Disrupting coupling within mycobacterial F-ATP synthases subunit ε causes dysregulated energy production and cell wall biosynthesis

The dynamic interaction of the N- and C-terminal domains of mycobacterial F-ATP synthase subunit ε is proposed to contribute to efficient coupling of H+-translocation and ATP synthesis. Here, we investigate crosstalk between both subunit ε domains by introducing chromosomal atpC missense mutations in the C-terminal helix 2 of ε predicted to disrupt inter domain and subunit ε-α crosstalk and therefore coupling. The ε mutant εR105A,R111A,R113A,R115A (ε4A) showed decreased intracellular ATP, slower growth rates and lower molar growth yields on non-fermentable carbon sources. Cellular respiration and metabolism were all accelerated in the mutant strain indicative of dysregulated oxidative phosphorylation. The ε4A mutant exhibited an altered colony morphology and was hypersusceptible to cell wall-acting antimicrobials suggesting defective cell wall biosynthesis. In silico screening identified a novel mycobacterial F-ATP synthase inhibitor disrupting ε’s coupling activity demonstrating the potential to advance this regulation as a new area for mycobacterial F-ATP synthase inhibitor development.

. Structural features of the F-ATP synthase and the mycobacterial coupling subunit ε. (a) The structural model of the mycobacterial F-ATP synthase was generated based on the cryo-EM model of E. coli F-ATP synthase (PDB ID: 5T4O) 47 and the structural model of the α 3 β 3 γε complex of the mycobacterial F-ATP synthase 7 . Subunits α (green), β (orange), γ (yellow), and ε (magenta) are from the M. smegmatis crystal structure (PDB ID: 6FOC) 7 , whereby the missing β-sheet elements of subunit γ were added by the respective subunit γ elements of the E. coli F-ATP synthase (PDB ID: 5T4Q). Mycobacterial subunit γ has a unique γ-loop 13 which is highlighted in red. The solution shape of α chi is displayed as green sphere 14 . The c-ring loop residues of M. phlei (wheat; PDB ID: 4V1G 9 ), proposed to interact with the rotating ε and γ stalks subunits are indicated. The modelled extended ε subunit (blue) of M. tuberculosis (PDB ID: 5YIO) reaches the DELSEEDregion in subunit α 10 . The low-resolution shapes of the related E. coli subunits a, b-dimer and δ are shown in the brown, light blue and red, respectively, and based on the EM map (PDB ID: 5T4O). (b) NMR solution structure of Mtε (PDB ID: 5YIO) 10 showing R105A, R111A, R113A, and R115A (magenta) in helix-2 of the C-terminal domain. (Inset) Closer view of the back surface revealing the interaction of R113 with E85 of the N-terminal domain, which is in close proximity to the hydrophobic cleft proposed to bind BDQ. We thank Dr. S. S. M. Malathy for the art work of (a).
Here, we used a complementary multidisciplinary approach to shed light on the essentiality of crosstalk between the NTD and CTD within mycobacterial subunit ε and the catalytic headpiece for ATP formation by introducing chromosomal missense mutations (R105A, R111A, R113A and R115A) in the C-terminal helix 2 of M. smegmatis mc 2 155 F-ATP synthase subunit ε (ε(R105A,R111A,R113A,R115A) mutant), called ε 4A throughout the text, predicted to interrupt inter-domain and subunit ε-α crosstalk and therefore coupling. The cell morphology of the mutant bacilli changed, and the overall cell length was only one-third of the size of wild-type bacteria (WT). Substitution of these amino acids caused a more than 10-fold drop in ATP synthesis and a moderate reduction in ATP hydrolysis of inverted membrane vesicles (IMVs) derived from M. smegmatis ε 4A mutant as well as a reduction in the intracellular ATP level of intact bacteria, when grown in minimal media. Both ETC-and carbon metabolism activities were increased in the mutant strain, which is discussed in context of how the TB drug BDQ works. Generating recombinant subunit ε mutants provided structural insights into alterations of the residues necessary for the mechanistic talks between the N-and C-terminal domains of this subunit. Finally, a new mycobacterial F-ATP synthase inhibitor was discovered.

Effect of epsilon regulation of F-ATP synthase on bacterial growth kinetics and cellular energetics.
In order to determine whether the arginine residues R105, R111, R113 and R115 with the NTD of mycobacterial subunit ε are essential for the crosstalk between both the NTD and the CTD, the M. smegmatis mc 2 155 F-ATP synthase mutant ε 4A with the substitution of the four arginine residues inside subunit ε to alanine was engineered. Targeted sequencing confirmed the presence of the desired mutations. Whole genome sequencing performed on both WT and the ε 4A mutant suggests that there are only several minor polymorphisms (Table S1). The ε 4A mutant and isogenic WT strain were first examined in rich 7H9 nutrient liquid medium supplemented with 0.2% glycerol. The maximum specific growth rate of the ε 4A mutant in this medium was similar to that of the WT strain (Fig. 2a). When grown in minimal medium containing 0.2% glucose, the mutant (doubling time of 8.9 h) grew significantly slower than the WT strain (doubling time of 6.4 h), and the final optical density was lower in the ε 4A mutant (Fig. S1a,b), which can be seen directly from their respective cell cultures at different time points (Fig. S1b). It is noteworthy that the mutant formed clumps in minimal media (Figs S1b and S2). Furthermore, slower bacterial growth correlated with a reduced intra-bacterial ATP level (Fig. S1c). In addition, when grown in Hartmans-de Bont (HdB)) minimal medium containing 0.2% glycerol (fermentable carbon source), the ε 4A mutant grew significantly slower than the WT strain, with doubling times of 4.1 h and 3.2 h, respectively (Fig. 2b). The growth was partially rescued in the ε 4A mutant complemented with the atpC gene from WT (Fig. 2b). Despite the ε 4A mutant growing slower on glycerol, no significant differences were observed between all three strains in terms of their dry weight, molar growth yield on glycerol, final external pH and optical density at 600 nm (OD 600 ) ( Table 1). Since glycerol is a fermentable carbon source, and the F-ATP synthase is involved in oxidative phosphorylation, the non-fermentable carbon source acetate was used to exclude any energy production contributed from substrate level phosphorylation during bacterial growth. Under this condition, the growth of ε 4A mutant was slower and significantly delayed when compared to the WT strain (7.9 h vs 4.4 h, respectively; Fig. 2c). A comparison of the stationary phase (growth at 72 h) for all three strains showed that the complemented strain grew to a final OD 600 that was close to that of the WT strain, while the ε 4A mutant had a ~30% reduction in final OD 600 (Fig. 2c). In this context, we do not exclude that there are possibilities that phenotypes those were not complemented by the expression of WT (i.e. Fig. 2b,c) were due to the effect of the background mutations.
Reduced growth correlated with a dramatically lower intracellular ATP level in the ε 4A mutant when compared to the WT strain (Fig. 2d). The ATP level of the complemented strain was restored to WT values (Fig. 2d). The molar growth yield on acetate [Y acetate ; dry weight of cells (g) produced with one mol of acetate utilized] was 5.78 ± 0.47 g/mol in the ε 4A mutant compared to 9.45 ± 0.78 g/mol for the WT strain ( Table 1). The external pH of the ε 4A mutant culture on acetate was more acidic (despite less biomass) compared to the WT strain (Table 1). These data demonstrate that carbon sources like acetate that are strictly coupled to the F-ATP synthase are metabolized less efficiently in the ε 4A mutant compared to the WT strain.
The growth rate and molar growth yield of the ε 4A mutant during growth on acetate was significantly lower than the WT strain suggesting that regulation by the epsilon subunit was crucial for efficient energy coupling. To assess the bioenergetic properties of the ε 4A mutant compared to the isogenic WT parent and complemented strain in more depth, we performed extracellular flux analysis (Fig. 3a-c). In this assay, oxygen consumption rate (OCR) represents the activity of the respiratory chain, while the extracellular acidification rate (ECAR) represents the activity of carbon metabolism and the TCA cycle 19 . Both respiration and carbon metabolism activity were significantly increased in the ε 4A mutant: the OCR of the mutant was 4.6-fold higher and the ECAR was 10.1-fold higher than WT cells (Fig. 3a). This profile would be indicative of an uncoupled cell with a depolarized membrane, as the uncoupler carbonyl cyanide m-chlorophenyl hydrazine (CCCP) causes an increase in both OCR and ECAR 19 . If true, CCCP should not stimulate OCR in the ε 4A mutant. However, the spare respiratory capacity (the degree of OCR stimulation by canonical uncouplers like CCCP) was increased by 4.5-fold in the ε 4A mutant (Fig. 3c). Taken together, these data suggest that the ε 4A mutant has upregulated respiratory and carbon metabolism pathways in order to maintain a fully energized membrane. We propose that the substitution of the four C-terminal arginines in the ε subunit leads to either increased uncoupled H + -pumping-and/or ATP synthesis activity of the F-ATP synthase, or dysregulated entry of protons into the cytoplasm: both of which require increased respiration to compensate (see below).
www.nature.com/scientificreports www.nature.com/scientificreports/ BDQ susceptibility is enhanced in the ε 4A mutant strain. Recently it has been shown that C-terminal residues of Mtε, including R109 and R115, undergo changes in the chemical shift upon binding of BDQ to the NTD of subunit ε, which appears to be due to interactions between the NTD and CTD 10 . To determine whether the amino acid replacement inside the ε 4A mutant alters BDQ susceptibility of the bacterium, growth inhibition dose-response curves were determined using the broth dilution method as described earlier 20 . As shown in Fig. S3a, the mutation of R105, R111, R113 and R115 to alanines render the mutant strain more sensitive to BDQ with a 3-fold shift in minimal inhibitory concentration (MIC 50 ) from 10.3 nM to 3.2 nM. Since BDQ is known to have delayed bactericidal activity during the first 3-4 days 20 , we measured the effect of BDQ on cell viability ( Fig. S3b-e). As displayed in Fig. S3b, BDQ showed very little bactericidal activity against WT M. smegmatis mc 2 155 during the first two days of culture, reflected by an increasing cell number of about 1.1 log 10 units in CFUs. Such increase in cell viability over the first two days was repeatedly not observed in the ε 4A mutant (Fig. S1c), indicating that the altered BDQ susceptibility went in part along with bacterial killing, and that subunit ε plays a critical role in it.
Effect of the ε 4A mutation on bacterial morphology and resistance to cell wall antimicrobials. Because of the growth differences between the ε 4A mutant and WT, the effect of the introduced mutation was further investigated in terms of colony size and cell morphology (Fig. 4). The WT M. smegmatis strain and the ε 4A mutant were plated on 7H10 agar plates and similar colony sizes were observed for both the WT and the mutant strain (Fig. 4a-c). Interestingly, the irregular opaque colonies of WT with their flat surface were replaced by a circular translucent granular feature with a slightly elevated center in the ε 4A mutant (Fig. 4a,b). This phenotype was partially restored when ε 4A mutant was complemented with the WT atpC gene (Fig. 4c). Surprisingly, when observed under a light microscope, the ε 4A mutant bacilli displayed a dramatically shorter cell length of 1.63 ± 0.38 μm, which is only one-third of the size of WT (4.60 ± 1.26 μm) (Fig. 4b). Due to the difference in , ε 4A mutant strain (red) and its complemented mutant strain (blue) in HdB minimal media containing (b) 0.2% glycerol or (c) 0.2% acetate. Growth was determined by measuring cell density using optical density at 600 nm (OD 600 ). The growth differences of all three strains could be observed in both growing conditions, in which the ε 4A mutant strain grew slower than WT strain, and the growth was partially rescued by introducing pMV262-atpC into the mutant strain. (d) Intracellular ATP levels measured in WT, ε 4A mutant and ε 4A complemented strain at mid-log phase in minimal media containing 0.2% acetate. All growth experiments are performed in three biological replicates, and the error bar represents standard deviations. colony morphology and cell size in ε 4A mutant, further investigation was done on the susceptibility of WT M. smegmatis, ε 4A mutant and ε 4A complemented mutant strains to the β-lactam antibiotic meropenem that targets cell wall synthesis of the bacteria 21 . Growth inhibition dose-response curves were determined using the broth dilution method as described earlier 21 . As shown in Fig. 4d, the ε 4A mutant revealed a reduced susceptibility by meropenem resulting in a minimum inhibitory concentration (MIC 50 ) of 10 μM. In comparison, WT and the complemented strain showed a similar inhibitory curve with a MIC 50 of 3 μM. The effect of meropenem on the cell viability of these mutants was determined (Fig. S4). While meropenem at a concentration of 3x MIC 50 did not affect cell viability against WT during the first four days of incubation (Fig. S4a,c), the same inhibitor concentration reduced moderately and reproducibly cell viability of the ε 4A mutant (Fig. S4b,d). In comparison, at 30x MIC 50 a drop of about 1.4 log 10 units and 2.8 log 10 units in CFUs was observed for the WT and ε 4A mutant strain, respectively ( Fig. S4a,b).

the crosstalk between the c-terminal helix and the ntD of subunit ε is critical for Atp synthesis and -hydrolysis.
To confirm that the reduction of intracellular ATP formation is related to the effect of the C-terminal substitutions of the ε 4A mutant, ATP synthesis was investigated using IMVs. As demonstrated in Fig. 5a, the IMVs of WT revealed an ATP synthesis activity of 3.24 ± 0.07 nmol min −1 (mg total protein) −1 . In contrast, when IMVs containing the F-ATP synthase mutant ε 4A were used, low ATP synthesis of only 0.23 ± 0.01 nmol min −1 (mg total protein) −1 was observed. To confirm that the substitution of the four C-terminal arginine residues of subunit ε causes the reduction of ATP synthesis activity, the IMVs of the complemented mutant was studied, revealing an ATP synthesis activity of 3.53 ± 0.08 nmol min −1 (mg total protein) −1 , which is similar to WT IMVs (Fig. 5a). Since the Western-blot analysis in Fig. 5b confirms that the amount of F-ATP synthases located in the ε 4A mutant vesicles was comparable with that of the WT and complemented mutant vesicles, and that subunit ε is present within the enzyme complex, the data demonstrate the crucial role of the mycobacterial subunit ε residues R105, R111, R113 and R115 in ATP synthesis.
Since we observed increased OCR and ECAR in the ε 4A mutant but reduced ATP synthesis of the ε 4A mutant IMVs, we tested whether ε 4A mutant IMVs allow H + to move freely across the membranes or whether the increased H + -pumping activity is uncoupled to ATP synthesis. Therefore, we firstly tested the leakiness of WT, ε 4A mutant and ε 4A complemented mutant IMVs to protons in the presence of the fluorescent dye 9-amino-6-chloro-2-methoxyacridine (ACMA). As shown in Fig. 5c, in the presence of ADP + P i , but absence of NADH and the uncoupler SF6847, neither the WT, nor the ε 4A mutant or ε 4A complemented IMVs showed a change in fluorescence indicating that the IMVs were intact, which was confirmed by the significant and fast quenching of ACMA after addition of NADH as well as the increase in fluorescence observed upon addition of SF6847. Interestingly, in the presence of NADH, WT-and ε 4A complemented mutant IMVs revealed a more drastic fluorescence quenching compared to the ε 4A mutant IMVs, reflecting an increased H + -conduction of the F O domain within the ε 4A mutant to the outside without coupled ATP synthesis.
In parallel, the effect of the ε 4A mutant on ATP hydrolysis was investigated using IMVs. As demonstrated in Fig. 5d,e, IMVs of WT M. smegmatis revealed an ATPase activity of about 41.15 ± 0.99 nmol min −1 (mg total protein) −1 , underlining recent results, which demonstrated that M. smegmatis hydrolyses ATP, albeit at a relatively low level 12,13 . The IMVs of the ε 4A mutant showed an ATPase activity of 24.9 ± 0.5 nmol min −1 (mg total protein) −1 . Thus, the mutant caused a 40% decrease in ATPase activity, when compared to the ATPase activity of the WT enzyme (Fig. 5e). Similar to WT IMVs, the complemented mutant revealed an ATP hydrolysis activity of 45.9 ± 1.3 nmol min −1 (mg total protein) −1 ), demonstrating that the mutation of the four C-terminal arginines to alanine causes the alteration of ATPase activity.
Alterations of a monomeric to an oligomeric form of Mtε due to arginine substitutions. To understand the effects of the ε mutant described above in more detail, the four Mtε mutants MtεR113A, MtεR111A,R113A, MtεR111A,R113A,R115A, and MtεR105A, MtεR111A,R113A,R115A were genetically engineered. The overall protein production and solubility of the four mutants were similar to recombinant WT Mtε. www.nature.com/scientificreports www.nature.com/scientificreports/ WT Mtε and its four mutant proteins were isolated via Ni-NTA affinity purification, followed by a size-exclusion (SEC) step. The SEC chromatogram of WT Mtε showed the peaks I and II at 9.5 ml and 11.5 ml, which correspond to two different oligomeric formations of Mtε, and the major peak III (13.8 ml), representing monomeric Mtε (Fig. 6a) 10 . In comparison, the single mutant MtεR113A revealed the higher oligomer peak I (9.5 ml) and a peak II at 12.5 ml with similar absorbance (Fig. 6b), indicating an additional lower oligomeric form of the protein and that the R113 to alanine substitution shift the equilibrium from a monomer to oligomeric forms. Interestingly, the elution profile of the double mutant MtεR111A,R113A showed that peak II shifted to even a smaller elution volume compared to peak II of mutant MtεR113A and therefore an increase of the oligomeric form (Fig. S5a). In case of the triple and quadruple mutants the ratio of peak I to peak II at 9.5 and 11.5 ml increased (Figs 6c, S5b), reflecting that the additional R to A substitution increase higher oligomer formation. The shifts in elution volume due to the mutations engineered are highlighted by the overlay of all five elution diagrams (Fig. 6d). These data demonstrated that substitution of R113 led to a shift of monomeric to oligomeric Mtε, which increased by the additional substitution of R115 to alanine. Finally, the formation of MtεR111A,R113A,R115A, and MtεR105A,R111A, R113A,R115A prevents the formation of any monomeric Mtε.
Structural changes of MtεR113A and MtεR105A,R111A,R113A,R115A. To understand whether structural changes occurred within the single mutant MtεR113A, which eluted in part as a monomer, and the quadruple mutant MtεR105A,R111A, R113A,R115A, NMR experiments were performed with both mutants and the WT protein (Fig. 7). As shown by the overlay of the 2D 1 H-15 N HSQC spectra of recombinant Mtε and the mutants, overall structural changes and the presence of a mixed oligomeric-(MtεR113A) or formation of a highly ordered oligomeric state (MtεR105A,R111A,R113A,R115A) were observed (Fig. 7a). Based on the assigned resonance information of WT Mtε, the residues of MtεR113A with significant change in chemical shift were identified and mapped (Fig. 7b,c). The residues D12-W16, L59-L70, the NTD-CTD interacting amino acids L48-V53, and the connecting region between the NTD and CTD (residues S83-S88) were strongly affected by the R113A mutation. In addition, the high chemical shift perturbations (CSPs) values of amino acids within the second helix of the CTD, reflect that the mutation might cause an overall structural rearrangement of the CTD of Mtε (Fig. 7b). The mixed or highly ordered oligomeric states of MtεR113A and MtεR105A,R111A,R113A,R115A indicate that the arginine residues of Mtε are strongly related to the structural stability of the Mtε compact form.

Identification of a novel mycobacterial F-ATP synthase inhibitor.
Because of the effects of the ε 4A on growth, morphology, ATP synthesis, ATP hydrolysis, structural alteration, as well as the recently described dynamics of the NTD and CTD within the Mtε solution NMR structure 10 , and the importance of the mycobacterial ε subunit in coupling 10,12,22 , a model was developed using the Mtε solution structure (PDB ID: 5YIO) 10 for in silico compound screening to identify a novel mycobacterial F-ATP synthase inhibitor (Fig. S6). Together with property filters, molecular docking scores and visual inspection of ligand interaction with the intended arginine residues, 19 compounds were prioritized for experimental characterization. The identified compound was epigallocatechin gallate (EGCG), which is one of the principle polyphenolic compounds found in the leaves of Camellia sinensis (tea) 23 . EGCG blocked NADH-driven ATP synthesis of IMVs of WT M. smegmatis with a half-maximal inhibitory concentration (IC 50 ) of 155.6 ± 1.2 nM (Fig. 8a) and was also a potent ATP synthesis inhibitor of the IMVs of the slow growing M. bovis bacillus Calmette-Guérin (IC 50 = 2.2 ± 0.3 µM; Fig. 8b). In comparison, BDQ was active with an IC 50 of 1.14 ± 0.2 nM (M. smegmatis IMVs) and 7.05 ± 1.32 nM (M. bovis BCG). Similarly, EGCG inhibited ATP synthesis of M. smegmatis IMVs in the presence of succinate (Fig. 8c), indicating that EGCG does not bind to the NADH-dehydrogenase. To look further into the EGCG-target enzyme, ATP synthesis of IMVs of the BDQ-resistant mutant I66M with the substitution in the M. smegmatis subunit c was used 24 . As revealed in Fig. S7a, the BDQ-resistant mutant I66M showed a 10-fold reduction in ATP synthesis inhibition by www.nature.com/scientificreports www.nature.com/scientificreports/ BDQ. Interestingly, the IC 50 of EGCG with IMVs of the I66M mutant was similar to WT IMVs. Even more, addition of 750 nM of EGCG reduced ATP synthesis of the I66M mutant IMVs significantly (Fig. S7b), confirming that the mycobacterial F-ATP synthase is the enzyme target and that the novel F-ATP synthase inhibitor EGCG shows no cross resistance to BDQ. Furthermore, EGCG also inhibited ATP hydrolysis activity by more than 50% at concentrations of 100 µM in a manner comparable to the known ATPase inhibitors quercetin or BDQ 13 (Fig. 8d). No uncoupling effect was observed on the WT IMV in the presence of 200 nM of EGCG (Fig. S7c). A clogP value of 1.49 was calculated for EGCG.

Mechanistic insights into EGCG-binding to Mtε.
To confirm binding of EGCG to Mtε and to shed light into the mechanism of action, we carried out NMR titration using the highly resolved and dispersed NMR-spectrum of Mtε (Fig. 8e). The titration of 15 N-labeled Mtε with EGCG was done in a molar ratio of 1:2 (Fig. 8e). Disappearance of some of the cross-peak intensities in the 1 H-15 N HSQC spectrum, major and slight changes of 15 N-and HN-resonances for some residues, and gradual line broadening with increasing molar ratios were observed, which demonstrate binding of EGCG with Mtε. A plot of CSPs after addition of EGCG to labeled Mtε at a molar ratio of  www.nature.com/scientificreports www.nature.com/scientificreports/ 1:2 are shown in Fig. 8f. Significant changes in chemical shift higher than average plus standard deviation (CSP > 0.1 ppm) were detected for the amino acids F24, T25, G30, V42, R62, D66, G68, F69, A84 and R115. Smaller CSPs over than average (0.05 ppm < CSP < 0.1 ppm) were identified for residues A2, V9, A10, K21, F22, T28, V29, A43-L45, V53, A64, V65 I79, E82, R105, R111-R113, V117, G118 and D121. Residues R26, T27, E31, G33, V46-R52, L61, G67, A116 and I120 showed disappearance of resonances. Most of the amino acid residues affected by EGCG binding are located at the K21-G30, V42-V53, L61-L70 and R111-D121 regions. When mapped onto the NMR solution structure of Mtε (Fig. 8g), most of the residues affected by EGCG have been proposed to be involved in the interaction of the MtεNTD with its second helix and/or in stabilizing the compact (ε c ) conformation of Mtε in solution 10 .
To corroborate our initial docking result, we have exploited the NMR HSQC data constraints to obtain bound conformation of EGCG-Mtε via molecular docking experiments and evaluated the stability of protein-ligand interactions over 200 nano-seconds by molecular dynamic (MD) studies. Our MD results suggest that EGCG maintains key hydrogen bonding contacts to amino acids D48, D60, and V65 over almost the entire time length of 200 nanoseconds (Fig. S8a). Ligand binding was further stabilized by hydrophobic interactions at A64, I90 and C-terminal residues such as A112 and A116 (Fig. S8b,c). Further, the 5-OH and 5′-OH groups on 2-phenyl and gallic ester were involved in solvent-mediated hydrogen bonding interaction with residues R52, D66 and R62.

Discussion
Metabolic adaptation and morphological plasticity of mutant ε 4A . Mycobacteria are obligate aerobes. They can also live under hypoxic and even anaerobic conditions in a non-replicating state 25,26 . It is proposed that they adapt to a slow growth rate by using a variety of different enzymes like the alternative dehydrogenases and hydrogenases, which keep the flow of reducing equivalents to the ETC when energy is limited. Such www.nature.com/scientificreports www.nature.com/scientificreports/ adaptation processes can go along with morphological changes as described for the reduction (7-fold) of the M. smegmatis cell length in a carbon-limited model 25,26 , reflecting that mycobacteria are able to generate small replicating cells. Here the M. smegmatis ε 4A mutant caused a significant change in colony morphology, a 3-fold reduction of the cell size, and decreased susceptibility of the antibiotic meropenem. The clump-formation in minimal media may reflect changes in the cell envelope and bacterial surface properties of the mutant strain, which may be related to the reduction in the energy currency ATP, which is important for bacterial cell envelope formation and optimal growth 3,27 .
Interestingly, in 7H9 nutrient media the M. smegmatis ε 4A mutant showed reduction in biomass compared to WT, reflecting the effect of the mutation on overall growth. Considering the high amount of energy needed to build the cell wall, the morphological plasticity shown by the reduction (3-fold) of the M. smegmatis mutant cell size may represent a bacterial strategy to keep producing large cell numbers in an energy limited situation. The lower amount of ATP generated by oxidative phosphorylation may in part be compensated by a higher turnover of the glycolysis pathway and the TCA cycle, in which substrate level phosphorylation provides ATP, and where the reducing equivalents NADH and FADH 2 will be generated for the subsequent use in the respiratory chain for ATP-formation. In line with this interpretation is the reduced growth of the mutant strain in minimal media with the non-fermentable carbon source acetate. This result is in line with the drop of intracellular ATP of the ε 4A mutant, which reached close to WT-level in the complemented strain. Since the TCA cycle has a major biosynthetic role in generating important intermediates for fatty acid-, steroid synthesis as well as amino acid, purines and pyrimidines precursors, a higher TCA cycle turnover would enable sufficient synthesis for the synthesis of proteins, DNA, and the cell wall. In addition, malate, formed within the TCA cycle, can be used to form pyruvate and further refill the pool of glycolytic intermediates by gluconeogenesis 28 . Such processes have been shown to be essential in TB infection in mice and dormancy 29 . Furthermore, the observed reduction of the ε 4A mutant in external pH under acetate conditions may alter expression of genes encoding enzyme of the glyoxylate pathway and/or TCA cycle as described for the isocitrate lyase of M. tuberculosis 30 . The enzyme provides a short-cut of the TCA cycle resulting in glyoxylate and succinate 28 , with the latter providing the electrons for oxygen reduction Mechanistic and structural influence of the subunit ε mutations. The recent Mtε structure showed connections between the amino acids A10-W16 with the regions L61-A64 and A81-I90, respectively 10 . Based on these as well as the atomic structure of the mycobacterial c-ring 9 and 15 N relaxation data of Mtε 10 , a structural and mechanistic model was described 10 , in which rotation of the c-ring would influence the connection between the c-loop residues R45, Q46 and E48 with residues F24 and K21 at the bottom of the MtεNTD, respectively. Subsequently, this interaction could translate proton-conduction via the MtεNTD epitopes consisting of amino acids V51-V53, L61-A64, I8-A10, and D12-N14 to the region A81-I90 which is near to the MtεCTD and includes E85. Here, amino acid F86 of the MtεNTD and E87-I90 of the -CTD form a hinge region. The latter would enable up-and down movements of the C-terminal domain (Fig. 1a), resulting in a transition of a compact (ε c ) to an extended (ε e ) state of this subunit. Considering the hydrogen-bond interaction between E85 of the MtεNTD and R113 of the -CTD, substitution of R113 to an alanine is expected to interrupt this link and altering the interaction of both domains, including the proposed sequential coupling steps described above. Interestingly, the most extensive chemical shift perturbations of the MtεR113A mutant identified were A12-W16, D48-V53, K59-L70, S83-88 and S99-D121 (Fig. 7b), which encompass the residues that were reported to be important for inter-domain interactions, namely D47-A49, E87-D91, and A108-V117, as well as those from the BDQ-binding epitope, A10-W16 10 . The latter sheds light on the observation that the M. smegmatis ε 4A mutant showed a 3-fold higher MIC 50 for BDQ compared to WT bacteria (Fig. S3a). Furthermore, the interruption of E85 and R113, leading to extensive chemical shift perturbations of the residues D48-V53 and S99-D121, would change the recently described inter-domain interactions between the C-terminal chains of A108 and A112 with D47-A49 10 , and finally the regulation of the C-terminal up-and down movements.
The R-to-A substitutions are not only affecting the interactions at the interface between the NTD and CTD of mycobacterial subunit ε but also the interaction of its CTD with the C-termini of the α 3 :β 3 -domain. Recent crosslinking studies demonstrated the close neighborhood of residue D121 of Mtε with the DELSEED-region of α 10. Together with a structural model, these data led to a proposed arrangement of R113 in helix α2 of Mtε to come in vicinity to another α-subunit via V117, thereby contributing to a connection with the DELSEED-part (Fig. 1a). Furthermore, R115 of helix α2 of subunit ε comes in vicinity to the DELSEED-part. At the same time, two β-subunits are proposed to interact with helix α2 of Mtε. These interactions of Mtε with the C-termini of α-β would provide sufficient contact to ATP synthesis and hydrolysis inside the catalytic α 3 :β 3 -headpiece. These connections would be disrupted in the case of IMVs of the M. smegmatis ε 4A mutant, causing a drastic drop of ATP synthesis (Fig. 5a). An inhibitory effect caused by the mycobacterial subunit ε would support recent data of the IMVs of the M. smegmatis ε 1-120 mutant, with a deletion of the very C-terminal amino acid D121 10 . In comparison, mutants defective of the very 10 35 , 45 36 or 51 37 C-terminal residues of the E. coli subunit ε displayed no significant difference in cell growths or ATP synthesis, highlighting the structural, mechanistic and regulatory specificities of the mycobacterial protein.
As shown by solution NMR, introduction of R-to-A substitutions into the recombinant Mtε (MtεR113A, MtεR111A,R113A, MtεR111A, R113A,R115A, and MtεR105A,R111A, R113A,R115A) disrupts the interactions at the interface between the NTD and CTD that are critical to its structural integrity (Fig. 7a-c). While the loss of contact upon mutating R113 may possibly be alleviated by nearby arginine residues, albeit marginally, substituting all arginine residues results in the unravelling of the protein, which manifests as oligomers. Considering the genes for the enzyme complex are arranged in an operon, and the fact that subunit ε as well as similar amounts of entire F-ATP synthase were detected within the IMVs of WT and the M. smegmatis ε 4A mutant (Fig. 5b), the oligomerization of recombinant MtεR111A,R113A,R115A, and MtεR105A, R111A,R113A,R115A may not occur in vivo (Figs 6, S5a,b). However, the R-to-A substitutions are enough to interrupt dramatically the function of the molecular engine in synthesis direction and to a lesser degree, to a reduction (40%) in hydrolysis of ATP. These data reflect that ATP synthesis and hydrolysis are not simply reverse mechanisms.
The presented data showing that genetic disruption of ε's coupling function causes substantial loss of ATP synthesis activity, due to uncoupled H + -transduction, provided valuable information pertaining to the potential of this subunit as drug target for TB drug discovery. This is supported by unravelling EGCG as a novel inhibitor against F-ATP synthases of both fast-and slow growing mycobacteria. EGCG has reported activity against TB 38,39 and was described to impact the integrity of the mycobacterial cell wall 40 . Since the ε 4A mutant displayed (2019) 9:16759 | https://doi.org/10.1038/s41598-019-53107-3 www.nature.com/scientificreports www.nature.com/scientificreports/ changes in cell size, morphology and meropenem sensitivity, reflecting alterations in the cell envelope and bacterial surface properties of the mutant strain, which is related to ATP reduction, the data presented indicate, that EGCG binds to ε of the mycobacterial F-ATP synthase and disrupts the crosstalk between ε's NTD and CTD with consequences in ATP-and cell wall formation. Its novel mechanism of action reflects the spring-like coupling mechanism of the central mycobacterial ε subunit, described recently 10 . These novel insights also fittingly complement the previous studies that established the binding of BDQ to mycobacterial subunit ε 10,41,42 . By inhibiting ATP synthesis of the BDQ-resistant M. smegmatis F-ATP synthase mutant I66M, the current work and the unravelled EGCG pave the way for additional options for inhibiting ATP synthesis in M. tuberculosis. Finally, this work supports metabolism as a mediator of tolerance 42 . Metabolic reprogramming of mycobacteria as shown by the increased killing of M. tuberculosis by BDQ when grown on non-fermentable energy sources, affects bactericidal activity of the drug and thereby antibiotic tolerance 43 .

Methods
The detailed descriptions of all methods can be found in Supporting Information.
Bacterial strains and growth condition. In brief, M. smegmatis mc 2 155 (ATCC 700084) was used in this study as the parental strain. For standard cultivation, all mycobacterial strains were grown at 37 °C in Middlebrook 7H9 broth or on Middlebrook 7H10 agar plates unless otherwise stated, and antibiotics were added to the culture media as needed. For growth curve in 7H9, mid-log-phase pre-cultures (OD 600 = 0.4-0.6) were diluted to OD 600 of 0.05 and OD was measured at various time points until the cultures reached stationary phase. Smear for each strain was made from mid-log-phase pre-cultures and acid-fast stained using a TB stain kit (BD, 212520) according to manufacturer's instructions. The minimal media used in the study were either prepared as previously described 44 or as Hartmans-de Bont (HdB) media 45,46 . 0.2% glycerol, glucose or sodium acetate was added as a sole carbon source, and antibiotics were added to the culture media as needed. For growth curve in minimal media, pre-cultures were diluted to OD 600 of 0.005 and OD was measured at various time points until the cultures reached stationary phase.
construction of the M. smegmatis atpC(R105A,R111A,R113A,R115A) mutant and the complemented strain. The detailed descriptions of methods can be found in Supporting Information. To generate a quadruple mutant with substitutions of arginine codons to alanine codons in the gene atpC: R105A, R111A, R113A and R115A (ε 4A mutant) in M. smegmatis F-ATP synthase, site-directed genomic mutagenesis by recombineering was carried out as described previously 42 . The final 1,219 kb double-stranded DNA (dsDNA) oligonucleotide that was used in recombineering was created in two steps using PCR, and then transformed into electro-competent M. smegmatis mc 2 155, which harbours the plasmid pJV53 that express mycobacteriophage Che9c recombineering genes gp60 and gp61, both being necessary for dsDNA homologous recombination. The obtained transformants were screened by MAMA (mismatch amplification mutation assay) colony PCR and a final verification of all four mutations in a single mutant (ε 4A mutant) was done by DNA sequencing. For complementation, a plasmid pMV262 containing the WT allele of atpC was electroporated into the ε 4A mutant of M. smegmatis. The complemented strain: ε 4A mutant/pMV262-atpC was confirmed by PCR and this plasmid was maintained in media containing kanamycin.