Synthesis and Kinase Inhibitory Potencies of Pyrazolo[3,4-g]isoquinolines

A new series of pyrazolo[3,4-g]isoquinoline derivatives, diversely substituted at the 4- or 8-position, were synthesized. The results of the kinase inhibitory potency study demonstrated that the introduction of a bromine atom at the 8-position was detrimental to Haspin inhibition, while the introduction of an alkyl group at the 4-position led to a modification of the kinase inhibition profiles. Altogether, the results obtained demonstrated that new pyrazolo[3,4-g]isoquinolines represent a novel family of kinase inhibitors with various selectivity profiles.


Introduction
Protein kinases are implicated in cellular signaling pathways involved in various pathologies such as cancer, neurodegenerative disorders or pain [1]. Therefore, these transferases, able to modulate protein targets by transferring a γ-ATP phosphate group to Ser/Thr and/or Tyr residues, are key targets for identifying new therapeutic strategies. As part of our ongoing studies aiming at identifying new heteroaromatic series with kinase inhibitory potential, we recently described pyridoquinazolines and a pyridoquinoline which were active toward Haspin (haploid germ cell-specific nuclear protein kinase) [2] ( Figure 1). Best compounds exhibited nanomolar potencies against Haspin. Haspin is an atypical serine/threonine kinase involved in the phosphorylation of Thr3 (threonine 3) of Histone H3 in mitotic cells. Due to its essential role in mitosis, Haspin appeared as an interesting target for cancer therapy [3]. However, as frequently reported (e.g., [4,5]), Haspin inhibitors cross-inhibited other protein kinases such as CLK1 (cdc2-like kinase 1), DYRK1A (dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A) or CDK9 (cyclin-dependent kinase 9). Fortunately, we identified lead compound I exhibiting a good activity toward Haspin (IC 50 Haspin = 50 nM) and selectivity (only 12% of kinase inhibited when tested on a large panel of 486 kinases) ( Figure 1). Moreover, we described the molecular interactions established between thio analog II and the Haspin ATP binding pocket, showing that inhibitor stabilization mainly involved hydrophobic interactions while a hydrogen bond was established between the pyridine ring and the kinase hinge region residues (PDB code 7OPS).
In this work, to enlarge our structure-activity relationship study and get an insight into the impact of the nature and size of ring A on the kinase inhibitory potency of the synthesized compounds, the 6-membered ring was replaced by a pyrazole nucleus (Figure 1).

Synthesis and Biological Activity of Pyrazolo[3,4-g]isoquinolines
In order to modify the size of the upper heterocycle, based on the synthesis of the previous series, we studied the reactivity of intermediate A [6,7] with hydrazine or methyl/ethylhydrazinium salts (oxalate or sulfate) (Scheme 1). In these conditions, pyrazoloisoquinolines 1a-1c were obtained in 62-70% yields. The regiochemistry of the pyrazole ring formation in the presence of alkylhydrazines was attested by 2D NMR experiments. As mentioned below in the biological evaluation part, the very low solubility of compound 1a led us to prepare the corresponding methanesulfonate salt 1a'. Amino counterparts 2a-2c were prepared in acceptable to good yields by catalytic hydrogenation of 1a-1c (Scheme 1).
The inhibitory potencies of new compounds 1b-1c and 2a-2c were studied toward a panel of eight protein kinases (Haspin, CLK1, DYRK1A, CDK9/Cyclin T, GSK-3β, CK1ε, CDK5/p25 and Pim1) ( Table 1). It should be noted that due to solubility issues, compound 1a could not be tested; only the corresponding methane sulfonate salt 1a' was evaluated. The percentage of residual kinase activity was evaluated at 10 µM and 1 µM compound concentrations. Haspin IC 50 values were determined for compounds with Haspin residual kinase activity <50% at 1 µM compound concentration. In order to assess the selectivity profile of the best inhibitors, IC 50 values were also measured for other kinases inhibited ≥50% at 1 µM. DYRK1A, which was more inhibited than CLK1. For 1a', SI was estimated to be >6.5 in regard to Haspin IC 50 value, and CLK1/DYRK1A IC 50 values were assumed to be >1 µM on the basis of % of inhibition at 1 µM (less than 50% for CLK1/DYRK1A). Assays were performed in duplicate using the ADP-Glo assay in the presence of 10 µM ATP. Typically, the standard deviation of single data points was <10%. Compounds I and II IC 50 values from reference [2]. Our results indicated that for nitro analogs, the most inhibited kinases were Haspin, CLK1, DYRK1A and CDK9. The most potent Haspin inhibitors were 1b and 1c with IC 50 values of 57 nM and 66 nM and selectivity index (SI) of 1.2 and 2.5, respectively. Best SI in favor of Haspin was observed for less potent inhibitor 1a', with Haspin being the sole kinase inhibited to more than 50% at 1 µM compound concentration. Concerning amino analogs, 2c was the best Haspin inhibitor with an IC 50 value of 62 nM and a selectivity index of 4 in favor of Haspin versus DYRK1A. Compound 2b, exhibiting a similar inhibitory profile, was slightly less active toward Haspin. Finally, compound 2a demonstrated the worst selectivity profile, with six out of eight kinases inhibited to more than 50% at 1 µM. Based on these results, due to better overall selectivity observed in the nitro series, we decided to study the impact of the introduction of various substituent at the 4-or 8-position of this tricyclic scaffold on their kinase inhibitory potency/selectivity. Nitro analog 1c with the best SI was selected as a starting point.

Cpds
Thus, compound 1c was reacted in the presence of Grignard reagents before aromatization by DDQ, leading to the corresponding diversely 4-substituted analogs 3a-3e. Compound 1c was also brominated using NBS in DMF to give 4 in order to study the impact of the substitution of the pyridine part via metallo-catalyzed coupling from a halogenated intermediate (Scheme 2) [8]. The position of the bromine atom was determined by 2D NMR experiments. However, given the low yield of this bromination reaction, this route was not further explored.
The kinase inhibitory potencies of these diversely substituted pyrazoloisoquinolines were studied toward the same panel of protein kinases as above. As indicated in Table 2, the introduction of a bromine atom at the 8-position was detrimental to Haspin inhibition, while the introduction of an alkyl group at the 4-position led to different kinase inhibition profiles. Actually, brominated analog 4 only inhibited Haspin by 23% at 1 µM. Compounds 3b (4-Et) and 3e (4-Bu) did not inhibit Haspin to more than 50% at 1 µM. Most potent Haspin inhibitor of this series was 3a bearing a methyl group (IC 50 = 167 nM); however, this analog was more potent towards CLK1 (IC 50 = 101 nM). Compounds 3c and 3d bearing a propyl or a cyclopropyl group were more active against CLK1/CDK9/GSK3 (IC 50 ranging from 218 to 363 nM) compared to Haspin. Altogether these results demonstrated that the alkylation of the 4-position led to a change in the kinase inhibition profile, with CLK1, CDK9 and GSK3 being preferentially inhibited over Haspin. Table 2. Percentage of kinases residual activity at 10 µM and 1 µM compound concentration. IC 50 values in nM (given in parentheses) were determined for Haspin when residual activity was <50% at 1 µM. For other kinases, IC 50 values were measured for most inhibited ones (Inhibition % ≥ 50% at 1 µM). Kinase inhibitory activities of 3a-3e and 4 were assayed in duplicate using the ADP-Glo assay in the presence of 10 µM ATP. Typically, the standard deviation of single data points was below 10%. In order to explain these different results, we undertook molecular modeling experiments to determine the putative binding mode of this new series within the Haspin ATP-binding pocket.

Molecular Modeling Experiments
Thus, the molecular interactions established between nanomolar inhibitor 1c, exhibiting a SI in favor of Haspin, as well as those of less potent 4-alkylated analogs 3a (SI in favor of CLK1) and 3e (only 22% of Haspin inhibition at 1 µM) were studied by docking. The docking experiments were performed using the Vina-1.2.1 hydrated method [9][10][11]. The Kinase ATP-binding site model was constructed from Protein Data Bank (PDB) 7OPS structure [2]. First of all, in order to validate the molecular modeling protocol used, docking was performed with a 7OPS ligand (II, Figure 1). This experiment concluded on the same binding mode as that determined by X-ray crystallography and therefore validated the procedure used.
As shown in Figure 2A, the tricyclic system of both series is mainly stabilized into the ATP-binding site by hydrophobic interactions. The pyridine moiety is oriented toward the hinge region and establishes an H-bond with GLY608 residue (Figure 2A,B). The main difference observed between compound II (with an upper pyrimidine ring), and this new series is the offset of the pyrazole ring position with respect to the VAL498 residue. Regarding 4-substituted derivatives (3a and 3e), the heteroaromatic scaffold adopted a similar pose ( Figure 2B). Alkyl groups (Me for 3a, Bu for 3e) are located in a highly hydrophobic pocket constituted by ILE490, ILE610 and PHE607 residues. However, these interactions failed to improve Haspin inhibitory potency compared to pyrazole derivatives not alkylated at the 4-position.

Conclusions
In summary, we synthesized a series of pyrazolo [3,4-g]isoquinoline derivatives diversely substituted at the 4-or 8-position. The results of the kinase inhibitory potency study demonstrated that better overall selectivity in favor of Haspin was observed in the nitro series for compounds 1b and 1c, with IC 50 values of 57 nM and 66 nM, respectively. On the other hand, the introduction of a bromine atom at the 8-position was detrimental to Haspin inhibition, while the introduction of an alkyl group at the 4-position led to different kinase inhibition profiles. Finally, docking experiments provided a putative binding mode for this new pyrazolo [3,4-g]isoquinoline series within the ATP-binding site. Altogether, the results obtained demonstrated that new pyrazolo [3,4-g]isoquinolines represent a novel family of kinase inhibitors with various selectivity profiles.

General
Starting materials were obtained from commercial suppliers and used without further purification. IR spectra were recorded on a Perkin-Elmer Spectrum 65 or Smart Orbit, Nicolet 5700 thermo electron FT-IR spectrometer (v in cm −1 ). NMR spectra, performed on a Bruker AVANCE 400 III HD ( 1 H: 400 MHz, 13 C: 101 MHz) or a Bruker AVANCE III HD 500 ( 1 H: 500 MHz, 13 C: 126 MHz), are reported in ppm using the solvent residual peak as an internal standard; the following abbreviations are used: singlet (s), doublet (d), triplet (t), quadruplet (q), quintuplet (quint), hexuplet (hex), heptuplet (hept), multiplet (m) and broad signal (br s). Coupling constants are expressed in Hertz. High resolution mass spectra were determined on a high-resolution Waters Micro Q-Tof or Thermo Scientific Q Exactive Q-Orbitrap apparatus (UCA START, Université Clermont Auvergne, Clermont-Ferrand, France). Chromatographic purifications were performed by column chromatography using 40-63 µm silica gel or by preparative TLC using silica gel-coated glass plates 60 F254 from Macherey Nagel. Reactions were monitored by TLC using fluorescent silica gel plates (60 F254 from Macherey Nagel). Melting points were measured on a Stuart SMP3 apparatus and were uncorrected.
The purity of all tested compounds was established by HPLC analysis using either a VWR Hitachi chromatograph (for 1a-1c, 1a', 2a-2c, 3e, 4) or an Agilent 1100 series G1315A (for 3a-3d) with DAD detector. A Macherey Nagel Nucleodur gravity column (4.6 mm × 250 mm, 5 µM) was used for all compounds. The flow rate was 0.5 mL/min, and the analysis was performed at 25 • C at 240 or 270 nm as the detection wavelength for each compound. Solvents were (A) water/0.1% formic acid, (B) Acetonitrile. Two methods were designed: method A was a gradient of 5:95 A/B for 5 min to 95:5 A/B in 25 min, whereas method B was an isocratic mode using 60/40 water/acetonitrile.
Magnesium turnings (100 mg, 4.1 mmol, 1 eq) in 2 mL of anhydrous THF were activated by adding a small amount of iodine. Then the suitable bromide (4.1 mmol, 1 eq) was added dropwise at 0 • C. Once the addition finished, the mixture was heated to reflux for 45 min. The concentration of the obtained organomagnesium solution was determined as follows: to a solution of iodine (100 mg) in anhydrous THF was added dropwise the organomagnesium solution until the complete disappearance of iodine color.
The reaction of organomagnesium/organolithium reagents with compound 1c: To a solution of compound 1c in anhydrous THF (0.05 mmol/mL) was added dropwise a solution of the corresponding organomagnesium derivative. The mixture was stirred at a suitable temperature under an argon atmosphere until all the starting material was consumed (TLC control, from 3 to 18 h). The mixture was then treated with a saturated aqueous NH 4 Cl solution, made alkaline with a saturated aqueous NaHCO 3 solution and extracted with EtOAc. Combined organic layers were dried over MgSO 4 and concentrated. The crude material was used in the next step without purification as follows: the residue was solubilized in a mixture of DCM/MeOH (3/2, 0.01 mmol/mL), then DDQ (2,3-dichloro-5,6-dicyano-p-benzoquinone) (1 eq) was added. The solution was stirred at room temperature until the TLC control showed that the additional product was totally oxidized (1 h). The reaction mixture was then concentrated under reduced pressure. The residue was purified by flash chromatography or/and preparative TLC to give the corresponding 5-substituted derivative.