Discovery and design of DNA and RNA ligase inhibitors in infectious microorganisms

Abstract

Background

Members of the nucleotidyltransferase superfamily known as DNA and RNA ligases carry out the enzymatic process of polynucleotide ligation. These guardians of genomic integrity share a three-step ligation mechanism, as well as common core structural elements. Both DNA and RNA ligases have experienced a surge of recent interest as chemotherapeutic targets for the treatment of a range of diseases, including bacterial infection, cancer, and the diseases caused by the protozoan parasites known as trypanosomes.

Objective

In this review, we will focus on efforts targeting pathogenic microorganisms; specifically, bacterial NAD⁺-dependent DNA ligases, which are promising broad-spectrum antibiotic targets, and ATP-dependent RNA editing ligases from Trypanosoma brucei, the species responsible for the devastating neurodegenerative disease, African sleeping sickness.

Conclusion

High quality crystal structures of both NAD⁺-dependent DNA ligase and the Trypanosoma brucei RNA editing ligase have facilitated the development of a number of promising leads. For both targets, further progress will require surmounting permeability issues and improving selectivity and affinity.

1. Introduction

The enzymatic process of polynucleotide ligation is carried out by RNA and DNA ligases, structurally and evolutionarily related members of the nucleotidyltransferase (NTR) superfamily ¹, which also includes eukaryotic GTP-dependent mRNA capping enzymes ¹. Ligase superfamily members share a conserved chemical mechanism (Fig. 1), which utilizes a nucleotide cofactor, either ATP or NAD⁺, to furnish an AMP moiety. In the first reaction step, attack of a conserved lysine nucleophile displaces either pyrophosphate (PPi) or nicotinamide mononucleotide (NMN), and forms a phosphoramidate bond between the lysine ε-nitrogen and the AMP moiety. In the second step, a nicked double-stranded polynucleotide binds, and the AMP is transferred to the 5′PO₄ end of the nick where it is connected via a phosphoanhydride linkage. The reaction is completed when the nicked 3′OH end attacks the phosphoanhydride, displaces AMP and joins the nicked-polynucleotide ends. In addition to mechanistic conservation, superfamily members share recognizably conserved structural elements ¹. Prominent among these is a nucleotide-binding domain (NTBD), which comprises the fundamental catalytic unit of the superfamily and may be an example of a stand-alone ancestral enzyme from which contemporary superfamily members evolved ^{1, 2}. Domain modules, which impart polynucleotide specificity, are covalently linked to both the N- and C-terminals of the NTBD, in a clamp-like arrangement, by peptide tethers. By virtue of these tethers, the clamp-like architecture is quite flexible ^{3, 4 5, 6, 7}, and progression through the catalytic cycle is facilitated by clamp opening and closing, which allows substrate ingress and product release ^{3, 5, 6}. One notable exception is the ATP-dependent RNA editing ligase from the pathogenic organism Trypanosoma brucei. In this organism, which will be discussed in greater detail in subsequent sections, modular domains associate non-covalently to the NTBD ⁸. Despite this difference, movement through the catalytic cycle is also thought to coincide with domain opening and closing ⁸.

Figure 1. Three-step nick joining mechanism by DNA and RNA ligases.

In step one, the active-site lysine attacks the alpha phosphate of NAD⁺ or ATP, displacing R (where R=PP_i in ATP-dependent DNA and RNA ligases and R=NMN in NAD⁺-dependent DNA ligases) and forming an enzyme-AMP intermediate. Following polynucleotide binding, the nicked 5′PO₄²⁻ attacks the enzyme-AMP intermediate, displacing the active-site lysine, forming a D(R)NA-AMP intermediate in step two. In step three, the nicked 3′OH attacks the D(R)NA-AMP intermediate, displacing AMP and joining the nicked polynucleotide strand.

In the repertoire of biochemical reactions, polynucleotide ligation, catalyzed by members of the NTR superfamily, is vitally important in a wide spectrum of cellular processes. For example, the 5′ to 3′ directionality of DNA replication results in a lagging strand whose synthesis is fragmented into discontinuous blocks, called Okazaki fragments. Lagging strand synthesis is completed when a DNA ligase joins the 5′PO₄ and 3′OH ends of apposing fragments. In addition to normal cellular process like DNA replication, the integrity of DNA may be interrupted by various exogenous assaults that result in oxidative damage ⁹, as well as alkylation ¹⁰ and deamination ¹¹. In order to preserve genomic integrity in the face of these assaults, cells have evolved various repair strategies, like long and short patch base excision repair ^{12, 13}, and nonhomologous end joining ^{14, 15}. Among other factors, the efficacy of these repair mechanisms is critically dependent upon nicked polynucleotide joining, which is generally the last step in these pathways ^16–18. Polynucleotide ligation is not, however, limited to nicked DNA, it is also an essential chemical reaction in the repair ¹⁹, splicing ²⁰ and editing processes of RNA ²¹. For example, following the post-transcriptional insertion or removal of polyuridylate tracts to, or from, nascent mitochondrial transcripts of the genus Trypanosoma ²², one of two RNA editing ligases joins the nicked ends of the mRNA, completing the RNA editing process ²³.

In light of the fundamental importance of polynucleotide ligation, it is not surprising that both DNA and RNA ligases have experienced a surge of recent interest as chemotherapeutic targets for the treatment of a range of disease, including bacterial infection ^{24, 25}, cancer ²⁶, and African sleeping sickness ²⁷. It is the purpose of this review to detail these recent advances. Specifically, we limit our attention to efforts targeting pathogenic microorganisms. In particular, we focus on bacterial NAD⁺-dependent DNA ligases, which are promising broad-spectrum antibiotic targets, and ATP-dependent RNA editing ligases from Trypanosoma brucei, the species responsible for the devastating neurodegenerative disease, African sleeping sickness. The defining structural features of each of these inhibitor targets will be discussed in turn, followed by discussion of the inhibitors reported to date for each target, paying particular attention to inhibitor development and subsequent in vitro and in vivo characteristics. The review will conclude with possible future directions.

2. NAD⁺-dependent DNA Ligases

DNA ligases were initially characterized by several labs in a flurry of activity spanning the years 1967 and 1968 ^{28, 29,30–33} as reviewed in ³⁴. DNA ligases are not only sentries of genomic integrity, participating in vital DNA repair pathways ^12–15, they also play a fundamental role during genetic recombination, a process that occurs during both meosis in eukaryotes, and V(D)J shuffling, which produces the diverse antibody specificity central to an effective immune response in vertebrates ³⁵. Furthermore, DNA ligases are essential in genome replication ³⁶. There are two principal subfamiles of DNA ligases whose members can be recognized according to whether they require an ATP or NAD⁺ nucleotide during the ligation reaction.

ATP-dependent DNA ligases span the greatest range of biological diversity. They have been characterized in all three kingdoms of life, as well as in bacterial and eukaryotic viruses ². NAD⁺-dependent DNA ligases, on the other hand, have nearly the same phylogenic distribution but are noticeably absent in Eukarya ². Also known as LigAs after the name of the E. coli gene that encodes them, NAD⁺-dependent DNA ligases are essential in Escherichia coli ³⁷ as well as several gram-positive and gram-negative organisms, including: Salmonella typhimurium ³⁸, Haemophilus influenza ³⁹, Bacillus subtilis ⁴⁰, Streptococcus pneumonia ⁴¹, Staphylococcus auerus ⁴² and Mycobacterium tuberculosis ^{43, 44}. The absence in eukaryotic organisms and essential function in bacteria make LigA an attractive target for the development of novel, broad-spectrum antibiotics. Indeed, several classes of LigA specific inhibitors have been reported. Before reviewing each inhibitor class, we will briefly detail the structural features of the ATP-dependent and NAD⁺-dependent DNA ligases. We will pay particular attention to their shared and unique structural elements, which in many cases have allowed LigA specific inhibitor development.

2.1 Defining structural features

Both the ATP-dependent and NAD⁺-dependent DNA ligases share a modular domain architecture forming a clamp-like structure (Fig. 2) whose opening and closing facilitates progression through the catalytic cycle ^{1, 3, 45}. A minimal catalytic core, which consists of the NTBD, and the oligonucleotide-binding domain (OBD) form the fundamental structural unit of the enzyme clamp. As the common structural feature of all NTR superfamily members, the NTBD houses features essential for nucleotide recognition and chemistry. For example, five of the six motifs, which are conserved across all members of the NTR superfamily¹ form a deep beta sheet cleft, which sandwiches the AMP between conserved aromatic and hydrophobic residues ^{1, 6, 46–50}. In both bacterial and ATP-dependent DNA ligases, AMP recognition occurs in the deep end of the binding pocket where it interacts with the protein backbone or side-chains, and the exocyclic-adenine substituents ^{6, 47, 48}. Following nucleotide recognition and binding, the catalytic lysine, contained in motif I (KXDG), attacks the nucleotide and forms the enzyme-adenylate intermediate. In ATP-dependent DNA ligases, the formation of the enzyme-adenylate intermediate is critically dependent on the OBD⁵¹. The OB fold, which defines the OBD, consists of a five-stranded anti-parallel beta-barrel and contains a sixth conserved motif, which is thought to position the triphosphate leaving group inline with the attacking lysine nucleophile⁶. In contrast, in NAD⁺-dependent DNA ligases, the OBD lacks conserved motif VI and is dispensable for enzyme-adenylate formation ^{52, 53}. Instead, a unique bacterial DNA ligase domain, domain Ia (Fig. 2), comprised of two helices that form a NMN binding pocket, is required. Crystal structures imply that domain Ia orients the NMN leaving group in a manner functionally homologous to the role played by the OBD in ATP-dependent DNA ligases⁵. Its essential function and uniqueness to bacterial LigAs make domain Ia an appealing structural feature to target during rational inhibitor design.

Figure 2. Modular domain architecture in DNA and RNA ligases.

A), B) and C) are ribbon diagrams of Human DNA ligase I (PDB ID 1X9N), E. coli DNA ligase (PDB ID 2OWO), and T. brucei RNA-editing ligase 1 (REL1) (PDB ID 1XDN). Each is shown in complex with their nicked polynucleotide substrates and homologous domains are indicated by color. The clamp-like domain arrangement that facilitates strand joining is clearly seen in (A) and (B). As discussed in the text, a second domain associates non-covalently with the NTBD to form the clamp in REL1. The enzyme-RNA complex in (C) is a homology model as described in Ref. 93. The nucleotide binding domain (NTBD) is shown in blue in (A), (B), and (C). In (B) and (C), the oligonucleotide binding domain (OBD) is shown in red. In (A) the DNA-binding domain (DBD) is shown in yellow, while the homologous helix-hairpin-helix domain (HhH) is shown in yellow in (B). Additional domains, unique to bacterial NAD⁺-dependent DNA ligases are shown in (B): domain Ia is colored magenta, while the structural-zinc-finger is colored green.

Following enzyme-adenylate formation in both bacterial and ATP-dependent DNA ligases, it is thought that large-scale isomerization events occur that move domain Ia and the OBD away from their respective NTBDs, a view consistent with available structural data ^{4, 5}. The motion opens the clamp and allows nicked, double-stranded DNA to bind. After DNA binding has occurred, the clamp shuts again and ligation may proceed. Prior to strand joining, clamp closure in both bacterial and ATP-dependent DNA ligases deforms their DNA substrates, imposing a RNA-like A conformation on nucleotides local to the nick ^{4, 45, 54, 55}. The distortion positions the opposing nicked ends such that nucleophilic attack may occur and seems to be a critical prerequisite to strand joining ^{45, 54–56}. DNA distortion is brought about by a set of homologous protein-substrate interactions as illustrated by a comparison of the crystal structures of both HuLig I and LigA in complex with the adenylated DNA substrate ^{54, 56}. In each structure, the NTBD and OBD engage the minor groove local to the nick, forming two sides to a substrate clamp that enforces the RNA A-like conformation local to the nick (Fig. 2). The key difference between the eukaryotic and bacterial crystal structures is the identity of the domain that forms the clamp base. In ATP-dependent DNA ligases, an alpha helical domain, called the DNA-binding domain (DBD), is covalently linked to the N-terminal of the NTBD and forms the clamp base ⁵⁴. The clamp base in NAD⁺-dependent DNA ligases, on the other hand, is formed by a set of four helix-hairpin-helix (HhH) motifs, which are linked to the N-terminal of the NTBD via a structural zinc-binding domain ^{50, 56}. Despite having different folds, the HhH domain and the DBD play analogous roles in strand joining, conveying the major DNA binding activity in representative members of each ligase class ^{45, 54, 57}.

While the domains that form the clamp work cooperatively to efficiently join nicked DNA strands in vitro ^{53, 54} additional clamp-C-terminal-domain extensions regulate activity in vivo. For example, in mammalian DNA ligase I, a nuclear localization signal and a PCNA binding motif extend from the N-terminal of the DBD and coordinate interactions with the DNA replication fork ^{36, 58}. The other two mammalian isoforms have modulating domains of similar function ⁵⁸. In bacterial LigAs, a highly mobile BRCT domain can be found extending from the C-terminal domain of the HhH domain. The structure of this domain has not been resolved by crystallography but has been solved independently in solution using NMR spectroscopy (PDB ID 1I7b). However, its importance in nicked strand joining remains a point of contention among scientists ^{52, 53, 57, 59, 60} and until a consensus has been reached, it is unlikely to command interest as a target receptor in rational inhibitor design.

2.2 Inhibitors

Although the BRCT domain is unlikely to serve as an inhibitor-binding site, Nandakumar, Nair and Shuman pointed out an exceptional opportunity for rational inhibitor design within the NTBD domain ⁵⁶. In the crystal structure of the step-two adenylate-DNA intermediate (2OWO), they note the presence of a hydrophobic tunnel extending from the exterior of the protein into the adenine-binding pocket of the NTBD domain where the N1, C2 and N3 edge of adenine is exposed ^{2, 56} (Fig. 3A). Although this pocket is present in all available LigA structures, it is absent in human DNA ligase I (HsLigI), where the space is blocked by a loop connecting motifs IV and V ^{54, 56}. As they note, inhibitors derived from C-2 substituted adenosine, AMP, or NAD⁺ could prove advantageous by simultaneously utilizing native substrate contacts, while tuning LigA specificity by adding alkyl or aryl substituents strategically positioned to occupy the hydrophobic tunnel ⁵⁶. Recently, this design strategy has been proven viable. The results of a recently reported high throughput assay describe 2-methylthio derivatives of ADP and ATP that inhibit E. coli LigA with IC₅₀ values of 0.5 and 2.1 μM respectively ⁶¹. While structural data illuminating the binding modes of these 2-methylthio derivatives was not reported, it seems likely that the bound state position of the ADP and ATP aromatic adenines overlap with that of the AMP-DNA intermediate adenine such that the 2-methylthio moieties extend into the hydrophobic tunnel. The plausibility that the positions of the bound state aromatic moieties overlap is bolstered by crystal structures of LigA bound to a pyridochromanone derivative, and three pyridopyrimidine derivatives, which are discussed in the next section.

Figure 3. Druggable cavities in the bacterial NAD⁺-dependent DNA ligase and the T. brucei ATP-dependent RNA editing ligase 1.

(A) The hydrophobic tunnel adjacent to the NAD⁺-binding pocket. AMP is shown occupying the pocket. Ribose and the N1, C2, N3 adenine atoms (from right to left) are visible. The surface representation of the protein is colored by residue type: blue-basic, red-acidic, green-polar, and white-hydrophobic. The image was constructed using PDB ID 2OWO. (B) Compound 2 shown protruding from the hydrophobic tunnel, PDB ID 3BA9. Protein coloring is as in (A). (C) Structures of pyridochromonone (1) (PDB ID 3BA8), and the three pryidopyrmidine compounds (2, 3, 4) (PDB IDs: 3BA9, 3BAA, 3BAB) which make use of the hydrophobic tunnel. The dashed ovals encompass the portion of the molecule that extends into the tunnel; the remainder of the molecule overlaps with the NAD⁺ binding pocket. Bold numbers are referenced in the text and tables 1 and 2. (D) The deep ATP binding pocket in T. brucei RNA editing ligase 1 is shown, coloring the same as in (A).

2.2.1 Pyridochromanones, Pyridopyrmidines

Preceding a forthcoming publication, Pinko and coworkers recently deposited four crystal structures into the PDB: 3BA8, which shows the LigA bound state of a fluorinated pyridochromanone derivative, (compound 1, Fig. 4) and a set of 3 pyridopyrimidine derivatives 2, 3, 4 (3BA9, 3BAA, 3BAB). In each crystal structure, the aromatic scaffold of the inhibitor occupies the binding site in a manner similar to the position taken by the DNA-adenylate adenine, illustrated by the crystal structure 2OWO. Furthermore, both the pyridochromanone and the pyridopyrimidine derivatives make use of the hydrophobic tunnel ² (Fig. 3B & 3C).

Figure 4.

Inhibitors of NAD+ dependent DNA ligases.

The occupancy of the fluorinated pyridochromanone derivative in the NAD⁺-adenine-binding pocket is consistent with the experimental characterization of a set of pyridochromanone derivatives carried out by Brötz-Oesterhelt et al ⁶². These compounds were identified in a high throughput screen and inhibited LigA via a NAD⁺ competitive binding mechanism. In two separate assays, carried out using both E. coli and S. pneumoniae LigA, IC₅₀ values ranged from 40 to 100 nM. Interestingly, compound 1, crystallized by Pinko et al., showed the best in vitro inhibition characteristics with an IC₅₀ value of 40 nM (Table 1). When compared to HsLigI inhibition effects, all of the pyridochromanone derivatives showed a roughly 1000-fold specificity toward LigA. Additionally, the pydridochromanone derivatives were active in vivo, but primarily against gram-positive bacteria. While MIC values reported for the gram-positive bacteria S. aureus and B. subtilis were in the low μg mL⁻¹ range, gram-negative E. coli, were unaffected until treated by the permeablizing agent polymycin B nonapeptide; this implies that gram-negative outer membrane permeability is low. Consequently, increasing membrane permeability while maintaining binding affinity and selectivity must be achieved before members of the current pyridochromanone derivative set can find use as broad-spectrum antibiotics. Nevertheless, in their current state, they boast nanomolar IC₅₀ values and bind LigA with 1000-fold selectivity compared to HsLigI, promising characteristics in a novel gram-positive antibiotic lead compound.

Table 1.

Compound no.	LigA IC50, (μM)	† HsLig1 IC50, (μM)	NAD+ competitive?	MIC (μg mL−1)
1	0.04	>75	Yes	*6
2–4	N/A	N/A	N/A	N/A
5	0.5	>100 (μg/mL)	Yes	64
6	53	>720 (T4)	No	2.5×104
7	63	>2000 (T4)	No	N/A
8	2.6 (n=0)	16 (n=0)	No	2.3×104
9	10.2 (n=0)	49 (n = 0)	No	N/A
10	46.2	320	Yes	*0.2
11	260	72	N/A	N/A
12	11.4	27	N/A	N/A
13	85	380	Yes	*0.1
14	225	94	N/A	N/A
15	13.5	212	Yes	*7

^†IC₅₀ values determined using the T4-phage ligase in the the place of HsLig1 are noted: (T4).

^*The membrane permeablyzing agent polymxin nonapeptide was added to the gram-negative bacterial growth media.

Pyridopyrmidine derivatives, similar to those crystallized in complex with LigA, have been reported as possible chemotherapies spanning a diverse range of afflictions from possible anti-diarrheal agents to bacterial drug efflux pumps ^63–67. Despite the diversity of their potential applications, a thorough characterization of their antibacterial potential has not been published, yet the LigA-bound crystal structures show promising features. In addition to aromatic moieties that overlap with the NAD⁺ adenine, each makes use of the hydrophobic tunnel adjacent to the NAD⁺ binding pocket (Fig. 3A, 3B & 3C), two features that lend themselves to competitive binding with NAD⁺ and selectivity toward LigA.

2.2.2 Pyrimidopyrmidines

Similar in structure to the pyridopyrimidine derivatives, a pyrimidopyrimidine derivative, 2,4-diamino-7-dimethylamino-pyrimido[4,5-d]pyrimidine, or DDPP, 5, was recently reported as a LigA inhibitor by researchers at Eli Lilly (Fig. 4). DDPP inhibits the Streptococcus pneumonia LigA with an IC₅₀ of 0.5 μM by competitively binding with NAD+ 68 (Table 1). Selectivity was determined by comparing LigA DDPP IC₅₀ values to those measured for T4-phage ligase and HsLigI. DDPP demonstrated 600 and 1000 fold selectivity toward LigA when compared to the phage and human ligase, respectively. Additionally, DDPP, like the pyridochromanone derivatives, demonstrated in vivo activity. MIC values reported for S. pneumoniae, S. aureus, H. influenza were 64, 128 and 128 μg mL⁻¹, respectively, indicating that the compound is effective against both gram-positive and gram-negative human pathogens ⁶⁸. As further validation of the compounds in vivo activity, S. pneumoniae was grown in the presence of tritium labeled thymidine both with and without DDPP. Incorporation of labeled thymidine into DNA was significantly reduced in the presence of DDPP. Similar incorporation inhibition was observed with novobiocin, a known DNA replication inhibitor ⁶⁸, consistent with the role of DDPP as a DNA replication inhibitor. While crystal structures of the LigA bound DDPP could not be obtained ⁶⁸, the similarity of the DDPP and pyridopyrimidine molecular scaffolds (compare 2, 3 and 4, to 5) makes it likely that DDPP shares a binding mode similar to that observed for the pyridopyrimidines. Furthermore, while the IC₅₀ of DDPP is roughly an order of magnitude greater than that of the pyridochromanone, DDDP demonstrates similar selectivity and better membrane permeability, as inferred by lower DDPP MIC values in gram-negative bacteria. Without the need to overcome membrane permeability issues, pyrimidopyrimidine derivatives, like DDPP, may serve as better leads to broad-spectrum antibiotics.

2.2.3 Arylamino compounds

While pyridochromanones, pyridopyrimidines, and pyrimidopyrmidines all bind competitively with NAD⁺, a third class of compounds exerts their inhibitory effects via a non-competitive mechanism. Arylamino compounds, including derivatives of chloroquine, a well-known anti-malarial agent suffering from emerging resistance issues ⁶⁹, and bisquinoline, a potential scaffold for new anti-malarial agents ^70–72, were the first compound class reported that specifically target LigAs ⁷³. Chloroquine, 6, and hydroxychloroquine, 7, both quinoline derivatives, showed IC₅₀ values of 53 and 63 μM respectively, with order of magnitude selectivity toward LigA when inhibitor activity was compared to that for HsLigI or the homologous ATP-dependent DNA ligase from T4 phage (Fig. 4 and Table 1). The C-6 (8) and C-8 (9) bisamide-linked bisquinolines, whose quinoline groups were substituted by C-4 diamino moieties, exhibit IC₅₀ values ranging from 2.6 to 10.2 μM, an improvement over the chloroquines. While various lengths of the bisamide linker were tested (n = 0, 6, 8 for 8; n = 0, 4 for 9), the inhibitory effects showed little variability, changing by no more than 0.8 μM, an amount only slightly larger than the reported experimental error. However, IC₅₀ values improved by roughly five-fold when the bisamide linker spanned the quinoline C-6 positions, rather than the C-8 positions. For example, compound 8, with n = 0, has a reported IC₅₀ of 2.6μM, while compound 9, with n = 0, has a reported IC₅₀ of 10.2μM. Inspection of the structures of chloroquine, hydroxychloroquine and the assayed bisquinoline derivatives shows a conserved diamino group extending from C-4 of the quinoline moieties. This diamino group was deemed an important mediator of inhibitor binding. When a chloro group was substituted for the diamino group at the C-4 position, a 1000-fold increase in measured IC₅₀ values were reported in strand joining assays. Without a crystal structure to confirm, one might speculate that the quinoline derivatives make use of the NAD⁺ binding site and compete with NAD⁺ binding. Despite this, and though a detailed kinetic analysis is absent, evidence suggests that quinoline derivatives exert the majority of their effects through DNA interactions ⁷³.

The arylamino compounds may suffer from poor membrane permeability, an issue that would hinder their development as broad-spectrum antibiotics. For example, based on in vitro IC₅₀ values, the in vivo activity was considerably less than expected. 25 mg mL⁻¹ of chloroquine were required to reduce log-phase growth of Salmonella typhimurium by ~50%, a concentration approximately three orders of magnitude larger than required for comparable in vitro inhibition (IC₅₀ = 53 μM or 27 μg mL⁻¹). One explanation for this, as the authors point out, is the weakly basic nature of the quinoline derivatives, which results in partial protonation at pH 7 ^{73, 74}. Charged compounds have reduced hydrophobicity and ability to partition into the outer membrane of gram-negative bacteria, which in turn may lead to the higher observed concentration requirements. While it is possible that membrane permeability could be improved with appropriate structural modifications, the majority of the in vitro activity would still likely be due to non-specific DNA interactions, which, at least until a ternary inhibitor-DNA-ligase structure is solved, eliminates structure-based inhibitor design as a viable alternative and increases the difficulty of the inhibitor refinement process.

2.2.4 Bisglycosyl diamines

Another class of compounds that shows potential to carry forward in the drug discovery pipeline is the bisglycosyl diamine derivatives, as reported by Srivastava and coworkers ^{75, 76}. Using in silico screening of an in-house compound database, two classes of bis-xylofuranosylated diamines were identified: those with variable length diamino alkyl spacers, e.g. compound 10, and those with phenylene carbomyl spacers, compounds 13 and 14 (Fig. 4). Diamino alkyl spacers of 3, 10 and 12 methylene units (compounds 10, 11, and 12) were synthesized and tested. The two phenylene carbomyl variants tested, 13 and 14, had fixed spacer lengths, but different oxidation states. Compounds 10, 11 and 12 were assayed in vitro for inhibitory effects and LigA selectivity using recombinant LigA from M. tuberculosis (MtuLigA). IC₅₀ values of 46.2, 260 and 11.4 μM, respectively, were reported for compound 10, 11, and 12 (Table 1). MtuLigA selectivity, while present, was modest. When compared to HsLigI, compounds 10, 11 and 12 showed 7.5, 2.4 and 4.5 fold selectivity, respectively, toward MtuLigA. In order to gain insight into the reported IC₅₀ values, AutoDock was used to dock each of the three compounds into the LigA crystal from M. tuberculosis (Mtu) ⁷⁵. The results demonstrated that the binding modes of compounds 10 and 12 were largely contained in a wire-mesh surface demarcating the binding volume of NAD⁺. On the other hand, the majority of compound 11 resided outside of the NAD⁺ binding volume ⁷⁶. A similar phenomenon was observed with the two tested variants of the phenyelene carbamoyl linked bis-xylofuranosylated diamino compounds, 13 and 14. Compound 13, which made the most effective use of the NAD⁺ binding volume, had an IC₅₀ value roughly 3 times lower than compound 14, whose binding mode showed little overlap with the NAD⁺ binding volume, i.e. IC₅₀ values of 85 and 225 μM, respectively. When MtuLigA and HsLigI inhibitory efficacy were compared, compound 13 selectively favored MtuLigA inhibition by roughly 4 fold. Interestingly, while compounds 10, 12 and 13 each made effective use of the NAD⁺ binding volume, none took advantage of the adjacent hydrophobic tunnel (Fig. 3A). The overlap of the AutoDock predicted inhibitor binding modes and the binding mode of NAD⁺ imply that the inhibitors bind competitively with NAD⁺. Kinetic analysis confirmed these predictions, indicating that compounds 10 and 13, the only two compounds tested, both bind LigA in a NAD⁺ competitive mechanism. Moreover, ethidium bromide displacement assays showed that these compounds did not intercalate DNA. While only a crystal structure can confirm the predicted detailed atomic inhibitor-receptor interactions, when considered collectively, the kinetic analysis and the ethidium bromide displacement assays support the AutoDock predicted overlap of the inhibitor and NAD⁺ binding volumes. In order to test whether the in vitro efficacy of the compounds held in vivo, MIC values were determined in E. coli and S. typhimurim. While MIC values were found to be in the low μg mL⁻¹ for both compounds 10 and 13, again the only two compounds tested, the membrane permeablyzing agent polymxin nonapeptide was used, indicating poor membrane permeability. In summary, while compound 10, has an IC₅₀ value approximately four times that of compound 12, it is a more effective inhibitor than compound 13. Furthermore, compound 10 shows greater LigA selectivity than both compounds 12 and 13. Those observations may make it the most promising broad-spectrum-antibiotic lead presented.

2.2.5 N-substituted tetracyclic indoles

In a more recent work by Srivastava and coworkers, in silico screens identified roughly half a dozen N-substituted tetracyclic indoles ⁵³. The initial set of in silico compounds were sorted for LigA selectivity by testing at high concentrations against MtuLigA and T4Lig. The high concentration selectivity screen resulted in the identification of compound 15, which was further experimentally characterized (Fig. 4). When joining assays were carried out in the presence of increasing concentrations of compound 15 using both recombinant MtuLigA and HsLigI, IC₅₀ values for MtuLigA were 13.5, and those for HsLigI were 204.0, a roughly 15 fold selectivity for LigA (Table 1). While extensive attempts at crystallizing the compound proved unsuccessful, independent docking studies performed using four different software packages (AutoDock, Gold, FlexX, and Affinity) were used to predict the compounds binding mode. The docking studies resulted in a consensus orientation of the compound, in which the tetracyclic indole occupied the adenine-binding cleft in the NTBD and the N-hexylpiperidine moiety extended toward domain Ia. Consistent with the predicted binding pose, kinetic studies indicated that compound 15 bound competitively with NAD⁺. The ability of compound 15 to inhibit bacterial growth was also independently determined in the gram-negative E. coli and S. typhimurim, with MIC values reported in the low μg mL⁻¹ range. While these values are the same order of magnitude as the in vitro IC₅₀ values, polymxin nonapeptide was used to permeablize the membrane indicating that membrane translocation may be a problem. The reported μM IC₅₀ value, 15-fold LigA selectivity compared to HsLigI, make compound 15 a promising candidate for further refinement.

3. ATP-dependent RNA ligases

The final promising therapeutic target that we will review here are the ATP-dependent RNA editing ligases from the pathogenic Trypanosoma species. The development of new and effective drugs to treat the diseases caused by these protozoan parasites has been relatively nonexistent. Each year, millions of people in the poorest countries in the world suffer needlessly from the infectious tropical diseases borne by these pathogens, including African sleeping sickness (T. brucei), Chagas disease (T. cruzi), and leishmaniasis (Leishmania major). These “orphan diseases” cause not only death, but also promote a crippling cycle of poverty within the Americas, Asia, and Africa, which collectively harbor a disproportionate amount of the neglected disease burden ⁷⁷. Existing drugs are too costly or difficult to deliver, induce debilitating or fatal side effects, and are showing increased signs of resistance ⁷⁸.

The targeting of RNA editing ligase in these species attacks a unique aspect of trypanosomal biology, which is the extensive editing of their mitochondrial mRNAs ²¹. Through the insertion and deletion of uridylates (U’s), the editing process transforms premature mitochondrial RNA (pre-mRNA) to mature mRNA in a multi-protein complex known as the editosome ^{22, 79}. The exact composition of the editosome complex has yet to be fully characterized, although 20S core complexes have a Mw of 1.6 MDa and appear to be comprised of 16–20 proteins ⁸⁰. It has recently been demonstrated that at least three different 20S editosomes of heterogeneous composition and distinct specificity are involved in the editing process ^{81, 82}, possibly reflecting compositional changes of this dynamic multicatalyst complex at different stages in the editing process ⁷⁹.

The mRNA editing process begins in the trypanosomal mitochondrial DNA, which consists of a topologically linked network of thousands of minicircles and dozens of maxicircles. It is the transcripts of these maxicircles, which encode components of respiratory complexes and energy transduction systems that undergo extensive RNA editing. The editing process begins when guide RNAs (gRNAs) are transcribed from the minicircles in the mt genome and subsequently base-pair with pre-mRNA sequences through a conserved “anchor sequence” ^{83, 84}. Endonucleolytic cleavage of the pre-mRNA strand occurs at a point of mismatch between the trans-acting gRNA and its cognate pre-mRNA, and the type of RNA mismatch determines which 20S editosome catalyzes the cleavage reaction. As specified by the gRNA sequence, U’s are then either added, by the terminal uridylyl transferase (TUTase) TbRET2, or deleted, by a U-specific 3′-exoribonuclease. The processed RNA fragments are then religated by one of two RNA ligases, RNA editing ligase 1 (TbREL1) or 2 (TbREL2). Interestingly, TbREL1 is essential for the process ²³ whereas TbREL2 is not ⁸⁵, implying that TbREL1 can complement for the loss of TbREL2, but not vice versa. This religation of the now completely base-paired double-stranded RNA strands occurs in a three-step process, which is mechanistically conserved among the NTR superfamily (Fig. 1).

3.1 Defining Structural Features

Many of the core features of the NTR superfamily structure are also conserved in TbREL1. TbREL1 is comprised of a catalytic N-terminal adenylation domain and a C-terminal domain that facilitates non-covalent interaction with another editosome protein, KREPA2 ⁸. In this system, the OBD, which is usually associated with DNA ligases and capping enzymes in cis, appears to be provided by KREPA2 in trans and has been predicted to act as a conformational switch regulating various steps in the editing process ⁸. In 2004, Deng et al. published a 1.2 Å resolution crystal structure of the adenylation domain, TbREL1, from T. brucei with the bound ATP ligand. This structure revealed a deep ATP binding pocket with three buried water molecules that could potentially be displaced by novel ligands (Fig. 3D). The closest known relative to TbREL1 is the RNA ligase 2 from T4 phage, which has crystal structures available for the AMP-intermediate, and RNA bound states ^{55, 86}. The established similarity between these two enzymes is important, as much of this information for the T4 phage system can be used to help interpret, understand, and direct strategic studies for the enzymatic activity of TbREL1. Steady-state and pre-steady-state kinetic analysis coupled with strategic mutagenesis of the T4 RNA ligase 2 has established functional roles for the highly conserved binding site residues and mapped many of the important interactions of the RNA ligase active site ^{86, 87}.

3.2 Inhibitors

Recently, Amaro et al. used an integrated computational and experimental approach called the “relaxed complex scheme” to identify a set of micromolar inhibitors of TbREL ²⁷. An initial virtual screen was first performed against the TbREL1 crystal structure. In order to account for protein-receptor flexibility, a molecular dynamics (MD) simulation of the target protein followed ⁸⁸. Each of the top hits from the initial virtual screen was then docked into multiple representative protein configurations extracted from the MD trajectory, yielding a spectrum of binding energies for each compound. For each compound, predicted binding energies were then averaged over the ensemble, giving an ensemble-average docking score that may better predict the ligand binding affinity than a single docking score based on the crystal structure alone. Measuring formation of TbREL1-[³²P]AMP by SDS-PAGE and autoradiography, the study was looking for inhibitors of the adenylylation reaction in step 1. The experimental assays were carried out at 10 μM compound concentration in the presence of 0.1% Triton X-100 as a first measure to select against any promiscuous, aggregate-based inhibitors ⁸⁹. Testing in round 1 identified three compounds representing two different chemical scaffolds, the most potent inhibitor 16 based on a polysulfonated naphtyl dye-like scaffold and compounds 17 and 18, which are sulfonated diketoanthracene based scaffolds (Fig. 5 and Table 2). A second Tanimoto-based similarity search identified three additional compounds 18, 19, 20 which all belong to the same family of dye-like compounds as compound 16. The top three compounds were tested for selectivity against the T4 phage and human DNA Lig IIIβ enzymes and were shown to have 6- to 36-fold selectivity for the trypanosomal target (Table 2). The compounds were not tested against whole cell assays, however, it is almost certain the charged sulfonic acid moieties will cause issues permeating cell membranes, and additionally, the azo linkage and sulfonate ester groups are subjects of metabolic concern. An additional challenge for the TbREL1 target is that crystallization is extremely challenging and so far only the ATP-bound receptor has been resolved.

Figure 5.

Inhibitors of ATP-dependent RNA ligase 1.

Table 2.

Compound no.	TbREL1 IC50, (μM)	HsLIG3β IC50, (μM)	ATP competitive?
16	1.95	27.49	Yes
17	~ 10 *	N/A	Yes
18	~ 100 *	N/A	Yes
19	1.95	36.89	Yes
20	1.01	6.59	Yes

^*IC₅₀’s were not determined; % activity at 10 μM inhibitor concentration showed 87.2% activity for compound 17 and 68.5% activity for compound 18

4. Expert Opinion Section

LigA is ideally suited for rational inhibitor design due in part to published high throughput assays ^{41, 61}, the availability of high quality crystal structures, unique structural domains, and the extent of biochemical characterization. For example, the work of Srivastava et al. targeted the NAD⁺ binding pocket formed by the AMP binding site in the NTBD and the NMN binding groove, housed in domain, Ia. By exploiting domain Ia, which is unique to bacterial LigAs, their target-structure based approach resulted in the discovery of LigA specific bisglycosyl diamines ^{75 76}, and the N-substituted tetracyclic indoles ⁵³. The results of Srivastava et al. proved that LigA specific inhibitors could be constructed using a structure based design strategy, yet the conserved hydrophobic tunnel adjacent to the NAD⁺ binding pocket (Fig. 3A) went unutilized. The usefulness of this tunnel in inhibitor design has been substantiated by four crystal structures (3BA8, 3BA9, 3BAA and 3BAB), recently deposited in the PDB by Pinko and colleagues, preceding a forthcoming publication. These structures excite the possibility of optimizing inhibitors, which primarily make use of the NAD⁺-binding pocket, by strategic substitutions that direct alkyl or aryl groups into the hydrophobic tunnel. Even in the absence of a bound inhibitor crystal structure, in silico docking could conceivably provide a qualitatively correct picture of the bound state that could be used during the optimization process.

Although utilizing the hydrophobic tunnel may result in inhibitors with greater LigA specificity, it may also stimulate resistance mutations more rapidly than inhibitors that specifically target essential residues within the NAD⁺ binding pocket. The recently proposed “substrate envelope hypothesis”, which offers an explanation of HIV-protease inhibitor resistance, is consistent with this view ^{90, 91}. In this hypothesis, the protease substrate occupies a “consensus volume”, which encompasses enzyme residues essential for catalysis. Inhibitor resistance mutations are more frequent in regions where the inhibitor falls outside of the consensus volume, contacting non-essential residues. Rapid generation time, and selective pressure are shared causes of antibiotic and antiviral resistance, making it likely that the principles entailed in the substrate envelope hypothesis are also applicable to antibiotic resistance. It follows that residues within the LigA hydrophobic tunnel may be more prone to resistance mutations when targeted by an inhibitor than residues essential for catalysis. Despite this, its size, and proximity to the NAD⁺-binding pocket make the hydrophobic tunnel the most promising structural element to target for optimized LigA inhibitor binding.

Targeting the RNA editing ligases in the pathogenic Trpanosoma species is another relevant and promising avenue for antiparasitic therapeutics. Economic factors are likely to play a role in the continued neglect of these targets by industry; yet, strides made in the academic arena are promising in that more attention is being paid to the orphan diseases thanks to shifts in the current funding climate as well as the increased participation of academic groups in drug discovery and design. A high throughput REL1 assay is under development ⁹² and will assist in diversifying the chemistry of known hits, which is an important next step before identification of any promising lead compounds. Although the REL1 enzymes do not have a human homologue, the high similarity to human ATP-dependent DNA ligases will require the simultaneous cross-optimization against these enzymes. In this regard, targeting REL1-RNA bound complexes may facilitate the development of inhibitors that take advantage of the RNA to achieve greater specificity over the related human enzymes. Targeting the hybrid complex would also allow the strategic development of high affinity compounds that will bind to and inhibit different complexes along the three-step ligation reaction pathway. Although high-resolution structures for REL1 in complex with RNA are not yet available, structural dynamics of homology models based on similarity to the available RNA-bound T4 phage RNA ligase crystal structure ⁵⁵ have been pursued ⁹³.

Abbreviations

NTR

Nucleotidyltransferase

PPi

pyrophosphate

NMN

nicotinamide mononucleotide

AMP

adenosine monophosphate

ATP

adenosine triphosphate

NTBD

nucleotide-binding domain

OBD

oligonucelotide-binding domain

gRNA

guide RNA

TbREL1

RNA editing ligase 1 from Trypanosma brucei

TUTase

terminal uridyl transferase

HsLigI

human DNA ligase I

DBD

DNA-binding domain

HhH

helix-hairpin-helix

LigA

NAD+-dependent DNA ligase