Distinct isoprenoid origins of cis- and trans-zeatin biosyntheses in Arabidopsis.

Plants produce the common isoprenoid precursors isopentenyl diphosphate and dimethylallyl diphosphate (DMAPP) through the methylerythritol phosphate (MEP) pathway in plastids and the mevalonate (MVA) pathway in the cytosol. To assess which pathways contribute DMAPP for cytokinin biosynthesis, metabolites from each isoprenoid pathway were selectively labeled with (13)C in Arabidopsis seedlings. Efficient (13)C labeling was achieved by blocking the endogenous pathway genetically or chemically during the feed of a (13)C labeled precursor specific to the MEP or MVA pathways. Liquid chromatography-mass spectrometry analysis demonstrated that the prenyl group of trans-zeatin (tZ) and isopentenyladenine is mainly produced through the MEP pathway. In comparison, a large fraction of the prenyl group of cis-zeatin (cZ) derivatives was provided by the MVA pathway. When expressed as fusion proteins with green fluorescent protein in Arabidopsis cells, four adenosine phosphate-isopentenyltransferases (AtIPT1, AtIPT3, AtIPT5, and AtIPT8) were found in plastids, in agreement with the idea that the MEP pathway primarily provides DMAPP to tZ and isopentenyladenine. On the other hand, AtIPT2, a tRNA isopentenyltransferase, was detected in the cytosol. Because the prenylated adenine moiety of tRNA is usually of the cZ type, the formation of cZ in Arabidopsis seedlings might involve the transfer of DMAPP from the MVA pathway to tRNA. Distinct origins of large proportions of DMAPP for tZ and cZ biosynthesis suggest that plants are able to separately modulate the level of these cytokinin species.

release of lateral buds from apical dominance, and delay of senescence. The biological activity, signal transduction, and metabolism of CKs have long been studied (1). However, it remains unclear how different classes of CKs are produced in plants and whether such classes of CKs play different roles in plant development. Most natural CKs are derivatives of N 6prenylated adenine. At least two CK species, trans-zeatin (tZ) and isopentenyladenine (iP), are considered to be active forms in Arabidopsis according to the specific recognition by a CK receptor that has recently been identified (2,3). By contrast, cis-zeatin (cZ) exhibits only low or no activity to this Arabidopsis receptor (2).
On the other hand, the tRNA-dependent CK biosynthesis pathway has also been proposed in plants, as some tRNA species contain an N 6 -prenylated adenine moiety, which, by hydrolysis, is capable of forming CKs (13). The prenylation of tRNA is catalyzed by tRNA isopentenyltransferase (tRNA-IPT; EC 2.5.1.8). In Arabidopsis, AtIPT2 encodes a tRNA-IPT (14). AtIPT9 is also similar to bacterial tRNA-IPT in sequence (8), though its enzyme activity has not been demonstrated. In plants, cZ and iP are generally the major components of prenylated tRNAs (13). Therefore, the prenylated tRNA has been considered as a possible source of cZ. In addition, the occurrence of zeatin cis-trans isomerase activity (15) suggests that the tRNA-mediated pathway might also contribute to the synthesis of tZ-type CKs through cZ-type CKs.
Plants have two possible biosynthesis pathways for the prenyl group of CKs, the methylerythritol phosphate (MEP) pathway in plastids and the mevalonate (MVA) pathway in the cytosol (16,17). Both pathways supply the common isoprenoid precursors isopentenyl diphosphate and DMAPP. Although the MEP and MVA pathways exist in separate subcellular locations, there is some exchange of common precursor(s) between * This study was supported in part by Grants-in-aid for scientific research (12142202 (to H. S.) and 15770035 (to K. T.)) from the Ministry of Education, Culture, Sports, Science, and Technology, Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  the two pathways (18,19). Therefore, administration of a MEP pathway precursor can partially suppress the growth inhibition caused by a block in the MVA pathway and vice versa (20). As is the case with other isoprenoid biosyntheses, the MVA pathway had been considered the sole route for providing DMAPP to CKs until the MEP pathway was uncovered recently. In fact, the incorporation of 14 C labeled MVA into the iP element of tRNA in vivo has been demonstrated in tobacco pith tissue (21). The incorporation of [ 13 C]MVA into iP and transzeatin riboside (tZR) in vitro was also reported in the endosperm of Sechium edule seeds (22). In addition, there are some reports that indicate that CK levels are reduced in plants when the MVA pathway is limited (23)(24)(25). On the other hand, the contribution of the MEP pathway to CK biosynthesis has never been studied previously. It should be noted that the incorporation of MVA does not exclude a potential role of the MEP pathway in the biosynthesis of CKs; it has been observed that isoprene units from both MEP and MVA pathways are incorporated into a single downstream isoprenoid (19,20).
To selectively label metabolites from the MEP and MVA pathways with 13 C in vivo, we have previously performed feeding of 13 C labeled 1-deoxy-D-xylulose (DX) or mevalonolactone (MVL) to Arabidopsis seedlings (19). DX is converted into the MEP pathway intermediate 1-deoxy-D-xylulose 5-phosphate by phosphorylation. Therefore, exogenous DX is able to complement the albino phenotype of the cla1-1 mutant (26), which is defective in 1-deoxy-D-xylulose-5-phosphate synthase in the MEP pathway. Similarly, the growth inhibition due to a block in the MVA pathway by mevastatin (an inhibitor of 3-hydroxy-3-methylglutaryl-CoA reductase) is rescued by exogenous application of MVL (19). Efficient 13 C labeling of metabolites from the MEP and MVA pathways was thus achieved by feeding 13 C labeled DX and MVL to the cla1-1 mutant and mevastatintreated plants, respectively, at a concentration that is sufficient to restore the phenotype nearly fully to wild type. This method allowed us to determine the contribution of the MEP and MVA pathways to the biosynthesis of gibberellins, another group of phytohormones, by gas chromatography-mass spectrometry (GC-MS) (19).
In this study, we address the biosynthesis route for the prenyl moiety of CKs using the 13 C labeled tracers in Arabidopsis seedlings. Our data demonstrate that the prenyl side chains of tZ-and iP-type CKs are mainly produced through the MEP pathway, whereas a large fraction of cZ derivatives is synthesized through the MVA pathway. We also show the subcellular location of AtIPTs produced as green fluorescent protein (GFP)-fusion proteins. Based on these data, we propose a crucial role of the plastid-localized MEP pathway in CK biosynthesis, and discuss how different classes of CKs are biosynthesized through the MEP and MVA pathways in plants.
Stable Transformants of Arabidopsis-Genomic DNA fragments of AtIPT3 and AtIPT7 containing the 5Ј-flanking regions (3.9 kb for AtIPT3, 4.1 kb for AtIPT7) and the full coding regions were ligated into pTH2 vectors (29) to produce chimeric genes fused to the amino terminus of the GFP gene. The chimeric genes were introduced into Arabidopsis by the floral dip method (30). More than five independent lines were obtained, and T3 plants were used for analysis.
Plant Hormone Analysis-Extraction and fractionation of CKs from Arabidopsis seedlings (3-4 g) were performed as described previously (31). The nucleotide CK fractions were further analyzed as the corresponding nucleosides after treatment with phosphatase (9). Each CK fraction was purified by immunoaffinity columns (32) except for the addition of anti-cis-zeatin riboside (cZR) antibodies to the original protocol. After desalting, the resulting samples were dissolved in H 2 O and analyzed with a liquid chromatography (LC)-MS system (model 2695/ ZQ2000MS, Waters). CKs were separated at a flow rate of 0.25 ml/min, with the gradients of solvents A (H 2 O), B (methanol), and C (0.1% acetic acid) set according to the following profile: 0 min, 95% A ϩ 5% C; 1 min, 95% A ϩ 5% C; 16 min, 45% A ϩ 50% B ϩ 5% C; 22 min, 25% A ϩ 70% B ϩ 5% C. Capillary voltage was 4.0 kV. Other conditions were described previously (9). The quantification of [MϩH] ϩ and ϩ1 signals were achieved at levels between 0.1 and 100 pmol/injection using a standard curve. Abscisic acid (ABA) was purified from the methanoleluted fraction from the MCX column (Waters) in CK purification by high performance liquid chromatography and analyzed by GC-MS as reported previously (33). The 13 C incorporation level was calculated using the fragment ion cluster between m/z 190 and 192 after subtraction of natural 13 C abundance (34). Campesterol (CAM) was eluted from the MCX column with CH 2 Cl 2 (3 ml x 3), after purification of CKs. CAM was further purified by high pressure liquid chromatography and analyzed by GC-MS as reported previously (19).

CK Levels in Arabidopsis
Seedlings-To study the metabolic origins of the prenyl group of CKs, we planned to conduct the 13 C labeling experiment that was employed previously to determine the contribution of the MEP and MVA pathways to gibberellin biosynthesis (19). To examine whether this method was feasible for CKs, we first analyzed endogenous CK levels by LC-MS in Arabidopsis seedlings under the same growth conditions, after a feeding of non-labeled DX and MVL. When DX was fed to the cla1-1 mutant to rescue its albino phenotype, all CKs analyzed were detectable by LC-MS. As shown in Table  I, the trend in the amount of individual CKs is as follows: nucleotide CKs Ͼ nucleoside CK Ͼ free-base CKs (e.g. transzeatin riboside monophosphate (tZRMP) Ͼ tZR Ͼ tZ). It was also noted that cZ-type CKs accumulated at much lower levels relative to tZ-and iP-type CKs. Similar accumulation patterns of different CK species were observed when the growth inhibition by mevastatin was rescued by MVL (Table I). These data allowed us to confirm that for all CKs examined, an accurate quantification of the molecular ion and its ϩ1 isotopomer ion by LC-MS would be possible in this system with the exception of cZ, which was detectable, but its ϩ1 isotopomer ion was below the range of reliable quantification.

Incorporation of [1-13 C]DX and [2-13 C]MVL into CKs-
To evaluate the role of the MEP pathway in providing DMAPP to CKs, [1-13 C]DX was fed to seedlings of the cla1-1 mutant. Likewise, to label metabolites from the MVA pathway [2-13 C]MVL was fed to wild type seedlings in the presence of mevastatin. In both systems, one 13 C atom would be incorporated into CKs if they were produced through the MEP or MVA pathways (Fig. 1). Table II shows the levels of 13 C incorporation into CKs determined by LC-MS. The incorporation of [1-13 C]DX was evident for all CKs listed in the table. For example, tZR has a molecular ion at m/z ϭ 352, whereas the corresponding ion for tZR from the [1-13 C]DX-treated cla1-1 seedlings was observed at m/z ϭ 353. A fragment ion of tZR was detected from the [1-13 C]DX-treated cla1-1 seedlings at m/z ϭ 221, which is 1 mass unit larger than that of authentic tZR (m/z ϭ 220). The presence of a 13 C atom in cis-zeatin riboside monophosphate (cZRMP) and cZR was clear in the mevastatin-treated plants fed with [2-13 C]MVL (Table II). Incorporation of [2-13 C]MVL into cZ was also indicated by an ion at m/z ϭ 221 ([MϩHϩ1] ϩ ), which appeared more abundant than [MϩH] ϩ of authentic cZ at m/z ϭ 220 (data not shown), but their relative intensities could not be quantified because of low levels of cZ in the samples (Table I). Fig. 2 shows relative levels of 13 C incorporation from [1-13 C]DX and [2-13 C]MVL into CKs, ABA, and CAM determined from the data sets given in Table II after the subtraction of natural 13 C abundance (34). It has been reported previously (35) that the MEP pathway mainly supplies precursors to the biosynthesis of ABA, which is produced through carotenoids in plastids. By contrast, CAM is a cytosolic phytosterol that is primarily synthesized via the MVA pathway. Our data show that CKs can be classified into two groups based on the 13 (Fig. 2). In marked contrast, about 75% of the prenyl moiety of cZ-type CKs (cZRMP and cZR) were labeled with [2-13 C]MVL, whereas the incorporation of [1-13 C]DX into cZ derivatives was significantly lower than the incorporation into tZ and iP derivatives.
As discussed previously (19), the levels of 13 C incorporation determined across separate feeding systems must be carefully interpreted, because the ratio of 13 C labels in products can be altered by the concentration of 13 C precursors in the media. Thus, the values in Fig. 2 do not immediately reflect the relative contribution of each isoprenoid pathway under normal growth conditions. Nevertheless, we conclude from the LC-MS data that tZ-and iP-type CKs in Arabidopsis seedlings are predominantly synthesized via the MEP pathway because of the following observations. First, the incorporation of [2-13 C]MVL into tZ and iP derivatives was consistently low (Ͻ20%), whereas nearly 90% of the isoprene units of CAM were labeled with 13 C in the same sample ( Fig. 2 and Table II). This observation indicates that the MVA pathway is not the primary route used to provide DMAPP to tZ-and iP-type CKs. Second, the level of 13 C incorporation into tZ and iP derivatives from [1-13 C]DX was as high as that into ABA, which is known to be biosynthesized mainly through the MEP pathway (35). On the other hand, we predict that the MVA pathway provides a greater proportion of DMAPP to the biosynthesis of cZ-type CKs in comparison to that of tZ-and iP-type CKs, because [2-13 C]MVL was introduced into these CKs nearly as efficiently as it was into CAM in the same feeding experiment. This idea is also supported by the significantly lower levels of 13 C incorporation into cZ-type CKs from [1-13 C]DX than that into ABA.
Subcellular Localization of AtIPTs-Because the prenyl group of tZ-and iP-type CKs is predominantly synthesized through the MEP pathway (Fig. 2), we speculated that some IPTs should use the DMAPP produced in plastids. Compared with the bacterial enzyme tmr, all AtIPTs except for AtIPT4 have an amino-terminal extension consisting of 30 -60 amino acids (7). The ChloroP 1.1 program (36) predicted that the amino-terminal regions of AtIPT1, -3, -5, and -8 might function to localize these AtIPTs to plastids. To obtain experimental evidence for the subcellular locations of IPT activities, we prepared DNA constructs designed to produce translational fusions of individual AtIPTs and GFP. Because some of the fusion proteins containing the full coding regions of AtIPTs tended to precipitate in cells (data not shown), the amino-terminal regions, which could function as transit peptides, were fused to GFP (for details, see "Experimental Procedures"). The plasmid

FIG. 2. Relative levels of 13 C incorporation into CKs, ABA, and CAM in [1-13 C]DX-fed Arabidopsis cla1-1 seedlings (A) and in [2-13 C]MVL-fed wild type seedlings in the presence of mevastatin (B).
The 13 C incorporation level was calculated using the molecular ion cluster shown in Table II after subtraction of natural 13 C abundance. 13 C incorporation levels were measured in three independent experiments. Error bars indicate S.D.
DNAs were introduced into both Arabidopsis leaf and root cells, and the transient expression of the chimeric genes was observed. The fluorescence of AtIPT1-, AtIPT3-, and AtIPT5-GFP in mesophyll cells was observed in plastids as the signal overlapped with autofluorescence derived from chlorophylls (Fig. 3,  A, B, D, and G). AtIPT8-GFP exhibited the same fluorescence pattern as AtIPT1 and -3 in root cells (Fig. 3, C, E, and H). However, the fluorescence of AtIPT4-and AtIPT2-GFP was distributed in the cytosol (Fig. 3, I, M, and N). It should be noted that an identical pattern of GFP fluorescence was observed when the GFP was fused to the carboxyl terminus of AtIPT4 (data not shown). The fluorescence of AtIPT7-GFP was observed in the mitochondria (Fig. 3, J and K). To avoid mislocation of the fusion proteins because of transient expression by a strong promoter, we next generated stable transformants of Arabidopsis expressing AtIPT3-and AtIPT7-GFP controlled by their native promoters. Consequently, the pattern of fluorescence produced by both of these constructs was essentially the same as the respective transiently expressed fusion proteins (Fig. 3, compare E with F and K with L). These results strongly suggest that four adenosine phosphate-IPTs, AtIPT1, -3, -5, and -8, localize in plastids in agreement with the idea that the MEP pathway primarily provides DMAPP for tZ and iP synthesis. As for tRNA-IPT, AtIPT2-GFP localized in the cytosol (Fig. 3, M and N).
We did not examine the location of AtIPT6 and -9, because AtIPT6 does not appear to encode a functional IPT in some Arabidopsis ecotypes owing to a point mutation (8) and because the enzyme activity of AtIPT9 could not be detected (data not shown).

DISCUSSION
Using a 13 C labeled precursor specific to the MEP and MVA pathways, we have demonstrated that both isoprenoid pathways can supply DMAPP to CKs in Arabidopsis seedlings (Fig.  2). In addition, subcellular localization of AtIPT-GFP fusion proteins has been determined in Arabidopsis cells. The results from GFP fluorescence observations suggest that IPT isozymes are distributed to multiple subcellular compartments (Fig. 4). These data now provide new insights into CK biosynthesis as discussed below.
The Origin of DMAPP for tZ-and iP-type CKs-The localization of several adenosine phosphate-IPTs to plastids (Fig. 3) is consistent with the role of the MEP pathway in providing DMAPP to tZ-and iP-type CKs (Fig. 2). The dominant role of the MEP pathway in the biosynthesis of iP-type CKs is in agreement with the recent finding that overexpression of the AtIPT8/PGA22 gene, the product of which was found in plastids when expressed as a GFP fusion protein (Fig. 3), drastically increased the level of iPRMP and iPR in Arabidopsis (37).
There are at least two possible explanations for the minor incorporation of [2-13 C]MVL into tZ-and iP-type CKs (Fig. 2). First, as has been suggested previously, there may be some exchange of common isoprenoid precursors (e.g. isopentenyl diphosphate or DMAPP) between the cytosol and plastids. This mechanism has been proposed to explain the minor incorporation of [2-13 C]MVL into ent-kaurene (a precursor for gibberellins), assuming that ent-kaurene synthesis occurs exclusively in plastids as supported by the subcellular localization of enzymes (38). However, in the case of the biosynthesis of tZ-and iP-type CKs, a small amount of 13 C label from [2-13 C]MVL can also be attributed to the function of another adenosine phosphate-IPT, AtIPT4, which is presumably present in the cytosol (Fig. 3I). In addition, the localization of AtIPT7-GFP in the mitochondria (Fig. 3, J-L) suggests that this enzyme may also use DMAPP from the MVA pathway, because ubiquinones, a group of isoprenoids in mitochondria, are synthesized principally through the MVA pathway (39). Therefore, a major role of the MEP pathway in the biosynthesis of tZ-and iP-type CK in Arabidopsis seedlings under the current growth conditions does not rule out a greater contribution of the MVA pathway to these CKs under different growth conditions, if the relative abundance of IPT isozymes is modulated. Thus, our present results do not contradict previous reports (22)(23)(24) indicating a role for the MVA pathway in providing DMAPP for tZ-type CKs. In this context, to fully understand the relative roles of the MEP and MVA pathways in CK biosynthesis, it will be informative to determine how individual AtIPT genes are regulated during plant development and under different environmental conditions.
Recently, Åstot et al. (24) measured the biosynthetic rates of tZRMP and iPRMP using in vivo isotope labeling and proposed an iPRMP-independent pathway for the biosynthesis of tZ-type CKs. In this model, a hydroxylated derivative of DMAPP is directly transferred to AMP (24). In light of this finding, our result showing the involvement of the MEP pathway in tZ biosynthesis is intriguing because hydroxymethylbutenyl diphosphate, the best hypothetical substrate in the iPRMPindependent pathway (24), has been known as an intermediate of the MEP pathway (Fig. 4) (40). In fact, it has been demonstrated recently (41) that TZS, an IPT from Agrobacterium, is capable of synthesizing tZRMP from hydroxymethylbutenyl diphosphate and AMP in vitro. To examine whether the iPRMP-independent pathway operates in plants, hydroxymethylbutenyl diphosphate should be tested as a substrate for AtIPT1, -3, -5, and -8 in vitro.
The Origin of DMAPP for cZ-type CKs-Based on the efficient 13 C incorporation of [2-13 C]MVL into cZ derivatives, relative to that into ABA and other CKs (Fig. 2), we postulate that a large portion of the prenyl group of cZ-type CKs originates from the MVA pathway. However, the level of [1-13 C]DX incorporation into cZ-type CKs was also significantly higher than incorporation into CAM, a cytosolic sterol that is mainly produced through the MVA pathway. Therefore, the MEP pathway appears to supply DMAPP to cZ derivatives at a level beyond the hypothesized exchange of isoprenoid precursors (Fig. 4). Unlike the possible participation of a P450 monooxygenase in the hydroxylation of iP to form tZ (12), the conversion of iP-type CKs into their corresponding cZ derivatives has not been re- ported. Therefore, this route cannot readily be hypothesized at the moment.
A probable explanation for the incorporation of MEP-derived DMAPP into cZ derivatives is isomerization of tZ-type CKs (Fig. 4). The presence of a cis-trans isomerase of zeatin has been reported in cell-free extracts from immature seeds of Phaseolus vulgaris (15). This enzyme catalyzes the isomerization of zeatin in both directions in favor of cZ to tZ and uses cZR and tZR as substrates as well. In our experimental conditions, the amounts of cZ-type CKs were much smaller than those of corresponding tZ-type CKs (Table I), and the majority of tZtype CKs were labeled with 13 C in cla1-1 seedlings fed with [1-13 C]DX (Fig. 2). Thus, even a small proportion of isomerization of tZ to cZ would account for the increase in 13 C incorporation from [1-13 C]DX into cZ-type CKs. In contrast, the rate of 13 C incorporation into tZ-type CKs from [1-13 C]DX would not be greatly affected by isomerization of cZ-type CKs because of their low levels in the current experimental system.
A role of the tRNA-dependent pathway has been proposed in the biosynthesis of cZ-type CKs, because the N 6 -prenylated adenine element of tRNA generally consists of iP-and cZ-type CKs but not tZ-type CKs (13). In addition, no hydroxylase activity has ever been shown to produce cZ derivatives from iP-type CKs. Our finding that the AtIPT2-GFP fusion protein localizes to the cytosol (Fig. 3) suggests that the N 6 -prenylation of tRNA takes place in this subcellular compartment in Arabidopsis and implies the use of DMAPP derived from the MVA pathway. In fact, our 13 C labeling experiments indicated that a large proportion of cZ derivatives are synthesized through the MVA pathway (Fig. 2), supporting the role of N 6 -prenylated tRNA in the biosynthesis of cZ-type CKs (Fig. 4). The incorporation of 14 C labeled MVA into the iP element of tRNA has been demonstrated previously in tobacco pith tissue (21).
In conclusion, the present study has demonstrated a critical contribution of the plastid-located MEP pathway in providing the prenyl precursor to tZ-and iP-type CKs in Arabidopsis seedlings. This conclusion is well supported by the localization of multiple AtIPT-GFP fusion proteins in plastids. In contrast, it is likely that the MVA pathway is mainly responsible for the biosynthesis of cZ derivatives, although our data do not exclude the possibility that a small fraction of cZ derivatives may be produced through isomerization of tZ-type CKs. It is still not clear whether cZ plays any critical role in plant development; however, the distinct origins of large fractions of DMAPP for tZ and cZ biosynthesis suggest that plants are able to separately modulate the level of these CK species. Further characterization of the biosynthesis pathway for cZ would be necessary to evaluate its physiological significance in plants. Additional work is required to uncover how multiple IPTs in varying subcellular compartments control the levels of different classes of CKs during plant development and in response to environmental cues.