Current biotechnological applications of L-amino acid deaminases for keto acids production

— There is a growing interest in the pharmaceutical and agrochemicals industries for the use of enantiomerically-pure amino acids. α-keto acids are commonly used in feeds, food additives, pharmaceuticals, and chemical manufacturing. Commonly, most α-keto acids are manufactured by chemical synthesis, but due to the increasing concern for environment-friendly approaches, microbial fermentation and enzymatic transformation are alternative processes for the production of keto acids. Regarding this, L-amino acid deaminase (LAAD), is a major enzyme for α-keto acids production and is only found in Proteus bacteria. In this review, we discussed the recent biological applications of the enzyme LAAD in the production of keto acids, and summarized the recent advancements in the biological production of six important a-keto acids; specifically, phenylpyruvate, α-ketoglutaric acid, pyruvate, α-ketoisocaproate, α-keto-γ-methylthiobutyric acid and α, -ketoisovaleric acid.

. These flavoenzymes catalyze an irreversible reaction (differently from aminotransferases) and do not require a specific step of cofactor regeneration, as otherwise required by the NAD-dependent dehydrogenases. However, because of the problems associated with overexpression of snake venom LAAOs in recombinant hosts and the limited substrate acceptance of the microbial counterparts, no appropriate LAAOs for biocatalysis are available (Pollegioni et al., 2013). L-Amino acid deaminases (LAADs) represent a suitable alternative to LAAOs. LAAD (first identified in the genera Proteus, Providencia, and Morganella) catalyzes the deamination of the L-isomer of amino acids, yielding the corresponding α-keto acids and ammonia without any evidence of H2O2 production. LAAD from Proteus myxofaciens (PmaLAAD), expressed in the Escherichia coli K12 strain, shows a preference for Lamino acids with aliphatic, aromatic, and sulfur-containing side chains (Pantaleone, Geller, & Taylor, 2001). This review focused on the applications of LAAD for the production of α-keto acids, and summarized the recent advancements in the biological production of six important a-keto acids; specifically, phenylpyruvate, α-ketoglutaric acid, pyruvate, α-ketoisocaproate, α-keto-γmethylthiobutyric acid, and α-ketoisovaleric acid.

EXPRESSION AND PURIFICATION OF RECOMBINANT LAADS
In order to produce LAAD in bulk for biochemical characterization and biotechnological purposes, this flavoenzyme must be expressed in different microbial hosts. Recombinant type-I L-Amino acid deaminase from Proteus myxofaciens (PmaLAAD) was expressed as a full-length protein in E. coli BL21(DE3) cells (at 28 °C), obtaining an enzyme with a specific activity of 0.6 U mg -1 protein on the substrate L-Phe (in the crude extract) (Motta, Molla, Pollegioni, & Nardini, 2016). Chromatographic steps were ineffective for enzyme purification due to the recombinant protein and its tight association with the bacterial membrane fragments. Thus, PmaLAAD could only be separated from soluble proteins by ultracentrifuging the crude extract Resuspension of the pellet formed by the membrane fraction (and containing the active enzyme) resulted in modest increases of a2-to 3.4-fold within specific activity (Table 1) (Motta et al., 2016;Pantaleone et al., 2001). The dual objectives of producing and preparing a pure PmaLAAD aided in designing a His-tagged fusion protein. The presence of the tag located either at the protein's Nor Cterminuses have no effect on the enzyme expression yields or its specific activity in the crude extract. A His-tagged PmaLAAD variant was purified to homogeneity by immobilized metal affinity chromatography as a soluble, yet inactive enzyme. Complete activity can only be restored (and reaching the crude extract's assayed value) after adding E. coli membranes to the purified enzyme sample (Motta et al., 2016). Type-I PmirLAAD in its homologous form was also produced as a His-tagged recombinant protein. In this case, the recombinant protein was expressed at 20 °C to prevent the overproduced protein from accumulating in inclusion bodies. Metal-chelating chromatology followed by solubilization with the detergent n-dodecyl-β-D-maltoside was used to purify the enzyme. Unfortunately, negatively affected enzyme activity was suboptimal; this could be due to the interaction of the latter with the bacterial membranes. The specific activity of the pure protein was 0.94 U mg -1 on L-Phe (Table 1). Therefore, due to partial loss of enzymatic activity and high detergent costs, the authors concluded that biocatalytic applications are better served by using whole cells to express the recombinant protein instead (Hou et al., 2015). Similar expression yields were obtained for the His-tagged Pm1LAADtoo. through preparation of a specific activity on the L-Phe of 1.54 U mg -1 protein (compared to 1.2 U mg -1 for PmaLAAD) was achieved after the ultracentrifugation purification step (Baek et al., 2011). Interestingly, by the optimized codon usage for E. coli expression and the gene was fused to the pelB leader peptide, the recombinant chimeric protein was expressed exclusively as inclusion bodies (Liu et al., 2013). This confirms that overoptimization of codon usage could induce an actual increase in protein expression but compromise of its cellular solubility. Nonetheless, the researchers were able to solubilize (to use 8 mol urea) and refold protein from the inclusion bodies (with a yield of 40 percent), obtaining an enzyme preparation with a high degree of purity but with a reduced specific activity (0.74 U mg-1 of pure Pm1LAAD) and a very high Km for L-Phe) (31.5 mM). This result indicates that the refolding approach facilitated partial recovery of the conformation/activity of the enzyme (Table  1) (Liu et al., 2013).As a general rule, the association of the enzyme to the bacterial membranes is fundamental for the catalytic competence of type-I LAADs. Consequently, when the putative N-terminal transmembrane α-helix of PmaLAAD was removed, the resulting ΔN-LAAD deletion variant is rendered nearly inactive because of its inability to accomplish membrane interaction (in the crude extract even) (Motta et al., 2016). In contrast, with type-II LAADs, the enzymatic activity is also (partially) retained in the deletion variants lacking the transmembrane α-helix. Regarding these enzymes, the insertion module is relatively sufficient for providing the necessary interaction of the protein with the bacterial membranes, even without the Nterminal transmembrane α-helix (Ju et al., 2016). Further debates have ensued since then, focusing on using whole cells to express LAAD vs. using recombinant purified enzymes. Whole cells are employed more consistently due to their production simplicity and the low cost; some authors assert that side reactions can occur, thus accumulating into small amounts of impurities and causing adverse effects on the overall yield (

ELEMENTAL SOLUTIONS OF AMINO ACIDS
In 2001, Pantaleone group was the first to propose employing LAAD in biocatalysis. The recombinant enzyme from P. myxofaciens oxidized the 20 natural L-amino acids, thus revealing its preference for aliphatic, aromatic, and Scontaining amino acids (Pantaleone et al., 2001) . This same group used a partially purified enzyme preparation to conduct specific activity on the L-Phe of 2.35 U mg -1 protein to determine its pH levels (which measured at 7.5). They also collaborated with Ian Fotheringham and Nick Turner the following year, using the same E. coli cells, which expressed the recombinant protein and 40 equivalents of ammonia-borane. The result was a deracemization of 2.5 mmol DL-Leu (in 50 mmol ammonium formate, pH 6.7), and a continuing production of the D-enantiomer, with a 90% yield and an ee> 99% (Alexandre et al., 2002). Together, these groups achieved a conversion higher than 85% obtained through racemic solutions of norleucine, methionine, O-benzyl serine, cyclopentyl-glycine (Alexandre et al., 2002), and isoleucine (Enright et al., 2003).

PHENYLPYRUVIC ACID PRODUCTION
In recent years, phenylpyruvic acid (PPA) has been utilized as a product reference of L-AAD catalyzed L-Phe conversion ( Table 2). PPA is a limiting raw material for the sweetener aspartame (production > 10,000 tons year-1) and employed primarily for patients with kidney diseases to reduce urea accumulation in their diets PPA producers were subsequently screened and selected using a variety of microorganisms such as P. Vulgaris, which was the most effective strain ( encoding aminotransferases. This led to increased PPA production on a higher scale (up to 10.0 g L -1 corresponding and reaching full substrate conversion) by utilizing a L-AAD variant that harbors the D165K/F263M/L336M substitutions, generated by error-prone PCR (Hou et al., 2016). The evolved L-AAD variant shows a 1.6-fold increase in maximal activity (2.63 μmol min-1 mg -1 protein), see (Table 2). Under optimized fed-batch conditions (i.e., feeding 4 g L -1 of L-Phe per hour), the maximal PPA production reached 21 g L -1 . However, PPA's current chemical synthesis methods have relatively low yield and result in unwanted environmental pollution. In this regard, biotransformation processes seem unsuitable as industrial alternatives.

PRODUCTION OF Α-KETOGLUTARIC ACID
The Jian Chen group used L-AAD from P. mirabilis KCTC to transform L-glutamic acid into 2-ketoglutaric acid (α-KG), which can be utilized in various applications such as pharmaceuticals, fine chemical, and the food and animal feed industries. It is believed that the biocatalytic process bypasses the multi-step chemical process regarding succinic acid and oxalic acid diethyl esters, thereby employing toxic compounds, generates toxic waste, and showing a much lower product yield in comparison. The ΔN-L-AAD variant enzyme was used after refolding (Liu et al., 2013). This preparation of enzymes showed the optimal activity at pH 8.0 and 45 °C and an apparent Vmax = 0.73 U mg -1 protein and Km for L-Phe of 31.5 mmol (Table 2 at air saturation). Studies have shown that 12.6% of α-KG can be produced in 6 hours, which is considered to be optimal conditions (12 g L -1 L-Glu, 0.1 g L -1 ΔN-LAAD, 5 mmol MgCl2, pH 8.0 and 40 °C) (Liu et al., 2013). Notably, the conversion was hampered by-product inhibition (Ki = 12.6 mmol for α-KG). Notably, product inhibition (Ki = 12.6 mmol for α-KG) significantly hampered the conversion process. Later on, the wild-type L-AAD from P. mirabilis KCTC 2566 was overproduced in Bacillus subtilis 168 cells (Table 2)  . This system produced 53.7 g α-KG L -1 when 100 g of monosodium glutamate was used and up to 89.1 g α-KG L -1 when the substrate was continuously fed at a constant rate of 6 g L -1 h -1 with an initial concentration of 50 g L -1 . Regardless of the advantage of reducing environmental pollution and the significant α-KG production reached by L-AAD biocatalysis, this bioconversion system is not ready to be turned into an industrial level.

PYRUVATE PRODUCTION
Currently, the production of pyruvate by the biotechnological process has become a fast-growing trend, and fermentation processes have been extensively . The genes encoding of these two alanine uptake transporters and one pyruvate uptake transporter were eliminated to minimize their usage by the cell system. These cells also succeeded in overproduction of both the wild-type and an isolated variant (30,000 clones generated by error-prone PCR) of Pm1LAAD. Pm1ep3, in its best form, contains 9 substitutions overall (at a great distance of > 15 Å from the active site) and displaying a 1.8fold increase in maximal activity (3.1 μmol min-1 mg -1 protein) with a ~ 2-fold decrease in Km for L-Ala (23.3 mmol) (Hossain et al., 2016). Interestingly, the enzyme can also be active on D-Ala (Vmax = 0.81 μmol min -1 mg -1 protein, Km= 31.5 mM), allowing further usage of the racemic amino acid to produce pyruvic acid. LAAD's ability to use the D-isomer of the amino acids as substrate require a more detailed evaluation. This is due to the fact that the four-location model for substrate binding discriminates the enantiomers for their ability for hydride transferal from their αC-H to the N (5) flavin position, see above. Under ideal conditions (i.e., 50 g L -1 DL-Ala, 20 g L -1 whole-cells expressing pm1ep3, pH 8.0, at 40 °C and 220 rpm) 14.6 g L -1 of pyruvate was achieved, corresponding to a biotransformation ratio of ~ 29% of in just 36 hours (Hossain et al., 2016).

T. GLABRATA FOR IMPROVING PYRUVATE PRODUCTION
The earliest microbial stain used for industrial-scale pyruvate production was the multi-vitamin auxotroph T. glabrata. The highest pyruvate production by T. glabrata reached 94.3 g/L within 82 h, and the yield on glucose substrate was about 0.635 g/g by a NaCl-tolerant strain (Lütke-Eversloh, Santos, & Stephanopoulos, 2007). Further investigation was focused on controlling the NADH level to enhance the rate of glycolysis. Two different NADH re-oxidation pathways were introduced to reduce NADH in the cytoplasm and in mitochondria. The highest pyruvate yield and productivity were increased to 38% and 21%, respectively (Qin, Johnson, Liu, & Chen, 2011). An optimal nutrient environment is also crucial to achieving a high pyruvate yield. With nitrogen optimization, the pyruvate concentration reached 85.9 g/L within 72 h with the optimal addition of thiamine, nicotinic acid, pyridoxine, biotin, and riboflavin (Yang et al., 2014).

CONSTRUCTING C. GLUTAMICUM FOR THE PRODUCTION OF PYRUVATE
C. glutamicum is well-established organic acid producer, it has been established as a microbial cell factory for pyruvate production. C. glutamicum strain was engineered by inactivating its pyruvate dehydrogenase complex, NAD +dependent l-lactate dehydrogenase, quinone oxidoreductase, and alanine transaminases AlaT and AvtA, and a reduced variant of acetohydroxyacid synthase was presented. Under low oxygen tension and in fed-batch fermentations, the strain formed 45 g/L of pyruvate, with a yield of 0.97 mol/ mol of glucose (Wieschalka et al., 2012;Zou, Hang, Chu, Zhuang, & Zhang, 2009). Even though this yield of pyruvate was minor than that of T. glabrata, the conversion rate from glucose was higher with C. glutamicum.

CONSTRUCTING S. CEREVISIAE FOR THE PRODUCTION OF PYRUVATE
S. cerevisiae does not usually produce organic acids in large scale, its tolerance to pH makes it a suitable producer of pyruvate.Van Maris et al., (2004) reported that the highest production of pyruvate (135 g/L) was obtained by a pyruvate decarboxylase-deficient mutant of S. cerevisiae TAM. Yet, supplementation of a C2 compound was required for the growth of this mutant because of the pyruvate decarboxylase deficiency. In another study, by using rational cofactor engineering and adaptive evolution, a pyruvate concentration of 75.1 g/L with a yield of 0.63 g pyruvate/g glucose was obtained (Wang et al., 2012). Furthermore, by regulating the thiamine biosynthesis genes, which are related to unstable cellular growth on sole carbon sources, a pyruvate amount of 8.21 ± 0.30 g/L was achieved after 96 h of cultivation (

A BIOLOGICAL TRANSFORMATION APPROACH TO IMPROVE PYRUVATE PRODUCTION
In current years, whole-cell and enzyme bioconversion methods have been developed. E.g., lactate dehydrogenase, glycolate oxidase, and lactate oxidase were used to convert lactate into pyruvate. In a previous study they used glycolate oxidase in Pichia pastoris to catalyze the conversion of pyruvate from L-lactate by the whole-cell transformation process (Payne et al., 1995). Acetaldehyde and carbon dioxide were converted into pyruvate by Rhodotorulagracilis d-amino acid oxidase (d-AAO), giving a 90% conversion rate (Miyazaki, Shibue, Ogino, Nakamura, & Maeda, 2001). In another study, a high pyruvate quantity of 111.9 g/L was obtained by expressing both glycolate oxidase and catalase together (Eisenberg et al., 1997). The catalase was needed to remove the effect of H2O2, which could convert the pyruvate to acetate.

PRODUCTION OF Α-KETOISOCAPROATE
ketoisocaproate Essential branched-chain amino acids L-Val, L-Ile and L-Leu, alongside the corresponding keto acids, have different applications in the feed, food and pharmaceutical sector: they are produced in quantities of up to 5,000 tons year-1 (Becker & Wittmann, 2012). In a previous study, in order to establish a successful fermentation mechanism, a recombinant Corynebacterium glutamicum strain was designed by metabolic engineering, and the maximal α-ketoisocaproate titter was reached 9.23 g/L (Bückle-Vallant, Krause, Messerschmidt, & Eikmanns, 2014). However, with the exception of the poor yield of alpha-ketoisocaproate, an auxotroph for branch-chained amino acids is still a barrier to industrial development owing to the deletion of ilvE. Another study has fabricated a plasmid-free C. glutamicum to produce 6.1 g/L αketoisocaproate (Vogt, Haas, Polen, van Ooyen, & Bott, 2015). Yet, the production of α-ketoisocaproate of C. glutamicum by metabolic engineering is still narrow by the growth reliant on the L-isoleucine. The whole-cell biosynthesis mechanism offers a bright path to the low-cost development process of α-ketoisocaproate. In a study, αketoisocaproate was prepared using the whole-cell transformation technique of Rhodococcusopacus DSM 43250, and α-ketoisocaproate titers reached 1275 mg/L (Yuhong Zhu et al., 2011). In another research, for the development of α-ketoisocaproate from leucine, an Escherichia coli BL21 (DE3) was constructed by wholecell biocatalyst with membrane-bound L-amino acid deaminase (LAAD) from Proteus vulgaris. The highest titter was reached 69.1 gL -1 (Song et al., 2015). In another study conducted, an even higher α-ketoisocaproate production of 86.55 gL -1 and a higher L-leucine conversion rate of 94.25 percent were achieved via three engineering strategies; altering the plasmid origin with various copy numbers, modulating the mRNA composition downstream of the initiation codon, and designing the ribosome bindingsite synthesis sequences (Song et al., 2017).

PRODUCTION OF Α-KETO-Γ-METHYLTHIOBUTYRIC ACID
E. coli cells overproducing pvLAAD managed to convert L-Met into α-keto-γ-methylthiobutyric acid, which is an indirect inhibitor of tumor cell growth and a methionine supplement in livestock feed. By using previously optimized conditions . 70 g L -1 L-Met was converted by 20 g L -1 whole-cell biocatalyst into α-keto-γ-methylthiobutyric acid with a 71.2% yield. The LAAD variant harboring the K104R/A337S substitutions, generated by error-prone PCR, possesses a lower Km for L-Met (decreased from 305 to 238 mM), and 63.6 g L -1 of α-keto-γ-methylthiobutyric acid could be produced in 24 hours. This single-step enzymatic process for production of α-ketoγmethylthiobutyric acid has a great potential at the industrial level compared with the traditional multi-step chemical systems.

Α-KETOISOVALERIC ACID
α-ketoisovaleric acid (α-KIV) is as a precursor in leucine and valine synthesis. It also serves as an initial compound in vitamin B5 biosynthesis (Chassagnole, Diano, Létisse, & Lindley, 2003). α-KIV is mainly synthesized via a multistep chemical method (i.e. the hydrolysis of azlactones and the Grignard reagents with diethyloxamates) (Cooper, Ginos, & Meister, 1983;Waters, 1947). These chemical processes are costly and complex which restrain the high industrial production of α-KIV. i et al. (2017) studied an alternative biotechnological way by expressing the L-AAD from Proteus myxofaciens ATCC 19,692 in E. coli BL21 (DE3) as a whole-cell biocatalyst system. Under the optimized conditions, the α-ketoisovaleric acid production with the wild type L-AAD was 2.014 g/L. Using the 3D structural model of L-AAD from P. myxofaciens and the simulation results when docking with the L-valine, key amino acid residues (N100, Q276, R316, and F318) were identified as potential target for site-saturation mutagenesis. The evolved L-AAD improved the biotransformation to 8.197 g/L after combining the mutated sites . the rational molecular engineering of the L-AAD using site-saturation mutagenesis improved the efficiency of the biocatalysis.  This amination/deracemization cascade began at an efficient pace by utilizing numerous electron-deficient cinnamic acids that produced D-phenylglycine derivatives with sufficient optical purity. This led to increased L-isomer production when DAAO replaced LAAD. The same group paired the deamination of L-arylalanines by LAAD with the consumption of the produced keto acid by an engineered Damino acid dehydrogenase to generate the corresponding Damino acid (to achieve NADP+ recycling through glucose dehydrogenase, ( This could be a more beneficial way of generating the Denantiomer when the corresponding L-or the DL-amino acid is commercially available in the future. Lysed E. coli cells expressing PmaLAAD converted natural L-amino acids into conforming α-keto acids, then asymmetrically reduced by L-or D-isocaproate reductases into (R)-or (S)-2-hydroxy acids (Fig. 4)   Under 1 bar O2 pressure, 50 -200 mM L-amino acids were fully converted into (S)-or (R)-hydroxy acids with a > 99% ee L-Tyrosine was transformed at 200 mM concentration (a value higher than its solubility) and 100 mg scale, and up to 0.5 g of L-isoleucine was converted (Busto et al., 2014). Among the production of 2-hydroxy acids, (R)-and (S)-4 hydroxy lactic acids are used to prepare various biologically active compounds such as Saroglitazar (Lypaglyn), which can treat type II diabetes.

III. CONCLUSION
Regardless, the fermentation method yielded higher amounts of pyruvate and a-KG, other a-keto acids yielded lower amounts. As a result, the genome-scale metabolic model of amino acid synthesis pathways should be investigated in order to accumulate more a-keto acids. Furthermore, because of its high conversion rate and simple and cost-effective separation process, the enzymatic bioconversion method may be a more effective approach for a-keto acid production. Enzymes with high substrate affinity and high conversion efficiency are expected to be developed in the future as new enzyme discovery tools and genetics information become available. Lastly, advances in metabolic engineering of whole-cell biocatalysts are propelling biocatalyst use to new heights. The ability to improve the overall performance of biocatalytic processes has been demonstrated by combining enzyme engineering with the modification of recombinant strain mechanism pathways. However, some work remains to be done, such as a better understanding of enzymes.