Contribution of the Mevalonate and Methylerythritol Phosphate Pathways to the Biosynthesis of Dolichols in Plants*

Plant isoprenoids are derived from two biosynthetic pathways, the cytoplasmic mevalonate (MVA) and the plastidial methylerythritol phosphate (MEP) pathway. In this study their respective contributions toward formation of dolichols in Coluria geoides hairy root culture were estimated using in vivo labeling with 13C-labeled glucose as a general precursor. NMR and mass spectrometry showed that both the MVA and MEP pathways were the sources of isopentenyl diphosphate incorporated into polyisoprenoid chains. The involvement of the MEP pathway was found to be substantial at the initiation stage of dolichol chain synthesis, but it was virtually nil at the terminal steps; statistically, 6–8 isoprene units within the dolichol molecule (i.e. 40–50% of the total) were derived from the MEP pathway. These results were further verified by incorporation of [5-2H]mevalonate or [5,5-2H2]deoxyxylulose into dolichols as well as by the observed decreased accumulation of dolichols upon treatment with mevinolin or fosmidomycin, selective inhibitors of either pathway. The presented data indicate that the synthesis of dolichols in C. geoides roots involves a continuous exchange of intermediates between the MVA and MEP pathways. According to our model, oligoprenyl diphosphate chains of a length not exceeding 13 isoprene units are synthesized in plastids from isopentenyl diphosphate derived from both the MEP and MVA pathways, and then are completed in the cytoplasm with several units derived solely from the MVA pathway. This study also illustrates an innovative application of mass spectrometry for qualitative and quantitative evaluation of the contribution of individual metabolic pathways to the biosynthesis of natural products.

Plant isoprenoids are derived from two biosynthetic pathways, the cytoplasmic mevalonate (MVA) and the plastidial methylerythritol phosphate (MEP) pathway. In this study their respective contributions toward formation of dolichols in Coluria geoides hairy root culture were estimated using in vivo labeling with 13 C-labeled glucose as a general precursor. NMR and mass spectrometry showed that both the MVA and MEP pathways were the sources of isopentenyl diphosphate incorporated into polyisoprenoid chains. The involvement of the MEP pathway was found to be substantial at the initiation stage of dolichol chain synthesis, but it was virtually nil at the terminal steps; statistically, 6 -8 isoprene units within the dolichol molecule (i.e. 40 -50% of the total) were derived from the MEP pathway. These results were further verified by incorporation of [5-2 H]mevalonate or [5,5-2 H 2 ]deoxyxylulose into dolichols as well as by the observed decreased accumulation of dolichols upon treatment with mevinolin or fosmidomycin, selective inhibitors of either pathway. The presented data indicate that the synthesis of dolichols in C. geoides roots involves a continuous exchange of intermediates between the MVA and MEP pathways. According to our model, oligoprenyl diphosphate chains of a length not exceeding 13 isoprene units are synthesized in plastids from isopentenyl diphosphate derived from both the MEP and MVA pathways, and then are completed in the cytoplasm with several units derived solely from the MVA pathway. This study also illustrates an innovative application of mass spectrometry for qualitative and quantitative evaluation of the contribution of individual metabolic pathways to the biosynthesis of natural products.
Polyisoprenoid alcohols together with sterols and quinone side chains constitute three main branches of terpene products originating from farnesyl diphosphate (FPP) 4 (1). These linear five-carbon unit polymers are divided into two groups, i.e. polyprenols and dolichols, according to the hydrogenation status of the ␣-terminal isoprene unit (dolichol structure is shown in Fig.  1). In cells, polyprenols and dolichols are always found as mixtures of prenologues, and data collected so far show polyprenols to be typical for bacteria and plants, whereas dolichols are generally attributed to animals and yeast (2). Nevertheless, it should be remembered that dolichols are the predominant form in some plant organs like roots (3). Data on the occurrence and functions of polyisoprenoids are summarized in recently published reviews (4,5). The formation of the polyisoprenoid chain, starting from the -end of the molecule (Fig. 1), proceeds in a biphasic manner with farnesyl-diphosphate synthase responsible for the synthesis of the all-trans-FPP (three isoprene units of -t 2 structure, t stands for trans-isoprene unit), and its further elongation by cis-prenyltransferase. The latter enzyme, cloned from several prokaryotic and eukaryotic organisms (see Refs. 6, 7 and references therein), including Arabidopsis thaliana (8,9) and Hevea brasiliensis (10), utilizes isopentenyl diphosphate (IPP) for elongation of FPP up to the desired chain length, thus producing a family of polyprenyl diphosphates (n isoprene units of -t 2 -c n-3 structure, c stands for cis-isoprene unit), which are subsequently converted to polyprenols or dolichols according to the "tissue-specific requirements" by a still unknown mechanism.
In plant cells two pathways are known to produce IPP utilized by numerous enzymes to finally give more than 50,000 different isoprenoid structures, the mevalonate pathway (MVA) and the mevalonate-independent methylerythritol phosphate pathway (MEP) (for reviews, see Refs. [11][12][13]. Both pathways are compartmentalized as follows: the MVA in the cytoplasm to provide sterols, the many sesquiterpenes, and the prenyl chains of ubiquinones, and the MEP one in the plastids giving hemi-, mono-, and diterpenes, carotenoids, and the side chain of plastoquinone. Although both pathways are thought to operate independently under normal conditions, interactions between them have been reported repeatedly. As a result of the exchange of intermediates derived from both pathways, "mixed origin" isoprenoids are described (reviewed in Ref. 12). More recently, the potential for the cross-talk between the MEP and MVA pathways has been directly proven by genetic modifications of pathway-specific enzymes and by application of pathway-specific inhibitors, namely mevinolin (14), also referred to as lovastatin, a highly specific inhibitor of 3-hydroxy-3-methylglutaryl-coenzyme A reductase in the MVA pathway, and fosmidomycin (15), a specific inhibitor of 1-deoxy-D-xylulose-5phosphate reductoisomerase in the MEP pathway. Both inhibitors have recently been used to perturb biosynthetic flux in hairy roots (16,17). The involvement of both pathways leading to the formation of a studied isoprenoid compound is most often estimated by the application of specifically labeled [ 13 C]glucose. A pathway-specific pattern of 13 C label within the isoprenoid residues ( Fig. 2) allows their origins via the MEP or MVA route to be discerned.
Using [ 3 H]mevalonate, it has been shown that in plants mainly cis-polyisoprenoid alcohols are synthesized via the MVA pathway (18,19), similarly to mammalian dolichols (20,21), although the possibility of an input from the alternative MEP pathway has not been addressed. Localization of polyprenols to plastids (22)(23)(24) and accompanying dolichols (a small fraction of total polyisoprenoid pool) to microsomes (25) might be the indication of their MEP and MVA origin, respectively. On the other hand, the MEP pathway has been found to contribute significantly to the synthesis of solanesol, an all-transpren-9 (26), whereas the solanesyl-like side chains of ubiqui-none turned out to be synthesized from IPP derived from the MVA pathway (27).
A hairy root culture of Coluria geoides was established for eugenol production (28); and because it also produces dolichols, it could be a useful model for studies of the early steps of dolichol biosynthesis in plants. The recently developed HPLC/ electrospray ionization-mass spectrometry methods (3) should greatly facilitate precise investigation of their structure. Despite many efforts, the biosynthetic origin of polyisoprenoid alcohols in plants remains unclear. Therefore, we have decided to analyze which pathway is the source of IPP built into dolichol.
Here we report that both pathways are involved in the biosynthesis of dolichols in hairy roots of C. geoides. The -terminal isoprene unit and several subsequent ones are synthesized with an involvement of both the MEP and MVA pathways, in contrast to the very last ␣-terminal and a few preceding units, where contribution of the MEP pathway is negligible. According to our MS data, on average 6 -8 isoprene units per dolichol molecule (ranging from 14 to 18 isoprene units) are formed in the MEP-dependent manner. A model is discussed, suggesting spatial regulation of dolichol synthesis and a unidirectional influx of IPP into plastids. C) at a 9:1 ratio (w/w). The maximal possible 13 C abundance in the products was thus 5.4% (see "Results"). For mass spectrometry either [ 13 C]glucose alone or, when indicated, native glucose supplemented with  [5-2 H]mevalonate (3.9 mg/flask) or 1-deoxy-[5,5-2 H 2 ]xylulose (4.6 mg/flask) was used. For inhibitor studies regular medium was supplemented either with mevinolin (30 M) or fosmidomycin (100 M) for the last 3 days of the culture, and no symptoms of toxicity were observed during these treatments. Two independent feeding experiments were carried out, each performed in duplicate. Nonsaponifiable lipids obtained from dried hairy roots were purified chromatographically as described earlier (3). For NMR analysis the dolichol-enriched fraction eluted from the silica gel column was further subjected to flash chromatography on RP-18 Lichroprep gel suspended in methanol. Pure dolichol mixture was eluted with acetone.

EXPERIMENTAL PROCEDURES
NMR Analysis-1 H and 13 C NMR spectra of metabolically 13 C-labeled dolichols (7 mM) were obtained with a Varian INOVA 400 MHz (Palo Alto, CA) spectrometer at 25°C in C 6 2 H 6 . Two-dimensional { 1 H, 13 C} gradient heteronuclear single quantum correlation experiments (31)(32)(33) were performed in proton-decoupled mode with a carbon spectral width of 25 kHz and 256 increments. One-dimensional 13 C experiment was performed with 64 K data points and a spectral width of 25 kHz in one-dimensional proton-decoupled mode. The measurement was carefully calibrated to achieve integrals for quantitative analysis. For CH 2 and CH 3 carbons of dolichol, T1 was measured yielding values between 1.1 and 2.7 s. The pulse sequence was optimized according to the literature (34,35); delay of 5.0 s, pulse width of 90°, and acquisition time of 0.2 s were applied avoiding nuclear Overhauser effects and presaturation. One hundred and twenty thousand scans (168 h) were collected yielding S/N ϭ 130. Spectra were calibrated against the chemical shift of benzene in 13 C spectra (128.0 ppm). Assignment of all 13 C NMR signals ( Table 2, for sequential numbering of carbon atoms, and see also supplemental Table 1 and supplemental Fig. 1) was done by combined use of twodimensional { 1 H, 13 C} gradient heteronuclear single quantum correlation spectra and literature data (36). Four residual signals were assigned to the traces of the organic solvent used for dolichol isolation, n-hexane (␦(C-␣ H 3 ), 14.4 ppm; ␦(C-␤ H 2 ), 23.1 ppm, and ␦(C-␥ H 2 ), 32.3 ppm), and to traces of "grease" accumulated during sample preparation (␦(CH 2 ), 30.2 ppm) as confirmed with 1 H, 13 C correlation spectra, whereas two low intensity signals (␦ 77.6 and 29.5 ppm) remained unassigned.
HPLC/ESI-MS analysis was performed as described for native dolichols (3) with the following modifications. Briefly, when indicated potassium acetate dissolved in solvent B was introduced post-column by a syringe pump (flow rate 5 l/min) through a T union into the LC flow before entering the mass spectrometer instead of the sodium salt. For statistical calculations, all the m/z data were normalized for [M ϩ Na] ϩ pseudomolecular ions. Because in all measurements only singly charged dolichol pseudomolecular ions were observed, the values of m/z were considered as molecular masses (in daltons) of pseudomolecular ions (adducts with sodium or potassium cations, as indicated).
Analysis of MS Spectra, Modeling of the Theoretical Envelope of MS Spectra-The theoretical distribution of isoptomers, ⌿(M), so-called "theoretical MS spectra," is mathematically given by a binomial distribution describing the probability of k successes (i.e. number of 13 C atoms replacing 12 C ones) in a hypothetical experiment of N ϭ n ϫ z tries, i.e. synthesis of Dol-n consisting of n isoprene units, each unit containing "z" carbon atoms which, theoretically, might be 13 C-enriched (see the Table 1) with the unitary probability of success, p, according to Equation 1, where M mi,n is the mass of monoisotopic [ 12 C]Dol-n; a binomial coefficient is given by Equation 2, and p is the probability of labeling of the specified carbon atom, which depends on the feeding conditions (summarized in Table  1). For the statistical analysis, the isotopomer distributions calculated for [1-13 C]glucose and [1,6-13 C 2 ]glucose experiments were additionally corrected for the natural 1.1% 13 C abundance at "nonenriched" positions (cf. Fig. 2). Analysis of MS Spectra, Estimation of the Average Molecular Mass of Dolichol-For each dolichol Dol-n, the average molecular mass was estimated as the location of the center of a Gaussian distribution fitted to the pattern of 10 -20 highest intensity signals recorded experimentally. The values were estimated with the standard deviation 0.1-0.3 Da.
Meta-analysis of MS Spectra, Linear Regression Analysis-The values of the estimated average molecular masses of dolichols were used to analyze the trends of the increase of dolichol molecular mass using two alternative linear regression procedures.
In the first linear regression procedure, the average molecular masses of dolichols, M(n,i), were analyzed as a function of the number of the isoprene units in the dolichol molecule, n, according to Equation 3, Thus, the average molecular mass of a dolichol molecule, consisting of n isoprene units M(n,i), is the sum of the masses of all the isoprenoid units, n ϫ m(i), and the intercept, called Dol 0 , which represents the mass difference obtained by subtraction of the mass of n repeating C 5 H 8 units from the total mass of the Dol-n pseudomolecular ion [M Dol-n ϩ Na] ϩ (see the chemical formula below Table 3). It is also worth stressing that Dol 0 contains no carbon atoms and consequently its mass cannot be dependent on feeding conditions. These calculations were performed separately for all the feeding conditions, enumerated by index i, and resulted in two sets of parameters, i.e. slopes m(i), which correspond to the average molecular mass of the isoprenoid unit specific for a given feeding experiment, and intercepts, Dol 0 . This approach permitted a qualitative estimation of the involvement of the metabolic pathways in the synthesis of individual isoprene units.
In the second linear regression procedure, the average molecular masses of dolichols were analyzed as a function of the theoretical isotopic enrichments, ⑀ MVA and ⑀ MEP , expected for the MVA and MEP pathways, respectively, according Equation 4; see Table 3 for the theoretical ⑀ values.
According to Equation 4, the average molecular mass of a Dol-n molecule is the sum of the following three components: the molecular mass of native dolichol, M nat (n), the mass increment from k units derived via the MEP pathway, k ϫ ⑀ MEP , and the mass increment from (n Ϫ k) units derived from the MVA pathway, (n Ϫ k) ϫ ⑀ MVA , and (⑀ MEP ‫؍‬ 2 ⁄ 3⑀ MVA ).
All the numerical analyses, including linear regression and fitting to the Gaussian distribution via conjugated gradient method (37), were performed with the aid of GnuPlot 4.0 software (©2004 Thomas Williams, Colin Kelley, available on line). [1-13 C]Glucose-Previous analysis of polyisoprenoid composition showed that hairy roots of C. geoides contained a family of dolichols (Dol-13 to Dol-29, with Dol-16 dominating) accompanied by traces of polyprenols of similar chain lengths (3). Because ϳ60% of dolichols accumulated in Coluria roots were in the form of esters with carboxylic acids (data not shown), alkaline hydrolysis was always performed prior to further analysis. Root cultures were grown for no longer than 3 weeks to keep glucose concentration in the medium sufficiently high to ensure that it was the main carbon and energy source.

C NMR Analysis of Polyisoprenoid Alcohols Labeled with
It is well established in the literature that after labeling with [1-13 C]glucose, carbon atoms derived from C-3 of IPP and DMAPP ( Fig. 2 and supplemental Fig. 1) should not be labeled either via the MVA or the MEP pathway. Indeed, analysis of the NMR spectrum of [ 13 C]-labeled dolichol ( Table 2) confirmed that their isotope abundances were the lowest of all the carbon atoms. Thus, the signals in 13 C-labeled dolichol (Dol) spectrum assigned to the IPP/DMAPP C-3 atoms were considered as reference signals of the natural 13 C isotope abundance, and the mean value of their integration calculated per a single carbon atom was considered as the natural abundance of ϳ1.1% and used for the subsequent normalization of the abundances of other signals.
High 13 C abundances in 13 C-labeled dolichol spectrum were recorded for C-2 and C-4 atoms of IPP and DMAPP; the calculated average abundance for C-2 and C-4 atoms was 3.0 Ϯ 0.4 and 3.0 Ϯ 0.2%, respectively. Somewhat higher values were observed for C-5 atoms of IPP and DMAPP (average value 3.5 Ϯ 0.3%). In contrast, the isotopic abundance of C-1 of IPP and DMAPP was considerably lower (average 1.3 Ϯ 0.4%) than those of C-2, -4, and -5. Notable differences between subgroups of the C-1 type atoms were recorded however. The isotopic abundance of the C-1 atoms located in the trans-isoprene units (ϳ2%) was notably higher than that of C-1 type atoms of the cis-isoprene units (ϳ1%), and in fact the latter one was at the level of the natural abundance. Nonetheless, it should be remembered that at the applied experimental conditions (10.9% isotopic 13 C abundance in the feeding medium), the maximal expected 13 C abundance in the products equals 5.4% because of glycolysis. In glycolysis, [1-13 C]glucose gives [3-13 C]glyceraldehyde 3-phosphate and [3-13 C]pyruvate, two precursors of the MEP pathway, and subsequently formed [2-13 C]acetyl-CoA, the direct precursor of the MVA pathway. Because of the isomerization of the triose phosphates, dihydroxyacetone phosphate and glyceraldehyde 3-phosphate are interconverted, and C-1 and C-6 of glucose are metabolically equivalent. Consequently, the 13 C abundance in the final product is ϳ50% that used in the supplied glucose, and the probability of labeling of each potentially labeled carbon atom is ϳ0.5. It should be noticed, however, that the expected 13 C abundance (5.4%) was never observed for any of the labeled positions, indicating possible activity of the pentose phosphate pathway that resulted in a partial loss of 13 C from C-1 of glu-cose, formation of 13 CO 2 , and consequently isotopic dilution. Nevertheless, the 13 CO 2 resulting from [ 13 C]glucose catabolism is not efficiently recycled in nonphotosynthetic tissues like roots, which in turn precludes scrambling in the labeling pattern in fact not observed in our labeling experiments (see "Discussion").
Conclusions Arising from 13 C NMR Spectrum of Dolichol-The observed labeling pattern was consistent with a dual pathway origin of dolichols; however, the intensity of labeling of selected carbon atoms was not constant throughout the length of the dolichol molecule. The lower isotopic abundances found for C-1, -2, and -4 of IPP and DMAPP than that of C-5 are well explained by their different labeling pattern (Fig. 2). Although C-5 is labeled via either pathway, C-1 will become labeled only when synthesized via the MEP one, whereas C-2 and -4 only via MVA. Consequently, if a dolichol molecule contains isoprene units derived from both pathways, its C-1, C-2, and C-4 positions will be labeled to a lesser extent than the C-5 positions of IPP/DMAPP because of isotopic dilution.
Thanks to the high resolution of the NMR spectra of dolichols, signals of carbon atoms could be assigned unambiguously to all the positions C-1 to C-5 of IPP/DMAPP and also could be differentiated between those deriving from the trans and the cis units (i.e. those near the terminus of the dolichol molecule and those near the ␣ terminus, respectively; see Fig. 1) ( Table 2 and supplemental Table 1). Keeping these data in mind, the results of the analysis of the 13 C NMR spectrum suggested that the -terminal part of the dolichol molecule contains both MEP-and MVA-derived isoprene units, in contrast to the ␣-terminal part exclusively derived from the MVA pathway (see supplemental Fig. 2). Simultaneously, the characteristic profile of this spectrum, especially no labeling of C-3 atoms of IPP, indicated that virtually no 13 C scrambling occurred during the labeling experiments. Because only approximate estimation of the relative contributions of both pathways to dolichol biosynthesis may be based on 13 C NMR analysis, 13 C-labeled dolichols were analyzed in parallel by mass spectrometry to further evaluate the relative involvement of both pathways to their biosynthesis.
MS Analysis of Polyisoprenoid Alcohols Labeled by [U-13 C 6 ]Glucose-As expected, the m/z values for pseudomolecular ions of dolichols obtained from roots grown on [U-13 C 6 ]glucose as the sole carbon source (99% isotope abundance) were increased compared with native dolichols and equaled the maximal m/z predicted for uniformly labeled [U-13 C 5n ]Dol-n (supplemental Fig. 3A). In addition to the signals of the labeled dolichols, signals of lower intensity of native Dol-n, originating from the inoculum, were recorded in the Isotopic abundance of [1-13 C]glucose used was 10.9% and thus maximal possible abundance was 5.4%, see "Results." Numbering of carbon atom positions is given in Fig. 2. Average isotopic abundance of C-3 signals was assumed to correspond to the natural 13 C abundance (ϳ1.1%) and was further used as a standard for calibration. ␣ and stand for OH-and hydrocarbon-terminal isoprene residues, respectively (see Fig. 1 and supplemental Fig. 1 and Table 1). Values in bold highlight 13 C labeling.  Fig. 3B) and were accompanied by signals corresponding to native Dol-16 with a maximum at 1132.3 Da. Spectra of several homologous dolichols (Dol-14 to Dol-19) were recorded (Fig. 3A); however, the latter one, because of its low intensity, was not further used for quantitative analysis. The observed shift of isotopomer distribution toward higher molecular mass indicated 13 C enrichment of the dolichol molecules. The complex distribution pattern of the signals, which mirrors the stochastic dispersion of the probability of formation of differentially labeled [ 13 C]dolichol isotopomers, is understandable bearing in mind that the probability of labeling of each potentially labeled carbon atom is ϳ0.5, as the result of stochastic mixing of triose pools during glycolysis (see "Results," " 13 C NMR Analysis of Polyisoprenoid Alcohols Labeled with [1-13 C]Glucose"). For each IPP molecule derived from the MEP pathway (C-1 and C-5 labeled, Fig. 2), the number of combinations of 13 C enrichment patterns is 2 2 , and the number of differently labeled IPP isotopomers is 3, with the 13 C distribution profile [ 13 C 2 ]:[ 13 C 1 ]:[ 13 C 0 ] in a 1:2:1 ratio. In the case of the MVA pathway (C-2-, -4-, and -5-labeled), the number of combinations is 2 3 , whereas the number of IPP isotopomers is 4, and the corresponding profile is [ 13 C 3 ]:[ 13 C 2 ]: [ 13 C 1 ]:[ 13 C 0 ] in a 1:3:3:1 ratio. The condensation of the differently labeled IPP molecules leading to the polyisoprenoid backbone is also further probabilistic, resulting in a complex mixture of differentially labeled dolichol isotopomers producing a complex pattern of m/z signals in MS analysis, as presented in Fig. 3. [1,[6][7][8][9][10][11][12][13] C 2 ]Glucose-To simplify the dolichol mass distribution patterns, [1,6-13 C 2 ]glucose was used for feeding. As expected,  JULY 25, 2008 • VOLUME 283 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 21029
the m/z values were higher than those obtained with [1-13 C]glucose for all dolichols studied ( Fig. 3B and supplemental Fig. 3C). Unexpectedly, however, the MS spectrum was still rather complex, which might result from the interference of two phenomena. First, 13 C abundance in the doubly labeled glucose molecule was different at C-1 and C-6 (99 versus 97%, respectively). Second, the 13 CO 2 released by the oxidative pentose phosphate pathway might have been reincorporated via the activity of the malic enzyme, yielding a small but noticeable redistribution of 13 C within the intermediates resulting in higher-than-expected formation of hypo-and hyper-labeled 13 Clabeled dolichol. Additionally, the effect of stochastic mixing of the pools of [ 13 C 2 ]IPP and [ 13 C 3 ]IPP, respectively, derived from the MEP and MVA pathways and incorporated into the polyisoprenoid chains, should also be considered.
Global Analysis of Mass Spectrometry Data-The complex mass spectra recorded for the dolichols provoked us to generate "theoretical spectra," i.e. the theoretical distributions of [ 13 C]dolichol isotopomers expected for dolichol synthesized exclusively via either the MEP or the MVA pathway. For this calculation, glycolysis was assumed as the only source of the intermediates for IPP formation. Representative theoretical spectra calculated separately for Dol-16 synthesized either via the MVA or the MEP pathways with [1-13 C]glucose feeding are shown in Fig. 4. The experimentally recorded Dol-16 spectrum was located between the two theoretical distributions, indicating a mixed biosynthetic origin of this dolichol. Similar correlation was obtained for [1,6-13 C 2 ]glucose labeling (not shown). This observation indicates that neither the MVA nor the MEP pathway was the sole source of IPP utilized for the formation of the dolichol molecule, which is in accordance with the NMR data described above. On the other hand, the complexity of the experimental MS spectra prompted us to apply numerical methods for their analysis.

Estimation of Average Molecular Masses of [ 13 C]Dolichols-
Because each dolichol species always occurs as a mixture of isotopomers, a knowledge of its average molecular mass was required for all the calculations described below. This estimation was done for the theoretical spectra generated above for each dolichol (Dol-14 -18) separately for each labeling conditions (supplemental Table 2). Thus for native or uniformly labeled glucose, for which the expected average molecular mass of Dol-n is the same regardless of the pathway used for its biosynthesis, 10 values of the expected average molecular masses of Dol-14 -18, 5 for either of these conditions, were obtained. For selectively labeled glucose, the theoretical average molecular masses expected for dolichols (Dol-14 -18) obtained exclusively either from the MEP or the MVA pathway gave two values for each dolichol, thus in total 10 values for [1-13 C]glucose and similarly 10 values for [1,6-13 C 2 ]glucose feeding were obtained (supplemental Table 2). The experimental spectra were treated in a manner analogous to that applied to the theoretical ones described above. The average experimental molecular masses were estimated separately for each dolichol labeled at all four labeling conditions (feeding by [1-13 C]-, [1,6-13 C 2 ]-, [U-13 C 6 ]-, and native glucose). Namely, for Dol-15-18, four mixtures of isotopomers each and three mixtures for Dol-14 were used, as shown in Fig. 3B; for technical reasons the spectrum of uniformly labeled Dol-14 obtained after [U-13 C 6 ]glucose feeding could not be recorded. A representative determination of the four values of average experimental molecular masses of Dol-16 (three labeling conditions overlaid) is shown in Fig. 5. As the final result a set of 19 experimental values of molecular masses of dolichols was obtained (supplemental Table 2). The location of the experimental values of the Dol-n average molecular masses between the two theoretical ones (Fig. 4) indicates that each Dol-n molecule was of a mixed biosynthetic origin, via both the MEP and MVA pathways. The obtained experimental average molecular masses of dolichols  were further analyzed to estimate the involvement of the two pathways in dolichol biosynthesis. It is worthwhile to emphasize that the value of the average experimental molecular mass of Dol-n permits to unambiguously determine the number of isoprene units derived from the MEP and the MVA pathway, respectively (for explanation see supplemental text).
Qualitative Analysis of Mass Spectrometry Data-Using the experimental average molecular masses of 13 C-labeled dolichols (supplemental Table 2), it was possible to estimate the average increase of the molecular mass between sequential dolichols, which corresponds to the average mass of a single isoprene unit. The mass of the average isoprene unit, m(i), will of course be dependent on the feeding condition. The values of the average molecular masses of dolichols (supplemental Table  2) were analyzed according to the linear regression method. The dolichol molecular mass is thus described by Equation 3. Individual trends estimated for each labeling experiment are presented in supplemental Fig. 4. The linear regression analysis leads to the estimation of four average masses of isoprene units and four values of Dol 0 , summarized in Table 3. The theoretically expected enrichment of the average molecular mass of a single isoprene unit was calculated based on the labeling pattern of the isoprene unit for [1-13 C]glucose and [1,6-13 C 2 ]glucose (Fig. 2). The experimental values of the enrichment of the mass of the isoprene unit (1.24 Ϯ 0.09 and 3.09 Ϯ 0.21 Da for the singly and doubly labeled glucose, respectively) were in better agreement with the theoretical values predicted for the MVA pathway (correspondingly 1.49 and 2.99 Da) than with those for the MEP pathway (0.99 and 1.49 Da, respectively). It should be stressed here that the average molecular mass of a single isoprene unit calculated above concerns isoprene units located in the proximity of the ␣ terminus of the Dol-n molecule because all the Dol-n species are synthesized by elongation of precursors from a common pool of polyprenyl diphosphates. Thus, Dol-n is formed directly from Dol-(n-1) via addition of a single isoprene unit to the ␣ terminus and consequently the estimated increments concern the ␣-terminal units.
On the other hand, the average molecular masses estimated experimentally for the entire Dol-n molecules (see above) were lower than those calculated for the MVA-derived dolichols, indicating that some part of these molecules must have been synthesized via the MEP pathway, which produces "lighter" isoprenoid units. It should be remembered that isoprenoid units derived from feeding by either [1-13 C]glucose or [1,6-13 C 2 ]glucose might be either "light," originating from the MEP pathway and so containing up to two 13 C atoms per unit, or "heavy," originating from the MVA pathway thus containing up to three 13 C atoms per unit. Too low a mass of the experimentally estimated average molecular mass of Dol-n is also mirrored by the lower-than-expected experimental values of Dol 0 , presented in Table 3 (further discussed in comments to supplemental Fig. 4). By elimination, because we know from the earlier considerations that the ␣-terminal isoprene units were derived from the MVA pathway, the MEP-derived ones could only have been used for the synthesis of the -proximal portion of the dolichol molecules. It should be also pointed out that good agreement of the values of the observed increase of the molecular mass between sequential dolichols with theoretical calculations clearly indicates that there was no sizeable loss of 13 C atoms along the biosynthetic pathway from mono-or doubly labeled glucose toward dolichol, which in turn indicates negligible 13 C scrambling.
Quantitative Analysis of Mass Spectrometry Data-We proceeded to estimate the number of the two types of isoprene units in the dolichol molecules. For the isoprene unit synthesized via the MVA pathway, the expected isotopic enrichment, ⑀ MVA , should be 1.49 and 2.94 Da for [1-13 C]glucose and [1,6-13 C 2 ]glucose, respectively, with 0.05-and 4.95-Da enrichment for native and uniformly labeled glucose (Table 3). For a dolichol molecule, Dol-n, synthesized entirely via the MVA pathway, its average molecular mass, M(n) after [1-13 C]glucose feeding should equal M nat (n) ϩ n ϫ 1.49, because it is the sum of the molecular mass of native Dol-n and the total enrichment of all the isoprene units. Similarly it should be M nat (n) ϩ n ϫ 2.94 for [1,6-13 C 2 ]glucose feeding and M nat (n) ϩ n ϫ 4.95 for feeding with uniformly labeled glucose. A graphic presentation of this linear increase of molecular mass plotted as a function of the enrichment expected for the MVA pathway is shown separately for each dolichol in Fig. 6 (solid lines). The experimental average molecular masses of dolichols were lower than those calculated for the MVA labeling, as discussed above. To estimate the number of the lighter (k) and heavier (n Ϫ k) isoprene units per Dol-n molecule, Equation 4 giving the average molecular mass of each dolichol species obtained with all three types of [ 13 C]glucose and native glucose feeding was used. Equations describing the individual trends estimated for each dolichol are presented in the supplemental text. Because ⑀ MEP ϭ 2 ⁄ 3⑀ MVA (see Table 3 for the theoretical ⑀ values), the slope value for the linear regression (Equation 4) against ⑀ MEP (n Ϫ 1 ⁄ 3 ϫ k) permitted estimation of k, the average number of isoprene units derived from the MEP pathway (Table 4). According to these calculations, six to eight (lighter) isoprene units per dolichol molecule are derived from the MEP pathway. In summary, the described meta-analysis of mass spectrometry data revealed that in dolichols longer than Dol-14 successive isoprene units used for elongation of the molecule (i.e. a single ␣-terminal unit for Dol-15; two units. ␣ and ␣ ϩ 1 for Dol-16; three ␣-terminal ones for Dol-17; four for Dol-18, see supplemental Fig. 1) are derived exclusively from the MVA pathway. The 13 C NMR spectroscopy data indicated that also the ␣-terminal unit of Dol-14 is of the MVA origin, similarly as in the longer dolichols. Other units, incorporated into the growing Dol molecule at earlier steps, may be derived from either pathway; on average between six and eight such units come from the MEP pathway and the balance (i.e. seven to five for the 13-unit oligoprenyl precursor) is from MVA. A model summarizing the involvement of the MVA and MEP pathways in the biosynthesis of a dolichol molecule is presented in supplemental Fig. 2. This conclusion on the involvement of both pathways in dolichol biosynthesis was verified by application of pathway-specific precursors and inhibitors.
Labeling of Polyisoprenoids with Pathway-specific Precursors-Supplementation of the feeding medium with deuterated compounds (i.e. either with [5,5- (Fig. 7A), whereas for [5-2 H]mevalonate all but the monoisotopic signals were enhanced (Fig. 7B) for each of the dolichols analyzed. In con-trast to mevalonate, feeding experiments with DX indicated incorporation of an even number of deuterium atoms per dolichol molecule, which is in agreement with the known maintenance of the two hydrogen atoms linked to C-5 of DX at the C-1 position (Fig. 2) of the isoprene unit (39,40). The observed enhancement of the odd ion [M ϩ 3] resulted from the incorporation of two deuterium atoms into a dolichol isotopomer containing one 13 C atom per molecule. The low incorporation of deuterium was because of the high dilution of the deuteriumlabeled precursors because they constituted only 0.06% (by mass) of glucose in the feeding medium. Effective incorporation of tritium from [ 3 H]mevalonate into both polyisoprenoid alcohols and sterols was also noted (supplemental Fig. 5). The highest labeling was found for 21-day-old cultures.
Effect of Pathway-specific Inhibitors on Isoprenoid Accumulation-Mevinolin (30 M), a specific inhibitor of 3-hydroxy-3-methylglutaryl-CoA reductase of the MVA pathway, efficiently inhibited accumulation of both polyisoprenoid alcohols and sterols in the oldest culture, resulting in a remarkable decrease of their content (by 85 and 82% for polyisoprenoids and sterols, respectively, see Table 5). A higher concentration of mevinolin (60 M) gave similar results (not shown); however, surprisingly, de novo synthesis of both lipids from [ 3 H]mevalonate was also inhibited suggesting a pleiotropic effect of this drug on isoprenoid metabolism. On the one hand this might be an indication of the inhibition of other enzymes of the pathway besides 3-hydroxy-3-methylglutaryl-coenzyme A reductase, as was earlier found for sesquiterpene cyclase (41) or in mammalian systems for geranylgeranyl diphosphate synthase (42). On the other hand, possible toxic effects of mevinolin should be also considered (43). Fosmidomycin, a specific inhibitor of 1-deoxy-D-xylulose 5-phosphate reductoisomerase decreased the accumulation of both lipids in the youngest culture (by 73 and 78% for dolichols and sterols, respectively), whereas an increased content of both lipids was found for 2-and 3-weekold cultures. The variable effects of fosmidomycin most proba-  Table  3). Solid lines follow the expected values calculated for masses of dolichols synthesized exclusively via the MVA pathway. The observed mass deficit of 7-15 Da found for mono-and di-labeled glucose experiments indicates substantial participation of the alternative MEP pathway in dolichol synthesis. bly mirror metabolic shifts between both pathways upon stress caused by the inhibitor.

DISCUSSION
In this study the biosynthetic origin of plant dolichols was investigated. We analyzed the contribution of the two alternative pathways known to produce IPP in plant cells, the MEP and the MVA pathway, to dolichol biosynthesis. Dolichol synthesis was studied by in vivo labeling with a general precursor (glucose), pathway-specific precursors (MVA or DX), and by inhibition of the metabolic flow with pathway-specific inhibitors. Our conclusion indicating a "mosaic" structure of plant dolichols is based on the following observations. First, isoprenoid units located at and near the -end of the dolichol molecules could be synthesized via either pathway as shown by the high intensities of the corresponding signals in the 13 C NMR spectrum. Second, isoprenoid units localized at the proximity of the ␣ terminus were exclusively of the MVA origin as shown by the numerical analysis of the 13 C-enriched masses of pseudomolecular ions and in parallel by NMR. Third, the calculated average number of lighter (i.e. derived from MEP) isoprenoid units per dolichol molecule was between 6 and 8 as based on the MS spectra, which was also supported by the 13 C NMR data. The efficient incorporation of 13 C atoms from uniformly labeled glucose into dolichols confirmed the applicability of Coluria roots as a model for biosynthetic studies. Dolichols were labeled almost exclusively by products derived from the glycolytic pathway as confirmed by the NMR spectrum (synchronized labeling of C-2 and C-4 carbon atoms, with proportionally higher labeling of C-5 atoms of IPP, see Table 2) and MS spectra (increase of the dolichol molecular mass, see Table 3). All these observations exclude any substantial scrambling of 13 C during the course of labeling. Yet the general conclusions drawn from such a model should take its limits into consideration. In vitro propagated organs might differ from their physiological equivalents. Heterotrophic growth of the tissue stops the influx of intermediates derived from photosynthesis as well as the lightdriven regulation of metabolism. Summarizing the validity of the meta-analysis of MS data applied here, it should be pointed out that by knowing the value of the enrichment of the molecular mass of a compound of interest and the number of isoprene units constituting its molecule one can estimate the relative input of both the MEP and MVA pathways to its biosynthesis (for details see supplemental text).
Finally, the proposed mechanism of dolichol biosynthesis is as follows. A mixture of IPPs derived from the MVA and MEP pathways is used in plastids for the initiation of the process and sequential condensations, and thus-formed oligoprenyl diphosphates (surely shorter than 14 isoprene units, see supplemental Fig. 2) are exported to the cytoplasm where they are finally elongated and terminated with IPP of exclusively MVA origin. Such a mechanism, besides compartmentalization of the sequential steps, also requires a unidirectional import of IPP from the cytoplasm to plastids. A spatial model of the organization of dolichol biosynthesis is presented in Fig. 8. Such a model assuming stochastic mixing of IPP molecules within the plastid compartment explains well the broad signal distribution pattern in the MS spectra obtained for [1,6-13 C 2 ]glucose feeding (Fig. 3B), which results from randomized insertion of the doubly (MEP-derived) and triply (MVA-derived) labeled IPP in the growing oligoprenyl chain. A further argument supporting the dualpathway biosynthetic origin of dolichol is given by the localization of the experimental distribution profile of [ 13 C]dolichol isotopomers between the theoretical profiles ( Fig. 4) calculated separately for each pathway. In line with the above model (Fig. 8) is a recent observation of an export of MEPderived IPP and possibly also geranyl diphosphate from plastids to the cytoplasm during sesquiterpene biosynthesis in carrot roots (44). The dolichol biosynthetic scenario described above raises the intriguing question of the nature of at least two independent cis-prenyltransferases which should be involved. Their existence seems plausible in the light of the in silico predicted occurrence of a family of six genes encoding cis-prenyltransferase in the Arabidopsis genome (45). However, the mechanism of the reaction remains unclear because the cytoplasm/endoplasmic reticulum localized cis-prenyltransferase should accept medium chain length oligoprenyl diphosphate as a substrate, although it is generally believed that only all-trans-FPP can serve as a starter for the polyisoprenoid chain.