Modification of an engineered Escherichia coli by a combinatorial strategy to improve 3,4-dihydroxybutyric acid production

3,4-Dihydroxybutyric acid (3,4-DHBA) is a multifunctional C4 platform compound widely used for the synthesis of various materials, including pharmaceuticals. Although, a biosynthetic pathway for 3,4-DHBA production has been developed, its low yield still precludes large-scale use. Here, a heterologous four-step biosynthetic pathway was established in recombinant Escherichia coli (E. coli) using a combinatorial strategy. Several aldehyde dehydrogenases (ALDHs) were screened, using in vitro enzyme assays, to identify suitable catalysts for the dehydrogenation of 3,4-dihydroxybutanal (3,4-DHB) to 3,4-DHBA. A pathway containing glucose dehydrogenase (BsGDH) from Bacillus subtilis, d-xylonate dehydratase (YagF) from E. coli, benzoylformate decarboxylase (PpMdlC) from Pseudomonas putida and ALDH was introduced into E. coli, generating 3.04 g/L 3,4-DHBA from d-xylose (0.190 g 3,4-DHBA/g d-xylose). Disruption of competing pathways by deleting xylA, ghrA, ghrB and adhP contributed to an 87% increase in 3,4-DHBA accumulation. Expression of a fusion construct containing PpMdlC and YagF enhanced the 3,4-DHBA titer, producing the highest titer and yield reported thus far (7.71 g/L; 0.482 g 3,4-DHBA/g d-xylose). These results showed that deleting genes from competing pathways and constructing fusion proteins significantly improved the titer and yield of 3,4-DHBA in engineered E. coli.

Currently, 3,4-DHBA is mainly produced using an H 2 O 2 -catalyzed reaction between glucose and alkali metal hydroxides conducted at 70°C for 24 h (Hollingsworth 1994), or from the hydrocyanation and hydrolyzation of R-3-chloro-1,2-propanediol (Inoue et al. 1991). However, these approaches usually generate low yields, in addition to involving harsh reaction conditions, complex product purification processes, high costs, and environmental pollution. Therefore, the biosynthesis of 3,4-DHBA from renewable sources has attracted considerable attention. For instance, efforts have been made to produce 3,4-DHBA by microbial fermentation.
There have been no reports of a natural pathway for 3,4-DHBA production. The first de novo biosynthesis of 3,4-DHBA, developed in E. coli, involved a six-step enzymatic reaction using glucose, and acyl-CoA as a substrate and acyl donor (Liu et al. 2009;Schweiger and Buckel 1984;Taguchi et al. 2008). After optimizations by Dhamankar et al., a yield of 0.7 g/L 3,4-DHBA (0.105 g 3,4-DHBA/g glucose) was achieved (Dhamankar et al. 2014). However, the numerous reaction steps, low catalytic efficiency, and low yield make this pathway unsuitable for industrial production.
D-xylose has a high proportion of lignocellulose, which is as the most abundant renewable biomass in energy production found in nature (Choi et al. 2014;Kawaguchi et al. 2016). A novel five-step biosynthetic pathway for the generation of 3,4-DHBA from Dxylose has been constructed in E. coli (Wang et al. 2017), in which xylose is catalyzed to 3,4-DHBA by double dehydrogenation followed by oxidation, dehydration and decarboxylation. In shake flask experiments, the titer of 3,4-DHBA has achieved 1.27 g/L (0.079 g 3,4-DHBA/g D-xylose), with 0.18 g/L (0.012 g 1,2,4-butanetriol/g D-xylose) of by-product of 1,2,4-butanetriol (BTO). This yield has remained low because of the lack of highly active enzymes to efficiently decarboxylate 2-keto-3-deoxy-D-xylonate and dehydrogenate 3,4-DHB, and because intermediate products are consumed by competing pathways.
In this study, a four-step biosynthetic pathway using D-xylose as a precursor was developed to produce 3,4-DHBA (Fig. 1). First, several aldehyde dehydrogenases (ALDHs) were screened and characterized to identify potential catalysts suitable for maximizing 3,4-DHBA accumulation. Second, competing pathways were disrupted by deleting xylA, ghrA, ghrB and adhP. Finally, a fusion construct containing PpMdlC and YagF was created to improve the titer of 3,4-DHBA. This study provided novel strategies for achieving a high titer and yield of 3,4-DHBA.

Methods and materials
Strains, plasmids and culture conditions The E. coli BL21(DE3) was used for expressing, purifying target proteins and producing 3,4-DHBA. Strains used in this study were showed in Table 1. The plasmids pET28a and pACYC184 were used for cloning and expressing the target genes. E. coli cells were cultured at 37°C in lysogeny broth medium added with appropriate concentration antibiotics. And plasmids in this study were showed in Table S1.

Construction of plasmids
The primers sequences used are listed in Table S2. pE01 harboring yagF from E. coli and PpmdlC from P. putida was constructed. Similarity, pE04, pE05 and pE06 carrying gdh from B. subtilis and different ALDH genes, respectively.
The gene segment of T7 promoter-PpmdlC was amplified from pET28a harboring PpmdlC. The backbone of pET28a harboring yagF with SpeI and SalI of multiple cloning site was obtained by PCR. A 5 nm rigid a-helical ER/K linker (Sivaramakrishnan and Spudich 2011) was selected to connected to these two segments to create a new plasmid which carrying a fusion construct (Fig. S1).

The disruption of competing pathways
The disruption of competing pathway was accomplished by CRISPR-Cas9. The donor DNA was connected by overlapping PCR from 500 bp upstream and downstream homologous arm. The E. coli BL21(DE3) carrying pCas vector was inoculated in 30°C and then added into 10 mM arabinose for inducing the expression of Cas9. When OD 600 reached 0.5-0.6, the E. coli cells were arranged on ice for 30 min. After washed by 10% glycerol sterile three times, it was split into 100 lL per tube. The competent cells carrying pCas were mixed with 100 ng pTargetT vector and 400 ng donor DNA to have an electrotransformation (2.5 kV, 200 X, 25 lF) with an electrical shock time within 5 ms. Then they were cultivated into 30°C and confirmed by colony PCR.
E. coli cells were cultured in 30°C with addition of 0.5 mM IPTG overnight to eliminate pTargetT series because IPTG could induce the other sgRNA, which  This study targets one PAM site of pTargetT. The pCas vector could be cured by cultivating it in 37°C to remove because its temperature sensitivity.

Protein expression and purification
E. coli cells were cultured at 37°C until the OD 600 was approximately 0.6-0.8, and then 0.2 mM IPTG was used to induce target proteins expression at 20°C for 16 h. Then the cells were collected by centrifugated at 8000 rpm for 10 min, washed by 0.9% NaCl twice and resuspended in appropriate volume PB solution at pH 7.5 (20 mM Na 2 HPO 4 and NaH 2 PO 4 ). The resuspended cells were centrifuged to collect the supernatants for purification after ultrasonication. The supernatants were filtered, loaded in the pre-equilibrated (20 mM imidazole, pH 7.5) Ni-chelating column with and then washed with different concentration of imidazole (20 mM and 50 mM, pH 7.5). The bound proteins including YneI, Gox0499 and Gox1122 would be eluted by 250 mM imidazole (pH 7.5). The protein samples were then put into dialysate (20 M PB solution, 200 mM NaCl and 5% glycerol, pH 7.5) overnight, and then identified by 12% SDS-PAGE.

Enzymes activity assay
The in vitro enzyme assays of the fusion protein and tandem enzyme reaction system containing YagF and PpMdlC were performed in a 50 mM PB solution (pH 7.5) containing 20 mM D-xylonate, 10 mM MgCl 2 , 1 mM TPP, and 1 mM NAD ? . The reaction mixture was incubated at 30°C for 5 min to allow the accumulation of 3,4-DHB produced by the five-fold excess of YagF, PpMdlC or the YagF and PpMdlC fusion protein. The absorbance of NADH at 340 nm was measured after 1 mM YneI was added to the reaction. The in vitro enzyme assay of NAD ? -dependent YneI, Gox0499, and Gox1122 activity was performed in a 50 mM PB solution (pH 7.5) containing 20 mM D-xylonate, 10 mM MgCl 2 , 1 mM TPP, 1 mM NAD ? , and a five-fold excess of YagF and PpMdlC. The reaction mixture was incubated at 30°C for 5 min to allow the accumulation of 3,4-DHB, after which 1 mM YneI, 1 mM Gox0499, and 1 mM Gox1122 were added. The NADH absorbance at 340 nm was then measured.
Whole-cell catalysis for 3,4-DHBA production Whole-cell catalysis was performed in 20 mL scale, containing 20 g/L of D-xylose, 50 mM PB solution (pH 7.5), 50 g/L of E. coli cells, 2 mM TPP, 2 mM NAD ? , 10 mM Mg 2? at 30°C. The products were sampled about 1 mL in every 12 h and analyzed via HPLC.

Analytical methods
Metabolite analysis was performed by HPLC on a Transgenomic 87H3 column, using refractive-index detection (RID). The mobile phase was 0.08 N H 2 SO 4 with a flow rate of 0.38 mL/min. The temperatures of the RID detector and column were 35°C. HPLC-MS was used to distinguishing 3,4-DHBA from the supernatants of reaction (Fig. S2). 3,4-DHBA (C 4 H 8 O 4 ) was corresponded to the peak of 119.03 Da under the negative ion mode.
Finally, YneI from E. coli is used to convert 3.4-DHB to 3,4-DHBA (Wang et al. 2017). However, YneI generates 3,4-DHBA and BTO as byproducts. To obtain highly active oxidases to minimize the production of BTO, we identified two NAD ? -dependent ALDHs (Gox0499 and Gox1122 from Gluconobacter oxydans) and compared their catalytic activity with that of YneI, using in vitro enzyme assays performed in cell lysates. Gox0499 had a higher specific activity (4.26 U/mg protein) than YneI (2.67 U/mg protein) and Gox1122 (2 U/mg protein).
To evaluate the catalytic activity of the enzymes in vivo, the plasmids pE01 (carrying yagF and PpmdlC), pA01 (carrying yneI and gdh), pA02 (carrying gox0499 and gdh), and pA03 (carrying gox1122 and gdh) were constructed to assemble the entire 3,4-DHBA biosynthetic pathway in E. coli BL21(DE3). Figure 2A shows the SDS-PAGE results of target proteins expressed by the strains E0-E01-A01, E0-E01-A02, and E0-E01-A03 after cultivation and target induction. The expression levels of BsGDH, YagF, and PpMdlC were the same among the three strains, but those of the ALDHs differed; Gox1122 levels were higher than those of YneI and Gox0499.

Disruption of competing pathways to improve 3,4-DHBA titer
To improve the titer of 3,4-DHBA synthesized from Dxylose, several genes in competing pathways were deleted using CRISPR/Cas9. Xylose isomerase, encoded by xylA, catalyzes the production of Dxylulose from D-xylose. D-xylulose then enters the pentose phosphate pathway, facilitating bacterial growth and metabolism. 2-keto-3-deoxy-D-xylonate aldolases, expressed by yagE and yjhG of the Dahms pathway, always consumed the intermediate product 2-keto-3-deoxy-D-xylonate in E. coli (Valdehuesa et al. 2014). A search of the NCBI database showed that the genome of E. coli BL21(DE3) did not contain yagE or yjhG. However, ghrA and ghrB of E. coli BL21(DE3) were found to encode glyoxylate reductase, which consumes the intermediate product 2-keto-3-deoxy-D-xylonate. In addition, the ALDH encoded by adhP has a high level of reduction activity, favoring the production of the by-product BTO from 3,4-DHB (Wang et al. 2017).
E1-E01-A02 produced a 56% higher 3,4-DHBA titer than E0-E01-A02 (Fig. 3). This showed that the deletion of xylA minimized the by-product formation of D-xylulose and increased the flux of the conversion of D-xylose to D-xylonate. The 3,4-DHBA titers produced by E2-E01-A02, E3-E01-A02, and E4-E01-A02 were 5.05, 5.24, and 5.38 g/L, respectively. This result indicated that the knockout of ghrA and ghrB increased the titer of 3,4-DHBA, but this effect was negligible. The 3,4-DHBA titer produced by the xylA, ghrA, ghrB and adhP quadruple-deficient strain E5-E01-A02 was 5.69 g/L, and the BTO titer decreased from 0.56 g/L to 0.33 g/L, suggesting that adhP disruption was essential for 3,4-DHBA production. E5-E01-A02 produced the highest titer, at 5.69 g/L (0.356 g 3,4-DHBA/g D-xylose), 87% higher than that produced by E0-E01-A02. We concluded that disruption increased the titer and yield of 3,4-DHBA. Construction of a fusion protein containing PpMdlC and YagF to increase carbon flux toward the 3,4-DHBA pathway PpMdlC was the rate-limiting enzyme in this pathway because of its lower catalytic activity (Wang et al. 2017;Zhong et al. 2019), which causes the accumulation of intermediate products and consumption by competing pathways. Fusion proteins that serve as multifunctional enzymes can be constructed (Chen et al. 2016); previous studies have shown that enhancing the spatial proximity of enzymes by constructing fusion proteins can increase catalytic efficiency and the titer of products in multi-enzyme cascade reaction (Albertsen et al. 2011;Lu et al. 2006;Ryosuke et al. 2018). Here, a fusion construct containing PpMdlC and YagF with a 5 nm rigid ahelical ER/K motif was created, and the resulting recombinant plasmid pE02 was co-transformed into E. coli BL21(DE3) among with pA02 to generate the engineered strain E5-E02-A02.
The protein expression of PpMdlC and YagF in E5-E02-A02 was lower than that in the engineered strain E5-E01-A02, which harbored plasmids pE01 and pA02 (Fig. 4A). Meanwhile, a new protein of over 120 kDa was observed, indicating that the fusion protein expressing PpMdlC and YagF was constructed successfully. The catalytic activity of PpMdlC and YagF was compared to that of a tandem enzymatic reaction system containing PpMdlC and YagF by monitoring the change in NADH absorbance at 340 nm, with D-xylonate used as a substrate and Gox0499 used to catalyze the subsequent reaction in the cascade. The catalytic activity of the fusion protein containing PpMdlC and YagF was only 1.09 U/g wet cell, compared to 2.53 U/g wet cell produced by the tandem enzymatic reaction system containing PpMdlC and YagF.

Conclusion
In this study, we developed a four-step enzymatic reaction pathway in E. coli to efficiently produce 3,4-DHBA from D-xylose. In addition to YagF from E. coli and PpMdlC from P. putida, used in the 3,4-DHBA biosynthetic pathway reported previously (Valdehuesa et al. 2014), BsGDH from B. subtilis was selected to catalyze the conversion of D-xylose to Dxylonate. Additionally, the ALDH Gox0499 from G. oxydans which exhibited higher activity than YneI from E. coli (Taguchi et al. 2008) was chosen to catalyze the conversion of 3,4-DHB to 3,4-DHBA, resulting in a higher titer of 3,4-DHBA and a lower titer of the by-product BTO. E. coli strains containing recombinant plasmids were modified by deleting genes in competing pathways and introducing fusion construct expressing PpMdlC and YagF. Whole cells of the engineered strain E5-E02-A02 produced 7.71 g/ L 3,4-DHBA (0.482 g 3,4-DHBA/g D-xylose), the highest titer and yield reported to date.