*2.3. Structural Studies*

2.3.1. Binding Modes of Guanine and 2,4-dioxopyrimidine Containing NAD+ Analogs to PARP-1 and PARP-2

Based on observed activity of 2,4-dihydroxypyrimidine nucleosides (including thymine and uracil derivatives) against PARP-1/2, one can assume that they can form canonical hydrogen bonds with an active site of PARP-1 and PARP-2, similar to those observed at the binding with natural substrate NAD+ (Figure 5). We carried out a molecular docking study to assess the efficiency of morpholino and ribonucleosides binding with the donor binding site of NAD+ (Figure 6A).

**Figure 5.** Structural model of the PARP-2 catalytic domain in complex with NAD+ at the donor binding site and ADP fragment at the acceptor binding site. The molecular surface illustrates ADP binding subsites of acceptor substrate (gray color), nicotinamide riboside fragment of donor NAD+ substrate (orange color) and the ADP fragment of donor NAD+ substrate (blue color). HD domain is not shown for simplicity. Substrates are shown in green color.

**Figure 6.** Predicted binding poses of **11IU** (**A**) and **5-I-Urd** (**B**) bound to the NA binding site of the PARP-2 catalytic domain. Hydrogen bonds are depicted as green dashed lines. Small molecules are shown in green color.

It includes the formation of hydrogen bonds with Gly863/Ser904 of PARP-1 or Gly429/Ser470 of PARP-2 and pi-stacking interaction with Tyr907 and Tyr473 of PARP-1 and PARP-2 correspondingly. The 5-halogen derivatives **11IU** (XP score −8.25 and −9.857 for PARP-1 and PARP-2 respectively (Figure 6A)) and **5-I-Urd** (XP score −8.328 and −8.902 for PARP-1 and PARP-2 respectively (Figure 6B)) were predicted to be among the top ranking binding scores due to enhanced hydrophobic interactions of iodine in the nicotinamide binding pocket, which is in agreement with experimental data.

Besides uracil-containing derivatives only guanine-containing compounds **Guo** (XP score −8.387 and −9.582 for PARP-1 and PARP-2 respectively) and **11G** (XP score −9.615 and −9.207 for PARP-1 and PARP-2 respectively) were able to form canonical hydrogen bonds in the nicotinamide binding subpocket of PARP-1/2, and have shown relatively high scoring function values. At the same time the presence of unsatisfied hydrogen bond acceptor groups of guanine residue located in the hydrophobic region of the nicotinamide binding subpocket likely leads to decreased binding affinity, observed in vitro. Lack of unsatisfied hydrogen bond groups in the hydrophobic nicotine amide subpocket is necessary according to analysis of all known co-crystallized PARP-1/2 inhibitors. Lack of this penalty likely leads to an overestimation of the scoring function value in case of PARP-1/2.

**11A**, **Ado**, **11C** and **Cyd** were unable to form canonical hydrogen bonds in the nicotine amide binding pocket with PARP1/2, and in agreement with experimental data have shown lowest scoring function values with dG > −7.7.

We found out that morpholino nucleoside derivatives in general have stronger inhibition activity to PARP-1/2 in comparison with corresponding ribonucleosides (Table 2). We could not correlate these observations with XP scoring functional values, but according to molecular docking predictions, the ribose fragment of ribonucleosides has a different localization in the PARP-1/2 catalytic site in comparison with the corresponding ribose fragment of the NAD+ substrate (Figure 5, Figure 6B). Presumably, it leads to reduced hydrophobic interactions in the subpocket formed by Tyr473, while the morpholine fragment occupies the binding site volume more efficiently (Figure 6A).

Overall, the conducted analysis shows that uracil-containing compounds, including thymine, can be used for the development of efficient PARP-1/2 inhibitors.

On the next step, we carried out a prediction of the binding modes of uracil-based NAD+ analogs synthesized in the current work, including **10T, 10IU, 4T** and **4IU**. We could not obtain the binding pose of the compound mimicking the natural substrate NAD+ binding pose using a molecular docking tool, expecting that uracil-containing derivatives would be bound to the nicotinamide binding subpocket, and adenosine would be bound to the adenine subpocket simultaneously (Figure 7A). One of the reasons could be due to the low propensity of the compound to fit the donor binding site or to the high number of conformational degrees of freedom of compounds. Therefore, we applied an alternative approach by combining the results of **5-I-Urd** and **11IU** molecular docking results and predicted the localization of the ADP fragment of the donor NAD+ nucleoside fragment based on the crystal structure of PARP-1 bound to BAD [38]. The predicted fragments were manually linked to each other, and molecular mechanics minimization using Schrödinger software with constraints on the heavy atoms of the PARP-1/2 and AMP fragment of compounds was carried out. By applying this protocol, we succeeded to obtain the molecular pose of compound **10IU** bounded to nicotinamide and ADP binding sites simultaneously in a similar way as the NAD+ substrate (Figure 7B). The visual inspection of the obtained binding poses shows the close proximity of the negatively-charged phosphate group and carbonyl oxygen of the pyrimidine group, potentially leading to a reduction of binding affinity. Another feature is the distant localization of the positively-charged group of the morpholine fragment from negatively-charged side chain of Glu558 of PARP-2. This side chain is involved in hydrogen bond formation with the ribose group of NAD+, while lack of this interaction can potentially lead to reduction of the binding affinity of **10IU.** The obtained results can explain why addition of the ADP group to **11IU** and **5-I-Urd** does not enhance its activity, and it also provides the basis for further compound optimization.

**Figure 7.** Predicted binding mode of **10IU** with the donor binding site of the PARP-2 catalytic domain. (**A**) The close-view of the donor binding site. (**B**) Structural alignment of the binding poses of **10IU** (green color) with NAD+ (cyan color) and **5-I-Urd** (black color). (**C**) Representative conformations of **10IU** from the molecular dynamics trajectory, the conformation with the intramolecular hydrogen bond with the non-bridging α-phosphate oxygen of ADP (bottom) and the conformation with the hydrogen bond with carbonyl oxygen of the heterocyclic base of the modified nucleoside (top) are shown. The molecular surface of the binding site is shown. Hydrogen bonds are depicted as green dashed lines.

As can be seen from Table 1, a substitution of P–O to P–N bond leads to an enhanced activity of the most tested NAD+ analogs. These effects can be explained by the formation of an intramolecular hydrogen bond that stabilizes pyrophosphate group conformation. To further study this phenomenon we carried out 100 ns molecular dynamics simulation of the complex of PARP-1 with the 5-I-uracil-containing ADP conjugate **10IU** with positional restraints on the heavy atoms of the adenine-ribose fragment of the compound and the CA-atoms of the protein. In this obtained MD trajectory, two types of hydrogen bonds of the P–NH group were observed: the hydrogen bond with the non-bridging α-phosphate oxygen of ADP (Figure 7C) and the hydrogen bond with the carbonyl oxygen of heterocyclic base of the modified nucleoside (Figure 7C). It has to be noted that in the first case the stabilization is independent from a type of heterocyclic base of morpholino nucleosides and is expected to be observed in all studied NAD+ analogs that is in agreement with in vitro data.
