Retrobiosynthetic NMR Studies with 13C-Labeled Glucose

The biosynthesis of gallic acid was studied in cultures of the fungus Phycomyces blakesleeanus and in leaves of the tree Rhus typhina. Fungal cultures were grown with [1-13C]glucose or with a mixture of unlabeled glucose and [U-13C6]glucose. Young leaves ofR. typhina were kept in an incubation chamber and were supplied with a solution containing a mixture of unlabeled glucose and [U-13C6]glucose via the leaf stem. Isotope distributions in isolated gallic acid and aromatic amino acids were analyzed by one-dimensional 1H and 13C NMR spectroscopy. A quantitative analysis of the complex isotopomer composition of metabolites was obtained by deconvolution of the13C13C coupling multiplets using numerical simulation methods. This approach required the accurate analysis of heavy isotope chemical shift effects in a variety of different isotopomers and the analysis of long range13C13C coupling constants. The resulting isotopomer patterns were interpreted using a retrobiosynthetic approach based on a comparison between the isotopomer patterns of gallic acid and tyrosine. The data show that both in the fungus and in the plant all carbon atoms of gallic acid are biosynthetically equivalent to carbon atoms of shikimate. Notably, the carboxylic group of gallic acid is derived from the carboxylic group of an early intermediate of the shikimate pathway and not from the side chain of phenylalanine or tyrosine. It follows that the committed precursor of gallic acid is an intermediate of the shikimate pathway prior to prephenate or arogenate, most probably 5-dehydroshikimate. A formation of gallic acid via phenylalanine, the lignin precursor, caffeic acid, or 3,4,5-trihydroxycinnamic acid can be ruled out as major pathways in the fungus and in young leaves of R. typhina. The incorporation of uniformly 13C-labeled glucose followed by quantitative NMR analysis of isotopomer patterns is suggested as a general method for biosynthetic studies. As shown by the plant experiment, this approach is also applicable to systems with low incorporation rates.

The biosynthesis of gallic acid was studied in cultures of the fungus Phycomyces blakesleeanus and in leaves of the tree Rhus typhina. Fungal cultures were grown with [1-13 C]glucose or with a mixture of unlabeled glucose and [U-13 C 6 ]glucose. Young leaves of R. typhina were kept in an incubation chamber and were supplied with a solution containing a mixture of unlabeled glucose and [U-13 C 6 ]glucose via the leaf stem. Isotope distributions in isolated gallic acid and aromatic amino acids were analyzed by one-dimensional 1 H and 13 C NMR spectroscopy. A quantitative analysis of the complex isotopomer composition of metabolites was obtained by deconvolution of the 13 C 13 C coupling multiplets using numerical simulation methods. This approach required the accurate analysis of heavy isotope chemical shift effects in a variety of different isotopomers and the analysis of long range 13 C 13 C coupling constants. The resulting isotopomer patterns were interpreted using a retrobiosynthetic approach based on a comparison between the isotopomer patterns of gallic acid and tyrosine. The data show that both in the fungus and in the plant all carbon atoms of gallic acid are biosynthetically equivalent to carbon atoms of shikimate. Notably, the carboxylic group of gallic acid is derived from the carboxylic group of an early intermediate of the shikimate pathway and not from the side chain of phenylalanine or tyrosine. It follows that the committed precursor of gallic acid is an intermediate of the shikimate pathway prior to prephenate or arogenate, most probably 5-dehydroshikimate. A formation of gallic acid via phenylalanine, the lignin precursor, caffeic acid, or 3,4,5-trihydroxycinnamic acid can be ruled out as major pathways in the fungus and in young leaves of R. typhina. The incorporation of uniformly 13 C-labeled glucose followed by quantitative NMR analysis of isotopomer patterns is suggested as a general method for biosynthetic studies. As shown by the plant experiment, this approach is also applicable to systems with low incorporation rates.
The genetic manipulation of phenylpropanoid-derived metabolites has been proposed as a prospective target for crop improvement (1). A fundamental understanding of the key enzymes of the various phenylpropanoid branching pathways is therefore of increasing interest.
Gallic acid (3,4,5-trihydroxybenzoic acid, compound 7, Fig. 1) serves as a fundamental precursor for gallotannins and ellagitannins, abundant classes of plant secondary metabolites. Gallic acid is also formed in substantial amounts by some fungi. A considerable number of studies on the biosynthesis of tannins has been reported (for review, see Gross (2) and Haslam (3)). Surprisingly, the biosynthesis of gallic acid, the phenolic unit of this important class of natural products, is still incompletely understood. Based mainly on radiolabeling studies, it has been proposed that gallic acid could be formed from phenylalanine (compound 5) via caffeic acid (compound 8), 3,4,5-trihydroxycinnamic acid (compound 9), or protocatechuic acid (compound 6, Fig. 1) (4 -9). This would imply that the carboxylic group of gallic acid is derived from the ␤ carbon atom of phenylalanine (compound 5), and ultimately from the enoylpyruvoyl group of chorismate (compound 3, Fig. 1). On the other hand, it has been argued that the carboxylic group of gallic acid is biosynthetically equivalent to the carboxylic group of shikimate (10 -15). This would imply that gallic acid is formed from an early shikimate intermediate, e.g. directly from 5-dehydroshikimate (compound 4) by dehydrogenation or via protocatechuic acid (compound 6) as an intermediate ( Fig. 1) (16). It was also proposed that the controversial results could indicate the existence of alternative biosynthetic routes in the same organism (17)(18)(19)(20).
This report describes studies with shake flask cultures of the fungus, Phycomyces blakesleeanus, and with young leaves of the tree, Rhus typhina, which were supplied with exogenous 13 C-labeled glucose. Gallic acid and amino acids were obtained by hydrolysis of cell material, and the 13 C distribution was analyzed by NMR spectroscopy. A central aspect of this approach was a retrobiosynthetic comparison of labeling patterns of gallic acid with those of amino acids. The results show unequivocally that gallic acid is biosynthesized in both organisms from an early intermediate of the shikimate pathway, but not via phenylalanine.  [1-13 C]glucose in 5 ml of water was added to the autoclaved medium (300 ml). Alternatively, a mixture of [U-13 C 6 ]glucose (320 mg) and unlabeled glucose (7.7 g) in 5 ml of water was added to the autoclaved medium (400 ml). Erlenmeyer flasks (1 liter) containing 400 ml of culture medium were seeded with one loopful of spores and were incubated with shaking at 25°C for 6 days. The mycelium was harvested by filtration and was washed with a small amount of water.

Materials-[U-
Plant Culture-The stems of young leaves of R. typhina (with a length of 3-4 cm) were immersed into a sterile solution of 1% (w/w) of glucose consisting of 96% of unlabeled glucose and 4% of [U-13 C 6 ]glucose. They were incubated in a dark compartment (Heraeus Vötsch climate chamber) at 18°C and 60% humidity for 33 days. Small sections of the stem were removed with a sharp razor at intervals of 1 or 2 days.
Isolation of Gallic Acid from P. blakesleeanus Culture-The culture medium from experiments with P. blakesleeanus was adjusted to pH 1, treated at 96°C for 6 h, and extracted with 150 ml of diethyl ether for 12 h using a liquid/liquid perforator. Fungal biomass was treated with 1 M sulfuric acid (100 ml) at 96°C for 6 h. The solution was filtered and was then continuously extracted with 150 ml of diethyl ether for 12 h using a liquid/liquid perforator. The organic phases obtained by extraction of culture medium and of cell mass hydrolysate of P. blakesleeanus were combined. The aqueous solution obtained by acidic treatment of biomass and subsequent extraction with ether was set aside for isolation of tyrosine and phenylalanine.
The organic phase was extracted twice with 75 ml of saturated sodium bicarbonate. The aqueous solution was adjusted to pH 1 by the addition of concentrated sulfuric acid and was again subjected to continuous extraction with 150 ml of diethyl ether. The organic phase was concentrated to a small volume (about 3 ml) under reduced pressure.
The solution was applied to a column of silica gel (Merck, 1.5 ϫ 35 cm). The column was developed with a mixture of diethyl ether/ethyl acetate/formic acid (50:40:10, v/v). Gallic acid was eluted with the colored solvent front. Fractions were combined and were brought to dryness under reduced pressure. Aliquots of fractions were applied to a TLC silica gel plate (Merck, Darmstadt), which was developed with a mixture of diethyl ether/ethyl acetate/formic acid (50:40:10, v/v). Gallic acid migrated at a R F of 0.82 and was detected by fluorescence quenching (254 nm).
Isolation of Gallic Acid from Plant Material-Acid treatment of biomass and isolation of gallic acid was performed as described for fungal cell mass.
High Performance Liquid Chromatography-Reversed phase high performance liquid chromatography was performed with a column of Nucleosil RP18 (4.5 ϫ 250 mm) which was developed with 23% aqueous methanol. The effluent was monitored photometrically (258 nm). Semipreparative separations were performed with a Nucleosil RP18 column (16 ϫ 250 mm). The retention volume of gallic acid was 48 ml.
Isolation of Amino Acids-Tryptophan was isolated from cell mass after alkaline hydrolysis as described earlier (21). Tyrosine and phenylalanine were isolated from the aqueous phase obtained after extraction of the biomass hydrolysate with diethyl ether (see above). The aqueous phase was adjusted to 6 M HCl and boiled under reflux for 24 h. The separation of the amino acids from the hydrolysate was performed as described earlier (22).
NMR Spectroscopy-1 H and 13 C NMR spectra were recorded at 500.13 and 125.76 MHz, respectively, with a Bruker DRX500 spectrom- eter. Gallic acid was dissolved in methanol-d 4 , tyrosine, and tryptophan in 0.1 M NaOD, and phenylalanine in 0.1 M DCl.
Determination of 13 C Enrichment-Relative enrichments for all carbon atoms of biosynthetic gallic acid, tyrosine, phenylalanine, and tryptophan were obtained by comparison of 13 C integrals with natural abundance standards. This standardization considers different relaxation behavior of carbon atoms and prevents false quantification of 13 C enrichments due to differences in relaxation times. Absolute 13 C enrichments of selected carbon atoms were obtained from the 13 C satellites in the 1 H spectra. The relative enrichments were subsequently referenced to these respective carbon atoms (23). Signal assignments were based on two-dimensional HMQC, HMBC, and INADEQUATE experiments (data not shown).
NMR Spectra Simulation-13 C-Coupling patterns were simulated with the program package NMRSIM (Bruker).

RESULTS
The present study was designed to determine unequivocally whether the carboxylic group of gallic acid originates from the carboxylic group of an early shikimate precursor (e.g. 5-dehydroshikimate) or from the ␤ carbon atom of the phenylalanine side chain. This question can be addressed by a comparison between the labeling patterns of biosynthetic gallic acid and phenylalanine or tyrosine, which are easily obtained by hydrolysis of cell protein. This retrobiosynthetic approach requires a labeling strategy conducive to significantly different labeling of the carboxylic group of early shikimate derivatives and of the ␤ carbon atom of phenylalanine or tyrosine. Since the carboxylic group of early shikimate derivatives and the ␤ carbon of phenylalanine are derived from different atoms (i.e. the carboxyl group and the methyl group) of pyruvate, this condition can be fulfilled by a variety of precursors, and unequivocal results were obtained with [1-13 C]glucose in P. blakesleeanus. In leaves of R. typhina, the incorporation rate of [1-13 C]glucose was too low to afford unequivocal results. However, this problem could be addressed by incorporation experiments using a mixture of [U-13 C 6 ]glucose and unlabeled glucose. It will be shown that the sensitivity of this technique is substantially better as compared with the single-labeled precursor technique.
Experiments with Cultures of P. blakesleeanus-A shake flask culture of P. blakesleeanus (300 ml) was supplied with 6 g of [1-13 C]glucose (99.5% enrichment). The culture was incubated for 6 days. Gallic acid (5 mg) was isolated from the culture fluid and from the cell hydrolysate as described under "Experimental Procedures." Phenylalanine (3 mg), tyrosine (4 mg), and tryptophan (2 mg) were obtained by hydrolysis of cell mass (8 g, wet weight).
Due to the symmetry of the aromatic rings of phenylalanine, tyrosine and gallic acid, the respective ring carbon atoms 2/6 and 3/5 yield only an averaged 13 C abundance value, although they have different biosynthetic origins (C-4 of erythrose 4phosphate/C-3 of phosphoenolpyruvate and C-1/C-3 of erythrose 4-phosphate, respectively). Tryptophan reflects the original, i.e. nonsymmetrical, labeling patterns of the shikimate ring system, and the labeling pattern of erythrose 4-phosphate (compound 2) can thus be reconstructed by a retrobiosynthetic approach on basis of the tryptophan biosynthetic pathway (Fig.   2). The labeling pattern of phosphoenolpyruvate (compound 1) can be deduced unequivocally from the labeling pattern of the side chains of phenylalanine and tyrosine (Fig. 2).
The observed labeling patterns are explained on basis of carbohydrate metabolic pathway. Label from [1-13 C]glucose is diverted to C-3 of triose phosphate type metabolites by the glycolytic pathway (compounds 14 and 15, Fig. 3). This results in the high 13 C enrichment of the ␤ position of phenylalanine and tyrosine reflecting C-3 of phosphoenolpyruvate (23.4% 13 C abundance, Fig. 2). Decarboxylation of [1-13 C]glucose 6-phosphate by the oxidative branch of the pentose phosphate pathway yields 13 CO 2 and unlabeled ribulose 5-phosphate (compound 13, Fig. 3). Erythrose 4-phosphate (compound 2) generated from the unlabeled ribulose 5-phosphate via the nonoxidative pentose phosphate pathway should be devoid of 13 C label. However, the observed 13 C labeling in the 4-position of erythrose 4-phosphate (18.5% 13 C abundance) can be explained by futile cycling via the glycolytic/glucogenetic pathways and/or via the mannitol pathway (Fig. 3). Specifically, FIG. 2. 13 C Abundance (in %) of gallic acid (compound 7), phenylalanine (compound 5), tyrosine (compound 10), and tryptophan (compound 11) from P. blakesleeanus cultured with [1-13 C]glucose. The carbon connectivities are shown schematically. The labeling patterns of erythrose 4-phosphate (compound 2) and phosphoenolpyruvate (compound 1) were reconstructed from the amino acids based on conventional mechanisms of the shikimate biosynthetic pathway (for details see Fig. 1). regeneration of glucose from triose phosphate isomerase could divert 13 C label to the 6-position of glucose. Similarly, the reversible conversion of glucose to mannitol (compound 16) could also divert label to the 6-position of glucose. The resulting [6-13 C]glucose 6-phosphate could then be converted to [4-13 C]erythrose 4-phosphate via the pentose phosphate pathway with or without contribution of oxidative decarboxylation. The observed labeling in C-4 of erythrose 4-phosphate indicates that at least two-thirds of exogenous glucose is distributed to the cellular hexose phosphate pool by futile glycolytic and/or mannitol cycling.
The labeling pattern of the benzenoid ring of gallic acid (compound 7) agrees closely with the ring labeling of phenylalanine (compound 5) and tyrosine (compound 10, Fig. 2). The carboxylic group of gallic acid has less 13 C abundance (2.9%) whereas the ␤ carbon atoms of tyrosine or phenylalanine are highly labeled (23.4%). It follows that the carboxylic group is not derived from the ␤ carbon of phenylalanine or tyrosine.
As shown in Fig. 1, the carboxylic atom of early shikimate precursors (e.g. 5-dehydroshikimate, compound 4) stems from C-1 of phosphoenolpyruvate (compound 1) which is virtually not labeled from [1-13 C]glucose (Fig. 2). Thus, a direct conversion of 5-dehydroshikimate to gallic acid would be in line with the experimental data.
Attempts to use the same experimental approach for the study of gallic acid biosynthesis in leaves of the tree, R. typhina, were unsuccessful. The amount of 13 C-label diverted to gallic acid and phenylalanine was so low (less than 2% 13 C abundance in the most highly labeled carbon atoms) that it was impossible to draw firm conclusions. This is due to the fact that exogenous glucose supplied to the intact leaves via the leaf stem is poorly metabolized, and the NMR spectra are dominated by the natural 13 C abundance of metabolites, which were formed prior to glucose application. A much more sensitive technique was therefore required. As shown below, the sensitivity problem can be overcome by the use of a mixture of [U-13 C 6 ]glucose and unlabeled glucose. This experimental ap-proach will be described first with the fungal culture and then with plant material.
A shake flask culture of P. blakesleeanus was supplied with a mixture of [U-13 C 6 ]glucose and unlabeled glucose at a ratio of 1:25. The NMR signals of metabolites isolated from this culture are complex multiplets as a consequence of 13 C 13 C coupling (Figs. 4 and 5). To resolve these multiplets, it was important to obtain a spectral resolution of at least 0.5 Hz. The central signals marked by asterisks represent molecules with a single 13 C atom, which were formed from the natural abundance material in the proffered glucose mixture. The complex satellites indicate the presence of various isotopomers that reflect biosynthetic contributions of the totally labeled glucose. A detailed analysis of the spectra indicates that the biosynthetic products are complex mixtures of different isotopomers.
The satellite patterns can be resolved by numerical simulation of the 13 C 13 C coupling patterns for each respective isotopomer in the mixture. As shown in Fig. 4, the chemical shift positions as well as amplitude modulations due to higher order coupling show perfect agreement between the simulated and experimental spectra. It should be noted that the lack of symmetry of the coupling with respect to the signal of the singlelabeled isotopomers is due to nonlinear coupling patterns and to chemical shift variations caused by heavy isotope shift effects. The heavy isotope shifts of individual nuclei in specific isotopomers as deduced from comparison of experimental signals with spectral simulations are summarized in Fig. 6. The various isotopomers in the biolabeled mixture are symbolized by bold lines, indicating multiple contiguous carbon atoms. A single adjacent 13 C atom typically results in a high field shift in the range of 2 Hz; two adjacent 13 C atoms are conducive to a high field shift of the observed carbon in the range of 4 Hz.
Single bond and multiple bond 13 C 13 C coupling constants are also summarized in Fig. 6. It should be noted that the chemical shift degeneracy of the ring carbon atoms 3 and 5 is broken by the heavy isotope shift in isotopomer d (Fig. 6) where C-3 has a heavy isotope shift of Ϫ4.9 Hz and C-5 has a heavy isotope shift of Ϫ2.0 Hz. This results in higher order coupling between C-3 and C-5 which are no longer homotopic.
Only the coupling patterns without the actual simulations are shown in Fig. 5. However, it should be noted that the intensity modulations due to higher order coupling contributions are again faithfully reflected by the simulated data.
The fraction of each isotopomer in the mixture can be obtained by integration of the signal groups representing each individual isotopomer. Some isotopomers can be estimated from different spectral patterns. Thus, the abundance of isotopomer d can be diagnosed independently from the signatures of C-2, C-3, and C-4. The quantitative data obtained independently from different parts of the spectrum are in very good agreement.
The isotopomer composition of the biosynthetic gallic acid (compound 7) is summarized in Fig. 7. The quantitative contribution of each isotopomer is indicated by the width of the line connecting contiguous 13 C atoms and also by numbers which indicate relative concentrations (mol %) of individual isotopomers. Isotopomer d indicates that approximately onethird of the erythrose 4-phosphate pool is formed via the pen-tose phosphate shunt or mannitol cycling (Fig. 3). Isotopomer e indicates that approximately two-thirds of the erythrose 4phosphate pool is synthesized via futile cycling of the hexose phosphate pool. This is in line with the erythrose 4-phosphate labeling pattern from [1-13 C]glucose (see above).
The complex multiplets of biosynthetic tyrosine (compound 10) obtained from the biomass grown on the mixtures of [U-13 C 6 ]glucose and unlabeled glucose was analyzed by the same approach. The isotopomer composition is again summarized in Fig. 7. It is immediately obvious that the isotopomer signatures of the benzene rings are closely similar in gallic acid and tyrosine. This is a conclusive proof for the origin of gallic acid from the shikimate pathway.
Virtually no isotopomers with contiguous 13 C atoms in ring atom C-1 and the ␤ carbon of the side chain are present in the amino acid. The conspicuous absence of this isotopomer type reflects the assembly of the amino acid from the ring atoms of shikimate and the side chain contributed by a second molecule, i.e. phosphoenolpyruvate.
On the other hand, the isotopomer composition of gallic acid is dominated by the [carboxyl-1,6-13 C 3 ]isotopomer b (Fig. 7). If gallic acid were derived from phenylalanine or tyrosine by subsequent cleavage of the side chain, the presence of this isotopomer could not be explained. It follows again that gallic acid is not formed via phenylalanine or tyrosine in the fungus.
Experiments with Young Leaves of R. typhina-Initial exper-  Fig. 6. The respective isotopomers are shown schematically by bold lines connecting contiguously adjacent 13 C atoms, and the observed carbon is marked by an arrow. The amplitudes of the simulated spectra were adjusted to reflect the relative abundance of each respective isotopomer in the mixture. iments mentioned above had shown that the incorporation of [1-13 C]glucose in leaves of R. thyphina is low. The poor utilization of exogenous nutrients had also hampered earlier studies in the literature and is probably the major reason for the conflicting reports on gallic acid biosynthesis. The preliminary experiments with fungal cultures indicated that the sensitivity problem could be addressed by the use of totally labeled glucose conducing to the formation of multiply labeled isotopomers. The signal contributions of the multiply labeled isotopomers are spread out in the frequency domain by 13 C 13 C coupling and are thus separated from the signals of molecules representing the high background of preformed gallic acid with natural 13 C abundance.
A large number of preliminary experiments were required to find experimental conditions affording appropriate transfer of 13 C from exogenous glucose into the plant system. Variations in the experimental set-up concerned (i) the age of the plant tissue used as judged by the size of the leaves and the degree of unfolding, (ii) the conditions of incubation, (iii) the length of the feeding period, and (iv) the 13 C abundance in the glucose used as supplement. These experiments will be described in some detail because similar considerations may be relevant in the context of other incorporation experiments with higher plants. (i) It was important to utilize very young leaves since the utilization of exogenous glucose decreased rapidly with increasing age and maturation of the tissue. (ii) Standardized conditions of temperature and humidity had to be maintained to extend the lifetime of the plant tissue as much as possible. We found 18°C and 60% relatively humidity most appropriate for R. typhina leaves. Incubation in the dark resulted in higher 13 C incorporation rates as compared with permanent or intermittent light. (iii) Both the utilization of glucose in R. typhina, and the biosynthetic capacity of the plant material after removal from the mother plant was relatively poor. It was therefore to extend the feeding period as much as possible. With appropriate care to minimize microbial growth, the leaves could be kept for more than 1 month. (iv) The most crucial feature is the labeling pattern of the precursor material. Using single 13 Clabeled glucose the maximum 13 C enrichment values observed in any metabolite studied was always less than 2% in these experiments. On the background of 1.1% natural 13 C abundance, these small enrichment increments contributed by the 13 C labeled precursor could not be evaluated with sufficient accuracy. Thus, the strategy with single-labeled glucose had to be abandoned.
Subsequent experiments were performed with [U-13 C 6 ]-, [1,2-13 C 2 ]-, and [2,3-13 C 2 ]glucose with 99% enrichment. Whereas the overall enrichment of metabolites was again low (Ͻ2% 13 C), the incorporation of the precursors could be conclusively demonstrated by the presence of 13 C 13 C coupling satellites in the spectra of gallic acid and amino acids. Actually, in the experiment with [U-13 C 6 ]glucose, 13 C 13 C coupling was extensive throughout the molecules analyzed. In light of the low enrichment of the metabolites formed, this presented a paradox. In retrospect, it is now clear that the cut leaves retained little biosynthetic capacity, irrespective of the incubation conditions. The vast majority of the isolated metabolites had already been synthesized prior to the feeding experiments. On the other hand, the system had been flooded with the labeled precursor, and this had resulted in the assembly of three 13 Clabeled precursor molecules for formation of phenylalanine. Thus, less than 1% of the total gallic acid had been synthesized from the labeled precursor during the incubation period but the FIG. 6. 13 C NMR parameters of gallic acid. The carbon connectivity of gallic acid is shown schematically. Top, heavy isotope shifts (Hz) in individual isotopomers (a-g) containing 2 or more 13 C atoms; contiguous 13 C-labeled isotopomers are shown by bold lines. Bottom, 13 C chemical shifts (ppm) and 13 C 13 C coupling constants are indicated by arrows (Hz, italic numbers). Coupling between C-3 and C-5 occurs when the chemical shift degeneracy of these carbon atoms is broken by isotope shift effects. small amount of newly formed gallic acid was produced almost exclusively from exogenous glucose.
On basis of these initial data, we performed an optimized experiment with 150 young leaves of R. typhina (60 g) which were supplied with a solution (300 ml) containing 250 mg of [U-13 C 6 ]glucose and 6 g of natural abundance glucose. Gallic acid (50 mg) and amino acids were isolated by the standard procedure, and high resolution NMR spectra of gallic acid and tyrosine were recorded. The complex 13 C NMR signals of the carboxylic carbon and ring atom C-1 of gallic acid are shown in Fig. 8. In contrast to the experiments with the fungal cultures (Figs. 4 and 5), the relative intensity of the coupling patterns is low as compared with the contribution of isotopomers carrying a single 13 C atom (marked by asterisks), which represent the contribution of natural abundance gallic acid. The large fraction of natural abundance material essentially represents gallic acid which had already been present at the beginning of glucose application, whereas a minor amount was contributed biosynthetically from unlabeled glucose in the feeding mixture. Fig. 8 also shows the carboxyl-and C-1 13 C NMR signals of natural abundance gallic acid. The asterisks mark the signals of the [carboxyl-13 C 1 ]-and [1-13 C 1 ]gallic acid. The plots were arranged to give identical amplitudes for the (truncated) signals of the biosynthetic sample and the natural abundance sample. The stochastic 13 C distribution in the natural abundance sample is conducive to a signal equivalent of 1.1% of the central signal for isotopomers with two adjacent 13 C atoms (such as [1,2-13 C 2 ]-, and [carboxyl-1-13 C 2 ]gallic acid). These signals are readily observed in the natural abundance sample. The stochastic contribution of isotopomers with 3 and 4 adjacent carbon atoms in natural abundance material is approximately 10 Ϫ4 and 10 Ϫ6 mol %, respectively. The relative abundances of these species are therefore far below the level of detection in the natural abundance gallic acid. On the other hand, signals pertaining to multiply labeled isotopomers are apparent in the biosynthetic sample and can be identified unequivocally on basis of the isotopomer deconvolution which has been described above (Fig. 8).
A quantitative analysis of the biosynthetic gallic acid spectra yields the isotopomer composition shown in Fig. 9. The abundance of isotopomers with 2 and more adjacent 13 C isotopomers was referenced to the satellites for the various [ 13 C 2 ]isotopomers in the natural abundance sample. The natural abundance background was then subtracted from the labeled sample, thus affording exclusively the composition of the material which was biosynthesized de novo from the proffered [U-13 C 6 ]glucose.
The isotopomer composition of biosynthetic tyrosine was determined by the same experimental approach and is also shown in Fig. 9. The isotopomer patterns of the aromatic rings of biosynthetic gallic acid and tyrosine are similar and leave no doubt that the aromatic rings of both compounds originate from the same (i.e. the shikimate) pathway. Gallic acid shows a significant amount of the triple-labeled [carboxyl-1,6-13 C 3 ]isotopomer in close analogy with the Phycomyces experiment. This isotopomer has no equivalent in tyrosine where the abundance of the [␤,1,2-13 C 3 ] is low in accordance with the expectations based on the shikimate pathway. In analogy with the Phycomyces experiment, it can be concluded that phenylalanine and tyrosine do not serve as major intermediates in the gallic acid biosynthetic pathway. DISCUSSION Feeding with totally 13 C-labeled fundamental precursors such as [U-13 C]carbohydrates or [U-13 C]lipid in conjunction with unlabeled material is conducive to biosynthetic formation of complex isotopomer mixtures. In earlier studies, we used two-dimensional NMR techniques to characterize multiply 13 Clabeled isotopomers (24,25). In the present study, this approach would have been difficult for reasons of sensitivity because the incorporation of glucose by leaves of R. typhina was low. However, we have shown that an even more detailed analysis of isotopomer mixture is possible by an in depth analysis of coupling multiplets in one-dimensional 13 C spectra. For this approach, it is important to obtain NMR spectra at high magnetic field strength to maximize the chemical shift dispersion. Moreover, it is important to work at the maximum obtainable resolution (better than 0.5 Hz) to maximize sensitivity and to minimize overlapping of lines. Integrals of each component of the complex signals can then be determined easily. Typically, the NMR signatures of an individual carbon atom will arise by the 13 C coupling multiplets of several different isotopomers as shown in Figs. 4, 5, and 8. The intuitive deconvolution of the complex multiplet superpositions is possible only in the most simple cases. Complex patterns require a stringent analysis by numerical simulation of the spectra for each contributing isotopomer.
As part of the deconvolution process, it is necessary to determine the heavy isotope shift effects and the coupling constants for each isotopomer from the complex spectra. It should be noted that molecular symmetries conducive to homotopy of nuclei can be broken by heavy isotope shift effects, and unexpected coupling patterns of higher order can result.
Whereas this process is laborious, it has the advantage to yield quantitative data for each isotopomer in the mixture. This isotopomer pattern can then be used efficiently for biosynthetic retroanalysis via comparison of isotopomer distribution patterns in different metabolites. For example, the comparison of isotopomer patterns of the benzenoid ring of gallic acid with those of phenylalanine and tyrosine leave no doubt whatsoever that gallic acid is a derivative of the shikimate pathway.
The experimental sensitivity of biosynthetic studies with single-labeled 13 C precursors is limited by the inherent sensitivity of NMR instrumentation, and, more important, by the presence of natural abundance material in the numerous cases where the de novo formation of metabolites from the proffered, isotopocally labeled precursor is low. The use of multiply labeled precursors and the subsequent, quantitative NMR assessment of biosynthetic isotopomers with multiple labels affords a large gain in experimental sensitivity because the relevant signals are separated from the natural abundance component via spreading in the frequency domain by 13 C 13 C coupling.
As mentioned above, the stochastic abundance of multiply labeled isotopomers in natural abundance material is very low (approximately 10 Ϫ4 mol % for triple-labeled and 10 Ϫ6 mol % for quadruple-labeled species). As a consequence, any observed signals of triple-and quadruple-labeled isotopomers can be unequivocally addressed as the result of de novo biosynthesis from the multiply labeled precursor.
In the biosynthesis of tyrosine and phenylalanine, two molecules of phosphoenolpyruvate are consumed for the biosynthesis of part of the ring system (i.e. C-6 and C-1) via shikimate and of the side chain via the enoylpyruvate side chain of chorismate. It is therefore surprising that the isotope contribution of [U-13 C 6 ]glucose to the side chain of tyrosine in R. typhina exceeds the level of 13 C-labeling in the ring (specifically, the [1,6-13 C 2 ]isotopomer) by a factor of about 2. Earlier studies had shown that plant tissues can contain substantial amounts of shikimate pathway intermediates such as shikimate and quinate (26). We conclude from the labeling data that preformed early shikimate derivates were converted to chorismate and further to tyrosine during the [ 13 C]glucose feeding period. This would result in the formation of the aromatic amino acid from a preformed ring system at natural 13 C abundance by the addition of a enoylpyruvate side chain derived from the proffered 13 C-labeled glucose mixture.
The data in this report show clearly that gallic acid is formed predominantly or entirely from an early intermediate (i.e. prior to prephenic or arogenic acid) of the shikimate pathway. Several authors have proposed that two different pathways, via an early shikimate intermediate and via one of the aromatic amino acid could be operative in the same organism (5,(17)(18)(19)(20). More specifically, it has been claimed that gallic acid is predominantly formed via dehydrogenation of shikimate in young leaves of Rhus succedanea, but via phenylalanine in older leaves of R. succedanea (17). Our data show that a biosynthetic pathway via phenylalanine could have contributed less than 2% of biosynthesized gallic acid in the experiments with P. blakesleeanus and less than 10% in the experiments with R. typhina leaves.
The biosynthesis of gallic acid (compound 7, Fig. 10) by dehydrogenation of 5-dehydroshikimic acid (compound 4) in both the fungus and the plant appears to be the major pathway. However, the data cannot exclude a pathway via protocatechuic acid (compound 6) by dehydration of 5-dehydroshikimic acid followed by a monooxygenase catalyzed reaction (Fig. 10) (4). Supporting this hypothesis, a conversion of 5-dehydroshikimic acid to protocatechuic acid has been observed with crude enzyme preparations from mung bean seedlings (16). Experiments to distinguish between the two possible routes are in progress.