Biosynthesis of isobutyraldehyde, the precursor for propane production
First, an isobutyraldehyde synthetic pathway was constructed based on the valine pathway of E. coli (green part of Fig. 1) [15, 16]. The alsS gene from B. subtilis was introduced to convert pyruvate into 2-acetolactate, which is the first metabolic intermediate of valine pathway. For the second and third steps, the ilvCD genes of E. coli were overexpressed to convert 2-acetolactate into 2-ketoisovalerate. In the last two steps, Kivd from Lactococcus lactis transformed 2-ketoisovalerate into isobutyraldehyde, which could be converted into propane further by the ADO from Prochlorococcus marinus MIT 9313. All genes mentioned above were cloned into corresponding vectors, resulting plasmids pBAD33-alsS-Kivd, pAL96-ilvCD, and pET-PMT1231 (Additional file 1: Figure S1). To verify the feasibility of this pathway, we transformed those plasmids into E. coli BL21(DE3), anticipating fulfilling the biosynthesis of propane. However, there was no propane detected in the gas sample of headspace vial. Analyzing the culture components suggested that the isobutyraldehyde produced by valine pathway was reduced into isobutanol by endogenous ALRs of E. coli. It is well known that E. coli harbors many endogenous ALRs which transform aldehydes into alcohols, and they generate a great obstacle for alkane production [15, 16]. Therefore, it is essential to delete ALRs of E. coli host strain for propane production.
A series of rationally targeted deletions of ALRs were carried out in E. coli BW25113, which had been proven well amenable for genetic engineering [17]. A two-step markerless recombination method was applied in this work [18, 19]. In light of the diverse specificity of ALRs on isobutyraldehyde and the functions that ALRs undertake in the metabolism of E. coli, we deleted 9 ALR genes, namely yqhD, adhE, adhP, eutG, yiaY, yjgB, fucO, yahK, and DkgA, following the rule that the deletions can decrease the ALR activity of E. coli and meanwhile do not inhibit cellular growth. The genetically engineered strain was verified by sequencing and named as BW25113 Δ13 (Additional file 1: Figure S2; Table S3). Given that 13 genes were deleted from the genome and some of those genes might have multiple functions, we inspected whether the strain could grow normally. The result indicated that the deletions did not hamper its growth (Fig. 2a). Although these gene deletions have no effect on cell growth, it is important to realize possible metabolic unbalance caused by the genetic manipulations here. For example, the fnr gene controls the transition between anaerobic and aerobic respiration by regulating a gene network which includes adhE, frdABCD, ldhA, and pflB [20]. In anaerobic condition, they use respective substrates as the terminal electron acceptors for oxidative phosphorylation, maintaining redox balance, and cellular growth [21]. The detailed examinations of these gene deletions on the physiological metabolism of the mutant strain are needed in the future study. We then evaluated the ability of BW25113 Δ13 to produce and accumulate isobutyraldehyde. Compared to the wild-type BW25113 which only accumulated 0.3 g/L isobutyraldehyde and 0.57 g/L isobutanol in 20 h, the engineered strain BW25113 Δ13 achieved an isobutyraldehyde titer up to 1.1 g/L, while very low amount of isobutanol was detected, indicating that 95 % activity of E. coli ALRs was eliminated (Fig. 2b). The results demonstrated that an E. coli strain nearly deprived of ALR activity had been successfully constructed, which could accumulate isobutyraldehyde as the major fermentation product.
Conversion of isobutyraldehyde into propane
In order to overexpress ADO with the pET system in the isobutyraldehyde-producing strain, the mutant strain BW25113 Δ13 was lysogenized by λDE3 lysogenization kit to build strain BW25113(DE3) Δ13. The plasmids pBAD33-alsS-Kivd, pAL96-ilvCD, and pET-PMT1231 were co-transformed into BW25113(DE3) Δ13, resulting in the strain BW25113(DE3) Δ13/Propane with an intact pathway for propane production (Additional file 1: Table S3). The strain was cultivated in TB media (30 g glucose/L) and induced with l-arabinose and isopropyl β-d-thiogalactopyranoside (IPTG) to produce propane. Finally, the propane was determined by gas chromatography (GC) (Fig. 3). The result showed that the strain BW25113(DE3) Δ13/Propane can successfully synthesize propane and no propane was formed in two control strains.
Improvement of propane production by modifying ADO substrate specificity toward isobutyraldehyde
Although ADO is capable of catalyzing a wide range of aldehydes (C3–C18) to produce alka(e)nes, it exhibits low activity with the short-chain aldehydes (C3–C5) [7, 22, 23]. Given that the engineered valine pathway has produced abundant isobutyraldehyde and may achieve higher isobutyraldehyde titer after further metabolic optimizations, poor activity of ADO on isobutyraldehyde emerges as the major obstacle for higher propane productivity. To address this obstacle, we sought to enhance catalytic efficiency of ADO on isobutyraldehyde through rational design based on ADO structural analysis.
Crystal structures of ADO (PMT1231) in complex with different substrates, some of which come from our previous studies, provide us an insight into substrate binding and catalytic mechanism of active center (Fig. 4a; Additional file 1: Figure S3) [9, 23–26]. We found that the substrate binding pocket of ADO capable of accommodating linear C18 aldehyde has too much redundant space for isobutyraldehyde (C4), while the width of substrate channel is relatively narrow for the branched methyl of isobutyraldehyde, especially near to the active center (Fig. 4a). It can be speculated that they are part of the reasons why ADO shows poor activity with the branched short-chain aldehydes. We therefore sought to engineer the substrate channel of ADO. First, residues Ala134 and Val41 were mutated into Phe or Tyr, aiming at that the introduced residues could shrink redundant space at the end of substrate binding pocket to fit the short chain (Fig. 4a; Additional file 1: Figure S3). Second, residues Ile127, Ala48, Ala131, Tyr135, Gln123, Gln123, Phe100, Ile40, and Ile37 were mutated into Gly or Ala to broaden substrate channel to accommodate the branched group (Fig. 4a; Additional file 1: Figure S3).
A whole-cell biotransformation method was used to assay the activity of the mutants [23]. Most mutants led to decline in propane production except the mutant I127G, which resulted in an 83 % increase in propane productivity compared to wild-type ADO (Fig. 4b; Additional file 1: Table S4). Based on this mutant, we subsequently investigated more double mutations (Additional file 1: Table S5). The mutants I127G/I37G and I127G/V41G increased upon propane productivity of I127G by 10.12 and 15.10 %, respectively. The propane productivity of the mutant I127G/A48G improved further and was more than two times greater than wild-type ADO (Fig. 4b; Additional file 1: Table S5). These results are corresponding with the in vitro enzymatic assay (Fig. 4c; Additional file 1: Table S6). Moreover, SDS-PAGE analysis showed that the mutant expression levels were consistent with wild-type ADO (Additional file 1: Figure S4).
We then introduced the mutations I127G and I127G/A48G into the strain BW25113(DE3) Δ13/Propane described above to investigate whether they could enhance de novo propane production. Compared to wild-type ADO, the pathway carrying mutant I127G and double mutant I127G/A48G yielded a 3-fold (267 μg/L) and 2.5-fold (233 μg/L) more propane, respectively (Fig. 4d). Although the double mutant I127G/A48G had better performance in whole-cell biotransformation and enzymatic assay experiments, it was the single mutant I127G that got highest propane productivity in de novo synthesis.