Overexpression of thermostable meso-diaminopimelate dehydrogenase to redirect diaminopimelate pathway for increasing L-lysine production in Escherichia coli

Dehydrogenase pathway, one of diaminopimelate pathway, is important to the biosynthesis of L-lysine and peptidoglycan via one single reaction catalyzed by meso-diaminopimelate dehydrogenase (DapDH). In this study, the thermostable DapDH was introduced into diaminopimelate pathway that increased the final titer (from 71.8 to 119.5 g/L), carbon yield (from 35.3% to 49.1%) and productivity (from 1.80 to 2.99 g/(L∙h)) of L-lysine by LATR12-2∆rpiB::ddhSt in fed-batch fermentation. To do this, the kinetic properties and the effects of different DapDHs on L-lysine production were investigated, and the results indicated that overexpression of StDapDH in LATR12-2 was beneficial to construct an L-lysine producer with good productive performance because it exhibited the best of kinetic characteristics and optimal temperature as well as thermostability in reductive amination. Furthermore, ammonium availability was optimized, and found that 20 g/L of (NH4)2SO4 was the optimal ammonium concentration for improving the efficiency of L-lysine production by LATR12-2∆rpiB::ddhSt. Metabolomics analysis showed that introducing the StDapDH significantly enhanced carbon flux into pentose phosphate pathway and L-lysine biosynthetic pathway, thus increasing the levels of NADPH and precursors for L-lysine biosynthesis. This is the first report of a rational modification of diaminopimelate pathway that improves the efficiency of L-lysine production through overexpression of thermostable DapDH in E. coli.

subspecies in L-lysine producer E. coli was beneficial to increase the L-lysine production, they neglected the differences in the optimal cultivated conditions between E. coli and C. glutamicum. For example, the temperature optimum for E. coli is 37 °C, whereas it is 30 °C for C. glutamicum. Note that the activity and stability of the intracellular enzymes in the host cell is changed with different conditions 9,24,25 . In this paper, we introduced a DapDH from different bacteria with different temperature optimum in E. coli to investigate its effect on L-lysine production; results indicated that the DapDH from thermophilic bacterium S. thermophilum (StDapDH) has the positive effects in improving the performance of L-lysine fermentation process by E. coli for the first time. In addition, the introducing mode and ammonium availability were also investigated, indicating that the co-existence of two pathways and sufficient ammonium availability are good for increasing the final titer of L-lysine with a high carbon yield and productivity in E. coli. These results reported here can serve as a general concept and guidance for breeding high-yielding strains and producing L-lysine in industry.

Results and Discussions
Overexpression, purification and function identification of His-tagged DapDH from different bacteria. The DapDH-coding gene ddh from different strain shows the huge difference of nucleotide and amino acids sequence identity among these strains (Fig. S1) 9,24,25 . According to previous reports, the DapDH from different strains exhibits different temperature optimum and substrate affinity 9,15 . In order to screen out the best DapDH for L-lysine production in E. coli, the six DapDHs from six representative bacteria [including C. glutamicum ATCC13032 (Cg2900; CgDapDH), Bacillus sphaericus IFO3525 (BAB07799; BsDapDH), C. therimocellum ATCC27405 (Cthe_0922; CtDapDH), B. fragilis YCH46 (Bf3690; BfDapDH), S. thermophilum IAM14863 (Sth1425; StDapDH) and U. thermosphaericus A1 (AB636161; UtDapDH)] was overexpressed in E. coli BL21 (DE3) using pET28a, and then used for investigating their functions and kinetic properties. According to the analysis of SDS-PAGE, the molecular mass of DapDH was about 40 kDa, which was nearly equal to the calculated molecular weights (data not shown). In addition, all of these DapDH orthologs are able to complement the meso-DAP auxotrophy of the E. coli ∆dapD/∆dapE (Fig. S2), indicating that these DapDHs are the functional forms of DapDH.
The purified enzymes were used for investigating the functions and kinetic properties. All of these DapDHs showed both the activities of oxidative deamination and reductive amination. However, all of these DapDHs catalyzed the reductive amination with higher efficiency than they catalyzed the oxidative deamination except the BfDapDH (Tables 1 and S1). It should be noted that different DapDHs showed a huge difference in the oxidative deamination and reductive amination (Table S1). Although DAP pathway is necessary for cell survival because it involves the peptidoglycan biosynthesis 26  possess the dehydrogenase and aminotransferase pathways 9 . Moreover, the kinetic analysis of these DapDHs again showed that different orthologs had different substrate affinity (K m ), thereby affecting the catalytic efficiency of enzyme ( Table 1). The K m of BfDapDH for THDPA (K m = 0.57 ± 0.14 mmol/L) was nearly five-fold higher than that of CtDapDH (K m = 0.11 ± 0.03 mmol/L). The kinetic constants were also determined for other DapDHs (including CgDapDH, BsDapDH, StDapDH and UtDapDH), indicating that they shared the similar values for THDPA and meso-DAP (within the ranges of BfDapDH and CtDapDH), but the kinetic constants towards + NH 4 of these orthologs were different ( Table 1). As can be seen from Table 1, BfDapDH exhibited a lowest K m for + NH 4 , followed by the BsDapDH, whereas the V max , K cat and K cat /K m of BfDapDH were not higher than the others. Although the CtDapDH exhibited a highest K m for + NH 4 , the V max , K cat and K cat /K m were ranked first (Table 1). It is noteworthy that different variants of DAP pathways exhibits the alterable roles on peptidoglycan and L-lysine biosynthesis in different strains under different cultural conditions 28 . All of these factors have contributed to the different activities of DapDH in different strains.
The effect of temperature on the reductive amination of THDPA was determined by assessing the enzyme activity at various incubation temperatures for 1 h. Consistent with the previous results 9, 24,25 , the DapDH from thermophiles shows the higher temperature optimum than that from enteric and soil species (Fig. 2a). For example, the purified CtDapDH exhibited a temperature optimum at 65 °C for reductive amination, whereas the temperature optimum was 33 °C for CgDapDH. For StDapDH and UtDapDH from thermophiles, the activity was stable over the temperature range of 40 °C to 60 °C and maintained at the high level (Fig. 2a). In addition, the effect of incubation time at 40 °C on the activity of the different DapDHs was also investigated. As shown in Fig. 2b, the activity of all DapDHs was decreased with increase of the incubation time, especially for CgDapDH and BsDapDH. Although the CtDapDH remained stable when incubated at 40 °C for 24 h, it exhibited the relatively low activity as compared with StDapDH and UtDapDH (Fig. 2b).

Inhibition of different DapDHs on reductive amination by nucleotide-cofactor, substrate and
product. DapDH is a bifunctional enzyme catalyzing the NADPH-dependent reductive amination to form meso-DAP with THDPA and + NH 4 as substrates and the NADP + -dependent oxidative deamination to form THDPA with meso-DAP as substrate 29 . In order to investigate whether nucleotide-cofactor, substrate and product involved in DapDH-catalyzed reaction regulate the activity of DapDH, the effects of nucleotide-cofactor, substrate and product on different DapDHs were studied on the reductive amination. For all of these DapDHs, the nucleotide-cofactor NADP + showed the competitive inhibition with NADPH in the presence of a high as well as constant THDPA and + NH 4 concentration, whereas it showed the noncompetitive inhibition with THDPA or + NH 4 in the presence of a high as well as constant NADPH and + NH 4 or THDPA concentration (Fig. S3). This is because DapDH is a bifunctional enzyme catalyzing the NADPH-dependent reductive amination and the NADP + -dependent oxidative deamination 29 , thus both NADP + and NADPH can be combined with the free form of DapDH 18 . However, the strength of inhibition on different DapDHs presented certain discrepancies (Fig. 3a). For example, the activity of CtDapDH and BsDapDH was dramatically decreased with the increase of the concentration of NADP + (K i = 7.3 ± 0.6 μmol/L and K i = 5.8 ± 0.3 μmol/L, respectively), whereas the CgDapDH showed the minimal changes (K i = 15.2 ± 1.3 μmol/L). The other nucleotide-cofactor NADPH was also tested for its regulating properties. As can be seen from  Fig. 3c). However, the activity of these DapDHs were firstly increased and then decreased with increasing (NH 4 ) 2 SO 4 (Fig. 3d). Especially for StDapDH, the activity was dramatically decreased and get closer to 20% of initial at 1 mol/L of (NH 4 ) 2 SO 4 when the concentration of (NH 4 ) 2 SO 4 was above 0.5 mol/L. It is well known that (NH 4 ) 2 SO 4 is a physiologically acid salt 30 . Therefore, excessive concentrations of (NH 4 ) 2 SO 4 changes the pH in the reaction system, thereby missing the optimal pH of DapDH. Kinetic studies were also carried out to test the products in the L-lysine biosynthetic pathway for inhibition of DapDHs, for example L-lysine, meso-DAP and L-isomer of DAP (i.e., L,L-DAP). Although L-lysine, as the end-product in pathway, controls multiple enzymes activity, including AK and DHDPS 3,31 , it has no inhibition on oxidative deamination and reductive amination (Fig. 3e). Consistent with the previous results 18,24,32 , the L,L-DAP inhibited only the deamination of meso-DAP. Conversely, meso-DAP inhibited slightly the amination of THDPA, especially for BfDapDH and CtDapDH (Fig. 3f). Fig. 1, the succinylase pathway is used as the only pathway for meso-DAP biosynthesis catalyzed by four enzymatic steps in E. coli. Previous studies 3,23 have suggested that introduction of the DapDH in L-lysine producer E. coli is beneficial to increase the L-lysine production. As mentioned above, six DapDHs from different bacterial are able to catalyze the biosynthesis of meso-DAP in E. coli ∆dapD/∆dapE (Fig. S3). However, different DapDHs had different temperature optimum and stability (Fig. 2). In addition, our previous work has indicated that the optimal fermentation temperature is 40 °C for producing L-lysine by E. coli LATR12 (Fig. S4). To investigate whether the introduction of DapDHs would improve the L-lysine productivity in LATR12, we compared the effects of these DapDHs on L-lysine production in the DapD-deficient strain LATR12-1. Expectedly, heterogeneous expression of DapDHs was able to complement the cell growth and L-lysine production of the LATR12-1 (Fig. 4a,b). However, heterogeneous expression of BfDapDH or CgDapDH had a certain negative role on glucose consumption, cell growth and L-lysine production, especially for BfDapDH. The low activity of BfDapDH is most likely due to its low expression 9 , whereas the inappropriate temperature may be contributed to the low activity of CgDapDH 32 . This speculation has been demonstrated in the analysis of the crude enzymatic activity (Table S2). Hudson et al. 9 pointed out that the low specific activity is an innate property of BfDapDH, whereas the activity of CgDapDH was decreased with increasing the incubation time at 40 °C (Fig. 2b). Conversely, the other DapDHs exhibited great momentum in improving the fermentative performance of LATR12 (Fig. 4).

Comparing the effects of the different DapDHs on L-lysine production in E. coli. As shown in
Overexpression of BsDapDH showed the best performance in the maximum specific growth rate (μ max ; 0.27 h −1 ), followed by the StDapDH (0.25 h −1 ), UtDapDH (0.25 h −1 ) and CtDapDH (0.21 h −1 ). Interestingly, LATR12-1(ddh St ) (10.3 ± 0.3 g/L) showed the highest production of L-lysine, whereas the L-lysine production of LATR12-1(ddh Bs ) (9.8 ± 0.5 g/L) was only slightly higher than that of LATR12 (9.3 ± 0.4 g/L) (Fig. 4b). DapDH catalyzes the biosynthesis of meso-DAP, which can be used as processor for the biosynthesis of peptidoglycan and L-lysine ( Fig. 1) 13 . However, the excessive increase in cell growth is not good for L-lysine production because more carbon source enter into the biosynthesis of peptidoglycan rather than L-lysine. To do this, we conceived that heterogeneous expression of StDapDh in E. coli is beneficial to construct an L-lysine producer with good productive performance.
optimizing the expression mode of StDapDH to enhance the carbon flux in diaminopimelate pathway. As stated above, StDapDH plays a positive role on improving L-lysine production by E. coli, but its catalytic efficiency is controlled by nucleotide-cofactor, substrate and product. In this study, we aimed to enhance the L-lysine productivity of LATR12 by optimizing the integrated mode of StDapDH-coding gene in LATR12 genome. The integrated modes included three dimensions: (1) the StDapDH-coding gene integrates at dapD loci of LATR12, resulted a LATR12∆dapD::ddh St ; (2) the StDapDH-coding gene integrates at rpiB loci of LATR12, resulted a LATR12∆rpiB::ddh St ; (3) the StDapDH-coding gene integrates at rpiB loci of LATR12-2 with weakened DapD, resulted a LATR12-2∆rpiB::ddh St . The original strain LATR12 and these recombinant strains were then used to investigate the efficiency of L-lysine fermentation process. Compared with LATR12, the disruption of rpiB (encoding ribose-5-phosphate isomerase B, a nonessential enzyme for growth of E. coli K12) 33 did not affect the cell growth and L-lysine production (Fig. S5). The data of glucose consumption and cell growth showed that the integrated mode of StDapDH-coding gene did not significantly change the glucose consumption and cell growth (Fig. 5a,b). However, the L-lysine production varied obviously with the change of integrated mode (Fig. 5c). The highest L-lysine production was observed for LATR12-2∆rpiB::ddh St (10.8 ± 0.6 g/L), followed by LATR12∆rpiB::ddh St (10.1 ± 0.4 g/L) and LATR12∆dapD::ddh St (9.9 ± 0.5 g/L). Previous studies 18,24,32 and our results (Fig. 3f,g) have proved that meso-DAP inhibits slightly the amination of THDPA, whereas L,L-DAP inhibits the deamination of meso-DAP, which are likely to cause more meso-DAP into decarboxylation catalyzed by diaminopimelate decarboxylase rather than into deamination because of exist of L,L-DAP. In addition, these results reconfirmed that heterogeneous expression of CgDapDh in E. coli is not better than that of StDapDh for L-lysine production (Fig. 5).
In order to know the reasons of change, we investigated the relative expression level of genes in L-lysine biosynthetic pathway from L-aspartate (i.e., lysC, metL, thrA, asd, dapA, dapB, ddh, dapD, dapC, dapE, dapF, lysA and lysP) in original strain and these recombinant strains (Fig. 5d). The levels of transcription of lysC, asd, dapA, dapB, lysA and lysP were significantly increased with introducing StDapDH in original strain LATR12. However, the relative expression levels of metL and thrA, as the bifunctional genes encoding aspartate kinase and homoserine dehydrogenase, were controlled by the expression levels of genes in succinylase pathway. As can be seen from Fig. 5d, metL and thrA exhibited an increasing expression level only by weakening or deleting dapD, whereas their expression levels suddenly decreased in LATR12∆rpiB::ddh St and LATR12-2∆rpiB::ddh St . Expectedly, the expression levels of genes in succinylase pathway (i.e., dapD, dapC, dapE and dapF) decreased even disappeared by weakening or deleting dapD. Interestingly, the expression levels of genes in succinylase pathway increased slightly with introducing StDapDH in original strain LATR12 (Fig. 5d).
Optimizing the availability of ammonium to improve the production efficiency of L-lysine in recombinant strains. In the course of L-lysine biosynthesis, the ammonium availability is one of greatest importance in either succinylase pathway or dehydrogenase pathway (Fig. 1). However, the ammonium concentration for stimulating the function of dehydrogenase pathway is higher than that of succinylase pathway 9,19 . Assuming that an increased ammonium availability could improve the fermentation performances of LATR12∆dapD::ddh St , LATR12∆rpiB::ddh St and LATR12-2∆rpiB::ddh St , we optimized the initial concentration of ammonium (i.e., (NH 4 ) 2 SO 4 ) in MS medium. As shown in Table 2, the maximum L-lysine production, cell growth and α obtained at the initial (NH 4 ) 2 SO 4 concentration of 20 g/L for LATR12-2∆rpiB::ddh St (12.3 ± 0.6 g/L of L-lysine) and LATR12∆rpiB::ddh St (10.9 ± 0.5 g/L of L-lysine), whereas the optimal (NH 4 ) 2 SO 4 concentration Although the maximal specific production rate of L-lysine (q Lys, max. ) was kept at a higher level at ≥15 g/L of (NH 4 ) 2 SO 4 , the L-lysine production, cell growth and α decreased with increasing the (NH 4 ) 2 SO 4 concentration. This is because that the high ammonium concentration inhibits the cell growth (Table 2) 34 . In order to understand the mechanism of ammonium uptake, we investigated the relative expression level of ammonium transporter (AmtB, encoded by amtB gene) and its regulatory protein (Uridylyltransferase; UTase, encoded by glnD gene) between without (NH 4 ) 2 SO 4 and with 20 g/L of (NH 4 ) 2 SO 4 by semiquantitative RT-PCR (Fig. S6). AmtB (encoded by amtB) is the main ammonium uptake system in E. coli 35 , but its function is regulated by UTase and a PII-type GlnK protein (for review, see Burkovsk et al.) 36 . Consistent with the previous results 37, 38 , the expression level of amtB was decreasing, whereas the expression level of glnD was increasing with the increase of (NH 4 ) 2 SO 4 (Fig. S6). Interestingly, the expression level of amtB in cells grown without (NH 4 ) 2 SO 4 was much higher than that of cells grown with 20 g/L of (NH 4 ) 2 SO 4 , especially for LATR12∆dapD::ddh St (186-fold). Conversely, the expression level of glnD in cells grown without (NH 4 ) 2 SO 4 was lower than that of cells grown with 20 g/L of (NH 4 ) 2 SO 4 (Fig. S6). These results showed that LATR12∆dapD::ddh St was more sensitive to ammonium concentration than the other test strains.
Changes of carbon flux in LATR12, LATR12∆dapD::ddh st and LATR12-2∆rpiB::ddh st . As mentioned above, introducing the StDapDH in DapD-deficient or DapD-attenuated strain increased significantly the performance of L-lysine production as compared with the original strain LATR12. To study the effects of StDapDH on L-lysine production, the changes of carbon flux in LATR12, LATR12∆dapD::ddh St and LATR12-2∆rpiB::ddh St were studied using GC-MS. More than 70 intracellular metabolites showed different levels in LATR12, LATR12∆dapD::ddh St and LATR12-2∆rpiB::ddh St . Among these 70 metabolites, 23 of them were closely related to L-lysine production in the biosynthetic pathway. To get a more detailed view of the changes in carbon flux caused by the introduction of StDapDH, the relative content of these 23 metabolites were determined in the post-logarithmic phase (Table S3). As shown in Fig. 6, the content of intermediates in pentose phosphate (PP) pathway including glucose-6-phosphate, frucose-6-phosphate and glyceraldehydes-3-phosphate were higher, but the content of phosphoenolpyruvate and pyruvate as the substrates of carbon anaplerotic reaction were slightly lower in recombination strains than in LATR12. It has been proven that 4 mol of NADPH is required for the production of 1 mol of L-lysine, and PP pathway is generally considered major pathway for NADPH formation 1 . This is why introduction of StDapDH led to elevated levels of PP pathway intermediates. The decrease of phosphoenolpyruvate and pyruvate could potentially be linked to the original strain used in the study (Table 3). MF disrupts the TCA cycle, and the MF-resistant mutants show a higher activity of phosphoenolpyruvate carboxylase 39 . However, the content of intermediates in TCA cycle was decreased during introduction of StDapDH in LATR12 except succinyl-CoA and oxaloacetate (OAA), which are the co-precursors for L-lysine biosynthesis (Fig. 6). Previous results indicated that the L-lysine biosynthetic pathway becomes even more efficient because of introduction of StDapDH 3,23 . From which we can infer that more carbon flux flow into OAA. In addition, Kind et al. 40 pointed out that succinyl-CoA serves as precursor for L-lysine biosynthesis via succinylase pathway. This may be the reason for the increase of succinyl-CoA in LATR12∆dapD::ddh St and LATR12-2∆rpiB::ddh St , in which dehydrogenase pathway is the main pathway for L-lysine biosynthesis. As another co-precursor, the content of L-glutamate was slightly higher in recombination strains than in LATR12. Predictably, the content of intermediates in terminal pathway of L-lysine biosynthesis was dramatically increasing in recombination strains except L-N-Succinyl-2-amino-6-ketopimelate and N-Succinyl-L,L-2,6-diaminopimelate, which are the intermediates in succinylase pathway. It should be noted that L-homoserine, as by-product of L-lysine production, in LATR12∆dapD::ddh St was higher than that in LATR1 and LATR12-2∆rpiB::ddh St (Fig. 6).
Fed-batch fermentation of LATR12 and LATR12-2∆rpiB::ddh st . The production performance of strains LATR12 and LATR12-2∆rpiB::ddh St was investigated in a fed-batch process. Figure 7 shows the time profiles of fed-batch fermentations in a 5-L jar fermenter. Fed-batch fermentation of LATR12-2∆rpiB::ddh St resulted in 119.5 ± 7.2 g/L of L-lysine with a productivity of 2.99 g/(L•h) and carbon yield of 49.1%. However, fed-batch fermentation of LATR12 resulted in 71.8 ± 5.2 g/L of L-lysine with a productivity of 1.80 g/(L•h) and carbon yield of 35.3%. Consistent with the results of GC-MS in shake flasks, the yield of phosphoenolpyruvate, pyruvate, α-ketoglutarate and succinate were lower, but the yield of succinyl-CoA and oxaloacetate were higher in LATR12-2∆rpiB::ddh St than in LATR12 (Table S4). Moreover, the yield of L-methionine was lower, whereas the yield of L-glutamate was slightly higher in LATR12-2∆rpiB::ddh St than in LATR12 (Table S4). Thus, the final strain LATR12-2∆rpiB::ddh St also allowed efficient L-lysine production under fed-batch fermentation.

Conclusions
For the first time, introduction of DapDH with high temperature optimum was identified as a critical factor for efficiently producing L-lysine in E. coli. It is clear from the study of the functions and kinetic properties of DapDHs that different DapDHs show a huge difference in the oxidative deamination and reductive amination, and show a higher catalytic efficiency for reductive amination than for oxidative deamination except the BfDapDH (Tables 1 and S1). In addition, different DapDHs show different responses for nucleotide-cofactor, substrate and product. For example, the activity of CtDapDH and BsDapDH was dramatically decreasing with the increase of the concentration of NADP + , whereas the CgDapDH showed the minimal changes. In addition, the integrated mode and ammonium availability were also investigated, indicating that the co-existence of two pathways and sufficient ammonium availability are good for increasing the final titer of L-lysine with a high carbon yield and productivity in E. coli. Fed-batch fermentation of the target strain LATR12-2∆rpiB::ddh St resulted in 119.5 ± 7.2 g/L of L-lysine with a carbon yield of 49.1% and productivity of 2.99 g/(L•h). These results indicated that overexpression of thermostable StDapDH to redirect diaminopimelate pathway has great potential to improve the efficiency of L-lysine production in E. coli. Although the efficiency of L-lysine production of LATR12-2∆rpiB::ddh St is relatively low so that it should not be used for the practical industrial application level,   . The PCR reaction system and procedure was set following our previous reports 5 . The transcriptional levels were normalized to the 16S rRNA from the same RNA samples. Each sample was analyzed in triplicate.
Analytical methods. A sample was taken from the shake flasks or fermenter every 2 or 4 h. A half of sample was used to measure the biomass concentration using a spectrophotometer at 600 nm or by gravimetric analysis. The correlation factor between dry cell weight (DCW) and OD 600 was determined as 0.277 (1 OD 600 = 0.277 g DCW). The other half of sample was diluted 100-fold, and then used to determine the glucose and l-lysine concentration using an SBA-40E immobilized enzyme biosensor (Shandong, China). The intracellular metabolites of different strains were analyzed by gas chromatography-mass spectrometry (GC-MS) according to the previous described 45 . By the end of fermentation, the fermentation liquors were also used to determine the concentration of by-products (including amino acids and organic acids) by high performance liquid chromatography (HPLC) according to the procedure described by Xu, et al. 46 . All data were collected from three independent culture samples, and then were analyzed statistically by Student's t test with a two-tailed distribution.