The chalcone synthase superfamily of type III polyketide synthases

Abstract

Covering 1970–2001

This review covers the functionally diverse type III polyketide synthase (PKS) superfamily of plant and bacterial biosynthetic enzymes, from the discovery of chalcone synthase (CHS) in the 1970s through the end of 2001. A broader perspective is achieved by a comparison of these CHS-like enzymes to mechanistically and evolutionarily related families of enzymes, including the type I and type II PKSs, as well as the thiolases and β-ketoacyl synthases of fatty acid metabolism. As CHS is both the most frequently occurring and best studied type III PKS, this enzyme's structure and mechanism is examined in detail. The in vivo functions and biological activities of several classes of plant natural products derived from chalcones are also discussed. Evolutionary mechanisms of type III PKS divergence are considered, as are the biological functions and activities of each of the known and functionally divergent type III PKS enzyme families (currently twelve in plants and three in bacteria). A major focus of this review is the integration of information from genetic and biochemical studies with the unique insights gained from protein X-ray crystallography and homology modeling. This structural approach has generated a number of new predictions regarding both the importance and mechanistic role of various amino acid substitutions observed among functionally diverse type III PKS enzymes.

---

Michael B. Austin

Mike Austin was born and reared in Springfield, Missouri. He studied philosophy briefly before engaging science, and received a B.S. in Chemistry from Southwest Missouri State University (SMSU) in 1998. Entering the Chemistry and Biochemistry Ph.D. program at the University of California at San Diego (UCSD), Mike joined Joe Noel's group in the Structural Biology Laboratory at the Salk Institute for Biological Studies. Mike's current interests include the use of such tools as X-ray crystallography, site-directed mutagenesis, and biochemical characterization to investigate the structural basis for the mechanistic functional divergence of evolutionarily related natural product biosynthetic enzymes, accompanied by the application of this knowledge to the engineering of both known and novel enzymatic activities of interest.

Joseph P. Noel

Joe Noel was born and raised in rural western Pennsylvania. Upon obtaining a B.S. degree in Natural Sciences with a concentration in Chemistry from the University of Pittsburgh at Johnstown, Joe entered the graduate program in Chemistry at the Ohio State University in the summer of 1985, and graduated from the Chemistry Department in 1990 with a PhD centered on the examination of the mechanistic enzymology of phospholipases with Professor Ming-Daw Tsai. Joe then completed his postdoctoral training with the late Paul B. Sigler in the Department of Molecular Biophysics and Biochemistry at Yale. While at Yale University, he undertook the X-ray crystallographic examination of heterotrimeric G-proteins. Joe and his group are now utilizing a combination of traditional mechanistic enzymology, molecular biology, plant biology, and tools in structural biology including protein X-ray crystallography and NMR to decipher the structure, function, and evolutionary lineage of a large number of enzymes that act in plant cells and many microorganisms to produce biologically active natural products including terpenes, polyketides, and alkaloids. Armed with the three-dimensional structure of the enzymes in plant cells responsible for the creation of this diverse array of bioactive compounds, his group is also working to engineer new specificities into these pathways to create novel products using a structurally-guided approach.

---

1 Introduction and background

Covering type III polyketide synthase (PKS) research between the discovery of chalcone synthase (CHS) in the1970s through the end of 2001, this review additionally touches on related and ancestral enzymes, and discusses the biological and medicinal applications of the natural products of type III PKS pathways. A recurring theme of this review integrates information from genetic and biochemical studies with the unique insights gained from both protein X-ray crystallography and homology modeling to formulate the structural basis for the functional diversity observed within the type III PKS superfamily of enzymes.

1.1 Metabolism and natural product biosynthesis

Natural products include a diverse collection of small molecules, which confer upon their hosts a multitude of attractive, defensive, communicative, or other biological advantages. An organism's collection of natural products often determines both intra- and inter-species interactions, thus contributing to both the identity and individuality of a species. Remarkably, most natural products are built up from a handful of simple building blocks, usually derived from one or more primary metabolic pathways. For example, each of the three types of PKSs catalyzes the sequential condensation of two-carbon acetate units derived from a malonate thioester into a growing polyketide chain, in a reaction sequence that mirrors the biosynthetic pathway of the fatty acid synthase (FAS) enzymes of primary metabolism. This review focuses on the third class of PKS (also known as CHS-like PKSs), traditionally associated with plants but recently discovered in a number of bacteria.

1.2 The chalcone synthase/stilbene synthase (CHS/STS) superfamily of plant and bacterial type III polyketide synthases

As a ubiquitous enzyme in higher plants, CHS provides the first committed step in flavonoid biosynthesis by catalyzing the sequential decarboxylative addition of three acetate units from malonyl-CoA to a p-coumaroyl-CoA starter molecule derived from phenylalanine via the general phenylpropanoid pathway. In the same active site, CHS then forms chalcone via the intramolecular cyclization and aromatization of the resulting linear phenylpropanoid tetraketide (Fig. 1). Downstream enzymes in branching pathways variously modify this chalcone scaffold to produce a number of biologically important compounds. These plant natural products are utilized for such diverse purposes as antimicrobial defense, anthocyanin flower pigmentation, UV photoprotection, pollen fertility, and induction of symbiotic root nodulation by Rhizobium bacteria leading to nitrogen fixation.

Fig. 1 Structure, CoA binding, and overall reaction of chalcone synthase (CHS). (a) Each monomer is colored in gold and blue, respectively, in this ribbon diagram. The N- and C-termini for each monomer are indicated. CoA is depicted as a stick diagram while the position of the active site cysteine is highlighted by *. The bottom panel shows the overall reaction catalyzed by CHS with the malonyl derived portions of chalcone shown in red. (b) Molecular surface representation of the CHS-CoA complex oriented as shown in (a). In the bottom panel, the two CHS monomers are separated and rotated slightly to highlight the flat dimerization interface along with the methionine side-chain and dyad related hole in the backside of the CHS active site.

Aside from the biosynthetic diversity provided by downstream tailoring enzymes, duplication and functional divergence of the CHS gene in plants has given rise to an expanding superfamily of homologous enzymes, many of which have only recently been cloned or functionally characterized. These type III PKS enzymes can differ from CHS in their preference for starter molecules, the number of acetyl additions they catalyze, and their mechanism of chain termination including alternative patterns of intramolecular cyclization, as is the case with the stilbene synthases (STSs). Because this important class of divergent CHS-like enzymes was discovered not long after CHS, known type III PKS enzymes are often referred to as the CHS/STS superfamily. Only a few years ago, less than five bacterial gene sequences in sequence databases were annotated as CHS-like. None of the enzymes encoded by these original depositions had been functionally characterized. Their scarcity, lower sequence homology, and the uncertain evolutionary history of type III PKS enzymes prompted speculation that these few bacteria had only recently acquired CHS-like genes via horizontal transfer from plants. However, the recent explosion of genome sequences has revealed many more bacterial CHS-like sequences, most sharing about 25% amino acid sequence identity with CHS and with each other. Only a small number of these bacterial enzymes have been functionally characterized to date, but this work has uncovered three new type III biosynthetic activities, and demonstrated the involvement of bacterial CHS-like enzymes in important secondary metabolic pathways. Given the rapid expansion in the number of deposited bacterial and plant CHS-like sequences, the initial elucidation of their biosynthetic capabilities, and the availability of high resolution structural information for some members of this family, increased attention has focused on these enzymes, resulting in a deeper appreciation for the CHS/STS superfamily and the general acceptance of the type III PKS designation.

While this review covers the essential features of several decades of type III PKS superfamily research, the reader is in many places referred to older or more focused reviews for more detailed information and additional primary references. This approach allows us to integrate recent insights into the structural basis of functional diversity within the type III PKS superfamily with a wealth of phytochemical and biological information. Clues obtained from studies of several functionally divergent type III enzymes, combined with mechanistic knowledge gained from recent crystal structures of CHS and related condensing enzymes, provide a fascinating example of the evolutionary process of metabolic divergence, an important contributor to the remarkable diversity of life on earth. Additionally, the comprehensively annotated primary sequence alignments of known divergent type III PKS enzymes presented here, viewed in light of this structural and functional information, should facilitate more reliable annotations and more appropriate selections of existing and newly discovered type III PKS gene sequences for biochemical analysis.

The remainder of Section 1 provides an introduction to related enzymes of both primary and secondary metabolism, and summarizes insights derived from various crystallographic and mechanistic studies of these PKS relatives. Section 2 covers CHS and summarizes the branching metabolic pathways associated with CHS products, while Sections 3 and 4 address other plant and bacterial type III PKS enzymes, respectively. A summary and concluding remarks are provided in Section 5.

1.3 PKS ancestors and related enzymes

1.3.1 Relevance of related condensing enzymes to type III PKSs. A better appreciation of the type III PKS superfamily is gained by comparing and contrasting these PKSs with the first two classes of PKSs as well as with the enzymes of fatty acid primary metabolism from which all three types of PKSs likely evolved, and with which they share an overall fold and mechanistic pathway (Fig. 2). While classification of both PKS and FAS enzymes is traditionally based upon genetic organization, recent crystallographic analyses have led to a more comprehensive understanding of the relevant differences between the architecturally similar domains of these sometimes remotely related enzyme families.

Fig. 2 Simplified E. coli fatty acid metabolism. Boxed labels denote enzymes possessing the αβαβα-fold. (a) Fatty acid biosynthesis. (b) Fatty acid degradation.

1.3.2 β-Ketoacyl synthases of fatty acid biosynthesis. While the essentials of fatty acid metabolism are found in any basic biochemistry text, several reviews provide a more comprehensive analysis.^1–4 A single round of fatty acid biosynthesis results in the elongation of the saturated hydrocarbon portion of an acyl substrate by two carbons. FAS systems complete each cycle by catalyzing one condensation and three reduction steps, with the help of a small handful of ancillary activities and protein domains (Fig. 2(a)). Substrates and intermediate products are maintained as thioester conjugates to one of two carrier molecules: either the small molecule coenzyme A (CoA) (Fig. 3), or the 9 kDa acyl carrier protein (ACP). Both carrier molecules utilize the same phosphopantetheine prosthetic group, whose terminal thiol participates in the thioester bond with the acyl substrate. Thioester bonds are utilized because they are weaker than similar bonds to carbon or oxygen. Their relatively high-energy state allows for facile isoenergetic transfer of substrates to catalytically essential active site cysteines, as well as energetically favorable formation of carbon–carbon bonds.

Fig. 3 (a) Structure and composition of malonyl-CoA. Note the phosphopantetheine arm also utilized by ACP (acyl carrier protein). (b) A NAC (N-acetylcysteamine) synthetic thioester.

While short chain acyl-CoAs such as acetyl-CoA are common end products of various degradative pathways, CoA is not the preferred carrier for most FAS biosynthetic enzymes. Substrates must first be activated through transfer to an ACP by an acyltransferase (AT) activity. This activity is sometimes labeled as a malonyl acyltransferase (MAT) to reflect its additional role in the transfer of the malonyl unit to ACP whereupon it is used for polyketide chain extension. Following the transfer of the substrate to the ketoacyl synthase (KAS or KS) domain's catalytic cysteine, this condensing enzyme catalyzes the addition of a two-carbon acetate unit to the enzyme bound thioester end of the fatty acid, via a decarboxylative condensation with malonyl-ACP. The resulting ACP-bound β-ketoacyl thioester is presented to an NADPH-dependent β-ketoacyl-ACP reductase (KR), which reduces the original substrate carbonyl (now the β-keto carbonyl) to an alcohol. A β-hydroxyacyl dehydratase (DH) catalyzes loss of water, leaving a carbon–carbon double bond. An NADH-dependent enoyl-ACP reductase (ER) module completes the reduction of the β-carbon, resulting in an acyl-ACP that resembles the original substrate, but with two additional methylene moieties. Most FAS complexes perform several cycles of elongation before a thioesterase (TE) activity releases the product as a free fatty acid. In vivo, it can be difficult to assess whether the final product length specificity of a FAS system depends more upon its thioesterase or its KAS activity.

Fatty acid biosynthetic systems are divided into two classes based upon genetic organization. Type I FAS systems, found in animals and fungi, encode the above activities in distinct domains on one or two multifunctional, multidomain protein chains. For example, while mammalian FAS activities are encoded in a single polypeptide that functions as a homodimer,⁵ the yeast FAS activities are distributed across two polypeptide chains (α and β), functioning as a multimeric complex incorporating six copies of each chain.¹ Conversely, plants and most bacteria instead have type II fatty acid biosynthetic systems. These pathways use similar enzymes to carry out the same functions as type I FAS systems. However, each of the type II FAS activities is catalyzed by a distinct and dissociable enzyme.³ Although some evidence from plant plastids suggests that type II FAS enzymes do associate in vivo, unequivocal stoichiometric information is not yet available.⁶

The FAS system of E. coli has been intensively studied, yielding many insights into type II fatty acid biosynthesis.³ Three different homodimeric KAS enzymes, varying in specificity for both substrate and product length, are found in E. coli (Fig. 2(a)). KAS I (FabB) elongates C₄ fatty acids to C₁₆ acyl chains, while KAS II (FabF) expands these C₁₆ fatty acids to C₁₈ acyl chains through the addition a single acetate unit. KAS III (FabH) initiates this process by elongating acetyl-CoA to a C₄ intermediate, thus activating C₂ fatty acids for entry into the iterative KAS I fatty acid biosynthetic cycle. All three E. coli KAS enzymes use decarboxylative condensations of malonyl-ACP to extend acyl substrates, but while KAS I and II utilize acyl-ACP starters as well, KAS III is unique in that it acts directly upon an acyl-CoA starter, but releases its product as an ACP thioester.

Interestingly, pathogenic Mycobacterium tuberculosis bacteria possess both type I and type II FAS systems. A conventional type I FAS produces C₁₆–C₂₄ long chain fatty acids, some of which are further extended to C₅₀–C₅₆ acyl products by a type II FAS system. The products of both FAS systems are combined in a high molecular weight α-alkyl-β-hydroxy fatty acid called mycolic acid (Fig. 4), which serves as a unique and critical component of the mycobacterium cell wall. As in E. coli, the initial biosynthetic step catalyzed by the type II FAS is carried out using a KAS III enzyme that utilizes malonyl-ACP. However, the mycobacterial KAS III is inactive against acetyl-CoA, and instead elongates a C₁₂ acyl-CoA product of the type I FAS.⁷

Fig. 4 Branched structure of a typical mycolic acid from Mycobacterium tuberculosis.

1.3.3 Thiolase family. Thiolases are divided into two functional classes: the 3-ketoacyl-CoA thiolases (thiolase-I) that are instrumental in fatty acid degradation, and the biosynthetic thiolases (thiolase-II), which exploit the reverse reaction in various biosynthetic pathways (Fig. 2(b)).⁸ Thiolases occur as dimers, tetramers, or hexamers, and in both homotypic and heterotypic forms.⁹

The iterative fatty acid β-oxidation degradative (FAD) pathway, although it uses CoA-instead of ACP-tethered substrates, is essentially the reverse of the FAS biosynthetic process, generating the same intermediate forms as each β-carbon is successively oxidized from a methylene to a ketone (Fig. 2(b)). However, rather than a KAS condensing enzyme, β-oxidation pathways utilize the distantly related 3-ketoacyl-CoA thiolase to eliminate the original carboxyl group and adjacent methylene of the oxidized β-ketoacyl-CoA, as an acetyl-CoA molecule. While 3-ketoacyl-CoA thiolase catalyzes both the forward and reverse reactions, thermodynamic equilibrium favors the degradative process.¹⁰

Biosynthetic thiolases maintain a much narrower substrate tolerance than the iterative FAD thiolases, converting two acetyl-CoA molecules into acetoacetyl-CoA, or vice versa. The condensing reaction of thiolase-II is utilized by eukaryotes for the biosynthesis of ketone bodies, cholesterol and steroid hormones, and by some prokaryotes for the production of their primary energy storage molecule, poly-3-D-hydroxybutyrate.^11,12 Biosynthetic thiolases differ from FAS and PKS condensing enzymes most notably in their method of generating the acetyl carbanion moiety used in the Claisen condensation reaction catalyzed by all three families of enzymes. While KAS and PKS enzymes achieve this via the decarboxylation of malonyl-CoA or malonyl-ACP, thiolases use a second active site cysteine in an anionic thiolate form to abstract a proton directly from acetyl-CoA.¹³

1.4 Three types of polyketide synthase

All PKSs, like their FAS ancestors, possess a β-keto synthase (KS) activity that catalyzes the sequential head-to-tail incorporation of two-carbon acetate units into a growing polyketide chain. However, whereas FAS systems perform reduction and dehydration reactions on each resulting β-keto carbon to produce an inert hydrocarbon, PKS systems omit or modify some of these latter reactions, thus preserving varying degrees of polar chemical reactivity along portions of the growing linear polyketide chain. Various PKS enzymes selectively exploit the reactivity of polyketide intermediates to promote intramolecular cyclization and π-bond rearrangements, generating an amazingly diverse collection of substituted monocyclic and polycyclic products from a simple acetyl building block. The major classes of PKSs are briefly described here, but the reader is referred to a number of outstanding and current reviews^14–20 for additional information regarding the extensive literature on polyketide biosynthesis and its comparison to fatty acid metabolism.²¹

PKSs have traditionally been divided into two classes by the same genetic organization criteria used to classify FAS systems. The enzymology and genetic organization of PKS types I and II, found in the biosynthetically prolific fungi and also in some bacteria, suggest that they evolved via FAS pathway gene duplication and partial loss of function, or perhaps, in some cases, more directly through partial gene duplication.

Type I PKSs resemble the yeast and animal FASs, consisting of multi-domain polyproteins that form large multi-functional biosynthetic complexes and act in either a modular or iterative fashion. Erythromycin biosynthesis in the bacterium Saccharopolyspora erythraea is the best-characterized example of a modular type I PKS enzymatic process, with two sets of most or all of the four core FAS domains (described in Section 1.3) encoded on three ∼350 kDa polypeptides, each linearly arranged set catalyzing a single round of elongation and full or partial reduction (reviewed in Refs. 14,17,19). The first module is preceded by AT and ACP domains, and the last module is followed by a TE domain for product off loading. Conversely, the biosynthesis of 6-methylsalicylic acid (6-MSA) in the fungus Penicillium patulum is carried out by an iterative, homotetrameric type I PKS, which uses a single copy each of KS, MAT, DH, KR and ACP to link four acetate units that cyclize via an intramolecular aldol condensation and aromatization (reviewed in Refs. 14,18,20). Interestingly, the KR and DH domains act upon only one of the four ketone moieties on the elongated polyketide.

In contrast, carbon flow through type II PKSs is directed by putative multienzyme complexes consisting of discrete, separable proteins, like the FAS type II systems found in plants and bacteria. Although enzymes of this second class are dissociable, and not always encoded in a linear fashion within the genome, they are believed to form in vivo complexes similar to those of the type I PKS systems. Type II PKS complexes lack the AT domains seen in FAS and type I PKS enzymes, but nonetheless possess acyltransferase activity. Actinorhodin biosynthesis in Streptomyces coelicolor (reviewed in Refs. 14,15) is catalyzed by a typical type II PKS, which uses six gene products to fashion eight acetate units into three fused, substituted six-membered rings (two aromatic carbon rings and a lactone). Three of these subunits, a KR, an aromatase (ARO), and a cyclase (CYC), are reaction-specific, while the remaining three subunits, an ACP, a KS, and a third domain with KS homology, comprise the iterative “minimal PKS” common to all type II PKS systems. This mysterious third domain was originally labeled a “chain length factor” (CLF), as swapping this domain between various type II minimal PKS complexes alters polyketide chain length. However, subsequent analysis noted that the missing catalytic cysteine is replaced by a glutamine in all known CLF subunits, a substitution that converts KS domains into efficient decarboxylases. This decarboxylation activity of the so-called CLF was subsequently proposed to provide the KS module with an acetyl starter, and the CLF was dubbed a “chain initiation factor” (CIF). A more recent and neutral label for this domain is KS-β, with the actual ketosynthase designated KS-α. KS-α and KS-β are translationally coupled and form a heterodimer. This is unusual among ketosynthase domains, as even the multidomain type I FAS and PKS polypeptides appear to associate in such a way as to allow a homodimeric arrangement of their KS domains.

NMR structures of ACP domains from E. coli and B. subtilis FAS^22–24 and Streptomyces PKS²⁵ have been solved, revealing subtle structural differences consistent with the observed inability of ACPs to substitute for one other.²⁵ Additionally, crystallographic efforts with individual PKS domains such as the macrocycle-forming thioesterase from erythromycin biosynthesis are beginning to bear fruit.²⁶ However, the lack of detailed information about the three-dimensional organization of the biosynthetic complexes of PKS types I and II limits our understanding of the structural determinants of substrate preference, polyketide chain length, and cyclization specificity in these large multi-enzyme complexes. Questions of PKS quaternary structure become even more complicated when enzymes are recruited from other biosynthetic pathways to act on polyketide intermediates, thus tailoring the chemical and biological properties of the final products.

Conversely, type III or CHS-like PKS enzymes maintain a simple architecture (a homodimer of identical KS monomeric domains—see Fig. 1), making them much more amenable to in vitro examination and manipulation as well as detailed structural analyses. Recent structural evidence,²⁷ summarized in Figs. 5 and 6, and discussed in Section 1.5, suggests that type III PKS enzymes emerged by gain of function from the structurally similar homodimeric KAS III. As was described in Section 1.3.2, KAS III initiates type II FAS biosynthesis by adding a single acetate unit to a small starter molecule, via a decarboxylative condensation with malonyl-ACP. In contrast, type III PKS enzymes have acquired multiple activities to catalyze varying numbers of iterative condensations utilizing a diverse set of much larger starter molecules. Moreover, these mechanistically complex type III PKSs have further developed three distinct polyketide cyclization mechanisms, all the while maintaining a simple homodimeric organization of KS monomers.

Fig. 5 Structural similarities and differences among αβαβα-fold enzyme families. The conserved αβαβα-folds are shown in rose and gold. Use of blue, green, red, or charcoal for additional highlighting facilitates comparisons by denoting specific insertion points into the primary sequence of the conserved fold. Blue asterisks mark the locations of conserved catalytic cysteines.

Fig. 6 Catalytic active site residues and overall reactions of αβαβα-fold enzyme families. Blue asterisks identify the conserved catalytic cysteines also denoted in Fig. 5. Red is used to track the incorporation or removal of acetate units in the various reactions.

In at least one case, a type III PKS enzyme (CHS) has recruited an NADPH-dependent monomeric ketoreductase²⁸ to specifically reduce to an alcohol one carbonyl position in either a linear polyketide intermediate or the cyclized trione. The resulting alcohol is eliminated as water in the subsequent aromatization that produces chalcone. This ketoreductase was called chalcone reductase (CHR), but the name polyketide reductase (PKR)²⁹ has been championed due to the enzyme's inactivity towards chalcone itself. While “chalcone reductase” is a somewhat misleading label, ketoreduction is also invoked in the proposed type III PKS catalyzed biosyntheses of several other plant natural products (see Section 3.3), suggesting that PKR is an overly general name for the CHS-specific ketoreductase. Although the synthesis of 6′-deoxychalcone by the combined action of CHS and CHR has been amply demonstrated in vitro, evidence supporting the formation of a CHS-CHR complex remains elusive, and CHR access to the buried CHS active site is improbable. Most likely, CHR acts on a diffusible linear CoA-thioester intermediate or an exchangeable and cyclic trione product (discussed in Section 2.2.2). Interestingly, CHR is more closely related to carbohydrate reductases than to the analogous KR domains of FAS and PKS types I and II.³⁰

1.5 Structural comparison of thiolase-fold enzymes illuminates type III PKS evolution and function

Pre-structural bioinformatic analysis successfully predicted the roles and identities of some catalytic residues and hinted as to the evolutionary relationships between condensing enzymes.³¹ The αβαβα fold, apparently common to all β-ketoacyl synthases and thiolases, was first observed in the 1994 structure of a homodimeric yeast 3-ketoacyl-CoA thiolase.⁹ Subsequently, in 1998, the crystal structure of E. coli KAS II (FabF) confirmed the conservation of the αβαβα fold in this condensing enzyme.³²

Along with a higher resolution structure of yeast thiolase,⁸ structures of a tetrameric biosynthetic thiolase from the prokaryote Zoogloea ramigera were elucidated, in various unliganded, CoA-bound, and cysteine-acetylated forms, thus facilitating improved modeling of thiolase reaction intermediates.^11,13 Multiple crystal structures of all three condensing enzymes from the model E. coli type II FAS system, namely KAS I (FabB),^33–35 KAS II (FabF),^32,36 and KAS III (FabH),^27,37,38 are also now available, as are the structures of Synechocystis KAS II³⁹ and the divergent Mycobacterium tuberculosis KAS III enzyme described in Section 1.3.2.⁴⁰ Both apo and complexed forms of most of these later enzymes are known, and some of the functional hypotheses suggested by individual structures have been tested by mutagenesis.^41–43

In 1999, our laboratory published the first crystal structure of a type III PKS, a CHS from alfalfa, in both apo and complexed forms.⁴⁴ Additional structures of various CHS mutants divulged further details of the condensation mechanism.⁴⁵ Subsequently, the structure of a 2-pyrone synthase (2-PS) from daisy sharing 74% amino acid identity with CHS confirmed anticipated alterations in the active site volume available to polyketide intermediates.⁴⁶ Most recently, the as-yet unpublished structures of two stilbene synthases (STSs) elucidated in our laboratory revealed the structural basis for these enzymes' divergent aldol-based cyclization mechanisms.

Comprehensive comparisons of thiolase-fold enzymes accompany two of the most recent structural studies.^34,39 Comparison of structural and mechanistic traits allows the grouping of structurally characterized thiolase-fold enzymes into three families (Figs. 5 and 6). One group encompasses KAS I and II, another group consists of KAS III and the CHS-like type III PKS enzymes, and the final group contains the biosynthetic and degradative thiolases. The features shared by all of these enzymes include conservation of the core αβαβα structural architecture, the location of the extensive dimerization interface, placement of the active site, and use of the same catalytic cysteine residue for covalent attachment of substrates and intermediates (Fig. 6). Differences include the extent and structure of the loops located on the opposite side of the active site cavity from the conserved αβαβα core, the number of active site residues contributed by the dyad-related monomer, the position and identity of key catalytic residues (excluding the universally conserved cysteine), and substrate preference for either CoA-linked or ACP-linked thioesters. The observed differences in the catalytic machinery of thiolase-fold enzyme families account for their varied susceptibility to inhibition by the antibiotic cerulenin, a feature historically exploited to help classify these related enzymes.³⁵

The homology-based structure/function analysis of divergent type III PKS enzymes presented in this review relies heavily on comparisons of both conserved and varied active site residues, both within the type III PKS enzyme family and between CHS-like enzymes and their closest structural and functional relatives, the ancestral KAS III enzymes of primary metabolism. For consistency, references to type III PKS residues throughout this review will utilize the numbering of the structurally characterized Medicago sativa CHS2 enzyme,⁴⁴ unless otherwise specified.

2 Chalcone synthase (CHS) and its biosynthetic pathways

2.1 CHS introduction

2.1.1 CHS history, reaction, and distribution. Because of its important role in initiating flavonoid biosynthesis (Fig. 7), it is not surprising that CHS was the first type III PKS enzyme to be discovered. Phytochemists noted the wide distribution of anthocyanin flower pigments and other flavonoid products in higher plants much earlier, and the incorporation of acetate units into the flavonoid skeleton was predicted in the 1950s.⁴⁷ The early characterization of CHS was done at the Institute for Biology II at the University of Freiburg, in Germany, using parsley CHS obtained from cell suspension cultures. In 1972, Kreuzaler and Hahlbrock used isotopic labeling to demonstrate the production of the flavanone naringenin from malonyl-CoA and the phenylalanine derivative p-coumaroyl-CoA.⁴⁸ Identification of the styrylpyrone natural product bis-noryangonin as a truncated side product of the reaction implicated the p-coumaroyl moiety as the initial substrate of the iterative elongation reaction (Fig. 8).⁴⁹ However, naringenin is formed as the released chalcone product of CHS undergoes an additional ring closure that occurs rapidly and stereospecifically in vivo through the catalytic action of chalcone isomerase (CHI) (Figs. 7 and 8). Because this Michael-type ring closure also occurs spontaneously but non-stereospecifically in aqueous solution within 30 min of chalcone's release from the CHS active site, researchers mistakenly concluded that CHS catalyzed the formation of naringenin rather than chalcone, and erroneously labeled the enzyme “flavanone synthase”. This error was corrected in 1980,⁵⁰ after publication of the purification and early characterization of CHS. In 1983, the first isolation of a chs gene was reported.⁵¹ As of March 2002, GenBank contained nearly 650 chs-like gene sequences.

Fig. 7 Plant pathways relevant to CHS and resulting classes of natural products. (a) The general phenylpropanoid pathway. (b) CHS and its importance in flavonoid biosynthesis. Note the different numbering conventions for chalcones and flavanones.

Fig. 8 CHS chemistry, reaction intermediates, and related products. (a) Our current understanding of CHS substrate loading, malonyl decarboxylation, and polyketide extension. (b) CHS reaction intermediates, truncation products, and possible routes to reduced chalcones are shown in the bottom panel. The related STS reaction (see text) is included for comparison.

While CHS is present in all gymnosperms and angiosperms examined thus far, the distribution and specificity of downstream tailoring enzymes varies dramatically across different species, within different tissues, and at different times during the life cycle of an individual plant. Due to this biochemical diversity, most plants maintain multiple copies of chs genes,⁵² which are expressed in various tissues at different developmental stages. Some of these isoenzyme genes are constitutively expressed, while others are transcriptionally induced by environmental stress including UV light, wounding, and pathogen infestation.^53–55 These diverse expression patterns reflect the wide range of biological roles fulfilled in plants by products of the downstream flavonoid metabolic grid.⁵⁶

2.1.2 Products and functions of CHS-related phenylpropanoid and flavonoid pathways. In this section, we present a brief overview of the upstream (general phenylpropanoid) and downstream (flavonoid) biosynthetic pathways relevant to CHS activity (Fig. 7), as well as the physiological and medicinal actions of their natural products. More detailed analyses can be found in a number of current reviews, many of which focus on different aspects of phenylpropanoid or flavonoid diversity, biosynthesis, or biological function.^56–62

Although chalcone synthase catalyzes the first committed step of the multi-branched flavonoid pathway, CHS is one of several enzymes using products of the upstream general phenylpropanoid pathway, which synthesizes p-coumaroyl-CoA from L-phenylalanine in three enzymatic steps (Fig. 7). Aside from flavonoid biosynthesis, intermediates and products of the general phenylpropanoid pathway are diverted to synthesize benzoic acid, salicylic acid, coumarins, and the monolignol precursors of lignin, an important structural and defensive polymer.^55,58 Evolution of lignin biosynthesis appears to have coincided with the movement of plants onto land; rudimentary flavonoid biosynthesis may have emerged soon after, co-opting phenylpropanoid substrates to protect the new land plants from harmful UV rays. Alternatively, early flavonoids may have first been used as signaling molecules or anti-microbial phytoalexins.⁶³

Plants have recruited a number of tailoring enzymes to modify the flavonoid scaffold (summarized in Fig. 7 and discussed below). Further diversity is achieved by various combinations of hydroxylation, O-methylation, glycosylation, acylation, prenylation, and conjugation at multiple positions, resulting in the identification so far of over 6000 naturally occurring flavonoids.⁶⁴ Flavonoid nomenclature can be confusing, as chalcones use a different numbering convention than other flavonoids. The carbons of chalcones' phenylpropanoid-derived rings are numbered 1–6 and the acetate-derived ring is numbered 1′–6′. Conversely, in the flavanone naringenin (and downstream flavonoids), the acetate-derived and middle rings are numbered 1–10, and the phenylpropanoid-derived ring is renumbered 1′–6′ (Fig. 7).

The chalcone product of CHS is stereospecifically isomerized by chalcone isomerase (CHI), forming the central heterocyclic ring of the flavanone (2S)-naringenin. The sequential oxidation, reduction, and dehydration of this central ring by flavanone 3-hydroxylase (F3H), dihydroflavonol reductase (DFR), and anthocyanidin synthase (ANS) [also known as leucoanthocyanidin dioxygenase (LDOX)], produces anthocyanidin (Fig. 7). Glycosylation of this aglycone yields the red pigment anthocyanin (reviewed in Ref. 59). Some species utilize additional B-ring hydroxylation of upstream intermediates, by flavonoid 3′ (or 3′5′) hydroxylase (F3′H and F3′5′H) P450s, to alter the color patterns of resultant anthocyanin pigments (Fig. 7). Anthocyanin pigments are transported to and accumulate in vacuoles, where pH and conjugation or complexation (with metals, malonic acid, or other flavones) enhances or modifies their colors to yield orange, pink, red, purple, or blue pigmentation. Flower color is an important determinant of a number of interspecies interactions: bees, for example, are attracted to blue flowers.⁵⁶

Other tailoring enzymes utilize intermediates in the anthocyanidin pathway to create additional classes of flavonoids (reviewed in Refs. 56,57 and summarized in Fig. 7). The CHI-catalyzed flavanone formation can be blocked by methylation of 6′-deoxychalcone's 2′-hydroxyl moiety. Alternatively, chalcone can be modified to produce aurones. Flavones are produced from flavanones by flavone synthase (FS or FNS) I and II, and in legumes isoflavones result from oxidation and a dramatic rearrangement of the flavanone carbon skeleton through the action of isoflavone synthase (IFS), which shifts the B ring from C-2 to C-3 on the lactone ring. Downstream modification of this isoflavonoid scaffold produces coumestrol and dihydropterocarpan natural products.

Flavonol synthase (FLS) converts dihydroflavanols into flavonols, and leucoanthocyanidin reductase (LCR) converts flavan-3,4-diols into flavan-3-ols. Both of these products are conjugated to each other forming condensed tannins (proanthocyanidins). It is important to note that the seemingly trivial introduction of a C2–C3 double bond by many of these enzymes transforms the non-planar flavanone and dihydroflavanol anthocyanidin pathway intermediates into planar molecules, and also completes the conjugation of the previously separate A- and B-ring π-systems.

In legumes, a branched pathway results from the regiospecific reduction of a CHS polyketide intermediate to an alcohol by the so-called chalcone reductase (CHR) enzyme introduced in Section 1.4. The resultant hydroxyl group is eliminated during the final aromatization step of the CHS reaction, producing 4,2′4′-trihydroxychalcone (commonly referred to as 6′-deoxychalcone). This product is also a candidate for methylation as well as the CHI and IFS reactions discussed above. (CHR is further discussed in the context of the CHS mechanism, in Section 2.2.2)

While CHI appears to possess a unique fold unrelated to any other enzyme of known function,⁶⁵ most of the other flavonoid tailoring enzymes belong to one of three common enzyme families: NADPH-dependent reductases, oxoglutarate-dependent dioxygenases, or cytochrome P450 hydroxylases.⁵⁷ Several lines of evidence implicate the co-expression and co-localization of most of these enzymes to the endoplasmic reticulum, as well as the in vivo channeling of intermediates ( i.e. flavonoid pathways are not dependent on a cytoplasmic diffusion of reactive or unstable intermediates). Further evidence (reviewed in Ref. 57) suggests the temporally regulated occurrence of loosely associated multi-enzyme complexes consisting of these predominantly cytoplasmic (soluble) proteins, perhaps on or near the membrane-associated P450 enzymes. However, no such complexes have yet been observed in vitro, even in the case of CHS and CHR, where the apparent exchange of reactive intermediates would suggest a particularly intimate interface.

The distribution and substrate specificity of many of these enzymes vary in different tissues and in different species. As a result, multi-branching pathways sometimes form metabolic grids that compete with each other for substrates, resulting in a diverse mixture of flavonoid products. For example, while DFR, FLS, F3′H, and F3′5′H might all potentially compete for the dihydroflavonol product of F3H, FLS is also able to accept the hydroxylated products of F3′H and F3′5′H as substrates. In some cases, mixtures of flavonoids are known to exhibit synergistically enhanced activities. On the other hand, some branch point enzymes are developmentally or constitutively expressed, while others are only transiently upregulated in response to a specific environmental cue.^58,66

Most flavonoids absorb harmful UV-B radiation (280–315 nm). Almost all green leaves contain epidermal flavonoids, especially flavonols and anthocyanin pigments (chlorophyll masks their color) (reviewed in Ref. 56), and mutant plants lacking these epidermal flavonoids are particularly sensitive to UV-B damage.⁶⁷ Besides acting as a sunscreen, various 3′4′-dihydroxyflavonoids (downstream of F3′H) are also effective free radical scavengers. Plants with UV-resistant genotypes or with sustained exposure to UV radiation generally maintain higher ratios of 3′,4′-hydroxylated to 4′-hydroxylated flavonoids.⁵⁶

Although a few biological or medicinal properties are confined to (or enhanced in) specific flavonoid subclasses, more often a particular activity or a set of related physiological activities will be represented by a distribution of variously modified natural products from different flavonoid subclasses. In other words, different species often arrive at different phytochemical solutions to the same or similar biological problems, in the form of varied flavonoid products exhibiting different combinations of chemical constituents. This kind of biological diversity also occurs within individual species, and is expected to confer an adaptive advantage to plants. Rapid elaboration of a plant's chemical repertoire makes intuitive sense, as many of the uses of flavonoids involve species–species interactions, often of a defensive nature.^54,56 Chemical defense strategies must continue to evolve, as pathogens or predators also continue to evolve countermeasures to battle with plant chemical defenses.

Many diverse flavonoids, including aurones, flavanols, isoflavans, 3-deoxyanthocyanidins, prenylated isoflavonoids, and pterocarpans, have significant antibacterial, antifungal, or antiviral activity against various pathogens (reviewed in Refs. 56,58). Some are inhibitors of the multidrug resistance (MDR) efflux pump implicated in the emergence of bacteria resistant to a wide range of antibiotics.^68,69 Conversely, several plant flavonoids, including chalcones, flavanones, flavones, and isoflavones, are known to promote symbiosis with nitrogen-fixing Rhizobia bacteria. Secreted from plant roots, these compounds act by inducing expression of the bacterial nod genes that allow Rhizobia to form symbiotic nodules on the plant root hairs, a relationship that provides the plant with usable nitrogen (reviewed in Refs. 61,62).

Plants also use various flavonoids, including flavones, flavonols, isoflavones, and condensed tannins, to control herbivory by both insects and animals, often with species–specific consequences (reviewed in Ref. 56). For example, the diets of primate species reflect their differing ability to tolerate condensed tannins.⁷⁰ In herbivores adapted to certain diets, flavonoids often play other roles. Some butterflies lay eggs only on leaves containing a particular combination of flavonol glycosides.⁷¹ Other butterflies sequester UV-absorbing flavonoids they ingest from their host plant in their wings; the amount and type of flavonoids sequestered influences mate selection.⁷¹

It is unknown to what extent flavonoids affect overall human health, as they are almost ubiquitously present in human diets. Nevertheless, a number of significant medicinal properties have been attributed to the flavonoids present in plant-rich diets. Given the functions of flavonoids in plants, their utility as free radical-scavenging antioxidants⁷² and antimicrobial agents ( e.g. antimalarials)⁷³ in humans is not surprising. However, flavonoids have also been shown to exhibit vasodilatory,⁷⁴ anti-cancer,⁷⁵ anti-mitotic,⁷⁶ anti-inflammatory,⁷⁷ anti-asthmatic,^78,79 and estrogenic⁸⁰ activities. Some flavonoids, particularly isoflavones, are called phytoestrogens due to their ability to mimic steroid hormones by binding to estrogen receptors and to type II receptors (which also bind estrogen).⁸⁰ Actually, the NodD receptor itself (the rhizobial target of flavonoids that some plants secrete to induce root nodulation leading to nitrogen fixation) may bear some structural as well as functional similarity to the ER-α estrogen receptor.⁸¹ Several flavonoids non-specifically inhibit tyrosine kinases,⁸² while others specifically inhibit⁸³ the Iκ-B kinases that activate the transcription factor NF-κB.⁸⁴ And finally, chalcone derivatives interfere with MDM2-mediated inhibition of the p53 tumor suppressor protein implicated in some types of tumors.⁸⁵ These important physiological interactions in humans are consistent with the observed health benefits of flavonoids. The differential effects of flavonoids on multiple vital signaling pathways, combined with their diversity and number, suggest that we have just begun to sample the medicinal potential of flavonoid natural products.

2.2 CHS structure and mechanism

2.2.1 CHS three-dimensional architecture. As discussed in Section 1.5, the five-layer αβαβα core and the location of the active site and dimerization interface is conserved among all thiolase-fold containing enzymes. The αβαβα fold appears to be the result of an ancient gene duplication event, and is made up of two pseudosymmetric αβα core domains, each composed of five β-strands and three α-helices. Each αβα core superimposes with an αβα domain in phosphoglucomutase.^9,86 The CHS dimerization interface contains both hydrophobic and hydrophilic residues, and is rather flat, except for a pair of N-terminal helices that lay entwined across what we will refer to as the “top” of the homodimer (Fig. 1). This interhelical packing contributes to the extensive (1580 Ų) surface area buried by the dimer interface,⁴⁴ but this helical region exists in multiple conformations in some crystal structures of CHS, as well as in other CHS-like enzymes such as 2-PS and STS. While these solvent-accessible helices are unlikely to affect the CHS reaction, they may, as in yeast thiolase, contain intracellular localization signals.⁸⁷ These amino-terminal helices might also undergo a conformational rearrangement to participate in the formation of transient multi-protein complexes with other enzymes that constitute part of one or more bifurcating phenylpropanoid biosynthetic pathways. Interestingly, while the structurally similar E. coli KAS III enzyme lacks these intertwined N-terminal extensions, the divergent KAS III from M. tuberculosis possesses similarly intertwining (but non-helical) N-terminal extensions.⁴⁰

The CHS homodimer contains two distinct bi-lobed active site cavities, situated at the bottom edge of each monomer's conserved αβαβα core (if the N-terminal helices are considered to be on the top). Identical six-residue loops from each monomer, which meet at the dimer interface, separate the active sites of the two monomers from each other. Each of these loops begins with residue Thr132 in the active site cavity and ends with a cis-peptide bond to Pro138. The adjacent side-chain of Met137 forms the only contribution of either polypeptide chain to the opposing monomer's active site cavity, and this thioether side-chain neatly plugs a hole in the other monomer (Fig. 1). A second series of larger loops specific to KAS III and type III PKSs surround the bottom half of the active site, and form an additional domain. Thus, the active site cavity is buried, except for a 16 Å CoA-binding tunnel opposite the dimer interface that connects the catalytic surface to the outside aqueous milieu. Interestingly, the width of this tunnel is slightly too narrow for the aromatic substrates and products which must pass through it, indicating that at least some dynamic mobility within and around this tunnel must occur in solution. Furthermore, in a single apo E. coli KAS III crystal structure, the lower domain that forms the floor of the CoA-binding tunnel is extensively disordered.²⁷ While this may be an artifact of crystallization, it might also indicate the presence of sufficient conformational flexibility to accommodate bulky substrate thioesters. However, no significant disorder in this lower domain has been observed in any of the other dozen or so available apo or complexed KAS III or type III PKS crystal structures.

2.2.2 CHS mechanism. Under the early “flavanone synthase” misnomer, parsley CHS was determined to have optimal activity at 30 °C, at pH 8, with a salt concentration between 0.1 and 0.3 M potassium phosphate, and without the use of cofactors.⁸⁸ Malonyl-CoA was shown to decarboxylate prior to its condensation with the p-coumaroyl starter molecule.⁸⁹ Further purification allowed researchers to confirm both the homodimeric nature of CHS and the absence of active site modifications such as pantetheinylation.⁹⁰ Once CHS was cloned, site-directed mutagenesis allowed the identification of the catalytic cysteine.⁹¹ Subsequently, Schröder's group also elucidated the independence of each monomer's active site, in elegant experiments using heterodimers obtained by tandem coexpression of both active and inactivated monomers.⁹²

The CHS reaction includes an acyltransferase activity that loads the p-coumaroyl starter moiety onto the catalytic cysteine, a decarboxylase activity that activates malonyl-CoA, an iterative condensing activity that couples the resulting acetyl anion to the growing ketide chain, a cyclase activity that forms the cyclized polyketide precursor of chalcone via an intramolecular Claisen condensation of the linear tetraketide intermediate, and finally an aromatase-like activity (Fig. 8). In CHS, the cyclization activity also performs the function of a thioesterase, by severing the polyketide's covalent attachment to the active site cysteine. The alfalfa CHS crystal structure⁴⁴ revealed the configuration of residues in the active site responsible for many of these activities, which led to a number of additional experiments to probe the CHS mechanism.

Examination of the CHS crystal structure led to the identification of three conserved residues including the catalytic cysteine. This catalytic triad is likely to participate in the multiple decarboxylation and condensation reactions necessary during the course of the CHS-catalyzed mechanism.⁴⁴ Subsequent mutational and kinetic analysis of these residues,⁴⁵ along with pH-dependent inactivation studies,^93,94 provided further clarification of the roles of these residues in the CHS mechanism.

The core chemical machinery of the type III PKS active site is a catalytic trio of residues consisting of Cys164, His303, and Asn336 (Figs. 6 and 8). Positioned at the top of the active site cavity, these residues act on substrates inserted into the active site by the pantetheine arm of CoA-linked molecules. Interestingly, the catalytic cysteine is conserved in all thiolase-fold enzymes, and is located at the N-terminal cap of one of a pair of buried amphipathic α-helices that make up the buried central α-layer of the thiolase-fold architecture (Fig. 5). A mechanistic role for these helices was proposed for thiolase⁸ and has since been invoked for other thiolase-fold enzymes. In model peptides, the placement of a cysteine near the N-terminus of an α-helix lowers its pK_a from 8.8 to 7.2.⁹⁵ The central thiolase-fold α-helices are presumed to exert the same effect, but this mechanistic inference has not been directly tested. The superimposability of the cysteine with the catalytic serine residue at the N-terminal cap of a similar helix in the unrelated serine protease family has also been cited as circumstantial evidence for this mechanistic proposal.³⁴

An “oxyanion hole” formed by protonated nitrogens on His303 and Asn336's adjacent side-chains positions the thioester carbonyl oxygen of CoA-bound substrates, and provides an electron sink to stabilize the tetrahedral transition state formed during both the transfer of starters to the catalytic cysteine and upon attack of the acetyl carbanion (derived from malonyl-CoA decarboxylation) on thioester-linked substrates. Furthermore, the activation of acetyl groups via the decarboxylation of malonyl-CoA is promoted through the stabilization of the enol tautomer by this oxyanion hole (Fig. 8(a)). This latter activity does not require the participation of the Cys164 nucleophile, which is within hydrogen bonding distance of His303. Both mutational and pH-dependent inactivation studies preclude a role as a general base for CHS's His303 Nε, which instead forms a stable imidazolium-thiolate ion pair with Cys164.^45,93,94 Mutations of CHS His303 to glutamine and alanine shift the cysteine's pK_a from 5.5 to 6.6 and 7.6, respectively.⁹³ The still somewhat acidic pK_a of the H303A mutant's cysteine most likely results from the helical dipole effect described above, but may also be influenced by other adjacent active site residues. Conversely, the Cys164 thiolate also modulates His303's decarboxylation activity. Loading of a starter molecule neutralizes the thiolate, which stimulates decarboxylation activity. Conversely, in the absence of a starter, the thiolate-imidazolium pair is less likely to catalyze the unproductive decarboxylation of the malonyl extender unit, in a rather elegant example of enzymatic self-regulation.⁹⁴ Notably, the H303Q mutant retains near wild-type catalytic activity. The amide nitrogen of the mutant's Gln303 side-chain superimposes perfectly with the Nε of the wild-type enzyme's His303 side-chain, and thus serves as a reasonable mimic of the hydrogen-bonding arrangement of His303 to Cys164.⁴⁵

The hydrophobic side-chain of another conserved active site residue, Phe215, is also thought to contribute to decarboxylation by encouraging the formation of neutral CO₂ from the charged malonyl-CoA terminal carboxylate.⁴⁵ In addition, it likely serves to reposition the sigma bond of malonyl-CoA that undergoes cleavage (Fig. 8(a)). The predicted perpendicular arrangement of this sigma bond is such that the extended π-systems of the keto-enol tautomers are parallel with the newly released electrons. Interestingly, Phe215 is one of two “gatekeeper” phenylalanines that block the lower portion of the opening between the CoA-binding tunnel and the active site cavity,⁴⁴ not unlike the swinging saloon doors in a cowboy movie (Fig. 9). The conformational flexibility of these residues, inferred from comparisons of several crystal structures, may confer upon CHS the ability to limit the access of water to the active site, while accommodating substrates and intermediates of varying size.

Fig. 9 Buried active site cavity of CHS with naringenin bound. View is from the outside, looking through the CoA-binding tunnel. The interface of the active site cavity and the CoA-binding tunnel is mediated by the “gatekeeper” phenylalanines at positions 215 and 265. Other active site residues discussed in the text are also shown.

Following transfer of the starter moiety to the catalytic cysteine and the subsequent decarboxylation of malonyl-CoA with loss of CO₂, the resulting acetyl-CoA carbanion, stabilized by the His-Asn oxyanion hole, is poised to attack the cysteine-bound starter's thioester carbonyl carbon using a classical Claisen condensation reaction. This results in the transfer of the starter to the CoA-linked acetyl group, and its simultaneous elongation by two carbons (Fig. 8). The negative charge found on the malonyl's carboxylate moiety is thus retained by acetyl-CoA upon decarboxylation, and then used to restore the negative charge of the thiolate leaving group. The resultant CoA-linked diketide group is once again poised for loading onto Cys164, and another round of decarboxylation and condensation. The Cys164 thiol thus cycles between an anionic thiolate form and a neutral acylated form, with no additional protonation or deprotonation required.

This cycle is repeated until the elongated polyketide completely fills the active site, which in CHS occurs after three acetyl additions. Active site mutants such as T197L, which impair CHS's Claisen cyclization activity but not its condensing activity, still make tetraketide products, indicating that termination of elongation is independent of the cyclization activity of CHS.⁴⁶ The location of active site mutations which impair cyclization, the scarcity of reactive residues in the CHS active site, and the relatively acidic nature of the tetraketide intermediate's methylene β-carbons make it likely that intramolecular Claisen cyclization of the linear intermediate occurs spontaneously, possibly through an internal proton transfer upon proper elongation, folding, and exclusion of bulk water within the CHS active site.⁴⁴ A number of inert active site residues, including Thr132, Ser133, Thr194, Thr197, Gly256, Phe265, and Ser338 (Fig. 9), are conserved in CHSs but vary in other type III PKSs, and are likely critical for providing the necessary steric guidance to the linear polyketide.⁴⁴ The contribution of polarity to these interactions is currently under investigation in our laboratory. Moreover, it is not established whether the proper cyclization conformation is achieved while the linear tetraketide's C1 carbon is attached to the active site cysteine or to the sulfhydryl group of enzyme-bound CoA. Nevertheless, the position of these C1-derived carbons observed in naringenin and resveratrol crystallographically determined complexes with CHS implicates the active site cysteine as the likely choice for final attachment prior to cyclization.⁴⁴

Except for thiolases themselves, which instead of decarboxylating a malonyl thioester employ a second catalytic cysteine to deprotonate acetyl-CoA, thiolase-fold enzymes use similar or analogous collections of lysines, histidines, or asparagines to interact with substrates and modify the conserved catalytic cysteine.³⁴ However, only KAS III enzymes, which catalyze a single addition of malonyl-ACP to a CoA-thioester starter, share the Cys-His-Asn catalytic triad used by CHS and other type III PKS enzymes. Despite this similarity, CHS and KAS III differ in their use of this catalytic machinery. While both enzymes use the His303-Asn336 oxyanion hole to bind and decarboxylate the malonyl moiety for polyketide elongation, KAS III utilizes an additional oxyanion hole, formed by the backbone amides of the catalytic cysteine and a nearby glycine residue, to orient and stabilize the carbonyl oxygen of the bound starter molecule. This additional backbone amide oxyanion hole is abolished in type III PKSs by an absolutely conserved proline (the only residue with a substituted backbone amide) at position 375 (Fig. 9).³⁴ Lacking this auxiliary oxyanion hole, CHS uses only the His303-Asn336 oxyanion hole for both loading starter molecules on Cys164 and for malonyl-CoA decarboxylation.⁹³ Looking down the CoA-binding tunnel from the surrounding solvent, KAS III orients starters to the right and towards the dimer interface, while in CHS starters are positioned to the left. Interestingly, even the divergent KAS III enzyme from M. tuberculosis that has evolved to utilize a bulky starter does so by creating a new binding pocket, downward and to the right, thus preserving its ability to use this second oxyanion hole.⁴⁰ The abolition of this second oxyanion hole by the presence of Pro375, rather than specificity for a particular carrier moiety or starter molecule, may well have been the watershed event in the evolutionary divergence of type III PKS enzymes from their KAS III ancestors.

Aside from a number of van der Waals contacts, CHS crystal structures determined in the presence of CoA reveal four residues that form hydrogen bonds with CoA.⁴⁴ Helix-2 forms the lower left side of the outer rim of the CoA-binding tunnel, and positions three solvent-exposed residues, namely Lys55, Arg58 and Lys62, for electrostatic interactions with two CoA phosphates. These basic residues are each highly conserved in plant CHS-like enzymes, but conservative substitution (or transfer of the basic side-chain to an adjacent residue) occurs in a small number of plant type III PKSs. A fourth highly conserved interaction with CoA forms between the backbone amide nitrogen of Ala308, found on the N-terminal end of helix-9, and the pantetheine moiety's outer carbonyl oxygen. In some type III PKS enzymes, the adjacent Pro307 is replaced with a lysine or an arginine, which can also contribute to CoA-binding. The subsequently determined crystal structure of 2-PS (see Section 3.2.1) demonstrates such an interaction, and also reveals a few other contacts with CoA not observed in CHS,⁴⁶ confirming that slight variations in CoA-binding are permitted within the type III PKS superfamily. The deletion or addition of CoA interactions may alter the kinetics of association and dissociation of CoA thioesters and thus influence the ultimate fate of type III PKS reaction intermediates.

While CoA binds to KAS III enzymes in a similar position, the E. coli FabH crystal structure reveals an alternate set of conserved residues used by this enzyme family to stabilize the bound cofactor. Additionally, the pantethiene arm of KAS III-bound CoA penetrates that enzyme's active site more deeply (by 1.8 Å) than does type III PKS-bound CoA. While these observed differences in CoA binding might reflect KAS III's need to maintain specificity for both CoA and ACP, the functional relevance of these changes has not been tested.

The intermediate CoA-linked polyketide products are believed to diffuse in and out of the CHS active site, based upon the following evidence. First, as the concentration of 2-mercaptoethanol or other thiol-bearing reducing agents is increased in in vitro assays, less chalcone and more CHS truncation products (discussed in Section 3.1.3) are produced.⁸⁸ Given the buried nature of the CHS active site, these products most likely form as a result of the diffusion of CoA-linked intermediates into the surrounding solution, where reducing agents can access and sever their CoA-thioester bonds. Likewise, linear polyketides form heterocyclic lactones when acidified in aqueous solution, and acidification of cyclization-impaired CHS mutant reaction mixtures results in a correspondingly increased yield of lactone products.⁹⁶ Additional supporting evidence for the exchangeability of reaction intermediates with the bulk solution comes from studies of the biosynthesis of reduced chalcone products, resulting from the combined action of CHS and CHR (introduced in Section 2.1.2). CHR is believed to reduce its target oxygen when presented with a CoA linked linear polyketide intermediate, prior to the CHS-catalyzed cyclization and aromatization that eliminates the resulting alcohol.²⁸ Alternatively, a CHR-mediated reduction of the unstable cyclized trione intermediate prior to aromatization would produce the same result (Fig. 8). The buried nature of the CHS active site precludes CHR from accessing the bound linear intermediate polyketide, which again suggests that CHS reaction intermediates can readily exchange with the bulk solvent. And finally, CHR is able to reduce only a portion of the chalcone produced in mixed assays with CHS,^28,30 as would be expected if further elongation, cyclization, and/or aromatization of polyketide intermediates competes kinetically with CHR-mediated reduction.

2.2.3 Remaining mechanistic questions. While a reasonable mechanistic view of chalcone biosynthesis now exists, a small but critical set of mechanistic questions remain to be addressed. These unanswered questions have thus far resisted suitable interpretations due to our present inability to effectively isolate intermediate steps of the CHS reaction. The multifunctional characteristics of individual catalytic residues in the CHS active site coupled with the iterative nature of the biosynthetic reaction complicate this process. While selective impairment of the enzyme by mutation or covalent modification has met with some success in the study of individual catalytic roles of residues and the isolation of particular mechanistic steps, most mutations impair all or none of the activities under examination. Steady-state kinetic analysis of wild type and mutant CHS enzymes provides important insight into CHS function,⁴⁵ yet these analyses, even when viewed in light of the CHS structures, inadequately address a number of remaining mechanistic questions. For example, are the three malonyl-mediated decarboxylation and condensation steps kinetically and mechanistically equivalent, or does acylation by mono-, di- or tri-ketide intermediates differentially affect these activities? This same question obscures issues related to the reactivity of CHS with alternate starters. For instance, is decreased activity with an alternate starter moiety always the result of differences in enzyme affinity for the CoA-linked thioester, or is the rate of extension influenced by the prior loading of some starters? Crystallographic analyses have thus far failed to trap bound ketide intermediates, and although crude single- or double-turnover experiments have been successful, a thorough and systematic structural and functional examination of intermediate biosynthetic steps has yet to be achieved. Impediments include the inability to trap and stabilize polyketide intermediates or effectively monitor the binding of various substrates in real time. Spectroscopic monitoring of CoA-binding, while suitable for comparing wild type and mutant enzymes, is inappropriate for direct comparisons of divergent native enzymes.

Another clouded issue concerns the movement of intermediate CoA-thioesters into and out of the active site. Evidence outlined in Section 2.2.2 suggests such movement occurs. However, the kinetics of these events, and the relationship of such on and off rates to additional chemical steps during the mechanistically complex series of biosynthetic steps catalyzed by CHS, remain elusive. Several experimental problems, in addition to the microscopic kinetic concerns discussed above, combine to obscure the issue of polyketide intermediate diffusion. The difficulty of stabilizing and separating these linear intermediates from the pool of available substrates in vitro contributes to mechanistic uncertainty. Acidification of reaction mixtures facilitates extraction of tri- and tetraketides via lactone formation, but whether such lactonization occurs in bulk solution or when bound to the enzyme has not been studied. For instance, what percentage, if any, of the lactonization of CHS's intermediate linear polyketides leading to truncation products such as bis-noryangonin and CTAL (see 3.1.3), or lactone products of divergent enzymes such as 2-PS (see Section 3.2.1) occur while linear intermediate-CoAs or Cys164-linked thioesters are bound in the active site? Furthermore, an understanding of the diffusion of intermediates based on the product distribution of in vitro reactions carried out in the presence of CHS and CHR is complicated by an incomplete appreciation of substrate channeling. While channeling probably occurs in vivo (see Section 2.1.2), CHS-CHR complexes have not been observed crystallographically or in solution, even at high concentrations, and therefore might require the presence of a particular enzyme isoform, additional scaffolding proteins, or other enzymes in the pathway.

Many of these and other fundamental mechanistic questions may be addressed using creative approaches that employ some combination of substrate, polyketide intermediate, or carrier moiety mimics amenable to spectroscopic analysis, the implementation of stop-flow and quenched-flow technology, or time-resolved structural techniques including NMR spectroscopy and X-ray crystallography.

3 Evolutionary divergence of plant type III PKSs

3.1 Introduction to type III PKS mechanistic divergence

3.1.1 CHS gene duplication and divergence. Although the details surrounding the evolutionary history of type III PKSs including their likely divergence from KAS III ancestors has not been fully elucidated, the complexity accompanying catalysis by chalcone synthase is unlikely to have been the first type III PKS activity to emerge. It is more likely that divergent catalytic steps such as the ability to utilize a phenylalanine-derived starter or the ability to iteratively elongate a typical KAS III starter arose earlier, with additional CHS-like activities evolving subsequently. Both of these scenarios are most likely to have produced either a linear product or a heterocyclic lactone. In fact, modern day type III PKSs appear to possess the biosynthetic vestiges of this early evolutionary history, as in the case of the CHS truncation products discussed in Section 3.1.3, or the 2-PS pyrone discussed in Section 3.2.1. The discovery of styrylpyrone natural products in the primitive plant Equisetum (horsetail) lends support to this hypothesis.⁹⁷ Nonetheless, a recently published study of Psilotum nudum reports a distribution of divergent and sophisticated type III PKS enzymes and activities, including CHS, STS, and phloroisovalerophenone synthase (VPS, see Section 3.2.7), in this most primitive surviving vascular plant.⁹⁸ As a result of this revealing Japanese study, it is now evident that both the emergence and divergence of type III PKS modern enzyme activities occurred earlier than expected in the obscure early lineage of terrestrial plants and their aquatic relatives.

However, CHS is by far the most common and widely distributed type III PKS in higher plants, and most or all of the divergent plant CHS-like enzymes characterized to date have arisen via extensive duplication and subsequent genetic variation of the chs gene.^99–101 This duplication provides CHS activity with functional redundancy, allowing the chs gene to mutate without endangering flavonoid biosynthesis. However, while bioinformatic analysis shows that grasses, for example, duplicate the chs gene approximately every 15–25 million years, these redundant chs genes have only a few million years to develop novel advantageous type III PKS phenotypes before random genetic drift introduces stop codons, dismantles the catalytic machinery, or interferes with proper protein folding.¹⁰² Bacterial CHS-like enzymes (Section 4), in contrast to their plant counterparts, exhibit so much sequence divergence that they may have evolved independently of CHS, either from more ancient type III PKS enzymes, or directly from KAS III ancestors. For the sake of clarity, however, all structure/function analyses of divergent enzymes presented here will be discussed in comparison to CHS.

3.1.2 Natural variation in initiation, elongation, and cyclization steps. The divergent CHS-like enzymes described below can differ from CHS in their preference for starter molecules, the number of acetyl additions catalyzed, and even in the mechanism of ring formation used to cyclize identical polyketide intermediates. Preferred type III PKS starter molecules include CoA thioesters of acetate, isovalerate, N-methylanthranilate, and more subtle variations of the normal p-coumaroyl starter moiety. The number of elongation steps varies between one and four, but most family members catalyze three extensions. Methylmalonyl-CoA appears to be incorporated in some cases (Section 3.2.10), but malonyl-CoA is by far the most common in vivo type III PKS source of acetate. Cyclization options for the extended polyketide include no cyclization, heterocyclic lactone formation, and carbon–carbon bond formation through intramolecular Claisen or aldol condensations. Divergent plant type III PKS reactions are summarized in Fig. 10.

Fig. 10 Comparison of the reactions and products of known divergent plant type III polyketide synthases (see text for details). The position and type of cyclization reaction (Claisen, aldol, or lactone) of each presumed linear polyketide intermediate is depicted. One reaction that is not catalyzed by a type III PKS, namely the reduction of benzalacetone to pHPB by benzalacetone reductase (BR), is also included. Other post-PKS modifications are shown in Fig. 13.

3.1.3 Truncation products of the CHS reaction. A small percentage of CHS reaction intermediates on the path to chalcone become permanently derailed, leading to the production of various side products. Identification of derailment products containing the coumaroyl moiety and less than three acetyl unit additions helped early researchers establish the order of catalytic steps during the CHS reaction.⁸⁸In vitro assays produce greater amounts of all of these truncated reaction products when excessive amounts of either CHS or reducing agent are employed, making it difficult to assess the extent, probable occurrence, and relevance of CHS derailment products in vivo. However, the bioactive potency of some flavonoid products is known to be enhanced when they are present as mixtures;⁵⁶ the presence of CHS “derailment” products in vivo may offer some similar advantage.

Derailed triketide and tetraketide intermediates are isolated as the heterocyclic lactones bis-noryangonin and coumaroyl triacetic acid lactone (CTAL), respectively (Fig. 8). A divergent type III PKS enzyme, coumaroyl triacetic acid synthase (CTAS), that produces coumaroyl triacetic acid as its major product in in vitro assays with standard substrates has been characterized,⁹⁶ and is discussed in Section 3.2.8. Acidification of CTAS reaction mixtures prior to organic extraction greatly increases the yield of the extracted CTAL lactone product. Unfortunately, this result is consistent with each of two mechanistic alternatives: either an acid-catalyzed lactonization of a linear product in solution, or the acid stabilization of a lactone originally formed within the PKS active site.

3.1.4 In vitro promiscuity versus in vivo substrate availability. Another potentially confusing feature of type III PKS enzymes is their rather promiscuous starter molecule specificity. CHS will elongate a wide variety of CoA thioester starters in vitro, including a number of substituted cinnamoyl, benzoyl, and short- and medium-length fatty acyl substrates.^103–107 However, the incorporation of a sub-optimal starter moiety often adversely affects the terminating cyclization reaction, yielding predominantly linear or lactone derailment products. Additionally, in vitro use of much larger starters often results in fewer acetyl unit additions, further demonstrating the interdependence of the various activities catalyzed within the type III PKS multifunctional active site cavity.

While the promiscuity of CHS-like enzymes complicates the assessment of in vivo biological function by making it difficult to categorize the significance of in vitro starter usage, overall trends (preference for larger or smaller, aromatic or aliphatic, straight-chain or branched starters) provide clues about the likely physiological substrate. Relative activities with a range of substrates must be weighed against both the in vivo availability of CoA thioesters and the cellular distribution of products. In fact, the promiscuity of CHS-like enzymes suggests that changes in the in vivo availability of particular substrates may initially play a more significant role than mutational variation in the emergence of new type III PKS activities, with subsequent mutational reinforcement of desirable activities. If true, this hypothesis implies that CHS's loose substrate specificity may be an evolutionary strategy for the rapid emergence of new activities. This is an attractive hypothesis, as many of the natural products synthesized by divergent plant type III PKSs provide resistance to fast-evolving microbial pathogens. Nonetheless, all divergent type III enzymes characterized to date exhibit one or more altered in vitro characteristics relative to CHS. These changes can include increased affinities for alternate starters, variable numbers of extension reactions, loss or gain of intramolecular aldol or Claisen cyclization activity, or a combination of these factors.

Interestingly, the ancestral KAS III enzymes are also quite promiscuous, accepting a broad range of linear and branched short chain fatty acyl substrates.¹⁰⁸ A similar change in in vivo substrate availability, perhaps due to the emergence of the general phenylpropanoid pathway or relocalization of the enzyme to the ER, might also have lead to the initial divergence of type III PKS enzymes from KAS III enzymes, prior to the alteration of starter positioning due to the aforementioned Pro375 mutation.

3.1.5 The “steric modulation” model for type III PKS functional diversity. Regardless of how type III PKS activity emerged from fatty acid biosynthesis, the modern-day type III PKS enzyme superfamily exhibits remarkable functional diversity. Comparison of the original crystal structure of alfalfa CHS to an alignment of related sequences immediately suggested a probable mechanism by which nature achieves this diversity (Fig. 11).⁴⁴ In all cases examined to date, the Cys-His-Asn catalytic triad inherited from the KAS III ancestor, as well as the type III PKS-specific Pro375 mutation that blocks KAS III-like starter positioning, are maintained. However, a number of relatively inert residues that define the shape and volume of the active site cavity vary substantially between functionally divergent superfamily members. At first glance, it appeared that steric modulation of the residues occupying these positions on the active site surface was all that was needed to achieve new substrate and product specificity (Fig. 11).

Fig. 11 (see previous page) Annotated sequence and secondary structure of alfalfa CHS. Key sections, highlighted in yellow, are aligned against the functionally divergent plant and bacterial type III PKSs, bacterial type III PKSs of unknown function (see text), as well as the evolutionarily related KAS III enzyme (FabH) from E. coli. CHS's catalytic triad, residues that contact bound CoA, and other residues important for functional diversity are highlighted in red, green, and blue, respectively. For clarity, only identical residues in the equivalent positions of the aligned sequences are highlighted, even though in some cases conservative substitutions may play an equivalent mechanistic role. Plant sequences:Ms_CHS = Medicago sativa (alfalfa) chalcone synthase, accession P30074; Gh_PYS (a.k.a. 2-PS) = Gerbera hybrida (daisy) methylpyrone synthase, accession CAA86219; Ps_STS = Pinus sylvestris (scots pine) pinosylvin-forming stilbene synthase, accession CAA43165; Bf_BBS = Bromheadia finlaysoniana (orchid) bibenzyl synthase, accession CAA10514; Hv_H/EDS = Hordeum vulgare subsp. vulgare (barley) homoeriodictyol/eriodictyol chalcone synthase, accession CAA70435; Rg_ACS = Ruta graveolens (common rue) acridone synthase, accession S60241; Ha_BPS = Hypericum androsaemum (tutsan) benzophenone synthase, accession AAL79808; Hl_VPS = Humulus lupulus (hop) valerophenone synthase (a.k.a. phloroisovalerophenone synthase), accession BAB12102; Hm_CTAS = Hydrangea macrophylla var. thunbergii (hydrangea) coumaroyl triacetic acid synthase, accession BAA32733; Rp_BAS = Rheum palmatum (rhubarb) benzalacetone synthase (a.k.a. p-hydroxyphenylbutenone synthase), accession AAK82824; Ps_CHS2 = Pinus strobus (white pine) CHS-like protein, accession CAA05214); Nt_ASCHSLE = Nicotiana sylvestris anther-specific CHS-like enzyme, accession CAA74847. Bacterial sequences: Sg_RppA = Streptomyces griseus 1,3,6,8-tetrahydroxynaphthalene synthase, accession BAA33495; Sc_THNS = Streptomyces coelicolor 1,3,6,8-tetrahydroxynaphthalene synthase, accession CAC01488; Sa_PhlD_homol = Streptomyces avermitilis putative PhlD homologue, accession BAB69299; Pf_PhlD = Pseudomonas fluorescens PhlD (see text), accession AAB48106, Ao_DhpA =Amycolatopsis orientalis dihydroxyphenylacetate synthase, accession T17474; Am_DhpA =Amycolatopsis mediterranei dihydroxyphenylacetate synthase, accession CAC48378; Mt_PKS10 = Mycobacterium tuberculosis CHS-like protein, accession CAB06631; Mt_PKS11 = Mycobacterium tuberculosis CHS-like protein, accession CAB09101; Bs_CHS-like = Bacillus subtilis CHS-like protein, accession NP_390087; Bh_CHS-like = Bacillus halodurans CHS-like protein, accession NP_241483; Dr_CHS-like = Deinococcus radiodurans CHS-like protein, accession AAF11641; Rc_CHS-like = Rhodospirillum centenum “CHSA” CHS-like protein, accession AAD43969. KAS III sequence: Ec_FabH = Escherichia coli KAS III (see text), accession NP_415609.

This steric modulation model was supported by the crystallographic elucidation of the structure of 2-pyrone synthase (2-PS; Section 3.2.1), a type III PKS that differs from CHS in starter usage, elongation steps, and cyclization chemistry. Just as a CHS-based homology model predicted,⁴⁴ the only significant structural difference between CHS and 2-PS, which share 74% sequence identity, is the reduction of active site volume in 2-PS by the introduction of bulkier side-chains at three key positions.⁴⁶ Notably, this reduction in volume is achieved without rearrangement of the protein main-chain. As further proof, introduction of these three mutations into alfalfa CHS fully converted that enzyme into a functional 2-PS.⁴⁶ To further test the effects of active site volume on specificity, the steric bulk of a single CHS side-chain (position 256) was modulated by site-directed mutagenesis, and the mutants assayed with various starters. As expected, the observed trends in starter usage and number of extensions correlated with active site volume.¹⁰⁹

While this work validates the simple steric modulation model for type III PKS functional elaboration, the repeated emergence of STS (Section 3.2.2) activity from CHS⁹⁹ does not fit this general paradigm, suggesting additional routes for the evolution of type III PKS mechanistic diversity. STS and CHS generate the same tetraketide intermediate, but differ in their cyclization mechanism and specificity. Alignments and CHS-derived homology models of STS failed to identify an STS consensus sequence or any structural basis for this divergent activity. Crystal structures of two STSs accompanied by mutagenic modulation of CHS cyclization specificity by the authors (manuscripts in preparation) confirm that alterations of the backbone position not predicted by homology modeling are the causative agent for cyclization changes. Brought about by amino acid substitutions remote from the active site, this backbone rearrangement subtly alters the position and hydrogen-bonding pattern of Thr132 and Glu192, two active-site residues conserved in both enzymes. Rather than using steric changes to achieve a different productive folding of the intermediate, as was previously assumed, these changes instead appear to drive the aldol-based cyclization mechanism of STS by facilitating a thioesterase-like hydrolysis activity.

Although 2-PS and STS are the only divergent type III PKS enzymes that have been structurally characterized to date, their examples illustrate both the value and pitfalls of using homology models to attempt to explain the structural basis for type III PKS functional diversity. However, the failure of STS homology models to predict important differences appears to be the exception rather than the rule. For each of the other divergent enzymes discussed below, sequence alignments and homology models highlight active site differences that are likely to affect specificity (Fig. 11). While these substitutions may not alone be sufficient to interconvert the various type III activities, they represent obvious starting points for mutagenic experimentation aimed at exploring the underlying evolutionary and mechanistic principles governing type III PKS functional divergence.

3.2 Plant type III PKS activities associated with divergent products

3.2.1 2-Pyrone synthase (2-PS). A CHS-like enzyme incapable of utilizing p-coumaroyl CoA as a starter molecule was cloned from the ornamental daisy Gerbera hybrida.¹¹⁰In vitro, GCHS2 was shown to synthesize 6-methyl-4-hydroxy-2-pyrone [methylpyrone or triacetic acid lactone (TAL)], by catalyzing two condensations of malonyl-CoA with an acetyl-CoA starter molecule (Fig. 10). If acetyl-CoA is not available, the enzyme can decarboxylate malonyl-CoA to make its own acetyl starter, but with ten-fold less efficiency.¹¹¹ Transformation of the daisy with an antisense construct of the gchs2 gene abolished the biosynthesis of gerberin and parasorboside, two natural products possessing a reduced methylpyrone scaffold (Fig. 12).¹¹¹ These glucoside compounds and their derivatives, found in various plant species, are known to inhibit bacterial and fungal growth, and also deter feeding by insects.¹¹¹ GCHS2 is expressed in both vegetative and reproductive tissues, and at different times than either of two known CHS enzymes in Gerbera.¹¹⁰

Fig. 12 Examples of several natural products mentioned in the text that are derived from plant type III polyketide synthases. PKS product-derived portions are highlighted in red.

Using gchs2 as a probe, a number of similar enzymes were cloned from G. hybrida and four other species. Comparison of these sequences to CHS indicates that this 2-pyrone synthase (2-PS) subfamily of type III PKS enzymes probably arose from a single chs gene duplication event, prior to the evolutionary diversification of the large Asteraceae family of angiosperms.¹⁰¹

The 2-PS enzyme encoded by the gchs2 gene, renamed g2ps1 to reflect its true biochemical function, has been extensively characterized, yielding new information about the mechanism of functional divergence in type III PKS enzymes. A number of small hydrophobic CoA-thioester starters, including propionyl-CoA, butyryl-CoA, and isovaleryl-CoA, are accepted in vitro by 2-PS. Although smaller than the rejected phenylpropanoid starter by only two carbons, benzoyl-CoA is also utilized by 2-PS in vitro, undergoing two acetyl additions followed by lactonization to produce 6-phenyl-4-hydroxy-2-pyrone. This latter compound is of interest as a scaffold for a family of HIV-1 protease inhibitors.¹¹¹

As discussed in Section 3.1.5, a 2-PS homology model based on the alfalfa CHS structure predicted a smaller active site cavity due to the increased steric bulk of three active site residues, with no other significant changes in structure, despite the hundred or so amino acid differences between CHS and 2-PS.⁴⁴ The atomic resolution crystal structure of the G. hybrida 2-PS was determined in our laboratory, alone and with acetoacetyl-CoA bound.⁴⁶ The structure confirmed the validity of the homology model's active site predictions by revealing the active site cavity of 2-PS to possess only a third of the volume observed in CHS. Otherwise, the 2-PS and CHS structures are remarkably similar, except for the minor rearrangement of a three-residue solvent-exposed loop at the mouth of the active site, which the 2-PS homology model failed to predict. The resulting interaction of this loop with the adenine ring and phosphates of CoA, as well as other interactions with CoA not observed in CHS, make CoA binding more extensive in 2-PS.⁴⁶

However, the mutation of three CHS active site cavity residues to their 2-PS counterparts is sufficient to make alfalfa CHS functionally identical, both in terms of specificity and kinetics, to 2-PS in in vitro assays.⁴⁶ This CHS triple mutant consists of the following changes at positions that form a triangle in the CHS active site cavity: T197L, G256L, and S338I. An increase in steric bulk at positions 197 and 338 effectively closes off the coumaroyl-binding pocket, excluding large substrates such as phenylpropanoid thioesters, while the increased bulk at the 256 position decreases the size of the putative cyclization lobe of the active site cavity.⁴⁶ The remarkable functional conversion of CHS into 2-PS by changing less then 1% of their differing residues supports an intuitively simple model of the steric modulation hypothesis, while discounting the relevance of the observed CoA-binding differences to 2-PS's functional divergence from CHS.

While the many differences between the 2-PS and CHS reactions complicate the analysis of this mutagenic conversion, single point mutations of alfalfa CHS at these three positions proved useful in the dissection of their relative contributions to the observed functional differences.⁴⁶ Not unexpectedly, each of the mutations comprising the triple mutant, when isolated, abrogate CHS's ability to make chalcone. More surprising is the ability of each of the T197L and S338I point mutants, but not the G256L mutant, to catalyze three acetyl extensions of p-coumaroyl-CoA. Paradoxically, these changes nearer the putative starter-binding pocket influence the terminating cyclization to a much greater extent than starter molecule specificity. Similarly, all three-point mutants exhibit increased utilization of acetyl-CoA as a starter, relative to wild-type CHS.⁴⁶ These results serve as an important reminder of the interconnectedness of specificity for starter usage, chain extension, and cyclization chemistry.

There are a few other important lessons to be learned from these experimental results. First, as suggested by bioinformatic analysis of the natural diversity within the CHS superfamily, the type III PKS condensing machinery is remarkably unaffected by modifications to surrounding active site residues. Moreover, the number of extensions catalyzed is more dependent upon the volume of the active site cavity than on its specific shape. This point is reinforced by the observation that assorted CHS point mutants retain the ability to form the appropriate tetraketide intermediate, but are unable to carry out the terminal intramolecular Claisen cyclization step. In keeping with the observed range of CHS products synthesized in vitro from a diverse collection of starters,^103–106 the correct choice of starter molecule seems to be crucial for achieving the conformation of the polyketide intermediate leading to the intramolecular Claisen condensation, but much less important for polyketide chain extension. And finally, the substrate preferences of 2-PS, CHS, and the intermediate CHS point mutants⁴⁶ demonstrate that type III PKS starter molecule specificity has both upper and lower size limitations. Exclusion of bulky phenylpropanoids from the small 2-PS active site is less surprising than the low affinity of CHS for small starters like acetyl-CoA. The ineffective transfer of small starters to CHS's catalytic cysteine might be, as previously suggested,⁴⁶ due to their increased conformational flexibility within the large p-coumaroyl-binding pocket. Alternatively, the inability of smaller substrates to exclude nucleophilic water molecules from the active site might be a source of the problem. For whatever reason, the various starter-binding pockets of type III PKS enzymes, like Goldilocks, seem to prefer substrates that are neither too large nor too small.

3.2.2 Stilbene synthase (STS). Not all type III PKS enzymes are as forthcoming with their architectural secrets as 2-PS. STS enzymes synthesize the same linear tetraketide intermediate (from p-coumaroyl-CoA and three malonyl-CoAs) as CHS. However, while CHS cyclizes this intermediate using an intramolecular Claisen condensation between carbons C6 and C1 (numbering from the cysteine thioester), STS makes the anti-fungal phytoalexin resveratrol by linking C2 and C7 via an intramolecular aldol condensation, accompanied by an additional decarboxylative loss of the C1 carbon as CO₂ (Figs. 8 and 10). The timing and mechanistic relevance of this latter decarboxylation to the STS aldol cyclization reaction remains unresolved.^92,112 Resveratrol and related stilbene natural products occur in a limited, yet widely dispersed subset of the plant kingdom.⁹⁹ Early feeding experiments with rhubarb confirmed that its rhapontigenin stilbene product was produced by oxidation and methylation of resveratrol's B ring, rather than by incorporation of a more substituted phenylpropanoid starter into the STS reaction.¹¹³ Conversely, pine STS enzymes produce pinosylvin instead of resveratrol, due to an unusual preference for unsubstituted cinnamoyl-CoA instead of p-coumaroyl-CoA.¹¹⁴ However, this pinosylvin-forming STS accepts p-coumaroyl-CoA as an efficient substrate in vitro, so substrate availability most likely explains this in vivo preference.

The first purification of STS from induced peanut cell cultures in 1984 confirmed that the STS and CHS reactions, although similar, are catalyzed by different enzymes.¹¹⁵ Cloning and further analysis revealed significant sequence homology between CHS and peanut STS.¹¹⁶ Apart from the divergent cyclization specificity, no significant mechanistic differences between CHS and STS were detected.^91,92 The subsequent cloning of STSs from pine¹¹⁷ (a gymnosperm) and grapevine¹¹⁸ (an angiosperm) facilitated a phylogenetic comparison of STS sequences with each other and with various CHSs, all of which are 60–90% identical at the amino acid level. This study indicated that STS activity has evolved from CHS on more than one occasion, but failed to identify the STS consensus sequence that could be linked functionally with the mechanistic basis for such divergent cyclization activity.⁹⁹ In contrast to the case of 2-PS, STS homology models based on the CHS structure reveal no significant differences. Although no evidence for a “steric modulation” explanation for the emergence of aldol cyclization specificity in STS enzymes exists, it had been widely assumed that divergence was achieved through folding of the tetraketide intermediate into an alternative conformation conducive to aldol condensation. Recently, using scaled-up in vitro assays and careful analysis of the resultant products, Sankawa and co-workers found that both STS and CHS produce small amounts (1–5% of major product yields) of each other's cyclization product. Additionally, this study demonstrated that STS, like CHS, also generates styrylpyrone side products.¹¹⁹

STSs and their products have recently attracted much attention, because of the agricultural and medicinal properties of their small molecule products. STS enzymes occur naturally in a subset of plants, but synthesize the antifungal phytoalexins resveratrol and pinosylvin using substrates common to all higher plants. Given this, STS encoding genes are an obvious target for transgenic crop enhancement. Resveratrol synthases from peanut and grapevine have been expressed in tobacco and alfalfa, respectively. In both cases, considerable resistance to pathogenic fungi was conferred upon the heterologous host plants.^120,121 In terms of disease prevention in humans, resveratrol has been shown to possess a number of beneficial medicinal activities, including copper chelation, anti-oxidant scavenging of free radicals, inhibition of both platelet aggregation and lipid peroxidation, anti-inflammatory activity, vasodilation, and anti-cancer properties.^122,123 Positive health effects of resveratrol are often attributed to the moderate consumption of red wine, the so-called “French paradox”.¹²⁴ Red wines generally contain higher amounts of resveratrol due to inclusion of the phytoalexin-rich grape skins and stems in the fermentation process for longer periods of time.

Very recently, these authors determined the crystal structures of both the pinosylvin-forming STS from P. sylvestris (pine) and the resveratrol-forming STS from A. hypogaea (peanut) (manuscripts in preparation). Both enzymes, although independently evolved from CHS through different sets of amino acid changes, exhibit similar main-chain conformational changes with respect to CHS. Structure-guided mutagenic conversion of CHS into STS confirmed that STS's alternative cyclization specificity is achieved by a conformational difference in the main chain of a short, buried loop that spans the dimer interface between the two active sites (residues 132–137). This movement is caused by various amino acid substitutions at buried positions adjacent to this loop, most notably the replacement of Val98 with a bulkier side-chain, and results in an altered active site hydrogen-bonding network around the slightly repositioned Thr132. This active site residue (conserved in both CHS and STS) changes its bonding interaction with Glu192 (conserved in all type III PKSs), and forms a hydrogen bond with a well-ordered water molecule adjacent to the catalytic cysteine. A three dimensional comparison of the CHS and STS active sites casts doubt upon models that attribute these enzymes' cyclization differences to alternative productive conformations of their shared linear tetraketide intermediate, achieved by steric differences between the CHS and STS active site cavities. Our preliminary results suggest instead that subtler active site changes, apparently of an electronic rather than steric nature, favor an alternative cyclization mechanism within the context of a quite similarly folded tetraketide intermediate. This latter model is also consistent with the cross-reactivity of CHS and STS previously observed by Ebizuka's group.¹¹⁹

Directed changes in three areas of primary sequence were required to achieve near-wild-type levels of STS activity in the converted CHS mutant. Considering that the separately evolved STS subfamilies vary considerably in each of these three areas of primary sequence, it is not surprising that previous (non-crystallographic) efforts in our laboratory and others have failed to identify all of these crucial changes, and have in some cases mistakenly identified spurious residues as mechanistically important. These examples contain valuable lessons for the interpretation of structure/function relationships. First, conclusions based on mutations of an STS/CHS chimera were reported in the same paper where the repeated evolution of STS was elucidated.⁹⁹ In this study, which predates any type III PKS crystal structure, a chimera consisting of the first 107 residues of S. alba CHS and 287 C-terminal residues of peanut STS was shown to lack activity. Mutating three residues in the N-terminal CHS-derived portion of this chimera to resemble STS (Q100E, V103M and V105R) resulted in very low STS-like activity, and the further mutation of G23T increased this activity to 25% that of wild-type STS. Position 103 corresponds to alfalfa CHS Val98, confirmed by our later structural work to be a crucial area of specificity determination for the conversion of CHS into STS. However, in retrospect the G23T mutation appears unrelated to any of the important architectural differences between STS and CHS, and is probably important only in the context of the folded and stabilized conformation of this artificial chimera. This illustrates an important point: within a conserved fold, the distribution and importance of some residues that confer rigidity or flexibility drifts in a compensatory fashion, as do various other compensatory changes in regions of primary sequence that associate in the folded protein. Such differences between divergent enzymes often reflect alternative ways to stabilize the same tertiary fold, and often have little to do with enzymatic differences. However, taken out of the context of the protein in which such changes evolved, these sequence differences can have seemingly important mechanistic effects, as the chimera's G23T change reflects.⁹⁹

In another example, Pro375, conserved in all type III PKSs, was mutated to glycine in both CHS and STS, resulting in impairment of both solubility and activity in both enzymes.¹²⁵ Mechanistic conclusions were argued for, citing the differential effects of this mutation on CHS and STS product distributions. Drawing conclusions about functional divergence from mutation of a conserved residue is risky, especially when overall activity or protein stability is greatly decreased by that mutation. The apparently greater effect of the P375G mutation on STS solubility as well as activity¹²⁵ is a warning that the extent to which this proline stabilizes the type III PKS fold has also diverged in these two enzymes.

3.2.3 Bibenzyl synthases (BBS) and dihydrostilbene biosynthesis. A related type III PKS activity is that of the bibenzyl synthases (BBS). These enzymes perform three acetyl elongations of reduced phenylpropanoid starters (dihydrocinnamoyl derivatives), and perform an STS-like (C2→C7) aldol cyclization, resulting in a dihydrostilbene product (Fig. 10). The double bond missing from the starter substrate is the same one formed by the PAL-catalyzed elimination of the amino group of phenylalanine (Fig. 7). The BBSs leading to dihydrophenanthrenes (Fig. 13) in orchids, as well as the putative dihydropinosylvin synthases from pine, will be considered sequentially here. Bibenzyl compounds also occur in liverworts, but their biosynthesis by liverwort STS-like enzymes has yet to be demonstrated. Liverwort bibenzyl-containing natural products are further discussed in Section 3.3.1, in the context of stilbenecarboxylic acid biosynthesis.

Fig. 13 The similar (post-PKS) cyclization reactions of three plant type III PKS natural products (see text for details).

Stressed or wounded orchid tissues, especially those infected by endomycorrhizal fungi, accumulate bibenzyl stilbenes and their tricyclic derivatives, 9,10-dihydrophenanthrenes (Fig. 13), presumably as antifungal agents.¹²⁶ An enzyme producing the bibenzyl product was purified, characterized, and subsequently cloned from a Phalaenopsis orchid, along with an O-methyltransferase enzyme found to be essential for the further conversion of bibenzyl into 9,10-dihydrophenanthrene.^127,128 This BBS clearly belonged to the CHS/STS superfamily, but preferred dihydro- m-coumaroyl-CoA as a starter over p-coumaroyl-CoA (6 vs. 0.2 pkat mg⁻¹ respectively, compared to 3 vs. 58 pkat mg⁻¹ for peanut STS), in in vitro assays.¹²⁸ Although the steady-state kinetics associated with type III PKS enzymes for their preferred substrates are usually quite similar, BBS's activity towards dihydro- m-coumaroyl-CoA is five-fold less than that of STS towards its preferred starter, p-coumaroyl-CoA.¹²⁹ These starters differ in both the position of ring hydroxylation and in the saturation of their propanoid moiety. Assays using substrate analogues demonstrated that BBS selects for each of these modifications over the more commonly used p-coumaroyl starter.¹²⁷

A study utilizing both mutant and BBS/STS chimeric enzymes attempted to elucidate the structural basis for this unusual substrate preference.¹²⁹ Chimeras were made by joining the C-terminal portion of grape STS to the N-terminal portion of BBS and vice versa, with residue 222 serving as the junction point. Although both chimeras exhibited 50-fold reductions in steady-state activity, likely due to mismatched pairing of drifting compensatory mutations as discussed in Section 3.2.2, the N-terminal segment appeared to dictate starter selection. These authors (in the absence of a three-dimensional structure) also made an STS mutant by replacing STS residues 230–233 with their BBS counterparts. In the context of a 25% loss of resveratrol synthase activity, this mutant exhibited a disproportionate increase in selectivity for the reduced starter. The mutant's rate of conversion of m-hydroxyphenylpropionyl-CoA into the bibenzyl product was 11% of its rate of conversion of p-coumaroyl-CoA into resveratrol (compared to only 0.35% in wild-type STS).¹²⁹ Notably, these surface-exposed loop residues are in fact located on the opposite side of the core αβαβα domain from the active site, and would seem unlikely to be important for determining specificity. The 25% loss of wild-type activity caused by these changes, which are remote from the active site, suggests that in the context of the STS protein they may be destabilizing to the proper, catalytically active fold.

However, a structure-based examination of the BBS sequence reveals a few residues that might directly influence starter molecule preference. CHS Val98 is replaced in BBS by an alanine, a substitution also found in other divergent type III PKS enzymes known to utilize starters possessing m-substituted phenyl rings (discussed in Sections 3.2.4 and 3.2.5). The substitution of CHS Met137 by leucine, also seen in pine STS, may work synergistically with Ala98. A third notable active site feature of BBS is the use of threonine in place of CHS's Gly211. This residue lines the front of the active site cavity, just underneath the gatekeeper phenylalanines located at the end of the CoA-binding tunnel. While this bulky substitution is unlikely to translate into increased specificity for m-substituted phenylpropanoid starters, it presumably changes the shape of BBS's starter-binding pocket. Substrates with reduced propanoid moieties possess increased conformational flexibility, which should give them an advantage over more rigid substrates in binding to a distorted starter-binding pocket. This advantage might be the basis of BBS's selectivity for reduced phenylpropanoid substrates.

The presence of dihydropinosylvins in pines is more difficult to explain at the current time, as a distinct dihydropinosylvin synthase (DPS) has not been conclusively identified. Dihydropinosylvins share the saturation of the bibenzyl linker with BBS products, but are derived from an unsubstituted dihydro-cinnamoyl starter, rather than an m-substituted starter. A putative DPS cloned from P. sylvestris and expressed in E. coli appeared to prefer dihydro-substrates,¹¹⁷ but better expression and purification revealed the enzyme to be a pinosylvin synthase (a pinosylvin-forming STS, Section 3.2.2).¹¹⁴ Moreover, the addition of P. sylvestris extract to the in vitro assay selectively inhibited the DPS activity of this enzyme even further.¹¹⁴ However, no further characterization of the reaction-modifying element contained in the pine extract accompanied these experiments, and the biosynthesis of dihydropinosylvins in pines remains a mystery.

Two more clues to the puzzle come from experiments designed to elucidate the cause of the stilbene cyclization specificity. First, the STS/CHS chimera with four additional mutations that exhibited 25% of the native STS enzyme's cyclization specificity, discussed in Section 3.2.2, preferred dihydro- p-cinnamoyl-CoA over p-coumaroyl-CoA, and was thus technically a dihydropinosylvin synthase rather than a resveratrol or pinosylvin synthase.⁹⁹ Since the STS template was from peanut, which makes only resveratrol stilbenes, this cannot be considered a physiologically relevant characteristic of the chimeric enzyme, but it may still offer insight into the basis of bibenzyl specificity. The second clue comes from another set of STS-probing experiments, where the −2 and −3 residues (relative to the catalytic Cys 164) were mutated to test their effect on cyclization specificity.¹³⁰ This 161–2 sequence is Gln-Gln in CHS (and in many plant type III PKSs), but Gln-His in pine STS and His-Gln in peanut STS. While mutagenic exchanges of these residues did not alter cyclization specificity as hoped, the peanut STS Gln-Gln mutant showed selectively reduced activity towards p-coumaroyl-CoA, while retaining dihydro-cinnamoyl-CoA activity. Thus, the peanut STS was effectively transformed into a dihydropinosylvin synthase by the H161Q mutation.¹³⁰ Interestingly, this residue is sandwiched at the dimer interface between the dimer's active site-spanning loops (residues 132–137) whose conformation differs in CHS and STS.

Taken together, these similar results with two different mutants of peanut STS implicate slight changes near the active site cavity that selectively reduce wild-type activity, with no harm to the dihydropinosylvin synthase activity. These results are consistent with the model proposed for BBS, where a flexible substrate compensates for a perturbed active site cavity. Aside from the issue of starter molecule binding, the additional flexibility of the tetraketide intermediate due to saturation of the starter molecule's propanoid moiety may help it achieve a folded conformation suitable for cyclization within a slightly contorted active site cavity.

3.2.4 Homoeriodictyol/eriodictyol synthase (HEDS or HvCHS). In addition to unsubstituted cinnamoyl-CoA and the 4-substituted p-coumaroyl-CoA, plants also commonly produce further substituted phenylpropanoid moieties, primarily used in lignin biosynthesis. These include caffeoyl-CoA, which has hydroxyls at both the 3 and 4 positions, and feruloyl-CoA, similar to caffeoyl-CoA but O-methylated at the 3 position (Fig. 14). Early research showed that CHS could use caffeoyl-CoA as a starter molecule, producing a 3-hydroxylated chalcone (leading to the 3′-hydroxylated flavanone eriodictyol).¹³¹ At CHS's optimum in vitro pH of 8, this activity with caffeoyl-CoA is kinetically less favorable than reactions using either p-coumaroyl- or cinnamoyl-CoA. Curiously, these same researchers found that the optimal pH for the caffeoyl-CoA reaction is 6.5. At this pH, CHS exhibits nearly equal levels of activity with either substrate (approximately two-thirds of the kinetic efficiency seen at pH 8 with p-coumaroyl-CoA). Previously, it was assumed that plants synthesized eriodictyol only via the (post-CHS) 3′-hydroxylation of naringenin.

Fig. 14 Phenylpropanoid CoA thioesters commonly found in plants (discussed in Sections 3.1.4 and 3.2.4).

Danish researchers have cloned an interesting chs-like gene (78% amino acid identity with the constitutively expressed barley CHS1) from a library prepared from Hordeum vulgare (barley) leaves infected with the fungus Blumeria graminis f.sp. hordei (bgh).¹³² Expression of the resulting HvCHS2 enzyme occurs 24–36 h after bgh inoculation, corresponding with a 500% increase in the eriodictyol-derived phytoalexin lutonarin. In vitro, HvCHS2 was shown to prefer the di-substituted caffeoyl- and feruloyl-CoA starters at physiological pH, with only minimal activity towards p-coumaroyl-CoA or cinnamoyl-CoA (Fig. 14).¹³² This unusual substrate specificity is the inverse of that shown by typical CHS enzymes. Activity profiles with all four of these CoA thioesters show that both enzymes exhibit decreasing activity towards substrates that are increasingly different in size from their preferred substrate (i.e. too large in CHS1 or too small in HvCHS2).¹³² Although HvCHS2 synthesizes pentahydroxychalcone from caffeoyl- and feruloyl-CoA (O-methylated at position 3 when feruloyl-CoA is the starter), the enzyme was labeled a homoeriodictyol/eriodictyol synthase (H/EDS) after the flavanone derivatives of these pentahydroxychalcone natural products.

These same authors, working before any type III PKS crystal structure was available, cited 27 positions where HvCHS2 differed from the CHS consensus sequence, and pointed out three changes that they found particularly notable: Q95H, A166G and G168A (alfalfa CHS numbering).¹³² Comparison with the CHS crystal structure, however, reveals two additional and perhaps more functionally important differences. First, substitution in the active site of an alanine for CHS's Thr197 very likely results in the larger starter-binding pocket needed to achieve HvCHS2's unusual specificity for bi-substituted phenylpropanoid-CoA starter molecules. This contrasts with 2-PS's use of a bulkier residue at this same position to decrease the 2-PS active site volume (Section 3.21).⁴⁶ While the T197A mutation is probably the most relevant change in HvCHS2, other mutations may reinforce the enzyme's preference for larger starters. One likely candidate is a V98A substitution, also observed in other plant type III PKSs that prefer meta- and ortho-substituted aromatic starter molecules (see Sections 3.2.3 and 3.2.5).

3.2.5 Acridone synthase (ACS). Although approximately 100 naturally occurring acridone alkaloids (Figs. 10 and 12) are known, they appeared to be restricted to plants of the Rutaceae family (including Citrus species), but have now also been reported in a single species of the Piperaceae family.^133,134 These plant defense alkaloid phytoalexins exhibit a number of other interesting biological activities as well, but are currently unsuitable for medicinal purposes due to their interference with DNA synthesis.¹³⁵

The three-ring acridone skeleton is made by acridone synthase (ACS), a CHS-like enzyme that catalyzes three condensations of malonyl-CoA to an N-methylanthraniloyl-CoA starter, followed by a CHS-like intramolecular Claisen cyclization of the tetraketide intermediate (Fig. 10).¹³³ Unlike the CHS reaction, formation of the heterocyclic middle ring is likely to occur prior to aromatization of the new phenolic ring possibly involving formation of a Schiff base, which is accompanied by elimination of a water molecule to produce the 1,3-dihydroxy-N-methylacridone product.¹³⁶ Initially purified from cell suspension cultures of Ruta graveolens (common rue),¹³³ several ACS isozymes have since been cloned from elicited or irradiated cell cultures and from immature flowers.^135–137 Although ACS has the unusual characteristic of eluting from gel filtration columns with an apparent molecular weight equal to between one and two monomers, sedimentation equilibrium experiments clearly show the active form to be a homodimer.¹³⁷

Interestingly, while neither Ruta CHS nor any other known CHS or STS can utilize N-methylanthraniloyl-CoA, wild-type ACS retains some CHS-like activity in vitro, producing naringenin chalcone from p-coumaroyl-CoA (with about 15% of the activity seen with N-methylanthraniloyl-CoA).¹³⁶ Like most divergent plant type III PKSs, the ACS protein sequence differs from CHS in about 100 places. Three of these changes (T132S, S133A, and F265V) are in the active site (Figs. 9 and 11), and were the subject of some recent and enlightening mutagenesis experiments. Mutation of these three amino acids in the ACS active site to their corresponding CHS residues seriously impaired the mutant ACS's ability to utilize its normal N-methylanthraniloyl-CoA starter, while greatly increasing CHS-like activity towards p-coumaroyl-CoA.¹³⁸ The corresponding reverse CHS triple mutant has not yet been reported, but the F265V mutation alone does not confer any ACS-like activity on alfalfa CHS.¹⁰⁷ When the single mutation of V265F was made in ACS, thus restoring the second CHS gatekeeper phenylalanine, ACS-like activity was impaired, but CHS-like activity did not improve.¹³⁸ Clearly, all three changes were important for the evolution of ACS from CHS, but since this ACS triple mutant still possesses marginal ACS-like activity, whereas wild-type CHS does not, additional differences must also be important for allowing the unusually wide N-methylanthraniloyl-CoA starter into the active site. Other ACS mutations made by these authors (“MS1” = R146K, M147L, N151R, M157F and I159M; and “MS2” = P203D, D204T, and A205H) did not significantly alter ACS specificity.¹³⁸

Some other second-tier residues not noted in this study may nonetheless be important for mechanistic differences between CHS and ACS. Interestingly, mutation of Phe215 (the other “gatekeeper” phenylalanine, absolutely conserved in both FAS KAS III and known type III PKS enzymes other than benzalacetone synthase, Section 3.2.9) to serine in alfalfa CHS unexpectedly results in the only known CHS mutant that can turn-over N-methylanthraniloyl-CoA (yielding a lactone product).¹⁰⁷ The ability of this non-physiological mutant to accept the bulkier N-methylanthraniloyl-CoA substrate is most likely due to the enlargement of the active site entrance (analogous to the F265V change in ACS). However, ACS does have an unusual substitution in the residue adjacent to position 215. Gly216 (Fig. 9), conserved in almost all other plant type III PKSs, is an alanine in ACS (Fig. 11). While the α-carbon of position 216 is buried and faces away from the active site cavity, insertion of a methyl group into this closely packed environment may subtly reposition the main-chain lining the active site surface. However, as mentioned in Sections 3.2.3 and 3.2.4, the as-yet untested second-tier difference most likely to affect ACS substrate specificity is the substitution of CHS Val98 for alanine, also observed in the other two type III PKS enzymes known to prefer meta- or ortho-substituted phenyl substrates.

3.2.6 Benzophenone synthase (BPS). Xanthones (Fig. 13) are tricyclic natural products found thus far predominantly in plants of the Gentianaceae and the Hypericaceae families.¹³⁹ In Hypericumandrosaemum cell cultures, xanthones accumulate both during cell growth and when induced by methyl jasmonate treatment.¹³⁹ While the physiological role of xanthones in plants has not yet been established, plant xanthones have been shown to possess antifungal properties.¹⁴⁰ Xanthones are formed by a regiospecific ring closure in benzophenone intermediates (Fig. 13), which are in turn synthesized by a type III PKS known as benzophenone synthase (BPS). BPS catalyzes three acetyl additions to a 3-hydroxybenzoyl-CoA starter molecule, followed by a CHS-like intramolecular Claisen cyclization and aromatization of the resultant tetraketide intermediate to form 2,3′,4,6-tetrahydroxybenzophenone (Fig. 13). A downstream enzyme, however, rather than BPS, catalyzes the additional, heterocyclic central ring closure to form the xanthone skeleton.^141a Although this downstream enzyme has not been identified, cell culture microsomal assays and inhibition studies indicate that ‘xanthone synthase’ is a cytochrome P450 oxidase, utilizing an oxidative phenol coupling mechanism.^141b In some cases, it appears that BPS may utilize an unsubstituted benzoyl-CoA starter to synthesize trihydroxybenzophenone, which is then hydroxylated at the 3′ position (by a benzophenone 3′-hydroxylase) prior to its conversion into xanthone.¹⁴² Study of these physiological pathways is further complicated in plants by the existence of multiple biosynthetic routes to benzoates. Benzoic acid can arise from intermediates of the shikimate pathway,¹⁴³ or from cinnamic acid from the general phenylpropanoid pathway (Fig. 7(a)), although the details of the phenylpropanoid pathway derivation remains obscure.¹³⁹

Examination of the only BPS sequence deposited in the database (from H.androsaemum) reveals a few unusual, yet subtle changes in the active site. Ser338, conserved in most plant type III PKS enzymes that catalyze Claisen or aldol cyclization reactions, is replaced in BPS with a glycine. CHS's second gatekeeper phenylalanine, Phe265, is conservatively mutated in BPS to a tyrosine. CHS's Gly256, known to reduce active site volume when mutated to a bulkier residue,^46,109 is a slightly larger alanine in BPS, and finally, position 216, adjacent to the other gatekeeper, Phe215, is occupied by a serine in BPS (Figs. 9 and 11). As discussed in Section 3.2.5 in the context of the ACS active site, introduction of a bulky side-chain on the buried 216 α-carbon may translate into a slight decrease in volume of the starter-binding pocket.

These minor active site changes alone seem unlikely to fully explain BPS's reported substrate specificity. Presumably these changes are complemented by additional mutations further from the active site cavity, yet sequence analysis does not suggest which changes might be functionally relevant. Unfortunately, while relative activities of BPS against various substituted and unsubstituted benzoyl-CoAs are published, the limited BPS literature fails to report a single assay using CHS's preferred phenylpropanoid-CoA starters. Conversely, at least one CHS enzyme has been shown in vitro to synthesize benzophenone when provided benzoyl-CoA as a starter.¹⁰⁵ These considerations raise the possibility that this BPS enzyme may in fact be a dual-purpose CHS, possessing relaxed starter specificity in order to accommodate two distinct physiological starter molecules, rather than being a fully committed BPS. While this scenario does not match the standard model for the functional divergence of type III PKS enzymes, where gene duplication provides redundancy prior to functional divergence (discussed in Section 3.1.1), the existence of separate cinnamoyl-, p-coumaroyl-, and benzoyl-CoA ligase enzymes in xanthone-producing plants¹³⁹ may provide the upstream regulation necessary to make a dual-purpose CHS enzyme physiologically feasible.

3.2.7 Phlorisovalerophenone synthase (VPS). Bitter acids such as humulone (Fig. 12) and lupulone are found in the cones of hop plants, which are used by beer brewers to impart a distinctive flavor to this ancient and much celebrated beverage. A divergent type III PKS, expressed in hop cone lupulin glands, was found to synthesize an intermediate leading to biosynthetic elaboration of these important natural products.^104,144,145 Phlorisovalerophenone synthase (VPS) accepts either isovaleryl- or isobutyryl-CoA in an otherwise CHS-like reaction sequence, consisting of three condensations with malonyl-CoA followed by an intramolecular Claisen cyclization to form phlorisovalerophenone or phlorisobutyrophenone, respectively (Fig. 10). Interestingly, while CHS enzymes have been shown to accept isovaleryl-CoA in vitro, producing a mixture of both phlorisovalerophenone and a triketide-derived lactone truncation product, hop VPS will not accept the CHS coumaroyl-CoA starter.¹⁰⁴

Comparison of the cloned hop VPS amino acid sequence with the CHS structure reveals a number of interesting differences near the starter-binding pocket. Two active site threonines at positions 132 and 197, conserved in CHS, are replaced with glycine and isoleucine in hops VPS. However, a second VPS has just been discovered in the primitive vascular plant Psilotum nudum (See Section 3.1.1). This enzyme uses serines at positions 132 and 197, and a valine at position 338 (Figs. 9 and 11). As discussed in the previous section, the residue at position 338 is a serine in CHSs, hops VPS, and all other plant type III enzymes known to catalyze the CHS-like intramolecular Claisen cyclization (other than BPS). As for the STS reaction (Section 3.2.2), there seems to be multiple active site configurations leading to the VPS reaction. Interestingly, the second-tier V98L mutation that our recent work has implicated in the conversion of CHS into STS is also seen in both VPS enzymes.

3.2.8 Coumaroyl triacetic acid synthase (CTAS). Various species of Hydrangea make unusual polyketide natural products. One such compound, hydrangic acid (3,4′-dihydroxystilbene-2-carboxylic acid), is a stilbenecarboxylic acid, apparently resulting from an STS-like C2-to-C7 intramolecular cyclization of the typical tetraketide intermediate, but with retention of the C1 carboxyl moiety on the new ring (Fig. 15).⁹⁶ Another Hydrangea product, hydramacroside B (Fig. 15), contains a linear tetraketide that could be derived from either the reduced phenylpropanoid dihydro-4-coumaroyl-CoA or from p-coumaroyl-CoA (if reduction of the starter moiety occurs after incorporation, as in the case of benzalacetone biosynthesis; see Section 3.2.9).⁹⁶ Researchers in Professor Ebizuka's laboratory cloned a type III PKS (HmS) from Hydrangea macrophylla var. thunbergii that was shown to produce an uncyclized p-coumaroyltriacetic acid (CTA) tetraketide that could be extracted by acidification of the in vitro assay mix to form CTA lactone (CTAL) (Figs. 10 and 15).⁹⁶ This study also confirmed the occurrence of CTA as a normal CHS truncation product (Fig. 8). These researchers hypothesized that CTAS may catalyze stilbenecarboxylic acid formation in vivo when complexed with a hypothetical cyclase.⁹⁶ However, both of these Hydrangea natural products appear to utilize tetraketide intermediates reduced at the C5 carbonyl; therefore, a ketoreductase (KR or CHR) is the most likely in vivo candidate for participation in a CTAS multienzyme complex. Alternatively, the reduction of the linear intermediate itself might be all that is needed to alter cyclization.

Fig. 15 Polyketide natural products found in Hydrangea (see Section 3.2.8).

Sequence analysis of CTAS reveals a substitution in the active site cavity that explains its unusual activity. An asparagine residue replaces the CHS-conserved Thr197 in CTAS (Figs. 9 and 11). We saw in Section 3.2.1 how an alfalfa CHS T197L point mutant displayed exactly the same mechanistic phenotype, producing CTAL as a major product when the reaction mixture was acidified.⁴⁶ As discussed in Section 2.2.2, Thr197 appears to be essential for achievement of the intramolecular Claisen cyclization in CHS. Another interesting feature of CTAS is a six-residue insertion (after CHS residue 290) relative to most plant type III PKS enzymes (Fig. 11). VPS and anther-specific CHS-like enzymes (Sections 3.2.7 and 3.2.11) have four- and two-residue insertions, respectively, in this solvent-exposed loop on the top of the monomer, remote from the active site. While unlikely to directly affect the mechanistic fate of enzyme-bound intermediates, this insertion is positioned to alter interactions between monomers through proximity to the adjacent and entwined N-terminal helices. In turn, such effects may influence dimerization and/or the formation of (putative) multi-enzyme complexes.

3.2.9 Benzalacetone synthase (BAS). The characteristic aroma of raspberries is conferred by the presence of p-hydroxyphenylbutan-2-one ( pHPB), also known as the raspberry ketone, whose two-step biosynthesis is initiated by an interesting type III PKS named benzalacetone synthase (BAS).¹⁴⁶ BAS catalyzes a single condensation of malonyl-CoA to the typical p-coumaroyl-CoA starter, but then decarboxylates the resulting diketide to form p-hydroxyphenylbut-3-ene-2-one ( p-hydroxybenzalacetone), resulting in the net addition of a single malonate-derived carbon (Fig. 10). Next, an NADPH-dependent benzalacetone reductase produces pHPB by eliminating the double bond on the linear portion of benzalacetone (Fig. 10).¹⁴⁶ BAS has been isolated from raspberries and from rhubarb, where its benzalacetone product is incorporated into the anti-inflammatory glucoside lindleyin (Fig. 12),¹⁴⁷ and may be present in turmeric and ginger as well.¹⁴⁸

The first purification of BAS from raspberries was complicated by extensive contamination by CHS, but enhanced activity in the presence of either β-mercaptoethanol or ethylene glycol (both inhibit CHS) and a higher affinity for the bulkier feruloyl-CoA than for p-coumaroyl-CoA (Fig. 14), convinced researchers that the BAS reaction was not merely a derailment of the normal CHS pathway.¹⁴⁶ The more recent cloning of BAS from rhubarb ( Rheum palmatum) revealed its amino acid sequence, and allowed better biochemical characterization after purification from CHS-free E. coli expression cultures.¹⁴⁷ Interestingly, the double bond that is eliminated by the downstream reductase is absent in the hydroxyphenylpropionyl-CoA starters used by the STS-like bibenzyl synthase (BBS) enzymes (see Section 3.2.3), but in vitro, BAS shows no activity with these reduced starters nor various aliphatic starters accepted in vitro by CHS.¹⁴⁷

The cloning of BAS followed the alfalfa CHS crystal structure, so Abe and coworkers were able to identify a curious substitution in a BAS active site residue. Phe215 (alfalfa CHS numbering), absolutely conserved in all other known type III PKS enzymes, is a leucine in BAS (Figs. 9 and 11). While normally this mutation would be considered conservative, this hydrophobic phenylalanine is proposed to assist in the malonyl decarboxylation and extension reactions, based upon its position in the active site, and conservation and mutational studies.^44,45 However, the presence of a second unorthodox BAS active site mutation not remarked upon in the literature makes the exact role of the F215L mutation unclear. CHS's Thr197, sterically altered in a number of divergent type III PKS enzymes, is exchanged for a reactive cysteine in BAS. While no other plant type III PKS uses a second active site cysteine, biosynthetic thiolases use such a cysteine to activate acetyl-CoA by the abstraction of a proton (Sections 1.3.2 and 1.5). Perhaps BAS utilizes its second active site cysteine to carry out its unusual decarboxylation of the diketide intermediate? This could involve cleavage of the diketide's thioester bond to CoA (likely to precede the decarboxylation reaction), or donation of a proton to the diketide to facilitate diketide decarboxylation. The nearby Leu215 residue may actively assist in this novel activity, but, conversely, the loss of the conserved gatekeeper phenylalanine might instead only minimize iterative CHS-like elongation of the diketide intermediate, by decreasing the enzyme's more typical decarboxylation activity toward malonyl-CoA, as observed in mutants of CHS Phe215.⁴⁵ Regardless of which mechanistic strategy BBS utilizes to execute its unusual reaction, it seems clear that chemical modifications to the type III PKS iterative machinery are involved, in contrast to the simple steric constrictions of the active site seen in both 2-PS (Section 3.2.1)⁴⁶ and CHS model studies.¹⁰⁹ Further study of this divergent mechanism is sure to produce valuable insight into the active site machinery of all type III PKS enzymes.

3.2.10 Synthesis of C-methylated chalcones in pine (PstrCHS2). C-Methylated chalcone natural products (Fig. 12) occur in Pinus strobus and a few other species, but flavonoid C-methyltransferase enzymes are unknown, and other methylated polyketides are known to derive their methyl groups by incorporating methylmalonyl-CoA.¹⁴⁹ A CHS-like PKS cloned from Pinus strobus, inactive with any of the normal phenylpropanoid or aliphatic type III PKS substrates, was found to turn-over a cinnamoyl diketide-N-acetylcysteamine (NAC) thioester substrate. (NAC is a simple, synthetically tractable pantetheine mimic used to activate unnatural starters or intermediate analogs for entry into PKS systems; see Fig. 3).¹⁴⁹ Further, this enzyme (PStrCHS2) was shown to utilize either methylmalonyl-CoA or malonyl-CoA to extend the diketide starter to a triketide (Fig. 10). Unexpectedly, a CHS from another (non-methylating) species was also shown to be capable of catalyzing a single extension of a diketide starter using methylmalonyl-CoA, but would complete the CHS reaction if given malonyl-CoA instead.¹⁴⁹ These authors proposed that homodimeric PStrCHS2 acts as a modular PKS in the in vivo biosynthesis of selectively C-methylated chalcone derivatives in P. strobus, in loose association with another homodimeric type III PKS to provide for substrate channeling. While the data are consistent with such an interpretation, neither the existence nor cooperative activity of such a complex has yet been demonstrated in vitro.

Specificity for both a diketide starter and a single turnover catalytic event, rather than use of methylmalonyl-CoA, are the in vitro characteristics that distinguish PStrCHS2 from typical type III PKS enzymes. These factors are independent of any hypothetical in vivo complex or channeling, and so must arise from some of the approximately 50 amino acid differences between PStrCHS2 and PStrCHS1, a closely related typical CHS cloned by the same authors.¹⁴⁹ One obvious active site difference in PstrCHS2 is the substitution of CHS Gly211 for a reactive Asp residue (Figs. 9 and 11). The iterative nature of BBS, the only other plant PKS enzymes with a residue of any significant size or polarity at this position (threonine, see Section 3.2.3), suggests that this change in PStrCHS2 is more likely to influence substrate selection than modularity. On the other hand, residue 211 is spatially positioned near the conserved Phe215, whose mutation to leucine in BAS is likely to prevent iterative activity by that enzyme (Section 3.2.9), so it is possible that this single change might be responsible for both of PstrCHS2's unusual in vitro characteristics. On the other hand, this mutation might be accompanied by reorientation of the side-chain at position 211 away from the active site cavity, in which case homology modeling of PstrCHS2 would be misleading.

3.2.11 Anther-specific CHS-like (ASCL) enzymes. A divergent sex-linked CHS-like enzyme, transiently expressed only in anthers during a specific phase of flower development, is conserved in a number of plant species and has a unique consensus sequence, forming yet another divergent type III PKS family. This anther-specific CHS-like (ASCL) enzyme family shares less than 40% amino acid identity with other plant type III PKS enzymes, but a number of conserved residues verify its membership in the type III PKS superfamily (reviewed in Ref. 150). The in vitro functional characterization of an anther specific CHS-like enzyme has not yet been reported, but it is one of many similarly expressed ORFs thought to participate in the poorly understood biosynthesis of exine (pollen cell wall).¹⁵⁰ Mutants lacking the genuine chs that is also expressed in anthers are deficient in flavonoid biosynthesis, further supporting a divergent metabolic role for the ASCL enzymes.^151,152

Comparison of the ASCL consensus sequence¹⁵⁰ to CHS reveals conservation of the catalytic triad and most of the other residues lining the active site, although the low overall sequence identity in surrounding areas increases the likelihood that second-tier interactions may subtly reshape the active site cavity (Fig. 11). However, CHS's conserved Thr197 is replaced by a glycine, a drastic difference similar to that observed in HEDS (see Section 3.2.4), where it is used for selection of larger (increasingly hydroxylated) phenylpropanoid starters. Overall, the ASCL active site thus seems most similar to HEDS, but a further substitution of Thr132 for a serine, as well as other unusual substitutions in the CHS/STS aldol cyclization specificity-determining area (see Section 3.2.2), preclude any exact homology-based prediction of ASCHSLE substrate or product specificity. Interestingly, the major component of exine is sporopollenin, an intractable polymer made up of both aliphatic and aromatic constituents, and exine is known to contain both p-coumaric acid and ferulic acid moieties.¹⁵³ The modeled ASCL active site appears capable of loading and elongating either starter. Thus it seems plausible, as Atanassov and co-workers suggest, that this CHS-like enzyme may synthesize a monomer for exine biosynthesis.¹⁵⁰ However, active site similarity to HEDS suggests that the substrates and intermediates (if not the product) of the anther-specific CHS-like enzymes may not be as novel as previously anticipated.

3.3 Plant natural product pathways likely to utilize unknown type III PKSs

Several other plant natural products are very likely to be synthesized by a CHS-like enzyme, but their in vitro biosynthesis has yet to be demonstrated. There are two possible reasons for this. The simplest explanation is that the responsible enzyme has yet to be isolated and characterized. Alternatively, the enzyme may be known, but might behave differently in vitro, thus obscuring its true biological function. In a few natural product cases, the latter option seems more likely, which leads us to consider the potential causes of such a discrepancy. The in vitro absence of some essential in vivo protein partner that modifies either the enzyme or a reaction intermediate seems the most plausible explanation. Thus far, the reduction of CHS intermediate products by chalcone reductase (CHR) is the only known example of another protein modulating a CHS-like reaction (aside from variability in the availability of starters). Although a long CoA-binding tunnel limits access to the buried type III PKS active site, the cavity's cyclization pocket is separated from the outside solution (orthogonal to the dimer interface) by a single, looped β-strand finger (Fig. 1). It is easy to imagine how formation of some multi-enzyme complex could cause a conformational change along this thin wall, thus reshaping the active site cavity. Interestingly, most of the following natural products appear to utilize a ketoreduction step during their biosynthesis.

3.3.1 Biosynthesis of various stilbenecarboxylic acids. While some stilbenecarboxylic acids are very similar to known type III PKS products (as mentioned in the context of STS and CTAS, Sections 3.2.2 and 3.2.8 respectively), others incorporate unusual medium- or long-chain fatty acids as starters. None of these stilbenecarboxylic acid biosynthetic pathways have been demonstrated in vitro. Stilbenecarboxylic acids found in hydrangea, liverworts, cannabis, and poison ivy will be discussed in turn.

In hydrangea, the stilbenecarboxylic product hydrangic acid (Fig. 15) appears to be derived from an STS-like cyclization of a reduced (at the C5 carbonyl group) tetraketide intermediate, but with retention (unlike STS) of the terminal C1 carboxyl moiety as a substituent of the new ring.⁹⁶ CTAS might catalyze the formation of hydrangic acid in the presence of the correct PKR to suitably modify the substrate, but sequence analysis of this gene viewed in light of our recent unpublished studies of STSs suggest that CTAS is more CHS-like than STS-like.

Besides dihydrostilbene natural products (bibenzyls, see Section 3.2.3) leading to three-ringed phenanthrenes (Fig. 13) and dimerized bis-bibenzyls, liverworts also make stilbenecarboxylic acids such as lunularic acid (Fig. 16).¹⁵⁴ Like hydrangic acid, lunularic acid appears to be synthesized from a C5 reduced tetraketide intermediate, but it also appears to derive from a reduced dihydro- p-coumaroyl starter, like the dihydrostilbenes. Prelunularic acid (Fig. 16), a cyclized precursor retaining the reduced C5 hydroxyl on its as yet non-aromatized new ring, has been isolated from these primitive plants,¹⁵⁴ but no liverwort type III PKS enzymes or putative ORFs have yet been reported in the sequence database or the literature. Thus, it is difficult to speculate on the mechanistic details of liverwort polyketide biosynthesis.

Fig. 16 Additional known stilbenecarboxylic acids discussed in Section 3.3.1 (see also Fig. 15) from: (a) liverworts, (b) marijuana, and (c) poison ivy and related species.

The psychotropic tetrahydrocannabinol (THC) natural products in Cannabis sativa (Fig. 16) are synthesized via the condensation of a stilbenecarboxylic acid (olivetolic acid) to an isoprenoid (geranyl pyrophosphate), followed by rearrangement and cyclization of the isoprenoid-derived portion (reviewed in Ref. 155). Feeding experiments in Cannabis with C13-labeled glucose implicate a polyketide synthase in the biosynthesis of the olivetolic acid precursor,¹⁵⁶ although the participation of a CHS-like enzyme has not yet been demonstrated. Nonetheless, a type III PKS mechanism involving three condensations of malonyl-CoA to a hexanoyl-CoA starter, followed by an STS-like cyclization (C2→C7) and aromatization, but with retention of the C1 carboxyl group as a ring substituent, would produce olivetolic acid (Fig. 16). Unlike the other stilbenecarboxylic acids discussed, olivetolic acid appears to be synthesized via an unreduced tetraketide intermediate, and also appears to aromatize spontaneously upon cyclization (no non-aromatic cyclized intermediates are known). Some THC-containing natural products lack the carboxylic acid derived from olivetolic acid, but in these cases decarboxylation seems to occur after olivetolic acid is joined to the isoprenoid moiety, as opposed to an STS-like loss of CO₂ concurrent with cyclization.¹⁵⁵

Various plants, predominantly from the Anicardiaceae family, synthesize natural products that cause contact dermatitis in pests (including humans).¹⁵⁷ Anacardic acids are stilbenecarboxylic acids resulting from cyclization and aromatization of three acetate additions (with one ketoreduction step) to a monounsaturated 16- or 18-carbon acyl-CoA starter (Fig. 16).¹⁵⁸ The sap of lacquer trees,¹⁵⁹ as well as the shells of cashew nuts,¹⁶⁰ contain these allergenic compounds. Urushiols, responsible for the allergenic effects of poison ivy, poison oak, and related Toxicodendron species, apparently result from the decarboxylation and hydroxylation of anacardic acids.¹⁵⁸ However, none of the responsible biosynthetic enzymes have been identified.

Anacardic acids are also found outside of the Anicardiaceae family, in geraniums. Geranium resistance to spider mites and aphids results from production of monounsaturated anacardic acids.¹⁶¹ Δ⁹-desaturation of a saturated 14-carbon fatty acid precursor leads to the unusual 16:1Δ¹¹ and 18:1Δ¹³ fatty acids that pest-resistant geraniums incorporate into anacardic acids.¹⁶²

3.3.2 Sorgoleone, an allelopathic sorghum phytotoxin. To preclude competition for resources, the roots of Sorghum bicolor exude sorgoleone (Fig. 17) (reviewed in Ref. 163), a potent phytotoxin that inhibits the growth of nearby weeds by competitive inhibition of photosynthetic electron transport.¹⁶⁴ The quinone head of sorgoleone presumably results from the action of a type III PKS utilizing an unsaturated long-chain fatty acid starter, but in vivo labeling with ¹³C-labeled acetate produced inconclusive results.¹⁶⁵ Further study is needed to elucidate the enzymes and details of sorgoleone biosynthesis.

Fig. 17 An allelopathic phytoalexin from Sorghum bicolor (see Section 3.3.2).

4 Bacterial type III PKS enzymes

4.1 Introduction to bacterial type III PKSs

Unlike plant type III PKS enzymes, which usually share 60–95% amino acid identity, bacterial CHS-like sequences typically share only 25–50% identity with each other and with plant type III PKS enzymes. Functional characterization of these extremely divergent bacterial enzymes has begun only recently, but some interesting and complicated reactions have already been discovered, and are the subject of a recent review.¹⁶⁶ While no bacterial type III PKS crystal structures are yet available, homology modeling predicts much more active site diversity than that seen in plant CHS-like enzymes, suggesting we still have much to learn about type III polyketide synthases.

4.2 Bacterial type III enzymes of known function

4.2.1 1,3,6,8-Tetrahydroxynaphthalene synthase (THNS). The cell walls of fungi contain the colored structural and UV-protectant biopolymer melanin, which is essential for the virulence against both plants and animals shown by many pathogenic fungal species.¹⁶⁷ Instead of the tyrosine-derived DOPA-based melanin produced by animals, many fungi synthesize melanin from 1,8-dihydroxynaphthalene (DHN) monomers.¹⁶⁸ DHN is derived by enzymatic reduction of 1,3,6,8-tetrahydroxynaphthalene (THN), which is itself formed in fungi from the condensation and cyclization of five acetate units by a large (approximately 2150 amino acids), ACP-utilizing multi-domain iterative type I PKS.^169,170

RppA, a type III PKS cloned from Streptomyces griseus by Horinouchi's group in 1995,¹⁷¹ and the first bacterial CHS-like protein to be functionally characterized, is also a THN synthase. This ∼85 kD homodimer was found to catalyze the formation of THN from five molecules of malonyl-CoA (Fig. 18).¹⁷²S. griseus knockouts of rppa are albino, and overexpression of RppA in E. coli confers pigment production ( via auto-oxidation of THN to form flaviolin) to this normally colorless heterologous host.¹⁷² THN is also thought to be incorporated into a family of prenylated naphthoquinone natural products found in various Streptomyces species.¹⁷² Consistent with this hypothesis, one of three chs-like sequences in S. coelicolor is highly homologous to S. griseus rppa, and also encodes a THN synthase.¹⁷³

Fig. 18 Known bacterial type III PKS enzyme reactions (see Sections 4.2.1 to 4.2.3). Type III PKS enzymes (THNS, PhlD, and DpgA) are shown in boldface. (a) Reaction of THNS leading to 1,3,6,8-tetrahydroxynaphthalene. Derivation of both flaviolin and 1,8-dihydroxynaphthalene natural products from 1,3,6,8-THN are also shown (see text). (b) PhlD's apparent role in the biosynthesis of monoacetylphloroglucinol (MAPG) and diacetylphloroglucinol (DAPG). (c) Reaction of DpgA leading to dihydroxyphenylacetate (DHPA). Additional enzymes further modify this product to form the unusual amino acid dihydroxyphenylglycine (DHPG), which is then incorporated into vancomycin-related antibiotics (see Fig. 19).

Two unusual features of the THNS reaction stand out. Unlike other CHS-like enzymes, THNS forms two new six-carbon rings, and does so using different intramolecular cyclization mechanisms. Although the intermediates of cyclization are unknown, comparison of the symmetric THN product to a linear polyketide implicates an STS-like aldol condensation followed by a CHS-like Claisen condensation. The second unusual aspect of the THNS reaction is its condensation of five acetate units. No other known type III PKS enzymes catalyzes more than three extension steps, presumably due to active site volume or the reactive nature of the resulting tetraketide intermediate. THNS may form the first of two rings prior to making a pentaketide, thus eliminating the potential for a number of unproductive side reactions. However, when provided with the non-physiological octanoyl-CoA starter, the THNS from S. griseus (RppA) makes a hexaketide, with lactone formation due to cyclization of only the final three acetate units.¹⁷⁴

A homology model of S. coelicolor THNS based on the alfalfa CHS crystal structure reveals a number of unusual changes in the active site. Conservation of the catalytic triad, both gatekeeper phenylalanines, and the absolutely conserved Pro375 speak to the reliability of the structural alignment, despite the low sequence similarity between the two enzymes (Fig. 11). In contrast to plant PKSs, which have a single active site cysteine (or two in the case of BAS, see Section 3.2.9), this model reveals four additional cysteines in the THNS active site. One of these, at CHS position 211, is a serine in the S. griseus THNS (RppA). The other three cysteines, at CHS positions 132, 194, and 197, are conserved among Streptomyces THNS enzymes, and occur at positions known to be important for modulating both substrate specificity and proper cyclization of polyketide intermediates (see Section 2.2.2) in plant type III PKS enzymes. It is a matter of speculation whether the potential reactivity of the cysteine rich active site in THNS plays a role in either promoting the proper folding of intermediate polyketides or in preventing formation of improper polyketide conformations that would otherwise lead to truncation products. A number of additional non-conservative substitutions, both in and out of the active site, further complicate our understanding of the THNS reaction.

Sequence gazing yields another interesting observation, concerning an arginine residue (at CHS position 308) on the surface of both known Streptomyces THNS sequences. The equivalent residue in the ancestral KAS III enzymes is also an absolutely conserved arginine, whose side-chain has been shown to facilitate KAS III ACP binding.¹⁷⁵ Likewise, the ACP-utilizing KS domains of both type I DEBS and type II actinorhodin PKS enzymes also possess this arginine residue. Conversely, CHS and all plant CHS-like enzymes possess an absolutely conserved alanine at this position, while all other known bacterial CHS-like sequences have a lysine here. It is tempting to speculate that some bacterial type III PKS enzymes, especially THNS, may have an as yet undiscovered specificity for ACP-thioester substrates. To the best of our knowledge, this possibility has not previously been considered for any CHS-like bacterial enzymes, since CoA specificity is a defining feature of the well-characterized plant enzymes. Two features of ACP-utilizing condensing enzymes that might obscure an unsuspected ACP specificity include the tendency of condensing enzymes to discriminate between various ACP isozymes within a species (see Section 1.5), as well as in vitro utilization of non-physiological CoA substrates.

4.2.2 Biosynthesis of 2,4-diacetylphloroglucinol (DAPG) in Pseudomonas. Some agricultural soils occasionally referred to as suppressive soils lower the incidence and severity of pathogen-induced disease in plants grown in them. In at least two cases (take-all disease in wheat and black root rot in tobacco), this resistance has been linked to the presence of certain Pseudomonas species producing 2,4-diacetylphloroglucinol (DAPG).^176,177 Furthermore, the presence of these antifungal-producing soil bacteria on the rhizoplane (root surface) appears to be selected for by maize,¹⁷⁸ suggesting the widespread agricultural importance of these plant-associated biological control agents (BCAs).¹⁷⁹

Bangera and Thomashow isolated and characterized four clustered genes ( phlACBD) in Pseudomonas fluorescens responsible for the biosynthesis of DAPG (Fig. 18).¹⁸⁰ Trios of predicted gene products from Pyrococcus furiosus and Methanococcus jannaschii show homology to the predicted gene products of PhlA, PhlC, and PhlB, but the CHS-like nature of the PhlD enzyme was unexpected. PhlA is homologous to E. coli FabH (KAS III), but like the KS-β domains of type II PKSs (Section 1.4) it lacks the active site cysteine. PhlC somewhat resembles thiolase, and PhlB shows no significant homology with any known protein. 2,4-Diacetylphloroglucinol is produced by transfer of an acetyl group to a monoacetylphloroglucinol (MAPG) intermediate, which could result from a CHS-like C6-to-C1 intramolecular cyclization of a tetraketide intermediate. Phenotypes of selective in vivo knockouts suggest that PhlA, PhlB, and PhlC function collectively in the acetylation of MAPG to form DAPG, and also demonstrate that all four gene products are necessary for the production of the MAPG intermediate (Fig. 18).¹⁸⁰ This latter result is surprising, as the synthesis and cyclization of a tetraketide from four malonyl-CoA molecules seems within the scope of an independent type III PKS homodimer. Bangera and Thomashow speculate that PhlA–C provide PhlD with an activated starter such as acetoacetyl-CoA. Another possibility is that PhlD does not function as an independent homodimer, but rather in some complex with PhlA–C. The in vitro characterization of these enzymes will undoubtedly answer these questions and provide further insights.

Like the Streptomyces THN synthases, sequence analysis of P. fluorescens PhlD suggests it has the same overall fold as the plant type III PKS enzymes, as well as conservation of the Cys-His-Asn catalytic trio and both gatekeeper phenylalanines in an otherwise quite divergent active site. Also like THNS, CHS's Ser338 is replaced by alanine, and additional cysteines are found in the active site, aligning with CHS threonines 132 and 197 (Fig. 11). Other putative PhlD active site residues differ from both THNS and CHS, presumably providing a unique steric shape to the PhlD active site cavity. On the basis of these differences, a Streptomyces avermitilis sequence in the database (accession BAB69299.1) that is annotated as a PhlD homologue appears instead to be another THNS.

4.2.3 (S)-3,5-Dihydroxyphenylglycine (DHPG) used in antibiotic biosynthesis. The third functionally characterized bacterial type III PKS is involved in the biosynthesis of an acetate-derived¹⁸¹ amino acid, (S)-3,5-dihydroxyphenylglycine (DHPG or DPG) (Fig. 18). DHPG is one of two unusual amino acids synthesized by various actinomycetes for incorporation into specific heptapeptides (by a non-ribosomal peptide synthetase), which are then cross-linked to form teicoplanin and vancomycin group glycopeptide antibiotics (Fig. 19).^182,183 These natural antibiotics are used medicinally to treat Gram-positive bacterial infections, but the unfortunate spread of vancomycin-resistant strains of bacteria has generated interest in harnessing the biosynthetic machinery of actinomycetes to produce novel antibiotics. This initiative led to the discovery of a CHS-like putative protein (ORF27) in the large gene cluster responsible for the biosynthesis of chloroeremomycin, a vancomycin-like antibiotic, in Amycolatopsis orientalis.¹⁸⁴ Subsequently, Li et al. showed that recombinant protein produced from ORF27 was responsible for the synthesis of 3,5-dihydroxyphenylacetate (DHPA), a precursor on the biosynthetic pathway to DHPG.¹⁸⁵

Fig. 19 Structure of vancomycin. Red denotes the location, in antibiotics of the vancomycin class, of atoms derived from the type III PKS natural product DHPA (see also Fig. 18).

Subsequently, an analogous type III PKS from Amycolatopsis mediterranei, which produces the similar antibiotic balhimycin, was shown to be one of five proteins (DpgA-D and Pgat) involved in the biosynthesis of intermediates leading to DHPG.¹⁸⁶ A knockout of the CHS-like dpgA abrogated balhimycin production in this organism, but it could be rescued by supplementation with dihydroxyphenylacetate (DHPA), a presumed intermediate in DHPG biosynthesis. Further, heterologous expression of DpgA in Streptomyces lividans resulted in the production of DHPA, whereas expression of DpgA–D in this host enhanced DHPA synthesis and further resulted in the production of another DHPG intermediate, 3,5-hydroxyphenylglyoxylate. Following heterologous expression in Streptomyces lividans, assays of extracts using either radiolabeled acetyl- or malonyl-CoA implicated malonyl-CoA as the only efficiently incorporated starter molecule for this reaction. Transamination of this latter intermediate by phenylglycine aminotransferase (Pgat), which is also involved in the biosynthesis of the other unusual amino acid incorporated into these antibiotics, completes the synthesis of DHPG.^186,187

Another group was able to purify soluble protein from the expression in E. coli of the analogous DpgA-D enzymes from A. orientalis, although the multimeric state of each purified enzyme and the tendency of mixtures to form complexes were unfortunately omitted from their report.¹⁸³ This study confirmed the ability of DpgA to synthesize DHPA from four molecules of malonyl-CoA, but found the reaction rate to be very slow with DpgA alone. Although neither DpgB nor DpgD show any activity with malonyl-CoA by themselves, DpgA's rate of DHPA production is increased 17-fold by the inclusion of DpgB in reaction mixtures, and the addition of DpgD to this mixture results in another two-fold increase in rate.¹⁸³ The exact role of DpgB and DpgD is not clear, but these crotonase superfamily enzymes appear to utilize their hydratase/dehydratase activities to facilitate the aromatization of the cyclized DHPA skeleton. However, this study firmly establishes DpgC's role in DHPG biosynthesis. DpgC oxidizes the methylene α-carbon of DHPA-CoA to a carbonyl, and also acts as a thioesterase, releasing the 3,5-dihydrophenylglyoxylate product as a free acid.¹⁸³

Uncertainty regarding the mechanistic details of the PKS's cyclization of the polyketide intermediate complicates analysis of the exact role of DpgB and DpgD. A C8-to-C3 intramolecular aldol cyclization of the linear intermediate has been presumed, as it produces the proper skeleton while maintaining the thioester at C1,¹⁸⁶ and use of CoA thioester substrate analogues by all three downstream Dpg enzymes supports this interpretation.¹⁸³ C8 of the proposed tetraketide derives from the first malonyl-CoA to bind the enzyme. If DpgA decarboxylates this starter before loading it, the resulting tetraketide's C8 is a methyl group, whose suitability as a nucleophile would be increased by a thiolase-like abstraction of a proton. As in other bacterial type III PKS enzymes, sequence alignments predict the presence of an additional cysteine in the DpgA active site (this time only at CHS position 194), making a thiolase-like methyl activation a distinct possibility. On the other hand, if the malonyl starter is loaded intact, the resulting linear polyketide possesses a free carboxylate group that can activate C8 as a nucleophile through a coupled decarboxylation mechanism. Precedence for this interpretation comes from mechanistic studies of plant type III PKS enzymes that show impaired decarboxylation of malonyl-CoA when no starter is bound to Cys 164.⁹⁴ Presumably, either of these mechanisms could outcompete the typical CHS-like C6-to-C1 Claisen cyclization, favored by the labile thioester bond at C1.

A final consideration is DpgA's apparent difficulty in aromatizing its product's new ring.¹⁸³ Interestingly, the linear portion of both the proposed cyclic intermediate and aromatized DHPA product lacks the π-bond conjugation with the new ring seen in most type III PKS products. Although sequence alignments reveal DpgA to have other unusual active site features that might explain its lack of aromatase activity, including an alanine at CHS positions 197 (Fig. 11), it is possible that this impaired aromatization instead reflects the unique nature of the C8-to-C3 cyclization product.

4.2.4 Bacterial type III genes of unknown function. Besides these three functionally characterized bacterial CHS-like enzymes, the genomes of a wide assortment of bacteria have revealed putative ORFs that clearly group with the bacterial type III PKS family (Fig. 11) (most recently reviewed in Ref. 166). Conservation of important catalytic and structural residues indicates that these ORFs represent functional type III PKS enzymes possessing the typical αβαβα fold. Nonetheless, significant variation at amino acid positions known to influence substrate and product specificity suggests that these CHS-like enzymes are functionally divergent from CHS and the three known bacterial type III PKSs. The degree of active site variation also indicates that these uncharacterized enzymes (unlike THNSs from various species of Streptomyces) are not isoenzymes, but probably catalyze a number of different biosynthetic reactions. Notably, at least two shared features distinguish these ORFs from the three known bacterial enzymes. None of them maintain CHS's second gatekeeper phenylalanine at position 265, nor appear to utilize multiple cysteines in the active site. Not enough is currently known about bacterial type III PKS active site structure and function to predict activity at this time, but the functional characterization of these sequences is likely to expand our understanding of type III PKS diversity.

5 Conclusions and future directions

Although comparison of CHS-like enzymes to their KAS III ancestors is informative, the details of type III PKS emergence remain obscure. And while much is known about CHS, some mechanistic questions remain unanswered (see Section 2.2.3). The last few years have revealed a number of new type III PKS superfamily members, and future efforts will undoubtedly discover more. Homology models suggest that much of the functional diversity of plant CHS-like enzymes has evolved by conservation of the decarboxylation and condensing machinery accompanied by simple steric modulation of other active site residues, as observed for 2-PS (Section 3.2.1). Bacterial CHS-like enzymes are much more divergent than their plant counterparts, however, and, like STS (Section 3.2.2), are likely to have accumulated some main-chain movements relative to CHS, thus complicating homology-based analyses of these latter gene products. The comparative analysis of type III PKS sequences given here, with the differences most likely to serve important mechanistic roles highlighted, should assist researchers to better annotate newly discovered CHS-like sequences. The mutagenic interconversion and in vitro characterization of type III PKS enzymes will continue to be a valuable tool for confirming functionally relevant differences within this superfamily, with crystallography providing answers where homology modeling fails to predict the structural and mechanistic consequences of crucial amino acid substitutions.

As our understanding of the biosynthesis of biologically and medicinally important natural products grows, the benefits of harnessing this awesome machinery become apparent. In Section 3.2.2, we saw how antifungal resveratrol biosynthesis was introduced into plants lacking this useful pathway. Altered flower color, nitrogen fixation, increased digestability of forage crops, and increased production of health-promoting natural products in food crops¹⁸⁸ are other goals of past and present flavonoid pathway engineering efforts (reviewed in Refs. 57,189). The combination of flavonoid pathway enzymes from different plants,¹⁹⁰ as well as the combinatorial application of enzymes for hydroxylation, methylation, prenylation, glycosylation and acylation,¹⁸⁸ promises to yield even greater chemical diversity. While these experiments, as well as extensive domain-swapping in various type I and II PKSs,¹⁹ demonstrate the utility of mixing and matching naturally evolved enzymes to achieve new pathways, our increasing understanding of individual enzymes will allow a more sophisticated approach. The ability to engineer substrate and product specificities (natural or unnatural) into native enzymes or domains will increasingly remove our dependence upon nature to provide desirable activities.

CHS-like enzymes are beginning to be appreciated for their natural functional diversity and architectural simplicity: two qualities that facilitate the elucidation of the determinants of that diversity, as examined in this review. The increasing ability to modulate specificity in type III PKS enzymes^{46,107,109,138} complements parallel efforts to exploit other aspects of flavonoid biosynthesis. Beyond controlling substrate and product specificity, a number of other type III PKS engineering projects seem worthwhile. Besides the use of substituted malonyl moieties to diversify acetate units, Schröder's characterization of the divergent methylmalonyl-CoA utilizing enzyme in Section 3.2.10 suggests the possibility of engineering modularity into these normally iterative enzymes. Using separate active sites for individual steps should allow increased diversity and optimization of each catalytic step. CHS's recruitment of the CHR ketoreductase implies that other enzymes (hydratases for example) could be selectively employed to further modify diffusable intermediates. Synthetic NAC thioesters (introduced in Section 3.2.10) have proven useful for introducing unnatural starters and intermediate analogs into various PKS systems.¹⁴⁹ CHS-like enzymes might be redesigned to optimize activities with these synthetic thioesters. Another approach is to re-engineer the upstream CoA ligase (Fig. 7) to catalyze the activation of unnatural starters. Alternatively, an ACP-utilizing type III PKS might allow the coupling of type III activities to the reactions of their larger and more-complicated cousins (PKS types I and II). Aside from the potential utility of these hypothetical divergent activities, experimentation toward these engineering objectives will undoubtedly shed even more light on type III PKS enzymes and mechanisms. However, these engineering flights of fancy should not distract us from appreciating the rich and beautiful tapestry of evolutionary and ecological interactions illuminated by our examination of the natural chemical diversity generated by the plant and bacterial type III PKS superfamily.

References

1. B. J. Rawlings, Nat. Prod. Rep., 1998, 15, 275.
2. C. C. DiRusso, P. N. Black and J. D. Weimar, Prog. Lipid Res., 1999, 38, 129.
3. C. O. Rock and J. E. Cronan, Biochim. Biophys. Acta, 1996, 1302, 1.
4. R. J. Heath and C. O. Rock, J. Biol. Chem., 1995, 270, 15531.
5. V. S. Rangan, A. K. Joshi and S. Smith, Biochemistry, 2001, 40, 10792.
6. P. G. Roughan and J. B. Ohlrogge, Plant Physiol., 1996, 110, 1239.
7. K. H. Choi, L. Kremer, G. S. Besra and C. O. Rock, J. Biol. Chem., 2000, 275, 28201.
8. M. Mathieu, Y. Modis, J. P. Zeelen, C. K. Engel, R. A. Abagyan, A. Ahlberg, B. Rasmussen, V. S. Lamzin, W. H. Kunau and R. K. Wierenga, J. Mol. Biol., 1997, 273, 714.
9. M. Mathieu, J. P. Zeelen, R. A. Pauptit, R. Erdmann, W. H. Kunau and R. K. Wierenga, Structure, 1994, 2, 797.
10. C. T. Walsh, Enzymatic Reaction Mechanisms, ed. A. C. Bartlett and L. W. McCombs, W. H. Freeman and Company, San Francisco, 1979.
11. Y. Modis and R. K. Wierenga, J. Mol. Biol., 2000, 297, 1171.
12. S. Masamune, C. T. Walsh, A. J. Sinskey and O. P. Peoples, Pure Appl. Chem., 1989, 61, 303.
13. Y. Modis and R. K. Wierenga, Struct. Fold Des., 1999, 7, 1279.
14. J. Staunton and K. J. Weissman, Nat. Prod. Rep., 2001, 18, 380.
15. B. J. Rawlings, Nat. Prod. Rep., 1999, 16, 425.
16. B. J. Rawlings, Nat. Prod. Rep., 2001, 18, 231.
17. B. J. Rawlings, Nat. Prod. Rep., 2001, 18, 190.
18. B. Shen, in Biosynthesis of Aromatic Polyketides, ed. F. J. Leeper and J. C. Vederas, Springer-Verlag, Berlin, New York, 2000 (Top. Curr. Chem., 209, 1–51).
19. C. Khosla, R. S. Gokhale, J. R. Jacobsen and D. E. Cane, Annu. Rev. Biochem., 1999, 68, 219.
20. R. Bentley and J. W. Bennett, Annu. Rev. Microbiol., 1999, 53, 411.
21. D. A. Hopwood and D. H. Sherman, Annu. Rev. Genet., 1990, 24, 37.
22. T. A. Holak, M. Nilges, J. H. Prestegard, A. M. Gronenborn and G. M. Clore, Eur. J. Biochem., 1988, 175, 9.
23. T. A. Holak, M. Nilges and H. Oschkinat, FEBS Lett., 1989, 242, 218.
24. G. Y. Xu, A. Tam, L. Lin, J. Hixon, C. C. Fritz and R. Powers, Structure (Camb), 2001, 9, 277.
25. M. P. Crump, J. Crosby, C. E. Dempsey, J. A. Parkinson, M. Murray, D. A. Hopwood and T. J. Simpson, Biochemistry, 1997, 36, 6000.
26. S. C. Tsai, L. J. Miercke, J. Krucinski, R. Gokhale, J. C. Chen, P. G. Foster, D. E. Cane, C. Khosla and R. M. Stroud, Proc. Natl. Acad. Sci. USA, 2001, 98, 14808.
27. X. Qiu, C. A. Janson, W. W. Smith, M. Head, J. Lonsdale and A. K. Konstantinidis, J. Mol. Biol., 2001, 307, 341.
28. R. Welle and H. Grisebach, FEBS Lett., 1988, 236, 221.
29. G. Forkmann, Genetics of Flavonoids, in The Flavonoids, ed. J. B. Harborne, Chapman & Hall, London, 1994, pp. 537–564.
30. R. Welle, G. Schröder, E. Schiltz, H. Grisebach and J. Schröder, Eur. J. Biochem., 1991, 196, 423.
31. M. Siggaard-Andersen, Prot. Seq. Data Anal., 1993, 5, 325.
32. W. Huang, J. Jia, P. Edwards, K. Dehesh, G. Schneider and Y. Lindqvist, Embo J., 1998, 17, 1183.
33. J. G. Olsen, A. Kadziola, P. von Wettstein-Knowles, M. Siggaard-Andersen, Y. Lindquist and S. Larsen, FEBS Lett., 1999, 460, 46.
34. J. G. Olsen, A. Kadziola, P. von Wettstein-Knowles, M. Siggaard-Andersen and S. Larsen, Structure (Camb), 2001, 9, 233.
35. A. C. Price, K. H. Choi, R. J. Heath, Z. Li, S. W. White and C. O. Rock, J. Biol. Chem., 2001, 276, 6551.
36. M. Moche, G. Schneider, P. Edwards, K. Dehesh and Y. Lindqvist, J. Biol. Chem., 1999, 274, 6031.
37. X. Qiu, C. A. Janson, A. K. Konstantinidis, S. Nwagwu, C. Silverman, W. W. Smith, S. Khandekar, J. Lonsdale and S. S. Abdel-Meguid, J. Biol. Chem., 1999, 274, 36465.
38. C. Davies, R. J. Heath, S. W. White and C. O. Rock, Struct. Fold Des., 2000, 8, 185.
39. M. Moche, K. Dehesh, P. Edwards and Y. Lindqvist, J. Mol. Biol., 2001, 305, 491.
40. J. N. Scarsdale, G. Kazanina, X. He, K. A. Reynolds and H. T. Wright, J. Biol. Chem., 2001, 276, 20516.
41. K. A. McGuire, M. Siggaard-Andersen, M. G. Bangera, J. G. Olsen and P. von Wettstein-Knowles, Biochemistry, 2001, 40, 9836.
42. D. Val, G. Banu, K. Seshadri, Y. Lindqvist and K. Dehesh, Struct. Fold Des., 2000, 8, 565.
43. A. Abbadi, M. Brummel, B. S. Schutt, M. B. Slabaugh, R. Schuch and F. Spener, Biochem. J., 2000, 345(Pt 1), 153.
44. J. L. Ferrer, J. M. Jez, M. E. Bowman, R. A. Dixon and J. P. Noel, Nat. Struct. Biol., 1999, 6, 775.
45. J. M. Jez, J. L. Ferrer, M. E. Bowman, R. A. Dixon and J. P. Noel, Biochemistry, 2000, 39, 890.
46. J. M. Jez, M. B. Austin, J. Ferrer, M. E. Bowman, J. Schröder and J. P. Noel, Chem. Biol., 2000, 7, 919.
47. H. Grisebach, in Proceedings of the IVth International Congress of Biochemistry, Wien 1958, Vol. II, pp. 56–70, Pergamon Press, London, 1959.
48. F. Kreuzaler and K. Hahlbrock, FEBS Lett., 1972, 28, 69.
49. F. Kreuzaler and K. Hahlbrock, Arch. Biochem. Biophys., 1975, 169, 84.
50. W. Heller and K. Hahlbrock, Arch. Biochem. Biophys., 1980, 200, 617.
51. U. Reimold, M. Kroeger, F. Dreuzaler and K. Hahlbrock, Embo J., 1983, 2, 1801.
52. R. E. Koes, C. E. Spelt, P. J. van den Elzen and J. N. Mol, Gene, 1989, 81, 245.
53. R. Wingender, H. Rohrig, C. Horicke, D. Wing and J. Schell, Mol. Gen. Genet., 1989, 218, 315.
54. C. J. Lamb, M. A. Lawton, M. Dron and R. A. Dixon, Cell, 1989, 56, 215.
55. B. Weisshaar and G. I. Jenkins, Curr. Opin. Plant Biol., 1998, 1, 251.
56. J. B. Harborne and C. A. Williams, Phytochemistry, 2000, 55, 481.
57. B. Winkel-Shirley, Plant Physiol., 2001, 126, 485.
58. R. A. Dixon and N. L. Paiva, Plant Cell, 1995, 7, 1085.
59. T. A. Holton and E. C. Cornish, Plant Cell, 1995, 7, 1071.
60. J. B. Harborne and C. A. Williams, Nat. Prod. Rep., 2001, 18, 310.
61. S. R. Long, Cell, 1989, 56, 203.
62. X. Perret, C. Staehelin and W. J. Broughton, Microbiol. Mol. Biol. Rev., 2000, 64, 180.
63. H. K. Stafford, Plant Physiol., 1991, 96, 680.
64. J. B. Harborne and H. Baxter, Handbook of Natural Flavonoids, Wiley-VCH, Weinheim, 1999.
65. J. M. Jez, M. E. Bowman, R. A. Dixon and J. P. Noel, Nat. Struct. Biol., 2000, 7, 786.
66. H. K. Dooner, T. P. Robbins and R. A. Jorgensen, Annu. Rev. Genet., 1991, 25, 173.
67. D. P. Ormrod, L. G. Landry and P. L. Conklin, Physiologia Plantarum, 1995, 93, 602.
68. N. R. Guz, F. R. Stermitz, J. B. Johnson, T. D. Beeson, S. Willen, J. Hsiang and K. Lewis, J. Med. Chem., 2001, 44, 261.
69. F. R. Stermitz, P. Lorenz, J. N. Tawara, L. A. Zenewicz and K. Lewis, Proc. Natl. Acad. Sci. USA, 2000, 97, 1433.
70. V. Reynolds, A. J. Plumptre, J. Greenham and J. B. Harborne, Oecologia, 1998, 115, 331.
71. M. Haribal and J. A. Renwick, Phytochemistry, 1996, 41, 139.
72. C. Ramirez-Tortosa, O. M. Andersen, P. T. Gardner, P. C. Morrice, S. G. Wood, S. J. Duthie, A. R. Collins and G. G. Duthie, Free Radical Biol. Med., 2001, 31, 1033.
73. R. Li, G. L. Kenyon, F. E. Cohen, X. Chen, B. Gong, J. N. Dominguez, E. Davidson, G. Kurzban, R. E. Miller, E. O. Nuzum, P. J. Rosenthal and J. H. McKerrow, J. Med. Chem., 1995, 38, 5031.
74. K. Yamaguchi, H. Honda, C. Wakisaka, A. Tohei and H. Kogo, Jpn. J. Pharmacol., 2001, 87, 67.
75. Y. Iwase, Y. Takemura, M. Ju-ichi, T. Mukainaka, E. Ichiishi, C. Ito, H. Furukawa, M. Yano, H. Tokuda and H. Nishino, Cancer Lett., 2001, 173, 105.
76. M. L. Edwards, D. M. Stemerick and P. S. Sunkara, J. Med. Chem., 1990, 33, 1948.
77. F. Herencia, M. L. Ferrandiz, A. Ubeda, I. Guillen, J. N. Dominguez, J. E. Charris, G. M. Lobo and M. J. Alcaraz, FEBS Lett., 1999, 453, 129.
78. M. E. Zwaagstra, H. Timmerman, A. C. van de Stolpe, F. J. de Kanter, M. Tamura, Y. Wada and M. Q. Zhang, J. Med. Chem., 1998, 41, 1428.
79. M. E. Zwaagstra, H. Timmerman, M. Tamura, T. Tohma, Y. Wada, K. Onogi and M. Q. Zhang, J. Med. Chem., 1997, 40, 1075.
80. M. E. Burow, S. M. Boue, B. M. Collins-Burow, L. I. Melnik, B. N. Duong, C. H. Carter-Wientjes, S. Li, T. E. Wiese, T. E. Cleveland and J. A. McLachlan, J. Clin. Endocrinol. Metab., 2001, 86, 1750.
81. Z. Gyorgypal and A. Kondorosi, Mol. Gen. Genet., 1991, 226, 337.
82. P. W. Groundwater, K. R. Solomons, J. A. Drewe and M. A. Munawar, Prog. Med. Chem., 1996, 33, 233.
83. G. W. Peet and J. Li, J. Biol. Chem., 1999, 274, 32655.
84. S. Ghosh, M. J. May and E. B. Kopp, Annu. Rev. Immunol., 1998, 16, 225.
85. R. Stoll, C. Renner, S. Hansen, S. Palme, C. Klein, A. Belling, W. Zeslawski, M. Kamionka, T. Rehm, P. Muhlhahn, R. Schumacher, F. Hesse, B. Kaluza, W. Voelter, R. A. Engh and T. A. Holak, Biochemistry, 2001, 40, 336.
86. M. Dai, Y. Feng and P. J. Tonge, J. Am. Chem. Soc., 2001, 123, 506.
87. R. Erdmann and W. H. Kunau, Yeast, 1994, 10, 1173.
88. F. Kreuzaler and K. Hahlbrock, Eur. J. Biochem., 1975, 56, 205.
89. F. Kreuzaler, R. J. Light and K. Hahlbrock, FEBS Lett., 1978, 94, 175.
90. F. Kreuzaler, H. Ragg, W. Heller, R. Tesch, I. Witt, D. Hammer and K. Hahlbrock, Eur. J. Biochem., 1979, 99, 89.
91. T. Lanz, S. Tropf, F. J. Marner, J. Schröder and G. Schröder, J. Biol. Chem., 1991, 266, 9971.
92. S. Tropf, B. Kärcher, G. Schröder and J. Schröder, J. Biol. Chem., 1995, 270, 7922.
93. J. M. Jez and J. P. Noel, J. Biol. Chem., 2000, 275, 39640.
94. D. Y. Suh, J. Kagami, K. Fukuma and U. Sankawa, Biochem. Biophys. Res. Commun., 2000, 275, 725.
95. T. Kortemme and T. E. Creighton, J. Mol. Biol., 1995, 253, 799.
96. T. Akiyama, M. Shibuya, H. M. Liu and Y. Ebizuka, Eur. J. Biochem., 1999, 263, 834.
97. C. Beckert, C. Horn, J.-G. Schnitzler, A. Lehning, W. Heller and M. Veit, Phytochemistry, 1997, 44, 275.
98. Y. Yamazaki, D. Y. Suh, W. Sitthithaworn, K. Ishiguro, Y. Kobayashi, M. Shibuya, Y. Ebizuka and U. Sankawa, Planta, 2001, 214, 75.
99. S. Tropf, T. Lanz, S. A. Rensing, J. Schröder and G. Schröder, J. Mol. Evol., 1994, 38, 610.
100. M. L. Durbin, G. H. Learn, Jr., G. A. Huttley and M. T. Clegg, Proc. Natl. Acad. Sci. USA, 1995, 92, 3338.
101. Y. Helariutta, M. Kotilainen, P. Elomaa, N. Kalkkinen, K. Bremer, T. H. Teeri and V. A. Albert, Proc. Natl. Acad. Sci. USA, 1996, 93, 9033.
102. V. Oberholzer, M. L. Durbin and M. T. Clegg, Gen. Genet. Syst., 2000, 75, 1.
103. G. Hrazdina, F. Kreuzaler, K. Hahlbrock and H. Grisebach, Arch. Biochem. Biophys., 1976, 175, 392.
104. K. W. Zuurbier, J. Leser, T. Berger, A. J. Hofte, G. Schröder, R. Verpoorte and J. Schröder, Phytochemistry, 1998, 49, 1945.
105. H. Morita, Y. Takahashi, H. Noguchi and I. Abe, Biochem. Biophys. Res. Commun., 2000, 279, 190.
106. I. Abe, H. Morita, A. Nomura and H. Noguchi, J. Am. Chem. Soc., 2000, 122, 11242.
107. J. M. Jez, M. E. Bowman and J. P. Noel, Proc. Natl. Acad. Sci. USA, 2002, 99, 5319.
108. K. H. Choi, R. J. Heath and C. O. Rock, J. Bacteriol., 2000, 182, 365.
109. J. M. Jez, M. E. Bowman and J. P. Noel, Biochemistry, 2001, 40, 14829.
110. Y. Helariutta, P. Elomaa, M. Kotilainen, R. J. Griesbach, J. Schröder and T. H. Teeri, Plant Mol. Biol., 1995, 28, 47.
111. S. Eckermann, G. Schröder, J. Schmidt, D. Strack, R. Edrada, Y. Helariutta, P. Elomaa, M. Kotilainen, I. Kilpelainen, P. Proksch, T. H. Teeri and J. G. Schröder, Nature, 1998, 396, 387.
112. J. Schröder and G. Schröder, Z. Naturforsch. C, 1990, 45, 1.
113. N. Rupprich, H. Hildebrand and H. Kindl, Arch. Biochem. Biophys., 1980, 200, 72.
114. S. Schanz, G. Schröder and J. Schröder, FEBS Lett., 1992, 313, 71.
115. A. Schoppner and H. Kindl, J. Biol. Chem., 1984, 259, 6806.
116. G. Schröder, J. W. Brown and J. Schröder, Eur. J. Biochem., 1988, 172, 161.
117. J. Fliegmann, G. Schröder, S. Schanz, L. Britsch and J. Schröder, Plant Mol. Biol., 1992, 18, 489.
118. F. Melchior and H. Kindl, FEBS Lett., 1990, 268, 17.
119. T. Yamaguchi, F. Kurosaki, D. Y. Suh, U. Sankawa, M. Nishioka, T. Akiyama, M. Shibuya and Y. Ebizuka, FEBS Lett., 1999, 460, 457.
120. R. Hain, H. J. Reif, E. Krause, R. Langebartels, H. Kindl, B. Vornam, W. Wiese, E. Schmelzer, P. H. Schreier, R. H. Stocker and K. Stenzel, Nature, 1993, 361, 153.
121. J. D. Hipskind and N. L. Paiva, Mol. Plant Microbe Interact., 2000, 13, 551.
122. L. Fremont, Life Sci., 2000, 66, 663.
123. G. J. Soleas, E. P. Diamandis and D. M. Goldberg, Clin. Biochem., 1997, 30, 91.
124. E. H. Siemann and L. L. Creasy, Am. J. Enol. Vitic., 1992, 43, 49.
125. D. Y. Suh, K. Fukuma, J. Kagami, Y. Yamazaki, M. Shibuya, Y. Ebizuka and U. Sankawa, Biochem. J., 2000, 350((Pt 1)), 229.
126. R. Gehlert and H. Kindl, Phytochemistry, 1991, 30, 457.
127. T. Reinecke and H. Kindl, Phytochemistry, 1994, 35, 63.
128. R. Preisig-Muller, P. Gnau and H. Kindl, Arch. Biochem. Biophys., 1995, 317, 201.
129. R. Preisig-Muller, R. Gehlert, F. Melchior, U. Stietz and H. Kindl, Biochemistry, 1997, 36, 8349.
130. G. Schröder and J. Schröder, J. Biol. Chem., 1992, 267, 20558.
131. N. A. M. Saleh, H. Fritsch, F. Kreuzaler and H. Grisebach, Phytochemistry, 1978, 17, 183.
132. A. B. Christensen, P. L. Gregersen, J. Schröder and D. B. Collinge, Plant Mol. Biol., 1998, 37, 849.
133. A. Baumert, W. Maier, D. Gröger and R. Deutzmann, Z. Naturforsch., C, 1994, 49, 26.
134. V. S. Parmar, S. C. Jain, S. Gupta, S. Talwar, V. K. Rajwanshi, R. Kumar, A. Azim, S. Malhotra, N. Kumar, R. Jain, N. K. Sharma, O. D. Tyagi, S. J. Lawrie, W. Errington, O. W. Howarth, C. E. Olsen, S. K. Singh and J. Wengel, Phytochemistry, 1998, 49, 1069.
135. K. T. Junghanns, R. E. Kneusel, A. Baumert, W. Maier, D. Gröger and U. Matern, Plant Mol. Biol., 1995, 27, 681.
136. K. Springob, R. Lukacin, C. Ernwein, I. Groning and U. Matern, Eur. J. Biochem., 2000, 267, 6552.
137. R. Lukacin, K. Springob, C. Urbanke, C. Ernwein, G. Schröder, J. Schröder and U. Matern, FEBS Lett., 1999, 448, 135.
138. R. Lukacin, S. Schreiner and U. Matern, FEBS Lett., 2001, 508, 413.
139. A. M. A. Abd El-Mawla and L. Beerhues, Planta, 2002, 214, 727.
140. L. Rocha, A. Marston, M. A. Kaplan, H. Stoeckli-Evans, U. Thull, B. Testa and K. Hostettmann, Phytochemistry, 1994, 36, 1381.
141. (a) L. Beerhues, FEBS Lett., 1996, 383, 264; (b) S. Peters, W. Schmidt and L. Beerhues, Planta, 1998, 204, 64.
142. W. Schmidt and L. Beerhues, FEBS Lett., 1997, 420, 143.
143. M. C. Wildermuth, J. Dewdney, G. Wu and F. M. Ausubel, Nature, 2001, 414, 562.
144. N. B. Paniego, K. W. Zuurbier, S. Y. Fung, R. van der Heijden, J. J. Scheffer and R. Verpoorte, Eur. J. Biochem., 1999, 262, 612.
145. Y. Okada and K. Ito, Biosci. Biotechnol. Biochem., 2001, 65, 150.
146. W. Borejsza-Wysocki and G. Hrazdina, Plant Physiol., 1996, 110, 791.
147. I. Abe, Y. Takahashi, H. Morita and H. Noguchi, Eur. J. Biochem., 2001, 268, 3354.
148. J. Schröder, Trends Plant Sci., 1997, 2, 373.
149. J. Schröder, S. Raiber, T. Berger, A. Schmidt, J. Schmidt, A. M. Soares-Sello, E. Bardshiri, D. Strack, T. J. Simpson, M. Veit and G. Schröder, Biochemistry, 1998, 37, 8417.
150. I. Atanassov, E. Russinova, L. Antonov and A. Atanassov, Plant Mol. Biol., 1998, 38, 1169.
151. I. E. Burbulis, M. Iacobucci and B. W. Shirley, Plant Cell, 1996, 8, 1013.
152. B. Ylstra, M. Muskens and A. J. Van Tunen, Plant Mol. Biol., 1996, 32, 1155.
153. J. Rozema, R. A. Broekman, P. Blokker, B. B. Meijkamp, N. de Bakker, J. van de Staaij, A. van Beem, F. Ariese and S. M. Kars, J. Photochem. Photobiol. B, 2001, 62, 108.
154. J. Gorham, in ‘Bryophytes: Their Chemistry and Chemical Taxonomy’, Proceedings of The Phytochemical Society of Europe, Vol. 29, pp. 171–200, Oxford University Press, 1990.
155. R. Mechoulam and S. Ben-Shabat, Nat. Prod. Rep., 1999, 16, 131.
156. M. Fellermeier, W. Eisenreich, A. Bacher and M. H. Zenk, Eur. J. Biochem., 2001, 268, 1596.
157. H. Baer, Clin. Dermatol., 1986, 4, 152.
158. P. M. Dewick, Medicinal Natural Products: a Biosynthetic Approach, Wiley, Chichester, 1997.
159. D. H. Hong, S. B. Han, C. W. Lee, S. H. Park, Y. J. Jeon, M. J. Kim, S. S. Kwak and H. M. Kim, Arch. Pharm. Res., 1999, 22, 638.
160. T. Rosen and D. B. Fordice, South Med. J., 1994, 87, 543.
161. R. Grazzini, D. Hesk, E. Yerger, D. Cox-Foster, J. Medford, R. Craig and R. O. Mumma, J. Am. Soc. Hortic Sci., 1995, 120, 343.
162. D. J. Schultz, E. B. Cahoon, J. Shanklin, R. Craig, D. L. Cox-Foster, R. O. Mumma and J. I. Medford, Proc. Natl. Acad. Sci. USA, 1996, 93, 8771.
163. B. E. Scheffler, S. O. Duke, F. E. Dayan and E. Ota, in Regulation of Phytochemicals by Molecular Techniques, Chapter 12, p. 257, Pergamon Press, Amsterdam, 2001.
164. C. I. Nimbal, J. F. Pederson, C. N. Yerkes, L. A. Weston and S. C. Weller, J. Agric. Food Chem., 1996, 44, 1343.
165. G. Fate, M. Chang and D. G. Lynn, Plant Physiol., 1990, 93, 201.
166. B. S. Moore and J. N. Hopke, ChemBioChem, 2001, 2, 35.
167. J. M. Henson, M. J. Butler and A. W. Day, Annu. Rev. Phytopathol., 1999, 37, 447.
168. M. J. Butler and A. J. Day, Can. J. Microbiol., 1998, 44, 1115.
169. I. Fujii, Y. Mori, A. Watanabe, Y. Kubo, G. Tsuji and Y. Ebizuka, Biochemistry, 2000, 39, 8853.
170. H.-F. Tsai, Y. C. Chang, R. G. Washburn, M. H. Wheeler and K. J. Kwon-Chung, J. Bacteriol., 1998, 180, 3031.
171. K. Ueda, K. M. Kim, T. Beppu and S. Horinouchi, J. Antibiot. (Tokyo), 1995, 48, 638.
172. N. Funa, Y. Ohnishi, I. Fujii, M. Shibuya, Y. Ebizuka and S. Horinouchi, Nature, 1999, 400, 897.
173. B. S. Moore, personal communication, 2001.
174. N. Funa, Y. Ohnishi, Y. Ebizuka and S. Horinouchi, J. Biol. Chem., 2002, 277, 4628.
175. Y. M. Zhang, M. S. Rao, R. J. Heath, A. C. Price, A. J. Olson, C. O. Rock and S. W. White, J. Biol. Chem., 2001, 276, 8231.
176. C. Keel, U. Schnider, M. Maurhofer, C. Voisard, J. Laville, U. Burger, P. Wirthner, D. Haas and G. Defago, Mol. Plant Microbe Interact., 1992, 5, 4.
177. J. M. Raaijmakers, D. M. Weller and L. S. Thomashow, Appl. Environ. Microbiol., 1997, 63, 881.
178. C. Picard, F. Di Cello, M. Ventura, R. Fani and A. Guckert, Appl. Environ. Microbiol., 2000, 66, 948.
179. U. F. Walsh, J. P. Morrissey and F. O'Gara, Curr. Opin. Biotechnol., 2001, 12, 289.
180. M. G. Bangera and L. S. Thomashow, J. Bacteriol., 1999, 181, 3155.
181. S. J. Hammond, M. P. Williamson, D. H. Williams, L. D. Boeck and G. G. Marconi, J. Chem. Soc., Chem. Commun., 1982, 344.
182. O. Choroba, D. Williams and J. Spencer, J. Am. Chem. Soc., 2000, 122, 5389.
183. H. Chen, C. C. Tseng, B. K. Hubbard and C. T. Walsh, Proc. Natl. Acad. Sci. USA, 2001, 98, 14901.
184. A. M. van Wageningen, P. Kirkpatrick, D. Williams, B. Harris, J. Kershaw, N. Lennard, M. Jones, S. Jones and P. Solenberg, Chem. Biol., 1998, 5, 155.
185. T.-L. Li, O. Choroba, H. Hong, D. Williams and J. Spencer, Chem. Commun., 2001, 20, 2156.
186. V. Pfeifer, G. J. Nicholson, J. Ries, J. Recktenwald, A. B. Schefer, R. M. Shawky, J. G. Schröder, W. Wohlleben and S. Pelzer, J. Biol. Chem., 2001, 276, 38370.
187. B. K. Hubbard, M. G. Thomas and C. T. Walsh, Chem. Biol., 2000, 7, 931.
188. S. R. Muir, G. J. Collins, S. Robinson, S. Hughes, A. Bovy, C. H. Ric De Vos, A. J. van Tunen and M. E. Verhoeyen, Nat. Biotechnol., 2001, 19, 470.
189. R. A. Dixon, C. J. Lamb, S. Masoud, V. J. Sewalt and N. L. Paiva, Gene, 1996, 179, 61.
190. X. Dong, E. L. Braun and E. Grotewold, Plant Physiol., 2001, 127, 46.

---