Phenylalanine-independent biosynthesis of 1,3,5,8-tetrahydroxyxanthone A retrobiosynthetic NMR study with root cultures of Swertia chirata Chang-Zeng Wang 1 , Ulrich H. Maier 1 , Michael Keil 2 , Meinhart H. Zenk 1 , Adelbert Bacher 3 , Felix Rohdich 3 and Wolfgang Eisenreich 3 1 Biozentrum-Pharmazie, Universita ¨ t Halle, Halle/Saale, Germany; 2 Boehringer Ingelheim Pharma KG, Ingelheim, Germany; 3 Lehrstuhl fu ¨ r Organische Chemie und Biochemie, Technische Universita ¨ tMu ¨ nchen, Garching, Germany Root cultures of Swertia chirata (Gentianaceae) were grown with supplements of [1- 13 C]glucose, [U- 13 C 6 ]glucose or [carboxy- 13 C]shikimic acid. 1,3,5,8-Tetrahydroxyxanthone was isolated and analysed by quantitative NMR analysis. The observed isotopomer distribution shows that 1,3,5,8- tetrahydroxyxanthone is biosynthesized via a polyketide- type pathway. The starter unit, 3-hydroxybenzoyl-CoA, is obtained from an early shikimate pathway intermediate. Phenylalanine, cinnamic acid and benzoic acid were ruled out as intermediates. Keywords: hydroxybenzoate; isotope labelling; NMR spectroscopy; retrobiosynthesis; xanthone. The shikimate pathway (Fig. 1) supplies building blocks for a wide variety of plant metabolites via phenylalanine (7), such as lignin, stilbenes and flavonoids, or via early pathway intermediates as specific precursors. Differentiation between these alternatives is possible by retrobiosynthetic analysis using general 13 C-labelled precursors (e.g. glucose; reviewed in [1]). Using this approach, we previously showed that the biosynthetic pathways of the tannic acid precursor, gallic acid (8), and the bitter compound, amarogentin (9), produced in Swertia chirata (Gentianaceae) branch off the shikimate pathway at a level before phenylpyruvate (6) (Fig. 1) [2,3]. We now exploit the retrobiosynthetic method to analyse the biosynthesis of a xanthone derivative. Xanthones are formed in at least 30 families of higher plants (e.g. Gentianaceae and Guttiferae) [4,5]. 1,3,5,8-Tetrahydroxy- xanthone (13, Fig. 2) is found in considerable amounts in the roots of S. chirata. A root culture of the latter plant has been used successfully in stable isotope incorporation experiments aimed at analysing the biosynthesis of amarogentin [3], and therefore this root culture appeared to be well suited for the present study. Early studies on the biosynthesis of xanthones suggested that the aromatic ring A (Fig. 2) is assembled via a polyketide-type pathway, whereas rings B and C are derived from a C 6 –C 1 benzoic acid moiety in a similar way to flavonoid biosynthesis [6]. More specifically, xanthones were proposed to be biosynthesized from hydroxybenzoyl- CoA and three molecules of malonyl-CoA (10)[7].An enzyme catalysing the condensation of 3-hydroxybenzoyl- CoA (11) and malonyl-CoA (10) to a benzophenone intermediate (12, Fig. 2A) has been isolated from Hyperi- cum androsaemum [8,9]. The biosynthetic origin of the hydroxybenzoyl-CoA precursor (11) of xanthones is subject to controversy. Thus, 14 C-labelled phenylalanine (7)wasreportedtobe incorporated into xanthones from Gentiana lutea, albeit with low incorporation rates [10], and label from cinnamic acid and benzoic acid was diverted to the xanthone, mangostin, in Garcinia mangostana [11,12]. Similarly, the benzoyl moiety of xanthone was claimed to be derived from L -phenylalanine in H. androsaemum via cinnamic acid, cinnamoyl-CoA, benz- aldehyde, benzoic acid and hydroxybenzoic acid [13,14]. These reports supported earlier data on the origin of C 6 –C 1 units via the phenylpropanoid pathway [15–17]. On the other hand, 14 C-labelled phenylalanine (7) and benzoic acid (15) were not incorporated into xanthones in cell cultures of Centaurium erythraea [13]. NMR experiments using 13 C-labelled samples of glucose and shikimate described in this paper show unequivocally that, in S. chirata, the xanthone precursor, 3-hydroxy- benzoyl-CoA, is formed from an early shikimate pathway intermediate (at a level before phenylpyruvate) and not from phenylalanine (7) via cinnamic acid or benzoic acid. Experimental procedures Materials [1- 13 C]Glucose and [U- 13 C 6 ]glucose were from Omicron (South Bend, IN, USA). [7- 13 C]Benzoic acid, [1- 13 C] bromoacetic acid and L -[U- 13 C 9 ]phenylalanine were from Correspondence to W. Eisenreich, Lehrstuhl fu ¨ r Organische Chemie und Biochemie, Technische Universita ¨ tMu ¨ nchen, Lichtenbergstr. 4, D-85747 Garching, Germany. Fax: + 49 89 28913363, E-mail: wolfgang.eisenreich@ch.tum.de Abbreviations: INADEQUATE, incredible natural abundance double quantum transfer experiment. (Received 14 March 2003, revised 30 April 2003, accepted 14 May 2003) Eur. J. Biochem. 270, 2950–2958 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03669.x Cambridge Isotope Laboratories (Andover, MA, USA). [U- 13 C 9 ]Cinnamic acid was prepared by treatment of L -[U- 13 C 9 ]phenylalanine with phenylalanine ammonia-lyase in 0.5 M sodium borate, pH 8.5. [carboxy- 13 C]Shikimic acid was synthesized from D -mannose and [1- 13 C]bromoacetic acid [18,19]. Incorporation experiments Root cultures of S. chirata were grown with supplements of [carboxy- 13 C]shikimate, [carboxy- 13 C]benzoate, [ring- 13 C 6 ]- cinnamic acid, [1- 13 C]glucose or a mixture of [U- 13 C 6 ]glu- cose and unlabelled glucose, as described previously [3]. Briefly, the cultures were grown in medium containing glucose instead of sucrose as carbon source without significant loss of viability or xanthone productivity. In the first experiment, the cultures were supplemented with a mixture of [1- 13 C]glucose and unlabelled glucose proffered at a ratio of 1 : 2.3 (w/w). Although this experiment could also have been performed with the labelled glucose as the only carbon source, the labelled compound was diluted with unlabelled glucose to reduce the cost. A second experiment was performed using a mixture of [U- 13 C 6 ]glucose and unlabelled glucose at a ratio of 1 : 20 (w/w). In this case, [U- 13 C 6 ]glucose as the only carbon source would afford uniformly 13 C-labelled products devoid of any biosynthetic information; therefore, [U- 13 C 6 ]glucose was diluted with unlabelled glucose (1 : 20) giving products with specific isotopomer compositions. The total concentration of glu- cose was 167 m M in these experiments. In further experi- ments, [carboxy- 13 C 1 ]shikimate, [carboxy- 13 C]benzoate or [ring- 13 C 6 ]cinnamic acid were proffered at concentrations of 0.5 m M , respectively, in medium containing 30 g glucose per litre. The cultures were incubated for 21 days. Isolation of 1,3,5,8-tetrahydroxyxanthone Plant material (fresh weight, 50 g) was pulverized under liquid nitrogen. The cold slurry was transferred to a flask and extracted three times with 200 mL methanol under a nitrogen atmosphere for 15 min. The slurry was filtered. The solution was concentrated to dryness under reduced pressure. The residue (500 mg) was applied to a column of silica gel (Silica Gel 60, 220–440 mesh, 20 · 1.8 cm; Merck, Darmstadt, Germany), which was developed with a mixture of chloroform and methanol (30 : 1, v/v). Fractions were combined and concentrated to dryness under reduced pressure. The residue was crystallised from methanol (yield, 30 mg). Fig. 2. Polyketide-type biosynthesis of 1,3,5,8-tetrahydroxyxanthone (13) with 3-hydroxybenzoyl-CoA (11) as starter unit. Fig. 1. Shikimate pathway as the source of phenylalanine and other plant metabolites. Equivalent positions originally derived from phos- phoenolpyruvate (1) and erythrose 4-phosphate (2)areindicatedbyred and green, respectively. The blue-coloured two-carbon fragment in phenylpyruvate (6) and phenylalanine (7) is obtained from a phos- phoenolpyruvate unit by decarboxylation of 5. Ó FEBS 2003 Biosynthesis of 1,3,5,8-tetrahydroxyxanthone (Eur. J. Biochem. 270) 2951 Isolation of amino acids In the experiments with 13 C-labelled glucose, the biomass, after extraction of the xanthone derivative, was hydrolysed, and phenylalanine was isolated as described previously [20]. The yield of phenylalanine isolated from 4 g root cells (dry weight) was 5 mg. NMR spectroscopy 1,3,5,8-Tetrahydroxyxanthone was dissolved in methanol- D 4 and phenylalanine was dissolved in 0.1 M DCl. 1 Hand 13 C NMR spectra were recorded at 500.13 MHz and 125.76 MHz, respectively, using a Bruker DRX500 spec- trometer. The data were processed with standard Bruker software ( XWINNMR 3.0). Two-dimensional incredible nat- ural abundance double quantum transfer experiments (INADEQUATE) [21] were performed with the Bruker pulse program INAD using a 135 ° read pulse (11.5 ls). Assessment of isotopomer composition The relative abundance of 13 C at specific positions of a given metabolite was calculated from the signal intensities in 1D 13 C NMR spectra (Fig. 3). Specifically, the signal integrals were determined for each 13 C-NMR signal of a metabolite from the labelling experiment and of the same compound at natural 13 C abundance [22]. The ratios of the signal integrals of the biolabelled compound and of the compound at natural abundance were then calculated for each respective carbon atom. Absolute 13 C abundances for certain carbon atoms (i.e. for carbon atoms with at least one attached hydrogen atom displaying a 1 H-NMR signal in a non- crowded region of the spectrum) were then determined from the 13 C coupling satellites in the 1 H-NMR spectra. As an example, the relative fraction of the 13 C-coupled satellites (J CH ¼ 160 Hz; Table 1) in the global intensity of the 1 H-NMR signal for H6 of 1,3,5,8-tetrahydroxyxanthone (13)(d ¼ 7.16 p.p.m.; Fig. 4) from the experiment with [U- 13 C 6 ]glucose accounted for 4.6% (Table 1). In other Fig. 3. 13 C-NMR spectrum of 1,3,5,8-tetrahydroxyxanthone. (A) Spectrum of a sample with natural 13 C abundance; (B) spectrum of a sample from the experiment with [1- 13 C]glucose. Asterisks indicate signals from impurities. Table 1. NMR data of 1,3,5,8-tetrahydroxyxanthone samples isolated from root cultures of S. chirata. Chemical shifts (p.p.m.) Coupling constants (Hz) Precursor [U- 13 C 6 ]Glucose [1- 13 C]Glucose [carboxy- 13 C]- Shikimate d 13 C d 1 H J HH J CC a J CH a % 13 C b % 13 C 13 C c % 13 C b % 13 C b 1 164.35 73.0(2), 62.8(8b) 4.6 28.9(2), 29.9(8b) 1.5 1.2 2 99.35 6.17(d) 2.2(4) 73.0(1), 66.3(3) 162(2) 4.2 31.1(1), 28.8(3) 6.2 d 1.1 3 168.05 66.8(4,2) 4.3 55.8(4, 2) 1.5 1.2 4a 159.36 74.7(4), 64.1(8b) 4.5 28.3(4), 31.2(8b) 1.4 1.2 4b 145.16 66.8(8a) 4.7 68.9(8a) 7.2 1.3 4 95.49 6.41(d) 2.2(2) 74.5(4a), 67.2(3) 168(4) 4.0 29.5(4a), 29.5(3) 6.0 1.0 5 138.33 69.0(6), 6.3(7) 4.3 16.8(6, 7) 1.9 1.2 6 124.68 7.16(d) 8.9(7) 69.2(5), 59(9) 160(6), 11(8) 4.6 15.5(5, 7), 45.2(7) 1.6 1.1 7 110.48 6.56(d) 8.8(6) 70.2(8), 59.5(6) 161(7) 4.7 11.2(8), 56.8(6, 8) 1.1 1.0 d 8 154.37 70.5(7) 4.3 65.8(7) 4.6 1.1 8a 108.71 67.0(4b), 55.5(9) 4.9 19.6(4b), 63.3(4b, 9) 1.2 1.1 8b 102.81 63.5(4a,1) 4.4 60.2(4a, 1) 5.5 1.1 9 185.80 55.7(8a) 4.5 56.4(8a) 1.5 14.3 a Determined from spectra of 13 C-enriched samples. Coupling partners are given in parentheses. b Absolute 13 C abundance. c Calculated as the fraction of 13 C coupled satellite pairs in the total signal intensity for a given carbon. Carbons coupled to the respective index carbon are indicated in parentheses. d Determined from 13 C satellites in the 1 H-NMR signal of the index atom. 2952 C Z. Wang et al.(Eur. J. Biochem. 270) Ó FEBS 2003 words, 4.6% of 13 isolated from the experiment with [U- 13 C 6 ]glucose contained 13 C at position 6. The relative 13 C abundances determined for all other positions in 13 were then referred to this value, thus obtaining absolute 13 C abundances for every carbon atom (% 13 CinTable1). In NMR spectra of multiply labelled samples displaying 13 C 13 C couplings, each satellite signal in the 13 C-NMR spectra was integrated separately. The relative fractions of each respective satellite pair in the total 13 C-NMR signal integral of a given carbon atom were then calculated (% 13 C 13 C in Table 1). These values were normalized to the 13 C abundances. For example, the signal intensities of the satellite signals reflecting [4b,8a- 13 C 2 ]-13 (sample from the experiment with [U- 13 C 6 ]glucose) accounted for  20% in the overall signal intensities for C-8a (Fig. 5 and % 13 C 13 C in Table 1). On the basis of the overall 13 C abundance of C-8a (4.9%), the molar contribution of [4b,8a- 13 C 2 ]-13 was calculated as 1.0 mol% (see Fig. 7A). Results As a prerequisite for the interpretation of labelling data by NMR spectrometry, unequivocal assignments of all signals are required. Although 13 C NMR signal assignments of 1,3,5,8-tetrahydroxyxanthone (13) were published on the basis of chemical-shift arguments [23], we independently assigned all signals by 2D carbon-carbon correlation experiments (INADEQUATE) (Fig. 6). This double-quan- tum filtered experiment reveals scalar couplings between adjacent 13 C atoms [21]; hence, only molecular species with adjacent 13 Catomsinthesamemoleculeareobserved. Because of the low abundance of doubly 13 C-labelled molecules in natural 13 C abundance samples (i.e. with two adjacent 13 C atoms in the same molecular species), the experiment is inherently insensitive. However, this draw- back can be overcome by the use of 13 C-enriched samples. Therefore, we analysed, by INADEQUATE spectroscopy, 1,3,5,8-tetrahydroxyxanthone from the experiment with [U- 13 C 6 ]glucose which had acquired intact blocks of 13 C-labelled atoms from [U- 13 C 6 ]glucose derived precursors (see below). Ten pairs of carbon signals detected in the spectrum provided a solid basis for the assignments summarized in Table 1. Some of the 1D 13 C-NMR signals of 1,3,5,8-tetra- hydroxyxanthone from the experiment with [U- 13 C 6 ]glucose are displayed in Fig. 5. The signals show satellites attributed to 13 C 13 C coupling involving one or two adjacent 13 C atoms, which indicate that metabolic precursors carrying two or more adjacent 13 C atoms have been incorporated into the biosynthetic product. Under the experimental conditions used, the proffered [U- 13 C 6 ]glucose is converted into a variety of multiply 13 C- labelled intermediary metabolites such as carbohydrate phosphates, pyruvate, and acetyl-CoA via the major glucose utilization pathways (glycolysis and pentose phosphate cycle). Simultaneously, intermediary metabolites are Fig. 4. 1 H-NMR signal for H-6 of 1,3,5,8-tetrahydroxyxanthone from the experiment with [U- 13 C 6 ]glucose. Satellite signals due to 1 H 13 C couplings are indicated. Fig. 5. 13 C-NMR signals of 1,3,5,8-tetrahydroxyxanthone from the experiment with [U- 13 C 6 ]glucose. Satellite signals due to 13 C 13 Ccou- plings are indicated. Fig. 6. Part of an INAEDQUATE spectrum of 1,3,5,8-tetrahydroxy- xanthone from the experiment with [U- 13 C 6 ]glucose. Ó FEBS 2003 Biosynthesis of 1,3,5,8-tetrahydroxyxanthone (Eur. J. Biochem. 270) 2953 produced with natural 13 C abundance from the natural abundance glucose proffered in large excess. Anabolic (biosynthetic) processes extract labelled and unlabelled molecules at random from the intermediary metabolite pools for utilization as building blocks. Natural products biosynthesized under the experimental conditions are therefore mosaics of labelled and unlabelled building blocks. Hence, they represent complex mixtures comprising a variety of 13 C-labelled isotopomers which can be present at relatively high abundance compared with their occur- rence in natural abundance material. A systematic decon- volution of the multiplets in the 13 C-NMR spectrum gives the molar fraction of each isotopomer that can be detected within the sensitivity limits of NMR spectroscopy. The multiply 13 C-labelled isotopomers of 1,3,4,8-tetra- hydroxyxanthone detected in the sample from the [U- 13 C 6 ]glucose experiment are summarized in Fig. 7A. Their relative abundance calculated from the fraction of each satellite pair in the global intensity of each respective signal (% 13 C 13 C in Table 1) can be referenced to the global absolute 13 C abundance for each carbon atom. This approach affords the molar fraction of each respective isotopomer. For the subsequent discussion, it is convenient to compress this information into a single icon (Fig. 7A) in which the labelling patterns of each observed multiply 13 C-labelled isotopomer are indicated by lines connecting the respective atom positions. The abundance in mol% is indicated numerically for each observed isotopomer and is also shown graphically by the relative width of the lines. The dots in Fig. 7A represent isotopomers with a single enriched carbon atom; the accompanying numbers indicate the mol% excess of the respective [ 13 C 1 ]-isotopomer above the natural 13 C abundance of 1.1%. A total of three single- labelled and 11 multiply 13 C-labelled isotopomers showed increased abundance compared with unlabelled material. The symmetric labelling pattern of ring A in the experiment with [U- 13 C 6 ]glucose implies that the biosyn- thetic pathway must involve a c 2 symmetric moiety which is free to rotate before giving rise to ring A. This result is in full accordance with the known polyketide-type pathway of xanthone biosynthesis [7–9] via the benzophenone inter- mediate 12 [8,9] comprising a c 2 symmetric trihydroxy- phenyl moiety (Fig. 2). A much simpler pattern of isotopomers was observed in xanthone biosynthesized from a mixture of [1- 13 C 1 ]glucose and natural abundance glucose. Signals of 1,3,5,8-tetra- hydroxyxanthone from the experiment with [1- 13 C]glucose are shown in Fig. 3B. The corresponding signals from a sample with natural 13 C abundance are shown for compar- ison in Fig. 3A. It is obvious that five carbon atoms (C8, C4b, C8b, C2 and C4) are significantly enriched in 13 C above the natural abundance 13 C level. Quantitative ana- lysis of the NMR signals (for details see Experimental procedures) afforded absolute 13 C abundance values for all carbon atoms of the target compound (Table 1, Fig. 7B). The 13 C abundances at C8, C4b, C8b, C2 and C4 varied between 4.6 and 7.2 mol%. The isotopomer compositions for proteinogenic phenyl- alanine from the experiments with [U- 13 C 6 ]glucose and [1- 13 C]glucose were determined as described above and are shown in Fig. 7D,E, respectively. The structural formulas have been oriented to match the labelling patterns of the amino acid with those of ring C of the xanthone derivative. In each experiment, it is immediately obvious that the labelling pattern of the aromatic ring in phenyl- alanine is closely similar to that of ring C in the xanthone Fig. 8. Reconstruction of the labelling patterns of phosphoenolpyruvate (1), erythrose 4-phosphate (2) and shikimate (3) from the observed labelling patterns of phenylalanine (7). (A) From the experiment with [1- 13 C]glucose. (B) From the experiment with [U- 13 C 6 ]glucose. The colours indicate equivalent positions biosynthetically derived from phosphoenolpyruvate (in red), erythrose 4-phosphate (in green) or the fragment of phosphoenolpyruvate after decarboxylation (in blue) (Fig. 1). Fig. 7. Labelling patterns of 1,3,5,8-tetrahydroxyxanthone (A–C) and phenylalanine (D and E). (A), (D) From the experiment with [U- 13 C 6 ]glucose; bold lines connect 13 C-labelled carbon atoms that were transferred from the same molecule of [U- 13 C 6 ]glucose; filled dots represent 13 C 1 -isotopomers with 13 C enrichments well above the nat- ural abundance contributions; numbers indicate 13 C enrichments in mol%.(B),(E)Fromtheexperimentwith[1- 13 C]glucose; filled dots indicate carbon atoms that acquired significant 13 C label; the numbers indicate 13 C abundances. (C) From the experiment with [carb- oxy- 13 C]shikimate; the filled triangle indicates the carbon atom that acquired significant 13 C label; the number indicates the 13 C abundance. 2954 C Z. Wang et al.(Eur. J. Biochem. 270) Ó FEBS 2003 derivative. It is also obvious that the similarity of the labelling does not extend further to include the b carbon of the tyrosine side chain and ring B of the tricyclic compound. As discussed in more detail below, the comparison between the isotopomer patterns shows con- clusively that the specific precursor of 1,3,5,8-tetra- hydroxyxanthone is a shikimate pathway intermediate before the level of phenylpyruvate. In line with that conclusion, the carboxylic group of shikimate is incorporated into the xanthone derivative as shown by the experiment with [carboxy- 13 C]shikimate. 1,3,5,8-[9- 13 C 1 ]Tetrahydroxyxanthone was found with an abundance of 14.3 mol% (Fig. 7C). No excess 13 C abun- dance was observed in 1,3,5,8-tetrahydroxyxanthone from experiments with [carboxy- 13 C]benzoate and [ring- 13 C 6 ]cin- namic acid as precursors (data not shown). Discussion The competing biosynthetic hypotheses proposed in the literature for the biosynthesis of ring C of 1,3,5,8-tetra- hydroxyxanthone can be subjected to a rigorous test by prediction of the 1,3,5,8-tetrahydroxyxanthone labelling pattern from the labelling patterns of central metabolic intermediates. In the present case, it is sufficient to derive the labelling patterns of erythrose 4-phosphate (2) and phos- phoenolpyruvate (1) by dissection of the phenylalanine labelling patterns according to the shikimate pathway, the universal source of the carbon skeletons of aromatic acids. The labelling patterns of phenylalanine (7) obtained by hydrolysis of cell mass from the experiment with [1- 13 C]glu- cose (Fig. 8A) or [U- 13 C 6 ]glucose (Fig. 8B) are qualitatively similar to those found with other plants [1,2]. The side chain reflects the labelling pattern of phosphoenolpyruvate from which it is biosynthetically obtained via the shikimate pathway of aromatic amino-acid biosynthesis (Fig. 1). The aromatic ring of phenylalanine reflects the labelling patterns of C2–C3 of phosphoenolpyruvate and of C1–C4 of erythrose 4-phosphate. Owing to the symmetry of the ring, the ortho and meta carbon atoms become pairwise homo- topic in the experiment with [1- 13 C]glucose. As a conse- quence, only an averaged value of 5.4% 13 C can be obtained for the ortho ring carbon atoms (reflecting C3 of phos- phoenolpyruvate and C4 of erythrose 4-phosphate, respect- ively). However, as the 13 C abundance for C3 of phosphoenolpyruvate can be gleaned from the b-carbon atom in the side chain (i.e. 5.8% 13 C), the enrichment for C4 of erythrose 4-phosphate can be determined as 5.0% 13 C from the average value. The deduced labelling patterns of the intermediary metabolites can then be used to reconstruct the labelling patterns of shikimate (3) in each respective experiment (Fig. 8A,B). Fig. 9. Labelling pattern of the B and C ring of 1,3,5,8-tetrahydroxyxanthone (13) from the experiment with [1- 13 C]glucose. (A) Prediction via phenylalanine (7), benzoic acid (14) and 3-hydroxybenzoic acid (15). (B) Prediction via an early shikimate pathway intermediate (e.g. 3); filled dots indicate 13 C label; numbers indicate predicted 13 C abundances; the 13 C abundances of phenylalanine carbon atoms were determined experimentally (see Fig. 7E); the 13 C composition of shikimate (3) was reconstructed from that of phenylalanine (7) by retrobiosynthetic analysis (see Fig. 8A). (C) Experimentally determined. For other details, see also legends to Figs 7 and 8. Ó FEBS 2003 Biosynthesis of 1,3,5,8-tetrahydroxyxanthone (Eur. J. Biochem. 270) 2955 The experimentally observed labelling patterns of phenyl- alanine (7) and the reconstructed labelling patterns of shikimate (3) were then used to predict hypothetical labelling patterns of 1,3,5,8-tetrahydroxyxanthone (13)via different biosynthetic routes. Specifically, Figs 9A and 10A show the predictions of the labelling patterns of the xanthone derivative via phenylalanine (7) and benzoic acid (14). On the other hand, Figs 9B and 10B show the prediction of the labelling pattern of the xanthone derivative 13 via an early shikimate intermediate (such as 3)actingas the specific biosynthetic precursor. The predicted labelling patterns via phenylalanine were at odds with the observed isotopomer compositions in 13 (Figs 9C and 10C), whereas the prediction via shikimate almost perfectly matched the detected patterns. Therefore, the phenylalanine hypothesis must be abandoned. Hence, 1,3,5,8-tetrahydroxyxanthone biosynthesized in S. chirata joins the growing list of secondary plant meta- bolites that are derived from an early shikimate derivative as opposed to a pathway via phenylalanine and cinnamate. A hypothetical mechanism for the conversion of shikimic acid (3) into 3-hydroxybenzoate (15) is shown in Fig. 11 [3]. We propose that the vinylogous elimination of phosphate from shikimic acid 3-phosphate (16) affords the dihydroxy-diene intermediate 17 which appears to be well suited for a subsequent dehydration yielding the aromatic ring system of 3-hydroxybenzoate (15). The CoA ester of 15 could then provide the starter unit for the downstream steps of the polyketide-type biosynthesis of 13. In line with this interpretation, the experiment with [carboxy- 13 C 1 ]shikimate shows that the carboxylic group, which is lost in the formation of phenylalanine and tyrosine (Fig. 1) is in fact incorporated into ring B of 1,3,5,8- tetrahydroxyxanthone (Figs 7C and 11). Phenylalanine, tyrosine and metabolites downstream from the amino acids would not have been advantageous to that labelling pattern. By comparison with the [ 13 C]glucose incorporation studies, the data structure in the experiment with [ 13 C]shiki- mate is simple. This prompts the question whether the incorporation experiments with [ 13 C]glucoses and the meti- culous deconvolution of the complex NMR spectra were really necessary to solve the biosynthetic problem. The answer lies in the difference between quantitative and qualitative analysis. On closer examination, the shikimate experiment shows only that at least a certain fraction of the xanthone derivative was obtained from a shikimate inter- mediate before the biosynthetic level of phenylpyruvate. On the other hand, the comparison of the labelling patterns of phenylalanine and the xanthone derivative from the experi- ments with 13 C-labelled glucose shows, on the basis of quantitative data, that the xanthone derivative is obtained predominantly or exclusively from an early shikimate derivative and not via phenylalanine. More specifically, in the experiment with [U- 13 C 6 ]glucose, several isotopomers of Fig. 10. Labelling pattern of the B and C ring of 1,3,5,8-tetrahydroxyxanthone (13) from the experiment with [U- 13 C 6 ]glucose. (A) Prediction via phenylalanine (7), benzoic acid (14) and 3-hydroxybenzoic acid (15). (B) Prediction via an early shikimate pathway intermediate, such as shikimate (3); bold lines connect 13 C-labelled carbon atoms that are transferred from the same molecule of [U- 13 C 6 ]glucose; filled dots represent 13 C 1 isotopomers with 13 C enrichment significantly above the natural abundance contributions; numbers indicate 13 C enrichments in mol%; the isotopomer composition of phenylalanine was determined experimentally (see Fig. 7D); the isotopomer composition of shikimate (3)was reconstructed from that of phenylalanine (7) (see Fig. 8B). (C) Experimentally determined. For other details, see legend of Fig. 8. 2956 C Z. Wang et al.(Eur. J. Biochem. 270) Ó FEBS 2003 13 predicted to be synthesized by a route via phenylalanine (7) could not be detected (for example [8a,8- 13 C 2 ]-13 and [4b,5,6- 13 C 3 ]-13) (Fig. 10A,C). On the basis of the sensitivity and the error limits of our NMR analysis, it can be concluded that at least 98% of the biosynthetic xanthone derivative was obtained from a shikimate pathway inter- mediate before the biosynthetic level of phenylpyruvate. This is confirmed independently by the experiment with [1- 13 C]glucose. Thus, the retrobiosynthetic method affords stringent, quantitative answers even in cases where a given metabolite can be biosynthesized via more than one metabolic pathway, which is by no means an esoteric or farfetched problem. The recent history of terpene biosyn- thesis research provides a striking example. Besides the well- known mevalonate pathway, plants were recently shown to operate a second pathway for the biosynthesis of isoprenoid precursors via 1-deoxyxylulose 5-phosphate (for review, see [24]). As a result of crosstalk between these pathways ([25] and references cited therein), plant terpenes typically comprise precursors from both pathways. Even though the contribution of the mevalonate pathway to the vast majority of the more than 20 000 plant terpenes is rather small, incorporation experiments with isotope-labelled mevalonate had been interpreted, incorrectly, in terms of a universal mevalonate origin for all plant terpenes. Only recently has retrobiosynthetic analysis shown that, in contrast with this assumption, most plant terpenes are obtained predominantly via the nonmevalonate pathway. Acknowledgements This work was supported by the Deutsche Forschungsgemeinschaft,the Fonds der Chemischen Industrie and the Hans-Fischer-Gesellschaft. The expert help of Angelika Werner and Fritz Wendling with the preparation of the manuscript is gratefully acknowledged. References 1. Bacher,A.,Rieder,C.,Eichinger,D.,Fuchs,G.,Arigoni,D.& Eisenreich, W. (1999) Elucidation of biosynthetic pathways and metabolic flux patterns via retrobiosynthetic NMR analysis. FEMS Microbiol. Rev. 22, 567–598. 2. Werner, I., Bacher, A. & Eisenreich, W. (1997) Formation of gallic acid in plants and fungi. A retrobiosynthetic study with 13 C-labelled glucose. J. Biol. Chem. 272, 25474–25482. 3. Wang, C Z., Maier, U.H., Eisenreich, W., Adam, P., Obersteiner, I., Keil, M., Bacher, A. & Zenk, M.H. (2001) Unexpected bio- synthetic precursors of amarogentin. A retrobiosynthetic 13 C NMR study. Eur. J. Org. Chem. 1459–1465. 4. Hostettmann, K. & Hostettmann, M. (1989) Xanthones. In Methods in Plant Biochemistry, Vol. 1. Plant Phenolics (Harborne, J.B., ed.), pp. 493–508. Academic Press, London. 5. Hegnauer, R. & Hegnauer, M. (1992) Chemotaxonomie der Pflanze,Vol.X,pp.463.Birha ¨ user-Verlag, Basel. 6. Floss, H.G. & Rettig, A. (1964) Biosynthesis with Gentianaceae. I. Biosynthesis of gentisin. Z. Naturforsch. 19b, 1103–1105. 7. Zenk, M.H. (1979) Recent work on cinnamoyl CoA derivatives. In Recent Advances in Phytochemistry,Vol.12Biochemistry of Plant Phenolics. (Swain, T., Harborne, J.B. & van Sumre, C.F., eds), pp. 139–176. Plenum Publishing Corporation, New York. 8. Beerhues, L. (1996) Benzophenone synthase from cultured cells of Centaurium erythraea. FEBS Lett. 383, 264–266. 9. Schmidt, W. & Beerhues, L. (1997) Alternative pathways of xan- thone biosynthesis in cell cultures of Hypericum androsaemum L. FEBS Lett. 420, 143–146. 10. Lewis, J.R. & Gupta, P. (1971) Biosynthesis of xanthones in Gentiana lutea. J.Chem.Soc.C, 629–631. 11. Bennet, G.J. & Lee, H H. (1988) The biosynthesis of mangostin: the origin of the xanthone skeleton. J. Chem. Soc. Chem. Com- mun., 619–620. 12. Bennet, G.J., Lee, H H. & Das, N.P. (1990) Biosynthesis of mangostin. Part 1: the origin of the xanthone skeleton. J. Chem. Soc. Perkin. Trans. I, 2671–2676. 13.AbdEl-Mawla,A.M.A.,Schmidt,W.&Beerhues,L.(2001) Cinnamic acid is a precursor of benzoic acids in cell cultures of Hypericum androsaemum L. but not in cell cultures of Centaurium erythraea RAFN. Planta 212, 288–293. 14. Abd El-Mawla, A.M.A. & Beerhues, L. (2002) Benzoic acid bio- synthesis in cell cultures of Hypericum androsaemum. Planta 214, 727–733. 15. Zenk, M.H. & Mu ¨ ller, G. (1964) Biosynthese von p-Hydroxy- benzosa ¨ ure und anderer Benzoesa ¨ uren in ho ¨ heren Pflanzen. Z. Naturforsch. 19b, 398–405. 16. Zenk, M.H. (1965) Biosynthesis von Vanillin in Vanilla planifolia Andr. Z. Pflanzenphysiol. 53, 404–414. Fig. 11. Hypothetical mechanism of 3-hydroxybenzoate formation in S. chirata. Label from [carboxy- 13 C]shikimate (3) is indicated by a filled triangle. Ó FEBS 2003 Biosynthesis of 1,3,5,8-tetrahydroxyxanthone (Eur. J. Biochem. 270) 2957 17. Zenk, M.H. (1967) Pathways of salicyl alcohol and salicin for- mation in Salix purpurea. Phytochemistry 6, 245–252. 18. Fleet, G.W.J., Shing, T.K.M. & Warr, S.M. (1984) Enantio- specific synthesis of shikimic acid from D -mannose: formation of a chiral cyclohexene by intramolecular olefination of a carbo- hydrate-derived intermediate. J. Chem. Soc. Perkin Trans. 1, 905–908. 19.Cho,H.,Heide,L.&Floss,H.G.(1992)Synthesisof D -())- [1,7- 13 C 2 ]shikimic acid. J. Labelled Compounds Radiopharm. 31, 589–594. 20. Eisenreich, W., Schwarzkopf, B. & Bacher, A. (1991) Biosynthesis of nucleotides, flavins, and deazaflavins in Methanobacterium thermoautotrophicum. J. Biol. Chem. 266, 9622–9631. 21. Bax, A., Freeman, R. & Frenkiel, T.A. (1981) An NMR technique for tracing out the carbon skeleton of an organic molecule. J. Am. Chem. Soc. 103, 2102–2104. 22. Eisenreich, W. & Bacher, A. (2000) Elucidation of biosynthetic pathways by retrodictive/predictive comparison of isotopomer patterns determined by NMR spectroscopy. Genetic Engineering, Principles and Methods (Setlow, J. K., ed.), Vol. 22, pp. 121–153. Kluwer Academic/Plenum Publisher, New York. 23. Zhou, H M., Liu, Y L., Blasko, G. & Cordell, G.A. (1989) Swertiabisxanthone-I from Swertia macrosperma. Phytochemistry 28, 3569–3571. 24. Eisenreich, W., Rohdich, F. & Bacher, A. (2001) The deoxy- xylulose phosphate pathway to terpenoids. Trends Plant Sci. 6, 78–84. 25. Schuhr, C.A., Radykewicz, T., Sagner, S., Latzel, C., Zenk, M.H., Arigoni, D., Bacher, A., Rohdich, F. & Eisenreich, W. (2003) Quantitative assessment of crosstalk between the two isoprenoid biosynthesis pathways in plants by NMR spectroscopy. Phy- tochem. Rev. in press. 2958 C Z. Wang et al.(Eur. J. Biochem. 270) Ó FEBS 2003 . Phenylalanine-independent biosynthesis of 1,3,5,8-tetrahydroxyxanthone A retrobiosynthetic NMR study with root cultures of Swertia chirata Chang-Zeng. the biosynthetic pathways of the tannic acid precursor, gallic acid (8), and the bitter compound, amarogentin (9), produced in Swertia chirata (Gentianaceae) branch off