Metabolic Engineering of Enterobacter aerogenes for Improved 2,3-Butanediol Production by Manipulating NADH Levels and Overexpressing the Small RNA RyhB

Biotechnological production of 2,3-butanediol (2,3-BD), a versatile platform bio-chemical and a potential biofuel, is limited due to by-product toxicity. In this study, we aimed to redirect the metabolic flux toward 2,3-BD in Enterobacter aerogenes (E. aerogenes) by increasing the intracellular NADH pool. Increasing the NADH/NAD+ ratio by knocking out the NADH dehydrogenase genes (nuoC/nuoD) enhanced 2,3-BD production by up to 67% compared with wild-type E. aerogenes. When lactate dehydrogenase (ldh) was knocked out, the yield of 2,3-BD was increased by 71.2% compared to the wild type. Metabolic flux analysis revealed that upregulated expression of the sRNA RyhB led to a noteworthy shift in metabolism. The 2,3-BD titer of the best mutant Ea-2 was almost seven times higher than that of the parent strain in a 5-L fermenter. In this study, an effective metabolic engineering strategy for improved 2,3-BD production was implemented by increasing the NADH/NAD+ ratio and blocking competing pathways.


INTRODUCTION
The biotechnological production of bulk chemicals is partially driven by the wish to reduce the reliance on fossil fuels because of limited reserves and increasing environmental concerns. The small molecule 2,3-BD is listed by the United States Department of Energy as a platform chemical with great potential for industrial applications (Song et al., 2019). The production of 2,3-BD by bacteria such as Escherichia sp. (Park et al., 2020), Bacillus sp. (Petrova et al., 2020), Klebsiella sp. (Sharma et al., 2018), and Enterobacter sp. (Thapa et al., 2019) has been investigated in depth in the last few years on account of its potential applications in the chemical industry for the production of polymers and solvents. As a versatile chemical feedstock and liquid fuel, 2,3-BD is extensively applied in the food, chemical, pharmaceutical, petrochemical, and aerospace industries (Ge et al., 2016). The development of biorefinery technology was further boosted due to concerns surrounding high energy costs and environmental problems (Ragauskas et al., 2006;Haveren et al., 2008). Most microbiological methods for the production of 2,3-BD rely on the biotransformation of expensive glucose. However, the utilization rate of glucose is still low, resulting in a pressing need for more efficient biological production methods for 2,3-BD.
Enterobacter aerogenes (syn. Klebsiella aerogenes) is a facultatively anaerobic Gram-negative bacterium that naturally produces 2,3-BD during fermentation at low O 2 concentrations. Due to its role in maintaining the intracellular redox balance of the pyridine nucleotide pool during glycolysis and biosynthesis, 2,3-BD was regarded as a classical end-product of anaerobic fermentation (Converti et al., 2003). The conversion of acetoin by 2,3-BD dehydrogenase regenerates NADH for continued glycolysis, and this reaction is increased to maintain the NAD + / NADH balance when there is a lack of oxygen (Blomqvist et al., 1993). Because NADH from glycolysis is regenerated via respiration under aerobic conditions, the switch from biomass synthesis to production of mixed acids and 2,3-BD is critical for maintaining the redox balance (Hung et al., 2011).
Many studies have demonstrated that small RNAs play critical roles in the metabolic regulation of both eukaryotes and bacteria (Beauchene et al., 2015;Machner and Storz, 2016). The small noncoding RNA RyhB undergoes base pairing with mRNAs to suppress gene expression (Massé et al., 2003). RyhB is related to a variety of crucial cellular functions such as iron homeostasis, antioxidant defenses, and TCA cycle activity in many bacteria, including Klebsiella pneumoniae and Escherichia coli (E. coli; Massé et al., 2003;Huang et al., 2012;Beauchene et al., 2015). In E. coli, formate dehydrogenase is downregulated by RyhB (Massé et al., 2005), suggesting that RyhB might regulate 2,3-BD production in E. aerogenes by suppressing formate production. Moreover, RyhB also influences the expression of NADH dehydrogenase I (Massé et al., 2005), which may influence the NAD + /NADH ratio during the production of 2,3-BD.
Enterobacter aerogenes can produce high titers of 2,3-BD from many different carbon sources, while producing different fermentation gases and soluble byproducts. Lactate is the main byproduct during the anaerobic fermentation process of E. aerogenes and blocking the lactate synthesis pathway can increase the yield of 2,3-BD due to their competition for NADH as a cofactor and pyruvate as a precursor. However, there are few reports on metabolic engineering in E. aerogenes. Nevertheless, a previous study improved hydrogen production in E. aerogenes by deleting a hydrogenase subunit and formate hydrogen lyase subunit, while also overexpressing a heterologous formate dehydrogenase and its positive regulator (Lu et al., , 2016Zhao et al., 2009). Similarly, Lu et al. (2010) investigated the effects of overexpressing formate hydrogenase and deleting lactate dehydrogenase (LDH) on hydrogen production. In a more recent study, CRISPR-Cas9 was used to knock out the nuoCD genes in the NADH dehydrogenase cluster of E. aerogenes IAM1183. The resulting double mutant Ea-1 (ΔnuoC/ΔnuoD) showed improved hydrogen production (Wu et al., 2017).
In this study, the gene ldh encoding LDH was deleted using the λ-Red recombination system to reduce byproduct accumulation, generating Ea-2 (ΔnuoC/ΔnuoD/∆ldh). To regulate the 2,3-BD production of IAM1183 via the small RNA RyhB, it was overexpressed in both strains Ea-1 and Ea-2, respectively, resulting in Ea-3 and Ea-4. We constructed a metabolic network model to quantify the metabolic fluxes of IAM1183 and its mutants and thereby analyze the changes in the metabolic pattern of the mutants. The best mutant exhibited a high 2,3-BD titer in both shake-flasks and batch fermentation, indicating its potential for scaled-up 2,3-BD production.
Fermentation experiments were performed in a 5-L bioreactor (Biostat A, Sartorius, Germany) with 3 L of fermentation medium, at 250 rpm and 37°C. The same medium was used for fermentation as the flask culture, and the carbon source was glucose. Chloramphenicol (50 mg/ml) was supplied to promote plasmid retention. The fermentation was conducted at 37°C and 250 rpm, with air supplied at 1 vvm. The pH was maintained in the range from 6.5 to 7.0 by adding 5 M NaOH (Jung et al., 2012). Genetic Manipulation Table 1 lists all primers used in this study. The ldh gene, which encodes LDH, was knocked out in the Ea-1 strain (E. aerogenes IAM1183 DnuoC/DnuoD) using a published λ-Red recombination method (Datsenko and Wanner, 2000), resulting in strain Ea-2 (E. aerogenes IAM1183 DnuoC/DnuoD/Dldh). Genomic DNA was isolated from E. aerogenes IAM1183 according to a published method (Miller, 1972) or using a commercial kit. The QIAprep Spin Miniprep kit and GenScript DNA gel extraction kit were used for plasmid isolation and DNA purification, respectively. The 50 μl PCR reactions contained: PCR reaction buffer (5 μl), dNTPs (10 mM each), primers (10 pmol each), 1 U of Pfu DNA polymerase or Taq polymerase, and 10-20 ng of the DNA template were conducted with a Bio-Rad thermocycler (Hercules, CA, United States). The correct products were confirmed by sequencing (Life Technologies, Shanghai Invitrogen, China). The purified plasmid pKK102-ryhB-cm, which was used to overexpress the small RNA RyhB, was introduced into Ea-3 by electroporation, and the resulting strain was named Ea-4 (E. aerogenes IAM1183 DnuoC/DnuoD/Dldh [pKK102-ryhB-cm]). RyhB expression was induced by adding 0.6 g/L l-arabinose (Wu et al., 2017).

NADH/NAD + Assay
The intracellular NADH/NAD + ratio was determined using a method described in a previous study .
The steady-state cells were harvested by centrifuging at 12,000 rpm, 4°C for 10 min. A colorimetric NAD/NADH Assay Kit (ab65348; Abcam, United Kingdom) was used to quantify the NADH/NAD + in cell extracts.

Analytical Methods
Cell density was assessed by measuring the optical density of the cultures at 600 nm (OD 600 ) using a UV-721G spectrophotometer. The dry cell weight (DCW) of E. aerogenes was subsequently calculated using the empirical formula: DCW (mg/ml) = 0.132 × OD 600 (Zhao et al., 2015). Analytes in the gas space, including 2,3-BD and ethanol, were measured using a GC-2010 Plus gas chromatography apparatus (Shimadzu, Japan) equipped with a thermal conductivity detector at 40°C and a Parapak Q column at 200°C, with He as the carrier gas. Samples comprising 10 ml of the fermentation cultures were centrifuged at 10,000× g and 4°C for 15 min, and the supernatant was then harvested for analysis of metabolites and residual sugar. The concentrations of formate, acetate, lactate, succinate, citrate, acetoin, and ethanol in the culture supernatant were determined by HPLC (LC-20A; Shimadzu, Japan), using a PREP-ODS (H) KIT C-18 column (Shimadzu) kept at 33°C, and a refractive index detector (RID-10A, Shimadzu). The mobile phase was composed of 0.2% aqueous phosphoric acid at a flow rate of 0.8 ml/min. Formate, ethanol, lactate, acetate,  citric acid, and 2,3-BD had retention times of 4.42, 5.12, 5.38, 5.67, 6.49, and 11.51 min, respectively.

RESULTS AND DISCUSSION
Effects of Knocking Out the nuoC, nuoD, and ldh Genes on 2,3-BD Production To examine the relationship between NADH dehydrogenase activity, 2,3-BD synthesis, and carbon flux distribution, CRISPR/ Cas9 precise editing technology (Wu et al., 2017) was used to delete the nuoC and nuoD genes, encoding the two subunits of NADH dehydrogenase, resulting in the E. aerogenes IAM1183 derivative Ea-1 ( Table 1). To increase the production of 2,3-BD, the ldh gene was further deleted using λ-Red recombination to limit lactate production in Ea-1, resulting in strain Ea-2 ( Table 1). To study the effects of the mutations on metabolite production and carbon source consumption, the strains were cultured in medium containing 40 g/L of glucose.
To measure 2,3-BD production, all strains were cultivated under aerobic conditions in 250-ml shake-flasks with 50 ml of glucose medium. The parental strain IAM1183 served as the control. After 20 h of cultivation, the cell density of both Ea-1 and Ea-2 was greater than that of IAM1183 ( Figure 1A). Compared with the exponential phase of wild-type IAM1183, which lasted 6 h, the constructed strains Ea-1 and Ea-2 showed corresponding phases that were ~4 and ~8 h longer, respectively ( Figure 1A). The pH of the culture broth of Ea-1 and Ea-2 decreased during 20 h of batch fermentation at a lower rate than that of IAM1183 ( Figure 1B). These results demonstrated that the deletion of nuoC, nuoD, and ldh reduced the acidification rate of Ea-1 and Ea-2, resulting in a higher plateau phase of the mutants (Figure 1). NADH availability and its reduction potential (the intracellular reduced NADH pool) has a stronger effect on the overall yield of 2,3-BD . Ji et al. (2010) constructed a Klebsiella oxytoca strain that does not accumulate 2,3-BD by deleting the aldehyde dehydrogenase gene aldA.
In this study, the 2,3-BD titer of Ea-1 reached 70.64 mM within 20 h of cultivation, which was much higher than the 45.72 mM of the wild type (Table 2), indicating that the removal of NADH dehydrogenase significantly improved 2,3-BD production in E. aerogenes IAM1138. As shown in Table 2, the double-knockout strain Ea-2 was obviously a superior 2,3-BD producer, with a 84.6% higher product titer than that of the wild-type IAM1183. Furthermore, the 2,3-BD yield from glucose improved 61% compared with the wild-type strain ( Table 2). Additionally, the increased growth rate of the Ea-2 mutant also led to an increase in 2,3-butanediol production. It was reported that when LDH is eliminated, the increased NADH availability and redistribution of carbon fluxes lead to an increase in 2,3-BD production (Jung et al., 2012). Metabolically engineered microbes often have lower growth rates than the wild type. The increased microbial growth of the double mutants in this study might be the result of reduced acidification of the medium. Guo et al. (2020) constructed a mutant K. pneumoniae strain by deleting three genes, ldhA, adhE (encoding alcohol dehydrogenase), and pta (encoding phosphotransacetylase), resulting in significantly increase the yield of 2,3-BD. Currently, undesirable byproducts (such as lactate, ethanol, and acetate) are also produced during the 2,3-BD fermentation process, which shunts the carbon flow and dramatically decreases the efficiency of 2,3-BD synthesis. Thus, blocking pathways that competitively consume NADH to redistribute carbon flux will be more conducive to the 2,3-BD synthesis.
Effect of Overexpressing the sRNA RyhB on 2,3-BD Production The plasmid pKK102-ryhB-cm, which overexpresses the small RNA RyhB, was introduced into Ea-1 and Ea-2 to generate Ea-3 and Ea-4, respectively. Based on the OD 600 , the final cell densities of Ea-3 and Ea-4 were lower than that of E. aerogenes IAM1183 by 13 and 24%, respectively (Figure 2A). Moreover, the two mutants Ea-3 and Ea-4 entered the plateau phase ~4-6 h later than the wild-type strain IAM1183, which reached the plateau at 6 h (Figure 2A). During 20 h of shake-flask cultivation, the pH of the Ea-3 and Ea-4 cultures decreased more slowly than that of IAM1183 ( Figure 2B). However, the final pH of the Ea-3 and Ea-4 cultures was lower than that of strains Ea-1 and Ea-2, respectively, implying that overexpressing the sRNA RyhB had an effect on acid production, which inhibited the growth of the strain (Figure 2).
We also found that overexpression of RyhB caused significant decrease in 2,3-BD production (18.3%) in Ea-3 compared with Ea-1, and a modest decrease in Ea-4 (14.6%) compared with Ea-2 when grown on glucose (Table 2; Figure 3). A previous study demonstrated that overexpression of RyhB could improve 2,3-BD production under anaerobic conditions, while we found that RyhB had no effect on aerobic 2,3-BD production (Wu et al., 2017). In E. coli, formate dehydrogenase is downregulated by RyhB under aerobic conditions (Massé et al., 2005). Indeed, the formate production of Ea-3 and Ea-4 was significantly lower than that of Ea-1 and Ea-2 (Table 2; Figure 3). Among the by-products, Ea-3 produced 7.17 mM lactate and 23.71 mM acetate, representing 39.5 and 79.9% increases compared to Ea-1, respectively. Furthermore, Ea-4 produced more acetate and lactate than Ea-2 (Table 2; Figure 3). The resulting acidification of the Ea-3 and Ea-4 cultures may explain their lower 2,3-BD titer. Interestingly, the production of acetate by Ea-3 was increased 2.67-fold compared with Ea-1, while Ea-4 produced 27% more acetate than Ea-2. Similarly, Kang et al. (2012) reported that overexpression of RyhB led to a threefold increase in acetate accumulation in E. coli, which indicated that RyhB had a comprehensive influence on the central glucose metabolism of E. aerogenes.

Effects of Different Carbon Sources on the Mutants
Enterobacter aerogenes can effectively utilize many different carbon sources (Perego et al., 2000). To detect the effects of different carbon sources on the mutants, shake-flask fermentations were conducted using 40 g/L of glucose, fructose, sucrose or galactose (sections "Effects of Knocking Out the nuoC, nuoD, and ldh Genes on 2,3-BD Production" and "Effect of Overexpressing the sRNA RyhB on 2,3-BD Production") With galactose and fructose, the production of 2,3-BD by Ea-1 was significantly increased, while the carbon source consumption rate was modestly decreased compared with IAM1183 (Table 3 Figure 4A), which was presumably caused by the lower NADH dehydrogenase activity in the two constructed strains, which conserved energy during the exponential growth phase. The concentrations of other fermentation byproducts, such as formate and acetate, all decreased >20% in Ea-1. However, there was a significant increase in lactate production, which, respectively, increased 28.9, 41.8, and 79.6% compared with IAM1183 with all carbon sources except for sucrose (Table 3), which was detrimental for the productivity and yield of 2,3-butanediol ( Figure 4A). When grown on galactose and fructose, the 2,3-BD production and carbon source utilization rate of Ea-2 also increased ( Table 3; FIGURE 3 | Metabolite profiles of IAM1183 and the mutants Ea-1 to -4 after 20 h of shake-flask cultivation with glucose as carbon source (n = 3). The data shown in Table 2 were used for the calculations. Figure 4A). Similar to glucose, these carbon sources are metabolized via the glycolysis pathway. Due to the likely presence of a secondary LDH or a side activity of an unrelated enzyme, Ea-2 still produced a small amount of lactate. The large decrease in lactate accumulation had positive impacts to 2,3-BD productivity, which, respectively, increased by 17.5 and 4.3% compared with Ea-1 on galactose and fructose (Table 3; Figure 4A). The slower decrease in the culture pH in the double mutant likely led to the increase in productivity (Booth, 1985). The double mutant showed an improvement in the consumption rates of all carbon sources. Notably, obvious ethanol accumulation was observed in the double mutant with sucrose as the carbon source, and this increase was accompanied by a reduced 2,3-BD titer. Glucose directly enters the glycolysis pathway and does not require any additional energy to convert it into a usable form (Öhgren et al., 2006). This explains why glucose was the most efficient carbon source, since sucrose requires an additional enzyme and energy input in order for it to be converted into glucose and processed in glycolysis. These processes may lead to the production of more by-products such as ethanol (Bonestroo et al., 1992).
The effect of overexpressing the sRNA RyhB was tested in four cultures with various carbon sources ( Table 4). The Ea-3 and Ea-4 strains showed a reduction in carbon source consumption, 2,3-BD production (Figure 4B), and yield with all carbon sources compared with Ea-1 and Ea-2, respectively. The accumulation of acidic by-products such as lactate and acetate reduced the 2,3-BD productivity (Table 4; Figure 4B). RyhB was reported  to have an even stronger effects on intermediary metabolism than on the TCA cycle enzymes, downregulating many different Fe-S-containing metabolic enzymes (Massé et al., 2005;Beauchene et al., 2015), which may explain the reduction in 2,3-BD production in Ea-3 and Ea-4. sRNAs are important regulators of gene expression and physiology in bacteria. RyhB is an iron-responsive sRNA well characterized in E. coli and conserved in other Enterobacteriaceae (Lillian et al., 2021). Zhang et al. (2018) have demonstrated that increased ATP levels were observed in the ryhB-knockout mutant of E. coli, which indicated that the decrease in ATP level has an impact on the lower 2,3-BD production of ryhB-overexpress mutant in our study. Interestingly, we found the 2,3-BD production of all mutants on sucrose was significantly decreased compared with the wild type (Figure 4), which was potentially caused by a redox imbalance due to the disruption of NADH dehydrogenase when sucrose was used as substrate (Jung et al., 2014).

The Intracellular NADH/NAD + Ratios of Different Strains
The effects of knocking out nuoC, nuoD, and ldhA on the intracellular NADH/NAD + ratio are shown in Figure 5. After 20 h of shake-flask cultivation, the NADH/NAD + ratios of strains Ea-1 and Ea-2, respectively, increased by 25.7 and 39.6% compared with the parental strain. This suggested that the redox balance of strains Ea-1 and Ea-2 was disrupted, which also explained the causes that significantly affect cell growth (Figure 1). As an intrinsic component of the respiratory chain, the mitochondrial NADH: ubiquinone oxidoreductase (complex I) mediates the transmembrane translocation of protons e in the first step of intracellular or mitochondrial NADH oxidation (or alternatively, NAD + reduction; Vinogradov, 2008). The knockout of nuoC and nuoD, encoding complex I, increased the NADH/NAD + ratio of Ea-1 and redistributed metabolic fluxes to increase 2,3-BD production. Similarly, a previous study showed that deletion of LDH increased NADH availability (Jung et al., 2012). In this study, the molar yield of 2,3-BD and NADH/NAD + ratio of Ea-2 was, respectively, 23 and 11% higher than in Ea-1 (Figure 5; Table 2). This suggests that the deletion of ldh also made additional NADH available for other pathways, especially the biosynthesis of 2,3-BD. The strain Ea-3 and Ea-4 showed significant downregulation in NADH/ NAD + ratio when compared with the Ea-1 and Ea-2, respectively ( Figure 5), which suggested that sRNA RyhB can significantly affect cell metabolism in addition to its role as a regulator of gene expression. Interestingly, a decrease in NADH/NAD + ratios can be observed upon ryhB activation in E. coli (Lyu et al., 2019). Thus, RyhB is not merely a genetic-level regulator, its function can also have profound impacts on cell metabolism. Calderón et al. (2013) reported that the NADH/NAD + ratio was higher in a ryhB-deletion strain than in the parental wild FIGURE 7 | Diagram of primary aerobic metabolic pathways for bacterial 2,3-BD production. Shown are three putative strategies to enhance 2,3-BD yield as outlined in this study.
Frontiers in Microbiology | www.frontiersin.org type-strain Salmonella typhimurium. We found that the NADH/ NAD + ratios of Ea-3 and Ea-4 were, respectively, decreased by 11.8 and 14.2% relative to Ea-1 and Ea-2 ( Figure 5). These results support a role of RyhB in regulating the NADH/NAD + ratio, suggesting that the redox balance of the E. aerogenes mutants was disturbed. A similar observation was made in E. coli by Lyu et al. (2019), which explains why RyhB can significantly affect cell metabolism in addition to its negative effect on 2,3-BD production. Since RyhB acts as a global regulator of glucose metabolism (Kang et al., 2012), it stands to reason that since several of the targets of RyhB encode proteins of the TCA cycle, the overexpression of RyhB could result in lower levels of NADH in E. aerogenes.

2,3-BD Production by Ea-2 in a Fermenter
Ea-2 exhibited the highest 2,3-BD productivity in shake-flask fermentations among all the tested strains, with a yield from glucose reaching 1.01 mol/mol (Figure 3). We therefore used Ea-2 for a scale-up experiment in a 5-L fermenter with 3 L of glucose-containing medium. Ea-2 displayed a more pronounced exponential phase (4 h longer than in IAM1183), reaching a 40% higher cell density after cultivation for 44 h (Figures 6C,F). More glucose was consumed by Ea-2 during the fermentation, which resulted in an almost 7-fold higher 2,3-BD titer compared with the wild type (Figures 6C,F). There was also an increase in acetate production (Figures 6A,D), while the lactate accumulation rate was clearly reduced in Ea-2 (Figures 6B,E).
The final titers of the major byproducts after 44 h of fermentation in the 5-L bioreactor were measured for both Ea-2 and WT, as summarized in Table 5. The WT exhibited higher final concentrations of formate and lactate, while the concentration of ethanol decreased. This was in agreement with the shake-flask fermentations. The mutants generated in this work offer a solid basis for further studies on the fermentative metabolism of E. aerogenes, with great application potential for industrial 2,3-BD production.

CONCLUSION
There are three metabolic pathways for the production of 2,3-BD in E. aerogenes (Figure 7). The present work demonstrates that the deletion of NADH dehydrogenase (ΔnuoC/DnuoD strain Ea-1), alone or in combination with the deletion of LDH (ΔnuoC/DnuoD/Dldh strain Ea-2), can significantly improve the 2,3-BD yield of E. aerogenes (Figure 7).
However, overexpression of the sRNA RyhB reduced the NADH/ NAD + ratio and 2,3-BD production (Figure 7). Metabolic flux analysis of the parental strain IAM1183 and mutants grown on four different carbon sources indicated that the impairment of dehydrogenase is beneficial for 2,3-BD production due to a reduction in acidic by-products, and also indicated that RyhB could redistribute metabolic fluxes in E. aerogenes. The 2,3-BD titer produced by the mutant strain Ea-2 in a 5-L fermenter exhibited an almost 7-fold improvement over the wild type. Hence, Ea-2 may be a basis for further improvement of 2,3-BD production via metabolic engineering, offering hope to finally achieve biotechnological 2,3-BD production on an industrial scale.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
HZ conceived and supervised the research, analyzed the results, and reviewed and edited the manuscript. HZ and YW designed the experiments. YW, QS, WC, and JY performed the investigations. YX, QS, HY, and FX validated the data. YW, WC, YL, PL, and KJ analyzed the results. YW wrote the original draft of the manuscript. WC modified the manuscript. All authors contributed to the article and approved the submitted version.