Table 1 lists the primers, plasmids, and strains used in this work. The Green Industry Biotechnology Laboratory of Tsinghua University kindly provided the E. aerogenes wild-type strain IAM1183. Escherichia coli DH5α (TaKaRa Biotechnology, Dalian, China) was used for genetic engineering. Kits for the isolation of genomic DNA, DNA gel extraction, and plasmid DNA purification were from GenScript (GenScript USA Inc. United States). All DNA-modifying enzymes, restriction endonucleases, and DNA polymerase were bought from New England BioLabs (Beverly, MA, United States). Tryptone and yeast extract were purchased from Thermo Fisher Biochemicals (Beijing Ltd.). Unless specified otherwise, all other reagents used in this study were purchased from Sigma-Aldrich (St. Louis, MO, United States).

Luria–Bertani (LB) broth including (g/L) 10 tryptone, 5 yeast extract, and 10 NaCl was used to culture E. coli for cloning and cryopreservation. The cultivation medium consisted of several reagents including (g/L) 6.8 Na₂HPO₄, 3 KH₂PO₄, 0.75 KCl, 0.28 Na₂SO₄, 5.35 (NH₄)₂SO₄, 34.2 ZnCl₂, 0.26 MgSO₄·7H₂O, 0.42 citric acid, 10 Casamino acids, 5 yeast extract, and 0.3ml trace element solution containing (g/L) 2.7 FeCl₃·6H₂O, 10 0.85 CuCl₂·2H₂O, MnCl₂·4H₂O, and 0.31 H₃BO₃ (Jung et al., 2012) used to produce 2,3-BD. Glucose, sucrose, galactose, and fructose (40g/L) were used as carbon sources. The 50ml cultures in 250-ml Erlenmeyer flasks were closed using silicone stoppers and incubated micro-aerobically at 37°C and 250rpm. Kanamycin (50mg/ml), chloramphenicol (25mg/ml), or ampicillin (50mg/ml) was added to the medium, respectively, when necessary.

Fermentation experiments were performed in a 5-L bioreactor (Biostat A, Sartorius, Germany) with 3L of fermentation medium, at 250rpm and 37°C. The same medium was used for fermentation as the flask culture, and the carbon source was glucose. Chloramphenicol (50mg/ml) was supplied to promote plasmid retention. The fermentation was conducted at 37°C and 250rpm, with air supplied at 1vvm. The pH was maintained in the range from 6.5 to 7.0 by adding 5M NaOH (Jung et al., 2012).

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 50μl PCR reactions contained: PCR reaction buffer (5μl), dNTPs (10mM each), primers (10pmol each), 1U of Pfu DNA polymerase or Taq polymerase, and 10–20ng 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 0.6g/L l-arabinose (Wu et al., 2017).

The intracellular NADH/NAD⁺ ratio was determined using a method described in a previous study (Yang et al., 2015). The steady-state cells were harvested by centrifuging at 12,000rpm, 4°C for 10min. A colorimetric NAD/NADH Assay Kit (ab65348; Abcam, United Kingdom) was used to quantify the NADH/NAD⁺ in cell extracts.

Cell density was assessed by measuring the optical density of the cultures at 600nm (OD₆₀₀) 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×OD₆₀₀ (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 10ml of the fermentation cultures were centrifuged at 10,000× g and 4°C for 15min, 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 0.8ml/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 11.51min, respectively.

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 40g/L of glucose.

To measure 2,3-BD production, all strains were cultivated under aerobic conditions in 250-ml shake-flasks with 50ml of glucose medium. The parental strain IAM1183 served as the control. After 20h 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 6h, the constructed strains Ea-1 and Ea-2 showed corresponding phases that were ~4 and ~8h longer, respectively (Figure 1A). The pH of the culture broth of Ea-1 and Ea-2 decreased during 20h 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 (Yang et al., 2015). 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 70.64mM within 20h of cultivation, which was much higher than the 45.72mM 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.

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 OD₆₀₀, 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 ~4–6h later than the wild-type strain IAM1183, which reached the plateau at 6h (Figure 2A). During 20h 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 7.17mM lactate and 23.71mM 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.

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 40g/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 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 effects of knocking out nuoC, nuoD, and ldhA on the intracellular NADH/NAD⁺ ratio are shown in Figure 5. After 20h 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 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.

Ea-2 exhibited the highest 2,3-BD productivity in shake-flask fermentations among all the tested strains, with a yield from glucose reaching 1.01mol/mol (Figure 3). We therefore used Ea-2 for a scale-up experiment in a 5-L fermenter with 3L of glucose-containing medium. Ea-2 displayed a more pronounced exponential phase (4h longer than in IAM1183), reaching a 40% higher cell density after cultivation for 44h (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 44h 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.