Assembly of linear methanol utilization pathway in E. coli
The linear methanol utilization pathway consists of two steps: oxidation of methanol into formaldehyde and carboligation of formaldehyde into DHA, which can be phosphorylated to dihydroxyacetone phosphate by dihydroxyacetone kinase and enter lower glycolysis (Fig. 1a). According to the calculation by eQuilibrator (Flamholz et al. 2012), the ΔrG′o values for the linear pathway and the RuMP pathway are 9.1 and − 3.4 kJ/mol, respectively, suggesting that the RuMP pathway is more thermodynamically feasible. However, the product of the linear pathway can enter glycolysis and be metabolized quickly, providing a strong driven force for methanol utilization. Therefore, the linear pathway is also supposed to be feasible. To assemble the linear pathway in vivo, NAD+-dependent MDH from Bacillus methanolicus and artificial FLS were overexpressed in E. coli. The multicopy plasmid pTrc99A with a strong trc promoter was used to achieve high-level expression of MDH and FLS since their specific activities are quite low (Krog et al. 2013; Siegel et al. 2015). Two recombinants Ec-ΔfrmA-mdh3
MGA3-fls and Ec-ΔfrmA-mdh2
PB1-fls carrying the fls gene and different mdh genes (Additional file 1: Table S3) were constructed. Enzyme activity assays demonstrated that both MDHs were functionally expressed. Strain Ec-ΔfrmA-mdh2
PB1-fls showed higher methanol oxidation activity that was approximately twice as high as the MDH activity of strain Ec-ΔfrmA-mdh3
MGA3-fls (Fig. 1b). FLS activity was determined by coupled reactions involving DHA formation from formaldehyde by FLS and DHA reduction by NAD+-dependent glycerol dehydrogenase. However, the reverse activity of MDH that could reduce formaldehyde to methanol with NADH consumption interfered with the NADH-dependent DHA reduction. Therefore, FLS activity was not determined here, whereas SDS-PAGE analysis indicated that MDHs and FLS were successfully expressed (Fig. 1c). Strain Ec-ΔfrmA-mdh2
PB1-fls was used in the subsequent experiments for higher MDH activity.
Bioconversion of methanol into biomass of the engineered E. coli
Despite the equipment of linear methanol assimilation pathway, the engineered strain could not initiate growth in M9 minimal medium with methanol (approximately 8 g/L, 1% v/v) as the sole carbon source. Similar phenomena were observed in previous studies on RuMP-based synthetic methylotrophs and undefined supplements such as yeast extract and tryptone were added to initiate cell growth on methanol (Whitaker et al. 2017). Thus, small amounts of yeast extract (1 g/L) were added in M9 minimal medium. Any improvements in cell growth in the presence of methanol might derive from the contribution of methanol assimilation. As controls, a ΔfrmA strain containing the empty pTrc99A vector (strain Ec-ΔfrmA-pTrc99A) was cultivated using the aforementioned media. A methanol evaporation control without inoculation was also conducted.
As shown in Fig. 2a, approximately 1 g/L methanol evaporated away during the cultivation. When the control strain Ec-ΔfrmA-pTrc99A was cultivated using yeast extract and methanol as co-substrate, no additional methanol consumption but slightly decrease in cell growth was observed (Fig. 2a, b). We speculated that methanol might be oxidized to toxic intermediate formaldehyde by the non-specific activities of alcohol dehydrogenases of E. coli, which affected the cell growth negatively. Regarding to strain Ec-ΔfrmA-mdh2
PB1-fls, no significant increase in biomass was observed when methanol was added (Fig. 2c), whereas slightly more methanol (1.45 g/L) was consumed compared to the evaporation control, suggesting methanol utilization of the engineered strain. To further validate methanol utilization, 13C-methanol-labeling experiment was performed. When 13C-methanol was used as a carbon source, 13C-labeled amino acids in biomass including alanine, aspartic acid, glutamic acid, phenylalanine, proline, glycine, lysine, serine, threonine, tyrosine, and 13C-labeled citric acid were detected (Fig. 3a; Additional file 1: Table S4). It has been reported that amino acids measurement could provide isotopic labeling information about eight crucial precursor metabolites in the central metabolism (You et al. 2012). The presented results showed that biosynthesis of key intermediates of glycolysis, PP pathway and TCA cycle, including 3-phosphoglycerate, phosphoenolpyruvate, pyruvate, acetyl-CoA, α-ketoglutarate, oxaloacetate, and erythrose 4-phosphate, withdrew carbon from 13C-methanol.
Adaptive evolution of the engineered E. coli to improve methanol utilization
To further improve the microbial performance in methanol medium and screen methanol-utilizing mutants, adaptive evolution based on GREACE (genome replication engineering assisted continuous evolution) was conducted (Luan et al. 2013). A proofreading-defective element of the DNA polymerase of E. coli (ε subunit encoded by dnaQ gene) was expressed in strain Ec-ΔfrmA-mdh2
PB1-fls to introduce random mutations into the genomic DNA during continuous passage cultivation in LB medium. For each passage, cells were transferred into M9 minimal medium supplemented with methanol to enrich potential mutants with improved cell growth on methanol. Mutants were then isolated from the culture and a mutant with the best cell growth on methanol was isolated and designated as Ec-ΔfrmA-mdh2
PB1-fls-M11 (Additional file 1: Figure S1). When mutant Ec-ΔfrmA-mdh2
PB1-fls-M11 was cultivated in M9 minimal medium supplemented with 1 g/L yeast extract and methanol, 2 g/L methanol was consumed, which was more than that consumed by its parent strain Ec-ΔfrmA-mdh2
PB1-fls. It was noticed that biomass of the mutant declined after 5 h and addition of methanol helps maintain the biomass (Fig. 2d). We speculated that such decline in cell growth was caused by the random mutations introduced by GREACE. 13C-methanol-labeling experiment was then conducted and the results validated that mutant Ec-ΔfrmA-mdh2
PB1-fls-M11 assimilated more 13C-methanol into biomass (Fig. 3b; Additional file 1: Table S5). The results demonstrated that coupling of metabolic engineering and adaptive evolution was an enabling strategy to endow microorganisms with the ability to utilize methanol.
Synthetic methylotrophs have been constructed by heterogenous expressing MDH and RuMP genes (Pfeifenschneider et al. 2017). RuMP cycle depends on regenerating the formaldehyde acceptor Ru5P, which requires high coordination of many enzymes involved in formaldehyde assimilation and PP pathway (Whitaker et al. 2015). On the contrary, the linear formaldehyde assimilation pathway used in this study only requires one enzyme FLS and directly produces C3 intermediate DHA, which could be a great advantage for pathway engineering. Previous and the present studies revealed that constructing synthetic methylotrophs was far more complicated than complementing metabolic pathways where several crucial factors need to be considered, such as how to keep the intracellular formaldehyde concentration below the toxicity threshold (Witthoff et al. 2015) and how to balance the reducing equivalent generated by methanol oxidation (Price et al. 2016). In this case, combining metabolic engineering and adaptive evolution could be an easy strategy to prepare a desirable mutant that assimilates methanol efficiently. By using such a combined strategy, improved methanol assimilation was obtained in the mutant Ec-ΔfrmA-mdh2
PB1-fls-M11. Whole-genome resequencing revealed that no mutation was introduced into the plasmid, which was consistent with the unchanged MDH activity (Fig. 1b). Meanwhile, 66 missense, synonymous, and intergenic mutations that covered amino acid transport and metabolism, signal transduction, cell wall/membrane/envelope biogenesis, etc. were discovered (Additional file 2: Table S6). Further investigation of these mutations will likely elucidate key factors of methanol utilization in synthetic methylotrophs.