Novel antimycobacterial compounds suppress NAD biogenesis by targeting a unique pocket of NaMN adenylyltransferase

Abstract

Conventional treatments to combat the tuberculosis (TB) epidemic are falling short, thus encouraging the search for novel antitubercular drugs acting on unexplored molecular targets. Several whole cell phenotypic screenings have delivered bioactive compounds with potent antitubercular activity. However, their cellular target and mechanism of action remain largely unknown. Further evaluation of these compounds may include their screening in search for known antitubercular drug targets hits. Here, a collection of nearly 1,400 mycobactericidal compounds was screened against Mycobacterium tuberculosis NaMN adenylyltransferase (MtNadD), a key enzyme in the biogenesis of NAD cofactor that was recently validated as a new drug target for dormant and active tuberculosis. We found three chemotypes that efficiently inhibit MtNadD at low micromolar range in vitro. SAR and cheminformatics studies of commercially available analogs point to a series of benzimidazolium derivatives, here named N2, with bactericidal activity on different mycobacteria, including M. abscessus, multidrug resistant M. tuberculosis, and dormant M. smegmatis. The on-target activity was supported by the increased resistance of a M. smegmatis strain overexpressing the target, and by a rapid decline in NAD(H) levels. A co-crystal structure of MtNadD with N2–8 inhibitor reveals that the binding of the inhibitor induced the formation of a new quaternary structure, a dimer-of-dimers where two copies of the inhibitor occupy symmetrical positions in the dimer interface, thus paving the way for the development of a new generation of selective MtNadD bioactive inhibitors. All these results strongly suggest that pharmacological inhibition of MtNadD is an effective strategy to combat dormant and resistant Mtb strains.

Graphical Abstract

Introduction

Despite extensive prevention and control measures in the last two decades, Tuberculosis yet represents a dramatic health issue with global impact. The World Health Organization (WHO) estimates over ten million active disease cases in 2016 and nearly 1.7 million deaths, 22% of which were among immunocompromised HIV-positive individuals¹. Mycobacterium tuberculosis (Mtb), which causes TB, undergoes a metabolic shift in the human host into a dormant state resulting in long-term persistent infection and phenotypic resistance to many otherwise successful drugs^2–4. Drug susceptible strains can be effectively treated with a 6 to 9-month regimen of multiple antibiotics, but the non-adherence to the therapeutic guidelines on a global level has exacerbated the selection and spreading of resistant strains in the last two decades⁵. Particularly alarming is the increase of multidrug-resistant tuberculosis (MDR-TB), resistant to isoniazid and rifampicin, two of the most powerful TB drugs, and extensively drug resistant (XDR-TB), which are additionally resistant to any fluoroquinolone, and to at least one of the second line drugs (amikacin, capreomycin or kanamycin)⁶.

Consensus is growing that novel, much needed antibiotics, should efficiently kill drug-resistant strains and dormant mycobacteria as well as shorten therapy via novel targets^{2, 3}.

Two novel classes of antibiotics, diarylquinolines and nitroimidazoles, have reached stage III of clinical trials, proving to be effective alone in treating latent tuberculosis (LTB) ^7–10. Bedaquiline (BDQ), the first-in-class compound of diarylquinolines, is the first new drug in decades to be clinically approved for TB treatment. In particular, due to its unique mechanism of action on mycobacterial ATP synthase, bedaquiline opened a new door for discovering anti-TB/LTB drugs using oxidative phosphorylation as a target system¹¹. However, resistance to BDQ has already been observed. Besides the selection of mutations of the atpE subunit of target ATP synthase, drug-responsive mechanisms involve activation of dormancy and metabolic remodeling ^{12, 13}. Notably, alternative ATP production from substrate-level phosphorylation appear to enable transient bacterial survival, as further confirmed by BDQ-enhanced killing of mycobacteria on non-fermentable energy sources¹³. Glycolysis not only produces ATP faster, albeit less efficiently, than oxidative phosphorylation but also supplies metabolic precursors required for macromolecular biosynthesis. Strikingly, glycolytic enzymes are upregulated in BDQ-treated mycobacteria, including lactate dehydrogenase that recycles NADH to NAD¹³. These observations emphasize the significance of targeting other aspects of energy metabolism that are essential for both actively replicating and dormant forms of mycobacteria.

Biogenesis and homeostasis of redox cofactors, most notably of NAD pool (NAD(H) and NADP(H)), is another crucial aspect of respiratory energy metabolism in both, replicating and dormant forms of mycobacteria ^{14, 15}.

NAD(H) is the entry electron donor in respiratory chain and the key oxidant driving the glycolysis. Thus, depletion of NAD(H) pool generates a glycolytic slowdown ¹⁶ and a rapid shutdown of electron transfer chain thereby affecting ATP synthesis from both sources.

In this view, NAD starvation, which represents alone a validated strategy to kill mycobacteria recalcitrant to current therapies^{17, 18}, may potentially show synergy with drugs targeting ATP synthesis such as bedaquiline.

The universally conserved and essential enzymes, NadD, NadE and NadF, which drive the last non-redundant steps of NAD(P) biosynthesis, were implicated by our previous genomics-driven studies as potential drug targets in numerous bacterial pathogens^19–22. Of these three enzymes, NadD is the most divergent from its human counterparts (NMNAT1–3²³), offering the opportunity of developing selective small-molecule inhibitors^{22, 24–27}. Our recent studies showed that induced degradation of NadD and NadE enzymes in a model system of M. smegmatis (Msm) led to rapid depletion of the NAD cofactor pool followed by cell death, even under non-replicative conditions¹⁷. Another group used a similar approach to show that induction of NadE degradation also suppressed acute and chronic Mtb infection in mice¹⁸. These findings provided ultimate validation of NAD metabolism as a target pathway for the development of new antimycobacterial therapies. Structure-function and inhibition studies of the key NAD(P) biosynthetic enzymes ^{22, 24, 25, 28, 29} including our recently published work on MtNadD²⁶ support their druggability.

We set out to identify inhibitors of MtNadD through a target-based screen of orphan compounds with mycobactericidal activity. Here, we report the identification as well as the biochemical and biological characterization of several potent inhibitors of NadD with bactericidal activity on mycobacteria, including M. abscessus, multidrug-resistant M. tuberculosis, as well as a prominent bactericidal effect under non-replicating conditions in M. smegmatis. To clarify in atomic detail the mode of action of these compounds and their structure-activity relationships, we also determined the crystal structure of MtNadD in complex with one of these inhibitors.

RESULTS AND DISCUSSION

Screening of bioactive antimycobacterial compounds affords novel MtNadD inhibitors.

Previous whole cell phenotypic screening has identified potent anti-mycobacterial compounds in a variety of tested conditions on replicating, non-replicating and intracellular forms of Mtb. These compounds were selected via several cell-based HTS campaigns as reported in PubChem ^{30, 31}, but their mechanisms and molecular targets remain unknown. We speculated that, given a limited space of potential targets (~200–300 essential genes in M. tuberculosis), at least some of these compounds could target the essential enzymes of NAD metabolism. For this reason, a collection of 1389 bioactive compounds, kindly provided by Global Alliance for TB Drug Development (TB Alliance) was tested against the recombinant M. tuberculosis NadD enzyme using colorimetric detection assay of released PPi (Figure 1). The primary assays, performed at fixed final concentration of 25 μM for each compound, and at subsaturating concentrations of ATP and NaMN, identified 16 compounds (~1% hit rate) inhibiting more than 30% of enzymatic activity, with three of these demonstrating >50% inhibition. Notably, this hit rate is comparable to our previous structure-based approach on NadD from Escherichia coli and Staphylococcus aureus ²⁴ which, however, used a more stringent concentration of the inhibitor (100 μM). These primary hits were repurchased from different vendors and their concentration-response profiling yielded 5 compounds with IC₅₀ values in the range of 6–22 μM (Supplementary Table 1). The three top-ranked compounds with significant structural diversity, termed here as N1, N2, and N3, were selected for further analysis, including cheminformatics and SAR studies by analogs’ testing, cell-based assays, and co-crystallization trials (Figure 1).

Figure 1. Flowchart of the whole-cell and target screening approach for delivering novel antitubercular NaMN adenylyltransferase inhibitors.

SAR for chemotypes N1, N2, and N3.

Before any further investigation on these chemotypes was pursued, we wanted to confirm their bioactivity as originally displayed in whole cell phenotypic assay. For example, we determined a minimal inhibitory concentration of 28 μM of the compound N2–11 for M. tuberculosis H37Rv. This result is consistent with the inhibition value reported in PubChem Bioassay 1626 (MIC of 16 μM). Overall, around 30 analogs of the 3 selected chemotypes were purchased from different vendors and tested in vitro against MtNadD (Figure 2 and Tables 1–3). Despite a limited chemical search exploration, 4 analogs show improved MtNadD inhibition relative to their parent compound, with up to 5-fold lower IC₅₀ (2.5 μM) for N1–11 analog, which represents the most potent inhibitor for MtNadD described so far ²⁶.

Figure 2. Chemical diversity of the 3 series of novel MtNadD inhibitors evaluated.

In red, molecules that were identified in our primary screen of the TB Alliance bioactive library against MtNadD (IC₅₀ < 20 μM). In black, novel analogues for each chemotype tested in this study. In blue, analogs that were not tested in our study, but for which a relevant activity against M. tuberculosis or a Mtb target enzyme has been reported in PubChem (see Tables 1–2 and main text for details).

Table 1.

Inhibitory properties of N1 class MtNadD inhibitors

ID	PubChem CID	Enzyme inhibition		Mycobacterium Growth inhibition
		IC50 (μM)	(%) at 1 μM	(%) at 50 μM	MIC (μM)
		MtNadDa	MtFBAb	smegmatis c	tuberculosis
					d	e
N1–1	1900457	–	54.81	–	–	>25
N1–2	3772903	–	54.28	–	–	50
N1–3	1898472	–	79.9	–	–	25
N1–4	1894659	–	97.0	–	–	25
N1–5	1898131	–	22.14	–	–	>25
N1–6	1119724	24.8±4	11.37	nd	–	nd
N1–7	3807641	17.4±3.4	14.68	nd	–	>25
N1–8	1896721	34.0±8.7	–	~50	–	–
N1–9	1896658	41.6±7.9	–	~20	–	–
N1–10	1203235	3.7±0.6	20.43	nd	<10	25
N1–11	3722630	2.5±0.3	–	~50	–	–
N1–12	3697907	29.6±6.2	–	nd	–	–
N1–13	1898213	13.3±2.9	38.2	nd	12.5	25
N1–14	1203236	14.3±2.2	–	nd	–	–
N1–15	1119726	5.3±1.8	–	nd	–	–
N1–16	1893392	76.1±16.4	41.76	~20	>100	nd
N1–17	1203237	>100	–	nd	–	–
N1–18	1899519	25.7±9.2	–	~50	–	–
N1–19	1896369	>50	–	~100	–	–
N1–20	1898247	97.8±47.8	14.6	~10	–	nd
N1–21	1122203	>100	–	~20	–	–
N1–22	1122208	>100	–	nd	–	–
N1–23	1122206	>50	–	nd	–	–
N1–24	1122209	>50	–	nd	–	–
N1–25	1898735	>50	–	nd	–	–
N1–26	1423254	>50	–	nd	–	–
N1–27	1897037	>100	–	nd	–	–
N1–28	3774654	>100	–	nd	–	–
N1–29	658395	–	19.8	–	>10	<25
N1–30	16682092	>100	22.17	~20	>100	nd
N1–31	3333039	–	54.21	–	–	>25

^aAssays performed in duplicate in at least 5 different inhibitors concentration. Standard errors are reported.

^bIn vitro inhibition of M. tuberculosis H37Rv fructose-bisphosphate aldolase (FBA), as reported in PubChem Bioassay AID 588726.

^cAntibacterial activity against M. smegmatis is expressed as percentage of growth suppression at fixed 50 μM concentration, except for N1–18 that was tested at 20 μM.

^dValues are deduced from PubChem Bioassay AID 1626 data.

^eValues obtained from PubChem Bioassay AID 449762 data. “–“, not assayed; nd, no detected effect. In red, compounds identified in the primary screening. In blue, analogs identified via cheminformatics search for which a bioactivity is reported, but not tested in our study.

Table 3.

Inhibitory properties of N3 class MtNadD inhibitors^a

ID	MtNadD IC50 (μM)	Growth inhibition, MIC (μM)
		M. smegmatis	M. tuberculosis	M. abscessus
N3–1	10.5±1.2	10	3.1	20.9
N3–2	96.4±25.9	>50	>100	>100
N3–3	>100	nd	–	–

^aAssays performed in duplicate in at least 5 different inhibitors concentration. Standard errors are reported. “–“, not assayed; nd, no detected effect (at 50 μM). ID color-coding is as in Tables 1 and 2.

In class N1, the replacement of a 3-fluoro benzyl group with the more hydrophobic 1-phenylethyl (N1–11) or benzyl (N1–10) moieties yielded the best improvement. More polar moieties such as five- or six-membered rings with two or more heteroatoms were poorly tolerated (IC₅₀ > 50–100 μM), while linear or branched hydrocarbon chains (N1–6 through N1–8), or the smaller cyclopropyl moieties (N1–9) were well-tolerated substitutions (15 μM < IC₅₀ < 40 μM), underlying the importance of the hydrophobic character in this inhibitor moiety.

Regarding class N2, is worth mentioning that three analogs were already identified in the primary screening, namely N2–7, N2–9, and N2–11. This class features a benzimidazolium core with a methyl and a 2-isopropyl-5-methylcyclohexyl-containing moiety at each N atom. All N2 tested analogs were active, although they did not significantly improve the parent compound potency. Of the two weaker analogs, compound N2–2, replacing a bulky methoxyphenyl or dimethoxyphenyl group in C-1 position of benzimidazole by a methyl group displays a substantially lower inhibitory activity (IC₅₀ ~100 μM). The replacement with a bromophenyl (N2–6) yields an intermediate IC₅₀ of 45 μM, pointing to the importance of the size of this moiety. The remaining analogs of this series, containing a methoxyphenyl group have similar activities compared to the parent compounds (5 μM < IC₅₀ < 20 μM) and yield a bland structure-activity profile. The only analog with a slightly improved activity (N2–1, IC₅₀ of 4.8 μM) has a distinct structure with two symmetric 2-isopropyl-5-methylcyclohexyl-containing moieties. The two N3 analogs tested in this study demonstrated a marked decrease in activity (Figure 2 and Table 3). We therefore did not follow up this compound in more detail.

At this stage, all tested analogs were selected only based on structural similarity, without any attempts of their rational improvement. Nevertheless, some analogs in series N1 and N2 displayed a moderate to significant improvement of inhibition efficacy, compared with the parent compounds, showing to be amenable to chemical optimization.

Effects of MtNadD inhibitors on mycobacteria cell growth.

Our original hits against MtNadD activity have potent antitubercular activities with reported EC₅₀ in the low micromolar range (see results in PubChem for bioassay AID 1626) and reconfirmed by us as shown hereafter. MtNadD IC₅₀ values are in the same order of magnitude (Tables 1–3), which is not unexpected since we demonstrated with a protein knock-down system that partial inhibition of MtNadD is sufficient to kill mycobacteria ¹⁷. To direct the choice of compounds for further evaluation, we next assessed their ability to also suppress the growth of M. smegmatis. Besides being a convenient nonpathogenic and rapidly growing laboratory strain, our choice of M. smegmatis as a surrogate of M. tuberculosis was justified by an observed tendency of Msm to display drug susceptibility profile closer to MDR than drug-sensitive M. tuberculosis ³² and by the conservation of NAD target pathway and key enzymes therein. MtNadD and MsNadD bear high degree of sequence identity (75%), rising to nearly 100% in the conserved regions of active site (Supplementary Figure 1).

N2 inhibitors display on-target killing activity in replicative and nonreplicative conditions.

Despite encouraging low micromolar IC₅₀ values, the N1 series resulted in poor M. smegmatis growth suppression (Table 1). At fixed concentration of 50 μM, the original N1 hit had no effect, while other N1 analogs produced at most a 50% growth suppression (Table 1). Moreover, the growth inhibition, when present, did not correlate with MtNadD inhibition, pointing to possible off-target effects of this series of compounds. Indeed, class N1 compounds are moderate to potent inhibitors of M. tuberculosis H37Rv fructose 1,6-biphosphate aldolase (FBA) as reported in PubChem Bioassay AID 588726. For instance, at 1 μM, primary hit N1–13 and analog N1–10 exert 40% and 20% inhibition of MtFBA activity, respectively, with prospective IC₅₀ values in the very low micromolar range. MtFBA has been proposed as a new pharmacological target of M. tuberculosis since it is upregulated in latent TB³³. Although an antibacterial molecule with dual-targeting mechanism presents obvious benefits such as, for instance, an expected low emergence of spontaneous resistance³⁴, the poor bioactivity in M. smegmatis led us to discard the N1 series for further characterization.

As opposed to N1 series, the N2 series of inhibitors display strong bactericidal activity against model M. smegmatis with MIC in the low micromolar range (Figure 3A–D and Table 2). Next, we reconfirmed a comparable bactericidal activity on M. tuberculosis, also for a few analogs that were not originally included in the previous HTS whole-cell bioassay (Table 2). For the most potent ones, we additionally measured MIC for Mycobacterium abscessus, a more divergent nontuberculous mycobacterium ³⁵ with a conserved NAD target pathway. The strong growth suppression of M. tuberculosis and M. smegmatis, and the less pronounced inhibition of M. abscessus, demonstrates N2 inhibitors’ specific targeting of NadD family in mycobacteria, and, in general, points to a protein target site potentially resilient to mutations. Furthermore, N2–7 and N2–11 caused rapid depletion of the NAD pool (Figure 3D, E), comparable to the effect of NadD “protein knockdown” as observed in our previous work¹⁷, corroborating their on-target activity.

Figure 3. Bactericidal, on-target effects of N2 inhibitors in replicating and non-replicating M. smegmatis.

(A-C) Growth suppression curves at different N2 series inhibitors’ concentrations underline their killing activity on replicating Msm at above 5 μM. DMSO at 1 % represents the “no inhibitor” control. Linezolid at 3 μM was used as a positive control. (D) Addition of the inhibitor at ~5× MIC concentration at log-phase (after 12.5 hours of growth) stops the growth of Msm and (E) rapidly deplete NAD pool. Residual levels of NAD were measured by a colorimetric detection kit and normalized by total protein in samples taken for 6 hours, in 2 hours’ time interval. (F) Cell viability by CFU assay after 1–4 days of incubation of Msm with 5xMIC inhibitor (25 μM) under carbon starvation conditions. In this case, Linezolid was used at 15 μM.

Table 2.

Inhibitory properties of N2 class MtNadD inhibitors evaluated

ID	PubChem CID	MtNadDa IC50 (μM)	Mycobacterium Growth inhibition, MIC (μM)
			smegmatis	tuberculosis			abscessus
					b	c
N2–1	2834410	4.8±0.7	5	46.7	–	–	61.4
N2–2	2834412	111±16	25	–	–	–	–
N2–3	2834423	–	–	–	I	I	–
N2–4	2834427	–	–	–	32.7	28	–
N2–5	2834425	–	–	–	I	I	–
N2–6	2834414	45.5±15.8	25	–	I	I	–
N2–7	2834416	12.6±2.8	5	–	16.6	–	–
N2–8	2834419	17.9±2.2	5	21.2	I	I	26.7
N2–9	16192954	18.2±2.6	5	38.5	10.4	I	26.2
N2–10	3908383	19.5±2.5	5	19.7	–	–	25.6
N2–11	16195070	6.3±1.1	5	28.2	15.9	I	26

^aAssays performed in duplicate in at least 5 different inhibitors concentration. Standard errors are reported. “–“, not assayed; nd, no detected effect.

^b,cValues deduced from PubChem Bioassay AID 1626 and AID 449762 data, respectively. For consistency, these values represent the minimal inhibitory concentration yielding >99% growth suppression and were obtained by fitting the reported EC₅₀ and Hill Slope values into ECanything built-in equation in Prism 7.0. I, compounds tagged as “Inconclusive” and which showed >98% inhibition at a single concentration of 25 μM”. No dose-response analysis was carried out. ID color- coding is as in Table 1.

Using a carbon starvation model for the non-replicating state in M. smegmatis^{17, 36}, we observed a bactericidal effect even for dormant mycobacteria (Figure 3F). Up to 4 logs in colony forming unit (CFU) reduction was observed upon 96 h incubation with inhibitor N2–11 (at 5× MIC dose), with almost no CFU reduction in parallel samples incubated with DMSO, linezolid (also at 5× MIC). The N2–8 analog shows a more delayed, but substantial, effect with over 2 logs CFU decrease. Taken together, these results provide strong evidence that proposed inhibitors of series N2 are effective against replicative as well as non-replicative mycobacteria.

N2–11 inhibitor suppresses drug-resistant Mtb.

Having demonstrated that MtNadD inhibitors suppress the growth of drug-tolerant mycobacteria, we next determined their activity on a panel of drug-resistant Mtb clinical isolates using a single dose assay in liquid media. We measured the effect of N2–11 (at a concentration of 25 μM, 6-fold the MIC for Mtb) on two Mtb ATCC strains (H37Rv and PZA monoresistant), and four clinical drug-resistant isolates: pyrazinamide resistant (PZA-R), MDR, XDR, and totally drug-resistant (TDR). Growth controls were run in parallel in the presence and absence of DMSO. The results (Supplementary Figure 2) showed that the inhibitor halted completely the growth of all clinical isolates. Despite being a single-dose assay, these data establish that class N2 of MtNadD inhibitors are promising to tackle drug-resistant tuberculosis.

Overexpression of NadD increases resistance to N2 series of inhibitors.

Overexpressing a gene encoding a drug target is expected to confer an increased resistance to that drug. Therefore, to genetically validate NadD as the molecular target of N2 inhibitors, we overexpressed MtNadD in M. smegmatis using a nonintegrative pVV16-derived plasmid (see the Methods section for cloning details). To this aim, we generated two different versions of MtNadD, wild-type NadD (NadD_WT) and NadD_Δ24, a non-functional version here used as a control ¹⁷. In the presence of selected N2 inhibitors, the NadD_WT strain consistently exhibited a larger MIC compared to NadD_Δ24 (~2-fold, Supplementary Figure 3). In addition, we included N3–1, a representative of the N3 series that we discarded earlier for its off-target mode of action. Coherently, with N3–1 no MIC shift was observed between NadD_WT and NadD_Δ24. Altogether, these data suggest that MtNadD may indeed represent a specific molecular target of the N2 series, antimycobacterial molecules identified in this study.

Structure of MtNadD in complex with inhibitor N2–8.

To elucidate the mechanism of MtNadD inhibition by N2 class of inhibitors, rationalize the observed SAR, and guide future rational inhibitor design, we determined 1.86 Å resolution crystal structure in complex with inhibitor N2–8.

The structure was refined to R_work of 0.172 and R_free of 0.192 with excellent overall geometry (Table 4). There are two molecules of MtNadD in the asymmetric unit (chain B and A in Figure 4A and Supplementary Figure 4A) that correspond to previously observed dimer in apo-structures of MtNadD ²⁶. Two MtNadD monomers have very similar structures with root mean square deviations (rmsd) of 1.2 Å over 186 Cα atoms (Supplementary Figure 4B). Analysis of the crystal contacts revealed the formation of dimer-of-dimers of MtNadD (Figure 4A). This new quaternary structure was induced by the N2–8 ligand binding, with two molecules of the ligand bridging the dimers together. The crystallographically identical ligand binding sites are formed by chains B and A’, and chains B’ and A, respectively. Analysis by the PISA server classifies the interface between two MtNadD dimers as stable. The clear electron density corresponding to inhibitor N2–8 allowed unambiguous modeling of the ligand (Figure 4 and Supplementary Figure 5). The inhibitor occupies a hydrophobic pocket formed by helices α5 and α6 of chain B and by helix a6 and strand b6 of chain A’ (Figure 4B). Overall, N2–8 exhibits high steric complementarity with the surface of the two MtNadD chains. The menthol moiety of the inhibitor makes van der Waals contacts with several hydrophobic residues of the pocket (Figure 4B and Supplementary Figure 6). Additional contacts include residues L151 and L158 from chain A’.

Table 4.

Data collection and refinement statistics.

	M. tuberculosis NadD in complex with N2–8 (PDB 6BUV)
Data collection
Wavelength (Å)	1.0000
Space group	H32
Cell dimensions:
a, b, c (Å)	163.73, 163.73, 153.65
α, β, γ (°)	90, 90, 120
Resolution (Å)	52.1–1.86 (1.91–1.86)a
R sym	0.077 (1.146)
CC1/2b	99.6 (54.7)
I / σI	11.08 (1.72)
Completeness (%)	99.3 (100.0)
Multiplicity	3.8 (3.8)
Refinement
Resolution (Å)	52.1–1.86
No. reflections (total / free)	65711 / 5923
Rwork / Rfree	0.172 / 0.192
Number of atoms:
Protein	2976
Ligand/ion	38
Water	296
B-factors:
Protein	43.7
Ligand/ion	60.7
Water	49.6
All atoms	44.4
Wilson B	33.4
R.m.s. deviations:
Bond lengths (Å)	0.008
Bond angles (°)	1.008
Ramachandran distributionc (%):
Favored	98.9
Allowed	1.1
Outliers	0
Rotamer outliersc (%)	0
Clashscored	2.53
MolProbity scoree	1.04

^aValues in parentheses are for the highest-resolution shell.

^bCC_1/2 correlation coefficient as defined in Karplus & Diederichs ⁴³ and calculated by XSCALE ⁴⁴.

^cCalculated using the MolProbity server (http://molprobity.biochem.duke.edu) ⁴⁵.

^dClashscore is the number of serious steric overlaps (> 0.4 Å) per 1000 atoms.

^eMolProbity score combines the clashscore, rotamer, and Ramachandran evaluations into a single score, normalized to be on the same scale as X-ray resolution ⁴⁵.

Figure 4. Structure of MtNadD in complex with inhibitor N2–8.

(A) Two N2–8 binding pockets induce the formation of dimer-of-dimers of MtNadD. The asymmetric unit dimer A-B is shown in shades of blue; the crystallographic symmetry dimer A’-B’ is shown in shades of green. (B) A close-up view of the N2–8 binding site. The hydrophobic residues lining up the inhibitor binding site are shown in sticks representation. σA-weighted 2F_O–F_C electron density map contoured at 1σ is shown as blue mesh. (C) Comparison of the closed active-site conformation of MtNadD•N2–8 complex (light blue) with MtNadD apo structure (yellow) (PDB ID, 4X0E) and MaNadD•NaAD complex structure (purple) (PDB ID, 5DEO). H-bonding between D109 and the adenosine moiety of the MaNadD•NaAD complex is shown as dashed line. The steric clashes of L164 with adenine ring, I113 with nicotinate ring, and G106 with the AMP ribose are marked by dotted circles.

Moreover, residue Glu160 contributes to the positioning of the positively charged benzimidazole core. The electron density for the phenoxymethyl part of the inhibitor is relatively poor compared with the rest of the inhibitor. The phenoxymethyl moiety could adopt multiple conformations, as it is oriented away from the protein. This is consistent with the SAR data on the N2 series of inhibitors that show that various sub-structures are permissible at R2 position (Figure 2 and Table 2). The inhibitor-binding site is located next to the active site of the enzyme (Figure 4C). The inhibitor-bound structure of MtNadD•N2–8 is similar to our apo-structures of MtNadD (PDB ID 4X0E and 4RPI) ²⁶ with rmsd of 1.1–1.6 Å over 177–188 Cα atoms (Figure 4C). This conformation represents an inactive state of the enzyme with a constricted active site that prevents the binding of both NaMN and ATP substrates. Here, helix α5 is displaced towards the active site in MtNadD•N2–8 as well as in apo-structures of MtNadD, compared to the structures of NadD bacterial homologs in product-bound state (Figure 4C). In particular, Ile114 lies in a clashing position with the nicotinic acid moiety of modeled NaAD. Additionally, Trp117 which should coordinate the nicotinic acid moiety of NaMN or NaAD is rotated outside of the active site. Moreover, Leu164, a key residue in ATP substrate recognition and stabilization ²⁶, is similarly oriented as in the apo-form and blocks the access of the ATP substrate (Figure 4C).

Besides freezing the enzyme in a non-productive conformation, the N2–8 inhibitor plays an active role in the active site destabilization. Most prominently, the side chain of Asp109 is flipped away from the active site towards to the inhibitor due to the electrostatic interaction with the positive charge on the benzimidazole ring (at a distance of about 4 Å) (Figure 4B, C). Asp109 is a strictly conserved residue for the entire NaMN/NMN-AT family ²⁶ which forms an H-bond with the 2’-OH of the AMP ribose³⁷ and is thought to help with essential Mg⁺² coordination during catalysis³⁸. Its essentiality has been recently demonstrated by site-directed mutagenesis in Plasmodium falciparum NMN-AT ³⁹. Finally, two salt bridges Asp29-Lys47 further stabilize the two dimers of this inhibitor-driven quaternary structure (Supplementary Figure 7). Notably, Lys47 is part of the PPH(K/R) loop motif involved in the recognition of NaMN phosphate, and found to be essential for catalysis in our previous study ²⁶. Thus, the newly formed electrostatic interaction may seize Lys47 in a nonfunctional state.

DISCUSSION

Over the past years, conventional target-directed drug discovery has not met expectations, with an increasing spending accompanied by less delivery of new drugs to the market ⁴⁰. Among well-known drawbacks of target-based strategies for antibacterial development are rapid evolution of resistance of single targets, the difficulty of developing broad-spectrum compounds, and the lack of whole cell activity of many potent biochemical hits. With respect to TB, the high resistance frequency of a single-target agent is less important, since combination drug therapy is always employed. Narrow-spectrum agents that do not affect the commensal microbiota would be also particularly valuable, since the lengthy treatment regimens required for the treatment of TB and MDR-TB have detrimental effect upon the host microbiome. Thus, a novel, target-based screen of potent hits resulted from phenotypic screen could be a promising strategy to develop smarter antitubercular drugs. NAD biosynthesis represents a validated pathway in Mtb and other bacterial pathogens for antibiotic development.

Here, we describe the identification and initial SAR analysis of MtNadD inhibitors with potent antimycobacterial activity through an initial target-based screen of whole cell active compounds. This approach yielded compounds that were 5-fold more potent against the purified enzyme and over 10-fold more effective against model M. smegmatis than those reported in our initial effort ²⁶.

These compounds maintain strong bactericidal effect against M. tuberculosis and some clinical isolates of MDR-, XDR-, and TDR-Mtb as well as a prominent bactericidal effect under non-replicating conditions in M. smegmatis. This represents the first evidence that pharmacological inhibition of MtNadD and, potentially, other essential NAD-related enzymes, is a valuable strategy to kill dormant and resistant Mtb strains which could be exploited alone or in combination with other drugs, e.g. targeting energy metabolism.

In our previous study, we proposed that in dormant cells MtNadD target is in a closed, inactive conformation. A shift into a catalytically active conformation would be induced by the increased ATP substrate levels reflecting the high respiratory activity of actively replicating cells. MtNadD enzymatic dormancy hinted for a previously unexplored strategy for development of small molecule inhibitors that would target and further stabilize the catalytically inactive closed form of the enzyme. Here, we report a crystal structure of MtNadD in complex with the benzimidazolium salt derivative N2–8 that locks the enzyme in a nonproductive conformation, thus setting the stage for future lead optimization efforts. The 3D analysis revealed that the inhibitor surprisingly interacts with a largely hydrophobic pocket at the dimer interface of ligand-induced, dimer-of-dimers quaternary structure. The fact that small molecule binding could involve new or perturbed quaternary structure is not unprecedented. Indeed, several studies, including our own work with a NadD homolog described the formation of a new dimeric assembly upon binding of inhibitor ^{25, 41}.

Various anti-mycobacterial compounds incorporate the benzimidazole substructure ⁴². We provided unequivocal evidence that N2 series of MtNadD inhibitors act on target by detecting (i) a 2-fold increase of MIC in a Mycobacterium strain overexpressing MtNadD and (ii) a rapid decline of the pathway end-product NAD. Furthermore, close inspection of the chemistry of N2 series of inhibitors reveals their uniqueness and novelty. Firstly, they have substitutions at both N atoms of benzimidazole core that are not tolerated for an anti-TB activity, as recently reviewed ⁴². By virtue of these substitutions, the compounds are positively charged benzimidazolium ions, as opposed to neutral anti-TB benzimidazole derivatives reported so far. The net positive charge of the benzimidazolium core is indeed critical for the inhibition of MtNadD target enzyme as it perturbs the essential Asp109 catalytic site residue and helps bridging the dimers of the induced quaternary structure by interacting with Glu160. Secondly, N2 inhibitors feature a bulky menthol acetate substituent which engages most contacts with the protein. Moreover, through a cheminformatics survey in the PubChem databases we identified analogs of some N2 inhibitors that lack the menthol acetate moiety, resulting in neutral benzimidazoles. Strikingly, such compounds were screened in as many as three independent M. tuberculosis whole-cell bioassays, and all resulted to be inactive (Supplementary Table 2).

Our study introduces a strategy for inhibiting the essential bacterial NaMN adenylyltransferase that targets a unique, unexplored pocket, independent from the active site. Previous screening campaigns have concentrated their efforts on targeting the enzyme active site or sub-sites, with modest outcomes. Indeed, this enzyme family features a large active site with an ATP binding pocket, and its targeting may lead to extensive cross-reactivity with nucleotide binding pockets on several different proteins. Thus, our approach targeting a unique novel pocket represents a significant step forward that can circumvent the abovementioned limitations. Future extensions of this work will include improvement of chemical activity of the lead compounds by further exploring chemical diversity space to bring them into a viable therapeutic range as potential stand-alone drugs or in combination with other TB drugs.

In conclusion, this study demonstrates the successful application of combined target-directed and phenotypic screen to arrive at novel scaffolds inhibiting mycobacterial NadD. Among the three identified scaffolds, the most promising one, containing a benzimidazolium core substituted with a menthol acetate group, has been characterized in detail revealing its potential for chemical optimization. Different analogs have shown to inhibit the growth of several Mycobacteria, including drug resistant M. tuberculosis, dormant M. smegmatis, and the pathogenic M. abscessus. These molecules inhibit the protein by displaying a novel mechanism and by binding to a unique site on the enzyme.

METHODS

Details of experimental procedures are provided in Supporting Information.