Computational-Designed Enzyme for β -Tyrosine Production in Lignin Valorization

: Lignin is an underutilized sustainable source of aromatic compounds. To valorize the low-value lignin monomers, we proposed an efﬁcient strategy, involving enzymatic conversion from trans - p -hydroxycinnamic acids to generate valued-added canonical and non-canonical aromatic amino acids. Among them, β -amino acids are recognized as building blocks for bioactive natural products and pharmaceutical ingredients due to their attractive antitumor properties. Using computational enzyme design, the ( R )- β -selective phenylalanine aminomutase from Taxus chinensis (TchPAM) was successfully mutated to accept β -tyrosine as the substrate, as well as to generate the ( R )- β -tyrosine with excellent enantiopurity (ee > 99%) as the unique product from trans - p hydroxycinnamic acid. Moreover, the kinetic parameters were determined for the reaction of four Y424 enzyme variants with the synthesis of different phenylalanine and tyrosine enantiomers. In the ammonia elimination reaction of ( R )- β -tyrosine, the variants Y424N and Y424C displayed a two-fold increased catalytic efﬁciency of the wild type. In this work, a binding pocket in the active site, including Y424, K427, I431, and E455, was examined for its inﬂuence on the β -enantioselectivity of this enzyme family. Combining the upstream lignin depolymerization and downstream production, a sustainable value chain based on lignin is enabled. In summary, we report a β -tyrosine synthesis process from a monolignol component, offering a new way for lignin valorization by biocatalyst modiﬁcation.


Lignin Valorization
The consistent depletion of finite petroleum resources has excited the rising eagerness for new renewable materials in recent years [1]. Lignocellulosic biomass, a low-priced and abundantly available feedstock, has piqued researchers' curiosity around the globe on the development of strategies for the conversion into renewable products [2][3][4]. The primary components of lignocellulosic biomass are cellulose, hemicellulose, and lignin. Among them, cellulose and hemicellulose can be degraded into carbohydrates by hydrolysis and further converted to ethanol and other fine chemicals [5]. In contrast to the polysaccharides in lignocellulosic biomass, lignin is formed by cross-linked aromatic monolignol precursors in a heterogeneous network to provide terrestrial plants with mechanical support, facilitating the transport of water and nutrients and defense against various pathogen attacks [6]. Because of the complexity of the structure and the recalcitrance to degradation, lignin is usually released as a waste in the pulp and paper mill effluent or is simply burned for the production of heat and power. Currently, over 50 million tons of lignin are produced annually in global paper and bioethanol factories [7]. Numerous studies have been carried out on the development of lignin valorization to generate aromatic value-added products [8].
However, lignin is primarily recognized as a macromolecule in valorization strategy based on the growing mature depolymerization techniques. Although this strategy has been accomplished with numerous promising examples, such as carbon fibers [9], adhesives [10], emulsifiers [11,12], binders [13,14], and lignin-based nanoparticles [15], the valorization of low-molecular-weight and high-value chemicals from lignin with aromatic potential is still in the initial stage and remains a huge challenge. Only a few industrial-scale examples in these areas have been reported until now, for instance, ferulic acid and vanillin production [8,16,17]. Currently, more rewarding strategies are being developed depending on the technological advances in lignin valorization, including recovery, depolymerization, and conversion. The establishment of a sustainable value chain has gained much attention in the area of lignin valorization [18,19].
The monolignols of lignin, differing in their degree of methoxylation: p-coumaryl-, coniferyl-, and sinapyl alcohol, are synthesized from the phenylpropanoid pathway, especially under drought [20], salinity [21], and heat [22] stresses. Besides, another monomer in the phenolic acid form like p-hydroxycinnamic acid also drives lignin polymerization and regulates seeding growth and germination [23]. The distribution of these monolignols varies among plant species and tissue types, and development stages and growth conditions. For instance, p-hydroxycinnamyl alcohol, representing the typical monolignol in grasses in a small percentage, offers significant potential to form p-hydroxycinnamic acid, which could also be produced directly from base-catalyzed depolymerized lignin liquors [24]. Notably, p-hydroxycinnamyl alcohol, comprising around 30% of total lignin, is one of the dominant compounds in curaua leaf fiber [25].

β-Amino Acids and Anticancer Drugs
Cancer was responsible for approximately 10 million deaths in 2020, according to the WHO report [26]. With the increase in mortality rates of cancer and the associated socioeconomic burden, one of the prime requirements of anticancer drug research is the establishment of a more economic and effective strategy for their industrial-scale synthesis.
β-amino acids, as building blocks in natural products with an attractive antitumor feature, have gained considerable interest in the field of oncology drug development over recent years. As a consequence, numerous bioactive products with high pharmacological potency that contain optically pure β-amino acid are already available as anticancer drugs. For instance, the notable β-amino acid-involved anticancer agent C-1027 chromoprotein, isolated from Streptomyces globisporus, is assembled by a unique β-amino moiety, (S)-3-chloro-4,5-dihydroxy-β-phenylalanine [27]. Another example in the treatment of various types of cancers is paclitaxel (Taxol ® ), which was initially extracted from the bark of the Pacific yew tree, Taxus brevifolia. Paclitaxel, as an effective microtubule-stabilizing drug to induce mitotic arrest of cells, carries a β-phenylalanine-derivative (2R,3S)-phenylisoserine on the functionally essential C13-side chain [28,29].
Phenylalanine aminomutase (PAM; EC 5.4.3.10), catalyzing the rearrangement of the (S)-α-phenylalanine and (R)-β-phenylalanine, is the principal enzyme in the committed step of the N-benzol phenylisoserinoyl side chain in the taxol biosynthesis pathway, while tyrosine aminomutase (TAM; EC 5.4.3.6) catalyzes the formation of C-1027 from L-tyrosine [27][28][29]. Both of these enzymes are members of the MIO (3,5-dihydro-5methylidene-4H-imidazole-4-one)-dependent aminomutase family. These so-called MIOdependent aminomutases recruit an unusual highly electrophilic moiety MIO for catalyzing the 2,3-amine shift in aromatic amino acids and present outstanding enantioselectivity. The members from this enzyme family are frequently applied for an asymmetric synthesis route without cofactor recycling to produce the optically pure β-amino acids [30,31]. On the other hand, the enzymatic synthesis of optically pure β-amino acids has been considered as an attractive strategy, compared to the chemical processes, which are limited by low yield or enantiopurity, and high cost of the waste treatment. Among these MIO-dependent enzymes, PAM is presumably evolved in ancient gymnosperms from the ancestry with phenylalanine ammonia-lyase (PAL; EC 4.3.1.24), according to our previous conclusion from phylogenetic analysis. Regulating the metabolite accumulation in the phenylpropanoid pathway, PAM could be employed as a connection point for conversion from monolignol to high-valued β-amino acids in the enzymatic lignin valorization strategy [32].
Enzyme design efforts relying on the screening of large mutant libraries have had noteworthy success in the last decades, but they are usually time-and cost-intensive by using wide-ranging random mutagenesis. Instead, computational methods have been developed to guide the selection of the target substrate of enzymes in a limited number. One of the most powerful model tools is PyRosetta, which was broadly used for enzyme design based on the binding energy calculation of small molecule-protein interaction [33]. Supported by the previous high-resolution X-ray structure analysis and site-directed mutagenesis studies, TchPAM was an enzyme with excellent enantioselectivity. Therefore, it was chosen for the biosynthesis investigation of β-tyrosine from trans-p-hydroxycinnamic acid. To avoid screening in the massive mutant libraries, TchPAM is designed using a computational program instead of random experimental evolution [34,35].

Computational Design of TchPAM for the Synthesis of β-tyrosine
Previous mutagenesis studies on MIO-dependent enzymes revealed that the responsible residues for the substrate selectivity were mainly placed around the aromatic side of the substrate [33]. The X-ray co-crystal structure of TchPAM with native substrate-bound (PDB ID: 4C5R) was investigated to identify the other potential residues for acceptance of β-tyrosine [34]. It was hypothesized that various residues in the active site of TchPAM orient and stabilize the aromatic ring of the substrate. Previous studies on the substrate spectrum of TchPAM revealed that this enzyme accepts a range of monosubstituted cinnamic acid derivatives, especially those with methyl groups at the para-position, such as 4-methoxycinnamic acid. Therefore, β-tyrosine with a 1,4-hydroxyl group is not accepted by this enzyme because of the electrostatic restriction instead of the steric constraint [36]. Excluding the essential residues for catalysis and stereoselectivity (Asn231, Gln319, Tyr322, Arg325, Asn355), the other residues were divided into two groups. Group one comprises residues like Phe86, Leu104, Cys107, Leu108, Leu179, Tyr424, Lys427, Ile431, and Glu455, which potentially interact with the phenolic side of β-tyrosine via a hydrogen bond. The second group includes residues, such as Asn458 and Gln459, which form a stable network with substrates. The former residues were customized to contain polar or electrically charged side chains, while the latter were randomly mutated in silico without any limitation ( Figure 1). In conclusion, a library of variants with multiple mutations was established using Rosetta enzyme design based on the penalty score of the constraints and visual structure screening of the designed enzymes.

Experimental Screening of the Designed TchPAM Mutants
Initially, the single point mutants from the mutant library were selected to explore the correlation between the phenolic group of β-tyrosine and the residues in the active site of the enzyme. Unexpectedly, residue Leu107, which was considered as the crucial position for the selectivity switch between tyrosine-and phenylalanine ammonia-lyase activity, was excluded in this computational mutant library [36,37]. To determine the substrate tolerance of the designed mutants, the specific activity in ammonia elimination reactions of (S)-α-phenylalanine, (R)-β-phenylalanine, (S)-α-tyrosine, and (R)-β-tyrosine were measured ( Figure 2). . The color coding corresponds to the position of residues in the active site: carboxyl binding pocket (red), aromatic binding pocket (blue), tyrosine towards pocket (green), essential functional residues (black), and all other residues without specified function (purple). The substrate (e.g., (S)-α-phenylalanine) is highlighted by a yellow background.

Experimental Screening of the Designed TchPAM Mutants
Initially, the single point mutants from the mutant library were selected to explore the correlation between the phenolic group of β-tyrosine and the residues in the active site of the enzyme. Unexpectedly, residue Leu107, which was considered as the crucial position for the selectivity switch between tyrosine-and phenylalanine ammonia-lyase activity, was excluded in this computational mutant library [36,37]. To determine the substrate tolerance of the designed mutants, the specific activity in ammonia elimination reactions of (S)-α-phenylalanine, (R)-β-phenylalanine, (S)-α-tyrosine, and (R)-β-tyrosine were measured ( Figure 2). Tyr424Asn, the best variant of the library, shows the highest specific activity toward β-tyrosine and served not only to confer tyrosine acceptance but also to enhance the phenylalanine ammonia-lyase activity. An analogous outcome was observed in Tyr424Cys, but with a lesser activity improvement, compared to the wild type and the other variants (Tyr424Asp and Tyr424Gln). In the wild type, it was hard to form an interaction due to the distance of above 8 Å between the ring centroids from Tyr424 and the aryl-ring on the substrates. The mutant Tyr424Cys has a similar but more flexible conformation in the active site. Although, the partial negatively charged sulfur group on the side chain could interact with the positive charge on the edge of the aromatic ring within a certain distance, there are fewer steric effects by cysteine than tyrosine at this position. This hypothesis is supported by their comparable activity toward the deamination of various substrates [38]. However, mutation of the residue at position 424 to smaller amino acids mainly affects the orientation of the neighboring residues, such as Phe86 and Lys427, during protein folding ( Figure 3). Asn or Gln, which were recognized as globally neutral amino acids, offered a partial positive charge on their formamide group when placed near the phenolic ring of the substrate and the residue Phe86. Therefore, these residues could provide an attractive interaction with the ring center ( Figure 3B,D) [38][39][40]. Notably, their amino groups were closer to the aromatic ring than their carbonyl groups, thereby increasing the electrostatic contribution. The functional group on position 424 resulted in a potential variety of conformation and geometrical constraints with its neighboring residues, especially in the case of Gln with a larger side chain [41]. Unlike the glutamine at this position, the mutant Tyr424Asn might have less or no influence on the substrate binding, considering the distance (5.7 Å) between the amide group and the centroid ring. Replacement of the tyrosine with negatively charged aspartic acid at this position generates a slightly repulsive force toward the negatively charged ring center in the substrates [42]. Therefore, the docking location of the substrate was likely more distant from this residue than the wild type ( Figure 3). Moreover, the carboxyl group of the residue Tyr424Asp might interact with the amino group of its neighboring residue Lys427, which was stabilized by the carboxyl group from the residue Glu455. It might be a possible reason for partial activity loss in mutant Tyr424Asp. In addition, both neighboring residues Lys427 and Glu455 were suggested to be catalytically essential positions in TchPAM, because their variants abolished the activity by using all four substrates. The other neighboring residue Ile431, however, seemed to strictly require a hydrophobic group since the variants harboring mutations to Asn or Glu lost complete activity. Consequently, the residue Tyr424 might influence the substrate positioning through the interaction of its side chain with the aryl ring on the aromatic amino acids. Moreover, the neighboring residues in this area play an important role in substrate positioning. Tyr424Asn, the best variant of the library, shows the highest specific activity toward β-tyrosine and served not only to confer tyrosine acceptance but also to enhance the phenylalanine ammonia-lyase activity. An analogous outcome was observed in Tyr424Cys, but with a lesser activity improvement, compared to the wild type and the other variants (Tyr424Asp and Tyr424Gln). In the wild type, it was hard to form an interaction due to the distance of above 8 Å between the ring centroids from Tyr424 and the aryl-ring on the substrates. The mutant Tyr424Cys has a similar but more flexible conformation in the active site. Although, the partial negatively charged sulfur group on the side chain could interact with the positive charge on the edge of the aromatic ring within a certain distance, there are fewer steric effects by cysteine than tyrosine at this position. This hypothesis is supported by their comparable activity toward the deamination of various substrates [38]. However, mutation of the residue at position 424 to smaller amino acids mainly affects  According to the amino acid sequence analysis, Phe86 was naturally replaced by tyrosine in the other MIO-dependent enzymes, which could accept tyrosine as substrates. However, the lack of activity in the Phe86Tyr variant indicated that the aromatic pair interaction between residue Phe86 and the benzyl or phenolic ring from the substrate probably induces a π-π interaction. This interaction helps to orient and stabilize the substrates in the wild type. Usually, this interaction could be observed at a pairing distance larger According to the amino acid sequence analysis, Phe86 was naturally replaced by tyrosine in the other MIO-dependent enzymes, which could accept tyrosine as substrates. However, the lack of activity in the Phe86Tyr variant indicated that the aromatic pair interaction between residue Phe86 and the benzyl or phenolic ring from the substrate probably induces a π-π interaction. This interaction helps to orient and stabilize the substrates in the wild type. Usually, this interaction could be observed at a pairing distance larger than 3.8 Å in slipped face-to-face stacking. The tyrosine variants increased the interaction by the withdrawing π-electron density from the hydroxyl group, resulting in a reduction of the electrostatic repulsion and the contact distance between the aromatic ring of residue Phe86Tyr and substrate [43,44]. As the other high-profile candidate residue Leu108, which was considered as the responsible residues for α-tyrosine acceptance in the enzyme, did not seem to have the equivalent substrate-switching ability toward its β-isomers [37]. By replacing the residue Leu108, the mutants Leu108Gln and Leu108Ser were considerably less active toward both phenylalanine and tyrosine. In the former, the residue glutamine with the large side chain was more sterically hindered than leucine in the wild type, which could explain the notable activity decrease. On the other hand, the hydroxyl group on residue serine contributed to the strength of interaction with the sulfur atom from its nearby neighboring residue Cys107, giving a variant Leu108Ser with much lower activity.
The Leu104Ser variant was found to be the most efficient enzyme to catalyze the deamination of (S)-α-phenylalanine, giving an impressive 4-fold enhancement of specific activity, while the Leu104Thr, with a shorter polar uncharged side chain, lost most activity. This was consistent with the previous structural analysis of Leu104Ala, which was predicted to have an unfavorable steric strain between the residue and substituted cinnamic acid derivatives, especially at the meta-position. The same was observed when expanding the substrate scope of the MIO-dependent PcPAL from Petroselinum crispum [45,46]. It could be suggested that the residue Leu104 might influence the enzymatic catalysis due not only to the steric effect but also to the hydrophobic interaction between the residue and aromatic ring from the substrate.

Kinetic Analyses of Designed Enzymes
For Tyr424Asn and Tyr424Cys, the variants with enhanced substrate tolerance toward β-tyrosine, the kinetic parameters in the deamination reaction were determined, compared to the wild type and the other variants (Tyr424Asp and Tyr424Gln), which were mutated at the identical position ( Table 1).
The kinetic data revealed that both variants with higher activity had an improved catalytic efficiency with tyrosine and a decreased catalytic efficiency with phenylalanine. Regarding β-tyrosine, the catalytic efficiency of Tyr424Cys and Tyr424Asn were 0.281 ± 0.023 and 0.234 ± 0.013 mM −1 s −1 , respectively. The variant Tyr424Cys displayed an approximately two-fold decreased K M value but a k cat value similar to that of the wild type. This is supporting the hypothesis that the replacement of tyrosine with a shorter side chain of the residue at position 424 could afford additional space for binding the phenolic ring on the substrate (Figure 3), whereas Tyr424Asn provided a similar binding affinity (K M ) and a superior turnover rate (around two-fold enhancement). This might be due to the interaction between the phenolic ring from the substrate and the partial positively charged amino group on the mutated residue when the β-tyrosine entered the catalytic site. Apart from the variant Tyr424Asp with a five-fold decline in K M for α-phenylalanine, all four mutations yielded a lower binding affinity for their natural substrates in both isomers. Especially, Tyr424Asn gave a massive 20-fold higher K M value when using βphenylalanine as the substrate. It could be suggested that the residue Tyr424 assisted to enhance phenylalanine acceptance because of its large enough and negatively charged side chain. Meanwhile, variants Tyr424Asn and Tyr424Cys with α-tyrosine showed a huge, approximately 7-fold decrease in K M value compared to the wild type. Therefore, the Tyr424 residue was confirmed to be a responsible position for conferring tyrosine or phenylalanine acceptance.

Enantioselectivity of Designed Enzymes
The wild type of TchPAM can produce (R)-β-phenylalanine with >99% enantiomeric excess by using trans-cinnamic acid as the substrate. This accounts for around 50% of the product mixture, where (S)-α-phenylalanine occupies the rest. However, the trans-p-hydroxycinnamic acid was not accepted by this enzyme [36]. As a rare (R)-selective phenylalanine aminomutase in nature, TchPAM presents outstanding potential for the production of enantiopure (R)-β-tyrosine. Until now, comparable enantioselectivity has not been observed in the wild type of another member from the MIO-dependent enzyme family in the bacterial kingdom, for instance, the enzyme KedY4 from Streptoalloteichus sp. and the enzyme CmdF Chondromyces crocatus [47,48]. In contrast, the OsTAM from Oryza sativa, the only example of (R)-selective tyrosine aminomutase with great enantioselectivity in the plant, which is the closest relative of TchPAM, is tricky to apply in trans-p-hydroxycinnamic acid production [49].
The previous X-ray crystal structure analysis and the site-directed mutagenesis studies in TchPAM gave the proposal of the isomerization mechanism for explaining the aminomutase activity, revealing two possible distinct binding modes of the carboxyl group on substrates. The ammonia 1,4-addition for the production of (S)-α-phenylalanine was carried out by forming a bidentate salt bridge between the carboxylate group on the substrate and the amino groups from residue Arg325. The Gln319 afforded an extra interaction with the hydrogen bond and contributes to the orientation of the substrate. In the case of conversion to the corresponding (R)-configured β-phenylalanine, the carboxyl group was stabilized by a network of hydrogen bonds, created by Asn231, Arg325, and Asn355. During the whole catalytic process, the aryl ring from trans-p-cinnamic acid was locked in the aromatic binding pocket, which contained the residues Phe86, Leu104, Leu107, and Leu108, and was identified to play a crucial role in substrate selectivity in previous studies [33]. From a similar catalytical point of view, two binding conformations could be postulated, when trans-p-hydroxycinnamic acid entered the active site as a starting substrate. Compared to the natural substrate trans-cinnamic acid, the ammonia addition of trans-p-hydroxycinnamic acid was difficult due to the substituted hydroxyl group of the benzenic ring. More electronegative oxygen atom shifts the electron to itself, stabilized by the negatively charged electron-withdrawing group, thus leading to the priming for nucleophilic attack at the β-position (Figure 4) [50]. Unlike the aryl ring from trans-cinnamic acid, which was stabilized in the aromatic binding pocket (Leu104, Leu108, and Leu179) by London dispersion force, the aryl ring from the trans-p-hydroxycinnamic acid oriented towards the residues with a polar or charged side chain (Cys107, Tyr424, Lys427, and Glu455) that could be attributed to its hydroxyl functional group ( Figure 5). Of these two mentioned binding conformations of the carboxylate group on the substrate, only one would be favored in general. In this situation, the negatively charged carboxylate on p-hydroxycinnamic acid was preferred to stabilize in the latter binding mode, involving the hydrogen bonding formed by Asn231, Arg325, and Asn355, leading to the production of (R)-β-tyrosine. It could be strongly supported by the huge difference in the binding affinity for (S)-α-tyrosine and (R)-β-tyrosine in the wild type ( Figure 5C,D). At this moment, the substrate backbone is placed closer to the carboxyl binding pocket, compared to the (S)-α-regioisomer ( Figure 5A,B).
During the whole catalytic process, the aryl ring from trans-p-cinnamic acid was locked in the aromatic binding pocket, which contained the residues Phe86, Leu104, Leu107, and Leu108, and was identified to play a crucial role in substrate selectivity in previous studies [33]. From a similar catalytical point of view, two binding conformations could be postulated, when trans-p-hydroxycinnamic acid entered the active site as a starting substrate. Compared to the natural substrate trans-cinnamic acid, the ammonia addition of trans-phydroxycinnamic acid was difficult due to the substituted hydroxyl group of the benzenic ring. More electronegative oxygen atom shifts the electron to itself, stabilized by the negatively charged electron-withdrawing group, thus leading to the priming for nucleophilic attack at the β-position (Figure 4) [50]. Unlike the aryl ring from trans-cinnamic acid, which was stabilized in the aromatic binding pocket (Leu104, Leu108, and Leu179) by London dispersion force, the aryl ring from the trans-p-hydroxycinnamic acid oriented towards the residues with a polar or charged side chain (Cys107, Tyr424, Lys427, and Glu455) that could be attributed to its hydroxyl functional group ( Figure 5). Of these two mentioned binding conformations of the carboxylate group on the substrate, only one would be favored in general. In this situation, the negatively charged carboxylate on phydroxycinnamic acid was preferred to stabilize in the latter binding mode, involving the hydrogen bonding formed by Asn231, Arg325, and Asn355, leading to the production of (R)-β-tyrosine. It could be strongly supported by the huge difference in the binding affinity for (S)-α-tyrosine and (R)-β-tyrosine in the wild type ( Figure 5C,D). At this moment, the substrate backbone is placed closer to the carboxyl binding pocket, compared to the (S)-α-regioisomer ( Figure 5A,B).     This hypothesis was backed up by the measurement of the conversion and the enantiomeric excess of the designed enzymes in ammonia addition reaction of trans-phydroxycinnamic acid, monitored by High-performance liquid chromatography (HPLC) assays. Both candidate mutants, where the residue was replaced by asparagine or cysteine with a smaller and different chemical functional side chain in position 424, presented the tyrosine ammonia-lyase activity (Table 2). Strikingly, the superior mutants Tyr424Asn and Tyr424Cys were designed for the generation of optically pure (R)-β-tyrosine. The specific activities of mutants Tyr424Asn and Tyr424Cys were 0.104 ± 0.002 and 0.141 ± 0.001 U/mg, respectively. The lack of tyrosine ammonia-lyase activity with the wild type might be ascribed to the steric clash between the aryl ring on the substrate and the residue Tyr424, while both β-selective variants provide extra space at this position. Taken together, these results strongly indicate the significance of the residues in a third binding pocket, including Tyr424, as well as its neighborhoods Lys427, Ile431, and Glu455, for substrate specificity. Table 2. Comparison of the specific activity and enantiomeric excess of wild type and mutants Tyr424Asn and Tyr424Cys in the ammonia addition reaction.

Enzyme
Specific Activity ee β

Conclusions and Perspectives
With excessive consumption of natural petroleum, the option to convert the underutilized lignin into high-value products has been recognized, based on its significant potential as a sustainable source of numerous aromatic compounds. In this paper, we reported an enzymatic conversion of trans-p-hydroxycinnamic acid into (R)-β-tyrosine with high enantiopurity as a valuable building block for pharmaceutical ingredients and drug candidates, such as kedarcidin [51]. However, the synthesis of (R)-β-tyrosine by natural MIO-dependent enzymes remains considerably challenging because of the low enantioselectivity of bacterial tyrosine aminomutase, as well as the poor tyrosine acceptance of the phenylalanine-metabolizing enzymes [47,48]. To form the cornerstone of the whole strategy, TchPAM was chosen for mutation via computational enzyme design to extend the substrate scope, according to the previous evidence of its great synthetic potential of (R)-β-phenylalanine. In conclusion, the variants Tyr424Asn and Tyr424Cys were found to yield (R)-β-tyrosine with high enantiopurity from trans-p-hydroxycinnamic acid. Using this variant, efficient and straightforward routes can be demonstrated by developing high-value chiral β-amino acids, as the building blocks in bioactive natural products, from lignocellulosic waste, which comprises corresponding aromatic compounds. This could be achieved through the multi-enzymatic cascade involving the designed TchPAM and the feruloyl esterases, which are able to release p-hydroxycinnamic acids from lignin, enabling sequential enzymatic conversion to be performed in a one-pot synthesis method [52]. Combining the enzymatic conversion with the further downstream synthesis of bioderived anticancer drug based on the enantiopure (R)-β-tyrosine, we pointed out the possibility of valorization of the underused lignin in a low-molecular-weight value-added product ( Figure 6). In summary, this study highlighted a lignin-based β-tyrosine synthesis process, employing the computational-designed TchPAM. This enzymatic conversion for the production of the (R)-β-tyrosine with excellent enantioselectivity and regioselectivity could connect with the upstream production process of p-hydroxycinnamic acid from lignin, as well as the drug development downstream in pharmaceutical industries, providing meaningful opportunities for future research in lignin valorization.

Computational Enzyme Design
To determine the appropriate residues to mutate, the computational design of TchPAM was processed by using RosettaScripts, a scripting language interface based on PyRosetta [53]. The β-tyrosine as the target substrate and the MIO moiety were manipulated by software OpenBabel 2.4 [54] and their transition state geometry optimized in Gaussian09 [55]. Moreover, the Biology and Chemistry Library Project (BCL) was used to generate the substrate and moiety conformers [56]. According to the staring coordinates, which are determined by the information of the PDB file, the natural substrate was replaced with the prepared target substrate in the binding pocket of the relaxed enzyme (PDB ID: 4C5R). Finally, the best designs were selected based on the calculation of catalytic constraints from PyRosetta by using a high-performance computer and the structure comparison of natural templates in the visualization system PyMOL [57] and Chimera [58].

Construction of TchPAM Mutants
According to the protein sequence from Taxus chinensis (NCBI, AAU01182.1), the pET28a plasmid, containing the codon-optimized pam gene, was synthesized by Proteo-Genix (Schiltigheim, France) and used as a template for site-directed mutagenesis with Q5 ® High-Fidelity 2X Master Mix (NEB, Ipswich, MA, USA). Apart from the mutation at residue Tyr424 with KLD Enzyme Mix from NEB (Ipswich, MA, USA), the touch-down PCR was performed for the rest of the mutants. All primers were ordered from Thermo Fischer Scientific (Waltham, MA, USA) and are listed in Supplementary Materials Table  1. The template strand was immediately removed by adding 1 µ L of DpnI for 1 h at 37 °C . Subsequently, the PCR reaction mixture was transformed into high-efficiency chemically In summary, this study highlighted a lignin-based β-tyrosine synthesis process, employing the computational-designed TchPAM. This enzymatic conversion for the production of the (R)-β-tyrosine with excellent enantioselectivity and regioselectivity could connect with the upstream production process of p-hydroxycinnamic acid from lignin, as well as the drug development downstream in pharmaceutical industries, providing meaningful opportunities for future research in lignin valorization.

Computational Enzyme Design
To determine the appropriate residues to mutate, the computational design of Tch-PAM was processed by using RosettaScripts, a scripting language interface based on PyRosetta [53]. The β-tyrosine as the target substrate and the MIO moiety were manipulated by software OpenBabel 2.4 [54] and their transition state geometry optimized in Gaussian09 [55]. Moreover, the Biology and Chemistry Library Project (BCL) was used to generate the substrate and moiety conformers [56]. According to the staring coordinates, which are determined by the information of the PDB file, the natural substrate was replaced with the prepared target substrate in the binding pocket of the relaxed enzyme (PDB ID: 4C5R). Finally, the best designs were selected based on the calculation of catalytic constraints from PyRosetta by using a high-performance computer and the structure comparison of natural templates in the visualization system PyMOL [57] and Chimera [58].

Construction of TchPAM Mutants
According to the protein sequence from Taxus chinensis (NCBI, AAU01182.1), the pET28a plasmid, containing the codon-optimized pam gene, was synthesized by Pro-teoGenix (Schiltigheim, France) and used as a template for site-directed mutagenesis with Q5 ® High-Fidelity 2X Master Mix (NEB, Ipswich, MA, USA). Apart from the mutation at residue Tyr424 with KLD Enzyme Mix from NEB (Ipswich, MA, USA), the touch-down PCR was performed for the rest of the mutants. All primers were ordered from Thermo Fischer Scientific (Waltham, MA, USA) and are listed in Supplementary Materials Table S1. The template strand was immediately removed by adding 1 µL of DpnI for 1 h at 37 • C. Subsequently, the PCR reaction mixture was transformed into high-efficiency chemically T7 express competent E. coli (NEB, USA), according to the manufacturer's protocol. Mutagenesis success was controlled by Sanger sequencing (GATC, Ebersberg, Germany) of purified plasmid from the well-isolated colonies.

Expression, Purification, and Concentration of TchPAM
For the expression of the wild type and the mutants of TchPAM, the cells were incubated in 100 mL of LB-medium with 50 µg/mL kanamycin at 37 • C and 120 rpm. Once the OD 600 reached 0.3-0.4, the cultures were induced with 0.1 mM IPTG at 25 • C with shaking at 120 rpm. After 22 h, cells were harvested by centrifugation, washed twice with 50 mM Tris-HCl buffer (pH 8.0), and resuspended in 5 mL of binding buffer (50 mM Tris-HCl, 150 mM NaCl, 10 mM imidazole, pH 8.0). Cell disruption was performed on ice by sonication (20 s pulse/30 s pause over 10 min, with 40 % amplitude, HD3100, Bandelin electronic GmbH & Co. KG, Berlin, Germany). Cell debris was removed by centrifugation (16,000× g, 20 min, 4 • C) (Megafuge 40R, Thermo Fisher Scientific Inc., Waltham, MA, USA). The histidine-tagged enzymes were obtained through purification with His SpinTrap columns (GE Healthcare Life Sciences, London, UK). The composition of the washing buffer (pH 8.0) was 50 mM Tris-HCl, 150 mM NaCl, and 20 mM imidazole, while the elution buffer had the same composition except for a higher concentration of imidazole (300 mM). Then, the purified enzymes were desalted and concentrated by the protein concentrators with a 10 K MWCO (Pierce ® Concentrator, PES, 0.5 mL, Thermo Fisher Scientific Inc., Waltham, MA, USA). Finally, the protein concentrations of the wild type and mutants of TchPAM were determined by using the Epoch TM Multi-volume Spectrophotometer System (BioTek Instruments, Inc., Winusky, VT, USA) with a Take3Plate (BioTek Instruments, Inc., Winusky, VT, USA), based on the BSA protocol. All purified enzyme solutions were diluted to 0.5 mg/mL as enzyme stock for subsequent activity measurement. The purity of the desalted enzymes was checked by SDS-PAGE (Supplementary Materials Figure S1).

TchPAM Activity Assays for Ammonia Elimination Reactions
The substrate solution of 1 mM substrate ((S)-α-phenylalanine, (R)-β-phenylalanine, (S)-α-tyrosine, or (R)-β-tyrosine) was prepared in Tris-HCl buffer (50 mM, pH 8.8) and pipetted in 96-well UV-Microplates. Then, 10 µL of purified wild type or mutant TchPAM enzyme were added to the preheated 100 µL of substrate solution. The ammonia elimination reactions were carried out in triplicate at 40 • C for 30 min, monitoring the product absorbance for trans-cinnamic acid at 300 nm, as well as for trans-p-hydroxycinnamic acid at 350 nm. The product absorbance was measured every 3 min by using the Epoch TM Spectrophotometer system (BioTek, Winusky, VT, USA).

Kinetics Measurements
The above-mentioned protocol was used to determine the kinetic parameters of the wild type and mutants of TchPAM. The measurements were performed in triplicate in 96-well UV-Microplates, with a varying substrate concentration range between 0.01 and 10 mM, until substrate saturation occurred. The maximal concentration of (S)-α-tyrosine was 2 mM due to its poor solubility in Tris-HCl buffer (pH 8.8). The kinetic parameters (K M and V max ) were determined by employing Michaelis-Menten nonlinear fitting using Origin (Pro 2019, OriginLab ® , Northampton, MA, USA).

HPLC Analysis in the Ammonia Addition Reaction
Substrate solution with 3 mM trans-p-hydroxycinnamic acids was prepared in ammonium sulfate solution (4 M, pH 10). First, 50 µL of purified enzymes were added to 500 µL of substrate solution. The reaction was performed at 40 • C. After the indicated reaction time (0, 1, 2, 3, 5, 7, and 24 h), 50 µL of sample were taken from the mixture and heated at 99 • C for 5 min. To determine the concentration of all four mentioned aromatic amino acids, which are possible products of the reaction, the samples were analyzed by HPLC (1200 Series, Agilent Technologies, Waldbronn, Germany). Then, 0.5 µL of sample were injected in the HPLC column (C18, 150 × 4.6 mm HyperClone 5 µm, Phenomenex Inc, Aschaffenburg, Germany) after derivatization with o-phthalaldehyde (OPA) and N-isobutyryl-L-cysteine (IBLC) in an automatic pre-column according to the protocol by Brucher et al. [59]. The compounds in the reaction mixture were eluted with 21% sodium phosphate buffer (40 mM, pH 6.5) and 79 % methanol at a constant flow rate of 1 min/mL. The detector was operated at 338 nm. The retention times of the produced aromatic amino acids were (S)-α-tyrosine 71.0 min, (R)-α-tyrosine 45.7 min, (S)-β-tyrosine 57.6 min, and (R)-β-tyrosine 42.7 min.  Figure S1: The SDS-PAGE gel of the crude protein lysate and purified enzymes. The band of TchPAM was observed between 63 and 75 kDa. WT: wild type.
Author Contributions: Conceptualization, J.R.; methodology, U.E. and F.P.; high performance computer, H.A.; computational enzyme design, screening the designed enzyme, characterization of the enzyme by UV spectroscopy and HPLC, F.P.; writing the manuscript, F.P.; revising the manuscript, J.R., U.E., A.D. and H.A.; supervision, J.R. All authors have read and agreed to the published version of the manuscript.
Funding: This work by author was supported by China Scholarship Council and is part of the bioeconomy graduate program BBW ForWerts. We gratefully thank the Open Access Publishing Fund of Karlsruhe Institute of Technology.

Conflicts of Interest:
The authors declare no conflict of interest.