Geranyl Diphosphate:4-Hydroxybenzoate Geranyltransferase from Lithospermum erythrorhizon CLONING AND CHARACTERIZATION OF A KEY ENZYME IN SHIKONIN BIOSYNTHESIS*

Two cDNAs encoding geranyl diphosphate:4-hy-droxybenzoate 3-geranyltransferase were isolated from Lithospermum erythrorhizon by nested PCR using the conserved amino acid sequences among polyprenyltransferases for ubiquinone biosynthesis. They were functionally expressed in yeast COQ2 disruptant and showed a strict substrate specificity for geranyl diphosphate as the prenyl donor, in contrast to ubiquinone biosynthetic enzymes, suggesting that they are involved in the biosynthesis of shikonin, a naphthoquinone secondary metabolite. Regulation of their expression by various culture conditions coincided with that of geranyltransferase activity and the secondary metabolites biosynthesized via this enzyme. This is the first established plant prenyltransferase that transfers the prenyl chain to an aromatic substrate. The prenylation reaction of aromatic substrates is involved in the biosynthesis of diverse molecules that play important roles in various biological activities in bacteria to mammals, e.g. electron via ubiquinone and plastoquinone In higher plants, largely the di-versification of aromatic secondary metabolites, to both chemical structures biological flavonoids among coq2 and its orthologs were used as forward and reverse primers, respectively. The latter sequence, 5 (cid:2) -GCIGTDSWYTTVATICC-3 (cid:2) , cor-responded to the amino acid sequence n-GIKSTA-c. The reaction mix- ture (0.3 (cid:1) l) was further used as the template for the second PCR, using AP-2 from the kit and another degenerate nucleotide (Rv-2), 5 (cid:2) -TCYT- GRTGDGCRTADATDGTRTC-3 (cid:2) (n-DTIYAHQD-c), as primers. The smear PCR products at about 800 bp were subcloned in pT7-blue (Amersham Biosciences Inc.) and randomly sequenced. Nine clones showed significant similarity with yeast 4HB:hexaprenyltransferase (4HPT), and could be separated into two independent groups. Both full-length clones were obtained by a 3 (cid:2) -rapid amplification of cDNA ends using a common forward primer 5 (cid:2) -CCCTTTGTGTTTGCTTA(C/ T)CCTCTC-3 (cid:2) . Finally, using the 5 (cid:2) and 3 (cid:2) sequences as specific prim- ers, the full-length cDNA clones were amplified with Pfu DNA polym-erase (Promega) to prevent an error in the former PCR with Taq DNA polymerase. The sequences of the cDNAs are in the DDBJ/GenBank TM DNA data bases under accession numbers AB055078 for LePGT-1 and AB055079 for LePGT-2, respectively. Gene The gene in by gene the Using pUG6 plasmid as a template, the geneticin (G418) resistance gene was amplified by PCR, in which the forward primer 5 (cid:2) -GTAAGGTTATCAGAAGGGCG-GAGTATACTATAGATTACAGTAGAA C-3 (cid:2) and the reverse primer 5 (cid:2) - GCACGCTATATCTACAAGAATCCA-AACAGTCTCAAGATGTAGTCG GCATAGGCCACTAGTGGATCTG-3 (cid:2) were used. In these sequences, the annealing sites to pUG6 are un-derlined, and the roman letters are the parts that are homologous to the yeast COQ2 gene. The PCR product was introduced into yeast strain W303-1A(a), and the disruption of COQ2 in G418-resistant clone was confirmed by genomic PCR. The disruptant (cid:3) coq2 showed the typical phenotype of a ubiquinone-defective mutant, in which it was capable of growing on minimal glucose medium but not on glycerol as the sole carbon source (26). Heterologous Expression in Yeast— Full-length LePGT-1 and LePGT-2 cDNAs were subcloned into yeast expression vector pDR196 (32), which was kindly provided by Dr. W. Frommer (University of Tu¨bingen), at Eco RI (5 (cid:2) ) and Xho I (3 (cid:2) ) sites. The sequences of the resulting plasmids, pDR-PGT1 and pDR-PGT2, were used to transform W303 (cid:3) coq2 , which were selected by SD medium ( (cid:4) uracil), by the lith- ium acetate method (33). For the expression of recombinant proteins, the yeast transformants cultured in 200 ml of YPAD medium were re-suspended in 10 ml of 20 m M KP i buffer (pH 7.4) containing 1.2 M sorbitol, and treated with Zymolyase Japan) for 60 min. After homogenization with a Downs glass homogenizer in 20 m M Tris-HCl buffer (pH 7.5) containing 0.6 M sorbitol, 1 m M phenylmeth- ylsulfonyl fluoride, and 10 m M dithiothreitol,

The prenylation reaction of aromatic substrates is involved in the biosynthesis of diverse molecules that play important roles in various biological activities in bacteria to mammals, e.g. electron transport via ubiquinone and plastoquinone (1). In higher plants, such prenylation largely contributes to the diversification of aromatic secondary metabolites, with regard to both their chemical structures and biological activities. For instance, some prenylated flavonoids act as phytoalexins that are involved in plant defense mechanisms (2,3), and have also been reported to be potential natural medicines (4 -6). Due to these interesting properties, intensive biochemical studies have been conducted to identify plant prenyltransferases which catalyze the transfer of prenyl diphosphate to an aromatic ring to generate a prenylated derivative, such as flavones (7,8), isoflavones (9 -12), coumarins (13,14), olivetolate (15), and 4-hydroxybenzoate (16). However, there has been no report on the identification of the genes that encode such prenyltransferases involved in the biosynthesis of these secondary metabolites.
Shikonin, a red naphthoquinone derivative, is a secondary metabolite that specifically occurs in boraginaceous plants, and is the active principle of the medicinal plant Lithospermum erythrorhizon (17). Since this compound and its derivatives exhibit antibacterial activity, their functions as phytoalexins have also been reported (18). The biosynthesis of shikonin includes a key prenylation step catalyzed by geranyl diphosphate (GPP) 1 :4-hydroxybenzoate (4HB) 3-geranyltransferase; i.e. coupling of the shikimate and mevalonate pathways (17,19,20). This enzyme plays a critical role in the regulation of shikonin biosynthesis in cell cultures of L. erythrorhizon, i.e. up-and down-regulation of this enzyme activity directly affects the production of shikonin (21,22). The enzyme activity is strongly suppressed by light irradiation, ammonium ion, and the synthetic auxin 2,4-dichlorophenoxyacetic acid (2,4-D) (23), whereas this activity is enhanced with the addition of oligogalacturonide (OG) (24) and methyl jasmonate (MJ) (25) to the medium. It has also been reported using a partially purified enzyme that this protein is ER membrane-bound and shows strict substrate specificity for geranyl (C 10 ) diphosphate for the chain length of the prenyl donor (16), which is in clear contrast to the mitochondrial polyprenyltransferase for ubiquinone biosynthesis (26). We attempted to isolate the cDNA of geranyltransferase from L. erythrorhizon as a model, since these characteristic properties should be advantageous for identifying the gene product.
In this report, we describe the isolation of two cDNAs for the prenyltransferase, designated LePGT-1 and -2 (L. erythrorhizon p-hydroxybenzoate:geranyltransferase) from L. erythrorhizon cultured cells, and functional expression of the gene products in yeast. A direct correlation between this gene expression and shikonin accumulation shows that they are involved in the secondary metabolism.

EXPERIMENTAL PROCEDURES
Plant Material, Reagents-Cell suspension cultures (strain M18) (27) were maintained in Linsmaier and Skoogs (LS) medium (28). They produced shikonin derivatives when cultured in M9 medium (29) in darkness (20 -23). In the present study, 4.5 ml of liquid paraffin was placed over 30 ml of M9 medium to extract shikonin derivatives produced by the cells (24), which interfere with the measurement of enzyme activity. GPP and dimethylallyl diphosphate were synthesized as previously described (30). Other chemical reagents were purchased from Wako Pure Chemicals (Japan), Nacalai (Japan), and Sigma.
Nested PCR and Rapid Amplification of cDNA Ends-A doublestranded cDNA library with an adapter sequence at both ends was synthesized from poly(A) ϩ RNA, which was prepared from the cells cultured in M9 medium for 4 days in the dark, with Marathon cDNA Amplification Kit (CLONTECH). In the first PCR, AP-1 from the kit and a degenerate primer (Rv-1) designed from the conserved sequences among coq2 and its orthologs were used as forward and reverse primers, respectively. The latter sequence, 5Ј-GCIGTDSWYTTVATICC-3Ј, corresponded to the amino acid sequence n-GIKSTA-c. The reaction mixture (0.3 l) was further used as the template for the second PCR, using AP-2 from the kit and another degenerate nucleotide (Rv-2), 5Ј-TCYT-GRTGDGCRTADATDGTRTC-3Ј (n-DTIYAHQD-c), as primers. The smear PCR products at about 800 bp were subcloned in pT7-blue (Amersham Biosciences Inc.) and randomly sequenced. Nine clones showed significant similarity with yeast 4HB:hexaprenyltransferase (4HPT), and could be separated into two independent groups. Both full-length clones were obtained by a 3Ј-rapid amplification of cDNA ends using a common forward primer 5Ј-CCCTTTGTGTTTGCTTA(C/ T)CCTCTC-3Ј. Finally, using the 5Ј and 3Ј sequences as specific primers, the full-length cDNA clones were amplified with Pfu DNA polymerase (Promega) to prevent an error in the former PCR with Taq DNA polymerase. The sequences of the cDNAs are in the DDBJ/GenBank TM DNA data bases under accession numbers AB055078 for LePGT-1 and AB055079 for LePGT-2, respectively.
Gene Disruption of Yeast-The COQ2 gene in yeast was disrupted by gene replacement with the homologous DNA (31). Using pUG6 plasmid as a template, the geneticin (G418) resistance gene was amplified by PCR, in which the forward primer 5Ј-GTAAGGTTATCAGAAGGGCG-GAGTATACTATAGATTACAGTAGAA CAGCTGAAGCTTCGTACG-C-3Ј and the reverse primer 5Ј-GCACGCTATATCTACAAGAATCCA-AACAGTCTCAAGATGTAGTCG GCATAGGCCACTAGTGGATCTG-3Ј were used. In these sequences, the annealing sites to pUG6 are underlined, and the roman letters are the parts that are homologous to the yeast COQ2 gene. The PCR product was introduced into yeast strain W303-1A(a), and the disruption of COQ2 in G418-resistant clone was confirmed by genomic PCR. The disruptant ⌬coq2 showed the typical phenotype of a ubiquinone-defective mutant, in which it was capable of growing on minimal glucose medium but not on glycerol as the sole carbon source (26).
Heterologous Expression in Yeast-Full-length LePGT-1 and LePGT-2 cDNAs were subcloned into yeast expression vector pDR196 (32), which was kindly provided by Dr. W. Frommer (University of Tü bingen), at EcoRI (5Ј) and XhoI (3Ј) sites. The sequences of the resulting plasmids, pDR-PGT1 and pDR-PGT2, were used to transform W303⌬coq2, which were selected by SD medium (Ϫuracil), by the lithium acetate method (33). For the expression of recombinant proteins, the yeast transformants cultured in 200 ml of YPAD medium were re-suspended in 10 ml of 20 mM KP i buffer (pH 7.4) containing 1.2 M sorbitol, and treated with Zymolyase (Seikagaku Kogyo, Japan) for 60 min. After homogenization with a Downs glass homogenizer in 20 mM Tris-HCl buffer (pH 7.5) containing 0.6 M sorbitol, 1 mM phenylmethylsulfonyl fluoride, and 10 mM dithiothreitol, the mixture was centrifuged at 10,000 ϫ g (5 min) and at 100,000 ϫ g (1 h) to pellet the microsome fraction. After washing, the membrane fraction was resuspended with 200 -500 l of 0.1 M Tris-HCl buffer (pH 7.5), which was used as an enzyme solution for the geranyltransferase assay described below. The reaction product was detected with high performance liquid chromatography when cold substrates were used. Substrate specificity for prenyl donor was examined using a radioactive assay.
Geranyltransferase Assay-Enzyme assay was carried out as described elsewhere (21). The reaction product, 3-geranyl-4-hydroxybenzoic acid (34), was quantitatively detected by high performance liquid chromatography with a Shimadzu LC-10A system: column, LiChrosphere 100RP-18 (Merck) 4 ϫ 250 mm; solvent system, methanol/H 2 O/ acetic acid (80:20:0.3); flow rate, 1 ml/min; detection, 230 -320 nm with a SPD6A photodiode array detector. The reaction product was identified by direct comparison with a standard sample. The microsomal fraction from the cultured cells of L. erythrorhizon was prepared as described previously (16). The values are the means of two replicates for each assay, and the assay was repeated more than three times. As marker enzymes for mitochondria and microsome, activities of cytochrome c oxidase and antimycin A insensitive NADPH-cytochrome c reductase were measured, respectively (35).
Radioactive Assay of Prenyltransferase by TLC-The enzyme assay mixture was the same as described above except that 24.2 M [universal- 14 C]4HB (specific activity 1.22 MBq/mol, Sigma) instead of cold 4HB and 400 M dimethylallyl diphosphate, GPP, farnesyl diphosphate, or geranylgeranyl diphosphate were used. Incubation was carried out in the same way as above, and the ethyl acetate-soluble portion was dissolved in 10 l of methanol, which was then loaded on a silica gel TLC plate (20 ϫ 20 cm). TLC was developed with benzene/ethyl acetate (8:2), and subjected to autoradiography to detect the reaction products. Their R F values are geranyl-4HB, 0.22; farnesyl-4HB, 0.28; gera-nylgeranyl-4HB, 0.34. These products were quantified with BAS2000 (Fuji film).
Northern Analyses of LePGT Expression-L. erythrorhizon cell cultures in M9 medium were agitated either in the dark or under continuous light (80 E/m 2 ⅐ s) with fluorescent lamps. Northern blot hybridization was carried out according to the standard protocol. The fragments to the 5Ј-flanking region, 680 bp for LePGT-1 and 756 bp for LePGT-2, were used as probes for the specific detection of each molecular species. The ␤-subunit of ATP-synthase was used as a control (36).
Inhibitors of geranyltransferase expression, NH 4 Cl (1 mM) and 2,4-D (1 M), were added to M9 medium and cells were harvested 7 days after inoculation, whereas positive regulators of shikonin production, MJ (10 M) (25) and OG (100 g/ml) (24), were added to LS medium supplemented with 1.2 M CuSO 4 . From intact plants of L. erythrorhizon, which had been cultivated in pots for 2 years, each organ was excised and RNA samples were prepared for Northern analyses.
Quantitative Analysis of Secondary Metabolites-Nearly all of the shikonin derivatives and dihydroechinofuran produced by the cells were secreted into the medium, and were then trapped with the paraffin layer. The amounts of these secondary metabolites in paraffin were estimated according to the standard method (24,37).

RESULTS
Cloning and Sequence Analysis of LePGT cDNAs-We attempted to isolate a cDNA encoding 4HB:geranyltransferase which is involved in the secondary metabolism of L. erythrorhizon by a molecular biological approach, since purification of this membrane-bound enzyme to homogeneity was unsuccessful (16). The enzymatic reaction of the geranyltransferase is functionally analogous to that of the prenyltransferase in ubiquinone biosynthesis, i.e. 4HB:polyprenyltransferase (Fig.  1A). Three genes for such prenyltransferases that accept 4HB and polyprenyl diphosphate as a prenyl acceptor and donor have been reported; 4HPT is the gene product of COQ2 of yeast (26), 4HB:polyprenyl diphosphate transferase (ppt1) of fission yeast (38), and 4HB:octaprenyltransferase (the gene product of ubiA) of Escherichia coli (39). Based on multiple alignment with these polyprenyltransferases and other eukaryotic orthologs in protein data bases, two degenerate PCR primers were designed from the conserved amino acid sequences. A library of cDNAs with adopter sequences was prepared from L. erythrorhizon cells cultured in M9 medium in the dark, in which the geranyltransferase was highly expressed. By nested PCR, a smear band was amplified, which was then subcloned and randomly sequenced. Nine DNA fragments that were significantly similar to yeast 4HPT were isolated and classified into two groups. Full-length clones corresponding to both groups were isolated by 3Ј-rapid amplification of cDNA ends, and were designated LePGT-1 and -2.
LePGT-1 (1,153 nucleotides) encodes an open reading frame corresponding to a protein of 307 amino acids, whereas LePGT-2 (1,240 nucleotides) has an open reading frame for 306 amino acids. The sequence identity between them was 76 and 93% on a nucleotide and amino acid basis, respectively. The amino acid sequence similarity to yeast 4HPT was only 35% for the full-length and 53% for the conserved region.
A multiple alignment among eukaryotic orthologs in the data bases is shown in Fig. 1B. Both LePGT clones possess the motif NDXXD for putative prenyl diphosphate binding (40), and a GX(K/Y)STAL sequence which seemed to be specifically conserved in this subfamily of 4HB:prenyltransferase, although the function of this domain is not yet known (Fig. 1B). Another highly conserved sequence, YDTIYAHQDK, is observed in the alignment among eukaryotic orthologs, but is not apparent in prokaryotic members. The polypeptide sequence of Arabidopsis has an extended leader peptide, which may be a mitochondrial signal according to the TargetP and ChloroP programs on CBS servers (www.cbs.dtu.dk/). LePGT-1 and -2 do not possess such a signal peptide for mitochondrial sorting, but are presumed to be sorted to the ER (41), which coincided the subcellular local-GPP:4HB Geranyltransferase in Shikonin Biosynthesis ization of native 4HB:geranyltransferase in L. erythrorhizon (42). Hydropathy analyses of LePGT polypeptides revealed that they are membrane-bound proteins, and the putative transmembrane domains are also depicted in Fig. 1B.
Functional Expression of LePGTs in a Yeast COQ2 Disruptant-To demonstrate the enzymatic activity of LePGT gene products, yeast was used for heterologous expression because of their high hydrophobicity. Wild type yeast (strain W303-1A), however, possesses 4HPT for ubiquinone biosynthesis localized in the mitochondria, which may raise the background level in the control experiment because of its broad substrate specificity for prenyl donor (43). Thus, the gene COQ2 was disrupted according to the method of Gü ldener et al. (31), and the disruptant W303⌬coq2, which is unable to use the nonfermentable carbon source glycerol, was used as a heterologous host to express LePGT cDNAs (Fig. 2). The growth defect of W303⌬coq2 on glycerol plate was complemented by re-introduction of COQ2 gene subcloned into a shuttle vector, pDR196, which had a PMA1 promoter to drive the heterologous gene in yeast cells. Full-length LePGT-1 and -2 subcloned into pDR196 were also introduced into W303⌬coq2, but they did not complement the growth on the glycerol medium.
3A), but the same fraction of disruptant did not show this enzyme activity. W303⌬coq2 transformed with full-length LePGT-1 and -2 were grown in YPAD medium and the enzyme activity was measured. The empty vector was used as a negative control. As depicted in Fig. 3A, both LePGT-1 and -2 showed an apparent activity of 4HB:geranyltransferase in the microsomal fraction using 4HB and GPP as the substrates, whereas the transformant of the empty vector gave almost no detectable activity. The activities of marker enzymes for mitochondria (cytochrome c oxidase) and ER (NADPH-cytochrome c reductase) in these fractions are also shown. Although the mitochondrial fraction contains ER membrane, the contamination of mitochondria in the microsomal fraction is very low. Geranyltransferase activity was undetectable in the supernatant (centrifugation at 100,000 ϫ g) of these two LePGT transformants. Using the method of Lineweaver and Burk (44), K m values of recombinant LePGT-1 and -2 for 4HB were determined as 10.3 and 53.8 M, whereas those for GPP were 5.1 and 45.9 M, respectively.
The most important criterion for determining whether the LePGTs are involved in shikonin biosynthesis is substrate specificity for the prenyl donor, since the geranyltransferase in the biosynthesis of shikonin and dihydroechinofuran is exclusively specific for GPP, whereas 4HPT for ubiquinone shows a broad substrate specificity regarding the chain length of prenyl diphosphate. Fig. 3B shows the substrate specificity of recombinant proteins of LePGT-1 and -2 in comparison with that of native mitochondrial 4HPT of wild type yeast. The yeast 4HPT has a wide preference for prenyl diphosphate of different chain lengths, e.g. it showed higher activity with geranylgeranyl diphosphate than with GPP as the substrate. In contrast, both gene products derived from L. erythrorhizon could accept only GPP and not dimethylallyl diphosphate, farnesyl diphosphate, or geranylgeranyl diphosphate as the substrate. This clear substrate specificity indicates that both LePGT-1 and LePGT-2 encode the geranyltransferases for secondary metabolism in this plant.
Expression of LePGT Genes in L. erythrorhizon-The activity of 4HB:geranyltransferase in L. erythrorhizon is dramatically induced when the cells are cultured in M9 medium in the dark, but not under light irradiation (21). In accordance with this enzyme activity, Northern analyses showed that both mRNAs were strongly induced in the dark whereas they were dramatically suppressed under illumination (Fig. 4A). These mRNA levels relative to those at day 0 are also shown in Fig. 4B, and are consistent with the accumulation pattern of shikonin derivatives in M9 medium (Fig. 4C).
Geranyltransferase also plays a critical role in the chemical regulation of shikonin production, i.e. this activity is up-regulated with MJ (25) as well as OG (24), whereas its expression is down-regulated by NH 4 ϩ and the auxin, 2,4-D (23). The effects of these chemical regulators on LePGT gene expression were analyzed by Northern hybridization. The effects of MJ and OG were investigated in LS medium in which geranyltransferase activity remained very low, but with the addition of both regulators, the LePGT mRNA levels clearly increased (Fig. 5, A and B, left) along with the enzyme activity (Fig. 5C, left). The increase in geranyltransferase activity reflected the accumulation of dihydroechinofuran, which is quickly biosynthesized from the enzymatic reaction product, 3-geranyl-4-hydroxybenzoic acid (Fig. 1A), and is therefore a more suitable metabolite than shikonin for monitoring the effect of geranyltransferase induction (24, 45) (Fig. 5D, left). The effects of inhibitors, which were observed in M9 medium, were also consistent with the mRNA level, enzyme activity, and secondary product accumulation, i.e. the accumulation of both LePGT mRNAs as well as the enzyme activity apparently decreased with NH 4 ϩ or 2,4-D, along with the dihydroechinofuran level (Fig. 5, right panels).
In the intact plant, shikonin is solely accumulated in underground parts (34), since its biosynthesis is completely inhibited under light conditions. The mRNAs of both LePGT genes were undetectable in aerial plant tissues, and were exclusively detected in root tissues, similar to shikonin accumulation, as shown in Fig. 6. Along with the strict substrate preference for GPP mentioned above, these expression studies support the notion that both cDNAs encode the geranyltransferases involved in shikonin biosynthesis.

DISCUSSION
In this study, we cloned and performed structural analyses of two cDNAs that encode geranyl diphosphate:4-hydroxybenzoate 3-geranyltransferase from cultured L. erythrorhizon cells (Fig. 1). Heterologous expression in yeast showed that both gene products exhibit enzyme activity to transfer the geranyl moiety to an aromatic substrate 4HB, and in particular they show exclusive specificity for the C 10 substrate GPP (Fig. 3), which are critical points for determining that these enzymes are responsible for shikonin biosynthesis (16).
Each organism has a specific chain length for prenyl moiety of ubiquinone molecule, and the side chain is usually penta-to decaprenyl residue, i.e. ubiqiunone-5 to -10. It is very unlikely that the diprenyl-4HB, which is the specific product of LePGTs, is converted to ubiquinone having geranyl residue as the prenyl chain. To exclude the possibility that LePGTs might be involved in ubiquinone biosynthesis in vivo, complementation studies were also done. The yeast COQ2 disruptant was transformed with LePGT-1 or -2, but the transformants failed to grow on glycerol medium, whereas the COQ2-containing plasmid, as a positive control, complemented the phenotype of disruptant (Fig. 2). Similarly, the plant ortholog AtPPT-1 also showed the same complementation in COQ2-disrupted fission

GPP:4HB Geranyltransferase in Shikonin Biosynthesis
yeast. 2 These results strongly support our conclusion that the LePGT cDNAs are not involved in ubiquinone biosynthesis in L. erythrorhizon.
The K m values of LePGT-1 to both substrates are apparently different from those of LePGT-2. As the K m values of native LePGT were reported to be 18.4 M for 4HB and 13.8 M for GPP (16), which are similar to those of LePGT-1 (10.3 and 5.1 M, respectively), but a direct comparison is not appropriate, because the K m values of native enzyme were determined with a partial purified protein, and thus it may be a mixture of LePGT-1 and -2. It is not clear whether these two gene products play different roles in vivo, but their expression patterns in L. erythrorhizon are very similar (Figs. 4 -6). They might be expressed in different cell types in root tissue.
LePGT polypeptides are predicted to have 8 or 9 transmembrane domains based on a hydropathy analysis by Kyte and Doolittle and a program on CBS servers (www.cbs.dtu.dk/services/TMHMM-1.0/). There are two obvious hydrophilic loops where the NDXXD motif and GX(K/Y)STAL motif are localized. They are presumed to be the binding sites for the substrate, i.e. the former is for prenyl diphosphate as seen in prenyl diphosphate synthases, and the latter is probably important for the aromatic substrate (39). More detailed analyses are needed to define the differences in the structure that contribute to the strict substrate specificity of LePGT for GPP.
The phylogenetic tree is shown in Fig. 7. Bacterial polyprenyltransferases tend to form a cluster, whereas eukaryotic polyprenyltransferases do not. This is at least partly attributable to the divergence of the mitochondrial sorting signal sequence in each molecular species. In particular, a plant ortholog of LePGT, Arabidopsis polyprenyltransferase, seems to be more divergent than those of unicellular eukaryotes and Caenorhabditis elegans (46). Although the mitochondrial targeting signal is absent in LePGT polypeptides, they may be derived from such prenyltransferase for ubiquinone biosynthesis. It is presumed that during evolution they lost the mitochondrial signal to localize in the ER membrane, which is advantageous both for the biosynthesis and secretion of shikonin out of cells, and gained substrate specificity for GPP. Since most organisms need ubiquinone to survive in the wild, Lithospermum cells should also have another 4HB:prenyltransferase localized in mitochondria. In fact, four to five bands were observed in the genomic Southern blot (data not shown).
Northern analyses revealed that the modes of regulation of gene expression were completely consistent with that of geranyltransferase activity using both positive and negative regulators (Figs. 4 -6). The other biosynthetic genes involved in shikonin biosynthesis, such as phenylalanine ammonia-lyase (45) and 4-coumarate CoA ligase (47) which may be also involved in lithospermic acid B production in this plant (48), did not show such a close correlation with shikonin accumulation. This indicates that LePGTs play a pivotal role in the regulation of shikonin biosynthesis at the gene expression level. Due to these attractive properties, LePGTs are suitable for use as model genes to study the molecular mechanism of dark-inducible secondary metabolism, and also its regulation by auxins and elicitors in higher plants.
In contrast to prenyltransferases involved in chain elongation of prenyl diphosphate, those involved in 4HB prenylation have not yet been subjected to molecular analyses. LePGTs isolated in this study are the first cDNAs that encode plant prenyltransferases which catalyze the transfer of the prenyl moiety to an aromatic substrate. They could be very powerful tools for characterizing prenylation of the aromatic moiety at a molecular level, e.g. identification of the 4HB-binding site. An interesting feature of these prenyltransferases is their clear preference for GPP as the substrate; this should be further characterized using chimeric enzymes with prenyltransferases 2 K. Okada, personal communication.   that show broad substrate specificity, such as 4HPT. From a biotechnological point of view, these cDNAs are expected to be useful for genetic engineering of shikonin production (49), since geranylation is a "bottle-neck" reaction in shikonin biosynthesis. Furthermore, as heterologous probes, these cDNAs may lead to breakthroughs in molecular genetic studies on the secondary metabolism of cannabinoid and prenylated flavonoids and coumarins, which involve prenyltransferase reactions.