Design and biological evaluation of substituted 5,7-dihydro-6H-indolo[2,3-c]quinolin-6-one as novel selective Haspin inhibitors

Abstract A library of substituted indolo[2,3-c]quinolone-6-ones was developed as simplified Lamellarin isosters. Synthesis was achieved from indole after a four-step pathway sequence involving iodination, a Suzuki-Miyaura cross-coupling reaction, and a reduction/lactamization sequence. The inhibitory activity of the 22 novel derivatives was assessed on Haspin kinase. Two of them possessed an IC50 of 1 and 2 nM with selectivity towards a panel of 10 other kinases including the parent kinases DYRK1A and CLK1. The most selective compound exerted additionally a very interesting cell effect on the osteosarcoma U-2 OS cell line.


Introduction
Marine products currently represent an underutilised source of leads for the pharmaceutical industry 1 . Besides their original and complex structures, they often offer new action modes and structural originality. Nevertheless, their low abundance and the presence of few analogues makes it difficult to obtain large libraries, perform full biological characterisation and achieve structure-activity relationship (SAR) exploration. Despite these difficulties, these products remain attractive due to their high valorisation potential, and their complex structures have prompted medicinal chemists to use disruptive strategies to intuitively isolate the pharmacophore elements that trigger biological activity.
For these reasons, some marine products and their synthetic analogues have emerged in drug discovery strategies and several of them have been reported for protein kinase inhibition 2 . Among them the most successful example is the polycyclic staurosporine I 3,4 . This lead compound has led from extraction, hemisynthesis and organic synthesis efforts to Lestaurtinib II 5,6 and simplified Enzastaurin III 7 , two potent drugs targeting VEGF receptors and kinases, which have entered clinical trials against leukaemia and cancer ( Figure 1). In this field, the chemical simplification of indolocarbazole scaffolds and caulersin IV have generated strong kinase inhibitors and cytotoxic agents [8][9][10][11][12][13][14] .
Among these bis-indole series, lamellarins, a group of pyrrole alkaloids, have emerged ( Figure 2) 15,16 . These compounds are a class of marine-derived natural products isolated from molluscs, ascidians and marine sponges. Nearly, 70 natural derivatives have been reported in this family which mainly contains a fused pentacyclic pyrroloisoquinoline lactone ring system 17 . Lamellarins have focussed the attention of medicinal chemists due to their diverse biological effects. Some have demonstrated cytotoxic activities and multidrug resistance (MDR) reversal in a number of cancer cell lines, as well as being confirmed inhibitors of topoisomerase I. Moreover, Lamellarins D (structure V, Figure 2), N and L have proved their ability to inhibit kinases such as GSK3b, DYRK1A and CDK5 in the nanomolar range 18 . Our first approach in this chemical series was based on the simplification of the synthetic model in order to discriminate the two activities, that is, diminish the topoisomerase I inhibition while retaining the kinase inhibition by fine-tuning the chemical structure. This objective was reached by replacing the pyrrole moiety with an indole skeleton and designing new chromeno [3,4b]indoles VI which revealed DYRK1A inhibition. In this structure, the rings of Lamellarin-D noted A, B and C were unchanged ( Figure 2). Despite structural modulations, the kinase inhibition remained mainly in the sub-micromolar range except for derivatives 1a and 1b, which exhibited a high selective inhibition of DYRK1A but revealed instability in basic media due to the lactone. We therefore decided to develop a more robust indoloquinolinone [19][20][21][22][23] series VII with the objective of creating a novel and druggable family of kinase inhibitors, as we envisioned that the presence of lactam combined with the NH of pyrrole would reinforce the hydrogen bond donor acceptor binding mode to the ATP active site.
Herein, we present access to the indolo [2,3-c]quinolone-6-one library and the evaluation of the family on a representative panel of kinases involved in CNS, inflammatory diseases and oncology. SAR are depicted and we demonstrate that in addition to retaining DYRK1A inhibition to a large extent, other molecules acting on CLK1 have also been designed. We additionally found that several compounds inhibit the Haspin kinase with an unprecedented selectivity. Molecular docking experiments were conducted to explain these results. Finally, screening on a cancer cell line was carried out and results showed that the compounds induced cellular effects and affect osteosarcoma cell lines in particular.

Chemistry
In order to introduce ring A on the indole, a Suzuki-Miyaura reaction appeared to be the most appropriate. The 3-iodoindole-2-carboxylic ethyl esters 10-13 were therefore prepared after 2 efficient steps. The first one consisted in the quantitative esterification of 2-4 using 10 mol% of sulphuric acid catalyst in ethanol, followed by iodination in presence of KOH to afford the attempted ethyl 3-iodo-1H-indolo-2carboxylates 24 . Noteworthy, the N-methylation of indole 11 with a    slight excess of iodomethane in presence of sodium hydride gave 14 in a near quantitative yield (Scheme 1). Next, we focussed on preparing the second partner for the cross coupling reaction, that is, the unavailable boronylated nitrobenzene derivatives (Scheme 2). When the starting 2-halogeno nitrobenzenes do not contain any acidic proton, the use of Grignard reagent is recommended in the literature 25 . The reaction was first carried out with phenyl magnesium chloride (1.2 equiv.) in presence of methylorthoborate as electrophile whereas a final acidic hydrolysis led to the desired 2-nitroaryl boronic acids 15-20 in fair good yields. When 2-halogeno nitrobenzenes bear an acidic proton, the Miyaura borylation reaction can be used with conditions involving PdCl 2 (dppf) as catalyst and potassium acetate as a base in dioxane. This method furnished pinacol boronic esters 21-24 with modest yields after 14 h of reaction at 80 C. Due to their sensitivity during the purification step, the boron derivatives were used in the cross coupling reaction as crude materials.
Assembly of the two building blocks was next performed using the Suzuki-Miyaura reaction in presence of PdCl 2 (dppf) and potassium carbonate as base in a refluxing mixture of water and dioxane. These conditions proved fully suitable to achieve all the envisioned cross couplings and the desired derivatives were isolated with yields ranging between 40 and 60% after 12 h of reaction (Table 1). Finally, we thought that the large library of final compounds 45-64 could be obtained by a "one pot" strategy (Table 2). To this end, the nitro derivatives 25-44 were also treated with iron powder in refluxing acetic acid and the in situ formed amine concomitantly formed the lactam ring by annelation with the nearby ethyl ester. An original library of substituted 5,7-dihydro-6H-indolo[2,3-c]quinolin-6-ones 45-65 was obtained with very high yields.

Kinase assays, SAR
We previously showed that analogues of Lamellarin D, chromeno [3,4-b]indoles, had the ability to inhibit the DYRK1A kinase 15 and it has been shown by different groups that DYRK inhibitors often cross-react with CLKs and with the mitotic kinase Haspin 26,27 . We therefore tested the inhibitory activity of the 22 synthesised derivatives on the HsHaspin, MmCLK1 and RnDYRK1A recombinant kinases (Table 3).
The selectivity of each derivative was also determined on a representative kinase panel including HsCdk2/Cyc A, HsCdk5/p25, HsCdk9/Cyc E, HsGSK3b and HsPIM1. Interestingly, more than half of the analogues displayed very high activity towards Haspin with a percentage of residual activity at 1 lM close to zero (ranging from 6% for compound 46 to 0% for compounds 61, 54, 55, and 62). Apart from 5 compounds (45, 58, 47, 53, 63, and 64), 17 novel indoloquinolinones showed an interesting selectivity towards Haspin, CLK1 and DYRK1A kinases with moderate activity on CDKs, GSK-3b and PIM1.
IC 50 for Haspin, CLK1 and DYRK1A were next calculated for 14 compounds showing residual kinase activity 25% on Haspin kinase at a concentration of 1 lM (Table 4). It clearly appeared that the lactam moiety favoured DYRK1A inhibition. While the hydroxylated derivatives 1a and 1b displayed good activity on DYRK1A, they are nevertheless the only molecules in their family to present this action whereas the 14 lactams reported in this study exhibited an inhibition below 300 nM. It is possible that due to the electro-donating lactam nitrogen atom (vs. the O of the lactone), the electron density on the carbonyl group is sufficiently modified to strengthen a favourable hydrogen bond in the active site (see molecular modelling studies). Moreover, the fcompounds 48, 49, and 51 (entries 2, 3, 5) appear to be the best DYRK1A inhibitors of the series with more potent IC 50 than the lactones 1a and 1b (entry 1).
The concomitant action of DYRK1A inhibitors with CLK1 was confirmed since most of the Haspin inhibitors were also active on DYRK1A and CLK1. The mode of interaction of 49, which is highly active on Haspin and exhibits strong activity on the other two kinases, was studied by molecular docking experiments (see Molecular docking studies section). Considering the two enzymes  we can say that compounds 48 and 49 inhibit DYRK1A and CLK1 almost equally in the nanomolar range. As regards to Haspin inhibition, the chemical series of type VII is of great interest. Eight compounds showed an IC 50 below 10 nM. Compound 45 (Table 3, entry 1), without any substituent on the aromatic parts, presented no kinase activity. To maximise the inhibition, at least one methoxy OCH 3 or hydroxy OH group is required in positions C-2 or C-3 (compounds 46, 48, 50, 51, 54, 55, 61), the amino derivatives or amides being less effective ( Table 4, entries 9 and 10).
Regarding the indole ring substituents, the absence of a substituent or the presence of an OCH 3 or a fluorine in C-10 position did not significantly affect the Haspin inhibition (products 46, 51 and 61, entries 1, 5, 13) when a functional group was also present on the C-2 or C-3 positions of the phenyl ring. Finally, the presence of ether in position C-9 (entry 11 vs. 5) or the methylation of indole (entry 14 vs. 5) made the compounds less effective.
The inhibitory activity of compound 49 on Haspin was about 12 times stronger than that of the other two enzymes. At 20 nM this molecule inhibited the 3 enzymes quantitatively. Compound 55 showed an even higher selectivity since the selectivity for CLK1 was identical, but against DYRK1A it rose to a factor of 65. At this stage, we can almost say that at a dose of 20 nM, this molecule shows a dual inhibition of Haspin and CLK1.
We further evaluated the selectivity profile of compound 55 on a larger panel of kinases (SelectScreen Whole Panel, Life Technologies). The inhibition profile of 55, evaluated at 1 lM, is depicted on a dendrogram on Figure 3 where kinases inhibited by a minimum of 80% are shown. A full list of the kinases tested is shown on Table S1 in supplementary information.
The inhibitory activities of compound 55 on Haspin, CLKs and DYRKs are well found in this new screening study. Product 55 is not specific but appears to be relatively selective since the compound 55 (AS-N14) presents a reasonable selectivity profile against a panel of 486 tested kinases since it inhibits 73 of the 486 kinases by >80%, and 159 of the 486 kinases by >50% at a dose of 1 mM.

Molecular docking studies
Molecular docking studies were carried out using Glide 28-30 from the Schr€ odinger Suite 31 , in order to compare putative binding  modes and explore interactions within the active sites of CLK1, DYRK1A and Haspin kinases. Active sites of the three crystal structures were superimposed and are shown in Figure 4. The residues engaging hydrogen bond interactions with docked ligands are highlighted in stick form.
One of the most active compounds, 49 was docked in each active site of the three kinases. The docking poses exhibited the same putative binding mode, highlighting a hydrogen bond between the acceptor atom O of the lactam ring of 49 and the backbone of the hinge Leu244 (CLK1), Leu241 (DYRK1A) and Gly608 (Haspin). In addition, 49 formed another interaction with the hinge region of Haspin through a hydrogen bond between NH of the lactam ring and the backbone of Gly609 ( Figure 4). In some docking poses, the molecule was flipped by 180 exposing the methoxy group of 49 towards the solvent area.
Derivatives 51 and 62 were next docked in order to investigate the studied binding mode in greater depth since 62 has an N-methyl group on the pyrrole moiety. The best docking poses of the two compounds 51 and 62 were similar to 49 ( Figure 5). No steric clash between the protein and the second methoxy group of the ligand was observed. Interestingly, 62 compared to 51 showed a weak H-pi interaction between the methyl group of the indole moiety and the gatekeeper, Phe605, of the kinase. The presence of the methyl group in 62 did not impact the binding mode of the compound, which explains the acceptable IC 50 of 62.
From the docking experiments, we predicted that the binding mode in each protein kinase, Haspin, CLK1 and DYRK1A, would be very similar. Nevertheless, as in most docking experiments, the docking score is not sufficient to predict the small variation in activity of the compounds and further intensive computational approaches such as free energy of binding (FEB) would be needed.

Cell assays
We next analysed the effects of selected compounds (Haspin IC 50 <15 nM) on the cell viability of several cell lines from osteosarcoma (U-2 OS), colorectal cancer (HCT116), breast cancer (MDA-MB231) and neuroblastoma (SH-SY5Y) as well as retinal fibroblast RPE-1 immortalised with hTERT ( Figure 6). In a primary screen, all compounds were tested in triplicate at 25 lM and viability was expressed as percentage of a DMSO control. U-2 OS and HCT116 cell lines appeared to be the most sensitive to our compounds and were even slightly more affected than the non-cancerous  Table 3. They were calculated from a dose-response curve for which each point was measured in duplicate, and reported in nM. Selectivity indexes (SI) were calculated as follows: IC 50 DYRK1A or CLK1/IC 50 Haspin.    be explained by the low solubility of the compounds, or their low affinity for lipidic plasma membrane or a high metabolisation rate in aqueous solution/cellular environment. On the other hand, several compounds such as 48, 61, 55, and 50 showed a reduction of equal or more than 75% of cell viability compared to the DMSO control. Hence, dose-response experiments were carried out on both U-2 OS and RPE-1 cell lines and EC 50 were calculated ( Table 5). The results confirmed the lack of efficacy of compounds 46, 49, 54, and 56 which showed EC 50 >25 lM on both cell lines regardless of their activity on Haspin kinase (IC 50 of 7, 1, 4 and 14 nM respectively). Derivative 51, despite no effect observed on RPE-1 cells (EC 50 >25 lM), showed a moderate activity on cell viability of the U-2 OS line with an EC 50 of 12.6 lM. The compounds were generally more efficient at inhibiting the viability of U-2 OS cancerous cells than that of normal RPE-1. Amongst the selected compounds, 55 and 50 displayed the strongest effect on the viability of U-2 OS cells (EC 50 of 3.4 and 4.3 lM, respectively). They were between 2 and 5 times more active on U-2 OS compared to RPE-1 cell viability.
Taken together, these results showed that some of our compounds such as 55 and 50 displayed interesting effects on cell viability of several cancerous cell lines.
We further examined the effect of our most efficient compound (55) on cancerous cells growing in 3D spheroids. U-2 OS and HCT116 cells spheroids were prepared and treated with different concentrations of either compound 55, CHR-6494 or SBS018 (compound 21 in ref Elie et al) 26 or with 0.5% DMSO for 7 days, after which, spheroids viability was evaluated (Figure 7). We observed a marked dose-dependent effect of compound 55 on both U-2 OS and HCT116 spheroid cell viability after 7 days of treatment. This effect was milder than the one observed with CHR-6494 and stronger than the one induced by SBS018 at similar concentrations.
We further characterised the functional effects of our most efficient compound 55 on endogenous Haspin in U-2 OS cells by immunofluorescence, quantifying the Haspin specific H3T3ph signal in early mitotic cells. The H3T3ph signal was measured on cells treated with 0.5 mM of compound 55 or CHR-6494 or SBS018 or with 0.1% of DMSO for 16 h (Figure 8). Our results showed that compound 55 could inhibit intracellular Haspin with a very similar efficiency to that off CHR-6494 or SBS018. These results further validate the functionality of our compound in cells, on the endogenous Haspin kinase activity.
We then characterised the effect of compound 55 on the cell cycle on U-2 OS cells. Cells were treated for 24 h with 1 mM of compound 55, CHR-6494, SBS018 or DMSO at 0.2% and their cell cycle profile was analysed by flow cytometry (Figure 9). Analysis of flow cytometry profiles showed, as expected, a strong increase in the percentage of cells in G2/M phase of the cell cycle with compound 55 as well as with the two references CHR-6494 and SBS018 compared to the DMSO control (20,17, and 14%, respectively vs. 6% for the DMSO control). Concomitantly, compound 55 further induced a reduction of cells in the G1 phase compared to the control (

Conclusion
We have synthesised a series of new Lamellarin analogues using the indolo[2,3-c]quinolone-6-one core. The analogues were obtained after a sequence involving (i) a palladium catalysed cross coupling reaction between 2-indolic esters and 2-nitrophenyl boronic acids as building blocks, and (ii) a cyclic lactam formation involving a reduction and an annelation. Twenty-two novel derivatives were synthesised and evaluated for their inhibitory activity on Haspin kinase and on a panel of 7 other protein kinases for selectivity assessment. Among this series, 8 compounds inhibited Haspin kinase with IC 50 below 10 nM. Docking studies showed a double hydrogen bond between the lactam and the hinge region of the kinase. The most active compounds 49 and 55 possess IC 50 of 1 and 2 nM respectively with selectivity towards the parent kinases DYRK1A and CLK1 between a 13 and 65-fold factor. Furthermore, the most selective compound 55 exerted an interesting cellular effect on the osteosarcoma U-2 OS cell line as well as on U-2 OS and colorectal carcinoma HTC116 spheroid viability. Additionally, we further validated the functionality of compound 55 on endogenous Haspin activity in cells. This interesting Haspin inhibitor will be used in further studies to develop efficient and selective Haspin inhibitors.

Chemistry
All reagents and solvents were purchased from commercial sources and used without further purification. 1 H NMR and 13 C NMR spectra were recorded on 400 MHz and/or 500 MHz Bruker FT-NMR spectrometers. All chemical shifts are given as d values (ppm) with reference to tetramethylsilane (TMS ¼ 0) as an internal standard. The peak patterns are indicated as follows: s, singlet; d, doublet; t, triplet; m, multiplet; q, quartette. The coupling constants, J, are reported in Hertz (Hz). UV detection at 210 nm. High resolution mass (MS) analysis was conducted using an LC/MSD TOF spectrometer system with electrospray ionisation (ESI). Reactions were monitored via thin-layer chromatography (TLC) carried out on commercial silica gel plates (GF254) under UV light.
General procedure A: preparation of 3-iodoindoles (11-13) Indole-2-carboxylic acid derivatives (3.1 mmol) were dissolved in Ethanol (50 ml) and conc. sulphuric acid (5 ml) was added. The solution was refluxed for about 12 h (monitoring with TLC) after completion of the reaction. The solution was poured into cold water (150 ml), and the white solid precipitate formed was collected by filtration. The solid material was washed with water and dried to produce quantitatively derivatives 6-9. Crude materials were next dissolved in a mixture containing crushed KOH (4.0 equiv.) pellets in DMF (15 ml) at room temperature. Next Iodine (1.5 equiv.) dissolved in DMF (3 ml) was added dropwise and the mixture stirred for 4 h at room temperature (monitored with TLC). After completion (TLC monitoring), the reaction mixture was poured onto a saturated aqueous solution of NaHSO 3 (15 ml), NH 4 OH (30%, 2 ml) and water (15 ml). The solid was filtered, dried under reduced pressure, and used directly without further purification. All spectral data for 10, 11, 13 are in agreement with previous reports [20,21]. Compounds 11-13 were engaged immediately in the next step.

Ethyl 3-iodo-5-methoxy-1-methyl-1H-indole-2-carboxylate (14)
Compound 11 (1.0 g, 2.9 mmol) was added to a stirred suspension of oil-free sodium hydride NaH (0.1 g, 4.36 mmol) in DMF (10 ml) at 0 C and the mixture was stirred for 10 min at this temperature. Then methyl iodide (0.49 g, 3.48 mmol) was added at 0 C and the whole mixture was stirred at room temperature (20 C) for 1 h. The solution was poured into ice cold water (50 ml) and extracted with CH 2 Cl 2 (2 Â 25 ml   concentrated in vacuo. The crude product was finally purified by silica gel flash chromatography to afford the desired product 25-44.

Docking parameters
Docking grids were centred and sized on crystalised ligands. Docking calculations were performed with extra precision. Ligand flexibility was considered and the option of sampling of ring conformation was activated. A maximum of 100 poses were generated and a post-docking minimisation was performed.

Conclusion
We have synthesised a series of new Lamellarin analogues using the indolo[2,3-c]quinolone-6-one core. The analogues were obtained after a sequence involving (i) a palladium catalysed cross coupling reaction between 2-indolic esters and 2-nitrophenyl boronic acids as building blocks, and (ii) a cyclic lactam formation involving a reduction and an annelation. Twenty-two novel derivatives were synthesised and evaluated for their inhibitory activity on Haspin kinase and on a panel of 7 other protein kinases for selectivity assessment. Among this series, 8 compounds inhibited Haspin kinase with IC 50 below 10 nM. Docking studies showed a double hydrogen bond between the lactam and the hinge region of the kinase. The most active compounds 49 and 55 possess IC 50 of 1 and 2 nM respectively with selectivity towards the parent kinases DYRK1A and CLK1 between a 13 and 65-fold factor. Furthermore, the most selective compound 55 exerted an interesting cellular effect on the osteosarcoma U-2 OS cell line as well as on U-2 OS and colorectal carcinoma HTC116 spheroid viability. Additionally, we further validated the functionality of compound 55 on endogenous Haspin activity in cells. This interesting Haspin inhibitor will be used in further studies to develop efficient and selective Haspin inhibitors.