2,3-Butanediol synthesis from glucose supplies NADH for elimination of toxic acetate produced during overflow metabolism

Overflow metabolism-caused acetate accumulation is a major problem that restricts industrial applications of various bacteria. 2,3-Butanediol (2,3-BD) synthesis in microorganisms is an ancient metabolic process with unidentified functions. We demonstrate here that acetate increases and then decreases during the growth of a bacterium Enterobacter cloacae subsp. dissolvens SDM. Both bifunctional acetaldehyde/ethanol dehydrogenase AdhE-catalyzed ethanol production and acetate-induced 2,3-BD biosynthesis are indispensable for the elimination of acetate generated during overflow metabolism. 2,3-BD biosynthesis from glucose supplies NADH required for acetate elimination via AdhE-catalyzed ethanol production. The coupling strategy involving 2,3-BD biosynthesis and ethanol production is widely distributed in bacteria and is important for toxic acetate elimination. Finally, we realized the co-production of ethanol and acetoin from chitin, the second most abundant natural biopolymer whose catabolism involves inevitable acetate production through the coupling acetate elimination strategy. The synthesis of a non-toxic chemical such as 2,3-BD may be viewed as a unique overflow metabolism with desirable metabolic functions.


Introduction
Under aerobic conditions, fast-growing organisms use fermentation instead of respiration for energy generation and excrete large quantities of acetate 1 . Recently, acetate overflow in Escherichia coli (E. coli) was identified as a result of proteome allocation to balance the conflicting proteomic demands of biomass synthesis and energy generation 2 . Acetate accumulation inhibits microbial growth, and acetate overflow is recognized as one of the major problems limiting the industrial applications of various bacterial strains [3][4][5][6] . Biotechnologists have attempted for more than 20 years to avoid overflowcaused acetate accumulation, but have achieved little success 7 .
2,3-Butanediol (2,3-BD) is a bacterial metabolite produced from pyruvate via several intermediates including α-acetolactate, acetoin (AC), and diacetyl 8,9 . The ability to generate 2,3-BD is widely distributed in bacteria, and the Voges-Proskauer reaction, which is related to intermediates of 2,3-BD production, is a fundamental test for bacterial classification 10 . 2,3-BD biosynthesis from pyruvate involves condensation of pyruvate to AC, and the reduction of AC to 2,3-BD, a step that requires an extra NADH. Thus, the production of 2,3-BD was proposed to prevent acidification by changing an acidic molecule to a neutral compound 10 and to consume excess reducing power 11 . However, 2,3-BD biosynthesis from glucose does not involve proton consumption and produces an extra NADH, which is contrary to the above theory. Until now, the physiological function of 2,3-BD biosynthesis in bacteria has not been clarified 12 .
In this study, a naturally existing strategy for toxic acetate elimination involving 2,3-BD biosynthesis and ethanol production was identified in Enterobacter cloacae subsp. dissolvens SDM (E. cloacae SDM). 2,3-BD biosynthesis from glucose was confirmed to supply the reducing power required for ethanol generation and acetate elimination. Based on the identified mechanism for acetate elimination, E. cloacae SDM was metabolically engineered to transform acetate produced during the Nacetylglucosamine (GlcNAc, the monomer of chitin) metabolism into ethanol. Co-production of ethanol and AC from chitin, the second most abundant natural biopolymer after cellulose, was realized.

Results
Acetate accumulates and then vanishes during the growth of E. cloacae SDM E. cloacae, a Voges-Proskauer-positive strain with the ability to produce 2,3-BD (Fig. 1a), is widely distributed in gut and rhizosphere soil 13 . 2,3-BD released by E. cloacae was reported to promote the growth of Arabidopsis thaliana 14,15 . In this study, E. cloacae SDM was firstly aerobically cultured with glucose as the carbon source. Similar to E. coli, E. cloacae SDM underwent acetate overflow metabolism, and acetate accumulated at the beginning of its growth (2 h) (Fig. 1b). Metabolic flux analysis indicated that 55% of utilized glucose was metabolized to acetate by E. cloacae SDM (Supplementary Fig.  S1a). Thereafter, acetate vanished (2−6 h) and carbon was redirected into 2,3-BD (73%) and ethanol (15%) ( Fig. 1b;  Supplementary Fig. S1b). To further identify the catabolic fate of acetate, 13 C-labeled sodium acetate (~10 mM) was added into the culture system of E. cloacae SDM at 2 h and the variation of 13 C-labeled acetate ( 13 C 1 -CH 2 COOH) was detected. Our results showed that exogenous 13 Clabeled acetate also vanished, and most of the 13 C-labeled acetate was reduced to 13 Table S1).
Interestingly, the production of ethanol also decreased, and acetate became the major metabolite of E. cloacae SDM (ΔbudABC) ( Fig. 1d; Supplementary Fig. S1e, f). Elimination of exogenously added 13 C acetate by these two mutant strains was also assayed. E. cloacae SDM (ΔbudABC) and E. cloacae SDM (ΔadhE) could not metabolize most of the 13 C-labeled acetate to 13 C-labeled ethanol as E. cloacae SDM (Fig. 1e-g), further supporting that 2,3-BD biosynthesis and ethanol production participate in the elimination of acetate.

2,3-BD synthesis is induced for acetate elimination
The expression of budA and adhE during the growth of E. cloacae SDM was assayed by quantitative real-time PCR (qPCR). Expression of adhE was initiated at the beginning of growth (2 h), and budA was induced at 4 h and then repressed when acetate was consumed (6−10 h) ( Supplementary Fig. S2a, b). When acetate was added to the medium, the expression of budA, but not adhE, was enhanced ( Fig. 2a, b; Supplementary Fig. S2c, d).
In Bacillus subtilis, the expression of the alsSD operon for AC (precursor of 2,3-BD) formation was regulated by a LysR-type transcriptional regulator AlsR [17][18][19] . BudR, a homolog of AlsR in B. subtilis, was annotated upstream of BudABC in the genome of E. cloacae SDM. Deletion of budR decreased the expression of budA, and both 2,3-BD production and acetate elimination were blocked in E. cloacae SDM (ΔbudR) (Supplementary Fig. S2e−g). However, there was no distinction in the expression of adhE between E. cloacae SDM (ΔbudR) and the wild-type strain (Fig. 2c). The expression of adhE was controlled by FNR, a global regulator in E. coli 20 . Mutation of FNR in E. cloacae SDM sharply reduced the expression of adhE but had no influence on the expression of budA (Fig. 2c). Thus, although both 2,3-BD biosynthesis and ethanol production are indispensable for acetate elimination in E. cloacae SDM, the expression of each of these two pathways is likely to be independent of the other, and only 2,3-BD synthesis was specially induced by acetate.
2,3-BD synthesis from glucose provides NADH for acetate elimination 2,3-BD synthesis is indispensable for acetate elimination, which involves the production of AdhE-catalyzed ethanol from acetyl-CoA. Independent regulation of budABC and adhE supports that the role of 2,3-BD synthesis in acetate elimination may provide a particular component required for the AdhE-catalyzed reaction. Since AdhE catalyzes NADH-mediated conversion of acetyl-CoA to ethanol, effects of 2,3-BD synthesis and ethanol production on the NADH/NAD + ratio were assayed. As shown in Fig. 2d, the NADH/NAD + ratio in E. cloacae SDM (ΔbudABC) decreased, while it increased in E. cloacae SDM (ΔadhE). Thus, both synthesis of 2,3-BD and the production of ethanol influenced the intracellular redox state. 2,3-BD synthesis provided NADH while ethanol production consumed NADH.
The above phenomenon led us to hypothesize that 2,3-BD synthesis is able to supply reducing power for AdhEcatalyzed acetate elimination. Further experiments involving redox state perturbation under aerobic conditions support our hypothesis. Decrease in the NADH/NAD + ratio through the expression of water-forming NADH oxidase (NOX) from Lactobacillus brevis CICC 6004 9 or using gluconic acid (an oxidation product of glucose) as a substrate reduced the efficiency of acetate elimination (Fig. 2e, f; Supplementary Fig. S2h−j). Glycolysis of one mole of glucose generates two moles of NADH and two moles of pyruvate. One mole of 2,3-BD synthesis consumes two moles of pyruvate and one mole of NADH. From the perspective of overall glucose metabolism, one mole of 2,3-BD synthesis may supply one mole of NADH. As shown in Fig. 2e, f, increasing the NADH/NAD + ratio by blocking BudC, which catalyzes the NADH-requiring reduction of AC to produce 2,3-BD, also improved the efficiency of acetate elimination. This result indicates that the partial pathway ending at AC is stronger than the 2,3-BD synthesis for NADH generation.
NADH is mainly generated by glycolysis and oxidized by fermentative metabolism under anaerobic conditions 21 . Synthesis of pyruvate dehydrogenase would be repressed in the absence of oxygen 22 . Formate-lyase (PFL) becomes the main source of acetyl-CoA under anaerobic condition 23 . Formate dehydrogenase (FDH) catalyzed Fig. 1 2,3-BD biosynthesis and ethanol production participate in the elimination of acetate produced by overflow metabolism. a Reactions related to ethanol production and 2,3-BD biosynthesis in E. cloacae SDM. AdhE, bifunctional acetaldehyde/ethanol dehydrogenase; BudB, αacetolactate synthase; BudA, α-acetolactate decarboxylase; BudC, 2,3-butanediol dehydrogenase; AC, acetoin; 2,3-BD, 2,3-butanediol. b Growth of E. cloacae SDM using glucose as carbon source. c Growth of E. cloacae SDM (ΔadhE) using glucose as carbon source. d Growth of E. cloacae SDM (ΔbudABC) using glucose as carbon source. e Consumption of exogenous added 13 C acetate by E. cloacae SDM and its derivatives. f Production of 13 C ethanol by E. cloacae SDM and its derivatives. g Yield of 13 C ethanol from 13 C acetate by E. cloacae SDM and its derivatives. 13 C sodium acetate at a concentration about 10 mM was added into the culture system of E. cloacae SDM, E. cloacae SDM (ΔadhE), and E. cloacae SDM (ΔbudABC) at 2 h. After 4 h of culture, the concentrations of 13 C ethanol and 13 C acetate were assayed. Experiments were carried out under aerobic conditions. Data shown are means ± SD (n = 3 independent experiments). ***P < 0.001 in two-tailed Student's t-test.
dehydrogenation of formate might generate extra NADH and influence the redox balance. However, there is no protein with a similarity higher than 30% to FDH of Candida boidinii in the genome of E. cloacae SDM, and no FDH activity was detected in the crude extract of E. cloacae SDM (Supplementary Fig. S3). Anaerobic ATP generation through glycolysis, PFL, and AdhE can convert 1 mol of glucose to 2 mol ethanol with an excess demand of 2 mol NADH. The production of 1 mol of AC from glucose could produce 2 mol of excess NADH, which means that the biosynthesis of one AC molecule supported the production of two ethanol molecules. Likewise, the production of 1 mol 2,3-BD could produce 1 mole of excess NADH, and therefore, biosynthesis of one 2,3-BD molecule supported the production of one ethanol molecule ( Supplementary Fig. S2k−m). As shown in Fig. 2g, the product proportions of E. cloacae SDM and E. cloacae SDM (ΔbudC) under anaerobic conditions coincide with the theoretical predictions, indicating that 2,3-BD synthesis participates in acetate elimination by providing excess NADH for the reactions catalyzed by AdhE.

Acetate elimination is widely distributed in bacteria
The distribution of BudA and AdhE in bacteria was studied by using the Protein BLAST program in the sequenced bacterial genomes from GenBank (updated December 3, 2018). Homologs of BudA are found in the genomes of 188 bacterial species, and homologs of AdhE are found in the genomes of 1389 bacterial species ( Fig. 3a; Supplementary Table S2). Among the 4333 completely sequenced bacteria, 135 species of bacteria, including E. cloacae, K. pneumoniae, and S. marcescens, have homologs of both BudA and AdhE (Fig. 3b).
K. pneumoniae ATCC 25955 was cultured aerobically in M9 medium supplemented with glucose as the carbon source. As shown in Fig. 3c, acetate accumulated at the first stage (0−4 h) and subsequently vanished (4−6 h) during the growth of K. pneumoniae ATCC 25955. K. pneumoniae ATCC 25955 (ΔadhE) lost the ability to produce ethanol and eliminate acetate (Fig. 3d). Additionally, K. pneumoniae ATCC 25955 (ΔbudABC) lost the ability to eliminate acetate, and the growth of the strain was inhibited by the accumulated acetate (Fig. 3e). These . c Relative β-galactosidase activity in E. cloacae SDM, E. cloacae SDM (ΔbudR) and E. cloacae SDM (Δfnr) containing pME6522-P budA or pME6522-P adhE at mid-log stage. The β-galactosidase activity in E. cloacae SDM is assumed to be 1. d Ratio of NADH/NAD + in E. cloacae SDM, E. cloacae SDM (ΔbudABC), and E. cloacae SDM (ΔadhE) at mid-log phase. e Ratio of NADH/NAD + in E. cloacae SDM, E. cloacae SDM-pETPc-nox, and E. cloacae SDM (ΔbudC). f Yield of 13 C ethanol from 13 C acetate in E. cloacae SDM, E. cloacae SDM-pETPc-nox and E. cloacae SDM (ΔbudC). 13 C sodium acetate at a concentration of about 10 mM was added into the culture system of these strains at 2 h. After 4 h of culture, the concentrations of 13 C ethanol and 13 C acetate were assayed. g Concentration of AC, 2,3-BD, ethanol, and the theoretical concentration of ethanol in E. cloacae SDM and E. cloacae SDM (ΔbudC) when cultured under anaerobic conditions. Experiments in a−f were carried out under aerobic conditions. Data shown are means ± SD. (n = 3 independent experiments). *P < 0.05 in two-tailed Student's t-test; **P < 0.01 in two-tailed Student's t-test; ns, no significant difference (P ≥ 0.05 in two-tailed Student's t-test). data demonstrate that 2,3-BD synthesis and ethanol production participate in the acetate elimination process of K. pneumoniae ATCC 25955.
E. coli K12 possesses only AdhE and could not produce 2,3-BD (Fig. 3b). As expected, it accumulated acetate during aerobic growth (Fig. 3f). When budABC from E. cloacae SDM was overexpressed, E. coli K12-pETRABC acquired the capability to produce 2,3-BD and eliminate the acetate produced by overflow (Fig. 3g). However, when adhE was knocked out, continuous accumulation of acetate during the growth of E. coli K12 (ΔadhE)-pET-RABC re-appeared (Fig. 3h). Therefore, the function of 2,3-BD synthesis to support acetate elimination can only be transferred into bacteria containing AdhE.
The genes budABC and adhE coexist in various bacterial species such as K. pneumonia, Vibrio cholera, and Serratia marcescens. Deletion of adhE or budABC resulted in higher acetate accumulation and decreased glucose utilization in both E. cloacae and K. pneumonia. Medium acidification and decreased survival due to acetate accumulation were also observed after deleting budABC in E. cloacae SDM (Supplementary Tables S3, S4). V. cholerae O1 has two biotypes, classical and El Tor 24 . The biotype El Tor, which possesses budABC, accumulated 2,3-BD without acetate accumulation, whereas the classical biotype, which lacks budABC, accumulated high concentrations of acetate. In addition, the 2,3-BD synthesis-positive biotype also showed higher survival rates, probably due to the absence of toxic acetate accumulation in the culture broth 25 . A similar phenotype was also observed for S. marcescens: this species showed a higher survival rate in the presence of budABC 26 . Thus, we can speculate that acetate elimination involving 2,3-BD synthesis and ethanol production is likely to be widely used in bacteria to contribute to rapid usage of carbon sources and cell survival. A total number of 4333 genomes were obtained by NCBI Blast online with the restriction of maximum identity level higher than 50%, query coverage of more than 90%, and E value lower than e −30 . b Organizations of 2,3-BD biosynthesis operon and adhE gene in different bacteria. c Growth of K. pneumoniae ATCC 25955 using glucose as carbon source. d Growth of K. pneumoniae ATCC 25955 (ΔadhE) using glucose as carbon source. e Growth of K. pneumoniae ATCC 25955 (ΔbudABC) using glucose as carbon source. f Growth of E. coli K12 using glucose as carbon source. g Growth of E. coli K12-pETRABC using glucose as carbon source. h Growth of E. coli K12 (ΔadhE)-pETRABC using glucose as carbon source. Experiments were carried out under aerobic conditions. Data shown are means ± SD (n = 3 independent experiments).

Elimination of acetate overflow requires the participation of Pta-AckA
Besides acetyl-CoA reduction to produce ethanol, activation of acetate to acetyl-CoA is also needed for acetate elimination 27 . Phosphotransacetylase (Pta), acetate kinase (AckA), and acetyl-CoA synthetase (AcsA) were reported to participate in acetyl-CoA production [28][29][30] . Assays using E. cloacae SDM with the knockout of AckA, Pta, and AcsA indicated that activation of acetate to acetyl-CoA is mainly executed by AckA and Pta, two enzymes expressed during acetate overflow (Fig. 4). As for the energy required for the AckA and Pta-catalyzed reaction, metabolic processes such as oxidative phosphorylation, pyruvate, 2,3-BD, and lactate synthesis could all produce ATP. However, only 2,3-BD synthesis can supply both ATP and NADH for acetate elimination with lower proteomic demands and without toxic acidic molecule production ( Supplementary Fig. S4a−d). The decreased ATP concentration resulting from the mutation of budABC in E. cloacae SDM can partially support the speculation that 2,3-BD synthesis also supplies energy for the activation of acetate ( Supplementary Fig. S4e). However, other experiments that could perturb ATP concentration without influencing the redox state are needed to identify the specific contribution of 2,3-BD synthesis in the energy requirements of acetate elimination.

Acetate generated from biomass chitin can be transformed into ethanol based on elimination mechanism
Besides being a toxic metabolite produced by overflow metabolism, acetate is also an unavoidable component in many renewable resources. Chitin, the second most abundant polysaccharide in the world after cellulose 31 , is a natural polymer of GlcNAc, with a global annual turnover of about 100 billion tons. Catabolism of chitin inevitably involves the production of acetate 32 . Acetate detoxification and coordinated glucose metabolism are two major problems that must be solved to make chitin utilization technologically viable. To identify whether the acetate elimination mechanism could be applied in chitin biorefinery, a combined catalytic and biotechnological approach for the production of AC (one of the 30 platform chemicals which are given the priority for development and utilization by the US Department of Energy) 33 and ethanol from chitin was proposed (Fig. 5a).
During the consumption of GlcNAc, E. cloacae SDM continuously accumulated acetate with a yield of 0.92 mol per molar of GlcNAc (0.92 mol/mol GlcNAc) (Fig. 5b). To support NADH requiring acetate elimination, two NADH-consuming enzymes lactate dehydrogenase (LdhA) and fumarate reductase (FrdA) were blocked. E. cloacae SDM (ΔldhAΔfrdA) can produce 2,3-BD with a lower acetate accumulation and a lower acetate yield (0.89 mol/mol GlcNAc). Besides meso-2,3-BD dehydrogenase encoded by budC, the (2R,3R)-2,3-BD dehydrogenase encoded by gdh can also consume NADH for AC reduction. Then, knockout of genes encoding these two 2,3-BD dehydrogenases and cultivation of mutant strains under anaerobic conditions were conducted to further supply NADH for acetate elimination and enable E. cloacae SDM producing AC and ethanol as its major products. Anaerobic cultivation of E. cloacae SDM (ΔldhAΔfrdAΔbudCΔgdh) eventually realized the coproduction of AC (0.95 mol/mol GlcNAc) and ethanol (0.89 mol/mol GlcNAc) using GlcNAc as the substrate ( Fig. 5b; Supplementary Fig. S5).
Chitin from the shrimp shell can be hydrolyzed by the chitinase of Bacillus licheniformis 34 . Then, anaerobic Fig. 4 Pta-AckA rather than AcsA is essential in the biotransformation of acetate to acetyl-CoA during acetate elimination. a Concentrations of 13 C-labeled acetate in E. cloacae SDM and its derivatives. b Concentrations of 13 C-labeled ethanol in E. cloacae SDM and its derivatives. 13 C sodium acetate at a concentration of about 10 mM was added into the culture system of strains at 2 h. After 4 h of culture, the concentrations of 13 C ethanol and 13 C acetate were assayed. Experiments were carried out under aerobic conditions. Data shown are means ± SD (n = 3 independent experiments). ***P < 0.001 in two-tailed Student's t-test; ns, no significant difference (P ≥ 0.05 in two-tailed Student's t-test).
cultivation of E. cloacae SDM (ΔldhAΔfrdAΔbudCΔgdh) was conducted using chitin hydrolysate as the substrate. As shown in Fig. 5c, E. cloacae SDM (ΔldhAΔfrdAΔ-budCΔgdh) produced 14.39 g L −1 AC and 8.10 g L −1 ethanol from chitin hydrolysate, with yields of 92% and 89% of the maximum theoretical yields, respectively. Discussion 2,3-BD is an important biochemical fuel with extensive industrial applications. Numerous strategies for microbial 2,3-BD production have been reported, whereas the metabolic function of 2,3-BD biosynthesis has not been well clarified. In this work, the physiological function of 2,3-BD biosynthesis was studied in E. cloacae SDM. A metabolic strategy in which 2,3-BD synthesis supports acetate elimination was proposed (Fig. 6). Acetate overflow leads to the initial accumulation of acetate in the early stages of E. cloacae SDM growth. Subsequently, acetate promotes the transcription of budABC and 2,3-BD synthesis supplies the reducing power required for AdhEcatalyzed ethanol production from acetyl-CoA. Compared with acetate catabolism through TCA and respiration, this strategy only requires three additional enzymes (BudA, BudB, and BudC) besides AdhE expressed during acetate overflow metabolism and can prevent the negative effects of acetate accumulation.
Acetate overflow is a bacterial allocation strategy to balance the proteomic demands of biomass synthesis and energy generation in order to achieve a high growth rate 2,35 . Acetate accumulation is one of the major problems restricting the industrial applications of some bacteria. However, simply blocking the acetate production (which would involve a decrease in energy generation) 36 or enhancing acetate catabolism through respiration (with an associated increase in proteomic demand) 37 cannot improve the fermentation performance of these bacteria. . NagE, N-acetylglucosamine specific PTS enzyme IIABC component; NagA, N-acetylglucosamine-6-phosphate deacetylase; NagB, glucosamine-6-phosphate deaminase; FrdA, fumarate reductase; LdhA, lactate dehydrogenase; PFL, pyruvate formate lyase; BudB, α-acetolactate synthase; BudA, α-acetolactate decarboxylase; BudC/GDH, 2,3-butanediol dehydrogenase; AckA, acetate kinase; Pta, phosphate acetyltransferase; AdhE, bifunctional acetaldehyde/ethanol dehydrogenase. b Yield of acetate, ethanol and AC produced by E. cloacae SDM and its derivatives under aerobic or anaerobic conditions using GlcNAc as the carbon source. c Batch fermentation using chitin hydrolysate as carbon source by E. cloacae SDM (ΔldhAΔfrdAΔbudCΔgdh) under anaerobic conditions. Data shown are means ± SD (n = 3 independent experiments).
Besides being a toxic metabolite produced by overflow metabolism, acetate is also an unavoidable component in many renewable resources. Chitin, the second most abundant polysaccharide in the world after cellulose, is a natural polymer of GlcNAc. Catabolism of chitin involves the inevitable production of acetate. Based on the acetate elimination mechanism identified in this work, we solved the problem of acetate accumulation during GlcNAc utilization by efficiently channeling reducing power to ethanol production. Two important chemicals, AC and ethanol could be co-produced from chitin, an abundant renewable resource. In addition, acetate is an inevitable toxic compound in hemicellulose hydrolysate. By coupling NADH-consuming ethanol production from acetate with NADH-producing xylose catabolism, S. cerevisiae can be engineered to convert xylose and toxic acetate in hemicellulose hydrolysate into ethanol 4 . It should be noted that the production of 2,3-BD might also lead to wastage of the carbon source. Exogenously supplying reducing power by electrochemical methods for acetate elimination might be another viable alternative.
The concept of overflow has traditionally been associated with the redirection of carbon towards excreted organic acids by heterotrophic microbes. Recently, organic acid overflow was reported in a mutant cyanobacterium Synechocystis sp. PCC 6803, which is unable to synthesize the carbon storage compound glycogen. A further study has demonstrated that both glycogen synthesis and organic acid overflow contribute to energy balancing in cyanobacteria 38 . Glycogen synthesis may also be viewed as a unique overflow metabolism in autotrophic microbes, helping to balance energy homeostasis under changing environmental conditions. In this study, acetate overflow in fast-growing E. cloacae SDM was also observed (Fig. 1b). The accumulation of acetate would result in acidification of the medium, which would be harmful to the bacterial strain. Then, biosynthesis of nontoxic 2,3-BD was activated, and this process was shown to provide NADH for the transformation of harmful acetate to ethanol (Fig. 6). The results demonstrate an innovative overflow strategy whereby the production of a non-toxic chemical can be exploited to drive desirable metabolic reactions.
In summary, we have revealed that the coupling between 2,3-BD synthesis and ethanol production establishes a naturally existing strategy for acetate elimination. Physiological characterization of naturally occurring metabolic processes, such as 2,3-BD synthesis, might provide alternative methods for solving problems that result in challenges for the industrial applications of bacteria. This study, therefore, stresses the significance of research in classic metabolic pathways with unidentified functions.  Mechanism of acetate-induced 2,3-BD biosynthesis supports ethanol production for acetate elimination. PDH, pyruvate dehydrogenase; Pta, phosphate acetyltransferase; AckA, acetate kinase; AdhE, bifunctional acetaldehyde/ethanol dehydrogenase; BudR, lysR family regulatory protein; BudA, α-acetolactate decarboxylase; BudB, α-acetolactate synthase; BudC, 2,3-butanediol dehydrogenase; Acetyl-P, acetyl phosphate.

Materials and methods
(China) and ThermoFisher (America), respectively. NAD + /NADH Quantification Colorimetric Kit was purchased from Biovision (K337-100, USA). EasyPure RNA Kit (used for RNA extraction) and TransStart Top Green qPCR SuperMix (used for qPCR) were purchased from Transgen (Beijing, China). HiScript II Q RT SuperMix for qPCR (+gDNA wiper, used for RNA reverse transcription) was purchased from Vazyme (R223-01, Nanjing, China). BacTiter-Glo™ Microbial Cell Viability Assay used for ATP detection was purchased from Promega (America). All other chemicals were of analytical grade and commercially available.
Bacterial strains, plasmids, media, and culture conditions All the strains and plasmids used in this study are listed in Supplementary Table S5. Lysogenic broth (LB) medium was used for the general culture of strains. M9 minimal medium 39 supplemented with 5 g/L yeast extract and various carbon sources, including glucose, gluconic acid was used for the culture of E. cloacae SDM. Sodium hydrosulfite (5 g/L) and resazurin (1 mg/L) were added in M9 minimal medium for the anaerobic culture of E. cloacae SDM. Ampicillin and kanamycin were used at a concentration of 100 and 50 μg/mL, respectively. Experiments were carried out under aerobic conditions unless otherwise stated.

Gene manipulation in bacterial strains
To construct the E. cloacae SDM (ΔbudABC) mutant strain, the homologous arms upstream and downstream of budABC were amplified using primers ΔbudABC-f1/ ΔbudABC-r2 and ΔbudABC-f3/ΔbudABC-r4 (Supplementary Table S6) from the genome of E. cloacae SDM. These two fragments were fused via PCR method using primers ΔbudABC-f1 and ΔbudABC-r4, which contained EcoRI and BamHI restriction enzyme sites, respectively. The infused fragment was linked with suicide plasmid pKR6K, which was also digested with EcoRI and BamHI. The resulting plasmid was designated pKR6K-ΔbudABC and transferred into E. coli S17-1 for conjugation with E. cloacae SDM. The single-crossover mutants with the integration of pKR6K-ΔbudABC into the chromosome were selected on an M9 agar plate supplemented with 20 g/L sodium citrate and 50 μg/mL kanamycin. The double-crossover mutants were selected on LB agar plates containing 15% (w/v) sucrose. All the constructed strains were validated by PCR and sequenced. The adhE, budR, fnr, budC, pta, ackA, and acsA mutants of E. cloacae SDM, budABC, and adhE mutants of K. pneumoniae ATCC 25955 were generated by using the same procedure. The primers used in this study were listed in Supplementary Table S6.
The knockout of adhE in E. coli K12 was conducted by the one-step inactivation method as described previously 39 . Briefly, E. coli K12-ΔadhE-f1/E. coli K12-ΔadhE-r2 (Supplementary Table S6) were used to directly amplify adhE mutant fragment from adhE::kan mutant strain purchased from The Coli Genetic Stock Center. The PCR products had~330 bp up and down homologous arms, respectively, outside of adhE for homologous recombination and an FRT-flanked kanamycin resistance cassette for target gene replacement and mutant strain screening.

Monitoring metabolic processes of strains
Unless specified, E. cloacae SDM, K. pneumoniae ATCC 25955, E. coli K12, and their derivatives were cultivated in M9 minimal medium supplemented with 5 g/L yeast extract and 40 g/L glucose as the sole carbon source at 37°C and 180 rpm. Experiments were conducted in 300 mL shake flasks containing 50 mL of medium. Samples were withdrawn periodically to determine the cell density, concentrations of glucose, 2,3-BD, acetate, ethanol, and other by-products. To monitor the acetate elimination of various bacterial strains, about 10 mM 13 Clabeled sodium acetate was added at the initial log-stage (at 2 h), and samples were withdrawn every 2 h to determine the concentrations of 13 C-labeled acetate and 13 Clabeled ethanol using gas chromatography-mass spectrometry (GC-MS, GCMS-QP2010 Plu, Shimadzu, Japan) and high-performance liquid chromatography (HPLC, Agilent 1100, America). Briefly, combined concentrations of labeled and unlabeled acetate were determined by HPLC. Concentrations of labeled acetate and unlabeled acetate were then detected by GC-MS. 13 C-labeled acetate ( 13 C acetate) was detected at m/z 61 and unlabeled acetate at m/z 60. The concentrations of 13 C-labeled ethanol and unlabeled ethanol were analyzed with the same procedure.

Quantitative real-time PCR
Total RNA was extracted from E. cloacae SDM or its derivatives using an EasyPure RNA Kit (Transgen, China). DNA contamination was eliminated by RNase-free DNase I (Vazyme, China), and the quality of extracted RNA was detected by 1.5% agarose gel electrophoresis and absorbance ratio at 260/280 nm. Then, cDNA was synthesized from total RNA by using HiScript II Q RT SuperMix (Vazyme, China).
The qPCR analysis was performed using TransStart Top Green qPCR SuperMix (Transgen, China) on a Light-Cycler 480 system (Roche). For absolute quantification of the copy numbers, standard curves were constructed for adhE and budA by serial dilutions of the recombinant plasmids harboring the same amplicons. Each reaction was performed in triplicate. Controls with no template or total RNA were included for each reaction on the same plate. The primers used for qPCR analysis are listed in Supplementary Table S6.

Construction of standard curves for qPCR
Partial budA gene sequences were amplified by PCR using primers budA-qPCR (SDM-f)/budA-qPCR (SDM-r) (Supplementary Table S6) and then cloned into pEASY-Blunt Simple Cloning Vector (Transgen, China) to form pEASY-Blunt-budA. This plasmid was extracted by using Plasmid Miniprep Kit (Biomiga, USA) and quantified by using NanoDrop ND-1000 (Thermo Scientific, USA). The copy number of the recombinant plasmids was calculated based on the molecular weight 40 . The standard curve for qPCR was constructed with serial 10-fold dilutions of a recombinant plasmid, ranging from 1 × 10 5 to 1 × 10 10 copies per μL. For each set, determined threshold cycle (C T ) values were plotted against the logarithm of their known initial copy number (per μL) using Origin software 9.0 (OriginLab, USA). PCR amplification efficiency (E) was calculated according to the equation: E = 10 −1/slope -1 as described previously 41 . The standard curve for qPCR of adhE was generated by using the same procedure. All of the reactions and controls with no template were run in triplicate.
Sample preparation for qPCR E. cloacae SDM and its derivatives were grown overnight in LB medium and then inoculated into M9 minimal medium supplemented with 5 g/L yeast extract and 40 g/L glucose as the sole carbon source. Cells of these strains were harvested every two hours for the quantification of mRNA. Cells from LB medium were termed as the sample at 0 h. To identify the effect of acetate on transcription of budA and adhE, the procedure was the same as above except that 40 mM sodium acetate was added into M9 minimal medium at the beginning.

β-Galactosidase assay
The promoter region of the adhE gene (P adhE ) was amplified from E. cloacae SDM genomic DNA using primers P adhE (SDM-f)/P adhE (SDM-r) ( Supplementary  Table S6). Then, the purified PCR product was ligated with pME6522 to generate pME6522-P adhE . In the same way, the promoter region of budA gene (P budA ) was cloned into pME6522 to generate pME6522-P budA . pME6522-P budA and pME6522-P adhE were transferred into E. cloacae SDM and its derivatives by electroporation.
E. cloacae SDM and its derivatives harboring pME6522-P budA or pME6522-P adhE were grown in M9 minimal medium supplemented with 5 g/L yeast extract and 40 g/L glucose as sole carbon source and harvested at the midlog stage by centrifugation at 8000 rpm for 5 min. The β-galactosidase activity was then determined using onitrophenyl-β-D-galactopyranoside as the substrate.
Detection of intracellular NADH/NAD + level E. cloacae SDM and its derivatives were cultured in an M9 minimal medium supplemented with 5 g/L yeast extract and 40 g/L glucose as the sole carbon source to mid-log phase. Then, E. cloacae cells were collected by centrifugation at 13,000 rpm for 1 min. The intracellular concentrations of NADH and NAD + were determined using the NAD + /NADH Quantification Colorimetric Kit (BioVision, K227-100, America) according to the manufacturers' instructions.
Detection of ATP level E. cloacae SDM and E. cloacae SDM (ΔbudABC) were grown in an M9 minimal medium supplemented with 5 g/ L yeast extract and 40 g/L glucose and harvested at midlog phase. The ATP levels were determined using the BacTiter-Glo™ Microbial Cell Viability Assay (Promega, America) according to the manufacturers' instructions.
Determination of survival rate E. cloacae SDM and E. cloacae SDM (ΔbudABC) were grown in M9 minimal medium supplemented with 5 g/L yeast extract and 40 g/L glucose. Cells were harvested at 6, 30, 54, 78, 102 h with centrifugation at 8000 rpm for 2 min. SYTO 9 and propidium iodide included in a LIVE/ DEAD BacLight Bacterial Viability Kits (Invitrogen, L7012) were used for the determination of survival rate at different times. The number of live/dead cells was counted by a quantitative imaging analysis flow cytometer (ImageStream x Mark II) with an exciting light at 480 nm.

Expression and purification of chitinase
The gene encoding chitinase was amplified from the genome of B. licheniformis DSM13 by PCR using primers chi-f/chi-r (Supplementary Table S6). The PCR product and plasmid pFLAG-CTS were digested by HindIII and XhoI and then linked by T4 DNA ligase to construct plasmid pFLAG-CTS-chi. Then, the plasmid pFLAG-CTS-chi was transformed into E. coli Top10 for chitinase expression.
Purification of chitinase was conducted as described by a group of Songsiriritthigul with slight modification 34 . In brief, the recombinant strain E. coli Top10-pFLAG-CTSchi was cultured in LB medium containing 100 µg/mL ampicillin at 37°C to an optical density at 600 nm of 1.5, then 1 mM IPTG was added to induce protein expression under 25°C for 18 h. After that, cells were harvested by centrifugation at 6000 rpm for 10 min and washed twice by 0.85% saline solution. The freshly prepared cell pellet was resuspended in lysis buffer (OD 600 reached approximately 30), then lysed on ice using a high-pressure crusher. Cell debris and intact cells were removed by centrifugation at 12,000 rpm and 4°C for 30 min. The resultant supernatants were immediately applied to a HisTrap HP column (5 mL) equilibrated with binding buffer (20 mM sodium phosphate and 500 mM sodium chloride, pH 7.4) and eluted with a gradient ratio of wash buffer (20 mM sodium phosphate, 500 mM imidazole, and 500 mM sodium chloride, pH 7.4) to obtain purified chitinase ( Supplementary Fig. S6).
Co-production of AC and ethanol from GlcNAc or chitin hydrolysate Chitin hydrolysate was prepared from colloidal chitin according to the method of Songsiriritthigul et al. 34 with some modification. Briefly, 50 g of chitin powder (Aladdin, China) was slowly added into 500 mL concentrated HCl with vigorous stirring for 3−4 h and then incubated overnight. The mixture was filtered through a cheesecloth, dropped slowly into 600 mL of ice-cold ethanol (50%) with rapid stirring on ice. The colloidal chitin was collected by centrifugation at 8000 rpm for 10 min and washed several times with tap water until the pH approached neutral. Then, recombinant chitinase of B. licheniformis was added into sterilized colloidal chitin (200 g/L) at a concentration of 0.5 U/mL and the mixture was incubated at 50°C and 180 rpm for 48 h.
For the co-production of AC and ethanol, E. cloacae SDM and its derivatives were cultured in M9 minimal medium supplemented with 5 g/L yeast extract and 40 g/L GlcNAc under aerobic or anaerobic condition. Alternatively, the chitin hydrolysate was fed into the M9 minimal medium to make the GlcNAc concentration at about 40 g/L. Anaerobic culture was conducted with 50 mL anaerobic bottles containing 40 mL of medium in the anaerobic incubator.

Analytical methods
Cell density was measured using a spectrophotometer (Lengguang-721, China) at a wavelength of 600 nm. The concentration of glucose was measured enzymatically by a bio-analyzer (SBA-40D, Shandong Academy of Sciences, China) after diluting to an appropriate concentration. The pH of the culture was measured by an Orion Star A211 pH meter. The concentration of 2,3-BD and AC were determined by GC (GC2014C, Shimadzu, Japan) using a capillary GC column as described previously 42 . Other metabolites e.g. acetate, ethanol, lactate, and succinate were measured using high-performance liquid chromatography (HPLC, Agilent 1100, America) as described 39 .

Statistical analysis
Software for initial data processing was Microsoft Excel 2013, and subsequent analysis was carried out using Origin (OriginLab), GraphPad Prism 5 (GraphPad). Venn diagrams of the distribution of AdhE and BudA were drawn by the webtool in Bioinformatics & Systems Biology. Analogous enzymes of AdhE and BudA were gained by NCBI Blast online. Detailed data analysis is described in the text.