Role of Glyoxylate Shunt in Oxidative Stress Response*

The glyoxylate shunt (GS) is a two-step metabolic pathway (isocitrate lyase, aceA; and malate synthase, glcB) that serves as an alternative to the tricarboxylic acid cycle. The GS bypasses the carbon dioxide-producing steps of the tricarboxylic acid cycle and is essential for acetate and fatty acid metabolism in bacteria. GS can be up-regulated under conditions of oxidative stress, antibiotic stress, and host infection, which implies that it plays important but poorly explored roles in stress defense and pathogenesis. In many bacterial species, including Pseudomonas aeruginosa, aceA and glcB are not in an operon, unlike in Escherichia coli. In P. aeruginosa, we explored relationships between GS genes and growth, transcription profiles, and biofilm formation. Contrary to our expectations, deletion of aceA in P. aeruginosa improved cell growth under conditions of oxidative and antibiotic stress. Transcriptome data suggested that aceA mutants underwent a metabolic shift toward aerobic denitrification; this was supported by additional evidence, including up-regulation of denitrification-related genes, decreased oxygen consumption without lowering ATP yield, increased production of denitrification intermediates (NO and N2O), and increased cyanide resistance. The aceA mutants also produced a thicker exopolysaccharide layer; that is, a phenotype consistent with aerobic denitrification. A bioinformatic survey across known bacterial genomes showed that only microorganisms capable of aerobic metabolism possess the glyoxylate shunt. This trend is consistent with the hypothesis that the GS plays a previously unrecognized role in allowing bacteria to tolerate oxidative stress.

The glyoxylate shunt (GS) 2 is known to be essential for utilizing acetate and fatty acids as carbon sources under physiological conditions requiring gluconeogenesis (1). The glyoxylate cycle is a variant of the tricarboxylic acid cycle and shares five of the eight enzymes; the glyoxylate cycle bypasses the carbon dioxide-generating steps of the tricarboxylic acid cycle, which are catalyzed by isocitrate dehydrogenase and ␣-ketoglutarate dehydrogenase (2). This shortcut for the tricarboxylic acid cycle has the potential to disturb cellular redox potentials because the primary function of the tricarboxylic acid cycle is the generation of reduced electron carriers (e.g. NADH and FADH 2 ) via oxidation of acetyl CoA. Bypassing the complete tricarboxylic acid cycle conserves carbon atoms for gluconeogenesis while simultaneously diminishing the flux of electrons funneled into respiration (3). The product of aceA, the first enzyme of the glyoxylate shunt, catalyzes the formation of glyoxylate and succinate from isocitrate; the next reaction, conversion of glyoxylate to malate, proceeds via malate synthase, which is encoded by glcB (or aceB in Escherichia coli). Isocitrate lyase competes with isocitrate dehydrogenase because they share the same substrate, isocitrate. Their relative activities are controlled by isocitrate dehydrogenase kinase/phosphatase (AceK) through isocitrate dehydrogenase inactivation (4).
The glyoxylate shunt is known to be up-regulated when acetyl-CoA is a direct product of a metabolic pathway, for example via degradation of acetate, fatty acids, and alkanes (5). However, a growing body of evidence suggests that the glyoxylate shunt also plays an important role in pathogenesis and the response to oxidative stress (6). Our research group has shown that hexadecane degradation by the oil-degrading bacterium Acinetobacter oleivorans DR1 generates oxidative stress, which induces expression of aceA, encoding an isocitrate lyase (7). Pectin degradation by Alishewanella spp. is also accompanied by oxidative stress, which triggers induction of the glyoxylate shunt (8). It seems likely that oxidative stress is an inevitable consequence of aerobic hydrocarbon degradation when there is a high demand for oxygen and possible uncoupling of oxygenase activity and in which side reactions can generate hydrogen peroxide (9). Bactericidal antibiotics commonly induce oxidative stress, which achieves cell death by promoting the formation of hydroxyl radicals through the Fenton reaction (10). Hydroxyl radical-mediated oxidative stress can be toxic by damaging proteins, membranes, and DNA (11). Mycobacterium tuberculosis was found to up-regulate its glyoxylate shunt when exposed to three different kinds of clinically used tuberculosis antibiotics: isoniazid, rifampicin, and streptomycin (12). E. coli experiencing superoxide stress also increased metabolic fluxes (48%) through the glyoxylate shunt (13). However, the molecular mechanism underlying the link between oxidative stress and the glyoxylate shunt remains poorly understood. Given its potential roles in oxidative stress defense, antibiotic resistance, and pathogenesis, the glyoxylate shunt is a promising target for development of new antibiotics. Furthermore, the glyoxylate shunt is widely distributed in bacteria, fungi, some protists, and plants (14) but is absent in most vertebrates.
Bacterial responses to oxidative stress and the accompanying bacterial physiological changes have been extensively investigated in many pathogenic microorganisms, including Pseudomonas aeruginosa, which invades the lungs of cystic fibrosis patients and causes a potentially lethal infection (15). Oxidative stress causes many phenotypic and physiological changes in P. aeruginosa, including changes in biofilm production and pathogenesis (16,17). P. aeruginosa can utilize acetate and fatty acids as sole carbon sources through the glyoxylate pathway. In P. aeruginosa, aceA (PA2634) is known to be important for assimilation of acyclic terpenes, leucine, ethanol, and acetate (18).
The aim of this study was to explore a new role for the glyoxylate shunt in contributing to tolerance to oxidative stress. We applied microarray, biochemical, and physiological analyses to wild-type and both aceA and glcB knock-out mutants of P. aeruginosa. Furthermore, a global bioinformatics analysis across sequenced bacterial genomes suggests that this role may extend broadly across the bacterial domain.

Experimental Procedures
Antibiotics, Culture Media, Bacterial Strains, and Growth Conditions-The wild-type and isocitrate lyase (aceA) and malate synthase (glcB) mutant strains of P. aeruginosa MPAO1 were purchased from the Washington University Genome Center. Bacteria were grown at 37°C in Luria-Bertani (LB) broth or M9 minimal medium with glucose (2 g/liter), shaking at 220 rpm. M9 media were prepared with different sole carbon sources: glucose (0.2%), succinate (10 mM), acetate (10 mM), and hexadecanoic acid (0.2%). Paraquat (PQ) was added at the final concentrations described in the figure legends. Growth was monitored by measuring the A 600 of the liquid cultures by using a BioPhotometer (Eppendorf).
Gene Expression Analysis Using Real-time RT-PCR (qRT-PCR)-Wild-type and mutant cells were grown to the exponential phase at 37°C in LB. The cells were treated with 1 mM PQ and then incubated for 15 min. Isolation of total RNA, cDNA synthesis, and PCR were performed as previously described (19). The qRT-PCR was performed using the iCycler iQ realtime PCR detection system (Bio-Rad). For quantification, the 16S rRNA gene was used to obtain reference expression data. Three independent PCR reactions from two independent biological replicates were performed to measure relative expression of denitrification genes and oxidative stress-related genes. Primers used in qRT-PCR assay were listed in supplemental Table S1.
Survival and Sensitivity of P. aeruginosa-For the sensitivity assay using paper disks, stationary-phase cells were added to each M9 agar plate with 0.2% glucose, then 20 l each of 10 mM or 100 mM kanamycin was loaded into the plates on paper disks. After 24 h of incubation, bacteria-free zones around the disks were measured. For the survival test, 2.5 mM PQ or 3 mM KCN was used. For the PQ survival test, exponential-phase cells were harvested and washed 3 times with PBS (pH 7.4). Approximately 10 7 cell/ml were inoculated into fresh PBS (2 ml) containing PQ. For the KCN survival test, 3 mM aliquots were added with or without 1 mM sodium nitrate to LB. At each time point the cells were harvested and washed in PBS. The number of viable cells was determined by measuring the cfu on LB plates.
Microarray Analysis-Cells were grown to the exponential phase (A 600 ϭ 0.4) at 37°C with aeration. Cells were then treated with 1 mM PQ for 15 min. Total RNA was isolated using the RNeasy Mini kit (Qiagen). Specific steps for further RNA isolation were conducted as previously described (20). cDNA probes for cDNA microarray analysis were prepared by the reverse transcription of total RNA (50 g) in the presence of aminoallyl-dUTP and 6 g of random primers (Invitrogen) for 3 h. The cDNA probes were cleaned up using a Microcon YM-30 column (Millipore) followed by coupling to the Cy3 dye (for the reference) or Cy5 dye (for the test samples) (Amersham Biosciences). The dried Cy3-or Cy5-labeled cDNA probes were then resuspended in hybridization buffer containing 30% formamide (v/v), 5ϫ saline-sodium citrate, 0.1% SDS (w/v), and 0.1 mg/ml salmon sperm DNA. The Cy3-or Cy5-labeled cDNA probes were mixed together and hybridized onto a microarray slide. The hybridization images on the slides were scanned using an Axon 4000B (Axon Instruments) and analyzed with GenePix Pro software (version 3.0, Axon Instruments) to determine gene expression ratios (control versus test sample). The microarray data were deposited in the NCBI GEO site (under accession no. GSE78230). Genes that showed changes of Ͼ1.5fold (up-regulated genes) were selected. Expression of representative genes was also confirmed by qRT-PCR.
Measurement of Oxygen Consumption Rate-Exponentialphase cells grown in LB were diluted in 20 ml of LB in serum vials (radius, 30 mm; height, 60-mm clear serum vials were used; the volume of all serum vials was 27 ml), and further inflow of oxygen was blocked. Under these conditions the consumption of dissolved oxygen in the sealed 27 ml was measured with a Fibox 3 LCD trace transmitter (PreSens). Measurements were taken every 5 s for 30 min and analyzed with LCDTRACEv204 software. The rate of oxygen consumption was calculated as the decrease in oxygen concentration measured at 10-min intervals. Measurements were adjusted according to cell numbers (cfu) present in each vial. Experiments were conducted in triplicate.
Detection of NO-The concentration of NO produced was estimated using the reagent 4,5-diaminofluorescein diacetate((Sigma), according to a previously described method (21). Exponential phase cultures (1 ml) were obtained and 10 M 4,5-diaminofluorescein diacetate was added. After 1 h of incubation, cells were washed twice with PBS, and fluorescence was measured under green fluorescent light (excitation wavelength, 495 nm; emission wavelength, 515 nm) using a multidetection microplate reader (Hidex).
Detection of N 2 O Concentration-Cells were diluted in 10 ml of LB in serum vials, and inflow of oxygen was blocked using a 20-mm silicon stopper that trapped gases generated during growth. After incubation, head-space gas was sampled through the stopper using a gas-tight syringe (Hamilton) at each time interval. The concentration of N 2 O was measured with a gas chromatograph (GC-2010; Shimadzu; ZP-Porapak Q 80/100 mesh column was used with the electron capture detector (ECD). The result of 4 h of incubation was depicted on a graph because nitrous oxide concentrations in 1-and 2-h samples were not enough for detection.
Biofilm Formation Assay-Polystyrene 96-well microtiter plates (BD Biosciences) were used as abiotic surfaces to study biofilm formation. Bacterial cultures were grown overnight, washed twice in PBS, and inoculated at concentrations of 10 6 cell/ml in LB broth. After 24 h of incubation at 37°C, biofilm formation was determined via crystal violet staining and quantified by measuring the absorbance at 595 nm, normalized by the absorbance at 600 nm.
Exopolysaccharide (EPS) Analysis-The EPS was isolated as follows. First, culture solutions of P. aeruginosa wild type (WT) and mutants grown in LB media were centrifuged for 30 min at 12,000 ϫ g at 4°C. Next, the supernatants were mixed with three volumes of cold ethanol and maintained overnight at 4°C. After centrifugation (5000 ϫ g, 15 min, 4°C), the pellets were suspended in 80% ethanol and centrifuged again followed by three rounds of washing. The final precipitates were dissolved in distilled water. To remove excess salts, the EPS was dialyzed (Centricon, Amicon). Then the extracted EPS was freeze-dried and weighed.
Confocal Laser Scanning Microscopy-Cells from biofilms were stained with 200 l of SYPRO Ruby biofilm matrix stain (Invitrogen) for 20 min. The cells were then observed by confocal laser scanning microscopy (LSM700; Carl Zeiss). Before the biofilm staining procedure, cells were incubated on the cover glass for 12 h at 37°C. Confocal images of SYPRO Rubystained biofilm cells were observed under red fluorescent light (excitation wavelength, 450 nm; emission wavelength, 610 nm) to evaluate the height and density of the biofilms (C-Apochromat 403/1.20 W Corr M27; Carl Zeiss). The confocal laser scanning microscopy images were analyzed using the Zen 2012 program.
Determination of ATP Concentrations-To measure intracellular ATP concentrations, the ENLITEN ATP assay system bioluminescence detection kit (Promega) was used in accordance with the manufacturer's instructions. To confirm ATP concentrations, exponentially growing cultures were treated with 0.5 mM or 1 mM PQ for 15 min, and cells were harvested for ATP extraction using 1% trichloroacetic acid buffer and treated as previously described (22). The ATP concentration is expressed as the molar concentration per mg of protein.
Measurement of the NADH/NAD ϩ Ratio-Nicotinamide adenine dinucleotide (NAD ϩ ) and NADH concentrations were measured using the EnzyChrom TM NAD ϩ /NADH Assay kit (BioAssay Systems) according to the manufacturer's instructions. Exponentially growing cultures were treated with 1 mM PQ for 15 min, and cells were collected for NADH and NAD ϩ extraction. Further procedures were performed as previously described (22). NAD ϩ and NADH concentrations were normalized by the amount of total protein.
Determination of Cellular Acid-soluble Thiols-Total acidsoluble thiols were measured using the method of Lawley and Thatcher (57), with some modifications. WT, ⌬aceA, and ⌬glcB cells grown in 50 ml of M9 media to the exponential phase were centrifuged and washed in phosphate buffer with 1 mM EDTA. The cells were lysed by sonication. Proteins were precipitated from the supernatant fractions by the addition of cold trichloroacetic acid (final concentration of 5%). After 10 min the acidified samples were clarified by centrifugation, and 20 l of clear supernatant was added to 180 l of a 5,5Ј-dithiobis-(2-nitrobenzoic acid) (DTNB) solution (200 g/ml in 0.2 M sodium phosphate buffer). The absorption at 410 nm was measured immediately. Glutathione (GSH) was used as the thiol standard, and the thiol content was normalized to the amount of total protein.
Bioinformatic Analysis of the Glyoxylate Shunt-To investigate the phylogenetic distribution of the glyoxylate shunt, isocitrate lyase and malate synthase sequences were retrieved using the Gene Search function in the integrated microbial genome (IMG) database (23), which contained 27,319 bacterial genomes as of July 2015. To reduce database redundancy, only one strain was arbitrarily selected for each species when genomic data were available for multiple strains of one species. The oxygen requirements of selected bacterial species were identified based on experimental evidence from published papers. Bacterial species harboring the glyoxylate shunt and experimentally identified physiology were subjected to cladogram analysis. Cladograms in the Newick format were generated using the program phyloT and visualized using iTOL (24).

Results
Physiological and Phenotypic Alterations in the Absence of the Glyoxylate Shunt-To ascertain the broad physiological role of the glyoxylate shunt in P. aeruginosa, PA2634 (⌬aceA) and PA0482 (⌬glcB) mutants were tested in a variety of assays. Based on the GS pathway stoichiometry, the aceA mutant is strongly committed to tricarboxylic acid cycle-related pathways that deliver reducing power (e.g. NADH and FADH 2 ) to respiratory chains, whereas the glcB mutant likely experiences elevated levels of glyoxylate when the GS is induced.
Growth of the two mutants was tested using minimal media containing different carbon sources (Fig. 1A). With succinate as sole carbon source, both mutants displayed growth curves similar to the WT. As expected, the aceA and glcB mutants showed growth defects in media requiring a functional glyoxylate shunt (acetate and hexanoic acid a carbon sources; Fig. 1A) (25). However, an unexpected growth promotion was observed in both mutants when grown in rich LB media that should boost oxidative stress, relative to the minimal medium. To further investigate this surprising growth-enhanced phenotype in the mutants, we raised the level of superoxide-induced oxidative stress in the cells by adding PQ to the LB medium; growth enhancement was retained in the aceA mutant, whereas the glcB mutant showed a severe growth defect (Fig. 1B). Note that a high intracellular glyoxylate concentration (expected in the glcB mutant) is likely to be toxic because glyoxylate can be transformed into toxic intermediates, such as glyoxal or glycolaldehyde (26).
The contrasting responses of the two mutants to PQ (Fig. 1B) prompted us to measure the expression of both genes in WT cells exposed to PQ and hydrogen peroxide (H 2 O 2 ); only aceA was highly up-regulated (Fig. 1C). Thus, aceA is clearly activated in response to oxidative stress. To confirm that our experimental procedures were valid, we performed a control experi-ment with an obligate aerobic strain of Pseudomonas putida, which showed that our system provided a truly anaerobic environment; without oxygen, aceA was not up-regulated in the presence of PQ (data not shown).
A survival assay in broth culture dosed with PQ (2.5 mM) also showed that aceA mutant cells were more resistant to PQ toxicity than WT cells and the glcB mutant (Fig. 1D). Moreover, a paper disc assay with PQ and kanamycin confirmed the trends found for PQ survival assay in broth (Fig. 1E). The above observations clearly establish that aceA and glcB have distinctive physiological impacts on the cell physiology of P. aeruginosa.
Transcriptomic and Biochemical Analyses of aceA Mutants-Microarray analyses of P. aeruginosa WT and glyoxylate shunt mutants were conducted with or without PQ. Up-regulation of genes involved in denitrification was immediately noticeable in mutants without PQ despite the fact that these assays were conducted under aerobic conditions (supplemental Table S2).
Nitrate present in LB medium would likely readily be available as an electron acceptor for denitrification. The nitrate reductase operon (narGHJI), the nitrite reductase gene (nirS), the nitric oxide reductase operon (norDBC), and the nitrous-oxide reductase gene (nosZ), required for denitrification, were all highly up-regulated in aceA mutants, whereas a lower level of induction was observed in glcB mutants. qRT-PCR largely confirmed the microarray data (Table 1): especially in the case of the aceA mutant ( Fig. 2A). Activation of nitrate-respiration genes implies alteration of cellular electron flow away from oxygen as the terminal electron acceptor. Oxygen consumption was measured simultaneously, and we found that the aceA mutant cells consumed less oxygen than WT cells (Fig. 2B), although the growth of aceA mutants was enhanced in LB medium (Fig. 1B). Potassium cyanide (KCN) is known to bind to cytochrome oxidase in cells, thereby interrupting oxygen transport and preventing aerobic respiration. The aceA   mutants (featuring up-regulated denitrification genes; Table 1) would be expected to show enhanced cyanide resistance (Fig.  2C). Also, as predicted, when nitrate was added to the medium, both mutants showed greater resistance to KCN (Fig. 2D), suggesting that glcB mutants can also direct electron flow toward denitrification despite the likely intracellular accumulation of toxic glyoxylate.
To confirm that up-regulated denitrification transcripts led to a respiratory shift toward denitrification, we quantified production of the denitrification intermediates, NO and N 2 O (Fig.  2, E and F). Genes encoding many membrane transporters were also highly up-regulated. Transporters in the major facilitator superfamily and the resistance-nodulation-division (RND) family showed the highest up-regulation in aceA mutants (supplemental Table S3). In particular, mexE (PA2493) showed a 54.454-fold increase in expression in aceA mutants compared with the WT. The up-regulation of this well known multidrug efflux pump system may also contribute to the tolerance of P. aeruginosa to toxic chemicals.
Enhanced Biofilm Formation and Pigment Production in aceA Mutants-Our data showed that aceA mutants showed increased resistance to oxidants including PQ and kanamycin (Fig. 1E). The degree of gene expression involved in oxidative stress is less in PQ-treated mutants compared with PQ-treated wild-type cells. Interestingly, no significant expression of oxidative stress-related genes was observed in aceA or glcB mutants without PQ, except up-regulation of katN (PA2185), which encodes a non-heme catalase ( Table 2, supplemental  Table S4). In cell binding assays to microtiter plates, the aceA and glcB-deficient mutants showed increased biofilm formation (Fig. 3A). We hypothesize that the thick EPS layer may be beneficial for aerobic denitrification and resistance to chemical stressors. In fact, enhanced EPS production was observed in aceA mutants (Fig. 3B). Confocal microscopic analysis confirmed increased biofilm promotion in aceA mutants (Fig. 3C). Interestingly, when the mutant cells were cultivated, a distinct color change was observed, resulting from increased pyoverdin and pyocyanin formation (data not shown). Microarray analysis supported these findings (supplemental Table S3).
Intracellular NADH Level with PQ Treatment -Exposure to PQ for 15 min increased ATP contents in all cells (Fig. 4A). This short term increase in ATP has been proposed to be caused by inhibition of energy-consuming processes by PQ rather than increased respiration (27). However, PQ-exposed aceA mutants had a higher basal ATP concentration than the two other cell types, which might be caused by the growth condition that preceded the assay (Fig. 4A). Both WT and glcB mutants, which showed a growth defect upon PQ addition, displayed lower ATP concentrations. The intracellular NADH/NAD ϩ ratio might be affected by the availability of electron donors and acceptors (28). Both mutants had a lower basal NADH/NAD ϩ ratio because of their higher rates of respiration and efficient growth in LB (Fig. 4A). The NADH/NAD ϩ ratio under PQ conditions was lower in the WT, likely due to the fact that the glyoxylate shunt bypasses two NADH-generating steps. Cells show an increased NADH/NAD ϩ ratio when NADH oxidation is slowed by limitation of electron acceptors (29). The addition  of PQ did not alter the NADH/NAD ϩ ratio in the aceA mutants, indicating that respiration in these cells was maintained, perhaps through a combination of pathways; lower oxygen consumption and higher denitrification-related events were observed (Fig. 2A). The level of NADH in PQ-treated WT cells dropped markedly; a similar observation has been made after antibiotic treatment (29). Our data suggested that WT cells experienced increased oxidative stress when treated with PQ, whereas mutants did not. In the mutants only a subtle change in the NADPH/NADP ϩ ratio was observed after PQ addition (Fig. 4A). Total Thiol Levels under ROS Stress-Like other ␣-keto acids (30), glyoxylate can react with H 2 O 2 , thus providing an indirect defense against ROS stress. Glyoxylate may also play a role in ROS defense by generating glycine via alanine-glyoxylate transaminase or serine-glyoxylate transaminase, which could lead to the production of glutathione, a tripeptide antioxidant composed of glycine-cysteine-glutamate (31). Interestingly, transaminase (PA0132) was one of the most highly up-regulated genes in PQ-treated WT cells (supplemental Table S5). Expression analysis combined with total thiol quantification showed that the level of glutathione, which composed the majority (Ͼ95%) of total thiols, was higher in WT than the two mutants (Fig. 4B). The gene encoding glutathione synthase (PA0407) was up-regulated 1.845-fold in PQ-treated WT cells, whereas aceA and glcB mutants showed no significant changes (supplemental Table S5). Genes encoding glutathione S-transferase (PA1655, PA1033, and PA2813) and those encoding glutathione peroxidases (PA2826 and PA0838), all of which are well known to function in ROS protection (32), were also up-regu-lated under PQ in the WT (supplemental Table S5). However, H 2 O 2 treatment increased the total thiol level in both WT and aceA mutants, suggesting that glutathione synthesis may proceed via an alternative pathway under such conditions (Fig. 4B).
Phylogenetic Distribution of the Glyoxylate Shunt in Bacteria-We next investigated the extent to which the glyoxylate shunt co-occurs with genes encoding aerobic respiration in bacteria. We investigated the phylogenetic distribution of the glyoxylate shunt across all available bacterial genomes. A search for glyoxylate shunt in the integrated microbial genomes database showed that it is present in 1538 bacterial genomes from 339 genera belonging to 7 phyla. To bioinformatically verify the completeness of the glyoxylate shunt, we confirmed the co-occurrence of isocitrate lyase and malate synthase. To determine the relationship between the glyoxylate shunt and various physiological or ecological characteristics, experimental data from previously published papers were compared. In total, 957 species were found to possess the phenotype in question, and the genomes of those species were used to construct a cladogram (Fig. 5). Proteobacteria (65.2%), Actinobacteria (20.9%), and Firmicutes (9.6%) were the major phyla found to possess the glyoxylate shunt, showing that this bypass is not randomly distributed among bacterial species. We observed that all species harboring the glyoxylate shunt were aerobes or facultative anaerobes, with the exception of the anaerobic species Thiorhodospira sibirica; bacteria with halophilic, acidophilic, alkaliphilic, or thermophilic lifestyles and habitats were not found to possess the shunt (33). The copy number of isocitrate lyase and malate synthase in most bacterial genomes was one or two. Thirty of the genomes were found to harbor more than three copies of the glyoxylate shunt; however, these appeared to be incorrect annotations. Thus, our bioinformatics analysis indicates that aerobic respiration and the glyoxylate shunt are linked. We hypothesize that protection from ROS is the mechanism of this linkage.

Discussion
Only one isocitrate lyase, encoded by the PA2634 gene, is present in the genome of P. aeruginosa (18). Little is known about the contribution of the glyoxylate shunt to bacterial physiology, except that it plays roles in acetate and fatty acid metabolism. However, it has been reported that the glyoxylate shunt, particularly isocitrate lyase, is required for the pathogenesis of M. tuberculosis, and aceA mutants showed significantly attenuated survival in macrophages (34). Similar findings have been reported for other pathogenic bacteria, such as P. aeruginosa and Rhodococcus equi, as well as for fungal strains (35)(36)(37). In the case of R. equi, the authors speculated that aceA could be involved in metabolism of host membrane lipid-derived fatty acids, which are potential carbon sources in macrophages (37). However, a full understanding of the relevance of the glyoxylate shunt under various stresses remains unclear (9,19,38).
The role and regulation of the glyoxylate shunt have mainly been studied in E. coli, in which the glyoxylate shunt is present in an operon, aceBAK (2). In P. aeruginosa, aceA (PA2634) and glcB (PA0482, malate synthase) are present at different genomic locations (18), and our global genomic analysis demonstrated that this is also the case in many other species of bacteria. In E. coli, IclR negatively regulates the aceBAK operon and appears to be directly regulated by glyoxylate and pyruvate, which have been shown to bind to the C-terminal domain of IclR (39,40). IclR is also indirectly regulated by FadR, a fatty acid biosynthesis regulator (41). Furthermore, positive regulation by both FruR and integration host factor (IHF) has been reported (42,43). The indirect regulation of aceA by RpoN has been documented in P. aeruginosa (44), but other mechanisms for regulation of the glyoxylate shunt have not been well studied. In particular, it remains unclear how ROS regulate this pathway. Our data show that aceA and glcB were not induced to the same degree in P. aeruginosa in the presence of added PQ (Fig. 1C). This different degree of up-regulation has also been shown in our previous studies (9,19,38). We observed that aceA and glcB mutants showed different responses to PQ (Fig.  1, B-E), which was supported by other research (45). It is worth noting that the growth defect of glcB mutants under PQ was restored by cysteine but not by other metabolites such as glutamate, glycine, glutathione, or glyoxylate (data not shown).
Contrary to our expectations, aceA mutants of P. aeruginosa had a growth advantage in rich media (Fig. 1B), in which elevated levels of ROS stress can be generated, as shown by our previous findings. 3 Growth in succinate-supplemented minimal medium was equivalent in all strains. Likewise, the observed resistance of aceA mutants to antibiotics and PQ could be due to decreased ROS stress under such conditions (Fig. 1, B-D). Here, we found that enhanced growth of the aceA  mutant in biofilms was linked to both flow of respiration toward denitrification and enhanced production of EPS (Figs. 2 and 3). The importance of EPS production, biofilm formation, and generation of outer membrane proteins, such as OprF, has been shown for P. aeruginosa infection of the cystic fibrosis lung environment, which has microaerobic oxygen conditions (46,47). Cells of aceA mutants showed similar characteristics to cystic fibrosis-adjusted cell features, indicating that aceA mutants had adapted to avoid ROS stress by using both respiration systems. Increased EPS production, which may aid in formation of conditions under which denitrification is possible, might have resulted from the Ͼ2.5-fold up-regulation of alginate synthase (PA3544) observed in aceA mutants (supplemental Table S3). EPS could restrict the passage of toxic chemicals, including PQ. This is supported by the fact that alginate production and biofilm formation are controlled by oxygen availability in P. aeruginosa (48). Aerobic denitrification in continuous cultures of P. aeruginosa has been demonstrated, even at high concentrations of dissolved oxygen (49). Interestingly, a linkage between the mexEF-oprN (PA2493-PA2495) operon and the nitrate respiratory chain has also been reported, even under aerobic conditions (50). Our data suggest that they may also play a role in aerobic denitrification in aceA mutants (supplemental Table S3).
Additionally, up-regulation of glutathione-related genes was observed in PQ-treated WT cells, which could play an important role in ROS stress defense (supplemental Table S4). Due to the differential expression of aceA and glcB observed under PQ-treated conditions, we hypothesized that the glyoxylate produced by isocitrate lyase could be used in other mechanisms (Fig. 1C). It has been reported that Chlamydomonas reinhardtii regulates isocitrate lyase by glutathionylation (51) and that a transaminase (PA0132), which was highly up-regulated in PQtreated WT cells (supplemental Table S5), can transform glyoxylate to glycine, a component of glutathione. Further experiments are warranted to investigate the role of glyoxylate in the regulation and production of glutathione in P. aeruginosa (Fig. 4B). Involvement of the glyoxylate shunt in denitrification was first identified in Thiobacillus versutus (52). Previous studies focused on the metabolic function of the glyoxylate shunt reported induction by specific carbon sources such as acetate. However, many of our experiments here were performed in rich media (LB) to exclude the influence of acetate metabolism. Interestingly, up-regulated expression of katN (PA2185), which encodes a non-heme catalase, was observed in mutants (Table  2); however, its role in P. aeruginosa is not fully understood (53). Increased siderophore production and decreased susceptibility to iron chelators were also observed in mutant cells (data not shown). These data indicate that iron requirement for WT and mutants would differ. It has been shown that inhibition of oxygen transfer under conditions of iron deprivation enhances secretion of iron chelators in P. aeruginosa (54,55). Taken together, these iron-related features are important for denitrification respiratory-induced cell conditions.
In Neisseria gonorrhoeae, the electron transfer chains of aerobic respiration and denitrification have been shown to be interwoven, allowing N. gonorrhoeae to co-express aerobic respiration and denitrification in microaerobic conditions (56).
The exact proportion of respiration that is performed via denitrification in aceA mutants remains unclear; however, we hypothesize that P. aeruginosa aceA mutants achieve a balance between the two respiratory pathways. Our bioinformatics analysis revealed that only aerobes or facultative anaerobes employ the glyoxylate shunt. This result implies that isocitrate lyase may also play a role in regulation of respiration and ROS management in many other species of bacteria (Fig. 5).
The glyoxylate shunt role in bacterial physiology has traditionally been associated with the need for gluconeogenesis, as induced by carbon-source limitation. Here, we have shown that the glyoxylate shunt in P. aeruginosa is induced by oxidative stress. This type of oxidative stress has implications for both colonization of surfaces (biofilm formation) and genomic evolution across the bacterial domain. We have also revealed several aspects of the mechanism by which the glyoxylate shunt governs a shift from aerobic to nitrate-based cellular respiration. Future research advancing a deeper understanding of the connections between oxidative stress response, the glyoxylate shunt, biofilm formation, cell pathogenicity, and development of microbial communities is warranted.