Pathway engineering of Propionibacterium jensenii for improved production of propionic acid

Propionic acid (PA) is an important chemical building block widely used in the food, pharmaceutical, and chemical industries. In our previous study, a shuttle vector was developed as a useful tool for engineering Propionibacterium jensenii, and two key enzymes—glycerol dehydrogenase and malate dehydrogenase—were overexpressed to improve PA titer. Here, we aimed to improve PA production further via the pathway engineering of P. jensenii. First, the phosphoenolpyruvate carboxylase gene (ppc) from Klebsiella pneumoniae was overexpressed to access the one-step synthesis of oxaloacetate directly from phosphoenolpyruvate without pyruvate as intermediate. Next, genes encoding lactate dehydrogenase (ldh) and pyruvate oxidase (poxB) were deleted to block the synthesis of the by-products lactic acid and acetic acid, respectively. Overexpression of ppc and deleting ldh improved PA titer from 26.95 ± 1.21 g·L−1 to 33.21 ± 1.92 g·L−1 and 30.50 ± 1.63 g·L−1, whereas poxB deletion decreased it. The influence of this pathway engineering on gene transcription, enzyme expression, NADH/NAD+ ratio, and metabolite concentration was also investigated. Finally, PA production in P. jensenii with ppc overexpression as well as ldh deletion was investigated, which resulted in further increases in PA titer to 34.93 ± 2.99 g·L−1 in a fed-batch culture.

phosphoenolpyruvate carboxylase (PPC) in Propionibacterium freudenreichii 22 , and propionyl-coenzyme A (CoA):succinate CoA transferase in Propionibacterium shermanii 23 . In our previous studies, the shuttle vector pZGX04 was constructed as an effective tool for engineering P. jensenii 24 , and PA synthesis was enhanced by overexpressing glycerol dehydrogenase and malate dehydrogenase 21,24 .
As shown in Fig. 1, the PA synthesis pathway in P. jensenii pyruvate carboxylase converts pyruvate to oxaloacetate, and PA is generated through seven intermediates, including malate, succinyl-CoA, and propionyl-CoA. Furthermore, lactate and acetate are by-products of pyruvate generation. PPC, although not found in propionibacteria, converts phosphoenolpyruvate to oxaloacetate by fixing CO 2 in some microorganisms 22 . Thus, carbon flux contributes to the earlier accumulation of PA without the reaction proceeding through pyruvate, and the carbon flux to lactate and acetate is eliminated at the same time. Several studies have shown that improvements in PPC activity enhance the production of organic acids such as succinate 25 and fumarate 26 in Escherichia coli.
In this study, the gene encoding PPC (ppc) from Klebsiella pneumoniae was overexpressed in P. jensenii by pZGX04, while the genes encoding lactate dehydrogenase (ldh) and pyruvate oxidase (poxB) were deleted from P. jensenii by pUC18. The effects of these genetic manipulations on cell growth, metabolic levels, and PA production were then investigated.

Results
Expression of ppc and deletion of ldh and poxB in P. jensenii ATCC 4868. The genome of K. pneumoniae was used as the template for ppc amplification because detailed genome information for P. jensenii is lacking and the genes of K. pneumoniae have been successfully expressed 20 . The expression vector pZGX04-ppc was constructed based on pZGX04 and transformed into E. coli JM110 for demethylation to improve the transformation efficiency of recombinant plasmids by eliminating the influence of the restriction-modification system in P. jensenii 23 . The engineered strain P. jensenii (pZGX04-ppc) was obtained by transforming the demethylated pZGX04-ppc into P. jensenii ATCC 4868. pUC18 was used as suicide plasmid for gene deletion in P. jensenii because it is unable to replicate and be expressed in propionibacteria. Similarly, the LDH deletion vector pUC18ldh-cm and POXB deletion vector pUC18-poxB-cm were transformed into P. jensenii ATCC 4868 after demethylation via E. coli JM110, and the engineered strains P. jensenii-Δ ldh and P. jensenii-Δ poxB were constructed through homologous recombination.
Effects of gene expression and deletion on intracellular NADH/NAD + ratio. Because ppc overexpression and ldh deletion have been proved effective for improving PA production, further investigation of P. jensenii (pZGX04-ppc) and P. jensenii-Δ ldh was undertaken. To verify the transcription of ppc and ldh in the corresponding engineered strains, we performed reverse transcription-qPCR, which introduced the 16S ribosomal RNA-encoding gene as the reference gene for normalization. As shown in Fig. 3A, compared with those in P. jensenii ATCC 4868, the transcription level of ppc in P. jensenii (pZGX04-ppc) increased 24.54 ± 1.34-fold, whereas that of ldh in P. jensenii-Δ ldh decreased 11.29 ± 0.91-fold. These results indicating that pZGX04-ppc and pUC18-ldh-cm were successfully transformed into in P. jensenii, ppc was transcribed to messenger RNA, and ldh was knocked out.
The specific activities of PPC and LDH in P. jensenii ATCC 4868 and the corresponding engineered strains were compared (Fig. 3B). The specific activity of PPC was 9.72 ± 0.44 U·g −1 DCW in P. jensenii (pZGX04-ppc)  but was not detected in P. jensenii ATCC 4868. The specific activity of LDH in P. jensenii-Δ ldh was 4.67 times lower than that in P. jensenii ATCC 4868. These results further demonstrated that the recombinant plasmids are effective for metabolic regulation in P. jensenii. The intracellular NADH/NAD + ratios of P. jensenii ATCC 4868, P. jensenii (pZGX04-ppc), and P. jensenii-Δ ldh at the middle exponential, later exponential, and stationary phases were analyzed. As shown in Fig. 3C, the intracellular NADH/NAD + ratios of the three strains decreased with the length of fermentation. Although the intracellular NADH/NAD + ratio in P. jensenii (pZGX04-ppc) (3.8 ± 0.30) was similar to that in P. jensenii ATCC 4868 (3.9 ± 0.36) at the middle exponential phase, it dropped rapidly during the later exponential phase (2.5 ± 0.31) and stationary phase (1.3 ± 0.22) of the culture and was significantly lower than those in P. jensenii ATCC 4868 (3.0 ± 0.28 and 1.6 ± 0.19, respectively). However, the intracellular NADH/NAD + ratios in P. jensenii-Δ ldh at all three phases were always higher than those in P. jensenii ATCC 4868.

Changes in intracellular metabolites.
To investigate the influence of expressing ppc and knocking out ldh on the intracellular metabolites of P. jensenii, we detected the intercellular metabolites of P. jensenii ATCC 4868, P. jensenii (pZGX04-ppc), and P. jensenii-Δ ldh at the exponential and stationary phases of fed-batch cultures. A total of nine key metabolites were selected for analysis: glycerate-3-phosphate, phosphoenolpyruvate, pyruvate, lactate, acetyl-CoA, oxaloacetate, citrate, malate, and propionyl-CoA. The content of these metabolites in the engineered strains relative to those of P. jensenii ATCC 4868 are shown in Fig. 4. PPC expression decreased phosphoenolpyruvate and increased oxaloacetate. As a result, the intercellular concentrations of glycerate-3-phosphate, pyruvate, lactate, and acetyl-CoA decreased. In particular, the level of pyruvate decreased significantly with the increase in PA synthesis, which was less than one-fifth that of the wild type at the stationary phase. Relatively, intercellular levels of citrate, malate, and propionyl-CoA increased, among which propionyl-CoA content increased 1.2-to 1.4-fold.
The deletion of ldh blocked the accumulation of lactate directly. At the exponential phase, the intercellular concentration of propionyl-CoA in P. jensenii-Δ ldh was substantially higher than that in the wild type-an increase of 89.0%. Simultaneously, the content of phosphoenolpyruvate, pyruvate, acetyl-CoA, oxaloacetate, and malate increased. Acetyl-CoA level increased by 75.1% in P. jensenii-Δ ldh at the stationary phase; however, the levels of other metabolites changed only slightly.

Discussion
P. jensenii is a species of dairy propionibacteria that has been used to produce PA via the dicarboxylic acid pathway with acetate and lactate as by-products 21,27 . Low yield, low productivity, and poor purity caused by feedback inhibition and high by-product accumulation have seriously restricted the industrialization of microbial PA production by propionibacteria 23 . Various strategies have been proposed to improve the fermentation process and enhance PA synthesis 3 . However, the majority of these efforts involve strain evolution 28,29 and fermentation optimization 16,19 . Metabolic engineering of propionibacteria has progressed slowly due to the lack of genome information, the shortage of available gene manipulation tools, the difficulties of transforming gram-positive bacteria, and the high GC content of genomes 5 . Recently, the genomes of several propionibacteria were fully sequenced, and in turn, the synthetic pathways of PA were elucidated [30][31][32] . The availability of this information provided the rationale and basis for the metabolic engineering of propionibacteria for improved PA production.
pZGX04, a Propionibacterium-E. coli shuttle plasmid, is the only reported gene expression vector available for P. jensenii 24 . A fermentation optimization strategy was developed for engineered P. jensenii in our recent study and integrates the use of a fed-batch culture with pH-shift control 18 . Using this strategy, we overexpressed ppc, knocked out ldh and poxB in P. jensenii ATCC 4868 by constructing P. jensenii (pZGX04-ppc), P. jensenii-Δ ldh and P. jensenii-Δ poxB, respectively, and further enhanced PA production. Pyruvate is a key metabolic node in the synthetic pathway of PA in propionibacteria 33 . A total of three branch pathways are derived from pyruvate: acetyl-CoA, which enters the tricarboxylic acid cycle or generates acetate; lactate formed by the catalysis of LDH; and oxaloacetate, which ultimately leads to PA (see Fig. 1). Methylmalonyl-CoA carboxyltransferase and pyruvate carboxylase convert pyruvate to oxaloacetate in certain propionibacteria through the transfer of carbon from methylmalonyl-CoA to pyruvate and the fixing of CO 2 , respectively. Furthermore, PPC converts phosphoenolpyruvate to oxaloacetate via CO 2 fixation in some microorganisms, although not in propionibacteria 22 . Therefore, in the engineered strains, the carbon flux flowed to oxaloacetate from phosphoenolpyruvate earlier, bypassing pyruvate, and the metabolic flux to acetyl-CoA and lactate was reduced as pyruvate was generated in smaller amounts (see Fig. 4A,B). As a result, the flux to oxaloacetate increased and the metabolism of the tricarboxylic acid cycle and PA synthesis was enhanced, as suggested by the improved levels of citrate, malate, and propionyl-CoA. In addition, the flow of P. jensenii (pZGX04-ppc) was more focused on the PA synthetic pathway, as demonstrated by the higher increases in malate and propionyl-CoA. Therefore, PA production was enhanced by overexpressing ppc in P. jensenii and decreasing the accumulation of by-products (acetate and lactate; see Fig. 2B).
NADH and NAD + play crucial roles in the intracellular chemical reactions and metabolism of microbial cells 34 . NADH is consumed along the metabolic pathway from pyruvate to PA, which changes the NADH/ NAD + ratio 19 . That is, the NADH/NAD + ratio is an indicator of the synthetic efficiency of P. jensenii for PA. Furthermore, it has been demonstrated in E. coli that a reducing intracellular environment is necessary for key cellular functions 35 . Compared with P. jensenii ATCC 4868, P. jensenii (pZGX04-ppc) had lower NADH/NAD + ratios at the later exponential phase and stationary phase, which indicated that the introduction of PPC accelerated PA synthesis in the late fermentation stage and consumed more NADH. These results coincided well with the higher PA production of P. jensenii (pZGX04-ppc).
Because lactate is one of the main by-products of glycerol metabolism in the fed-batch culture of P. jensenii, ldh was knocked out in P. jensenii ATCC 4868. Compared with that of P. jensenii ATCC 4868, the NADH/NAD + ratio of P. jensenii-Δ ldh was higher because NADH is consumed during lactate synthesis from pyruvate. The activities of many intracellular enzymes were influenced by the variation in NADH/NAD + ratios, and thus the distribution of intracellular metabolic flow was influenced 36 . Therefore, the higher reducibility maintained in P. jensenii-Δ ldh has a negative effect on microbial activity, which results in slow cell growth and an extended fermentation period. Simultaneously, the lack of a lactate pathway leads to an increase in metabolic flow from pyruvate to PA and acetate, which was shown by the increased concentrations of acetyl-CoA, oxaloacetate, malate, and propionyl-CoA in P. jensenii-Δ ldh.
The growth of P. jensenii was strongly inhibited as poxB deleted. It is supposed that low accumulation of acetate affects the metabolism of acetyl phosphate, which has been proved to be a regulator of signal transduction in the two-component regulatory systems in some bacteria 37 . Presecan-Siedel et al. also found that Bacillus subtilis did not grow when acetate kinase was inactivated 38 . Therefore, the unnormal metabolism of acetyl phosphate may be responsible for the growth defect of P. jensenii-Δ poxB.
In this study, we enhanced PA production in P. jensenii by constructing the engineered strains P. jensenii (pZGX04-ppc) and P. jensenii-Δ ldh. The overexpression of ppc decreased the accumulation of by-products and accelerated PA biosynthesis by adjusting the distribution of metabolic flow, and the production and productivity of PA in P. jensenii (pZGX04-ppc) increased to 33.21 ± 1.92 g·L −1 and 7.59 ± 0.46 g·L −1 ·g −1 DCW, respectively. The deletion of ldh altered the redox state of the cells and reduced the cell growth rate, and PA production and productivity in P. jensenii-Δ ldh increased to 30.50 ± 1.63 g·L −1 and 9.08 ± 0.37 g·L −1 ·g −1 DCW, respectively. Furthermore, by expressing ppc and deleting ldh simultaneously, we increased the production of PA to 34.93 ± 2.99 g·L −1 , an increase of 29.61% over production by wild-type P. jensenii ATCC 4868. Since overexpression of glycerol dehydrogenase, malate dehydrogenase, fumarate hydratase, and PPC, and deletion of ldh have been proved to be effective on improving PA production of P. jensenii 21 , further studies should be conducted to investigate the combinatorial effect of them on PA production in the future work.

Materials and Methods
Strains, plasmids, and culture conditions. P. jensenii ATCC 4868 and K. pneumoniae subsp. pneumoniae ATCC 12657 were purchased from the American Type Culture Collection (ATCC) and used as templates to amplify genes or as hosts for gene expression and deletion. E. coli JM110 (Stratagene, La Jolla, CA) was used for plasmid cloning and maintenance. pUC18 (TaKaRa, Dalian, China) was used as suicide plasmid to knock out genes in P. jensenii, and the shuttle vector pZGX04 was constructed in our previous study 24 . pRS303 was maintained in our laboratory to amplify the hygromycin B resistance gene (hygB). P. jensenii was cultured anaerobically at 32 °C in sodium lactate broth medium (10 g·L −1 yeast extract, 10 g·L −1 tryptic soy broth, and 10 g·L −1 sodium lactate), and E. coli and K. pneumoniae were grown at 37 °C in Luria-Bertani medium (10 g·L −1 NaCl, 10 g·L −1 peptone, and 5 g·L −1 yeast extract) with a shaking speed of 220 rpm. When necessary, 10 μ g·mL −1 chloramphenicol and 50 μ g·mL −1 hygromycin B were added to the P. jensenii cultures, and 100 μ g·mL −1 ampicillin was added to the E. coli cultures.
Construction of plasmids and recombinant strains. The genomic DNA of K. pneumoniae and P. jensenii were isolated with an E.Z.N.A. TM Bacterial DNA Kit (Omega Bio-Tek, Doraville, GA, USA). ppc was amplified via polymerase chain reaction (PCR) from the genome of K. pneumoniae, and the PCR products were purified with a TaKaRa MiniBEST Agarose Gel DNA Extraction Kit (TaKaRa). The expression vector pZGX04-ppc was constructed by replacing orf2 in pZGX04 with ppc. The front and back sequences flanking ldh and poxB were amplified from the genome of P. jensenii, and the chloramphenicol resistance gene (cm) and hygB were amplified from pZGX04 and pRS303, respectively. The ldh/poxB deletion frame was constructed by fusing the three fragments via PCR and inserting them into pUC18 to obtain the deletion vectors pUC18-ldh-cm and pUC18-poxB-hygB. The primers used for PCR are listed in Table 2. pZGX04-ppc, pUC18-ldh-cm, and pUC18-poxB-hygB were first transformed into E. coli JM110 for demethylation and then transformed into P. jensenii ATCC 4868 to construct P. jensenii (pZGX04-ppc), P. jensenii-Δ ldh, and P. jensenii-Δ poxB 24 . The construction of P. jensenii-Δ ldh (pZGX04-ppc) was carried out by transforming pZGX04-ppc into P. jensenii-Δ ldh with the same method. The deletion of ldh/poxB and expression of ppc were verified through colony PCR and sequencing. The expression of ppc was controlled by the original promoter of orf2 in pZGX04.
Ten-milliliter aliquots were removed from the fermenter every 12 h to measure the cell density and concentrations of PA, lactic acid, acetic acid, and glycerol according to previously published methods 16 . The influence of sampling volume was negligible, and the change in volume caused by Ca(OH) 2 solution addition was removed for calculations.

Primer
Nucleotide sequence (5′-3′)  Reverse transcription-quantitative PCR (RT-qPCR). Cells were harvested in the exponential phase when the optical density at 600 nm reached 0.6-0.7 in sodium lactate broth medium. Total RNA was isolated by using an RNAprep pure Cell/Bacteria kit (TIANGEN, Beijing, China). Complementary DNA was obtained by using a PrimeScript II first-strand complementary DNA synthesis kit (TaKaRa), and qPCR was conducted with SYBR Premix Ex Taq (TaKaRa) on a LightCycler 2.0 system (Roche, Basel, Switzerland). The fold changes were determined by the 2 −ΔΔCt method normalized to the 16S rRNA gene.
jensenii-Δ ldh were collected at the exponential phase (48 h) and centrifuged at 7,000 x g for 10 min. The PPC activity of the cells was measured according to the method of Ammar et al. 22 , and the LDH activity of the cells was measured according to the method of Zheng et al. 39 .
Determination of intracellular metabolites. Cells were sampled and quenched at the middle exponential phase, later exponential phase, and stationary phases for intracellular metabolite detection. The extraction and detection of metabolites were performed by using the procedures described in our previous study 19 .
Statistical analysis. Unless otherwise noted, at least three independent experiments were performed for each conclusion, and the average values with standard errors are reported. Student's t-test analysis was performed with SPSS 12.0 software (SPSS Inc., Chicago, USA) to investigate statistical differences. P values of less than 0.05 were considered statistically significant.