Escherichia coli Increases its ATP Concentration in Weakly Acidic Environments Principally through the Glycolytic Pathway

Acid resistance is an intrinsic characteristic of intestinal bacteria in order to survive passage through the stomach. Adenosine triphosphate (ATP), the ubiquitous chemical used to power metabolic reactions, activate signaling cascades, and form precursors of nucleic acids, was also found to be associated with the survival of Escherichia coli (E. coli) in acidic environments. The metabolic pathway responsible for elevating the level of ATP inside these bacteria during acid adaptation has been unclear. E. coli uses several mechanisms of ATP production, including oxidative phosphorylation, glycolysis and the oxidation of organic compounds. To uncover which is primarily used during adaptation to acidic conditions, we broadly analyzed the levels of gene transcription of multiple E. coli metabolic pathway components. Our findings confirmed that the primary producers of ATP in E. coli undergoing mild acidic stress are the glycolytic enzymes Glk, PykF and Pgk, which are also essential for survival under markedly acidic conditions. By contrast, the transcription of genes related to oxidative phosphorylation was downregulated, despite it being the major producer of ATP in neutral pH environments.


Introduction
Both commensal and pathogenic enteric bacteria have adapted to resist several hours of exposure to the acidic environment of the human stomach, which has an average pH of 2.0, before entering the more hospitable intestinal tract. Multiple genes and pathways play a role in acid resistance (AR) and the acid tolerance response (ATR) in bacteria [1]. At least four AR systems have been identified in Escherichia coli [2][3][4][5][6], which make use of enzymatic processes to consume protons or produce metabolites to buffer the pH and protect against stress [7,8]. AR1 (oxidative system) requires the sigma factor RpoS [9,10] and the cyclic AMP receptor protein CRP [11]. AR2-AR4 function through amino acid decarboxylation [4,12,13], with AR2, AR3 and AR4 reliant on glutamate (glutamate decarboxylase system) [12,13], arginine (arginine decarboxylase system) and lysine (lysine decarboxylase system) [14][15][16][17], respectively. AR2 is the most effective amino acid-dependent system, requiring two glutamate decarboxylases (GadA and GadB), which are active at an acidic pH, and the antiporter GadC [18][19][20]. The glutaminase and adenosine deaminases were also shown to contribute to AR in E. coli, and these enzymes use their respective substrates to release NH 3 into the cytoplasm and raise the intracellular pH [21,22].
Interestingly, we previously published findings demonstrating a relationship in E. coli between intracellular ATP levels and the environmental pH [23]. We found that a high concentration of ATP within the bacterial cell is necessary for survival under extremely acidic conditions [23]. Although ATP levels decreased rapidly when the media was adjusted from a near-physiological pH of 7.5 to 2.5, ATP concentrations were actually found to increase only when cells were pre-adapted to a weakly acidic environment (pH 5.5), which was still permissible to E. coli growth [23].
The mechanism and purpose of this observed effect remain unclear, although there have been several proposed explanations. One possibility is that under stress at low pH, metabolic processes that would otherwise consume ATP are intentionally suppressed or forcibly inhibited by suboptimal enzymatic conditions. Bacterial growth was in fact noted to be reduced at pH 5.5. Alternatively, the production of ATP in weakly acidic environments may be augmented. ATP is produced in E. coli through both oxidative phosphorylation and glycolysis when glucose is present as the carbon source.
In oxidative phosphorylation, F 1 Fo-ATPase catalyzes the synthesis of ATP from ADP and inorganic phosphate using the electro-chemical gradient of protons across the cellular membrane. The greater pH gradient between the intracellular and extracellular compartments at pH 5.5 could boost the membrane potential and drive further ATP synthesis. Interestingly, when compared to levels at pH 7.5, both the pH gradient and membrane potential between intracellular and extracellular compartments were found to be minimal at a low pH (pH 2-2.5) [2,3,24]. Our previous work had shown that ATPase mutants could still elevate the level of ATP at pH 5.5 [25]. Foster et al. speculated that the ATPase may function as a proton pump and consume ATP to regulate the intracellular pH in E. coli [7]. During glycolysis, one mole of glucose is fermented and at least two moles of ATP are synthesized, which can either be used for biosynthesis or be hydrolyzed by F 1 Fo-ATPase [26]. These results suggest the possibility that the ATPase consumes ATP to pump out protons in acidic environments, and that another mechanism independent from the ATPase is used to generate ATP.
In glycolysis, glucose is converted to pyruvate through a 10-step pathway with multiple intermediate metabolites ( Figure 1). Steps 1 and 3 consume ATP, and steps 7 and 10 catalyze ATP synthesis, resulting in a net balance of 2 ATP per molecule of glucose. The initial step of the glycolysis pathway is catalyzed by glucokinase (Glk), a hexokinase isozyme that facilitates phosphorylation of glucose to glucose-6-phosphate. Step 7 is catalyzed by 3-phosphoglycerate kinase (Pgk) [27,28]. Two isozymes of pyruvate kinase, PykF and PykA, oversee the final step, where phosphoenolpyruvate is converted into pyruvate [29][30][31][32]. Pyruvate is then used in the tricarboxylic acid (TCA) cycle under aerobic conditions to produce additional NADH and succinate for further processing in oxidative phosphorylation. Citrate synthase (GltA) catalyzes an early step in the TCA cycle to condense oxaloacetate and acetyl coenzyme A into citrate and coenzyme A [33,34].
In this study, we attempt to clarify which metabolic pathway or genes encoding enzymes are responsible for the increase in intracellular ATP concentration in E. coli grown in weakly acidic media with glucose as the sole carbon source. This will in turn reveal the mechanisms used by E. coli for adaptation to acidic environments. The expression of genes encoding enzymes responsible for ATP synthesis at different pH levels was analyzed, and the effects on AR and ATP concentration were investigated.

Strains, Plasmids, Culture Media and Reagents
The bacterial strains and plasmids used in this study are listed in Table 1. All strains were initially grown at 37 °C in LB medium, then tested in EG medium, which is E medium [35] containing 0.4% glucose [36]. The medium pH was adjusted by the addition of NaOH or HCl. Antibiotics were used at the following concentrations: ampicillin, 100 µg/mL; kanamycin, 30 µg/mL; chloramphenicol, 30 µg/mL. Antibiotics, L-arabinose, N, N'-dicyclohexylcarbodiimide (DCCD) and carbonyl cyanide 3-chlorophenylhydrazone (CCCP) were obtained from Sigma-Aldrich (St. Louis, MO, USA).
In this study, we attempt to clarify which metabolic pathway or genes encoding enzymes are responsible for the increase in intracellular ATP concentration in E. coli grown in weakly acidic media with glucose as the sole carbon source. This will in turn reveal the mechanisms used by E. coli for adaptation to acidic environments. The expression of genes encoding enzymes responsible for ATP synthesis at different pH levels was analyzed, and the effects on AR and ATP concentration were investigated.

One-Step Inactivation of Chromosomal Genes in E. coli
Gene knock-out mutants in the glycolysis pathway in E. coli were created using conventional Red-mediated recombination [37]. A linear fragment containing the kanamycin or chloramphenicol resistance gene flanked by about 60 base pairs of the target metabolic gene was introduced through electroporation into E. coli strain BW25113, which already carried the Red helper plasmid pKD46. The primers used to amplify these genes are shown in Table 2. L-arabinose (1 mM) was added to the medium and bacteria were incubated at 37 • C for 1 h after electroporation. The mutants were selected on plates of LB medium containing the relevant antibiotic. DNA from individual transformants was isolated and tested by PCR amplification to confirm the integration of the resistance gene, and the Red helper plasmid pKD46 was cured by growth at 42 • C [38]. Primers used for confirmation are shown in Table 2.

P1 Transduction
The knock-out genes were transferred from BW25113 to W3110 by P1 phage as previously described [39]. The transductants of W3110 were selected by antibiotics.

Acid Tolerance Response (ATR) and Acid Resistance (AR) Test
For the logarithmic phase ATR test, the survival of wild-type and mutant strains was measured as previously described [23,25]. After overnight culture in LB medium with antibiotics where necessary, the cells were diluted 1000-fold with EG medium at pH 7.5 and cultured at 37 • C until the optical density at 600 nm (OD 600 ) reached 0.3 to 0.4. Cells were collected by centrifugation at 5000× g for 4 min and pellets were suspended in a 2-fold volume of EG medium at pH 5.5 before incubation under micro-aerobic culture conditions for 4 h without shaking. The OD 600 reached 0.2-0.3 after a 4 h adaptation at pH 5.5. The adapted cells were washed with fresh EG medium at pH 5.5 and then diluted 100-fold in EG medium at pH 2.5 [40]. After incubation at 37 • C for 1 h or 2 h, the cells were spun down, resuspended in phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 , and 1.4 mM KH 2 PO 4 , pH 7.4) and spread on LB agar plates for overnight culture. Colonies were counted, and viability was expressed as the percentage of viable cells out of the total cell number before the acidic challenge. The tests were performed in at least three independent experiments with biological triplicates.
For stationary phase AR assays, cells were grown overnight in LB buffered with either 100 mM morpholinepropanesulfonic acid (MOPS, pH 8) or 100 mM morpholineethanesulfonic acid (MES, pH 5.5) [10]. Cultures were grown in 4 mL of the buffered medium in 15-mm test tubes with shaking (220 rpm) at 37 • C to stationary phase (22 h) followed by 1:1000 dilution into prewarmed (37 • C) pH 2.5 EG. Viable-cell counts were determined at 2 h post-acid challenge by diluting cells in PBS, plating cells onto LB agar, and incubating plates at 37 • C before counting the number of colonies. Values given are representative of the results of triplicate experiments reproducible to within 50%.

ATP Content Measurement
After culturing as above, the cells were chilled on ice and then centrifuged at 10,000× g for 5 min at 4 • C. The pellets were treated with a solution containing 20 mM Tris-HCl, 50 mM MgSO 4 , 4 mM EDTA and 50% methanol at pH 7.4 for 30 min at 70 • C and then were centrifuged at 10,000× g for 5 min. The ATP content of the supernatant was measured as described previously by a luminometer (Turner Designs, Inc., Sunnyvale, CA, USA) [23,41,42]. Luciferase and standard ATP were purchased from Sigma.
The measurement was repeated at least three times independently, and the mean value and the standard deviation were calculated.

RNA Extraction and RNA-seq Analysis
The cells were cultured at pH 7.5 in EG medium until the OD 600 reached 0.3-0.4, then spun at 10,000× g for 5 min and transferred into the same volume of fresh EG medium at either pH 7.5 or pH 5.5. After half hour incubation, total RNA was isolated using RNA-Solv reagent (Omega, Norcross, GA, USA) according to the manufacturer's protocol. The concentration and purity of RNA were determined using the GeneQuant RNA/DNA Calculator (Pharmacia-Biotech, Cambridge, UK). The RNA samples were stored at −80 • C for the RNA-seq assay. The isolated RNA was sequenced on an Illumina HiSeq 2000 at Ribobio Co (Guangzhou, China). Reads contaminated by adapter sequence were removed from the pair end (PE) reads (30 bp) when either of the PE reads was polluted. Low quality reads (phred quality ≤19 in more than 15% of all the read bases) were further filtered out. Any reads that aligned to rRNA sequence were also filtered out. In addition, we removed reads that contained more than 5% Ns. The cleaned reads were aligned to the reference sequences in the National Center for Biotechnology Information (NCBI) database by Rockhopper v2.03 using default parameters. HTSeq v0.6.0 was utilized to calculate the co gen units of each e, and gene expression levels were represented by the Reads per Kilobase per Million reads (RPKM) method, as described by Mortazavi et al. [43]. The RNA-seq analyses were repeated in two independent experiments using three biological replicates. Differentially expressed genes (DEGs) at either pH 5.5 or pH 7.5 were identified using the DEGseq method (software: DESeq2), using a cut off of fold change > 2 and p value < 0.01.

Gene Ontology Annotation and KEGG Pathway Analysis
Gene Ontology (GO) is comprised of three aspects to describe gene functions: biological process (BP), molecular function (MF) and cellular component (CC). When performing functional enrichment analysis on the DEGs, we considered the BP branch. The Kyoto Encyclopedia of Genes and Genomes (KEGG) PATHWAY database was referenced for their maps of molecular interactions and reaction and relation networks for metabolism.
We used the online web tool DAVID [44] for functional enrichment analysis of GO using the KEGG pathways. EASE score was used to evaluate whether the DEGs were significantly enriched for a specific gene function. Benjamini-Hochberg (BH) method was used to adjust p-values for multiple comparisons. The R software programs fisher.test and p.adjust was also used. The enrichment threshold for p-value significance was set to 0.01.

Other Methods
The growth curves of all strains were measured by Bioscreen C (Oy Growth Curves Ab Ltd. Helsinki, Finland). Complementation plasmids were cloned as described previously [23]. After overnight culture in LB medium with antibiotics where necessary, the cells were diluted 500-fold with EG medium at pH 7.5 or pH 5.5 and cultured at 37 • C. Protein concentrations were measured by absorption at 595 nm using the Bradford Protein Assay (BioRad, Bradford, UK), with bovine serum albumin as the standard. The mRNA level was measured by qPCR following methods described in previous publications [21]. Primers for qPCR are listed in Table 2.

Statistics
Data were reported as mean ± standard error of the mean. Statistical significance of difference was determined by using unpaired two-tailed Student's t-test. A value of p < 0.05 was considered to be statistically significant. One-way ANOVA(Analysis of Variance) followed by Bonferroni's post hoc test or F-test was used to determine significant differences (p < 0.05) between groups.

Carbonyl Cyanide 3-Chlorophenylhydrazone (CCCP) and N,N -Dicyclohexylcarbodiimide (DCCD) Affect the ATP Level and Acid Resistance under Different Conditions
We had previously reported that E. coli mutants DK8 (deletions in all subunit genes of the FoF1-ATPase), atpE (deleted subunit c of the FoF1-ATPase) and atpD (deleted subunit β of the FoF1-ATPase) did not show significant changes in intracellular ATP concentrations under weakly acidic conditions [25]. CCCP is a protonophore and an uncoupler of oxidative phosphorylation that disrupts the cell membrane potential [45]. To test the survival of E. coli under different acidic conditions when ATP production via oxidative phosphorylation is inhibited, E. coli strain W3110 was grown in EG medium at pH 7.5 to an OD 600 of 0.3 to 0.4. The cells were then steadily adapted to increasing acidic conditions by growing at pH 5.5 for 4 h and then challenging for 1 h at pH 2.5, with the addition of CCCP (30 µM) at different time points (Figure 2A). No viable cells were detected when W3110 was given CCCP during the exposure to a pH of 2.5, whether or not CCCP had also been given at pH 5.5. However, there was no significant effect on survival when CCCP was only added to cells at a pH of 5.5 ( Figure 2A). Interestingly, the ATP level was only significantly decreased after the addition of CCCP at pH 7.5, with no significant decreases noted at pH 5.5 ( Figure 2B). We had previously reported that E. coli mutants DK8 (deletions in all subunit genes of the FoF1-ATPase), atpE (deleted subunit c of the FoF1-ATPase) and atpD (deleted subunit β of the FoF1-ATPase) did not show significant changes in intracellular ATP concentrations under weakly acidic conditions [25]. CCCP is a protonophore and an uncoupler of oxidative phosphorylation that disrupts the cell membrane potential [45]. To test the survival of E. coli under different acidic conditions when ATP production via oxidative phosphorylation is inhibited, E. coli strain W3110 was grown in EG medium at pH 7.5 to an OD600 of 0.3 to 0.4. The cells were then steadily adapted to increasing acidic conditions by growing at pH 5.5 for 4 h and then challenging for 1 h at pH 2.5, with the addition of CCCP (30 µM) at different time points (Figure 2A). No viable cells were detected when W3110 was given CCCP during the exposure to a pH of 2.5, whether or not CCCP had also been given at pH 5.5. However, there was no significant effect on survival when CCCP was only added to cells at a pH of 5.5 ( Figure 2A). Interestingly, the ATP level was only significantly decreased after the addition of CCCP at pH 7.5, with no significant decreases noted at pH 5.5 ( Figure 2B).  (A) W3110 cells were adapted for 4 h at pH 5.5 before challenging for 1 h at pH 2.5. CCCP was added at adaptation (pH 5.5) and challenge (pH 2.5) as indicated (N.D., not detected). (1) Control: W3110 without CCCP at both pH 2.5 and pH 5.5; (2) CCCP was added at pH 5.5, then cells were collected, washed twice and transferred to pH 2.5 for challenge; (3) CCCP was added both at pH 5.5 and pH 2.5; (4) CCCP was added only at pH 2.5. (B) The ATP content was measured after W3110 cells were incubated at pH 5.5 and pH 7.5 for 4 h with and without CCCP (viable cell number~5 × 10 7 /Ml). (C) W3110 cells were adapted for 4 h at pH 5.5 before challenging for 1 h at pH 2.5. DCCD was added at different concentrations as indicated. (D) The ATP content was measured after W3110 cells were incubated at pH 5.5 and pH 7.5 for 4 h with and without DCCD. The average values and standard deviations were obtained from three experiments. Comparisons are made between samples with and without CCCP or DCCD for both survival and ATP content after 1h of challenge. (** p < 0.01; * p < 0.05).
DCCD, an inhibitor of F 0 F 1 -ATP synthase, binds to the H + -transporting acidic residue of the F 0 c subunit, preventing the ATP synthesis activity of F 1 [46]. We found that the survival of E. coli under extremely acidic conditions (pH 2.5) was decreased by the addition of increasing amounts of DCCD ( Figure 2C). Interestingly, intracellular ATP levels were decreased at pH 7.5, but not significantly affected by DCCD at pH 5.5, except at the highest concentration (0.5 mM) ( Figure 2D). Recently, it was reported that DCCD inhibits F 0 F 1 -ATPase activity during the fermentation of glycerol through a process dependent on pH and potassium ions [26,47].
These results suggest that under weakly acidic conditions, ATP production via oxidative phosphorylation is either minimal or can easily be compensated through other processes.

Global Identification and Functional Inference of Acid-Regulated Genes, and Transcription Analysis of Known Glycolysis Genes
We analyzed the changes in the transcription levels of E. coli genes and transcripts, which include most of those with known roles in functional and metabolic pathways, both before and after the adaptation of cells to mildly acidic media [5,13,48]. Our RNA-seq data showed that the majority of gene transcription was upregulated (379 genes) rather than downregulated (261 genes) in E. coli undergoing acid adaptation at pH 5.5. A total of 155 genes were induced by at least two-fold (p < 0.01, FDR < 0.05), and interestingly, 26 of those genes were induced to almost four-fold higher levels than their baseline ( Figure 3A, Supplementary Table S1). By contrast, the transcription of 69 genes was decreased by two-fold or more (p < 0.01, FDR < 0.05) ( Figure 3A, Supplementary Table S1). Among the induced genes, several belonged to metabolic pathways known to be associated with the AR2, AR3 and AR4 systems ( Figure 3B, Supplementary Table S1 and Figure S1). Genes whose transcription was downregulated included ones associated with fatty acid oxidation, the electron transport chain, "energy derivation by oxidation of other organic compounds" and "generation of precursor metabolites producing energy", indicating that under weakly acidic conditions, E. coli may not rely significantly upon these other systems for energy production ( Figure 3C). This includes the sdhC gene, which encodes succinate dehydrogenase complex subunit C, a major component of the TCA cycle (Supplementary Figure S2). The expression of genes encoding glycolytic enzymes such as the acetyltransferase component of the pyruvate dehydrogenase complex (AceF) was upregulated greater than two-fold, and pyruvate kinase I (PykF) was upregulated almost two-fold (Supplementary Table S1). The expression of genes The expression of genes encoding glycolytic enzymes such as the acetyltransferase component of the pyruvate dehydrogenase complex (AceF) was upregulated greater than two-fold, and pyruvate kinase I (PykF) was upregulated almost two-fold (Supplementary Table S1). The expression of genes in the glycolysis pathway was further tested by qPCR. The results showed that under weakly acidic conditions, pykF and glk expression was increased 2-and 1.5-fold, respectively, compared to at a neutral pH ( Figure 3D). Other genes in the glycolysis pathway (pykA, pgi, pgk and pfo), and the first TCA cycle gene (gltA), were also upregulated under weakly acidic conditions ( Figure 3D).
Altogether, the gene transcription data suggest that the expression of several glycolysis pathway genes increases in weakly acidic environments, and that the increased ATP content is less likely due to the induction of other metabolic pathways or reduced energy consumption.

Characteristics of E. coli Knockout Mutants in Glycolysis and TCA Cycle Genes
As glycolysis gene transcription was upregulated under weakly acid conditions, we created knockout mutants of key enzymes in the glycolytic pathway whose reactions directly catalyze ATP synthesis (pgk, pykA and pykF), as well as other glycolysis genes (pgi, glk and pfo) and the enzyme responsible for initiating the TCA cycle (gltA) [49]. Using conventional Red-mediated recombination, these genes in E. coli strain BW25113 were replaced with chloramphenicol or kanamycin resistance genes, and clones were selected on plates with the appropriate antibiotic. Individual knockouts were confirmed through PCR amplification.
The growth kinetics of these mutants under different conditions was compared. The ∆gltA and ∆glk strains grew slower than the other mutants in EG medium ( Figure 4A) at pH 7.5. Although ∆gltA exhibited the slowest growth rate at a neutral pH in EG ( Figure 4A,C) and LB media [49], ∆glk also grew to a low optical density at 600nm (OD 600 ) in EG media at both pH 7.5 and 5.5 ( Figure 4A,B). At pH 5.5, The glk mutant reached an OD 600 of only 0.2 with the lowest growth rate (µ) of all the mutants, likely because this key gene is responsible for the first step in the glucose metabolism pathway. The ∆pykA and ∆pykF mutants still grew at pH 7.5 and pH 5.5 in EG medium (Figure 4), but a double-knockout mutant of both genes did not (Figure 4). ∆pykA grew slower than ∆pykF at pH 7.5 ( Figure 4A,C), however, this trend was reversed at pH 5.5 ( Figure 4B,D). These data indicate that the two pyruvate kinases are necessary for E. coli growth in EG medium, in which glucose is the sole energy source, and that pykF and pykA are particularly active in acidic and neutral pH environments, respectively. In EG medium at pH 5.5, ∆pgk reached steady-state slowly, with a final optical density of around 0.4, almost as low as that of ∆pykF, which also shared the same µ value ( Figure 4B,D). These results suggest that GltA may be important for E. coli growth. Under mildly acidic conditions, components of the glycolysis pathway (Glk, Pgk and PykF) take on a more important role. Other mutants such as ∆pgi and ∆pfo showed no effect on the growth of E. coli compared to wild type strain. Our results indicate that the metabolism of glucose by E. coli in different pH environments may be mediated by different glycolytic enzymes. Comparisons are made between the individual mutants and WT(wild type), asterisk (*) indicates pvalue < 0.05, two asterisk (**) indicates p-value < 0.01, three asterisk (***) indicates p-value < 0.001.

Comparisons of ATP Levels between Different Mutant Strains under Weakly Acidic Conditions
It was noted that the ATP concentration in E. coli was increased after adaptation to weakly acidic conditions [23]. To investigate which glycolysis pathway genes are important for this process, we compared the ATP content in our knockout mutants and the wild-type strain W3110 at pH 7.5 and pH 5.5 in EG medium ( Figure 5A). The increase in ATP after adaptation to pH 5.5 was slightly higher in ΔpykA compared to W3110. The gltA mutant showed a slightly higher ATP content after acid adaptation, however, it did not reach the level seen in W3110.

Comparisons of ATP Levels between Different Mutant Strains under Weakly Acidic Conditions
It was noted that the ATP concentration in E. coli was increased after adaptation to weakly acidic conditions [23]. To investigate which glycolysis pathway genes are important for this process, we compared the ATP content in our knockout mutants and the wild-type strain W3110 at pH 7.5 and pH 5.5 in EG medium ( Figure 5A). The increase in ATP after adaptation to pH 5.5 was slightly higher in ∆pykA compared to W3110. The gltA mutant showed a slightly higher ATP content after acid adaptation, however, it did not reach the level seen in W3110. In contrast to the effect seen in W3110, adaptation to pH 5.5 resulted in a drop in ATP level for deletion mutants Δglk, ΔpykF and Δpgk. The Δglk mutant also showed significantly lower ATP content compared to W3110 at pH 7.5. Complementation using plasmids harboring the pykF and pgk genes resulted in wild-type ATP levels in their respective deletion mutants ( Figure 6A). In contrast to the effect seen in W3110, adaptation to pH 5.5 resulted in a drop in ATP level for deletion mutants ∆glk, ∆pykF and ∆pgk. The ∆glk mutant also showed significantly lower ATP content compared to W3110 at pH 7.5. Complementation using plasmids harboring the pykF and pgk genes resulted in wild-type ATP levels in their respective deletion mutants ( Figure 6A). These results indicate that the glycolysis pathway is the main source of ATP production in E. coli in acidic environments. In addition, the enzymes encoded by pgk and pykF may be important for increasing ATP synthesis under weakly acidic conditions. It is possible that an alternative metabolic system produces ATP in pgk and pykF mutants, but it may be less active or fail to increase ATP under acidic conditions.

Comparisons of Acid Resistance between Different Mutant Strains
As ATP content is important for E. coli survival in extremely acidic environments, these mutants were also tested for acid resistance. Survival after the acid challenge was reduced in all strains, but compared to the wild-type, the effect was particularly significant in glk, pgk and pykF mutants, with a smaller difference seen in the gltA mutant ( Figure 5B). Complementation in mutants, Δpgk and ΔpykF with their respective plasmids restored the level of AR to that of the wild-type strain ( Figure  6B). Together with our previous results, these data show that the glycolysis genes glk, pgk and pykF participate in regulating ATP levels in E. coli to enable their survival upon acid challenge. We next examined whether these genes also affect acid resistance for cells in the stationary phase of growth. When the mutants were grown to stationary phase at pH 8 and then challenged for 2 h at These results indicate that the glycolysis pathway is the main source of ATP production in E. coli in acidic environments. In addition, the enzymes encoded by pgk and pykF may be important for increasing ATP synthesis under weakly acidic conditions. It is possible that an alternative metabolic system produces ATP in pgk and pykF mutants, but it may be less active or fail to increase ATP under acidic conditions.

Comparisons of Acid Resistance between Different Mutant Strains
As ATP content is important for E. coli survival in extremely acidic environments, these mutants were also tested for acid resistance. Survival after the acid challenge was reduced in all strains, but compared to the wild-type, the effect was particularly significant in glk, pgk and pykF mutants, with a smaller difference seen in the gltA mutant ( Figure 5B). Complementation in mutants, ∆pgk and ∆pykF with their respective plasmids restored the level of AR to that of the wild-type strain ( Figure 6B). Together with our previous results, these data show that the glycolysis genes glk, pgk and pykF participate in regulating ATP levels in E. coli to enable their survival upon acid challenge. We next examined whether these genes also affect acid resistance for cells in the stationary phase of growth. When the mutants were grown to stationary phase at pH 8 and then challenged for 2 h at pH 2.5, only ∆pykF showed significantly decreased survival (p < 0.05) ( Figure 5C). When the mutants were grown to stationary phase at pH 5.5 before challenging for 2 h at pH 2.5, both the pykF and pgk mutants showed lower survival. We noted that wild-type cells normally had a higher ATP content at pH 5.5 than that at pH 8 at the stationary phase. However, the intracellular ATP level of ∆pykF was significantly decreased in the stationary phase at both pH 8 and pH 5.5 ( Figure 5D), while the pgk mutant showed a significantly decreased level only when grown to stationary phase at pH 5.5. These results confirmed that the glycolysis pathway mainly functions to supply ATP at a weakly acidic pH.
To further investigate whether these mutants primarily affect AR and ATP levels through the glycolysis pathway, fructose and pyruvate were used to bypass the glk and pykF mutations. The data showed that 0.2% fructose added to 0.2% glucose could restore ATP levels and cell survival of glk mutants under extremely acidic conditions ( Figure 6C,D). However, 0.4% pyruvate added to 0.2% glucose as a carbon source was unable to rescue pykF mutants, and even wild-type W3110 showed lower ATP levels and lower survival ( Figure 6D). These results confirm that genes in the glycolysis pathway, particularly those with a role in ATP generation, are responsible for keeping ATP levels elevated in weakly acidic environments.

Discussion
Gut bacteria primarily enter the human intestinal tract through the stomach, whose acidic environment might initially be buffered to a pH of up to 6.0 depending what other contents are being digested. After approximately 4 h, the pH gradually decreases to 2.5, granting the bacteria time to activate their acid resistance genes and adapt [50]. During this period of adaptation to weakly acidic conditions, E. coli was found to have an elevated ATP level, which is believed to play an important role in AR [23].
The concentration of ATP in E. coli cells is known to increase in response to certain stresses, such as the change in osmotic pressure and environmental pH [23,51]. The pykA gene was reported to negatively regulate ATP levels under anaerobic conditions [32]. The availability of ATP during adaptation to acidic conditions becomes increasingly more important for survival as the pH decreases [23]. This could be explained by an increased demand for energy as new mechanisms are activated to counteract the hostile environment and keep the cytoplasmic space at a constant pH [52,53]. However, these systems no longer fully compensate after the pH drops below 6, as the intracellular compartment will begin to acidify at that point [53]. For this reason, we selected a pH of 5.5 as our experimental condition for growth under a mildly acidic challenge [54].
Whether the increase in ATP levels during acid adaptation is a direct result of increased production (and if so, by which pathway), or through limiting the energy consumption of nonvital systems, was previously unclear. Our analysis showed that gene transcription was upregulated rather than downregulated in E. coli undergoing acid adaptation, suggesting that the mechanism is not via a reduction in ATP use. Furthermore, genes related to transcription were more upregulated than those related to translation under these conditions. The ATP content also correlated directly with the growth of different mutants and the wild type strain (Figures 4 and 5).
Our results show that the glycolytic enzymes encoded by pykF and pgk provide the primary supply of ATP to E. coli under weakly acidic conditions. In addition, the gene expression level of AceF, a component of the pyruvate dehydrogenase complex, was also increased more than two-fold during acid adaptation. Similarly, the levels of enzymes involved in glucose metabolism pathways, including glucokinase (Glk), as well as the expression of pyruvate ferredoxin oxidoreductase (Pfo), were increased under acidic conditions. The deletion of glk showed significant effects on both AR and the ATP level, likely because Glk mediates the initial step of the glycolysis pathway and is a key enzyme. Its deletion may also affect the downstream ATP production from PykF and Pgk. By contrast, when we treated E. coli with CCCP to disrupt the membrane potential, the ATP level decreased only slightly during acid adaptation, revealing that oxidative phosphorylation is not a major supplier of ATP under these conditions, despite the fact that at near-neutral pH values, it is the main producer of ATP. E. coli treated with DCCD also showed no significant change in ATP level at pH 5.5, which is in agreement with our ATPase mutant data [25]. This was also supported by our RNA-seq data showing that the transcription of genes associated with respiration was downregulated in weakly acidic growth conditions. The transcription of genes in the fatty acid and lipid oxidation pathways, as well as those related to energy derivation by oxidation of other organic compounds or the generation of precursor metabolites, were also downregulated during adaptation to weakly acidic media ( Figure 3C).
Our data suggest that glycolysis is the main pathway that supplies ATP at pH 5.5 and the metabolism of glucose by E. coli in different pH environments may be mediated by different glycolytic enzymes. ∆pykA grew slower than ∆pykF at pH 7.5 ( Figure 4A), but, this trend was reversed at pH 5.5 ( Figure 4B). ∆glk grew at the slowest rate in EG media at both pH 7.5 and 5.5, yet it may be supported by other glucose metabolism pathways. For example, Roseman et al. reported that when glucose is taken up by E. coli, it can be converted to glucose 6-phosphate by phosphotransferase [55]. However, this enzyme activity may be inhibited by a low pH.
Our experiments also showed that the addition of pyruvate, the end-product of glycolysis, could not rescue pykF mutants or even wild-type W3110 at pH 5.5 even when 0.2% glucose was present as a carbon source ( Figure 6C,D). In fact, it may be that pyruvate itself negatively impacts E. coli survival under extremely acidic conditions. Pyruvate is normally further metabolized to lactic acid, acetic acid, alanine, or acetyl-CoA through the TCA cycle [56,57]. We showed that adaptation to mildly acidic media (pH 5.5) resulted in a two-fold upregulated expression of the aceF pyruvate dehydrogenase gene and the lactate dehydrogenase ldhA gene (Supplementary Table S1). Acetyl-CoA is converted into acetate, coenzyme A, and ATP via the enzymes phosphotransacetylase (Pta) and acetate kinase (AckA). One study showed that the deletion of ackA and pta increased the survival of E. coli [58]. Others have found that when excess glucose is converted to acetate (as also happens with the addition of pyruvate or during phosphate starvation), it can partially uncouple and deplete the proton motive force (PMF), causing cell death under extremely acidic conditions [59][60][61]. In E. coli, The ATPase may pump out protons to regulate the intracellular pH [2], which could also compensate for the decrease in the PMF otherwise mediated by the respiratory chain. In our own experiments, we also observed that the addition of acetate decreased the survival of E. coli in extremely acidic conditions. It remains unclear whether this negative impact on survival is related to decreased efficiency in ATP production or to its direct effects on the PMF. GadY, which encodes a small RNA, has been reported to decrease acetate production [62] and our results showed that GadY was upregulated under weakly acidic conditions (Supplementary Table S1). These data could suggest that pyruvate metabolism under acidic conditions may be through the lactate pathway rather than the TCA cycle or acetate pathway.
It is generally accepted that amino acid decarboxylation can regulate intracellular pH in acidic environments [1,18]. Cells containing more ATP showed a higher intracellular pH (pHi) and greater cell survival in pH 2.5 medium than cells with a low ATP concentration [23]. Amino acids could help maintain the intracellular ATP level and increase cell survival [23]. Proton consumption by hydrogenase-3 in E. coli has been implicated in the ability to survive under extremely acidic and anaerobic conditions [63]. ATP may have multiple effects on acid resistance, with the different ATPases hydrolyzing it and regulating the pHi under different conditions [2,25,26]. We previously reported on a DNA repair system which uses ATP as the substrate, and is necessary for E. coli survival at a very low pH [23]. While the mechanism by which ATP helps E. coli survive extremely acidic environments remains unknown, the importance of maintaining an elevated level while growing under such conditions is clear.
Our analysis revealed that other metabolic processes can also be affected by acidic environments. Apt was highly upregulated, suggesting that adenosine nucleotides may be synthesized more actively under these conditions. Interestingly, this is in line with our previous study showing that PurA and PurB are important for AR [23]. The expression of genes associated with biofilms (MinD) and quorom sensing (LuxS) was also increased (Supplementary Table S1) [64,65], indicating that these processes may also be affected by environmental pH. The expression level of gadX [66] was increased at a weakly acidic pH, which was regulated by RpoS (Supplementary Table S1).

Conclusions
In this study, we demonstrated that glk, pykF and pgk were necessary for the rise in ATP under weakly acidic conditions and for survival in markedly acidic environments (typically pH 1.5-3.5). In contrast, the metabolic pathways related to oxidative phosphorylation, fatty acid oxidation and energy derivation by oxidation of other organic compounds were downregulated. The inhibition of oxidative phosphorylation did not affect the ATP increase at pH 5.5. These results showed that glycolysis is the primary source of the elevated ATP levels seen in E. coli grown under weakly acidic conditions. Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4425/11/9/991/s1, Figure S1: The up-and down-regulated genes in amino acid metabolic pathway, Figure S2: The up-and down-regulated genes in the TCA cycle, Table S1: Folds change of differential expression genes between pH 7.5 and pH 5.5.
Author Contributions: Y.S. designed the study and wrote the manuscript. W.Z., W.S., and T.N. performed the experiments. X.C. analyzed the RNA-seq data. N.Q. revised the manuscript. All authors contributed to read and approved the revised manuscript.