Structure, function, and inhibition of adenylosuccinate lyase (ADSL) from Mycobacterium tuberculosis

Abstract

Adenylosuccinate lyase (ADSL), encoded by the purB gene, is an essential enzyme in the purine biosynthesis pathway of Mycobacterium tuberculosis (Mtb), making it a promising target for antimicrobial drug development. Here, we report the expression, purification, kinetic characterization, high-throughput screening (HTS), and structural analysis of Mtb ADSL. We developed a highly sensitive and scalable bioluminescent assay using a PPDK-luciferase coupling system to quantify ADSL enzymatic activity via AMP detection. This assay enabled reliable kinetic analysis and successful pilot HTS of a small-molecule library, identifying bithionol and tetraiodothyroacetic acid (Tetrac) as inhibitors of Mtb ADSL. Inhibitory activity was confirmed using an orthogonal fluorescence polarization (FP) assay and further validated using the AMP-Glo^™ luminescence assay. Specificity was evaluated using human ADSL (huADSL) to confirm that the compounds selectively inhibited Mtb ADSL while sparing the human enzyme. Thermal shift and gel-based protein stability assays demonstrated direct binding of bithionol and Tetrac to Mtb ADSL. Furthermore, bithionol and Tetrac displayed antibacterial activity against M. tuberculosis strains H37Ra and H37Rv, with moderate to low cytotoxicity toward human cells. Supplementation with exogenous AMP restored the growth of M. tuberculosis H37Ra inhibited by bithionol and Tetrac, confirming that both compounds act through on-target engagement of MtbADSL. The phagocytosis assay demonstrated that the compounds retained intracellular efficacy against M. tuberculosis. Finally, we determined the crystal structures of Mtb ADSL in two apo forms at high resolution (1.78 Å and 2.1Å), revealing conserved tetrameric architecture with distinct active-site features that differentiate Mtb from human ADSL. Modeling suggested that both compounds bind to an allosteric site adjacent to the active site. These findings provide a framework for structure-guided development of selective ADSL inhibitors as potential anti-tubercular agents.

Tuberculosis, caused by Mycobacterium tuberculosis (Mtb), remains the second leading cause of human death from infectious diseases globally,^{1, 2} and has re-emerged as the leading cause of death from a single infectious agent following the COVID-19 pandemic.^{3, 4} Although several drugs are currently available to treat tuberculosis, the emergence of multidrug-resistant Mtb strains and the side effects associated with existing therapies have created an urgent need to identify new and effective drug targets.

One promising area for drug discovery is the de novo purine biosynthesis pathway, which is essential for the systemic pathogenicity of microorganisms in the bloodstream.⁵ A recent study published in Nature by Lamprecht et al. highlighted the potential of developing anti-tuberculosis agents by targeting PurF,⁶ the amidophosphoribosyltransferase that catalyzes the first committed step in the purine biosynthesis pathway. Purine nucleotide synthesis in both eukaryotes and prokaryotes occurs via two primary pathways: (1) the de novo pathway, which begins with metabolic precursors such as amino acids, ribose 5-phosphate (R5P), carbon dioxide (CO₂), and ammonia (NH₃); and (2) the salvage pathway, which recycles free purine bases and nucleosides.

Genomic analyses of Mtb have revealed that nearly all enzymes involved in the de novo purine synthesis pathway, except PurT, are essential for the bacterium's survival.^7-11 Among these, the purB gene encodes adenylosuccinate lyase (ADSL), an enzyme that catalyzes two key steps in the purine biosynthesis pathway: the reversible conversion of 5-aminoimidazole-4-(N-succinylcarboxamide) ribonucleotide (SAICAR) to 5-amino-1-ribosyl-4-imidazolecarboxamide 5′-phosphate (AICAR), and the conversion of succinyl-adenosine monophosphate (SAMP) to adenosine monophosphate (AMP). Because ADSL is essential for Mtb survival and growth, it represents a promising target for antimicrobial drug development.^{7, 12}

ADSL belongs to the argininosuccinate lyase/fumarase C superfamily of enzymes, which includes ADSL/δ2-crystallin, class II fumarase, L-aspartase, and 3-carboxy-cis,cis-muconate lactonizing enzyme.¹³ Structurally, ADSL forms homotetramers, with each monomer comprising three distinct domains. The enzyme is essential for the production of purine nucleotides through the de novo biosynthetic pathway and plays a central role in cellular metabolism.

The enzymatic mechanism of ADSL involves the β-elimination of fumarate from SAMP. This begins with the abstraction of the Cβ-proton from SAMP, generating a carbanion intermediate with two negative charges. A catalytic acid then donates a proton to the N1 or N6 atom of the substrate, resulting in cleavage of the Cα─N bond and release of fumarate and AMP.^14-16 In humans, ADSL deficiency has been associated with severe health consequences, including intellectual disability, muscle wasting, and epilepsy.¹⁷

Although ADSL plays an essential role in the purine biosynthesis pathway of mycobacteria, its structural and biochemical characterization has remained limited. Previous work by Banerjee et al. provided biochemical insights into ADSLs from M. tuberculosis and M. smegmatis (Msm) and reported the crystal structure of the M. smegmatis enzyme;¹² however, the structure of MtbADSL has not been determined. In this study, we report the expression, purification, crystallization, and X-ray crystal structure determination of MtbADSL at high resolution. To the best of our knowledge, this is the first structural characterization of Mtb ADSL. In addition to structural studies, we established an innovative and sensitive bioluminescent assay to evaluate ADSL enzymatic activity and identify small-molecule inhibitors. These efforts enabled the discovery and validation of candidate inhibitors with low micromolar potency, supported by orthogonal biochemical and biophysical assays. Together, our work lays the groundwork for structure-guided drug discovery targeting ADSL and highlights its potential as a selective and tractable therapeutic target in the fight against tuberculosis.

RESULTS

Expression and Purification of Mtb ADSL

To characterize Mtb ADSL and identify potential inhibitors, the purB gene was first codon-optimized for expression in Escherichia coli, synthesized, and cloned into the pET28a expression vector to include an N-terminal His-tag. Recombinant His-tagged Mtb ADSL was then expressed in E. coli and purified using Ni-NTA affinity chromatography followed by size exclusion chromatography. The purified protein showed high purity (>95%) as determined by SDS-PAGE (Figure 1A), with a typical yield of approximately 5 mg of purified protein per liter of bacterial culture.

Figure 1. Purification and CD spectrum analysis of recombinant Mtb ADSL.

(A) SDS-page analysis of purified recombinant Mtb ADSL. (B) Far UV spectrum CD of Mtb ADSL.

Circular dichroism (CD)

The quality and secondary structure of the recombinant Mtb ADSL protein were assessed using CD spectroscopy. CD spectral analysis revealed that the purified protein is predominantly α-helical, comprising approximately 69.7% α-helix (Figure 1B). This α-helical content closely aligns with that reported for ADSLs of human¹⁸ and M. smegmatis,¹² indicating that Mtb ADSL shares a conserved secondary structure with its human counterpart. These results suggest a high degree of structural conservation across species, consistent with ADSL’s essential cellular function.

Crystal structure

We next determined the crystal structures of the apo form of M. tuberculosis ADSL in two distinct crystal forms, resolved at 2.1 Å and 1.78 Å, respectively (Supplementary Table S1). Although the protein crystallized in different lattice systems, each crystal form contained a single ADSL dimer in the asymmetric unit (Figure 2A). The application of crystallographic symmetry revealed that two dimers assemble into a homotetramer, which represents the biologically functional unit of the enzyme (Figure 2B). Notably, the active site of ADSL is formed by contributions from three monomers within the tetrameric assembly, emphasizing the importance of the quaternary structure for enzymatic function. Consistent with its essential role, the overall architecture of the Mtb ADSL homotetramer closely mirrors that of ADSL homologs from other species, including the human enzyme, underscoring its structural conservation across organisms (Figure 2B).

Figure 2. Crystal structure of Mtb ADSL.

(A) Superimposition of crystal structure of the Mtb ADSL dimer in form 2 with form 1 (left) and with human ADSL (right). The monomers of crystal form 2 were colored in green and cyan respectively. The ADSL dimer in crystal form 1 was in red. Human ADSL was in magenta (PDB: 2VD6). (B) Reconstituted homotetramer of ADSL through crystallographic symmetry (left) and comparison with the human ADSL tetramer (right). (C) Structural comparison of substrate-binding residues between Mtb and human ADSLs. Only residues that are different between the two enzymes and directly contact the substrate SAMP or the products AMP and fumarate are shown in stick representation. Atomic coloring: oxygen, red; nitrogen, blue; phosphate, orange; sulfur, light orange; carbon atoms of SAMP and human ADSL residues, gray; carbon atoms of Mtb ADSL residues are colored yellow or dark orange. Cartoons of Mtb ADSL monomers A, B, and D and human ADSL are shown in green, cyan, dark orange, and grey, respectively. Residue labels for human ADSL are in black, and those for Mtb ADSL are shaded and shown in yellow or dark orange.

The crystal structures of Mtb ADSL dimer obtained in two distinct forms are nearly identical, with a root-mean-square deviation (RMSD) of just 0.2 Å upon structural superimposition (Figure 2A, left panel). Given this high degree of similarity, all subsequent structural analyses and discussions are based on the higher-resolution structure (1.78 Å, form 2). Among known homologs, the Mtb ADSL structure most closely resembles that of M. smegmatis (Msm) ADSL (PDB ID: 4NLE),¹² with an RMSD of 0.5 Å, consistent with their high sequence identity of approximately 84.5%. This underscores the overall structural conservation of ADSL across mycobacterial species.

More broadly, ADSL appears to be highly conserved across species. Superimposition of the Mtb ADSL dimer and human ADSL (huADSL) tetramer structures (PDB ID: 2VD6)¹⁸ revealed an RMSD of 1.55 Å, further supporting a conserved overall fold (Figure 2A, right panel). Compared to the ligand bound form of human ADSL, the N-terminal domain of apo Mtb ADSL is more outwards away from the bound SAMP in human ADSL. The slightly different orientation could result from crystal packing, as this part in the apo form of Msm ADSL is closely resembling the human counterpart.

Despite the high overall structural conservation between mycobacterial and human ADSL, several notable differences are present in the active site, particularly among residues involved in binding of substrate and products (Figure 2C and Supplementary Figures S1 and S2). These subtle variations could be leveraged to design selective inhibitors that target the mycobacterial enzyme while sparing human homolog.

On one side of the SAMP-binding pocket (Figure 2C and Supplementary Figure S2A, S2D), which accommodates the succinate/fumarate moiety, residue M89 in human ADSL is substituted by K79 in both Mtb and Msm ADSLs. In the human ADSL-SAMP and ADSL-AMP-fumarate complex, M89, alongside the conserved H86, lies near the carboxyl group of the succinate moiety of SAMP and fumarate. In contrast, the substitution of M89 with the positively charged lysine (K79) in mycobacterial ADSL, together with H76 (the equivalent of human H86), may allow additional hydrogen bonding or the formation of salt bridges with the substrate or fumarate product. These interactions could modify substrate/product recognition or stabilization. Furthermore, T201 in human ADSL, which contributes to polar interactions with SAMP or fumarate, is replaced by P189 in mycobacterial ADSL (Figure 2C and Supplementary Figure S2C, S2D). This substitution introduces a nonpolar residue into the binding pocket, potentially altering the surface characteristics of the active site and affecting the binding of the substrate or product. Together, these variations may influence substrate affinity and specificity and provide a structural basis for designing inhibitors that selectively target mycobacterial ADSL while minimizing interaction with the human enzyme.

On the opposite side of the SAMP/AMP-binding pocket, which accommodates the ribose and phosphate groups, residue M299 in human ADSL (from a neighboring monomer within the tetramer) is replaced by R287 in mycobacterial ADSL (Figure 2C and Supplementary Figure S2A, S2B, S2D). This change introduces a positively charged side chain that may alter the local electrostatic environment, further supporting selective inhibitor design. Additionally, a conservative substitution of L331 in human ADSL to V319 in mycobacterial ADSL may slightly expand the pocket, allowing accommodation of bulkier or more flexible inhibitor scaffolds. Similarly, an A335-to-V323 from human to mycobacterial ADSL substitution at the phosphate-binding region increases hydrophobicity and may subtly reshape the pocket, offering additional avenues for differentiation. Additionally, substituting D333 in human ADSL with C321 in mycobacterial ADSL introduces a unique nucleophile for designing selective covalent inhibitors.

In the adenine-binding region, a pair of substitutions, C113 and R329 in the human enzyme replaced by R101 and G317 in mycobacterial ADSL, respectively, are also noteworthy (Figure 2C and Supplementary Figure S2A, S2D). These changes are likely to affect the binding mode of the adenine moiety of SAMP/AMP and may introduce novel opportunities for designing mycobacteria-specific inhibitors that discriminate against human homolog.

Collectively, these active site differences between Mtb and human ADSL provide a structural basis for the rational design of selective inhibitors targeting the mycobacterial enzyme, while minimizing off-target effects on the human counterpart.

Steady-state kinetics using a PPDK-based coupling luminescence assay

We explored multiple assay platforms to characterize the enzymatic activity of the Mtb ADSL. Several methods have previously been reported to quantify ADSL activity, including (i) monitoring absorbance changes at 282 nm during conversion of SAMP to AMP;^{12, 19} (ii) high-performance liquid chromatography (HPLC) to quantify SAMP and AMP;²⁰ and (iii) liquid chromatography–mass spectrometry (LC-MS)-based detection of reaction products.²¹ However, these techniques are limited by low throughput and poor signal-to-background (S/B) ratios (typically <1), making them unsuitable for HTS of inhibitors.

To overcome the limitations of previously reported low-throughput or low-sensitivity methods for quantifying ADSL enzymatic activity, we evaluated a bioluminescent assay based on a previously developed firefly luciferase-coupled platform designed to monitor the enzymatic activity of pyruvate phosphate dikinase (PPDK).²² PPDK catalyzes a reversible reaction between AMP, phosphoenolpyruvate (PEP), and inorganic pyrophosphate (PPi) to generate ATP, pyruvate, and phosphate. Given that AMP is also the product of the ADSL-catalyzed reaction, we hypothesized that the AMP generated by ADSL could be coupled to the PPDK-luciferase reaction, enabling sensitive and HTS-compatible quantification of ADSL activity via luminescence (Figure 3A).

Figure 3. PPDK-luciferase coupling ADSL enzyme assay.

(A) Principle of PPDK-luciferase coupling ADSL enzyme assay by AMP detection. (B) Determination of linear range of AMP detection by PPDK-luciferase coupling assay. For all PPDK-based assay, the following conditions were used unless specified: SAMP (20 μM or concentration series), PEP (3 mM)), PPi (42 μM), PPDK (50 nM), firefly luciferase (50 nM), luciferin (1.5 mM), CHAPS (0.005%). The reaction was performed in 40 μl volume in a buffer containing 5 mM MgSO₄, 7.5 mM (NH₄)₂SO₄, 1.0 mM dithiothreitol, 1.0 mM EDTA, 0.37% Sucrose, and 50 mM HEPES (pH 7.0). (C) Kinetics analysis of PPDK-based Mtb ADSL assay by titrating Mtb ADSL and SAMP. (D) Dose response curve of inhibition of the Mtb ADSL activity by concentration series of 4-HNE. N=3. For all dose-response curve fittings, data were normalized using DMSO-treated wells with MtbADSL set as 100%, and wells without MtbADSL set as 0%. (E) Well-to-well relative luminescence in 384-well format with and without MtbADSL; N=32. (F) HTS parameters for the optimized PPDK-luciferase coupling MtbADSL assay in a 384-well plate, by comparing the RLU generated with and without MtbADSL. N=32. ****, p<0.0001.

To establish this assay, we first codon-optimized, synthesized, cloned, expressed, and purified full-length Clostridium symbiosum (Cs) PPDK and firefly luciferase using the same approach employed for Mtb ADSL. Using purified enzymes, we determined the linear detection range of AMP via steady-state kinetic analysis of the PPDK-luciferase coupled reaction in the absence of ADSL. This assay demonstrated a robust linear response to AMP concentrations up to approximately 6 μM (Figure 3B), confirming its suitability for detecting low-micromolar levels of AMP.

We then applied the PPDK-luciferase assay to determine the steady-state kinetic parameters of ADSL (Figure 3C). The resulting values, K_m = 21.6 μM, k_cat = 1.8 s⁻¹, and k_cat/K_m = 0.083 μM⁻¹·s⁻¹, represent ~10-, ~18-, and ~120-fold improvements, respectively, over those obtained using traditional spectrophotometric method that measures SAMP absorbance at 282 nm.¹² These improvements underscore the superior sensitivity and dynamic range of the PPDK-luciferase coupled assay relative to conventional methods.

Inhibition of Mtb ADSL by 4-hydroxynonenal

To determine whether the assay could also be used for inhibitor screening, we evaluated the inhibitory effect of 4-hydroxynonenal (4-HNE), a lipid peroxidation product previously reported to inhibit human ADSL.²³ Given that 4-HNE is a reactive electrophile capable of covalently modifying proteins through Michael addition,^23-25 we hypothesized that it might also inhibit Mtb ADSL. Indeed, our results demonstrated dose-dependent inhibition of Mtb ADSL by 4-HNE, with an IC₅₀ value of 2.6 μM (Figure 3D), validating the use of the PPDK-luciferase assay for assessing inhibitor efficacy against ADSL.

Finally, we optimized the assay in a 384-well plate format to assess its suitability for HTS applications. The assay showed excellent performance metrics, including a Z’ factor of 0.75, a signal-to-background (S/B) ratio of 17, a signal window (SW)²⁶ of 10, and a coefficient of variation (CV) of 11.4% (Figure 3E,3F). These values meet or exceed NIH guidelines for HTS assay validation,²⁶ establishing the PPDK-luciferase assay as a robust and scalable platform for Mtb ADSL characterization and inhibitor screening.

Pilot HTS

To assess the utility of the PPDK-luciferase coupling assay in a high-throughput format, we performed a pilot screen using the NINDS small molecule library, which contains 960 compounds dispensed across three 384-well assay plates. The electrophilic compound 4-HNE was included as a known ADSL inhibitor and used as a positive control. The assay demonstrated excellent performance, yielding an average Z’ factor of 0.66, a S/B ratio of 19, a SW of 10.5, and a CV of 11.4% (Figure 4A). These parameters fall within the acceptable thresholds for HTS assays as defined by the NIH, which recommends Z’ ≥ 0.4, SW ≥ 2, and CV ≤ 20%.²⁶

Figure 4. Pilot high-throughput screening.

(A) Z’ factor, S/B ratio, and CV over plates of a pilot HTS against the NINDS Collection II compound library in three 384-well plates. (B) Screening flowchart. (C) Compound structures of hits. (D) Representative kinetic readings of dose-response inhibition of ADSL by a concentration series of bithionol. (E) Dose response inhibition of ADSL activity by bithionol and Tetrac. N=3.

From this primary screen, 16 compounds at a test concentration of 20 μM exhibited greater than 40% inhibition of Mtb ADSL enzymatic activity (Figures 4B and 4C, Supplementary Figure S3). After cheminformatic grouping and triage, including PAINS (pan-assay interference compounds) filtering,^{27, 28} seven representative candidate inhibitors were selected for further analysis. These compounds were repurchased as powder stocks for validation and further investigation.

Subsequent dose–response analysis confirmed that two of the seven repurchased compounds, bithionol (BT) and tetraiodothyroacetic acid (Tetrac or TT), displayed reproducible, concentration-dependent inhibition of Mtb ADSL, with IC_50-Mtb-PPDK values in the low micromolar range (Figures 4C-4E; Table 1).

Table 1.

Characterization of bithionol and Tetrac

	IC50-Mtb± (μM)				IC 50-human	SI #	MIC μg/ml (μM)*		CC50 (μM)$			TI ≠	ΔTm
Cpd	PPDK	FP	ADSL-free	AMP-Glo	PPDK		HR37a	HR37v⊥	A549	HEK	STTG1		(°C)
BT	1.6	6.2	21.6	1.2	14.7	9.2	3.1 (8.7)	9.9 (28)	32.4	19.3	139	1.2	1.8
TT	4.8	13.0	24.5	1.6	22.4	4.7	6.2 (8.3)	22 (30)	47.8	98	>200	1.6	0.3

Note: Cpd, compound; BT, bithionol; TT, Tetrac; ±, IC₅₀, defined as compound concentration required to inhibit enzyme activity by 50%; #, SI, selectivity index, defined as IC_50-human/IC_50-Mtb determined using the PPDK-coupling assay; *, numbers in parentheses are in μM; ⊥, MIC for H37Rv was estimated by EC₉₀ (effective concentration to inhibit bacterial growth by 90%). $, CC₅₀, cell cytotoxicity defined as compound concentration required to reduce cell viability by 50%; ≠, TI, therapeutic index defined as CC_50-A549/MIC_H37Rv; HEK, HEK293; STTG1, CCF-STTG1.

Development of a secondary assay based on fluorescence polarization (FP).

Because primary screening assays, particularly those involving coupled enzyme systems, are prone to artifacts due to compound aggregation, off-target inhibition, or interference with assay components, it is critical to validate hits using orthogonal approaches. To this end, we attempted to develop a secondary assay. Previously, we established an FP assay using the Transcreener^® AMP²/GMP² FP detection system (Bellbrook Labs) to quantitatively monitor AMP production during the UBA5-mediated UFMylation reaction.²⁹ Given that AMP is also the direct product of the ADSL-catalyzed reaction, we hypothesized that this FP assay could be repurposed to assess ADSL enzymatic activity. Supporting this idea, we observed a marked decrease in FP signal in reactions containing ADSL compared to control reactions lacking the enzyme (Figure 5A), indicating successful AMP production by ADSL. This established that the FP assay reliably detects ADSL enzymatic activity through AMP quantification. The FP-based assay showed strong performance, with a Z’ factor of 0.5, an S/B ratio of 3.7, a SW of 5.7, and a CV of 8.2% (Figure 5A), again meeting NIH HTS assay validation standards. ²⁶

Figure 5. Validation through secondary, ADSL-free, AMP-Glo^™ assays and selectivity assay against human ADSL.

(A) Assay performance for FP-based ADSL assay with and without ADSL. N=3. ****, p<0.0001. (B) Dose response inhibition of ADSL activity by bithionol and Tetrac using the FP assay. N=3. (C) Assay performance for ADSL-free PPDK-luciferase coupling assay with and without the AMP substrate. N=3. ****, p<0.0001. (D) Fitting of dose-response curve of inhibition of the coupling enzymes in the PPDK-luciferase coupling ADSL-free assay by compounds. AMP, 10 μM. Data were normalized using DMSO-treated wells with AMP set as 100%, and wells without AMP set as 0%. N=3. (E) Assay performance for AMP-Glo^™-based ADSL assay with and without ADSL. N=3. ****, p<0.0001. (F) Dose response inhibition of ADSL activity by bithionol and Tetrac using the AMP-Glo^™ ADSL assay. N=3. (G) Assay performance for human ADSL assay with and without huADSL. N=3. ****, p<0.0001. (H) Dose response inhibition of huADSL activity by bithionol and Tetrac using the PPDK assay. N=3.

We then applied the FP-based assay to retest the inhibitory activity of bithionol and Tetrac against Mtb ADSL. Both compounds demonstrated dose-dependent inhibition of Mtb ADSL activity in this orthogonal assay, yielding IC_50-FP values within a similar micromolar range (Figure 5B; Table 1). Minor differences in IC₅₀ values between the PPDK and FP assays are likely due to differences in assay sensitivity and enzyme concentrations, with the FP platform requiring a slightly higher ADSL concentration for optimal signal detection.

ADSL-free coupling enzyme assay

To evaluate whether the identified inhibitors directly affect the enzymes in the PPDK-luciferase coupling system, we implemented a ADSL-independent control assay. In this setup, AMP was introduced as the substrate for the PPDK and luciferase enzymes, alongside all necessary cofactors such as PEP and inorganic pyrophosphate (PPi), but without the inclusion of ADSL. As shown in Figure 5C, the PPDK-luciferase system clearly distinguished between the presence and absence of AMP, generating a strong luminescent signal only when AMP was included.

We subsequently evaluated the two hit compounds using an ADSL-free assay format to determine whether they interfered with the coupling enzymes, PPDK and luciferase. As shown in Figure 5D and summarized in Table 1, both compounds exhibited only modest inhibition of the coupling system. Notably, the IC₅₀ values for bithionol and Tetrac in the ADSL-free format (IC_{50-ADSL-free-PPDK}) were more than fivefold higher than their corresponding values in the presence of ADSL (IC_50-ADSL-PPDK). These findings indicate that the primary inhibitory effects of bithionol and Tetrac are directed toward ADSL rather than the enzymes in the coupling assay.

Validation using AMP-Glo^™ kit

To further validate the identified hits, we performed an additional experiment using the commercially available AMP-Glo^™ kit (Promega), which also utilizes a firefly luciferase-coupled system to detect ATP generated from an enzymatic cascade. The key difference between the PPDK-based luminescence assay and the AMP-Glo assay lies in the enzymatic steps used to convert AMP to ATP.³⁰ In the AMP-Glo assay, AMP is first converted to ADP by adenylate kinase (AK) in the presence of ATP through the reaction: ATP + AMP → 2 ADP. Following this conversion, an ATP-depleting enzyme such as adenylate cyclase is introduced to eliminate any residual input ATP by converting it to 3′,5′-cAMP and pyrophosphate. To prevent this enzyme from interfering with the next step, an adenylate cyclase inhibitor is subsequently added. This ensures that the newly generated ADP can then be converted to ATP by pyruvate kinase in the presence of PEP without interference, allowing accurate luminescence-based quantification of AMP.

One limitation of the AMP-Glo^™ kit is its reliance on an ATP-depleting enzyme, such as adenylate cyclase, to remove residual input ATP required for the AK reaction. This step is essential to distinguish between luminescent signals derived from pre-existing ATP and those generated de novo from AMP. However, if any test compounds interfere with the activity of the ATP-depleting enzyme, it may lead to false-negative results. In contrast, the PPDK-based luminescence assay does not require input ATP, making it inherently less susceptible to compound-induced interference with assay components.

Despite this limitation, the AMP-Glo^™ kit remains a useful alternative for evaluating compounds that do not affect the ATP-depletion step. Therefore, we employed this commercial assay as an orthogonal method to further validate the inhibitory activity of bithionol and Tetrac against ADSL. Our results showed that the AMP-Glo assay could robustly differentiate reactions containing ADSL from those without, exhibiting approximately a 514-fold increase in luminescence in the presence of ADSL (Figure 5E). Using this system, we confirmed that both bithionol and Tetrac inhibited ADSL in a dose-dependent manner, with IC₅₀ values closely matching those obtained from the PPDK-coupled assay (Figure 5F, Table 1).

Selectivity against human ADSL (huADSL)

To evaluate compound selectivity and potential off-target effects, we cloned, expressed, and purified human ADSL (huADSL) and performed inhibitory assays against the human enzyme. The assay showed strong discriminatory power, yielding ~35-fold higher luminescence in the presence of huADSL compared to enzyme-free controls (Figure 5G, Table 1). Both compounds showed moderate inhibition of huADSL (Figure 5H); however, their IC₅₀ values against the human enzyme were much higher than those observed for Mtb ADSL. This difference resulted in selectivity indices (SIs) of 9.2 for bithionol and 5.1 for Tetrac, indicating that both compounds exhibit moderate but meaningful selectivity for the mycobacterial enzyme over the human homolog.

MIC determination and Cytotoxicity

To evaluate the antibacterial activity of bithionol and Tetrac, we determined their minimum inhibitory concentrations (MICs) against both the attenuated M. tuberculosis H37Ra strain (ATCC 25177) and the virulent H37Rv strain, the most extensively characterized model for M. tuberculosis pathogenesis and drug susceptibility. The MIC was defined as the lowest compound concentration that completely inhibited visible bacterial growth compared to the untreated control, or as the concentration required to achieve a 90% reduction in growth. Inclusion of both strains allowed evaluation of potential differences in compound potency arising from variations in membrane and carrier proteins between H37Ra and H37Rv, which have been reported to affect antibiotic susceptibility, such as with rifampicin.³¹

As shown in Table 1, bithionol and Tetrac exhibited MIC values of 3.1 μg/ml and 6.2 μg/ml, respectively, against M. tuberculosis H37Ra, demonstrating their potential as anti-tubercular agents. Against the wild-type virulent strain M. tuberculosis H37Rv, both compounds showed reduced potency, with MIC values of 9.9 μg/ml for bithionol and 22 μg/ml for Tetrac (Supplementary Figure S4).

To further evaluate the therapeutic potential of these compounds, we assessed their cytotoxicity in multiple cell lines. Since M. tuberculosis primarily infects the lungs but can also disseminate to the central nervous system (CNS), causing meningoencephalitis, intracranial tuberculomas, and cerebral abscesses,^32-34 we selected three representative cell types: human embryonic kidney (HEK293), lung carcinoma (A549), and astrocytoma-derived CCF-STTG1 cells. Cells were treated with increasing concentrations of bithionol and Tetrac, up to 200 μM, and viability was measured using the Cell Counting Kit-8 (CCK-8) assay, as described in our previous studies (Table 1, Figure 6).^35-38

Figure 6. Cell viability against HEK293, A549 and CCF-STTG1 cells.

HEK293, A549 and CCF-STTG1 cells (ATCC) cells were incubated with various concentrations of compounds, and then viability was assayed at 48 h of incubation, using the WST assay. Data were normalized by setting the absorption readings of cells treated with DMSO to 100% and the readings from cell-free wells treated with DMSO to 0%. N = 2.

Bithionol exhibited moderate cytotoxicity, with CC₅₀ values of 19.3 μM and 32.4 μM against HEK293 and A549 cells, respectively. In contrast, Tetrac was less cytotoxicity, with CC₅₀ values of 98 μM and 47.8 μM in the same cell lines (Table 1). Notably, both compounds showed minimal toxicity toward astrocyte-like CCF-STTG1 cells, with CC₅₀ values of 139 μM for bithionol and >200 μM for Tetrac.

Using A549 lung epithelial cells as a reference model for the pulmonary tropism of M. tuberculosis, the therapeutic indices (TIs) of bithionol and Tetrac varied between the attenuated H37Ra and virulent H37Rv strains. For H37Ra, the CC₅₀ values were approximately 3.7–5.8-fold higher than their respective MICs, indicating a favorable selectivity window. In contrast, against H37Rv, the CC₅₀ values were only 1.2–1.6-fold higher than the MICs, reflecting a more modest therapeutic margin. At the MICs determined for H37Rv, bithionol reduced A549 cell viability by ~45%, whereas Tetrac treatment maintained ~71% viability. These results indicate that both compounds display on-target antimycobacterial activity with moderate selectivity, providing a promising foundation for further optimization to enhance safety and therapeutic potential.

Investigation of mechanism via AMP rescue assays

To understand the mechanism of inhibition and determine whether blocking the de novo AMP biosynthesis pathway could be circumvented through activation of the purine salvage pathway, we carried out MIC assays in the presence of exogenous AMP. Supplementing mycobacterial cultures with AMP allowed us to assess whether external purine sources could rescue growth from compound-mediated inhibition. This approach provides a direct measure of on-target engagement, since reversal of growth inhibition by AMP supplementation would indicate that the compounds act within the AMP biosynthetic pathway rather than through nonspecific toxicity. Using this strategy, we evaluated the extent to which exogenous AMP could modulate the inhibitory effects of bithionol and Tetrac at their MIC concentrations (Figure 7A).

Figure 7. AMP rescue and phagocytosis assaya.

(A) Rescue assay using AMP supplementation in M. tuberculosis. Mtb H37Ra was treated with Bithionol and Tetrac at MIC in the presence of increasing concentrations of AMP (0–5 mM). Bacteria were cultured in 7H9 complete medium and incubated at 37°C for 7 days. Media only and untreated bacterial cultures served as controls. Growth rescue was assessed relative to inhibitor-only controls. Experiment was done in triplicate. Cpds, compounds. (B) Bacterial load in RAW 264.7 macrophage cell lysate. The compound treatment groups were compared with the DMSO control by one-way ANOVA using GraphPad Prism. **, p < 0.001; ***, p < 0.0001. N=2. (C) Cell viability against RAW264.7 cells. Data were normalized by setting the absorption readings of cells treated with DMSO to 100% and the readings from cell-free wells treated with DMSO to 0%. N = 2.

Our results showed that supplementation with AMP significantly mitigated the growth-inhibitory effect of Tetrac, with mycobacterial growth largely restored at 1.25 mM AMP. In contrast, bithionol’s inhibitory effect remained robust, with substantial rescue only observed when AMP concentrations reached 5.0 mM. The results were consistent with our findings that bithionol was more potent than Tetra in inhibition of MtbADSL and mycobacterial growth.

Overall, because extracellular AMP needs to be imported and processed via salvage pathways, these observations support that both compounds act on-target by inhibiting the de novo AMP biosynthesis pathway. Tetrac is more readily bypassed by the salvage pathway, while bithionol’s stronger inhibition requires significantly higher AMP levels to overcome, consistent with its superior efficacy.

Phagocytosis assay

M. tuberculosis is a facultative intracellular pathogen, with alveolar macrophages and other innate immune phagocytes serving as the primary cellular defense in the lungs. Thus, targeting and reducing intracellular M. tuberculosis within macrophages is a critical strategy for controlling tuberculosis infection.

In this study, we employed an intracellular infection model using RAW264.7 murine macrophages to evaluate the efficacy of bithionol and Tetrac to suppress intracellular M. tuberculosis replication. RAW264.7 cells were allowed to phagocytose opsonized M. tuberculosis and were subsequently treated with inhibitors or a DMSO vehicle control. As shown in Figure 7B, both compounds suppressed intracellular bacterial growth in a dose-dependent manner, demonstrating greater potency in inhibition of intracellular bacterial growth than against extracellular bacteria alone. Bithionol was particularly effective, reducing intracellular Mtb growth by ~2.3 log units at 5 μM and achieving a 4.6-log reduction at 15 μM. Tetrac, while less potent, still reduced intracellular burden by ~1 log at 5 μM and ~2.5 log at 15 μM.

To exclude the possibility that these effects resulted from host cell toxicity, we assessed cytotoxicity in RAW264.7 cells under the same treatment conditions. At 15 μM, RAW264.7 macrophages treated with bithionol maintained approximately 75% viability, whereas those treated with Tetrac remained fully viable, comparable to untreated controls (Figure 7C). These results confirm that the observed antibacterial effects were not due to compound-induced cytotoxicity.

Together, these results demonstrate that bithionol and Tetrac effectively suppress intracellular M. tuberculosis replication within macrophages, supporting their potential as therapeutic agents to limit bacterial survival and dissemination during tuberculosis infection.

Protein thermal shift assay

To investigate the mode of action of the compounds, we examined whether they directly bind to ADSL using a protein thermal shift assay (TSA), as previously described.³⁹ The TSA results revealed differential thermal stabilization of MtbADSL by the two compounds. While Tetrac induced a modest increase in the melting temperature (T_m) of ADSL by 0.7 °C, bithionol at 10 μM caused a more pronounced shift of 1.3 °C (Figure 8A). Given that a change of ≥0.5 °C in T_m is generally considered indicative of significant binding, these results suggest that both bithionol and Tetrac interacts directly with ADSL, with bithionol exhibiting stronger binding affinity to MtbADSL than Tetrac, consistent with our enzyme inhibition data.

Figure 8. Protein thermal shift assay.

(A) Protein TSA using real-time PCR system. N=3. BT, bithionol; TT, Tetrac. (B) Evaluation of protein thermal stability using a gel-based assay.

To further validate these findings, we performed a gel-based thermal shift assay, which confirmed dose-dependent stabilization of MtbADSL by both compounds, thereby reinforcing their direct binding to the enzyme. In this format, both bithionol and Tetrac conferred dose-dependent protection of MtbADSL from heat-induced denaturation and aggregation (Figure 8B). Notably, although Tetrac showed only a moderate shift in T_m in the TSA, the gel-based assay clearly demonstrated its ability to stabilize MtbADSL, supporting its direct interaction with the enzyme. These results validate the TSA findings for bithionol and reinforce Tetrac’s binding activity through orthogonal biophysical evidence.

Competition with 4-HNE

To further explore the inhibition mechanism, we examined whether the two identified hits compete with 4-HNE for binding to MtbADSL using thermal shift assays (Figure 8A). The results showed that 4-HNE alone produced only a modest increase in the T_m of MtbADSL (0.4 °C). When co-incubated with bithionol, however, the T_m increased by 2.1 °C, slightly greater than the additive effect of each compound alone, suggesting that bithionol and 4-HNE likely bind to distinct sites on MtbADSL and may act cooperatively. In contrast, co-incubation of 4-HNE with Tetrac reduced the T_m of MtbADSL by ~0.6 °C, indicating an antagonistic interaction. These observations suggest that while bithionol and 4-HNE may stabilize MtbADSL through non-overlapping binding modes, Tetrac and 4-HNE appear to interfere with one another’s interactions. Collectively, these findings provide new insight into the binding mechanisms of bithionol and Tetrac and their potential interplay with known inhibitors of MtbADSL.

Docking studies

To elucidate the binding mode and potential mechanism of inhibition, in silico docking of bithionol and Tetrac was performed using Schrödinger Maestro based on the crystal structure of MtbADSL. SiteMap analysis identified ten potential binding pockets with site scores above 0.8, corresponding to five unique sites duplicated by the tetrameric symmetry of MtbADSL (Supplementary Figure S5A-S5F). Both compounds preferentially docked to a high-scoring pocket 1 (site score = 0.9) located near the AMP-binding site at the interface of chains A and D (Supplementary Figure S5A, S5B).

Induced-fit docking revealed that bithionol forms a hydrogen bond between the hydroxyl group of its dichlorophenol ring and W20, and a halogen bond between the opposite chlorine atom and R326 (Figure 9A,9B). In contrast, Tetrac forms a salt bridge between its carboxylate group and R83, and halogen bonds between its diiodophenol moiety and residues E108 and R326 (Figure 9C,9D).

Figure 9. Predicted binding models of bithionol and Tetrac with MtbADSL based on induced-fit docking.

(A, C) Three-dimensional binding models showing polar interactions between bithionol (A) or Tetrac (C) and MtbADSL. Atom colors: nitrogen, blue; oxygen, red; sulfur, yellow; carbon, purple (bithionol), orange (Tetrac), and gray (MtbADSL). The MtbADSL structure is shown in ribbon representation, while ligands are displayed as sticks. Salt bridges are depicted as light purple dashed lines, hydrogen bonds in yellow, and halogen bonds in dark purple. (B, D) Two-dimensional interaction maps for bithionol (B) and Tetrac (D). Salt bridges are shown in blue–red gradients, halogen bonds in yellow, and hydrogen bonds in purple, with arrows indicating donor or acceptor roles. (E) Superposition of predicted bithionol and Tetrac binding models with the human ADSL–AMP–fumarate complex (PDB ID: 2VD6). The human complex is shown in light gray, the predicted bithionol binding model in plum, and the predicted Tetrac binding model in orange.

In human ADSL, when complexed with the substrate or the AMP/fumarate products, the domain spanning residues S31–N126 (corresponding to S21–R116 in M. tuberculosis ADSL) adopts a closed conformation (Figure 9E, Supplementary Figure S6A).¹⁸. In contrast, in the absence of product or substrate binding, the corresponding domain in M. tuberculosis ADSL remains in an open conformation. Comparison of the substrate-bound and apoenzyme structures reveals that upon AMP binding, residues R75, H76, and D77 undergo conformational shifts of approximately 6.3 Å, 4.7 Å, and 3.5 Å, respectively, facilitating closure of the AMP-binding domain. In the bithionol-bound model, however, the compound appears to introduce unfavorable close contacts that prevent domain closure and thus disrupt formation of the AMP-binding pocket. Moreover, the Cζ atom of R326, which normally shifts rightward to stabilize the phosphate group of AMP, instead moves toward the AMP-binding site in the bithionol-bound conformation, producing a steric clash likely to hinder AMP association. These structural observations suggest that bithionol inhibits MtbADSL by stabilizing an open, catalytically inactive conformation of the enzyme.

Similarly, Tetrac binding also prevents the domain from adopting the closed conformation required for catalysis, thereby blocking formation of the AMP-binding pocket. In addition, residue V323 shifts rightward by approximately 1.5 Å in the Tetrac-bound model, a displacement that may further impede AMP accommodation and contribute to enzyme inhibition (Supplementary Figure S6B).

Discussion

Purine biosynthesis is a fundamental metabolic pathway required for DNA, RNA, and cofactor synthesis, making it essential for cellular survival and proliferation. In Mtb, the de novo purine biosynthetic pathway is critical for pathogenesis and survival in host environments.^7-10 Among the enzymes in this pathway, ADSL catalyzes two key steps: the conversion of SAICAR to AICAR and SAMP to AMP, which are crucial for purine nucleotide formation in bacteria.⁴⁰ Genetic studies have confirmed that purB is essential for Mtb growth and virulence, underscoring its potential as a drug target.⁷

Despite its importance, ADSL has not been extensively explored for drug development in tuberculosis. Our study addresses this gap by combining assay development, inhibitor screening, and structural analysis to characterize Mtb ADSL and enable structure-guided inhibitor discovery.

We also solved two high-resolution crystal structures of apo Mtb ADSL (1.78 Å and 2.1 Å), which revealed a conserved tetrameric structure and active-site architecture. Despite this conservation, structural alignment with human ADSL highlighted key residue differences at substrate-binding sites, including K79, R101, P189, R287, G317, V319, C321, and V323 in Mtb, which are promising targets for species-selective inhibitor design. These differences offer a foundation for developing inhibitors that can selectively target Mtb ADSL without affecting human homolog.

We also developed a highly sensitive, high-throughput-compatible luminescence-based assay using a PPDK-luciferase coupling system to quantify AMP production, a direct readout of ADSL enzymatic activity. This assay enabled robust kinetic analysis and pilot HTS, identifying bithionol and Tetrac as low micromolar inhibitors.

Bithionol, a clinically approved anthelmintic agent originally used for the treatment of fascioliasis and other parasitic infections,⁴¹ has also been recognized for its broad-spectrum antimicrobial properties. Notably, it has shown potent activity against Mycobacterium abscessus, with reported MICs ranging from 0.625 to 2.5 μM.⁴² Previous studies have observed that bithionol induces abnormal cell morphology, membrane disruption, and cell lysis in mycobacteria, suggesting a bactericidal effect; however, the precise molecular mechanism underlying its antimycobacterial activity remained unclear.⁴² In the present study, we identified Mtb ADSL as a direct molecular target of bithionol. Our biochemical and biophysical analyses demonstrate that bithionol binds to and inhibits Mtb ADSL in vitro, providing the first evidence of a defined mechanism of action. This finding not only sheds light on the antibacterial activity of bithionol but also highlights its potential for repurposing as a selective inhibitor targeting the purine biosynthesis pathway in M. tuberculosis.

Tetrac is a deaminated analog of L-thyroxine that functions as an antagonist at the integrin α_vβ₃ receptor, thereby inhibiting the proliferative and proangiogenic actions of thyroid hormones.⁴³ While Tetrac has been extensively studied for its anti-cancer properties, particularly in blocking tumor cell proliferation and angiogenesis, there is limited information regarding its antimycobacterial activity. In our study, Tetrac was identified as a low micromolar inhibitor of Mtb ADSL. Although the exact mechanism by which Tetrac exerts its antimycobacterial effects remains to be fully elucidated, our findings suggest that its inhibition of ADSL contributes to its activity against M. tuberculosis. This represents a novel mechanism of action for Tetrac and highlights its potential for repurposing as an anti-tubercular agent.

To confirm that the observed inhibitory activity was due to direct action on ADSL and not on components of the coupling assay, we employed orthogonal validation methods. These included an FP assay, ADSL-free assay, and the AMP-Glo^™ detection system, all of which supported the initial screening results. Bithionol and Tetrac demonstrated notable inhibitory effects against both Mtb and human ADSLs in the PPDK assay, with a 4.7- to 9.2-fold difference in potency, favoring inhibition of the mycobacterial enzyme. This preferential inhibition underscores their potential as selective anti-tubercular agents while reducing the likelihood of off-target effects on the human enzyme.

In support of on-target engagement, rescue experiments with exogenous AMP, the direct product of the MtbADSL reaction, restored the growth of M. tuberculosis treated with either compound, confirming that their antibacterial activity derives from inhibition of the de novo AMP biosynthesis pathway. Notably, rescue required relatively high AMP concentrations: growth inhibition by Tetrac was reversed at ≥1.25 mM AMP, while suppression by the more potent bithionol persisted until AMP reached 5.0 mM. Since endogenous AMP levels in human plasma and cells are typically in the nanomolar to low micromolar range,⁴⁴ they are far below the threshold needed to bypass the inhibitory effects of these compounds. This disparity strongly suggests that physiological AMP levels would not rescue M. tuberculosis in vivo, providing a favorable therapeutic window.

Mtb is a facultative intracellular pathogen that can persist and replicate within host phagocytic cells, particularly macrophages. It has evolved sophisticated mechanisms to survive within the hostile intracellular environment, including manipulation of host immune responses and disruption of cellular processes such as mitochondrial function and apoptotic signaling.^{45, 46} In our study, RAW264.7 murine macrophages infected with M. tuberculosis were treated with bithionol and Tetrac. A marked, dose-dependent decrease in intracellular bacterial burden was observed at concentrations of the inhibitor that did not induce cytotoxicity in the RAW264.7 cells. This indicates that the reduction in bacterial load was not a secondary effect of host cell death. Microscopic analysis of the treated macrophages confirmed the absence of morphological signs of toxicity. Since the lungs are the primary site of M. tuberculosis infection, candidate therapeutics must reach pulmonary tissues while minimizing toxicity toward lung cells. In A549 epithelial cells, bithionol and Tetrac displayed CC₅₀ values of 32.4 μM and 47.8 μM, respectively, corresponding to only ~1.2–1.6-fold higher concentrations than their MICs against H37Rv. These results suggest modest selectivity, indicating that while both compounds exhibit clear on-target antimycobacterial activity, further optimization will be important to improve the therapeutic index and ensure safety in lung tissue contexts. Notably, both compounds showed differential cytotoxicity across cell types: although moderately toxic to A549 and HEK293 cells, their CC₅₀ values against astrocyte-like cells were much higher (~139 μM for bithionol and >200 μM for Tetrac). This variability highlights cell-type–specific responses and suggests that, if these compounds were to cross the blood–brain barrier, they may offer a more favorable safety margin in the CNS compared to lung or kidney-derived cells.

Biophysical studies further supported direct target engagement. Protein TSA revealed that bithionol significantly increased the melting temperature of MtbADSL, consistent with direct binding. A gel-based thermal stability assay further supported this conclusion.

In summary, we established robust biochemical assays, solved crystal structures, and validated bithionol and Tetrac as chemical leads targeting the purine biosynthesis pathway in M. tuberculosis. The combination of structural insights, biochemical selectivity, AMP rescue validation, and favorable cytotoxicity profiles positions these compounds as promising scaffolds for further optimization. These advances are particularly timely given the global rise in drug resistant tuberculosis and the urgent need for novel therapeutic strategies.

Experimental Section

Materials purchased

Phospho(enol)pyruvic acid monosodium salt hydrate (PEP) (Millipore Sigma, P0564); sodium pyrophosphate decahydrate (PPi) (ThermoFisher Scientific, 448061000); adenyl succinic Acid (ammonium salt) (SAMP) (Caymen Chemical, 34334).

Expression, purification, and characterization of M. tuberculosis and human ADSLs

The Mtb purB gene with a C-terminal His-tag was codon-optimized, synthesized, and cloned into the pET-26b(+) plasmid using NdeI and EcoRI restriction sites (GeneUniversal). The expression plasmid MtbADSL-p26 was transformed into E. coli Rosetta (DE3) cells, which were then cultured in Luria broth medium supplemented with 100 μg/mL kanamycin at 37 °C until the optical density (OD) at 600 nm reached 0.6. The cells were induced with 0.3 mM IPTG and further incubated with shaking at 150 rpm and 16 °C for 18 hours. After incubation, the cells were collected by centrifugation at 8000 rpm for 20 minutes. The cell pellets were resuspended in lysis buffer (20 mM HEPES, pH 7.5, 200 mM NaCl), and lysozyme along with phenyl methyl sulfonyl fluoride (PMSF) was added. Cells were lysed using sonication and then centrifuged at 15,000 rpm at 4 °C for 30 minutes. The supernatant was loaded onto a pre-charged Ni-NTA affinity column (Qiagen) and washed with the resuspension buffer containing 30 mM imidazole. His-tagged MtbADSL was eluted using lysis buffer containing 80mM - 200mM imidazole. The MtbADSL was further purified by size exclusion chromatography using a 75 Superdex column, and peak fractions were collected and pooled. The purified MtbADSL was stored in a buffer containing 25 mM HEPES, pH 7.5, 150 mM NaCl, and 2 mM DTT.

The human ADSL gene was codon-optimized, synthesized and cloned into the pET-28_3C plasmid using BamH1 and EcoRI restriction sites (GeneUniversal). The expression and purification of ADSL was performed using the same protocol described above for the Mtb ADSL.

Expression and purification of pyruvate phosphate dikinase (PPDK) and firefly luciferase

The PPDK gene of Clostridium symbiosum (Cs) was codon-optimized, synthesized, and cloned into the pET-28a plasmid using NdeI and EcoRI restriction sites (GeneUniversal). Full-length firefly luciferase was generated by reassembling previously constructed split luciferase fragments using a megaprimer-based PCR strategy, following established protocols.^{47, 48} The expression and purification of CsPPDK and firefly luciferase were performed using the same protocol described above for Mtb ADSL.

Circular dichroism (CD)

Circular dichroic spectra were measured in a Chirascan^™ circular dichroism spectrometer. Far ultraviolet spectra were measured from 260 to 200 nm, bandwidth of 1 nm, using 100 ml of solution in a 0.1 mm path length cuvette. The protein concentration was measured spectrophotometrically and used at 0.2 mg/ml. Data were obtained in milli degrees (mdeg) and converted into Delta Epsilon (De) for estimation of secondary structure using CDNN software⁴⁹ using a mean residue weight of 113, as we described previously.^{50, 51}

Crystallization, data collection and structure determination

Initial crystallization conditions for ADSL were identified using commercial screens from Hampton Research, including the Index, Crystal Screen, and Wizard kits. Based on these initial hits, conditions were further optimized to improve crystal quality. Crystals of form 1 were grown at room temperature using the hanging drop vapor diffusion method by mixing 2 μL of protein solution (17.8 mg/mL) with 2 μL of reservoir solution containing 0.2 M Li₂SO₄, 0.1 M Tris-HCl buffer (pH 8.5), and 30–35% PEG 4000. Crystals of form 2 were grown at room temperature using the hanging drop vapor diffusion method by mixing 2 μL of protein solution (13 mg/mL) with 1 μL of reservoir solution containing 32-35% Tacsimate, pH 7.0, 3-4% MPD. Prior to data collection, crystals were transferred to reservoir solution supplemented with 25% glycerol for cryoprotection, flash-cooled in a nitrogen stream at 100 K, and stored in liquid nitrogen. X-ray diffraction data for crystal forms 1 and 2 were collected at the beamlines NYX (19-ID) of Brookhaven National Laboratory and BL9-2 of Stanford Synchrotron Radiation Laboratory (SSRL) at 100 K to resolution of 2.1Å and 1.78Å, respectively. Diffraction data for crystal form 1 were processed and scaled using autoPROC,⁵² whereas data for crystal form 2 were processed with HKL2000⁵³ (Supplementary Table 1).

The crystal structures were determined by using the molecular replacement method with the crystal structure of ADSL of M. smegmatis (PDB: 4NLE¹²) as a search model using the PHENIX program suite.⁵⁴ The structure refinement was carried out using the PHENIX program suite,⁵⁴ and monitored and modeled using Coot.⁵⁵

Steady-state kinetics of PPDK-based luminescence AMP-detection assay

The PPDK-based luminescence AMP-detection assay was modified according to the protocol developed by Sakakibara et al.²² Steady-state enzyme kinetics were assessed at RT using a reaction buffer A composed of 5 mM MgSO₄, 7.5 mM (NH₄)₂SO₄, 1.0 mM dithiothreitol, 1.0mM EDTA, 0.37% Sucrose, and 50 mM HEPES (pH 7.0).

To establish the PPDK-AMP luminescence assay for Mtb ADSL, steady-state kinetic analyses were performed to determine the enzyme's kinetic parameters. Reactions were assembled in solid white-bottom 384-well plates. MtbADSL at varying concentrations was incubated with different concentrations of the following reagents: SAMP, PPDK, PEP, PPi, luciferase, luciferin, and 0.005% CHAPS. For each experiment, a single reagent was varied across different concentrations while all other components were maintained at fixed concentrations. This design enabled the evaluation of the effect of each reagent on MtbADSL activity under controlled conditions. Luminescence was measured in real-time at 37 °C using a BioTek H1 Reader with a 2-second integration time, full light emission, top optics, and a gain setting of 200. Reaction velocities at each component concentration were calculated and fitted to the Michaelis–Menten equation using GraphPad Prism 9. All experiments with quantitative data were conducted in triplicate unless otherwise stated.

Pilot high-throughput screening using the PPDK-AMP assay

For high-throughput screening, we utilized the NINDS Collection II library, which contains 1,024 pharmacologically active compounds distributed across fifteen 96-well plates. To facilitate screening in a high-density format, the first 960 compounds were reformatted and transferred to three 384-well plates. These plates were then screened using the PPDK-AMP luminescence assay described above, with the aim of identifying small-molecule inhibitors of Mtb ADSL. The HTS was performed with the following components: SAMP (20 μM), PEP (3 mM), PPi (42 μM), PPDK (50 nM), firefly luciferase (50 nM), luciferin (1.5 mM), CHAPS (0.005%) in buffer A. Each compound was tested at a fixed concentration of 20 μM during the screening. DMSO served as a negative control for each plate screening, while 4-HNE at 20 μM was employed as a positive control for inhibition. The quality of the HTS assay was assessed by analyzing the Z’-factor, S/B ratio, SW, and CV for each plate.

Fluorescence polarization (FP) Assay

AMP levels in the samples were quantified using the Transcreener^® AMP/GMP FP Assay Kit (Bellbrook), following the manufacturer’s instructions. Reactions were set up in a total volume of 20 μL in a 384-well black plate, comprising 15 μL of enzymatic reaction mixture and 5 μL of detection reagent. MtbADSL (1 μM) was preincubated with varying concentrations of test compounds (0–100 μM) for 1 hour. Subsequently, SAMP (30 μM) was added, and the reaction proceeded for 45 minutes at room temperature. After incubation, 5 μL of the Transcreener^® AMP/GMP FP Detection Reagent containing the fluorescent probe was added. Plates were incubated for an additional 1.5 hours to develop the fluorescent signal. Fluorescence polarization was measured using a fluorescence plate reader, with the degree of polarization directly reflecting AMP production. IC₅₀ values for the compounds were determined using GraphPad Prism 9.

ADSL-free coupling assay

Test compounds were evaluated at various concentrations. The PPDK-luciferase reagents were utilized for this assay with AMP as the substrate in the absence of MtbADSL. Following a 30-minute incubation, PPDK (50 nM), luciferase (50 nM), luciferin (1.5 mM), CHAPS (0.005%) were added to the reaction. AMP added to a final concentration of 10 μM. Plates were read at 37 °C with a luminescence endpoint, an integration time of 2 minutes, full light emission, top optics, and gain set to 150. Appropriate controls were included. The experiment was performed in triplicate.

AMP detection assay using the AMP-Glo^™ kit

The AMP Glo^™ Assay Kit from Promega was utilized to detect AMP in samples following the kit's instructions. MtbADSL (0.5 μM) was incubated for 1 hour with various concentrations of compounds (0-25 μM). Following this, SAMP (10 μM) was added and incubated for an additional 45 minutes. Then, 5 μL of AMP Glo^™ Reagent I was added to each well and incubated at RT for 1 hour. The reaction was completed by adding 10 μL of AMP Glo^™ Reagent II and incubating for 30 minutes at RT, enabling the enzyme reaction. Luminescence was measured in each well using a BioTek H1 Reader, and the data were analyzed relative to a standard curve to determine AMP concentration. The IC₅₀ values of the compounds were calculated using the GraphPad Prism 9.

Activity against human ADSL using PPDK-AMP assay

Reactions were carried out in 384-well plates with a final volume of 40 μL per well, consisting of 20 μL enzymatic mix and 20 μL detection reagent. huADSL (0.3 μM) was preincubated for 1 hour with test compounds at concentrations ranging from 0 to 100 μM. Each reaction included SAMP (20 μM), PEP (3 mM), PPi (42 μM), PPDK (50 nM), firefly luciferase (50 nM), luciferin (1.5 mM), and CHAPS (0.005%) in buffer A. Luminescence was monitored in real-time at 37 °C using a BioTek H1 plate reader with top optics, full emission capture, 2-second integration, and a gain setting of 200. IC₅₀ values were determined using GraphPad Prism 9. All assays were performed in triplicate.

Protein thermal shift assay (TSA) using Real-Time PCR System

The TSA was conducted using an Applied Biosystem 7500 Fast Real-Time PCR System (ThermoFisher Scientific) from 4 to 90 °C, with methods we described previously.³⁶ Briefly, the MtbADSL at a final concentration of 3 μM in 1× PBS was mixed with each compound or each compound together with 4-HNE to attain a 10 μM final concentration for 30 min at 25 °C. TSA was carried out in the MicroAmp Fast Optical 96-Well Reaction Plate (ThermoFisher Scientific). Thermal denaturation was monitored using SYPRO Orange dye (5X) (Life Technologies) according to the manufacturer’s manual. The T_m was calculated, using a Derivative model using the Protein Thermal Shift Software v1.0 (ThermoFisher Scientific). ΔT_m was defined as the difference between T_m-DMSO and T_m-Compound.

Gel-based TSA

Thermostability of MtbADSL was monitored for the detection of binding events between MtbADSL and compounds by following the protocol as described earlier.⁵⁶ Briefly, 5 μM MtbADSL per reaction (in 1xPBS) was incubated with 15 μM or 30 μM of compounds. Protein in control reactions was incubated with an equivalent volume of DMSO. Following incubation, reaction mixtures were heated on Mastercycler^™ Nexus Thermal Cycler (Eppendorf^™) at different temperatures ranging from 4°C to 90°C for 10 min and then cooled down to room temperature for 4 min. Following centrifugation at 16,800g for 40 min at 4°C, SDS-PAGE of supernatants was performed.

Cytotoxicity

Cytotoxicity and cell viability were measured using the cell counting kit-8 (CCK-8) (GLPBIO) according to the manufacturer’s protocol, with minor modifications. HEK293, A549 CCF-STTG1, and RAW264.7 cells were cultured in DMEM media supplemented with 10% FBS at 37 °C in a humidified atmosphere of 5% CO₂. A total of 1.5 × 10⁵ HEK293, A549, CCF-STTG1 and RAW264.7 cells were seeded in duplicate in a 96-well plate and incubated for 24 hours in a CO₂ incubator. Cells were treated with MtbADSL inhibitors at concentrations ranging from 0.32 to 200 μM for 48 hours at 37 °C. Following treatment, 10 μL of CCK-8 was added to each well and incubated at 37 °C for 1 to 4 hours. DMSO served as a control. Absorbance was measured at 460 nm using a BioTek Synergy HI microplate reader. The cytotoxic concentration (CC₅₀) was determined via nonlinear regression on dose-response curves using GraphPad Prism 9. Data was normalized by setting the absorption readings of cells treated with DMSO to 100% and the readings from cell-free wells treated with DMSO to 0%. N = 2.

Antibacterial susceptibility assay

Susceptibility testing was performed using a two-fold serial broth dilution method⁵⁷ in 96-well flat-bottom microtiter plates was assessed against the live-attenuated Mycobacterium tuberculosis H37Ra. Experiments were conducted in sterile Middlebrook 7H9 broth supplemented with 10% ADC (BD Biosciences, USA), 0.2% glycerol, and 0.05% Tween-80. The bacteria were cultured to mid-log phase at 37 °C with shaking in the same medium. Test samples were serially diluted to achieve final concentrations ranging from 50 μg/mL to 0.39 μg/mL. All wells, except for the negative controls, were inoculated with bacterial suspensions at a final density of 5 × 10⁵ CFU/mL. Plates were then incubated at 37 °C for one week. Rifampicin, prepared over the same concentration range, served as a positive control. Plates were incubated at 37 °C in a 5% CO₂ atmosphere for one week. Bacterial growth inhibition was assessed by the addition of 20 μL AlamarBlue HS Cell Viability Reagent and plates were re-incubated for 2-4 h. The MIC was determined based on the concentration of compounds that prevented the color change from pink (growth) to blue (no growth).

M. tuberculosis H37Rv was inoculated into 20 mL of Middlebrook 7H9 + 10%OADC + 0.5% glycerol + 0.05% Tween-80 from glycerol stock and incubated at 37 °C for 1 week until mid-log phase. Two-fold serial dilutions of Bithionol (BT) or Tetrac (TT) were prepared in 2 mL 7H9 10%OADC + 0.5% glycerol + 0.05% Tween-80, and concentrations ranging from 32 μM to 1 μM, including 0 μM as a no-drug control. To each of 2 mL serially diluted compound containing medium, 200 μL of mid-log phase H37Rv (OD₆₀₀ of 0.4) were added, to result in a culture with ~OD of 0.04. Cells were incubated with each compound for 10 days. The OD₆₀₀ was measured at each drug concentration and analyzed using GraphPad Prism. MIC was estimated as EC₉₀ by fitting the experimental data. The experiment was performed in triplicate.

Rescue assay using AMP supplementation in Mtb

To assess whether exogenous AMP can rescue the growth inhibitory effect of MtbADSL-targeting compounds, a rescue assay was performed using Mtb H37Ra treated with the bithionol and Tetrac at MIC, in the presence of different concentrations of AMP (0-5 mM). M. tuberculosis was cultured in Middlebrook 7H9 broth supplemented with 10% OADC, 0.05% Tween-80, and 0.5% glycerol until mid-log phase (OD₆₀₀ ≈ 0.6–0.8). Bacterial cultures were diluted to an OD₆₀₀ of 0.05 in fresh 7H9 complete medium. The hit compounds were added to the wells at its predetermined MIC concentration. AMP was added to the cultures at final concentrations of 0-5 mM. Vehicle-only (VC), no cells, and compound only (0 mM AMP) served as controls. Plates were incubated at 37°C for 7 days. 20 ul AlamarBlue HS Cell Viability Reagent was added to each well and plates were re-incubated for 2-4 h before scanning and visulization. The degree of rescue was evaluated by comparing growth in MtbADSL inhibitor + AMP conditions relative to the inhibitor-only group. The experiment was performed in triplicates.

Phagocytosis Assay

RAW264.7 murine macrophage-like cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin-streptomycin (Gibco) at 37°C in a humidified 5% CO₂ incubator. Cells were seeded at a density of 2 × 10⁵ cells per well in 24-well tissue culture plates allowed to adhere overnight. Separately, Mtb H37Ra was cultured in Middlebrook 7H9 broth supplemented with 10% OADC, 0.05% Tween-80, and 0.5% glycerol to mid-log phase. Bacterial cultures were harvested by centrifugation at 3,000 × g for 10 minutes at room temperature and washed twice with sterile phosphate-buffered saline (PBS, pH 7.4) to remove residual media. Bacteria were resuspended in antibiotic-free DMEM. RAW264.7 cells were infected with the treated or control M. tuberculosis at a multiplicity of infection (MOI) of 10:1 and incubated at 37°C for 2 hours to allow phagocytosis. After the infection period, cells were washed three times with warm PBS to remove non-adherent bacteria, followed by incubation in DMEM containing streptomycin (100 μg/mL) for 1 hour to kill extracellular bacteria. Cells were then washed thoroughly with PBS to remove residual antibiotics. Cells were treated with 5 and 15 μM concentrations of bithionol and Tetrac for 24-48 hrs. For bacterial load quantification, macrophages were lysed using 0.05% SDS in sterile water for 5 minutes at room temperature. Lysates were serially diluted in PBS and plated on Middlebrook 7H10 agar plates supplemented with 10% OADC. Plates were incubated at 37°C for 3–4 weeks, and colony-forming units (CFUs) were enumerated to assess intracellular bacterial survival. Experiment was done in duplicate.

Molecular Docking

Bithionol and Tetrac were docked into the crystal structure of MtbADSL using Schrödinger Maestro (version 14.6, 2024-7 release, Windows-x64). The protein structure was imported and preprocessed using the Protein Preparation Wizard with default settings. The preprocessing steps included capping termini, assigning bond orders from the CCD database, adding hydrogens, filling missing loops and side chains with PRIME, and generating protonation states using EPIK at pH 7.4 ± 2.0. Hydrogen-bond optimization was performed with PROPKA (pH 7.4), sampling water orientations to minimize steric clashes. A restrained minimization was conducted using the OPLS4 force field to converge heavy atoms to an RMSD of 0.3 Å. Waters located within 5 Å of ligands or forming fewer than three hydrogen bonds to non-water atoms were deleted.

Potential ligand-binding sites were predicted using SiteMap, identifying ten pockets with site scores > 0.8. Ligands were prepared using LigPrep to generate ionization states at pH 7.4 ± 2.0 (EPIK), remove salts and solvents, and retain specified chiralities. Induced-fit docking (IFD) was performed using standard protocols. Receptor grids were centered on all SiteMap-predicted pockets within a 20 Å box. Glide docking employed van der Waals scaling factors of 0.5 for both the receptor and ligands, sampling ring conformations within a 2.5 kcal mol⁻¹ energy window and generating up to 20 poses per ligand. Residues within 5 Å of ligand poses were refined, and ligands were redocked into the top 10 receptor conformations within 30 kcal mol⁻¹ of the lowest-energy structure using standard precision mode. The top-scoring poses were selected for visualization and analysis.