To Gibberellins and Beyond! Surveying the Evolution of (Di)terpenoid Metabolism

To bicycle or not to bicycle

Notably, diterpenoid biosynthesis can be initiated by either of two distinct classes of reactions. While both involve carbocationic cascades, these are triggered in very different ways. The reactive allylic diphosphate ester bond present in GGPP invariably undergoes lysis/ionization to trigger one such carbocationic cascade, in reactions catalyzed by class I diterpene synthases (EC 4.2.3.x). However, this can be preceded by a protonation-initiated (bi)cyclization reaction (Figure 3), catalyzed by class II diterpene cyclases (EC 5.5.1.x), which leaves the allylic diphosphate ester bond intact for ionization by a subsequently acting class I diterpene synthase. From the trivial labdane name assigned to the most commonly observed hydrocarbon skeletal structure resulting from such class II bicyclization, the derived polycyclic natural products have been termed the labdane-related diterpenoids (65). Such metabolism is universally found in vascular plants due to the requisite production of GAs, whose biosynthesis proceeds through such a sequential pair of class II and class I cyclization reactions.

To bicycle first: The labdane-related diterpenoids

Of the more than 12,000 diterpenoids ~7,000 fall into the labdane-related super-family, highlighting the widespread diversification of such biosynthesis. The vast majority of these natural products are found in plants, where the requisite production of gibberellins (GAs) seems to have provided a genetic reservoir for derivation of more specialized labdane-related diterpenoids, particularly stemming from duplication of the genes encoding the two diterpene synthases/ cyclases, as will become evident in the following sections. Notably, these also are ancestral to all the plant (class I) terpene synthases involved in hemi-, mono-, sesqui-, as well as di- terpenoid biosynthesis (7), which form a moderately sized gene family (20+ members) in vascular plants (10). These enzymes produce the hydrocarbon backbones that define structurally related families, and represent the initial step in such terpenoid biosynthesis. Accordingly, GAs provided not only the origins of the diterpenoid metabolism discussed here, but to some extent that of the smaller terpenoid natural products as well. This broader evolutionary scenario has been recently reviewed (17), and almost certainly only applies to plants, not to microbes as suggested elsewhere (61), as will become evident in the discussions below.

Gibberellins: The ancestral labdane-related diterpenoids

GA metabolism has been comprehensively reviewed in several recent publications (5; 28; 66; 121). Here only the early steps in GA biosynthesis are described (Figure 4). In particular, those leading up to formation of the characteristic 6-5-6-5 ring structure, as it is duplication of these upstream enzymes that has led to the observed diversity of labdane-related diterpenoid natural products. GA biosynthesis is initiated by bicyclization of GGPP to copalyl/labdadienyl diphosphate (CPP), with enantiomeric stereochemistry (as defined relative to that found in cholesterol), in reactions catalyzed by class II diterpene cyclases then termed CPP synthase (CPS). The resulting ent-CPP is then further cyclized to the ent-kaurene, with this class I diterpene synthase then termed kaurene synthase (KS), which catalyzes a complex bicyclization and ring rearrangement reaction (Supplemental Movie). This tetracyclic 6-6-6-5 olefin is further transformed by a multiply reactive CYP, kaurene oxidase (KO), that converts C19 from a methyl group to carboxylic acid, via series of hydroxylation reactions, with intervening dehydration (54). Notably, while CYPs are typically found associated with the outer membrane of the endoplasmic reticulum (ER), at least the KO from Arabidopsis thaliana is found on the outer membrane of the plastid instead (31). This alternative sub-cellular localization presumably provides preferential access to the plastid-derived ent-kaurene. The resulting ent-kaurenoic acid is then finally converted to the gibberellane skeletal structure by another CYP, kaurenoic acid oxidase (KAO).

Not surprisingly, the amount of bioactive GAs is tightly regulated in plants. There clearly is tight transcriptional regulation of the genes for both anabolic and catabolic enzymes (28; 121). Notably, this includes transcription of CPS, whose activity can be considered to initiate GA biosynthesis (91). In addition, it has been suggested that the CPS involved in GA biosynthesis are subject to inhibition exerted by physiologically relevant levels of the divalent magnesium ion (which varies in plant plastids in response to light), and that also serves as an enzymatic cofactor for such class II diterpene cyclases, while those dedicated to more specialized metabolism are exempt from such inhibition (72). Susceptibility to this co-factor inhibition effect has been associated with the identity of a single amino acid residue, with a histidine found at this position in all CPS unambiguously associated with GA biosynthesis, while this is an arginine (or glutamine) in other class II diterpene cyclases. Strikingly, exchange of these residues is sufficient to ‘switch’ enzymatic susceptibility to such inhibition (49). Thus, it seems likely that this cofactor inhibition is a physiologically relevant regulatory mechanism. Recent work lends itself to speculation about the evolution of plant GA biosynthesis. This is based in part on the availability of genome sequences for the moss/bryophyte Physcomitrella patens and the lycophyte Selaginella moellendorffii, whose divergence from the angiosperm lineage is nearly as ancient (3; 79). It has been shown that S. moellendorffii produces and responds to GA, using genes homologous to those identified in angiosperms, while P. patens does not produce or respond to GA (34; 124). On the other hand, P. patens does produce ent-kaurene and ent-kaurenoic acid using enzymes homologous to those found in vascular plants (27; 52), and these diterpenoids further seem to have some physiological role in the moss (2; 26). Thus, it seems likely that the production of ent-kaurenoic acid arose early in plant evolution, before the divergence of bryophytes and the vascular plants some 450 million years ago (MYA), with the more specific evolution of GA biosynthesis arising in the next 50 million years – i.e., before separation of the lycophyte lineage from that of the angiosperms that occurred some 400 MYA (Figure 5).

Regardless of evolutionary history, it is clear that GAs are required for normal plant growth and development. Notably, this includes not only stem elongation and flowering (as shown in Figure 6), but also seed germination, such that plants with severe GA deficiency are sterile. Hence, the enzymatic genes underlying GA biosynthesis must be present in all plant genomes, providing the aforementioned genetic reservoir that can be drawn upon by either individual gene or whole genome duplication to derive more specialized labdane-related diterpenoid metabolism, as well as the more distantly descended class I terpene synthases, as will be discussed in the following sections.

Conifer resin acids: Early labdane-related diterpenoids

An abietadiene synthase involved in resin acid biosynthesis from grand fir, Abies grandis was the first diterpene synthase from more specialized labdane-related diterpenoid metabolism to be cloned (94). This bifunctional enzyme, AgAS, catalyzes both class II and I cyclization reactions (Figure 7), and was clearly related to both the CPS from A. thaliana (AtCPS) and maize (Zea mays), and the KS from pumpkin (Cucurbita maxima), which also had been cloned by that time (4; 96; 122). As noted in this original report, AgAS contained motifs in common with both. Specifically, a DxDD motif in common with the CPSs, and a DDxxD motif in common with the KS, as well as other known class I (di)terpene synthases. However, it also was noted that AgAS was similar in length to both the CPSs and KS, which are similar in length to each other, and all these shared at least some regions of significant similarity throughout their sequences (94). While the evolutionary relationship between all of these terpene synthases/cyclases was somewhat enigmatic at that time, later work demonstrated that at least some early diverging (non-vascular) plants contain a similarly bifunctional CPS/KS (27; 40). This led to the suggestion that GA biosynthesis may have originally relied upon such a bifunctional CPS/KS, arising from fusion of separate CPS and KS from a bacterial origin, which then underwent an early gene duplication, allowing sub-functionalization to the separate CPS and KS observed in vascular plants (53). Intriguingly, the genome of the lycophyte S. moellendorffii contains bifunctional diterpene synthases also involved in more specialized metabolism (47; 95), but these appear to have a separate evolutionary origin from those involved in conifer resin acid biosynthesis such as AgAS (10). While a number of bifunctional diterpene synthases have since been identified from gymnosperms (43; 50; 80; 86; 125), it should be pointed out that conifers seem to have monofunctional CPS and KS for GA biosynthesis (42), and also recently have been shown to have monofunctional class I diterpene synthases involved in more specialized metabolism as well (21). Thus, work on resin acid biosynthesis has both provided key insights into the evolutionary origins of labdane-related diterpenoids, particularly of the key diterpene synthases/cyclases, and the manifold ways in which such evolution unfolds.

In addition, AgAS has served as a model diterpene synthase/cyclase for investigation of terpene synthase structure-function relationships. This includes work with labeled substrates and analogs of high energy intermediates (70; 71; 73; 75–78), as well as extensive mutagenesis (49; 68; 69; 73). One of the key findings from this work was demonstration that the class I and II active sites were separate, and catalytically independent (albeit structurally interdependent), opening up the possibility that these were located in distinct domains (67; 71). Another was the demonstration that broadly conserved residues in both the class I and II associated domains (active sites) played important roles in catalysis (68; 69). More recently, a crystal structure has been determined for AgAS (127), revealing a tri-domain (γβα) protein structure wherein the class II active site sits at the interface between the N-terminal γ and β domains, while the class I active site falls within the C-terminal α-domain (Figure 7). These observations not only verified the modular nature predicted for the diterpene synthases/cyclases, but also the existence of an extensive interface between the domains associated with each activity (i.e., consistent with the previously demonstrated structural interdependence), particularly the β and α domains. Moreover, this atomic level structure has already led to additional insights into the enzymatic mechanism of at least the class II cyclization reaction as well (13).

Cereals: a model system for investigation of labdane-related diterpenoids

Expanded and diversified diterpene synthase gene families have been observed in angiosperms as well, most notably in the cereal crop plant family. Rice (Oryza sativa) has been known to produce more specialized labdane-related diterpenoids for over 40 years (38), and the relatively early availability of its genome sequence enabled application of a functional genomics approach to investigation of the underlying biosynthetic enzymes (64; 100; 123). This led to biochemical characterization of its full complement of both CPS and KS-like (KSL) enzymes, which significantly expanded the range of reactions for which the relevant enzymes were molecularly identified, as previously only those for the production of ent-kaurene and abietadiene were known (Figure 8). Moreover, despite extensive phytochemical investigations, the rice KSL (OsKSL) actually exhibited greater diversity than was expected from the known natural products, with the resulting diterpenes easily detectable in planta (55; 56). This demonstrated the ability of such a functional genomics approach to elucidate metabolism, as well as highlighting the extensive elaboration of labdane-related diterpenoid metabolism in rice.

Characterization of the OsKSL family led to detailed insight into the plasticity of class I terpene synthases. The alleles for OsKSL5 from ssp. indica and japonica were found to catalyze distinct cyclization reactions with ent-CPP, with that from japonica producing tricyclic ent-pimaradiene (37), while that from indica produced tetracyclized and rearranged ent-isokaurene (119). Comparison of their sequences led to identification of single residue responsible for this dramatic change in product outcome (Figure 9), which was found to be applicable to not only these alleles and the closely related OsKSL6, but also to even distantly related KS (120). It further was possible to apply this single residue ‘switch’ to increase the complexity of the reaction catalyzed by OsKSL4, from production of a tricyclic syn-pimaradiene to a rearranged tetracycle (57), as well as apply a similar ‘switch’ in AgAS (112). Moreover, these results, coupled with additional mechanistic investigation (82), highlighted role for electrostatic interactions, particularly with the diphosphate anion co-product, in such class I terpene synthase catalysis (128).

Molecular evidence of a role for more specialized labdane-related diterpenoids in maize (i.e., cloning and characterization of a second, inducible CPS), actually preceded identification of such natural products in this cereal crop plant (24; 87). Subsequent release of the maize genome sequence (88), revealed the presence of not only the already reported expanded CPS, but also KSL, gene families. In addition, wheat (Triticum aestivum) recently has been shown to contain similarly expanded CPS and KSL gene families, with diverse biochemical function (117; 129). Intriguingly, investigation of OsKSL substrate range demonstrated some stereochemical promiscuity (i.e., of CPP), with two that react with the endogenous syn-CPP also able to react with CPP of normal stereochemistry, although this is not found in rice (55). However, the wheat ortholog of the rice syn-CPP producing OsCPS4 (i.e., TaCPS2), produces normal CPP instead (117), with similar normal and syn- CPP stereochemical promiscuity observed with the wheat TaKSL as well (129). Such metabolic plasticity may underlie the observed diversification of labdane-related diterpenoid metabolism in the cereal crop plant family, and potentially even more broadly.

Labdane-related diterpenoids in dicots: Examples from the Salvia genus

Although labdane-related diterpenoid biosynthesis has not been as extensively investigated in dicotyledonous plants, it is clear that at least some have diverse such natural products. The most extensive molecular information is available for species from the Salvia genus, although other dicots have been investigated as well (15; 84). Examples from Salvia include identification of the diterpene synthases responsible for biosynthesis of sclareol from Salvia sclarea (6; 19; 85). These are noteworthy for their illustration of the ability of terpene synthases, of both classes, to introduce water during the course of the catalyzed reaction (Figure 10), the importance of the resulting hydroxyl groups for increasing solubility and providing hydrogen-bonding potential will be discussed in more detail below. Another example stems from investigation of tashinone biosynthesis in the Chinese medicinal herb Danshen (Salvia miltiorrhiza), which provides insight into the production of phenolic diterpenoid natural products more generally. In particular, the relevant diterpene synthases have been identified and shown to produce a cyclohexa-1,4-diene isomer of abietadiene that is poised for aromatization (16). Strikingly, the various Salvia KSLs, although clearly most closely related to other KS(L)s, are more similar in size to prototypical plant class I terpene synthases, having lost the N-terminal γ-domain (33), the significance of which will be discussed in more detail below.

Not to bicycle: Class I diterpene synthases acting directly on GGPP

There are many diterpenoids whose biosynthesis is initiated by class I diterpene synthases that react directly with GGPP (i.e., rather than a bicyclic derivative such as CPP). However, given that the complexity of the end product is a function of the number of double-bonds in the precursor (11), surprisingly few such class I diterpene synthases have been identified, particularly by comparison to the multitude of functionally diverse mono- and sequi- terpene synthases already known (14). While a casbene synthase from castor bean (Ricinus communis) and a taxadiene synthase from a yew species (Taxus brevifolia) were among the first plant terpene synthases to be cloned (51; 114), only a few others are known. These include cembrenol and neocembrene synthases (44; 106), whose product only slightly differs from casbene (Figure 11). Interestingly, the three class I diterpene synthases producing casbene, cembrenol or neocembrene (respectively) are more closely related to mono- and sesqui- (class I) terpene synthases than angiosperm KS(L) – e.g., all of these fall into the TPS-a sub-family, rather than the KS(L) TPS-e/f sub-family (10). In addition, these three class I diterpene synthases are significantly smaller than the KS(L), being similar in size to the prototypical plant mono- and sesqui- terpene synthases. This reflects loss of the γ-domain (Figure 12), which forms half of the class II diterpene cyclase active site, and this loss appears to have occurred early in angiosperm evolution given the extensive and independent diversification of the relevant TPS sub-families (10). The β-domain that forms the other half of the class II diterpene cyclase active site is retained, which may reflect its extensive interface with the class I α-domain noted above. Indeed, loss of the γ-domain has been observed to independently occur at least two other times (33), suggesting that this is relatively facile. By contrast, it does not appear that further loss of the β-domain has occurred. While class I terpene synthases composed of only an α-domain have been recently reported from the lycophyte S. moellendorffii, these are more closely related to microbial terpene synthases rather than other plant terpene synthases, indicating a separate evolutionary origin (45), not derivation from the βα bi-domain terpene synthases.

Much of the chemical diversity generated by sesqui- and particularly mono- terpene synthases requires rearrangement of the initial primary allylic diphosphate ester bond to a tertiary position, which removes the geometric barrier imposed by the trans 2,3-double bond to direct 1,6-bond formation of a cyclohexene ring (12). With GGPP the corresponding intermediate would be geranyllinalyl diphosphate. Accordingly, the discovery of a geranyllinalool synthase suggested the potential for analogous rearrangement of GGPP (32). This was substantiated by the more recent report of a rhizathalene synthase (103), which almost certainly proceeds through the corresponding rearranged geranyllinalyl diphosphate as an initial intermediate en route to initial cyclization to a 1,6-cyclohexene containing intermediate and, hence, to the more complex polycyclic rhizathalene (Figure 13). This then suggests that much more chemical diversity can be expected to arise from the action of class I diterpene synthases, which is supported by the diversity observed with the characterized microbial enzymes (92; 100).

Further transformations: Elaborating upon the hydrocarbon backbone

The diterpene synthases/cyclases described above most often produce a hydrocarbon olefin, which is, not surprisingly, highly hydrophobic (e.g., their partition coefficient or logP is ≥ 8.5). Accordingly, these must almost invariably be further elaborated by the introduction of oxygen, typically catalyzed by CYPs (EC 1.14.13.x), which are integral membrane proteins in eukaryotic organisms such as plants (62), enabling access to the presumably membrane embedded diterpenes. As discussed for GA biosynthesis above, these mono-oxygenases can catalyze multiple reactions with the same substrate, leading the incorporation of several atoms of oxygen. In addition, again from GA biosynthesis, it already is evident that these enzymes can alter the hydrocarbon backbone (e.g., the ring contraction catalyzed by KAO; Figure 4), leading to assignment of the resulting natural products to new structural families (e.g., gibberellins are not considered to be kauranoids per se, despite their common biosynthetic origins). However, most often the CYPs involved in diterpenoid natural products biosynthesis catalyze the hydroxylation reactions that are prototypical of these enzymatic super-family (62).

While the addition of a single hydroxyl group does increase solubility (e.g., decreasing logP by > 1), the resulting compounds resemble membrane components such as cholesterol, and most likely remain within the membrane. Accordingly, in order to provide solubility in a biological setting, it seems likely that two spatially separated hydroxyl groups are required, consistent with examination of the known bioactive diterpenoid natural products. At least in a few cases, this has been shown to decrease logP to ≤ 5 (116), which incidentally matches Lipinski’s rules for desirable pharmaceutical properties (46). Regardless, the incorporation of such polar groups also provides hydrogen-bonding potential, which would be expected to significantly increase the ability of the resulting natural product to specifically bind biological macromolecular targets. However, it should be noted that only in very few cases has it been established how multiple CYP operate in such biosynthesis, either order of action or even identification of the relevant CYP. Beyond GA, although several CYP have been identified from taxol biosynthesis, their order of action remains uncertain (18), and only for the relatively simple case of rice orzyalexin D & E biosynthesis, requiring only two CYPs, is such information available for diterpenoid natural products (116).

The relationship among CYPs can be inferred to some extent from the formal nomenclature. In particular, by the original definition, CYPs sharing more than 40% amino acid sequence identity are grouped into families, as designated by the number immediately following the CYP super-family designation, and CYPs sharing more than 55% amino acid sequence identity are grouped into sub-families designated by letter(s) following the family number, with individual CYP then finally identified by a concluding number, although some merging of (sub-)families has inevitably occurred as more molecular information becomes available (58). As an example of the nomenclature, the KO involved in GA biosynthesis from A. thaliana is CYP701A3 (30), with all the known KOs falling within the CYP701 family, and all those from angiosperms within the CYP701A sub-family (58). However, even such sub-family membership can not be used to assign biochemical function, as it has recently been shown that CYP701A8 from rice does not catalyze the prototypical KO reaction, but instead carries out hydroxylation at the neighboring C3α position in a variety of labdane-related diterpenes (110), consistent with assignment of KO activity in rice to the distinct paralog CYP701A6 (36; 83). The KAOs also involved in GA biosynthesis fall into the CYP88A sub-family (29), with no characterized family members yet found to operate in more specialized diterpenoid metabolism, although at least one member of different CYP88 sub-family does function in more specialized triterpenoid biosynthesis (89).

A number of other CYP sub-families are known to act in diterpenoid biosynthesis, with several members of the CYP725A sub-family assigned roles in taxol production (18), and at least two members of the CYP720B sub-family serving as promiscuous, yet multiply reactive oxidases in conifer resin acid metabolism (23; 81). In addition, two CYP families seem to contain multiple sub-families that function in diterpenoid natural products biosynthesis (Figure 14), albeit these are among the most extensive CYP families in plants. First, from the exceptionally large CYP71 family, a tobacco (Nicotiana tabacum) CYP71D sub-family member functions as a C6 hydroxylase in production of cembrendiol (107), although most of the member of this sub-family function in mono- or sesqui- terpenoid biosynthesis (22). At least two CYP71Z sub-family members seem to operate in rice labdane-related diterpenoid biosynthesis, both catalyzing C2-hydroxylation, with CYP71Z6 acting upon ent-isokaurene, presumably for production of the derived oryzadiones/oryzalides, while CYP71Z7 does so in phytocassane biosynthesis (115). In addition, at least one member of CYP99A, which is actually a sub-family of the CYP71 family (98), operates in momilactone biosynthesis, catalyzing the transformation of C19 from a methyl to the carboxylic acid necessary for 19,6β-lactone ring formation (109). Second, from the large CYP76 family, a number of rice CYP76M sub-family members have been shown play a role in more specialized labdane-related diterpenoid biosynthesis as well (108). For example, CYP76M7 catalyzes C11α-hydroxylation in phytocassane biosynthesis (98), while CYP76M6 and 8 catalyze C9β and C7β hydroxylation of 3α-hydroxy-ent-sandaracopimaradiene to produce oryzalexins E and D, respectively (116). Moreover, a CYP76AH sub-family member from Danshen seems to operate in tanshinone production, catalyzing C12-hydroxylation in formation of the intermediate ferruginol (20). Finally, members of the CYP714 family are known to function in GA metabolism (48; 126; 130), but no family members have yet been assigned roles in more specialized diterpenoid biosynthesis. Given the relatively few identified CYP from diterpenoid metabolism, it seems likely that other CYP (sub-)families will be found to operate in such natural product biosynthesis.

Beyond the increased solubility and hydrogen-bonding potential imparted by the introduction of oxygen(s) catalyzed by the CYPs, these also offer opportunities for additional elaboration. This can be simply further oxidation, as catalyzed by short-chain dehydrogenases/reductases (SDR) such as that involved the final step of rice momilactone A biosynthesis (90), but perhaps more interesting is the addition of various functional groups, such as that catalyzed by the characterized transferases from taxol biosynthesis (18; 105). However, even less is known about the enzymes involved in such further elaboration than the initially operating CYPs.

A range of roles for diterpenoid natural products drives their evolution

As described above, the GA phytohormones play critical roles in normal plant growth and development. Consistent with their prominent role in promoting stem elongation, as discussed above GA biosynthesis seems to have evolved during the differentiation of vascular plants, coupling function to evolution. As has been illustrated above, (di)terpenoid biosynthesis has evolved from that of GA, but these more specialized metabolites must have a selectable function in order to drive functional specialization and retention of the underlying enzymatic genes. In a few cases, biological roles have been assigned more specialized diterpenoid metabolites (Figure 15), providing some insight into the selective pressure leading to evolution of the relevant biosynthetic pathway. For example, conifer resin acids seem to function in defense against bark beetles, undergoing oxidative cross-linking following extrusion and volatilization of the mono- and sesqui- terpene components of the rosin to form a physical barrier, as well as exhibiting antifeedant and antibiotic activity (41). However, in the absence of genetic evidence, it is difficult to definitively assign more specific roles and/or physiological function. Hence, the focus here will be on diterpenoid natural products for which such genetic evidence is available. For example, it is clear from knock-out lines of the relevant diterpene synthase that rhizathalene acts as a semi-volatile antifeedant against root herbivory by the fungus gnat (Bradysia ssp) in A. thaliana (103). Interestingly, while the function of cembrendiol is not clear, suppression of the relevant CYP leads to accumulation of the cembrenol precursor instead, with the resulting tobacco plants exhibiting resistance to aphid colonization (107). In rice, analysis of knock-out lines of the upstream OsCPS4 and more specific OsKSL4 indicates that the primary function of the momilactones is allelochemical activity (i.e., suppression of the growth of other plant species)(118). Despite being the first rice phytoalexins isolated against the devastating fungal blast pathogen (Magneportha oryzae)(8; 9), these momilactone deficient knock-out lines are no more susceptible to infection by M. oryzae than their corresponding wild-type (parental) lines (118). However, suppression of OsCPS4 in a different genetic background does lead to increased susceptibility, as well as loss of allelopathy (101), perhaps due to a difference in timing of phytoalexin accumulation, which has been correlated with susceptibility (25). Indeed, alleochemical activity is attributable to momilactone B, which is selectively secreted from the roots, while momilactone A preferentially accumulates within the plant (e.g., upon infection), consistent with a role as a phytoalexin (39). These results highlight the complexity inherent in determining physiological function, as well as the differential activity of even closely related natural products. For example, momilactone B may be derived from momilactone A by a single biosynthetic step (64). On the other hand, such differential function then provides the selective pressure necessary for evolution of the relevant gene encoding the enzyme catalyzing this additional biosynthetic step. By contrast to the examples discussed above, despite the complexity and length of the relevant biosynthetic pathway, it is unclear what physiological function taxol serves in yew trees, and such uncertainty is prevalent among not only the diterpenoids, but all natural products. Nevertheless, the cases where such information is available demonstrates that diterpenoid activity spans plant-insect, plant-plant and plant-microbe interactions (Figure 15). According, these natural products serve in variety of roles, providing selectable advantages that drive continuing evolution of diterpenoid chemical diversity.

The yin and yang of biosynthetic gene clusters in plants

Finally, as observed in a so-far limited number of cases for plant natural products more generally (63), there also are diterpenoid biosynthetic gene clusters, investigation of which has offered some insight into the nature and assembly of these in plants versus microbes (where analogous gene clusters are shaped by horizontal gene transfer, which does not appear to be relevant in plants). In particular, it became evident at an early stage in the functional genomics based investigation of rice diterpenoid biosynthesis that both OsCPSs involved in more specialized metabolism were closely linked to OsKSL that acted on their enzymatic product – i.e., genes encoding sequentially acting enzymes were close together in the rice genome (74; 113). Moreover, the only other genes in these regions encoded CYP and, in one case, also an SDR (83). Given that all of the genes were clearly of plant origin and that their closest homologs are paralogs found elsewhere in the rice genome, these clusters do not appear to have arisen from any sort of horizontal gene transfer, but assembly during evolution of the rice genome via the usual vertical transmission. Nevertheless, later work demonstrated that, consistent with their physical linkage to the upstream OsCPS4 and OsKSL4, the nearby CYP99A2 and/or 3, as well as SDR, were involved in production of the momilactones, forming a dedicated biosynthetic gene cluster on rice chromosome 4 (Figure 16)(90). On the other hand, while the cluster on chromosome 2 has been primarily associated with phytocassane biosynthesis (98), CYP from this cluster seem to also participate in not only biosynthesis of the oryzadiones, for which the relevant OsKSL6 is found within the cluster, but also that of oryzalexins (108; 115; 116). Involvement in oryzalexin biosynthesis highlights the patchwork nature of this cluster, as the relevant OsKSL10 is found elsewhere in the rice genome. In addition, the KO paralog CYP701A8 also seems likely to function in phytocassane and oryzalexin biosynthesis, but is found elsewhere in the rice genome as well (albeit CYP701A8 is tightly linked in a tandem gene array with CYP701A6, encoding the essential KO required for GA production, ensuring its inheritance in any case)(110). While the patchwork coverage of biosynthetic pathways by these gene clusters is consistent with the absence of horizontal gene transfer, it is then somewhat unclear what drove their assembly. Although co-regulation via localized effect [e.g., on chromosomal state (111)] may provide pressure for maintenance of a cluster, the genes from the diterpenoid biosynthetic gene cluster on rice chromosome 2 are differentially regulated, with no obvious correlation to linear arrangement, arguing against a role for co-regulation in even cluster maintenance, at least in this case (98). More critically, it has been pointed out that co-regulation does not provide a driving force for assembly, which was suggested to require both positive selection (e.g., for production of a diterpenoid providing advantageous activity) and negative selection against incomplete inheritance (e.g., toxicity associated with a metabolic intermediate or derivative thereof)(99). Evidence for such dual push-pull selection pressure is evident from genetic dissection of the rice momilactone biosynthetic gene cluster, where the associated physiological functions in allelopathy and plant microbial defense provide positive selection pressure, but a striking 2-fold decrease in seed germination rate was observed with the OsKSL4 knock-out line, demonstrating a strong negative selection pressure against inheritance of OsCPS4 in the absence of OsKSL4 (118). Thus, the observed patchwork nature of natural product biosynthetic gene clusters in plants may result from the lack of negative selection against the loss of certain genes (i.e., those not found in the relevant cluster, such as OsKSL10, loss of which presumably does not have a significant deleterious effect).