High-yield production of 1,3-propanediol from glycerol by metabolically engineered Klebsiella pneumoniae

Background Glycerol is a major byproduct of the biodiesel industry and can be converted to 1,3-propanediol (1,3-PDO) by microorganisms through a two-step enzymatic reaction. The production of 1,3-PDO from glycerol using microorganisms is accompanied by formation of unwanted byproducts, including lactate and 2,3-butanediol, resulting in a low-conversion yield. Results Klebsiella pneumoniae was metabolically engineered to produce high-molar yield of 1,3-PDO from glycerol. First, the pathway genes for byproduct formation were deleted in K. pneumoniae. Then, glycerol assimilation pathways were eliminated and mannitol was co-fed to the medium. Finally, transcriptional regulation of the dha operon were genetically modified for enhancing 1,3-propanediol production. The batch fermentation of the engineered strain with co-feeding of a small amount of mannitol yielded 0.76 mol 1,3-PDO from 1 mol glycerol. Conclusions Klebsiella pneumoniae is useful microorganism for producing 1,3-PDO from glycerol. Implemented engineering in this study successfully improved 1,3-PDO production yield, which is significantly higher than those reported in previous studies. Electronic supplementary material The online version of this article (10.1186/s13068-018-1100-5) contains supplementary material, which is available to authorized users.


Strains, plasmids, and primers
The strains, plasmids, and primers used in this study are listed in Table 2 and Additional file 1: Table S1. All K. pneumoniae strains were derived from the wild-type strain KCTC 2242 (Korean Collection for Type Culture) [17]. KMK-01, KMK-02, and KMK-05 strains have been reported in our previous study [12]. To construct additional deletion mutants, λ red recombination was used with pRedET-transformed strains [18]. The plasmid pKD4 was used to synthesize an antibiotic-resistant gene flanked by FLP recognition target sites, while 707-FLP was used to eliminate the resistance cassette. The KMK-11, KMK-12, KMK-21, KMK-22, KMK-23, and KMK-46 strains ( Table 2) were constructed using this method. Deletions of target genes were confirmed by agarose gel electrophoresis (Additional file 2: Figure S1) and sequencing their PCR products generated with genomic DNA of the strains and confirmation primers, indicated by 'con' in their name in Additional file 1: Table S1.  [14]. KMK-23M, KMK-46, and KMY on the right-hand side represent the engineered strains (Table 2) (See figure on next page.) Decreased mtlA (Gene ID: 12547935 [Genbank]) gene expression in KMK-23 was achieved by mutation of its 5′-untranslated region (UTR). The sequence of 5′-UTR of mtlA has been designed by RBS calculator [19,20] to have 62% mtlA expression compared to the parental strain. The sequence 5′-TAG ACA GAG TCT AAC AGA CCA TCG AGG AAC GTATG-3′, which consists of bases − 32 to − 1 from the mtlA translation start codon, was chosen and introduced to the 5′-UTR region using genome editing with CRISPR/Cas9 [21,22]. The plasmids, pZA-Cas9 and pZS-CRISPR, for genome editing were used as reported in a previous study [23]. The DNA fragments, mtlA CRISPR F and R (Additional file 1: Table S1), were synthesized, annealed, and inserted into pZS-CRISPR at BsaI to generate crRNA. The resulting plasmid was designated pZS-CRISPR mtlA. The DNA fragments, and mtlA rescue F and R (Additional file 1: Table S1) were synthesized, annealed, and used as rescue DNA. The plasmid pZS-CRISPR mtlA and rescue DNA were transformed into a pZA-Cas9 containing KMK-23 strain. The genome-edited strain, KMK-23M, was confirmed by DNA sequencing.
The dhaL gene was cloned into the pZS21MCS plasmid using the Gibson assembly method [24,25]. The plasmid pZS21::MCS was cut using restriction enzymes, HindIII and MluI, and assembled with amplified dhaL gene fragment generated with primers dhaL F and R (Additional file 1: Table S1) by NEBuilder Assembly Tool (http:// nebui lder.neb.com). The resulting plasmid, pZS21dhaL, was transformed into KMK-46 to construct KMY. The engineered strains and plasmids were confirmed by DNA sequencing.

Analytical methods
The cell density was monitored by UV visibility spectrophotometer (Shimazu UV mini 1240; Shimadzu, Tokyo, Japan) at 600 nm (OD 600 ). Culture broth (1 mL) was transferred to microcentrifuge tubes and centrifuged at 13,000 rpm for 10 min at 4 °C. Supernatants were transferred to new microcentrifuge tubes and used for analysis of metabolites. To measure concentrations of 1,3-PDO, 2,3-BDO, acetate, lactate, succinate, glycerol, mannitol, glucose, xylose, and galactose, high-performance liquid chromatography (Younglin Instrument ACME-9000, Seoul, South Korea) with a Sugar SH1011 column (Shodex, Tokyo, Japan) maintained at 70 °C, an RI detector maintained at 45 °C, and 5 mM H 2 SO 4 as the mobile phase with a flow rate of 0.6 mL min −1 was used. NADH and NAD + assay was conducted by NAD/NADH-GloTM Assay G9071 (Promega, Madison, WI, USA) according to the manufacturer's protocol. Each strain for the assay was cultured to the mid-log phase (OD 600 = 1) with 30 mL working volume in 100-mL Erlenmeyer flask and harvested for the assay.

Real-time RT-PCR
The strains were cultured in flask and harvested in early exponential phase (OD ~ 1

Pathway engineering to reduce byproduct formation
For removing pathogenicity of the strain, we deleted wabG gene from K. pneumoniae KCTC 2242, which is responsible for making a lipopolysaccharide [26]. The resulting strain, KMK-01 is a parental strain in this study. K. pneumoniae KMK-01 produces 1,3-PDO as a major product from glycerol. At the same time, it also produces a significant number of byproducts, including 2,3-BDO, lactate, acetate, and succinate ( Table 3). Deletions of ldhA and pflB have been reported significantly reducing the amounts of these byproducts when glucose is fed as a carbon source [12]. When KMK-02, the KMK-01 ΔldhA mutant, was cultured in glycerol, lactate production was reduced from 0.06 to 0 mol mol −1 . KMK-05, the KMK-02 ΔpflB mutant, produced no acetate, similar to previous results observed for glucose [12]. 1,3-PDO production increased from 17.94 to 20.14 g L −1 and 20.11 g L −1 in KMK-02 and KMK-05, respectively, compared to the parental strain, while the molar yield from glycerol was almost the same. Meanwhile, 2,3-BDO production increased in these mutants. Therefore, budA encoding 2,3-butanediol dehydrogenase, was deleted from the KMK-05 strain and the resulting strain was designated KMK-12. KMK-12 produced less 2,3-BDO than the parental strain, 0.04 compared to 0.23 mol mol −1 , respectively. Despite the reduction in byproduct formation by KMK-12, the molar yield of 1,3-PDO from glycerol, 0.47 mol mol −1 glycerol, did not notably change (Table 3). Meanwhile, deletion of budA caused significant decreases in glycerol consumption and biomass production, resulting in a reduction in 1,3-PDO titer, as previously reported [13].

Elimination of glycerol assimilation pathways to improve 1,3-PDO yield
To improve the molar yield of 1,3-PDO from glycerol, glycerol assimilation pathways were eliminated on K. pneumoniae. Two glycerol assimilation pathways, which start from glycerol and lead into glycolysis, exist in K. pneumoniae (Fig. 1a). Glycerol kinase (GlpK, NCBI-Protein ID: AEK00501) is responsible for the conversion of glycerol to glycerol 3-phosphate, while glycerol dehydrogenase (DhaD, NCBI-Protein ID: AEJ99991) converts glycerol to dihydroxyacetone with NADH production [27][28][29]. GlpK is active under aerobic conditions, while DhaD is active under anaerobic conditions [29]. Strains KMK-21 and 22 were constructed by deleting dhaD and glpK, respectively, from KMK-12. Strain KMK-23 was generated by deleting both dhaD and glpK from KMK-12. When these strains were cultured for 24 h in flasks containing glycerol, there were decreases in 1,3-PDO productions and molar yields in KMK-12 and KMK-23, while a marginal increase in 1,3-PDO production was in KMK-22 (Additional file 3: Table S2). It is because DhaD-DhaKLM pathway provides 1 mol of NADH for 1,3-PDO production, while GlpK-GlpD pathway does not (Fig. 1a). Therefore, dhaD deletion resulted in decreases in 1,3-PDO production, while glpK deletion did not. To prove this, NADH/NAD + ratios were measured in KMK-12, KMK-21, KMK-22, and KMK-23 strains. The results also supported that the strains with dhaD deletion, KMK-21 and KMK-23 produced less NADH, resulting in low NADH/NAD + ratios ( Fig. 2a and b). To enhance 1,3-PDO yield, glycerol flux towards assimilation should be restricted further. Therefore, glucose was supplied as co-substrate to replace the glycolytic flux generated from glycerol assimilation. When glycerol and glucose were supplied together, the molar yield of 1,3-PDO from glycerol in KMK-21 and KMK-23 significantly increased (Table 4). Since glucose was not converted to 1,3-PDO (Additional file 4: Table S3), the molar yield of 1,3-PDO was calculated based on glycerol consumed. Deletion of dhaD (strain KMK-21) improved molar yield from 0.56 to 0.77 mol mol −1 compared to the parental strain (KMK-12). When dhaD was deleted, glycerol consumption was restricted because DhaD is active under microaerobic conditions and, therefore, this mutant assimilated very little glycerol. Deletion of glpK (KMK-22) improved 1,3-PDO production from 4.87 to 6.09 g L −1 compared to KMK-12, with a marginal improvement in yield. This may be because deletion of glpK enhances DhaD activity by more than twofold according to Ashok et al. [30]. However, it also enhances the flux into glycerol assimilation pathways, draining glycerol for energy and cell mass production. Therefore, no gain in 1,3-PDO molar yield from glycerol was observed for KMK-22. When both the glpK and dhaD genes were deleted, the molar yield of 1,3-PDO from glycerol improved from 0.56 to 0.84 mol mol −1 glycerol, with an improvement in titer from 4.87 to 5.27 g L −1 . This implies that elimination of glycerol assimilation enhanced carbon flux from glycerol towards 1,3-PDO production, while energy and cell mass production were supported by flux from carbon to glucose. Compared to the previously reported molar yield of 1,3-PDO from glycerol, 0.60 mol mol −1 glycerol, [15,16], the molar yield obtained by KMK-23 was much higher, 0.84 mol mol −1 glycerol. Despite that, we observed increased acetate production from engineered strains. Acetate production could be reduced by keeping glycerol or glucose concentration low during fed-batch fermentation [8,31].
However, when mannitol was used as a co-substrate, its consumption over 24 h was much higher than consumption of other co-substrates (Fig. 3). To reduce fermentation cost, mannitol consumption should be restricted. Here, we attempted to solve it by decreasing  the expression level of mannitol-specific transporter, MtlA (NCBI-Protein ID: AEK00431) by genome editing [32][33][34]. The 5′-UTR sequence of the mtlA was determined by RBS calculator and then was edited by CRISPR/Cas9 method as described in "Methods". The resulting strain was confirmed by sequencing and designated as KMK-23M. The expression level of mtlA in KMK-23M was decreased 25% compared to KMK-23 (Additional file 5: Figure S2A). Mannitol consumption of KMK-23M until 24 h was reduced from 9.68 to 6.28 g L −1 compared to that of KMK-23, while cell mass and 1,3-PDO production both increased substantially by 31 and 34%, respectively (Fig. 4). No reduction was observed in molar yield of 1,3-PDO from glycerol. This suggests that for KMK-23, the glycolytic flux supplied by mannitol exceeded the optimal point. Therefore, reducing glycolytic flux in KMK-23M benefited both cell growth and the conversion of glycerol to 1,3-PDO. , which belong to the dha operon [14]. According to a study by Bächler et al. [35], the promoter of the dhaKLM operon in Escherichia coli, PdhaK, is negatively regulated by DhaK and DhaM, and positively regulated by DhaL. Dephosphorylated DhaL can bind the sensing domain of DhaR (NCBI-Protein ID: AEJ99990) and activate DhaR. DhaR positively regulates dhaT expression, but negatively regulates dhaB [36]. Negative regulation of dhaB could be beneficial to 1,3-PDO production because the product of DhaB, 3-hydroxypropionaldehyde, is a toxic intermediate [37]. DhaR is repressed by DhaK, while activated by dephosphorylated DhaL that is inactivated by DhaM-mediated phosphorylation [35,38]. In other words, dhaK and dhaM negatively regulate and dhaL positively regulates dhaT expression via DhaR. We deleted dhaKLM on KMK-23M strain genome and named the resulting strain KMK-46. Then, the dhaL gene-encoding activator was overexpressed and the resulting strain was designated KMY, expecting DhaR activated by DhaL transcriptionally upregulates dhaT expression (Fig. 1b). To demonstrate the expression level changes of dhaR, dhaL and dhaT quantitative RT-PCR was conducted in KMK-23M, KMK-46, and KMY strains. The expression levels of dhaR were not changed significantly in those engineered strain (Fig. 5a). On the other hand, dhaL expression levels confirmed that dhaKLM deletion and dhaL overexpression were properly conducted in KMK-46 and KMY strains (Fig. 5b). As expected, dhaT expression in KMY was increased threefold than KMK-23M and KMK-46 (Fig. 5c). Interestingly, KMK-46, dhaKLM deleted strain, did not produce 1,3-PDO at all, while dhaL overexpression in that strain increased 1,3-PDO production by 26.4% compared to KMK-23M (Fig. 6a-d). The molar yield of KMY was also 20% higher than for KMK-23M (Fig. 6e). This is contrast to the result of overexpression of dhaR (Additional file 5: Figure S2B) in the KMK-46 strain, resulted in no 1,3-PDO production (data not shown). In addition, When three strains, KMK-01 (wild type, black bar), KMK-23 (dashed bar), and KIK-23M (genome-edited strain, gray bar), were cultured in the medium containing 40 g L −1 glycerol and 20 g L −1 mannitol, mannitol and glycerol consumption, 1,3-PDO production, microbial growth (OD 600 ), and 1,3-PDO production per unit mannitol consumption for 24 h were compared when dhaT was overexpressed in KMK-23M and KMK-46 strains, 1,3-PDO production was much less than that of KMY (data not shown). These results suggest that the regulation of dha operon is not simple and has not been fully understood yet. In this study, the developed KMY strain with deletion of dhaKLM and overexpression of dhaL partially proved their regulatory functions on dhaT expression.

Engineering of transcriptional regulation of the dha operon
Batch fermentation was performed with the resulting strain, KMY. A total of 20.59 g L −1 1,3-PDO was produced in 24 h with a molar yield of 0.76 mol mol −1 glycerol (Fig. 6f ). The 1,3-PDO yield from the carbon sources in total, including mannitol, was 0.54 g g −1 . The molar yield in this study was superior to that reported in other published studies; these studies reported molar yields of around 0.6 mol mol −1 glycerol [15,16]. The increase in the 1,3-PDO titer from batch fermentation compared to culturing in flasks may be due to the pH control during the process. During the entire fermentation, dissolved oxygen level was maintained at near zero percent, because the microaerobic condition with low aeration has been known to be beneficial for 1,3-PDO production [5]. The developed strain produced significantly lower amounts of byproducts with a much higher molar yield of 1,3-PDO from glycerol. This could be extremely helpful in reducing the cost of product separation [11]. The high-molar yield of 1,3-PDO was enabled by eliminating genes in the glycerol assimilation pathway and feeding bacteria a co-substrate for cell mass and maintenance energy. Regulators influencing expression of dhaT were also helpful in increasing 1,3-PDO production and yield compared to the parental strain. With continuous metabolic engineering, a cost-effective biochemical process to produce 1,3-PDO using K. pneumoniae would be possible near future.

Conclusions
In this study, we minimized the production of byproducts during 1,3-PDO synthesis by deleting the pathway genes from the K. pneumoniae genome. Next, the glycerol assimilation pathway was eliminated and mannitol was fed as a co-substrate to improve glycerol flux towards 1,3-PDO production. The 5′-UTR of mtlA was edited to reduce mannitol consumption. Finally, 1,3-PDO production was enhanced by modifying dha regulation through deletion of dhaKLM and overexpression of dhaL. In

Additional file
Additional file 1. Table S1: Oligonucleotides used in this study.
Additional file 3. Table S2: Fermentation data of KMK-12 and its mutants after 24 hrs of flask cultivation with 40 g L −1 glycerol as a sole carbon source.
Additional file 4. Table S3: Fermenation data of KMK-12 and its mutants after 24 hrs of flask cultivation with 40 g L −1 glucose as a sole carbon source.
Additional file 5. Figure S2: Relatively gene expression levels of A mtlA in KMK-23 and KMK-23M when glucose or mannitol was used as a cosubstrate and B dhaR when the gene was overexpressed in KMK46 strain. The gene expression levels were detected by quantitative RT-PCR.