DsrA Modulates Central Carbon Metabolism and Redox Balance by Directly Repressing pflB Expression in Salmonella Typhimurium.

Bacterial small RNAs (sRNAs) function as vital regulators in response to various environmental stresses by base pairing with target mRNAs. The sRNA DsrA, an important posttranscriptional regulator, has been reported to play a crucial role in defense against oxidative stress in Salmonella enterica serovar Typhimurium, but its regulatory mechanism remains unclear. The transcriptome sequencing (RNA-seq) results in this study showed that the genes involved in glycolysis, pyruvate metabolism, the tricarboxylic acid (TCA) cycle, and NADH-dependent respiration exhibited significantly different expression patterns between S. Typhimurium wild type (WT) and the dsrA deletion mutant (ΔdsrA strain) before and after H2O2 treatment. This indicated the importance of DsrA in regulating central carbon metabolism (CCM) and NAD(H) homeostasis of S. Typhimurium. To reveal the direct target of DsrA action, fusion proteins of six candidate genes (acnA, srlE, tdcB, nuoH, katG, and pflB) with green fluorescent protein (GFP) were constructed, and the fluorescence analysis showed that the expression of pflB encoding pyruvate-formate lyase was repressed by DsrA. Furthermore, site-directed mutagenesis and RNase E-dependent experiments showed that the direct base pairing of DsrA with pflB mRNA could recruit RNase E to degrade pflB mRNA and reduce the stability of pflB mRNA. In addition, the NAD+/NADH ratio in WT-ppflB-pdsrA was significantly lower than that in WT-ppflB, suggesting that the repression of pflB by DsrA could contribute greatly to the redox balance in S. Typhimurium. Taken together, a novel target of DsrA was identified, and its regulatory role was clarified, which demonstrated that DsrA could modulate CCM and redox balance by directly repressing pflB expression in S. Typhimurium. IMPORTANCE Small RNA DsrA plays an important role in defending against oxidative stress in bacteria. In this study, we identified a novel target (pflB, encoding pyruvate-formate lyase) of DsrA and demonstrated its potential regulatory mechanism in S. Typhimurium by transcriptome analysis. In silico prediction revealed a direct base pairing between DsrA and pflB mRNA, which was confirmed in site-directed mutagenesis experiments. The interaction of DsrA-pflB mRNA could greatly contribute to the regulation of central carbon metabolism and intracellular redox balance in S. Typhimurium. These findings provided a better understanding of the critical roles of small RNA in central metabolism and stress responses in foodborne pathogens.

INTRODUCTION

Salmonella is a Gram-negative, facultative anaerobe and generalist pathogen that is capable of causing bacteremia, gastroenteritis, and systemic infection in a range of different hosts (1). During its infectious cycle, Salmonella is recognized by macrophages, neutrophils, and dendritic cells, all of which internalize and comprise the bacterium in a Salmonella-containing vacuole (SCV), where Salmonella faces several stresses, including imbalanced reactive oxygen/nitrogen species (ROS/RNS), low pH, and iron deficiency (2, 3).

Oxidative stress response is essential to microbial adaptability and pathogenicity (4). An elaborate ROS defense system enables bacteria to survive in a multitude of environmental stresses (5). These defense mechanisms include scavenging enzymes involved in detoxification of reactive radicals and repair enzymes used for restoring cellular physiology (6, 7). NADH, generated primarily in the tricarboxylic acid (TCA) cycle, plays an important role in cell synthesis and detoxifying ROS. Conversely, NADH is the source of energy generation in the respiratory chain reactions, which produce ROS. Thus, bacteria need to redistribute the flux of central carbon metabolism (CCM) to rebalance redox reaction by modulating the multilevel regulation machinery (8). For instance, CCM could generate more reduced equivalent (NADH) to reduce oxidative stress by modulating the fluxes (8). Also, many enzymes involved in CCM, such as succinate dehydrogenase (Sdh), fumarate reductase (Frd), and aconitase (Acn), contain flavins or iron-sulfur clusters for univalent redox reactions (7, 9). Without this protection, a number of crucial cellular components in Salmonella, including DNA, RNA, proteins, and membrane lipids, would be structurally or functionally deterred from augmentation of ROS (10–12).

Small RNAs (sRNAs) have significant effects on gene expression by reinforcing transcriptional regulation and providing links between different regulatory modules (13). Bacterial sRNAs have been reported to play an important role in response to various environmental stresses, such as, carbon starvation (14), virulence (15, 16), acid stress (17), oxidative stress (18, 19), and antibiotics (20, 21). Since most sRNAs bind to their mRNA targets by imperfect complementarity, they are often found to regulate multiple genes and rewire complex regulatory networks (13, 22). However, even with multiple validated targets, sRNA usually has much more target candidates based on prediction. These candidates, under given conditions, are found not to interact with the sRNA directly (23, 24). This raises the notion that some of them could be regulated by the sRNA under alternative conditions.

In our previous work, DsrA was found to play an important role in oxidative stress resistance of Salmonella enterica serovar Typhimurium (25). However, the regulatory mechanism of DsrA in defending against oxidative stress remains unclear. To reveal the target of DsrA action, in this study, we performed a whole-genome transcriptome analysis, characterized the in silico predicted DsrA targets, and investigated the direct interaction between DsrA and its targets by site-directed mutagenesis.

RESULTS

Overview of RNA-seq.

In pursuit of the biological role of DsrA, gene expression before and after a 30-min H2O2 treatment was compared in the ΔdsrA strain and in the WT as well. Among 12 samples, a range of 4,042 to 4,119 genes per sample were detected, which accounted for 71.8% to 73.3% of the whole genome (5,623 genes) in S. Typhimurium. In the H2O2-treated WT compared to an untreated sample, 1,389 genes were significantly (P < 0.05) differentially expressed, including 597 upregulated genes and 792 downregulated genes (Fig. 1A). In the ΔdsrA strain after the treatment, 530 and 586 genes were significantly (P < 0.05) up- and downregulated, respectively (Fig. 1B).

FIG 1

Overview of RNA-seq. (A) Pie graph of DEGs in wild type. (B) Pie graph of DEGs in the ΔdsrA strain. (C) Venn diagram showing DEGs in wild type (WT) and the ΔdsrA strain (MT). Genes with a fold change of ≥2 and Bonferroni-corrected P value (Padj) of <0.05 were determined to be DGEs.

Comparative analysis of differentially expressed genes (DEGs) in the WT and ΔdsrA strain was shown in Fig. 1C. It was found that seven genes (srlA, srlE, slrB, tdcB, nrdD, treC, and sdaC) were downregulated in WT while upregulated in the ΔdsrA strain, and an additional 346 genes were exclusively downregulated in WT but showed no apparent changes in the ΔdsrA strain, suggesting that DsrA might negatively regulate the expression of these genes. However, two genes (ypeC and clpA) were upregulated in WT but downregulated in the ΔdsrA strain, and 276 genes were exclusively upregulated in WT while not changed in the ΔdsrA strain, suggesting that DsrA might positively regulate the expression of this set of genes.

The DEGs were also subjected to functional categorization using the KEGG database (see Fig. S1 in the supplemental material). It was found that there were more upregulated and less downregulated genes in both the carbohydrate metabolism pathway and global metabolism pathway in the ΔdsrA strain than those in WT (Fig. S1). The differences of DEG profile in these pathways between WT and mutant suggested DsrA might have a nonnegligible role in metabolic regulation in S. Typhimurium.

Validation of RNA-seq data via RT-qPCR.

The H2O2-induced changes in transcript expression were validated by quantitative real-time PCR (RT-qPCR). A number of representative genes were randomly selected from the differentially expressed genes identified in transcriptome sequencing (RNA-seq). In WT, six genes (iroB, sufA, soxR, dps, ahpF, and katG) with increased transcript abundances and four genes (nuoH, srlA, bssR, and tdcB) with decreased transcript abundances were selected for RT-qPCR analysis. In the ΔdsrA strain, 10 differentially transcribed genes, including six upregulated (sufA, soxR, dps, srlA, tdcB, and ahpF) and four downregulated (cadA, lamB, bssR, and pduA) genes were selected for the same purpose. The RT-qPCR results showed a good correlation with the RNA-seq data in both WT (R = 0.997, P < 0.0001) and the ΔdsrA strain (R = 0.979, P < 0.0001) (Fig. 2), supporting the reliability and validity of the RNA-seq data.

FIG 2

RT-qPCR validation of RNA-seq data for selected differentially expressed genes. The relative transcription of genes found to be differentially regulated in the RNA-seq analysis of the wild-type strain (A) or ΔdsrA strain (B) after H2O2 treatment was examined by RT-qPCR. 16S rRNA was used as a reference gene. Mean log2 fold change (FC) in the transcription of genes in six independent RT-qPCR experiments were plotted against the respective log2 FC determined by RNA-seq. The coefficient of determination (R) and P value were calculated in Microsoft Excel.

Regulatory effect of DsrA on central carbon metabolism.

The expression of the genes related to NAD-dependent oxidation in CCM, including glycolysis, tricarboxylic acid (TCA) cycle, and pyruvate metabolism, was analyzed in this study. Figure 3A shows the genes with H2O2-altered transcript expression in the ΔdsrA strain or WT involved in glycolysis. After H2O2 treatment, the pyruvate kinase gene (pykA) was downregulated (2.15-fold) in WT but slightly upregulated (1.31-fold) in the ΔdsrA strain. Another pyruvate kinase gene (pykF) was transcribed higher in both WT (2.82-fold) and the ΔdsrA strain (6.84-fold) after H2O2 treatment, whereas upregulated magnitude was substantial in the ΔdsrA strain (Fig. 3A). The ackA-pta operon, which converts acetyl coenzyme A (acetyl-CoA) to acetate (Fig. 4), was upregulated in WT (3.31-fold) but not changed in the ΔdsrA strain. Similar to pykF, adhE encoding alcohol dehydrogenase was expressed higher in both WT (2.43-fold) and the ΔdsrA strain (4.58-fold) after H2O2 treatment (Fig. 4). The upregulated magnitude in the ΔdsrA strain was more significant.

FIG 3

Heat map of genes involved in the major metabolic pathways in the wild-type and ΔdsrA strains after H2O2 treatment. (A) Glycolysis. (B) Pyruvate metabolism. (C) TCA cycle. (D) NADH-dependent respiration. Each row represents an individual gene. Log2 FC of genes were labeled in the corresponding grid. The scale of this heat map is given as log2 FC ranging from −7 (blue) to +7 (red).

FIG 4

Biochemical pathways involved in NAD(H) cycle in S. Typhimurium. The red star represents the verified target of DsrA in this study. Black stars represent the predicted proteins that could interact with PflB.

Our results showed that the expressions of aceE, aceF, pdhR, and lpdA, which were involved in pyruvate metabolism (Fig. 4), were significantly upregulated in both WT (19.45-, 17.93-, 11.91-, and 2.02-fold, respectively) and the ΔdsrA strain (15.32-, 10.02-, 2.34-, and 2.70-fold, respectively) after H2O2 treatment, indicating pyruvate was actively metabolized under oxidative stress (Fig. 3B). The expression of pflB was downregulated (2.72-fold) in WT while slightly upregulated (1.49-fold) in the ΔdsrA strain after H2O2 treatment (see Fig. 3B and Fig. 6A). In contrast to pflB, the expression of ldhA was significantly increased (5.32-fold) in WT and slightly decreased (1.43-fold) in the ΔdsrA strain after the treatment, suggesting the coupling between NADH production and consumption in the pyruvate-to-lactate and pyruvate-to-formate reactions (Fig. 4).

Under H2O2 treatment, the TCA cycle activity was generally repressed in WT, while the genes involved in the TCA cycle showed different trends of expression in the ΔdsrA strain (Fig. 3C). Notably, acnA encoding aconitate hydratase (Fig. 4), which is involved in oxidative stress response, showed a significantly high expression (3.72-fold) in WT but minor change in the ΔdsrA strain. In addition, the isocitrate dehydrogenase gene (icdA) expressed slightly higher (1.13-fold) in WT and lower (1.45-fold) in the ΔdsrA strain. The expression of sucB encoding α-ketoglutarate dehydrogenase was not changed in WT but significantly increased (2.54-fold) in the ΔdsrA strain (Fig. 4). Intriguingly, sucC (3.01-fold) and sucD (2.99-fold) showed a similar expression pattern to sucB, most likely because sucCD and sucB are in the same operon (Fig. 4).

Regulatory effect of DsrA on cell respiration.

NADH generated from the TCA cycle can be oxidized by NADH dehydrogenase (NDH) in the respiratory chain reactions (Fig. 4). The expression of genes encoding NDH was analyzed in this study. The ndh gene encoding NDH-2 showed an increasing expression both in WT (7.2-fold) and the ΔdsrA strain (6.17-fold). In addition, the entire nuo operon encoding NDH-1 was downregulated in WT after H2O2 treatment, of which nuoHIJKLMN had a significantly lower expression (Fig. 3D). Despite sharing the same operon, the nuo operon in the ΔdsrA strain showed different expression patterns with slightly decreased nuoABC but increased nuoE-nuoN transcripts after H2O2 treatment (Fig. 3D).

Predicted mRNA targets of DsrA in S. Typhimurium.

To find out the target of DsrA, comparative analysis of the whole-genome expression in WT and the ΔdsrA strain was performed, and in silico prediction was run in the CopraRNA and IntaRNA webserver (26). Six predicted genes (acnA, srlE, tdcB, nuoH, katG, and pflB) were selected as target candidates based upon the output of the query. For preliminary evaluation of the predictions, expressions of the candidate genes in the presence and absence of DsrA were compared by RT-qPCR analysis. As shown in Fig. 5A, the mRNA levels of pflB, nuoH, srlE, and tdcB were significantly higher in the ΔdsrA strain or ΔdsrA-pZE0 strain than those in WT or the ΔdsrA-pdsrA strain, suggesting the negative regulatory effect of DsrA on the four genes. On the contrary, the expression of acnA and katG in the ΔdsrA strain or ΔdsrA-pZE0 strain was significantly lower than that in WT or the ΔdsrA-pdsrA strain, indicating the positive regulation of DsrA on these two genes.

FIG 5

The regulatory effect of DsrA on six target candidates. (A) The relative expression of pflB, nuoH, srlE, tdcB, acnA, and katG in the wild-type, ΔdsrA, ΔdsrA-pZE0, and ΔdsrA-pdsrA strains was determined by RT-qPCR. 16S rRNA was used as a reference gene. The expression of genes in the wild type was regarded as 1. (B) Fluorescence measurement of target expression fusion with GFP. The 5′ UTRs or intergenic regions of candidate target mRNAs were cloned into pXG-10sf or pXG-30sf vectors, respectively. GFP expression in the LB liquid medium was quantified by a microplate reader and normalized by OD600. pZE0 and pXG1 were used as negative control plasmids for DsrA expression and target gene fusion plasmid, respectively. hns::gfp was constructed as the positive control for DsrA regulation. **, P < 0.01

To further evaluate the prediction results, six target gene fusions were constructed by using an established two-plasmid reporter system as previously described (27, 28), in which DsrA was expressed in the pZE12-luc plasmid and a predicted mRNA target was expressed as an sfGFP-fusion protein in the pXG reporter plasmid. pZE0 expressing a 10-nt nonsense nucleotide and pXG1 were used as the negative control plasmids for sRNA expression and mRNA target expression fusion, respectively. The fusion of hns::gfp was also constructed in parallel as the positive control since hns is a well-known target repressed by DsrA (28). Of the six genes tested, the fluorescence of pflB::gfp in the ΔdsrA-pdsrA strain was significantly lower than that in the ΔdsrA-pZE0 strain, indicating that DsrA could repress the expression of pflB (Fig. 5B).

Evidence for DsrA-pflB mRNA interaction by direct base pairing.

Based on in silico prediction, DsrA can base pair with the mRNA sequence of pflB to form a 12-bp RNA-RNA duplex (Fig. 6A and C). This base pairing was experimentally confirmed by mutating the nucleotide sequences in DsrA and pflB::gfp to disrupt the pairing and by compensating these mutations to restore the pairing (Fig. 6C). Overexpression of a mutant variant of DsrA (pdsrA*) carrying 2-nucleotide substitutions in the predicted base-pairing region resulted in failure to repress the expression of wild-type pflB::gfp fusion (Fig. 6D). Consistently, introducing a compensatory mutation in the pflB::gfp (pflB*) by a mutation completely abolished the repression by wild-type DsrA. However, the compensatory mutation was efficiently suppressed by overexpression of pdsrA* (Fig. 6D). Taken together, these results suggested that translational repression of pflB by DsrA was mediated by direct base pairing.

FIG 6

The regulatory effect of DsrA on pflB. (A) Alignment of pflB gene from various Salmonella species. Asterisks indicate the conserved nucleotides. The RBS region was marked in green. The start codon was marked in red. The RNase E sites were marked in blue. The DsrA-pflB interaction regions were highlighted. The symbol “:-:” indicates a 165-bp insertion. The symbol “::” indicates a 40-bp insertion. (B) Heat map of pflB expression in the wild type and ΔdsrA strains before and after H2O2 treatment. (C) Predicted interactions of Salmonella DsrA with pflB mRNA. Mutated nucleotides were indicated by red letters. The numbers represented the position of nucleotides in the pflB CDS region from start codon or DsrA RNA from transcription start site. (D) DsrA regulates the pflB mRNA by base pairing mechanism. The ΔdsrA strain was transformed by combinations of pXG plasmids along with control plasmid (pZE0), DsrA expression plasmid (pdsrA), or the DsrA mutant expression plasmid (pdsrA*) as indicated. GFP expression in the LB liquid medium was quantified by a microplate reader and normalized by OD600. (E) Regulatory effect analysis of DsrA on pflB in the RNase E thermosensitive rne-3071 strain (rne-Ts) and the control allele strain (rne-Ctr) at 37°C. Bacterial cells were grown at 37°C to an OD600 of 0.3 and then treated with 3 mM H2O2 incubated at 37°C for 30 min, followed by total RNA extraction and gene expression analysis. (F) Regulatory effect analysis of DsrA on pflB in the RNase E thermosensitive strain rne-3071 (rne-Ts) and the control allele strain (rne-Ctr) at 44°C. Bacterial cells were grown at 30°C to an OD600 of 0.3 and then treated with 3 mM H2O2 incubated at 44°C for 30 min, followed by total RNA extraction and gene expression analysis. 16S rRNA was used as a reference gene. The expression of pflB in the rne-Ctr wild type or rne-Ts wild type was regarded as 1. Error bars indicate standard deviations (n = 3). Statistical significance is as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, non-significant.

The repression effect of DsrA on pflB was RNase E dependent.

Alignment of pflB gene sequences revealed its high conservation among various Salmonella species (Fig. 6A). Notably, the sequence AAATT around the DsrA-pflB mRNA interaction site matched a consensus motif for RNase E in Salmonella (RN↓WUU) (Fig. 6A) (29). As a result, we then checked whether DsrA affected pflB mRNA in an RNase E-dependent manner using a thermosensitive RNase E mutant (rne-Ts) (30, 31). The dsrA deletion mutant (ΔdsrA strain), dsrA wild-type expression strain (ΔdsrA-pdsrA strain), and dsrA site mutant expression strain (ΔdsrA-pdsrA* strain) of rne-Ctr strain and rne-Ts strain were constructed, respectively, and the expressions of pflB in different strains were analyzed. As shown in Fig. 6F, the negative regulatory effect of DsrA on pflB was found in the rne-Ctr strain; however, it was abrogated in the rne-Ts strain at 44°C where RNase E was inactivated. Intriguingly, there were no significant differences in pflB expression between the ΔdsrA-pdsrA and ΔdsrA-pdsrA* strains of the rne-Ctr strain at 44°C.

In order to be consistent with the RNA-seq condition, the pflB expression analyses in the rne-Ctr strain and rne-Ts strain was also performed under 37°C, where we thought that RNase E activity might be partially inhibited. As shown in Fig. 6E, pflB expression in the Ctr-ΔdsrA or Ctr-ΔdsrA-pZE0 strains was prominently higher than that in the Ctr-WT or Ctr-ΔdsrA-pdsrA strain. In addition, the pflB expression in the Ctr-ΔdsrA-pdsrA* strain was also significantly higher than that in the Ctr-ΔdsrA-pdsrA strain, which indicated DsrA could directly repress the expression of pflB in the presence of RNase E. However, pflB expression showed no considerable differences among the Ts-ΔdsrA, Ts-ΔdsrA-pZE0, Ts-ΔdsrA-pdsrA, and Ts-ΔdsrA-pdsrA* strains. These results suggested that the repression effect of DsrA on pflB was RNase E dependent. The base-pairing of DsrA might recruit RNase E to degrade pflB mRNA and reduce the stability of pflB mRNA.

Effect of the repression of pflB by DsrA on the redox balance in Salmonella.

To evaluate the physiological effect of DsrA on pflB, a pflB-overexpression strain (WT-ppflB) and a pflB-dsrA cooverexpression strain (WT-ppflB-pdsrA) were constructed. The plasmid pHM1 was used as the negative control plasmid for pflB expression. The NAD+/NADH ratio was measured in the WT, ΔdsrA, ΔdsrA-pZE0, ΔdsrA-pdsrA, WT-pHM1, WT-ppflB, and WT-ppflB-pdsrA strains. As shown in Fig. 7A, both the ΔdsrA and ΔdsrA-pZE0 strains exhibited a considerably higher NAD+/NADH ratio than WT, with a more marked effect in the presence of H2O2, suggesting that the redox balance could be altered in the ΔdsrA strain and severely perturbed under oxidative stress. Likewise, the NAD+/NADH ratio was significantly higher in the WT-ppflB strain than that in WT or WT-pHM1, regardless of the presence or absence of H2O2, indicating the redox imbalance in the WT-ppflB strain. It was found that PflB could interact with several proteins related to NAD-dependent reaction, which echoed the role of pflB in disturbed redox balance (Fig. 7B). However, the WT-ppflB-pdsrA strain showed a significantly lower NAD+/NADH ratio than the WT-ppflB strain, indicating that overexpression of DsrA could prominently improve the redox balance in the WT-ppflB strain. Taken together, the repression of DsrA on pflB significantly contributed to the redox balance in S. Typhimurium.

FIG 7

The effect of DsrA-pflB mRNA interaction on the redox balance in S. Typhimurium. (A) Intracellular NAD+/NADH ratio of S. Typhimurium before and after H2O2 treatment. (B) STRING software prediction of the PflB-protein interaction network. PflB was highlighted in orange. The proteins involved in central carbon metabolism were highlighted in green. Error bars indicate standard deviations (n = 3). Statistical significance is as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, non-significant.

DISCUSSION

DsrA regulates central carbon metabolism.

sRNA-mediated fine-tuning within metabolic pathways is another major functional theme in enterobacteria because they encounter many different metabolic niches inside and outside of their hosts (32, 33). Modulation of CCM is crucial for cells to rebalance the redox ratio and keep ROSs at harmless levels (8). Our RNA-seq study showed that the number of downregulated genes involved in metabolism in the ΔdsrA strain after H2O2 treatment was much less than that in the WT (see Fig. S1 in the supplemental material), which implied an increased metabolic activity in the ΔdsrA strain under oxidative stress compared to WT. For example, the increased expression of pykA and pykF in the ΔdsrA strain indicated a high activity of pyruvate kinase and an increased pyruvate synthesis in the ΔdsrA strain under oxidative stress. Under oxidative stress conditions, pyruvate usually accumulates by activating pyruvate kinase Pyk (34) and undergoes nonenzymatic decarboxylation with ROSs to produce acetate (35). Conversely, the “high-activity” pyk led to increased growth but decreased oxidative stress resistance in yeast (36), from which we can infer that “high-activity” of Pyk could decrease the resistance of the ΔdsrA strain to oxidative stress.

Salmonella possesses the following two main pyruvate dissimilation pathways: (i) conversion to acetyl-CoA through dehydrogenase (AceE, AceF, and PdhR), dihydrolipoamide dehydrogenase (LpdA), and pyruvate-formate lyase (PflB) with formate as a by-product; and (ii) conversion to lactate by lactate dehydrogenase (LdhA) (Fig. 4). Significant upregulation of aceE, aceF, lpd, pta, and ackA in WT suggested that more pyruvate was involved in the nonenzymatic reaction to eliminate ROS and produce acetate (Fig. 4). Conversely, AdhE catalyzes the conversion of acetyl-CoA to ethanol, accompanied by the production of two molecules of NAD+ (Fig. 4). The pflB and adhE genes were significantly upregulated in the ΔdsrA strain, while pta-ackA had little change. This could lead to the following: (i) inefficient conversion of pyruvate to acetyl-CoA coupled with the production of formate as a by-product, and/or (ii) efficient conversion of acetyl-CoA to ethanol coupled with the production of oxidative equivalent NAD+, both of which were not conducive to the ΔdsrA strain in defense against oxidative stress.

The TCA cycle is usually repressed under oxidative stress to increase the formation of α-ketoglutarate (37, 38). α-Ketoglutarate is another ketoacid that can relieve oxidative stress by nonenzymatic reactions (35). The expression patterns of icdA and sucB genes found here suggested an accumulation of α-ketoglutarate in WT and a degradation of α-ketoglutarate in the ΔdsrA strain under oxidative stress, which would lead to less resistance to oxidative stress in the ΔdsrA strain than in the WT.

Broad spectrum of DsrA-target interaction.

Similar to transcriptional factors, sRNA can simultaneously target multiple mRNAs (39). It was reported that Escherichia coli could express approximately 300 sRNAs and develop over 2,000 unique sRNA-mRNA target interactions (40–42). Salmonella is also able to express approximately 300 sRNAs (42, 43), which implies that Salmonella might also harbor a complex network of sRNA-mRNA interactions. The secondary structure of DsrA contains three stem loops, denoted as SL1, SL2, and SL3, and a long linker between SL1 and SL2 (44). The conformational flexibility makes DsrA more accessible to multiple mRNA binding. Currently, five targets have been determined for DsrA, namely, rpoS, hns, mreB, rbsD, and lrp, among which only rbsD is targeted in the coding DNA sequence (CDS) region far downstream of the start codon, and the other four are targeted in the region near the start codon (44–48).

Targeting the CDS region of mRNA to induce mRNA degradation is an important mechanism for small RNA regulation (49–51). Many sRNAs have been reported to function in this way, such as the CoaR/tcpI system, GcvB/asnA system, and MicC/ompD system (16, 50, 52). sRNA MicC binds the ompD mRNA far downstream of the ribosome binding site (RBS) (+67 to +78 nucleotide from the start codon) and induces RNase E cleavage at the +83 position of ompD mRNA in conjunction with Hfq (52). The tricomplex of sRNA, Hfq, and RNase E binds the target sequence via the MicC “seed sequence,” since RNase E-mediated mRNA degradation requires specific structural features containing an A/U-rich sequence and an adjacent secondary structure (53–55). In this study, the sequence “AAATT” downstream of the DsrA-pflB mRNA interaction site matched a consensus motif for RNase E in Salmonella (RN↓WUU) (29). The DsrA-pflB duplex and the “AAATT” sequence provided conditions for RNase E digestion. In addition, an AAN-repeat sequence “AACAAA” was found upstream of the DsrA binding site in pflB mRNA, which was consistent with the distal face-binding motif of Hfq (A-A-N)n and might allow Hfq to bind the mRNA to recruit RNase (56, 57). All discussed above indicated that the binding of DsrA with pflB could promote the formation of the DsrA-Hfq-RNase E complex and induce the RNase E cleavage on pflB mRNA.

Among six target candidates in this study, only pflB::gfp was repressed by DsrA, which highlighted the strength and weakness of in silico prediction. Site-directed mutation analysis was performed to explore the interaction between DsrA and pflB. Intriguingly, the inhibition of pflB was not completely abolished by mutation of DsrA, and the regulation of pflB* by DsrA* was also just partially restored, which implied that DsrA could not only directly regulate pflB expression but also indirectly regulate pflB expression through other means (Fig. 6D and E). In addition, pflB expression in the Ctr-ΔdsrA-pdsrA* strain was significantly higher than that in the Ctr-ΔdsrA-pdsrA strain at 37°C, while showing no significant differences compared with that in the Ctr-ΔdsrA-pdsrA strain at 44°C (Fig. 6E and F). Similarly, it was reported that most sRNA partners in RIL-seq data would not be predicted by CopraRNA, or their gene expressions were not affected when tested with reporter fusions (58), indicating that sRNAs, including DsrA, could regulate specific genes under specific conditions.

DsrA regulated the redox balance by repressing pflB in S. Typhimurium.

pflB is the second gene in the focA-pflB operon that encodes a metabolic pathway converting pyruvate to acetyl-CoA with formate as a by-product (Fig. 4) (59). The evolution of PflB is important for bacteria to tolerate oxidative stress (7). We found that the pflB gene was highly conserved in various Salmonella species (Fig. 6A) but less conserved among different microbial species, such as Salmonella and E. coli (data not shown), which may result from the evolutionary differences among species. Consistent with our results, a double deletion mutant of pflB and ldhA in E. coli showed an extremely low intracellular NAD+/NADH ratio, and this NAD+/NADH imbalance was mainly caused by the inactivation of PFL rather than the inactivation of LDH (60–62), which highlighted the importance of PflB in maintaining cellular redox homeostasis.

Several proteins (i.e., LdhA, AceE, AceF, Pta, AckA, and AdhE) in pyruvate metabolism pathway were predicted by STRING software to interact with PflB (Fig. 7B). This could explain, to some extent, the difference in pyruvate metabolic flux between the WT and ΔdsrA strains (Fig. 4 and 8). AdhE was reported to regulate the activity of pyruvate formate lyase (PFL) (63). Thus, to inhibit PflB activity in the ΔdsrA strain, cells might activate the expression of AdhE. The high expression of AdhE would also generate excess NAD+ and enhance the imbalance of redox reaction (Fig. 8). If the expression of dsrA was restored, the above circumstances would be improved (Fig. 8). Notably, the increased magnitude of the NAD+/NADH ratio in the ΔdsrA strain compared to WT was much higher than that in the WT-ppflB strain (Fig. 7A), indicating the more significant effect of DsrA on redox homeostasis than pflB. As an important posttranscriptional regulator, DsrA might act at the center of the regulatory network. It could simultaneously interact with multiple targets and regulate the redox balance of S. Typhimurium through multiple mechanisms in addition to repression of pflB.

FIG 8

Proposed model for the regulation of NAD(H) homeostasis by DsrA-pflB mRNA interaction. Red arrows represent the metabolic pathways with increased flux. Blue arrows represent the metabolic pathways with reduced flux.

In this study, the critical role of DsrA in regulating CCM and NAD(H) homeostasis in S. Typhimurium was elucidated by transcriptome analysis. A novel target of DsrA, pflB encoding pyruvate-formate lyase, was identified from the studies of PflB-GFP fusion proteins in the presence and absence of DsrA. Direct base pairing between DsrA and pflB mRNA sequences was predicted in silico and confirmed by mutational analysis. Moreover, phenotype analysis of intracellular redox homeostasis showed that the interaction between DsrA-pflB mRNA could greatly contribute to the redox balance in S. Typhimurium (Fig. 8).

MATERIALS AND METHODS

Bacterial strains and growth conditions.

S. Typhimurium strain ATCC14028 is referred to as wild type (WT) and was used for mutant construction. Bacterial cells were grown at 37°C with reciprocal shaking at 220 rpm in LB broth. Where appropriate, the medium was supplemented with antibiotics at the concentrations of 100 μg/mL ampicillin (Amp), 34 μg/mL chloramphenicol (Cm), 40 μg/mL spectinomycin (Sm), and 100 μg/mL hygromycin B (HYG).

For H2O2 treatment, a single colony was inoculated into fresh LB broth and grown at 37°C with rotation at 200 rpm until an optical density at 600 nm (OD600) of 1.0 was achieved. Each culture was then inoculated at a dilution of 1:100 into 5 mL fresh LB and grown to mid-log phase. This was followed by the treatment of 3 mM H2O2 for 30 min. After that, the cells were collected for subsequent experiments.

Strain construction.

The ΔdsrA and ΔdsrA-pdsrA strains were obtained from the same source as indicated in our previous work (25). Plasmid pZE0, which expresses a 10-nt nonsense RNA, was the negative control vector for sRNA expression plasmids. All of the strains used in this study are summarized in Table 1. A complete list of plasmids and oligonucleotides is included in Tables S1 and S2 in the supplemental material.

TABLE 1

All strains used in this study

Strain	Description	Plasmid(s)	Reference
Wild type	S. Typhimurium strain ATCC14028
ΔdsrA strain	dsrA-deletion strain		25
ΔdsrA-pZE0 strain	Negative control of ΔdsrA-pdsrA strain	pZE0	This study
ΔdsrA-pdsrA strain	The complemented strain of dsrA	pZE-dsrA	25
ΔdsrA-pdsrA* strain	dsrA mutant (dsrA*) overexpressing in ΔdsrA strain	pZE-dsrA*	This study
WT-pHM1	Negative control of WT-ppflB	pHM1	This study
WT-ppflB	pflB overexpressing strain	pHM-pflB	This study
WT-ppflB-pdsrA	pflB and dsrA overexpressing strain	pHM-pflB, pZE-dsrA	This study
pZE0-phns::gfp	hns::gfp expression fusion in ΔdsrA-pZE0 strain	pXG-hns::gfp; pZE0	This study
pdsrA-phns::gfp	hns::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-hns::gfp, pZE-dsrA	This study
pZE0-pacnA::gfp	acnA::gfp expression fusion in in ΔdsrA-pZE0 strain	pXG-acnA::gfp	This study
pdsrA-pacnA::gfp	acnA::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-acnA::gfp, pZE-dsrA	This study
pZE0-pnuoH::gfp	nuoH::gfp expression fusion in ΔdsrA-pZE0 strain	pXG-nuoH::gfp; pZE0	This study
pdsrA-pnuoH::gfp	nuoH::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-nuoH::gfp, pZE-dsrA	This study
pZE0-psrlE::gfp	srlE::gfp expression fusion in ΔdsrA-pZE0 strain	pXG-srlE::gfp; pZE0	This study
pdsrA-psrlE::gfp	srlE::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-srlE::gfp, pZE-dsrA	This study
pZE0-ptdcB::gfp	tdcB::gfp expression fusion in ΔdsrA-pZE0 strain	pXG-tdcB::gfp; pZE0	This study
pdsrA-ptdcB::gfp	tdcB::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-tdcB::gfp, pZE-dsrA	This study
pZE0-pkatG::gfp	katG::gfp expression fusion in ΔdsrA- pZE0 strain	pXG-katG::gfp; pZE0	This study
pdsrA-pkatG::gfp	katG::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-katG::gfp, pZE-dsrA	This study
pZE0-ppflB::gfp	pflB::gfp expression fusion in ΔdsrA-pZE0 strain	pXG-pflB::gfp; pZE0	This study
ΔdsrA-pdsrA-ppflB::gfp	pflB::gfp expression fusion in ΔdsrA-pdsrA strain	pXG-pflB::gfp, pZE-dsrA	This study
ΔdsrA-pdsrA*-ppflB::gfp	pflB::gfp expression fusion in ΔdsrA-pdsrA* strain	pXG-pflB::gfp, pZE-dsrA*	This study
pZE0-ppflB*::gfp	pflB mutant (pflB*) expression fusion in ΔdsrA- pZE0 strain	pXG-pflB*::gfp; pZE0	This study
pdsrA-ppflB*::gfp	pflB mutant (pflB*) expression fusion in ΔdsrA-pdsrA strain	pXG-pflB*::gfp, pZE-dsrA	This study
pdsrA*-ppflB*::gfp	pflB mutant (pflB*) expression fusion in ΔdsrA-pdsrA* strain	pXG-pflB*::gfp, pZE-dsrA*	This study
pZE0-pXG1	GFP control vector in ΔdsrA-pZE0 strain	pXG1; pZE0	This study
ΔdsrA-pdsrA-pXG1	GFP control vector in ΔdsrA-pdsrA strain	pXG1; pdsrA	This study
ΔdsrA-pdsrA*-pXG1	GFP control vector in ΔdsrA-pdsrA* strain	pXG1; pdsrA*	This study
rne-Ctr-WT	Salmonella Typhimurium SL1344 (rluC-rne) IG::cat		30, 31
rne-TS-WT	Salmonella Typhimurium SL1344 (rluC-rne) IG::cat rne-3071 (TS)		30, 31
rne-Ctr-ΔdsrA	dsrA-deletion strain of rne-Ctr		This study
rne-Ctr-ΔdsrA-pZE0	dsrA-deletion strain of rne-Ctr harboring pZE0	pZE0	This study
rne-Ctr-ΔdsrA-pdsrA	The complemented strain of dsrA of rne-Ctr	pZE-dsrA	This study
rne-Ctr-ΔdsrA-pdsrA*	dsrA mutant (dsrA*) overexpressing in ΔdsrA strain of rne-Ctr	pZE-dsrA*	This study
rne-TS-ΔdsrA	dsrA-deletion strain of rne-TS		This study
rne-TS-ΔdsrA-pZE0	dsrA-deletion strain of rne-TS harboring pZE0	pZE0	This study
rne-TS-ΔdsrA-pdsrA	The complemented strain of dsrA of rne-TS	pZE-dsrA	This study
rne-TS-ΔdsrA-pdsrA*	dsrA mutant (dsrA*) overexpressing in ΔdsrA strain of rne-TS	pZE-dsrA*	This study

DsrA deletion strains of rne-Ctr and rne-TS were constructed using λ-Red recombinase one-step inactivation method (64). Briefly, the upstream fragment of dsrA, HYG-resistant fragment, and the downstream fragment of dsrA were linked by overlapping PCR to produce the ΔdsrA::hph fragment, followed by gel purification. Then 100 ng of ΔdsrA::hph fragments was transformed into Salmonella wild-type carrying the pKD46 helper plasmid. Phage22 was used to transduce the ΔdsrA::hph fragment to the rne-Ctr WT or rne-Ts WT strain (65). The correct insertions of the HYG marker gene were verified by PCR using dsrA-F1&R2.

To construct the pflB overexpressing strain, the entire CDS region of pflB was amplified using the primer set of pflB-F&R followed by gel purification. The PCR products and pHM1 vector were digested using SalI/EcoRI. The digested products were ligated for 1 h at 22°C to generate a recombinant plasmid pHM-pflB. Subsequently, pHM-pflB was transferred into the WT strain and selected on Sm plates to yield the pflB overexpressing strain (WT-ppflB). To construct the pflB and dsrA double overexpressing strain, pflB overexpressing plasmid (pHM-pflB) and dsrA overexpressing plasmid (pZE-dsrA) were simultaneously transformed into WT and selected on the plates containing Amp and Sm to yield the WT-pdsrA-ppflB strain.

RNA-seq sample preparation.

Cells of WT, ΔdsrA, and ΔdsrA-pdsrA strains before and after H2O2 treatment were harvested, 2/5 volumes of the ice-cold “stop solution” (19% ethanol and 1% acidic phenol [pH 4.3]) was added into the cell suspension. After incubating on ice for 30 min to prevent RNA degradation (66), total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol.

cDNA library construction and RNA-seq.

The total RNA from the WT and ΔdsrA strains were treated with specific biotinylated oligonucleotides to delete rRNA. A strand-specific cDNA library was constructed using Illumina TruSeq stranded mRNA kit. The obtained cDNA libraries were sequenced using Illumina HiSeq 4000 by the BGI Group (Shenzhen, Guangdong, China). After that, the adaptors were removed, and the quality of RNA-seq data was assessed using SOAP software (67). High-quality reads were mapped to the genome of Salmonella strain ATCC14028s using HISAT software (68). Relative expression of each individual gene was calculated by the number of fragments per kilobase of transcript per million mapped reads (FPKM) in each sample using Bowtie2 and RSEM software packages (69, 70). The DESeq2 (71) method was used to calculate the differentially expressed genes (DEGs). The genes with a fold change of ≥2 and Bonferroni-corrected P value (Padj) of <0.05 were determined to be DGEs.

RT-qPCR.

Total RNA extracted from H2O2-treated and -untreated samples was reverse-transcribed into cDNA using the PrimeScript RT reagent kit with gDNA Eraser (TaKaRa, Dalian, China). Subsequently, the mRNA levels of the genes were determined by quantitative real-time PCR (RT-qPCR). In parallel, 16S rRNA was included as a reference for normalization of the gene expression data. Results were analyzed using the comparative critical threshold method (2−ΔΔCT) (72). If the level of relative gene expression is increased or decreased by at least 2-fold, the altered expression is generally regarded as upregulation or downregulation (73).

Construction of gene expression fusion.

A gene-gfp fusion was constructed as described previously (27, 28). For example, a DNA fragment containing the last 25 codons of the focA gene, 5′ untranslated region (UTR) and CDS region of the pflB gene was PCR amplified using the primer set of pflBg-F&R. After gel purification and NsiI/NheI digestion, the digested fragment was ligated with pXG30 for 1 h at 22°C to generate a recombinant plasmid pXG-pflB. The pXG-pflB plasmid was subsequently transferred into the ΔdsrA-pZE0 strain and the complemented ΔdsrA-pdsrA strain, and then selected on Amp/Cm plates respectively. As a positive control, hns::gfp was also constructed in parallel since it is well known that DsrA represses the expression of hns (28). The empty plasmid pXG-1 was used as a negative control and transferred in ΔdsrA-pZE0 and ΔdsrA-pdsrA strains, respectively.

GFP fluorescence quantification.

Single colonies of the ΔdsrA strain constructs harboring GFP fusions and DsrA expression plasmids were inoculated in triplicates into 1 mL LB broth containing Amp and Cm and grown overnight at 37°C. One hundred microliters of overnight culture was dispensed into each well of 96-well optical bottom black microtiter plates for the measurement of OD600 and fluorescence (excitation at 476 nm and emission at 510 nm, using an emission cutoff filter of 495 nm) using Tecan i-control (Molecular Devices, Austria).

Intracellular NAD+/NADH quantification.

Intracellular concentrations of NAD+ and NADH in WT, ΔdsrA, ΔdsrA-pZE0, ΔdsrA-pdsrA, WT-pHM1, WT-ppflB, and WT-ppflB-pdsrA strains were measured using an NAD+/NADH cell-based assay kit (Beyotime Biotechnology, Shanghai, China). Briefly, cells treated and untreated with 3 mM H2O2 for 30 min were harvested by centrifugation at 12,000 × g for 5 min at 4°C and resuspended in an extracting buffer to obtain the total NAD+ and NADH, named NADHTotal. A portion of NADHTotal was heated at 60°C for 30 min to degrade NAD+. Thus, only NADH remained in the sample, named NADHNADH. The rest of NADHTotal was placed at 37°C for 10 min to convert all NAD+ into NADH. Cell debris was removed from the extract by centrifugation at 12,000 × g for 10 min at 4°C. The amount of NADH in NADHTotal and NADHNADH samples was measured separately. According to a range of NADH standards (0 to ∼200 pM), NADH concentrations in the cell extracts were determined.

Sequence alignments.

Nucleotide sequences of the pflB gene from various Salmonella samples were downloaded from the NCBI database using the following genomes: S. Typhimurium 14028S (CP001363), Salmonella enterica serovar Typhi Ty2 (AE014613), Salmonella enterica serovar Enteritidis SE81 (CP050721), Salmonella enterica serovar Indiana JT 01 (CP028131), Salmonella enterica serovar Newport VNSEC031 (CP039436), Salmonella enterica serovar Derby CVM 30155 (CP053048), Salmonella enterica serovar Infantis VNSEC002 (CP039443), Salmonella enterica serovar Pullorum CFSAN022642 (CP075018), Salmonella enterica serovar Dublin USMARC-69838 (CP032449), and Salmonella enterica serovar Kentucky PU131 (CP026327). Alignments were generated with ClustalW software.

Data analysis.

All the experiments were technically repeated at least three times with three biological replicates per assay. Statistical analysis of the triplicate data set was performed using the two-tailed unpaired t test (α = 0.05) by Microsoft Excel 2016 (Microsoft Inc., Redmond WA, USA).

Data availability.

Gene expression data have been deposited with NCBI Gene Expression Omnibus (GEO) under accession number GSE180425.