Overview on Multienzymatic Cascades for the Production of Non-canonical α-Amino Acids

The 22 genetically encoded amino acids (AAs) present in proteins (the 20 standard AAs together with selenocysteine and pyrrolysine), are commonly referred as proteinogenic AAs in the literature due to their appearance in ribosome-synthetized polypeptides. Beyond the borders of this key set of compounds, the rest of AAs are generally named imprecisely as non-proteinogenic AAs, even when they can also appear in polypeptide chains as a result of post-transductional machinery. Besides their importance as metabolites in life, many of D-α- and L-α-“non-canonical” amino acids (NcAAs) are of interest in the biotechnological and biomedical fields. They have found numerous applications in the discovery of new medicines and antibiotics, drug synthesis, cosmetic, and nutritional compounds, or in the improvement of protein and peptide pharmaceuticals. In addition to the numerous studies dealing with the asymmetric synthesis of NcAAs, many different enzymatic pathways have been reported in the literature allowing for the biosynthesis of NcAAs. Due to the huge heterogeneity of this group of molecules, this review is devoted to provide an overview on different established multienzymatic cascades for the production of non-canonical D-α- and L-α-AAs, supplying neophyte and experienced professionals in this field with different illustrative examples in the literature. Whereas the discovery of new or newly designed enzymes is of great interest, dusting off previous enzymatic methodologies by a “back and to the future” strategy might accelerate the implementation of new or improved multienzymatic cascades.


INTRODUCTION
Chemically, an amino acid (AA) is any molecule containing a carboxylic acid and an amino group. This family of compounds is hugely heterogeneous, and includes any linear or cyclic molecule containing both substituents. The amino moiety can be situated at the α-, β-, γ-, δ-, etc. position with respect to the carbonyl group of the acid. AAs can be further functionalized with other substituents, additionally heterogenizing this family of compounds. Nonetheless, the L-isomer of α-AAs occupies a central role in biosciences, since these molecules were evolutionarily chosen to become the building blocks of one of the macromolecular scaffolds sustaining life. The 22 genetically encoded AAs present in proteins (the 20 standard AAs together with selenocysteine and pyrrolysine) are commonly referred as proteinogenic amino acids (PAAs). On the other hand, the IUPAC recommended the terminology "common amino acids" for genetically-encoded AAs and other AAs with known biological functions, such as L-DOPA (IUPAC, and IUb, 1984). By the end of the seventies, more than 140 natural AAs and AA-derivatives had been proposed in natural proteins (Uy and Wold, 1977;Hunt, 1985); more than 900 NcAAs have been suggested in plants (Yamane et al., 2010). Thousands of synthetic NcAAs not occurring naturally have been chemically synthesized by different approaches (Xue et al., 2018). In this sense, the "proteinogenic" or "non-proteinogenic" classification of AAs can result imprecise, as non-coded AAs are also commonly found in significant quantities in proteins, such as hydroxylysine and hydroxyproline. Several D-α-AAs are also widely described in peptides and proteins (Martínez-Rodríguez et al., 2010c;Grishin et al., 2020), and the emerging role of its metabolism in the innate defense has been recently suggested (Sasabe and Suzuki, 2018). On the other hand, AAs not included in the above-mentioned group have also been referred to as non-protein, non-natural, unnatural, non-canonical, nonproteinogenic or unusual, among other terminologies (Hunt, 1985;Fan et al., 2015;Blaskovich, 2016;Agostini et al., 2017;Baumann et al., 2017). We will refer to them as "non-canonical amino acids" (NcAAs).
The occurrence and biological activity of NcAAs in vivo is hugely diverse, and some of their properties have been reflected in different papers (Wagner and Musso, 1983;Hunt, 1985;Nunn et al., 2010;Yamane et al., 2010;Walsh et al., 2013;Blaskovich, 2016;Zou et al., 2018;Narancic et al., 2019;Hedges and Ryan, 2020). Broadly, several NcAAs are utilized as intermediates in primary metabolic pathways (e.g., homoserine, ornithine, citrulline,. . .). Free NcAAs also possess antimicrobial, antiproliferative, anti-inflammatory, or other biologically relevant activity (Martínez-Rodríguez et al., 2010c;Blaskovich, 2016). They serve as building blocks for many different small bioactive peptide scaffolds (Walsh et al., 2013), including hormones (Martínez-Rodríguez et al., 2010c), or confer specific structural properties to proteins (i.e., collagen or insulin). The activity of many NcAAs have been proposed not directly related with the physiology of the organism itself, but with its relationship with other organisms in its environment, suggesting physiological (toxic), deterrent (pheromonal) or other modifying roles external to the responsible species (Hunt, 1985). Microorganisms are a clear example on how evolution has taken advantage of using NcAAs over competing organisms (Nunn et al., 2010), and microbial antibiotics have been widely described (Martínez-Rodríguez et al., 2010c;Walsh et al., 2013;Blaskovich, 2016). Many of the isolated NcAAs or peptide-containing NcAAs also serve as plant-defense against predators, pathogens or other organisms competing for the same resources (Yamane et al., 2010;Rodgers, 2014). Furthermore, some NcAAs have also been shown to be toxic for humans and domestic animals (Bell, 2003;Nunn et al., 2010;Rodgers, 2014), and have been suggested as potentially linked to neurodegenerative diseases (Rodgers, 2014).

Biomedical and Biotechnological Applications of NcAAs
Enantiopure NcAAs are of considerable economic importance because of their broad industrial applications (Patel, 2013;Narancic et al., 2019). The global AA market increased from 0.7 million tons in 1985 to 9.3 Million Tons in 2019, and is expected to reach a volume of 11.9 million tons in 2025 (Ikeda and Takeno, 2013;IMARC group, 2019). A "Compound Annual Growth Rate" (CAGR) of 5.6% is expected from 2015 to 2022 (Sánchez et al., 2018). Amino acid global sales reached $20 billion in 2014 and are predicted to exceed $35 billion by 2022 (Yan and Wang, 2019). Besides the advances focused on the million-ton scale bulk PAAs L-glutamate and L-lysine, the Industry continues widening its interest toward other specialty NcAAs (Wendisch, 2020), and there is strong commercial interest in developing new amino acid applications (Sánchez et al., 2018). In fact, some NcAAs have proven key for the production of many of the 200 topgrossing pharmaceuticals (Boville et al., 2018b). Some examples are sitagliptin (Merck, $5.91 billion), pregabalin (Pfizer, $4.62 billion) or glecaprevir (AbbVie, $3.44 billion) (Belk, 2018).
From the biotechnological and biomedical point of view, NcAAs have found applications as a significant expansion of the building-block repertoire and/or as organocatalysts (Agirre et al., 2019), also in the manufacture of a wide range of pharmaceuticals (Patel, 2013;Narancic et al., 2019), or as linear and cyclic peptides (Blaskovich, 2016;Martin et al., 2018). Some of the commercial applications directly rely on the natural properties of peptide-containing NcAAs, such as several antibiotics. Norine, an online database, contains updated and important information on different NcAAs present in non-ribosomal peptides (Flissi et al., 2020). Besides the direct commercial applications of NcAAs or derived NcAAcompounds, they have also been used into structure-activity relationships (SAR) peptides (Blaskovich, 2016); incorporation of NcAAs into protein and peptides using ribosomes is an increasing field with many possibilities (Martin et al., 2018). In vivo incorporation of NcAAs into antimicrobial peptides is also an enlarging field of study due to the potential discovery of novel antibiotics to broaden the human pharmacological barriers toward microbes. Genetic engineering methodologies allow gene libraries of 10 1 -10 8 variants and thus, random incorporation of NcAAs in their sequence; sampling can thus be carried out by high throughput approaches, greatly shortening the time needed for new antimicrobial development (Baumann et al., 2017). Among them, the possibility to engineer proteins to incorporate site-specifically different NcAAs has had a major impact the development of this field (Agostini et al., 2017;Zhao et al., 2020). This technology has removed the constraints imposed by nature on the use of the 22-genetically encoded AAs, allowing to insert new properties and/or improved functionalities on protein scaffolds. Some examples are improvements in enzyme catalytic efficiency, chemical and/or thermal stability, substrate scope, enantio-and stereoselectivity or inhibition. Also, bioorthogonal functionalization, spectroscopic probes, photocross-linking, metal-chelating or post-translational modificationmimicking has been reported (Agostini et al., 2017;Almhjell and Mills, 2018;Rezhdo et al., 2019;Zhao et al., 2020). Thus, the huge advances on the standardization of methodologies for the production of enzymes incorportating NcAAs (Smolskaya and Andreev, 2019;Hammerling et al., 2020) are also boosting the application and economic significance of NcAAs.
Among the numerous L-α-NcAAs with biomedical interest (Table 1), L-homophenylalanine is a precursor for the preparation of valuable angiotensin-converting enzyme (ACE) and renin inhibitors (e.g., enalapril, lisinopril, quinapril, ramipril, trandolapril or benazepril, among others; Ahmad et al., 2009). L-α-aminobutyric acid (L-ABA, homoalanine) is a building block for the synthesis of important drugs such as ethambutol (antituberculosis drug), or levetiracetam and brivaracetam (antiepileptic drugs; Zhu et al., 2011). L-DOPA (levodopa) and several derivatives -including carbidopa-are among the most useful drugs for Parkinson's disease treatment (Min et al., 2015;Gupta et al., 2019). L-5-hydroxytryptophan therapeutic effects include treatment of depression, chronic headache and insomnia (Hara and Kino, 2013). L-and Dbiarylalanine-containing compounds have found different applications (e.g., dipeptidyl peptidase 4-, botulinum toxin-or amyloid-β-peptide aggregation-inhibitors; Ahmed et al., 2015). On the other hand, D-para-hydroxyphenylglycine (D-pHPG) and D-phenylglycine (D-PheGly) are utilized in the semisynthesis of many different antibiotics (Martínez-Rodríguez et al., 2010c and references therein) ( Table 2). Among these antibiotics, amoxicillin, cephalexin and ampicillin are included in the World Health Organization's list of essential medicines. D-Phenylalanine is used in the preparation of nateglinide, a drug for the treatment of type 2 diabetes (Bettini et al., 2020). D-valine is used in the synthesis of tau-fluvalinate, a pyrethroid pesticide . D-amino acid mixtures have been shown to trigger biofilm disassembly in some bacterial species, promoting antibiotic sensitivity (Kolodkin-Gal et al., 2010;Dawe et al., 2017). As highlighted by the Nobel Prize Frances Arnold, important Protein Engineering efforts are also been paid to develop new 'NcAA synthases' for efficient, environmentally friendly production of valuable NcAAs . Nonetheless, many different enzymatic approaches have already been described in the past for NcAA production, which might give us some lessons from the past; these are interesting starting points for improvement, allowing new designs utilizing the immense enzymatic knowledge gained by the development of Green Chemistry.

MULTIENZYMATIC CASCADES FOR THE PRODUCTION OF NcAAs
NcAAs chemical synthesis continues receiving at present huge attention in the literature due to the relevance of these compounds (Xue et al., 2018;Mei et al., 2020;Zou et al., 2020). On the other hand, the search for sustainable processes to decrease the environmental impact of industrial processes (Wenda et al., 2011;Sheldon and Brady, 2018) is probably the main reason boosting the development of multienzymatic cascade (MEC) reactions. The unprecedented development in bioinformatics, metagenomics and de novo design coupled with protein engineering (i.e., directed evolution and highthroughput screening) during the last decade have resulted in a massive diversification on the enzymes available for synthetic Chemistry and Biology (Devine et al., 2018). These advances have accelerated the arrival of the "Fourth Wave of Biocatalysis" (Bornscheuer, 2018) or the so-called "Golden Age of Biocatalysis" (Devine et al., 2018); combinations of enzymes -whether in cascade reactions or via metabolic engineering-brings together many beneficial features which might convert this strategy as the 'first choice' to advance in the biotransformation field (Bornscheuer, 2018). The huge research efforts in this field are reflected by different reviews from the last decade on general operational and functional aspects on different MECs, some of them also including disseminated information on AA production using chemoenzymatic and multienzymatic systems (Lopez-Gallego and Schmidt-Dannert, 2010;Hall and Bommarius, 2011;Ricca et al., 2011;Oroz-Guinea and García-Junceda, 2013;France et al., 2017;Quin et al., 2017;Devine et al., 2018;Schrittwieser et al., 2018;Sperl and Sieber, 2018;Wu and Li, 2018;Cutlan et al., 2019;Hwang and Lee, 2019;Giannakopoulou et al., 2020). The interest on MECs in biotransformation and biomedical engineering is thus clear, being an attractive alternative for the production of biofuels, pharmaceuticals and fine chemicals (Hwang and Lee, 2019). A short summary of different MECs developed for the production of NcAAs during the last decades can be consulted in Table 3.
The plausible combinations of MEC are huge and their operational aspects have been reviewed in detail in the last lustrum. Firstly, MEC can be developed using purified proteins, whole cells or cell-free extracts. The spatial organization of MECs have also received special attention, since it affects important operational parameters such as substrate/product diffusion, the availability of cofactors or the probable interference or inhibition of cofactors among the different enzymes used in the reactions (Quin et al., 2017;Bugada et al., 2018). In this sense, many diverse multienzyme arrangements have been proposed, such as fusion proteins (e.g., including a peptide linker), nucleic acid-based or protein-based scaffolds, co-immobilization, Vesicle-based or protein-based encapsulation, or even repurposed cellular organelles (Bugada et al., 2018;Hwang and Lee, 2019). Also, different recent reviews have described different approaches and materials suitable for MEC immobilization (Zdarta et al., 2018;Hwang and Lee, 2019;Ren et al., 2019;Romero-Fernández and Paradisi, 2019). Since many MECs rely on the regeneration of cofactors for continuous operation, its selection is also important; several regeneration systems have been studied from the middle of 20th century, and different alternatives are available, such as formate dehydrogenase (FDH), glucose dehydrogenase (GDH) or NADH Oxidase (NOX) among many others (De Wildeman et al., 2007;Hall and Bommarius, 2011;Tassano and Hall, 2019). This complex scenario makes difficult to efficiently categorize MECs, but the reader is referred to the seminal work by Kroutil's group to envision different parameters which can be used for Irreversible inhibitor of the enzyme cystathionine γ-lyase Asimakopoulou et al., 2013;Weiser et al., 2015 β-(1-Azulenyl)-L-alanine Spectroscopic probe for investigating protein dynamics and protein-protein interactions Watkins et al., 2020 L-theanine Taste-enhancing properties and probable health benefits, approved as GRAS ingredient by the FDA. Mu et al., 2015Mu et al., , 2019Yang et al., 2020 18    Substituted L-pyridylalanines Found in anticoagulants, dipeptidyl peptidase 1 inhibitors, leukocyte adhesion inhibitors, azaindoline anticancer agents, antidiabetics, or as structures for organocatalysts Ahmed et al., 2016 L-Tryptophan analogs Starting materials for chemical syntheses as well as probes for chemical biology. e.g., 4-Nitro-Trp, biosynthetic and chemical precursor to thaxtomin A, potentially useful agrochemical and a chemical precursor to the tumor-promoter indolactam VA; 4-cyano-tryptophan, fluorophore for imaging studies in vitro and in vivo; β-Alkyl Tryptophan Analogs, frequent components of useful natural products, biochemical probes, and therapeutics Herger et al., 2016;Romney et al., 2017;Boville et al., 2018a,b;Dick et al., 2019 Metal-chelating NcAAs for metalloprotein engineering e.g., (2,2 -bipyridyn-5-yl)alanine and 2-amino-3-(8-hydroxyquinolin-3-yl)propanoic acid Almhjell and Mills, 2018 Further NcAAs can be consulted in Blaskovich (2016) classification of biocatalytic artificial cascades (e.g., number of steps/catalysts, chronology, topology, types of catalysts used; Schrittwieser et al., 2018).

Hydantoinase Process
The "Hydantoinase Process" is a cheap and environment-friendly enzymatic cascade for the potential production of virtually any enantiopure α-AA. This process is known for more than four decades, and received its name from the ability of Dhydantoinases (dihydropyrimidinase, E.C. 3.5.2.2) to hydrolyze a wide spectrum of D,L-5-monosubstituted hydantoins. The latter compound was converted till the corresponding N-carbamoyl-D-α-AA, which could be afterward hydrolyzed chemically to the corresponding D-α-AA. D-hydantoinase was afterward coupled with a stereospecific D-carbamoylase (E.C. 3.5.1.77) to obtain the corresponding D-α-AA starting from a racemic mixture of 5-monosubstituted hydantoins ( Figure 1A), taking advantage of the spontaneous racemization of these substrates under certain conditions (see below; Grifantini et al., 1998;Slomka et al., 2017). Its principal application was the production of D-pHPG and D-PheGly (precursors of Ampicillin and Amoxicillin, Table 2), but it has been applied industrially for the production of different enantiopure AAs by DSM, Evonik, Kanegafuchi or Recordati (Bommarius et al., 1998;Wilms et al., 2001;Wenda et al., 2011). Despite its relevance, DKRs accomplished through this process were initially limited to 5-monosubstituted hydantoins for which a fast spontaneous racemization was favored since racemization of most hydantoins is usually a very slow process; chemical racemization is highly dependent on the pH, temperature and other factors (such as bulkiness) of the substituent in the 5-position of these precursors (Pietzsch and Syldatk, 2002). This enzymatic tandem was enhanced to increase its substrate scope, allowing the production of additional D-α-AAs by inclusion of a Modulator of L-phenylalanine-mediated amyloid formation, proposed as a therapeutic molecule in phenylketonuria. Used to obtain nateglinide for the treatment of type 2 diabetes; also used as analgesic and anti-stress agent Singh et al., 2014;Bettini et al., 2020 D-Val Precursor of different compounds. Fluvalinate (tau-fluvalinate) pyrethroid pesticide; it also forms part of D-penicillamine, actinomycin D, fungisporin and valinomycin (pharmaceutical drugs in clinical. Actinomycin D is employed clinically as chemotherapeutics for the treatment of highly malignant tumors Substituted-D-arylalanines Synthesis of biarylalanines through chemoenzymatic reaction (botulinum toxin inhibitors, amyloid-β-peptide aggregation inhibitors, kinesin-14 motor protein KIFC1 inhibitors or and reverse cholesterol transport facilitators Ahmed et al., 2015 Substituted D-Trp derivatives Building blocks for mitragynine, or inhibitors of breast cancer resistance protein or necrostatins Amino acid ester racemase/esterase system ACLR-homolog with AAER activity + esterase -An illustrative example is shown, but there are other possible combinations. It is advisable that the reader consults the specific example in the corresponding section to determine whether (i) isolated, enantiopure and/or enantioenriched AAs are obtained, and (ii) the MEC is a general method suitable for the production of different NcAAs or specific NcAAs. Additional steps might also be necessary, such as separation of other compounds produced in the reactions, or chemical steps to convert the systems into chemoenzymatic DKR methods. DHYD, D-hydantoinase; HR, hydantoin racemase; NSAR, N-succinyl-amino acid racemase; DCAR, D-carbamoylase; LCAR, L-carbamoylase; L-NxAH, L-N-Substituted amidohydrolase (e.g., L-acylase, LCAR or L-succinylase); D-NxAH, D-N-Substituted amidohydrolase (e.g., D-acylase, DCAR, or D-succinylase); L-AMID, L-amidase (e.g., L-proline amidase or L-amidase); ACLR, α-amino -caprolactam racemase; D-AMID, D-amidase (e.g., D-aminopeptidase); DAAO, D-amino acid oxidase; L-TA, L-enantioselective transaminase; LAADH, L-amino acid dehydrogenase (e.g., LeuDH, GluDH); CRS, cofactor recycling system; LAAO, L-amino acid oxidase; TrpS, tryptophan synthase; DAADH, D-amino acid dehydrogenase; PAL, phenylalanine ammonia lyase; LAAD, L-amino acid deaminase; D-TA, Denantioselective transaminase; L-PheDH, L-phenylalanine dehydrogenase; TD, toluene dioxygenase; TGDH, toluene cis-glycerol de-hydrogenase; TPL, tyrosine phenol lyase; AR, alanine racemase; AAER, amino acid ester racemase.
This D-hydantoinase/L-carbamoylase system was further expanded by coupling with an N-succinyl-amino acid racemase (NSAR, E.C.3.1.5.87), since the latter enzyme allows the racemization of the D-N-carbamoyl-α-AA produced by hydantoinase. Thus, in situ conversion producing L-Ncarbamoyl-α-AA in the reaction medium occurs, which can be further hydrolyzed by L-carbamoylase to the corresponding L-α-AA ( Figure 1D). However, this system still depends on the L-residual enantioselective activity of D-hydantoinase for those substrates whose chemical racemization is not favored. Different reports on the hydantoinase/NSAR/L-carbamoylase system have been reported Lo et al., 2009). Engineered dihydropyrimidinase from Brevibacillus agri (L159V mutant), Bacillus kaustophilus L-carbamoylase and Deinococcus radiodurans NSAR were purified, and allowed the production of L-homophenylalanine (90% yield, 5 h, Lo et al., 2009). This L-enantiospecific MEC system has further been expanded by inclusion of a hydantoin racemase, speeding up the process by racemization of the remaining L-5-monosubstituted hydantoin for substrates whose chemical racemization is not favored (Rodríguez-Alonso et al., 2015, 2016, 2017; Figure 1D). Using purified and immobilized enzymes, the proposed "doubleracemase Hydantoinase Process" was efficiently applied for the synthesis of different NcAAs (L-norvaline, L-norleucine, L-ABA, L-homophenylalanine). The immobilized system could be reused 15 times, retaining 80% of the initial activity (Rodríguez-Alonso et al., 2017).
The Amidohydrolase Process is an example of the expansion possibilities of well-stablished MECs (Martínez-Rodríguez et al., 2020). As way of example, oxyfunctionalized AAs [L-methionine-(S)-sulfoxide and different γ-hydroxy-AAs] were obtained by using a NSAR/L-acylase tandem coupled with the stereoselective isoleucine dioxygenase from Bacillus thuringiensis (Supplementary Figure S1, Enoki et al., 2016). L-methionine-(S)-sulfoxide with 97% yield and 95% d.e. was produced starting from racemic N-acetyl-Met. Hydroxylation of AAs is one of the numerous theoretical MEC expansion possibilities for the production of oxo-functionalized AAs Smirnov et al., 2012;Busto et al., 2014;Peters and Buller, 2019).
Coupling of D-aminopeptidase from Ochrobactrum anthropi and ACLR from Achromobacter obae allowed the production of D-Ala, D-ABA, D-Ser, and D-Met (Figure 3B, Asano and Yamaguchi, 2005b;Yamaguchi et al., 2007); As way of example, 45 mM L-alanine amide was converted to D-alanine (7 h, yield > 99.7%; Asano and Yamaguchi, 2005b). Coupling of purified ACLR from Achromobacter obae mutant (L19V/L78T) obtained by directed evolution, together with a thermostable mutant of D-amino acid amidase from Ochrobactrum anthropi SV3 allowed for the biosynthesis of different enantio-enriched D-PheGly and D-Phe derivatives ( Figure 3C, Yasukawa and Asano, 2012). On the other hand, the optical purity of this system varied greatly for aromatic substrates, with e.e. values ranging 16-99%. These results differed to the results presented previously on aliphatic substrates ( Figure 3C, Yamaguchi et al., 2007), suggesting enantioselectivity of D-amino acid amidase, as it was observed with D-succinylase ( Figure 2E). Production of enantiopure L-Ala, L-Leu, and L-Met (100% yield, >99% e.e.) was reported by coupling ACLR from Achromobacter obae with L-Amino acid amide hydrolase from Pseudomonas azotoformans (Yamaguchi et al., 2007). This strategy was afterward proved useful for the production of different L-α-AAs. By overexpression of mutated ACLR together with L-amino acid amidase from Brevundimonas diminuta in E. coli, efficient production of various enantioenriched (S)-phenylalanine derivatives was achieved ( Figure 3D, 99% yield, 90-98% e.e.). This system allowed production of L-homophenylalanine, which continuously precipitated in the reaction mixture (>99% yield, 98% e.e., 12 h; Yasukawa and Asano, 2012). On the other hand, lower e.e. values were also obtained for other aromatic compounds.

Amino Acid Oxidase-Based MECs
Amino acid oxidases (AAOs) are important biotechnological flavoenzymes catalyzing the oxygen-dependent oxidative deamination of D-or L-α-AAs, resulting in α-keto acids, ammonia and hydrogen peroxide (through an imino acid intermediate which can decompose spontaneously to the corresponding α-keto acid and NH 3 , Figure 4A; Pollegioni et al., 2008Pollegioni et al., , 2013Asano and Yasukawa, 2019). These enzymes have a wide variety of biomedical and biotechnological applications, including the production of different α-keto acids, important intermediate building blocks. Both LAAOs (EC 1.4.3.2, Pollegioni et al., 2013) and DAAOs (EC 1.4.3.3, Pollegioni and Molla, 2011) have been described. Inclusion of a catalase in AAO-based biotransformations is a common general strategy to avoid the toxicity of the H 2 O 2 produced during the recycling of the FAD coenzyme necessary for AOO activity; thus, this AAO/catalase basic scaffold can be directly used for the KR of AA racemates, producing a mixture composed of 50% of enantiopure AA and 50% of the corresponding α-keto acid ( Figure 4A; Pollegioni et al., 2008). Turner's group proposed a preparative chemoenzymatic method for deracemization of NcAAs by inclusion of a non-selective chemical reductant, transforming back the intermediate imino acid produced by D-and L-AAOs till a racemic mixture of the original AA, thus allowing 100% conversion of the initial AA racemate (Figure 4A; Alexandre et al., 2002;Beard and Turner, 2002).
Frontiers in Bioengineering and Biotechnology | www.frontiersin.org 13 August 2020 | Volume 8 | Article 887 FIGURE 4 | Continued from porcine heart, catalase from bovine liver and β-tyrosinase (β-Tyr, tyrosine phenol lyase) purified from Citrobacter intermedius were successfully applied. The radio-labeled atom is highlighted by an asterisk (Bjurling et al., 1990). (C) Enzymatic synthesis of 1-14 C-or 3-14 C-tryptophan (R = H) or 1-14 C-or 3-14 C 5-hydroxytryptophan (R = OH). It was conducted with a MEC consisting of DAAO, Glutamic-pyruvic transaminase (GPT), catalase and tryptophanase (Tpase, L-tryptophan indole-lyase). An asterisk highlights the relative positions of the radio-labeled atoms (Pająk et al., 2018). (D) Synthesis of NcAAs starting from primary amines; a MEC comprising mutant porcine kidney DAOO and nitrilase AY487533 (NHase) was used (Kawahara et al., 2017). (E) Two-step chemoenzymatic syntheses of halogenated D-Trp derivatives. Combination of biocatalytic halogenation by tryptophan halogenase (TrpH) or Trp synthase (TrpS) with L-Trp oxidase (LTrpO), followed by non-selective reduction allowed to obtain D-Trp derivatives (Schnepel et al., 2019). Bjurling et al., 1990). A similar MEC was applied for the synthesis of 14 C-L-Trp and 5-OH-Trp starting from racemic 1-14 Cor 3-14 C-alanine, replacing β-tyrosinase by tryptophanase (EC 4.1.99.1; Figure 4C, Pająk et al., 2018). Bienzymatic conversion of primary amines to NcAAs has been proposed using an engineered DAAO from porcine kidney and an R-enantioselective nitrilase (GenBank Acc. No AY487533, uncultured organism; Figure 4D). Moderate to low yields (62-73%) and e.e. (40%) were achieved with this methodology (Kawahara et al., 2017). Despite the numerous KR applications proposed for LAAOs, their recombinant production in heterologous hosts has proven difficult (Pollegioni et al., 2013). Recent studies have greatly increased the knowledge on these enzymes (e.g., L-Trp-, L-Lys-, Gly-or L-Arg oxidases) which might boost the applications of LAAOs in the production of D-AA derivatives (Francis et al., 2017;Asano and Yasukawa, 2019). An L-Trp oxidase from L. aerocolonigenes has been successfully applied using two alternative chemoenzymatic MECs for the synthesis of different halogenated D-Trp derivatives ( Figure 4E; Schnepel et al., 2019). In a first step, Trypthophan halogenase or Tryptophan synthase (TrpS), are used for the production of halogenated L-Trp derivatives. The L-enantiomer can then be stereo-destroyed by LAAO till the corresponding α-imino acid. Following the general use of non-selective chemical reductant applied for AAOs (Figure 4A), the imine intermediate can be transformed back till a racemic mixture of L-and D-α-AA. Whereas the D-α-AA will not be recognized by LAAO, the regenerated L-isomer enters back to the reaction, converting the system in a chemoenzymatic DKR ( Figure 4E). Conversions achieved approximately 90%, with e.e. >92% (Schnepel et al., 2019). This MEC shows an interesting approach, since CLEAs obtained by cross-linking of precipitated tryptophan halogenase, flavin reductase and alcohol dehydrogenase were used, converting this crystalline precipitate into a multifunctional and recyclable MEC for the production of halogenated L-Trp derivatives in the gram scale (Frese and Sewald, 2015). Since Trypthophan halogenase/cofactor recycling CLEAs also recognized L-Trp, coupling of this system with an LAOO might use cheaper racemic mixtures of D,L-Trp as starting point.
Besides the above applications, AAOs are also pivotal enzymes in other MECs, and in fact, their use has been mainly linked to those applications; coupling of AAOs to amino acid dehydrogenases, ammonia-lyases or aminotransferases avoids the necessity of chemical transformation or further processing of the "undesired" α-keto acid produced by D-or L-AAO, allowing total conversion of the initial AA racemate till the corresponding D-or L-α-AA (see sections "Amino Acid Dehydrogenase-Based MECs, " "Ammonia Lyase-Based MECs, " and "Transaminase-Based MECs").
L-and D-ABA have received huge attention in the literature, and has been produced in many different reports using ThrD. As way of example, ThrD was coupled with LeuDH and GDH or FDH cofactor-recycling systems (Tao et al., 2014). A scale-up of the process (30 L of reaction in a 50-L fermenter) allowed the production of 29.2 mol L-ABA (97.3% theoretical yield), with a productivity of 6.9 g·L −1 ·h −1 (Tao et al., 2014). This MEC module has also been used to generate a heterologous biosynthetic pathway leading to the production of L-ABA in Saccharomyces cerevisiae (Weber et al., 2017). Metabolic engineering allowed the expansion of the latter system for the production of S-2-aminobutanol. D-ABA was produced by a tri-enzymatic cascade with cell-free extract or purified enzymes, composed of ThrD, D-amino acid dehydrogenase and FDH, starting from L-Thr (>95% yield and >99% e.e.; Chen et al., 2017). More than 15 different enantioenriched and enantiopure D-α-AAS -including several Dphenylalanine derivatives-have been produced coupling a "real" L-AAD to engineered enantioselective meso-diaminopimelate dehydrogenase and a cofactor-recycling system (conversions from 45.3 to >99%, e.e. values ranging 52.1 to >99%; Figure 6B, Parmeggiani et al., 2016;Zhang et al., 2019). Racemic mixtures might be used as well as pure L-AAs, turning it into a more efficient and cheaper system. A scarcely described strategy utilized immobilized PLP-dependent L-methioninase (L-methionine γ-lyase, EC 4.4.1.11) together with GluDH on polyacrylamide and chitosan (with no regeneration system) for the production of L-ABA, starting from L-methionine (Figure 6C, El-Sayed et al., 2015).
An alternative strategy allowing in situ production of α-keto acids for further conversion by AADHs consist in the oxidation of α-hydroxy acids. This is the case of the MEC combining mandelate racemase, mandelate dehydrogenase and L-AADHs, allowing the production of L-phenyglycine from mandelic acid; this system further uses an elegant internal cofactor recycling system (Resch et al., 2010). This 3-step one-pot reaction has been efficiently applied in a whole cell system, reaching a production of 79.70 g·L −1 ·d −1 (Tang et al., 2020). It was also used for the production of different L-PheGly derivatives starting from different mandelic acid derivatives; conversions ranging 49-97% were achieved, with e.e., >97% (Figure 6D, Fan et al., 2015). Further expansion of this system allows in situ production of an α-hydroxy acid starting from fatty acids. This strategy was shown for the production of L-norleucine starting from hexanoic acid. This MEC consisted in a combination of a P450 peroxygenase, two stereocomplementary L-and D-hydroxyisocaproate dehydrogenases along with LeuDH; no extra recycling system was needed, since this MEC also provide internal cofactor regeneration ( Figure 6E). L-norleucine conversion was lower than 35% with up to 5 mM substrate concentrations; e.e. > 97%; L-ABA production was also reported (Dennig et al., 2018).

Ammonia Lyase-Based MECs
Ammonia-lyases (ALs, EC 4.3.1.X, defined as carbon-nitrogen lyases that release ammonia as one of the products) comprise a heterogenous enzymatic group catalyzing the reversible cleavage of C-N bonds, typically of α-AAs, producing an unsaturated (or cyclic) derivative and ammonia. More than 30 different EC subclasses of ALs are reported, showing remarkable structural, functional and mechanistic differences, which can be broadly grouped into seven main classes Viola, 2020). Besides of the interest on ALs for the production of β-AAs and other APIs (Sariaslani, 2007;Turner, 2011;Xue et al., 2018), applications for the synthesis of NcAAS have also been reported, being aromatic amino acid ALs [phenylalanine AL (PAL), histidine AL (HAL) and tyrosine AL (TAL)] the most relevant and studied enzymes, showing a marked L-enantioselectivity (Turner, 2011;Parmeggiani et al., 2015;Zhu et al., 2019). Although as an isolated case, the acyclic amino acid propargylglycine was recognized by a PAL enzyme (Weiser et al., 2015).
As for other enzymes, their reversible catalytic properties together with their enantioselective character allows the use of isolated ALs both for (i) KR processes for deracemization of AAs (obtaining a mixture of D-α-AAs and α,β-unsaturated acid, Figure 7A, Poppe and Retey, 2003;Turner, 2011;Tork et al., 2019) or (ii) asymmetric synthesis of L-α-AAs (Figure 7B), using high concentration of ammonia to shift PAL equilibrium toward the amination reaction starting from achiral α,β-unsaturated acids (Poppe et al., 2012;Tork et al., 2019). Other outstanding feature is the non-necessity of expensive cofactors or recycling systems, thus providing a cost-effective and easier application. It is important to highlight that aminomutase-like activity has been detected in PALs ; whereas this fact might be a drawback for general application in the synthesis of enantio-enriched α-AAs, it might also open up new biotechnological properties of PALs.
from parsley and ammonia ( Figure 7C). The starting materials were arylaldehydes, which were synthesized in situ from the corresponding α,β-unsaturated acids by the Wittig reaction (Paizs et al., 2006). Coupling of PAL together with LAAD allowed the conversion of cinnamic acid derivatives into the corresponding enantio-enriched D-α-AA by a chemoenzymatic process ( Figure 7D) Zhu et al., 2019). Although PAL is reported to be mainly L-enantioselective, PAL-catalyzed amination of the cinnamic acids were reported to lead to the formation of significant levels of the D-enantiomer 4 (in particular, for cinnamic acids with an electron-deficient aromatic ring; Parmeggiani et al., 2015). Taking advantage of the unexpected production of both enantiomers over the time, coupling of this system with a DAAO, together with the use of a non-selective reductant, produced different L-α-AAs starting from α,β-unsaturated acids (conversions 66-82%; e.e. values > 99%; Parmeggiani et al., 2015; Figure 7D). Labeling of L-Tyr with carbon and hydrogen isotopes was achieved by coupling of PAL and L-phenylalanine 4 aminooxygenase (Pająk et al., 2018); this strategy might be expanded for the production of L-DOPA using p-hydroxyphenylacetate 3-hydroxylase (Min et al., 2015), including tetrahydropterin-and NADH-recycling systems (Hara and Kino, 2013) (Supplementary Figure S4).

Transaminase-Based MECs
Transaminases (TAs, EC 2.6.1.X, also known as aminotransferases) are a heterogeneous group of enzymes catalyzing the transfer of an amino group between two different molecules. In this kind of enzymes, different enantioselective PLP-dependent TAs catalyze the (reversible) transfer of an amino group between an amino donor and an amino acceptor (in general, a carbonyl group such as α-ketocarboxylic acids or ketones), yielding chiral amines with a new stereocenter (e.g., AAs, Figure 8A, Guo and Berglund, 2017). During TA catalysis, PLP-recycling is accomplished, which is an advantage in enzymatic synthesis since no additional cofactor-recycling system is needed (Figure 8A). TAs are well documented enzymes for the synthesis of NcAAs (Meiwes et al., 1997;Taylor et al., 1998;Li et al., 2002;Rozzell and Bommarius, 2002), but their 4 Since (i) reduction of the e.e. value of the produced amino acid was observed over time, (ii) the reaction conditions used pHs above 9, and (iii) chemical racemization of AAs is known to occur at high pHs, the formation of the opposite enantiomer might be the result of chemical racemization. biotechnological interest is far beyond the production of these compounds, since they allow the production of other important molecules [e.g., β-AAs (Rudat et al., 2012) or amines (Höhne and Bornscheuer, 2009)]. TAs continue receiving huge attention, and have been reviewed extensively during the last decade due to their huge biotechnological interest (e.g., Mathew and Yun, 2012;Simon et al., 2014;Guo and Berglund, 2017;Slabu et al., 2017;Patil et al., 2018;Xue et al., 2018;Cutlan et al., 2019). Protein engineering strategies have been broadly used to evolve ω-TAs (e.g., Park et al., 2013a,b;Walton et al., 2017), and a database on sequences and structures of biotechnologically relevant engineered ω-TAs is available (Buß et al., 2018).
TAs 5 with potential application in α-NcAA production can be broadly grouped according to the position of the transferred amine group with respect to the carboxylic moiety (when the substrate/reaction product is an AA). α-TAs catalyze the transfer of the amino group at the α-carbon, whereas ω-TAs (also referred to as amine-TAs) transfer the amino group to a carbon further away from the carboxylic group 5 (Guo and Berglund, 2017;Slabu et al., 2017). Both R-and S-enantioselective and promiscuous TAs have been reported (as natural or engineered enzymes, Taylor et al., 1998;Koszelewski et al., 2010); a unique L to D-stereoinverting hydroxyphenylglycine aminotransferase has also been reported (Müller et al., 2006;Walton et al., 2017). Since the enantioselectivity of TAs can be greatly affected by the reaction conditions, enantio-enriched compounds might be produced in some cases (Koszelewski et al., 2010); TAs are also described to suffer substrate and/or product inhibition (Guo and Berglund, 2017), and produce by-products which need to be eliminated or separated; these aspects need to be taken into account when designing TA-based MECs. Since shifting the reaction equilibrium to the desired product of the reaction is required to maximize the productivity of TAs, different strategies have been already proposed to overcome some of TAs potential drawbacks (e.g., distillation, use of biphasic reaction systems, recycling of the carbonyl compounds, degradation/transformation of by-products of the reaction (Koszelewski et al., 2010;Guo and Berglund, 2017;Patil et al., 2018). As way of example, enzymatic methods to remove pyruvate from the reaction medium (if alanine is used as amino donor) can be accomplished with acetolactate synthase, lactate dehydrogenase reduction (together a nicotinamide recycling system such as FDH, GDH, or NOX), pyruvate decarboxylase (or phenyl pyruvate decarboxylase with phenyl alanine as amine FIGURE 7 | AL-based production of α-AAs. (A) General scheme of PAL-mediated deracemization of racemic mixtures of AAs to enantio-enriched or enantiopure D-α-AAs mediated by deamination of L-α-AAs. (B) Asymmetric amination of α,β-unsaturated acids for the production of enantio-enriched or enantiopure L-α-AAs. α,β-unsaturated acid (α,β-UA) (Tork et al., 2019). (C) Production of L-Phe and other L-α-NcAAs using a chemoenzymatic approach. The starting arylaldehydes were transformed into the corresponding α,β-unsaturated acids by the Wittig reaction; further application of porcine liver esterase (PLE), PAL from parsley and ammonia allowed transformation into L-phenyl-, L-4-chlorophenyl-, L-3-fluorophenyl-or L-thiophen-2-yl-alanine (Paizs et al., 2006). (D) Chemoenzymatic synthesis of L-and D-phenylalanine derivatives using PAL/LAAD and PAL/DAOO systems . donor) or alanine dehydrogenase (Koszelewski et al., 2010;Simon et al., 2014).
One of the enzymatic combinations that can be carried out for shifting TA reaction equilibrium accounts on coupling two different TAs in a one-pot two-step procedure (Figures 8B-E). In this general strategy, a "primary" TA converts an α-keto acid to the corresponding α-AA. The "secondary" TA (in general a different α-TA or a ω-TA) transforms the α-keto acid back, replenishing the initial amino donor (Figures 8B,D) Ager et al., 2001;Li et al., 2002;Gefflaut et al., 2012;Park et al., 2013a). The enantiomer obtained in these systems will depend on the enantioselectivity of the TAs used, and thus, both D-or L-α-AAs can be obtained; the use of ω-TAs in these systems presents many advantages, since the secondary reaction can be conducted using amino donors different to AAs, reducing the number of possible interferences among the reactivity of both TAs. Enzymatic synthesis of the herbicide L-phosphinothricin FIGURE 8 | TA-based production of α-AAs. (A) General scheme for reversible α-TA and ω-TA catalysis. When R 2 = COOH, the reaction correspond to an α-TA; ω-TA do not necessarily precise a carboxylic moiety in this position, although they usually also recognize α-AAs as substrates. Both the direct reaction (reduction, dashed line) and the reverse reaction (oxidation, full line) can be used for the production of enantiopure AAs (i.e., starting from keto/keto acids or from amines/AAs). (B) General scheme for production of enantiopure or enantio-enriched α-AAs using two different α-TAs. The selected example (C) consists on the enzymatic synthesis of the herbicide L-phosphinothricin using 4-aminobutyrate:2-ketoglutarate (AK-TA) transaminase from E. coli and glutamate:oxalacetate transaminase (GO-TA) from Bacillus stearothermophilus (24 h reaction, 0.5M substrate concentration). L-Glu is used as the amino donor for AK-TA, and transformed to a-ketoglutarate; the latter compound is transformed back to L-Glu by GO-TA, which uses L-Asp as amino donor and converts it to oxaloacetic acid. Oxaloacetic acid decarboxylate spontaneously to pyruvate, driving the AK-TA reaction till the product of the reaction (Bartsch et al., 1996;Ricca et al., 2011). (D) General scheme for the production of enantiopure α-AAs using α-TA/ω-TA combination. The selected example (E) consists on the enzymatic synthesis of different L-α-AAs using S-selective branched-chain transaminase (BCTA) from E. coli and ω-TA from Ochrobactrum anthropi (OATA) (Park et al., 2013b). was achieved by using 4-aminobutyrate:2-ketoglutarate transaminase from E. coli and glutamate:oxalacetate transaminase from Bacillus stearothermophilus (Figure 8C, 24 h reaction, 0.5M substrate concentration; Bartsch et al., 1996;Ricca et al., 2011). Different L-α-AAs were obtained by using S-selective branched-chain transaminase from E. coli and ω-TA from Ochrobactrum anthropi (Figure 8E, Park et al., 2013a).
As mentioned above, shifting of α-TAs reaction equilibrium can be achieved by coupling with enzymes allowing the degradation/transformation of the by-products of degradation (e.g., pyruvate if alanine is used as amino donor). The use of acetolactate synthase (AceS) was one of the first proposed strategies to remove the by-products of the reaction, by condensation of two pyruvate molecules to acetolactate, which spontaneous decarboxylates to acetoin ( Figure 9C) . α-TA/AceS system also allowed the production of NcAAs such as L-tert-Leu or L-and D-ABA or D-Glu Taylor et al., 1998). Since α-keto acids can be the substrates of TAs (Figures 8, 9), enzymatic conversion of different compounds till the corresponding α-keto acids constitute a general strategy to expand TA-based MECs; AAOs, AADH or deaminases (sections "Amino Acid Oxidase-Based MECs" and "Amino Acid Dehydrogenase-Based MECs") are thus candidates to be coupled with TAs for utilization of AAs as initial substrates. Inclusion of L-amino acid deaminases 6 into the α-TA/AceS system allows the synthesis of different AAs, depending on the enantioselectivity of the enzymes. As way of example, L-ABA was obtained from L-Thr; whole-cell systems containing aromatic TA, ThrD and AceS were used efficiently for the production of L-ABA, although L-Ala was obtained as a by-product of the reaction (pyruvate is formed by spontaneous decomposition of oxaloacetate during the reaction, and it can be aminated back to L-Ala if not removed from the reaction) ( Figure 9D; Fotheringham et al., 1999;Ager et al., 2001). In an analogous approach, NSC Technologies (Monsanto) produced different D-α-AAs using D-aminotransferase from Bacillus sp. and AceS, with a capacity in the multiton·a −1 scale (Liese et al., 2006).
Removal of L-Ala by-product formed by TA as a result of secondary amination of pyruvate formed by decomposition of OAA (Figure 9D) was possible by coupling the ThrD/TyrTA/AceS MEC with DAOO and alanine racemase 6 Inclusion of a LAAO yields an analogous MEC from the theoretical point of view. Figure S2). This expanded system also allowed the production of L-ABA, lowering the concentrations of L-alanine by-product in the reaction (Zhu et al., 2011). This five enzyme-MEC allowed a large scale-preparation of L-ABA by combining in a whole cell system (reaction volume 1500 L in a 2000-L jar fermenter), resulting in an L-ABA concentration of 25.38 g·L −1 (246 mM) at the second stage of the reaction (Zhu et al., 2011).

(Supplementary
Most of the problems arisen from these initial TA systems (such as secondary amination of the α-keto acids produced during the reaction, e.g., pyruvate) can be generally overcome by substitution of α-TA by ω-TA. Simplification of previous multienzymatic systems utilized for the synthesis of ABA (Figures 9C,D and Supplementary Figure S2) was achieved by using a ThrD together a ω-TA. With this configuration, both Land D-ABA were synthesized by using S-or R-ω-TA, respectively. L-Thr (0.3 M) was transformed in less than 24 h, with conversions over 99% and e.e. > 99% (Figure 9E, Park et al., 2010Park et al., , 2013b. A similar system allowed for concomitant production of D-Thr and D-or L-ABA starting from racemic Thr (>99% e.e.; Han , and Shin, 2015). An alternative approach for the production of D-or L-ABA consisted in coupling an L-methioninase with a D-TA from Bacillus sp or an L-TA from E. coli, using methionine as substrate (Silva et al., 2019).
Recently, a "real" promiscuous LAAD from Proteus mirabilis together a engineered D-TA from Bacillus sp. YM-1 has been applied for the synthesis of different D-Phe derivatives starting from L-or D,L-α-AAs using whole cells (Figure 10A, Walton et al., 2017). Besides these derivatives, the authors reported that the D-TA variants shown in this work also displayed increased activity for D-α-AAs with aliphatic or polar side chains. Furthermore, they also suggested that the disadvantage of using D-Glu as amino donor could be overcome by generating the donor substrate in situ with Glu-or Asp-racemase and by replacing LAAD from Proteus mirabilis with that from Proteus myxofaciens, shown to not deaminate L-Glu or L-Asp (Walton et al., 2017). Amino acid racemases have been included for in situ regeneration/production of the amino donor of the TA reaction. As way of example, D-TA coupled to GluDH-FDH-glutamate racemase recycling system allowed the production of D-Phe or D-Tyr (48 and 60 g·L −1 , respectively. e.e. > 99%, 35 h) from the corresponding α-keto acids ( Figure 10B, Bae et al., 1999Bae et al., , 2002. If amino acid racemases are included in these MECs, it is advisable to know whether the AR can recognize the product of the reaction. AAOs have also been coupled with TAs to obtain different NcAAs. DAAO/catalase, together with an L-aspartate amino transferase (α-TA) were used to produce L-2-naphthyl alanine with a 98% yield and 99.5% e.e. (Figure 10C, Caligiuri et al., 2006). The use of L-CSA as amino donor allows shifting of the reaction by spontaneous degradation of the thio-α-keto acid to pyruvate. A NADH-dependent lactate dehydrogenase was added in this system to transform the produced pyruvate, allowing to follow the course of the reaction spectrophotometrically; on the other hand, addition of this enzyme could also become a general idea to remove the pyruvate formed, avoiding putative transformation to L-Ala by α-TA (see above). An alternative Selected examples for the production of L-α-AAs (R 1 ) and D-α-AAS (R 2 ) produced by D-amino acid transaminase from Bacillus sphaericus (D-α-TA) and S-selective-ω-TA Ochrobactrum anthropi or Paracoccus denitrificans (S-ω-TA) (Park and Shin, 2014). (B) Coupling of enantiocomplementary L-α-TA/R-ω-TA for deracemization of AA mixtures, yielding two different enantiopure/enantioenriched α-AAs of opposite chirality. Selected examples for the production of L-α-AAs (R 1 ) and D-Glu (R 2 ) by branched-chain TA from Escherichia coli (L-α-TA) and an engineered variant of (R)-selective ω-TA from Arthrobacter sp (R-ω-TA) (Park and Shin, 2014). (C) General scheme for the preparation of different D-or L-α-NcAAs using D-TAs or L-TAs coupled to acetolactate synthase (AceS). D-or L-Asp are used as amino donors, producing oxaloacetate (OAA) as by-product of the reaction. OAA decarboxylate spontaneously to pyruvate, which is converted to acetolactate by AceS, followed by spontaneous decarboxylation to acetoin (AceS reaction sequence inside a dashed box). AceS thus allows shifting the reaction toward the acetoin, and also prevents pyruvate from being aminated to L-alanine . (D) Preparation of L-ABA in whole cell E. coli system starting from L-Thr. The system comprised threonine deaminase from E. coli (ThrD), aromatic TA from E. coli (TyrTA), and AceS. Starting from 0.5 M of L-Thr and L-Asp, concentrations of the products of the reaction after 24 h were 27.71 mg·mL -1 (L-ABA), 1.23 mg·mL -1 (L-Ala) and 1.01 mg·mL -1 (L-Asp) (Fotheringham et al., 1999). (E) General scheme for the biosynthesis of L-ABA (continuous line) or D-ABA (dashed line) using TD*/ω-TA systems (Park et al., 2013a).

FIGURE 10 | Continued
After 4 h of reaction e.e. values ranging 90 to >99% were achieved (Walton et al., 2017). (B) Production of D-Phe and D-Tyr using D-TA from Bacillus sp. YM-1 (D-TA) coupled to in situ regeneration/production of the amino donor of the reaction by [GluDH/FDH/Glutamate Racemase (GluR)] (Bae et al., 1999). (C) Multi-step enzyme catalyzed deracemization of D,L-2-naphthylalanine (2-NA) using DAAO/catalase/TA MEC (98% yield and 99.5% e.e.). *Lactate dehydrogenase was used to monitor the reaction course, although its inclusion in the system might help to avoid pyruvate transforming till L-Ala (Caligiuri et al., 2006). (D) Conversion of racemic AAs into D-α-AAs using AlaDH, NOX and engineered ω-TA (mut D-ω-TA) MEC (Han and Shin, 2018). (E) Biosynthetic pathway for D-PheGly production. Whole cells of E. coli were used, overexpressing hydroxymandelate synthase (HmaS) and hydroxymandelate oxidase (Hmo) (both from A. orientalis or S. coelicolor) and D-(4-hydroxy)-phenylglycine aminotransferase from Pseudomonas putida (HpgAT) (Müller et al., 2006). strategy was proposed afterward by inclusion of a LAAO together a D-TA, using D-aspartate as the amino donor, which can be generated with an aspartate racemase (Tessaro et al., 2008). L-ABA was produced using an engineered DAAO/catalase module together with ω-TA (Seo et al., 2012). Together with the ω-TA from Vibrio fluvialis JS17, a fusion protein was created with DAAO from Rhodotorula gracilis and Vitreoscilla hemoglobin, since it was earlier reported to significantly enhance the DAAO activity in the production of D-fluoroalanine. Using whole cells together with a biphasic system (to partly overcome ω-TA inhibition by benzaldehyde), 500 mM racemic ABA were transformed to 485 mM L-homoalanine (>99% e.e.; Seo et al., 2012). Racemic glufosinate was also converted to the Lenantiomer by application of a DAAO and different TAs (Green and Gradley, 2017). When coupled to TAs, AADHs also can provide a deracemization process starting from AA racemates. An L-AlaDH/NOX system, together with an engineered D-ω-TA from Arthrobacter sp. was shown as an effective MEC for deracemization of AA racemates ( Figure 10D). Synthesis of D-alanine with a 95% conversion yield (starting form 100 mM solutions) and >99% e.e. was achieved after 24 h (Han and Shin, 2018). Coupled D-TA and L-PheDH from Lysinibacillus sphaericus (together with a recycling system using ethanol and alcohol dehydrogenase) were applied for the production of different para-halogenated derivatives (Br-, Cl-, and F-) of Phe, as well as Tyr, via stereo-inversion of the D-enantiomers to the L-isomer; e.e. > 99% (Khorsand et al., 2017).
Recently, coupling of different enantioselective SAMdependent α-keto acid methyltransferases (MT) with a halide methyltransferase (HMT) and a ωor L-α-TA has been successfully applied for the synthesis of several enantio-enriched D-and L-β-methyl-α-AAs (Supplementary Figure S3, Liao and Seebeck, 2020). An S-adenosylhomocysteine nucleosidasedeficient E. coli strain was necessary for this approach, using CH 3 I as the alkylating agent. Conversions ranging 39 to >95% were achieved for different substrates. Methylation resulted in a high stereoselectivity, with 3R:3S methyl ratios ranging 92:8 to 99:1.
Natural pathway engineering has allowed the production of PheGly derivatives. As way of example, D-PheGly was produced in recombinant E. coli cells using phenylpyruvate as substrate, by a three-step route composed of hydroxymandelate synthase (HmaS), hydroxymandelate oxidase (HmO) and a stereoinverting hydroxyphenylglycine TA (HpgAT) (Figure 10E, Müller et al., 2006). The same strategy was used to produce L-PheGly, but using L-4-hydroxyphenylglycine transaminases from A. orientalis and S. coelicolor (Liu et al., 2014). Another synthetic biology-derived D-Phg operon has been recently developed on the basis of the natural lpg operon from S. pristinaespiralis; substitution of the natural PglE TA by the stereoinverting HpgAT allowed the creation of different plasmids which were utilized for the synthesis of D-PheGly in different engineered actinomycetal expression strains (Figure 11A, Moosmann et al., 2020). Wu et al. reported different complementary enzymatic modules for one-pot conversion of styrenes to the corresponding (S)-α-hydroxy acids, (S)-amino alcohols, and (S)-α-amino acids in high yields and e.e. (>98%, Wu et al., 2016). Combination of 3 out of the 4 modules proposed in E. coli (Module 1: epoxidase and epoxide hydrolase; Module 2: alcohol dehydrogenase and aldehyde dehydrogenase. Module 4, hydroxy acid oxidase, Lenantioselective α-TA, catalase and glutamate dehydrogenase) allows for the production of different L-PheGly derivatives ( Figure 11B). Besides the production of these compounds, this work represent a valid proof of concept of elegantly tackled multimodular whole cell factories, which is a general idea to construct different MECs (Farnberger et al., 2017;France et al., 2017). The previous modular strategy ( Figure 11B, Wu et al., 2016) was further expanded for the enantioselective synthesis of D-PheGly derivatives starting from racemic mandelic acids, styrenes, or L-Phenylalanine derivatives. Firstly, in an analogous MEC to that presented previously ( Figure 6D, Resch et al., 2010), mandelate racemase (MR), (S)-mandelate dehydrogenase (SMDH), D-PheGly aminotransferase (DpgAT) and GluDH built a MEC module to produce different D-PheGly derivatives, starting from different mandelic acids ( Figure 11C); this was possible thanks to the unique stereoinverting L-to D-activity of DpgAT . Conversions ranging 58-94% were achieved, with e.e. values ranging 93-98%. This module was adapted by removal of the MR and joined to two of those presented previously (modules 1 and 2, Figure 11B), in order to obtain D-PheGly derivatives starting from styrenes ( Figure 11D). Finally, this MEC was further extended by inclusion of another module converting L-phenylalanine into styrene, by inclusion of PAL and phenylacrylic acid decarboxylase (PAD), turning into a MEC using nine different enzymes ( Figure 11E, Zhou et al., 2017). D-PheGly was produced with this 8-step cascade, demonstrating they high efficiency of a long non-natural enzyme cascade/pathway. Different D-Trp derivatives have been produced from substituted indoles coupling a tryptophan synthase from Salmonella enterica, LAAD from Proteus species, together with and an engineered D-alanine TA, with e.e. values ranging from 94 to >99% (Figure 11F, Parmeggiani et al., 2019b). One  (Moosmann et al., 2020). (B) Conversion of styrenes to L-PheGly derivatives with E. coli multimodular systems. SMO, styrene monooxygenase from Pseudomonas sp. VLB120; SpEH, epoxide hydrolase from Sphingomonas sp. HXN-200; AlkJ, alcohol dehydrogenase from P. putida GPo1; ALDH, phenylacetaldehyde dehydrogenase from E. coli; HMO, hydroxymandelate oxidase from S. coelicolor A3(2); CAT, catalase from E. coli; αTA, branch chain amino acid transaminase from E. coli. GluDH, glutamate dehydrogenase from E. coli (Wu et al., 2016). (C) Production of different D-PheGly-derivatives starting from α-hydroxy-acids. MR, mandelate racemase; SMDH, (S)-mandelate dehydrogenase; DpgTA, D-PheGly aminotransferase; GluDH, glutamate dehydrogenase . (D) Production of different D-PheGly-derivatives starting from styrenes. SMO, styrene monooxygenase; SpEH, epoxide hydrolase; AlkJ, alcohol dehydrogenase; ALDH, phenylacetaldehyde dehydrogenase; SMDH, (S)-mandelate dehydrogenase; DpgTA, D-PheGly aminotransferase. (E) Production of D-PheGly starting from L-Phe. Expansion of the previous system starting from styrenes was used, by addition of two additional enzymes. PAL: phenylalanine ammonia lyase; PAD, phenylacrylic acid decarboxylase . (F) One-Pot biocatalytic synthesis of D-Tryptophan derivatives from indoles coupling tryptophan synthase (TrpS), LAAD and engineered D-alanine TA (engDATA; Parmeggiani et al., 2019b). example of expansion of TA-based MECs toward the production of functionalized AA derivatives is provided by combination of an acetaldehyde-dependent aldolase (hydroxy-3-methyl-2keto-pentanoate aldolase, HPAL) from Arthrobacter simplex together a branched-chain TA from Bacillus subtilis (BCAT). Aldol condensation of α-ketobutyrate and acetaldehyde followed by transamination allowed enzymatic synthesis of 4-hydroxy-Lisoleucine (4HIL) (Smirnov et al., 2007). Whereas the final yield of 4HIL was not high, the authors proposed that this problem might be overcome by shifting the equilibrium of the reaction by separation of the reaction products and recirculation of the unreacted substrates.

Other MECs for the Synthesis of NcAAs
Other multienzymatic or chemoenzymatic strategies have been proposed for the synthesis of NcAAs, although they have not found yet such a biotechnological interest as those reported above. On the other hand, it is important to bear in mind that besides the more than 20 different enzymes comprised in the different MECs presented in this paper, many enantioselective or stereospecific "free" enzymes are useful for NcAAs synthesis [e.g., proteases (Nakao et al., 1996;Miyazawa, 1999), 2-oxoglutaratedependent oxygenases (Peters and Buller, 2019), aldolases (Xue et al., 2018;Fesko, 2019) or other different PLP-enzymes (Di Salvo et al., 2020)].

Lipase-Containing MECs
Lipases are important enzymes from the industrial point of view, due to its huge versatility and robustness in the production of different compounds. Despite their outstanding promiscuous properties, one drawback that lipases (and proteases) can show is the lack of a perfect enantioselectivity, but several enantiopure compounds can be prepared using them. A bienzymatic method composed of Lipozyme R (Mucor miehei) and Alcalase R was applied for the production of enantiopure L-tert-leucine (99.5% e.e., Figure 12A) (Turner et al., 1995). A similar system was proposed starting from different 5(4H)-oxazolone derivatives ( Figure 12B); P. cepacia lipase hydrolyzes different oxazolone derivatives yielding optically active N-benzoyl-L-α-AA methyl esters. Subsequent methyl ester hydrolysis by proteases (prozyme 6 and protease N; pH 6.8), might yield N-benzoyl-L-α-AAs of high enantiomeric purity ( Figure 12B; Crich et al., 1993;Miyazawa, 1999).

Tyrosine Phenol Lyase-Containing MECs
Tyrosine phenol lyase (TPL, β-tyrosinase) has been applied on different MECs for the production of interesting pharmaceutical compounds, including the enantioselective synthesis of L-Phe/L-Tyr derivatives (Martínez-Montero et al., 2019). Coupling with DAAO was proposed three decades ago ( Figure 4B, section "Amino Acid Oxidase-Based MECs"). Park et al. (1998) designed a hybrid pathway using toluene dioxygenase, toluene cis-glycerol de-hydrogenase, and TPL for the production of L-DOPA using benzene as substrate ( Figure 12C). This strategy presented toxicity drawbacks since it was used in whole-cell systems (Park et al., 1998). A somehow similar system obtained by coupling of monooxygenase P450-BM3 with a NADPH regeneration system an engineered TPL allowed one-pot synthesis of L-Tyr derivatives starting from monosubstituted benzenes, pyruvate, and ammonia ( Figure 12D, Dennig et al., 2015). L-lactate oxidases coupled to a catalase and a designed thermophilic TPL (TTPL) was applied for one-pot synthesis of L-tyrosine derivatives ( Figure 12E, Li et al., 2020). The best lactate oxidase studied (AvLOX) was used for in situ generation of pyruvate starting from L-lactate (also obtained with another enzyme cascade, and thus, they might be coupled); different o-phenol acceptors were used for TTPL. Using 28-36 h of reaction, recovery yields ranging 51-84% were obtained for the different L-Tyr derivatives .

Tryptophan Synthase-Containing MECs
Tryptophan synthase (TrpS) has proved an interesting tool of expansion of NcAA MECs, since it has been engineered to accept a wide range of compounds, such as nitroalkanes, nitroindoles or 3-substituted oxindoles for the production of many different NcAAs (Buller et al., 2015;Herger et al., 2016;Romney et al., 2017Romney et al., , 2019Boville et al., 2018a,b), leading even to new quaternary stereocenters (Dick et al., 2019). TrpS has also been engineered to accept L-Thr as a substrate for the synthesis of β-methyltryptophan derivatives, which could be further halogenated by a halogenase (Francis et al., 2017).
TrpSs have been included in AAO-and TA-based MECs (sections "Amino Acid Oxidase-Based MECs" and "Transaminase-Based MECs"; Parmeggiani et al., 2019b;Schnepel et al., 2019). On the other hand, a series of NcAAs including L-Trp and L-Cys derivatives were synthesized using a D-threonine aldolase/TrpS/Alanine Racemase system in high conversion and excellent enantioselectivity (e.e. > 99%) starting from Gly and paraformaldehyde. In a first step, D-threonine aldolase from Arthrobacter sp transformed the substrates into D-serine with high yield (240 g/L). Racemization into D,L-Ser was carried out with alanine racemase from Bacillus subtilis; L-ser was then transformed into different L-NcAAs by coupling with various indoles/thiols using L-tryptophan synthase from E. coli (Yu et al., 2019).
identified, thus allowing to propose a putative AAER/esterase system for deracemization of AA esters, analogous to the "Amidase Process" (Figure 12F).

Other MECs Taking Advantage of Metabolic Pathways
Besides some of the examples shown in this work, many additional metabolic natural or engineered MECs exist in the literature. A recent seminal and must read review on biosynthetic pathways to NcAAs has been published (Hedges and Ryan, 2020), and the reader is referred to it for further information on these systems. As isolated examples, the L-Trp biosynthesis pathway from E. coli has been expanded by inclusion of engineered L-Phenylalanine 4-hydroxylase from C. violaceum (Figure 12G, Hara and Kino, 2013) or engineered aromatic amino acid hydroxylase from Cupriavidus taiwanensis (Mora-Villalobos and Zeng, 2018); whole cell systems for 5-HTP depend on the elimination of the E. coli tryptophanase gene, and inclusion of recycling of cofactors. Approx. 2-4 mM 5-HTP was produced using those systems. Natural routes for PheGly derivative synthesis have been reported, which might be further exploited for future MEC development (Al Toma et al., 2015). In this sense, special consideration should be paid to natural secondary metabolic routes conducting to NcAAs (or using them) to expand established MECs for their synthesis. Metabolic engineering of microorganisms have also allowed production of NcAAs such as L-norvaline/L-norleucine (Anderhuber et al., 2016), L-citruline (Eberhardt et al., 2014), hydroxyproline (Falcioni et al., 2015), L-ornithine ) different D-α-AAs (Mutaguchi et al., 2018) or different N-Alkylated AAs (Mindt et al., 2019(Mindt et al., , 2020van der Hoek et al., 2019). Thus, isolated enzymes comprised in biosynthetic metabolic routes for the production of different compounds (Hedges and Ryan, 2020) might find application on some of the existing MECs described in this paper.

CONCLUSION
In this review, we have brought together different established MECs and other less-used approaches for the biosynthesis of several NcAAs, providing a general overview on different methodologies available in the literature. Some of the enzymes described in this review have also been evolved to alter their substrate scope, widening their application, and inviting to revisit the old methodologies for further development. Enzymes from biosynthetic pathways to NcAAs (Hedges and Ryan, 2020) are clear candidates for MECs expansion. On the other hand, information of specific enzymes for functionalization of AAs or AA transformation into other important pharmaceutical compounds is also accumulating in the literature (e.g., Smirnov et al., 2012;Busto et al., 2014;Hönig et al., 2017;Hyslop et al., 2018;Parthasarathy et al., 2018;Peters and Buller, 2019;McDonald et al., 2019;Hedges and Ryan, 2020;Mindt et al., 2020;Song et al., 2020;Wendisch, 2020;Zhang et al., 2020;Zhao et al., 2020). These enzymes are, in principle, complementary to well stablished NcAA-production MECs. As way of example, promiscuous L-tryptophan decarboxylases have been applied for the synthesis of different triptamines (McDonald et al., 2019), and might be coupled to TprS-containing MECs for their expansion. Different proteases have also been applied to produce amino acidester derivatives starting from amino acids (Song et al., 2020), which might be coupled to several MECs proposed in this review. Thus, whereas the discovery of new or newly designed enzymes continues being of great interest, a "back and to the future" strategy might also speed up the "Fourth Wave of Biocatalysis" by dusting off previous enzymatic methodologies. We expect that potential readers in the field find the information contained in this paper helpful to speed up the generation of new or improved MECs.

AUTHOR CONTRIBUTIONS
SM-R designed and directed the idea of this review. All the authors participated in literature search, writing, and comments on the manuscript.

FUNDING
SM-R thanks the University of Granada for the support provided by project PPJI2017-1 and the European Cooperation in Science and Technology (COST Action CA15133). Authors are also grateful to the Andalusian Regional Government through Endocrinology & Metabolism Group (CTS-202).