2.3.2. Micromolar Inhibition of PARP-2 Activity by Targeting Acceptor Substrate Binding Site

As mentioned above, the dependence of compounds **4** efficacy on the heterocyclic base looks like Cyt < Ura < Ade < 5-Cl-Ura < Gua < Thy < 5-Br-Ura < 5-I-Ura for PARP-1 and (Cyt, Ura) < 5-Cl-Ura < Gua < (5-Br-Ura, Thy) < Ade < < 5-I-Ura for PARP-2. The main discrepancy between PARP-1 and PARP-2 series is the position of the Ade-containing conjugate **4A**: this compound does not influence on the activity of PARP-1, but weakly inhibits PARP-2. For P–N-containing compounds **10**, these series turn into Ura < Cyt < Gua < Ade << Thy << 5-I-Ura for PARP-1 and (Cyt, Ura) < Gua < Thy < 5-I-Ura << Ade for PARP-2. We noticed that these series were almost the same for PARP-1, but they had significant difference for PARP-2 in the position of the Ade-containing inhibitor. Namely, compound **10A** weakly inhibits PARP-1 (similar to its P–O analog **4A**), but proves to be the most effective for PARP-2. These results are counterintuitive due to the low activity of **10A** and **Ado** compounds predicted in silico and observed in vitro. Indeed, **10A** and **Ado** cannot be efficiently bound

by the nicotinamide subpocket according to the molecular docking predictions. In particular, this is due to a lack of pharmacophores necessary for the formation of critical hydrogen bonds with the NA site, including lactam or carboxamide groups [20].

Therefore, first we assumed that the adenine-MorXpp fragment of **10A** can compete for binding with the ADP fragment of NAD+. It has to be noted that NAD+ can be bound to PARP-1/2 at least at two binding sites of PARP-1/2, including donor (NAD+) and acceptor (PAR) binding ones (Figure 5). On the first step we evaluated the efficiency of adenine-MorXpp binding to the donor ADP binding site using molecular docking simulation. Molecular docking predictions show that adenine-MorXpp is not expected to bind more efficiently than Ado-50 -pp or NAD+ in the donor binding site, based both on bindings score values and visual inspection of high-ranked binding poses. Visual inspection indicates that the morpholine group does not form hydrogen bonds with residues His862 and Ser864 of PARP-1 or His428 and Ser430 of PARP-2, while the hydroxyl group of ribose in Ado-50 -pp does. The conducted analysis strongly suggests that **10A** cannot bind to donor NA and ADP binding sites.

On the next step we evaluated the hypothesis whether **10A** and **4A** can compete for binding with the acceptor substrate. We found that the adenine-MorXpp fragment of these compounds can efficiently mimic the interactions of the ADP fragment of the acceptor substrate. The latter was predicted based on the crystal structure of PARP from *Gallus gallus* (red junglefowl, PDB identifier 1A26; [78]). Moreover, the NH<sup>+</sup> moiety of the morpholine ring of the **10A** compound can form a salt bridge with Glu988 or Glu558 residues of PARP-1 or PARP-2, respectively. In this case, the morpholine ring can mimic an interaction of the 20–OH group of adenosine and expected to further enhance binding affinity in comparison with natural acceptor substrate (Figure 8A). Visual inspection of binding poses shows that PO→PN substitution leads to the formation of an additional intramolecular hydrogen bond with phosphate oxygen leading to an increased stability of interaction, which is supported by an enhanced docking score and enhanced in vitro activity. In accordance with the analyzed activity of the **10IU** compound we suggest that PO→PN substitution can be used as a general strategy to stabilize the active conformation of the compounds containing diphosphate groups.

It has to be noted that the binding mode of **10A** with the acceptor binding site of PARP-2 is characterized by the high solvent exposure of the compound and multiple polar interactions, which include at least 10 hydrogen bonds. The hydrophilic nature of the stabilizing interactions with the PARP-1/2 of the compound can explain the moderate activity of the **10A** compound, and provides further strategies of compound optimization.

The observed selectivity of **10A** to PARP-2 can be explained by the variable region of PARP-1/2 in proximity to the acceptor binding site. In particular, loops of PARP-1 (978–986) and PARP-2 (544–556) have a distinct conformation and amino acid composition (Figure 8A–C), while according to the structural model, Asn555 of PARP-2 is involved in the formation of a hydrogen bond with phosphate oxygen of **10A** and replaced by Leu985 in the case of PARP-1 (Figure 8B,C). Additionally, PARP-1 lacks stabilizing interactions with Tyr552 due to shortening of the corresponding loop. This is supported by the lower XP binding score for PARP-1 that was −7.975 and −10.574 in case of PARP-2.

Furthermore, lack of activity of **10A** against PARP-3 supports the before suggested mechanism of action. Indeed, PARP-3 structure was reported to be different from PARP-1/2 and characterized as mono(ADP-ribose) transferase [79,80]. In accordance with this data, molecular docking predicted that **10A** does not have the same mode of binding to PARP-3 as to PARP-1/2 nor other MorXppA compounds. Analysis of the binding site revealed that this could be due to steric hindrance caused by Arg408 and Lys421 residues, as well as the hindrance of the hydrogen bond acceptor and donor groups of Gln512 and Arg408, correspondingly, by small molecule (Figure 8D).

**Figure 8.** Predicted binding pose of **10A** with the acceptor binding site of the PARP-1/2 catalytic domain. (**A**) The structural alignment of PARP-1 and PARP-2. Variable loops are indicated in red and yellow colors for PARP-1 and PARP-2, respectively. Detailed view of **10A** interaction with PARP-1 (**B**) and PARP-2 (**C**) is shown. (**D**) PARP-3 acceptor binding site with superimposed binding pose of **10A** from PARP-2/**10A** complex**.** Steric clashes are shown with red disks. Unsatisfied hydrogen bond donor and acceptor atoms of the PARP-3 acceptor binding site hindered by ligand are shown as spheres. Hydrogen bonds are depicted as dashed lines. HD domain is not shown for simplicity.

On the whole, structural analysis of the binding modes of **10A** together with experimental data suggests that this compound acts by a different mechanism than known PARP-1/2 potent inhibitors and binds to the acceptor substrate binding site, but not to the donor NAD+ binding site. Additionally, the presence of variable regions in proximity of the acceptor binding site may lead to enhanced PARP-2 specificity of **10A**.
