Turbine enzyme’s structure in the crosshairs to target tuberculosis

Tuberculosis (TB) is the deadliest infectious disease worldwide, killing at least 1.5 million people yearly (1). An increasing challenge has been to treat multidrug-resistant (MDR) and extensively drug resistant (XDR) TB cases (MDR-TB and XDR-TB, respectively), estimated at 0.5 million in 2017. This has required up to 2 y of treatment with drugs that can have harsh side effects. Of the estimated people infected with drug-resistant TB, <15% have completed treatment and appear to be cured. However, some experts state that we are at a “tipping point in revolutionizing the care of patients with MDR- and XDR-TB” (2). A major factor in this was the development of bedaquiline (BDQ), the first anti-TB drug in more than 40 y that has a distinct new target of action (3). BDQ is a diarylquinoline, a class of compounds first identified as effective growth inhibitors for Mycobacterium smegmatis, a nonpathogenic relative of the main TB culprit, Mycobacterium tuberculosis. The target of BDQ in both these species was identified as the F-type ATP synthase (3), a key enzyme in cellular energy metabolism that is essential for viability of these obligate aerobes (4, 5). However, the core structure and functions of F-type ATP synthases are conserved in mitochondria, the “powerhouse” organelles in humans and all eukaryotes (6). Fortunately, although BDQ binds at functionally conserved sites in the transmembrane domain (F_o) (7) (Fig. 1, Lower Left, red atoms on c-ring), chemical optimization of BDQ yielded >10,000-fold selectivity for inhibiting the mycobacterial enzyme vs. the mitochondrial enzyme (8). However, this exquisite selectivity made BDQ ineffective against other genera of pathogenic bacteria (3). Modified diarylquinolines can kill gram-positive bacteria like Staphylococcus aureus [the cause of methicillin-resistant S. aureus (MRSA) infections], but reduce the selectivity to just 10-fold vs. the mitochondrial enzyme (9). This strongly suggests that future success in developing antibacterials against this target may depend on identifying bacteria-specific features of the enzyme. This is the promise of the study by Zhang et al. (10) in PNAS that has determined an initial structure for the F₁-catalytic domain of the ATP synthase from M. smegmatis (designated herein as Msm-F₁). Although of modest resolution, the structure offers some tantalizing insights on possible bacteria-specific inhibitory mechanism(s) that could better guide our future aim at this important target for TB and hopefully for other challenging drug-resistant pathogens.

Fig. 1.

Ribbon models of structures determined for an ATP synthase (Left) and two catalytic F₁ complexes from bacteria are oriented by a conserved “core” region of their γ-subunits (18). F₁ subunits α (red), β (yellow), γ (blue), and ε (green) are colored as in ref. 10, with the εCTD in a lighter green. For F_oF₁ of P. denitrificans (Left) (15), all other subunits of F₁ (δ), F_o (abb′c₁₂), and parts of the inhibitor zeta are colored distinctly and labeled. The Pd-ε has a typical εCTD, but only the εNTD was resolved. The red spheres on the c-ring subunits identify the conserved H⁺ binding residues. For the F₁ structures from M. smegmatis (10) and Bacillus PS3 (19), α-subunits are not shown; each β-subunit is labeled according to its usual adenine nucleotide occupancy in prior F₁ structures (DP, diphosphate; E, empty; TP, triphosphate). The β_E-subunit of Msm-F₁ shows possible ADP and phosphate bound, as in the similar structure of F₁ from C. thermarum (21). The β_E-subunit of PS3-F₁ has a bound sulfate. The example of the ε-subunit down state is an isolated ε-subunit of E. coli (25). A homology model of Msm-ε in the up state (magenta) was made with I-TASSER (29), using the PS3 ε-subunit as template; it is shown separately and superimposed with the ε-subunit in PS3-F₁ (Right). Mass-weighted ellipses are shown (Lower Right) for the β-subunits of each F₁, as viewed from “below,” for P. denitrificans (yellow), M. smegmatis (mesh), and Bacillus PS3 (green). All images were produced with Chimera (30).

In aerobic bacteria and in mitochondria, energy from respiration is used to pump protons (H⁺) across a membrane, generating a proton-motive force. In turn, the ATP synthase uses the proton-motive force to drive synthesis of ATP, an essential energy “currency” used to fuel many cellular processes (11). The enzyme works as a rotary nanomotor: the “downhill” flow of H⁺ ions across the membrane through F_o turns the central c-ring (Fig. 1, Lower Left); this is directly coupled to drive rotation of an asymmetric central stalk (γ- and ε-subunits) inside the peripheral catalytic complex (F₁), which then drives conformational changes in the catalytic sites to achieve sequential synthesis of ATP at each site (6). The enzyme is reversible so that if the proton-motive force is too low, it can hydrolyze ATP and pump H⁺ ions in the opposite direction across the membrane. Also, F₁ can be separated from membranes in vitro and works only as an ATPase. Some bacteria, such as the nonrespiring pathogen Streptococcus pneumoniae, use F_oF₁ mainly as an ATPase-driven H⁺ pump, and it is essential for maintaining membrane potential and pH homeostasis (12). However, many organisms rely on F_oF₁ to generate most cellular ATP so that when thermodynamics would favor ATP hydrolysis, there are apparent mechanisms to reduce futile hydrolysis of ATP (6, 13).

Some bacteria, like Escherichia coli, can use F_oF₁ in ATPase mode during fermentative growth, and their isolated membranes have significant F-ATPase activity. In contrast, mycobacterial membranes have highly latent ATPase activity that can be revealed by in vitro treatments (14). Zhang et al. (10) confirm that this latency is intrinsic to Msm-F₁, which has ATPase activity ∼500× lower than that of E. coli F₁ (Ec-F₁). F₁ in some α-proteobacteria, such as Paracoccus denitrificans (Pd-F₁), is inhibited by a unique protein, zeta (Fig. 1, Left), that may be the progenitor of mitochondria’s specific inhibitor, IF₁ (15). But the enzyme’s most ancient inhibitor protein is likely the C-terminal domain (CTD) of its own small ε-subunit (16, 17). Only ε-subunit’s N-terminal domain (εNTD) is an essential part of the rotor connecting F₁ to F_o in all species (Fig. 1, Left), and the εCTD is absent in a few bacterial species. However, in diverse bacteria and in chloroplasts, the εCTD can inhibit F₁ and F_oF₁ to varying degrees. As seen with isolated ε-subunits and in several structures of F₁ or F_oF₁, ε-subunits can adopt a state in which its NTD and CTD pack together and the εCTD does not contact other F₁ subunits; biochemical studies have shown that this ε-subunit “down” state (Fig. 1) allows the enzyme to be active. The ε-subunit down state also occurs in its homolog in mitochondrial F₁ (mito-F₁), but a mitochondria-specific subunit locks it in that state, and there is no evidence that the ε-subunit homolog inhibits mito-F_oF₁ (6). So far, inhibitory states of ε-subunits have been seen in Ec-F₁ (18) and in F₁ of Bacillus PS3 (PS3-F₁) (19) (Fig. 1, Right). In each case, the εCTD swings away from the εNTD, and a final segment of the εCTD extends “up” into the central cavity of F₁, jamming the rotary motor in a transiently inactive state. The mycobacterial εCTD is approximately one-third shorter than in most bacteria, but isolated ε-subunit of M. tuberculosis can fold in the down state, but with shorter α-helices in the εCTD (20). Zhang et al. (10) now observe Msm-F₁ with the ε-subunit in a similar down state (Fig. 1, Center). The ε-subunit down state was also seen in structures of another latent F₁-ATPase from an extremophile, Caldalkalibacillus thermarum (Ct-F₁) (21). This suggests that their latent ATPase activities are not due to inhibition by the εCTD. However, a prior study truncated or mutated the εCTD of C. thermarum and indicated that the εCTD is involved in the inhibited state of the wild-type enzyme (22). Now, Zhang et al. (10) use a protease, trypsin, to activate ATPase of Msm-F₁ up to 100-fold.

Zhang et al. whet the appetite for more structural details on the ATP synthase from mycobacteria and other bacterial pathogens, with the promise of identifying better ways to selectively target this important bioenergetic enzyme for development of new antibacterials.

While further studies could identify another cause of proteolytic activation, only degradation of the ε-subunit appears to correlate with the rapid activation period. Prior studies showed that trypsin activates Ec-F₁ and Ec-F_oF₁ by cleaving the εCTD (23–25). Homology modeling suggests that Msm-ε could form an up state (Fig. 1, Center, magenta up model), but it would not penetrate the cavity of F₁ as deeply as seen in Ec-F₁ or PS3-F₁ (Fig. 1, Right). Thus, it seems possible that an ε-inhibited state could be predominant in Msm-F₁ in solution, but the ε-subunit down state is preferred in the crystallization conditions used (10). One possible test would be to express and study Msm-F₁ without the ε-subunit, as done routinely with PS3-F₁. If confirmed, ε-inhibition could provide a promising new means to target F_oF₁ in M. tuberculosis, since it does not occur in mitochondrial ATP synthase. Since ε-inhibition is seen with the E. coli and PS3 enzymes, it could certainly be investigated for potential importance in related species. For instance, Salmonella can grow without F_oF₁ in laboratory cultures, but F_oF₁ function is critical for its capacity to survive and replicate within macrophages (26).

As the structure determined shows the ε-subunit in the down state, Zhang et al. (10) focus on similarities between Msm-F₁ and the structure of another latent enzyme, Ct-F₁ (21) to consider the cause of such strong inhibition. First, these structures align closely with each other at a specific rotary angle (i.e., the position of the γ-subunit vs. the three surrounding β-subunits). In Fig. 1, Lower Right, it is clear that Msm-F₁ is at a distinct angle vs. ε-inhibited PS3-F₁, but Msm-F₁ is much closer in angle to zeta-inhibited Pd-F₁ and to several structures of mito-F₁ not shown here (18). The other similarity between Msm-F₁ and Ct-F₁ is more intriguing: Both appear to have the usually empty β_E-subunit site occupied by ADP and phosphate but without Mg²⁺, even though the crystallization medium contained excess Mg²⁺. Interestingly, one mito-F₁ structure has ADP, but not Mg²⁺ or phosphate, on the β_E-subunit (27), and this structure is at the same rotary angle as ε-inhibited Ec-F₁ and PS3-F₁ (28). This could imply that Msm-F₁ and Ct-F₁ fall into an inhibited state because they have lost Mg²⁺ at the “wrong” angle in the rotary cycle.

Clearly, more functional studies will be needed to discern what mechanisms actually control the highly latent ATPase of mycobacterial F_oF₁. But Zhang et al. (10) whet the appetite for more structural details on the ATP synthase from mycobacteria and other bacterial pathogens, with the promise of identifying better ways to selectively target this important bioenergetic enzyme for development of new antibacterials.