Imidazopyridine-fused [1,3]diazepinones: modulations of positions 2 to 4 and their impacts on the anti-melanoma activity

Abstract A series of 19 novel pyrido-imidazodiazepinones, with modulations of positions 2, 3 and 4 of the diazepine ring were synthesised and screened for their in vitro cytotoxic activities against two melanoma cell lines (A375 and MDA-MB-435) and for their potential toxicity against NIH-3T3 non-cancerous cells. Selected compounds were also evaluated on the NCI-60 cell line panel. The SAR study revealed that the molecular volume and the cLogP of compounds modified at position 2 were significantly correlated with the activity of these compounds on melanoma cell lines. Moreover, introduction of a heterocyclic group at position 2 or an azido-alkyl chain at position 4 led to compounds displaying a significantly different activity profile on the NCI-60 cell line panel, compared to phenyl-substituted compounds at position 2 of the diazepinone. This study provides us crucial information for the development of new derivatives active against melanoma.


Introduction
Melanoma is the most aggressive skin cancer and remains a therapeutic challenge. Its global incidence is increasing continuously, with 287,723 new cases in 2018, growing up to 301,694 in 2020 and 466,914 in 2040 1,2 . The prognosis of melanoma depends on how early the diagnosis is. While surgical resection of local forms is usually curative, metastatic melanoma (MM) requires systemic treatment and is associated with very poor prognosis. Indeed, the extreme chemoresistance of MM makes it less sensitive to conventional chemotherapy. Since 2011, new therapeutic strategies have emerged and offer promising results 3 . Small molecules, so-called "targeted" therapies, have been developed to specifically inhibit the V600E mutated BRAF kinase found in 50% of MM (Vemurafenib, Dabrafenib, Encorafenib) 4 . However, the rapid emergence of drug resistance to initially responsive cancers has led to only marginal patient benefit. To limit these resistances, mainly due to the reactivation of the MAPK pathway, MEK inhibitors (Trametinib, Cobimetinib, Binimetinib) are used in combination 5 . Nevertheless therapeutic escape are inevitably encountered 6 . In parallel, immunotherapies (inhibitors of CTLA4, PD1 and PDL1) have shown very interesting results, since some treated patients have not presented any relapse to date 7 . Unfortunately, no test can identify potentially responders and these therapies are not free of toxicity 8 . Finally, association between targeted therapies and immunotherapies confers a longer progression-free survival and duration of response, but led to a higher rate of serious toxicity 9 . Therefore, the development of new drugs that could be used in combination with mentioned therapies is still in demand to overcome the extreme chemoresistance of metastatic melanoma.
Recently, we reported the discovery of JMV5038, an original pyrido-imidazodiazepinone derivative, using a screening on the NCI-60 cancer cell line panel 10 . This compound showed a promising activity on several cancer cell lines, with a growth inhibition of 50% (GI 50 ) in the low micromolar range on melanoma cells and with no significant cytotoxic activity on NIH-3T3 fibroblasts, even at 100 mM. Further investigations revealed that JMV5038 increased apoptosis in subG1 phase and microscopic observations of treated cells showed morphological changes, associated with an increase of the membrane permeability and disruption of the actin network. We then started a structure-activity relationship (SAR) study on these series 11 . Whereas positions 3 and 5 of the diazepine ring could not be modified without loss of the cytotoxic activity, positions 2 and 4 of the diazepine cycle tolerated several modifications ( Figure 1). Concerning position 2, we showed that large hydrophobic groups, such as a 1-or a 2-naphthyl ring, as well as halogenophenyl groups, did not modify the biological activity of JMV5038. For position 4, several modifications were tolerated with preservation of JMV5038 cytotoxic activity, like short elongation of the alkyl chain or introduction of bulky groups such as a tert-butyloxymethyl substituent (Figure 1). To further explore the SAR on these series, we decided to pursue the modulations of position 2 by introducing different hydrophobic groups and, of the position 4, with the introduction of different alkyl chains. Among others, we decided to study the introduction of azido or alkyne groups on these two positions, such groups could be of interest to investigate, in the future, further modulations using click chemistry. For example, such approach could permit to introduce a biotin residue to identify the pharmacological target of our compounds, using "clickable" cell-permeable activity-based probe 12 .
In this paper, we report the synthesis of new analogues of JMV5038. Their cytotoxic activity on melanoma cell lines and their potential toxicity against NIH-3T3 non-cancerous cells were evaluated. Subsequently, a statistical study was carried out to highlight the physicochemical parameters influencing the activity profile of our molecules. Selected compounds were then evaluated by the NCI on the sixty cancer cell lines panel. Further in vitro functional evaluations were finally performed on the most relevant compounds.

Instruments and reagents
Commercially available reagents and solvents were used without further purification. Reactions were monitored by an analytical Waters Alliance 2690 HPLC instrument, equipped with a photodiode array and an analytical Chromolith Speed Rod RP-C18 185 Pm column (50 Â 4.6 mm, 5 mm), at a flow rate of 3.0 ml/min, and gradients of 100/0-0/100 eluents A/B for 5 min (eluent A ¼ H 2 O/0.1% TFA and B ¼ CH 3 CN/0.1% TFA). Detection was performed at 214 nm. Retention times (Tr) are reported in min. 1 H and 13 C NMR spectra were recorded on Bruker spectrometers (300 or 400 MHz) at room temperature in deuterated solvents. Chemical shifts (d) are expressed in parts per million (ppm), relative to the resonance of CDCl 3 ¼ 7.26 ppm for 1 H (77. 16 ppm for 13 C), CD 3 OD ¼ 3.31 ppm for 1 H (49.00 ppm for 13 C) or DMSO d 6 ¼ 2.50 ppm for 1 H (39.52 ppm for 13 C). The following abbreviations were used: s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), br (broad). Analytical thin-layer chromatography (TLC) was performed using 60 F254 aluminium-backed silica gel plates coated with a 0.2 mm thickness of neutral aluminium oxide. Column chromatography was performed on silica gel 60 (70-230 mesh) or on alumina gel (activated, neutral, Brockmann I). RP-preparative HPLC purification was performed on a Waters HPLC 4000 instrument, equipped with a UV detector 486 and Waters Delta-Pack 40100 mm, 15 Å, 100 lm, reversed-phase column. Standard conditions were eluent system A (H 2 O/0.1% TFA), system B (CH 3 CN/0.1% TFA). A flow rate of 50 ml/min and a gradient of (0-40)% B over 30 min were used, detection was performed at 214 nm. The LC/MS system consisted of a Waters Alliance 2695 HPLC, coupled to a Micromass (Manchester, UK) ZQ spectrometer (positive electrospray ionisation mode, ESIþ). All the analyses were carried out using a Merck Chromolith Speed rod C18, 25 Â 4.6 mm reversed-phase column. A flow rate of 3 ml/min and a gradient of (0-100)% B over 3 min (or over 15 min) were used. Eluent A: water/0.1% HCO 2 H; eluent B: acetonitrile/0.1% HCO 2 H. Retention times (RT) are given in minutes. Nitrogen was used for both the nebulising and drying gas. The data were obtained in a scan mode ranging from 100 to 1000 m/z in 0.1 s intervals; 10 scans were summed up to get the final spectrum. High-resolution mass spectrometry analyses were performed with a Waters Synapt G2-S time-of-flight mass spectrometer fitted with an electrospray ionisation source. All measurements were performed in the positive ion mode. Melting points (Mp) are uncorrected and were recorded on a Stuart capillary melting point apparatus SMP3.

Synthesis of 4-benzoyl-benzaldehyde 25
To a solution of 4-benzoylbenzoic acid (1 g, 4.42 mmol) in 30 ml of ethanol and 30 ml of cyclohexane, 2 ml of sulphuric acid were added. The solution was refluxed for one night. After cooling to room temperature, the solution was washed with a saturated NaHCO 3 aqueous solution (2 Â 30 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo to offer ethyl 4-benzoylbenzoate as a white powder (1.09 g, 97% yield), with spectral data in agreement with the literature 16 . Ethyl 4-benzoylbenzoate (1 g, 3.94 mmol) was dissolved in 10 ml of dry THF and cooled to 0 C. LiAlH 4 (388 mg, 10.2 mmol, 2.6 equiv) was added by fractions and the mixture was stirred at 0 C for 3 h. 50 ml of water were added and the solution was extracted with ethyl acetate (3 Â 50 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo. The residue was then dissolved in 10 ml of chloroform and 4 equiv. of manganese(II) oxide were added. The mixture was stirred at reflux for 4 h. After cooling to room temperature, the mixture was filtered through a pad of celite and washed with chloroform. The filtrate was then evaporated in vacuo. The residue was purified by chromatography on silica gel, eluted by DCM/ EtOH 99/1 v/v to offer 4-benzoyl-benzaldehyde 25 as a white solid (145 mg, 18% yield), with spectral data in agreement with the literature 17  To a solution of 6-quinolinecarboxylic acid (1 g, 5.71 mmol) in 40 ml of ethanol, 2 ml of sulphuric acid were added. The solution was refluxed for one night. After cooling to room temperature, 100 ml of water were added and the solution was basified using K 2 CO 3 . The solution was then extracted with dichloromethane (3 Â 30 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo to offer ethyl 6-quinolinecarboxylate 25 (1.1 g, 95% yield) as a white powder. Spectral data were in agreement with the literature 18 .
Ester 25 (1.1 g, 5.47 mmol) was then dissolved in 10 ml of dry THF and cooled to 0 C. LiAlH 4 (375 mg, 10.9 mmol, 2 equiv) was added by fractions and the mixture was stirred at 0 C for 2 h. 50 ml of water were added and the solution was extracted with ethyl acetate (2 Â 50 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo. The residue was then dissolved in 10 ml of chloroform and 4 equiv. of manganese(II) oxide were added. The mixture was stirred at reflux for 4 h. After cooling to room temperature, the mixture was filtered through a pad of celite and washed with chloroform. The filtrate was then evaporated in vacuo. The residue was purified by chromatography on silica gel, eluted by DCM/nHexane 9/1 v/v to offer 6-quinolinecarboxaldehyde 26 as a white solid (343 mg, yield 30%); mp: 73-75 C (litt. 19
(2S)-2-Boc-amino-1-(2-aminoimidazo[1,2-a]pyridin-3-yl)-hexan-1one (  General procedure for the synthesis of diazepinones 4-20 Method A (for compounds 5-13 and 15-20). A solution of compound 2a-i (0.33 mmol) in a 50/50 v/v mixture of trifluoroacetic acid and dichloromethane (5 ml) was stirred at room temperature for 1 h. After completion of the reaction (HPLC analysis), the solution was concentrated to dryness and co-evaporated several times with acetonitrile to completely remove TFA. Then, were added THF (8 ml), K 2 CO 3 (182 mg, 4 equiv.), iodine (251 mg, 3 equiv.) and the appropriate aldehyde (1 equiv.). The mixture was stirred under heating at 70 C for one night. THF was then removed under reduce pressure and 30 ml of dichloromethane were added. The organic phase was successively washed with a 10% m/v sodium thiosulphate aqueous solution (2 Â 30 ml) and then with a saturated NaHCO 3 aqueous solution (2 Â 30 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo. The residue was purified by chromatography on alumina, eluted by DCM, DCM/EtOH 99/1 v/v and finally DCM/ EtOH 98/2 v/v. Method B (for compounds 4 and 14). A solution of compound 2a or compound 2e (0.33 mmol) in a 50/50 v/v mixture of trifluoroacetic acid and dichloromethane (5 ml) was stirred at room temperature. After completion of the reaction (HPLC analysis), the solution was concentrated to dryness and co-evaporated several times with acetonitrile to completely remove TFA. Then, 5 ml of water were added and the solution was basified by K 2 CO 3 and then extracted by chloroform (3 Â 10 ml). The organic layer was dried over Na 2 SO 4 , filtered and the solvent was evaporated in vacuo. The crude was then dissolved in 20 ml of chloroform and 1 equiv of the appropriate aldehyde was added. The solution was stirred at 80 C for 6 h. After cooling, DDQ (0.44 mmol, 2 equiv) was added to the reaction mixture. The solution was stirred at room temperature for 6 h (monitoring by TLC). The solution was washed with aqueous saturated sodium hydrogen carbonate solution (2 Â 30 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo. The residue was purified by chromatography on alumina, eluted by DCM and then DCM/ EtOH 99/1 v/v for compound 4 or by preparative HPLC for compound 14 (gradient of 100% H 2 O/0.1% TFA to 40% CH 3 CN/ 0.1% TFA).      [1,3]diazepin-5-one (as trifluoroacetate salt) (14).    [4,5-  To a solution of compound 19 (20 mg, 0.04 mmol) in 5 ml of DCM was added at 0 C a mixture of HBr 33% in acetic acid (5 ml). The solution was then stirred at room temperature. After completion of the reaction (HPLC analysis), the solution was concentrated to dryness and the crude product was dissolved in 2 ml of DMSO and purified by preparative HPLC (gradient of 100% H 2 O/0.1% TFA to 40% CH 3 CN/0.1% TFA). The fractions containing the product were collected. The ACN was evaporated in vacuo. The aqueous residue was basified by NaHCO 3 and extracted with DCM (3 Â 25 ml). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was evaporated in vacuo to offer compound 21 as a white solid ( (22). To a solution of compound 23 as a bromhydrate salt 11 (300 mg, 0.6 mmol) in 15 ml of DCM, hydroxybenzotriazole monohydrate (101 mg, 0.66 mmol, 1.1 equiv.), EDCI (126 mg, 0.66 mmol, 1.1 equiv.), 4-pentynoic acid (59 mg, 0.6 mmol, 1 equiv.) and triethylamine (184 mL, 1.3 mmol, 2.2 equiv.) were added at 0 C. The solution was stirred at 0 C for 30 min and then at room temperature for one night. The solution was washed with an aqueous saturated NaHCO 3 solution and then with brine. The organic layer was dried over Na 2 SO 4 , filtered and the solvent was evaporated in vacuo. Neutral red uptake (NRU) viability assay A viability assay was used to evaluate the cytotoxic activity on A375 and MDA-MB-435 melanoma cells as previously described 20 .
According to OCDE/ICCVAM guidance n 129, NRU assay was also used to evaluate the toxicity on the non-cancerous NIH-3T3 cells 21 . Cells were seeded at a final concentration of 2000 cells/ well (A375), 3000 cells/well (NIH-3T3) or 7500 cells/well (MDA-MB-435) in 96-well microtiter plates for 24 h. Then, cells were exposed to two negative controls (1% DMSO in culture medium (vehicle) and culture medium alone), or to different compounds freshly dissolved in DMSO, in a range of concentration from 0.01 to 40 mM (six wells per concentration). After 72 h of incubation, cells were stained with new medium containing 50 mg/mL of neutral red (3amino-7-dimethylamino-2-methylphenazine hydrochloride, Sigma Aldrich) at 37 C for 2 h. The neutral red medium solution was then removed and the wells were washed with PBS. Finally, 100 mL of solution containing ethanol, acetic acid and water (50/1/ 49 v/v/v respectively) were added into the wells and the absorbance was measured at 540 nm using the microplate reader 352 Multiscan MS (Labsystems). Cell survival rates were normalised to the absorbance values of untreated control cells. IC 50 was obtained using a logistic regression model on Graphpad Prism 5.0 software. Results are given as the mean ± standard deviation (SD) of three independent experiments. Statistical analysis was performed using the Student's t-test.

Sulforhodamine B (SRB) assay
Primary anticancer assay was performed on a panel of 60 human cancer cell lines derived from nine neoplastic diseases, in accordance with the protocol of the Drug Evaluation Branch, National Cancer Institute, Bethesda, Maryland, USA 22 . Cells were incubated with tested compounds at 10 mM for 48 h. The growth percentage was evaluated spectrophotometrically using a protein binding dye, SRB. Results for each tested compound were reported as the percent of growth of treated cells when compared to untreated control cells.

Cell cycle analysis
The cell cycle analysis was performed as previously described to compare data with our previous results 10 . Flow cytometric analysis was performed on 350,000 MDA-MB-435 cells seeded in culture dishes (60 mm diameter) and allowed to grow for 48 h. Cells were then treated with compound 5 or 8 at 5 mM concentration for 16 h and 48 h. After treatment cells were harvested and fixed with 70% ethanol. The fixed cells were then incubated with 10 mg/mL RNase A (Sigma Aldrich) and 1 mg/mL propidium iodide (Sigma Aldrich), in the dark, for 24 h at 4 C. Finally, DNA content of the cells was analysed using Gallios Beckman Coulter flow cytometer with FlowJo software v1.0.

Immunofluorescent staining for actin localisation
To evaluate the effect of compound treatment on cell morphology and stress actin fibres, an immunofluorescence imaging of cells was performed as previously described 10 . MDA-MB-435 cells were seeded on cover slips in petri dishes and allowed to grow for 24 h. After cell growth, cells were treated with the vehicule (control cells) or with JMV5038, compound 5 or 8 at 5 mM for 20 h. Then cells were fixed using Antigenfix for 20 min and permeabilized using 0.2% Triton X-100 for 4 min at room temperature. Actin was stained using a primary human anti-actin antibody (made from mouse) and a secondary Alexa Fluor 568-conjugated goat anti-mouse antibody. Both of them were incubated 1 h at room temperature. Nuclei were counter-stained using Hoechst 33342. Representative images were obtained using a Zeiss LSM 780 inverted confocal microscopy. Confocal fluorescence microscopy was performed on fixed cells under a 543 nm wavelength excitation for AlexaFluor568, and 405 nm for nuclei, and visualised at 620 nm (red) and 450 nm (blue) respectively.

Statistical analysis
Data set of physicochemical properties (cLogP, TPSA and molecular volume, Table 2) were calculated using the online available molsoft cheminformatic tool 23 . IC 50 values of the compounds included in the test are presented in Table 1 for the correlations between IC 50 and physicochemical properties of compounds modified in position 2 (Figure 2), and in Table S1 (SI) for the correlations between in vitro anti-cancer activity of JMV5038, compound 5 and compound 7 on the NCI cell line panel (Figure 4). Pearson's correlation coefficient r and level of significant p values were calculated using Origin Pro 2019. Diamines 3a-i were then successively reacted with a set of aldehydes, using two strategies. For compounds bearing an alkyne group (compounds 4 and 14), a 3-step procedure was used. Firstly, the trifloroacetate salts of the diamines were neutralised and converted into their corresponding free bases. Secondly, these latter were then refluxed successively in chloroform with the adequate aldehyde to lead to the corresponding gem-diamine intermediate, which was then oxidised using dichloro-dicyano-1,4benzoquinone (DDQ). Finally, pyrido-imidazodiazepinones 4 (Scheme 3) and 14 (Table 1) were isolated in low yields (5-18%), after purification by chromatography on alumina gel or by preparative chromatography. Indeed, the presence of the alkyne group led to several unidentified by products, which complicated the purification step. For other derivatives, diamines 3a-i, as trifluoroacetate salts, were directly reacted successively with a set of aldehydes with potassium carbonate and iodine at 70 C overnight. Initially, tertbutanol was used as the solvent. However, we found that tetrahydrofurane (THF) gave the same results and as it is easier to remove, this solvent was preferred for the cyclisation step. Pyrido-imidazodiazepinones 5-13 and 15-20 were finally isolated in 8-98% yield, after purification by chromatography on alumina gel. The introduction of a chiral center alpha to the carbonyl group might lead to partial racemization. However, using chiral HPLC analysis, we previously demonstrated that no epimerization occured during the synthesis of diazepine derivatives 13 .

Results and discussion
For several compounds, the necessary aldehydes used in the cyclisation step were synthesised (Scheme 2). For compounds 6 and 10, where R 2 is a 4-benzophenyl or a 6-quinolinyl substituent respectively, the corresponding aldehydes 24 and 26 were prepared, using a 3-step procedure, starting from their respective carboxylic acids.
Briefly, 4-benzoyl-benzoic acid or 6quinolinecarboxylic acid were esterified using ethanol in acidic medium. The ethyl esters 23 or 25 were then reduced into their corresponding alcohols, using lithium aluminium hydride. These latter were then oxidised using manganese (II) oxide to generate 4-benzoyl-benzaldehyde 25 and quinoline-6-carbaldehyde 26. For compound 4 0 bearing an alkyne group, the TMS protected 4-formylphenylacetylene 27 was prepared from 4-bromobenzaldehyde, using a Sonogahira reaction, according to the procedure described by Cummings et al. (Scheme 3) 24 . Boc protecting-3-indolecarboxaldehyde 28 was prepared from 3-indolecarboxaldehyde and di-tertbutyl dicarbonate, using the procedure described by Pelcman et al. (Scheme 3) 25 . After cyclisation and oxidation, compounds 4 0 and 9 0 were then unprotected, using respectively a solution of tetrabutylammonium fluoride in THF or a mixture of TFA/DCM, according to Scheme 3.
For 4-azidobutyl derivatives 16-18, introduction of the azido group was studied using two strategies. Initially, the azido group was introduced in the final step of the sequence by diazo transfer on the free amino derivative 23, using imidazo-1-sulfonyl azide in presence of potassium carbonate and copper sulfate (Scheme 4). The imidazolyl reactant was preferred to the classical sodium azide, because this reagent is more stable and easier to handle 26 . The diazo transfer reaction proceeded smoothly and led to several by-products, which complicated the purification step. Compounds 16-18 were finally isolated in 10% yields from compound 1 (5step procedure). In a second approach, Boc-6-azido-Nle-OH 29 was prepared from Boc-Lys-OH, according to the procedure described by Shoonen et al. 15 Acylation of 1 with compound 29 led to compound 3 h in 71% yield. The Boc group was then deprotected, using a mixture of trifluoroacetic acid/dichloromethane (50/50 v/v) and the resulting diamine was cyclized with 4bromobenzaldehyde to offer compound 16 in 30% yield. Compared with the initial 5-step sequence, the 4-step procedure led to a better global yield from compound 1 (21% vs 10%) and compounds were easier to purify. Finally, alkyne derivative 22 was prepared in 94% yields by acylation of the amine 23 with pentynoic acid (Scheme 4).

Structure-activity relationships
The in vitro cytotoxic activity of diazepine derivatives was measured towards A375 and MDA-MB-435 melanoma cell lines, using neutral red uptake viability assay. The half maximal inhibitory concentrations (IC 50 ) values are reported in Table 1.
We started our investigation by studying the position 2 of the diazepine ring. From our previous studies, we observed that 4bromophenyl, 4-iodophenyl, 2,4-dichlorophenyl and 1-or 2-  Table 1. naphthyl substituents were the most preferred groups in this position 11 , which could suggest that such substituents occupy a large hydrophobic pocket in the target. To further study this position, we investigated the substitution of this position by several other hydrophobic groups.
Firstly, an alkyne, an azido group or a phenylketone were introduced in para position of the phenyl substituent, because we previously demonstrated that the para position was the most favourable position for the phenyl substitution 11 . The results are reported in Table 1 (entries 2-4). The activities were comparable for the two tested melanoma cell lines, even if the MDA-MB-435 cell line was more sensitive to compounds 4 0 and 5, as also observed with JMV5038 (Table 1, entry 1). Whereas the substitution of the bromine atom of JMV5038 by an azido group (compound 5) or, in a lesser extent, by an alkyne group (compound 4 0 ) maintained the cytotoxic activity in the two melanoma cell lines tested, the benzophenone derivative 6 was totally inactive, probably due to the steric hindrance brought by the supplementary phenyl ring. Indeed, the molecular volume of compound 6, predicted using the molsoft cheminformatic tool, is about 20% greater than that of JMV5038 (397 vs 317 Å 3 , Table 2), whereas for compounds 4 0 and 5, the molecular volume increase is only 2% and 4% respectively (324 and 330 Å 3 ), suggesting that the molecular volume of the substituent occupying the position 2 of the diazepine ring is a crucial parameter for the biological activity.
Introduction of [6 þ 6] or [6 þ 5]-heterocycles was then investigated at position 2. Surprisingly, whereas the MDA-MB-435 cell line was more sensitive to JMV5038 and to compounds 4 0 and 5 than the A375 cell line (see entries 1-3, Table 1), these compounds were around 3 times less active on MDA-MB-435 than on A375 cells (see for example, compounds 7, 10 or 11, entries 5, 8 and 9, Table 1). This result could suggest that the molecular mechanism involved in the cytotoxic activity of compounds 7, 10 and 11 is not totally similar to the one involved in the activity of JMV5038. For [6 þ 6] derivatives, the coumarinyl group totally abolished the activity (compound 12), whereas the 2-or 6-quinolinyl groups were tolerated, even if the biological activities were weaker than those obtained with their previously described naphthyl analogues (IC 50 ¼ 35.49 and 30.71 mM for compounds 10 and 11 respectively, compared to IC 50 of 1.4 mM for 1-or 2-naphthyl derivatives on MDA-MB-435 cell line 11 ). Indeed, compound 12 possesses a large molecular volume (370.59 Å 3 ), which could be detrimental for the activity, as observed with compound 6. Moreover, in comparison with their naphthyl analogues, compounds 10, 11 and 12 possess a greater polar surface (Topological Polar Surface Area, TPSA) and a weaker cLogP ( Table  2). Whereas an increase of the polar surface was well tolerated in the case of the azido compound 5 (TPSA increase of 185% for compound 5 compared to JMV5038) suggesting that this parameter is not crucial for the activity of these series of compounds, the lipophilicity seems to be more important. Indeed, the cLogP of compounds 10 to 12 (2.43, 2.53 and 1.84 respectively) were weaker or slightly weaker than those calculated for JMV5038 and its 1-or 2-naphthyl analogues (2.83 and 3.31/3.43, Table 2). Concerning the [6 þ 5] derivatives, best activities on A375 cell line were obtained with the benzothiophene derivative 8 (IC 50 ¼ 7.97 mM), followed by the benzofuran derivative 7 (IC 50 ¼ 8.47 mM). Indolyl derivative 9 0 was totally inactive on the two melanoma cell lines. Among the three [6 þ 5] analogues tested, which are known to be bioisosters of the naphthyl group 27 , indole derivative 9 0 is slightly more hydrophilic (cLogP ¼ 2.20) than the benzofuran and the benzothiophene derivatives 7 and 8 (cLogP ¼ 2.47 and 2.93 respectively, Table 2), confirming that the lipophilicity is a crucial parameter for the choice of substituents to introduce in position 2.
To further evaluate the influence of some physicochemical parameters on the cytotoxic activity of compounds modified at position 2 (i.e. compounds 4 0 212), their IC 50 values were graphed as a function of three commonly used molecular descriptors (Figure 2), namely cLogP, molecular volume and TPSA. Whereas no correlation were observed between TPSA and IC 50 (correlation coefficient of 0.22 and 0.13 for A375 and MDA-MB-435 cell lines respectively, p values > 0.1, data not shown), a statistically significant correlation was obtained when the IC 50 values were compared to the molecular volume of compounds 4 0 212 and JMV5038, with a Pearson correlation coefficient of 0.75 (p values ¼ 0.01) for A375 cell line and 0.65 (p values ¼ 0.04) for MDA-MB-435 cell line. Thus, to obtain some activity on the two melanoma cell lines, the molecular volume of the diazepinone derivative need to be less than 350 Å 3 (except for compound 9 0 , see below) and with a cut off of around 330 Å 3 to obtain an IC 50 less than 10 mM (Figure 2(A,B)). Thus, a large substituent like a benzophenone (compound 6) or a coumarinyl group (compound 12) led probably to a steric clash with the target and to the loss of the activity. Then, the influence of the cLogP was investigated. After removing compounds 6 and 12 for a too high molecular volume, a statistically significant correlation was obtained when the IC 50 were compared to the cLogP, with a correlation coefficient of À  (Figure 2(C,D)). Thus, a cLogP greater than 2.4 was required to obtain a cytotoxic activity (IC 50 10 mM) within these series of compounds. This result could explain that the indole derivative 9 0 , even if it possesses a molecular volume less than 350 Å 3 , presents a cLogP slightly too low to be active (namely 2.20, Table 2).
Finally, a 3D diagram was built to visualise the influence of these two physicochemical parameters on the cytotoxic activity of compounds modified at position 2 (mean of the IC 50 obtained on the two melanoma cell lines). Figure 3 shows clearly that a cLogP between $2.5 and 2.8, associated with a molecular volume between 310 and 350 Å 3 are necessary to obtained active compounds in these series of compounds.
To resume the RSA study of the position 2, compound 5 showed the best activity of all derivatives modified in this position, with cytotoxic activities against the two tested melanoma cell lines closed to those obtained with JMV5038. Azido group is known to exhibit some properties in common with halogens and in particular this group resembles a bromo substituent in terms of approximate size, polarity and electronic properties 28 , which could explain that compound 5 presented a similar profile of activity to JMV5038.
Then, we turned our attention to the SAR study of the position 4 of the diazepine ring, keeping constant the 4-bromophenyl group at R 2 (i.e. compounds 13-17 and 19-22). We previously demonstrated that position 4 could tolerate short linear alkyl chains (methyl, ethyl or hydroxymethyl), but also bulky groups like tert-butyl, but not too close to the diazepine ring 11 . To further study this position, several other chains were introduced into position 4. n-Butyl and n-azidobutyl chains are tolerated (compounds 13 and 16 respectively), with a good activity obtained for compound 16, close to the activity obtained with JMV5038 (mean of IC 50 values in MDA-MB-435 cells are 4.12 mM for compound 16 compared to 1.30 mM for JMV5038, entries 1 and 14, Table 1). Surprisingly, we observed that the stereochemistry is not so crucial for the activity when an n-azidobutyl chain was introduced at position 4 of the diazepine ring. Whereas a R-methyl group at position 4 led to a totally inactive compound 10 , a R-azidobutyl chain led only to a slight decrease of the activity on A375 cell line, and to a more significant decrease for the MDA-MB-435 cell line (see entries 14 and 15, Table 1). This result could suggest that the n-azidobutyl derivatives were not positioned exactly in the same way as the methyl analogues in the target or that the mechanisms of action involved in the cytotoxicity are not totally similar. However, as also observed with the 4-methyl derivatives, the presence of a bromophenyl substituent at R 2 is still necessary, because compound 18 is totally inactive (entry 16, Table 1). Introduction of a ramified isobutyl chain (compound 15) led also to a loss of the activity (entry 13, Table 1). This result is not surprising, because we already reported that an isopropyl chain in this position is unfavourable for the activity 11 , suggesting that substituents borne   Table S1, SI. The difference between the growth level of a specific cell line and the mean growth is represented by a bar. Cell lines that were more sensitive are displayed as bars that project to the right of the mean. Cell lines that were less sensitive are displayed with bars projected to the left. by position 4 of the diazepine ring occupied a relatively narrow pocket in the target. In the same way, a less flexible chain bearing an alkyne group (compound 14) was not tolerated, as well as a long chain (compound 22) ( Table 1, entries 12 and 20 respectively). Surprisingly, some activity was recovered with the bulky benzyl ester derivative (compound 19, entry 17, Table 1), whereas compounds with chains bearing a phenyl ring (namely a benzyl or a -(CH 2 ) 4 NHCbz substituent) were previously determined to be inactive 10,11 . This result could suggest that compound 19 could be a pro-drug. Indeed, compound 19 was partially converted into its corresponding acid 21, with a hydrolysis rate of 65.5%, determined by HPLC-MS, after 72 h in the presence of melanoma cells and culture medium. The acid derivative 21 was prepared and tested, but this compound revealed to be inactive (Table 1, entry  19), which seems to invalidate the pro-drug hypothesis. At this time we could not exclude that the main problem with compound 21 is not its activity towards the target, but its increase of polarity (cLogP of 2.1 for compound 21, compared to 3.97 for the ester 19); this parameter revealing to be important for the biological activity. However, contrary to what was observed with modulations of the position 2 of the diazepine ring, no significant correlation could be highlighted between the activity of the 4modified compounds and their cLogP or their molecular volume (data not shown).
Finally, the introduction of a fused tetrahydropyrrolo cycle between positions 3 and 4 (compound 20) led to the loss of the activity. This result confirms that the position 3 of the diazepine ring needs to be unsubstituted, as already observed with the N3methyl derivative of JMV5038 11 .
Then, all active compounds on the two melanoma cell lines (IC 50 < 40 mM) were assessed on the NIH-3T3 mouse fibroblast cell line, in order to check the toxicity of our compounds on non-cancerous cells. Globally, the diazepine compounds were not toxic on this cell line (IC 50 > 40 mM), except for compounds 16 and 17 bearing an n-azidobutyl chain, where a not negligible toxicity was noted (IC 50 around 10 mM, Table 1, entries 14 and 15). As a possible modification of the mechanism of action was hypothesised for these two derivatives (see above), this phenomenon could probably be also correlated with the cytotoxicity observed on the NIH-3T3 cell line.
In vitro antiproliferative activities on the NCI-60 cancer cell line panel From the SAR study of the position 2 of the diazepine ring conducted on A375 and MDA-MB-435 cell lines, it appeared that the sensitivity of the two melanoma cell lines tested towards several compounds was quite different.
Whereas a substituted phenyl group in position 2 of the diazepine (compounds 4 0 and 5) led to a similar response than JMV5038 (i.e. more active on MDA-MB-435 than on A375 cell lines, Table 1, entries 2-3), introduction of [6 þ 6] or [6 þ 5]-heterocycles in this position led to compounds more active on A375 than on MDA-MB-435 cell lines (compounds 7-8, 10-11, Table 1, entries 5-6 and 8-9). Even if the mechanism of action and the target of JMV5038 are not elucidated at this moment, the specificity of the activity profiles on different cell lines could suggest that these new derivatives did not share exactly the same mechanism involved in the cytotoxic activity. To better compare the activity profile of these two series of compounds, compounds 5 and 8 were selected and then evaluated towards a panel of sixty cancer cell lines corresponding to nine different cancer types, i.e. leukaemia, melanoma, as well as lung, colon, central nervous system, ovarian, renal, prostate and breast cancers (NCI evaluation). Indeed, it is now well established that the growth inhibitory patterns of anticancer drugs against the NCI-60 cell line panel correlated well with their modes of action 29 . For example, compounds that bind to tubulin have similar growth inhibition patterns regardless of which site on tubulin they bind 30 . Compounds 5 and 8 were chosen because they are the most active compounds in the two series of 2-modified derivatives ( Table 1, entries 3 and 6). Compounds were tested using the sulforhodamine B (SRB) assay that provides a direct cell density index. The amount of dye incorporated by the cells increases with the increase of the total protein biomass. Cells were incubated for 48 h with compounds at 10 mM concentration, then cell growth was evaluated spectrophotometrically and presented as the percent of untreated control cells. Values of the most sensitive cell lines were compared to those obtained with JMV5038 ( Figure 4, for all data see SI).
As suggested by our results obtained on A375 and MDA-MB-435 cell lines, the activity profile on the NCI-60 cell line panel revealed a great difference of sensitivity between the two tested compounds. Compound 5 presented a profile similar to that obtained with JMV5038 (statistically significant correlation, r ¼ 0.72, p values < 0.01), with the best activity obtained against MDA-MB-435 melanoma cell line (growth inhibition percentage of 87%), as already observed with JMV5038 and several other analogues 11 . On the other hand, the activity profile obtained with compound 8 was very different from that of JMV5038 and no correlation was obtained (r ¼ 0.07, p values ¼ 0.63), confirming that the mechanism of action involved in the cytotoxic activity is probably not totally comparable to that of JMV5038.

DNA cell cycle analysis
Compounds 5 and 8 were then selected for further characterisation, because these two compounds showed different activity profiles (similar to that of JMV5038 for compound 5 and different for compound 8). Thus, compound 5 could probably shared a common mechanism of action with JMV5038, contrary to compound 8. DNA cell cycle analysis was performed using MDA-MB-435 cells treated with either compound 5 or compound 8, and compared to that obtained with cells treated with vehicle (DMSO) as a negative control 10 . Cell cycle pattern changes of MDA-MB-435 cells were investigated using fluorescence-activated cell sorting (FACS) ( Figure 5 and Figure S1, SI). In this experiment MDA-MB-435 cells were treated with or without 5 mM concentration of compound 5 or 8 for 16 h and 48 h.
After 16 h, compound 5 increased the proportion of cells in G2/M phase (33.1% compared to 14.3% in control). At 48 h, this increase was still present, but less marked (18.4% vs 14.3%). A slightly increase of apoptosis in subG1 phase was also observed (2.4% compared to 1.2% in control). These induced changes in the cell cycle were quite comparable with those already observed in cells treated with JMV5038 10 . For compound 8, a slightly increase of the proportion of cells in G2/M phase was also observed (20.8% at 16 h and 15.4% at 48 h vs 14.3% in control). However, compound 8 increased the proportion of cells in phase G1 at 16 h (53.0% compared to 47.1% in control). This profile was opposite from that observed with compound 5, which led to a decrease of cells in G1 phase at 16 h. The activity profile of compounds 5 and 8 on the NCI-60 cancer cell lines was quite different and this dissimilarity was also observable through the analysis of the cell cycle, with a slightly shifted distribution of DNA, particularly in the G1 phase.

Morphological changes of cells
In our previous study, we reported that JMV5038 led to changes in the cell morphology, concomitant with a disruption of the actin network 10 . Morphological changes in MDA-MB-435 cells induced by compounds 5 or 8 were investigated and compared to those induced by JMV5038. For this purpose, fixed and actin immunostained MDA-MB-435 cells were imaged by confocal microscopy ( Figure 6). Control cells treated with DMSO (20 h, 5 mM concentration) were spread and the tensile forces of actin stress fibres stretched the cell-adhesive phenotype. In contrast, cells treated with compounds 5 or 8 (20 h, 5 mM concentration) showed the most recognisable actin-morphological features, comparable with those obtained with cells treated with JMV5038. Even if compounds 5 and 8 showed a different profile of activity on the two tested melanoma cell lines and on the NCI-60 cell line panel, the morphological changes induced by the three tested compounds were very similar. A diffuse distribution of non-polymerized actin, associated with a disorganisation of the actin cytoskeleton and altered malignant melanoma cell morphology were observed.

Conclusion
Nineteen new pyrido-imidazodiazepinone derivatives were synthesised, bearing various substituents at positions 2, 3 and 4 of the diazepine ring. Their cytotoxic activities against A375 and MDA-MB-435 melanoma cell lines were evaluated, as well as their potential toxicity against NIH-3T3 fibroblast cells. Introduction of a 4-ethynyl-, a 4-azidophenyl substituent was tolerated onto the position 2 of the diazepine ring and led to compounds with a similar in vitro activity on melanoma cells and with no apparent toxicity on NIH-3T3 cells. On the other hand, introduction of a bicyclic and heteroaromatic group on this position led to compounds possessing a significant difference of activity towards the two tested melanoma cell lines. These results were confirmed by comparing the activity profile of compounds 5 and 8 on the NCI-60 cell line panel, indicating that the mechanisms of action of these two compounds are not exactly similar. This dissimilarity was also observed through a slightly shifted distribution of the cell cycle phases. However, both compounds led to comparable modifications on cells morphology. SAR studies revealed also that the molecular volume of the substituent introduced onto the position 2 of the diazepine ring, as well as the cLogP of the molecule, were two crucial parameters to maintain the activity. Thus, a molecular volume less than 350 Å3 and a cLogP greater than $2.5 were necessary to maintain the activity. These results will guide the choice of the substituents to be introduced at this position in next studies. Then, modulations of the position 4 of the diazepinone revealed that a butyl chain, an azidobutyl chain or an ester group were tolerated, even if the azido-alkyl group led to some toxicity on the NIH-3T3 cells. However, such azido compounds could be particularly interesting to introduce further modulations using click-chemistry. In particular, introduction of a biotin or a fluorescent moiety on these compounds could be helpful to determine their cellular localisation as well as their pharmacological target using a proteomic approach.