Multiple acyl-CoA dehydrogenase deficiency kills Mycobacterium tuberculosis in vitro and during infection

The human pathogen Mycobacterium tuberculosis depends on host fatty acids as a carbon source. However, fatty acid β-oxidation is mediated by redundant enzymes, which hampers the development of antitubercular drugs targeting this pathway. Here, we show that rv0338c, which we refer to as etfD, encodes a membrane oxidoreductase essential for β-oxidation in M. tuberculosis. An etfD deletion mutant is incapable of growing on fatty acids or cholesterol, with long-chain fatty acids being bactericidal, and fails to grow and survive in mice. Analysis of the mutant’s metabolome reveals a block in β-oxidation at the step catalyzed by acyl-CoA dehydrogenases (ACADs), which in other organisms are functionally dependent on an electron transfer flavoprotein (ETF) and its cognate oxidoreductase. We use immunoprecipitation to show that M. tuberculosis EtfD interacts with FixA (EtfB), a protein that is homologous to the human ETF subunit β and is encoded in an operon with fixB, encoding a homologue of human ETF subunit α. We thus refer to FixA and FixB as EtfB and EtfA, respectively. Our results indicate that EtfBA and EtfD (which is not homologous to human EtfD) function as the ETF and oxidoreductase for β-oxidation in M. tuberculosis and support this pathway as a potential target for tuberculosis drug development.

M aintenance of an energized membrane is essential for Mycobacterium tuberculosis (Mtb) to grow and survive periods of non-replicating persistence 1 . This need has driven tuberculosis (TB) drug development efforts towards Mtb's energy metabolism. These efforts are supported by the first anti-TB drug approved in over 40 years-the ATP synthase inhibitor bedaquiline 2 .
Mtb's energy related pathways exhibit varying degrees of vulnerability to inhibition. Uptake of its main carbon sources in vivo is performed by specialized transporters: the multi-subunit Mce1 complex transports fatty acids 3 and the Mce4 complex facilitates the uptake of cholesterol 4 . Inactivation of Mce1 can reduce intracellular growth 5 , although it has also been reported to cause hypervirulence 6 , while Mce4 was conclusively shown to be essential for survival during the chronic phase of infection 4 . LucA, which acts as a regulator of both Mce1 and Mce4, is also required for wild type levels of Mtb virulence 3 , further supporting that inhibiting the ability to import host lipids affects Mtb's pathogenicity.
Cholesterol degradation yields multiple products, including acetyl-CoA, propionyl-CoA, pyruvate, and likely succinyl-CoA that can be used for energy generation or lipid biosynthesis [7][8][9][10][11] . Deletion of genes encoding cholesterol oxidation enzymes attenuated growth 12 and caused survival defects of Mtb in mice 13,14 . A screen against Mtb residing in the phagosomes of macrophages identified several compounds that target cholesterol metabolism 15 . In contrast, fatty acids are degraded solely through β-oxidation. Mtb's genome encodes multiple enzymes for each step of β-oxidation, including 34 putative acyl-CoA ligases, 35 putative acyl-CoA dehydrogenases, 22 putative enoyl-CoA dehydratase, five putative β-hydroxyacyl-CoA dehydrogenase, and six putative thiolases 16 . Some of these enzymes are necessary for infection, but they also have been shown to play roles in other pathways, such as complex lipid biosynthesis 17 and cholesterol degradation 18 . Due to the apparent redundancy of Mtb's fatty acid β-oxidation machinery, this metabolic pathway has thus been presumed to be invulnerable to chemical inhibition.
In the present study, we define an enzyme complex that has previously not been recognized as required for fatty acid degradation in Mtb. It consists of an electron transfer flavoprotein composed by two subunits-FixA (Rv3029c) and FixB (Rv3028c) -and a membrane oxidoreductase (Rv0338c), which we propose to re-name as EtfB Mtb , EtfA Mtb , and EtfD Mtb based on their human counterparts. Deletion of Mtb's EtfD causes multiple acyl-CoA dehydrogenase deficiency, which prevents utilization of fatty acids as carbon sources and can kill Mtb in vitro and during mouse infection.

Results
A possible role for EtfD Mtb in Mtb's energy metabolism. EtfD Mtb (Rv0338) is a membrane protein of unknown function predicted to be essential for growth of Mtb on agar plates 19,20 . To experimentally determine its topology, we fused E. coli alkaline phosphatase PhoA to EtfD at specific residues of the predicted transmembrane helices. Transport of PhoA outside of the cytoplasm enables reactivity with the chromogenic substrate 5bromo-4-chloro-3-indolyl phosphate p-toluidine (BCIP). Based on this assay, the soluble portion of EtfD Mtb faces the cytoplasm ( Supplementary Fig. 1a), and this topology agrees with the prediction generated by the MEMSAT3 algorithm 21 ( Supplementary  Fig. 1b).
Next, we performed in silico analysis, which placed EtfD Mtb into the cluster of orthologue groups 0247 (COG0247) composed of Fe-S oxidoreductases involved in energy production and conversion (Fig. 1a). COG0247 includes a variety of enzymes, including lactate dehydrogenase and the methanogenic-related heterodisulfide reductase; however, most proteins in COG0247 remain uncharacterized and are of unknown function. EtfD Mtb is predicted to be a chimeric enzyme assembled from four domains: an N-terminal domain, which is similar to the gamma subunit of nitrate reductases and putatively binds a cytochrome b; a central 4Fe-4S di-cluster domain similar to succinate dehydrogenases; and two C-terminal cysteine-rich domains (CCG), which are often found in heterodisulfide reductases (Fig. 1b). This analysis led us to hypothesize that EtfD is a component of Mtb's energy metabolism.
EtfD Mtb is linked to fatty acid metabolism and it is essential in vivo. To analyze the impact of EtfD Mtb depletion on Mtb's growth in vitro, we generated a TetOFF strain, in which EtfD levels are controlled by anhydrotetracycline (ATC)-inducible proteolysis 22 . Depletion of EtfD Mtb inhibited Mtb's growth in regular medium supplemented with oleic acid, albumin, dextrose and catalase (OADC), which is consistent with its predicted essentiality (Fig. 1c, d). Curiously, when the same medium was supplemented with a fatty acid free enrichment (albumin, dextrose and sodium chloride, ADN), depletion of EtfD Mtb did not impact growth to the same extent (Fig. 1c, d). This is similar to the fatty acid sensitive phenotype observed in response to inactivation of Mtb's type II NADH dehydrogenases 23 , which allowed the genetic deletion of Ndh-2 by growth in a fatty acid free medium. We applied the same strategy to etfD Mtb and generated a deletion strain (ΔetfD) (Supplementary Fig. 2a). The ΔetfD strain was confirmed by whole genome sequencing (WGS) to not contain additional polymorphisms known to affect growth (Supplementary Table 1). This knockout strain also phenocopied the fatty acid sensitivity observed with the knockdown mutant ( Supplementary Fig. 2b, c).
We next sought to investigate the importance of etfD Mtb for pathogenesis in an aerosol model of TB infection in mice. After aerosol infection, ΔetfD was unable to grow in mouse lungs and declined in viability from day 14 onwards (Fig. 2a and Supplementary Fig. 3a). In agreement with the CFU data, gross lung pathology showed no lesions in the mice infected with ΔetfD ( Fig. 2b and Supplementary Fig. 3b). The attenuation was even more pronounced in spleens, where no ΔetfD CFU were recovered at any time point ( Fig. 2a and Supplementary Fig. 3a). All phenotypes were rescued by reintroducing an intact copy of etfD Mtb .
Mtb requires EtfD Mtb to consume fatty acids and cholesterol as a carbon source and to prevent toxicity of long-chain fatty acids. The fatty acid sensitivity of ΔetfD suggested that this protein is, directly or indirectly, required for fatty acid metabolism. To test this further, we grew strains in media with different single carbon sources. ΔetfD was able to grow in both glycolytic (glycerol) and gluconeogenic (acetic acid and propionic acid) carbon sources, although at a slower rate than the wild type (Fig. 3a). However, longer chain fatty acids (butyric acid, palmitic acid and oleic acid) did not support detectable growth of ΔetfD (Fig. 3b). Fatty acids with four carbons or more in length require functional β-oxidation to be utilized, hence these results indicated that ΔetfD might display a defect in β-oxidation. Mtb oxidizes the cholesterol side-chain through β-oxidation. Enzymes involved in cholesterol side-chain degradation are essential for Mtb to grow with cholesterol as single carbon source 7,12 . Similarly, ΔetfD was not able to grow with cholesterol as a single carbon source (Fig. 3c).
To distinguish between two possible interpretations, namely 1) ΔetfD simply failed to use fatty acids as carbon source and 2) ΔetfD was intoxicated by fatty acids, we tested if glycerol could rescue the impaired growth of ΔetfD with fatty acids. Glycerol was able to restore growth in medium with butyric acid as carbon source, but it was not able to restore growth with long-chain fatty acids. Cholesterol partially inhibited growth in this mixed carbon source setup ( Supplementary Fig. 4). This argued for a toxic effect and led us to evaluate the impact of fatty acids and cholesterol on the viability of ΔetfD. We found that butyric acid and cholesterol were not toxic, while long-chain fatty acids were bactericidal to ΔetfD (Fig. 3d). Glycerol did not rescue the bactericidal effect of long-chain fatty acids, and partially rescued growth with cholesterol in the medium ( Fig. 3d and Supplementary Fig. 4).
Mtb acyl-CoA dehydrogenase activity requires EtfD Mtb . We applied metabolomics to understand why the consumption of fatty acids is prevented in the absence of EtfD Mtb . We focused these studies on 13 C 4 -labeled butyric acid to isolate the impact of etfD disruption on fatty acid consumption from potentially additional confounding effects associated with the toxicity of longer chain fatty acids. Inspection of labelled metabolites in central carbon metabolism confirmed the presence of butyric acid in WT, ΔetfD, and the complemented mutant ( Fig. 4 and Supplementary Fig. 5). All three strains were thus able to import 13 C 4 -labelled butyric acid. c Spot assay on solid media with an EtfD-TetOFF strain, where protein levels are controlled by anhydrotetracycline (ATC). Serial dilutions (10 6 down to 1 bacteria) were incubated for 14 days. These results are representative of three independent experiments. d Quantification of spot assays (c). This quantification aggregates the data from the three independent experiments. To quantify the spot assays, we picked the first dilution where isolated CFUs could be counted and divided it by the optical density (OD) of the corresponding bacterial suspension. The discrepancy between the OD of the bacterial suspension and the corresponding CFUs, which is due to sensitivity to lipids in the solid media, is reflected by this ratio. Data are averages with individual data points depicted. Error bars correspond to standard deviation. Statistical significance was assessed through paired two-tailed t-test. *P < 0.05; **P < 0.01. To be catabolized through β-oxidation, butyric acid needs to be transformed by an acyl-CoA ligase into butyryl-CoA. Strikingly, butyryl-CoA accumulated in ΔetfD approximately 45-fold relative to WT and the complemented mutant. The remaining intermediates of butyric acid β-oxidation were not detectable in any of the strains, but the pool size of the end-product acetyl-CoA was approximately 2-fold lower in ΔetfD than in WT. Labelled TCA cycle intermediates, with the exception of succinyl-CoA, were also partially depleted in ΔetfD ( Fig. 4 and Supplementary Fig. 5). The metabolomic profile of ΔetfD, specifically the accumulation of butyryl-CoA, strongly suggested that inactivation of EtfD interfered with the function of acyl-CoA dehydrogenases (ACAD), which in turn impaired fatty acid catabolism.
EtfD Mtb interacts with an electron transfer protein. We immunoprecipitated EtfD Mtb and identified putative interacting proteins by mass spectrometry. Among the 49 total hits (Supplementary Table 2), 41 were located or predicted to be located at the membrane/cell wall, while the remaining eight were cytoplasmic proteins. To investigate possible links to β-oxidation-a cytoplasmic process-we focused on the cytoplasmic interactors, Fig. 3 EtfD is necessary for the utilization of fatty acids that require β-oxidation. a-c Strains were grown in media with single carbon sources that a do not require and b require β-oxidation for catabolism, as well as cholesterol (c). Carbon sources were used at the following final concentrations: glycerol 25 mM, acetic acid 25 mM, propionic acid 2.5 mM, butyric acid 2.5 mM, palmitic acid 250 μM, oleic acid 250 μM, and cholesterol 250 μM. To sustain growth and avoid toxicity palmitic acid and oleic acid were replenished every 3-4 days for the first 14 days of culture. Cholesterol was replenished every to 3 to 4 days for the first 14 days to minimize precipitation. Data are averages of three replicates and are representative of three independent experiments. Error bars correspond to standard deviation. "Comp" stands for complemented. d Viability of ΔetfD in media with single carbon sources (butyric acid, palmitic acid, and oleic acid), or in mixed carbon sources (fatty acids and glycerol) at the same concentrations used in a and b was assessed at days 7 and 14. Data are averages of three replicates and are representative of three independent experiments. Error bars correspond to standard deviation. "Comp" stands for complemented. LOD stands for limit of detection.
which consisted of two flavoproteins EtfB Mtb and Rv1279, the sigma factor SigA, a putative helicase Rv1179c, an exopolyphosphatase Ppx1, the phthiocerol dimycocerosate (PDIM) biosynthesis enzyme PpsC, the protease ClpP2, and Rv1215c, a protein with putative proteolytic activity (Fig. 5a). We were especially interested in the flavoprotein EtfB Mtb , because of its homology (30% identity; 87% coverage) with the beta-subunit of the human electron transfer flavoprotein (ETF). Moreover, etfB Mtb (annotated as fixA-rv3029c) forms an operon with etfA Mtb (annotated as fixB-rv3028c), which shares homology with the alpha subunit of the human ETF (41% identity: 98% coverage). In humans, ETF 24 interacts with a cognate membrane oxidoreductase 25 (EtfD) and both are required to re-oxidize the FAD co-factor of multiple ACADs. This led us to hypothesize that Mtb might display a similar activity. Although EtfD Mtb is not a homolog of the human EtfD, our working model predicted that EtfBA Mtb and EtfD Mtb constitute a complex necessary for the activity of ACADs in Mtb (Fig. 5b).
To further assess a functional connection between EtfD Mtb and EtfBA Mtb , we asked if these proteins co-occur across bacterial proteomes. A BLASTp search against a database of 6240 bacterial proteomes (identity cutoff of >30% and coverage cutoff of >75%) identified 469 EtfD Mtb , 473 EtfB Mtb , and 472 EtfA Mtb homologs, 98% of which occur in actinobacteria, with the spirochete Leptospira interrogans-the causative agent of leptospirosis-as a notable exception ( Fig. 5; and Supplementary Data 1). EtfD, EtfB and EtfA showed a strong co-occurrence (P-value <10 −10 ), which was suggestive of a functional connection. Curiously, similar to Mtb, L. interrogans is proposed to use host-derived fatty acids as the primary carbon sources during infection 26 .
These results support the hypothesis that EtfD Mtb serves as oxidoreductase for EtfBA Mtb , which in turn is necessary for the activity of ACADs.
EtfBA Mtb and EtfD Mtb are required for acyl-CoA dehydrogenase activity. If EtfBA Mtb and EtfD Mtb participate in the same biochemical pathway then inactivation of EtfBA Mtb should also impair the use of fatty acids and cholesterol as single carbon sources by Mtb. To test this prediction, we first isolated a knockout strain for etfBA in fatty acid free medium (Supplementary Fig. 6) and confirmed its genetic identity through WGS (Supplementary Table 1). We then grew wild type, ΔetfBA and complemented ΔetfBA in media with different carbon sources. ΔetfBA was able to grow with glycerol, albeit slower than WT (Fig. 6a). In contrast, we did not detect growth in butyric acid, oleic acid, or cholesterol as single carbon sources, corroborating our prediction (Fig. 6b-d).  4 Stable isotope tracing reveals a block in β-oxidation at the level of acyl-CoA dehydrogenases. Strains were grown on filters on top of solid medium permissible to ΔetfD growth for 7 days and then transferred to solid media with 13 C 4 -labelled butyric acid (2.5 mM) as single carbon source for 24 hrs. Levels of the indicated 13 C-labeled metabolites (total 13 C pool, except for 13 C 4 -butyrate) were quantified by LC-MS analysis. Data are averages of three replicates and are representative of two independent experiments. Error bars correspond to standard deviation. "Comp" stands for complemented. Statistical significance was assessed by one-way ANOVA followed by post hoc test (Tukey test; GraphPad Prism). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. ns not significant. # metabolites with no statistically significant difference between wild-type and ΔetfD in the second independent experiment. To independently test the role of EtfBA Mtb and EtfD Mtb in βoxidation, we assessed if an acyl-CoA oxidase (ACO) could allow both ΔetfBA and ΔetfD to grow with fatty acids. ACOs are peroxisomal enzymes that catalyze the same reaction as ACADs using molecular oxygen to re-oxidize FAD rather than ETF/ ETFD 27 . The gene pox3, which encodes the extensively characterized ACO Pox3 from the yeast Yarrowia lipolytica 28,29 (Fig. 6d), was codon adapted for Mtb and expressed under the control of a strong, constitutive promoter in both ΔetfBA and ΔetfD. Pox3 has a high affinity for fatty acids with six and eight carbons chain length and low affinity for most other fatty acids 28 ; thus, we grew strains in a medium with octanoic acid (high affinity), butyric acid (low affinity), or oleic acid (low affinity) as single carbon sources. ΔetfD was not able to utilize octanoic acid as single carbon source, further confirming its inability to oxidize multiple fatty acids, while the growth rate of ΔetfD::pox3 was similar to that of wild type and complemented strain (Fig. 6e). A similar result was obtained with butyric acid, although it took longer for ΔetfD::pox3 to reach wild type density ( Supplementary  Fig. 7). As expected, based on the affinity profile and function of Pox3, pox3 did not rescue ΔetfD growth with oleic acid as single carbon source and it did not alter the growth rate on glycerol. Similarly, ΔetfBA did not grow on octanoic acid as single carbon source, while ΔetfBA::pox3 was able to grow in the same medium, although reaching a lower final optical density than WT and complemented mutant. ΔetfBA::pox3 was able to grow with butyric acid entering stationary phase 21 days after wild-type and complemented strain, it did not grow with oleic acid and it conferred a slight advantage on glycerol when compared with ΔetfBA ( Fig. 6e and Supplementary Fig. 7). Acyl-CoA oxidase activity was confirmed in the cell lysates of Pox3 expressing strains ( Supplementary Fig. 8).
These results confirmed that EtfBA Mtb and EtfD Mtb constitute a complex necessary for the activity of fatty acid β-oxidation ACADs.

Discussion
Mtb's energy metabolism displays a remarkable plasticity which supports its adaptation to a multitude of host microenvironments 30 . Its unusual domain structure suggested that the membrane protein encoded by rv0338c could be an unrecognized, essential component of Mtb's energy metabolism. Our studies identified Rv0338c as a member of a short electron transfer pathway essential for ACADs activity. Based on the similarity of this system to the human ETF system we propose to rename the respective Mtb genes as etfD Mtb (rv0338c), etfB Mtb (rv3028c), and etfA Mtb (rv3029c).
The increased susceptibility to fatty acids of an EtfD Mtb TetOff strain suggested a possible connection with fatty acid metabolism. Accordingly, our data showed that an etfD Mtb deletion mutant was not capable of utilizing fatty acids with four carbons or more as single carbon sources, thus indicating impairment in βoxidation. We also found that ΔetfD cannot utilize cholesterol. These metabolic defects were associated with different outcomes regarding viability: butyric acid (short-chain) was not toxic, while palmitic acid and oleic acid (long-chain) were bactericidal with or without glycerol and cholesterol negatively impacted growth in mixed carbon source medium (glycerol + cholesterol). Importantly, ΔetfD did not grow and presented a survival defect in mice. Long-chain fatty acids, including oleic acid, are common components of human macrophages 31 and carbon sources for Mtb. Similarly, cholesterol is a carbon source for Mtb during infection 4 . Hence, the inability to utilize fatty acids and cholesterol, together with the increased susceptibility to long-chain fatty acids likely explain the in vivo essentiality of EtfD Mtb . The mechanism of long-chain fatty acid toxicity to Mtb is still poorly understood, even though it was first described in the 1940's 32 . In other bacteria, several mechanistic explanations have been proposed for the bactericidal activity of long-chain fatty acids, including membrane potential disruption 33,34 , oxidative stress induction 35 , or fatty acid biosynthesis inhibition 36 . Blocking Mtb cholesterol side chain β-oxidation was previously shown to impair cholesterol utilization and to result in the accumulation of toxic intermediates, which negatively impacted growth even in the presence of alternative carbon sources 7,12 . We have found that ΔetfD displays a similar phenotype, further reinforcing a function of EtfD in β-oxidation.
The metabolome of ΔetfD in medium with butyric acid as sole carbon source revealed an accumulation of butyryl-CoA, which indicated an impairment in β-oxidation. This can explain the inability of ΔetfD to oxidize fatty acids and connects EtfD Mtb with the β-oxidation enzymes that act on acyl-CoAs-the ACADs. None of Mtb's thirty-five annotated ACADs are essential in vivo 37 and there are no reports showing a specific ACAD being essential in vitro for the utilization of any fatty acid. Thus, ACADs are likely to be functionally redundant. However, this redundancy was not sufficient to support growth of ΔetfD in media with fatty acids as single carbon sources, strongly suggesting that EtfD Mtb is necessary for the activity of multiple, if not all, ACADs.
Immunoprecipitation using EtfD Mtb as bait revealed multiple possible interacting proteins, suggesting that EtfD Mtb might integrate several pathways in Mtb. We were especially interested in the interaction with the cytoplasmic protein EtfB Mtb , which together with EtfA Mtb constitute a putative electron transfer flavoprotein (ETF). Based on the data available for EtfB and EtfA homologs 38 , these two proteins form a heterodimer. The fact that only EtfB was identified in the co-immunoprecipitation assay suggests that EtfD interacts more strongly with this subunit of Mtb's ETF. The human homologs (EtfAB) form an enzyme that re-oxidizes the FAD co-factor of multiple ACADs and transfers the electrons to the membrane bound oxidoreductase electron transfer flavoprotein oxidoreductase (EtfD), which then reduces the electron carrier ubiquinone, hence contributing to the generation of energy 39 . Mutations rendering defects in EtfAB or EtfD lead to a metabolic disease named multiple acyl-CoA dehydrogenase deficiency, which among other outcomes is characterized by the inability to oxidize fatty acids 40 . This led us to hypothesize that Mtb's EtfBA Mtb and EtfD Mtb might work together in a similar pathway. That EtfD Mtb and EtfBA Mtb show a strong pattern of co-occurrence across bacterial proteomes and ΔetfBA was unable to utilize fatty acids and cholesterol as single carbon sources were strong indications in favor of our hypothesis. Nevertheless, for a yet unclear mechanism, it was notable that ΔetfBA grew more slowly than ΔetfD. The expression of the acyl-CoA oxidase Pox3 28,29 , an enzyme that catalyzes the same reaction as ACADs, but uses molecular oxygen as an electron acceptor 27 , was able to rescue the growth of both ΔetfD and ΔetfBA in fatty acids as single carbon source, showing that in both cases β-oxidation was impaired at the ACAD step and confirming our proposed hypothesis. Although we do not have direct evidence, it is very likely that the same phenomenon explains the inability of ΔetfD and ΔetfBA to utilize long chain fatty acids and cholesterol. Additionally, both mutants, with ΔetfBA displaying a stronger phenotype, presented a growth defect even with glycerol as single carbon source. It is possible that defective β-oxidation might interfere, for example, with mycolic acid recycling 41 . Also, ACADs participate in other metabolic pathways such as branched-chain amino acids catabolism 42 , which might also have contributed to the growth defect in glycerol.
We are reporting a function for EtfD/EtfBA in β-oxidation, but it is possible that these proteins may have additional roles in Mtb. It has recently been reported that both EtfBA Mtb and EtfD Mtb are essential for resisting toxicity in media containing heme 43 . This further strengthened the functional relation between these proteins and suggested a role in iron metabolism. Interestingly, transcript levels of etfD Mtb , but not etfBA Mtb , respond to the iron content of the medium in an IdeR (iron metabolism transcriptional regulator) dependent manner 44,45 . It is also noteworthy, that the investigation of the Mtb transcriptional response to a series of compounds that inhibit EtfD 46 suggest a role in redox sensing. Curiously, it was shown that transcription of etfD is repressed by hydrogen peroxide 47 . It is thus possible that EtfD might also act as a redox sensor. Whether EtfBA Mtb and/or EtfD Mtb complex have a direct or indirect role in heme utilization and redox homeostasis is a question worthwhile to be addressed in the future.
In conclusion, we have identified a complex composed of EtfBA Mtb and the cognate membrane oxidoreductase EtfD Mtb that is required for the function of multiple ACADs. This complex constitutes a previously unsuspected vulnerable component of Mtb's β-oxidation machinery. EtfD Mtb has no structural homologs in humans and was recently proposed as the target of a series of compounds that are bactericidal against Mtb 46 , which suggests that it could be a potential target for TB drug development. The presence of EtfD Mtb and EtfBA Mtb homologs in Leptospira interrogans suggests that this pathway might be relevant for other pathogens that rely on host fatty acids as carbon sources 26 .

Methods
Culture conditions. For cloning purposes we used Escherichia coli as a host, which was cultured in LB medium at 37°C. Mtb was cultured at 37°C in different media: Middlebrook 7H9 supplemented with 0.2% glycerol, 0.05% tyloxapol, and ADNaCl (0.5% fatty acid free BSA from Roche, 0.2% dextrose and 0.85% NaCl) or in Middlebrook 7H10 supplemented with 0.5% glycerol and 10% oleic acid-albumindextrose-catalase (OADC), and in a modified Sauton's minimal medium (0.05% potassium dihydrogen phosphate, 0.05% magnesium sulfate heptahydrate, 0.2% citric acid, 0.005% ferric ammonium citrate, and 0.0001% zinc sulfate) supplemented with 0.05% tyloxapol, 0.4% glucose, 0.2% glycerol, and ADNaCl with fatty-acid-free BSA (Roche). Modified Sauton's solid medium contained 1.5% bactoagar (BD) and glycerol at a higher concentration (0.5%). For single or mixed carbon source cultures we have used glycerol 25 mM, sodium acetate 2.5 mM, propionic acid 2.5 mM and butyric acid 2.5 mM. Octanoic acid, palmitic acid, oleic acid, and cholesterol were dissolved in a solution of tyloxapol and ethanol (1:1) and added to the medium at a final concentration of 250 μM to modified Sauton's minimal medium with 0.5% fatty acid free BSA (Roche) and 0.85% NaCl. Since fatty acids of eight carbons or more are toxic in the mM range, cultures were replenished every 3-4 days for the first 14 days, as described elsewhere 48 . We also used this strategy for cholesterol medium, due to its poor solubility. Tyloxapol and ethanol (1:1) was also added to the cultures with glycerol, acetic acid, propionic acid and butyric acid in the same quantities, to control for a possible impact on Mtb's growth. Mycobacterium smegmatis MC 2 155 was cultured in Middlebrook 7H10 supplemented with 0.2% glycerol at 37°C. Antibiotics were used at the following final concentrations: carbenicillin 100 μg/ml, hygromycin 50 μg/ml and kanamycin 50 μg/ml. Mutant construction. Mtb H37Rv etfD Mtb conditional knockdown was generated using a previously described strategy that control expression through proteolysis 49 .
Briefly, a Flag tag and DAS + 4 tag were added to the 3′ end of etfD Mtb , at the 3′ end of the target gene. This strain was then transformed with a plasmid expressing the adapter protein SspB under the control of TetR regulated promoters. In the presence of anhydrotetracycline (ATC) 500 μg/ml, TetR loses affinity to the promoter and sspB expression is de-repressed. SspB acts by delivering DAS + 4-tagged proteins to the native ClpXP protease. Hence, when ATC is added to the culture, SspB expression is induced and EtfD Mtb -DAS-tag is degraded, working as a TetOFF system (EtfD Mtb -TetOFF).
We have obtained deletion mutants through recombineering by using Mtb H37Rv expressing the recombinase RecET. Constructs with the hygromycin resistant gene (hygR) flanked by 500 bp upstream and downstream of the target loci were synthesized (GeneScript). In the case of etfD, since the flanking gene aspC is in the same orientation and it is essential for growth 50 , we have included in the construct the constitutive promoter hsp60 to avoid polar effects. Mutants were selected in modified Sauton's with hygromycin. The plasmid expressing recET was counterselected by growing the deleted mutants in modified Sauton's supplemented with sucrose 10 %. For complementation we have cloned etfD and etfBA under the control of the promoter phsp60 into a plasmid with a kanamycin resistant cassette that integrates at the att-L5 site (pMCK-phsp60-etfD and pMCK-phsp60-etfBA) and transformed the deleted mutants. The gene pox3 from Yarrowia lipolytica was codon adapted for Mtb use, synthesized (GeneScript), cloned into a plasmid under the control of the promoter pTB38, with a kanamycin resistant cassette that integrates in the att-L5 site (pMCK-pTB38-pox3) and transformed into both etfD Mtb and etfBA Mtb deletion mutants. All generated strains and plasmids are listed in Supplementary Tables 3 and 4, respectively.
Whole genome sequencing. The genetic identity of ΔetfD and ΔetfBA was confirmed by whole genome sequencing (WGS).
Between 150 and 200 ng of genomic DNA was sheared acoustically and HiSeq sequencing libraries were prepared using the KAPA Hyper Prep Kit (Roche). PCR amplification of the libraries was carried out for ten cycles. 5-10 × 10 6 50-bp paired-end reads were obtained for each sample on an Illumina HiSeq 2500 using the TruSeq SBS Kit v3 (Illumina). Post-run demultiplexing and adapter removal were performed and fastq files were inspected using fastqc 51 . Trimmed fastq files were then aligned to the reference genome (M. tuberculosis H37RvCO; NZ_CM001515.1) using bwa mem47. Bam files were sorted and merged using samtools48. Read groups were added and bam files de-duplicated using Picard tools and GATK best-practices were followed for SNP and indel detection49. Gene knockouts and cassette insertions were verified for all strains by direct comparison of reads spanning insertion points to plasmid maps and the genome sequence. Reads coverage data was obtained from the software Integrative Genomics Viewer version 2.5.2 (IGV) [52][53][54] .
PhoA fusion assay. Truncated versions of EtfD Mtb were fused with the E. coli alkalyne phosphatase PhoA. This enzyme requires the oxidative environment of the periplasm to be active and degrades the substrate BCIP generating a blue precipitate. We have fused phoA at different residues located in the transmembrane domains. Positive control consisted in PhoA fused with the antigen 85B, while the negative control was PhoA alone 55 . All plasmids were transformed in M. smegmatis MC 2 155 and the assay was performed in LB plates with and without BCIP.
Mouse infection. Mouse experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, with approval from the Institutional Animal Care and Use Committee of Weill Cornell Medicine. Forty-eight female, eight-week-old Mus musculus C57BL/6 (Jackson Labs) were infected with~100-200 CFU/mouse using an Inhalation Exposure System (Glas-Col). Strains were grown to mid-exponential phase and single-cell suspensions were prepared in PBS with 0.05% Tween 80, and then resuspended in PBS. Lungs and spleen were homogenized in PBS and plated on modified Sauton's medium to determine CFU/organ at the indicated time points.
Metabolomics. Strains were grown in modified Sauton's until an OD 580nm of 1 and 1 ml of culture was used to seed filters 56 placed on top of solid modified Sauton's medium. Bacteria grew for 7 days, after which the filters were transferred to solid modified Sauton's medium with butyric acid or 13 C-labelled butyric acid (Cambridge Isotope Laboratories, Inc) at a final concentration of 2.5 mM for 24 h. For metabolite extraction bacteria were disrupted by bead beating three cycles, 50 s (Precellys 24, Bertin technologies) in a solution of acetonitrile:methanol:water (4:4:2). The relative abundances of butyric acid, CoA species and TCA intermediates were determined using an ion-pairing LC-MS system, as previously described 57 . In brief, samples (5 uL) were injected onto a ZORBAX RRHD Extend-C18 column (2.1 × 150 mm, 1.8 µm; Agilent Technologies) with a ZORBAX SB-C8 (2.1 mm × 30 mm, 3.5 μm; Agilent Technologies) precolumn heated to 40°C and separated using a gradient of methanol in 5 mM tributylamine/5.5 mM acetate. Post-column, 10% dimethyl sulfoxide in acetone (0.2 ml/min) was mixed with the mobile phases to increase sensitivity. Detection was performed from m/z 50-1100, using an Agilent Accurate Mass 6230 Time of Flight (TOF) spectrometer with Agilent Jet Stream electrospray ionization source operating in the negative ionization mode. Incorporation of 13 C was quantitated and corrected for natural 13 C abundance using Profinder B.08.00 (Agilent Technologies).
Immunoprecipitation. We transformed ΔetfD with a plasmid expressing flag tagged EtfD under hsp60 promoter (pMEK-Phsp60-etfD Mtb -flag) and used WT Mtb expressing only the flag tag as a control. Mtb whole-cell lysates were collected from 120 ml log phase culture in butyric acid single carbon source Sauton's medium, incubated with 1% DDM for 2 h on ice, followed by anti-Flag beads (Sigma) overnight incubation with gentle rotation. Beads were collected on the second day, washed with lysis buffer (50 mM Tris-HCl, 50 mM NaCl, pH 7.4), and eluted with 100 ng/μl Flag peptide. The eluates were resolved on SDS-PAGE before mass spectrometry.
For mass spectrometry analysis, the total spectrum count (TSC) from biological duplicates were summed. We calculated the ratio of summed TSC from EtfD Mtb -Flag vs. Flag control and used a cut-off of ≥10.
In silico analysis. Transmembrane domain topology of EtfD Mtb was performed in MEMSAT3 21 . Domain architecture of EtfD Mtb , EtfB and EtfA was based on HHPred 60 and XtalPred-RF 61 . Eggnog 62 was used for the cluster of orthologous groups analysis (COG). The members of COG247 that include etfD were used to generate a rootless phylogenetic tree in iTOL (version 6) 63 .
To analyze the presence or absence of EtfD Mtb , EtfB Mtb , and EtfA Mtb homologs across bacterial species, we obtained the set of UniProt reference bacterial proteomes, which are are selected both manually and algorithmically by UniProt as landmarks in (bacterial) proteome space 64 . We discarded proteomes with no taxonomic labels and performed the analysis on a final set of 6240 bacterial reference proteomes. Using EtfD Mtb , EtfB Mtb , and EtfA Mtb as query protein sequences, we used the following protein BLAST (BLASTp) parameter values: identity cutoff of >30%, coverage cutoff of >75%, e-value cutoff of 10-3. Visual representations of phylogenies with surrounding color-coded rings were generated using the software tool GraPhLan 65 , with the phylogenetic try built from the taxonomic categorization of the 6240 UniProt bacterial reference proteomes. To evaluate the statistical significance of co-occurrence of EtfD Mtb and EtfBA Mtb in Actinobacteria, we performed a hypergeometric test to evaluate the probability of observing k species with EtfD and EtfBA homologs, given M Actinobacterial species, n species with EtfD homologs, and N species with EtfBA homologs.
Quantification and statistical analysis. Generation of graphics and data analyses were performed in Prism version 9.0 software (GraphPad).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Whole genome sequencing data generated in this study were deposited in NCBI's Sequence Read Archive (SRA) under BioProject PRJNA670664. LC-MS data generated in this study were deposited in the MetaboLights database 66 under accession code MTBLS2374. To identify possible homologues of EtfD, EtfB, and EtfA, we used the subset of Uniprot reference proteomes corresponding to bacteria (https://www.uniprot.org/ proteomes/?query=*&fil=taxonomy%3A%22Bacteria+%5B2%5D%22+AND +reference%3Ayes). Source data for figures are provided as a Source Data file. Source data are provided with this paper.