Discovery of a novel allosteric inhibitor scaffold for polyadenosine-diphosphate-ribose polymerase 14 (PARP14) macrodomain 2

Graphical abstract


Introduction
Poly-(ADP ribose) Polymerases (PARPs) are ADP-ribosyl transferase enzymes which post-translationally modify substrate proteins. 1 Of at least 17 human family members of PARPs a sub-set, referred to as mono(ADP-ribose)transferases (mARTs), are capable of transferring on a single ADP unit to a given substrate. 2 PARP14 (ARTD8) is the largest of the mARTs and contains multiple domains including an ADP ribose transferase domain (ART), a WWE domain, two (RNA binding) RRM repeats and three (ADP-ribose binding) macrodomains. 3 PARP14 was found to be highly expressed in B-cell lymphoma and hepatocellular carcinoma and has been associated with poor patient prognosis. 4 Furthermore PARP14 has been linked to inhibition of pro-apoptotic kinase JNK1 which activates pyruvate kinase M2 isoform (PKM2) which in turn promotes a higher rate of glycolysis in cancer (Warburg effect) 5 shown in some contexts to be regulated by high MYC expression. 6 Despite links with cancer pathogenesis 5,7 and inflammatory diseases, 1b,c,7,8 only a few small molecule PARP14 inhibitors have been reported and many have suffered from a lack of selectivity. 9 Most examples of PARP inhibitors have targeted the catalytic domain (ART) 10 such as a recent example by Upton and coworkers who identified moderately selective PARP14 inhibitors, 10e however to date no PARP14 modulators targeting other domains such as the macrodomains have been reported until recently. 11 PARP14 contains three macrodomain modules (MD1, MD2 and MD3); biophysical characterisation of macrodomain:ADP-ribsose peptide binding was carried out revealing MD2 as the most potent ADP ribsosyl peptide binding domain and therefore the most likely to deliver a functional effect through small molecule inhibition (PARP14 MD1/ADP-ribose peptide K D 137 ± 7 lM, PARP14 MD2/ ADP-ribose peptide K D 6.8 ± 0.1 lM, PARP14 MD3/ADP-ribose peptide K D 15 ± 0.9 lM, Supp. Info Fig. 1 (Fig. 1 An initial medium throughput screen (50 k compounds) revealed compound GeA-69 (1) as a sub-micromolar inhibitor of PARP14 MD2 ADP-ribose binding as measured by AlphaScreen TM , ITC and BLI. 11a A co-crystal structure of closely related sulfonamide derivative 2 with PARP14 MD2, which was obtained in the course of the project revealed a unique allosteric binding mode for this inhibitor (PDB ID 5O2D). Overlay of this structure with bound ADP-ribose from a previously published co-crystal structure of PARP14 MD2 (PDB ID 3Q71) 12 showed that compound 2 occupied a novel pocket adjacent to the binding site for ADPribose ( Fig. 2A). 11a Carbazole 2 engages PARP14 MD2 in a pocket adjacent to the ADP-ribose binding site and the interaction is characterised by a H-bond between the carbazole N-H and backbone carbonyl of Pro1130 (N-O distance 2.8 Å), an H-bond between one sulfonamide carbonyl and the backbone N-H of Ile1132 (O-N distance 2.8 Å), and an H-bond from the sulfonamide N to a water molecule in the binding pocket. A comparison of the two structures rationalises inhibitory activity as carbazole 2 induces a shift in the loop region adjacent to Pro1130 which consequently moves into the ADPribose binding site ( Fig. 2A). Evaluation of the co-crystal structure of carbazole 2 with PARP14 MD2 also revealed the possibility of extending the methanesulfonamide motif into larger substituents exploring peripheral regions of this newly identified allosteric site.

Systematic SAR studies of screening hit GeA-69 (1)
The screening hit GeA-69 (1) was part of a focused library from the Bracher lab, originally designed for the improvement of kinase inhibitors derived from the 1-(aminopyrimidyl)-b-carboline alkaloid annomontine. 13 The SAR studies on screening hit GeA-69 (1) are described in the following compound library generated as potential PARP14 MD2 inhibitors (Fig. 3). In this library, the b-carboline ring system was replaced by its deaza analogue carbazole, and a number of aromatic and heteroaromatic rings were attached to position 1 (Scheme 1) using Suzuki-Miyaura cross coupling reac-tions of known 1-bromocarbazole 14 with commercially available or synthesised boronic acids and esters to give compounds 3-12 (Scheme 1).
Unfortunately none of these analogues (compounds 3-14) showed any inhibition of PARP14 MD2. Only a few further modifications of the 1-aryl substituent were performed, whereby all new compounds contained the acetylamino moeity, which was recognised as important for activity in this early stage of the project.
The aza analogue 15 was obtained from N-SEM protected 1-bromocarbazole by Masuda borylation at C-1, directly followed by Suzuki-Miyaura cross-coupling with 4-amino-3-bromopyridine, subsequent N-acetylation and SEM deprotection, as previously described. 11a This compound has virtually identical size as the active compound 1, but interestingly was found to be completely inactive at inhibiting PARP14 MD2 presumably due to the differences in electronics of both molecules. Consequently, this compound could serve as a useful negative control in biochemical experiments. The pyridyl-isomers 16 and 17 were obtained in the same manner using 3-amino-2-chloro-and 3amino-4-chloropyridine in the cross-coupling reaction (Fig. 5). Furthermore, using Suzuki-Miyaura cross-coupling reactions, the acetylaminophenyl residue was attached to position 1 (Scheme 1) of the b-carboline ring system 15 in order to obtain a ring A aza-analogue 18 and to the canthin-4-one 19 and desazacanthin-4-one 16 20 ring systems in order to give analogues bearing tetracyclic core structures (Fig. 5).
An analogue of GeA-69 (1) with the acetamido group shifted from the ortho to the meta position at the phenyl ring 21 was prepared by Suzuki-Miyaura cross-coupling of 1-bromocarbazole with 3-aminophenyl boronic acid, followed by N-acetylation. Additionally, the complete acetylaminophenyl residue was shifted from C-1 to N-9, whereby in one example a rigid isomer 22 was obtained,  and in the other, by means of a methylene spacer, a product 23 in which by appropriate rotation both the phenyl and the acetamido group can adopt positions that are very similar to those these groups have in the lead structure GeA-69 (1). Compound 22 was obtained by N-arylation of carbazole with 2-fluoro-1-nitrobenzene, 17 subsequent reduction of the nitro group, and N-acetylation.
As modifications of the central pyrrole ring (ring B) of GeA-69 (1) N-methyl and N-benzyl analogues 24 and 25 were prepared starting from corresponding N-substituted 1-bromocarbazoles via Suzuki-Miyaura cross-coupling with 2-aminophenylboronic acid and subsequent N-acetylation. Dibenzofuran analogue 26 and dibenzothiophene analogue 27 were obtained in a similar manner from commercially available 4-bromodibenzofuran and known 4iododibenzothiophene (Fig. 7). 18 These experiments were performed before we obtained the crystal structure of PARP14 MD2 with inhibitor 2, which demonstrated the relevance of the pyrrole NH-group (Fig. 2).
In order to replace the NH group of ring B with either an alternative hydrogen bond donor (hydroxy group) or a hydrogen bond acceptor (carbonyl group), known 1-iodofluorenone 19 was coupled   in the established manner to give the 1-arylfluorenone 28 which was easily reduced to the racemic fluorenol 29 with sodium borohydride (Fig. 7).
Controlled mono-acetylation of 2,2 0 -diaminobiphenyl with equimolar amounts of acetic anhydride gave monoamide 30 in moderate yield. Monoamide 30 was then used to access the seco analogue 31 and the acridone analogue 33. Buchwald-Hartwig arylation of the unsubstituted anilino group with iodobenzene to give biaryl 31 and with methyl 2-iodobenzoate to give biaryl 32, respectively, was accomplished with the BINAP/Pd 2 (dba) 3 catalyst system. Ester 32 was hydrolysed to give the corresponding carboxylic acid, which was converted into the acridone 33 by polyphosphoric acid-mediated intramolecular acylation (Scheme 2). 20 Further, a series of modifications of ring A was performed. Ringsubstituted analogues 37-39 were obtained in two steps from readily available 1,2,3,4-tetrahydrocarbazol-1-ones 21 34-36 in two steps. Treatment of the ketones with POBr 3 in anisole gave the corresponding 1-bromocarbazoles under bromination/dehydrogenation conditions in moderate to poor yields. Subsequent standard Suzuki-Miyaura cross-coupling gave the desired arylcarbazoles 37-39 (Scheme 3).
Analogue 44 bearing a partially hydrogenated A-ring was obtained from the corresponding brominated tetrahydrocarbazole 23 via Suzuki-Miyaura cross-coupling. A truncated analogue, the 7-aryl-3-isopropylindole 45, in which ring C is replaced by an isopropyl group, was obtained by Suzuki-Miyaura cross-coupling of the respective 7-bromoindole. The 6-aza-5,6,7,8-tetrahydro analogue 47 was prepared in a similar manner from known intermediate 46. 24 Improved yields were obtained, if the secondary amine was protected with the Boc group prior to the cross-coupling reaction (Scheme 5).
A screening of the above presented compounds on PARP14 MD2 clearly demonstrated that lead structure GeA-69 (1) is very sensitive to structural modifications. Carbazoles bearing (hetero)aromatic residues different from the acetylaminophenyl residue of GeA-69 (1) (Figure 4) were found to be inactive. Analogues with almost identical shape albeit very different electronically (aza analogues in the rings A, C and D) are completely or virtually (b-carboline 18, IC 50 30 lM) inactive. Any changes in the central pyrrole ring (ring B) eliminated inhibitory activity as well. The NH group was found to be essential, it can not be replaced by another hydrogen bond donor, as demonstrated by the inactive fluorenol analogue, 29. Surprisingly, the dibenzothiophene analogue 27 showed considerable inhibition (IC 50 2.5 lM), whereas the dibenzofuran, 26 and the acridone, 33 were inactive. The same holds for the (deaza)compounds having tetracyclic canthin-4-one backbones (canthin-4-one 19, deazacanthin-4-one 20). The seco analogue of GeA-69 (1), biaryl 31, was completely inactive, demonstrating that not only the presence of the functional groups of the lead structure, but also their fixation by the carbazole backbone is most important.
The tetrahydro-analogue 44 showed only a slight loss in activity (IC 50 1.1 lM) compared to GeA-69 (1), whereas its 6-aza analogue 47 bearing a polar aliphatic amino group in ring A, was inactive. Lipophilic chlorine substituents at ring A (compounds 37-38) were fairly tolerated (IC 50 1.4 and 3.0 lM), but the 6-methoxy analogue 39 was inactive. These observations can be rationalised by the hydrophobic environment in the binding region of ring A consisting of residues V1032, V1092, M1108, I111, I1112, F1129, I1132 (Fig. 2). Removal of the N-acetyl residue from GeA-69 (1), conversion of the acetamide into a tertiary amide 54 or into the proposed trifluoroalkyl bioisoster 51, as well as reduction of the amide moiety to an amine 53 resulted in complete loss of activity, the thioamide 52 was an order of magnitude less active (IC 50 10.5 lM) than GeA-69 (1).
In conclusion, these data confirm a very narrow structure-activity relationship for rings A-C (Fig. 3), and for further optimisation of the screening hit GeA-69 (1) only modifications of either the N-acyl residue or ring D were deemed promising.

SAR studies of ring D and N-acyl residues
Initial construction of the carbazole series was performed using 1-bromo-9H-carbazole and a series of pinacol boronic esters which were coupled under standard Suzuki-Miyaura conditions, furnishing biaryl products in moderate to good yields (Scheme 1). A number of these compounds were then converted to the corresponding acetamides or methanesulfonamides and profiled for their binding activity with PARP14 MD2. Whilst binding activity was not improved, additional substituents on ring D such as methyl, fluoro and cyano were tolerated maintaining single digit lM activity (compounds 55-57, Table 1). As previously observed a comparison of these compounds with the inactive non-acetylated and non-sulfonylated anilines (eg compounds 59-61, Table 1) showed the requirement of this group for binding activity.
Further modification of biaryl-amine 48 to the corresponding amides or sulfonamides (Scheme 7) was carried out. The corresponding amides and sulfonamides 62-108 were then profiled for their PARP14 MD2 binding affinity (Tables 1 and 2).
Compounds were profiled for binding activity with PARP14 MD2 through a competitive (AlphaScreen TM ) binding assay measuring the displacement of ADP-ribose peptide from PARP14 MD2. 11a Promising compounds were additionally profiled by biophysical assays such as Bio-Layer Interferometry or Isothermal Titration Calorimetry as previously described. 11a As previously described the parent carbazole GeA-69 (1) was profiled for its broader selectivity over 12 other human macrodomains, showing exquisite selectivity for MD2 of PARP14. 11a Furthermore a representative selectivity screen of 46 kinases in a Differential Scanning Calorimetry assay did not reveal any significant activity of carbazole GeA-69 (1) at 10 lM. 11a

Discussion
The binding activities of synthesised PARP14 MD2 inhibitors are summarised in Tables 1 and 2. Despite comprehensive SAR studies of the A-C rings of this carbazole series, no points for the development of more potent ligands were discovered, a number of deriva- tives were synthesised functionalising ring D (Figure 3). Only small additional substituents to the ring were tolerated (e.g. compounds 55-57, Table 1). Interestingly, elaboration of the sulfonamide in compound 2 into the homologated ethane-, propane-and butane-sulfonamides analogues (compounds 62-64, Table 1) furnished equipotent compounds. Further elaboration of the acetamide in GeA-69 (1) mostly retained single digit lM binding activity (eg compounds 66,67). Interestingly the n-pentanoyl ana-logue 68 was seemingly inactive, which may be due the entropic penalty associated with longer alkyl substituents or a steric clash with the protein. However, guided by the apparent tolerance of some larger substituents in place of the acetamide in GeA-69 (1) and methanesulfonamide in compound 2, the 2-phenylacetamide and phenylmethanesulfonamide of compounds 78 and 79 (IC 50 7.6 ± 0.3 and 3.6 ± 0.3 lM respectively, Table 1) were chosen for further development as they enabled rapid access to diversity and provide a suitable vector for binding pocket exploration. A number of hetero-and substituted-aromatics were appended onto the biaryl core (examples 83-108, Table 2). Moderately flat SAR was observed for both 2-and 4-substituted phenylacetyl and phenylmethanesulfonamide groups. It was found that introduction of a 3-cyano substituent in the phenylmethanesulfonamide series provided a slight improvement in binding activity compared with GeA-69 (1). Carbazole 108 displays sub-micromolar activity for a Data was not successfully obtained due to solubility issues in the AlphaScreen assay with this example.
Scheme 7. Synthesis of amide and sulfonamide derivatives of aniline 48. PARP14 MD2 (IC 50 660 ± 30 nM). Notably, by comparison the corresponding 3-cyanophenylacetamide 107 displays diminished binding activity relative to sulfonamide 108, potentially due to the greater tolerance of the sulfonamide to maintaining H-bond acceptor interactions as shown in the PARP14 MD2:compound 2 co-crystal structure (Fig. 2B). The 3-cyanobenzyl group of compound 108 may make interactions with adjacent hydrophobic residues M1108, L1137 and F1144. Although we were unable to obtain a co-crystal structure of compound 108 to confirm these interactions, we performed docking studies to examine possible binding modes of the larger compound compared to compound 2. Simple minimisation of compound 108 in PARP14 MD2 is unable to find a binding pose due to clashes between the larger 3-cyanophenyl group and the protein. To account for potential side chain rotations that would be necessary to accommodate this group, we performed SCARE docking (SCan Alanines and Refine) using ICM. 27 The optimised pose for compound 108 shows a rotation of the side chain of F1144 to open up space so that the 3-cyanophenyl group can make interactions with M1108 and L1137 in addition to a pistacking interaction with F1144 (Fig. 8). However, it is not obvious from this docking study why the 3-cyanophenyl group would be preferred to other hydrophobic groups such as in compounds 79 and 83-107. Sub-micromolar PARP14 MD2 affinity of carbazole 108 was also confirmed by BioLayer Interferometry (BLI) providing a calculated K D of 550 nM ± 220. Whilst lead compound 108 is larger and a less ligand efficient inhibitor of PARP14 MD2 than original hit compound 1, owing to the more tolerant SAR around it represents an attractive chemical starting point for future development. Additional examples similar to compound 108 (see SI, compounds 109-116) have been explored and work to improve the binding activity and physicochemical properties of this lead molecule will be reported in due course.

Summary
We herein report the development of a novel class of allosteric modulators of the second macrodomain of PARP14. Initial identification of carbazole GeA-69 (1) as a submicromolar inhibitor of PARP14 MD2 was made following a medium throughput screen. 11a Inhibitory activity can be rationalised through a PARP14 MD2 cocrystal of a similar derivative, sulfonamide 2 (PDB ID 5O2D). Investigation into this carbazole series was then made revealing new opportunities for ligand elaboration. Systematic analysis of SAR demonstrated a very narrow structure activity relationship for rings A-C (carbazole scaffold), and for further optimisation of the screening hit 1 only modifications of either the N-acyl residue or ring D showed promise. A number of carbazole containing compounds were tolerated in this newly identified allosteric site of PARP14 MD2 including a 3-cyano substituted phenylmethanesulfonamide 108. Carbazole 108 displays submicromolar activity binding to PARP14 MD2 by AlphaScreen (IC 50 0.66 lM) which was also confirmed by BLI (K D 0.55 lM).
This lead molecule along with others in this series are useful chemical starting points in the development of chemical probes for this poorly understood epigenetic target.