Three-component Castagnoli-Cushman reaction with ammonium acetate delivers 2-unsubstituted isoquinol-1-ones as potent inhibitors of poly(ADP-ribose) polymerase (PARP)

Abstract An earlier described three-component variant of the Castagnoli-Cushman reaction employing homophthalic anhydrides, carbonyl compound and ammonium acetate was applied towards the preparation of 1-oxo-3,4-dihydroisoquinoline-4-carboxamides with variable substituent in position 3. These compounds displayed inhibitory activity towards poly(ADP-ribose) polymerase (PARP), a clinically validated cancer target. The most potent compound (PARP1/2 IC50 = 22/4.0 nM) displayed the highest selectivity towards PARP2 in the series (selectivity index = 5.5), more advantageous ADME prameters compared to the clinically used PARP inhibitor Olaparib.


Introduction
Poly(ADP-ribose) polymerase (PARP) enzymes are regarded as important targets for the development of anticancer drugs owing to the clinical success of PARP inhibitors Olaparib, Talazoparib, Niraparib and Rucaparib approved for cancer treatment 1 . PARP1 and PARP2 are two key enzymes that are critical for repairing single-strand breaks ("nicks") in the DNAa mechanism which is critical for the survival of both normal and cancer cells 2 . It is generally accepted that, despite the non-selectivity of the approved PARP inhibitors, it is the inhibition of PARP1 that is responsible for the manifestation of their clinical efficacy 3 . Normal cells do not divide as frequently as cancer cells, which allows them to survive PARP1 inhibition. However, tumour cells with certain mutations that are synthetically lethal with PARP1 inhibition 4 are efficiently killed by the drugs of this class. This notion motivated the development of selective PARP1 inhibitors such as NMS-P118 reported by Nerviano Medical Sciences 5 .
The majority of PARP1 inhibitors, including the above-mentioned "paribs", were designed to mimic the nicotinamide moiety of NAD þ (from which the adenine ribose unit of poly(ADP-ribose) originates) with which the inhibitors compete for the NAD þ -binding site of PARP1. This mimicry is achieved via the use of either a rotationally constrained primary benzamide (as in Niraparib and NMS-P118) or a benzamide motif embedded in a ring (as in Olaparib, Talazoparib and Rucaparib). Another characteristic feature noticeable in some of the advanced PARP1 inhibitors is the presence of a fluorine atom (the "magic fluorine" highlighted in blue) in the meta-position of the NAD þ -mimicking benzamide moiety ( Figure 1). This "magic fluorine" has been shown to enhance binding to the target 5 .
Recently, we discovered 1-oxo-3,4-dihydroisoquinoline-4-carboxamides 1 as a novel chemotype capable of delivering potent PARP inhibitors. After screening the amide residues around the basic core, we identified compound 1a as a lead structure for further development. It displayed a potent inhibition profile towards PARP1 (IC 50 ¼ 156 nM) and PARP2 (IC 50 ¼ 70.1 nM). Moreover,  compound 1a displayed a better microsomal and plasma stability in vitro compared to clinically used PARP1 inhibitor Olaparib 6 . Encouraged by this finding, we continued looking for ways to improve the potency profile of the 1-oxo-3,4-dihydroisoquinoline-4-carboxamide series. Carboxamide 1a bore no substituents in position 3 of the 1-oxo-3,4-dihydroisoquinoline core. Therefore, we considered exploring the structure-activity relationships within the series 2 where the 3-mono-and 3,3-disubstituted scaffolds could be screened while the 1,4 0 -bipiperidine carboxamide would remain unchanged. Access to derivatives 2 was envisioned via the amidation of carboxylic acids 3 which, in turn, could be prepared via the recently reported three-component Castagnoli-Cushman reaction (3 C-CCR) of homophthalic anhydride (HPA), carbonyl compounds and ammonium acetate 7 ( Figure 2). Herein, we describe the results obtained in the course of realising this strategy.
In order to probe for the minimal substitution at position 3, we aimed to synthesise 3-methyl-substituted 1-oxo-3,4-dihydroisoquinoline-4-carboxylic acid 3 m. Unfortunately, the 3 C-CCR protocol did not furnish the desired product, likely due to the high volatility of acetaldehyde. Thus, for the preparation of 3 m, we resorted to the alternative approach recently reported by Shaw and coworkers 8 .
nitrobenzene)sulfonyl (4-nosyl) imine of acetaldehyde from a-aminosulfone 4 9 on treatment with equimolar amount of base. N-Sulfonyl imine 4 thus generated can be reacted with HPA in the Castagnoli-Cushman fashion. Indeed, treatment of 4 with 1 equiv. of DBU followed by the addition of HPA furnished 4-nosyl CCR adduct which was esterified to give ester 5 in modest yield. Removal of the 4-nosyl protecting group followed by ester hydrolysis furnished the target acid (3 m) which was immediately introduced in the amidation reaction with 1,4 0 -bipiperidine to give compound 2 m in modest 26% yield over 3 steps (Scheme 2). Compounds 2a-m (confirmed to retain the trans-configuration based on the values of 3 J(H 3 -H 4 ) coupling constants observed in their 1 H NMR spectra) were tested for inhibitory activity towards PARP1 and PARP2 using the commercially available colorimetric activity assay kit from BPS Bioscience (San Diego, CA) in full accordance of the supplier's method description 10-11 . The initial screening was performed against PARP1 at 1 mM concentration of each compound, in triplicate (n ¼ 3) measurements. Compounds which displayed over 80% inhibition of the enzyme activity were in dose-response mode (n ¼ 3) against PARP1 and PARP2 using Olaparib as the reference inhibitor in order to determine the compounds' half-maximal inhibitory concentration (IC 50 ) and assess the isoform selectivity (Table 1).
It is apparent from the data presented in Table 1 that some small alkyl groups introduced in position 3 of the 1-oxo-3,4-dihydroisoquinoline-4-1,4 0 -bipiperidine carboxamide scaffold increase compounds' potency towards PARP1 and PARP2 (cf. 2c, 2e (the most potent inhibitor in the series) and 2 m vs. 1a) while bulkier (cyclo)alkyl groups do not change (2d) or ablate (2f and 2 l) the inhibitor's potency. Similarly, disubstituted (2 h) and spirocyclic (2i-k) analogs do not inhibit PARP1 in the concentration range relevant for drug development 12 . In light of the bulky substituents at position 3 being detrimental to the inhibitor's activity, it came as a surprise that 3-aryl-substituted analogs 2a-b displayed better potency compared to their unsubstituted counterpart (1a).
We attempted to rationalise some of the observed structureactivity relationships by performing docking simulation of the inhibitors' binding to the PARP1 active site. Figure 3 displays the binding poses of the most potent compound 2e as well as its active (2 b) and inactive (2 h and 2 l) analogs. Compounds 2e and 2 b displayed a favourable network of lipophilic contacts in the PARP-1 active cavity (Figure 2(A,B)). The 3,3-dimethyl substitution in 2 h causes a transition to another metastable conformation with the loss of hydrophobic interactions with Val762 and destabilisation of electrostatic contacts with Glu763/Asp766 ( Figure  2(C)). In the case of 2 l, the cyclopentyl moiety appears to induce a conformational rearrangement of molecule. The resulting reorientation of the 1,4 0 -bipiperidine moiety leads to the loss of the hydrophobic contacts with Val762 and p-stacking interaction with Tyr907 (Figure 2(D)).
To our delight, when tested for inhibition of PARP1 and PARP2, compound 6 indeed displayed an improved potency profile towards both enzymes (PARP1 IC 50 ¼ 22.0 ± 3.2 nM, PARP2 IC 50 ¼ 4.3 ± 0.5 nM, SI ¼ 5.1). Moreover, this compound can be regarded as a relatively rare PARP2-selective inhibitor which can enrich the toolbox of compounds needed for the investigation of the physiological role of PARP2 inhibition [14][15][16] .
We were curious to compare the physicochemical and ADME properties of the newly discovered lead compound (6) to those of more potent, clinically used PARP1/2 inhibitor Olaparib. Indeed, as was noted previously, the sheer potency of pharmacological agents is not a sole determinant of the pharmacodynamic efficacy and should be considered in combination with the overall candidate's profile 17 .
While, of course, the approved PARP1 inhibitor Olaparib is well within the limits of druglikeness as defined by Lipinsky 18 (and so is compound 6), the two compounds are quite similar in terms of molecular weight and lipophilicity. Quite reassuringly, compound 6 displayed similar stability in plasma to that of Olaparib. However, the metabolic stability of the former is significantly higher ( Table 2).   In summary, we have further optimised the inhibitory potency of an earlier discovered lead 1a based on a 1-oxo-3,4-dihydroisoquinoline-4-carboxamide scaffold by exploring various substitutions at position 3. This exploration was reliant on the earlier described three-component variant of the Castagnoli-Cushman reaction of ammonium acetate. The profiling of various 3-monoand 3,3-disubstituted analogs of compound 1a for PARP1 and PARP2 inhibition allowed establishing structure-activity relationships and led to the identification of 3-cyclopropyl substituent as the preferred periphery pattern. Introduction of a fluorine atom in the position 7 of the 1-oxo-3,4-dihydroisoquinoline core (inspired by the presence of this "magic fluorine" in several clinical candidates and approved drugs) further improved the potency profile and led to 1-oxo-7-fluoro-3,4-dihydroisoquinoline-4-carboxamide 6 as the new lead compound. It displayed 5.5-fold selectivity towards PARP2, good metabolic and plasma stability as well as better plasma protein binding, compared to Olaparib.