Metabolic engineering of Escherichia coli based on adaptive evolution and omics technology for highly efficient l-valine production under oxygen-limited conditions Authors: Hengshuai Zhang, Zhirui Ha, Aiying Wei, Chunguang Zhao, Gang Meng, Xin Li, Faqing Wu, Jingui Liu, Heyun Wu, Qian Ma, Xixian Xie Published: 2026 ABSTRACT AbstractIn industrial production, the yield of desired targets derived from carbon sources is frequently diminished by the competitive influence of cellular metabolism within microbial cell factories. The bioproduction of l-valine exemplifies a classic process that is confronted with such a dilemma, substantially hindering its economic industrial-scale production. In this study, we aim to engineer a cell factory capable of efficiently synthesizing l-valine with high yield by minimizing the consumption of its precursor pyruvate through the TCA cycle under oxygen-limited conditions. Metabolic engineering-based adaptive laboratory evolution (ALE) under oxygen-limited conditions resulted in the development of an evolved strain ALE2-40 with better cell growth and enhanced l-valine yield. Through comparative omics analysis and validation experiments, it was uncovered that during the ALE process, both pyruvate dehydrogenase activity and NADH availability were significantly improved. Moreover, beneficial targets have the potential to contribute to the NADH and ATP pools, thereby further promoting l-valine synthesis. Based on these results, reverse engineering of the evolved strain ALE2-40 was further conducted. Ultimately, the final strain VAL19 demonstrated remarkable performance, achieving an impressive l-valine titer of 93.7 g/L within 28 h in a 5-L bioreactor under oxygen-limited conditions, with a remarkable yield of 60.4% from glucose—equivalent to 92.9% of the theoretical yield—and a productivity of 3.35 g/L/h. These results set a new benchmark for the fermentative production of l-valine, with the highest yield and productivity reported so far.Keywords: l-valine, Escherichia coli, Adaptive evolution, Metabolic engineering, Oxygen-limited fermentation Introduction In nature, microorganisms exist in environments that may or may not contain oxygen. Consequently, they usually rely on either the respiratory chain or the substrate-level phosphorylation to generate the energy necessary for sustaining viability. In comparison, the efficiency of ATP production via the respiratory chain usually significantly surpasses that achieved through substrate-level phosphorylation [1]. Nevertheless, the operation of the respiratory chain necessitates the availability of oxygen, and meanwhile, the tricarboxylic acid (TCA) cycle becomes active, which poses a significant competing pathway for the synthesis of chemicals that utilize pyruvate as a precursor [2]. Thus, for the production of pyruvate or derivatives using facultative anaerobic bacteria such as Escherichia coli and Corynebacterium glutamicum, efforts have been made to shift the fermentation process from aerobic conditions to oxygen-limited or anaerobic conditions, aiming to minimize the consumption of pyruvate via the TCA cycle, and to increase the yield of target products from carbon sources [[3], [4], [5]]. Under oxygen-limited or anaerobic conditions, the re-oxidation of nicotinamide adenine dinucleotide (NADH) generated during glycolysis is vital for maintaining cofactor balance [6], otherwise, cell growth and the biosynthesis of target products may be impeded due to the accumulation of fermentative byproducts such as acetic acid, lactic acid and succinic acid [6]. A critical challenge in the efficient production of pyruvate and derivatives lies in rewiring intracellular metabolic pathways to facilitate high-yield production under oxygen-limited or anaerobic fermentation conditions [7]. The rewiring of the metabolic pathway necessitates achieving an equilibrium between the central carbon metabolism and the synthetic pathway, as well as ensuring cofactor balance [8]. l-valine, an important branched-chain amino acid that uses pyruvate as its precursor, finds significant applications across diverse fields such as feed, food, and medicine. The fermentative production of l-valine has been mainly focused in C. glutamicum [8] and E. coli [9]. Currently, as high as 103 g/L l-valine has been produced in C. glutamicum under aerobic fermentation, yet the yield is merely 35.0% from glucose [8], which is significantly lower than the theoretical yield of 65%. Improving the yield of l-valine has been a persistent challenge limiting the application of these microbial cell factories. To tackle this issue, two-stage fermentation strategies have been explored [2,9]. This approach decouples cell growth from efficient production by transitioning from aerobic fermentation to oxygen-limited or anaerobic fermentation, thereby achieving a balanced trade-off between biomass formation and l-valine production. In a recent study, Lang et al. [9] have reported the production of 90.6 g/L of l-valine via engineered E. coli in a 3-L bioreactor using a two-stage aerobic-microaerobic fermentation approach, achieving an impressive productivity of 3.24 g/L/h. However, the l-valine yield from glucose stood at 48.0%, which remains unsatisfactory and is likely owing to the pyruvate consumption via TCA cycle during the aerobic-stage fermentation. Hasegawa et al. [6] have reported the high-yield of 57.0% from glucose in C. glutamicum under oxygen deprivation conditions, which also required the first stage of cell cultivation under aerobic condition followed by harvesting cells for production under oxygen deprivation condition. The yield attained stands as the highest ever reported; however, the complicated operation process extends the production duration and increases costs. In their study, the engineering of the NAD-preferring mutant acetohydroxy acid isomeroreductase (AHAIR) and NAD-specific leucine dehydrogenase from Lysinibacillus sphaericus to achieve cofactor balance is the crucial strategy that guarantees the efficient l-valine production under oxygen deprivation conditions [6]. Similarly, through systems metabolic engineering including redox cofactor balancing, we previously succeeded in achieving high-yield production of l-valine in an engineered E. coli by a two-stage fermentation process in a 5-L bioreactor [2]. Unlike the approach taken by Hasegawa et al. [6], the method we adopted eliminated the need for cell harvesting. Instead, by simply turning off the aeration when cell grew to an OD600 value around 100 and maintaining a stirrer speed of 200 rpm, the aerobic fermentation was seamlessly transitioned into the second stage of oxygen-limited fermentation. Accordingly, a titer of 84 g/L l-valine with a yield of 41.0% from glucose and a productivity of 2.33 g/L/h was obtained [2]. Considering that the aerobic phase poses the primary challenge to achieving high-yield production of l-valine, we undertook the initiative to transition from the two-stage fermentation to a single-stage, oxygen-limited fermentation. In this study, rational metabolic engineering was combined with adaptive laboratory evolution under oxygen-limited conditions, which led to an evolved high-yield strain ALE2-40. Further genomic and transcriptomic analysis between the evolved strain and the starting strain revealed several potential targets that closely related to the NADH and ATP availability. Through reverse engineering grounded in omics data, a highly efficient l-valine producing strain, VAL19, was ultimately developed. This strain demonstrated remarkable production capability, yielding 93.7 g/L of l-valine within 28 h in a 5-L bioreactor, with a yield of 60.4% from glucose, representing 92.9% of the theoretical yield. The productivity also reached a record-breaking high value of 3.35 g/L/h, thereby exhibiting immense promises for the industrial-scale production of l-valine. The strategies employed in this study have the potential to serve as a valuable reference for the engineering of microbial cell factories for analogous chemicals utilizing pyruvate as a precursor. Materials and methods Microorganisms and plasmids The strains and plasmids used in this study are listed in Table 1. The primers used in this study are listed in Supplementary Table S1. The starting strain VAL0 was previously constructed in our lab. E. coli DH5α was used as the cloning host for plasmid construction, and plasmid pSTV28 was used for the overexpression of genes. The plasmids pREDCas9 and pGRB were generously provided by Prof. Tao Chen of Tianjin University and were applied for gene manipulation [10]. Table 1Strains and plasmids used in this study.Table 1Strains/PlasmidsCharacteristicsSourcesStrainsE. coli DH5αHost for cloningThis labVAL0E. coli MG1655, ΔlacI, ΔldhA, ΔadhE, ΔpflB, ΔfrdABCD, yneO::Ptrc-alsS, lafU::Ptrc-alsS, ilvC::Ptrc-ilvCL67E,R68F,K75E, ycgH::Ptrc-ilvCL67E,R68F,K75E, ilvE::Ptrc-bcd, yjiV::Ptrc-bcd, yjiT::Ptrc-brnFEThis labVAL01VAL0, ΔpoxBThis studyVAL02VAL01, ΔackAThis studyVAL03VAL02, ΔtdcDThis studyVAL04VAL03, ΔptaThis studyVAL05VAL04, rph::Ptrc-pgkThis studyVAL06VAL04, rph::Ptrc-gapAThis studyALE1-1-ALE1-48The mutant strains derived from a first round of successive passages of adaptive laboratory evolution under anaerobic conditions with strain VAL06 as the starting strainThis studyALE1-45The evolved strain with optimal fermentation performance obtained during the first round of adaptive laboratory evolutionThis studyALE2-1-ALE2-48The mutant strains derived from the second round of successive passages of adaptive laboratory evolution under anaerobic conditions with strain ALE1-45 as the starting strainThis studyALE2-40The evolved strain with optimal fermentation performance obtained during the second round of adaptive laboratory evolutionThis studyVAL07VAL06 carrying plasmid pSTV28This studyVAL08ALE2-40 carrying plasmid pSTV28This studyVAL06-yfdHVAL06 carrying plasmid pSTV28-PJ23105-yfdHThis studyVAL06-yfdH∗VAL06 carrying plasmid pSTV28-PJ23105- yfdH∗This studyVAL06-fimHVAL06 carrying plasmid pSTV28-PJ23105-fimHThis studyVAL06-fimH∗VAL06 carrying plasmid pSTV28-PJ23105-fimH∗This studyVAL06-rplPVAL06 carrying plasmid pSTV28-PJ23105-rplPThis studyVAL06-rplP∗VAL06 carrying plasmid pSTV28-PJ23105-rplP∗This studyVAL06-lpdVAL06 carrying plasmid pSTV28-PJ23105-lpdThis studyVAL06-lpd∗VAL06 carrying plasmid pSTV28-PJ23105-lpd∗This studyVAL06-tpiAVAL06 carrying plasmid pSTV28-PJ23105-tpiAThis studyVAL06-ynfMVAL06 carrying plasmid pSTV28-PJ23105-ynfMThis studyVAL06-yeeNVAL06 carrying plasmid pSTV28-PJ23105-yeeNThis studyVAL06-azoRVAL06 carrying plasmid pSTV28-PJ23105-azoRThis studyVAL06-ybcLVAL06 carrying plasmid pSTV28-PJ23105-ybcLThis studyVAL06-napFVAL06 carrying plasmid pSTV28-PJ23105-napFThis studyVAL06-yihAVAL06 carrying plasmid pSTV28-PJ23105-yihAThis studyALE2-40-ΔiadAALE2-40, ΔiadAThis studyALE2-40-ΔargIALE2-40, ΔargIThis studyALE2-40-ΔargFALE2-40, ΔargFThis studyALE2-40-ΔxynRALE2-40, ΔxynRThis studyALE2-40-ΔyjiHALE2-40, ΔyjiHThis studyALE2-40-ΔyjiGALE2-40, ΔyjiGThis studyALE2-40-ΔpotHALE2-40, ΔpotHThis studyALE2-40-ΔrsxEALE2-40, ΔrsxEThis studyALE2-40-ΔfimGALE2-40, ΔfimGThis studyALE2-40-ΔspeCALE2-40, ΔspeCThis studyVAL09ALE2-40, yciQ::PJ23105-ybcLThis studyVAL10ALE2-40, ylbE::PJ23105-yihAThis studyVAL11ALE2-40, gapC::PJ23105-napFThis studyVAL12ALE2-40, ΔrsxEThis studyVAL13VAL12, ΔfimGThis studyVAL14VAL12, ΔspeCThis studyVAL15VAL12, ΔyjiGThis studyVAL16VAL13, ΔspeCThis studyVAL17VAL13, ΔyjiGThis studyVAL18VAL17, ΔspeCThis studyVAL19VAL18, aceE::PfliC-aceEThis studyPlasmidspREDCas9Sper, Cas9 and λ Red recombinase expression vector[10]pGRBAmpr, gRNA expression vector[10]pSTV28The expression vector, CmrThis labpSTV28-PJ23105-yfdH∗pSTV28 derivative for yfdHI248T overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-yfdHpSTV28 derivative for yfdH overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-fimH∗pSTV28 derivative for fimHG87C overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-fimHpSTV28 derivative for fimH overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-rplP∗pSTV28 derivative for rplPI73L overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-rplPpSTV28 derivative for rplP overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-lpd∗pSTV28 derivative for lpdA321T overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-lpdpSTV28 derivative for lpd overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-tpiApSTV28 derivative for tpiA overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-ynfMpSTV28 derivative for ynfM overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-yeeNpSTV28 derivative for yeeN overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-azoRpSTV28 derivative for azoR overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-ybcLpSTV28 derivative for ybcL overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-napFpSTV28 derivative for napF overexpression controlled by the promoter PJ23105, CmrThis studypSTV28-PJ23105-yihApSTV28 derivative for yihA overexpression controlled by the promoter PJ23105, CmrThis study Strains and plasmids used in this study. CRISPR/Cas9-mediated genome editing For strain engineering, chromosomal deletion or integration by CRISPR/Cas9 system was conducted as described by Li et al. [10]. Here, we present the knockout of gene poxB as an example. Briefly, a pair of 20 bp complementary sequences was designed using CRISPR RGEN Tools (http://www.rgenome.net/cas-designer/) and then synthesized in primers and annealed to form a dsDNA. The annealed dsDNA with flanking sequences homologous to the pGRB backbone was ligated into the linearized pGRB plasmid using the ClonExpress® II One Step Cloning Kit (Vazyme, China) to construct a gRNA-expressing plasmid (pGRB-poxB). Two homologous arms were ligated by overlapping PCR, generating a donor DNA fragment (DNA-poxB). Plasmid pGRB-poxB and the donor DNA-poxB were co-transformed into electrocompetent cells of target strains harboring the plasmid pREDCas9. A concentration of 0.1 mM IPTG was used to induce the λ-Red recombinase expression. LB plates with respective 50 μg/mL ampicillin and spectinomycin were used for positive colony selection. Positive colonies were further verified by colony PCR and DNA sequencing. Validated colonies were cultivated in LB containing 0.2% l-arabinose overnight to cure the plasmid pGRB-poxB. The curing of plasmid pREDCas9 was performed via culturing at 42 °C overnight. Oxygen-normal fermentation in shaking flasks Shake-flask fermentation of the engineered strains constructed prior to adaptive laboratory evolution (ALE) was performed under oxygen-normal conditions at a shaking rate of 220 rpm. These engineered strains were first inoculated into LB liquid medium from solid plates and cultured at 37 °C with shaking at 220 rpm for about 12 h. Subsequently, the strains were inoculated into 30 mL of LB medium, which served as the seed culture, within a 500-mL shaking flask. They were then cultivated at 37 °C under a shaking rate of 220 rpm for approximately 12 h. Following the seed-culture phase, the seed was transferred to a 500-mL baffled shaking flask with a final fermentation volume of 30 mL, using a 10% inoculum volume. The fermentation process was carried out at 37 °C, with a shaking rate of 220 rpm for 24 h. The fermentation medium (per liter) contained 20 g glucose, 1 g yeast extract, 4 g citrate, 6 g K2HPO4, 1 g MgSO4·7H2O, 60 mg FeSO4·7H2O, 5 mg MnSO4·H2O, 0.5 mg VH, a few drops of antifoam, and 8 mg phenol red. Throughout the fermentation process, the pH was carefully regulated and maintained at approximately 7.0 by the controlled addition of 25% (v/v) ammonia solution (NH3·H2O) using a micro-injector, with the color change of phenol red serving as an indicator. Once the glucose in the fermentation medium was depleted, 1 mL of 60% (w/v) glucose solution was introduced to sustain the fed-batch fermentation. Adaptive laboratory evolution under oxygen-limited conditions To enhance the strain's tolerance to oxygen-limited conditions and promote high-yield l-valine production, we employed ALE to optimize its growth performance when subjected to the stress of oxygen-limited condition. A tightly sealed reagent bottle filled with fermentation medium was employed to establish an oxygen-deprived environment. The culture medium used in ALE contained 20 g/L glucose, 1 g/L yeast extract, 4 g/L citrate, 6 g/L K2HPO4, 5 g/L (NH4)2SO4, 1 g/L MgSO4·7H2O, 60 mg/L FeSO4·7H2O, 5 mg/L MnSO4·H2O, 0.5 mg/L VH, 20 g/L MOPS to sustain pH stability, and 8 mg/L phenol red. During serial-passage, a 1-mL aliquot of the cell culture was transferred to a fresh reagent bottle. The culture was then incubated at 37 °C, with continuous shaking at a rate of 220 rpm. Once the strain exhibited a markedly enhanced growth performance following multiple serial-passages of evolutionary culture, the culture was diluted and evenly spread across an LB solid plate. A total of 48 single colonies were randomly chosen for initial screening by 96-deep-well plate fermentation. Subsequently, the strains that demonstrated markedly enhanced growth performance during this preliminary screening were subjected to a secondary screening via shake-flask fermentation under oxygen-limited fermentation. Following shake-flask fermentation under oxygen-limited conditions, the strain with the highest l-valine titer was introduced into a second round of ALE. Oxygen-limited fermentation of the ALE-derived engineered strains After the ALE experiment, an initial fermentation screening of the evolved strains was conducted in 96-deep-well plates. Single colonies were picked using sterile toothpicks from LB solid plates and then cultured in 96 deep-well plates containing 450 μL LB medium for seed culture. The seed culture was carried out at a temperature of 37 °C and a shaking rate of 220 rpm for 12 h. Following which, the seed culture was transferred to a new 96 deep-well plate containing 450 μL fermentation medium and added glucose solution (60%, w/v) to a final concentration of 10 g/L. The 96 deep-well plate was sealed with a sealing mat to establish an oxygen-limited condition. The fermentation process was conducted at 37 °C with a shaking rate of 220 rpm. For oxygen-limited fermentation in shaking flasks, the seed and fermentation media were the same as those used under oxygen-normal conditions, except that the shaking rate during fermentation phase was reduced from 220 rpm to 80 rpm to create an oxygen-limited environment. Fed-batch fermentation under oxygen-limited conditions in a 5-L bioreactor Cells cultured on agar slants were inoculated in a 5-L bioreactor (Biotech-5BG, Baoxing, China) containing 2-L fermentation medium. The fermentation medium was the same as that used in the shake-flask fermentation, except for the omission of phenol red as pH indicator. The fermentation temperature was maintained at 37 °C, while the pH was kept at approximately 7.0 through automated addition of 25% (v/v) ammonia solution (NH3·H2O). Once the glucose in the initial medium was exhausted, an 80% (w/v) glucose solution was introduced at an optimal rate to ensure that the residual glucose concentration remained below 2 g/L. To establish an oxygen-limited environment, the stirrer speed was set to a mere 300 rpm, and aeration was set to 0.2 L/min. After approximately 8 h of fermentation, the DO dropped to 0 and remained at this level until the end of the fermentation process. Analysis of the intracellular NADH/NAD+ ratio The cells were harvested at 16 h of shake-flask fermentation and washed with prechilled PBS buffer, followed by swift lysis using a specialized lysis buffer. Next, intracellular NADH levels were quantified utilizing an Enhanced NAD+/NADH Assay Kit (WST-8 method, S0176S, Beyotime Biotechnology, Shanghai, China). The concentrations of NADH and NAD+ were then standardized using the NADH standard curve, enabling the calculation of the NADH/NAD+ ratio. Analysis of intracellular ATP level The cells harvested at 16 h of shake-flask fermentation were washed with prechilled PBS buffer and rapidly lysed with the lysis buffer. Subsequently, the intracellular ATP levels were determined using a luciferase-based Enhanced ATP Assay Kit S0027 (Beyotime Biotechnology, Shanghai, China), according to the method described in previous study [11]. The ATP concentrations were normalized using the ATP standard curve and exhibited as μmol/DCW. Analysis of PDH activity assay The cells were harvested at 16 h of shake-flask fermentation, washed with prechilled PBS buffer and lysed by ultrasonic disruption. The PDH activity assay was determined using a Pyruvate Dehydrogenase (PDH) Activity Assay Kit BC0385 (Solarbio, Beijing, China). The PDH activity assay was exhibited as U/DCW. Analytical methods for the determination of biomass and metabolites The cell biomass was measured by the absorbance at 600 nm (OD600) with an ultraviolet spectrophotometer. The concentration of glucose was measured using an SBA-40C biosensor (Shandong Province Academy of Sciences, China). The concentration of l-valine after derivatization was detected by HPLC (Prominence LC-20A, Shimadzu, Japan) coupled with a UV detector at a wavelength of 360 nm. The acetonitrile/water mixture (50:50, v/v) with 50 mM sodium acetate used as the eluent at a flow rate of 1 mL/min. Organic acids were detected by HPLC (SHIMADZU LC-20AT) using an Aminex® HPX-87H column (7.8 mm×300 mm) coupled with an RID-20A refractive index detector was used for separation and detection. The column temperature was set at 37 °C. Sulfuric acid solution with a concentration of 5 mM was used as the eluent at a flow rate of 0.5 mL/min. Statistical analysis Three independent replicates were conducted in shake-flask fermentation experiments. Data were presented as means with standard deviations. Statistical analysis was performed using one‐way analysis of variance, followed by Student's t‐test. For the 5-L bioreactor fermentation, a minimum of three batches for each strain were carried out, with a representative batch presented in detail. Results Metabolic engineering of the starting strain to build a suitable chassis for adaptive evolution under oxygen-limited conditions In our previous study, an l-valine-hyperproducing strain designated as VAL0 was constructed, which demonstrated the capacity for efficient l-valine production through a two-stage fermentation process, with similar performance as our previously reported VHY18 strain [2]. Specifically, during the first-stage fermentation, aerobic conditions were required to facilitate fast cell growth. Subsequently, during the second-stage oxygen-limited fermentation, l-valine was produced with high yield. Although the decoupling of cell growth and production resulted in an enhanced l-valine yield compared to the conventional aerobic fermentation, the two-stage fermentation approach still encountered challenges, including a prolonged fermentation duration and a relatively low yield during the aerobic cell growth phase. To address these problems, we aim to employ adaptive laboratory evolution (ALE) to establish a single-stage, oxygen-limited fermentation process and curtail the bypass consumption of the precursor pyruvate via the TCA cycle. However, several key issues must be taken into consideration before conducting ALE. Firstly, to facilitate the transition from aerobic fermentation to oxygen-limited fermentation, it is imperative to block the acetic acid synthesis pathway in the starting strain. This is because E. coli strains inherently possess a tendency to accumulate acetic acid as a by-product when cultivated under anaerobic or oxygen-limited conditions with abundant glucose. Specifically, gene poxB encoding the pyruvate dehydrogenase, gene ackA encoding the acetate kinase, gene tdcD encoding the propionate kinase, and pta encoding the phosphate acetyltransferase, were sequentially knocked out in VAL0, constructing strains VAL01, VAL02, VAL03, and VAL04, respectively. The fermentation characteristics of these strains under oxygen-normal conditions were compared in shake-flask fermentation with VAL0 as the control, and the results are shown in Fig. 1. As shown in Fig. 1A, the accumulation of acetic acid gradually decreased with the sequential knockout of the acetic acid synthesis genes. The titer of acetic acid in strain VAL01-VAL04 gradually decreased to 4.8, 3.5, 2.4, and 1.5 g/L, which were 7.7%, 32.7%, 53.8%, and 71.2% lower than that in the starting strain VAL0. Meanwhile, the corresponding titer of l-valine increased to 38.8, 39.6, 41.1, and 42.2 g/L, which were 0.3%, 2.3%, 6.2%, and 9.1% higher than that in the starting strain VAL0, respectively. There was no significant growth damage among these strains. The above results suggested that the knockout of acetic acid synthesis genes in VAL04 markedly reduced the accumulation of acetic acid, concurrently enhancing l-valine synthesis by conserving pyruvate that would otherwise have been converted to acetic acid. Engineering of the starting strain before ALE under oxygen-limited conditions. A: Engineering of the acetic acid synthesis pathway by sequentially deleting poxB, ackA, tdcD, and pta in the starting strain VAL0. B: The effect of overexpression of pgk and gapA on cell growth and l-valine synthesis. Another critical issue that is ubiquitously encountered by microorganisms under anaerobic or oxygen-limited conditions is the inadequate provision of ATP and imbalanced cofactor regeneration. Regarding the biosynthesis of l-valine under oxygen-limited conditions, sufficient ATP supply via the EMP pathway is required to guarantee cell growth. As the cofactor requirement for l-valine synthesis had already been changed from NADPH to NADH during the construction of the starting strain VAL0, sufficient NADH replenishment is necessary for efficient l-valine synthesis. In the EMP pathway, glyceraldehyde-3-phosphate dehydrogenase encoded by gene gapA catalyzes the conversion of glyceraldehyde-3-phosphate and NAD+ into 1,3-bisphosphoglycerate and NADH. Subsequently, 1,3-bisphosphoglycerate is converted to 3-phosphoglycerate by the catalysis of phosphoglycerate kinase encoded by gene pgk, generating one molecule of ATP. Thus, we further enhanced the expression of gapA and pgk to obtain an l-valine microbial workhorse more adaptable for the oxygen-limited condition. The pgk and gapA genes were respectively overexpressed under Ptrc promoter in strain VAL04, constructing strains VAL05 (VAL04, rph::Ptrc-pgk) and VAL06 (VAL04, rph::Ptrc-gapA). The oxygen-normal shake-flask fermentation characteristics of strains VAL04, VAL05, and VAL06 were compared, with the results shown in Fig. 1B. The l-valine titer and OD600 of strain VAL05 were 42.4 g/L and 23.9, respectively, showing no significant difference compared to the control strain VAL04. It indicates that the overexpression of the pgk gene had no substantial effect on l-valine biosynthesis or cell growth, possibly because phosphoglycerate kinase was not a rate-limiting enzyme in EMP pathway. For strain VAL06, the l-valine titer reached 44.5 g/L, representing a 6.0% increase over the control strain VAL04; but its cell OD600 decreased to 20.0, with a 16.7% decline compared to the control strain. This result may be attributable to the overexpression of gapA, which potentially elevated the intracellular NADH level and enhanced the metabolic driving force, thereby promoting carbon flux toward the NADH-consuming l-valine biosynthesis pathway. The intracellular NADH content and NADH/NAD+ ratio determination results confirm this deduction (Fig. S1). Consequently, the carbon flux into the TCA cycle was reduced, which in turn compromised cell growth, leading to increased l-valine production with reduced biomass formation. In other words, the yield of l-valine from glucose was significantly improved as the result of the enhancement of NADH replenishment. Thereby, strain VAL06 was readily prepared as a suitable chassis for subsequent ALE experiment. Adaptive laboratory evolution of VAL06 under oxygen-limited condition ALE experiment under oxygen-limited conditions was then carried out using VAL06 as the starting strain. Specifically, the tightly sealed reagent bottles filled with fermentation medium were used to create an oxygen-limited environment for ALE (Fig. 2A). Using limited oxygen as a selection pressure, improved cell growth and l-valine production was expected through the conduct of ALE. Every seven days, 1 mL of cell culture was transferred into another 100-mL reagent bottle that was filled with fresh fermentation medium. After six rounds of the above serial-passage culture, the strain exhibited improved growth performance. The evolved culture was then diluted and spread on agar plates to isolate single colonies. Selected clones were cultured in sealed 96-well plates under oxygen-limited conditions for 20 h. The results are shown in Fig. 2B. Two successive rounds of ALE conducted on strain VAL06 under oxygen-limited conditions. A: Schematic diagram illustrating the workflow of ALE under oxygen-limited conditions. B: Fermentation results of the potential positive mutant strains screened in the first-round ALE in a sealed 96-well microplate. C: Verification of positive mutant strains isolated from the first-round 96-well microplate screening in oxygen-limited shake-flask fermentation. D: Fermentation results of the potential positive mutant strains selected in the second-round ALE in a sealed 96-well microplate. E: Verification of positive mutant strains isolated from the second-round 96-well microplate screening in oxygen-limited shake-flask fermentation. As shown in Fig. 2B, the OD600 of evolved strains ALE1-1, ALE1-5, ALE1-23, ALE1-32, ALE1-33, ALE1-43, ALE1-45, ALE1-46, and ALE1-47 were significantly higher than others, indicating that these strains have potentially acquired improved adaptation to oxygen-limited condition. These strains were subsequently subjected to oxygen-limited shake-flask fermentation, and the results are presented in Fig. 2C. As shown in Fig. 2C, the l-valine titers of ALE1-1, ALE1-5, ALE1-23, ALE1-32, ALE1-33, ALE1-43, ALE1-45, ALE1-46, and ALE1-47 were 45.6, 45.2, 47.1, 46.9, 46.7, 49.0, 49.0, 47.0, and 46.4 g/L, respectively; with OD600 values of 19.2, 20.7, 19.3, 19.8, 18.9, 19.4, 20.0, 18.5, and 18.5, respectively. Compared to the starting strain VAL06, most evolved strains showed increased l-valine production, with ALE1-43 and ALE1-45 exhibiting the most notable improvements. However, only the OD600 values of ALE1-5 and ALE1-45 were higher than the control strain, while other strains displayed varied degrees of growth reduction. Although ALE1-5 demonstrated significantly improved growth, its l-valine production didn't show notable improvement. After comparison, ALE1-45 exhibited a significant increase in l-valine titer with improvement in cell growth under oxygen-limited conditions. Since the aim of this study was to identify strains with viable cell growth and enhanced production performance under oxygen-limited conditions, ALE1-45 was selected for the second round of ALE. The transfer interval was shortened from seven days to three days as cell growth gradually got better, and after eleven rounds of serial-passage evolutionary culture, strain growth was further enhanced. The evolved culture was diluted and plated again to isolate single colonies, which were then cultured in sealed 96-well plates under oxygen-limited conditions for 16 h. As shown in Fig. 2D, the OD600 values of evolved strains ALE2-5, ALE2-9, ALE2-11, ALE2-20, ALE2-29, ALE2-35, and ALE2-40 were significantly higher than those of the other strains. These selected strains were further evaluated via oxygen-limited shake-flask fermentation, and the results are presented in Fig. 2E. The l-valine titers of ALE2-5, ALE2-9, ALE2-11, ALE2-20, ALE2-29, ALE2-35, and ALE2-40 were 41.0, 41.5, 48.4, 50.8, 51.0, 48.5, and 51.8 g/L, respectively; and the OD600 values reached 19.5, 23.6, 19.4, 22.2, 22.0, 20.1, and 22.6, respectively. Compared to the control strain ALE1-45, strains ALE2-20, ALE2-29, and ALE2-40 exhibited improvements in both l-valine production and cell growth. In contrast, ALE2-5 showed a higher OD600 but a lower l-valine titer than the control strain. Notably, ALE2-40 achieved the highest level of l-valine production while maintaining good growth under oxygen-limited conditions, making it a promising candidate for further evaluation. Fermentation assessment of the evolved strain in a 5-L bioreactor The fermentation performance of the evolved strain ALE2-40 was further assessed in a 5-L bioreactor under oxygen-limited conditions, with the starting strain VAL06 as the control under the same conditions (Fig. 3). As depicted in Fig. 3, throughout the overall fermentation process, both the cell growth and l-valine production of ALE2-40 demonstrated a marked enhancement in comparison to VAL06, achieving a final l-valine titer of 85.4 g/L at 28 h (Fig. 3B), whereas, at the same time, VAL06 attained an l-valine titer of 78.5 g/L (Fig. 3A). Strain ALE2-40 attained its peak OD600 value of 25.9 at 20 h, whereas strain VAL06 reached a comparable OD600 level of 25.1 up to 28 h of fermentation. Despite of the fact that enhanced cell growth would unavoidably consume more carbon sources, the l-valine yield derived from the total consumed glucose utilized by ALE2-40 still surpassed that of VAL06. To be precise, ALE2-40 attained an l-valine yield of 58.0% from glucose, marking a 3.8% increase over the yield of 55.9% observed in VAL06. Furthermore, no notable accumulation of frequently encountered organic acids, such as acetic acid and succinic acid, was detected. These results clearly indicate that, as a result of the ALE experiment, the enhanced cell growth exhibited by strain ALE2-40 significantly contributed to the efficient l-valine production under oxygen-limited conditions. As shown in Fig. 3B, l-valine crystals began to precipitate from the fermentation broth of ALE2-40 at 28 h, indicating the end of the fermentation process to prevent imposing an additional burden on the following extraction procedure. Thus, for the 5-L bioreactor fermentation in this study, the fermentation period was set at 28 h. In contrast to our previous two-stage (aerobic-oxygen limited) fermentation approach, strain ALE2-40 constructed in this study adopted a single-stage oxygen-limited fermentation mode, which greatly reduced the fermentation duration from 36 h [2] to a mere 28 h. This innovative strategy not only boosted the l-valine titer but also markedly enhanced its yield from glucose, substantially elevating the fermentation performance of the strain. Comparative fed-batch fermentation analysis of the evolved strain ALE2-40 and the starting strain VAL06 under oxygen-limited conditions in 5 L bioreactors. A: Fermentation results of the starting strain VAL06. B: Fermentation results of the evolved strain ALE2-40. Comparative genomic analysis between the evolved strain and the starting strain To further investigate the mechanisms for the high l-valine production in strain ALE2-40 under oxygen-limited conditions, comparative genomic analysis between ALE2-40 and the starting strain VAL06 was first conducted under oxygen-limited conditions (Fig. 4A and B). There are four single nucleotide variants (SNVs) identified in ALE2-40 compared to the starting strain, involved in four genes: G961A in lpd, which encodes lipoamide dehydrogenase; T743C in yfdH encoding bactoprenol glucosyl transferase; A217C in rplP encoding 50S ribosomal subunit protein L16; and G259T in fimH encoding type 1 fimbriae d-mannose specific adhesin. Comparative genomic analysis of the evolved strain ALE2-40 and the starting strain VAL06. A: Genomic sequencing results of ALE2-40. B: Gene sequencing analysis of sequencing reads. C: Validation of candidate targets identified through comparative genomic analysis. To assess the contribution of these mutations to l-valine production, reverse metabolic engineering was employed. Each of these four mutated genes and the corresponding original genes were expressed in the starting strain VAL06 using a medium-copy plasmid pSTV28, controlled by the promoter PBBa_J23105 with medium transcriptional strength. The results of the shake-flask fermentation test are shown in Fig. 4C. Among these four mutations, the yfdH mutant and lpd mutant led to the most significant improvement in l-valine titer, while mutations in fimH and rplP showed no notable effect. According to the functional annotation in KEGG, yfdH encodes bactoprenol glucosyl transferase that catalyzed the transfer of glucose to undecaprenyl phosphate, forming undecaprenyl-phosphate glucose, a key precursor in bacterial cell wall biosynthesis. The mutation within the yfdH gene might potentially change the allocation of glucose flux among undecaprenyl-phosphate glucose synthesis, the central carbon metabolism, and l-valine biosynthesis pathway. However, to confirm the underlying mechanism, further experimental validation in the future work is essential. The lpd gene encodes lipoamide dehydrogenase, which catalyzes the transfer of electrons to the ultimate acceptor, NAD+. Lipoamide dehydrogenase is a critical component of the pyruvate dehydrogenase complex (PDHc), and 2-oxoglutarate dehydrogenase, which play critical roles in pyruvate metabolism and TCA cycle, respectively. The activity of PDHc is greatly inhibited under anaerobic conditions due to its high sensitivity towards NADH inhibition [12]. It has been reported that the NADH/NAD + ratio of an anaerobic E. coli cell is several-fold higher than that of an aerobic cell, which further leads to the inhibition of PDHc from high level of NADH [12]. The mutation in lpd has been reported effective in reducing the NADH sensibility of PDHc, which allows for increased PDH activity to facilitate cell viability of E. coli under anaerobic culture [13]. In this study, the activity of PDH in strain VAL06-lpd∗, ALE2-40, and VAL06 was detected and compared to explore the mechanism of lpd mutation acquired during oxygen-limited evolution, and the results are presented in Fig. S2A. As shown in the figure, the activity of PDH in strains ALE2-40 and VAL06-lpd∗ was found to be 70.8% and 70.6% higher, respectively, compared to that in the VAL06 strain. This result indicated that the evolutionary process indeed enhanced the functionality of PDH; concurrently, the lpd mutation in VAL06 elevated PDH activity to a level comparable to that observed in ALE2-40. In addition, the NADH level and the ratio of NADH/NAD+ in these three strains also presented similar tendencies (Fig. S2B). The above results suggested that the mutation in lpd effectively mediated the sensitivity of PDHc to NADH level under oxygen-limited conditions, allowing for viable cellular metabolism with balanced distribution of pyruvate among cell growth and l-valine synthesis. Comparative transcriptomic analysis between the evolved strain and the starting strain Subsequently, comparative transcriptomic analysis was conducted on both the starting strain VAL06 and the evolved strain ALE2-40 to further explore the evolutionary mechanism from transcription level under oxygen-limited conditions (Fig. 5A). Samples were collected at 12 h (mid-exponential growth stage) and 25 h (stable stage) of fermentation for comparative analysis. A total of 4240 genes were identified using the Fastp platform, with a Q20 quality score above 95.7% after quality control filtering. The average genome mapping rate of the samples was above 97.0%, indicating that the sequencing data was of high quality and suitable for further analysis. Comparative transcriptomic analysis of the evolved strain ALE2-40 and the starting strain VAL06. A: Scatter plot analysis of the number of differentially expressed genes at different time points. B: Venn diagram analysis of the number of differentially expressed genes at different time points. C: GO functional enrichment analysis of differentially expressed genes. D: Metabolic pathway enrichment analysis of differentially expressed genes. E: Shake-flask fermentation validation of upregulated genes identified through comparative transcriptomic analysis. F: Shake-flask fermentation validation of downregulated genes identified through comparative transcriptomic analysis. G: The intracellular NADH levels, NAD+ levels and NADH/NAD+ ratios in the starting strain, evolved strain and engineered strains. H: The intracellular ATP levels in the starting strain, evolved strain and engineered strains. The ∗ symbol represents 0.01 < P < 0.05, and the ∗∗ symbol represents P < 0.01. All statistical analysis data were compared to the control strains (VAL07 or ALE2-40). A diverse array of comparative analysis was performed on the transcriptomic data, encompassing an examination of gene transcriptional levels in VAL06 and ALE2-40 across various fermentation time points (Fig. 5A and B), a comparison of gene ontologies (Fig. 5C), and an analysis of functional enrichment (Fig. 5D). At 12 h of fermentation, 117 genes were upregulated, and 59 genes were downregulated in strain ALE2-40 compared to the starting strain. At 25 h of fermentation, 372 genes were upregulated, and 81 genes were downregulated in strain ALE2-40 compared to the starting strain (Fig. 5A). Venn diagram analysis of gene expressions at different time points revealed a total of 592 differentially expressed genes between the starting strain VAL06 and the evolved strain ALE2-40 (Fig. 5B). 17 differentially expressed genes were consistently identified at both sampling points (Fig. 5B). These 17 genes were subjected to further analysis, revealing that 7 genes were significantly upregulated and 10 genes were significantly downregulated. These 17 genes with significant expression changes were selected for further functional studies. Since the upregulated gene targets have already shown up-regulation in the ALE2-40 strain, we decided to individually overexpress these genes in the starting strain VAL06. We employed the plasmid pSTV28 for this purpose, and the gene transcription was controlled by the promoter PBBa_J23105. Through this strategy, we aimed to explore the effects of these upregulated genes on cell growth and the biosynthesis of l-valine. The results of shake-flask fermentation experiments, as illustrated in Fig. 5E, reveal that the overexpression of the genes yihA, ybcL, and napF notably enhanced the titer of l-valine. Among these genes, ybcL, which encodes a protein whose precise function remains elusive, exerted the most significant influence, elevating the l-valine titer by 14.6%, albeit with a concurrent 9.8% reduction in cell density measured at OD600. Notably, overexpression of the yihA gene, which is responsible for encoding a GTP-binding protein [14], and is essential for normal cell division, also led to a substantial 9.8% surge in l-valine production, albeit accompanied by a 15.2% decline in cell density measured at OD600. The overexpression of napF led to an 8.4% increase in l-valine titer without affecting cell growth, which encodes an iron-sulfur protein that is part of the periplasmic nitrate reductase (Nap) operon crucial for anaerobic growth in low-nitrate environments [15]. As a key component of the periplasmic nitrate reductase system, NapF probably benefits the energy generation under oxygen-limited conditions, potentially contributing to improved l-valine biosynthesis. For the validation of the downregulated genes, we decided to check the effect of gene knockout for verification. Given that deletion of genes can easily disrupt metabolic balance and adversely affect cell growth or the synthesis of l-valine, we opted to validate the deletion of downregulated targets in ALE2-40, which already acquired a more balanced cellular metabolism compared to VAL06 from ALE under oxygen-limited conditions. The results of shake-flask fermentation are shown in Fig. 5F. Among the tested genes, the respective knockout of five genes including rsxE, speC, yjiG, argI, and fimG, resulted in significant enhancement of l-valine titer, with increases of 9.6%, 9.6%, 9.1%, 7.9%, and 3.8%, respectively, without apparent cell growth damage in comparison with strain ALE2-40. The rsxE gene encodes a protein that is part of a membrane-bound complex that couples electron transfer with translocation of ions across the membrane, which is required to maintain the reduced state of SoxR. It probably transfers electron from NAD(P)H to SoxR. Thereby, the deletion of this gene probably saves NADH for l-valine synthesis. The speC gene encodes ornithine decarboxylase, a key enzyme in polyamine biosynthesis, catalyzing the formation of putrescine. Given that the formation of ornithine is essential for l-arginine biosynthesis to maintain cellular metabolism, which is a high-ATP demanding process, deletion of speC may improve the ATP utilization efficiency for essential amino acid biosynthesis via reducing the unnecessary expenditure. Although the precise function of the yjiG gene, which encodes an inner membrane protein, remains elusive, it has been identified as a protein containing a nucleoside transporter/FeoB GTPase gate domain in E. coli O1:K1/APEC. Similar function might also be present in the E. coli chassis used in this study, potentially linking it to the ATP pool through variations in the GTP pool. The argI gene encodes ornithine carbamoyltransferase, which catalyzes the synthesis of l-citrulline by transferring the carbamoyl group from carbamoyl phosphate to l-ornithine, following which step, l-arginine could be synthesized with l-citrulline as its precursor. Given that ArgF and ArgI could catalyze the same reaction, the knockout of argI probably also benefited ATP availability via slightly attenuated l-arginine synthesis. The fimG gene was involved in the assembly and adhesive function of type I fimbriae in E. coli, contributing to host adhesion and environmental adaptation. It has been reported that the fim genes including fimF, fimG, and fimH play a major role in fimbrial morphology as longitudinal modulators, among which fimF and fimG have interchangeable functions [16]. Thus, the deletion of fimG probably reduces the energy consumption associated with fimbriae elongation, effectively redirecting this saved energy towards the biosynthesis of l-valine. In summary, the five downregulated targets are either associated with NADH-related processes or with mechanisms for conserving ATP, both of which are pivotal for the synthesis of l-valine. To confirm the above deduction, the intracellular NADH/NAD+ ratios and ATP levels in strains VAL06, ALE2-40, ALE2-40-ΔrsxE, ALE2-40-ΔspeC, ALE2-40-ΔyjiG, ALE2-40-ΔargI, and ALE2-40-ΔfimG, were measured. As the results shown in Fig. 5G, the ratio of NADH/NAD+ in the evolved strain ALE2-40 significantly increased to 3.44-fold compared with the starting strain VAL06, suggesting improved NADH availability acquired in the evolutionary process. The knockout of rsxE, speC, yjiG, argI, and fimG in ALE2-40 all led to increased NADH/NAD+ ratios, with respective increases of 69.0%, 60.3%, 147.1%, 3.6% and 8.1%. Notably, the knockout of argI and fimG exerted only a relatively subtle influence on the NADH/NAD+ ratio. In contrast, the effects of the knockout of rsxE and yjiG were most prominent, suggesting important function in enriching intracellular NADH. Given that the function of rsxE is intricately linked to NADH consumption, the consequences of its deletion are reasonable; however, yjiG emerges as a novel target that exhibits a close association with the intracellular NADH pool. As depicted in Fig. 5H, the intracellular ATP content in strain ALE2-40 remained relatively stable, showing no significant difference when compared with that in strain VAL06. When contrasted with ALE2-40, the knockout of the rsxE gene or argI gene yield a slight improvement in the intracellular ATP level. The knockout of the yjiG gene led to a 19.3% increase in the intracellular ATP content. Notably, the knockout of the fimG or speC gene both resulted in substantial increases in the ATP content, with rises of 188.8% and 91.3%, respectively. It seems reasonable to suggest that the downregulation of gene expression observed in ALE2-40, stemming from the ALE experiment, exerted a comparatively modest influence on boosting the intracellular ATP pool. In contrast, if the genes fimG or speC were to be completely knocked out, their effects on enhancing the ATP pool would probably be significantly amplified. Transcriptomics-guided reverse metabolic engineering Through comparative transcriptomic analysis, differentially expressed genes within strain ALE2-40 were identified, and their impact on l-valine production was rigorously validated. The findings revealed that the upregulation of genes such as ybcL, yihA, and napF, as well as the downregulation of rsxE, speC, yjiG, argI, and fimG, significantly enhanced l-valine titer. Building on these insights, reverse metabolic engineering strategies were implemented to further enhance the fermentation performance of strain ALE2-40. Specifically, the PJ23105 promoter was employed to overexpress the upregulated genes, while downregulated genes underwent systematic pairwise deletion analyses to identify optimal combinations for enhancing l-valine production synergistically. First, the overexpression of previously identified upregulated genes was carried out in strain ALE2-40. Specifically, the ybcL, yihA, and napF genes were individually overexpressed in the genome of ALE2-40, resulting in the construction of strains VAL09 (ALE2-40, yciQ::PJ23105-ybcL), VAL10 (ALE2-40, ylbE::PJ23105-yihA), and VAL11 (ALE2-40, gapC::PJ23105-napF). The results of oxygen-limited shake-flask fermentation for these strains are depicted in Fig. 6A. The l-valine titers recorded for VAL09, VAL10, and VAL11 were 15.9, 27.6, and 27.0 g/L, respectively, revealing no significant enhancement in comparison to the control strain ALE2-40, which yielded 26.1 g/L l-valine. Notably, overexpression of the ybcL gene resulted in a marked reduction in both biomass (by 40.1%) and l-valine titer (by 39.1%), implying that the expression level of ybcL was already optimal, and any further upregulation probably disrupted the metabolic equilibrium of the strain, and compromised the fermentation performance. Conversely, VAL10 and VAL11 demonstrated biomass and l-valine production levels comparable to those of the control strain, suggesting that additional overexpression of these upregulated genes did not confer any beneficial effects. These observations collectively indicate that, following oxygen-limited adaptive evolution, the expression of these genes had attained a state of equilibrium, which might potentially be disrupted by a too strong expression of the same genes. The effects of omics-guided reverse metabolic engineering in ALE2-40. A: Effects of overexpression of the upregulated genes on the production performance of the strains. B: Effects of the knockout of the downregulated genes on the production performance of the strains. C: Comparison of the intracellular NADH levels, NADH/NAD+ ratios, ATP levels and PDH enzyme activity in engineered strain VAL18 and the evolved strain ALE2-40. D: Effects of the replacement of native promoter of gene aceE with PfliC on the production performance, intracellular NADH/NAD+ ratio and PDH enzyme activity in strain VAL19. E: Fed-batch fermentation results of strain VAL18 under oxygen-limited conditions in a 5 L bioreactor. F: Fed-batch fermentation results of strain VAL19 under oxygen-limited conditions in a 5 L bioreactor. Subsequently, the identified downregulated genes were further engineered in strain ALE2-40 via sequential pairwise deletion. Given that the deletion of potential essential genes would greatly affect cellular metabolism, cell growth and l-valine production of the engineered strains were closely examined. Following initial pairwise comparisons, the strain exhibiting superior l-valine production capacity was selected for subsequent combinatorial testing with additional downregulated gene targets in a secondary screening round. The rsxE gene, the deletion of which exhibited highest increase in boosting l-valine production (Fig. 5F), was first deleted in ALE2-40 to construct strain VAL12 (ALE2-40, ΔrsxE). Based on VAL12, we further tested the pairwise deletion of rsxE and other downregulated target genes. Specifically, genes fimG, speC, argI, and yjiG were respectively knocked out in VAL12, constructing strains VAL13 (VAL12, ΔfimG), VAL14 (VAL12, ΔspeC), and VAL15 (VAL12, ΔyjiG). However, despite multiple rounds of attempts, we were unable to successfully obtain the argI deletion in the rsxE knockout strain. The likely cause of this phenomenon was severe disturbance in cellular metabolism due to the combined deletion of rsxE and argI. Consequently, the combination involving the deletion of both argI and rsxE was abandoned. Oxygen-limited shake-flask fermentation comparison of the engineered strains was conducted, and the results are shown in Fig. 6B. The l-valine titers of VAL12–VAL15 were 27.4 g/L, 33.0 g/L, 31.9 g/L, and 27.2 g/L, respectively. In comparison to strain VAL12, strain VAL13 and VAL14 demonstrated 20.4% and 16.4% increases in l-valine titer, respectively, indicating that the co-deletion of rsxE with fimG (in VAL13), or rsxE with speC (in VAL14) both conferred significant metabolic advantages for l-valine biosynthesis. In contrast, deletion of rsxE with yjiG in VAL15 slightly reduced l-valine production. Thus, strain VAL13 with highest increase in l-valine production was chosen for a second round of pairwise deletion test. Subsequently, the speC and yjiG genes in the VAL13 strain were respectively deleted, thereby generating two engineered derivatives: VAL16 (VAL13, ΔspeC) and VAL17 (VAL13, ΔyjiG). The results of shake-flask fermentation are illustrated in Fig. 6B. The l-valine titers in VAL13, VAL16, and VAL17 were 33.0, 32.3, and 32.0 g/L, respectively. Although VAL16 and VAL17 exhibited marginally lower l-valine titers compared with VAL13, both mutants demonstrated a significant reduction in glucose consumption. This metabolic shift resulted in substantially improved yields of 49.2% (VAL16) and 51.0% (VAL17) from glucose, representing 11.6% and 15.6% increases over the VAL13 strain (44.1%). These findings suggest that the further deletion of downregulated genes steered the metabolic state of the strains towards a more efficient biosynthesis of l-valine. Considering the paramount importance of production yield in industrial bioproduction, strain VAL17 with the highest l-valine yield from glucose was selected as the optimal candidate for further metabolic engineering due to its superior l-valine productivity. Subsequently, we tested the combinatorial deletion of both the speC and yjiG genes in VAL17, resulting in the construction of strain VAL18 (VAL17, ΔspeC). As depicted in Fig. 6B, strain VAL18 attained an impressive l-valine titer of 38.6 g/L, along with a yield of 55.2% from glucose. Notably, after integrating all the advantageous deletions of downregulated genes, we observed marked enhancements in l-valine titer and yield. The above results suggested that the reverse metabolic engineering guided by the transcriptomic analysis further enhanced the synthesis of l-valine. To comprehensively elucidate the mechanism underlying the enhanced l-valine production in strain VAL18, a comparative analysis was conducted on the ATP content, NADH/NAD+ ratio, and pyruvate dehydrogenase activity between VAL18 and ALE2-40. As shown in Fig. 6C, when these four target genes were sequentially knocked out, notable rises in NADH/NAD+ ratio, PDH activity, and ATP content were observed in VAL18, with respective increases of 70.0%, 74.3%, and 53.5% in comparison to that in ALE2-40. The possibility exists that the effect of the downregulation of these genes, which developed through adaptive evolution, could be substantially amplified upon their complete deletion. Reducing futile ATP consuming has proven effective in reserving ATP for the bioproduction of chemicals with high ATP demand [11]. Although the synthesis of l-valine is not directly high-ATP-demanding, ATP supply is essential for the engineered strain under oxygen-limited conditions. Given that the mutation of the lpd gene resulted in an elevation of the enzymatic activity of pyruvate dehydrogenase (PDH), the function of which not only confers advantages to cell viability and growth under oxygen-limited conditions but also contributes to a higher level of pyruvate consumption. It was therefore proposed that, if we can further fine-tune the expression of the aceE gene, which encodes pyruvate dehydrogenase E1, for instance, by dynamically down-regulating it during fermentation, it is likely that the production efficiency of l-valine could be further benefited. Consequently, the expression of the aceE gene in VAL18 was subjected to further refined control through the utilization of a growth-coupled promoter PfliC. PfliC functions as the native promoter for the fliC gene, which encodes the structural protein of the flagellar filament during the growth phase. In such a design, the transcription of aceE, regulated by PfliC, would be dynamically deactivated during the stationary phase of the culture. Accordingly, strain VAL19 was obtained. The fermentation capability of VAL19 was compared with VAL18 (Fig. 6D). The results indicated that the regulation of aceE controlled by PfliC slightly decreased the activity of PDH and benefited the l-valine production in VAL19. It is noteworthy that the cell growth of strain VAL19 exhibited a slight decrease in comparison to that of VAL18. This observation suggests the potential for an enhanced yield of l-valine from glucose in VAL19. Fed-batch fermentation of the engineered strains in a 5-L bioreactor under oxygen-limited conditions To conduct a thorough assessment of the fermentation performance of the engineered strain VAL18 and VAL19, 5-L bioreactor fermentation was carried out. An oxygen-limited fermentation condition was created with the stirrer speed setting to mere 300 rpm and aeration setting to as low as 0.2 L/min. The results are shown in Fig. 6E (VAL18) and Fig. 6F (VAL19). Owing to the limited solubility of l-valine, the fermentation process was terminated at 28 h, when the concentration of l-valine reached its saturation point and began to precipitate (Fig. 6E). It is noteworthy that the cell growth of strain VAL19 outperformed that of VAL18 at 12 h and 16 h, while exhibited a decline in comparison to that of VAL18 after 24 h. The final biomass levels (measured as OD600) for strains VAL18 and VAL19 were 22.5 and 20.5, respectively. This phenomenon is most probably attributable to the fact that, during the cell growth phase, PfliC facilitates transcriptional initiation with a certain strength, and as cell growth progresses into the late phase, it exerts a suppressive effect on transcriptional initiation. Consequently, a partial decoupling of cell growth from l-valine synthesis was achieved, contributing to the enhanced synthesis of l-valine from glucose in VAL19. The l-valine titer in strain VAL19 reached a remarkable 93.7 g/L. This value represents a 3.4% increase compared to the l-valine titer of strain VAL18, which stood at 90.6 g/L. Moreover, referring to Fig. 3, when compared with VAL06 (78.5 g/L) and ALE2-40 (85.4 g/L), the l-valine titer in VAL19 was 19.4% and 9.7% higher, respectively. In addition, strain VAL19 demonstrated a yield as high as 60.4%, which corresponds to 92.9% of the theoretical yield of 65% from glucose, outperforming VAL18's yield of 59.2%. In summary, the engineered strain VAL19, derived from the evolved strain ALE2-40, demonstrated simultaneous improvements in both l-valine titer and yield following omics-guided reverse metabolic engineering. These findings suggest that the synergistic integration of irrational ALE with rational reverse engineering strategies offers significant potential for developing high-performance microbial cell factories tailored for industrial bioproduction. Discussion In the construction of microbial cell factories, the quest for achieving high titer, yield, and productivity stands as the most prevalent objective within laboratory settings. However, in the actual industrial application, high titer does not invariably represent the optimal target. For instance, certain chemicals exhibit low solubility, and an excessively high titer can result in crystal precipitation. Such precipitation poses a significant obstacle to downstream purification processes. In such scenarios, a combination of higher yield and productivity, along with an appropriate titer, proves to be more applicable and economically efficient. The production of l-valine is a typical process that is confronted with such issue, as an elevated titer is often accompanied by the issue of l-valine crystallization. Under such circumstances, the increase of yield derived from glucose serves as a critical bottleneck and a primary concern for reducing production costs. Due to the function of TCA cycle under aerobic conditions, the divert of pyruvate to TCA cycle has been a main factor limiting the yield of l-valine. In this study, the adaptive evolution of an engineered l-valine producing strain in sealed bottle filled with fermentation medium improved the cell growth under oxygen-limited conditions and significantly benefited the yield of l-valine from glucose. It demonstrated that the “live or die” stress constitutes a direct and highly effective driving force for evolution. After the ALE experiment, strain ALE2-40 acquired both better cell growth and enhanced l-valine production. To our surprise, the genomic mutations identified were limited to merely four single-nucleotide variants. Notably, the mutation in the lpd gene has been validated to effectively elevate both the NADH/NAD+ ratio and the activity of pyruvate dehydrogenase. This, in turn, enhances cell viability under oxygen-limited conditions and supplies an increased amount of NADH for the synthesis of l-valine. In our opinion, this feature is a key contributor to the robust and efficient production of l-valine by the evolved strain ALE2-40 under oxygen-limited conditions. In contrast to the limited number of targets identified via genomics analysis, a far greater number of targets have been revealed through transcriptomics analysis. Interestingly, upon further augmenting the expression levels of the beneficial upregulated genes in the evolved strain ALE2-40, no incremental enhancement in l-valine production was observed. In contrast, the stepwise deletion of the beneficial downregulated genes, specifically rsxE, fimG, yjiG, and speC, resulted in a marked elevation in l-valine yield (strain VAL18), which was underscored by an increased l-valine titer with a slight decline in cell growth. It is intriguing to note whether these targets are knocked out individually (Fig. 5G and H) or collectively (Fig. 6C), they collectively exert a synergistic impact on the enrichment of intracellular NADH and ATP. While the individual deletion of each of these four genes has been demonstrated to be advantageous for l-valine synthesis, the outcomes of their combined knockout present intriguing findings. In the first-round combination, strain VAL13, characterized by the concurrent deletion of rsxE and fimG—genes that respectively enhance NADH and ATP availability—exhibits markedly superior performance when compared to other combinations. Subsequently, upon the deletion of fimG, which was anticipated to enhance ATP availability, the subsequent integration of the yjiG deletion (resulting in strain VAL17), known for its potential to elevate NADH availability, outperformed the extra boost in ATP supply via the deletion of speC (strain VAL16). The benefits of speC deletion (strain VAL18) became evident only after the yjiG deletion had already enhanced the intracellular NADH level. The above-mentioned results indicate that achieving a balance in the supply of NADH and ATP, rather than allowing their levels to become excessively high, holds the utmost significance for the efficient synthesis of l-valine. Growth dependent promoters have been widely used for dynamically fine-tuning cellular metabolism. Drawing inspiration from the influence of the lpd mutant on both the activity of PDH and the production of l-valine, we employed the growth-dependent promoter Pflic, renowned for its ability to shut down gene transcription during the late-stage fermentation. In the construction process, we also tested the utilization of constitutive weak promoters for the control of aceE, however, the application of weak promoters greatly damaged cell growth and l-valine production (data not shown). In contrast, the utilization of Pflic appropriately mediated the balance between cell growth and l-valine production. Ultimately, leveraging metabolic engineering based on adaptive evolution and omics technology culminated in the development of a highly efficient microbial cell factory for l-valine production under oxygen-limited conditions, which is currently the most efficient l-valine producing strain (Table 2). Table 2Advances of the bioproduction of l-valine in representative microorganisms.Table 2MicroorganismStrainsTime (h)Titer (g/L)Yield (%)Productivity (g/L/h)Fermentation modeReferencesBrevibacteriumflavumB. flavum MDVR2-215670.1391.10Two-stage fermentation[17]KlebsiellaoxytocaK. oxytoca VKO-95612258.72.18Aerobic fermentation[18]C. glutamicumC. glutamicum aceE A16 Δpqo Δppc (pJC4 ilvBNCE)5586.5231.57Aerobic fermentation[19]C. glutamicum BNGECTMDLD/ΔLDH48227354.73Cell catalysis[4]C. glutamicum Val-924150576.25Cell catalysis[6]C. glutamicum Val1938103352.67Aerobic fermentation[8]E. coliE. coli VHY183684412.33Two-stage fermentation[2]E. coli VAL384892341.92Aerobic fermentation[20]E. coli VAL322890.6483.24Two-stage fermentation[9]E. coli 0454629 batches71.6572.05Two-stage fermentation[21]E. coli VAL192893.760.43.35Oxygen-limited fermentationThis study Advances of the bioproduction of l-valine in representative microorganisms. The findings of this study emphasize the importance of ATP supply and cofactor balance for both the viability of microorganisms under oxygen-limited conditions and efficient production, and they also demonstrate the powerful potential of combining rational and irrational strategies in constructing microbial cell factories. In future work, enhancing ATP generation by introducing a heterologous ATP-generating pathway or further optimizing NADH regeneration could further improve l-valine synthesis efficiency. Conclusion In this study, we successfully developed a highly efficient strain capable of producing l-valine with an extraordinary yield, leveraging a combined approach involving adaptive laboratory evolution (ALE) and reverse engineering guided by genomic and transcriptomic data. Initially, we engineered an optimal chassis for ALE by enhancing glycolysis and reducing acetic acid accumulation. Subsequently, ALE conducted under oxygen-limited conditions significantly boosted l-valine production. Further, comparative genomic and transcriptomic analysis between the evolved strain and the control strain identified potential targets for enhancing l-valine production, which guided our subsequent reverse engineering endeavors. The final strain, VAL19, achieved an impressive l-valine titer of 93.7 g/L, with a yield of 60.4% from glucose and a productivity of 3.35 g/L/h, under oxygen-limited fermentation conditions in a 5-L bioreactor. To the best of our knowledge, this represents the highest yield for l-valine production in E. coli, marking a significant advancement in the development of efficient microbial cell factories for industrial bioproduction. CRediT authorship contribution statement Hengshuai Zhang: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation. Zhirui Ha: Writing – original draft, Visualization, Investigation. Aiying Wei: Visualization, Formal analysis. Chunguang Zhao: Visualization, Formal analysis. Gang Meng: Visualization, Formal analysis. Xin Li: Investigation, Data curation. Faqing Wu: Methodology, Investigation. Jingui Liu: Visualization, Supervision, Methodology. Heyun Wu: Visualization, Supervision. Qian Ma: Writing – review & editing, Visualization, Supervision. Xixian Xie: Writing – review & editing, Supervision, Funding acquisition, Conceptualization. Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Zhirui Ha, Aiying Wei, Chunguang Zhao and Gang Meng are currently employed by Ningxia Eppen Biotech Co., Ltd. Acknowledgement This research was supported by the National Key Research and Development Program of China (2022YFA0911800), the Key Research and Development Program of Ningxia Hui Autonomous Region (2025BEE02021), National Natural Science Foundation of China (22378315, 32200038) and Natural Science Foundation of Tianjin (24JCYBJC00830, 23JCYBJC00860). Footnotes Peer review under the responsibility of Editorial Board of Synthetic and Systems Biotechnology. Supplementary data to this article can be found online at https://doi.org/10.1016/j.synbio.2026.05.011. Contributor Information Qian Ma, Email: qianma1987@tust.edu.cn. Xixian Xie, Email: xixianxie@tust.edu.cn. Supplementary data The following is the Supplementary data to this article. Multimedia component 1 References Associated Data Supplementary Materials Multimedia component 1