A genome-wide screen in Escherichia coli reveals that ubiquinone is a key antioxidant for metabolism of long-chain fatty acids

Long-chain fatty acids (LCFAs) are used as a rich source of metabolic energy by several bacteria including important pathogens. Because LCFAs also induce oxidative stress, which may be detrimental to bacterial growth, it is imperative to understand the strategies employed by bacteria to counteract such stresses. Here, we performed a genetic screen in Escherichia coli on the LCFA, oleate, and compared our results with published genome-wide screens of multiple non-fermentable carbon sources. This large-scale analysis revealed that among components of the aerobic electron transport chain (ETC), only genes involved in the biosynthesis of ubiquinone, an electron carrier in the ETC, are highly required for growth in LCFAs when compared with other carbon sources. Using genetic and biochemical approaches, we show that this increased requirement of ubiquinone is to mitigate elevated levels of reactive oxygen species generated by LCFA degradation. Intriguingly, we find that unlike other ETC components whose requirement for growth is inversely correlated with the energy yield of non-fermentable carbon sources, the requirement of ubiquinone correlates with oxidative stress. Our results therefore suggest that a mechanism in addition to the known electron carrier function of ubiquinone is required to explain its antioxidant role in LCFA metabolism. Importantly, among the various oxidative stress combat players in E. coli, ubiquinone acts as the cell's first line of defense against LCFA-induced oxidative stress. Taken together, our results emphasize that ubiquinone is a key antioxidant during LCFA metabolism and therefore provides a rationale for investigating its role in LCFA-utilizing pathogenic bacteria.

Long-chain fatty acids (LCFAs) 7 are carboxylic acids with an unbranched aliphatic chain of 12-20 carbon atoms. Several bacterial pathogens such as Mycobacterium tuberculosis, Pseudomonas aeruginosa, and Salmonella typhimurium metabolize LCFAs derived from host tissues, which enable their survival in harsh environments and contribute to their virulence (1)(2)(3). Much of our understanding of LCFA metabolism is obtained from studies in Escherichia coli. LCFA metabolism is mediated by proteins encoded by the fad (fatty acid degradation) genes that transport LCFAs inside the cell and subsequently degrade it to acetyl-CoA via the ␤-oxidation pathway. Acetyl-CoA feeds into the tricarboxylic acid (TCA) and glyoxylate cycles to generate metabolic intermediates for growth. Reduced cofactors generated during ␤-oxidation and TCA cycle are oxidized by the electron transport chain (ETC) resulting in production of ATP. The fad genes are controlled at a transcriptional level by three regulatory systems: (i) positive regulation by the global cyclic AMP receptor protein-cyclic AMP complex (catabolite repression), (ii) negative regulation by the transcriptional regulator, FadR, repression of which is relieved by binding to acyl-CoA, a metabolic intermediate in LCFA degradation, and (iii) negative regulation by the ArcA-ArcB two-component system (reviewed in Refs. 4 and 5).
Growth on LCFAs requires the presence of either oxygen as terminal electron acceptor (aerobic metabolism) or alternative electron acceptor such as nitrate (anaerobic metabolism). Thus, LCFAs are non-fermentable carbon sources, which unlike glucose, a fermentable carbon source, require optimal functioning of ETC (6,7). Components of ETC are located in the inner membrane of E. coli. During aerobic metabolism NADH and FADH 2 are oxidized by NADH dehydrogenases (Ndh and Nuo) and succinate dehydrogenase (Sdh), respectively, and the electrons are transferred to the lipid-soluble electron carrier, ubiquinone. The reduced form of ubiquinone, ubiquinol, in turn donates electrons to the terminal oxidases, cytochrome bo (Cyo) and cytochrome bd (Cyd), which further transfer electrons to molecular oxygen reducing it to a water molecule. During the flow of electrons through ETC, Nuo, Cyo, and Cyd generate proton motive force that further drives ATP synthesis through ATP synthase (8,9). Consistent with the increased requirement of ETC on non-fermentable carbon sources, in contrast to growth on glucose, mutants of several ETC components exhibit severe growth defects on succinate, acetate, lactate, and malate (9 -12). However, the importance of ETC components for growth on LCFAs has not been examined. Different non-fermentable carbon sources enter into the TCA cycle at varied steps and are theoretically expected to generate different amounts of reduced cofactors, thus even among these carbon sources there could be a difference in the requirement of ETC components. Hence, the requirement of individual components of the ETC in LCFA metabolism should be investigated.
In addition, a recent study has shown that LCFAs induce oxidative stress in E. coli (13), however, the mechanism by which bacteria combat this stress is not known. The enzymatic scavengers (peroxidases, catalases, and superoxide dismutases) are known to be the major oxidative stress combat players in bacteria (14). On the other hand, the role of the ETC component, ubiquinone, as an antioxidant in bacteria is underappreciated. There is only one report in E. coli that suggested ubiquinone to function as an antioxidant based on oxidative stress phenotypes of mutants defective in ubiquinone biosynthesis (15). But, how ubiquinone counteracts reactive oxygen species (ROS), what is the physiological condition under which ubiquinone plays a predominant role, and what is the relative contribution of ubiquinone to the overall oxidative stress response remains to be assessed. The remarkable use of LCFAs by pathogens such as M. tuberculosis, P. aeruginosa, and S. typhimurium (1-3) warrants a detailed investigation of the strategies employed by bacteria to mitigate LCFA-mediated oxidative stress.
With the motive to fill in the above-mentioned knowledge gaps in LCFA metabolism, using the Keio deletion library of E. coli (16), we performed a quantitative genetic screen on oleate (C18:1 cis-9) and compared our LCFA dataset with recent high-throughput genetic screens of additional carbon sources (17). This comparative analysis revealed that among all aerobic ETC components only genes involved in the biosynthesis of ubiquinone are more important for growth in LCFAs when compared with other non-fermentable carbon sources. Our detailed studies show that of the various oxidative stress combat players in E. coli, ubiquinone is the key player that mitigates elevated levels of ROS generated by LCFA degradation, however, its role as an antioxidant cannot be described solely by its known electron carrier function in the ETC.

A chemical-genomic screen reveals the pathways used by E. coli to aerobically metabolize long-chain fatty acids
We tested the ability of 3994 mutants from the Keio deletion library to grow on oleate as the sole carbon source. To measure fitness defects during growth on oleate, we pinned the Keio library onto M9 minimal agar plates containing either oleate or glucose as the sole carbon source. Because oleate was solubilized in the detergent Brij-58, the growth medium containing glucose was also supplemented with Brij-58. We took images of the plates at a single time point and quantified colony size using image analysis software (18). Replicate colony size measurements were reproducible in the screen (Fig. 1A, R ϭ 0.86). We then assigned a fitness score to mutants in the oleate condition by calculating the statistical significance of the difference in colony size between oleate and the glucose control using an established analysis pipeline (19) with slight modifications ("Experimental procedures"). To facilitate the comparison of oleate to other carbon sources, we reanalyzed data from a previous screen (17), comparing both fermentable (glucosamine, N-acetylglucosamine, and maltose) and non-fermentable (acetate, succinate, and glycerol) carbon sources to glucose as a control. The fitness scores we report thus represent the statistical significance of a change in colony size on a particular carbon source as compared with growth on glucose with positive and negative fitness scores representing increased and decreased colony size, respectively. A full list of fitness scores of Keio library strains in different carbon sources (normalized to a glucose control) is available in supplemental Dataset S1A.
The screen identified a large number of mutants with severe growth defects on oleate. To better understand the physiological basis for these growth defects, we performed a global analysis using gene set enrichment analysis (GSEA) and gold-standard biological pathways (20 -22), to find pathways that play a significant role in growth on oleate (supplemental Dataset S1B). Importantly, global analysis highlighted the ␤-oxidation pathway (FDR q value: 1.1%), critical for the utilization of LCFAs as an energy source, as a unique signature of growth on oleate. This result emphasizes the robustness of our high-throughput genetic screen and our statistical approach. In addition, oleate, acetate, and succinate shared a significant enrichment in the TCA and glyoxylate cycles (FDR q value Ͻ5%), a result consistent with these pathways being critical for generation of reduced cofactors and metabolites for growth on non-fermentable carbon sources (4,5). Furthermore, oleate, acetate, and succinate utilization exhibited enrichment in multiple pathways for electron transfer activity (FDR q value Ͻ5%), underscoring the importance of ETC for energy generation during growth on non-fermentable carbon sources. We also performed GSEA analysis for protein complexes with 3 or more proteins (20,22). Oleate, acetate, and succinate utilization were enriched for ETC complexes, NADH dehydrogenase I (Nuo) (FDR q value Ͻ1%) and succinate dehydrogenase (FDR q value Ͻ7%) (supplemental Dataset S1C). The enriched metabolic pathways and complexes for oleate utilization are depicted in Fig. 1B.
Although glycerol is considered to be a non-fermentable carbon source, this substrate did not show a significant enrichment of any of the ETC pathways or complexes in our GSEA analysis. We speculate that because there is ATP generation (by substrate-level phosphorylation) after the entry of glycerol in the later part of glycolysis (23), this ATP might allow for growth on glycerol. Moreover, the fermentative utilization of glycerol has also been suggested (24). Thus, for detailed studies on the requirement of ETC components we further focused on oleate, acetate, and succinate. To circumvent problems of highthroughput screens (incorrect strains, suppressors, and cross- p-and q-represent nominal p value and FDR q value, respectively, obtained from GSEA analysis of pathways and complexes in E. coli. *FadE is a flavoprotein that reduces FAD to FADH 2 during ␤-oxidation. It has been speculated that FadE itself might oxidize FADH 2 to FAD by transferring electrons from its dehydrogenase domain to the ETC (61). However, there is no experimental evidence for the same. C, fitness scores were calculated for oleate and other carbon sources as compared with glucose control. Fitness scores of components of the ␤-oxidation pathway and ETC are shown.

Role of ubiquinone in LCFA metabolism
contamination), before a detailed follow-up analysis of ETC mutants, we verified phenotypes using transductants. Furthermore, because oxygen availability to cells in a colony varies with the size of colony (25), in our subsequent experiments we also validated the requirement of ETC components in aerobic metabolism by assessing their phenotypes in liquid medium.

Requirement of ubiquinone does not correlate with the energy yield of non-fermentable carbon sources
A comparison of our LCFA dataset with GSEA analysis for additional carbon sources clearly showed that ETC is critical for growth of E. coli on non-fermentable carbon sources. However, there was a difference in the requirement of various ETC components. Of the two NADH dehydrogenases involved in ETC, Ndh is known to be more important during aerobic respiration, whereas Nuo is essential for dimethyl sulfoxide and fumarate (anaerobic) respiration (26,27). However, we noted that for aerobic metabolism of oleate, acetate, and succinate, Nuo was important, whereas Ndh was dispensable (Fig. 1C). Early studies had reported the growth defect of nuo mutants in acetate, however, they did not assess the relative requirement of Ndh and Nuo (10,28). We validated our results by comparing the growth profile of ⌬nuo (the entire nuo operon is deleted) and ⌬ndh in liquid medium; only ⌬nuo showed significant growth defect in oleate, acetate, and succinate (Fig. 2). We suggest that because Nuo couples NADH oxidation to proton translocation (H ϩ /e Ϫ ϭ 2), whereas Ndh does not (H ϩ /e Ϫ ϭ 0) (8,26), it is likely that growth on non-fermentable carbon sources is dependent on Nuo for ATP production. Furthermore, among oleate, acetate, and succinate, Nuo was more significantly required for growth in acetate (Figs. 1C and 2) that has the worst net ATP yield of these three carbon sources (4). Although ⌬nuo did not grow in acetate, the strain exhibited an extended lag in oleate and succinate, and then grew at the same rate and to the same yield as wild-type (WT) (Fig. 2). To rule out the possibility that the ⌬nuo strain was picking up suppressors in oleate and succinate, we subcultured cells from stationary phase. We again observed extended lag in succinate and oleate. Similar to Nuo, Cyo, which is the major terminal oxidase during aerobic metabolism and also generates proton motive force in the ETC (H ϩ / e Ϫ ϭ 2) (8), was maximally required in acetate (Fig. 1C). Collectively, our results suggest that the quantitative contribution of Nuo and Cyo to growth on non-fermentable carbon sources is inversely correlated with the energy yield of these substrates.
In contrast to Nuo and Cyo, we observed that the requirement of ubiquinone was more in oleate compared with other non-fermentable carbon sources (Fig. 1C). In E. coli, the biosynthesis of ubiquinone is known to involve at least 12 ubi genes (9, 29) (Fig. 1B). The data for only seven ubi deletion strains were available from previous genetic screens for comparison with our oleate dataset (17) (Fig. 1C). In independent studies the growth phenotype of null mutants of five ubi genes (ubiE, ubiF, ubiH, ubiI, and ubiX) have been reported in succinate (30 -33). In our analysis, only the phenotype of ubiF, ubiH, and ubiI mutants agreed with their known phenotypes in succinate (30,31) (Fig. 1C). We thus constructed several transductants of all five ubi mutants and assessed their growth at a candidate level (supplemental Fig. S1). The phenotype of all ubi mutants corroborated with their known phenotype in succinate: ubiE, Figure 2. Among NADH dehydrogenases, only Nuo complex is required for growth on non-fermentable carbon sources. WT, ⌬nuo, and ⌬ndh strains were grown in minimal medium containing one of the carbon sources, and OD 450 was measured. Each medium condition had Brij-58. The experiment was done 4 times; each experiment had 2 technical replicates. A representative dataset with average and S.D. of technical replicates is shown.

Role of ubiquinone in LCFA metabolism
ubiF, and ubiH mutants exhibited no growth (30,32), ubiX mutant showed a growth defect (33), and the ubiI mutant displayed wild-type growth (30,31). In acetate, all ubi mutants had a phenotype similar to that in succinate. In oleate, ubi mutants either did not grow at all (ubiE, ubiF, and ubiH) or showed a significant growth defect (ubiI and ubiX) (supplemental Fig.  S1). Upon relating the growth profile of ubi deletion strains observed in our study with their ubiquinone levels reported in the literature we find that mutants with no detectable ubiquinone (ubiE, ubiF, and ubiH) (30,32) do not grow in any of the three non-fermentable carbon sources, whereas a ubiI mutant, which produces reduced levels of ubiquinone (30,31), exhibits growth defect only in oleate. UbiI catalyzes C5-hydroxylation of the 3-octaprenylphenol intermediate in the ubiquinone biosynthesis pathway and the residual level of ubiquinone in a ⌬ubiI strain is attributed to the suboptimal C5-hydroxylase activity of a C6-monooxygenase, UbiF (31).
For the last several decades, the increased requirement of ubiquinone for energy generation during growth with succinate compared with glucose has been the rationale for identifying genes involved in ubiquinone biosynthesis (34,35). Despite this, it was surprising that a recent study showed that ⌬ubiI, where the ubiquinone level is reduced to only ϳ10 -15% of WT level, exhibited normal growth in succinate (30,31). Data from genetic screens and candidate studies in solid media reinforce that even ϳ10 -15% ubiquinone is sufficient for growth in succinate, whereas, this ubiquinone level is suboptimal for growth in LCFAs.
We investigated further the differential requirement of ubiquinone on non-fermentable carbon sources. We compared the growth profile of ⌬ubiH and ⌬ubiI strains in liquid media. Whereas ⌬ubiH showed a growth reduction on glucose, the strain did not grow at all with oleate, acetate, and succinate ( Fig.  3A). On the other hand, ⌬ubiI showed a significant growth defect only in oleate (Fig. 3A); the growth defect was complemented by ubiI cloned on plasmid (supplemental Fig. S2). Taken together, our analysis of the growth phenotypes of ETC mutants on various carbon sources highlights a specific role of ubiquinone in LCFA metabolism.

Ubiquinone is maximally required in oleate to mitigate elevated levels of ROS
E. coli grown in oleate is reported to accumulate higher levels of ROS compared with cultures grown in glucose (13). In a separate study, ubiquinone was suggested to function as an antioxidant in E. coli (15). A ubiCA knock-out, which produces no detectable ubiquinone (36), was shown to exhibit several oxidative stress phenotypes in LB: accumulation of superoxide and H 2 O 2 in membranes, hypersensitivity to oxidative stress inducing agents, and up-regulation of catalase (15). In the present work, whereas ubi deletion strains that produce no detectable ubiquinone did not grow at all in oleate, acetate, and succinate, ubiI mutant that produces decreased levels of ubiquinone showed significant growth defect only in oleate (Figs. 1C and 3A, and supplemental Fig. S1). Collectively, these observations led us to argue that of the above three non-fermentable carbon sources, E. coli generates the highest ROS lev-els when cultured in oleate and the increased requirement of ubiquinone in oleate is to counteract elevated levels of ROS.
Consistent with the above proposal, glutathione and thiourea, widely used chemical antioxidants in E. coli (37)(38)(39), partially recovered the growth defect of ⌬ubiI in oleate (Fig. 3, B and C). We did not observe a complete recovery because another factor responsible for the poor growth of ⌬ubiI in oleate would be reduced energy generation due to lowering of ETC function. Interestingly, the WT strain also grew better in the presence of glutathione and thiourea consistent with a previous report that cells suffer oxidative stress during growth in oleate (13) (Fig. 3, B and C). Furthermore, we tested whether E. coli indeed generates the highest ROS levels when cultured in oleate in comparison to acetate and succinate. For this, we measured intracellular ROS levels in WT and ⌬ubiI cultured in Tryptone broth (TB) supplemented with various carbon sources by a colorimetric assay based on the reduction of nitro blue tetrazolium (NBT) dye by superoxide. WT had the highest ROS levels in TB-Ole; ϳ1.5-fold higher than basal medium (Fig.  3D). This fold-change was comparable with a previous report where ϳ2-fold higher ROS levels were observed in WT cells grown in minimal medium supplemented with oleate compared with cells grown in glucose-supplemented medium (13). Moreover, ROS levels further increased in ⌬ubiI in all media conditions with maximal ROS levels again in TB-Ole-grown cells (Fig. 3D). Brij-58 alone did not result in increased ROS. Importantly, when ubiquinone-8 was supplied exogenously, ROS levels decreased by ϳ15-25% in WT grown in TB-Ole and ⌬ubiI grown either in TB or TB-Ole (Fig. 3E), reiterating that ubiquinone relieves oxidative stress.
NBT enters inside the cells and has been extensively used in E. coli and other Gram-negative bacteria to measure intracellular superoxide (40 -43). We independently validated that the reduction of NBT reports on intracellular superoxide levels by showing that overexpressing superoxide dismutase, SodA, from plasmid decreases the NBT signal in all tested media conditions (TB, TB-Brij, and TB-Ole; supplemental Fig. S3, A and B). We further confirmed our results from the NBT assay by using a fluorescent dye, dihydroethidium, which also detects superoxide (supplemental Fig. S3C) (44). Experiments involving measurement of ROS levels were carried out in TB to support the growth of ubiI knock-out, which shows growth defects in minimal medium containing oleate as the sole carbon source (supplemental Fig. S1 and Fig. 3A). Furthermore, because TB causes mild catabolite repression (45), supplemented carbon sources are expected to be coutilized with carbon components of TB. As a representative, the increase in biomass and transcript levels of fadL (outer membrane transporter for LCFAs) and fadE (acyl-CoA dehydrogenase involved in ␤-oxidation of LCFAs) in WT grown in TB-Ole confirmed the co-utilization of oleate with TB (46, 47) (supplemental Fig. S4, A and B). Because the maximum difference in ROS levels was observed in the stationary phase (time point T4 in supplemental Fig. S4, A and C), for single time point experiments cultures were sampled in this growth phase.

Ubiquinone is a major antioxidant during oleate metabolism
We tested whether in addition to ubiquinone other oxidative stress response players counteract ROS generated by oleate. In TB medium, ROS levels were ϳ1.3 to 1.7-fold higher when strains lacked either the ubi genes or other players ((alkyl hydroperoxide reductase subunit, AhpC; oxidative stress regulator, SoxR; superoxide dismutase, SodA; catalase, KatE; and an enzyme involved in glutathione biosynthesis, GshB) (14, 15)) ( Fig. 4A). In contrast, in TB-Ole medium, ROS levels increased only in ubi deletion strains (Fig. 4A). We considered two possibilities for the TB-Ole results: (i) there is redundancy of enzymatic scavengers and their regulators, and (ii) as long as ubiquinone is present, it does not allow ROS to build-up further in TB-Ole thereby reducing dependence on other players. Con- Figure 3. The increased requirement of ubiquinone for growth in oleate is to mitigate elevated levels of ROS. A, ⌬ubiI shows significant growth defect in liquid medium containing oleate as the sole carbon source. WT, ⌬ubiH, and ⌬ubiI strains were grown in minimal medium containing one of the carbon sources, and OD 450 was measured. Each medium condition had Brij-58. B and C, the growth defect of ⌬ubiI in oleate is partially recovered by glutathione (B) and thiourea (C). WT and ⌬ubiI were grown in minimal medium containing oleate with or without 1 mM glutathione (GSH) or 1 mM thiourea, and OD 450 was measured. 1 mM urea was included as control for thiourea.

Role of ubiquinone in LCFA metabolism
sistent with the second possibility, the enzymatic scavengers, katG and ahpC were induced (ϳ2-fold) by oleate only in a ⌬ubiI strain (Fig. 4, B and C), and this increased expression was reduced by ϳ25% upon exogenous supplementation of ubiquinone-8 (Fig. 4D).
Our above data suggests that ubiquinone is a major player that mitigates LCFA-mediated oxidative stress. We examined whether this defense system is induced by LCFAs. E. coli ubiquinone is designated as Q 8 and exists in two redox states in the cell, ubiquinone and ubiquinol (9). We extracted lipids from WT and several fad knockouts, separated lipids by HPLC, and measured total Q 8 content (ubiquinone and ubiquinol). Peaks for ubiquinone and ubiquinol in the samples were assigned based on the elution time of pure standards, and the reduction

Role of ubiquinone in LCFA metabolism
of ubiquinone and ubiquinol peaks in ⌬ubiI (supplemental Fig.  S5). In WT cells, Q 8 content was ϳ1.8-fold higher in TB-Ole compared with TB (Fig. 4E). Q 8 levels did not increase in cells grown in TB-Brij. Importantly, increase in the Q 8 levels was dependent on oleate utilization because Q 8 levels did not increase in TB-Ole in fad mutants defective in LCFA transport and ␤-oxidation (⌬fadL, ⌬fadD, and ⌬fadE) (Fig. 4E). Interestingly, the increase in ROS levels in TB-Ole was also due to oleate utilization; compared with TB, the ROS level did not exhibit a considerable increase in TB-Ole in fad mutants (Fig.  4F). The ROS level in TB was higher in the ⌬fadD strain possibly due to accumulation of endogenous free fatty acids (48,49).

⌬ubiI⌬ubiK double mutant produces no detectable ubiquinone
While this manuscript was under preparation, an uncharacterized gene, yqiC (renamed as ubiK), was identified as a ubiquinone biosynthesis player based on the reduction of ubiquinone levels to ϳ20% of WT in the ⌬ubiK strain (29), however, growth of the ⌬ubiK strain on non-fermentable carbon sources has not been evaluated. Importantly, in our comparative analysis, we observed that similar to ⌬ubiI, ⌬ubiK showed a statistically significant fitness defect only in oleate (supplemental Dataset S1, FDR Ͻ5%; Fig. 1C).
To further strengthen our proposal that reduced levels of ubiquinone are suboptimal for growth in oleate thus leading to oxidative stress, we independently compared ubiquinone levels in ⌬ubiI and ⌬ubiK, validated the growth defect of ⌬ubiK in candidate studies, and determined ROS levels in ⌬ubiK. Q 8 levels were reduced to ϳ15-20% in both ⌬ubiI and ⌬ubiK, and a peak corresponding to an intermediate (elution time: ϳ15 min) was observed in both strains ( Fig. 5A and supplemental  Fig. S5). Except oleate, ⌬ubiK did not show growth defect on any other non-fermentable carbon source (Fig. 5B). The growth defect on oleate was complemented by ubiK cloned on plasmid (Fig. 5C), concomitant with restoration of ubiquinone levels (supplemental Fig. S6). Furthermore, similar to ubi deletion strains, compared with WT in TB, ROS levels were ϳ2.5-fold higher in ⌬ubiK in TB-Ole (Fig. 5D).
The related phenotypes of ⌬ubiI and ⌬ubiK prompted us to examine the phenotype of the ⌬ubiI⌬ubiK double mutant. In contrast to the normal growth of ⌬ubiI and ⌬ubiK in LB medium, the double mutant formed tiny colonies on LB (Fig.  5E). To determine ubiquinone levels in the ⌬ubiI⌬ubiK strain, we resorted to growing the cultures in a rich LB-glucose medium, where there is reduced dependence on ubiquinone for growth (50). Interestingly, there was no detectable ubiquinone in the ⌬ubiI⌬ubiK double mutant (Fig. 5F and supplemental Fig. S7), suggesting redundancy in the ubiquinone biosynthesis pathway. Importantly, similar to single ubi deletion strains (ubiE, ubiF, and ubiH) that produce no detectable ubiquinone, the ⌬ubiI⌬ubiK double mutant showed growth defect in glucose but did not grow at all with oleate, acetate, and succinate (Fig. 5G). Taken together, our data convincingly establishes that there is a differential requirement of ubiquinone among non-fermentable carbon sources, requirement being maximal in oleate to alleviate oxidative stress.

Quantitative contribution of Nuo and Cyo to growth is inversely correlated with the energy yield of non-fermentable carbon sources
The important functions of ETC are ATP production and maintenance of redox balance. We find that Nuo is the major NADH dehydrogenase for aerobic growth on oleate, acetate, and succinate (Datasets S1, A and C; Figs. 1C and 2). Because nuo mutants would be unable to oxidize NADH efficiently this would increase NADH/NAD ϩ ratios, which might further inhibit TCA and glyoxylate cycle enzymes resulting in a decrease in cellular metabolites (28). Besides, because Nuo is a proton pump (H ϩ /e Ϫ ϭ 2) (8), its deletion would lead to a decrease in ATP synthesis in the cell. Therefore, in nuo mutants grown on non-fermentable carbon sources ATP would be produced mainly from FADH 2 oxidation.
The ⌬nuo strain exhibited no growth in acetate (Fig. 2). We speculate that in acetate, in addition to the increased NADH/ NAD ϩ ratio and decrease in cellular metabolites, the nuo mutant cannot produce energy to support growth because only 1 molecule of ATP will be generated per acetate molecule considering the ATP:FADH 2 ratio to be 1 (8). However, 1 molecule of ATP is also expended to activate acetate to acetyl-CoA (4), thus in a nuo mutant there will be no net ATP gain. On the other hand the nuo mutant exhibited an extended lag in oleate and succinate (Fig. 2). We suggest that because nuo mutant can generate energy from oxidation of FADH 2 produced during ␤-oxidation of oleate and in the first step of succinate utilization (4,51), this allows cells to re-adjust its metabolism to maintain redox balance resulting in the same growth rate and growth yield as WT cells. Furthermore, unlike acetate, activation of succinate is not required for its metabolism and although conversion of oleate to oleoyl-CoA is required for oleate metabolism, only 1 molecule of ATP is consumed per 18 carbon atoms (4,51). Reduced energy generation might also be the reason for the significant growth defect of ⌬cyo strains in acetate (Fig. 1C). Because Cyo, the major terminal oxidase during aerobic metabolism generates proton motive force (H ϩ /e Ϫ ϭ 2) (8), disruption of the Cyo complex would be expected to result in a more compromised growth on carbon sources such as acetate that have poor net ATP yield.

Ubiquinone relieves oxidative stress generated by LCFAs
It was surprising that unlike other ETC components, requirement of ubiquinone, the electron carrier of ETC, did not correlate with the energy yield of non-fermentable carbon sources. Our detailed analysis shows that ubiquinone was maximally required in oleate to mitigate elevated levels of ROS. Compared with WT cells, ROS levels were ϳ1.5-fold higher in a ⌬ubiI strain grown in glucose, acetate, or succinate but there was no difference in the growth profile of WT and ⌬ubiI in these carbon sources (Fig. 3, A and D). The utilization of oleate resulted in a ϳ1.5-fold increase in ROS levels in WT cells compared with other carbon sources that further increased to ϳ2.5-fold in a ⌬ubiI strain (Fig. 3D). Importantly, this elevated level of ROS (ϳ2.5-fold) was deleterious as evident from the significant growth defect of the ⌬ubiI strain in oleate and that the growth

Role of ubiquinone in LCFA metabolism
defect could be partially recovered by chemical antioxidants (Fig. 3). These data clearly indicate that in oleate-utilizing cells optimum levels of ubiquinone are required to manage ROS below a toxic threshold. Furthermore, we find that among various oxidative stress combat players, ubiquinone is the key antioxidant during LCFA metabolism. This is supported by the observation that strains deleted for oxidative stress combat players other than ubi genes do not exhibit an increase in ROS in oleate-utilizing cells (Fig. 4A). Moreover, whereas ubiquinone accumulates in the presence of oleate, other players are induced only in a mutant defective in ubiquinone biosynthesis (Fig. 4, B, C, and E). Interestingly, oleate degradation generates ROS and also provides a signal for ubiquinone accumulation (Fig. 4, E and F). These results suggest a feedback loop that prevents excessive ROS formation during growth in LCFAs.
An earlier study has shown that ubiquinone is present in excess over flavins and cytochromes in the E. coli inner membrane (52). Thus under normal conditions ubiquinone is not limiting for its electron transfer function. Considering this, ϳ2-fold increase in ubiquinone levels in cells utilizing oleate could bring a significant physiological response. Few ubi genes are regulated by the ArcA-ArcB two-component system and catabolite repression (53)(54)(55)(56)(57). It will be interesting to investigate the mechanisms that regulate ubiquinone levels during LCFA degradation. We suggest that ROS itself might not be the signal for up-regulation of ubiquinone, because despite exhibiting a higher level of ROS the ⌬fadD experiments. B, ⌬ubiK shows significant growth defect in oleate. Dilutions of the cultures were spotted on minimal medium containing one of the carbon sources. Each medium condition had Brij-58. ⌬fadL was used as a control. The experiment was repeated 3 times. A representative dataset is shown. C, ubiK cloned on plasmid complements the growth defect of ⌬ubiK in oleate. Dilutions of WT and ⌬ubiK carrying either empty plasmid (pACYC184) or pACYC184 with ubiK (pSA4) were spotted on minimal medium containing oleate as the sole carbon source. ⌬fadL transformed with pACYC184 was used as a control. The experiment was repeated 2 times. A representative dataset is shown. D, ⌬ubiK strain has increased ROS levels. WT and ⌬ubiK were grown either in TB or TB-Ole, and ROS levels were determined by NBT assay. Data were normalized to the ROS level of WT in TB and represent average (Ϯ S.D.) of 3 independent experiments. E, ⌬ubiI⌬ubiK shows a synthetically sick phenotype in LB. WT, ⌬ubiI, ⌬ubiK, and ⌬ubiI⌬ubiK strains were streaked on LB and incubated overnight. F, ubiquinone is not detected in the ⌬ubiI⌬ubiK double mutant. Total Q 8 level in lipid extracts from WT, ⌬ubiI, ⌬ubiK, and ⌬ubiI⌬ubiK cells grown in LB-glucose was determined. Q 8 levels were normalized to the Q 8 level of WT in LB-glucose and represent the average (Ϯ S.D.) of 3 independent experiments. G, ⌬ubiI⌬ubiK shows a synthetic lethal phenotype in non-fermentable carbon sources. Dilutions of the cultures were spotted on minimal medium containing one of the carbon sources. Each medium condition had Brij-58. ⌬fadL was used as a control. The experiment was repeated 2 times. A representative dataset is shown.

Role of ubiquinone in LCFA metabolism
strain had basal ubiquinone levels in TB medium (compare Fig. 4, E and F).
Succinate has been extensively used for screening ubi genes (34,35). We suggest that oleate is a better carbon source than succinate for identifying ubi players especially ones that have a partial effect on ubiquinone levels. The unique requirement of ubiI and ubiK for growth in oleate validates this suggestion (Figs. 1C, 3A, and 5B, and supplemental Fig. S1). UbiK is a small protein (Ͻ100 amino acids) that belongs to the BMFP (Brucella membrane fusogenic protein) superfamily. In S. typhimurium, it is involved in the regulation of flagella and fimbriae expression, motility, colonization, and virulence of the organism (29,58,59). E. coli UbiK physically interacts with a non-enzymatic ubiquinone biosynthesis player, UbiJ, and is proposed to be an assembly factor for additional Ubi proteins (29). Our results that a double mutant of ubiK and an enzyme involved in ubiquinone biosynthesis, ubiI, shows synthetic sick/lethal phenotype with no detectable ubiquinone provides strong genetic evidence of the interaction between UbiK and other Ubi proteins (Fig. 5, E-G, and supplemental Fig. S7).

Proposed mechanisms by which ubiquinone might counteract LCFA-mediated oxidative stress
Several mechanisms have been suggested for ROS formation that includes extraction of electrons from reduced metal centers of certain enzymes by molecular O 2 , leakage of electrons during oxidation-reduction cycles of ETC promoting the reaction of free electrons with O 2 , and autoxidation of flavoproteins (14,15,60). Søballe and Poole (15) first demonstrated the role of ubiquinone in counteracting ROS in bacteria and proposed two mechanisms to explain its antioxidant function. First, ubiquinone limits ROS formation due to its ability to rapidly transfer electrons from upstream respiratory dehydrogenases to terminal oxidases thereby decreasing the chance of single-electron donation to oxygen. Second, the reduced form of ubiquinone (ubiquinol) can scavenge ROS (15). Fig. 6 shows the probable sites of ROS formation during growth of E. coli in LCFAs and the mechanisms by which ubiquinone might counteract LCFA-induced oxidative stress. Regarding the site of ROS formation, we speculate that during LCFA catabolism, high NADH/NAD ϩ and FADH 2 /FAD ratios would increase the electron flow in the ETC thereby increasing   (1)(2)(3). It will be interesting to examine whether ubiquinone participates in managing LCFAmediated oxidative stress in these pathogens. In fact ubiB, ubiE, ubiJ, and ubiK mutants of S. typhimurium are impaired for intracellular proliferation in macrophages (29,63). Because S. typhimurium utilizes fatty acids in macrophages during chronic infection (2), it is possible that ubiquinone is required by this intracellular pathogen to combat oxidative stress generated by LCFAs.

Strains and plasmids
Chemical genomics screen used the Keio deletion library derived from BW25113 (16). The majority of the follow-up experiments were conducted in BW25113 background, and either both independent clones from the library and/or fresh transductants were analyzed to rule out genetic errors. Strains, plasmids, and primers used for plasmid construction are listed in supplemental Table S1.

Media composition and growth conditions
Media had the following composition: M9 minimal medium was 5.3 g/liter of Na 2 HPO 4 , 3 g/liter of KH 2 PO 4 , 0.5 g/liter of NaCl, 1 g/liter of NH 4 Cl, 1.2 g/liter of MgSO 4 , 2 mg/liter of biotin, 2 mg/liter of nicotinamide, 0.2 mg/liter of riboflavin, and 2 mg/liter of thiamine; Lysogeny broth (LB) was 5 g/liter of yeast extract, 10 g/liter of Bacto-Tryptone, and 5 g/liter of NaCl; and Tryptone broth (TB) was 10 g/liter of Bacto-Tryptone, and 5 g/liter of NaCl. Unless otherwise specified, when required, media were supplemented with one of the following carbon sources at a final concentration of 5 mM: glucose or sodium salt of acetate, succinate, or oleate. Stock of oleate (50 mM) was prepared in 5.0% Brij-58 (64). Media were solidified using 1.5% (w/v) bacto agar. For chemical genomics screen, minimal medium was supplemented either with 5 mM oleate or 0.2% glucose with 0.5% Brij-58. To facilitate comparison of our screen on oleate with published screens on other carbon sources (17), the same concentration of glucose was used in our control condition.
Cultures were incubated at 37°C. For experiments in liquid medium, unless indicated otherwise, primary cultures were grown in 3 ml of TB, which were further re-inoculated either in TB or TB supplemented with desired carbon source to an initial OD 600 of ϳ0.01. These secondary cultures were grown for defined time periods. For the detection of Q 8 in the ⌬ubiI⌬ubiK double mutant, strains were grown in LB supplemented with 0.2% glucose.

Library screening and data processing
The chemical genomics screen was performed using the same methodology as reported previously with slight modifications (17). Briefly, the Keio deletion library was arrayed in 1536format and pinned onto plates containing minimal medium agar supplemented either with oleate or glucose with Brij-58, using a Singer Rotor robot. Plates were incubated at 37°C for 21 h for glucose with Brij-58 and 42 h for oleate. Time points were chosen such that fitness differences were apparent but growth had not saturated.
Pictures of the plates were taken using a Canon G10 digital camera. Colony size was quantified from plate images using the HT Colony Grid Analyzer software package (18). Colony sizes were filtered and normalized using established methods for chemical genomics in E. coli K-12 (19). To account for potential effects of Brij-58 on growth, fitness scores for the oleate condition were generated by directly comparing colony size between oleate and glucose with Brij-58 control using the same statistical test as the S-score (18).
To facilitate comparison of the chemical genomic profile of oleate to other carbon sources, raw data from minimal media conditions of a previous large-scale chemical genomic screen (17) were reanalyzed using the same workflow above and M9 minimal with 0.2% glucose as a normalization. Finally, fitness score distributions for each condition were scaled to have a standard inter-quartile range of 1.35 (17) to minimize bias in any downstream analysis from conditions with a large variance. The MATLAB code and raw data for this analysis are available online (https://github.com/AnthonyShiverMicrobes/fitness_lcfa). 8

Growth curves
Overnight cultures grown in LB were pelleted, washed, and re-suspended in M9 minimal medium. Cells were re-inoculated in 200 l of M9 minimal medium containing the desired carbon 8 Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site.

Role of ubiquinone in LCFA metabolism
source to a starting OD 450 of ϳ0.03, in 96-well plates, using a robotic liquid handling system (Tecan). When required, antibiotic, glutathione, thiourea, or urea at a desired concentration was added to the medium. Plates were incubated in a shaker at 37°C, and OD 450 of the cultures was measured at the designated time intervals (Tecan Infinite M200 monochromator). The incubator shaker and microplate reader were integrated with the liquid handling system, and the transfer of plates between shaker and reader was automated.

Dilution spotting
Overnight cultures grown in LB were pelleted, washed, and re-suspended in M9 minimal medium. Several dilutions of cultures were spotted on M9 minimal medium containing the desired carbon source. Antibiotic was added whenever required. Plates were incubated and imaged at various time intervals using the Gel Doc XR ϩ imaging system from Bio-Rad. A representative image where growth differences were apparent is shown in the figures.

NBT reduction assay
Secondary cultures (3 ml) were grown in culture tubes (27 ml capacity) for 16 h. ROS levels were determined by a NBT reduction assay following the protocol described in Ref. 42 with slight modifications. Cells were pelleted and washed with Hanks' balanced salt solution (HBSS). Washing was required to remove growth medium to avoid interference of Brij-58. 10 9 cells were re-suspended in 0.2 ml of HBSS and split in two aliquots: 0.5 ml of NBT (1 mg/ml) was added to one aliquot and the other aliquot was left untreated. Both aliquots were incubated at 37°C for 30 min. 0.1 ml of 0.1 M HCl was added and samples were centrifuged at 18,400 ϫ g for 15 min. Supernatant was discarded, and pellet was treated with 0.4 ml of dimethyl sulfoxide (DMSO) to dissolve reduced NBT (formazan blue), followed by addition of 0.8 ml of HBSS. Formazan blue was quantified at 575 nm. To determine the absorbance corresponding to formazan blue, absorbance of the aliquot without NBT was deducted from absorbance obtained for the NBT-treated sample. For experiments where ROS levels were determined in different phases of growth, 15 ml of secondary cultures were grown in 125-ml flasks. Independent cultures (from the same primary culture) were set-up for each time point.

␤-Galactosidase assay
␤-Gal assays were performed in MC4100 background. Secondary cultures (15 ml) were grown in 125-ml flasks for 14 -16 h. Cells were pelleted, washed at least 4 times with Z-buffer, and diluted to OD 450 ϳ0.5. Several washings of the culture were critical to remove Brij-58 because the detergent interferes with ␤-gal assay. katG and ahpC promoter activity was measured by monitoring ␤-gal expression from single-copy katG-lacZ and ahpC-lacZ transcriptional fusion, respectively as described (65).

Preparation of ubiquinol standard
Ubiquinol-8 was prepared by reduction of ubiquinone-8 (Avantis Polar Lipids) following the procedure used for reduction of ubiquinone-10 (66). Briefly, 19 ml of hexane was added to 1 ml of ubiquinone solution (1 mg/ml in hexane). 1 ml of methanol and 200 mg of sodium borohydride were added (solution was covered to avoid light), stirred, and kept for 5 min. 2 ml of water was added and mixed thoroughly to dissolve sodium borohydride. The mixture was centrifuged at 2050 ϫ g for 5 min. Colorless organic supernatant was transferred to a fresh tube and stored at Ϫ20°C. The conversion of yellow colored ubiquinone-8 to colorless ubiquinol-8 indicated complete reduction.

Quinone extraction and detection by HPLC-photodiode array detection analysis
Quinones were extracted using the protocol described in Ref. 31 with slight modifications. 15-ml secondary cultures were grown in 125-ml flasks for 16 h. Equal numbers of cells (ϳ3 ϫ 10 10 cells) for all cultures were pelleted and the pellet mass was determined. Pellets were re-suspended in 100 l of 0.15 M KCl and to the re-suspension 200 l of glass beads (acid washed Յ 106 m, Sigma), 600 l of methanol, and 12 g of ubiquinone-10 standard (Sigma) in hexane (used as internal control for extraction efficiency) were added. Samples were vortexed for 15 min, and 400 l of hexane was added. Samples were again vortexed for 3 min and then centrifuged at 3380 ϫ g for 1 min. The upper hexane layer was transferred to a fresh microcentrifuge tube. 100 l of this hexane layer was completely dried under vacuum and re-suspended in 100 l of mobile phase, an isocratic solution constituted by using 40% ethanol, 40% acetonitrile, and a 20% mixture of 90% isopropyl alcohol and 10% lithium perchlorate (1 M). Lipid extracts were separated by reverse-phase HPLC with a C18 column (Waters Sunfire 5 m column, 4.6 ϫ 250 mm) at a flow rate of 1 ml/min using mobile phase. Quinones were detected using a Photodiode array detector. To detect total Q 8 (ubiquinone-8 ϩ ubiquinol-8) in samples, max of ubiquinone-8 and ubiquinol-8 standards was determined by scanning from 240 to 399 nm. max of ubiquinone-8 was 275 nm, and that of ubiquinol-8 was 290 nm. Peaks for ubiquinone-8 and ubiquinol-8 in the effluent were assigned based on the elution time of pure standards, and reduction of ubiquinone-8 and ubiquinol-8 peaks in the ⌬ubiI strain. For each sample, the Q 8 peak area per unit mass was calculated, and to account for the difference in extraction efficiency between samples, the Q 8 peak area per unit mass was divided by ubiquinone-10 peak area.
Author contributions-R. C., S. A., and K. J. designed the study. R. C. performed the high-throughput genetic screen. A. L. S. performed analysis of the high-throughput data. S. A., K. J., and H. B. performed follow-up experiments. T. P. standardized initial experiments. R. C., K. J., S. A., A. L. S., and H. B. wrote the manuscript. All authors reviewed and edited the manuscript. R. C. supervised the project.