Archaebacterial Ether-linked Lipid Biosynthetic Gene EXPRESSION CLONING, SEQUENCING, AND CHARACTERIZATION OF GERANnGERANn-DIPHOSPHATE SYNTHASE*

of isopentenyl diphosphate with allylic diphosphates to produce GGPP which is the important precursor of archaebacterial ether-linked lipids. We developed an expression screening method for cloning the GGPP synthase gene, which utilizes the carotenoid biosynthesis genes of Erwinia uredouora to visualize a clone expressing GGPP syn- thase, and then screened a genomic DNA library from S. acidocaldarius for the GGPP synthase gene by using this method. Positive clones were shown to contain GGPP synthase gene by the use of an in vitro assay.

payment of page charges. This article must therefore be hereby marked indicate this fact.
(1). Archaebacterial lipids are mainly ether-linked lipids in place of the usual membrane ester-linked phospholipids of eubacteria and eukaryotes and enable the organisms to inhabit extreme environments. These lipids are composed of a glycerol group (or more complex polyols, i.e. calditol) and saturated isoprenoid moieties with 20, 25, or 40 carbon atoms. I t is proposed that geranylgeranyl diphosphate, which is synthesized by a cytosolic GGPP' synthase, is attached to glyceryl phosphate, followed by saturation of the geranylgeranyl moieties (phytanylation), head to head dimerization to produce C,, ether-linked lipids, and a variety of other modifications (2-4).
The archaebacterium Sulfolobus acidocaldarius, which grows at temperatures up to 90 "C at pH 2, contains three major classes of ether-linked lipids, diphytanyl glycerol diethers, glycerol dibiphytanyl glycerol tetraethers, and a glycerol dibiphytanyl nonitol tetraether (1, 5). Lipids containing cyclopentane rings in the phytanyl moiety are also found, and the degree of cyclization increases with increasing growth temperature. The structural properties and stability of membranes composed of these lipids are very interesting, but nothing is known about the genes related to the biosynthesis of these lipids. Geranylgeranyl diphosphate is a precursor for the etherlinked lipids, and the biosynthesis of this compound is essential for archaebacteria. GGPP plays a different role in organisms other than archaebacteria. In plants and some bacteria, it is the precursor of carotenoids. In eukaryotes, GGPP is also the precursor for the C,, lipid moiety in prenylated proteins (6-141, which has attracted special interest recently with respect to signal transduction and protein sorting in cells. GGPP synthase, which belongs to a family of enzymes classified as prenyltransferases, catalyzes the head-to-tail condensation between isopentenyl diphosphate and an allylic diphosphate to give an amphiphilic molecule containing 4 isoprene units. The GGPP synthase genes only related to carotenoid biosynthesis were cloned from a photosynthetic bacteria (15), phytopathogenic bacteria (16), Neurospora crassa (171,and Capsicum annuum (18). In Erwinia uredovora, carotenoid biosynthesis genes (GGPP synthase (crtE), phytoene synthase (crtB), phytoene desaturase ( c r t l ) , lycopene cyclase (crtY), 6-carotene hydroxylase ( c r t z ) , and zeaxanthin @-glucosidase (crtX)) exist in the form of a gene cluster and cells of EScherichia coli transformed with this gene cluster produced carotenoids to become yellow (16).
To facilitate research of archaebacterial membranes, we developed an expression screening method for the GGPP syn-FPP, farnesyl diphosphate; GGPS, geranylgeranyl-diphosphate syn- The abbreviations used are: GGPP, geranylgeranyl diphosphate; thase; ORF, open reading frame; GlcNAc-l-P-transferase, UDP-Glc-NAc-dolichyl-phosphate N-acetylglucosaminephosphotransferase; Mur-NAc-l-P-transferase, phospho-N-acetylmuramoyl-pentapeptide-transferase; kb, kilobase pair(s). thase gene, which utilizes the carotenoid biosynthetic genes of E. uredovora to visualize a GGPP synthase expressing clone. A gene for thermostable archaebacterial GGPP synthase was isolated using this color selection method. This is a first report of a gene for biosynthesis of archaebacterial membrane. General Procedures-Restriction enzyme digestions, transformations, and other standard molecular biology techniques were carried out as described by Sambrook et al. (19).
Construction of S. acidocaldarius Genomic Library-S. acidocaldarius was grown in 1 liter of ATCC medium 1723 a t 70 "C and harvested according to the method described in the ATCC catalogue. The genomic DNA was isolated and purified as described by Ausubel et al. (20). Genomic DNA was partially digested with Sau3AI. The resulting restriction fragments were then fractionated by 0.5% agarose electrophoresis. The region of the gel between 3 and 6 kb was cut out, and the DNA in the gel was isolated by the NaI method. The Sau3AI fragments (2.7 pg) were ligated with 1.4 pg of BamHI cut dephosphorylated pUC119 and transformed into E. coli DH5a. The clones were stored at Construction of Competent Cells Containing pACYC-IB-Plasmid pCAR25, which contained the gene cluster of carotenoid biosynthesis, was digested with SnaBI and HpaI. The 2.8-kb fragment, which contained the crtl and crtB genes, was recovered, ligated with a n EcoRI linker, and digested with EcoRI. The resulting DNA was ligated with EcoRI cut dephosphorylated pACYC184 to construct pACYC-IB and used to transform E. coli DH5a. Competent cells were obtained from DH5dpACYC-IB by the CaC1, method.
Expression Cloning of S. acidocaldarius GGPP Synthase Gene-Plasmids were isolated from the Sulfolobus genomic library described above by the alkaline method. The competent cells, which contained pACYC-IB, were transformed with 10 ng of DNA from the genomic library and spread on LB plates containing tetracycline (50 pg/ml) and ampicillin (50 pg/ml). Since the competent cells express phytoene synthase (crtB) and phytoene desaturase (crtl), those transformed by a plasmid that contains the GGPP synthase gene produce lycopene and become red. Ten red positive clones were obtained from about 4,000 transformants. The pUC119 plasmids that contained inserts of S. acidocaldarius DNA fragment were recovered from the red-colored cells, and each clone was analyzed.
DNA Sequence Analysis-All DNA sequences were determined by the dideoxy chain termination method. Computer analysis and comparison of DNA sequences were performed using GENETYX genetic information processing software.
Measurement of GGPP Synthase Actiuity-E. coli DH5a cells were transformed with the pUC119 derivative described above. The colonies were used to inoculate 100 ml of LB medium containing ampicillin (50 pg/ml) and were incubated at 37 "C overnight. The cells were harvested and disrupted by sonication in 4 ml of 50 mM Tris-HC1 buffer, pH 7.0, containing 10 mM 2-mercaptoethanol and 1 mM EDTA. The homogenate was heated at 55 "C for 60 min and then centrifuged at 10,000 x g for 10 min. The supernatant was used to assay for GGPP synthase activity. The assay mixture contained, in a final volume of 1 ml, 0.48 nmol of [l-'4Clisopentenyl diphosphate (1.92 GBq/mmol), 25 nmol of (al1-E)-FPP, 5 pmol of MgCl,, 20 pmol of Tris-HC1 buffer, pH 6.8, and 0.3 mg of enzyme. This mixture was incubated at 55 "C for 30 min, and the reaction was stopped by chilling the reaction mixture in a n ice bath. The mixture was shaken with 3.0 ml of 1-butanol which had been saturated with H,O. The 1-butanol layer was washed with water, and radioactivity in the butanol layer was determined.
-70 "C. Products Analysis-Polyprenyl diphosphates produced by the enzymatic reaction were treated with acid phosphatase according to the method of Fujii (21). The hydrolysates were extracted with pentane, and the pentane-soluble products were analyzed by reversed phase thin layer chromatography using LKC-18 developed with acetone/H,O (9/1) and normal phase thin layer chromatography using Kieselgel60 developed with benzene/ethyl acetate (9/1). Authentic standard alcohols were visualized with iodine vapor, and the distribution of radioactivity was detected by scanning the plate with an Aloka radiochromatoscanner and by autoradiography.
Partial Purification of Cloned GGPP Synthase-The extract obtained from E. coli DH5dpGGPS3 was heat-treated, and the supernatant fraction was precipitated with 3040% saturation of (NH,),SO,. The precipitated protein fraction was dialyzed and chromatographed on a DEAE-Toyopearl650M column (1.0 x 16 cm) equilibrated with Buffer A (10 mM Tris-HC1 buffer, pH 7.7, 1 mM EDTA). Elution was performed with a linear gradient from 0 to 0.85 M of NaCl in Buffer A. Fractions containing GGPP synthase were collected and dialyzed against Buffer A. The dialysate applied to a Mono Q column (5 x 50 mm) equilibrated with Buffer A. Elution was performed with a linear gradient of 0-0.85 M of NaCl in the same buffer. The GGPP synthase fraction was analyzed by SDS-polyacrylamide gel electrophoresis (10%) after staining with Coomassie Brilliant Blue.

Construction of a Red-White Screening System for the GGPP Synthase Gene-A new expression method was constructed in
order to screen for the GGPP synthase gene. A genomic library of S. acidocaldarius was used to transform an E. coli strain that expressed the genes for phytoene synthase and phytoene desaturase. Concurrent expression of the gene for GGPP synthase from S. acidocaldarius produces lycopene, and the normally white transformants becomes red. In order to express the genes for phytoene synthase and phytoene desaturase, a DNA fragment from E. uredovora containing crtB and crtl, which coded phytoene synthase and phytoene desaturase (161, respectively, was ligated into the EcoRI site of pACYC184 (Fig. 1). Plasmid pACYC184 is a multicopy cloning vector that is able to co-exist with pUC vectors with a ColEl origin and carries the chloramphenicol resistance gene (Cm) and the tetracycline resistance gene (Tc). We transformed E. coli DH5m with pA-CYC-IB and made competent cells from the transformant. Cells carrying pACYC-IB were white. Upon transformation with pSF21, which contained the E. uredouora GGPP synthase gene (crtE) and directed synthesis of the enzyme (data not shown), the cells became red.
Expression Cloning of S. acidocaldarius GGPP Synthase Gene-A genomic DNA library from S. acidocaldarius (ATCC 33909) was constructed by inserting partially Sau3AI-digested DNA into the BamHI site of pUC119. Cells carrying pACYC-IB were transformed with the library. Approximately 4,000 recombinants were screened, resulting in identification of 10 red colonies.
Analysis of the Clones-Plasmids from the red colonies were examined and found to contained inserts ranging from 4.7 to 6.5 kb (Fig. 2) ). Therefore, in order to distinguish the thermostable GGPP synthase from thermolabile prenyltransferases derived from the E. coli host, cell homogenate was heated at 55 "C for 60 min prior to the enzyme assay. The prenyltransferase activities were assayed by a butanol extraction method. This procedure measures activity for No heat treatment was carried out in preparation of the extracts. The enzymatic activity was measured at 37 "C. several prenyltransferases, including FPP synthase, GGPP synthase, hexaprenyl-diphosphate synthase, and cis-polyprenyl-diphosphate synthase that were expected to reside in S. acidocaldarius (1). All clones that derived from the red colonies produced butanol-extractable products (Table I).
Product Analysis-Although red colonies of E. coli co-transformant indicated that the pUC-derived plasmids contained the GGPP synthase gene, other prenyltransferases, for example, FPP synthase, hexaprenyl-diphosphate synthase, might produce geranylgeranyl diphosphate as a minor component. Therefore the reaction products of the enzymes produced in E. coli harboring the pUC-derived plasmids were hydrolyzed and analyzed. As shown in Fig. 3, radiochromatographic analysis clearly indicates that the product alcohol derived from pGGPS3 is (all-E)-geranylgeraniol. The product alcohol derived from other plasmids was also analyzed and determined to (all-E)-             acidocaldarius;N.c.,N. crassa(17);E.u.,E. uredovora(16);E.c.,E. coli(25);S.c.,Saccharomyces cerevisiae (26,27). geranylgeraniol (data not shown). Hence the cloning of the gene obtained from S. acidocaldarius was confirmed as GGPP synthase gene.

E I L F A I L A V Y F Q T V T I T I *
Nucleotide Sequence-The nucleotide sequence of the 2.3-kb Hind111 fragment in pGGPSl was determined by the dideoxy chain termination method according to the strategy shown in Fig. 2. There are two ORFs (designated as ORF-1 and ORF-2) in same strand. The nucleotide sequence of the HindIII fragment is shown in Fig. 4. ORF-1 begins with an ATG codon at position 40 and terminates with a TAA codon at position 1029.
A putative Shine-Dalgarno sequence, which was also reported with respect to S. acidocaldarius (241, was found at nucleotides 31-36 (GAGAAG). ORF-1 contains 990 nucleotides, which encodes a 330-amino acid protein with a calculated relative molecular weight (M,) of 36,873.ORF-2 begins with an ATG codon at position 1032 and terminates with a TAG codon at position 2016. A putative SD sequence also exists at the upstream of the start codon. ORF-2 contains 984 nucleotides, which encodes a 328-amino acid protein with a calculated relative molecular weight (M,) of 35,470.
Amino Acid Sequence Homology-The deduced amino acid sequences of proteins encoded by ORF-1 and ORF-2 were compared with those for other prenyltransferases (15,17,(25)(26)(27). The protein encoded by ORF-1 has significant homology with other prenyltransferases (Fig. 5). The protein exhibits 25% identity and 61% similarity with GGPP synthase from E. uredouora and also shows homology with GGPP synthase from Cyanophora paradoxa, FPP synthase from Bacillus stearothermophilus and E. coli, and hexaprenyl-diphosphate synthase from yeast. To confirm whether ORF-1 encodes GGPP synthase, plasmid pRV11-1 was constructed by deletion of ORF-2 region with ExoIII. The activity of GGPP synthase in DH5d pRV11-1 transformant was determined (Table I). The activity was almost the same as those obtained from the original plasmids. Thus, ORF-1 encodes GGPP synthase. With respect t o ORF-2, data bank searches revealed that the protein encoded by ORF-2 had homology with UDP-N-acetylglucosamine-dolichyl-phosphate N-acetylglucosaminephosphotransferase (Glc-NAc-1-P-transferase) (EC 2.7.8.15) and phospho-N-acetylmunucleotides are numbered from the end of HindIII site. The ribosomebinding sites for GGPP synthase and ORF-2 are underlined. The initiation codons of ORF-1 and ORF-2 are boxed. ramoyl-pentapeptide-transferase (MurNAc-1-P-transferase) (EC 2.7.8.13) (data not shown). Partial Purification of the Cloned GGPP Synthase-The cloned GGPP synthase was partially purified in three steps. The specific activity of the partially purified enzyme is 8.7 nmol/min/mg of protein.
Substrate Specificity of Cloned GGPP Synthase-The substrate specificity of GGPP synthase was studied using five allylic diphosphates. As shown in Table 11, dimethylallyl diphosphate, geranyl diphosphate, and (all-E)-FPP are good substrates, with maximal activity from dimethylallyl diphosphate. The reaction products were analyzed as shown in Fig. 6. In any case, the product with the longest chain length was geranylgeraniol.
Effect of Temperature on Enzyme Stability-The enzyme was heated at the indicated temperature and then assayed for the activity (Fig. 7). Most of the activity for GGPP synthase remained even after heating a t 60 to 70 "C for 100 min, although 90% of the enzyme activity was lost on incubation at 90 "C for 10 min.
Effect of pH-The enzyme activity was measured in various pH from 4 to 11. It is difficult to assay below pH 4 because of the instability of the allylic diphosphates due to a spontaneous metal ion-dependent hydrolysis at low pH. The optimal pH is 5.8. This value is slightly smaller than those of other previously characterized prenyltransferases, FPP synthase (silkworm, optimal pH 7 (2811, cis-polyprenyl-diphosphate synthase (rat, pH 8.5 (29); B. subtilis, pH 8.5 (30)).

DISCUSSION
In the present study, we report a new expression screening method for the geranylgeranyl diphosphate synthase gene and the cloning of the gene from S. acidocaldarius. In expression screening, it is essential that the gene introduced into a host cell is transcribed and translated to produce active enzyme. Although it was established that the crtl and crtB reporter genes from E. uredovora were expressed in E. coli, it was not clear if the GGPP synthase gene derived from S. acidocaldarius was also expressed in E. coli. It has been reported that S. acidocaldarius genes include Shine-Dalgano-like sequences (241, but not the eubacterial promoter sequences (31). Promoters in S. acidocaldarius gene are constructed with two conserved sequence elements, a distal promoter element located between positions -38 and -25 and a proximal promoter element located between positions -11 and -2. The distal promoter element is TATA-like and more essential than the proximal promoter element. Although the arrangement of the distal promoter element and proximal promoter element resembles that of the two defined eubacterial promoter elements (-35 region and -10 region), there is no sequence similarity between these elements. Promoters in archaebacteria are more similar to those of eukaryotes rather than eubacteria (31). However some archaebacterial protein-encoding genes have been cloned (W), and 100 "C (A) for a indicated period and then the remaining enzyme activity was determined as described under "Experimental Procedures." by exploiting their ability to direct protein synthesis and to complement mutations in E. coli (32)(33)(34). In this study, we obtained several clones that showed GGPP synthase activity. The gene seems to be transcribed from the E. coli promoter-like sequence in S. acidocaldarius genome. ORF-1 of pGGPS2, -3, and -4 is in the reverse direction from the lacZ promoter in the parental pUC vector. The mRNA is translated by using the Shine-Dalgarno-like sequence that is found in upstream of GGPP synthase coding region. This system for expression screening should also apply for cloning other genes for GGPP synthase, including mammalian GGPP synthase, whose product is attached to the heterotrimeric G-protein y subunit and other small molecule G-proteins.
Comparisons between the deduced amino acid sequence for S. acidocaldarius GGPP synthase and the sequences of other cloned prenyltransferases, including FPP synthase, GGPP synthase, and hexaprenyl-diphosphate synthase, reveals significant identity and overall homology. In previously sequenced prenyltransferases, there are two highly conserved aspartaterich domains (I,L,V)XDDXX(D,N), where X encodes any amino acid. It has been proposed that they are involved in enzyme catalysis, where one domain serves as a binding site for the metal ion salt of the diphosphate moiety of isopentenyl diphosphate and the other for the allylic diphosphate. S. acidocaldarius GGPP synthase also has two aspartate-rich domains. However the seventh position of the second domain of S. acidocaldarius GGPP synthase is glycine instead of aspartic acid (Fig. 5 ) . Poulter et al. (35) recently examined the role of the second aspartate-rich domain in rat FPP synthase by using site-directed mutagenesis. The substitution of glutamate for the aspartic acid at position 7 did not result in altered enzyme activity. These data indicate that the seventh position of the second aspartate-rich domain is not involved in catalytic center of prenyltransferase.
The cloned Hind111 fragment contains two open reading frames. One is GGPP synthase and the other has homology with GlcNAc-1-P-transferase and MurNAc-1-P-transferase. GlcNAc-1-P-transferase catalyzes the first step of a N-linked glycoprotein biosynthesis and MurNAc-1-P-transferase acts in a peptidoglycan biosynthesis. Both GlcNAc-1-P-transferase and MurNAc-1-P-transferase utilize UDP sugar and cis-polyisoprenyl phosphate. In archaebacteria, except for some methanobacteria, cell membrane glycoproteins work to maintain the shape of cell instead of the peptidoglycans found in eubacteria (36). The cell wall from S. acidocaldarius contains two glycoproteins of molecular weight 40,000 and 100,000 as major subunits with glucose and mannose as major carbohydrate components (37). The sugar chain structure of the cell surface glycoproteins and lipid intermediates were only determined for halophiles (3840). N-Acetylgalactosamine or glucose instead of N-acetylglucosamine directly linked to the asparagine residue of the protein and dolichyl compound, whose a-isoprene unit was saturated, worked as a sugar carrier lipid. Therefore, the enzyme encoded by ORF-2 seems to catalyze the reaction between UDP sugar and dolichyl phosphate. The region containing ORF-1 and -2 may be a n operon of membrane biosynthesis genes.
The cloned GGPP synthase catalyzes the consecutive oligomerization C, + C,, * C,, 4 C20. Dimethylallyl diphosphate is the favorite substrate. In the case of dimethylallyl diphosphate and geranyl diphosphate, the products with shorter chain lengths than GGPP were obtained. We characterized prenyltransferase activities in s. acidocaldarius. Cell homogenate from S. acidocaldarius showed activities to produce FPP, GGPP, hexaprenyl diphosphate, and decaprenyl diphosphate. In the purification procedure, FPP synthase activity always associated with GGPP synthase activity, and another prenyltransferase that produced only FPP was not observed (data not shown). Since squalene, which is derived from farnesyl diphosphate, is obtained from archaebacteria, GGPP synthase seems to produce FPP as an intermediate.
In summary, segments of the S. acidocaldarius GGPP synthase gene were cloned by isolating red-colored cells that produced lycopene, followed by identification of the enzyme activity. These results will allow for a detailed analysis of the structure and function of the enzyme in the future by standard biochemical techniques and may help to provide a better understanding of consecutive condensation of prenyltransferase, and the DNA fragments cloned in this work will be useful in the studies of archaebacterial membrane biosynthesis and its regulation.