Biosynthesis of Nucleotides, Flavins, and Deazaflavins in Methanobacterium thermoautotrophicum*

The biosynthesis of deazaflavins, flavins, ribonucle- otides, and selected amino acids was studied in Methanobacterium thermoautotrophicum by incorporation of I3C-labeled acetate and pyruvate. 13C enrichments were monitored by I3C and 'H NMR spectroscopy. The biosynthesis of ribonucleotides follows the standard pathway. The xylene ring of riboflavin is formed from two pentose moieties in agreement with studies in yeasts and eubacteria. The pyrimidine ring and the ribityl side chain of the deazaflavin chromophore of coenzyme FdZ0 are derived from the purine nucleotide pool. The phenolic ring and C-5 of the deazaflavin system are supplied by the shikimate pathway. A hy- pothetical mechanism for the assembly of the deazaflavin chromophore from 5-amino-6-ribitylamino- 2,4( lH,3H)-pyrimidinedione and 4-hydroxyphenyl-pyruvate is proposed.

The biosynthesis of deazaflavins, flavins, ribonucleotides, and selected amino acids was studied in Methanobacterium thermoautotrophicum by incorporation of I3C-labeled acetate and pyruvate. 13C enrichments were monitored by I3C and 'H NMR spectroscopy. The biosynthesis of ribonucleotides follows the standard pathway. The xylene ring of riboflavin is formed from two pentose moieties in agreement with studies in yeasts and eubacteria. The pyrimidine ring and the ribityl side chain of the deazaflavin chromophore of coenzyme FdZ0 are derived from the purine nucleotide pool. The phenolic ring and C-5 of the deazaflavin system are supplied by the shikimate pathway. A hypothetical mechanism for the assembly of the deazaflavin chromophore from 5-amino-6-ribitylamino-2,4( lH,3H)-pyrimidinedione and 4-hydroxyphenylpyruvate is proposed.
Methanogenic bacteria generate energy and cell mass by the reduction of CO, with H2 as electron donor. The pathway of CO, assimilation is well established Stupperich, 1978, 1980). Briefly, the reduction of CO, yields active methyl groups that can be utilized for the formation of acetyl-coA Fuchs, 1984a, 1984b). Further reductive carboxylation yields pyruvate, which can serve as a precursor for a wide variety of metabolic pathways.
A variety of unusual coenzymes was detected relatively recently in methanogenic bacteria (for review, see Wolfe, 1985). The green fluorescent coenzyme F420 (Fig. 1) was detected by Wolfe and his co-workers in various species of methanogenic bacteria (Cheeseman et al., 1972), and its structure ( I , Fig. 1) was elucidated by Eirich et al. (1978). The chromophoric moiety of the coenzyme was assigned the structure of 7,8-didemethyl-8-hydroxy-5-deazariboflavin (2, factor FJ.' Whereas flavocoenzymes can participate in one-and *This work was supported by grants from the Deutsche Forschungsgemeinschaft and the fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
two-electron redox steps, the deazaflavins are limited to twoelectron reactions and can be considered as functional analogs of pyridine nucleotides (Jacobson and Walsh, 1984). In methanogenic bacteria, the deazaflavin coenzyme F420 is involved in key steps of methanogenesis and carbon assimilation (Zeikus et al., 1977). Deazaflavin coenzymes have also been found in several other microbial species such as Streptomyces, mycobacteria, and nonmethanogenic Archaebacteria, in which they serve special functions such as the biosynthesis of tetracycline and as cofactor of DNA photolyase (Daniels et al., 1985;Eker et al., 1980;McCormick and Morton, 1982). Jaenchen et al. (1984) observed the incorporation of  guanine into coenzyme F420 by M. thermoautotrophicum and concluded that the pyrimidine ring of the coenzyme is derived from a purine precursor. Scherer et al. (1984) observed the incorporation of label from [1-'3C]glycine into a carbon atom of coenzyme F420 which is characterized by a I3C NMR signal a t 162.7 ppm. The authors assumed that this signal represents C-loa of the deazaflavin chromophore and concluded that the incorporation should have occurred via a purine intermediate. The biosynthesis of riboflavin has been studied extensively in fungi and eubacteria. Early studies documented that the pyrimidine ring of the flavin chromophore is derived from a purine precursor (for review see Young, 1986). More recently, it was shown that the xylene ring i s derived from two pentose moieties Bacher, 1988, 1990). The biosynthesis of riboflavin in Archaebacteria has not been studied up to now. This paper reports on incorporation studies with "C-labeled acetate and pyruvate which yield information on the biosynthetic pathways of deazaflavins, flavins, purines, pyrimidines, and amino acids in M. thermoautotrophicum.

Microorganism-M. thermoautotrophicum
Marburg (DSM no. 2133) was obtained from the Deutsche Sammlung von Mikroorganismen, Gottingen, Federal Republic of Germany. The strain was subcultured a t weekly intervals in serum bottles under an atmosphere of H,/C02 (4:1, v/v) which were incubated a t 65 "C.
Serum bottles (100 ml) equipped with butyl rubber septa and containing 20 ml of medium were gassed with a mixture of Hz/C02 (4:1, v/v) to a pressure of 2 bars and autoclaved a t 121 "C. Culture medium in fermentor vessels (0.3-20 liters) was autoclaved and gassed immediately after removal from the autoclave with a mixture of Hz/ COP (4:1, v/v) a t atmospheric pressure. Fermentation-". thermoautotrophicum was grown in a thermostatted vessel containing 300 ml of medium at 65 "C with stirring. The medium was gassed with a mixture of HZ/C02 (4:1, v/v, 150 ml/ min). When the culture had reached late exponential growth phase, it was transferred into a 20-liter fermentor containing 18 liters of medium. The fermentor was equipped with a p H electrode and was gassed with the H2/C02 mixture (4:1, v/v) a t a rate of 3 liters/min. The cell suspension was stirred a t 65 "C. A sterile solution of 5% sodium sulfide was added constantly a t a rate of 2 ml/h. When the cell density had reached a value of 0.4 OD (660 nm, 1 cm), the pH was adjusted to 6.0 by the addition of 1 M HCl. Subsequently the p H was kept at this value by automated addition of 20% NazCO,. 13C-Labeled sodium acetate was added as a sterile solution to a final concentration of 4 mM when the cell density had reached a value of 0.5 OD. '%Labeled sodium pyruvate was added to the culture in two portions at cell densities of 1.2 and 1.8 OD to a final concentration of 2.1 mM. The fermentor was kept in the dark to avoid photodecomposition of coenzymes. Fermentation was continued until early stationary phase. Cells were harvested aerobically by centrifugation and frozen at -20 "C. The medium was set aside for isolation of coenzymes.
Isolation of Flavins and Deazaflavins from the Culture Medium-The culture fluid was passed through a column of Dowex 1-X8 (formate form, 20-50 mesh, 5 X 30 cm). The column was washed with water. The combined effluents were passed through a column of charcoal (35-50 mesh, 3 X 7 cm). The column was washed with water. Fluorescent material was eluted from the Dowex column by 10 liters of 2 M formic acid. The eluate was concentrated to dryness under reduced pressure. Fluorescent material was also eluted from the charcoal column by 3 liters of 0.8 M NH,OH containing 50% ethanol. The eluate was concentrated to dryness under reduced pressure. The material eluted from both columns was dissolved in water and combined. The solution was placed on a column of QAE Sephadex A-25 (HCO; form, 3 X 41 cm). The column was developed with a linear gradient of 0-1.5 M NH,HCO, (total volume, 2 liters). Fractions of 20 ml were collected. Elution of flavins and deazaflavins was monitored by photometry and by HPLC analysis. Riboflavin was eluted at 0.15 M NH,HCOs, FMN at 0.7 M NH4HC03, and factor F, a t 0.75 M NH4HC0,. The respective fractions were pooled and concentrated to dryness under reduced pressure.
Isolation of Flavins and Deazaflavins from Bacterial Cell.-Cell paste of M. thermoautotrophicum (100 g) was suspended in 50% aqueous acetone (400 ml) a t -10 "C. The suspension was stirred at 4 "C for 30 min. The suspension was centrifuged, and the residue was again extracted as described until the supernatant was colorless (approximately 10 times). The residual cell mass was suspended in 80% aqueous methanol (400 ml) and heated at 80 "C for 20 min in order to extract tightly bound FMN. The suspension was centrifuged, and the cells were again extracted as described (two to three times). The supernatants from the extraction procedures were combined and concentrated to dryness under reduced pressure. The residue was dissolved in 80 ml of water and placed on a column of QAE Sephadex A-25 (HCOi form, 3 X 41 cm). The column was developed with a linear gradient of 0-1.5 M NH,HCO, (total volume, 2 liters) followed by 1.5 M NH,HCO:, (1 liter). Fractions of 20 ml were collected. FMN was eluted a t 0.7 M NH,HCO:$, FAD at 0.75 M NH,HCO:3, and coenzyme F4,, at 1.5 M NH,HCO,. The respective fractions were concentrated to dryness under reduced pressure. Corresponding fractions obtained from cell extract and cell culture medium were combined.
Purification of Riboflavin-Crude FAD was dissolved in 0.1 M phosphate buffer (pH 7) and treated with phosphodiesterase (Naja naja sputatrin venom, 100 pg) for 24 h at room temperature. The resulting FMN was combined with FMN obtained from culture fluid and cell extract and dissolved in 5 ml of 0.1 M NH4HC03 containing 0.5 mM MgClZ, and the pH of the solution was adjusted to 10.4. Alkaline phosphatase (7 units) was added, and the mixture was kept at 37 "C for 2 h. The resulting riboflavin was combined with crude riboflavin obtained from the culture fluid. Purification of riboflavin was achieved by preparative reversed phase HPLC using a column of Lichrosorb RPll (16 X 250 mm) and an eluent containing 30% methanol. Riboflavin was monitored by photometry a t 365 nm. The retention volume was 260 ml. Fractions were collected and concentrated to dryness under reduced pressure.
Purification of Deazaflauins-Factor Fd2, was purified by preparative HPLC on a column of Lichrosorb RP18 (16 X 250 mm) with an eluent containing 27% MeOH and 30 mM formic acid. The effluent was monitored photometrically (405 nm). The retention volume was 200 ml. Fractions were combined and concentrated to dryness under reduced pressure.
Coenzyme Flpo was hydrolyzed by treatment with 1 M HC1 a t 110 "C for 1 h. The solution was adjusted to p H 9 with 13% NH,OH and concentrated to dryness under reduced pressure. The resulting 5"phosphate of factor FO was dissolved in 0.1 M ammonium bicarbonate buffer (pH 10.4) containing 0.5 mM MgCl, and treated with alkaline phosphatase (10 units) a t 37 "C for 20 min. The solution was adjusted to p H 6 by the addition of formic acid and was concentrated under reduced pressure.
The samples of factor F, obtained as described above were combined and dissolved in 30 ml of 0.1 M phosphate buffer (pH 10). Insoluble material was removed by centrifugation. The pH of the yellow colored supernatant was adjusted to 6 by the addition of concentrated formic acid. The solution was kept overnight a t 4 'C. The precipitate of factor Fo was harvested by centrifugation and washed twice with ice water. The supernatant was placed on a preparative HPLC column of Lichrosorb RP18 (16 X 250 mm) which was developed with 27% methanol. The retention volume of factor F, was 240 ml. Fractions were concentrated to dryness under reduced pressure. When necessary, factor Fo was recrystallized from a water/ dimethyl sulfoxide mixture (4:1, v/v).
Isolation of Nucleosides-The residue remaining after repeated acetone and methanol extraction of bacterial cell mass was suspended in a mixture of ethanol/ether (3:1, v/v, 200 ml) and boiled under reflux for 1 h. The suspension was filtered, and the residue was washed with ether and dried a t room temperature. The residue was suspended in 100 ml of 1 M NaOH and was stirred for 24 h a t room temperature. Concentrated hydrochloric acid (10 ml) and 25% trichloroacetic acid (10 ml) were added, and the suspension was centrifuged. The supernatant was adjusted to pH 8.0 by the addition of 25% NH,OH. Barium acetate (50 mg) and ethanol (650 ml) were added. The mixture was kept at 4 "C for 12 h, and the precipitate of nucleotide barium salts was harvested by centrifugation. Barium acetate (200 mg) was added to the supernatant, which was again kept a t 4 "C for 12 h and then centrifuged. The precipitates were combined and dissolved in 50 ml of 50 mM HCl. A concentrated solution of sodium sulfate (2 ml) was added, and the precipitate of BaS0, was removed by centrifugation. The supernatant was placed on a column of Dowex 5OW-XS (H' form, 200-400 mesh, 2 X 34 cm). The column was developed with 160 ml of 50 mM HCl and subsequently with water. Fractions of 20 ml were collected and analyzed by HPLC. The retention volumes were as follows: UMP, 100 ml; GMP, 300 ml; CMP and AMP, 1.000 ml. CMP and AMP were separated by preparative HPLC using a column of Lichrosorb RPls (16 X 250 mm) and an eluent containing 100 mM ammonium formate and 100 mM formic acid. The nucleotides were monitored a t 254 nm. 2'-AMP and 3'-AMP had retention volumes of 260 and 560 ml, and 2'-CMP and 3'-CMP had a retention volume of 100 ml, respectively. Fractions were concentrated to dryness under reduced pressure. The isolated nucleotides were dissolved in 0.1 M ammonium bicarbonate buffer (pH 10.4) containing 0.5 mM MgCI, and were treated with alkaline phosphatase (30 units) at 37 "C for 7 h. Final purification of all nucleosides was performed by preparative HPLC on a column of Lichrosorb RP,, (16 X 250 mm). Water was used as eluent for the purification of uridine, cytidine, and guanosine. Adenosine was eluted with 5% methanol. The retention volumes were 135, 145, 420, and 530 ml for uridine, cytidine, guanosine, and adenosine, respectively. Fractions were collected and concentrated to dryness under reduced pressure.
Isolation of Amino Acids-The residue obtained after alkali treatment of bacterial cell mass (see above) was dialyzed against water and lyophilized. The product was suspended in 200 ml of 6 M HC1 containing 4% thioglycolic acid. The mixture was stirred and heated under reflux a t 110 "C for 24 h. The solution was filtered, and the filtrate was lyophilized. The residue was dissolved in water and placed on a column of Dowex 50W-X8 (NH: form, 200-400 mesh, 3.1 X 28 cm). The column was developed with 3.5 liters of 0.2 M ammonium formate (pH 4.5) followed by 1.5 liters of 0.2 M ammonium formate (pH 6.5), and finally 1 liter of 0.5 M ammonium formate (pH 6.5). Fractions were analyzed by automated amino acid analysis or by thin layer chromatography (cellulose plates, 3% ammonium chloride, visualization with ninhydrin). Retention volumes were as follows: neutral and acidic amino acids, 350-850 ml; tyrosine and phenylalanine, 950-1,200 ml; lysine, 3,200 ml; histidine, 4,000 ml; arginine, 5,500 ml. Fractions were pooled and concentrated to dryness under reduced pressure. Tyrosine and phenylalanine were obtained in pure form by preparative HPLC on a column of Lichrosorb RP,, (16 X 250 mm) using water as eluent and a UV detector a t 280 nm. The retention volumes of tyrosine and phenylalanine were 85 and 160 ml, respectively. Fractions were concentrated under reduced pressure. Tyrosine was recrystallized from water. Glutamic and aspartic acid were separated from the neutral amino acids by anion exchange chromatography on Dowex 1-X8 (formate form, 200-400 mesh, 1.5 X 34 cm). Glutamic and aspartic acid were eluted with 0.01 M formic acid. Further purification was achieved by preparative HPLC on a Nucleo-si1 SB 10 column (formate form, 16 X 250 mm) with an eluent containing 10 mM formic acid. The amino acids were detected photometrically a t 206 nm or by refractometry. The retention volumes of glutamic acid and aspartic acid were 95 and 205 ml, respectively.
The neutral amino acids were separated by cation exchange chromatography on Dowex 50W-X8 (H' form, 200-400 mesh, 2.3 X 40 cm). The column was developed with a linear gradient of 0-3 M HCl (total volume, 2 liters). Fractions of 20 ml were collected. Retention volumes were as follows: serine, 730 ml; threonine, 750 ml; glycine, 865 ml; alanine, 880 ml. Fractions were pooled, concentrated to dryness under reduced pressure, and lyophilized. Serine and threonine were purified further by preparative HPLC on a Nucleosil SB 10 column (formate form, 16 X 250 mm) using water as eluent. The retention volumes of serine and threonine were 40 ml.
N M R Spectroscopy-NMR Spectra were recorded at 8.46 tesla on a Bruker AM360 spectrometer a t 300 K. Factor Fo, riboflavin, and guanosine were measured in Me2SO-&, tyrosine, and aspartate in 0.1 M NaOD, glycine, alanine, phenylalanine, and serine in 0.1 M DC1, and cytidine in 0.1 M deuterophosphate buffer (pH 6.6, uncorrected glass electrode reading).
All ',IC NMR spectra were recorded under identical conditions. Data acquisition and processing parameters were as follows: 64 K data set, 30 pulse width (2 p s ) , 2.5s scan interval, 14.7-kHz spectral range, composite pulse decoupling, 1-Hz line broadening.
During all two-dimensional experiments, the samples were not rotated.
Data acquisition and processing parameters for two-dimensional experiments were as follows: ( a ) COSY, 32 scans/tl increment, 1.8-s relaxation delay, 500 X 2,056 raw data matrix size zerofilled to 2 K in f, and processed with 2 Hz Gaussian in the tl dimension and 90" shifted squared sine bell filtering in the t 2 dimension; ( b ) HOHAHA, 16 scans/t, increment, 1.0-s relaxation delay, 63-ms MLEV-17 mixing period preceded and followed by 2.5-ms trim pulses; 90" pulse width, 60 ps; 256 X 1,024 raw data matrix size, zerofilled to 1 K in f, and processed with 2 Hz Gaussian in the tl dimension and 90" shifted squared sine hell filtering in the t2 dimension; (c) ROESY, 64 scans/ tl increment, 1.1-s relaxation delay, 200-ms continuous wave spin lock period 90" pulse width, 60 ps; 450 X 1,024 raw data matrix size, zerofilled to 1 K in fl and processed with 2 Hz Gaussian in the tl dimension and 60" shifted sine bell filtering in the t2 dimension; ( d ) 'H-13C HMQC, 96 scans/tl increment, 1.0-s relaxation delay, 3.5-ms delay period for evolution of 'JCH corresponding to a coupling constant of 145 Hz; 128 X 2,048 raw data matrix size, zerofilled to 512 words and processed with 90" shifted sine bell filtering in t2; ( e ) 'H-13C HMBC, 192 scans/t, increment, 1.0-s relaxation delay, 3.5-ms delay for suppression of ' J C H ; 62-ms and 125-ms delay periods for evolution of long range couplings, corresponding to coupling constants of 8 and 4 Hz, respectively; 128 X 2,048 raw data matrix size zerofilled to 512 words in f, and processed with 90" shifted sine bell filtering in tl; ( f ) INADEQUATE, 64 scans/t, increment, 2.3-s relaxation delay, Ernsttype phase cycle; 3.6-ms delay for evolution of 'Jcr; 512 X 2,048 raw data matrix size, zerofilled to 2 K in fl and processed with 60" sine bell filtering in t,.

RESULTS
M. thermoautotrophicum has a highly specialized metabolism which enables it to grow autotrophically on COz and Hz.
Complex organic nutrients are utilized rather inefficiently, probably for lack of appropriate uptake systems. Simple organic nutrients such as acetate and pyruvate are at least metabolized to some extent and can be used for 13C incorporation studies. Predictably, such simple precursors will contribute 13C label to many different positions in virtually all complex metabolic products. These complex labeling patterns can be assessed with high accuracy by 13C NMR spectroscopy. A rigorous interpretation of the 13C labeling obtained with simple two-or three-atomic precursors can be achieved by comparison among the labeling patterns of a variety of metabolic products from the same fermentation. With this aim, we have analyzed quantitatively the labeling patterns of purine and pyrimidine nucleotides as well as amino acids. These data subsequently served as the background for the interpretation of the 13C labeling pattern in the coenzymes, deazaflavin and riboflavin.
Huster and Thauer (1983) observed that the utilization of pyruvate in M. thermoautotrophicum is optimal at pH 6. The same is true for acetate as shown by preliminary experiments with [14C]acetate (data not shown). Based on these data, the microorganism was grown at a constant pH of 6 in a mineral salts medium under an atmosphere of H,/COz (4:1, v/v). Labeled acetate (1-"C, 2-''CC, or 1,2-"Cc,) was added to a final concentration of 4 mM.
Pyruvate was unstable in long time fermentations. After the addition of sodium [l-14C]pyruvate to a culture of M. thermoautotrophicum, 50% of the radioactivity was lost within 15 h, probably as 14C02. [1-"C]Pyruvate was therefore added sequentially in two portions to a rapidly growing culture.
In all feeding experiments, incubation was continued until early stationary phase. The cells were then separated from the culture medium. Riboflavin, FMN, and factor Fo were isolated from the culture fluid. Coenzyme F420, FAD, and FMN were isolated from the cell mass. The flavocoenzymes were converted to riboflavin, and coenzyme F420 was converted to factor Fo by hydrolysis. Purine and pyrimidine nucleosides  " ''3C-1'3C couplings observed in the spectrum of the biosynthetic sample from the [1,2-l3Cz]acetate fermentation ' Percent coupling calculated as percentage of total signal for a given carbon atom involved in 1''C-13C coupling.
indicating the incorporation of intact acetate units.
Carbon atom arbitrarily assigned a relative "C enrichment of 1.0. ND, not determined.

6
were isolated after hydrolysis of RNA by alkali treatment of cell mass, and amino acids were isolated after acid hydrolysis of cellular proteins. '"C enrichments of the isolated metabolites were determined by I 3 C NMR spectroscopy. Integrals of individual signals were calibrated using the spectra of samples with natural "'C abundance which were measured under identical experimental conditions. In each compound under study, the carbon atom with the lowest relative enrichment was assigned a relative enrichment value of 1.0. I 7 When possible, 13C enrichments of selected carbon atoms were also determined by evaluation of the 13C satellites in 'H NMR spectra. This approach was limited to carbon atoms attached to protons with simple NMR signatures, preferably singlets. The enrichment values derived for a given carbon atom from 13C and 'H NMR spectroscopy were generally in good agreement.
The labeling patterns of guanosine (5, Fig. 2) and cytidine (7, Fig. 3) isolated from RNA are summarized in Table I. The labeling pattern of adenosine was identical to that of guano-sine, and the labeling pattern of uridine was identical to that of cytidine (data not shown). The labeling pattern of the ribose moiety was the same in all four nucleosides studied and can be easily rationalized on the basis of known metabolic pathways in M. thermoautotrophicum (Fig. 4). It is well established that pyruvate (8) is formed by reductive carboxylation of acetate (Fuchs and Stupperich, 1980). The labeling pattern of pyruvate is directly reflected by alanine (9)  The labeling pattern of ribose phosphate was deduced from measurements of the nucleosides 5 (Fig. 2) and 7 (Fig. 3). This approach avoids the problem of anomeric 13C signal splitting. For details see pyruvate by transamination (Table 11). Starting from pyruvate, glucogenesis leads to a symmetrical duplication of the labeling pattern. Pentose formation by the oxidative branch of the pentose phosphate cycle then yields the observed labeling pattern of ribose (Fig. 4). The incorporation of intact acetate units into the positions 4 and 5 of the ribose moiety was clearly shown by l3C-I3C coupling observed in the samples from the fermentation with double labeled acetate.
The pathway of purine biosynthesis in bacteria and fungi is well documented and has been reinvestigated recently in yeast by Kozluk and Spenser (1987). The observed labeling pattern of purines in M. thermoautotrophicum is well in line with the standard pathway as shown in Fig. 2. Similarly, the labeling pattern of pyrimidine nucleotides (Fig. 3) is in line with the respective standard pathway.
The biosynthetic origin of the xylene ring of riboflavin has been established recently by studies with yeast and Bacillus subtilis (Bacher et al., 1983Le Van et al., 1985;Bacher, 1988, 1990). The loss of C-4 from ribulose 5phosphate (16, Fig. 6) by way of an intramolecular rearrangement yields the novel carbohydrate, ~-3,4-dihydroxy-%butanone 4-phosphate (I 7). Condensation with 5-amino-6ribitylamino-2,4(1H,3H)-pyrimidinedione (18) yields 6,7-dimethyl-8-ribityllumazine (20). Subsequent dismutation of the lumazine yields riboflavin ( 2 1 ) as shown earlier by Plaut (1960Plaut ( ,1963. Thus, the xylene ring of the vitamin is ultimately  acetate. The spectrum was processed by exponential multiplication with a line broadening factor of 0.2. Satellite signals from coupling with '"C are indicated by arrows. formed from two molecules of the intermediate 17. Based on the established labeling pattern of the pentose pool, the labeling pattern can be predicted as shown in Fig. 6. The experimental findings are in full agreement with the expectation. It should be noted that an intact acetate unit is not incorporated into the xylene ring as shown by the absence of I:iC""c coupling. This feature is the direct consequence of the elimination of carbon 4 from the pentulose precursor. The data leave no doubt that the biosynthesis of riboflavin in M. thermoautotrophicum follows the same pathway as in yeasts and eubacteria. The NMR spectrum of factor Fo has not been assigned previously. We first analyzed the 'H NMR spectrum of factor Fo by COSY, HOHAHA, and spin locked NOE (ROESY) experiments. The HOHAHA experiment using a relatively long mixing period of 68 ms showed two scalar coupled spin networks encompassing the protons of the ribityl side chain and the protons of the phenolic ring system, respectively. In combination with the COSY data which revealed the directly coupled spins and the NOE data obtained from the ROESY experiment, all proton signals could be assigned unambiguously (Table IV). The lRC NMR spectrum of factor Fo was subsequently assigned by DEPT spectroscopy and inverse 'H-"'C correlation experiments (Table V). Quaternary carbon atoms were assigned by multiple quantum multiple bond correlation experiments optimized for 'JcH and 'JCH. ',"C NMR spectra of factor Fo samples from different fermentations are shown in Fig. 7. The presence of strong satellites in the sample from [1,2-"C2]acetate is due to 13C-13C coupling resulting from the incorporation of several intact acetate units. These are best observed in the INADEQUATE experiment shown in Fig. 8. Moreover, this experiment gave additional confirmation to the 13C NMR signal assignments.
A portion of the 'H NMR spectrum obtained with factor Fo from [1,2-"C2]acetate is shown in Fig. 9. The protons at C-5 and C-6 give rise to a singlet and a doublet, respectively. The 13C enrichments can be calculated from the "C-coupled satellites of the proton signals. As shown in Table VI for selected examples, the enrichment values derived from 13C and 'H NMR measurements are in close agreement. The combined results of 13C incorporation into factor Fo are summarized in Table VI1 and Fig. 10.
A comparison of the labeling patterns of factor Fo, riboflavin, and guanosine confirms immediately the common origin of the pyrimidine ring in all of these compounds. Not surprisingly, the ribityl chain of the deazaflavin cofactor shows the same labeling pattern as the ribityl chain of riboflavin and the ribose moiety of nucleosides. On the other hand, it is immediately obvious that the carbocyclic rings of riboflavin and deazaflavin have entirely different origins. The labeling pattern of the phenolic ring in the deazaflavin is characterized by a mirror symmetry that is indicated in Fig. 10 by a dashed line. This labeling pattern virtually duplicates that of tyrosine, thus indicating that the deazaflavin chromophore is derived from the shikimate pathway.
The question arises as to whether we can identify the specific shikimate derivative serving as the committed precursor. The labeling of the phenolic ring in factor Fo would be consistent with tyrosine ( 1 5 ) , 4-hydroxyphenylpyruvate (I4), or 4-hydroxybenzoate (23) as precursor. Prephenate (13, Fig. 5) is less likely as a precursor; as a consequence of its prochiral character, the labeling of factor Fo resulting from this precursor should not show symmetric label distribution in the carbocyclic ring.
Some additional information can be obtained from the analysis of the labeling pattern of C-5 of the deazaflavin chromophore. C-5 is labeled from [2-13C]acetate and from [1,2-13C2]acetate but not from the other precursors studied. Although the labeling pattern of 4-hydroxybenzoate (23) could not be observed directly, it is obvious from the shikimate pathway that its carboxyl group should reflect the labeling of the carboxyl group of shikimate (12, Fig. 5) and not C-3 of pyruvate. It follows that C-5 of the deazaflavin chromophore cannot be derived from the carboxylic group of shikimate via 4-hydroxybenzoate. However, C-5 could well be derived from the aliphatic side chain of 4-hydroxyphenylpyruvate or tyrosine.

DISCUSSION
Complex labeling patterns result from feeding experiments with simple two-and three-carbon compounds as a consequence of the considerable number of reaction steps between the precursor and the product. Progress in NMR techniques and instrumentation enables the accurate experimental assessment of these complex labeling patterns. As shown in this paper, a rigorous interpretation is possible on the basis of comparison with the labeling patterns of a sufficiently large number of different metabolites.
The pyrimidine moieties of both riboflavin and deazaflavin are derived from a purine precursor. Moreover, we could show by studies with [l'-'4C]aden~~ine that the pyrimidine carbon atoms of the purine nucleotide precursor are transferred to- a "'C-"C couplings observed in the spectrum of the biosynthetic sample from the [1,2-13C2]acetate fermentation I, Percent coupling calculated as percentage of total signal for a given carbon atom involved in "C-'"C coupling.
' Carbon atom arbitrarily assigned a relative '% enrichment of 1.0.
indicating the incorporation of intact acetate units.
Average due to multiplet overlapping.  gether with the attached ribose unit which yields the ribityl side chain of the coenzymes (Schwarzkopf et al., 1990). This suggests that the biosynthesis of the two cofactors shares a common pathway. It is well established that the biosynthesis of riboflavin starts from GTP (19, Fig. 6). Opening of the imidazole ring (Foor and Brown, 1975), reduction of the ribityl side chain, and deamination of the pyrimidine ring lead to 5-amino-6-ribitylamino-2,4( lH,3H)-pyrimidinedione 5"phosphate (Burrows and Brown, 1978;Nielsen and Bacher, 1981), which is subsequently dephosphorylated to yield 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione (18) (Neuberger and , the obligatory precursor of 6,7-&methyl-8-ribityllumazine (20). The divergence between the riboflavin and the deazaflavin pathways could occur at the level of 18 or its 5"phosphate. Whereas the dephosphorylation of 5amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione 5'-phosphate is an obligatory step in the biosynthesis of flavins, it is conceivable that the condensation with a shikimate intermediate could proceed at the level of the ribitylaminopyrimidine phosphate. In this case, the biosynthetic pathway would lead directly to the FMN analog of the deazaflavin series. The benzenoid ring of the deazaflavin chromophore is derived from a shikimate derivative. 4-Hydroxybenzoate (23, Fig. 10) can be ruled out on the basis of the labeling pattern of C-5 of the chromophore. Prephenate appears unlikely as a precursor since one would expect nonsymmetric label distribution in the product. Thus, 4-hydroxyphenylpyruvate (14) and tyrosine (15) must be considered as potential precursors.
It should be noted that 4-hydroxyphenylpyruvate in aqueous solution can undergo a spontaneous cleavage reaction yielding 4-hydroxybenzaldehyde (Doy, 1960). Conceivably, this reaction could be catalyzed enzymatically, and the aldehyde could serve as an intermediate in deazaflavin biosynthesis. However, 4-hydroxyphenylpyruvate could also serve directly as precursor.
The pyrimidine 18 (Fig. 11) is easily dehydrogenated. Wood and co-workers have observed the formation of a pyrimido[5,4-g]pteridine derivative (24, Scheme 1) from 18 in the presence of oxygen (Cresswell et al., 1960). This dimerization could proceed via oxidative formation of a quinoid product 25. We propose that such a quinoid species could react with the thermodynamically favorable enol of 4-hydroxyphenylpyruvate (26), resulting in the formation of the adduct 27. The subsequent elimination of NH3 and oxalate could generate the mesomeric system 28. This intermediate could be cyclized to the deazaflavin chromophore by a two-electron oxidation step. This proposed sequence of events is as yet hypothetical and requires further confirmation.