Chapter 40

Biosynthesis of Riboflavin

ADELBERT BACHER, SABINE EBERHARDT, and GERALD RICHTER

INTRODUCTION

Riboflavin is biosynthesized by plants and many microorganisms. Various bacteria and fungi can overproduce the vitamin to a considerable extent (29). The potential of these flavinogenic organisms for the technical production of the vitamin was recognized early, and the first biosynthetic studies were performed in an effort to improve fermentation yields. Thus, early fermentation studies with the flavinogenic ascomycete Eremothecium ashbyii suggested a biosynthetic connection between purines and riboflavin (51). The terminal intermediate, 6,7-dimethyl-8-ribityllumazine (compound 8, Fig. 2), was also found early in the course of studies with the flavinogenic E. ashbyii (53). A tentative pathway was subsequently proposed on the basis of studies with Saccharomyces cerevisiae (9, 10, 60). The biosynthesis of riboflavin has been reviewed repeatedly, and the reader is directed to these articles for an in-depth discussion of the earlier work (2, 5, 24, 25, 65, 67).

Fig. 2Biosynthesis of riboflavin and of flavocoenzymes. It should be noted that the intermediate, 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione (compund 5), is generated de novo (see Fig. 1) and by recycling via riboflavin synthase (see Fig. 3).

Source:

Organisms most frequently used in biosynthetic studies were the flavinogenic ascomycetes Ashbya gossypii and E. ashbyii, various yeasts such as Saccharomyces cerevisiae and the flavinogenic Pichia (Candida) guilliermondii, and Bacillus subtilis. Escherichia coli was used only occasionally in the earlier studies. Riboflavin auxotrophs of E. coli were isolated only relatively recently (17, 18, 68, 75). They require very high concentrations of riboflavin (about 200 mg/liter) for growth, and this was probably the main reason for their late discovery. The approximate map positions of three rib loci had been reported on the E. coli chromosome at 28, 66, and 56–59 (later changed to 40) min on the basis of mating and transduction experiments (15, 18, 78), but the phenotypes of the respective mutants had not been unambiguously assigned. Recently, the sequences and map positions of all six genes involved in the biosynthesis of riboflavin and of flavocoenzymes in E. coli have been determined, and their functions have been established unequivocally (Table 1) (70, 71; K. Kitasuji, S. Ishino, S. Teshiba, and M. Arimoto, European Patent 0542240 A2 92119308.2, May, 1993; A. Bacher, S. Eberhardt, M. Fischer, S. Mörtl, and G. Richter, unpublished data).

Table 1Genes and enzymes involved in biosynthesis of riboflavin and flavocoenzymes

Studies in Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) are virtually absent. The presence of six unlinked rib genes has been reported, but their functions remain unknown (83).

The biosynthesis of riboflavin is summarized in Fig. 1 and 2. The elucidation of this pathway proceeded over a period of more than three decades and was completed only relatively recently. Briefly, C-8 of GTP (compound 1, Fig. 1) is released as formate by the action of GTP cyclohydrolase II, and the opening of the imidazole ring is accompanied by the release of pyrophosphate, yielding the product 2,5-diamino-6-ribosylamino-4(3H)-pyrimidinone 5'-phosphate (compound 2) (32). This reaction is followed by hydrolytic deamination of the pyrimidine ring and by reduction of the ribosyl side chain. In bacteria, the deamination of the ring precedes the reduction step yielding the intermediate 3 (Fig. 1), which is subsequently reduced (26). In fungi, the sequence of reactions is reversed, and the pathway proceeds via the intermediate 2,5-diamino-6-ribitylamino-4(3H)-pyrimidinedione 5'-phosphate (56, 60). Both pathways thus lead to the formation of 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione 5'-phosphate (compound 4, Fig. 1), albeit via different intermediates. After hydrolysis of the phosphoric ester, compound 4, the condensation of the product, compound 5, with l-3,4-dihydroxy-2-butanone 4-phosphate (compound 7) yields 6,7-dimethyl-8-ribityllumazine (compound 8) (79). Dismutation of the lumazine yields riboflavin (compound 9) and the pyrimidine compound 5, which can be recycled in the pathway (65, 79). Riboflavin can be converted to riboflavin 5'-phosphate (flavin mononucleotide [FMN]) and flavin adenine dinucleotide (FAD) by a bifunctional riboflavin kinase/FAD synthetase in E. coli (Kitatsuji et al., European Patent, 1993). These reactions are considered in more detail below.

Fig. 1Biosynthesis of the riboflavin precursor, 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione (compound 5). The E. coli genes specifying each step are indicated. Their positions to the nearest 0.1 min are derived from the physical map (chapter 109).

Source:

GTP CYCLOHYDROLASE II

Extensive work performed in the 1950s and 1960s had established the close connection between the biosynthetic pathways of purines and of riboflavin (for review see references 2, 5, 24, 25, 65, 67, and 87). Experiments with a purine mutant of E. coli established that a purine derivative serves as the direct precursor of the pyrimidine ring of the vitamin (36). Work with a purine mutant of Aerobacter aerogenes (12) and with Corynebacterium sp. (19) showed that the committed precursor is a compound at the biosynthetic level of guanine. The origin of the ribityl side chain of riboflavin from the ribosyl side chain of the purine precursor was established by work with a triple purine mutant of S. typhimurium (13).

These preliminary studies were confirmed by the isolation of an enzyme from E. coli which catalyzes the release of formate from the imidazole ring of GTP and pyrophosphate, apparently in a simultaneous reaction (32, 33). The enzyme was designated GTP cyclohydrolase II (EC 3.5.4.25) in order to distinguish it from GTP cyclohydrolase I (EC 3.5.4.16), which catalyzes the first committed step in the biosynthesis of tetrahydrofolate (for review see chapter 41, this volume). GTP cyclohydrolase II requires magnesium as cofactor. The enzyme is highly specific for GTP as substrate. Pyrophosphate acts as an inhibitor (32).

The gene for GTP cyclohydrolase II in E. coli has been cloned by a marker rescue strategy (70) and corresponds to the ribA locus, which had been mapped previously at 28 min on the gene linkage map by transduction analysis (18, 78). The enzyme is not homologous to GTP cyclohydrolase I.

PYRIMIDINE DEAMINASE/PYRIMIDINE REDUCTASE

Early work with riboflavin mutants of Saccharomyces cerevisiae and Aspergillus nidulans had indicated the formation of the pyrimidinedione, compound 5 (Fig. 1), by the reduction of the ribityl side chain of compound 2, followed by the deamination of the pyrimidine ring and dephosphorylation of the side chain (2, 5).

A somewhat different sequence of reactions leads to the formation of compound 5 from compound 2 in E. coli as well as B. subtilis. Burrows and Brown (26) obtained an enzyme from cell extracts of E. coli that catalyzes the deamination of compound 2, yielding 5-amino-6-ribosylamino-2,4(1H,3H)-pyrimidinedione 5'-phosphate (compound 3, Fig. 1). The protein was purified about 200-fold, and a mass of 80 kDa was estimated on the basis of gel filtration experiments.

Burrows and Brown (26) also purified an E. coli enzyme catalyzing the reduction of compound 3 to yield compound 4. The reductase was purified about 200-fold. It required NADPH as cofactor, and a molecular mass of 37 kDa was estimated from gel filtration experiments. The dephosphorylated forms of compounds 2 and 3 could not be used as substrates by the respective enzymes.

Recent experiments showed that the deaminase and the reductase are specified by a single gene, ribD, in E. coli. A single gene, ribG, coding for both enzyme activities, has also been found in B. subtilis. Since this was unexpected in light of the enzyme studies described above, the unpublished genetic data will be discussed in some detail.

The sequence of the entire rib operon of B. subtilis has been established independently by two different groups (54; J. B. Perkins, J. G. Pero, and A. Sloma, European Patent Application EP 405370 A1 910102, January, 1991). The operon contains five open reading frames designated ribG, ribB, ribA, ribH, and ribT. Perkins and Pero (64) noted that the 5' part of the ribG gene shows homology to deoxycytidylate deaminase and therefore suggested that this gene could specify the enzyme responsible for the deamination of compound 2.

More recently, it became apparent that the 3' end of the ribG gene is homologous to the RIB7 gene of Saccharomyces cerevisiae (I. Gerstenschläger, personal communication), which is supposed to catalyze the reduction of the intermediate compound 2, thus suggesting that the ribG gene of B. subtilis could code for a bifunctional deaminase/reductase. This hypothesis was confirmed by hyperexpression of ribG in a recombinant E. coli strain (M. Fischer, G. Richter, and A. Bacher, unpublished data). The recombinant protein catalyzes both reaction steps, i.e., deamination of compound 2 and reduction of compound 3. Deletion analysis showed that the N-terminal part and C-terminal part correspond to domains with deaminase and reductase activity, respectively.

An open reading frame similar to the ribG gene of B. subtilis is located in close proximity to the nusB gene of E. coli at 442 kb on the chromosome (77). This gene (ribD) has now been hyperexpressed in a recombinant E. coli strain (S. Eberhardt, G. Richter, C. Krieger, and A. Bacher, unpublished data) to yield a bifunctional enzyme with deaminase and reductase activity.

THE ELUSIVE PYRIMIDINE PHOSPHATASE

The initial steps of riboflavin biosynthesis lead to the formation of the phosphoric acid ester, compound 4 (Fig. 1). However, it has been shown that compound 4 is at best a very poor substrate for lumazine synthase (55). Moreover, the 5'-phosphate of the lumazine, compound 8, does not act as substrate for riboflavin synthase (34). It follows that the intermediate 4 should be dephosphorylated enzymatically prior to its conversion to 6,7-dimethyl-8-ribityllumazine (compound 8). However, no evidence for the existence of a specific phosphatase has been obtained. Thus, it appears possible that the dephosphorylation of compound 4 may be catalyzed by a phosphatase of broad substrate specificity.

3,4-DIHYDROXY-2-BUTANONE-4-PHOSPHATE SYNTHASE

6,7-Dimethyl-8-ribityllumazine (compound 8, Fig. 2), the terminal intermediate in the biosynthesis of riboflavin, has been discovered in culture filtrates of E. ashbyii (2, 5, 24, 25, 53, 65, 67, 87). The formation of the lumazine from the pyrimidine compound 5 appeared to require a 4-carbon precursor for the formation of the pyrazine ring. The nature of this 4-carbon compound remained elusive for more than three decades. A considerable number of experimental reports have been published, and a variety of potential precursors have been proposed. The older literature on this controversial subject has been reviewed (2, 5) and will not be discussed. The suggested formation of compound 8 by dismutation of compound 5 in the absence of a second substrate by an enzyme of E. coli is no longer tenable (35).

In vivo isotope incorporation studies with A. gossypii and B. subtilis suggested that the carbon atoms of the elusive 4-carbon precursor could be formed from the carbon atoms 1, 2, and 3 and, surprisingly, carbon atom 5 of a pentose precursor (7, 8, 47). In agreement with these experiments, Shavlovsky and his coauthors reported the enzymatic formation of riboflavin in cell extracts of C. guilliermondii from the pyrimidine compound 4 in the presence of various carbohydrate phosphates (49, 50). Subsequent studies showed that the in vitro isotope incorporation pattern using C. guilliermondii cell extract was in full agreement with the in vivo experiments (57).

The formation of the 4-carbon precursor from a pentose phosphate was then confirmed by the isolation of an enzyme from C. guilliermondii which catalyzes the formation of a novel carbohydrate, 3,4-dihydroxy-2-butanone 4-phosphate (compound 7), from ribulose 5-phosphate (compound 6) (80). The enzyme-catalyzed reaction involves the release of C-4 of ribulose phosphate as formate. Carbon 5 of the substrate is reattached to C-3 by an intramolecular rearrangement yielding the product compound 7. The enzyme requires Mg²⁺ for activity, but no other cofactors. The reaction mechanism has been studied in some detail (81). All data are consistent with a sequence of tautomerization reactions.

The gene coding for 3,4-dihydroxy-2-butanone-4-phosphate synthase in E. coli has been cloned by marker rescue (71) and is identical with the ribB locus previously mapped at 66 min (18, 78). Initially, the gene product had been incorrectly addressed as a heat shock protein, and the cognate gene had been designated htrP by Raina et al. (69). Probably these authors were dealing with a temperature-sensitive mutation of the 3,4-dihydroxy-2-butanone-4-phosphate synthase. The gene had also been sequenced and mapped independently by Yang and Depew, but its function had not been determined (85, 86). The ribB gene codes for a 23.3-kDa peptide of 217 amino acids. A molecular mass of 47 kDa was obtained by analytical ultracentrifugation, thus suggesting a dimer structure (G. Richter and A. Bacher, unpublished data).

Sequence homology suggests that the ribA gene of the B. subtilis rib operon codes for a bifunctional protein with GTP cyclohydrolase II and 3,4-dihydroxy-2-butanone-4-phosphate synthase activity. Thus, the initial steps of both converging branches of the riboflavin pathway appear to be catalyzed by a single protein in B. subtilis. This hypothesis is confirmed by the observation that ribA of B. subtilis can complement both ribA and ribB mutations of E. coli (H. Ritz and A. Bacher, unpublished data).

6,7-DIMETHYL-8-RIBITYLLUMAZINE SYNTHASE

6,7-Dimethyl-8-ribityllumazine synthase catalyzes the condensation of 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione with 3,4-dihydroxy-2-butanone 4-phosphate (55). More recently, the reaction mechanism has been studied in some detail by Kis et al. (39).

In B. subtilis, the enzyme forms a complex with the terminal enzyme of the pathway, riboflavin synthase. This enzyme complex has been studied in considerable detail (4, 6, 11, 41, 42, 43). The 1-MDa protein consists of 60 lumazine synthase subunits (designated β subunits) which form an icosahedral capsid enclosing three riboflavin synthase subunits (designated α subunits). An X-ray structure of the icosahedral capsid has been obtained (42, 43). The containment of the riboflavin synthase module inside the lumazine synthase capsid allows for efficient substrate channeling between the active sites of the riboflavin synthase and the lumazine synthase, resulting in an unexpectedly high overall velocity at low substrate concentrations (38a).

An open reading frame coding for a protein homologous to the lumazine synthase of B. subtilis has been found in close proximity to the nusB gene at 443 kb of the E. coli chromosome (77). This gene has now been hyperexpressed in E. coli, and the recombinant protein showed the expected lumazine synthase activity (S. Mörtl, M. Fischer, and A. Bacher, unpublished data). However, the protein has not yet been studied in detail. We propose to designate the gene coding for lumazine synthase as ribE. Mutants with defects of the ribE gene have not been reported hitherto in E. coli. The genes ribD, ribE, and nusB, as well as an open reading frame of unknown function, may be under the control of a common promoter (77).

RIBOFLAVIN SYNTHASE

Riboflavin synthase (EC 2.5.1.9) catalyzes the formation of riboflavin from 6,7-dimethyl-8-ribityllumazine. Plaut and his coworkers showed that the enzyme catalyzes an unusual dismutation involving the transfer of a 4-carbon unit between two molecules of the substrate compound 8 (Fig. 2). This implicates the formation of one molecule of riboflavin (compound 9) and one molecule of the pyrimidine compound 5 from two molecules of compound 8 (82). The elegant work by Plaut and his coworkers on the mechanism of riboflavin synthase has been reviewed repeatedly (2, 65, 67).

The gene coding for riboflavin synthase in E. coli has been cloned recently by marker rescue (30). It complements the ribC mutation, which had been mapped previously at 40 min (15). The peptide sequence of riboflavin synthase shows marked homology between the N-terminal and the C-terminal half. An alignment of the first 97 amino acids of the E. coli enzyme against the rest yields 25 identical residues and 32 conservative replacements. This may indicate that the protomer folds into two structurally similar domains (30, 74). In light of the dismutation mechanism, which requires the binding of two identical substrate molecules in close proximity at the active site of the enzyme (62, 66), it appears possible that the two homologous domains each provide the binding site for one respective substrate molecule at the active site of the protein.

The second product of riboflavin synthase, 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione (compound 5, Fig. 1), is also a substrate of lumazine synthase in the biosynthetic pathway (see above). Thus, this pyrimidine can be formed both de novo from GTP and by the action of riboflavin synthase. The overall stoichiometry implicates the formation of one molecule of riboflavin from one molecule of the pyrimidine, compound 5, and two molecules of the carbohydrate, compound 7. Ultimately, all carbon atoms of the xylene ring are derived from two molecules of the carbohydrate. As a consequence, every second molecule of compound 5 which is formed de novo has to be processed twice by lumazine synthase and riboflavin synthase (Fig. 3).

Fig. 3Intermediate recycling in the terminal steps of riboflavin biosynthesis.

Source:

STRUCTURAL HOMOLOGS OF RIBOFLAVIN SYNTHASE INVOLVED IN BIOLUMINESCENCE

Proteins with close sequence homology to riboflavin synthase have been found in various luminescent bacteria, where they serve as optical transducers in bioluminescence (46, 59). Lumazine protein is a monomeric, highly fluorescent protein binding the riboflavin precursor compound 8 with high affinity (K_d = 50 nM) which has been found in various Photobacterium species (58). Yellow fluorescent protein from Vibrio fischeri (specified by luxY) is also homologous to riboflavin synthase and binds FMN as fluorophore (16).

Lumazine protein and yellow fluorescent protein have no known enzymatic function. However, they shift the wavelength of bioluminescence emission and increase the quantum yield due to their interaction with bacterial luciferase (46).

BIOSYNTHESIS OF FLAVOCOENZYMES

In order to perform its metabolic role, riboflavin must be transformed to the coenzyme forms, FMN (riboflavin 5'-phosphate) and FAD. The phosphorylation of the 5' hydroxy group of riboflavin yielding FMN is catalyzed by riboflavin kinase (EC 2.7.1.26). The enzyme requires ATP as substrate. FMN can be further converted to FAD by FAD synthetase (EC 2.7.7.2). ATP serves as substrate, and the reaction implicates the release of inorganic pyrophosphate. In most organisms studied, riboflavin kinase and FAD synthetase are separate proteins (for review see reference 3).

Recent evidence indicates that a bifunctional flavokinase/FAD synthetase is present in E. coli. The structural gene for this protein is located between the genes rpsT and ileS, and its sequence has been determined by Kamio et al. (37). The gene could be overexpressed, and the cognate protein of unknown function was designated protein X. Kitatsuji et al. (European Patent, 1993) have now found that this protein is the bifunctional flavokinase/FAD synthetase. Site-directed mutagenesis experiments suggest that the protein consists of two folding domains. The riboflavin kinase activity is associated with the C-terminal part, and the FAD synthetase activity is associated with the N-terminal part. Recombinant strains overexpressing this protein can be used to prepare the coenzymes FMN and FAD. We propose the designation ribF for the gene specifying the bifunctional flavokinase/FAD synthase.

A bifunctional enzyme with riboflavin kinase and FAD synthetase activity had been purified earlier from Brevibacterium ammoniagenes (52; R. Blake, European Patent Application EP146,871 Cl. C12N9/12A, Chem. Abstr. 103, 84146h, July, 1985). It has also been tentatively suggested that a bifunctional flavokinase/FAD synthetase could be present in B. subtilis. This enzyme accepts dihydroriboflavin as substrate (38).

TRANSPORT

As mentioned above, riboflavin auxotrophs of E. coli require very high concentrations of riboflavin (about 200 mg/liter) for growth (17). This suggests that E. coli and probably other members of the Enterobacteriaceae are devoid of a transport system for the vitamin.

Other bacteria such as B. subtilis and Lactobacillus casei can utilize riboflavin efficiently at low concentrations. An uptake system has been partially characterized in B. subtilis (38).

TRANSCRIPTION AND REGULATION

In E. coli, the genes ribD and ribE, the antitermination factor nusB, and an open reading frame of unknown function appear to be expressed under the control of a single promoter (77). A systematic analysis of the regulation of riboflavin biosynthesis in E. coli has not been performed. Metabolic studies using riboflavin auxotrophs gave no evidence for regulation of any type (75). It is therefore conceivable that the expression of the riboflavin biosynthetic enzymes in E. coli is constitutive. Since riboflavin cannot be absorbed from the environment, owing to the absence of a transport system, it may be appropriate for the microorganism to produce the compound continuously at a low rate.

In contrast to the situation in E. coli, the rib operon of B. subtilis can be regulated over a wide range (14, 21, 22, 23, 27). The involvement of a transcription antitermination mechanism has been proposed (64). A regulatory protein is encoded by the ribC locus at 145° on the chromosome (40).

In several luminescent bacteria, putative riboflavin biosynthesis genes have been found as part of the luciferase operon (44, 45, 76). Luciferase utilizes FMN as a cofactor, and the activation of the luciferase operon may involve an increased requirement for this cofactor. Thus, the joint control of riboflavin biosynthesis and bioluminescence may be advantageous.

BIOTECHNOLOGY OF RIBOFLAVIN

Riboflavin is commercially prepared in bulk quantities by chemical synthesis and by fermentation for use in human and animal nutrition. Overproducers of the vitamin occur naturally in a variety of microbial taxons such as ascomycetes, yeasts, and bacteria. The early history of riboflavin production by fermentation has been reviewed (29). More recent studies on the production of riboflavin have focused on B. subtilis (A. Y. Kukanova, V. G. Zhdanov, A. I. Stepanov, and V. A. Panova, USSR SU 908092 A1 841230, December, 1984; Perkins et al., European Patent Application, 1991; A. I. Stepanov, V. G. Zhdanov, A. Y. Kukanova, M. Y. Khaikinson, P. M. Rabinovich, J. Iomantas, and Z. M. Galushkina, Fr. Demande FR 2546907 A1 841207, December, 1984), Candida famata (R. B. Bailey, G. W. Lauderdale, D. L. Heefner, C. A. Weaver, M. J. Yarus, L. A. Burdzinski, and A. Boyts, PCT Int. Appl. WO 92/01060 A1 920123, January, 1992; D. L. Heefner, C. A. Weaver, M. J. Yarus, L. A. Burdzinski, D. C. Gyure, and E. W. Foster, PCT Int. Appl. WO88/09822 A1 881215, December, 1988), A. gossypii (R. Kurth, Ger. Offen. DE 4037441 A1 920527, May, 1992), and E. ashbyii (K. Tachibana, M. Takahashi, Y. Matoba, S, Yamauchi, and T. Toda, Jpn. Kokai Tokkyo Koho HP 05076386 A2 930330 Heisei, March, 1993). A fermentation yield of 15 g/liter has been reported for a genetically engineered strain of B. subtilis (Perkins et al., European Patent Application, January, 1991). Yields of 21 g/liter and productivities of 0.17 g liter^–1 h^–1 have been reported for Candida famata (Bailey et al., PCT Int. Appl., 1992; Heefner et al., PCT Int. Appl., 1988). Yields of 3 to 5 g/liter have been described for Saccharomyces cerevisiae (A. Matsuyama, T. Nikaido, S. Kageyama, and K. Kawai, Jpn. Kokai Tokkyo Koho JP 63112996 A2 880518 Showa, May, 1988[a]; A. Matsuyama, T. Nikaido, S. Kageyama, and K. Kawai, Jpn. Kokai Tokkyo Koho JP 63112997 A2 880518 Showa, May, 1988[b]; S. Takao, T. Nikaido, A. Matsuyama, S. Kageyama, and K. Kawai, Jpn. Kokai Tokkyo Koho JP 63098399 A2 880428 Showa, April, 1988).

ENZYMES OF RIBOFLAVIN BIOSYNTHESIS AS POTENTIAL ANTIMICROBIAL TARGETS

Various members of the Enterobacteriaceae are unable to absorb extracellular riboflavin. Thus, riboflavin auxotrophs of E. coli require at least concentrations of 200 mg/liter for growth (17). Very high concentrations of riboflavin are also required for the growth of Saccharomyces cerevisiae and C. guilliermondii (48, 61), but studies with pathogenic Candida strains are not available. Since riboflavin is absolutely required for growth and cell division in these microorganisms, inhibitors of riboflavin biosynthesis could potentially serve as antimicrobial agents for the therapy of enterobacterial and yeast infections. Several inhibitors for the last enzyme in the pathway, riboflavin synthase, have been prepared (1, 2, 84). These compounds had no antimicrobial activity in vivo, probably owing to insufficient uptake into the microbial cells.

BIOMIMETIC FORMATION OF RIBOFLAVIN

The terminal steps in the riboflavin pathway can be mimicked by chemical reactions in the absence of enzyme catalysis. Thus, 6,7-dimethyl-8-ribityllumazine has been obtained by boiling an aqueous solution containing 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione and ribulose 1,5-bisphosphate (28, 31; C. J. Strupp, thesis, ETH Zürich, Zurich, Switzerland, 1992). More recently, the lumazine compound 8 (Fig. 2) was obtained by spontaneous, nonenzymatic reaction between the biosynthetic intermedates 5 and 7 under very mild conditions (K. Kugelbrey, K. Kis, and A. Bacher, unpublished data).

The lumazine, compound 8, can be converted to riboflavin by boiling of the aqueous solution at neutral or acid pH (20, 63, 72, 73). The uncatalyzed and the enzyme-catalyzed reaction show the same regiospecificity. Thus, the later steps of riboflavin biosynthesis can proceed in the absence of enzyme catalysis in aqueous solutions at moderate temperature. These observations may be relevant for the prebiotic evolution of flavin-type coenzymes. However, it is as yet unknown whether the biomimetic formation of the intermediate compound 5 is also possible.

ACKNOWLEDGMENTS

Work in our laboratory has been supported by the Deutsche Forschungsgemeinschaft and the Fonds der Chemischen Industrie. The expert help of Angelika Kohnle with the preparation of the manuscript is gratefully acknowledged.