Structure-Driven Discovery of α,γ-Diketoacid Inhibitors Against UL89 Herpesvirus Terminase

ABSTRACT: Human cytomegalovirus (HCMV) is an opportunistic pathogen causing a variety of severe viral infections, including irreversible congenital disabilities. Nowadays, HCMV infection is treated by inhibiting the viral DNA polymerase. However, DNA polymerase inhibitors have several drawbacks. An alternative strategy is to use compounds against the packaging machinery or terminase complex, which is essential for viral replication. Our discovery that raltegravir (1), a human immunodeficiency virus drug, inhibits the nuclease function of UL89, one of the protein subunits of the complex, prompted us to further develop terminase inhibitors. On the basis of the structure of 1, a library of diketoacid (α,γ-DKA and β,δ-DKA) derivatives were synthesized and tested for UL89-C nuclease activity. The mode of action of α,γ-DKA derivatives on the UL89 active site was elucidated by using X-ray crystallography, molecular docking, and in vitro experiments. Our studies identified α,γ-DKA derivative 14 able to inhibit UL89 in vitro in the low micromolar range, making 14 an optimal candidate for further development and virus-infected cell assay.

INTRODUCTION
Human cytomegalovirus (HCMV), one of the eight known human herpesviruses, is widespread with a high prevalence throughout the human population.1 HCMV causes severe diseases and life-threatening conditions in immune-compro- mised individuals, such as acquired immunodeficiency syn- drome, and organ transplant patients, neonates, and young children.1 It is the leading congenital infection in developed countries, causing a number of neural abnormalities including hearing loss, mental retardation, and microcephalia. The drugs licensed so far for the treatment of HCMV infection are the antisense oligonucleotide fomivirsen and four viral DNA polymerase inhibitors: ganciclovir (GCV), valganciclovir, cidofovir (CDV), and foscarnet (FOS).2 There are several disadvantages associated with the use of these drugs including(i) poor bioavailability, (ii) serious side effects induced byprolonged treatment, and (iii) emergence of resistant viralof antiviral compounds able to target viral entry, transcription, protein synthesis and processing, genome replication, or maturation have been identified.2,6−13 Among these, a promising strategy is to block the HCMV terminase machinery14 which mediates DNA binding, cleavage, and packaging because these processes do not occur in mammalian cells and therefore specific terminase inhibitors could have fewer undesirable side effects.HCMV replicates its genome as concatamers that are large DNA molecules with several copies of the genome. Afterward, these are cut into single unit-length genome and packaged inside the capsids by the HCMV terminase machinery.15 HCMV terminase is a protein complex composed of UL56, UL89, and UL51 responsible for binding, cleavage, and translocation of viral DNA inside the procapsids, respectively.

The three- dimensional (3D) structure of the nuclease domain of UL89 wasstrains.3 Development of drug resistance is associated with andetermined by X-ray crystallography, whereas the structures ofaggressive clinical course, organ dysfunction, and mortality.4 Several cases of multidrug resistance to GCV, CDV, and FOS have been reported.5 Therefore, there is an urgent need for the development of new drugs with a different mode of action, good bioavailability, and a safer pharmacological profile. Novel classesLetermovir, a 3,4-dihydro-quinazoline-4-yl-acetic acid deriv- ative, a non-nucleoside inhibitor with a potent antiviral activity, specific against the viral terminase subunit pUL56,19,20 has been developed by AiCuris GmbH & Co KG and is currently in a phase III clinical study.21 An alternative approach consists of targeting UL89, another subunit of the terminase complex,drug discovery program against HCMV. We have focused our efforts on elucidating the 3D structural data of UL89 inhibitors with UL89-C.18 In our previous report, we found that pUL89-C endonuclease activity was inhibited by the HIV inhibitor raltegravir (1, Figure 1), suggesting that inhibiting pUL89-C is possible using a similar metal-chelating pharmacophore.18which contains two domains, the C-terminal region (UL89-C, 418-674), which has proved nuclease activity,16,18,22 and the N- terminal part (UL89-N, 1-417), which has putative adenosine- triphosphatase activity.23−25 UL89 inhibitors may be designed to block the adenosine 5′-triphosphate (ATP) binding site on the UL89-N region; amongst them, there are 2-bromo-5,6-dichloro-1-β-D-ribofuranosyl benzimidazole (Figure S1), 2,5,6- trichloro-1-(β-D-ribofuranosyl)benzimidazole, 2-bromo-5,6-di-chloro-1-β-D-ribopyranosyl-1H-benzimidazole (GW275175X), and 3-hydroxy-2,2-dimethyl-N-[4([[5-(dimethylamino)-1- naphthyl]sulfonyl]amino)-phenyl]propanamide (BAY 38- 4766), which interact with the ATP binding site and induce mutations on the UL89-N domain.

However, in spite of the efforts, none of them have reached the market.2,6,7 The presence of the highly conserved ATP binding motif in many different proteins could impede to obtain specific inhibitors for the viral UL89-N.33Another approach is focused on finding HCMV inhibitors able to target the nuclease activity of UL89-C. The structure of UL89-C, which is structurally related to the human immunodeficiency virus (HIV) integrase,18 exhibits a central eight-stranded mixed β-sheet, with parallel and antiparallel strands, flanked on both sides by six α-helices. The nuclease active site of UL89-C is located in a cleft formed by three loops β2−β3, β5−α3, and α5−α6. Three acidic residues Asp463, Asp651, and Glu534 chelate two Mn2+ ions, which are essential for UL89-C DNA cleavage. An UL89-C nuclease inhibitor should interact with the two Mn2+ ions and the amino acidslocated in the loops to block DNA interaction and cleavage.Most of the compounds against UL89-C tested so far are divalent metal-sequestering inhibitors, such as 1-(3-chloro-4- fluorobenzyl)-5-hydroxy-4-oxo-1,4-dihydropyridine-3-carbox- ylic acid34 (Figure S1), an hydroxypyridonecarboxylic (HPC) derivative, and 3-hydroxy-1-methyl-6-((4′-methyl-[1,1′-biphen- yl]-4-yl)amino)dihydropyrimidine-2,4(1H,3H)-dione35 (Fig- ure S1), a 3-hydroxypyrimidine-2,4-dione (HPD) derivative. HPC and HPD compounds inhibit the nuclease activity of UL89-C in the low micromolar range (IC50 = 2.2 and 1 μM, respectively), HCMV replication, and infectious virus produc-tion in cell culture.

Therefore, there is a correlation between blocking of pUL89 endonuclease activity and inhibition of HCMV replication and viral genome cleavage. Recently, 2,7- dihydroxy-4-(propan-2-yl)cyclohepta-2,4,6-trien-1-one (Figure S1) with metal-chelating motifs was identified as an inhibitor of pUL15C of herpes simplex viruses 1, a homologue of pUL89-C.36 Crystallographic studies of 2,7-dihydroxy-4-(propan-2- yl)cyclohepta-2,4,6-trien-1-one binding to the nuclease active site of bacteriophage terminase gp2 nuclease domain revealed that the interaction is mainly mediated by the two metal ions (PDB code: 5C2F).37Previous studies have reported the structure−activity relation- ships (SARs) of UL89 inhibitors explained by moleculardocking simulation, using the apo form of UL89, without the possibility to validate docking binding mode because there is no UL89-inhibitor complex available. Clearly, it is important to have an understanding of the interactions of UL89 inhibitors with UL89 and offer detailed information to drive a target-basedOn the basis of computational molecular docking studies of 1 (Figure S2A), the pivotal chemical regions of 1 binding pUL89- C endonuclease domain active site were identified. Upon definition of the chemical features, a new set of compounds was designed to improve the inhibition activity. Herein, we present the synthesis and evaluation of a library of small organic compounds able to inhibit UL89-C nuclease activity in an in vitro enzymatic assay. The ability of these compounds to bind UL89-C was tested by thermal denaturation. Furthermore, the X-ray structure of a protein−ligand complex unveiled the precise binding mode of the ligand in the enzyme active site, allowing afurther structure-driven antiviral development against UL89-C activity.

RESULTS AND DISCUSSION
Rational Design and Chemistry. Our rational design began by simplifying the structure of 1, a HIV integrase inhibitor, by selecting chemical features important for UL89-C binding. Computational docking studies of 1 in the active site of UL89-C highlighted a number of molecular interactions that contribute to binding affinity (Figure S2A). The pivotal regions of 1 binding the pUL89-C active site are the metal chelator group and the benzyl group. The three coplanar oxygen atoms of the diketo acid (DKA) moiety chelate Mn2+ ions in the active site, explaining the requirement for metal ions for drug binding. The central oxygen, O(H), bridges both metal ions, whereas two O (carbonyl) coordinates the metal ions (Figure S2A). Additionally, the terminal phenyl ring interacts with Asn536 (distance = 3.74 Å), Met579 (3.30 Å), and Thr537 (4.66 Å). These interactions could help position the inhibitor near the divalent metal cations in the active site. The five-membered oxadiazole moiety and isopropyl group are not directly involved into interactions with the UL89-C binding site (distance oxadiazole-Phe466 > 6.24 Å). On this basis, we designed 2 containing an α,γ-DKA motif linked to a 4-fluorophenyl ring (Figures 1 and 2A). The α,γ- DKA core, a privileged structure found in thousands of HIV integrase inhibitors with IC50 values in low micromolar concentrations,38−45 is able to chelate Mn2+46 (Figure S2A) and able to disrupt DNA−protein interactions.The α,γ-DKA motif has also been successfully used as a starting point to obtain inhibitors against (1) hepatitis C virus NS5B polymerase,50 (2) HIV-reverse transcriptase RNase H,51 (3) hepatitis C virus NS5b polymerase,52 and (4) influenza endonuclease.53 The 4-fluorophenyl ring is a common motif found in 1 and other inhibitors.

To validate our binding pose predictions and to facilitate further structural modifications, we determined the crystal structure of UL89-C bound with 2 (PDB code 6EY7). The elucidation of the binding mode of 2 in the UL89-C binding site by X-ray crystallography (see below and Figure 2A) and the finding of 2 able to inhibit UL89-C prompted us to enlarge the α,γ-DKA series (3−17, Schemes 1 and 2). The substituents onthe aromatic ring (3−4), amide (5), carboxylic acid (6−7), and benzylic (8−17) positions were studied to determine SAR of the series. The fluoride atom of 2 was substituted with iodine (3) and bromine (4). In 5, the methyl group of the amide portionwas exchanged for a bulkier alkyl substituent. Furthermore, the carboxylic functionality of 2 was converted to an amide (6−7). The 4-fluorobenzyl and the N-methyl-4-fluorobenzylamine groups of 2 were exchanged with a CH3 group to give 8 and 9, respectively. Different planar (hetero)aromatic rings, variously decorated, have been added to the α,γ-DKA chelatingcore (10−17) to promote a π−π stacking interaction with Phe466.On the basis of the remarkable profile shown by α,γ-DKA derivatives, 18, lacking the α,γ-DKA core required to achieve chelation, was used as the negative control. The synthesis of compounds 2−5 and 8 was straightforward and was accomplished in three reaction steps. In the first step, the primary amine (2c−5c, Scheme 1) was acetylated understandard conditions to afford the intermediate 2b−5b. Thepreparation of intermediates 2a−5a was achieved by treating 2b−5b with dimethyl oxalate. Then, 2a−5a were hydrolyzed to afford the corresponding acid derivatives (2−5).44,54,55 Compounds 6 and 7 were obtained by coupling methylamine and 4-fluorobenzylamine, respectively, with 2. Compounds 8−9 and 13 were prepared by basic hydrolysis of 8a−9a and 13a (Scheme 2).Finally, on the basis of the structure of a potent HIV integrase inhibitor MK-2048 (Figure S3A), we substituted the α,γ-DKA core with a β,δ-DKA chelating group (19−20 and Figure S3A). The synthesis of 19−20 (Scheme 3) started from the formation Scheme 3.

Reagents and Conditions: (a) HC(OEt)3, CO(NH2)2, Refluxing Xylenes, 4 h (57% Yield); (b) 23, EtBr, K2CO3, DMF, 80 °C, 14 h (34%); (c) 22 (1 equiv), KOH (1equiv), EtOH, rt, Overnight (65%); (d) 21, N-(4- Fluorobenzyl)-N-metilamine or N-4-Fluorobenzylamine, EDCI, HOAt, DCM, rt, Overnight (35−36%); and (e) LiOH 1 N, MeOH, 40 °C, 1 h (43−97%)of the dihydropyridone core (23), which was subsequently ethylated giving 22 and then partially hydrolyzed obtaining21.56−58 Afterward, 19b and 20b were obtained by coupling benzylamines with 21, using amide coupling reagents. Finally,ester derivatives 19b and 20b were converted to 19 and 20, respectively.UL89-C/2 Complex Structure. To study the binding mode of 2, we determined the 3D crystal structure of the UL89-C/2 complex (PDB code 6EY7). Diffraction data collected from crystals of UL89-C soaked with Mn2+ and 2 were solved by molecular replacement, using the solved inhibitor-free UL89-Cas the initial model.18 Strong electron density within the active site was observed adjacent to the Mn2+ ions (Figure S2B).The UL89-C/2 structure was refined to 2.9 Å resolution (Table 2). The UL89-C/2 crystal also showed anomalous scattering, confirming the presence of two Mn2+ ions in the active site. As expected, 2 bound through the α,γ-DKA core to the two Mn2+ ions in the active site (Figure 2B).

The three adjacent oxygen atoms of the α,γ-DKA fragment chelated the two Mn2+ ions in a pairwise fashion such that the central oxygen atom was bridging the two ions (Figure 2B). Thus, the Mn2+ ions were coordinated by two acidic residues (Glu534-Asp463 or Asp463-Asp651) from UL89-C and two oxygen atoms of 2 (Figure 2B) in a tetrahedral geometry. An identical coordination pattern has been observed between the active site of LACV L- protein and an α,γ-DKA derivative.59 All four compounds in the asymmetric unit of the UL89-C/2 crystals showed clear electron density for the orientation displayed in Figure 2A. The complex UL89-C/2 also revealed that the 4-fluorobenzyl group blocks a region crucial for DNA/UL89-C recognition. To further confirm the presence of 2 in the UL89-C/2 complex, we exchanged the fluorine atom for iodine (3) or bromine (4) atom because both are able to originate measurable anomalous scattering effects. Indeed, the anomalous difference maps confirmed the presence of the iodine atom of 3 in the UL89- C/3 complex (Table 2). The position of the iodine atom on the UL89-C/3 complex overlaps with the fluorine atom of 2 found in UL89-C/2 (Figure S2C). On the basis of the crystallographic studies on UL89-C/2 and UL89-C/3, the α,γ-DKA chelating core served as the metal-binding motif and the 4-fluorophenyl ring afforded additional interactions with the hydrophobic cleft formed by Asn536 and Met579 (Figure 2B,C).

The methyl group of 2, pointing outside the binding site, may be substituted with a bulky alkyl group to get an additional hydrophobic interaction with the close Phe466 (Figure 2B,C). The carboxylic acid interacted with Ser473 by hydrogen bonds with a distance of 3.6 Å (Figure 2C).Inhibition of UL89-C Nuclease Activity. A quantitative UL89-C nuclease assay was designed to evaluate the inhibition potency of 1−20. These compounds passed through a pan assay interference compounds filter (http://zinc15.docking.org/ patterns/home/).60Linear DNA and UL89-C were incubated at 37 °C for 30 min in the presence of 1−20 at a concentration of 50 μM. Linear DNA was run on an agarose gel electrophoresis gel, stained with Syto 60, and quantified by fluorescence analysis. The compounds revealed a percentage of inhibition ranging from totally inactive (0%) to almost complete inhibition (97%) (Table 1). At 50 μM, 1 and 2 showed similar inhibition (36.6 and 56.2%) but with an improvement of the percentage efficiency index (PEI) (defined as the percentage of inhibition at a given concentration/molecular weight) from 0.08 to 0.15 because of the lower molecular weight of 2 (Figure 1).Using a novel enzyme-linked inmunosorbent assay format as a screening assay, Wang et al. found that 1 inhibited pUL89-C activity with a IC50 of 96 μM. These results are in a good agreement with the inhibitory response in our assay.34The comparable activity of 1 and 2 corroborated our hypothesis that an α,γ-DKA scaffold linked to a 4-fluorophenyl group was the minimum motif required to modulate the UL89- C nuclease activity. We studied the importance of each individual group, α,γ-DKA and 4-fluorophenyl, in the contribution to the inhibition of the activity of UL89C by analyzing the activity of both fragments separately, 2b and 8 (Figure S3B), and we confirmed our hypothesis that they were unable to inhibit UL89-C activity (Table 1).

On the basis of the UL89-C/2 complex structure, several α,γ- DKA compounds decorated with different (hetero)aromatic groups spanning a wide range of size, aromaticity, and H-bond capability (3−17) were selected for further studies. The substitution of the fluorine atom of 2 for iodine (3) or bromine(4)slightly improved the activity (47.7 and 51.5%, respectively). The substitution of the methyl group of 2 for a cyclopropyl ring(5)showed an improvement of potency from 37.4 to 76.3%. Theimproved activity of 5 might be related to the additional interaction between the cyclopropyl fragment and Phe466. Derivatization of the carboxylic acid functionality to an amide (6 and 7) led to inactive compounds, confirming the crucial role of the electrostatic interaction between the carboxyl group with Mn2+. The replacement of the aromatic ring of 2 with a methyl (8 and 9, Scheme 2) resulted in a complete loss of activity as well, pointing to a possible role for the aromatic substituent in the recognition process. Altogether, these preliminary results suggested that an α,γ-DKA scaffold linked to a bulky aromatic substituent might contribute to UL89-C inhibition. Starting from this assumption, we were keen to examine the effects of different (hetero)aromatic substituents on the α,γ-DKA system. To this end, derivatives 10−17 (Scheme 2) were assayed.Derivatives bearing an electron-rich aromatic ring such as 4-methylphenyl (12), phenylpiperidine (14), or a heteroaromatic ring (17) exhibited high inhibitory activity (89.1−97.2%). Instead α,γ-DKA compounds bearing an electron-deficient aromatic ring such as a 4-fluorophenyl (11), 4-trifluoromethyl- phenyl (13), or 4-pyridyl (15) showed low inhibition at 50.0 μM. Structural changes introduced in the metal-chelating motif, as in 18, greatly affected the ability of the inhibitor to bind to the active site. On top of that, the alteration of the α,γ-DKA chelating core by an α,δ-DKA (19 and 20) made the compounds unable to chelate the metal ions in the active site.

To estimate the IC50 values, the concentrations of 1-5, 10, 12, and 14−17 were lowered to 25, 12.5, 6.2, and 3.1 μM, respectively, in the fluorescent UL89-C nuclease assay. At a lower concentration (25 μM), 1−4 showed low activity, meanwhile compound 5, bearing a cyclopropyl group on nitrogen atom, showed the highest activity between derivatives 1−5; this might be due to an additional hydrophobic interaction with Phe466. To note, the halogen atom (fluorine, iodine, or bromine) attached to the phenyl group (2−4) does influence the inhibitory potency of α,γ-DKA compounds to UL89-C. Among 10−17, we observed that α,γ-DKA compounds bearing a pyridine ring (15), a 4-fluorobenzene (11), or a 4- trifluoromethyl-benzene (13) have low inhibitory activity against UL89-C. This might be due to the substituent effect on the electrostatic potential of the benzene ring. A strong dependence on the electrostatic potential of the aromatic ring and UL89-C affinity is suggested. Compounds bearing a benzene ring decorated with substituents such as CF3 or F have shown to be less potent than compounds bearing electron- donating groups such as NR2 or CH.61,62 As expected, lowering further the concentration to 12.5 and 6.25 μM, 5, 10, 12, and 16 showed a reduced percentage of inhibition (22.6−32.7%, Table 1), whereas 14 and 17 still showed good potency against UL89- C (65 and 80%, respectively). The estimated IC50 values for 14 and 17 are 4.5 and 9.2 μM, respectively (Figure S4A,B). Furthermore, 14 showed a PEI of 0.35, four times greater than that of 1 (PEI = 0.08, Figure 1).

The high activity of 14 might be due, as predicted by docking simulation, to a strong π−π interaction between Phe466 and the electron-rich aromatic ring of 14 (Figure 2D). From the focused library of α,γ-DKA, we identified that the minimal structural determinant for the UL89- C activity is an α,γ-DKA group linked to an electron-rich aromatic ring (Figure S15).Thermal Shift Assay. Protein melting temperature (Tm)determined by thermal shift assays (TSAs) is used to monitor ligand effects on protein stability.63 The interaction of a ligand with the protein usually causes an increase in Tm. TSA was used to study the thermal stabilization of UL89-C upon binding of 14,17, and 18, and the latter was used as a negative control (Figure 3). Tm of ligand-free UL89-C was 52.7 °C (in the presence ofdimethyl sulfoxide (DMSO), Tm was 52.0 °C). A significant rise in Tm was observed for 14 (Tm = 55.7 °C) and 17 (Tm = 55.5°C), confirming their ability to bind UL89-C and stabilizing the protein. 18 did not stabilize the protein (Tm = 52.5 °C, Figure 3), thus confirming lack of binding to UL89-C. These results further suggested that the α,γ-DKA core was a crucial motif required for UL89-C inhibition.

Molecular docking studies were conducted to elucidate the binding mode of UL89-C inhibitors in the active site for the noncrystallized complexes. The atomic coordinates of the UL89-C active site were used for computational docking of 1, 2, 14, and 17 (Figures S2A, 2D, and S2D). The binding pose of 1 showed that (i) the chelating motif was positioned close to the two Mn2+ cations, (ii) the 4-fluorobenzyl substituent interacted by hydrophobic interaction with Asn536 and Met579, and (iii) the oxadiazole fragment is not directly involved in the interactions with the UL89-C binding site. The predicted binding pose of 2 showed the same orientation of groups i and ii as 1. The root-mean-square deviation (rmsd) between the docked and crystallographic poses of 2 on the UL89-C active site is 0.9 Å. Molecular docking simulation of 14 and 17 demonstrated that the α,γ-DKA group chelated the two Mn2+ cations and the electron-rich aromatic ring interacted with the adjacent Phe466 by π−π interaction (Figures 2D and S2D).

CONCLUSIONS
UL89 HCMV terminase is a promising antiviral target because of its key role in the processing and packaging of the viral DNA inside the capsid, during viral replication. Blocking this crucial step could avoid the spread of the infection. We have focused our efforts on developing potential inhibitors against UL89-C because its nuclease activity is exclusive to the virus. On the basis of the structure of 1, a weak inhibitor of UL89-C, we designed 2. Compound 2, bearing an α,γ-DKA scaffold linked to a 4-fluorophenyl group, is able to inhibit UL89-C activity. Here, we report the X-ray structure of UL89-C bound to two α,γ-DKA derivatives (2 and 3) as well as the in vitro thermal shift and nuclease assays. Our studies showed that the alterations of the α,γ-DKA scaffold proved to be deleterious to the activity and the potency was influenced by substituents on the α,γ-DKA core. Among them, 14, bearing an α,γ-DKA group linked to an aromatic electron-rich moiety, inhibited UL89-C at low micromolar range. The molecular binding of 14 to UL89-C was confirmed by a fluorescence-based thermal stability assay and its binding interactions were predicted by molecular docking simulation. Considering the strong correlation found with UL89 inhibitors against UL89 enzyme and antiviral potency,34,35 14 is an optimal candidate to be tested in HCMV-infected cells. Moreover, several DKA derivatives have exhibited low cytotoxicity in vitro64−69 so we expect that 14 would share this feature in vitro as well. The structural information gained from the UL89-C/2 complex and the identification of the chemical features determining the UL89-C inhibition will allow further development and optimization of α,γ-DKA inhibitors against the HCMV UL89-C.

Synthesis of Inhibitors. Synthetic and chemical character- ization for all compounds can be found in the Supporting Information General Chemistry Materials and Methods section. X-ray Crystallography. UL89-C was purified and crystal- lized as previously described.18 Crystals of UL89-C were grown by mixing the protein solution with a solution containing 10% Mes 1 M pH 6.5, 6% PEG8000, and 150 mM calcium acetate in sitting drop plates at 20 °C for 1 week. The crystal complex between the UL89-C and the inhibitor was formed by soaking the protein crystals in a solution containing 1 mM of 2 or 3, 2 mM of MnCl2, 10% Mes 1 M, 6% PEG8000, and 150 mM calcium acetate. 2 was soaked for one week, and 3 was soaked for 24 h. All crystals were cryoprotected in reservoir buffer containing 1 mM of 2 in 10% PEG400 or 3 in 20% glycerol and flash-cooled in liquid nitrogen prior to diffraction analysis. Diffraction data were recorded from cryocooled crystals (100 K) at the ALBA synchrotron in Barcelona (BL13-XALOC beam- line). Data were integrated and merged using XDS70 and scaled using XSCALE.71 The structure of the complexes was solved by molecular replacement with the CCP4 supported program MOLREP72,73 using the UL89-C structure (PDB code 3N4P) as a starting model and refined with Refmac74 and manually checked with Coot.75 2 was placed (LigandFit, PHENIX supported program)76 and refined (phenix.refine, PHENIX supported program)77 in the additional electron density present at the active site pocket. For the UL89-C/3 complex, anomalous data were collected at wavelength 0.9194 Å and the difference anomalous map was calculated (Figure S2C). Molecular topologies for 2 and 3 were generated with PRODRG server.

Data collection and refinement statistics are shown in Table 2. In Vitro Nuclease Assay. Purified UL89-C at a concen- tration of 3 μM dissolved in a buffer containing 30 mM Tris pH 8.0 and 50 mM NaCl was mixed with 3 mM MnCl2 and different concentrations of inhibitors 1−20 (50−3.2 μM) in DMSO. The concentration of DMSO was 5% in all reactions. The nuclease activity was started by addition of 100 ng of linear pOPINM vector, digested with KpnI−HindIII. The reaction was incubated for 30 min at 37 °C. After that time, the reaction was stopped by addition of ethylenediaminetetraacetic acid at a final concentration of 30 mM. Then, 0.2 μM of Syto 60 stain in a dye-free loading buffer was added. The samples were incubated for 10 min and loaded on an agarose gel (1%). Gel electrophoresis was run at 90 V for 45 min in the dark in 1× TAE buffer. The image of the agarose gel was obtained from the LI-COR Odyssey Infrared Imaging equipment. The non- digested DNA in the reaction was quantified using Odyssey Infrared Imaging system version 3.0 software, using a control DNA for the measurements. The percentage of inhibition was obtained from the amount of DNA present at different concentrations of inhibitors. Thermal Shift Assay. The aliquots were prepared on ice in a 96-well polymerase chain reaction (PCR) plate (Bio-Rad) sealed with optically clear PCR sealers (Bio-Rad) and contained 15 μM UL89-C, 30 mM Tris-HCl (pH 8.0), Inhibitor Library 50 mM NaCl, Sypro Orange protein (Invitrogen), and 0.5 mM MnCl2. Then, 14, 17, and 18 dissolved in DMSO were added at a final concentration of 25 μM. For the control, the same volume of DMSO was added to reach a final concentration of 5%. The samples were heated from 20 to 95 °C at a rate of 0.5 °C/min, and the change in fluorescence (λex = 490 nm; λem = 575 nm) was monitored over time in an iCycler iQ Real Time PCR Detection system (Bio-Rad). All measurements were carried out in triplicate. Data evaluation and melting point determination were performed using Bio-Rad Optical System Software.