Exploring the Evolvability of Plant Specialized Metabolism: Uniqueness Out Of Uniformity and Uniqueness Behind Uniformity

Abstract The huge structural diversity exhibited by plant specialized metabolites has primarily been considered to result from the catalytic specificity of their biosynthetic enzymes. Accordingly, enzyme gene multiplication and functional differentiation through spontaneous mutations have been established as the molecular mechanisms that drive metabolic evolution. Nevertheless, how plants have assembled and maintained such metabolic enzyme genes and the typical clusters that are observed in plant genomes, as well as why identical specialized metabolites often exist in phylogenetically remote lineages, is currently only poorly explained by a concept known as convergent evolution. Here, we compile recent knowledge on the co-presence of metabolic modules that are common in the plant kingdom but have evolved under specific historical and contextual constraints defined by the physicochemical properties of each plant specialized metabolite and the genetic presets of the biosynthetic genes. Furthermore, we discuss a common manner to generate uncommon metabolites (uniqueness out of uniformity) and an uncommon manner to generate common metabolites (uniqueness behind uniformity). This review describes the emerging aspects of the evolvability of plant specialized metabolism that underlie the vast structural diversity of plant specialized metabolites in nature.


Introduction: From the Origin
Plant specialized metabolism, also called plant secondary metabolism, is a generic term for the biochemical mechanism that biosynthesizes an array of metabolites; many of these metabolites are valuable for the environmental interaction of land plants with other organisms (Arimura et al. 2009), while some other metabolites, in the form of medicines or spices, are beneficial to humans.The biosynthetic pathways of plant specialized metabolites are characterized by their biosynthetic origins that typically branch out from the biosynthesis of primary core metabolic pathways for amino acids, nucleic acids and lipids that are highly conserved across the plant kingdom (Weng 2014, Weng et al. 2021).In contrast to the apparent commonalities in the chemical structures and the occurrence of such biosynthetic precursors across various plant species, plant specialized metabolites sharing core structures usually occur in a lineage-specific manner.However, these metabolites exhibit vast diversity in their overall structures and biological activities (Table 1) (de Vries et al. 2021), both of which have likely contributed to increased hidden ecological fitness in diverging and fluctuating environments.
Most plant specialized metabolites are composed of lowmolecular-weight organic compounds (∼1,000 m/z) and are likely associated with enzymes, transporters and other biosynthetic machineries, which are typically composed of polypeptides (∼100 kDa).Size distributions of metabolites in the KNAp-SAcK database and proteins with catalytic sites in the UniProt Knowledgebase have a median molecular weight of 402.2 m/z and 51.8 kDa, respectively (Afendi et al. 2012) (Fig. 1).The notable similarity in the two molecular size distribution curves suggests that the metabolite sizes are restricted by the enzyme sizes.
The structural uniqueness of specialized metabolites suggests that their evolutionary origins are relatively recent compared with conserved and common metabolites such as primary metabolites and plant hormones.In general, specialized metabolites are more abundant than phytohormones that share common precursors (Table 1).For example, both glucobrassicin, a specialized indole glucosinolate, and indole acetic acid, a phytohormone auxin, are commonly derived from tryptophan.However, their relative abundances differ by approximately 2,000-fold in the aerial parts of 2-week-old seedlings of Arabidopsis thaliana (Sugahara et al. 2020).Moreover, the Table 1 Summary of contrasting properties between core and specialized metabolisms levels of specialized steviol glycosides and the plant hormone gibberellins, both of which are geranylgeranyl pyrophosphatederived structurally related diterpenoids, are estimated to differ in abundance by around 10,000-fold in the leaves of Stevia rebaudiana (Brahmachari et al. 2011).These significant differences in the abundances of phytohormones and specialized metabolites raise an intriguing question regarding whether there is a correlation between the quantity of metabolites and the elapsed time since metabolic evolution because the amounts of metabolites required to perform given biological functions would have been optimized in terms of cellular cost.This is analogous to catalytic optimization in which the high expression of a low-activity enzyme is replaced by the low expression of a highly active enzyme.Whether low metabolic flux levels emerged as such from the beginning of the metabolic pathway or whether highly active metabolic pathways have evolved suddenly remain an open question.In all likelihood, both the activity and quantity of enzymes would have been continuously optimized during metabolic evolution with respect to cellular cost.
On the other hand, low levels of metabolites might also reflect their biochemical nature, either as pathway intermediates that are rapidly metabolized by highly catalytic enzymes or as toxic and/or highly reactive compounds that are too detrimental to be stably accumulated in plants.The latter scenario is exemplified by an adaptation to the dilemma of evolving rattlesnake venom toxins and avoiding self-poisoning by tailoring proteins called auto-inhibitors (Ukken et al. 2022).Plants have successfully avoided self-intoxication by toxic metabolites (autotoxicity) via conjugation (e.g.glycosylation, methylation and acylation), sequestration to specialized organs or intracellular organelles or secretion into extracellular spaces.Each of these methods enables the stable accumulation of specialized metabolites and avoids autotoxicity.Other than detoxification, plants can adapt to highly toxic metabolites.For example, the adaptive evolution of topoisomerase I with tolerant mutations against the specialized alkaloid camptothecin, which exerts anticancer effects via the inhibition of topoisomerase I, is an Fig. 1 The size distribution of metabolites and enzymes.(A) The size distribution of plant metabolites calculated from 61,520 metabolites listed in the KNApSAcK database (http://www.knapsackfamily.com/KNApSAcK_Family/)(Afendi et al. 2012).The median size of metabolites is 402.2 with a 95% confidence interval (CI): 400.4 < median < 403.9 (minimum: 27.0, 1/4: 308.0, 3/4: 556.1 and maximum: 928.2). (B) The size distribution of 9,006 enzymes (i.e.proteins with catalytic sites) from A. thaliana, Nicotiana tabacum and Sesamum indicum deposited in the UniProt Knowledgebase (https://www.uniprot.org/uniprotkb?query=*).The median size of the enzymes is calculated to be 51.8 kDa with a 95% CI: 51.3 < median < 52.3 (minimum: 10.1, 1/4: 39.6, 3/4: 69.8 and maximum: 115.2) using the following formula: protein molecular weight (kDa) = amino acid length (a.a.) × 120 (average amino acid molecular weight).The upper and lower graphs indicate a histogram and box plot for each distribution, respectively.example of an alternative strategy for avoiding the problem of autotoxicity (Sirikantaramas et al. 2008).
The biological relevance of plant specialized metabolism has often been ascribed to enhancing ecological fitness against various biotic and abiotic stresses, rather than to the autonomous control of cell and tissue development of the plants (Huang et al. 2019).This is partly because it has generally been challenging to relate the developmental control of primary metabolism, plant hormones and other biological mechanisms that are highly conserved in the plant kingdom with plant specialized metabolites that are found only in selected plant species.
However, increasing evidence reveals that, in some instances, specialized metabolism can exert control of plant tissue development.For example, the enzymatic activity of thalianol synthase and thalianol acyltransferase 2, which are involved in the biosynthesis of thalianol and other triterpenes in A. thaliana, has been shown to control root development (Bai et al. 2021).The sequential emergence of plant hormones during evolution already suggests that plant hormones are conserved in plants as only an evolutional consequence of the selection pressures for their possession in a common ancestral plant and that small molecules that are currently categorized as 'specialized metabolites' are eligible to become new plant hormones in the future.In this context, the births of lineage-specific metabolic branches and extensions from central metabolism, which have repetitively and frequently occurred during evolution, have been the frontier of metabolic evolution.This in turn suggests that the current 'border' between primary and specialized metabolism is defined merely by a tentative snapshot during metabolic evolution that is more dynamic than expected (Table 1).
Genomic approaches have accelerated the complete identification of the whole set of biosynthetic enzymes involved in various classes of plant specialized metabolites and have revealed that these enzyme genes can be grouped into a couple of classes despite the diversity in the structures of plant specialized metabolites (Lau and Sattely 2015, Qu et al. 2019, Hong et al. 2022, La Peña R et al. 2023).More specifically, specialized metabolism is a flow of enzyme-catalyzed structural changes in substrate molecules through oxidation, reduction and conjugation modification (glycosylation, methylation, acylation, etc.), where cytochrome P450 monooxygenases (CYP), 2-oxoglutarate-dependent dioxygenases (2ODD/DOX), UDP sugar-dependent glycosyltransferases (UGTs), S-adenosylmethionine-dependent O-methyltransferases (OMTs) and acyltransferases are universally involved as the common catalytic units (Kawai et al. 2014, Wilson and Tian 2019, Hansen et al. 2021).Thus, it is likely that gene multiplication and the neofunctionalization of a limited set of family enzymes have acted together to generate specialized metabolites for structural uniqueness.
These enzyme genes constitute superfamilies with dozens to hundreds of structurally related genes in a given plant genome.Although there is still much to debate around the putative molecular mechanisms of how such superfamilies of enzyme genes evolved, the combination of whole-genome duplication, which multiplies all the genes in a genome, and local tandem duplication, which multiplies a selected region in the genome, are likely to be critical for the lineage-specific specialization of metabolic pathways (Chae et al. 2014, Zhan et al. 2022).Enzyme genes in a given superfamily usually have different catalytic specificities owing to differences in spatiotemporal expression and amino acid sequences responsible for the catalytic specificity.The high copy number variations of enzyme genes suggest not only the biochemical diversity and the resilience of metabolism but also the versatility of these superfamily genes based on their plasticity and multiplicity.In this review, we focus on new aspects of metabolic evolvability, which enable an increase in the structural diversity of specialized metabolites in nature.

Metabolic Convergence in Phylogenetically Distant Plants
Oxidation and glycosylation reactions frequently occur sequentially in plant hormone metabolism and specialized metabolism, thereby increasing the structural diversity and the water solubility of metabolites in plant cells (Kawai et al. 2014).For example, it has been revealed that the specialized glycoalkaloid metabolism of tomatoes and potatoes in the Solanaceae family diverged from a common triterpene pathway by recruiting different enzymes (Akiyama et al. 2021b(Akiyama et al. , 2022)).This is a typical example of lineage-specific specialized metabolism formed by lineage-specific enzyme evolution, also known as divergent evolution, which forms a chemotaxonomic group sharing similar specialized metabolites.
Possible molecular mechanisms behind the convergent evolution of identical specialized metabolites in phylogenetically distant plants include the gene repertories involved, enzymatic plasticity through gain or loss of genes, functional differentiation by spontaneous mutation and gene recombination under the physicochemical constraints of organic compounds, and biochemical constraints of catalysis.

Catalytic Convergence in Enzymes
Sesamin, a specialized lignan that accumulates highly in dietary sesame seeds (Sesamum spp.), is observed not only in phylogenetically related species of Lamiales such as Paulownia spp.but also in phylogenetically distant Cuscuta of Solanales, Zanthoxylum spp. of Sapindales, Houttuynia spp.and Piper spp. of Piperales, Magnolia spp. of Magnoliales in the angiosperm and Ginkgo in the gymnosperm (Afendi et al. 2012, Kumar et al. 2022) (Fig. 2, Table 2).Although the physio-ecological functions of sesamin in plants are mostly unknown, this widespread sporadic presence of sesamin suggests untapped usefulness in increasing the fitness of the producing plants.Notably, CYP81Q1 from Sesamum spp.and its related gene from Lamiales are the only examples of a sesamin biosynthetic enzyme that catalyzes two sequential methylenedioxy bridge formations for the substrate pinoresinol, which is derived from two coniferyl alcohol molecules (Ono et al. 2006, Noguchi et al. 2014).The apparent lack of CYP81Q-related genes in sesamin-producing plant species other than Lamiales suggests that non-CYP81Q sesamin biosynthetic genes have evolved independently in other sesamin-producing plants by convergent metabolic evolution (Nelson and Werck-Reichhart 2011).The frequent occurrence of sesamin in unrelated lineages might reflect the fact that only relatively small numbers of catalytic steps would be required for sesamin biosynthesis from phenylpropanoid derivatives, such as coniferyl alcohol, which are highly conserved throughout the plant kingdom.This is also the case with caffeine, a specialized purine alkaloid sporadically found in many phylogenetically distant plant lineages (Huang et al. 2016) that does not require a long biosynthetic pathway.
However, it is surprising that the biosynthetic enzymes of an antitumor lignan, (deoxy)podophyllotoxin, were recently identified in phylogenetically distant plants as its biosynthesis requires much greater complexity and many more reaction steps than sesamin and caffeine biosynthesis, even though it has been acquired in completely different lineages: cow parsley (Anthriscus sylvestris of Apiales), hinoki-asunaro (Thujopsis dolabrata var.hondae of Coniferae) and mayapple (Podophyllum peltatum of Ranunculales) (Fig. 2, Table 2).From these plants, there are three distinct OMTs that have been identified to catalyze methylation at the same position of their substrate lignans (Lau andSattely 2015, Yamamura et al. 2023).Importantly, corresponding OMTs from those three different plant species share common amino acid substitutions in their substrate pockets.The data suggest that these OMTs were convergently incorporated into the specialized lignan metabolism under the functional constraint of recognizing the common substrate structure with common specificity.The common structural basis of the independently evolved metabolic enzymes with common specificity is known not   only for the regiospecificity of OMTs but also for the regiopromiscuity of CYPs and the sugar-donor selectivity of UGTs (Ohgami et al. 2015, Christ et al. 2019, Ono et al. 2020).These studies highlight the necessity of functional plasticity and the genetic multiplicity of the enzyme for driving convergent metabolic evolution.2).All aurone synthases are classified as polyphenol oxidases (PPOs), but their molecular lineages, substrate specificity and subcellular localization are different, revealing the parallel evolution of aurone biosynthesis-based functional plasticity of PPO enzymes (Supplementary Fig. S1).
The biosynthesis of cyanogenic glucosides is a thoughtprovoking example in considering specialized metabolic evolution.Their biosynthetic pathways in Sorghum (a grass), Lotus (a legume), Manihot (cassava) and Eucalyptus (of the Myrtaceae family) are commonly catalyzed by CYP79, CYP71 and UGT85 paralogs despite their phylogenetic distance (Kannangara et al. 2011, Takos et al. 2011, Thodberg et al. 2018, Hansen et al. 2022) (Table 2).Importantly, these paralogs are involved in the biosynthesis of glucosinolates, which are Brassicaceae-specialized metabolites known as 'mustard oil bomb' .In the parallel evolution scenario, these structurally similar enzymes are only paralogous, not orthologous, and have been independently recruited from the same enzyme family.The exceptional reactions specifically catalyzed by CYP736 in Lotus and CYP706 and UGT87 in Eucalyptus, respectively, which are apparently lineage-specific recruited enzymes (Takos et al. 2011, Hansen et al. 2022), support the parallel evolution of common metabolisms.Alternatively, the similarity of these enzymes, which would have been involved in ancestral common metabolism, may be remnants of enzymes lost in many other present lineages.
Similarly, the biosynthesis of diosgenin, a specialized steroidal saponin derived from triterpenes, is known to undergo convergent evolution in different lineages-dicotyledonous plants, including Leguminosae and Solanaceae, and monocotyledonous plants including Meliaceae and Dioscoreaceae (Table 2).It has been reported that in both monocots and dicots, cholesterol dihydroxylation reactions are commonly catalyzed by enzymes of the CYP90 family, but the subsequent monohydroxylation and cyclization reactions are catalyzed by distinct subclasses of CYP monooxygenases (Christ et al. 2019).The dihydroxylation of cholesterol is a reaction similar to brassinosteroid biosynthesis, suggesting that the dihydroxylases evolved independently from a common ancestral CYP gene for brassinosteroids.
These pioneering works described earlier indicate that convergent metabolic evolution has likely occurred repeatedly in various classes of enzyme genes.Metabolic evolution is basically achieved by duplicating and modifying preexisting enzymes as the de novo design of enzyme genes to catalyze specific reactions for specific substrates is extremely difficult; it is much easier to change the substrate specificity, regiospecificity or sugar-donor selectivity of preexisting enzymes.Given the cellular ability to copy genes with catalytic preferences, it is reasonable that altering the function of copied genes increases the variation and therefore the chance for convergence in metabolic evolution.However, lineage-specific recruited enzymes partly reflect differences in genetic availability, presumably owing to the specialized genomic preset associated with speciation.

Multivalent Metabolites
Specialized metabolites generally refer to phylogenetically specific small-molecule compounds.A broad spectrum of structural diversity in specialized metabolites has long been regarded as the basis for the unique biological activities exhibited by each specialized metabolite with unique chemical structures.However, emerging evidence shows that there are cases where metabolites that are conserved across kingdoms display distinct biological activities in species from different kingdoms.
We propose to define such metabolites as 'multivalent metabolites' .It is noteworthy that a group of small molecules in animals, long known as neurotransmitters, are often widely conserved in plants and microorganisms and have characteristic bioactivities outside the animal kingdom.Examples of such molecules include amino acids and biogenic monoamines.In plants, amino acids and biogenic monoamines are often part of the specific response to disease, mechanical wounding and drought stress.For example, the nonprotein amino acid gamma-aminobutyric acid, which is an inhibitory neurotransmitter in the central nervous system of mammals, has been suggested to be a signaling molecule involved in drought stress and disease response in plants (Bown and Shelp 2016).Moreover, the amino acid glutamate, which also functions as a neurotransmitter in animals, acts as a long-range signaling agent across organs in plants (Toyota et al. 2018).Acetylcholine (ACh), which was identified as the first example of a neurotransmitter, is also ubiquitous in plants.The biological function of ACh in plants has been largely attributed to responses against environmental stresses including (but not limited to) salinity (Qin et al. 2021).A. thaliana is capable of biosynthesizing ACh in planta, and the exogenous application of ACh has been shown to promote root hair development (Murata et al. 2015).The higher accumulation of ACh in Solanum melongena (eggplant) and Phyllostachys spp.(bamboo) than mammalian neuronal tissues implies that ACh might have untapped biological functions in these plant species (Horiuchi et al. 2003).In addition, biogenic monoamines constitute a group of specialized metabolites sporadically yet broadly found in all three kingdoms of life and exhibit biological activities that are unique to each species.Serotonin, known as the 'happy hormone' in animals, was shown to be incorporated into the cell wall and participates in the containment of the pathogen Bipolaris oryzae in Oryza sativa (rice) (Ishihara et al. 2008, Fujiwara et al. 2010).Melatonin, which has long been known as a neuronal hormone that regulates the sleep-wake cycle in mammals, is also found in plants, and its biological function has often been associated with disease resistance (Back 2021).Histamine, which functions both as a neurotransmitter and as a signaling molecule in the local immune response in animals, has been shown to accumulate extensively in the stinging hairs of Urtica spp.(nettle) and other selected plant species and is thought to protect the plants from herbivores by inducing itching and inflammation in animals (Iwamoto et al. 2014, Ensikat et al. 2021).Notably, the discovery of histamine and ACh was not achieved from animal tissues but from rye infected with ergot fungi (Barger and Dale 1910).Short-chain fatty acids (SCFAs) are another class of small molecules that can be categorized as 'multivalent metabolites' .Recent studies show that SCFAs might serve more than just as intermediates in carbon metabolism and additionally as environmental signaling molecules that are monitored by various organisms.For example, butyrate produced by gut microbes triggers macrophage-mediated immune responses under anaerobic conditions in the mammalian intestine (Chang et al. 2014).In plants, endogenous acetic acid promotes tolerance against drought stress, and isovaleric acid from Bacillus spp.has been shown to trigger growth inhibition (Kim et al. 2017, Murata et al. 2022).SCFAs are often converted into esters of coenzyme A (CoA) and then serve as substrates for the acylation of lysine residues of histone in eukaryotes, showing that SCFAs are indirectly involved in epigenetic regulation (Nitsch et al. 2021).The CoA ester of 2-hydroxyisobutyrate, among other SCFAs, has been associated with the dark-induced metabolic shift in sugar metabolism in plants (Zheng et al. 2021).As the perception of free-acid forms of SCFAs by plants is largely unknown, it is unclear whether SCFAs exhibit their regulatory activities either directly as free acids or indirectly as substrates, for example, for histone modification regulators.Amino acids, biogenic amines and SCFAs are universally present across kingdoms and thus have not been necessarily classified as specialized metabolites from an authentic metabolite categorization.However, the emerging biological roles of these metabolites in plants suggest that, in addition to small molecules with unique structures and biological functions, small molecules with highly conserved structures across kingdoms exhibiting lineage-specific physiological functions in plants could also be included in the list of plant specialized metabolites.Therefore, the definition of plant specialized metabolites might require minor modification.
While the chemical structures of multivalent metabolites are highly conserved in animals and plants, the biosynthetic mechanisms of those metabolites are unique to plants in some instances.For example, serotonin in animals is formed through the hydroxylation of tryptophan at the C5 position, followed by decarboxylation of the carboxyl group.In contrast, plants biosynthesize serotonin through the decarboxylation of tryptophan, followed by hydroxylation at the C5 position (Fujiwara et al. 2010).ACh, in contrast, is generated by transferring an acetyl group from acetyl-CoA to choline by choline-Oacetyltransferase (ChAT) in animals.However, no genes with substantial sequence similarity to the previously identified animal ChAT have been found in plants, leaving the biosynthetic origin of plant ACh ambiguous.Instead, an esterase that hydrolyzes ACh was isolated from maize and has been shown to belong to a plant-specific GDSL lipase family (Sagane et al. 2005) that is likely to be phylogenetically specific, as the corresponding enzyme activity was not detected from its orthologous gene in A. thaliana (Muralidharan et al. 2013).Similarly, the sporadic occurrence of serotonin and melatonin in plants can be explained primarily from the findings that only selected plant species have genes encoding tryptophan decarboxylase (TDC), a branch-point enzyme that converts tryptophan into tryptamine, which is a central precursor to serotonin, melatonin and many other specialized metabolites with an indole moiety, including a spectrum of monoterpenoid indole alkaloids (Negri et al. 2021).Notably, genes encoding plant TDC do not show apparent sequence similarity to those of animals and have so far been functionally identified only in selected species including Catharanthus roseus (vinca) and Ophiorrhiza pumila (De Luca et al. 1989, Yamazaki et al. 2003).Moreover, the sequence similarity among those functionally characterized plant TDC genes is approximately 70% at the amino acid level (Negri et al. 2021).Collectively, these reports indicate that, as in the evolutionary scenario of 'authentic' plant specialized metabolites, both convergent evolution and lineage-specific parallel evolution likely play key roles in the sporadic occurrence of multivalent metabolites in various plant species.

Co-presence and Metabolon
The biosynthesis of specialized metabolites generally involves multiple enzymes for the completion of the pathway.Thus, in addition to the catalytic specificities of enzymes involved, (i) the synchronous spatial and temporal gene expression and (ii) the co-localization of enzymes are essential for achieving the proper operation of plant specialized metabolism (Chae et al. 2014).The transcriptional co-expression of enzyme genes in plant specialized metabolism is partly supported by gene clustering that has occurred as a consequence of gene multiplication.However, the co-expression mechanisms of enzyme genes that are located in a discrete locus of the genome remain largely elusive.In turn, the orthotopic nature of enzyme gene expression in plant specialized metabolism allowed Arabidopsis thaliana trans factor and cis-element prediction database II (ATTED-II) and other coexpression analyses to be implemented for identifying a number of enzyme genes (Itkin 2013, Shang et al. 2014, Lau and Sattely 2015, Ono et al. 2020, Obayashi et al. 2022).
The continuous reactivity of oxidation and glycosylation likely contributed to the enhanced coupling of CYP/DOX and UGT.As reported in Solanaceous tobacco, a defect in a UGT leads to deglycosylation and the accumulation of toxic aglycones, resulting in impaired growth (Heiling et al. 2021).Therefore, adaptation to endogenous metabolic disorders while coping with fluctuating environmental cues is a central issue in the development of new metabolism; the molecular evolution of UGT as partner enzymes with CYPs and DOXs is likely to be key to minimizing the potential risks of autotoxicity from newly synthesized metabolites (Li et al. 2021).Indeed, along with the evolution of plant genomes, the copy number of CYP/DOX and UGT superfamily enzyme genes would have increased coordinately in seed plants (Kawai et al. 2014), suggesting that avoiding autotoxicity is a key constraint to the specialized metabolic evolution.
In contrast, metabolons that are putative enzyme complexes have garnered increasing attention as a possible mechanism for optimizing sequential metabolic reactions (Pandey et al. 2017).A metabolon is essentially an enzyme-protein complex that is integrated into organelle membranes and is thought to channel sequential enzyme reactions, thereby facilitating and optimizing the overall metabolic reaction cascades while avoiding the reactivity and toxicity of intermediate metabolites by metabolic channeling.The physical interaction is the most apparent case of the co-presence of proteins in the vicinity, enabling catalytic cooperation and substrate channeling through micro-locally enriched metabolic fluxes in subcellular compartments (Zhang and Fernie 2021).The formation of metabolons is known in conserved core metabolic pathways of prokaryotic and eukaryotic cells, including glycolytic bodies in which the enzymes of the glycolytic system assemble under low oxygen conditions, the tricarboxylic acid cycle, Calvin-Benson cycle and nucleotide synthesis by the purinosome (Schmitt and An 2017, Zhang et al. 2017, Pareek et al. 2021).
Metabolons play a role not only in core metabolism but also in specialized metabolism (Nakayama et al. 2019, Zhang andFernie 2021).Metabolons that catalyze the biosynthesis of cyanogenic glycosides and lignan glycosides (Ono et al. 2020) are additional examples of metabolons composed of CYP and UGT with CYP reductase (CPR) (Laursen et al. 2016).During the metabolic pathway, specialized metabolites are often oxidized and further decorated with sugar moieties to form O-glycosides, many of which are thought to be catalyzed by CYP and UGT.Generally, the glycosylation of aglycones contributes to enhanced water solubility and reduced substrate reactivity.The rich repertoire of CYP and UGT genes in plant genomes serves as the physical and biochemical core of metabolons and should allow for evolving metabolons with new catalytic activity.
Given that CYP and UGT are highly multiplied enzymes frequently involved in specialized metabolism (Kawai et al. 2014), we speculate that they would have been partly optimized as ready-to-interact as metabolon components during plant metabolic evolution.Notably, the formation of metabolon between CYP and UGT is also observed in mammalian phase II xenobiotic metabolism; the regio-selectivity of morphine glucuronidation by UGT2B7 was found to be altered by specific interaction with CYP3A4 (Ishii et al. 2007).The physical interaction between CYPs and UGTs commonly observed in plant and animal kingdoms suggests the biological importance of the interaction between CYP and UGT in continuous reactions of oxidation followed by glycosylation.Therefore, the interaction of metabolic enzymes is a prerequisite-not only in avoiding autotoxicity by providing physical channeling of substrate-binding pockets and rapid detoxification by attaching sugar moieties but also in biochemical cooperation in their catalysis (Tatsis et al. 2017, Heiling et al. 2021, Li et al. 2021).
More recently, more studies have reported that pathway enzymes are not the only components of metabolons; there are also proteins that provide scaffolds for enzyme complexes and maintain and control the structural integrity of metabolons (Gou et al. 2018, Waki et al. 2020).Understanding the importance of metabolons should lead to a better overview of the molecular mechanism that drives the biochemical diversity in specialized metabolism during evolution.
It has also been recently recognized that soluble enzymes and small molecules can be confined to a micro-environment by liquid-liquid phase separation without the presence of a membrane system, thereby improving the efficiency of enzyme reactions (Dahmani et al. 2023).Notably, when expressed in yeast and hetero-multimerized through an optogenetic approach, violacein biosynthetic cluster genes VioC and VioE preferentially catalyzed the formation of an antibacterial and antifungal alkaloid, deoxyviolacein, from protodeoxyviolaceinate, which is otherwise easily oxidized nonenzymatically to prodeoxyviolacein (Zhao et al. 2019).The results clearly indicate the feasibility of liquid-liquid phase separation as an alternative mechanism to metabolons for enhancing the metabolic reactions.Although the optimization of the metabolic pathway likely depends on metabolons for membrane-associated enzymes and their interactors, liquid-liquid phase separation might play an indispensable role in the efficacy of the metabolic pathway exclusively involving soluble enzymes.In both cases, the reconstitution of the reactions and the establishment of a quantitative assessment system for efficacies of enzyme reactions are essential for the proof-of-concept studies.

Specialized Metabolism in Domestication
Why do plants produce structurally diverse specialized metabolites?A general explanation is that specialized metabolites increase the ecological fitness of plants.For example, flower pigments, scents and toxins help to attract or repel certain species (Arimura et al. 2009).The biological activities of specialized metabolites have been studied within an ecological context in nature and were considered to help maximize the chance of survival following various biotic interactions with pollinators, seed dispersers, pathogens and commensal fungi, as well as experiencing drought, UV or other abiotic stresses.However, the biological functions of specialized metabolites in plants, especially under the laboratory setup and in experimental fields, are often ambiguous.This is because the biological functions of metabolites have been originally optimized for plants in natural habits.Therefore, metabolites can easily become less valuable under artificially controlled environments wherein the fluctuation in various biotic and abiotic environmental parameters is less than that in natural ecosystems.There are pioneering examples of metabolite functions from an ecological context associated with pollinator preference shift via the biosynthesis of various specialized metabolites (e.g.anthranilates, flavonoids, carotenoids and alkaloids) affecting recognizable phenotypes in floral petals, anthers and nectars, which likely affect reproductive isolation and eventually speciation (Lüthi et al. 2022, Roy et al. 2022, Liang et al. 2023, Mori et al. 2023).Nevertheless, the vast majority of the 'extended' phenotypes would not have been verified unless the plants were placed in the correct ecological context to exert specialized functions.
In contrast to wild species living in natural habitats, cultivated crops are organisms that have been developed to support human life.Together with productivity, which is the most important agronomic trait, specialized metabolites that contribute to commercial traits in quality-color, aroma, taste or storage durability extending shelf life-have also been intensively selected during domestication and modern breeding.Historically, flower color variants of horticultural plants, e.g.tulips and morning glories, have been bred and collected.Collections of germplasms of snapdragon (Antirrhinum) and morning glory (Ipomoea) contributed to the understanding of the flavonoid biosynthetic pathway and acted as a genetic resource for identifying various flavonoid biosynthetic genes (Rausher et al. 1999, Bradley et al. 2017).
Similarly, specialized metabolites causing bitterness, astringency and toxicity, which were often found in the edible parts of ancestors of modern crops, have been selectively removed by sensory screening.For example, genomic comparisons of the biosynthetic gene clusters in cucumber cucurbitacins and almond cyanogenic glycosides revealed that the corresponding functional biosynthetic genes present in wild species are absent from modern crops (Shang et al. 2014, Zhou et al. 2016, Sánchez-Pérez et al. 2019).Similarly, as the terminal sugar modification in group A soyasaponins is associated with strong bitterness and astringency, the removal of the responsible UGT genes has been a breeding target for improving the aftertaste of soybean products (Sayama et al. 2012).
Moreover, the levels of capsaicinoids (pungent components in chili pepper) and caffeine (alkaloid stimulant in tea) have been manipulated (Ogino et al. 2019, Tanaka et al. 2019).In the case of red wine, astringency is a positive trait suitable for longterm aging.The Tannat cultivar of red wine grapes is famous for an extremely high polyphenol content; accordingly, the cultivarspecific polyphenol biosynthetic enzyme gene is enriched in the genome compared with the Pinot Noir cultivar (Da Silva et al. 2013).This is thought to be a result of artificial selection of an anthocyanin-and tannin-rich cultivar.
In the case of tomato, enzyme genes that are involved in the formation of a bitter substance, tomatine, and a smoky volatile, guaiacol, have been either deleted or modified to accumulate as static glycosides by the introduction of new hydroxylation and sugar modifications (Tikunov et al. 2013, Cárdenas et al. 2019, Kazachkova et al. 2021, Akiyama et al. 2021a, 2022).These reports show that bitterness and toxicity have been reduced in various crops during domestication and that specialized metabolism can be readily altered with sufficient selective pressure.Even prior to the age of molecular biology and functional genomics, without modern transgenic technologies, humans modified the specialized metabolites of domesticated crops by selecting favorable sensory traits for dietary foods and beverages.These specialized metabolites would have contributed to ecological fitness in the native environment, but their functions are evaluated in an agricultural context and appear to be partly substituted or enforced by pesticides and fertilizers under artificial cultivation.

Evolvability of Specialized Metabolism
The molecular mechanism of how enzymes have acquired new functions is poorly understood.The escape of adaptive conflict theory predicts that the emergence of the novel catalytic activity that accepts originally unaccepted metabolites and produces minor metabolites precedes gene duplication (Des Marais and Rausher 2008); thus, the newly acquired catalytic activity will be enhanced when the minor metabolites enhance an individual's fitness.Gene duplication appears not only to solve the physicochemical dilemma of functional constraints between the original and the new activity of a single progenitor enzyme and allow the new activity to be rewired within a reasonable spatiotemporal framework but also to secure a molecular basis for the swift evolution of metabolic pathways (Lanier et al. 2023).Indeed, it is known that the catalytic activity of an enzyme often increases when the promiscuous catalytic activity toward multiple substrates becomes specific to a single substrate owing to the negative trade-off between catalytic promiscuity (generalism) and specificity (specialism) (Des Marais andRausher 2008, Khersonsky andTawfik 2010).Thus, the latent and promiscuous activities of enzymes are crucial seeds for metabolic evolution.It is important to note that when the duplicated genes are rewired to be expressed in different spatiotemporal locations, biochemical adaptation of such genes is placed in new metabolic contexts.This would liberate the enzymes from biochemical constraints for maintaining originally assigned catalytic activities and allow them to readily become promiscuous until acquiring new biochemical functions.However, given the limited information on known catalytic activities of the vast majority of enzymes, the catalytic modulation by tissue-specific physical interactions of catalytic enzymes, noncatalytic scaffold proteins and redox partners and allostery by protein-metabolite interactions (Tatsis et al. 2017, Baker and Rutter 2023, Zhao et al. 2023), we understand only a very small part of the metabolic evolvability that enables the development of the huge chemical diversity of specialized metabolites in nature.
Many specialized metabolic genes are frequently duplicated in tandem at specific genomic locations (Chae et al. 2014) although some are known to be duplicated by retroposition from transcripts into the genome (Matsuno et al. 2009).Importantly, there have also been reports of metabolic evolution in eukaryotes via horizontal gene transfer (HGT) (Kirsch et al. 2022).Phenylalanine ammonia lyase, the enzyme catalyzing the first committed step in the phenylpropanoid pathway leading to lignin, lignan and flavonoids, was acquired ancestrally via HGT during symbiosis with soil bacteria (Emiliani et al. 2009).Thus, it is of particular interest whether HGT has played indispensable roles in the evolution of specialized metabolism in plants.
Older genes, such as those involved in central metabolism, are intertwined with many intermolecular optimizations, and it has been reported that spare genes are rarely retained before they acquire new gene functions due to high molecular entanglements (Kuzmin et al. 2020).The functional differentiation of duplicated genes has been studied, but the evolution of new functions has been reported to be highly related to the low potency of the underlying gene.This is likely because many enzymes that mediate specialized metabolism are lineagespecific (i.e.recently) multiple superfamily genes that successfully developed different catalytic activities.Because groups of genes that have formed relatively recently, such as those for specialized metabolism, are usually distant from genes involved in central metabolism, it is unlikely that they have experienced a high degree of intermolecular optimization with other genes compared to central metabolic genes.Therefore, they may be more likely to undergo functional innovation.The unique evolutionary context located at the periphery of metabolism allows the emergence of highly specialized metabolic functions via the low entanglement of catalytic units (Table 1).
Steviol glycosides that are widely used as natural sweeteners are derived from the metabolism of diterpenes, which share their biosynthetic origins with the phytohormone gibberellin.Likewise, both triterpene saponins/steroidal glycoalkaloids and brassinosteroids are derived from shared sterol precursors.Moreover, auxins and glucosinolates are derived from tryptophan, whereas strigolactones originate from carotenoids.The biosynthesis and metabolism of all of these phytohormones are also mediated by many oxygenation and glycosylation reactions, practically sharing the involvement of CYP, DOX and UGT genes with specialized metabolism, as discussed in this review.We speculate that these genes were unlikely recruited to various metabolic pathways on the basis of the high gene multiplicity in plant genomes but rather that they became multiple genes as the common catalytic units for the evolutionary latency to assemble metabolism at multiple levels of co-presence; common transcription factors (Shoji 2019), protein-protein interactions (Nakayama et al. 2019, Zhang andFernie 2021) and subcellular compartmentation by phase separation (Dahmani et al. 2023) should have played indispensable roles in the molecular evolution of catalytic units with unique biochemical properties.In other words, these catalytic units are specialized in that they are prone to reorganize new metabolism.The uniformity in cooperative and low entangled catalytic units would ultimately allow for thrifty natural selection, avoiding the much costly de novo synthesis of such units (Fig. 3).Therefore, it is feasible that particular gene families have been expanded through feedforward interactions, in which the repurposed units are functionally optimized (Table 1).Metabolism is actually a continuous process, with no clear boundaries between categories, but the features described here provide a new perspective for understanding metabolic evolution in plants.
Plants are constantly updating their specialized metabolism to increase ecological fitness for their survival in nature.Therefore, specialized metabolic evolution is an arms race for adaptive chemical traits by diversifying common enzyme genes, which is analogous to the race against pathogen effectors via the diversification of multiple nucleotide-binding domains and leucine-rich repeat-containing gene (NLR)-mediated plant immunity (Jones et al. 2016).They are comparable in terms of the race for diversification of interacting molecules that function at the boundary between organisms.Even the biological relevance of the metabolites that are currently crucial for a plant will be biochemically updated as the environmental context changes.However, the manner of updating the bioactivities of the metabolites might not significantly change compared with the ongoing evolution of specialized metabolisms.The accumulation of this structure-activity relationship will Fig. 3 Two conceptual modes that support the evolvability of plant specialized metabolism.(A) The 'uniqueness out of uniformity' concept.Biosynthesis of plant specialized metabolites generally starts from highly central and highly conserved (i.e.across the plant kingdom) core metabolites as precursors.Therefore, the vast structural diversity of plant specialized metabolites depends largely on an array of catalytic properties exhibited by specific enzymes.However, such enzymes often feature common catalytic units that drive metabolic divergence.(B) The 'uniqueness behind uniformity' concept.On the other hand, there are examples of common plant specialized metabolites that are biosynthesized by a specific set of enzyme-coding genes that do not share an apparent common evolutionary origin.The trajectories to assemble specialized metabolisms can be likened to "bricolage".
help predict enzyme activity and the convergent evolution of enzymes in specialized metabolism (Yang et al. 2018, Fukushima andPollock 2022).

Perspectives from the Underground
In this review, we described possible mechanisms of how the convergent evolution, co-presence and evolvability of specialized metabolisms have been achieved.Driving force to convergently develop the identical metabolites currently remain unknown.We speculate that the biological activities of convergent specialized metabolites in analogous tissues and organs in different plants are likely to be associated with the adaptations for disease resistance, microbial symbiosis, pollinator attraction and herbivore avoidance that are commonly indispensable among various plant species.However, this is not the case when the sites of accumulation of the identical metabolites are distinct in different plant species.Aurone found in nonflowering liverworts is speculated to contribute to UV tolerance on land.However, in flowering plants, aurone pigments may also contribute to pollinator attraction via coloration in floral organs (Davies et al. 2020).Thus, specialized metabolites may have different physiological functions in different evolutionary contexts.For example, the in planta role of sesamin has long been enigmatic.However, the recent discovery of soilborne microorganisms that have acquired sesamin-degrading enzymes from sesame fields suggests that sesamin is consumed by selected microorganisms (Kumano et al. 2016).Similarly, caffeine-degrading microbes have been identified (Summers et al. 2015).It would be particularly interesting to clarify whether these microbes that evolved to assimilate specialized metabolites are enriched in cultivated fields of crops producing the responsible specialized metabolites for the sake of hidden biological interactions.
Specialized metabolites secreted from roots include triterpenoid acids and coumarins, which are thought to participate in interactions with soil microorganisms and insects (Nakayasu et al. 2022, Stringlis et al. 2019;Zhong et al. 2022).Findings have emerged that plants promote specific microbiota formation, via specialized metabolites, to obtain water, minerals, and organic compounds.Future studies will clarify whether specialized metabolites with unknown functions play important roles in symbiosis and co-evolution with other underground organisms as in the case of thalianol and other specialized metabolites that are secreted from the roots of A. thaliana (Huang et al. 2019).Furthermore, several intestinal microorganisms that metabolize ingested plant specialized lignans into enterolignans have been reported in the human gut microbiome (Bess et al. 2020), providing a new ecological perspective on external biological interactions via specialized metabolites with the microbiome beyond the internal physiology of plants producing specialized metabolites.Thus, specialized metabolism is expected to expand into the large field of the metabolite-mediated interplay between multiple organisms, from ecology and agriculture to human health.

Fig
Fig. 2 (Continued) A well-known example of the convergent evolution of specialized flavonoids occurs in the biosynthesis of flavones.While the oxidation of flavanone to flavone is catalyzed by CYP in general, a DOX enzyme, flavone synthetase I, catalyzes the corresponding reaction in the Apiaceae family (Martens and Mithöfer 2005).Other notable examples of convergent Fig. 2 (Continued)

Table 2
Examples of sporadically distributed common specialized metabolites in plants (Afendi et al. 2012AcK database(Afendi et al. 2012).b Estimated divergent time