable of completely inhibiting the growth of KMS-12-BM xenografts in a 16-day efficacy model and was well tolerated. Compound 9c represents the first example of a macrocyclization strategy leading to improved Pim kinase inhibitors, and we have demonstrated that 9c has properties suitable for the in vivo study of Pim kinase inhibition.
We thank Wes Barnhart, Kyung Gahm, and Sam Thomas for chiral SFC separation of intermediates and final compounds; Heather Eastwood for high throughput solubility determination; Chris Scardino for ICP (inductively coupled plasma) metal analysis; Ed Lobenhofer for comparative biology and safety science support; and Iain Campuzano for high resolution mass spectrometry.
All authors have given approval to the final version of the manuscript.
The authors declare no competing financial interest.