Lysine Decarboxylase with an Enhanced Affinity for Pyridoxal 5-Phosphate by Disulfide Bond-Mediated Spatial Reconstitution

Lysine decarboxylase (LDC) catalyzes the decarboxylation of l-lysine to produce cadaverine, an important industrial platform chemical for bio-based polyamides. However, due to high flexibility at the pyridoxal 5-phosphate (PLP) binding site, use of the enzyme for cadaverine production requires continuous supplement of large amounts of PLP. In order to develop an LDC enzyme from Selenomonas ruminantium (SrLDC) with an enhanced affinity for PLP, we introduced an internal disulfide bond between Ala225 and Thr302 residues with a desire to retain the PLP binding site in a closed conformation. The SrLDCA225C/T302C mutant showed a yellow color and the characteristic UV/Vis absorption peaks for enzymes with bound PLP, and exhibited three-fold enhanced PLP affinity compared with the wild-type SrLDC. The mutant also exhibited a dramatically enhanced LDC activity and cadaverine conversion particularly under no or low PLP concentrations. Moreover, introduction of the disulfide bond rendered SrLDC more resistant to high pH and temperature. The formation of the introduced disulfide bond and the maintenance of the PLP binding site in the closed conformation were confirmed by determination of the crystal structure of the mutant. This study shows that disulfide bond-mediated spatial reconstitution can be a platform technology for development of enzymes with enhanced PLP affinity.


Introduction
L-lysine is an essential amino acid and industrially important material used in animal feed and food and dietary supplements. It can be synthesized from aspartate [1,2]. The aspartate is converted into L-aspartate semialdehyde (ASA) by the consecutive reaction of two enzymes. ASA is a precursor for the biosynthesis of various amino acids such as L-threonine, L-isoleucine, Lmethionine, and L-lysine. On the L-lysine biosynthetic pathway, ASA is condensed with pyruvate to generate dihydrodipicolinate (DHDP). DHDP reductase reduces DHDP to produce tetrahydrodipicolinate (THDP). Currently, four different pathways for the biosynthesis of Llysine that branch out from THDP have been reported in bacteria [3]: the succinylase pathway, the acetylase pathway, the mDAP dehydrogenase pathway, and the recently discovered PLOS ONE | DOI: 10.1371/journal.pone.0170163 January 17, 2017 1 / 16 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 BL21(DE3)-T1 R , which was grown in 1 L of LB medium containing 100 μg/mL ampicillin at 37˚C until the OD 600 reached 0.7. After induction with 1.0 mM 1-thio-β-D-galactopyranoside (IPTG), the culture medium was maintained for a further 20 h at 18˚C. The culture was then harvested by centrifugation at 4,000 × g for 20 min at 4˚C. The cell pellet was resuspended in buffer A (40 mM Tris-HCl, pH 8.0) and disrupted by ultrasonication. The cell debris was removed by centrifugation at 13,500 × g for 30 min and the lysate was applied to a Ni-NTA agarose column (Qiagen). After washing with buffer A containing 30 mM imidazole, the bound proteins were eluted with 300 mM imidazole in buffer A. Finally, trace amounts of contaminants were removed by size-exclusion chromatography using a Superdex 200 prep-grade column (320 mL, GE Healthcare) equilibrated with buffer A. All purification experiments were performed at 4˚C. SDS-polyacrylamide gel electrophoresis analysis of the purified proteins showed a single polypeptide of 44.0 kDa that corresponded to the estimated molecular weight of the SrLDC monomer. The purified protein was concentrated to 65 mg mL -1 in 40 mM Tris-HCl, pH 8.0. Site-directed mutagenesis experiments were performed using the Quick Change site-directed mutagenesis kit (Stratagene). The production and purification of the SrLDC mutants were carried out by the same procedure employed for the wild-type protein. Primers used for cloning and site-directed mutagenesis are listed in S1 Table. UV/Vis spectroscopy The amount of PLP covalently bound to the SrLDC proteins was monitored using UV/Vis absorbance spectroscopy [21,22]. The final concentration of 45.42 μM of SrLDC proteins was used. Spectra of the SrLDC proteins between 300 and 500 nm were recorded on a SHIMADZU UV-VIS Spectrophotometer UV-1800 in 0.1 M potassium phosphate (pH 6.0) buffer.

Isothermal titration calorimetry
The binding affinity between the SrLDC proteins and PLP was measured using a Nano ITC model (TA Instruments) at 20˚C. Protein samples at 100 μM were prepared in 40 mM Tris, pH 8.0. For titrations of SrLDC with PLP, 1.96 μl of 2.5 mM PLP was injected 25 times and each injection was monitored at 200-second intervals. The protein solution in the ITC reaction cell was stirred at 250 RPM. The heat of cofactor dilution into the buffer was subtracted from the reaction heat. Calculated heat area per injection was used for fitting in the independent binding model for calculation of the thermodynamic parameter (K d ).

LDC activity assay
To measure the lysine decarboxylase (LDC) activity of the SrLDC proteins, 56.78 μM of purified enzyme was added to 200 μl of reaction mixture containing 0.1 M potassium phosphate, pH 6.0, 50 μM L-lysine, and various concentrations of PLP. The reaction mixtures were incubated at 37˚C for 30 sec. The reaction was stopped by heating the solution at 90˚C for 5 min. After centrifugation at 13,500 × g for 5 min, the remaining L-lysine was detected using the lysine oxidase/peroxidase method. Lysine oxidase converts the remaining L-lysine into 6amino-2-oxohexanoate, NH 3 , and H 2 O 2 , and the H 2 O 2 is then reduced by peroxidase using 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) (ABTS). After the LDC reaction, equal volume of 2x lysine oxidase/peroxidase solution (0.1 unit ml -1 lysine oxidase, 1 unit ml -1 peroxidase, and 3.6 mM ABTS in 0.1 M potassium phosphate buffer, pH 8.0) was added to the LDC reaction mixture. The amount of oxidized ABTS was detected by measuring absorbance at 412 nm. To investigate the effect of pH on the LDC activity, the LDC activity assay was carried out at a pH range from 5 to 10. For thermal stability experiments, enzymes were preincubated at 37˚C and 60˚C, and the pre-incubated SrLDC proteins were used in the LDC activity assays. All experiments were performed in triplicate.

Cadaverine conversion assay
The cadaverine conversion assay was performed using a procedure similar to that described for the LDC activity assay. The conversion mixture of 5 mL contained 0.5 M potassium phosphate buffer (pH 6.0), 0.5 M L-lysine, 4.54 μM of purified enzyme, and various concentrations of PLP. The conversion mixture was then incubated at 37˚C for 30 min. Samples were periodically collected and the cadaverine conversion was monitored by measuring the amount of the remaining L-lysine. Crystallization, data collection and structure determination of SrLDC A225C/T302

Melting temperature (T m ) measurement
Crystallization of the purified SrLDC A225C/T302C mutant protein was initially performed with commercially available sparse-matrix screens from Rigaku and Molecular Dimensions by using the hanging-drop vapor-diffusion method at 20˚C. Each experiment consisted of mixing 1.0 μL of protein solution (65 mg mL -1 in 40 mM Tris-HCl, pH 8.0) with 1.0 μL of reservoir solution and equilibrating the drop against 0.5 mL of reservoir solution. The SrLDC A225C/T302C mutant was crystallized in the condition of 1.4 M ammonium sulfate, 0.1 M sodium cacodylate, pH 6.5, and 0.2 M sodium chloride. The crystals were transferred to a cryo-protectant solution composed of the corresponding condition described above and 30% (v/v) glycerol, fished out with a loop larger than the crystals, and flash-frozen by immersion in liquid nitrogen. All data were collected at the 7A beamline of the Pohang Accelerator Laboratory (PAL, Pohang, Korea), using a Quantum 270 CCD detector (ADSC, USA). The SrLDC A225C/T302C crystals diffracted to 1.8 Å resolutions. All data were indexed, integrated, and scaled using the HKL-2000 software package [23]. The SrLDC A225C/T302C crystals belonged to the space group With two molecules of SrLDC A225C/T302C per asymmetric unit, the crystal volume per unit protein mass was 2.7 Å 3 Da -1 , indicating that the solvent content was 54.0% [24]. The structure of SrLDC A225C/T302C mutant was determined by molecular replacement with CCP4 version of MOLREP [25] using the structure of refined SrLDC as a model. Model building was performed manually using the program WinCoot [26] and structure refinement was performed with CCP4 refmac5 [27]. Data collection and refinement statistics are summarized in Table 1. The structure of SrLDC A225C/T302C mutant was deposited in the Protein Data Bank with PDB codes of 5GJO.

Strategy for the introduction of disulfide bond in SrLDC
In the previous study, we determined the crystal structures of SrLDC in several different forms and revealed that the protein has a highly flexible PLP site [20]. We also reported that, due to the flexible PLP binding site, the protein undergoes an open/closed conformational change at the PLP binding site depending on the PLP binding ( Fig 1A). Especially, two loops located in the vicinity of the PLP binding site, the PLP stabilization loop (PS-loop) and the regulatory loop (R-loop), undergoes a significant structural movement depending on the PLP binding ( Fig 1A). In the open conformation of SrLDC, the PS-loop and the R-loop move away from  Most of other PLP-dependent enzymes contain highly stable active site and they hold its cofactor strongly with high affinities for PLP [28][29][30][31][32][33][34]. These structural and biochemical studies indicate that the high stability at the PLP binding site contributes to the high affinity for PLP. In order to develop a SrLDC mutant with an enhanced PLP affinity and thus make the production of cadaverine using the enzyme be a cost-effective process, we performed the structure-based protein engineering. The rationale of the protein engineering is to increase the stability of the PLP binding site and consequently maintain the site in the closed conformation. Based on the structural conformation of the closed form of SrLDC, we decided to introduce an internal disulfide bond between the R-loop and its neighboring region. We selected two target residues, Ala225 and Thr302 located at the R-loop and its neighboring region, respectively, and generated the SrLDC A225C/T302C mutant by replacing these residues to cysteine (Fig 1B and 1C). These two residues contribute to the stabilization of the R-loop in the closed conformation by forming a hydrogen bond and are considered to be located in a suitable distance to form a disulfide bond each other when mutated to cysteine. We expect that the closed conformation of the R-loop by the introduced disulfide bond might subsequently cause the PS-loop to be maintained in the closed conformation ( Fig 1C). This spatial reconstitution by introduction of the artificial disulfide bond might consequently make the PLP binding site be maintained in the closed conformation and increase the affinity for PLP. To prepare the negative control, we also designed a mutant that maintains the PLP binding site in the open conformation ( Fig 1D). In the open conformation of SrLDC, the R-loop is stabilized by interaction with the N-terminal region, and we selected two residues, Gly227 located at the R-loop and Lys2 at the N-terminal region, and generated the SrLDC K2C/G227C mutant by replacing these residues to cysteine (Fig 1B and 1D). We expect that the PLP binding site of the mutant to be maintained in the open conformation by the introduced disulfide bond and exhibits much lower PLP affinity than the wild-type.
Enhanced PLP affinity of the SrLDC A225C/T302C mutant We first purified the three recombinant SrLDC proteins, SrLDC WT , SrLDC A225C/T302C , and SrLDC K2C/G227C . Surprisingly, the SrLDC A225C/T302C mutant showed a clear yellow color compared with SrLDC WT and the SrLDC K2C/G227C mutant, indicating that the SrLDC A225C/T302C mutant tightly holds the PLP cofactor (Fig 2A). To compare the amount of PLP bound to the three SrLDC proteins, we then performed the UV/Vis absorption spectra scanning from 300 to 500 nm. The SrLDC A225C/T302C mutant showed absorption peaks at 327 and 418 nm, a characteristic of internal aldimine formed between PLP and enzyme ( Fig 2B). On the other hand, SrLDC WT and the SrLDC K2C/G227C mutant showed no detectable absorption spectra (Fig 2B).
These results indicate that the SrLDC A225C/T302C mutant retained much more PLP in its binding site than SrLDC WT or SrLDC K2C/G227C mutant during the expression and purification procedures. We also performed the isothermal titration calorimetry experiments and compared the K d value of the SrLDC A225C/T302C mutant with that of SrLDC WT . SrLDC WT and the SrLDC A225C/T302C mutant showed K d values of 72 and 21 μM, respectively, indicating that the K d value of the SrLDC A225C/T302C mutant was increased in 3.4 fold compared with that of SrLDC WT (Fig 2C). These results suggest that introduction of the disulfide bond indeed increased the PLP affinity of the SrLDC A225C/T302C mutant.  concentration of PLP (Fig 3A). More dramatic differences were observed with low or no PLP supplement. Although SrLDC WT and the SrLDC K2C/G227C mutant showed no detectable activity without PLP supplement, the SrLDC A225C/T302C mutant exhibited LDC activity corresponding to 65% of SrLDC WT activity in the presence of 0.2 mM PLP ( Fig 3A). Moreover, the activity of SrLDC A225C/T302C mutant in the presence of 0.01 mM PLP was almost identical to that of SrLDC WT in the presence of 0.2 mM PLP (Fig 3A). In order to investigate conversion to cadaverine in a condition that is close to the real cadaverine production, we measured the conversion rate to cadaverine by addition of 0.5 M of lysine. In the presence of 0.2 mM PLP, the SrLDC A225C/T302C mutant showed almost 100% cadaverine conversion, whereas SrLDC WT and the SrLDC K2C/G227C did only 60% and 30%, respectively (Fig 3B). Similar to the activity comparison described above, more dramatic differences were observed with low or no PLP supplement. The SrLDC A225C/T302C mutant converted 25% of lysine into cadaverine even without PLP supplement, whereas both SrLDC WT and the SrLDC K2C/G227C mutant showed no detectable cadaverine conversion (Fig 3B). Moreover, the SrLDC A225C/T302C mutant in the presence of only 0.02 mM of PLP converted more lysine into cadaverine than SrLDC WT did in the presence of 0.2 mM of PLP (Fig 3B). Measuring the time course of cadaverine conversion showed that the difference in cadaverine conversion between SrLDC WT and SrLDC A225C/T302C mutant was even more conspicuous at the initial phase of the conversion period (Fig 3C). With 0.02 mM of PLP, the SrLDC A225C/T302C mutant exhibited even higher cadaverine conversion than SrLDC WT with 0.2 mM of PLP. We suspect that the enhanced LDC activity and cadaverine conversion of mutant SrLDC A225C/T302C , in particular with low or no PLP supplement, resulted from an enhanced stabilization of the PLP binding site by the disulfide bondmediated spatial reconstitution.
In addition, we also investigate whether the SrLDC A225C/T302C mutant shows resistance against pH increase. LDCs are known to have a low optimum pH and show reduced activity when the pH is increased. Because one proton ion is consumed by each decarboxylation reaction, leading to an increase in pH, LDCs tend to lose their activity as the reaction proceeds. We measured the LDC activity of the SrLDC A225C/T302C mutant at pH ranging from 5 to 10 and compared with that of SrLDC WT . As we expected, both the SrLDC A225C/T302C mutant and SrLDC WT exhibited the highest activity at pH 6.0, and the activity of both enzymes gradually decreased while pH increased (Fig 3D). However, it is important to note that the SrLDC A225C/T302C mutant showed a higher activity than SrLDC WT did throughout the whole pH range, and the difference in activity between the two enzymes was even more pronounced at a higher pH (Fig 3D). When the LDC activity of the mutant was measured in the presence of 0.01 mM PLP, the resistance of the SrLDC A225C/T302C mutant against pH increase was more apparent than that shown in the presence of 0.2 mM PLP. At pH 10, the activity of the mutant enzyme in the presence of only 0.01 mM PLP was up to two-fold higher than that of SrLDC WT in the presence of 0.2 mM PLP (Fig 3D). These results indicate that the SrLDC A225C/T302C mutant, compared with SrLDC WT , maintains its relatively high activity even when pH increases as a result from prolonged reactions.
Enhanced thermal stability of the SrLDC A225C/T302C mutant In general, introduction of internal disulfide bonds in proteins results in an increase in thermostability of the engineered proteins [35][36][37]. Although we introduced the internal disulfide bond to increase the stability of the active site, we expected an enhanced thermostability of the SrLDC A225C/T302C mutant. We then measured the melting temperature (T m ) of the mutant enzyme and compared with that of SrLDC WT . Interestingly, the SrLDC A225C/T302C mutant showed a T m of 56.9˚C, which is higher than the T m of 52.27˚C observed for SrLDC WT ( Fig  4A). Next, we investigated how prolonged incubation at 37˚C affected the activity of the proteins. Only 50% of SrLDC WT activity remained after incubation for one hour, and an almost complete loss of activity was observed after four hours of incubation (Fig 4B). In contrast, the activity of the SrLDC A225C/T302C mutant remained relatively high, with almost half of it present even after four hours of incubation (Fig 4B). Even after four hours of incubation, the SrLDC A225C/T302C mutant showed almost identical activity to that of SrLDC WT without incubation (Fig 4B). We also measured the activity of the proteins after incubation at 60˚C. SrLDC WT showed an immediate decrease in activity during incubation at this temperature, and an almost complete loss of activity after an incubation of two min (Fig 4C). In contrast, the SrLDC A225C/T302C mutant retained a relatively high activity up to two min of incubation at 60˚C, and still showed activity after three min of incubation (Fig 4C). These results indicate that the disulfide bond-mediated spatial reconstitution resulted in not only enhanced stability at the active site but also in increased thermostability of the enzyme.
Crystal structure of the SrLDC A225C/T302C mutant To structurally confirm the enhanced stability of the PLP binding site in mutant SrLDC A225C/T302C is derived from the introduced disulfide bond, we determined its crystal structure at 1.8 Å resolution (Fig 5A and Table 1). As expected, a disulfide bond was observed between A225C and T302C, and the PS-loop and the R-loop of the mutant showed the closed conformation (Fig 5B and 5C). Moreover, When we compared the B-factors of the This observation indicates that the engineered disulfide bond also enhanced the substrate binding affinity. Interestingly, we observed the clear electron density for PLP cofactor even without addition of PLP in the crystallization solution, indicating that the SrLDC A225C/T302C mutant retained the PLP cofactor during purification and crystallization procedures (Fig 5C). Taken together, the engineered disulfide bond seems to result in the maintenance of the R-loop in the closed conformation, and in turn leads to the PS-loop to be retained in the closed conformation as well, which consequently affects the stability of the AS-loop.

Discussion
Enhanced stability of enzymes with industrial applications may improve production yield and lead to expanded operational environments such as temperature and pH. The introduction of disulfide bonds has been used as a powerful engineering tool to increase protein stability [35,36,38]. Recent studies have also shown that disulfide bond engineering can be used in a wide range of applications such as kinetic stability and functional modification of proteins [39][40][41]. However, not all engineered disulfide bonds produce an improved enzyme due to the difficulty in predicting the conformation and thermodynamics of an engineered disulfide bond. Our successful protein engineering on SrLDC has two unique features compared with the previously reported works. First, we used disulfide bond-mediated spatial reconstitution at the cofactor binding site to increase cofactor affinity rather than increase the stability of protein folding itself. Second, we achieved the equivalent of killing two birds with a single stone; one engineered disulfide bond enhanced both the enzymatic activity and the resistance to pH and temperature of the target protein. In many cases, engineered disulfide bonds lead to the increase of either enzymatic activity or enzyme stability [42]. Extensive structural analysis seem to enable this protein engineering to be successful.
Here, it is worth to notice that appropriate flexibility at the active site is important for maximum enzymatic activity. As described above, the SrLDC A225C/T302C mutant shows a stabilized AS-loop and forms the closed substrate binding site. Based on these findings, we introduced an additional disulfide bond between the AS-loop and the PS-loop in the SrLDC A225C/T302C mutant and generated SrLDC K143C/L185C/A225C/T302C mutant (S4 Fig). We anticipate that the more stable active site results in the improved LDC activity of the mutant. As expected, the purified quadruple mutant showed a yellow color similar to the SrLDC A225C/T302C mutant. However, its enzymatic activity was only 65% even compared with that of SrLDC WT (S4 Fig). We speculate that the low activity of the quadruple mutant is caused by a high rigidity of the substrate biding site. Based on these results, we suggest that the maintenance of an appropriate flexibility at the active site is one of the important factors to be considered when we design a structure-based protein engineering.
PLP is a cofactor required in a variety of enzyme reactions such as transamination and decarboxylation [43]. Especially, enzymes involved in the biosynthesis of amino acids and their derivatives utilize PLP as an essential cofactor. For this reason, PLP has been one of the critical control factors for the production of industrially important bioproducts and many attempts have been also made to reduce the cost by a continuous supplement of PLP. As a platform technology, this approach could be used for the development of highly efficient PLPdependent enzymes that would allow more cost-effective production of valuable bioproducts.