Using rice (Oryza sativa) as a model system, FNSII was identified to be the primary enzyme generating the flavone nucleus for tricin biosynthesis in grasses (Figure 5; Lam et al., 2014). Recombinant OsFNSII catalyzes direct conversions of flavanones, i.e., naringenin and eriodictyol, into apigenin and luteolin, respectively, in vitro (Brazier-Hicks and Edwards, 2013; Lam et al., 2014). In addition, over-expression of OsFNSII in Arabidopsis resulted in the accumulation of flavones (apigenin, luteolin, and chrysoeriol) O-glycosides which are normally not present in wild-type plants (Lam et al., 2014). Further analyses of the rice OsFNSII knockout mutant revealed substantial depletion of soluble tricin O-conjugates as well as tricin-lignin in cell walls, demonstrating the direct and predominant involvement of OsFNSII in the generation of both soluble and lignin-integrated tricin in rice (Lam et al., 2014, 2017). Moreover, the OsFNSII mutant accumulated soluble naringenin but not the other flavanones, e.g., eriodictyol (Lam et al., 2014), and generated altered lignins incorporated with naringenin (Lam et al., 2017), indicating that the in planta substrate of OsFNSII is primarily naringenin.

OsFNSII, or CYP93G1, is a P450 enzyme belonging to the same CYP93G subfamily as OsF2H, or CYP93G2. Using naringenin as a common substrate, OsFNSII and OsF2H are the branch-point enzymes for the biosynthesis of tricin O-conjugates and flavone C-glycosides, respectively (Figures 3, 5). Phylogenetic analysis revealed that OsFNSII and OsF2H form two separate clades, each containing highly conserved sequences from the grass family (Figure 6A; Lam et al., 2017). Hence, sub-functionalization of CYP93G members probably preceded lineage divergence within Poaceae, resulting in the widespread distribution of the two classes of flavone-derived metabolites in grasses today. It is noteworthy that grass FNSIIs and F2Hs have a different phylogenetic origin from dicot FNSIIs and F2Hs, all of which exclusively belong to the CYP93B subfamily (Figure 6A; Kitada et al., 2001; Martens and Mithöfer, 2005; Zhang et al., 2007; Fliegmann et al., 2010; Wu et al., 2016; Zhao et al., 2016; Jiang et al., 2019). Noteworthily, grass species do not contain any CYP93B members and dicots do not have CYP93G members (Du et al., 2016).

Functionally, redundant enzymes other than FNSII are likely to be involved in tricin biosynthesis in grasses. For example, the rice OsFNSII mutant still accumulated soluble tricin and other flavones in anthers albeit at reduced levels compared with wild type (Wang et al., 2020a), while it shows substantial depletion of soluble tricin O-conjugates and tricin-lignin in vegetative tissues (Lam et al., 2014, 2017). In fact, two rice FNSIs were shown to catalyze the conversion of naringenin into apigenin in vitro (Kim et al., 2008; Lee et al., 2008b). In addition, maize possesses an FNSI (ZmFNSI-1) which shows in vitro FNS activities and results in the accumulation of flavones when over-expressed in Arabidopsis (Ferreyra et al., 2015; Righini et al., 2019).

In the plant kingdom, 3',5'-substituted flavonoids are patchily distributed, because F3'5'Hs, the enzymes responsible for catalyzing 5'-hydroxylation, are only present in isolated plant lineages (Tanaka and Brugliera, 2013). This is in contrast to the ubiquitous nature of flavonoid 3'-hydroxylases (F3'H; exclusively members of the CYP75B subfamily) that gives rise to the prevalence of 3'-substituted flavonoids (Tanaka and Brugliera, 2013). There have been strong interests for the investigation of F3'5'Hs as they are the key enzymes for the generation of delphinidin-derived anthocyanins, which confer blue or violet coloration in plant tissues, such as flowers and fruits (Tanaka and Brugliera, 2013). For ornamental purposes, transgenic expression of foreign F3'5'Hs has been employed to engineer novel blue or violet color in roses (Rosa hybrida), chrysanthemums (chrysanthemum morifolium), and carnations (Dianthus caryophyllus), all of which naturally lack delphinidin-derived anthocyanins (Katsumoto et al., 2007; Brugliera et al., 2013; Noda et al., 2013; Tanaka and Brugliera, 2013).

The canonical F3'5'Hs are CYP enzymes belonging to the CYP75A subfamily (Tanaka and Brugliera, 2013). Apparently, CYP75A-encoding genes have been lost repeatedly or became non-functional in many lineages during evolution (Tanaka and Brugliera, 2013). In rice, the only CYP75A member (CYP75A11) did not show any F3'5'H functions in in vitro enzyme assays or in CYP75A11 over-expressing transgenic Arabidopsis plants (Lam et al., 2015). On the other hand, a rice CYP75B member (CYP75B4) solely contributes to the 5'-hydroxylation activity during tricin biosynthesis, as evidenced by in planta metabolite analysis. For example, the CYP75B4 T-DNA knockout mutant is completely devoid of soluble selgin and tricin O-conjugates in vegetative tissues (Lam et al., 2015) and tricin-lignin in cell walls (Lam et al., 2019a). In addition, transgenic Arabidopsis co-expressing CYP75B4 and OsFNSII accumulates O-conjugates of selgin and tricin (Lam et al., 2015). Meanwhile, apigenin produced by OsFNSII using naringenin as a preferred in planta substrate was initially expected to undergo sequential B-ring hydroxylations to form tricetin as a tricin precursor. However, while CYP75B4 3'-hydroxylates apigenin to luteolin, it fails to 5'-hydroxylate luteolin to tricetin (Lam et al., 2015). Instead, CYP75B4 catalyzes 5'-hydroxylation of chrysoeriol to produce selgin (Lam et al., 2015). Chrysoeriol could be generated by 3'-O-methylation of luteolin, whereas selgin could undergo 5'-O-methylation to generate tricin. The flavonoid B-ring O-methylation reactions are known to be catalyzed by several O-methyltransferases in rice (see B-ring O-methylations below). Collectively, tricin biosynthetic pathway in rice has been re-established as: naringenin → apigenin → luteolin → chrysoeriol → selgin → tricin (Figure 5; Lam et al., 2015, 2019a). Meanwhile, chrysoeriol O-linked derivatives accumulates in rice vegetative tissues (Galland et al., 2014; Lam et al., 2015, 2019a; Eloy et al., 2017), whereas tricetin and its O-linked derivatives (e.g., O-conjugates) are rarely detected in grasses (Galland et al., 2014; Lam et al., 2015, 2019a; Eloy et al., 2017), supporting that chrysoeriol, instead of tricetin, is an intermediate along the tricin biosynthetic pathway.

CYP75B4 is a flavone-specific bifunctional B-ring hydroxylase in rice. It displays very weak 3'-hydroxylase activity toward naringenin while converting apigenin to luteolin readily (Lam et al., 2015; Park et al., 2016). In addition, its 5'-hydroxylation activity was restricted to chrysoeriol, but not any other 3'-methoxylated or 3'-hydroxylated flavonoids (Lam et al., 2015). Hence, the enzyme is now dedicated as apigenin 3'-hydroxylase/chrysoeriol 5'-hydroxylase (A3'H/C5'H). The dual catalytic activities have also been demonstrated in the highly conserved orthologs in sorghum (CYP75B97) and switchgrass (CYP75B11; Figure 6B), indicating that similar enzymology and intermediates were recruited for tricin biosynthesis in the grass family (Lam et al., 2019a). Further evidence indicated that the 3'-hydroxylation reaction (apigenin → luteolin) for tricin biosynthesis is also predominantly contributed by A3'H/C5'H. Thus, the rice CYP75B4 mutant accumulates elevated amounts of soluble apigenin metabolites along with the incorporation of apigenin into cell wall lignins (Lam et al., 2015, 2019a). On the other hand, CYP75B3, the only other CYP75B member in rice, is a canonical F3'H which catalyzes in vitro 3'-hydroxylation of a wide range of flavonoids including apigenin (Shih et al., 2008; Lam et al., 2015, 2019a; Park et al., 2016). However, CYP75B3 loss-of-function mutants are preferentially deficient in 3'-substituted flavone (luteolin and chrysoeriol) C-glycosides, while their production of soluble and lignin-integrated tricin remains unaffected (Lam et al., 2019a). Apparently, CYP75B3 primarily functions together with OsF2H along the separate biosynthetic pathway for flavone C-glycosides (Figure 3).

The highly conserved A3'H/C5'Hs in grasses are distinctive from other F3'5'Hs with regard to their phylogeny and catalytic properties. They are phylogenetically distant from CYP75A F3'5'Hs and were likely recruited through neofunctionalization of an ancestral CYP75B F3'H protein (Figure 6B). Similarly, several Asteraceae species had acquired CYP75B F3'5'Hs independently through convergent evolution, leading to delphinidin-derived anthocyanin pigmentation (Seitz et al., 2006; Seitz et al., 2015). In addition, the grass A3'H/C5'Hs are substrate specific for both 3'-hydroxylation (apigenin) and 5'-hydroxylation (chrysoeriol), while CYP75A and Asteraceae CYP75B F3'5'Hs could utilize a variety of non-substituted, 3'-hydroxylated and 3'-methoxylated flavonoids as substrates. Intriguingly, the unique catalytic properties of A3'H/C5'Hs are reminiscent of the bifunctional phenylpropanoid meta-hydroxylase (CYP788A1) required for syringyl (S)-lignin biosynthesis in the spikemoss Selaginella moellendorffii. CYP788A1 is involved in both 3- and 5-hydroxylations of phenylpropanoids, but it could only catalyze 5-hydroxylation after 3-O-methylation (Weng et al., 2008, 2010).

Several cation-independent OMTs in grasses were found to catalyze in vitro O-methylation of flavones in grasses (Kim et al., 2006; Lin et al., 2006; Zhou et al., 2006, 2008, 2009). Interestingly, these enzymes have been annotated as caffeic acid O-methyltransferases (COMT) or 5-hydroxyconiferaldehyde O-methyltransferases (CAldOMT) as they also show in vitro O-methylation activities toward 5-hydroxyconiferaldehyde, 5-hydroxyferulic acid, and caffeic acid, which are intermediates in the monolignol pathway; hence, they are also involved in S-lignin biosynthesis (Figure 7A; Collazo et al., 1992; Piquemal et al., 2002; Ma and Xu, 2008; Sattler et al., 2012; Koshiba et al., 2013).

Recently, knockout and knockdown mutant analyses have demonstrated that grass COMT/CAldOMTs are actually bifunctional enzymes required for both tricin and S-lignin biosynthesis (Fornalé et al., 2016; Eudes et al., 2017; Daly et al., 2019; Lam et al., 2019b). Rice and sorghum deficient in COMT/CAldOMT accumulated reduced levels of soluble tricin but increased levels of selgin (mono-methoxylated) and luteolin (non-methoxylated) when compared with wild-type controls (Lam et al., 2015; Eudes et al., 2017). In addition, maize, rice, and sorghum plants deficient in COMT/CAldOMT were depleted in both tricin-lignin and S-lignin (Fornalé et al., 2016; Eudes et al., 2017; Lam et al., 2019b). Apparently, the highly conserved grass COMT/CAldOMT orthologs (Figure 6C) have likely evolved dual catalytic functions for the two parallel biosynthetic pathways of flavonoids and monolignols, contributing to the widespread occurrence of soluble and lignin-integrated tricin metabolites in the grass family nowadays.

Based on the revised tricin biosynthetic pathways and the new findings in the COMT/CAldOMT-deficient grass plants, the catalytic activities of COMT/CAldOMTs were re-examined. Recombinant COMT/CAldOMTs in rice and sorghum were found to catalyze 3'-O-methylation of luteolin and 5'-O-methylation of selgin (Kim et al., 2006; Lin et al., 2006; Zhou et al., 2006; Eudes et al., 2017; Lam et al., 2019b), which are the substrates of COMT/CAldOMTs in the tricin biosynthetic pathway (Figure 5). Meanwhile, rice OsCAldOMT1 shows comparable catalytic efficiencies toward selgin and 5-hydroxyconiferaldehyde, which are the substrates of COMT/CAldOMTs in tricin and monolignol biosynthetic pathway, respectively (Figures 5, 7A), further suggesting the bifunctional roles of COMT/CAldOMTs in tricin and monolignol biosynthesis in grasses (Lam et al., 2019b).

Functionally redundant OMTs other than COMT/CAldOMTs appear to be present for the biosynthesis of tricin in grasses as tricin-derived metabolites, including tricin-lignin, are not completed depleted in the COMT/CAldOMT loss-of-function mutants in maize, sorghum, and rice (Lam et al., 2015, 2019b; Fornalé et al., 2016; Eudes et al., 2017). In fact, several cation-dependent caffeoyl-CoA O-methyltransferase (CCoAOMT)-related enzymes could catalyze 3',5'-O-methylation using various flavone substrates (Lee et al., 2008a), but their involvement in tricin biosynthesis in planta requires further investigations.

Based on the types of soluble tricin metabolites detected in grasses, O-glycosylations and O-conjugations with monolignols and their acylated derivatives represent the predominant structural modifications of tricin (Dong et al., 2014; Lan et al., 2016a; Eloy et al., 2017; Peng et al., 2017). These modifications occur after the formation of tricin aglycone (Hong et al., 2007; Jiang et al., 2016; Lan et al., 2016a).

O-Glycosylations of flavonoids are usually catalyzed by uridine diphosphate (UDP)-dependent glycosyltransferases (UGT; family 1 glycosyltransferases 1; GT1; Ko et al., 2006; Yonekura-Sakakibara and Hanada, 2011; Kim et al., 2015), which utilize UDP sugars as sugar donors (Yang et al., 2018). A number of UGTs from rice (Ko et al., 2006, 2008; Hong et al., 2007; Luang et al., 2013; Chen et al., 2014; Peng et al., 2017) and wheat (Shi et al., 2020) are capable of catalyzing the conjugation of sugars, usually glucose, to one or multiple hydroxyl groups of tricin in vitro and/or when over-expressed in transgenic plants. Single-nucleotide polymorphisms (SNPs) in several putative UGTs were also found to be directly associated with the variations of flavone O-glycoside accumulation in different natural cultivars and/or recombinant inbred lines of rice (Chen et al., 2014; Dong et al., 2014; Peng et al., 2017; Li et al., 2019) and wheat (Shi et al., 2020). The different O-glycosylations could enhance solubility and stability, and might be involved in regulating storage, transport, and detoxification of tricin (Yonekura-Sakakibara and Hanada, 2011).

In addition to sugars, tricin conjugates with monolignols and their derivatives, leading to the formation of soluble tricin-oligolignols along with insoluble tricin-lignin in the cell walls. The soluble tricin-oligolignols in grasses have been found to be either optically active (Wenzig et al., 2005; Xiong et al., 2011) or inactive (racemic; Lan et al., 2016a). The optically active tricin-oligolignols, which have been often referred to as “flavonolignans” (Begum et al., 2010; Chambers et al., 2015; Csupor et al., 2016), may be formed by oxidative radical coupling of tricin with monolignols or their derivatives with the assistance of dirigent proteins, similar to the biosynthesis of lignans (Davin and Lewis, 2003; Umezawa, 2003; Paniagua et al., 2017), in which dirigent proteins serve as auxiliary proteins for guiding the regioselective and stereoselective coupling of phenoxy radicals from monolignols and their analogs. For example, the absolute configuration of a diastereomeric pair of β–O–4 neolignan-type flavonolignans, threo-(−)-guaiacylglycerol-β-tricin ether [(−)-salcolin A], and erythro-(−)-guaiacylglycerol-β-tricin ether [(−)-salcolin B] isolated from Sinocalamus affinis (Poaceae) were determined as 7''S,8''S and 7''R,8''S, respectively (Xiong et al., 2011). This strongly suggests that the coupling between tricin and coniferyl alcohol radicals to form 4'–O–8'' bond proceeds enantioselectively, probably mediated by a dirigent protein, giving rise to the optically active quinonemetide, which are then attacked by water non-stereoselectively, giving rise to both (−)-(7''S,8''S)-salcolin A and (−)-(7''R,8''S)-salcolin B (Figure 8). This is in line with the recent findings that a dirigent protein, AtDIR12/AtDP1, was involved in the formation of arylglycerol-β-aryl ether (β–O–4) type neolignans in Arabidopsis (Yonekura-Sakakibara et al., 2021). However, from Avena sativa, (−)-salcolin A and (+)-salcolin B were isolated (Wenzig et al., 2005). In this case, the diastereomers should have opposite absolute configuration at 8'' position, forming (−)-(7''S,8''S)-salcolin A and (+)-(7''S,8''R)-salcolin B (Figure 8). During their formation, the radical coupling should afford racemic quinonemethide in terms of 8'' position, and the following water addition at 7'' position should be diastereoselective to give rise to the optically active diastereomers (Wenzig et al., 2005). On the other hand, optically inactive tricin-oligolignols are generated solely by radical coupling (Figure 7B) and are considered to exist at least partially as the precursors for the generation of tricin-lignin polymers (see Tricin-lignin formation below; Lan et al., 2016a).

Tricin is incorporated into lignin polymers in grass cell walls by radical coupling (Lan et al., 2015), essentially the same way lignification takes place solely with monolignols (coniferyl alcohol, sinapyl alcohol, and p-coumaryl alcohol) in typical non-grass vascular plants (i.e., gymnosperms, dicots, and non-grass monocots). The compatibility of tricin with radical coupling was demonstrated by biomimetic oxidations of tricin with monolignols using peroxidase/hydrogen peroxide and silver (I) oxide as oxidants (Lan et al., 2015). Tricin was found to cross-couple to monolignols exclusively via the 4'–O–β-coupling mode (Figure 7B; del Río et al., 2012; Lan et al., 2015), probably because the radical from the 4'-hydroxyl group of tricin is more stabilized than the other possible radicals as supported by a density functional theory study (Elder et al., 2020). Thus, in plant cell walls, it is expected that tricin is first oxidized by phenol oxidases, presumably laccases (LAC) and/or peroxidases (PRX; Figure 7A; Tobimatsu and Schuetz, 2019), and then coupled with monolignol radicals or acylated monolignol radicals to form tricin-(4'–O–β)-linked phenylpropane units in the lignin polymers (Figure 7B). As tricin is unable to undergo dehydrodimerization, and it does not cross-couple directly with growing lignin polymers, tricin predominantly incorporates into the starting ends of the final lignin polymer chains (Lan et al., 2015). Thus, tricin is expected to serve as a nucleation site for lignification (Lan et al., 2015; Berstis et al., 2021).

Lignin-integrated tricin content in grasses was estimated to be around 0.5–7mg/g whole cell wall or 2–33mg/g lignin by thioacidolysis (Lan et al., 2016b). These contents are several folds higher than extractable tricin content (Lan et al., 2016b), suggesting that the majority of tricin synthesized in grasses is incorporated into lignin polymers in cell walls.

In contrast to their prevalence in grasses and other monocot lineages, tricin-derived metabolites are only sporadically distributed in dicots. Metabolomics studies have reported their occurrences in several dicot lineages, spanning from basal dicots, like individual Ranunculus spp. (Li et al., 2005; Aslam et al., 2012), to two core dicot lineages: rosids [e.g., Agelaea pentagyna (family: Connaraceae), Medicago legumes and Trigonella foenum-graecum (family: Fabaceae); Kuwabara et al., 2003], and asterids [e.g., Artemisia vulgaris (family: Asteraceae), Leucas cephalotes (family: Lamiaceae), and Lonicera japonica (family: Caprifoliaceae)] (Miyaichi et al., 2006). Meanwhile, tricin-lignin is only detected in leaves of alfalfa, albeit at much lower quantity than those in grasses (Lan et al., 2016b). Intriguingly, although tricin is restricted to certain dicot lineages, its flavone precursors, including apigenin, luteolin, and/or chrysoeriol, are widely distributed in non-tricin-accumulating dicots (Harborne, 1974). Hence, the occurrences of tricin derivatives are probably resulting from independent and convergent recruitment of novel enzyme activities in those isolated tricin-accumulating dicot lineages.

Three possible types of dicot enzymes, FNSIs, FNSIIs, and F2Hs, have been described for flavone nucleus formation (Martens et al., 2001; Martens and Mithöfer, 2005; Zhang et al., 2007; Ferreyra et al., 2015; Li et al., 2020a), but their contribution to tricin biosynthesis remains elusive in tricin-producing dicots. Both FNSIs (Britsch, 1990; Martens et al., 2001; Miyahisa et al., 2006; Yun et al., 2008) and FNSIIs (Kitada et al., 2001; Fliegmann et al., 2010; Wu et al., 2016; Zhao et al., 2016; Jiang et al., 2019) catalyze direct desaturation of flavanones into flavones, whereas F2Hs converts flavanones to 2-hydroxyflavanones which were proposed to be intermediates for generating the flavone skeleton (Akashi et al., 1998; Zhang et al., 2007).

Initially identified in parsley (Petroselinum crispum), FNSIs were long presumed to be confined to Apiaceae (Britsch, 1990; Martens et al., 2001; Yun et al., 2008). However, they were subsequently isolated from other dicots, including Arabidopsis (Ferreyra et al., 2015) and Morus notabilis (Li et al., 2020). Interestingly, angiosperm FNSIs outside Apiaceae are apparently phylogenetically unrelated to FNSIs in Apiaceae and non-vascular plants; thus, FNSIs were probably evolved convergently in distant plant lineages (Li et al., 2020). Meanwhile, all the known dicot FNSIIs and F2Hs are CYP enzymes belonging to the CYP93B subfamily (Kitada et al., 2001; Martens and Mithöfer, 2005; Zhang et al., 2007; Fliegmann et al., 2010; Wu et al., 2016; Zhao et al., 2016; Jiang et al., 2019). FNSIIs are present in most flavone-accumulating dicots, such as Gerbera hybrids (Martens and Forkmann, 1999), Lonicera japonica, L. macranthoides (Wu et al., 2016), Glycine max (Fliegmann et al., 2010; Jiang et al., 2010), Glycyrrhiza echinate (Akashi et al., 1999), Salvia miltiorrhiza (Deng et al., 2018), and Scutellaria baicalensis (Zhao et al., 2016). On the other hand, F2Hs were only reported in a few dicot species, including G. echinata (Akashi et al., 1998), Chrysanthemum indicum (Jiang et al., 2019), and M. truncatula (Zhang et al., 2007). It remains to be investigated whether FNSI, FNSII, and/or F2H are required for tricin biosynthesis which is restricted to isolated dicot lineages, such as the Medicago legumes.

Considerable knowledge about the 3'- and 5'-hydroxylation reactions required for tricin biosynthesis in Medicago legumes has come to light recently (Lui et al., 2020). Canonical CYP75A F3'5'Hs are not involved in the B-ring modifications, but instead, a group of Medicago-unique CYP75B proteins, including M. truncatula MtFBH-4 as well as alfalfa (M. sativa) MsFBH-4 and MsFBH-10, are utilized. In in vitro enzyme assays, these CYP proteins catalyze 3'-hydroxylation of different flavonoid classes (flavanone, flavone, and flavonol) and 5'-hydroxylation of their 3'-methoxylated derivatives which include chrysoeriol. Furthermore, apigenin is converted to 3'- and 5'-substituted flavones (i.e., luteolin, chrysoeriol, selgin, and tricin) when these CYP75B proteins are transiently expressed in Nicotiana benthamiana leaves. Consistent with these findings, M. truncatula MtFBH-4 knockout mutants are completely depleted in tricin O-glycosides, hence establishing an indispensable role of MtFBH-4 in tricin biosynthesis. Basically, the same reaction steps that occur in grasses (Figure 5) have been acquired independently by the Medicago legumes to produce tricin.

The Medicago-unique CYP75B enzymes required for tricin biosynthesis are distinct from the grass A3'H/C5'Hs with regard to their catalytic properties and phylogenetic origins (Lam et al., 2015; Lui et al., 2020). For example, the 5'-hydroxylase activity is restricted to chrysoeriol for the grass enzymes but is extended to other 3'-methoxylated flavonoids for the Medicago enzymes. Interestingly, the Thr-to-Gly substitution in the substrate recognition site 6 domain is critical for these Medicago enzymes to catalyze the 5'-hydroxylation reactions (Lui et al., 2020). On the other hand, the equivalent position is replaced by a Leu residue in the grass A3'H/C5'Hs (Lui et al., 2020), but it is unknown whether this could account for their more specific substrate preference for 5'-hydroxylation. Meanwhile, the Medicago-unique CYP enzymes have likely acquired the novel 5'-hydroxylase activities through neofunctionalization of redundant CYP75B F3'Hs following the divergence of the Medicago genus from other lineages in the legume family (Lui et al., 2020). Convergent evolution of CYP75B F3'5'H had also occurred independently in several Asteraceae lineages for the generation of delphinidin-derived blue/violet pigments (Seitz et al., 2006, 2015). By sharp contrast, A3'H/C5'Hs are highly conserved amongst grasses, consistent with prevalence of tricin in the grass family (Lam et al., 2019a). It would be intriguing to decipher the enzymology and evolution of B-ring hydroxylations for tricin biosynthesis in other isolated dicot lineages.

The enzymes responsible for the 3'- and 5'-O-methylation reactions remain elusive for tricin biosynthesis in dicots. It is possible that they are also COMT/CAldOMT enzymes, as in the case for the grass bifunctional OMTs. In fact, Arabidopsis knockout mutant analyses demonstrated the dual roles of COMT/CAldOMT in the production of monolignols and flavonoids (Do et al., 2007; Tohge et al., 2007; Nakatsubo et al., 2008). However, there is no tricin accumulation in Arabidopsis, presumably due to the absence of F3'5'H enzymes. Meanwhile, the expression of an endogenous COMT gene is upregulated in transgenic alfalfa over-expressing the gene encoding N-acetylserotonin O-methyltransferase (MsASMT1), which catalyzes the final step in melatonin biosynthesis (Cen et al., 2020). In addition to increased melatonin formation, the transgenic alfalfa plants produced elevated amounts of various soluble chrysoeriol- and tricin-derived metabolites (Cen et al., 2020), which might be resulting from increased COMT activities. However, FgCOMT1 isolated from the tricin-accumulating legume fenugreek (Trigonella foenum-graecum; Kuwabara et al., 2003) could O-methylate 5-hydroxyferulic acid but not quercetin (a 3'-hydroxylated flavonol) or tricetin in vitro (Qin et al., 2012). Over-expression of FgCOMT1 in Arabidopsis atomt1 knockout mutant only partially restored the accumulation of sinapoyl aldehyde and sinapic acid (intermediates of the monolignol biosynthetic pathway) but not isorhamnetin (a 3'-methoxylated flavonol; Qin et al., 2012).