Functional Diversification and Structural Origins of Plant Natural Product Methyltransferases

In plants, methylation is a common step in specialized metabolic pathways, leading to a vast diversity of natural products. The methylation of these small molecules is catalyzed by S-adenosyl-l-methionine (SAM)-dependent methyltransferases, which are categorized based on the methyl-accepting atom (O, N, C, S, or Se). These methyltransferases are responsible for the transformation of metabolites involved in plant defense response, pigments, and cell signaling. Plant natural product methyltransferases are part of the Class I methyltransferase-superfamily containing the canonical Rossmann fold. Recent advances in genomics have accelerated the functional characterization of plant natural product methyltransferases, allowing for the determination of substrate specificities and regioselectivity and further realizing the potential for enzyme engineering. This review compiles known biochemically characterized plant natural product methyltransferases that have contributed to our knowledge in the diversification of small molecules mediated by methylation steps.


Introduction
Methylations, a process universal within all organisms, are responsible for extensive cellular modulations. These reactions are typically catalyzed by methyltransferases that require cofactor S-adenosyl-L-methionine (SAM), resulting in the formation of the methylated product and S-adenosyl-L-homocysteine (SAH) [1]. More commonly known for their roles in epigenetics, methyltransferases are also involved in natural product biosynthesis in plants. Plant natural product methyltransferases (PNPMTs) contribute to the diversification of specialized metabolites with functions such as pigments [2,3], antioxidants [4], signal transducer, and plant defense response [5][6][7][8]. Phenolic compounds, specifically, support plant growth and provide additional advantages to tolerate environmental stresses against insects and microbial invasion [9,10].
PNPMTs and their subsequent methylated products also display pharmaceutical properties for humans. Compared to their non-methylated counterparts, methylated plant natural products exhibit different chemical properties, and thereby alter biological activity. For instance, methylated resveratrol was shown to decrease its genotoxicity level compared to non-methylated resveratrol, which has been reported to cause chromosome aberration [11]. Furthermore, methylated resveratrol displays improved bioavailability and overall bioactivity [12,13]. Methylation also influences the binding of small molecules to a human neurotransmitter receptor. For example, as an intermediate in the morphine pathway, the O-methylated benzylisoquinoline (BIA) thebaine exhibits stimulatory effects and weaker analgesic properties compared to morphine, which lacks O-methylated groups [14]. Similarly, the methylated monoterpene indole alkaloid (MIA) ibogaine is less polar than

Classification of PNPMTs Based on Methyl Acceptor
Methyltransferases are categorized based on the substrate atoms O-, N-, C-, S-, and Sethat accept the methyl group. Though rare, a few reported PNMPTs were noted for their ability to transfer methyl groups to more than one acceptor, such as those methylating halides and thiocyanates [22]. Among all plant natural product methyltransferases, O-directed methyltransferases (OMTs) are the most abundant to date [1]. OMTs target the hydroxyl groups of small molecules, such as alkaloids, lignin precursors, simple phenols, phytoalexins, phytohormones, and flavonoids, to produce their methylated form. This also includes the hydroxyl moiety of the carboxyl group, which can be found in salicylic acid [5,23], jasmonic acid [7] and iridoid loganic acid [24]. OMTs are also further characterized into two subclasses based on their dependence on or independence of a divalent cation. The cation-dependent OMTs, also known as caffeoyl-CoA OMTs (CCoAOMTs), typically have a lower subunit molecular weight of 26-30 kDa. CCoAOMTs are mainly involved in lignin, phytohormone, and scent metabolism [16,25]. Comparatively, the cationindependent OMTs, also known as caffeic acid OMT (COMT), range between 37 and 43 kDa and are primarily active in phenylpropanoid and alkaloid biosynthesis [26].
The abundance of OMTs is also seen through the incredible diversity of their methoxylated products (Figure 1). For instance, plant alkaloid OMTs have been well-characterized, including the monoterpene indole alkaloid in Catharanthus roseus to synthesize vindoline, which is then dimerized with catharanthine to form anticancer vinblastine [27] and monoterpene isoquinoline alkaloid in Psychotria ipecacuanha [28]. Notably, out of all known alkaloid OMTs to date, the majority of them are involved in benzylisoquinoline alkaloid biosynthesis (Table 1). Further characterization of plant alkaloid OMTs, known as norbelladine 4 OMTs, has been reported in Amaryllidaceae alkaloid biosynthesis from Narcissus spp and Lycoris aurea [29,30]. Although heterocyclic nitrogenous methoxypyrazines are not characterized as alkaloids, the OMTs involved in their pathway have been reported [31][32][33].
The position and number of methyl groups determine the structural diversity of flavonoids, though these are restricted by the placement of hydroxyl groups [136]. Omethylation can occur at any position theoretically, however, 7-and 4 -methoxylation are most prevalent among flavonoids; interestingly, a number of polymethoxylated flavonoids have also been reported [136]. Although the majority of flavonoid-OMTs are active towards aglycone substrates, anthocyanin 3 OMT and 5 OMT were later characterized and displayed strong preference for glycosylated substrates [138].
While OMTs are generally regiospecific, plant OMTs involved in phenylpropanoid and flavonoid biosynthesis are less selective and catalyze sequential methylations of almost identical substrates [1,65,107,139]. A small group of plant OMTs has demonstrated strong enzymatic specificity for a single structure, but overall, plant OMTs tend to accept a wide span of substrates, with varying substrate preferences, including flavonoids, lignin precursors, and alkaloids [48,140]. For example, an alkaloid OMT, 10-hydroxycamptothecin OMT in Camptotheca acuminata, was strongly preferred to methylate 10-hydroxycamptothecin but was also capable of methylating flavonoids, stilbenes, and caffeic acids [48].
Though not as abundant as the OMTs, N-directed methyltransferases (NMTs) still offer a respectable diversity of natural products. The N-methylated phytochemicals consist of primary amines, secondary amines found in indoles and imidazoles, tertiary amines, or more complex alkaloids [1] (Figure 2). Most N-directed methyltransferases have been cloned from alkaloid-producing plants. In particular, the caffeine biosynthetic pathway in xanthosine involves a series of N-methylation steps that serve as a model to study alkaloid biosynthesis [151,152]. The biosynthetic genes involved were cloned from Camellia sinensis [152,153], Coffea arabica [54,55,57,58,153] Paullinia cupana [59], Theobroma cacao [154], and Citrus sinensis [151]. Remarkably, caffeine and its xanthine precursors have evolved independently in a number of flowering plants using either of the two biochemical pathways through caffeine synthase or xanthine methyltransferase-like enzymes, as exhibited in chocolate, citrus, and guarana plants [151]. Based on coding sequences, xanthine NMTs were determined to belong to the SABATH (salicylic acid, benzoic acid, theobromine synthase) family of methyltransferases, which exclusively utilize the "proximity and desolvation effects" [1,5], rather than general acid-base catalysis or cation-dependent mechanisms. Since SABATH methyltransferases generally methylate oxygen atoms, it is likely that a recent change in the function of xanthine alkaloid-producing species occurred [16]. A recent study hypothesized that despite stemming from different lineages, xanthosine methyltransferases have convergent evolutionary histories [16,151]. A number of NMTs involved in monoterpene indole alkaloid biosynthetic pathways have been characterized, including those involved in vindoline production in Catharanthus roseus [49], N-methylation of picrinine from the Apocynaceae family [51], and ajmaline synthesis [52]. NMTs have contributed significantly to the diversification of benzylisoquinoline alkaloids (BIA), and several NMTs have been characterized in nine Ranunculales species [16].
S-directed methyltransferases (SMTs) are generally involved in the production of volatile halogen and sulfur compounds as well as thiocyanate [1]. A few examples of characterized SMTs in plants have been reported, including those from C. roseus [132]. A thiocyanate methyltransferase found in Arabidopsis thaliana that synthesizes methylthiocyanate [124] was later structurally analyzed [22]. Halide/bisulfide methyltransferase activity was also detected in the leaves of Brassica oleracea [155]. Some MTs exhibit high specificity for halides or bisulfides, known as halide ion methyltransferases (HMTs) and halide/thiol methyltransferases (HTMTs). HMTs and HTMTs are found in coastal trees, grasses, and several agricultural plants, with particularly high levels of activity in Raphanus sativus (daikon radish), Oryza sativa (paddy rice) Triticum aestivum (wheat), and Cyathea lepifera A number of NMTs involved in monoterpene indole alkaloid biosynthetic pathways have been characterized, including those involved in vindoline production in Catharanthus roseus [49], N-methylation of picrinine from the Apocynaceae family [51], and ajmaline synthesis [52]. NMTs have contributed significantly to the diversification of benzylisoquinoline alkaloids (BIA), and several NMTs have been characterized in nine Ranunculales species [16].
S-directed methyltransferases (SMTs) are generally involved in the production of volatile halogen and sulfur compounds as well as thiocyanate [1]. A few examples of characterized SMTs in plants have been reported, including those from C. roseus [132]. A thiocyanate methyltransferase found in Arabidopsis thaliana that synthesizes methylthiocyanate [124] was later structurally analyzed [22]. Halide/bisulfide methyltransferase activity was also detected in the leaves of Brassica oleracea [155]. Some MTs exhibit high specificity for halides or bisulfides, known as halide ion methyltransferases (HMTs) and halide/thiol methyltransferases (HTMTs). HMTs and HTMTs are found in coastal trees, grasses, and several agricultural plants, with particularly high levels of activity in Raphanus sativus (daikon radish), Oryza sativa (paddy rice) Triticum aestivum (wheat), and Cyathea lepifera (fern) [125]. These HMT and HTMT genes were later cloned [156,157]. HMTs have also been biochemically characterized in broccoli with high substrate specificity toward homocysteine [158]. Lyi and colleagues also reported a selenocysteine methyltransferase that produces Se-methylselenocysteine [135]. Interestingly, these two genes play roles in sulfur and selenium metabolism in broccoli. Another SAM-dependent HTMT was cloned from Raphanus, and upon biochemical testing, exhibited high specificity for iodide, bisulfide, and thiocyanate [125]. Plant SMTs have also been characterized in the production of brassinin in cruciferous vegetables [131,159]. A comprehensive list of biochemically characterized PNPMTs with some structurally characterized features is detailed in Table 1.

The Identification of Plant Natural Product Methyltransferase (PNPMT) Genes
All known Class I PNPMTs consist of an alternating α-helix/ß-strand structure with associated monomeric molecular masses of 25 to 55 kDa [1]. Bioinformatic analyses to identify putative PNPMT genes are primarily based on a series of conserved motifs, with heavy reliance on the core Rossmann fold [1,25]. Within the Rossmann fold, a nine-residue amino acid incorporating a glycine-rich structure "GxGxG" signature sequence corresponding to a SAM-binding motif is found in SAM-dependent MTs (red boxes in Figure 3). Although these consensus sequences (V/I/L)(L/V)(D/E)(V/I)G(G/C)G(T/P)G [160] contain substitutions, the amino acid sequences are typically better aligned within the same subclasses (cation-dependent, cation-independent, SABATH, BIA). produces Se-methylselenocysteine [135]. Interestingly, these two genes play roles in sulfur and selenium metabolism in broccoli. Another SAM-dependent HTMT was cloned from Raphanus, and upon biochemical testing, exhibited high specificity for iodide, bisulfide, and thiocyanate [125]. Plant SMTs have also been characterized in the production of brassinin in cruciferous vegetables [131,159]. A comprehensive list of biochemically characterized PNPMTs with some structurally characterized features is detailed in Table 1.

The identification of Plant Natural Product Methyltransferase (PNPMT) Genes
All known Class I PNPMTs consist of an alternating α-helix/ß-strand structure with associated monomeric molecular masses of 25 to 55 kDa [1]. Bioinformatic analyses to identify putative PNPMT genes are primarily based on a series of conserved motifs, with heavy reliance on the core Rossmann fold [1,25]. Within the Rossmann fold, a nine-residue amino acid incorporating a glycine-rich structure "GxGxG" signature sequence corresponding to a SAM-binding motif is found in SAM-dependent MTs (red boxes in Figure  3). Although these consensus sequences (V/I/L)(L/V)(D/E)(V/I)G(G/C)G(T/P)G [160] contain substitutions, the amino acid sequences are typically better aligned within the same subclasses (cation-dependent, cation-independent, SABATH, BIA).  Table 1. Plant NPMTs include cation-dependent SbCCoAOMT, cation-independent MtHI4′OMT, SABATH CrLAMT, and CcDXMT, and BIA TfPavNMT, in comparison to the bacterial MT within Class I bacterial LiOMT (Leptospira interrogans, WP_000087781), and non-Class I bacterial SvAviRbMT (Streptomyces viridochromogenes, WP_003998250). The red boxes indicate the glycine-rich Rossmann fold consensus sequences with 0-3 mismatches. The amino acid residues highlighted in blue are highly conserved, those in green are conserved, and those in yellow are moderately conserved. "*" represents every 10 th amino acid after each numerical indicator.  Table 1. Plant NPMTs include cation-dependent SbC-CoAOMT, cation-independent MtHI4 OMT, SABATH CrLAMT, and CcDXMT, and BIA TfPavNMT, in comparison to the bacterial MT within Class I bacterial LiOMT (Leptospira interrogans, WP_000087781), and non-Class I bacterial SvAviRbMT (Streptomyces viridochromogenes, WP_003998250). The red boxes indicate the glycine-rich Rossmann fold consensus sequences with 0-3 mismatches. The amino acid residues highlighted in blue are highly conserved, those in green are conserved, and those in yellow are moderately conserved. "*" represents every 10th amino acid after each numerical indicator.
Amino acid sequences of PNPMTs from separate subclasses of cation-dependent, cation-independent, SABATH, and BIA MTs were chosen from the list of biochemically characterized plant natural product MTs (Table 1) to construct a phylogenetic tree (Figure 4). The genes within the same subclasses share high sequence similarity, while several of these highly homologous MTs have been reported to be involved in the biosynthesis of similar classes of metabolites, including clusters of BIA OMTs/NMTs and alkaloid biosynthetic genes (Ca10OMT, IpeOMT1, Vm16OMT, Cr16OMT) [16,48]. Sequence analysis and the phylogenetic results confirmed the functional classification of cation-dependent, cationindependent, and SABATH families. Class I cation-dependent bacteria OMT is included to compare with plant cation-dependent OMTs (Figure 4). Amino acid sequences of PNPMTs from separate subclasses of cation-dependent, cation-independent, SABATH, and BIA MTs were chosen from the list of biochemically characterized plant natural product MTs (Table 1) to construct a phylogenetic tree (Figure 4). The genes within the same subclasses share high sequence similarity, while several of these highly homologous MTs have been reported to be involved in the biosynthesis of similar classes of metabolites, including clusters of BIA OMTs/NMTs and alkaloid biosynthetic genes (Ca10OMT, IpeOMT1, Vm16OMT, Cr16OMT) [16,48]. Sequence analysis and the phylogenetic results confirmed the functional classification of cation-dependent, cation-independent, and SABATH families. Class I cation-dependent bacteria OMT is included to compare with plant cation-dependent OMTs (Figure 4).

Structural Biochemistry of PNPMTs
Protein X-ray crystallography has allowed for the extensive structural elucidation of numerous Class I methyltransferases. The resulting X-ray crystal structures have offered great insight into the general structures of these methyltransferases and have determined that they remain highly consistent, despite the primary amino acid sequences having low similarity [162]. Nevertheless, the primary amino acid sequences are still necessary to conduct phylogenetic analysis as they yield clarity into the evolutionary relationships of all PNPMTs. Furthermore, increased genomic data and the availability of coding sequences facilitate the molecular characterization or determinants of function [16,151]. For example, characterized active sites that accept xanthine have distinguishing features that promote attachment with xanthine molecules [152,154]. In particular, the hydrophobic pocket and specific residues affecting substrate specificity, encourage such xanthine attachment, later confirmed by site-directed mutagenesis [16,151,154,163]. In plant OMTs, structural analysis often involves dimers. Most OMTs form homodimers [92], though a BIA OMT was reported to form a heterodimer [164]. Typically, a dimerization interface is present as several helices actively participate in substrate binding. This dimerization interface also constitutes a hydrophobic "back wall" [92], though its participation in substrate binding varies among plant OMTs [16]. The analysis of primary amino acid sequences revealed that the N-terminus domain contributes to the dimerization with partial involvement in substrate-binding within the hydrophobic pocket, while the C-terminus domain is involved in SAM-binding via the Rossmann fold.
Class I PNPMTs are characterized by a conserved domain known as the Rossmann fold. The Rossmann fold is defined by a seven-stranded ß-sheet, commonly ordered 3 2 1 4 5 7 6, with the seventh strand in an antiparallel position to the remaining strands and three α-helices on each side of the ß-sheet [17,165].
Recognized as the defining features of all SAM-dependent methyltransferases, six sequentially ordered motifs (I-VI) either contribute to the formation of the SAM-binding pocket or have interactions with SAM itself. [1,18,25,166] (Figure 5). Motif I consists of ß-strand I and the adjoining loop, and located at the N-terminus is a glycine-rich sequence (GxGxG) with one or two glycines interacting with SAM [18]. Generally, this glycine-rich sequence remains conserved throughout Class I, but there is a minority of cases where alanine residues replaced the glycines [18]. Motif II consists of ß-strand II and the following helix, and there is either an aspartate or glutamate residue located at the C-terminus [18,166]. The helix is structurally conserved and directly contributes to the SAM-binding pocket [18]. Motif III is involved in SAM binding. This motif possesses a hydrophilic amino acid at the N-terminus of ß-strand III, and a partially conserved glycine residue at the C-terminus [18]. Motif IV is composed of ß-strand IV and the adjoining loops. Located at the C-terminus is a partially conserved acidic residue [1]. Motif V consists of a helix following strand IV, although this motif does not typically interact with SAM [18]. However, in a few cases, this motif may act as a site for hydrophobic side chains, securing SAM's adenine moiety [166]. Finally, motif VI, very uncommonly interacting with SAM, is characterized by strand V and its preceding loop, as well as a conserved glycine residue, followed by two hydrophobic residues at the beginning of the strand [18]. Interestingly, among all subclasses, the residues of the motifs were highly conserved, although SAM's position with its binding pocket varied depending on the subclass [18,25,166] (Figure 5).  Table 1. The colored beta sheets and alpha helices represent Motifs I-VI, which make up the characteristic MT Class I Rossmann fold. The red strand represents Motif I, the light blue strand and helix represents Motif II, the orange strand  Table 1. The colored beta sheets and alpha helices represent Motifs I-VI, which make up the characteristic MT Class I Rossmann fold. The red strand represents Motif I, the light blue strand and helix represents Motif II, the orange strand represents Motif III, the green strand represents Motif IV, the yellow helix represents Motif V, and the purple strand represents Motif VI. The darker blue structure is either SAM or its demethylated form, SAH. The lighter blue structure is the substrate.
The Rossmann fold is a multifunctional domain of the PNPMT and other Class I methyltransferases alike-not only is the Rossmann fold able to house the methyl donor SAM, but it also has influences on the substrate-binding pocket, and thus, the orientation of the methyl acceptor atom [5]. While the Rossmann-fold domain is widely conserved across all Class I methyltransferases, structural tweaks allow for some degree of promiscuity. In contrast to DNA methyltransferases, which are designed to identify the general features of macromolecular substrates due to their widely conserved active sites [19], PNPMTs generally exhibit higher levels of specificity [1]. A number of important amino acid residues that contribute to this specificity have been identified via structural-based analysis and confirmed by site-directed mutagenesis in several cases, including in caffeoyl-CoA methyltransferases [103], chalcone and isoflavone methyltransferase [92], and benzylisoquinoline [16,167].
In some alkaloid biosynthesis studies, heterodimers have been suggested to participate in the catalysis of new substrates. From Thalictrum tuberosum, the heterodimers of four OMTs accepted substrates, such as catechols, hydroxycinnamates, and alkaloids, exhibit far more promiscuity than their homodimer counterparts [47]. A missing step in noscapine biosynthesis in Papaver somniferum was filled with a then newly discovered heterodimer consisting of PsSOMT2 and PsSOMT/Ps6OMT [164]. Due to the degree of promiscuity within alkaloid heterodimers, understanding their functionalization has the potential to uncover new biochemical processes [136]. Overall, the advances in the structural biochemistry of PNMPTs have highlighted the degree of specificity that both conserves the SAM-binding pocket and provides structural determinants of functions and their involvements in plant natural product biosynthesis.

Engineering Plant Natural Product Methyltransferases (PNPMTs)
Methyltransferases are attractive enzymes to manipulate for the purpose of altering metabolic pathways. The regioselectivity of PNPMTs is relatively more favorable in comparison to organic synthesis in which protecting groups are often added to produce specific methylated products with a further disadvantage of low yield [1]. The structure-function analysis has been extensively applied for enzyme engineering, which involves mutagenesis approaches to substitute their amino acids to obtain the desired methylated products. For example, a poplar flavonoid-OMT mutant with amino acid Asp257Gly was reported to methylate the 3-hydroxyl group compared to its native selectivity in methylating the 7-hydroxyl group [168]. In another case, variants of Vitis vinifera resveratrol OMT with point mutations were generated and resulted in modified native substrate specificity to increase the yield of pinostilbene (monomethylated resveratrol), rather than to produce dimethylated resveratrol [169]. In addition to shifting in the regiospecificity of methylation and/or modified substrate specificity, the promiscuity of PNPMTs has been proven to be more versatile, as the biosynthetic pathway can be altered with a possibility of yielding an unknown molecule via de novo pathways, and thus, increasing the diversity of plant natural products [1,170]. In most cases, beneficial mutations are identified by in-silico methods combined with a high-throughput screening assay to accelerate the engineering process. Zhao and colleagues rationally designed Peucedanum praeruptorum bergaptol O-methyltransferase to increase the production of bergapten with improved pharmacological properties, especially for increasing pigmentation levels [171]. One Val320Ile mutant protein was particularly reported to be promising for further applications in treating depigmentation disorder [171]. In the absence of crystal structures, computer-guided analysis to analyze the active site and the catalytic mechanisms of plant natural product methyltransferases can provide alternative manufacturing strategies via metabolic engineering approaches in microbial systems.
A number of de novo biosynthesis pathways that produce methylated natural products have been successfully designed in microbial systems. For instance, dimethylated resveratrol (pterostilbene) was produced in an engineered Escherichia coli strain expressing Arabidopsis caffeic acid/resveratrol O-methyltransferase. The intracellular pool of L-tyrosine, a precursor molecule in the stilbene biosynthetic pathway, was improved in this engineered strain grown in L-methionine-rich media to increase the SAM supply [172]. This approach overcomes a central issue in engineering methyltransferases, which is cofactor regeneration, as SAH is a known inhibitor for methyltransferase activities and its accumulation can be toxic for the microbial hosts [1,170]. Although the incorporation of native SAM recycling pathways involving multiple enzymatic reactions can be challenging to integrate within the host systems, deregulation of the SAM pathway was shown to successfully increase the production of O-methylated products. MetJ or Met repressor in bacteria is often a target for silencing or deletion in order to regulate methionine production. Silencing of the MetJ gene via CRISPRi enhances the production of O-methylated anthocyanin in Escherichia coli [173]. Similarly, deleting MetJ was also effective in shuttling the pool of SAM in an engineered Escherichia coli strain that overexpressed methionine biosynthetic pathway genes, especially in improving vanillate production [174].
The engineering of methyltransferases has advanced beyond its ability in transferring a methyl group. The development of SAM analogs has allowed for the diversification of small molecules. In particular, a mutated halide methyltransferase from Arabidopsis thaliana efficiently transferred ethyl-, propyl-, and allyl-moieties to SAH in order to produce SAM analogs with the respective alkyl groups [175]. Recently, efforts in developing SAM analogs to diversify active biomolecules have been accelerated after the discovery of the naturally occurring carboxy-S-adenosylmethionine pathway involved in tRNA modification [176]. Interestingly, coclaurine N-methyltransferase from Coptis japonica was selected to test the possibility of plant methyltransferases catalyzing the reaction using an alternative cofactor of carboxy-SAM [177]. By comparing its SAM-binding residues with the crystal structure tRNA carboxymethyltransferase, which highly prefers carboxyl-SAM, coclaurine N-methyltranferase was successfully engineered to accept carboxy-SAM over SAM to produce carboxymethylated tetrahydroisoquinoline [177]. While the use of SAM analogs in the diversification of plant small molecules seems likely to be effective at smaller scales, more innovations are still required to increase efforts in their regeneration and efficient synthesis.
Although microbial systems remain excellent hosts for engineering plant methyltransferases, plant hosts are still relevant to consider, especially in obtaining natural products for industrial and specific agricultural applications. Engineered monolignol 4-O-methyltransferase, for example, was useful for reducing lignin content for the effective production of liquid biofuels, a desired utilization of cellulosic fiber, while it allows for the diversification of possible active compounds in transgenic plants [178]. Engineered plant host systems also provide additional platforms for extending the natural product biosynthetic pathways beyond methylated metabolites, as novel compounds with unique biological activities might be accumulated. Overall, the versatility of plant natural product methyltransferases can inspire future efforts in enzyme and metabolic engineering for improved capabilities in widening the diversity of bioactive compounds with pharmacological properties.

Conclusions and Future Directions
The structural elucidation of PNPMTs has heavily augmented our knowledge of how SAM-dependent methyltransferases contribute to the diversification of small molecules in the plant kingdom. The increasing number of structure determinations, along with the elucidation of their architecture, has significantly provided insights into their molecular evolution. A comprehensive representation of biochemically characterized PNPMTs demonstrates not only the structural diversity of Class I of SAM-dependent methyltransferases but also provides a point of reference to search for other PNPMTs that have yet to be discovered. With the continuous advances in next-generation sequencing and progress in -omics-based technologies, we anticipate more PNPMT genes will be discovered. While only less than 20% of the PNPMTs that we compiled have been structurally elucidated, we expect more reports on the protein crystal structures of PNPMTs. This knowledge of structural biochemistry will benefit the development of machine learning algorithms for more advanced prediction of protein folding, gene annotation, and functional validation. By delineating the substrate specificity and structural attributes of PNPMTs, we have provided an overview of the recently expanded strategies to facilitate the development of enzyme and metabolic engineering. We also anticipate that all these advancements will give rise to further diversification and increased production of methylated natural products with enhanced pharmacological properties as human therapeutics. Ultimately, the complete knowledge of molecular and structural determinants of PNPMT function will reveal important aspects of the catalytic mechanism and inform metabolic engineering efforts, thus generating novel biocatalysts for wide applications in biotechnology.