Current mechanistic understanding of thiamin

Current mechanistic understanding of thiamin diphosphatedependent enzymatic reactions Frank Jordan Department of Chemistry and the Program in Cellular and Molecular Biodynamics, Rutgers, the State University, Newark, New Jersey, 07102 USA. E-mail: [email protected]; Fax: ⫹1 973 353 1264 Received (in Cambridge, UK) 22nd October 2002 First published as an Advance Article on the web 17th January 2003 Covering: 1990–2002 The mechanism of thiamin diphosphate-dependent enzymatic reactions is discussed, concentrating on two enzymes involved in decarboxylating pyruvic acid, the yeast pyruvate decarboxylase and the pyruvate dehydrogenase multienzyme complex from Escherichia coli. The availability of high-resolution X-ray structures for several thiamin diphosphate-dependent enzymes, the use of site-specifically substituted protein variants (resulting from site-directed mutagenesis), the development of model reactions for the various putative intermediates, and the application of new mechanistic tools in solution have all contributed to a much better understanding of the role of the protein component in catalysis. Perhaps the most important advance in our understanding of these mechanisms concerns the role of the 4⬘-aminopyrimidine component of the coenzyme, widely ignored prior to the publication of the X-ray results. The current view is that the two aromatic rings both contribute to catalysis, perhaps carrying out an intramolecular proton transfer to initiate the various reactions, an ability that makes this coenzyme virtually unique among coenzymes. 1 2 2.1 2.1.1 2.1.2 2.1.3 2.2 2.2.1 2.2.2 3 3.1 3.2

Introduction What have we learnt from the three-dimensional structure of ThDP enzymes Conserved structural features The thiamin diphosphate fold: the Mg(II) environment The “V” coenzyme conformation Conserved hydrogen bonds Non-conserved structural features Acid-base groups at the active center Additional cofactors Properties and observation of thiamin-bound covalent intermediates on and off the enzymes The 1⬘4⬘-iminopyrimidine intermediate The C2-carbanion or ylide/carbene—ionization of a weak acid

3.3 3.4

The substrate-ThDP complex C2
-hydroxyethylidene thiamin diphosphate— the enamine or second carbanion, and C2
-hydroxyethylthiamin diphosphate, the product-ThDP complex 3.4.1 Acid-base properties 3.4.2 Direct observation of the enamine 3.4.3 Pathways with oxidation of the enamine 4 ThDP-bound intermediates are chiral; applications to chiral organic synthesis 4.1 Coenzyme-bound intermediates 4.2 Carboligase reactions 4.3 Synthetic applications of benzoin-type condensations 5 Examples of the assignment of roles to specific residues in YPDC and PDHc-E1

Frank Jordan was born in Budapest, Hungary and immigrated to the United States in 1957. He received his undergraduate training at Drexel University in Philadelphia (BSc 1964 in Chemistry) and his PhD at the University of Pennsylvania’s Department of Chemistry in 1967 working with Edward R. Thornton. This was followed by a NATO fellowship in the laboratory of the late Bernard Pullman at the Sorbonne in Paris (1967–68) on problems in quantum biochemistry, and by an NIH fellowship in the laboratory of Frank H. Westheimer at the Department of Chemistry at Harvard, carrying out research on a model study for the enzyme acetoacetate decarboxylase (1968–70). In 1970, he joined the Department of Chemistry at Rutgers in Newark as an assistant professor, rising to the rank of full professor in 1979 and to his current title as Rutgers University Board of Governors Professor of Chemistry in 1997. He served as chair of his department from 1985–91 and 1994–2000 and has been director of a NSF-funded multidisciplinary research training program in Cellular and Molecular Biodynamics since 1994. He was a recipient of: (1) the Rutgers University Board of Trustees Award for Excellence in Research in 1983; (2) a Johnson and Johnson Research Discovery Fellowship in 1988–90; (3) the Honor Scroll of the American Institute of Chemists, New Jersey Section in 1995; (4) the 1998 Excellence in Education Award of the North Jersey Section of the American Chemical Society. He was elected a Fellow of the American Association for the Advancement of Science (AAAS) in 1995. Since 1998, he has served at first as an ad hoc then as full member of the Physical Biochemistry Study Section of the National Institutes of Health. In addition to studies described in this review, he has a long-standing interest in the structure, mechanism and folding of serine proteases. He is the author or co-author of more than 175 publications and reviews and 3 patents, and has graduated 46 PhD students from his laboratories. Frank Jordan

184

Nat. Prod. Rep., 2003, 20, 184–201 This journal is © The Royal Society of Chemistry 2003

DOI: 10.1039/b111348h

5.1 5.2 5.3 5.4 6 6.1 6.2 6.3 7 8 8.1 8.2 8.3 8.4 9 10

E51, a conserved YPDC residue hydrogen bonded to the N1⬘ ThDP atom D28, a residue conserved on pyruvate decarboxylases only E477, a residue located over the thiazolium ring of pyruvate decarboxylases H407-E1 from E. coli, a residue not visible in the structure of the wild-type enzyme Rate limiting steps Substrate and solvent kinetic isotope effects Chemical quench of reaction intermediates Kinetic resolution of enzyme-bound intermediates Examples of new ThDP enzymes Summary: unifying mechanistic features of ThDPdependent decarboxylations Role of the V conformation-strain may drive decarboxylation A “solvent” effect stabilizes zwitterionic intermediates Intramolecular uphill proton transfer to initiate the reaction Acid-base catalysis Acknowledgements References

Abbreviations used ThDP, thiamin diphosphate; PDC, pyruvate decarboxylase; YPDC, yeast pyruvate decarboxylase; POX, pyruvate oxidase; TK, transketolase; PDHc, the pyruvate dehydrogenase multienzyme complex; PDHc-E1, the first (ThDP-dependent) subunit of PDHc; E. coli, Escherichia coli; PFOR, pyruvate ferredoxin oxidoreductase; AHAS, acetohydroacid synthase; BFD, benzoylformate decarboxylase; HEThDP, C2α-hydroxyethylthiamin diphosphate, the acetaldehyde.ThDP adduct; HBThDP, C2α-hydroxybenzylthiamin diphosphate, the benzaldehyde.ThDP adduct; LThDP, C2α-lactylthiamin diphosphate, the pyruvate.ThDP adduct. 1

Introduction

The availability of the high-resolution three-dimensional structure of their cognate enzymes has greatly aided the elucidation of the role of the protein component in the reaction mechanisms of coenzymes derived from water-soluble vitamins. This information needs to be supplemented by the corresponding model systems, so that comparisons can be made between the enzymatic and model chemical systems for the same reaction. Such comparisons enable us to test current theories of enzyme catalysis, revealing how the protein provides catalysis over and above that provided by the coenzyme. In this review recent developments are summarized in our understanding of how one of the best-known water-soluble vitamins, thiamin (vitamin B1; the coenzyme is thiamin diphosphate, ThDP), carries out its functions. With the structures of several key members of this family of enzymes reported during the past decade, there is a better opportunity to start assessing the common and divergent features of such enzymes. The mechanistic understanding of this family of enzymes is useful in several respects. Some of these enzymes and their variant forms (created by site-directed mutagenesis) have been found to produce chiral compounds with excellent enantiomeric excess, making them usable perhaps even on an industrial scale. Also, novel metabolic pathways are being uncovered at a steady pace with some reactions requiring ThDP, and the mechanistic understanding can be applied to these new enzymes as well. The reactions of ThDP include: non-oxidative reactions (reactions B, G, Scheme 1), such as those of pyruvate decarboxylase (PDC) 1–6 producing acetaldehyde and benzoylformate decarboxylase (BFD) 7 producing benzaldehyde; and oxidative

pathways (reactions C–F, Scheme 1), such as the pyruvate oxidases (POX) using flavin as the oxidant to produce acetate 8 or acetyl phosphate,9 the pyruvate dehydrogenase multienzyme complex (PDHc) and the entire family of 2-oxoacid dehydrogenase multienzyme complexes which utilize lipoic acid as the oxidant to produce acetylCoA or acyl-CoA;10 and pyruvateferredoxin oxidoreductases (PFOR) which use Fe4S4 cluster chemistry to produce acetylCoA.11 A yet different view of ThDP reactions classifies reactions into “ligating” and “nonligating” types. This is a useful notion since the central enamine intermediate on all ThDP enzymatic pathways is an excellent nucleophile. There are some reactions of ThDP where this nucleophilic character leading to ligation is obvious, as with the ketolases (such as transketolase TK in the pentose shunt pathway),12 and acetohydroxyacid synthase (AHAS),13 an enzyme important in the biosynthesis of branched chain amino acids in plants and other organisms (with the notable exception of animal cells). But, in virtually all ThDP reactions, carboligase side products (derived from the reaction of the central enamine intermediate with acetaldehyde or pyruvate, resulting in the formation of acetoin or acetolactate, respectively) are commonly observed. As will be suggested below, conceptually, it is also useful to view the reactions of reductive acyl transfer between the E1 and E2 subunits of the 2-oxoacid dehydrogenase multienzyme complexes as such ligation reactions. An excellent review by Kluger 1 covered the topics here discussed through the 1980s, prior to the publication of any highresolution structural data. During the intervening years, many advances were made not only with the publication of a large number of three-dimensional structures of ThDP enzymes, but also concerning model chemistry for the structure and reactivity of the key intermediates. Although the mechanisms drawn in Schemes 2 and 3 have not changed in a fundamental way, the influence of the 4⬘-aminopyrimidine ring and of the activecenter amino acids in catalysis is information not available in the past and this information must guide our current thinking about the mechanism of ThDP-dependent enzymes. Examples in this review will emphasize the systems studied in the author’s laboratory, yeast pyruvate decarboxylase (YPDC) and the E1 subunit of the pyruvate dehydrogenase multienzyme complex from Escherichia coli, especially its ThDP-dependent E1subunit (PDHc-E1), while references to other work will be quoted as needed. Also, though there are very important issues concerning enzyme regulation with several of the enzymes, such issues are outside the scope of this review. The structures of the compounds discussed in this review are given in Table 1. 2 What have we learnt from the three-dimensional structure of ThDP enzymes 2.1 2.1.1

Conserved structural features The thiamin diphosphate fold: the Mg(II) environment

It was proposed by Perham and co-workers, prior to the appearance of the X-ray structures,14 that there is a conserved region in all ThDP enzymes with the sequence GDGX26N(C)N, as exemplified on YDPC with D444 and N470, N471 (Fig. 1). In all structures, these residues appear to serve as ligands to the required Mg(II) ion, which provides the platform for binding the diphosphate side-chain. Binding of the Mg(II) ion probably precedes binding of ThDP. Substitution into the ThDP fold of D444 or of corresponding residues is typically fatal to the enzyme. Substitution of even the adjacent glycine in PDHc-E1 from E. coli showed that binding to this fold also impacts on the hysteretic activation of the enzymes by ThDPⴢMg(II), a common feature of all ThDP enzymes. As seen in Fig. 2, the Mg(II) site and its relationship to the ThDP site are nicely conserved between YPDC and PDHc-E1. Nat. Prod. Rep., 2003, 20, 184–201

185

Scheme 1

2.1.2

The “V” coenzyme conformation

One of the most striking results reported for all three of the first X-ray structures (YPDC, POX and TK) was the presence of the so-called “V” coenzyme conformation (see Fig. 3 for illustration on YPDC). The conformation describes the disposition of the planar aromatic 4⬘-aminopyrimidine and thiazolium rings with respect to the bridging methylene group.15 This V conformation is only seldom found in thiamin derivatives in the absence of protein and is clearly not at a low potential energy.16,17 The immediate question is what forces support this conformation, and what are the consequences of this enforced conformation. Under the V conformation, almost poised to support it, there is always a large hydrophobic side-chain. In a paper by Guo et al., the I415 residue located under the V conformation in the YPDC was systematically substituted to smaller and smaller side-chains, indicating that the size of the 186

Nat. Prod. Rep., 2003, 20, 184–201

side-chain was indeed important.18 Using computational approaches, it was suggested that the flexibility of thiamin is reduced by the isoleucine side-chain, and it also contributes to the “solvent” effect near the active center, long thought to contribute to the catalytic rate acceleration by lowering the local dielectric constant.19 Importantly, it was also shown that the methionine found at the corresponding position in POX, or the leucine found in TK, when placed into position 415 in YPDC, reduces the activity significantly. Apparently, each active center has evolved to accommodate one of these large side-chains. 2.1.3

Conserved hydrogen bonds

In all X-ray structures reported to date, there appear to be highly conserved hydrogen bonds to all three of the nitrogen atoms of the 4⬘-aminopyrimidine ring. On YPDC (Fig. 1), these are E51 to N1⬘, the main chain NH of I415 to N3⬘, and

Scheme 2

Scheme 3

the main chain C ᎐ O of G413 specifically on the N3⬘ side of the 4⬘-aminopyrimidine ring to the N4⬘-HN3 (the subscript denoting the side of the ring). The hydrogen bonds to N1⬘ and N4⬘ appear to be stronger, approximately 2.6 Å long over several structures. Almost certainly, these three conserved hydrogen bonds have importance in catalysis, not only in binding the coenzyme. Early on, Schellenberger 2 and his co-workers had shown that the ThDP analogues lacking any of the three nitrogen atoms in the 4⬘-aminopyrimidine ring would have much-impaired activities, especially the 1⬘-deaza and 4⬘-deaza analogues. It is also noteworthy that the three hydrogen bonds help to modulate the state of ionization and influence the rotation around the C4⬘–N4⬘ bond, both factors affecting the tautomerization, while there are no obvious corresponding conserved interactions with the thiazolium ring. 2.2 2.2.1

Fig. 1 Amino acid side-chains surrounding thiamin diphosphate and Mg(II) and YPDC.

Non-conserved structural features Acid-base groups at the active center

There are a number of steps on each ThDP-dependent pathway that demand the presence for multiple ionizable groups (please see Scheme 2 for YPDC and BFD and Scheme 3 for PDHc-E1). Surprisingly, the constellation of amino acids around the ThDP while conserved between PDCs from the yeasts Saccharomyces uvarum 5 and Saccharomyces cerevisiae 6 and the bacterium Zymomonas mobilis 20 (Fig. 1), is different from that on BFD,7 even though the reactions are the same except for C6H5 in the Nat. Prod. Rep., 2003, 20, 184–201

187

Table 1

Compounds discussed in this Review

latter replacing CH3 in PDC. While both PDC and BFD must protonate an enamine intermediate, it is not even clear that the residue carrying out this reaction is even of the same type; it could be a His on BFD and an Asp/Glu 21–23 on PDC. Nor is the residue carrying out the deprotonation of the hydroxyl group the same, it is an Asp on YPDC 22,23 and is likely to be a His on BFD.7,24 The only acid-base groups highly conserved are an Asp as a ligand on the Mg(II) ion, and a Glu across from the N1⬘ atom of ThDP. An interesting three-dimensional conservation of the constellation of active center acid-base residues between TK 12 and PDHc-E1 25 has become evident from the X-ray structures. The conservation has been explained by the similarity of the ligation reaction carried out in the second half of both reactions. A histidine was identified in a position in which it can protonate a very weak base in the acceptor; H263 is perhaps responsible for protonating the aldehyde carbonyl group in TK, while H407 probably protonates one of the sulfur atoms of the dithiolane ring of lipoamide (see Section 5). 2.2.2

Additional cofactors

In a significant fraction of ThDP enzymes (perhaps the most important ones in metabolism), oxidation of the enamine 188

Nat. Prod. Rep., 2003, 20, 184–201

follows the decarboxylation reaction. The oxidizing agent can be flavin (POX), the dithiolane ring of lipoic acid (PDHc and related enzymes) and Fe4S4 cluster in PFOR. A rather important issue concerning these reactions is whether the oxidation proceeds by a free radical intermediate (only proven so far for PFOR). 3 Properties and observation of thiamin-bound covalent intermediates on and off the enzymes Owing to the presence of several possible covalent intermediates formed between the ThDP and the substrate and product, as well as the so-called enamine, along with the emerging evidence for different states of ionization and tautomerization of ThDP itself, the study of ThDP reactions on and off the enzymes has long been a treasure trove for both bioorganic chemists and mechanistic enzymologists. 3.1

The 1⬘4⬘-iminopyrimidine intermediate

In the paper on the structure of the YPDC from the yeast Saccharomyces cerevisiae,6 and on the basis of the structure of other ThDP enzymes,7,9,11,12,25 we pointed out that one of the properties of the V conformation is to bring to within the range

Fig. 4 Rapid-scan stopped-flow on mixing pyruvate with the E477Q YPDC variant; A310 is assigned to the 1⬘,4⬘-imino tautomer of ThDP; assignment of A350 is incomplete.

Fig. 2 Amino acid side-chains surrounding ThDP and Mg in PDHcE1 from E. coli.

Fig. 5 Circular dichroism spectra of the E477Q YPDC variant with benzoylformic acid in the absence and presence of pyruvamide. The inset shows the “raw” initial and final spectra, while the larger spectra represent difference spectra.

Fig. 3 Conformation of ThDP on YPDC with I415 as a conformational pivot.

of 3.0–3.4 Å the N4⬘ and C2 atoms—a most intriguing distance. This observation tempted others and us to suggest that there is intramolecular proton transfer between these two atoms (reaction A, Scheme 1). It was also suggested that, for at least the shorter of these two distances, it would be very difficult to fit a proton at C2 and two protons at N4⬘ simultaneously, giving credence to the idea that the 4⬘-aminopyrimidine ring cycles between the 4⬘-amino and 1⬘,4⬘-imino tautomeric forms during the reaction sequence. This rare imino tautomer is stabilized by three highly conserved hydrogen bonds to N1⬘, N3⬘ and N4⬘H3⬘ (denoting the proton bonded to N4⬘ on the N3⬘ side of the pyrimidine ring). The participation of the 4⬘-aminopyrimidine in this manner had been suggested more than 20 years ago.26–28 In a recent paper from the author’s group, spectroscopic evidence, both rapid-scan stopped-flow UV and circular dichroism (CD), was presented indicating the presence of a hitherto unreported absorption between 300 and 310 nm on certain YPDC variants in the presence and even in the absence of the pyruvate substrate (Figs. 4 and 5).29 Such an absorption could

also be generated by adding a base to N1-methylpyrimidinium salts in either water or in aprotic organic solvents. On the basis of this model system, we suggested that the absorption on YPDC between 300 and 310 nm pertains to the 1⬘,4⬘-iminopyrimidine tautomer of ThDP. Concurrently, a group at Moscow State University suggested that the broad negative CD signature centered around 320–330 nm, and observed for three decades on TK and on mammalian PDHc’s is indeed pertinent to the 1⬘,4⬘-imino ThDP tautomer.30 Importantly, we have identified conditions under which this negative CD signature centered at 320–330 nm could be observed on YPDC (Fig. 6) and on PDHc-E1 from E. coli (to be published) for the first time. However, these observations typically required so-called “ligation” conditions, i.e., the presence of an acceptor for the enamine. This raises the interesting further issue: should this negative CD band at 320–330 nm indeed be associated with the 1⬘,4⬘-imino ThDP tautomer, why and how does the protein stabilize it under these particular conditions? In yet to be published work, the positive CD band has now also been observed on PDHc-E1 in the presence of the phosphonolactyl-ThDP, a stable analogue of LThDP, strongly suggesting that with this analogue the ThDP exist in its imino tautomeric form. At the same time, we have also shown that the positive CD band centered at 305–310 nm and the negative one near 320–330 nm exist under different conditions.31 While we are confident of the assignment of the positive band at 305– 310 nm to the fixed V conformer of iminoThDP, we are less certain of the origins of the negative band at 320–330 nm. It is Nat. Prod. Rep., 2003, 20, 184–201

189

there appeared a report by Arduengo and co-workers, reporting for the first time generation of the C2 carbanion/ylide/carbene in non-aqueous medium using KH as base. The 13C chemical shift of this species was found to be 254 ppm.34 Work in the author’s laboratory has not only confirmed these results, but also showed that the species is very reactive with benzaldehyde,35 confirming Breslow’s proposals for thiamin action.36 Model studies also showed that the ylide/carbene reacts with benzaldehyde via a nucleophilic addition, and no evidence of reactions typical of carbenes, such as insertion, could be found.37,38 In view of the unexpected 13C chemical shift of the C2 carbanion compared to that of the same nucleus in the thiazolium ion precursor (a nearly 100 ppm deshielding experienced on ionization), more experiments are needed to confirm the state of ionization at this crucial position on a variety of ThDP enzymes, and under various conditions such as pH. 3.3

Fig. 6 Circular dichroism spectra of the E477Q YPDC variant in the presence of acetaldehyde in the absence and presence of pyruvamide. Enzyme was dissolved to a concentration of 8.6 mg mL⫺1 (or 143 µM active sites). Acetaldehyde was added to 0.3 M concentration at 25 ⬚C. Pyruvamide was added to a final concentration of 20 mM. Upper panel represents actual spectra; difference spectra are shown in the lower panel.

clear, however, that both bands are associated with bound ThDP since there is little else conserved in the active centers of the four enzymes in which they have been observed. 3.2 The C2-carbanion or ylide/carbene—ionization of a weak acid Dissociation of the thiazolium C2H to the conjugate base is an essential first step for all ThDP reactions. Assuming that the thiazolium C2H is abstracted by the imino tautomer of ThDP acting as the base, one should ask whether the relevant pKas are balanced for rapid proton transfer. In model systems the pKas for the conserved Glu (across from the N1⬘ atom) and at N1⬘ are likely to be similar, while the pKa for ionization of the amino group at N4⬘H5⬘ (denoting the proton bonded to N4⬘ on the C5⬘ side of the pyrimidine ring) once the N1⬘ atom is protonated is ca. 12,26,29 and for ionization of C2H is 17–19.32 The special environment of the V coenzyme conformation present in all of these enzymes 18 assures that for a reasonable distance between N4⬘ and C2, proton transfer would take place at a rate exceeding the turnover number for many such enzymes, i.e. 60–100 s⫺1. So far, the only experiment to address the state of ionization at C2 of enzyme-bound ThDP was carried out by 13C NMR on YPDC, and reported a 13C chemical shift of 157 ppm, identical to that found for ThDP and thiazolium salt models, along with a line width very similar to that of free pyruvamide in the same solution. The results were interpreted to mean that C2H of the bound ThDP is undissociated at pH 6.0.33 In the same year, 190

Nat. Prod. Rep., 2003, 20, 184–201

The substrate-ThDP complex

The first ThDP-bound intermediate for the decarboxylation reactions is the result of a nucleophilic attack at the carbonyl carbon of the substrate and produces C2α-lactylthiamin diphosphate or LThDP. This would be a logical intermediate to draw, since the driving force for decarboxylation will be the electrophilic (electron withdrawing) effect of the thiazolium ring beta to the departing carboxylate. This intermediate is highly reactive (especially in its zwitterionic form) and its chemistry has been well established by Kluger and co-workers.39 Interestingly, Kluger and Smyth’s experiments suggested that this compound could not be decarboxylated by PDC.40 It was therefore important to establish the intermediacy of LThDP. Two methods have been recently reported that indeed indicate that this intermediate is distinct. In an imaginative study, Tittmann and Hübner and their group of collaborators carried out chemical (acid) quench of a mixture of pyruvate, ThDP and PDC, showing that one can observe both LThDP and HEThDP under these conditions,41 strongly suggesting that both are on the pathway. In the author’s laboratory, a different approach was developed. First, Haijuan Li showed that the E91D YPDC variant could form a stable apo enzyme, which could be reconstituted with virtually any ThDP derivative, including LThDP.42 Later, Min Liu 22 and Zhen Zhang 23 found that LThDP is partitioned on apo-YPDC, both to pyruvate (in the reverse direction) and is indeed decarboxylated according to the measurement of enamine and acetaldehyde. This evidence strongly supports the intermediacy of LThDP as a distinct chemical entity on pyruvate decarboxylases. In results being reported by the author and crystallographer collaborators, the phosphonolactyl analogue of LThDP (a C2 adduct of methylacetylphosphonate and ThDP which cannot be decarboxylated) was co-crystallized with PDHc-E1 from E. coli, providing a wealth of information regarding the reorganization taking place upon formation of the substrate-ThDP covalent complex.43 Two loops not seen in the absence of this inhibitor are clearly seen, and some residues move away from ThDP, some others move into position for catalysis. This is the first structure of a substrate-ThDP covalent complex analogue visualized on a ThDP enzyme and it underlines the importance of attempts to obtain further such structures. It is clear from this study that site-directed mutagenesis studies can lead to erroneous interpretations regarding the essential or nonessential nature of residues near the coenzyme site in the absence of structures such as this. 3.4 C2
-hydroxyethylidene thiamin diphosphate— the enamine or second carbanion, and C2
-hydroxyethylthiamin diphosphate, the product-ThDP complex These two intermediates are indeed at equilibrium with each other both on and off the enzymes, hence it is convenient to discuss them together.

The enamine intermediate could properly be called a central intermediate given its likely presence on all ThDP-dependent catalytic pathways. The author’s group has spent the past 20 years elucidating the properties of this intermediate in the absence and in the presence of ThDP enzymes. These are very important comparisons if we are to understand the contribution of the protein to individual steps. 3.4.1

Acid-base properties

To examine the rates of the steps leading to and from the C2α-carbanion/enamine in Scheme 2, in this laboratory the enamine was generated by the addition of base to a C2α-hydroxybenzylthiazolium or C2α-hydroxyethylthiazolium salt,44,45 and then either the proton transfer or oxidation rates were measured. First, the pKa at C2α(reaction B, Scheme 1) was measured in pure DMSO.46 By direct observation of the enamine generated in a stopped-flow spectrophotometer, the rate constants for reversible proton dissociation at the C2α position could also be measured in water. The pKa is between 15 and 16 47,48 for C2α-hydroxybenzylthiazolium salt, and near 15 in 32–37 mol% DMSO for C2α-hydroxyethylthiazolium salt (extrapolates to approximately 18 in water).49 There is a substantial primary deuterium kinetic isotope effect for the deprotonation reaction (4–6); the rate constant for reprotonation of the enamine/C2α carbanion by water is many orders of magnitude below diffusion control. It was concluded that YPDC and BFD assist in the protonation of the enamine to afford rate constants commensurate with enzymatic turnover numbers.48,49 How do ThDP enzymes solve this high pKa problem? The following experiment provides some answers to this riddle. When the E91D variant of apo-YPDC (EC 4.1.1.1) was exposed to C2α-hydroxybenzylThDP (HBThDP, Scheme 2), this putative intermediate was partitioned on the enzyme between release of the benzaldehyde product (evidenced by regeneration of active enzyme), and dissociation of the proton at C2α to form the enamine/C2α-carbanion intermediate (evidenced by the appearance of the visible spectrum of the intermediate). While the pKa for this dissociation is ∼15.4 in water, formation of the enamine at pH 6.0 on the enzyme indicates a greater than 9 unit pKa suppression by the enzyme environment.50 The fluorescence emission properties of thiochrome diphosphate, a fluorescent ThDP analogue and a competitive inhibitor for YPDC, when YPDC-bound resemble that observed in pentan-1-ol and hexan-1-ol, suggesting an apparent dielectric constant of 13–15 for the YPDC active center. Such a low effective dielectric constant could account for much of the observed >9 unit pKa suppression at the C2α position for ionization of HBThDP. The dramatic stabilization of this (and presumably other) zwitterionic intermediate(s) is sufficient to account for as much as a 109-fold rate acceleration on YPDC, providing the bulk of the rate acceleration by the protein over and above that afforded by the coenzyme. Similar experiments have been carried out with HEThDP,23 confirming the ability of YPDC to partition this intermediate as well, to acetaldehyde in the forward and the enamine in the reverse direction (the enamine in this case has a λmax near 295 nm, so that it could only be detected by indirect oxidative methods). With this intermediate, whose pKa is even higher, perhaps 18, the pKa suppression induced by the enzyme is even more impressive. At the same time, the enamine could also be generated from HEThDP by PDHc-E1. This result suggests that the PDHc-E1 also possess an active center that can stabilize zwitterions. 3.4.2

Direct observation of the enamine

The enamine has been detected directly by Vis spectroscopy on YPDC, when derived from conjugated 2-oxo acids 51 (XC6H4CH ᎐ CHCOCOOH and XC6H4COCOOH, where X is o-, m- or p- from the 2-oxo acid substituent). It could be demonstrated

that decarboxylation is quite fast for these alternate substrates,52 but once decarboxylated they become inhibitors, i.e., they are mechanism-based inhibitors. More recently, the p-nitroXC6H4COCOOH was used with BFD, and not only the enamine near 400 nm, but also a charge-transfer band centered at 620 nm, and attributed to the interaction of a protein sidechain with substrate-ThDP adduct (p-nitromandelyl-ThDP), could be monitored by rapid-scan stopped-flow methods, enabling the authors to estimate several specific rate constants for this alternate substrate.24 This and earlier results, including model studies from the author’s laboratory, strongly suggest that the structure of the enamine is dominated by the planar uncharged resonance contribution. Recently, Schneider and co-workers reported X-ray structural data interpreted as evidence for a TK-bound enamine derived from decarboxylation of 3-hydroxypyruvic acid,53 presumably leading to 1,2-dihydroxyethylidene-ThDP. This indeed is a remarkable observation in view of the length of time needed to complete the experiment during which the enamine would have to survive on the enzyme without being protonated. The authors claimed that indeed the C2–C2α environment is planar, as expected for the enamine resonance contribution. This is important evidence since a pyramidal environment would suggest that there is a hydrogen atom bonded to the C2α atom, or that it reflects more the resonance contributions from a dipolar zwitterion. 3.4.3

Pathways with oxidation of the enamine

In the initial model system in the author’s laboratory, the enamine was first generated under electrochemical conditions. According to both cyclic voltammetry and bulk electrolysis (enabling isolation of the products), there is a one-electron oxidation via a radical cation intermediate 54 [reaction C, Scheme 1]. The caveat with these initial experiments is that instead of the hydroxyl (or oxide) group almost certainly present on enzymes, the alcohol at the C2α atom was first converted into the methyl ether so that a strong base could be utilized for generation of the enamine. Subsequently, in a series of papers from the group of Fukuzumi, electrochemical experiments starting with thiazolium compounds and aldehydes were reported in which the radical cation could be observed directly by ESR.55 The enzyme pyruvate ferredoxin oxidoreductase PFOR carries out decarboxylation of pyruvic acid by ThDP, followed by oxidation via Fe4S4 clusters, and finally forms acetylCoA from the acetyl equivalent and CoASH. It has been known for two decades that this reaction proceeds by free radical chemistry, the radical perhaps corresponding to the thiazolium cation radical generated in the models. Much of the historical background on this topic is covered in a recent review by Ragsdale.56 Fontecilla-Camps and colleagues reported the first X-ray structure for a PFOR,11 and followed it up with a remarkable structure of the enzyme with a stable radical signal (ESR).57 These authors found that the putative acetyl-ThDP radical (a one-electron oxidation product of the enamine/ 2α-hydroxyethylidene) has an unexpected distribution of the free electron. The electron spin is found mostly around the now non-planar thiazolium ring. The ring itself may have undergone tautomerization of a proton from the C4-methyl group to C5, while the double bond between C4 and C5 had migrated to the C4–C4exomethylene position, Yet another interesting feature is the remarkably long (>1.8 Å) C2–C2α bond. These are very exciting preliminary findings that certainly will be examined in greater detail. The crystal showing radical activity was grown at pH 9.0, a pH at which thiamin undergoes reversible ring opening reactions and it would be desirable to confirm the findings at pH values closer to neutrality. The enzyme pyruvate oxidase (POX) has dual functions: it produces acetyl phosphate in Lactobacillus plantarum 9 and Nat. Prod. Rep., 2003, 20, 184–201

191

acetic acid in Escherichia coli.8 Models for flavin-catalyzed oxidation of the central enamine (reaction D, Scheme 1) at the C2α position by flavin have yielded the following information: (1) There is a need for both a base and for a free hydroxyl group at the C2α position. (2) Surprisingly, the model reaction can proceed by both a one-electron 58 and two-electron pathway since a 5-deazaflavin analogue also acts as an oxidizing agent, presumably proceeding by hydride transfer, albeit much more sluggishly than flavin.49 This latter reaction is more analogous to NAD⫹-dependent oxidations. (3) The second-order rate constant for oxidation of the enamine with flavin is >6000 s⫺1 M⫺1, requiring a relatively modest “effective concentration” on the enzyme.58 The principal function of the protein is to bring FAD and ThDP into close proximity on POX. The crystal structure of POX 9 indicates that the arrangement of the cofactors is inappropriate for hydride transfer; the isoalloxazine and thiazolium rings are not stacked, as is the case for example for glutathione reductase or dihydrolipoamide dehydrogenase, both of which carry out hydride transfer between FADH2 and NAD⫹. One is therefore compelled to conclude that POX carries out the reaction by two consecutive one-electron oxidations of the enamine to acetyl-ThDP, but that there is little energetic advantage to the radical versus the hydride pathway. In addition, for this enzyme, the protein has little to contribute to accelerate the redox process, other than bringing the cofactors into close enough proximity for efficient electron transfer. In the family of 2-oxoacid dehydrogenase multienzyme complexes, there are three proteins (named E1, E2 and E3) charged with the overall reaction generating acyl-coenzyme A and NADH as products. It is generally accepted that lipoyl-E2 (lipoic acid is covalently amidated onto a lysine residue of E2) is the oxidizing agent for the enamine product of decarboxylation which is non-covalently bound to E1. Incisive experiments from Frey’s laboratories 59 suggested that the redox process and acyl transfer are distinct steps in the mechanism. An early model for lipoic acid-catalyzed oxidation of the central enamine (reaction E in Scheme 1) demonstrated the reaction between the enamine and linear disulfides, but not with lipoic acid.60 Our first model based on pre-generation of the enamine 61 indicated that: (1) The enamine was virtually unreactive with lipoic acid in aprotic media. (2) Addition of Hg compounds (presumably to trap the sulfide) would lead to a modest reactivity. (3) A C2α-methoxyl derivative was reactive under these circumstances.61 It was concluded that the protein must play an important function, probably by the addition of an electrophile, such as a proton via a general acid, to one of the sulfur atoms of the lipoic acid, thereby shifting the equilibrium to the products. Next, methyl S-methyllipoate (forming thiosulfonium salts) was synthesized to mimic the positive charge created by protonation of lipoic acid. The diastereomeric pair of methyl S-methyllipoates (from ,-lipoic acid methyl esters) was then added to the enamine generated from C2α-methoxybenzylthiazolium salts giving the following information: 62 (1) There is a tetrahedral intermediate formed between the S-methyllipoate and the enamine (evidenced by NMR and mass spec); and (2) most importantly, the second order rate constant for the reaction is 6.6 × 104 s⫺1 M⫺1, perhaps 108-times faster than with unactivated lipoic acid. The results led us to speculate that, when the enamine-E1 complex reacts with lipoamide-E2, there is a tetrahedral intermediate formed, for an instant cross-linking E1 and E2, with the assistance of a general acid catalyst located on E1 near the ThDP site. Since the model studies were reported, there are now three relevant E1 structures in the literature.24,63,64 On the basis of the models here discussed, likely histidine residues have been identified that could mediate the reductive acetylation. On the PDHc-E1 from E. coli, the residue H407 has been identified as being important for this function (see Section 5 below). An important conclusion from the chemical model studies on FAD versus lipoic acid is that the protein surrounding the 192

Nat. Prod. Rep., 2003, 20, 184–201

cofactor must activate lipoic acid, while FAD is reactive enough to need no additional activation by the protein. Important further questions remain as to whether the intermediates on POX and PFOR are one and the same substratederived radical, and concerning the electronic structure of the radical. Importantly, the generality of the existence of a 2-acylThDP intermediate on the pathway of 2-oxoacid dehydrogenase complexes also needs to be reexamined in view of the emerging structural data. 4 ThDP-bound intermediates are chiral; applications to chiral organic synthesis 4.1

Coenzyme-bound intermediates

As can be seen in Scheme 2, depicting the stepwise mechanism of PDC, there are three ThDP-bound intermediates, of which LThDP and HEThDP possess a chiral center. One must therefore at least entertain the possibility that only one of the two enantiomers will be formed on the enzyme. There is another factor that may create an additional source of chirality, the existence of the conformationally rigid “V” ThDP. So, there may be additional non-obvious chiral species, such as the iminoThDP tautomer and the enamine. The HEThDP isolated from enzymatic reaction mixtures 35 years ago was shown to be optically active.65 Kluger and co-workers reported resolution of C2α-hydroxyethylthiamin, enabling them to relate the sign of the CD signal to an absolute configuration.1 With the availability of larger amounts of pure enzymes, several groups have examined the near-UV CD spectra (240– 350 nm) of potential enzyme-bound intermediates. Historically, one of the early observations was made on TK, which in the presence of the acceptor molecule gives rise to a very pronounced broad negative CD peak centered at 320–330 nm.66 This has been confirmed by several groups over three decades and was also seen, albeit more weakly, on the mammalian pyruvate dehydrogenase E1 component. Recently, we have observed several chiral intermediates on both YPDC (Figs. 5 and 6) and PDHc-E1 from E. coli. On the PDHc-E1, we have the clearest evidence to date of the cofactor ThDP being bound in a chiral environment. The potent inhibitor thiamin-2-thiothiazolone diphosphate (a very close structural analogue of ThDP in which the C2H is replaced by a C᎐᎐S)) gives a positive CD band centered at 330 nm and the band was shown to pertain to an enzyme-bound species (Fig. 7).67 The band is displaced by a second even more potent inhibitor thiamin-2-thiazolone diphosphate (the C2H is replaced by a C᎐᎐O) or large excess of ThDP. With a different inhibitor analogous to LThDP, the covalent complex formed between ThDP and methylacetylphosphonate, there was a

Fig. 7 Circular dichroism spectra of parental E1 titrated with thiamin 2-thiothiazolone diphosphate. The parental E1 (9.0 µM) in 10 mM KH2PO4 buffer (pH 7.0) was titrated with thiamin 2-thiothiazolone diphosphate at concentrations ranging from 0.49 to 50 µM and 1 mM MgCl2.

different positive CD signal induced, centered at 305–310 nm,31 which we believe pertains to the iminoThDP tautomer on the basis of model experiments and also observations on YPDC.29 Addition of pyruvate to such a system gave a larger negative band centered between 320 and 330 nm, reminiscent of that reported on TK. It is important to note that there is no free ThDP present hence there is no turnover, clearly signaling that there is a pyruvate-binding site in addition to that on ThDP, and, most importantly, the CD signals must pertain to ThDP itself in some ionization state/tautomeric state. On YPDC, there was a positive CD signal observed between 300 and 310 nm in a variant with the E477Q substitution and attributed to the iminoThDP (also seen by rapid-scan stoppedflow in the UV spectrum). Addition of acetaldehyde this time induced the broad negative CD band centered between 320 and 330 nm (Fig. 6).28 On the basis of experiments with PDHc-E1 and YPDC (also supported by model experiments), we suggest that the positive CD signal between 300 and 310 nm on any ThDP enzyme can be assigned to the 1⬘,4⬘;-iminoThDP tautomer and serves as a very convenient signature for this newly discovered species. We suspect that it has been in the spectra all along but was overlooked since it is rather weak, it is very near to the major peaks of the protein, and its quantification requires careful difference measurements and modern instrumentation. At the same time, there still appears some uncertainty regarding the assignment of the negative CD band centered at 320– 330 nm. Its presence has now been documented on several ThDP enzymes (TK, PDHc-E1 and YPDC), but there is no model system yet for this broad CD signal. What can be concluded from the cumulative data is that this negative CD band is strongest in the presence of an “acceptor” for the ligation reaction (discussed in the next paragraph) in all of these systems. The presence of this CD band is essentially independent of the amino acids surrounding ThDP since there is no significant active center conservation between YPDC, on the one hand, and TK and PDHc-E1, on the other. 4.2

Carboligase reactions

The seminal studies of Breslow, and many other studies since then, clearly showed that ThDP (indeed thiazolium salts with no substitution at the C2 atom) can carry out benzoin condensations, in which two carbonyl carbon atoms form a new bond (to be contrasted with the more widely known aldol-type condensations). This reaction is sometimes called an acyloin reaction, although historically those involved free radical rather than ionic mechanisms: 2 C6H5CHO

C6H5CH(OH)C(᎐᎐O)C6H5

(1)

Similar reactions can be envisioned with aliphatic 2-oxo acids and with aliphatic aldehydes: While such condensation (ligation) reactions are the essence of the chemistry of transketolase (TK) and acetohydroxyacid synthase (AHAS), such carboligation side products have been observed with virtually all classes of ThDP enzymes. The notion that TK 13 and AHAS 68 would produce products with essentially 100% enantiomeric excess (ee) is not surprising in view of 50 years of experience with the stereospecificity of enzymatic reactions. More surprising, and of potentially great synthetic utility, is the finding that these side products in the non-ligating enzymes are also produced with excellent ee. As an example of these side reactions, it was found that acetoin produced by the YPDC is optically active.69 More recently, it was shown that the E477Q active center variant of YPDC while

much impaired in acetaldehyde production becomes an acetoin synthase, while the D28N variant also impaired in acetaldehyde production becomes an acetolactate synthase. Equally significantly, both products are produced with significant enantiomeric excess (there is a single chiral center in each compound).70 The observed enantioselectivity suggests that when the enamine intermediate forms a bond with the aldehyde or pyruvic acid at the carbonyl carbon it preferentially reacts with the re or si face of the electrophile (Fig. 8) perhaps determined by the shape of the channel that the electrophile traverses en route.

Fig. 8 Accounting for formation of different enantiomers of carboligase products depending on whether the enamine is approached by the si or re face of the “acceptor” molecule.

4.3

Synthetic applications of benzoin-type condensations

The fact that such reactions can be driven in the carboligation direction starting with acetaldehyde and condensing onto it benzaldehyde, thereby producing phenylacetylcarbinol (PAC), makes this of potential use for producing important pharmaceutical intermediates, α-ketols, with high ee. EnzymeⴢThDP ⫹ CH3CHO Enzyme-Enamine ⫹ C6H5CH(OH)C(C᎐᎐O)CH3 C6H5CHO (phenylacetylcarbinol, PAC)

(3)

For several years, the group at Juelich, Germany has made great strides in using ThDP enzymes as synthetic tools. They have used PDC, BFD and most recently benzaldehyde lyase (which carries out a retro-benzoin condensation) in very imaginative ways to synthesize many classes of compounds using these enzymes. In some cases, enzyme engineering was used to improve the ee, and most recently they have also used combinatorial genetic methods such as gene shuffling to select for higher ee.There are excellent recent reviews by this group of this emerging and potentially very fruitful field.71–75 5 Examples of the assignment of roles to specific residues in YPDC and PDHc-E1 Assignment of specific roles to particular amino acids surrounding the ThDP has turned out to be quite challenging. There is considerable experience now with multiple activecenter variants on several related enzymes to try to assign function. Interestingly, and disappointingly, very thorough steady-state kinetic analysis covering the entire pH range of activity yielded relatively modest information regarding specific steps in which a residue may participate. An exception to this is (2) Nat. Prod. Rep., 2003, 20, 184–201

193

Fig. 9 Left: A single YPDC subunit; the protein is represented as a ribbon drawing whereas the ThDP cofactor is shown as a space filling model. Right: the substrate activation pathway from C221 to ThDP.

perhaps the delineation of the substrate activation pathway in which the Hill treatment with the value of the Hill coefficient provided excellent guidance. In a series of studies over the past decade, the signal transduction pathway for substrate activation on YPDC was mapped out residue by residue (Fig. 9).76–85 Most frustrating is perhaps the finding that even YPDC active-center residues whose substitution leads to a 100–1000-fold decrease in kcat or kcat/Km caused, in virtually all cases, only modest shifts in rate-pH profiles.13,21

N1⬘-methylThDP to the wild-type YPDC or to the E51A variant led to no observable activity, but it was found that N1⬘methylThDP bound to Apo-E51A some 18-times better than did ThDP itself. We conclude from this that all three forms, including the protonated ThDP and both tautomers, must coexist on YPDC, but this is impossible for the N1⬘-methylThDP, which in fact is bound in a satisfactory manner (Fig. 10). This

5.1 E51, a conserved YPDC residue hydrogen bonded to the N1⬘ ThDP atom As suggested above, on the basis of the X-ray structure alone, one would surmise that this highly conserved residue will have a dramatic effect in all ThDP enzymes since there is a good hydrogen bond formed between the N1⬘ atom and the carboxylate oxygen atom. Yet, at the resolution of the X-ray structures currently available for ThDP enzymes, the hydrogen positions are not defined, so that neither the state of ionization nor the tautomeric state of ThDP could be deduced with any certainty. This is generally true for all acid-base residues so that modeling with the assumption of a particular state of ionization, or of the tautomeric state, would be futile, given the uncertainty of the effects of the microenvironment on the pKas. On YPDC, the E51Q, E51D, E51N and E51A substitutions all led to greatly diminished kcat and kcat/Km values.86 The E51D substitution turned out to be informative, since at low substrate concentrations (steps starting with free enzyme and culminating in decarboxylation) the logkcat/S0.5-pH plots displayed an acid-shift for the entire curve, showing that the distance of the negative charge from N1⬘ influences this curve. A plausible explanation is that the pKa of the N1⬘-protonated pyrimidinium ring is reduced (i.e., the ring is more difficult to protonate), this in turn making it more difficult to catalyze the tautomeric equilibration, which in turn would reduce the rate at which the iminoThDP can abstract the thiazolium C2H to form the ylide/ carbanion in the required first step. It had been reported that the E51Q substitution reduced the rate of C2H to D exchange (as a measure of the rate of the first step ylide formation) significantly, but even for that variant C2H dissociation may not have become rate limiting.33 A different type of experiment was designed to test the E51A variant (low but detectable activity) with ThDP and N1⬘methylThDP in parallel. The premise of the experiment is shown in Scheme 4, testing whether the activity requires both tautomers, and the N1⬘-protonated pyrimidinium intermediate, an obligatory species for the interconversion of the two tautomers. The N1⬘-methylThDP is an electrostatic mimic for the N1⬘-protonated intermediate, and it can be converted into the imino tautomer as we have shown, but, it precludes existence of the 4⬘-aminopyrimidine form. In fact, addition of 194

Nat. Prod. Rep., 2003, 20, 184–201

Fig. 10 Top: Tautomerization of the 4⬘-aminopyrimidine. Bottom: Forms of N1⬘-MethylThDP.

suggests that the conserved residue is needed to catalyze the tautomeric equilibration, but in its absence the tautomerization is still possible, as all of the variants at this position displayed residual activity. Assuming flexibility of the enzyme in the region of E51, in theory, a water molecule could also catalyze the tautomeric equilibration, albeit at a slower rate. In accord with this idea, incubation of the E51A (but not of the other E51-substituted variants) with ThDP overnight increased the activity significantly. Curiously, the extent of impairment of catalysis experienced by substitution at this highly conserved residue varied with the enzyme, the only correlation being that those enzymes with

Scheme 4

Parallels in the mechanisms of TK and PDHc to account for the similarities in their active centers.

alkaline-shifted pH optima suffered less activity reduction by alteration of this residue. As an example, PDHc-E1 from E. coli with the corresponding E571A substitution exhibited only 50-fold reductions in activity, albeit the ThDP binding was clearly damaged according to several criteria.87 Hence, the function of the conserved acidic residue is difficult to dissect even with the cumulative evidence at hand. Apparently, positioning of the 4⬘-aminopyrimidine at the active center by the three conserved hydrogen bonds is very important, but only the hydrogen bond from N1⬘ is to a carboxylic acid, while the other two are to peptide backbone donor or acceptor. With the pKas at N1⬘ for 4⬘-aminopyrimidine and in glutamate being so similar in models, one would surmise that catalysis, in addition to binding, is an important role for this conserved residue. 5.2

D28, a residue conserved on pyruvate decarboxylases only

In the structures of the pyruvate decarboxylases from yeast 5,6 and Zymomonasmobilis,20 this residue is located above the 4(-aminopyrimidine ring witha water molecule nearby. While both kcat and kcat/Km-type terms are significantly reduced in the YPDC D28A and D28N variants, its clear intervention in a step post-decarboxylation became evident from studies of the carboligase side reaction. Remarkably, though these substitutions led to greatly diminished acetaldehyde production, they had no adverse effect on the carboligase side reactions; in fact, this was the only substitution identified so far that converted the enzyme into a fairly respectable acetolactate synthase.70 From these observations, we deduced that the state of ionization is D28COOH through formation of LThDP, but D28 transfers a proton thereafter to E477, thus becoming D28COO⫺, so as to repel the second pyruvate in the wild-type enzyme. This would account for the fact that with wild-type enzyme the carboligase side products constitute less than 1% of the acetaldehyde product. The strongest evidence for a post-decarboxylation role was generated from studies in which we partitioned HEThDP on the apo-E91D YPDC variant according to eqn. (4):

Enamine

HEThDP

Acetaldehyde

(4)

[Parenthetically, the E91D YPDC variant is used for such studies because of its ability to be converted into apo-enzyme then reconstituted with any ThDP analogue.42] We have methods to detect partitioning in both directions. We then tested partitioning of HEThDP with the doubly substituted D28A/E91D variant and found that this substitution only (among D28, E477, H114 and H115) allowed virtually no formation of acetaldehyde.22,23 We therefore conclude that a major role of residue D28 is to help deprotonate the C2α–OH for release of acetaldehyde. Curiously, at least so far as the wildtype enzymes are concerned, the position of D28 in YPDC is substituted by a serine in BFD.7 It is very difficult to envision how a serine could carry out this function. The residue D28 on YPDC also appears to have a role in the protonation of the enamine, perhaps in conjunction with residue E477 according to partial reduction in the rate of enamine formation from HEThDP by the D28A/E91D variant. Consistent with this suggestion, in careful difference CD spectra (Fig. 11), the D28A and E477Q variants appear to give rise to some of the same and one different long-lived chiral intermediate, both different from that observed with the wild-type YPDC. The positive CD signal

Fig. 11 CD difference spectra of YPDC-bound intermediates on wildtype and variant enzymes: on the left, protein ± substrate; on the right, results of subtracting the spectra on the left from each other.

Nat. Prod. Rep., 2003, 20, 184–201

195

of this intermediate was centered at 290 nm, very similar to the wavelength determined for the enamine derived from pyruvic acid (295 nm) in models, and consistent with the idea that this residue helps to protonate the enamine. 5.3 E477, a residue located over the thiazolium ring of pyruvate decarboxylases The E477Q substitution on YPDC led to greatly diminished acetaldehyde release with very much reduced kcat and kcat/Km.21 However, the carboligase reaction leading to acetoin was not only not impaired, to the contrary, its rate was even faster than with the wild-type enzyme; effectively the E477Q substitution converted this variant into an acetoin synthase.70 This behavior would immediately suggest a role subsequent to the decarboxylation step (enamine formation) since, through decarboxylation, acetaldehyde and acetoin formation share a common pathway. More insight into the behavior of this residue was gleaned from rapid-scan stopped-flow and difference CD spectroscopic measurements, indicating the build-up of an intermediate with λmax near 310 nm, and a positive CD signal centered at 305, respectively. These observations suggest the presence of the iminoThDP tautomer. According to the intermediate partitioning experiments mentioned above, addition of HEThDP to the apo-E91D/E477Q variant allowed normal release of acetaldehyde but impaired the rate of enamine formation. This evidence suggests that E477 at least contributes to the enamine protonation step. As suggested by Fig. 8, there is also enamine buildup according to the CD experiment with the E477Q YPDC variant. For this variant, there is a difference peak somewhat broader than with the D28A variant and at a slightly longer wavelength, since it probably represents a superposition of a signal for the enamine at 295 nm and the one for the iminoThDP at 305 nm (Fig. 11). On the basis of the electronic spectroscopic data indicating that with the E477Q YPDC variant the lifetime of the iminoThDP is increased it appears that the E477 may be required for formation of the iminoThDP, perhaps deprotonating the 4⬘-amino group, possibly with the intermediacy of a water molecule. This could be one of the most important and unique functions of residue E477 and it may explain why the steadystate kinetic data imply that the residue E477 participates both in reactions starting with free enzyme through decarboxylation, and those starting with decarboxylation and culminating in product release. We recently suggested that the iminoThDP has functions throughout the reaction pathway.88 5.4 H407-E1 from E. coli, a residue not visible in the structure of the wild-type enzyme As a final example, the fascinating story of the H407 residue of PDHc-E1 will be discussed since it taught the author an important lesson regarding structure-function correlations. In the recently published 3D structure of the PDHc-E1 from E. coli, the first example of the large class of bacterial PDHcs with α2 quaternary structures,25 three regions of the enzyme of 886 residues could not be seen: the N-terminal 1–55 residues, residues 401–413 and 541–557. A BLAST search of the sequence data bank indicated that, among the ThDP enzymes, only TK shows significant homology to this PDHc-E1. If the structures of TK 12 and PDHc-E1 from E. coli are overlaid, the overall fit is reasonable (both are homodimers) and the shape similarity is evident. When the ThDPs of the two structures are superimposed, a loop bearing H263 of TK overlaps the active center of PDHc, and sequence alignments suggest that H407 of PDHc-E1 corresponds to H263 of TK in function. Therefore, the H407A variant was prepared; it was overexpressed in E. coli so that its properties could be studied. Assays carried out with the H407A-E1 variant reconstituted with E2-E3 subcomplex signaled an activity 196

Nat. Prod. Rep., 2003, 20, 184–201

approximately 1/1000th of the parental E1. But, assaying the H407A-E1 variant for E1-specific activity with the external oxidant 2,6-dichlorophenolindophenol indicated a less than 10-fold decrease of the activity. Since this oxidant enters the sequence post-decarboxylation (it oxidizes the enamine), it could be concluded that the steps through decarboxylation were only modestly affected by the substitution. The obvious next possibility, namely that the residue H407 affects the communication between the E1 and E2 subunits, was tested by MALDI-TOF mass spectrometry. In particular, E1 and pyruvate carry out reductive acetylation of E2, specifically at its lipoamide, a bound form of lipoic acid (which has a reactive dithiolane ring system, a five-membered saturated ring with two adjacent sulfur atoms). During acetyl transfer the mass of the E2 increases by 44 mass units, and this increase could be monitored very conveniently by MALDI-TOF mass spectrometry. Using this tool, it was demonstrated that indeed reductive acetyl transfer was greatly impaired by the H407A PDHc-E1 substitution.89 In parallel with these studies, a structural study was undertaken in which crystals were grown of the PDHc-E1 to which the covalent adduct formed between ThDP and methylacetylphosphonate was added. This complex is coined phosphonolactyl-ThDP, denoting its relationship to LThDP. The structure of this complex is the first LThDP analogue visualized on a ThDP enzyme and its refinement is nearing completion,43 but several results already obvious are highly relevant to this review. Most importantly, there are major shifts among active center residues: (1) Residue E522 is no longer near the phosphonolactyl-ThDP, rather residue Y599 takes its place forming a hydrogen bond to the phosphonic acid thereby explaining why the E522A substitution produced only modest reduction in activity; and (2) the loop bearing H407 is now visible (presumably due to its reduced mobility) and is indeed interacting with one of the two oxygen atoms of the phosphonic acid. The conclusion from this is that H407 participates in events through decarboxylation, presumably stabilizing the LThDP adduct by hydrogen bonding to one of the carboxyl oxygen atoms. This stabilization is not of great energetic significance (less than 1.4 kcal mol⫺1) since steps through decarboxylation are only modestly affected by the H407A substitution. However, the structure implies a whole lot more. Since we now have experimental evidence showing that reductive acetylation of E2 is also affected by this substitution at E1, and in a most dramatic fashion, we can make some statements concerning the post-decarboxylation phase of the reaction as well. (1) Very likely the CO2 departs and the dithiolane ring of E2 enters from the same side of the enamine intermediate, i.e., the reaction takes place by retention of configuration at the C2α atom. (2) The H407 side-chain is in a position to interact with the lipoic acid during the reaction as an electrophile by hydrogen bonding to a sulfur atom, or outright protonating it concomitantly with the reaction of the second sulfur atom with the C2α atom of the enamine, as suggested by the previous model study.62 The results with this H407A variant provide a hypothesis to answer the question as to why there is such high conservation of active centers between TK carrying out a ligation reaction and PDHc-E1. Let us consider the simplest covalent mechanism for the reductive acetyl transfer from E1 to lipoyl-E2 as a nucleophilic attack by the C2α-carbanion/enamine on the S8-atom of the dithiolane ring with concomitant ring opening. This reaction can also be viewed as a “ligation” reaction (ligating the C2α atom of the enamine to S8 of the lipoic acid), and, most importantly, it may involve electrophilic assistance by protonation or hydrogen bonding of an electrophile to a weak base, a carbonyl oxygen on TK, and the S6 atom on the lipoamide-E2. The conservation of the constellation of amino acid residues at the active centers, and especially of the position of the loop bearing H407 in PDHc-E1 (corresponding to the H263 in TK; both of these residues are highly conserved), suggests that the

same solution to this difficult problem may have evolved in both classes of reactions (Scheme 4). 6 6.1

Rate limiting steps Substrate and solvent kinetic isotope effects

As in any multi-step enzymatic reaction, the issue of which step is rate limiting is always of interest. Several approaches have been used over the years to study this issue. The earliest approach was to determine 13C/12C ratios at natural abundance in the release of 45CO2/44CO2 with an isotope ratio mass spectrometer resulting from decarboxylation of pyruvate by YPDC.90 These isotope ratios can lead to an estimate of the 13C/12C kinetic isotope effects (KIE) on the steady-state kinetic term Vmax/Km, usually interpreted in terms of rate-limiting steps among steps starting with the free enzyme through the first irreversible step, decarboxylation. A model for the intrinsic KIE for the pure decarboxylation step was also determined as 1.050, which is consistent with other such findings.90 The KIE for YPDC was 1.008, suggesting that decarboxylation (Scheme 2) is only partially rate limiting on the enzyme; in other words, the LThDP is converted into the enamine at a rate at least 5-times faster than it reverts back to free enzyme. In an extensive study of YPDC, Schowen and co-workers obtained similar values.91 With the availability of substituted YPDC variants, this method has been put to use again and Huskey and co-workers have studied a number of such variants with substitution at both the catalytic and at the regulatory sites.92 An examination of Scheme 2 may tempt one to assume that YPDC would show significant solvent KIEs (SKIE; determining the rate of the reaction under the same conditions in light and heavy water while paying attention to the issue of shifting pKas in the two solvents) given the number of reaction steps that may require proton transfers. A complete such study for YPDC is complicated by: (1) the need for many substrate concentrations to fully describe the sigmoidal Michaelis-Menten plot resulting from the substrate activation phenomenon; and (2) the bell-shaped Vmax/Km-pH and Vmax-pH profiles requiring measurements at many pH (pD) values. Earlier, Schowen and co-workers also reported values for wild-type YPDC over a limited pH range.91 With the availability of active-center substituted variants, several such studies were carried out.22,83,84 Again, due to the substrate activation process, analysis is also more complex: it is simplest at lowest substrate concentrations where Vmax/Km conditions apply, and under saturating substrate concentrations where Vmax conditions apply. Simply, results at substrate concentration approaching zero (Km) pertain to transition states starting with decarboxylation and culminating in product release. On the basis of results with the E51D, E477Q, C221A, C221D and C221E variants, along with those on the wild-type YPDC, we could conclude that under no circumstances examined (either under Vmax/Km or Vmax conditions) does one observe rate-limiting proton transfers, since the SKIE is never greater than 1.5. Rate-limiting proton transfer would have a theoretical SKIE near 3.0. What is observed instead is that the SKIE is always lower at low substrate than at high substrate—so clearly there are significant changes in hydrogen bonding strengths between the two limits. While the significantly inverse SKIE at low substrate concentrations (ca. 0.4–0.5) had earlier been attributed to the regulatory site cysteine ionization,91 subsequent studies with that cysteine 221 substituted to alanine or glutamate/aspartate ruled that idea out,83,84 as did the state of ionization of that cysteine—cysteine 221 is dissociated at pH

6.0 the optimum pH for YPDC.78 So, with the help of the SKIEs we can conclude that at high substrate concentrations no proton transfer is fully rate limiting (from decarboxylation through product release), and that very similar steps are rate limiting in YPDC, PDC from Zymomonas mobilis 93 and BFD.94 At low substrate concentrations on YPDC, the strength of hydrogen bonding changes at the active center once the substrate occupies the substrate activation site, presumably C221. In other words, the hydrogen bonding at the active center becomes weaker as the enzyme progresses from the free to the substrate-activated state. The E51D active-center YPDC substitution is the only one so far showing an effect on SKIE under Vmax conditions (Dkcat = 0.89 ± 0.01). Given the location of the residue E51, this evidence suggests that E51 and, indirectly, the iminoThDP tautomer are involved in some step starting with decarboxylation and culminating in product release. 6.2

Chemical quench of reaction intermediates

Recently, Tittmann et al., reported a novel method to determine forward rate constants for the decarboxylation pathway in Scheme 2 by mixing pyruvate with PDC from Zymomonas mobilis or YPDC in a chemical quench instrument within ms then quenching with strong acid.41 Under these conditions, LThDP and HEThDP are stable and their concentrations can be used to estimate forward rate constants. The method is premised on the observation that the chemical shifts of the C6⬘ proton of ThDP and the ThDP-related covalent complexes are distinct. A limitation of the method is that it requires reasonably large concentrations to be detected by NMR (albeit very little protein) and it cannot determine reverse rate constants. This method is quick and can be used under various conditions with many enzyme variants. This method is also useful to pinpoint the role of particular amino acids in the mechanism. As an example, the D27A Zymomonas mobilis variant was shown to have a role in the release of product, as the HEThDP intermediate was found to accumulate, in accord with results on the corresponding residue D28 in YPDC discussed in Section 5. 6.3

Kinetic resolution of enzyme-bound intermediates

In favorable cases where there are chromophoric intermediates, one may be able to monitor the kinetic fate of the intermediates directly. The ThDP-dependent enzyme BFD uses benzoylformate as substrate. It was found in the author’s laboratories that the enamine resulting from decarboxylation of this substrate on ThDP should have a λmax of 380 nm. Guided by this information, the interaction of p-nitrobenzoylformic acid with BFD was studied by rapid-scan stopped-flow spectroscopy. Two transients with λmax at 620 and 410 nm were found to interconvert, the first transient was attributed to the substrate-ThDP covalent complex, the one at 410 nm to the enamine.24 Several rate constants could be extracted from the pre-steady-state measurements for this alternate substrate, providing the values given below for the wild-type BFD. One could then turn to substituted BFD variants and ask which step(s) was affected by the substitution.24,95 As seen for the wild-type BFD, product release is rate limiting for this alternate substrate. With the H70A active-center substitution, k1 is diminished by a factor of > 3000. While the precise nature of the interaction producing the charge transfer band at 620 nm is not certain, we have now observed such charge transfer bands on both YPDC and BFD, and from very different types of chromophoric substrates. These bands kinetically precede the enamine so they most likely

Nat. Prod. Rep., 2003, 20, 184–201

197

pertain to the substrate ThDP covalent complex, analogous to LThDP in Schemes 2 and 3. 7

Examples of new ThDP enzymes

With the advances in gene mapping methods, there are many new proteins being discovered at a rapid pace. In theory, ThDP enzymes could be identified by the GDGX26NN or GDGX26NCN motif mentioned above which characterizes the Mg(II)diphosphate binding fold. A few enzymes relatively recently explored will be highlighted as examples. Townsend and co-workers identified in Streptomyces clavuligerus an enzymatic activity that condenses glyceraldehyde3-phosphate with -arginine in a ThDP-dependent reaction leading to N 2-(2-carboxyethyl)arginine,96 a key intermediate on the clavulanic acid biosynthetic pathway (eqn. (5)).

(5)

Acetoin dehydrogenase (eqn. (6)) is a multienzyme complex, very similar to the 2-oxo acid dehydrogenase multienzyme complexes.97 It consists of three proteins, E1 (again, it has ThDP), while E2 and E3 are analogous to those found in E2 and E3 in the 2-oxo acid dehydrogenases, i.e. with lipoic acid and FAD, respectively. The reaction couples the reverse of reaction G to reaction E in Scheme 1. The initial phase is a retro-benzoin condensation, where the enamine is oxidized by lipoyl-E2.

(6)

A very interesting activity was reported to be carried out by 2-hydroxyphytanoylcoenzyme A lyase 98 shown in eqn. (7).

(7) The enzyme indolepyruvate decarboxylase is similar in activity to pyruvate and benzoylformate decarboxylases and its structure was shown to be similar as well 99 (eqn. (8)).

(8)

Benzaldehyde lyase 100,101 (eqn. (9)) carries out the retrobenzoin condensation of benzoin to benzaldehyde and promises to be useful as a biocatalyst.

(9)

An excellent summary has been published of these and other novel ThDP enzymes with synthetic promise.75 8 Summary: unifying mechanistic features of ThDP-dependent decarboxylations 8.1 Role of the V conformation-strain may drive decarboxylation The universality of the “V” coenzyme conformation observed in the large number of structures of ThDP enzymes now 198

Nat. Prod. Rep., 2003, 20, 184–201

known, and the knowledge that this is the less stable conformation found in the absence of enzymes, certainly suggests that there is some strain imposed on binding of substrate to coenzyme. There are two examples reported of the strain sensed in the thiazolium ring: (1) the thiazolium cation radical 57 (perhaps its rearranged version) and (2) the complex between PDHc-E1 from E. coli and the phosphonolactyl-ThDP bound to it.43 If we adopt the proposal of Dunathan 102 regarding pyridoxal action which suggested that the scissile bond be nearly perpendicular to the aromatic plane for decarboxylation, that conformer of lactate in LThDP which places the carboxylate group perpendicular to the thiazolium ring should be the reactive one. This indeed is seen in the alignment of the carboxyl surrogate in the free phosphonolactylThDP structure;103 however, the perturbations of the aromatic thiazolium ring system when enzyme-bound were more unexpected. The suggestion is that binding of the pyruvate in the covalent LThDP adduct introduces significant strain to the system, and the release of this strain will be an important driving force to the decarboxylation reaction. The evidence with the active site residues does not suggest that a negative charge at the active center is involved to create electrostatic repulsion of the carboxylate thus driving the decarboxylation. The finding of the non-planar thiazolium intermediates strongly supports steric distortion as a driving force. Of course, indirectly it also suggests that the aromaticity of the thiazolium ring may not be that large (compared to say benzene with 36 kcal mol⫺1) if the ring can be distorted by the energy gained from binding substrate and coenzyme. In addition, the strain found in these enzyme-bound intermediates might also provide the structural explanation for the “alternation of active sites with a functional dimer” mechanism suggested by kinetic studies on both YPDC 88 and BFD.24 8.2

A “solvent” effect stabilizes zwitterionic intermediates

It had been reported by Lienhard and co-workers 19 and confirmed by a variety of studies that model reactions for ThDPcatalyzed decarboxylations take place considerably faster in ethanol than in water. This could be rationalized on the basis of a common electrostatic feature of key intermediates such as the ylide and the enamine, in addition to LThDP and the conjugate base alcoholate of HEThDP, namely a dipolar or zwitterionic charge distribution. On the basis of reaction kinetic theories, conversion of an ion into a zwitterion (partial or full charge neutralization on going to the transition state) should be faster in a solvent of lower dielectric constant. It is of course more difficult to demonstrate such effects on the enzymes themselves. Nevertheless, as early as 1970, Wittorf and Gubler showed that the active center environment behaved as one of lower effective dielectric constant according to the behavior of a fluorophore.104 The availability of the E91D variant of the YPDC and apo-E1 of PDHc enabled the author’s group to test this hypothesis yet once more. In regarding the YPDC mechanism, enamine should by all rules be uphill the reaction HEThDP energetically. In fact, in a series of model studies (quoted earlier), the pKa for this process was found to be ca. 15–16 for C2α-hydroxybenzyl ThDP and at least 17 for the C2α-hydroxyethylThDP. Since YPDC has a rather promiscuous active center it will accept benzoylformic acid as substrate, and apparently it will also accept HBThDP as a probe. Reconstitution of the apo-E91D YPDC variant with HBThDP was shown to: (a) release benzaldehyde in the forward direction (using an aldehyde assay) and (b) form the enamine in the reverse direction according to two spectroscopic tools that can oxidatively trap the enamine.50 This was a remarkable finding, given the pKa mentioned above. More recently, it was shown beyond any doubt that both YPDC and PDHc-E1 could also convert HEThDP into the enamine! For YPDC we have pretty good evidence that it is probably the residue D28, very likely with the

assistance of E477 (and perhaps even with the assistance of the iminoThDP tautomer), that carries out the reversible protonation-deprotonation at the C2α carbon atom. [Invoking the principle of microscopic reversibility, one can study this reaction from either direction since the same catalyst group should be responsible for catalysis in both directions in a reversible reaction.] However, HEThDP is almost certainly not on the pathway of the PDHc-E1, hence on the latter enzyme there is no specific residue charged with the protonation-deprotonation at the C2α carbon atom. Therefore, most likely, the observations imply that the environment favors shifting the equilibrium on the enzyme towards the zwitterionic intermediates, hence lowering the effective pKa by 9–10 units—a most impressive pKa lowering, perhaps among the most dramatic reported so far. This further gives us a relative barrier lowering of as much as 13–14 kcal mol⫺1. A new method was devised to test the active center environment of ThDP enzymes. Thiochrome diphosphate is a fluorescent alkaline oxidation product of ThDP and is a competitive inhibitor of YPDC. The fluorescence emission spectrum of this fluorophore was measured on YPDC and in the series of alkan-1-ols, methanol to hexan-1-ol, and in water. The wavelength of emission maximum against solvent dielectric constant gave a respectable linear plot for the solvent series. The wavelength of emission maximum for the fluorophore probe attached to YPDC placed the “effective protein dielectric constant” between those of hexan-1-ol and pentan-1-ol, the interpolated dielectric constant being between 13 and 15. This effective dielectric constant compared to that of water would be sufficient to account for much of the observed pKa lowering.50 It had been estimated 91 that the YPDC protein accelerates the rate of acetaldehyde formation by as much as 1012–1013. We suggest that perhaps as much as 108–109 of that rate acceleration would be achieved by virtue of this “solvent” effect. While such estimates of protein dielectrics are fraught with complications (and are controversial to boot), we believe that the significant pKa suppression observed on both YPDC and PDHc-E1 is beyond any doubt, although the magnitude of the exact suppression is hard to judge. It is tempting to generalize further that such effects will make a significant contribution to rate accelerations on other (perhaps all) ThDP enzymes.

branched-chain dehydrogenase E1 subunit.110 So far, substitution of no residue at the active centers of YPDC or PDHc-E1 has abolished the activity, or even diminished it, by more than ca. 104. Several substitutions do indeed make purification more challenging, and several substitutions impair ThDP binding. That said, however, and on the basis of these two examples, the author believes that none of these substitutions is fatal to the activity, further confirming that ThDP carries out the bulk of the catalysis. Interestingly, the solvent kinetic isotope effects appear to sense this as well, since no SKIE greater than 1.5 has been reported so far under Vmax conditions for any ThDP enzyme, suggesting that proton transfers are not fully rate limiting under any conditions and with any variant so far tested. Of course these acid-base groups are significant to provide optimal rates by, for example, aligning the substrate for addition to ThDP, aligning the reactive conformer of LThDP, assuring that the cryptic stereochemistry, where relevant, is correct, etc. But, with nearly a decade of reports on these enzymes providing a better understanding of structurefunction relationships, the overwhelming importance of the cofactor has gained more and more credence, making thiamin diphosphate virtually unique in the coenzyme field, considering the multiple intramolecular reactions implied by the reports. The evidence should now convince the biochemistry community of the catalytic importance of the 4⬘-aminopyrimidine ring in ThDP reactions, although this is clearly not yet reflected in most current texts and general reviews of enzyme mechanisms. Given that this enzyme-bound “V” ThDP conformation may affect the very first step in all ThDP reactions, and has not yet been mimicked in chemical models, one can safely conclude that modeling of ThDP enzymes, and studies of the enzymes themselves, both remain fertile fields for further research.

8.3

10

Intramolecular uphill proton transfer to initiate the reaction

The initial proton transfer to generate the ylide/C2-carbanion probably involves transfer from the thiazolium C2 position to the iminoThDP tautomer (either directly or via the intervention of some other protonic site). Models from the author’s laboratory and from Washabaugh and Jencks suggest that this proton transfer is uphill since the 4⬘-amino group has a pKa of 12–13 (on N1⬘ protonation) and the thiazolium C2H a value of ca. 17–19; in other words, the C2-carbanion is about 5–6 pKa units stronger as a base than the 4⬘-imino nitrogen. Any “solvent” effects of the type discussed in Section 8.2 would lower both pKas, albeit by perhaps somewhat different magnitudes. Assuming that the proton transfer rate is near diffusion controlled, the 5–6 pKa unit uphill barrier would reduce the rate constant by ca. 105–106, leaving ample speed so that the proton transfer for most ThDP enzymes would not be rate limiting, given that the kcat for such enzymes seldom exceeds 102 s⫺1 per subunit. In addition, as mentioned earlier, the pKas at the N1⬘ site of the 4⬘-aminopyrimidine are well balanced (that for the N1⬘-protonated ThDP is very near 5.0). 8.4

Acid-base catalysis

As of writing, the principal residues have been mapped out for YPDC 18,21,86 and PDHc-E1,105 BFD,95 TK,106 pyruvate decarboxylase from Zymomonas mobilis 107–109 and a mammalian

9

Acknowledgements

It is a pleasure to acknowledge both former and current students and postdoctorals, and valued colleagues and collaborators whose names appear in the references. Supported by NIHGM-50380, NIH-GM-62330, the Rutgers Busch Biomedical Fund, the Rutgers Board of Governors Fund and NSF-BIR94/ 13198. References

1 R. Kluger, Chem. Rev., 1987, 87, 863. 2 A. Schellenberger, Biochim. Biophys. Acta, 1998, 1385, 177. 3 F. Jordan, N. Nemeria, F. Guo, I. Baburina, Y. Gao, A. Kahyaoglu, H. Li, J. Wang, J. Yi, J. R. Guest and W. Furey, Biochim. Biophys. Acta, 1998, 1385, 287. 4 F. Jordan, FEBS Lett., 1999, 457, 298. 5 F. Dyda, W. Furey, S. Swaminathan, M. Sax, B. Farrenkopf and F. Jordan, Biochemistry, 1993, 32, 6165. 6 P. Arjunan, T. Umland, F. Dyda, S. Swaminathan, W. Furey, M. Sax, B. Farrenkopf, Y. Gao, D. Zhang and F. Jordan, J. Mol. Biol., 1996, 256, 590. 7 M. S. Hasson, A. Muscate, M. J. McLeish, L. S. Polovnikova, J. A. Gerlt, G. L. Kenyon, G. A. Petsko and D. Ringe, Biochemistry, 1998, 37, 9918. 8 B. Bertagnolli and L. P. Hager, J. Biol. Chem., 1991, 266, 10168. 9 Y. A. Muller and G. E. Schulz, Science, 1993, 259, 965. 10 L. Reed, Acc. Chem. Res., 1974, 7, 40. 11 E. Chabriere, M. H. Charon, A. Volbeda, L. Pieulle, E. C. Hatchikian and J. C. Fontecilla-Camps, Nat. Struct. Biol., 1999, 6, 182. 12 Y. Lindquist, G. Schneider, U. Ermler and M. Sundstrom, EMBO J., 1992, 11, 3273. 13 D. Chipman, Z. Barak and J. V. Schloss, Biochim. Biophys. Acta, 1998, 1385, 401. 14 C. F. Hawkins, A. Borges and R. N. Perham, FEBS Lett., 1989, 255, 77. 15 W. Shin, J. Pletcher, G. Blank and M. Sax, J. Am. Chem. Soc., 1977, 99, 1396. 16 F. Jordan, J. Am. Chem. Soc., 1974, 96, 3623.

Nat. Prod. Rep., 2003, 20, 184–201

199

17 F. Jordan, J. Am. Chem. Soc., 1976, 98, 808. 18 F. Guo, D. Zhang, A. Kahyaoglou, R. S. Farid and F. Jordan, Biochemistry, 1998, 37, 13379. 19 J. Crosby, R. Stone and G. E. Lienhard, J. Am. Chem. Soc., 1970, 92, 2891. 20 D. Dobritzsch, S. Konig, G. Schneider and G. Lu, J. Biol. Chem., 1998, 273, 20196. 21 M. Liu, E. A. Sergienko, F. Guo, J. Wang, K. Tittmann, G. Hübner, W. Furey and F. Jordan, Biochemistry, 2001, 40, 7355. 22 M. Liu, PhD Dissertation, Rutgers University Graduate Faculty at Newark, NJ, 2002. 23 Z. Zhang, PhD Dissertation, Rutgers University Graduate Faculty at Newark, NJ, 2002. 24 E. A. Sergienko, J. Wang, L. Polovnikova, M. S. Hasson, M. J. McLeish, G. L. Kenyon and F. Jordan, Biochemistry, 2000, 39, 13862. 25 P. Arjunan, N. Nemeria, A. Brunskill, K. Chandrasekhar, M. Sax, Y. Yan, F. Jordan, J. R. Guest and W. Furey, Biochemistry, 2002, 41, 5213. 26 F. Jordan and Y. H. Mariam, J. Am. Chem. Soc., 1978, 100, 2534. 27 F. Jordan, G. Chen, S. Nishikawa and B. Sundoro-Wu, Ann. N.Y. Acad. Sci., 1982, 378, 14. 28 F. Jordan, J. Org. Chem., 1982, 47, 2748. 29 F. Jordan, Z. Zhang and E. Sergienko, Bioorg. Chem., 2002, 30, 188. 30 M. V. Kovina, I. A. Bykova, O. N. Solovjeva, L. E. Meshalkina and G. A. Kochetov, Biochem. Biophys. Res. Commun., 2002, 294, 155. 31 F. Jordan, N. Nemeria, Y. Yan and S. Zhang, under review. 32 M. W. Washabaugh and W. P. Jencks, Biochemistry, 1988, 27, 5044. 33 D. Kern, G. Kern, H. Neef, K. Tittmann, M. Killenberg-Jabs, C. Wikner, G. Schneider and G. Hübner, Science, 1997, 275, 67. 34 A. J. Arduengo, J. R. Goerlich and W. J. Marshall, Liebigs Ann./ Recueil, 1997, 365–374. 35 (a) A. Brown, PhD Dissertation, Rutgers University Graduate Faculty at Newark, NJ, 1999; (b) A. Brown, I. Estevez-Ikemoto and F. Jordan, in preparation. 36 R. Breslow, J. Am. Chem. Soc., 1957, 79, 1762. 37 Y-T. Chen and F. Jordan, J. Org. Chem., 1991, 56, 5029. 38 Y-T. Chen, G. L. Barletta, K. Haghjoo, J. T. Cheng and F. Jordan, J. Org. Chem., 1994, 59, 7714. 39 R. Kluger, J. Chin and T. Smyth, J. Am. Chem. Soc., 1981, 103, 384. 40 R. Kluger and T. Smyth, J. Am. Chem. Soc., 1981, 103, 1214. 41 K. Tittmann, R. Golbik, K. Uhlemann, L. Khailova, G. Schneider, M. Patel, F. Jordan, D. M. Chipman, R. G. Duggleby and G. Hübner, under review 2002. 42 H. Li, W. Furey and F. Jordan, Biochemistry, 1999, 38, 9992. 43 P. Arjunan, A. Brunskill, S. Zhang, Y. Yan, N. Nemeria, F. Jordan and W. Furey, in preparation. 44 F. Jordan, Z. H. Kudzin and C. B. Rios, J. Am. Chem. Soc., 1987, 109, 4415. 45 F. Jordan, in The Chemistry of the Enamines, ed. Z. Rappoport, John Wiley and Sons, Chittenden, UK, 1994, pp. 1253–1302. 46 F. G. Bordwell, A. V. Satish, C. B. Rios, A. Chung and F. Jordan, J. Am. Chem. Soc., 1990, 112, 792. 47 G. Barletta, W. P. Huskey and F. Jordan, J. Am. Chem. Soc., 1992, 114, 7607. 48 G. Barletta, W. P. Huskey and F. Jordan, J. Am. Chem. Soc., 1997, 119, 2356. 49 Y. Zou, PhD Dissertation, Rutgers University Graduate Faculty at Newark, NJ, 1999. 50 F. Jordan, H. Li and A. Brown, Biochemistry, 1999, 38, 369. 51 D. J. Kuo and F. Jordan, J. Biol. Chem., 1983, 258, 134l7. 52 S. Menon-Rudolph, S. Nishikawa, X. Zeng and F. Jordan, J. Am. Chem. Soc., 1992, 114, 10110. 53 E. Fiedler, S. Thorell, T. Sandalova, R. Golbik, S. Konig and G. Schneider, Proc. Natl. Acad. Sci. USA, 2002, 99, 591. 54 G. Barletta, A. C. Chung, C. B. Rios, F. Jordan and J. M. Schlegel, J. Am. Chem. Soc., 1990, 112, 8144. 55 I. Nakanishi, S. Itoh, T. Suenobu and S. Fukuzumi, Chem. Commun., 1997, 1927. 56 S. Ragsdale, Chem. Rev., 2003, 103, in the press. 57 E. Chabriere, X. Vernede, B. Guigliarelli, M. H. Charon, E. C. Hatchikian and J. C. Fontecilla-Camps, Science, 2001, 294, 2559. 58 C. C. Chiu, K. Pan and F. Jordan, J. Am. Chem. Soc., 1995, 117, 7027. 59 P. A. Frey, D. S. Flournoy, K. Gruys and Y. S. Yang, Ann. N.Y. Acad. Sci., 1989, 373, 21.

200

Nat. Prod. Rep., 2003, 20, 184–201

60 W. H. Rastetter, J. Adams, J. W. Frost, L. J. Nummy, J. E. Frommer and K. B. Roberts, J. Am. Chem. Soc., 1979, 101, 2752. 61 C. C. Chiu, A. Chung, G. Barletta and F. Jordan, J. Am. Chem. Soc., 1996, 118, 11026. 62 K. Pan and F. Jordan, Biochemistry, 1998, 37, 1357. 63 A. Aevarsson, A. K. Seger, S. Turley, J. Sokatch and W. G. J. Hol, Nat. Struct. Biol., 1999, 6, 785. 64 A. Aevarsson, J. L. Chuang, R. M. Wynn, S. Turley, D. T. Chuang and W. G. J. Hol, Structure, 2000, 8, 277. 65 J. Ullrich and A. Mannschreck, Eur. J. Biochem., 1967, 1, 110. 66 L. E. Meshalkina, C. Wikner, U. Nilsson, Y. Lindqvist and G. Schneider, in Biochemistry and Physiology of Thiamin Diphosphate Enzymes, ed. H. Bisswanger and A. Schellenberger, A.u.C. Intemann, Wissentschaftlicher Verlag, Prien, Germany, 1996, pp. 532–542. 67 N. Nemeria, Y. Yan, Z. Zhang, A. M. Brown, P. Arjunan, W. Furey, J. R. Guest and F. Jordan, J. Biol. Chem., 2001, 276, 45969. 68 D. H. G. Crout, in Biosynthesis of Branched Chain Amino Acids, ed. Z. Barak, D. M. Chipman and J. V. Schloss, Verlag Chemie, Weinheim, 1990, pp. 199–242. 69 G. C. Chen and F. Jordan, Biochemistry, 1984, 23, 3576. 70 E. A. Sergienko and F. Jordan, Biochemistry, 2001, 40, 7369. 71 M. Pohl, in Advances in Biochemical Engineering/Biotechnology, ed. T. Scheper, Verlag Springer, Heidelberg, 1997, vol. 58, pp. 15–43. 72 U. Schörken and G. A. Sprenger, Biochim. Biophys. Acta, 1998, 1385, 229. 73 H. Iding, P. Siegert, K. Mesch and M. Pohl, Biochim. Biophys. Acta, 1998, 1385, 307. 74 G. A. Sprenger and M. Pohl, J. Mol. Catal., B, 1999, 6, 15. 75 M. Muller and G. A. Sprenger, in Thiamine: Catalytic Mechanisms and Role in Normal and Disease States, ed. F. Jordan and M. Patel, Marcel Dekker, New York, 2003. 76 X. Zeng, B. Farrenkopf, S. Hohmann and F. Jordan, Biochemistry, 1993, 32, 2704. 77 I. Baburina, Y. Gao, Z. Hu, S. Hohmann, W. Furey and F. Jordan, Biochemistry, 1994, 33, 5630. 78 I. Baburina, D. J. Moore, A. Volkov, A. Kahyaoglu, F. Jordan and R. Mendselsohn, Biochemistry, 1996, 35, 10249. 79 I. Baburina, H. Li, W. Furey, B. Bennion and F. Jordan, Biochemistry, 1998, 37, 1235. 80 I. Baburina, G. Dikdan, F. Guo, G. Tous, B. Root and F. Jordan, Biochemistry, 1998, 37, 1245. 81 H. Li, W. Furey and F. Jordan, Biochemistry, 1999, 38, 9992. 82 H. Li and F. Jordan, Biochemistry, 1999, 38, 10004. 83 J. Wang, R. Golbik, B. Seliger, M. Spinka, K. Tittmann, G. Hübner and F. Jordan, Biochemistry, 2001, 40, 1755. 84 W. Wei, M. Liu and F. Jordan, Biochemistry, 2002, 41, 451. 85 E. A. Sergienko and F. Jordan, Biochemistry, 2002, 41, 3952. 86 Y. Gao, PhD Dissertation, Rutgers University Graduate Faculty at Newark, NJ, 2001. 87 Y. Yan, N. Nemeria and F. Jordan, under review 2003. 88 E. A. Sergienko and F. Jordan, Biochemistry, 2001, 40, 7382. 89 N. Nemeria, P. Arjunan, A. Brunskill, F. Sheibani, W. Wei, Y. Yan, S. Zhang, F. Jordan and W. Furey, Biochemistry, 2002, 41, in the press. 90 F. Jordan, D. J. Kuo and E. U. Monse, J. Am. Chem. Soc., 1978, 100, 2872. 91 F. Alvarez, J. Ermer, G. Hübner, A. Schellenberger and R. Schowen, J. Am. Chem. Soc., 1995, 117, 1678. 92 W. P. Huskey, L. Chan and Y. Yuan, submitted for publication. 93 S. Sun, R. G. Duggleby and R. L. Schowen, J. Am. Chem. Soc., 1995, 117, 7317. 94 P. M. Weiss, G. A. Garcia, G. L. Kenyon, W. W. Cleland and P. F. Cook, Biochemistry, 1988, 27, 2197. 95 L. S. Polovnikova, M. J. McLeish, J. T. Burgner, N. L. Anderson, E. A. Sergienko, F. Jordan, G. L. Kenyon and M. S. Hasson, under review. 96 N. Khaleeli, R. Li and C. A. Townsend, J. Am. Chem. Soc., 1999, 121, 9223. 97 F. B. Opperman, B. Schmidt and A. Steinbüchel, J. Bacteriol., 1991, 73, 757. 98 V. Foulon, V. D. Antonenkov, K. Croes, E. Waelkens, G. P. Mannaerts, P. P. van Veldhoven and M. Casteels, Proc. Natl. Acad. Sci. USA, 1999, 96, 10039. 99 J. Koga, Biochim Biophys Acta, 1995, 1249, 1. 100 A. S. Demir, M. Pohl, E. Janzen and M. Muller, J. Chem. Soc., Perkin Trans. 1, 2001, 633. 101 A. S. Demir, O. Sesenoglu, E. Eren, B. Hosrik, M. Pohl, E. Janzen,

102 103 104 105

D. Kolter, R. Feldmann, P. Dunkelmann and M. Müller, Adv. Synth. Catal., 2002, 344, 96. H. C. Dunathan, Adv. Enzymol., 1971, 35, 79. A. Turano, W. Furey, J. Pletcher, M. Sax, D. Pike and R. Kluger, J. Am. Chem. Soc., 1982, 104, 3089. J. H. Wittorf and C. J. Gubler, Eur. J. Biochem., 1970, 14, 53. F. Jordan, N. Nemeria, S. Zhang, W. Wei, Y. Yan, Y. Park, P. Arjunan, A. Brunskill and W. Furey, in Thiamine: Catalytic Mechanisms and Role in Normal and Disease States, ed. F. Jordan and M. Patel, Marcel Dekker, New York, 2003.

106 C. Wikner, U. Nilsson, L. Meshalkina, C. Udekwu, Y. Lindqvist and G. Schneider, Biochemistry, 1997, 36, 15643. 107 G. Schenk, F. J. Leeper, R. England, P. F. Nixon and R. G. Duggleby, Eur. J. Biochem., 1997, 248, 63. 108 J. M. Candy and R. G. Duggleby, Biochim. Biophys. Acta, 1998, 1385, 323. 109 A. K. Chang, P. F. Nixon and R. G. Duggleby, Biochem. J., 1999, 339, 255. 110 R. M. Wynn, R. Ho, J. L. Chuang and D. T. Chuang, J. Biol. Chem., 2001, 276, 4168.

Nat. Prod. Rep., 2003, 20, 184–201

201