Further evidence for an alternative, nonmevalonoid pathway of terpenoid biosynthesis was obtained by Arigoni and coworkers (6), who observed that Menyanthes trifoliata incorporates [2-¹⁴C]mevalonate into triterpenes and sterols, but not into the monoterpene glycoside, loganin. Additional observations that could not be accommodated within the framework of the mevalonate pathway have been reported by several research groups. Various interpretations, such as compartmentalization of nonexchanging acetate pools as well as terpene biosynthesis pathways via valine or leucine as precursors, have been proposed (7–11).

More recent studies by Arigoni and his research group established 1-deoxy-d-xylulose (hereafter designated deoxyxylulose) as the committed precursor of the alternative pathway in plants and bacteria (12, 13, §). The carbohydrate was incorporated into menaquinone with high efficacy by Escherichia coli and was also shown to serve as precursor for ginkgolides in Ginkgo biloba, whereas steroids were biosynthesized in the same plant via the mevalonate pathway. Deoxyxylulose had been isolated earlier from Streptomyces hygroscopicus (14, 15) and had been shown to serve as a biosynthetic precursor for thiamin (16, 17) and pyridoxal (18, 19).

Labeling data of the ginkgolides and of the E. coli terpenoids obtained in vivo from labeled glucose could be explained by formation of the deoxypentulose (or its 5-phosphate) from glyceraldehyde and active acetaldehyde generated in the thiamin pyrophosphate catalyzed decarboxylation of pyruvate (12, 13, §). The labeling patterns were identical with those detected independently by Rohmer, Sahm, and coworkers for the formation of hopanoids and other terpenes in various bacteria (20, 21).

Meanwhile, it has been shown that the deoxyxylulose pathway is also involved in higher plants in the biosynthesis of diterpenes in Salvia miltiorrhiza (A. Cartayrade, M. K. Schwarz, and D.A., unpublished data), of taxoids in Taxus chinensis (22), of essential oils in Mentha piperita (23), of chlorophyll and carotenoids in Lemna gibba, Hordeum vulgare, Daucus carota (24), and Scenedesmus obliquus (25), and of isoprene in higher plants (26).

This paper describes studies with single and multiple ¹³C-labeled deoxyxylulose using a photosynthetically active culture of Catharanthus roseus. The data show that the pentulose is efficiently incorporated into carotenoids and phytol, and that the pathway involves a skeletal rearrangement rather than a fragmentation/reassociation process.

Plant Material. Photomixotrophic suspension cultures of C. roseus were maintained for more than 10 years on Linsmaier and Skoog medium (27) containing 3% glucose instead of sucrose. The cells were grown under continuous light conditions (2,400 lx) on a rotary shaker (100 rpm) at 28°C. Ten 1-liter flasks, each containing 250 ml of medium, 11 mg of labeled deoxyxylulose, and 2.5 g (dry weight) of C. roseus cells were incubated for 7 days.

Preparation of Labeled Precursors. [U-¹³C₆]Glucose, [2-¹³C]pyruvate, and [3-¹³C]pyruvate (99% ¹³C abundance) were purchased from Isotec. [U-¹³C₃]Glyceraldehyde was prepared from [U-¹³C₆]glucose (28). [1-¹³C]1-Deoxy-d-xylulose was prepared by enzymatic condensation of [3-¹³C]pyruvate with unlabeled d-glyceraldehyde (29, 30). [2,3,4,5-¹³C₄]Deoxyxylulose was prepared similarly from [2-¹³C]pyruvate and [U-¹³C₃]d-glyceraldehyde. The following ¹³C NMR data were obtained for [2,3,4,5-¹³C₄]deoxyxylulose in dimethyl sulfoxide-d₆: C-1, 26.7 ppm; C-2, 211.6 ppm (¹³C coupling to C-3, 42.3 Hz; ¹³C coupling to C-5, 2.4 Hz); C-3, 77.0 ppm (¹³C coupling to C-4, 39.8 Hz); C-4, 72.5 ppm (¹³C coupling to C-5, 42.5 Hz); C-5, 61.9 ppm.

Isolation of Terpenoids. C. roseus cells (1 kg wet weight) were harvested, frozen, thawed, and extracted exhaustively with acetone and a mixture of acetone/hexane (1:1 vol/vol). The extract was reduced to a volume of 200 ml, and the resulting two-phase mixture was extracted with hexane and chloroform. This final organic phase was then evaporated to a volume of 5 ml under reduced pressure. The solution was applied to a column of silica gel 60 (220–440 mesh; column size, 40 × 6 cm), which was subsequently developed with a mixture of hexane/ethyl acetate (1:1 vol/vol). Fractions containing β-carotene, lutein, chlorophyll, and phytosterols, respectively, were combined and taken to dryness.

The crude carotene fraction was dissolved in a mixture of hexane/chloroform (5:1 vol/vol) and chromatographed on a column of silica gel 60 (column size, 30 × 3 cm) using the same solvent mixture as eluent. The fraction containing β-carotene was collected and taken to dryness (yield, 3 mg).

The crude lutein fraction was chromatographed on a column of silica gel 60 (column size, 30 × 3 cm) using ligroin/acetone/chloroform (5:5:4 vol/vol) as eluent. The lutein fractions were combined and taken to dryness (yield, 7 mg).

The fractions containing chlorophyll a, chlorophyll b, and phytosterols were combined. The solution was taken to incipient dryness. The residue was boiled under reflux in methanol containing 3% KOH for 30 min. The reaction mixture was extracted with petrol ether (40–60°C). The petrol ether phase was washed to neutrality with water, dried over sodium sulfate, taken to dryness, and applied to a column of silica gel 60 (column size, 24 × 3 cm). The column was developed with a mixture of CH₂Cl₂/ethyl acetate (9:1 vol/vol). Fractions containing phytosterols were combined and taken to dryness (yield, 95 mg). The phytol fraction was further resolved on a second column of silica gel 60 (column size, 15 × 2.5 cm) using hexane/ethyl acetate (85:15 vol/vol) as eluent. The phytol fractions were combined and taken to dryness (yield, 38 mg).

NMR Spectroscopy. ¹H and ¹³C NMR spectra of phytol, β-carotene, lutein, and sitosterol were recorded in CDCl₃ using a Bruker DRX 500 spectrometer equipped with an ASPECT station. ¹³C NMR spectra were measured as follows: 45° pulse (3 μs); repetition time, 3.2 s; spectral width, 29 kHz; data set, 64 kilo-words; temperature, 17°C; zero-filling to 128 kilo-words before to Fourier transformation; Gaussian apodization; ¹H decoupling by WALTZ 16 during acquisition and relaxation. Two-dimensional experiments and data processing routines were performed according to standard Bruker software (xwinnmr 1.1).

¹³C signal assignments were confirmed by two-dimensional INADEQUATE experiments performed with metabolites obtained from the incorporation experiment with [2,3,4,5-¹³C₄]deoxyxylulose. ¹³C abundance in terpenoids was analyzed by quantitative NMR spectroscopy. ¹H-decoupled ¹³C NMR spectra of samples from incorporation experiments and of samples with natural ¹³C abundance (1.1% ¹³C) were recorded under identical conditions. Relative ¹³C abundance of individual carbon atoms was then calculated by comparison of ¹³C signal integrals between ¹³C-labeled and unlabeled material. To determine absolute ¹³C abundance, ¹³C-coupled satellite signals in ¹H NMR spectra were analyzed. The absolute ¹³C enrichment was then used to standardize the relative ¹³C abundance of all carbon atoms. The fraction of multiply labeled isotopomers was calculated from ¹H-decoupled ¹³C NMR spectra as the fraction of ¹³C¹³C-coupled satellites in the global ¹³C signal of the respective carbon atom.

A cell culture of C. roseus was grown in presence of 0.3 mM [1-¹³C]- or [2,3,4,5-¹³C₄]deoxyxylulose. The growth rate and the biosynthesis of terpenoids were not affected by the deoxyxylulose supplement. β-Carotene, lutein, and sitosterol were isolated as described and were analyzed by one- and two-dimensional NMR spectroscopy. Chlorophyll a and b were hydrolyzed to yield trans-phytol for NMR analysis.

The ¹³C abundance of individual carbon atoms in the terpenoid metabolites was determined as described. In the experiment with [1-¹³C]deoxyxylulose, the methyl carbons C-17, C-18, C-19, and C-20 of trans-phytol showed high ¹³C enrichment (Table 1; average ¹³C abundance, 17.7 ± 1.2%; labeled positions shown as solid circles in Fig. 1), whereas the other atoms were virtually not labeled (average ¹³C abundance, 1.2 ± 0.2%).

Table 1 13C NMR analysis of trans-phytol from a cell culture of C. roseus supplied with [1-13C]deoxyxylulose or[2,3,4,5-13C4]deoxyxylulose

Figure 1 Structure (A) and 13C labeling pattern (B) of trans-phytol obtained by saponification of chlorophyll. Solid circles represent carbon atoms that acquired label from [1-13C]deoxyxylulose. Bold lines connect carbon atoms that were transferred (more ...)

It is well established that phytol is formed from isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) via geranylgeranyl pyrophosphate. Each of the labeled atom positions C-17, C-18, C-19, and C-20 is biosynthetically equivalent to C-5 of IPP or DMAPP. It follows that C-1 of deoxyxylulose is specifically incorporated into C-5 of IPP and DMAPP, respectively. It should be noted that the carbon atoms C-4, C-12, and C-16 show an average ¹³C abundance of 1.6 ± 0.1%, well above the average abundance of the other “unlabeled” carbon atoms (1.1 ± 0.1%). For the corresponding C-8, the signal could not be evaluated. The three carbon atoms with slightly elevated ¹³C abundance are biosynthetically equivalent to C-4 of IPP/DMAPP. The increased ¹³C abundance of these carbon atoms reflects the somewhat imperfect stereocontrol of the enzyme, IPP/DMAPP isomerase.

Label from [1-¹³C]deoxyxylulose was also specifically incorporated into all but two methyl positions of β-carotene (average ¹³C abundance of labeled positions, 16.3 ± 0.3%) and lutein (average ¹³C abundance of labeled positions, 17.2 ± 1.8%). The labeled methyl groups are shown in Fig. 2C as solid circles. The labeled positions were again biosynthetically equivalent to C-5 of IPP and DMAPP. In light of the relatively low concentration of deoxyxylulose in the culture medium (0.3 mM), the high ¹³C enrichment values in trans-phytol and carotenoids suggest that deoxyxylulose is a precursor with close metabolic proximity to IPP and DMAPP.

Figure 2 Structure of β-carotene (A) and lutein (B). The same labeling pattern (C) was obtained for both compounds. Arrow indicates the carbon atom documented by the NMR patterns in Fig. 3 B and C. For additional details see the legend to Fig. 1.

In the experiment with [2,3,4,5-¹³C₄]deoxyxylulose, ¹³C label was diverted to all positions of phytol (Table 1) and carotenoids (data not shown), with the exception of those that were labeled in the experiment with [1-¹³C]deoxyxylulose (Figs. 1 and 2). The average ¹³C enrichment of the labeled positions was 8.7 ± 0.4% in trans-phytol, 7.1 ± 0.7% ¹³C in β-carotene, and 7.0 ± 0.7% ¹³C in lutein. As shown in Fig. 3, ¹³C signals of trans-phytol, β-carotene, and lutein were characterized by intense ¹³C-coupled satellites (approximately 85% of signal intensity relative to the overall ¹³C signal intensity). As shown below, these satellite signals arise by ¹³C¹³C coupling resulting from joint incorporation of ¹³C atoms from [2,3,4,5-¹³C₄]deoxyxylulose into C-1, C-2, C-3, and C-4 of IPP and DMAPP. The uncoupled central signals marked by asterisks in Fig. 3 represent isotopomers with a single ¹³C atom, which were formed from the natural ¹³C abundance glucose in the culture medium.

Figure 3 NMR patterns of metabolites derived from [2,3,4,5-13C4]deoxyxylulose. (A) C-2 of phytol. (B) C-14 of β-carotene. (C) C-14 of lutein. (D) C-5 of sitosterol. Signals of natural 13C abundance isotopomers biosynthesized from proffered (more ...)

The complex satellites were analyzed by deconvolution of the multiplets using ¹³C¹³C coupling constants that had been unequivocally established by two-dimensional INADEQUATE spectroscopy using the metabolites from the [2,3,4,5-¹³C₄]deoxyxylulose experiment. As shown in Fig. 3A, the NMR signature of C-2 of phytol (arrow in Fig. 1B) consists of nine lines. This carbon atom is biosynthetically equivalent to C-2 of the terpenoid precursors. The satellite signals arise by one bond coupling of the observed carbon to two adjacent carbon atoms (coupling constants of 72.7 and 47.3 Hz, respectively, Table 1) and by long-range ¹³C¹³C coupling via two bonds (2.4 Hz). The two-dimensional INADEQUATE (not shown) reveals unequivocally that the long-range coupling is to carbon 4. All four contiguous carbon atoms are part of one isoprenoid moiety (Fig. 1). The coupled signals are not arranged symmetrically with respect to the signal from the natural abundance material marked by an asterisk in Fig. 3A. The substantial upfield shift (4.6 Hz) of the coupled signals is caused by heavy isotope shifts of the neighboring ¹³C atoms (α shifts by carbons 3 and 1 and β shift by carbon 4). It should be noted that only two isotopomers are observed in Fig. 3A for the segment under investigation, namely [2-¹³C₁]phytol (natural abundance metabolite from unlabeled glucose) and [1,2,3,4-¹³C₄]phytol. Other isotopomers, if present, would have yielded additional signals at the positions marked by arrows in Fig. 3A.

The arguments used for interpretation of Fig. 3A can likewise be applied for the interpretation of Figs. 3 B and C. The carbon atoms under specific investigation in carotene and lutein correspond to the one marked by an arrow in Fig. 2C. Each of these carbon atoms is biosynthetically equivalent to C-2 of IPP/DMAPP. The observed carotenoid carbons in Figs. 3 B and C show coupling to three other ¹³C atoms, which are all biosynthetically derived from the same isoprenoid precursor molecule. In close agreement with the observations on phytol, the data confirm that only the [1,2,3,4-¹³C₄]isotopomers of DMAPP and IPP have been formed biosynthetically from the proffered [2,3,4,5-¹³C₄]deoxypentulose.

An isoprenoid unit of β-carotene is also shown in the two-dimensional INADEQUATE spectrum in Fig. 4. Carbon atoms 16, 1, 2, and 3 are shown to be connected via single bond ¹³C¹³C coupling. The signals for C-1 and C-2 appear as doublets as a consequence of their simultaneous coupling to two adjacent ¹³C atoms via single bonds (coupling constants, approximately 35 Hz).

Figure 4 Section of INADEQUATE spectrum of β-carotene biosynthesized from [2,3,4,5-13C4]deoxyxylulose. (Inset) The relevant part of the β-carotene structure. Contiguous 13C atoms are connected by bold lines.

According to the data in Fig. 3D and Table 2, [2,3,4,5-¹³C₄]deoxyxylulose was also incorporated into sitosterol as indicated in Fig. 5. The ¹³C signal of C-5 of sitosterol, which is biosynthetically equivalent to C-2 of IPP/DMAPP, comprises four satellite signals, whereas the carbon atoms in phytol (Fig. 3A) and carotenoids (Fig. 3 B and C), which are biosynthetically equivalent to C-2 of IPP/DMAPP, comprise eight satellite lines. The long-range couplings observed in Fig. 3 A–C are transferred via sp² carbon atoms conducing to relatively large two-bond coupling constants. On the other hand, the C-5 of sitosterol is connected to C-1 via single bonds resulting in long-range coupling constants of less than 0.5 Hz, which cannot be resolved experimentally. However, it is obvious from the analysis of the other signals that each IPP/DMAPP unit contributed from [2,3,4,5-¹³C₄]deoxyxylulose carried four contiguous ¹³C atoms (Table 2).

Table 2 13C NMR analysis of sitosterol from a cell culture of C. roseus supplied with [1-13C] deoxyxylulose or[2,3,4,5-13C]deoxyxylulose

Figure 5 Structure (A) and 13C labeling pattern (B) of sitosterol. For details see Fig. 1. Arrow indicates the position documented by the NMR spectrum in Fig. 3C.

As determined from the intensities of the satellite lines relative to the uncoupled natural abundance signal (marked by asterisks in Fig. 3D), the diversion of deoxyxylulose to sitosterol was approximately 15 times less efficient than in the case of phytol and carotenoids. A similar dilution of the isotopic label was also detected in the experiment with [1-¹³C]deoxyxylulose where the maximum ¹³C abundance observed was at best 1.5%.

In summary, a total of 22 spectroscopically nonequivalent isoprenoid units could be analyzed by the detailed study of four metabolites yielding highly redundant information on the labeling patterns of the isoprenoid intermediates, IPP and DMAPP. These data show unequivocally that only one respective isotopomer of IPP and DMAPP had been formed from [2,3,4,5-¹³C₄]deoxyxylulose, namely [1,2,3,4-¹³C₄]DMAPP and [1,2,3,4-¹³C₄]IPP. These labeled precursors were diverted with high and closely similar efficacy to carotene, lutein, and phytol. They were also diverted to sitosterol, but with 15 times lower efficacy.

The combined labeling data from studies with ¹³C-labeled glucose, pyruvate, and deoxyxylulose (12, 13, 20–26, §) require that the biosynthesis of the isoprenoid intermediates proceeds by breaking of the C-3/C-4 bond of deoxypentulose and that a new bond is established between C-2 and C-4 of the original deoxypentulose intermediate (Fig. 6). This could occur by rearrangement or by fragmentation and reassociation of the fragments.

Figure 6 The skeletal rearrangement conducive to the observed labeling patterns of IPP as reconstructed from the terpenoid compounds studied. 13C labels from [1-13C]deoxyxylulose are shown as solid circles. Labeling from [2,3,4,5-13C4]deoxyxylulose (more ...)

We have now shown that [2,3,4,5-¹³C₄]deoxyxylulose is transformed into quadruple-labeled isotopomers of IPP and DMAPP (i.e., [1,2,3,4-¹³C₄]IPP and [1,2,3,4-¹³C₄]DMAPP, respectively) in cells of C. roseus. The exclusive formation of quadruple-labeled isotopomers from [2,3,4,5-¹³C₄]deoxyxylulose proves the strictly intramolecular nature of the process. A fragmentation/reassociation reaction would have resulted in the formation of isotopomers with fewer than four contiguous ¹³C atoms as a consequence of stochastic recombination between fragments arising from the isotope-labeled precursor and deoxyxylulose molecules formed in vivo from the natural abundance glucose that had been proffered in large excess.

Our interpretation is based on the analysis of more than a dozen isoprenoid moieties forming part of the metabolites under study. The close similarity of the enrichment values observed by biosynthetically equivalent carbon atoms in different isoprenoid units of each respective metabolite documents the high accuracy of the experimental method.

The relatively high isotope enrichments obtained in the presence of a low concentration of ¹³C-labeled deoxyxylulose show that the precursor is used with high efficiency for biosynthesis of phytol and carotenoids. It is an open question whether the free carbohydrate or a derivative such as a phosphoric acid ester acts as the committed precursor.

In our experimental system, the incorporation of deoxyxylulose into sitosterol is about 15-fold lower than in phytol and carotenoids. This is well in line with earlier studies by Arigoni and his coworkers in the plant, G. biloba, where ginkgolides were derived preferentially from deoxyxylulose, whereas steroids were derived preferentially from the acetate pool via mevalonate (13, §). In both cases, it seems that the mevalonate pathway is operative in the cytosol of the plant cells and the deoxyxylulose pathway is segregated in plastids, with some flux of metabolites between the different isoprenoid pools.

Preliminary data obtained with ¹³C-labeled acetate indicate that the mevalonate pathway is operative in the C. roseus cell line and that its metabolic products are preferentially channeled into the biosynthesis of sitosterol (Fig. 7). A more detailed analysis of metabolite partitioning of the two pathways in C. roseus is needed.

Figure 7 Hypothetical partitioning of the mevalonate pathway and the alternative pathway in C. roseus. R = H or PO3H2.

We thank N. Hohenstein (Munich) for excellent technical assistance and F. Wendling, A. Werner, and H. Wagner (Garching) for help in the preparation of the manuscript. This work was supported by grants from the Deutsche Forschungsgemeinschaft, Bonn (SFB 369), and the Fonds der Chemischen Industrie.