Structural Elements Involved in ATP Hydrolysis Inhibition and ATP Synthesis of Tuberculosis and Nontuberculous Mycobacterial F-ATP Synthase Decipher New Targets for Inhibitors

ABSTRACT

The F₁F_O-ATP synthase is required for the viability of tuberculosis (TB) and nontuberculous mycobacteria (NTM) and has been validated as a drug target. Here, we present the cryo-EM structures of the Mycobacterium smegmatis F₁-ATPase and the F₁F_O-ATP synthase with different nucleotide occupation within the catalytic sites and visualize critical elements for latent ATP hydrolysis and efficient ATP synthesis. Mutational studies reveal that the extended C-terminal domain (αCTD) of subunit α is the main element for the self-inhibition mechanism of ATP hydrolysis for TB and NTM bacteria. Rotational studies indicate that the transition between the inhibition state by the αCTD and the active state is a rapid process. We demonstrate that the unique mycobacterial γ-loop and subunit δ are critical elements required for ATP formation. The data underline that these mycobacterium-specific elements of α, γ, and δ are attractive targets, providing a platform for the discovery of species-specific inhibitors.

INTRODUCTION

Mycobacteria can be separated into tuberculosis (TB)-causing mycobacteria (Mycobacterium tuberculosis, M. bovis, M. africanum, M. microti, M. canetti, and M. pinnipedii [1]) and nontuberculous mycobacteria (NTM). NTM infections are increasing in their global prevalence, morbidity, and mortality, with M. abscessus (Mab) as one of the most frequently identified rapidly growing NTM species (2, 3). Mab is regarded as one of the most antibiotic-resistant bacteria. Together with the emergence of multidrug-resistant (MDR) TB and the high incidence of latent TB infection (23% of world population), mycobacterial diseases present a formidable global challenge for public health (2, 4). Besides mutations in drug targets, resistance is caused by low permeability of the cell wall, biofilm formation, deficient drug-activating enzymes, target modifications, drug metabolism, or induction of drug efflux pumps (2, 5). Such pumps are either ATP- or proton-motive force (PMF) driven energy forms, generated by the electron transport chain (ETC) and the ATP forming F₁F_O-ATP synthase. The process of oxidative phosphorylation contributes mainly to the pathogens’ synthesis of ATP, making the F₁F_O-ATP synthase (subunits α₃:β₃:γ:δ:ε:a:b:b’:c₉) essential for the bacteria (6 to 8). Its H⁺-translocating F_O domain (subunits a:c₉) uses the PMF to drive rotation of the central stalk subunits γ-ε. The latter causes sequential binding, entrapment, and phosphorylation of ADP to ATP within the nucleotide-binding and catalytic α₃:β₃-headpiece. The peripheral stalk subunits b-δ:b’ smoothen transmission of power between the rotary c-ring and the α₃:β₃:γ:ε domain (9 to 12).

A special feature described for F-ATP synthases of the mycobacterial model systems, M. smegmatis and M. bovis, is their inability to establish a significant proton gradient during ATP hydrolysis due to their latent ATPase activity, preventing ATP wastage and maintaining ATP homeostasis (13 to 16). Regulative elements contributing to low ATPase activity in M. smegmatis can be attributed to the subunit ε (17, 18) and the mycobacterium-specific C-terminal domain (αCTD) of the nucleotide-binding subunit α (11, 15, 16). However, information about latent ATP hydrolysis of any NTM F₁-ATPase is missing. Importantly, both pathogens can survive in the absence of oxygen, when trapped within certain hypoxic niches in patients, where the pathogen slows its growth (19). This makes the bacteria exquisitely sensitive to any further ATP depletion and thus susceptible to drugs targeting ATP synthesis as demonstrated for the anti-TB drug bedaquiline (BDQ), being active against Mtbs (20), and with lower potency toward Mab’s F-ATP synthase (21), as well as GaMF1, which kills both bacteria by targeting the mycobacterium-specific loop of the rotary γ-subunit (22, 23). This 12- to 14-amino-acid loop is absent in the human homologue and proposed to be an essential feature for catalysis (14, 22). However, so far, no detailed study exploring the role of this γ-loop of any recombinant mycobacterial F-ATP synthase has been described. Previously, the crystallographic structure of the M. smegmatis F₁-ATPase published at 4.0 Å resolution proposed an MgADP inhibitory state (24). More recently, the electron cryo-microscopy (Cryo-EM) structures of the BDQ-inhibited M. smegmatis F-ATP synthase have been resolved (11, 25).

Here, we describe the cryo-EM structure of the M. smegmatis F₁-ATPase (MsF₁-ATPase) and the F₁F_O-ATP synthase (MsF-ATP synthase) at defined nucleotide-occupation states. The mycobacterium-specific modifications of the F-ATP synthase, namely, the αCTD (15), an inserted δ-domain (10), or the extra γ-loop (14), respectively, are visualized. Rotary dynamics studies of the recombinant M. smegmatis ε-free complex and its mutant with the deletion of the αCTD provide first insights into the chemo-mechanical coupling and regulation mechanisms. Mutational studies of the subunits α and γ within the recombinant M. smegmatis F₁-ATPase and F-ATP synthase demonstrate unequivocally the importance of the αCTD and the γ-loop for latent ATP hydrolysis and ATP formation, respectively. Movements of the peripheral stalk subunit δ visualized different states of the M. smegmatis F-ATP synthase and underpin its function as a transfer element of elastic energy during ATP formation. Finally, the first recombinant Mab F₁-ATPase and its αCTD mutants highlight the overall phenomena of ATP hydrolysis inhibition and reveal differences of this domain among mycobacteria, making it a specific target for novel inhibitors to bind.

RESULTS AND DISCUSSION

Cryo-EM analysis of the MsF₁-ATPase.

The MsF₁-ATPase used for cryo-EM studies was purified according to Wong et al. (18). No exogenous nucleotides were added prior to vitrification. We collected 2,316 movies at an accelerating voltage of 300 kV, yielding a final class of 120,114 for reconstruction (Fig. S1 in the supplemental material). The MsF₁-ATPase cryo-EM structure was resolved to a nominal resolution of 3.5 Å (Fig. 1A; Fig. S1B to D; Table S2A to C). Although the density of the ε-subunit is present, the low-quality density map led to a final model of the α₃:β₃:γ complex (Fig. 1B). The presented structure reveals an F₁-ATPase where the indirect binding change mechanism proposed by Boyer was observed (26). In the three noncatalytic α-subunits, MgATP could be well modeled in the density maps (Fig. 1C). No additional density could be observed in the empty catalytic β-subunit, while MgADP occupied the remaining β-subunits (Fig. 1C). The asymmetric nucleotide occupation of the α₃β₃-hexamer is further illustrated in the conformational changes between the αβ pairs. In the tight catalytic state (αβ_DP), the αβ pairs forms a close interface, which further widens from the loose (αβ_TP) to the empty state (αβ_E) (Fig. 1D, Movie S1). Comprehensively, the ₃₉₂βDELSEED₃₉₈ region was observed to shift by approximately 29.5 degrees from the β_DP to β_E state (Fig. 1D, bottom inset).

FIG 1.

Structural analysis of M. smegmatis F₁-ATPase. (A) Superimposition of the 3.5 Å resolved cryo-EM density map and atomic model of the MsF₁-α₃:β₃:γ. Weak density corresponding to the ε-subunit was removed for clarity. (B) The structural model of the F₁ α₃:β₃:γ complex, illustrated in forest green, orange, and yellow, respectively. For clarity, the extended α C-terminus residues, D534 to P545, are illustrated in green spheres. (C) In the F₁-ATPase, the nucleotide binding sites of all three α-subunits were occupied by the Mg-ATP, while in the catalytic β-subunits, no nucleotide was found in the βE subunit, and Mg-ADP was found in both βDP and βTP subunits. (D) Comparison of conformational states of the β-subunits of MsF₁-ATPase performed with superimposition on the N-terminal β-barrels. In brief, the C-terminus of the different states of the β-subunits were observed to undergo slight rotation, particularly in the β_E state compared to the β_DP and β_TP states. The rotation angular shifts in the phosphate-binding loop (₁₆₀GXXXXKT₁₆₇) and the β ₃₉₂βDELSEED₃₉₈ region are represented in the top and bottom insets, respectively.

No significant rotation was observed between the β_DP and β_TP DELSEED region (rotation angle = 4.02 degrees), as well as the phosphate-binding loop (₁₆₀GXXXXKT₁₆₇) in the different states (Fig. 1D, top inset). The nucleotide occupation of the presented F₁-ATPase differs from the 4.0 Å crystallographic structure of the MsF₁-ATPase (PDB 6FOC) (24), in which crystals have been grown in the presence of 1 mM ADP. This crystallographic structure reveals density in the β_E site, which could reflect either phosphate or sulphate. The nucleotide occupancy was interpreted as a possible ADP-dependent ATPase inhibitory mechanism (24).

The catalytic states of the F₁-ATPase correlates to the position of the antiparallel coiled α-helices in the γ-subunit. The latter drives the conformational changes observed in the catalytic α₃β₃-hexamer (27, 28). Apart from the antiparallel α-helices that protrude into the α₃β₃-hexamer (amino acids 1 to 35 and 256 to 304), the remainder of the γ-subunit forms a globular fold comprising both α-helices and β-sheets (Fig. 2A). In the current cryo-EM structure, the α-helices and β-sheets including the Rossmann fold were built (Fig. 2A). The density could be assigned to the highly polar γ₁₆₆DDEGDDAGADGILG₁₇₉ loop. This mycobacterium-specific γ-loop has been described as a novel anti-TB and anti-Mab compound target, leading to the inhibitor GaMF1 (22). To test whether this loop affects ATP hydrolysis of the MsF₁-ATPase, the MsF₁-ATPase Δγ166-179-mutant was engineered (Fig. 2B; Table S3). As shown in Fig. 2C, the deletion of the γ-loop did not cause any significant change in ATPase activity.

FIG 2.

Analysis of the M. smegmatis Δγ166-179 loop. (A) Structure comparison of this cryo-EM resolved (yellow) and the crystallographic structure (PDB 6FOC) (21) (blue) γ-subunit. The globular domain and the γ166-179 loop (red) are resolved in this cryo-EM structure model. (B) SDS gel of MsF₁-ATPase (lane 1) and its mutant Δγ166-179 (lane 2). M, marker. (C) ATP hydrolysis of the MsF₁-ATPase and the F₁-ATPase Δγ166-179 mutant represented in bar chart. Calculated specific activities are 0.139 ± 0.002 and 0.147 ± 0.002 μmol min⁻¹ (mg protein)⁻¹, respectively. ns indicates P = 0.2217. Statistical analysis of the data was performed with unpaired Student's t test, two-tailed with n = 3.

Self-inhibition mechanism of ATP hydrolysis of the mycobacterial F₁-ATPase.

A second mycobacterium-specific structural feature includes the extended 3.5-kDa C-terminus of subunit α (αCTD, residues 514 to 549 in Mtb [Mtb αCTD] and 514 to 548 in M. smegmatis [Ms αCTD]) (15). Amino acids 514 to 549 and 540 to 549 of the Mtb αCTD have been predicted to form a random coil, whereby residues 526 to 539 were predicted and demonstrated by NMR spectroscopy to form an α-helix in Mtb (15), while former cryo-EM structures reveal a β-sheet in M. smegmatis (11). The recently described chromosomal deletion mutation of the α_CTD mutant Δα514-548 stimulated ATP hydrolysis of inverted membrane vesicles (IMVs) of M. smegmatis (15), while fusing the Mtb α_CTD at the C-terminus of subunit α of the Bacillus PS3 F₁-ATPase (PS3 F₁) decreased ATPase activity of the hybrid enzyme (14). Even more, previous mutational studies of MsF₁-ATPase mutants have demonstrated that deletion of the entire C-terminal 35 amino acids enhanced ATP hydrolysis 60-fold, which is similar to the activity of the PS3 F₁-ATPase (15), and that residues α514-522 and α538-549 contributed to the major suppression of ATPase activity (16). In the MsF₁-ATPase cryo-EM structure presented here (Fig. 1B), additional densities in the γ-subunit can be assigned to the C-terminal residues α534 to 545 of one α subunit forming a catalytic pair with β_Dp. Amino acids αE534, αE536, αK539, αR541, and αK542 of this stretch interact with the γ residues γE207, γE209, and γE213 (Fig. S2).

This α–γ interaction is further stabilized by residues αE534 and αE536 with γR201 and γR202 (Fig. S2) and proposed to form the hook (αCTD) of a rachet (see details below) and thereby prevent ATP hydrolysis by blocking anticlockwise rotation of the central stalk but allowing rotation in ATP synthesis direction (11, 25). In contrast, the CTDs of the remaining two α subunits are disordered, implying that α527-545 shifts from an unstructured element into a β-strand when interacting with γ210-212. Such structural alterations have also been described for the inhibitory protein IF of the mitochondrial F₁-ATPase (29), or subunit ζ in ɑ-proteobacteria (30), whose N termini shift from an intrinsically disordered region to an inhibitory ɑ-helix when interacting with their regulatory partners in the catalytic sites.

αCTD’s interaction with γ causes intramolecular friction during rotation.

To further elucidate the effect of the αCTD on rotation of the engine, we employed a single-molecule rotation assay. We used the recently generated M. smegmatis ε-free complex MsF₁-αβγ (18), which in contrast to the wild-type (WT) MsF₁-ATPase showed an increase in specific ATPase activity (Fig. 3). Cysteines were introduced in γG111 and γL231 to facilitate binding of the biotinylated beads (here labeled F₁-αβγ_cys; Fig. 3A). This provided a comparable system to the well-studied TF₁ complex (αβγ). A second mutant with the deletion of the entire MsαCTD (here labeled F₁-α_ΔCTDβγ_cys) with an ATP hydrolysis activity of 3.27 ± 0.31 μmol min⁻¹ (mg protein)⁻¹ was generated to investigate the effect of αCTD on rotation (Fig. 3B and C).

FIG 3.

(A) Cysteine introductions into the MsF₁-αβγ. (B) No significant differences in ATP hydrolysis rates were observed in the mutants. The specific activities for MsF₁-αβγ, MsF₁-αβγ_cys, MsF₁-α_ΔCTDβγ, and MsF₁-α_ΔCTDβγ_cys are 1.22 ± 0.07, 1.14 ± 0.33, 3.21 ± 0.29, and 3.27 ± 0.31 μmol min⁻¹ (mg protein)⁻¹, respectively. Statistical analysis was performed with one-way ANOVA (analysis of variance) test, with n = 3. Adjusted P values are P < 0.0004 (***), ns (1) = 0.9906, and ns (2) = 0.9944. (C) SDS-PAGE gel of the purified mutants. The αβγ subunits are prominent at approximately 60, 55, and 35 kDa, respectively. In the deleted αCTD mutants, the α subunit runs lower (*). (D and E) Rotation of F₁-αβγ_cys and F₁-α_ΔCTDβγ_cys. (D) Pause and rotation of F₁-αβγ_cys (gray lines) and F₁-α_ΔCTDβγ_cys (red lines) at 2 mM MgATP. (E and F) Time courses of rotation of F₁-αβγ_cys (gray lines, n = 5 molecules) and F₁-α_ΔCTDβγ_cys (red lines, n = 10 molecules).

First rotation assays with 40 nm gold colloid under nonviscous friction revealed insufficient F₁-αβγ_cys rotating molecules compared to F₁-α_ΔCTDβγ_cys, probably due to the intrinsic autoinhibitory nature of the CTD. Therefore, magnetic beads with a diameter of about 200 to 400 nm (31) have been used in the presence of 2 mM MgATP with a recording time of 30 frames per second (fps). Smooth rotation of F₁-αβγ_cys (Movie S2; 5 molecules) and F₁-α_ΔCTDβγ_cys (Movie S3; 10 molecules) rather than stepping rotation was observed (32), which may be due to the high viscous friction of the magnetic beads. Both molecules showed slow transitions between continuous rotation and intervening pauses (Fig. 3D), as previously found for TF₁ (33). The mean times for continuous rotation and pause should correspond to the lifetimes of a catalytically active and inhibition state, respectively. As shown in Table S3, the mean time for rotation and pause was 14 to 23 s and 43 to 52 s, respectively, in both molecules, revealing that a significant difference was not observed in the slow transition between rotation and pausing state of both complexes. Importantly, F₁-αβγ_cys and F₁-α_ΔCTDβγ_cys differed in the rotation velocity during continuous rotation (Fig. 3E and F, Table S4, Movie S2 and S3) with a mean rotation velocity of 3.1 ± 2.2 rps and 5.2 ± 2.2 rps for F₁-αβγ_cys and F₁-α_ΔCTDβγ_cys, respectively. Since the rotation velocity is in principle limited by the hydrodynamic friction on rotating magnetic beads, the rotation velocity is a measure of torque that the F₁ molecule exerts. Therefore, the observed difference suggests that F₁-αβγ_cys including the αCTD generates lower torque. However, considering that the C-terminal extension is not directly involved in the torque generation process, it would be more reasonable to consider that the αCTD acts by causing intramolecular friction when interacting with the rotating γ subunit. Although the αCTD–γ interaction within the MsF₁ATPase structure seems to be stable (Fig. S2), the present results suggest that the transition between inhibition state by the αCTD and the active state is a rapid process such that the rotation is seemingly slowed uniformly over all rotational angles without showing distinctive pauses.

αCTD inhibition is also conserved in NTM F₁-ATPases.

Although the αCTD extension is unique to mycobacteria, it differs in amino acid composition in particular to the three Mab subspecies, underlined by the differences in the predicted α-helical and β-sheet content compared to other mycobacterial αCTDs (Fig. 4A and B).

FIG 4.

Characterization of the auto-inhibition element αCTD of Mab’s F₁-ATPase. (A) Amino acid sequence alignment of the αCTDs from different mycobacterial organisms showing significant differences in amino acid composition. The percentage of identity is presented in darker to lighter shades of blue, representing the most homologous to the least homologous. (B) Secondary structure predictions, using PREDATOR (39), of the C-terminal region of α subunit in Mtb, M. smegmatis, and Mab, revealing three secondary structure elements: two β-sheets formed by residues numbered 519 to 524 and 538 to 542, and one α-helix formed by residues numbered 525 to 531. Non-highlighted black lines represent random coils prediction. (C) Left, 12% SDS gel displaying the first recombinant Mab F₁-ATPase (α₃:β₃:γ:ε). Middle, representative negative-stained electron micrograph of Mab F₁-ATPase revealing well-dispersed and individual particles. Right, 2D classes of the Mab F₁-ATPase showing the typical hexameric α₃:β₃ fashion with additional densities corresponding to the central stalk subunit γ. (D) ATP hydrolysis assay of the Mab WT F₁-ATPase and its α mutants. The adjusted P values for the ATPase activity differences between Mab F₁-α_Δ514-548βγε, F₁-α_Δ532-548βγε, and F₁-α_Δ542-548βγε to the WT enzyme was determined to be P < 0.0001, P = 0.0428, and P = 0.0045, respectively. Comparison between the Mab F₁-α_Δ532-548βγε and F₁-α_Δ542-548βγε revealed a P value of ns = 0.0979. All statistical analysis was performed using one-way ANOVA test, where n = 3. (E) The purified proteins were assayed on a 12% SDS-PAGE gel for comparison to the WT enzyme (lane 1). The major band indicated by an * of Mab F₁-α_Δ514-548βγε was analyzed by MALDI TOF (Table S4), confirming that this major band carries the α_Δ514-548 deletion, leading to a further migration in the SDS gel compared to the α-band of the WT F₁-ATPase.

To explore whether NTM F₁-ATPases also uses the same regulative αCTD, we generated the first recombinant Mab subsp. abscessus F₁-ATPase (MabF₁-ATPase; α₃:β₃:γ:ε) and visualized its α₃:β₃-domain with subunit γ in top views of 2D projections of negative-stained electron micrographs (Fig. 4C). The MabF₁-ATPase displayed low ATP hydrolytic activity of 0.04 μmol min⁻¹ (mg protein)⁻¹ (Fig. 4D). Based on the secondary structure prediction (Fig. S3B), three deletion mutants, MabF₁-α_Δ514-548βγε, F₁-α_Δ532-548βγε, and -F₁-α_Δ542-548βγε, were generated for assessment of their roles in the inhibition of Mab’s ATPase activity (Fig. 4D and E).

Deletion of the entire αCTD in the F₁-α_Δ514-548βγε mutant enhanced ATP hydrolysis by 32-fold (1.28 ± 0.048 μmol min⁻¹ (mg protein)⁻¹), while the ATPase activity of the F₁-α_Δ532-548βγε and F₁-α_Δ542-548βγε mutants increased by a factor of 9 (0.36 ± 0.01 μmol min⁻¹ (mg protein)⁻¹) and 13.5 (0.54 ± 0.05 μmol min⁻¹ (mg protein)⁻¹), respectively, compared with the WT complex (Fig. 4D). The results illustrate that the unique αCTD is the major ATP hydrolysis inhibitory element of F-ATP synthases of both TB-causing mycobacteria and NTM. Furthermore, as shown for the M. smegmatis enzyme (11), the Mab αCTD with its specific residues and its interacting γ subunit may become an interesting inhibitor target in future, since it is not present in the human counterpart and would thus ensure lack of on-target toxicity.

Cryo-EM analysis of MgATP-saturated M. smegmatis F-ATP synthase.

While the cryo-EM structures of the BDQ-inhibited MsF-ATP synthase have been published (11, 25), we explored the structure of the MsF-ATP synthase saturated with 1 mM MgATP. The MsF-ATP synthase was incubated with 1 mM Mg-ATP for 5 min on ice prior to vitrification. We collected 3,663 movies, yielding a final particle set of 59,576 particles (Fig. S3A) that were subsequently sorted into three rotational states (Fig. 5A). The different rotational states correspond to the orientation of the subunit γ, alluding to different rotary catalysis positions. The final resolutions of the rotational states 1 to 3 are 4.4, 4.7, and 7.3 Å, respectively (Fig. S3A to F). The overall percentage of particles were favored toward rotational states 2 (44.2%), 1 (37.6%), and 3 (18.2%). The densities of all F-ATP synthase subunits can be visualized clearly from the maps in states 1 and 2. In state 3, the densities of the c-ring cannot be resolved properly in view of the low resolution. The model of MsF-ATP synthase was built into the map of state 1 (Fig. 5B), using the starting model of the recently described M. smegmatis structure (PDB 7JG5) (11). The quality of the optimized model obtained after refinement was then evaluated using Molprobity, yielding a Molprobity score of 1.52 (Table S2C). This derived model was used as a reference for rigid body fitting into the maps of the two remaining rotational states. Like in other bacterial F-ATP synthases, the central stalk rotates by about 120^° from one rotational state to another (Movie S4).

FIG 5.

Structural analysis of M. smegmatis F-ATP synthase. (A) Cryo-EM density maps of the MsF-ATP synthase in three major rotational states determined based on the rotation of the central stalk (γ, yellow cartoon; ε, magenta cartoon), with the resolutions of 4.4, 4.7, and 7.3 Å, respectively. (B) Structural model generated from cryo-EM map in the rotational state 1. (C) Cryo-EM density map of MsF-ATP synthase in state 1, with the cross-sections of (blue box) nucleotide binding site (red surface) and (red box) C-terminal domain of αβ-hexamer. In the F-ATP synthase, the nucleotide binding sites of all three noncatalytic α subunits were occupied by the Mg-ATP, while in the catalytic β subunits, ADP was assigned in the β_E subunit and Mg-ATP in both β_DP and β_TP.

Nucleotide occupancies were observed in all nucleotide-binding sites (Fig. 5C, blue box). Like in the α subunits of the cryo-EM MsF₁-ATPase (Fig. 1C) and the recently described MsF-ATP synthase structures (PDB ID 7JG5 [11] and 7NJK [25]), all three α subunits bind MgATP (Fig. 5C). In the present structure, ADP was assigned in the β_E subunit and adopted a “forced open” conformation for its αβ-interface (Fig. 5C, red box, and Fig. S4), while in the so-called β_TP and β_DP subunits, MgATP occupied both catalytic sites and adopted a “loose” and “tight” conformation, respectively (Fig. 5C, red box). This nucleotide occupation pattern is different from the pattern determined by Guo et al. (PDB ID 7JG5) (11) with a phosphate being modeled in β_E and β_DP, while β_TP was occupied by MgATP. Such occupation may have been caused by the missing exogenous nucleotides during enzyme preparation. The occupancy of nucleotides in the β subunits within our structure also differs from the most recent structure (PDB ID 7NJK) (25), in which ADP, MgADP, and MgATP have been found in β_E, β_TP, and β_TP, respectively (Movie S5).

The αCTD and γ-loop inhibit ATP hydrolysis and synthesis, respectively.

The density attributed to the extension of the α subunit was also evident in this F-ATP synthase structure. This extension was observed in α_DP from E531 to P545 (state 1) and from E534 to P545 (state 2), but not in the map of rotational state 3 (Fig. S5). No additional density could be observed in the α_E and α_TP. So far, studies using recombinant mycobacterial F₁-ATPase (reference 16 and studies above) or F₁F_O-ATP synthase (11) to explore the unique αCTD have focused on its role in self-inhibition of ATP hydrolysis. To understand if the release of this auto-inhibition affects proper function of the mycobacterial F-ATP synthase in synthesis, the recombinant MsF-ATP synthase with a deletion of α514-548 (F₁F_O Δα514 to 548) was generated and purified (Fig. 6A).

FIG 6.

Characterization of the M. smegmatis F-ATP synthase and its mutants. (A) SDS-PAGE gel displays the subunit compositions of the purified protein complexes, including recombinant MsF-ATP synthase (lane 1) and its mutants Δα514-548 (ΔαCTD) (lane 2) and Δγ166-179 (lane 3). In lane 2, the major band indicated by an * was analyzed by MALDI-TOF (Table S2 A–C) and revealed that this major band is the correct α_Δ514-548 deletion mutant, leading to a further migration in the gel compared to the α-band of the WT F₁-ATPase. In lane 3, the subunit γ with Δγ166-179 migrated slower than WT γ, largely due to a highly charged amino acid composition from 166 to 179. M, marker. (B) ATP synthesis of the M. smegmatis F-ATP synthase (wild type), F₁F_O Δα514-548, and F₁F_O Δγ166-179 represented in bar chart with SD. **, P < 0.0092; ***, P < 0.0001. (C) ATP hydrolysis activity of M. smegmatis F-ATP synthase (wild type), F₁F_O Δα514-548, and F₁F_O Δγ166-179 represented in bar chart. Calculated specific activities are 0.112 ± 0.003, 1.846 ± 0.035, and 0.106 ± 0.002 μmol·min⁻¹·(mg protein)⁻¹, respectively. ****, P < 0.0001; ns, P = 0.4199. Statistical analysis of the data is performed with unpaired Student's t test, two-tailed, where n = 3.

We used our established reconstitution protocol to detect ATP synthesis (Fig. 6B) (12). The outcome revealed a 12% decrease (29.4 ± 0.9 nmol min⁻¹ (mg protein)⁻¹) in ATP synthesis in the reconstituted liposomes compared to the WT enzyme (33.3 ± 0.2 nmol min⁻¹ (mg protein)⁻¹; Fig. 6B) underpinning the concept of the recently described ATP synthesis inhibitor AlMF1 by targeting the αCTD and its interaction region within the γ-subunit (34). In contrast, a major increase of ATP hydrolysis was observed in the mutant complex (Fig. 6C), showing a 16.5-fold increase compared to WT complex, consistent with previous studies (15). A so-called “fail-safe” device (25), constituted by the extra mycobacterial γ166-179 loop and also visualized in the density of the presented structure (Fig. S5B), was discussed as a second inhibitory element of ATP hydrolysis by which the peripheral stalk subunit residue b’R72 would form a salt bridge with either γ-loop residue D170 or D171 (25).

Although the MsF₁-ATPase Δγ166-179-mutant shows low ATP hydrolysis activity (Fig. 6C), we tested whether a Δγ166-179 deletion in the recombinant MsF-ATP synthase with its peripheral b’ subunit would show enhanced ATPase activity. As demonstrated in Fig. 6C, the generated MsF-ATP synthase Δγ166-179 mutant revealed no difference in ATPase activity compared to the WT enzyme, highlighting that the γ-loop is not critical in ATP hydrolysis inhibition. Importantly, a decrease of 43.2% (14.4 ± 1.0 nmol min⁻¹ (mg protein)⁻¹; Fig. 6B) in ATP synthesis was observed for the recombinant F-ATP synthase Δγ166-179 mutant, which is in line with the 54% ATP synthesis reduction measured for the chromosomal Δγ166-179 deletion mutation of M. smegmatis IMVs determined preciously (14).

Taken together, the results suggest that while the mycobacterium-specific αCTD prevents the central stalk mainly to rotate in the ATP hydrolysis direction, the mycobacterial γ-loop insertion affects rotation in the ATP synthesis direction, possibly via the proposed positively charged b’R72 and the negatively charged γD170/γD171, a pattern conserved across all mycobacterial species.

Recently, the anti-TB F-ATP synthase inhibitor GaMF1, which interacts with several residues of the γ-loop (22), has been repurposed as an anti-Mab F-ATP synthase inhibitor that is bactericidal and revealed enhanced anti-Mab activity in combination with the anti-Mab drugs clofazimine, rifabutin, or amikacin (23), demonstrating that the mycobacterial γ-loop is an attractive mycobacterial-wide drug target.

Movements of subunit δ, a mycobacterial-wide inhibitor target.

As demonstrated for the subunits α and γ, differences in amino acid composition of the mycobacterial enzyme, compared to other prokaryotic or human counterparts, paved avenues for the discovery of molecules interfering with various regulative mechanisms of this F-ATP synthase (see above and references 22, 23, and 34). In this context, the mycobacterial peripheral stalk subunit δ displays a unique N-terminal 111-residue extension (10, 35).

Mutational studies of subunit δ in the recombinant M. smegmatis F-ATP synthase have highlighted the importance of this subunit for effective catalysis (12) and has indicated the critical communication for smooth transfer of elastic energy between the fused N-terminal residues of δ and the C-terminus of the peripheral stalk subunit b, as well as the N-terminus of subunit α during rotation. Superimposition of the three models with alignments performed on subunit a showed slight movements of the peripheral stalk (Fig. S8A, side view) and significant movements of subunit δ (Fig. 7A and B, top view) with its noncanonical N-terminal domain (NTD) (helices ɑ3 to ɑ8), the central domain (CD; helices ɑ9 to ɑ14), and the CTD (β1 to β4 and ɑ15 to ɑ16). Such movement also includes α-helices 3 and 4 with residues δR171, δR171G, δR177Q, and δQ178R demonstrated to be important for ATP synthesis (12). This area has recently been identified as a novel compound target, leading to the F-ATP synthase hit molecule DeMF1, which inhibits NADH-driven ATP synthesis of M. smegmatis IMVs with a half-maximal inhibitory concentration (IC₅₀) of 0.5 μM (12). Its inhibitory potency has also been demonstrated in the reconstituted recombinant MsF-ATP synthase (12). Targeting of the δ subunit demonstrates the potential to advance subunit δ’s flexible coupling as a new area for the development also of NTM F-ATP synthase inhibitors, which is confirmed by the 43% ATP synthesis inhibition of Mab IMVs in the presence of 500 μM DeMF1 (Fig. 7C).

FIG 7.

(A) The side views (left, middle) and top view (right) of superimposed peripheral stalks from the rotational state 1 (slate blue), state 2 (green), and state 3 (pink) show slight movement of the peripheral stalk when the central stalk rotates from one state to the next state. (B) Top view of superimposed peripheral stalks from the rotational state 1 (subunit b: light violet; subunit δ: slate blue) and state 2 (subunit b: light green; subunit δ: green). α-helix3 (H3) can be seen from the top view, while subunit δ is composed of 13 α-helices (H4 to H16) and four β-sheets lying between H14 and H15. (C) Inhibition of ATP synthesis by DeMF1 using M. abscessus and M. smegmatis IMVs with NADH as substrate represented in bar chart with SD. The gray bar shows drug free. ****, P < 0.0001. Statistical analysis of the data was performed with t test with n = 3. (D) Amino acid sequence alignment of the fused b–δ subunits from different mycobacterial species revealing significant differences in amino acid composition. The calculation of the percentage of identity was performed and showed in darker to lighter shades of blue, representing the most homologous to the least homologous. While DeMF1 interacting residues δR177, -L180, -E366, -R371, -R372, -E374, and -R403 are similar in Ms and Mab (*), amino acid polymorphisms were observed for δR171, -R173, -Q178, -H273, and -G373 (*).

Although with 1,000-fold lower potency, this result underlines the druggability of mycobacterial subunit δ. The sequence alignment of the mycobacterial δ subunit (Fig. 7D) reveals that the DeMF1’s interacting residues δR177, -L180, -E366, -R371, -R372, -E374, and -R403 are similar in Ms and Mab. Amino acid polymorphisms were observed for δR171, -R173, -Q178, -H273, and -G373, explaining the 1,000-fold lower efficiency in Mab IMVs. These data pave the way for future medical chemistry efforts for new leads targeting mycobacterial subunit δ with specificities to the class of TB-causing mycobacteria and NTM.

Conclusion.

The cryo-EM structures of the MsF₁-ATPase and the -F₁F_O-ATP synthase revealed different nucleotide occupation within the catalytic sites, going along with movements of the C-terminal domain of the catalytic β subunits and the central stalk element. To maintain PMF and ATP homeostasis, TB-causing mycobacteria and NTM “invented” the αCTD. Deletion of this domain resulted in stimulated ATPase activity of the motor and minor inhibition in ATP synthesis, highlighting its role in preventing wastage of ATP and maintaining PMF. The αCTD binds to the rotating γ-subunit, leading to intramolecular friction (Fig. 8) and, together with subunit ε, inhibits ATP hydrolysis. Rotational studies indicated that the transition between the inhibition state by the αCTD and the active state is a rapid process in which rotation slows uniformly over all rotational angles. In comparison, the unique mycobacterial γ-loop does not affect ATP hydrolysis inhibition either in the F₁-ATPase or in the entire F-ATP synthase but is a fundamental element for ATP formation (Fig. 8). Movements of the peripheral stalk, in particular subunit δ including the mycobacterium-specific noncanonical NTD, provide the smooth transmission of power caused be the rotary c-ring and central stalk, which is essential for ATP synthesis (Fig. 8). The αCTD (16), the γ-loop (14), and the δNTD (12) have been described as novel anti-TB targets with new anti-TB inhibitors to bind (Fig. 8) (12, 22, 34). Although different in amino-acid composition and predicted secondary structure like in the case of αCTD (Fig. 4A and B), the data presented underline that these three mycobacterium-specific elements are important for F-ATP synthase’s regulation and mechanisms, and ideal anti-NTM inhibitor targets, providing the platform for species specific medical chemistry efforts.

FIG 8.

Summary cartoon of the data presented. The cryo-EM structures of the MsF₁-ATPase and the MsF₁F_O-ATP synthase have been resolved and showed different nucleotide occupation within the catalytic sites. TB-causing mycobacteria and NTM “invented” the αC-terminal extension (left) binding to the central and rotating subunit γ leading to intramolecular friction and inhibition of ATP hydrolysis. The cryo-EM structure of the MsF₁F_O-ATP synthase reveals the unique mycobacterial γ-loop, which is a fundamental element for ATP synthesis. Movements of the peripheral stalk subunit δ including the mycobacterium-specific noncanonical NTD have been resolved. Such movements provide the smooth transmission of power caused be the rotary c-ring and central stalk. The mycobacterial δNTD is being discussed as an ideal anti-TB and anti-NTM inhibitor target.

MATERIALS AND METHODS

Generation of M. smegmatis F₁ αβγ mutants for single-molecule rotational assay.

To allow binding of the biotinylated beads, G111 and L231 in the γ-subunit were mutated to cysteines, generating the MsF₁-αβγ_G111C/L231C mutant (labeled as the MsF₁-αβγ_cys here). To observe rotational differences between the latent MsF₁-αβγ_cys and the enzymatically active αΔ514-548 complex, a deletion from amino acids 514 to 548 was introduced to the MsF₁-αβγ_cys mutant, generating the MsF₁-α_Δ514-548βγ_G111C/L231C (labeled as the MsF₁-α_ΔCTDβγ_cys here). Both mutants were generated via site-directed mutagenesis, with their relevant primers and templates listed in Table S2. The respective plasmids were amplified using KAPA HiFi DNA polymerase (Kapa Biosystems; Wilmington, MA, USA). PCR products were subsequently treated with DpnI prior to transformation in E. coli TOP10 cells. Plasmids were extracted using GeneJET Plasmid Miniprep kit (ThermoFisher Scientific; Waltham, MA, USA), and DNA sequencing (Bio Basic Asia Pacific, Singapore) was performed to confirm the integrity of the plasmid. Plasmids carrying the gene of interest were then transformed into M. smegmatis mc²4517 electrocompetent cells for large-scale protein production and purification (18). Protein purification was performed as previously described (18). For downstream biotinylation assays, 5% glycerol was added to the purified sample and subsequently concentrated to 5 μM with a 100-kDA Centricon (Millipore; Burlington, WA, USA) prior to usage.

Single-molecule rotation assay.

The introduced cysteines on the γ-subunit of MsF₁-αβγ_cys and MsF₁-α_ΔCTDβγ_cys were biotinylated to attach magnetic beads as an optical probe. The rotation assay was performed as described previously using magnetic beads with slight modifications (31). Briefly, basal buffer (50 mM MOPS, pH 7.0, 50 mM KCl, 2 mM MgCl₂, 5 mg/mL BSA) was infused into the flow cell. After 10 min incubation, ~100 nM MsF₁-αβγ_cys or MsF₁-α_ΔCTDβγ_cys dissolved in basal buffer was infused into the flow cell. After 10 min incubation, unbound enzyme was washed out with basal buffer, and magnetic beads were infused in the rotation assay buffer. After 10 min of incubation, unbound beads were washed out with rotation buffer (50 mM MOPS, pH 7.0, 50 mM KCl, 2 mM MgCl₂, indicated concentrations of ATP) containing an ATP-regenerating system (2 mM phosphoenolpyruvate, 100 μg/mL pyruvate kinase). The rotations were visualized on a phase-contrast microscope (IX-70; Olympus) with a 100× objective lens at 30 fps (FC300M; Takex). The rotation assay was performed at 23 ± 2°C.

Generation of Mab WT F₁-ATPase and its mutants.

The atp operon encoding Mab F₁-ATPase was amplified from the genomic DNA of Mycobacterium abscessus subsp. abscessus (strain ATCC 19977; GenBank: CU458896.1) using the primers listed in Table S3. The operon was incorporated as previously described (18). In brief, overlapping regions of the vector pYUB1049 and the Mab F₁-ATPase (atpAGDC) were designed for PCR amplification. Amplified fragments of both the genomic DNA and the vector were then assembled with NEBuilder HiFi DNA Assembly. Thereafter, the plasmid (pYUB1049-MabF₁) was transformed into E. coli TOP10 cells for plasmid amplification. Plasmid extraction and DNA sequencing were performed to confirm the integrity of the plasmid. After obtaining the plasmid with the right gene of interest, a His₆ tag was introduced in the N-terminus of the β-subunit via site-directed mutagenesis. The primers utilized for His₆ tag insertion were 5′-AAT Gca tca cca tca cca tca tAC CGC ACC AAC TGC TAA CAA GAC-3′ and 5′-GTa tga tgg tga tgg tga tgC ATT ACT CGT CTC TTC TCG TTC TCT AGT G-3′. After PCR amplification, the amplified product was treated with DpnI to remove leftover template. Finally, treated DNA samples were transformed into E. coli TOP10 cells for plasmid amplification, as mentioned before. Plasmids carrying the inserted His₆ tag were then verified through DNA sequencing.

Mutational studies involving the α C-terminus deletion were performed utilizing the template pYUB1049-MabF₁αβγε. Site-directed mutagenesis was performed to generate the Mab F₁-αΔα514-548βγε, -αΔ532-548βγε, and -αΔ542-548βγε mutants using primers listed in Table S3. Amplified PCR products were treated like in the case of the insertion of the His₆ tag. Verification of plasmids carrying the desired mutations was then performed with DNA sequencing.

Purification of Mab WT F₁-ATPase and its mutants.

Plasmids verified to carry the desired gene of interest were transformed into electrocompetent M. smegmatis mc²4517. Transformed cells were plated on 7H10 agar plates supplemented with 50 μg/mL of both kanamycin and hygromycin. Single colonies were picked and inoculated into an autoinduction medium (ZYP-5052) (18) supplemented with 50 μg/mL of both kanamycin and hygromycin. Cells were incubated at 37°C, 180 rpm for 72 h, prior to harvesting. For harvesting, cells were spun down via centrifugation at 4°C, 3,648 × g for 20 min.

Harvested cells were resuspended in buffer A (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 15% glycerol, 2 mM Pefabloc). Resuspended cells were lysed via sonication 3 times at 25% with 1 min intervals. After sonication, cells were further ruptured with LM20 Microfluidizer (Microfluidics, Westwood, MA, USA) at 13,000 lb/in², 3 times. Cell lysate was then clarified using ultracentrifugation at 150,000 × g, for 45 min. Next, the supernatant was incubated with Ni-NTA Agarose (Qiagen Science, Germantown, MD, USA) for 1 h at 4°C. His₆-tagged proteins were eluted with increased step-gradient imidazole concentrations from 0 mM to 250 mM in buffer A. Eluted fractions containing the desired protein was pooled and applied on an ion-exchange column (ResQ, 1 mL; GE Healthcare, Chicago, IL, USA) with the following buffers: 50 mM Tris, pH 7.5, 10% glycerol (vol/vol) and 50 mM Tris, pH 7.5, 1 M NaCl, and 10% glycerol. Linear gradient salt elution was performed from 150 to 500 mM NaCl at a flow rate of 1 mL/min. For the α C-terminal deletion mutants, a step-gradient salt elution was utilized to resolve the elution peaks. Samples were concentrated, wherever necessary, with a 100 kDa-cutoff Centricon (Millipore, Burlington, WA, USA) at 4000 × g centrifugal speed.

Generation of the M. smegmatis F-ATP synthase mutants for reconstitution in liposomes.

M. smegmatis F-ATP synthase Δα514-548 (identified as Ms F-ATP synthase ΔαCTD here) and Ms F-ATP synthase Δγ166-179 were generated via site-directed mutagenesis with primers and its template listed in Table S3. Detailed description of the generation of the template pYUB1049-MsF-ATP synthase can be found in Saw et al. (36). The respective plasmids were amplified using KAPA HiFi DNA polymerase. PCR products were subsequently treated with DpnI prior to transformation in E. coli TOP10 cells. Plasmids were extracted using GeneJET Plasmid Miniprep kit (ThermoFisher Scientific, Waltham, MA, USA), and DNA sequencing (Bio Basic Asia Pacific, Singapore) was performed to confirm the integrity of the plasmid. Plasmids carrying the gene of interest were then transformed into M. smegmatis mc²4517 electrocompetent cells for large-scale protein production and purification (37). Protein purification was performed as previously described (36).

Continuous ATP hydrolysis assay.

Continuous ATP hydrolysis assay performed on all mutants described above was performed as previously described (38). For all ATP hydrolysis measurements, no sample dilution was performed, and data were collected in triplicates. Detailed description of the analysis of specific activities of the enzymes may be found in Wong et al. (18).

Reconstitution and ATP synthesis assay of the recombinant M. smegmatis F-ATP synthase and its mutants.

Reconstitution experiments and ATP synthesis assays for the Ms F-ATP synthase and its mutants ΔαCTD and Δγ166-179 were performed as previously described by Harikishore et al. (12).

Preparation of Mab IMVs and ATP synthesis.

To purify inverted membrane vesicles (IMVs) of Mab for ATP synthesis, cells were grown overnight at 37°C in 7H9 medium supplemented with 10% Bovine albumin, dextrose and catalase (ADC), 0.2% glycerol, and 0.05% Tween 80 until they reached an attenuance at 600 nm of 0.6 to 0.7. The culture was expanded in 200 mL supplemented 7H9 medium and grown in 1-L shake flasks (180 rpm) until it reached an attenuance at 600 nm of 0.6 to 0.7. This culture was used to inoculate a 500-mL culture that was then grown overnight in 2-L shake flasks (180 rpm) until it reached an attenuance at 600 nm of 0.6 to 0.7. Approximately 5 g (wet weight) of WT Mab was resuspended in 20 mL membrane preparation buffer (50 mm MOPS, 2 mm MgCl₂, pH 7.5) containing EDTA-free protease inhibitor cocktail (one tablet per 20 mL buffer, Roche) and 1.2 mg · mL⁻¹ lysozyme. The suspension was stirred at room temperature for 45 min and additionally supplemented with 300 μL of 1 M MgCl₂ and 50 μL DNase I. Stirring was continued for another 15 min at room temperature. All subsequent steps were performed on ice. Cells were lysed by three passages through an ice-cooled microfluidizer (Microfluidics, Westwood, MA, USA) at 18,000 lb/in². The suspension containing lysed cells was centrifuged at 4,200 × g at 4°C for 20 min. The supernatant containing the membrane fraction was further subjected to ultracentrifugation at 45,000 × g at 4°C for 1 h. The supernatant was discarded, and the precipitated membrane fraction was resuspended in membrane preparation buffer containing 15% glycerol, separated into aliquots, snap-frozen, and stored at −80°C. The concentrations of the proteins in the vesicles were determined by the bicinchoninic acid assay (BCA; Pierce, Waltham, MA, USA).

ATP synthesis was measured in flat-bottomed white 96-well microtiter plates (Corning, Corning, NY, USA). The reaction mix (50 μL) comprised assay buffer (50 mm MOPS, pH 7.5, 10 mM MgCl₂) containing 10 μM ADP, 250 μM P_i, and 1 mM NADH. The concentration of P_i was adjusted by addition of 100 mM KH₂PO₄ to the assay buffer. ATP synthesis was started by adding Mab IMVs at a final concentration of 5 μg · mL⁻¹. The reaction mixture was incubated at room temperature for 30 min before adding 50 μL CellTiter-Glo reagent, followed by incubation for another 10 min in the dark at room temperature. The luminescence produced, which correlates with the amount of ATP synthesized, was measured by an Infinite 200 Pro plate reader (Tecan; DKSH, Zurich, Switzerland), using the following parameters: luminescence; integration time, 500 ms; attenuation, none.

Data availability.

All relevant data are available from the authors.

Structural data that support the findings of this study are openly available in PDB at https://www.rcsb.org and EM Data Bank at https://www.ebi.ac.uk/emdb/. M. smegmatis F₁-ATPase: PDB ID 7Y5A, EMDB ID EMD-33614. M. smegmatis F-ATP synthase structures: PDB ID 7Y5B, EMDB ID EMD-33615 (rotational state 1); PDB ID 7Y5C, EMDB ID EMD-33616 (rotational state 2); and PDB ID 7Y5D, EMDB ID EMD-33617 (rotational state 3).