Optimization of Selective Mitogen-Activated Protein Kinase Interacting Kinases 1 and 2 Inhibitors for the Treatment of Blast Crisis Leukemia

ABSTRACT: Chronic myeloid leukemia (CML) is a myeloproliferative disease caused by bcr-abl1, a constitutively active tyrosine kinase fusion gene responsible for an abnormal proliferation of leukemic stem cells (LSCs). Inhibition of BCR-ABL1 kinase activity offers long-term relief to CML patients. However, for a proportion of them, BCR-ABL1 inhibition will become ineffective at treating the disease, and CML will progress to blast crisis (BC) CML with poor prognosis. BC-CML is often associated with excessive phosphorylated eukaryotic translation initiation factor 4E (eIF4E), which renders LSCs capable of proliferating via self-renewal, oblivious to BCR-ABL1 inhibition. In vivo, eIF4E is exclusively phosphorylated on Ser209 by MNK1/2. Consequently, a selective inhibitor of MNK1/2 should reduce the level of phosphorylated eIF4E and re-sensitize LSCs to BCR-ABL1 inhibition, thus hindering the proliferation of BC LSCs. We report herein the structure−activity relationships and pharmacokinetic properties of a selective MNK1/2 inhibitor clinical candidate, ETC-206, which in combination with dasatinib prevents BC-CML LSC self-renewal in vitro and enhances dasatinib antitumor activity in vivo.

INTRODUCTION
Messenger RNA (mRNA) translation is a highly regulatedprocess pervasive to many biological processes, in which eukaryotic translation initiation factor 4E (eIF4E), a cap- dependent translation factor, plays a crucial role. eIF4E, one of the five MAP kinase interacting serine/threonine kinases 1 and 2 (MNK1/2) substrates,1 is phosphorylated in vivo only bycentral nervous system disorders,6 and malignancies such as glioblastoma,7−9 breast cancer,10−12 pancreatic ductal adeno- carcinoma,13,14 and leukemias.15,16 Abnormally high mRNA translation of oncogenic transcripts is thought to fuel the growth of cancers, including chronic myeloid leukemia (CML), where tumorigenic pathways are activated through eIF4E overexpression and/or phosphorylation.17−22 Specifically, blastMNK1/2 on Ser209, and accordingly, phospho-eIF4E is not detected in Mnk1/2 knock-out mice.2 Overexpression of phosphorylated eIF4E has been linked to inflammation,3−5aReagents and conditions: (i) acetamide or benzamide, N,N-dimethylethylenediamine, K2CO3, CuI, Toluene, 90 °C, 12 h; (ii) NBS, CH2Cl2, 1 h;(iii) (4-hydroxyphenyl)boronic acid, Na2CO3, Pd(OAc)2, P(PPh3)4, 1,4-dioxane, 90 °C, 12 h; (iv) (4-hydroxyphenyl)boronic acid, Na2CO3, Pd(PPh3)4, 1,4-dioxane, 90 °C, 12 h; (v) phenylboronic acid, Pd(dppf)2Cl2, Cs2CO3, DMF, H2O, 140 °C, 30 min.Scheme 2. Synthesis of Imidazopyrazine Analogues 19−33aaReagents and conditions: (i) (4-(ethoxycarbonyl)phenyl)boronic acid, Pd(dppf)2Cl2, Cs2CO3, toluene, 90 °C, 12 h; (ii) NBS, MeOH, CCl4 or NIS, DMF; (iii) LiOH, H2O, THF, CH3OH; (iv) R1R1′NH, HATU, NMM, DMF; (v) arylboronic acid, Pd(PPh3)4, K3PO4, Na2CO3, H2O, 1,4-dioxane,90 °C, 12 h; (vi) DIPEA, HOBt, EDCI·HCl, O-(tetrahydro-2H-pyran-2-yl)hydroxylamine, CH2Cl2, DMF, rt, 12 h; (vii) 1M aq HCl, CH3CN,MeOH, rt, 4 h.crisis chronic myeloid leukemia (BC-CML) is a dire stage of CML, where patients are no longer responsive to treatment with BCR-ABL1 kinase inhibitors23 such as imatinib, dasatinib, or nilotinib,24,25 that in most cases ensure long-term survival to patients in chronic phase (CP) CML. BC-CML patients have high levels of nuclear β-catenin,16,17,26 MNK1/2, eIF4E, and phosphorylated eIF4E27,28 that are believed to promote not only resistance to treatment with BCR-ABL1 inhibitors, but also stimulate a proliferation of leukemic stems cells (LSCs)16,29 via self-renewal.30 Lim et al.28 have shown that phospho-mimetic eIF4E (eIF4E-209D) elicits LSC self-renewal ability whereas the un-phosphorylatable eIF4E (eIF4E-209A) does not.

They have, along with others, also shown that inhibition of β-catenin impairs the ability of primary BC-CML cells to self-renew to form colonies16,26 and that thetumorigenic MNK/eIF4E/β-catenin axis plays an important role in BC-CML as well as in other cancers.28,31,32The current best and only option to cure BC-CML is an allogenic bone marrow transplantation33 (only beneficial to a small proportion of patients33), and consequently, long-term treatment of patients with BC-CML remains an unmet medical need.Pharmacologic and knock-down inhibition of eIF4E phosphorylation by MNK1/2 prevents β-catenin activa- tion28,31,32 and suppresses the ability of LSCs to self-renew,28 indicating that selective MNK1/2 kinase inhibition by a small molecule could be an avenue to treat patients with BC-CML. In addition, selective MNK1/2 kinase inhibition may be associated with minimal adverse effects, as Mnk knockout mice2 are viable, healthy, and have normal hematopoiesis.aReagents and conditions: (i) (4-(4-methylpiperazine-1-carbonyl)phenyl)boronic acid, Pd(PPh3)4, toluene, EtOH, K2CO3, 140 °C, microwave reactor, 30 min; or morpholino(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)methanone, Pd(dppf)2Cl2, DMF, H2O, Cs2CO3, microwave reactor, 140 °C, 30 min; (ii) 4-boronobenzoic acid, Pd(dppf)2Cl2, DMF, H2O; Cs2CO3, 90 °C, 18 h; (iii) tert-butyl piperazine-1-carboxylate, HATU, NMM, DMF, rt, 18 h; (iv) TFA, CH2Cl2, rt; (v) N,N-dimethylpiperidin-4-amine, HATU, NMM, DMF; (vi) (4-cyanophenyl)boronic acid, K2CO3, Pd(PPh3)4, 1,4-dioxane, H2O, 90 °C, 6 h.Several inhibitors of MNK1/2 kinases have recently been reviewed,34,35 notably CGP5738036−38 and cercosporamide.39 Both cercosporamide and CGP57380 lack kinase selectivity, and the latter is a weak MNK1/2 inhibitor with potency in the low micromolar range.

Theodosia et al. have revealed MNK inhibitors with anti-proliferative activity against AML cell lines.40,41 Likewise, merestinib, a nonselective MEK inhibitor, was also shown to inhibit MNK1/2 in AML progenitors with anti-leukemic effects.42 The effectiveness of pharmacologic co- inhibition of MNKs and mTORC1 against BC-CML cells wasrecently published.43,44 Han et al.45 have disclosed potent MNK1/2 inhibitors but revealed no disease relevant biological data. Recently, two MNK inhibitors, BAY1143269 and eFT508, have progressed to clinical trial. BAY1143269,46,47 a predom- inantly MNK1 inhibitor, was administered in a phase I clinical study to patients with advanced solid tumors.48 eFT508, a potent and selective MNK1/2 inhibitor, is in phase II for the treatment of lymphoma, colorectal cancer, triple negative breast cancer, and hepatocellular carcinoma.49−51To date there are little efficacy data and very few publicationssupporting the use of small-molecule MNK inhibitors to treat BC-CML. We have recently published our discovery efforts to generate a dual MNK and BCR-ABL1 inhibitor for thetreatment of BC-CML52 with a single agent, and we report herein a complementary approach with the design, synthesis, and structure−activity relationships of very selective MNK1/2 inhibitors. These compounds intentionally lack BCR-ABL1 inhibitory activity, and therefore can potentially be combined with the appropriate tyrosine kinase inhibitor for the treatment of BC-CML patients. Starting from fragments, small-molecule inhibitors of MNK1/2 were optimized, and a nominated preclinical development candidate was tested in a K562 mouse xenograft model.The compounds described herein fall into three classesfollowing the three central bicyclic rings explored, imidazopyr- azine, imidazopyridazine, or imidazopyridine.Imidazopyrazines.

Compounds 4 and 5, as shown in Scheme 1, were prepared in three steps starting from commercially available 6-bromoimidazo[1,2-a]pyrazine 1, which was submitted to a copper mediated amination leading to 6-amide imidazopyrazines 2a,b. The amide derivatives were brominated using NBS to give 3-bromo-6-amide imidazopyr- azines 3a,b, which were subjected to a Suzuki cross coupling reaction to afford 4 and 5, respectively. Compound 8 wasprepared in two successive Suzuki coupling steps, as depicted in Scheme 1, with 66% overall yield.Imidazopyrazine analogues 19−33, shown in Scheme 2, were prepared in a similar manner from 1. Beginning with a Suzuki coupling to introduce a para-benzoyl ester at the imidazopyr-azine 6-position to give 9, which was halogenated at C-3 with either NBS or NIS to afford compounds 10 and 11. The esters were hydrolyzed to carboxylic acids 12 and 13 that, in turn were converted to amides 14−18. Arylation at C-3 using asecond Suzuki cross coupling afforded 19−25 and 27−33.Carboxylic acid 25 was converted into a protected hydroxamic acid intermediate, which was treated with an aqueous solution of hydrochloric acid to afford 26.Compounds 36 and 37 were synthesized in three steps comprising a Suzuki coupling, ester hydrolysis, and amide formation, as depicted in Scheme 3.Imidazopyridazines. As illustrated in Scheme 4, com- pounds 39 and 40 were both prepared in one step from 38. Compound 42 was synthesized in two steps starting with a Suzuki cross coupling between 38 and 4-boronobenzoic acid to give 41, which was converted to amide 42. Compound 45 was also prepared in two steps from commercially available acid 43, which was converted to amide 44 followed by a Suzuki cross- coupling reaction to give 45.Imidazopyridines. The preparation of 47 and 48 is depicted in Scheme 5.aReagents and conditions: (i) (4-methylpiperazin-1-yl)(4-(4,4,5,5- tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)methanone, (A- Phos)2PdCl2, NaHCO3, DMF, H2O, 90 °C, 1 h; or morpholino(4- (4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)methanone, Pd- (PPh3)4, K3PO4, 1,4-dioxane, H2O, 90 °C, 19 h.

RESULTS AND DISCUSSION
Fragment inhibitors of MNK1 published by Oyarzabal et al.53offered a diversity of scaffolds among which, a few imidazopyridazine analogues displayed high ligand efficiencies ranging from 0.32 to 0.52 kcal/mol per heavy atom. We found the imidazopyridazine scaffold to be an attractive starting point to generate novel inhibitors of MNK1/2 with drug-like properties, suitable for proof of concept in man. Accordingly, we gave it the highest priority.Docking studies suggested that interactions with MNK1/2 could be improved by varying the bicyclic core and by modifying the substituents at C-3 and at C-6, whose alterations would considerably widen the structure−activity relationships (SAR) scope. Bicyclic cores were only selected, if modeling proposed a binding mode to the kinases similar to that of the imidazopyridazine. All these taken together suggested a need for SAR exploration of bicyclic analogues bearing substituents at both C-3 and C-6. Accordingly, our strategy was to divide this scaffold into three parts as outlined in Figure 1, where eachpart was investigated and the best of each were combined to give an optimized MNK1/2 inhibitor.Modeling. We relied on molecular modeling to understand how our bicyclic-derived inhibitors interact with MNK1/2 and to guide optimization of their potency. A common structural feature in both MNK kinases is the DFD54 motif, in which an aspartic acid Asp193 and Asp228 for MNK1 and MNK2, respectively, replaces the glycine of the DFG motif that is more common to kinases. A DFD-out structure of MNK1 (PDB ID: 2HW6)55 is available, while MNK2 has been crystallized in both the DFD-in (PDB ID: 2HW7)55 and DFD-out (PDB ID: 2AC3)54 conformations.

To validate our modeling, we searched the PDB for a co-crystal structure with inhibitors having the same scaffold. We found a structure (PDB ID: 4YMJ) of TRK kinase in the DFG-out conformation with a 4-(imidazo- pyridazin-3-yl)-benzonitrile inhibitor.56 This compound has the same core as some of our inhibitors, and docking studies supported that it binds similarly to the TRK and MNK kinases. Consequently, we used the MNK DFD-out conformation in our modeling studies. The ATP binding sites of MNK1 and MNK2 share high sequence homology,57 and modeling indicates that the compounds bind similarly to both kinases. Specifically, the bicyclic core acceptor nitrogen (N-1) hydrogen bonds with the amide backbone of the kinase hinge residues Leu127 in MNK1 and Met162 in MNK2. The C-3 substituents reside in the hydrophobic pocket formed by residues Phe124, Val63, Leu108, and Cys190 in MNK1 and the homologous residues Phe159, Val98, Leu143, and Cys225 in MNK2. The substituents at the core 6-position occupy the phosphate binding pocket close to the ATP ribose binding pocket. A major part of the inhibitor is buried inside the protein except for the substituents at the 6-position, which interact with the glycine-rich P-loop and the loop connecting the hinge region and helix αD of the C-terminal lobe. Compounds were docked into MNK1 and MNK2 using the automated docking program Glide 2017-3 release58 using standard settings, and showed that top-ranked poses had the same binding mode as when manually modeled. The compounds described in this article interact mostly with the same amino acid residues in both MNK1 and MNK2. Furthermore, they inhibit both MNK1 and MNK2 with similar potency (less than 2-fold difference in all cases).

In light of the similarities of the amino acids engaged in compound binding to MNK1/2 and the similarity between the inhibitory activities of the compounds against both MNK1 and MNK2,we will only discuss the modeling of inhibitors bound to MNK2.The inhibitory activity of all compounds against MNK1/2 was assessed in a biochemical assay52 with ATP at Km concentration of 1.2 mM and 250 μM for MNK1 and MNK2 assays, respectively. Compounds with a potency of 1 μM or lower against either MNK1 or MNK2, were assayed in a HeLa cell line52 for their ability to inhibit the phosphorylation of eIF4E by MNK1/2. This functional assay relies on a sandwich immunoassay to quantify phosphorylated eIF4E. In both enzymatic and cell-based assays, each compound was tested in a dose-dependent manner, in duplicate, using eight points from a 3-fold serial dilution. IC50 values were determined by nonlinear regression using GraphPad prism, and the values reported are the average of at least two experiments.C-6 Substituent Variation. A quick probing of bicyclic scaffolds led to the imidazopyrazine derivatives shown in Table1. The acetamide of 4 was replaced by a benzamide as in 5aIn vitro assay where a compound’s ability to inhibit MNK1 or MNK2 phosphorylation of a peptide substrate is determined as IC50 values from the average of at least two measurements. bIn vitro assay where a compound’s ability to inhibit MNK1/2 phosphorylation of eIF4E on Ser209 and IC50 values were determined as the average of at least two measurements. n.a. = not available.which resulted in potency loss. The inhibitory activity toward MNK1/2 was restored by reversing the benzamide in 5 resulting in 19. The primary amide function was removed to yield compound 8, which was slightly less potent than compound 19, suggesting that the amide of 19 contributes to potency.Modeling of 4 (Table 1) into the ATP site of MNK2 shows that the imidazole N-1 hydrogen bond acceptor forms a hydrogen bond to the kinase hinge (Figure 2), very similar to that established by the N-1 of the adenine base of ATP. This is the backbone NH hydrogen bond donor of residues Leu127 in MNK1 and Met162 in MNK2.

This hydrogen bond is stable during the molecular dynamics (MD) simulations (Supporting Information, MD file compound 4). The 2-position of the imidazopyrazine is pointing toward the oxygen of the backbone carbonyl of Glu160, and the 8-position is pointing toward the oxygen of the backbone carbonyl of Met162.Interactions between the aromatic hydrogen at the 2- and 8- positions and the backbone carbonyls are expected to be favorable to inhibitor binding59 to the kinases. When either the 2- or 8-position is substituted all activity is lost (Supporting Information, Table S1).The phenyl in the 3-position has extensive hydrophobic interactions with the kinase. It is sandwiched between the gatekeeper residue Phe159 and the DFD-motive residue Phe227 and engages in van der Waals contacts with the side chains of Val98, Ala111, Leu143, Cys225, and Leu212. The hydroxyl of 4 (Table 1) forms a stable hydrogen bond per MD simulations, with the backbone carbonyl of Asp226 from the DFD motif. The acetamide is solvent exposed on one side and binds in a largely hydrophobic area, where it interacts with the side chains of Leu90, Leu212, and Phe227. During the MD simulations, the nitrogen or carbonyl of the amide (4, 5) periodically hydrogen bonds with the backbone NH or side chain of Ser166 either directly or water mediated. However, these interactions are not stable during the MD simulations and probably do not contribute much to the binding affinity.

The substitution of the methyl of acetamide (4) for a phenyl(5) led to a drop in potency, possibly due to a steric clash between the phenyl and Leu212 of the glycine-rich loop. Removing the amide linker favors a better interaction between the phenyl directly connected to C-6 (8, 19) and residues Leu90, Leu212, and Phe227. Furthermore, the backbone NH of Ser166 is pointing toward the face of the phenyl resulting in an attractive electrostatic interaction between the negative π cloud of the phenyl and the partial positive charge of the NH donor (Figure 3). These observations are in favor of the improvement in potency observed when a phenyl occupies the C-6 position. The amide of 19 marginally improved potency relatively to 8, probably due to a hydrogen bond between the carbonyl and the side-chain hydroxyl of Ser166. Both compounds 8 and 19 inhibit eIF4E phosphorylation in cells with similar potencies, and 19 exhibited good permeability and favorable in vitro ADME properties, as shown in Table 2.The primary amide of 19 was replaced by larger amides to yield N-methylpiperazine benzamide 20 and morpholine benzamide 32 (Table 3), both equipotent with 19, but with improved permeability and solubility. The inhibitory activities of 19, 20, 32, and 36 showed that the substituents represented by the R1 equal to O, NCH3, NAc (Table 3) had a moderate impact on inhibitory potency toward MNK1/2. Modeling (Figure 3) shows that the phenyl at 6-position makes van der(22) or meta (23) positions resulted in a 5- to 30-fold loss in potency. The following para substituents illustrate how this part of the binding pocket can accommodate both hydrophobic and hydrophilic groups. The OH was replaced by CH3 (24), COOH (25), CONHOH (26), CN (27), Cl (28), CF3 (29),CHF2 (30), and CONH2 (31) and led to derivatives that inhibit MNK1 and MNK2 with potencies similar to that of 20.

The hydroxamic acid derivative 26 showed a substantial improvement in potency relative to 20, but 26 as well as the primary amide (31) and acid (25), all had poor permeability and consequently lacked cell-based activity. The carboxylic acid was replaced by a tetrazole (Supporting Information, Table S3) to give an equipotent compound, but with a bad permeability and practically no cell activity. The benzonitrile 27, with a LogD of 1.67 (Table 3) is less lipophilic than analogues of similar potency (28, 29, 30) and displayed excellent solubility and permeability properties. Accordingly, the benzonitrile group was selected as the preferred substituent at C-3. At this stage, the improved compounds, exemplified by 27 and 33, exhibited an imidazopyrazine with a benzonitrile at C-3 andmethylpiperazine or morpholine benzamide at C-6.MNK2 X-ray structures were examined to understand why the para position of the C-3 phenyl tolerates both hydrophobic and hydrophilic groups. Figure 2 represents 4 docked into MNK2 and shows that the para position is surrounded by hydrophobic residues Val98, side chain methylenes of Lys113, Phe227, and Phe159. Consequently, compounds with small hydrophobic para substituents like Me (24), Cl (28), CF3 (29), and CHF2 (30) have more hydrophobic interactions with the kinase than compound 21 the unsubstituted analogue. On the other hand, compounds with more polar substituents such as 4- OH (20), 4-CO2H (25), 4-CONHOH (26), 4-CN (27), and4-CONH2 (31) also have good inhibitory activity that could bethe result of hydrogen bonding to the backbone carbonyl ofCyp 3A4/2D6 IC50 (μM)>20/>20Asp226 for hydrogen bond donating substituents (Figure 2) oraAverage of at least two determinations, except for MW. bDistributioncoefficient between 1-octanol and aqueous phosphate buffer at pH 7.4. cParallel artificial membrane permeability assay at pH 7.4. dHalf-life in human and mouse liver microsomes. eIC50 for Cyp 3A4 and 2D6 inhibition.Waals contacts with the side chains of Leu90, Leu212, Ser166 β-carbon, and Gly165 α-carbon.

In addition to the favorable interaction between the NH of Ser166 and the phenyl at C-6 described above, the piperazine substituent makes van der Waals contacts with the methylenes of Glu92 and Glu209. However, these residues are on the edge of the binding site and are solvated. Any interaction between the piperazine basic nitrogen and either acid is not likely to contribute much to the binding affinity. The methylpiperazine benzamide 20 and morpholine benzamide 32 amides appear to be the best overall C-6 substituents.C-3 Substituent Variation. We next kept the methyl- piperazine benzamide, our best solubilizing group, constant at C-6, and explored derivatization at C-3. Moieties such as pyridine derivatives, pyrimidine, bicyclic heterocycle, carbonyl bridged heterocycles failed to improve potencies relative to a phenyl substituent (Supporting Information, Table S2). These observations are in agreement with modeling studies, that support the hydrophobic nature of the binding site occupied by the phenyl at C-3. Consequently, our SAR investigations were then narrowed to only substituting the para position of phenyl at C-3. Removing the hydroxyl (21) or moving it to the orthowith the basic side chain of Lys113 for acidic or hydrogen bond accepting groups as shown in Figure 3.Modeling studies showed that the cyano group of 27 forms a hydrogen bond with Lys113, and during an MD simulation (Supporting Information, MD file compound 27), this hydrogen bond was present for 60% of the time. The increased potency of 26 could be the result of two hydrogen bond interactions between the NH and OH of the hydroxamic acid group to Asp226 and Lys113, respectively.C-6 Substituent Variation. We re-examined the suitability of benzamide moieties at C-6 to verify that the SAR established with the C-3 phenol derivatives holds true for the C-3 benzonitrile analogues. Accordingly, two benzonitrile deriva- tives, morpholino (33) and N,N-dimethylamino piperidinyl(37) were generated. 27 and 33 are equipotent toward MNK1/ 2 inhibition, just as the phenolic derivatives 20 and 32 are.

Analogues 27, 33, and 37 were all of similar potency toward MNK1 inhibition, but for MNK2, 37 is 3- and 4-fold more potent than 27 and 33, respectively.Compounds 27, 33, and 37 were found to have favorable ADME properties as shown in Table 4. They were administered orally to mice, and 27, with a bioavailability of 48%, was the fastest absorbed, with a Cmax and AUC 4−6 times higher than those of 33 and 37 (Supporting Information, Table S4). Compound 27 achieves an excellent selectivity and inhibits only four kinases at more than 50% out of 104 kinases when tested at a concentration of 10 μM (Supporting Information, Table S5). Overall, 27, combines excellent physicochemicalproperties (Table 3) with good in vitro and in vivo ADME properties and excellent PK properties and kinase selectivity profile. Thus, compound 27 was selected as the lead compound for the lead optimization phase.Multiparameter Lead Optimization. Increasing Po- tency: Swapping Cores. Alteration at C-3 and C-6 of the imidazopyrazine core produced 27 with an inhibitory potency of 0.5 μM against MNK1 and MNK2, in both enzymatic and cellular assays. At this stage, the core is the only part of the scaffold that was not modified, and we thought that its alterations could modulate potency and/or ADME properties as done by others.60−62Accordingly, in addition to imidazopyrazine, six bicyclic ringswere investigated (Table 5 and Supporting Information, Table S6). Modeling studies have shown that their binding mode to MNK1/2 remain similar to that of imidazopyrazine analogues;all derived analogues have in common the N-1 binding to hinge Ser166 and the CN hydrogen bonding to Lys223, and exhibited previously described hydrophobic interactions for phenyl groups at both C-3 and C-6.Briefly, in comparison to the imidazopyrazine 27, [1,2- a]pyrimidine and 1H-imidazo[4,5-c]pyridine analogue (Sup- porting Information, Table S6) were less potent, and 1H- benzo[d]imidazole derivative (Supporting Information, Table S6) was equipotent to 27. In contrast, the pyrazolo[1,5- a]pyrimidine analogue was 10-fold more potent than 27 (Supporting Information, Table S6), and compounds from pyrazolo[1,5-a]pyrimidine series showed SAR, physicochem- ical, and PK properties similar to those of the imidazopyr- idazine analogues.

Therefore, we focused on the imidazopyr- azine, imidazopyridazine, and imidazopyridine cores, and derived analogues as shown in Table 5.Similarly, imidazopyridazine analogues 39, 40, and 42 are more potent than imidazopyrazine analogues 27 and 33. This trend is also observed when the imidazopyrazine core is replaced by an imidazopyridine with an improvement in potency, of 6- to 9-fold for MNK1 and of 5- to 7-fold for MNK2, seen when comparing imidazopyrazines 27 and 33 to imidazopyridines 47 and 48.In HeLa cells, inhibitory data of eIF4E phosphorylation shown in Table 5 are clustered between 0.22 and 0.44 μM. This inhibition is often weaker by 3- to 47-fold in whole-cell as compared to cell-free assays, probably due to permeability issues (39, 40, 42), nonspecific binding or a combination of both. Highly permeable compounds (27, 47, 48) exhibit a smaller potency difference between enzymatic and cellular data, and 27, the most permeable compound, is equally potent in both assays. Compound 45 is 47-fold less potent in cellular assays despite a good PAMPA permeability. However, 45 has a high Caco-2 effiux ratio of 13 (Table 4) that could explain the loss of potency.To rationalize the impact of N-7 on potency, we examined the structural models which revealed that the imidazopyrazine N-7 (Figures 2 and 3) is not involved in any hydrogen bond in any of the docked complexes studied. On the contrary, it resides in a hydrophobic environment and has van der Waals interactions with Gly165 α carbon and Leu90 side chain, as visualized in Figures 2 and 3. This may account for part of the approximately 5- to 10-fold gain in potency observed when this N-7 nitrogen is replaced by a CH (compare 33 to 48).

Indeed, our previous modeling studies63 have also shown the important contribution of hydrophobic interactions to the potency of the compounds.It may be surprising that compounds with an imidazopyr- idazine scaffold were also substantially more potent than the imidazopyrazine analogues as both cores have two nitrogen atoms with a lone pair that can be solvated. The N-5 of the imidazopyridazine is also surrounded by hydrophobic residues (Leu90 and Leu212). We believe that the difference in exposure of the lone pairs plays a determining role in the influence of the core on potency. In the imidazopyrazines the lone pair is shielded by the C-6 phenyl where as in the imidazopyridazines the lone pair is shielded by both the C-3and C-6 phenyls. This impacted the interaction between the ligand and protein as well as the solvation energy.DFT calculations indicate that the desolvation penalty for the imidazopyrazine N-7 may be much higher than that of the imidazopyridazine N-5, and it costs 10.8 kJ/mol more to remove the water from the imidazopyrazine complex in Figure 4.Figure 4. A single water molecule was placed in front of the imidazopyrazine N-7 (A) and imidazopyridazine N-5 (B). The complexes were minimized using DFT-B3LYP 6-31G**.In contrast to the benzamide’s minor influence on binding in imidazopyrazine series, the 4-N,N-dimethylamino piperidine 45 in the imidazopyridazine series is substantially more potent than piperazine and morpholine benzamide analogues 39 and 40, respectively.The piperidine-dimethylamine (45) in Figure 5 is sandwiched between hydrophobic residues Gly91, Leu168, and Phe227, while the base may interact with the acids Glu92 and Glu209. Figure 5 shows the interactions of the dimethylamino piperidine group 45 with Glu209 and Phe227 of MNK2. In the MD simulations, the basic nitrogen interacts with the side chain of Glu209 during most of the simulation.

However, when bound to MNK2, this part of the inhibitor is solvent exposed; solvent effects are likely to shield the contributions of any electrostatic interactions, and they may not contribute much to the binding affinity.We speculate that the N,N-dimethylamino piperidine interacts better with Phe227 and Glu209 than the N- methylpiperazine. During the MD simulations (Supporting Information, MD file compound 45), the interactions withthese residues were stable for 45 while they were only observed during 50% of the simulations for 20.ADME and PK. Modulation of ADME Properties: Correcting Issues Related to N-Alkylated Benzamides. A selection of compounds, shown in Table 4, were evaluated in vitro to assess their metabolic stability, Cyp inhibition, and protein binding properties. It appears that substituting the imidazopyrazine core effectively led to more potent derivatives 39 and 47 (Table 5), but with an increased metabolic liability (Table 4). Methylpiperazine 39 and 47 are unstable in human and mouse liver microsomes, and this is most likely attributable to N-demethylation. To address that issue morpholine 40, piperazine 42, and N,N-dimethylamino piperidine 45 deriva- tives were generated and were found to be metabolically stable. Unfortunately, both piperazine 42 and dimethylamino piperidine 45 derivatives were subject to a high effiux ratio (Table 4) which might negatively impact their PK properties. Morpholine derivatives 40 and 48 showed low metabolic liability in mouse, rat, dog, and human microsomes, and low Cyp 3A4 and 2D6 inhibition, and they exhibited a higher cell permeability as per the Caco-2 assay (Table 4) compared to the basic piperazine and piperidine analogues.Pharmacokinetic Properties.

In spite of the high effiux ratio of 45, it was orally administered to mice at 5 mg/kg and gave a poor systemic exposure with a Cmax of 69 ng/mL and an AUC0‑inf of 347 ng·h/mL. In contrast, 40 and 48 exhibited higher exposure and substantially improved PK properties relative to the lead 27 when comparing AUC and Cmax values (Table 6). Indeed, compared to 27, compounds 40 and 48 exhibited several characteristics favorable to good exposure, including lower basicity, higher Caco-2 permeability, and a lower Caco-2 effiux ratio.Kinase Selectivity. Compounds 40 and 48 were tested against a panel of 104 kinases together with representative MNK inhibitors such as CGP57380,36−38 cercosporamide,39 and AST-487.52,64 These known MNK inhibitors are often used in biological studies but their selectivity profiles were not readily available. CGP57380 is a multikinase inhibitor65 as it was found to inhibit ABL1 (E255 K), ABL1 (WT), MEK1, MEK2, MNK1, MNK2, and RET at more than 99% (Supporting Information, Table S5).Table 7 shows that 40 and 48 are selective, and specifically, imidazopyridine 48 inhibits only two kinases at ≥65%, sixkinases fewer than imidazopyridazine 40. The bicyclic core of 48 had an unexpected influence on its kinase selectivity, and the difference in the torsion between the bicyclic ring and the phenyl at C-3 could be a plausible explanation of why 48 is more selective than 40. There is a 29° versus 42° torsion variation in DFT-B3LYP 6-31** minimized structures. The more coplanar 40 probably has a shape that allows it to fit better to a greater number of kinases.Cell-Based Anti-proliferative Activity. The anti-proliferative effects of 40 and 48 were assessed in vitro, using CellTiter-Glo viability assay against 25 hematological cancer cell lines including the K562 cell line that overexpresses eIF4E (K562 o/e eIF4E)28 (Supporting Information, Table S7). The IC50’s were in general in the micromolar range similar to published data.

Compound 40 appears to be more potent than 48 with submicromolar IC50, especially against GK-5, DOHH2, AHH-1, and P3HR-1 cell lines. We speculate that the superior activity of 40 against these cell lines could be the result of the combination of its potency and off-target inhibition of kinasessuch as PDGFRA and FLT3, against which 40 gave an inhibitory potency of 0.25 and 0.88 μM, respectively. MNK1/2 inhibitors that show strong anti-proliferative effects in vitro have been known to often inhibit other kinases.34,39In contrast, 40 and 48 alone or in combination with imatinib or dasatinib were both able to suppress the ability of leukemic stem cells from BC-CML patient to form colony (data to be published elsewhere). This was previously described for nonselective MNK1/2 inhibitors AST-487, CGP57380, and cercosporamide, and it was shown that imatinib and dasatinib alone do not inhibit the self-renewal activity of LSCs.28Selection of Compound 48 as a Preclinical Candidate. A comparison of the two optimized compounds, 40 and 48, shows that 40 is the most potent in the enzymatic assay, but that potency did not carry over to eIF4E phosphorylation inhibition in HeLa cells, where 48 is slightly more potent. Compound 48 has a greater solubility, permeability and an excellent kinase selectivity. In maximum tolerated dose studies, oral administration of 48, showed no weight loss and no clinical adverse effects in healthy mice up to the tested dose of 200 mg/ kg.To avoid any ambiguities in interpreting efficacy studies, a compound which inhibits fewer kinases will be preferred for investigating the impact of the inhibition of MNK1/2 on biological processes.

In light of these data, compound 48 was ranked higher and was selected as the preclinical candidate.aKinase inhibition data are from a competition assay with a proprietary ligand to determine the compound’s binding affinity against a panel of 104 kinases.In Vivo Efficacy. We have reported that dual specific MNK1/2 and BCR-ABL1 inhibitors potently inhibited the proliferation of five BC-CML lines in vitro, and these inhibitors elicited a dose-dependent anti-tumor activity in an in vivo mouse xenograft model.52 As MNK1/2 and BCR-ABL1 inhibitory activities were combined into one molecule, it was challenging to distinguish the contribution of MNK1/2 inhibition in the observed biological response. Now, we have shown that selective MNK1/2 inhibitors 40 and 48 are much less anti-proliferative against the five BC-CML cell lines (Supporting Information, Table S7). However, there areemerging data67−69 that support that MNK1/2 inhibition effectin vivo could be greatly enhanced by the tumor microenviron-ment or the stroma without which, the amplitude of MNK1/2 inhibitors anti-proliferative effect on BC-CML cell lines in in vitro cellular assays could appear to be much weaker. Accordingly, compound 48 was selected for further evaluation in vivo in a mouse xenograft model. The antitumor effect of compound 48 was then assessed in a K562 e/o eIF4E mouse xenograft model after oral administration at 25, 50, or 100 mg/kg alone or in combination with a 2.5 mg/kg fixed dose of dasatinib throughout the study (Figure 6).Figure 6A shows that the treatment was well tolerated, and the weight loss observed with the vehicle or compound 48 administered alone is due to the burden of the tumor growth on mice.Dasatinib at 2.5 mg/kg elicited a tumor growth inhibition (TGI) of 88% (Figure 6B) with one tumor-free animal. In contrast, 48 alone only yields a maximum TGI of 23% at the highest administered dose of 100 mg/kg, which did not impede tumor growth, and was similar to the nontreated animals.

Combining 48 with 2.5 mg/kg of dasatinib not only increased tumor growth inhibition in a dose-dependent manner but, more importantly led to 2, 5, and 8 out of 8 tumor-free animalsat 25, 50, and 100 mg/kg, respectively (Supporting Information, Table S8). The combination of 48 and dasatinib inhibited tumor growth at all tested doses, and no weight loss was recorded (Figure 6A). Both the combination of 48 and dasatinib and, on the other hand, the dual MNK1/2 and BCR- ABL1 inhibitors52 prevented tumor growth in the same mouse xenograft model. The combination study clearly showed, that aselective MNK1/2 inhibitor plays a role in TGI by enhancing the antitumor effect of dasatinib.Biomarker Study. Western blot analysis of tumor samples excised from mice administered with 2.5 mg/kg of dasatinib and 100 mg/kg of 48 (Supporting Information Figure S1) revealed a maximum reduction in phosphorylated CRKL, a BCR-ABL1 substrate, from 4 to 8 h post dose. The reduction of phosphorylated eIF4e was significant after 1 h post dose and was sustained at 4 and 8 h post dose. These observations are consistent with the tumor growth inhibition induced by the combination of 48 and dasatinib and shows that both MNK1/2 and BCR-ABL1 kinases are inhibited in vivo.

CONCLUSION
Starting from bicyclic fragment hits, selective and potent MNK1/2 inhibitors were designed and synthesized. Substitu- ents at both C-3 and C-6 and the bicyclic moiety were optimized. The imidazopyridazine and imidazopyridine de- rivatives, which required a nonbasic benzamide for optimal PK properties, also showed that these cores were equally effective at increasing potency for MNK1/2 inhibition and at modulating kinase selectivity profiles. Imidazopyridine deriva- tive 48 displayed excellent kinase selectivity and significantly enhanced the anti-tumor activity of dasatinib in a BC-CML mouse xenograft model. This is in agreement with our previous anti-tumor data obtained with single compound dual MNK/ BCR-ABL1 inhibitors, for which the impact of MNK inhibition could not be clearly established. Biomarker studies show that the combination effectively reduced the ability of both MNK1/ 2 and BCR-ABL1 kinases to phosphorylate their substrates eIF4E and CRKL, respectively. Treatment of patients resistant to dasatinib with a combination of MNK1/2 inhibitor and dasatinib should overcome the MNK-eIF4E axis-mediated resistance and restore sensitivity to dasatanib treatment. In addition, the combination might require a lower dose of dasatinib to eFT-508 achieve efficacy, alleviating dasatinib toxicity. This combination therapy with MNK1/2 inhibitors could be extended not only to other BCR-ABL1 inhibitors current or future, but also to other therapies using kinase inhibitors that have become ineffective due to an overexpression of phosphorylated eIF4E.