Dimer Formation of Octaprenyl-diphosphate Synthase (IspB) Is Essential for Chain Length Determination of Ubiquinone*

Ubiquinone (Q), composed of a quinone core and an isoprenoid side chain, is a key component of the respiratory chain and is an important antioxidant. In Escherichia coli, the side chain of Q-8 is synthesized by octaprenyl-diphosphate synthase, which is encoded by an essential gene, ispB. To determine how IspB regulates the length of the isoprenoid, we constructed 15 ispB mutants and expressed them inE. coli and Saccharomyces cerevisiae. The Y38A and R321V mutants produced Q-6 and Q-7, and the Y38A/R321V double mutant produced Q-5 and Q-6, indicating that these residues are involved in the determination of chain length. E. colicells (ispB::cat) harboring an Arg-321 mutant were temperature-sensitive for growth, which indicates that Arg-321 is important for thermostability of IspB. Intriguingly, E. coli cells harboring wild-type ispB and the A79Y mutant produced mainly Q-6, although the activity of the enzyme with the A79Y mutation was completely abolished. When a heterodimer of His-tagged wild-type IspB and glutathioneS-transferase-tagged IspB(A79Y) was formed, the enzyme produced a shorter length isoprenoid. These results indicate that although the A79Y mutant is functionally inactive, it can regulate activity upon forming a heterodimer with wild-type IspB, and this dimer formation is important for the determination of the isoprenoid chain length.

and S. cerevisiae (6). Each of these organisms has a specific isoprenoid side chain length as part of the ubiquinone molecule, e.g. Q-6 for S. cerevisiae, Q-8 for E. coli, Q-9 for rat, and Q-10 for S. pombe and human. For this reason, ubiquinone species have been used for classification in microbial taxonomy (7). The length of the side chain of ubiquinone is precisely defined by the action of polyprenyl-diphosphate synthases, but not by 4-hydroxybenzoate-polyprenyl-diphosphate transferases, which catalyze the condensation of 4-hydroxybenzoate and polyprenyl diphosphate (8). When various polyprenyldiphosphate synthase genes, such as the mutant GGPP synthase gene from Sulfolobus acidocaldarius, the hexaprenyldiphosphate synthase gene (COQ1) from S. cerevisiae, the heptaprenyl-diphosphate synthase gene from Haemophilus influenzae, the octaprenyl-diphosphate synthase gene (ispB) from E. coli, the solanesyl-diphosphate synthase gene (sdsA) from Rhodobacter capsulatus, and the decaprenyl-diphosphate synthase gene (ddsA) from Gluconobacter suboxydans, are expressed in an S. cerevisiae COQ1 mutant, each transformant produced mainly Q-5, -6, -7, -8, -9, and -10, respectively (9). When COQ2, which encodes 4-hydroxybenzoate-hexaprenyldiphosphate transferase in S. cerevisiae, was expressed in an E. coli ubiA mutant cell line, the transformant produce Q-8, but not Q-6 (10). These results indicate that prenyl-diphosphate synthase determines the chain length of ubiquinone and that 4-hydroxybenzoate-polyprenyltransferases can accept the various isoprenoid chains as a substrate.
In E. coli, ispB is an essential gene, responsible for the biosynthesis of both ubiquinone and menaquinone (11,12). An E. coli ubiA mutant, which does not produce Q-8, is not able to grow on a non-fermentable carbon source, but can grow on glucose (10). However, an E. coli ubiA Ϫ menA Ϫ mutant, which lacks both ubiquinone and menaquinone biosynthesis genes, can grow only when a small amount of Q-8 is still produced by leakiness of the mutations. Thus, ubiA Ϫ menA Ϫ mutants with an absolute lack of production of Q-8 and menaquinone-8 cannot be isolated.
Long-chain polyprenyl-diphosphate synthases (C 40 , C 45 , and C 50 ) catalyze the condensation of FPP, which acts as a primer, and IPP to produce each prenyl diphosphate with various chain lengths. These enzymes possess seven conserved regions including two DDXXD motifs that are binding sites for the substrates in association with Mg 2ϩ (13,14). Short-chain polyprenyl-diphosphate synthases (C 15 and C 20 ), such as FPP and GGPP synthases, have been identified in organisms ranging from bacteria to mammals (15), and the mechanisms that determine the chain length have been reported (16). Site-directed mutagenesis of farnesyl-diphosphate synthases from Bacillus stearothermophilus (16) and certain avian species (17) and geranylgeranyl-diphosphate synthase from S. acidocaldarius (16) was used to determine which amino acids are important for the determination of chain length of short-chain prenyl diphosphates. These amino acids were those at the fourth and fifth positions before an aspartate-rich motif in region II or one amino acid at the fifth position before this motif and two amino acids in region II. Recently, we reported that substitution of glycine for alanine before the first DDXXD motif in decaprenyldiphosphate synthase allowed the enzyme to synthesize products with longer chain lengths (18). Thus, the fifth amino acid before region II of long-chain polyprenyl-diphosphate synthases plays an important role in the mechanism of chain length determination (18).
Generally, polyprenyl-diphosphate synthases are known to function as a dimer. The medium-chain polyprenyl-diphosphate synthases (C 30 and C 35 ) from Micrococcus luteus BP26, B. stearothermophilus, and Bacillus subtilis are composed of heterodimers (19,20). GGPP synthase purified from bovine brain forms a homo-oligomer (150 -195 kDa) (21). However, the subunit structure of long-chain polyprenyl-diphosphate synthases remains to be determined. Recently, geranyl-diphosphate synthase isolated from spearmint (22) was found to form a heterodimer. One subunit has similarity with known prenyltransferases, and the other has similarity with the Arabidopsis GGR protein (23), but the aspartate-rich motifs are not conserved.
In this study, we describe the mutational analysis of octaprenyl-diphosphate synthase (IspB) from E. coli. From the analysis, we found that IspB forms a homodimer that is important for the determination of isoprenoid chain length.
Strains and Plasmids-E. coli strains DH10B and JM109 were used in the general construction of plasmids (24). KO229 (ispB::cat) (11), which is the ispB-defective mutant of E. coli harboring pKA3 (ispB), was used as a host strain to express IspB mutants and for ubiquinone extraction. YKK6 (COQ1::URA3) (8), which is the COQ1-defective mutant of S. cerevisiae, was used for complementation analysis and ubiquinone extraction. The plasmids pBluescript KS(Ϫ)/SK(ϩ) and YEp13M4 were used as vectors (24,25). The strains and plasmids used in this study are listed in Table I.
Construction of IspB Mutants by Site-directed Mutagenesis-Sitedirected mutagenesis by PCR was performed following the method of Ito et al. (26). Four oligonucleotide primers (MUT, R1 (for each mutational primer), T7, and T3) (see Table II) were used in amplifications. pKO56, which contains the open reading frame and downstream region of ispB, was used as template in PCRs. First, PCR was performed with the MUT and T3 primers in one reaction and with the R1 and T7 primers in another. An aliquot of each of the reaction mixtures was mixed in a new tube to form the heteroduplex ispB template, and full-length mutant ispB was amplified with the T7 and T3 primers by PCR. This DNA fragment was digested with EcoRI and HindIII and cloned into pBluescript KS(Ϫ). This construct was transformed into E. coli DH10B and KO229 (ispB::cat) for analysis of enzyme activity and ubiquinone production.
Complementation by Mutant ispB in an E. coli ispB Disruption Mutant and S. cerevisiae COQ1-defective Mutant-E. coli KO229 (ispB::cat) harboring pKA3 (11) was transformed with the plasmid containing mutant ispB, which produced transformants that were resistant to spectinomycin and ampicillin. The transformants were subcultured five times in LB medium containing 50 g/ml ampicillin and plated on LB agar medium containing ampicillin. The resulting colonies were then replicated on LB medium containing ampicillin or spectinomycin. Spectinomycin-sensitive and ampicillin-resistant strains, which had the mutant ispB plasmid, but not pKA3, were selected and used for ubiquinone analysis.
To express the mutant ispB genes in S. cerevisiae YKK6 (COQ1::URA3), a COQ1-ispB fusion gene was constructed. The S2 and A3 primers were used to amplify the ispB gene by PCR. The fragment was digested with EcoRI and HindIII and cloned into pSA1 (8), which has 53 amino acids of the Coq1 mitochondrial import signal with the COQ1 promoter. The BamHI-HindIII fragment also was cloned into the yeast shuttle vector YEp13M4 (25). YKK6 was transformed with both plasmids by the lithium acetate method (27) and was selected on Synthetic Complete (0.67% (w/v) yeast nitrogen base, 2% (w/v) glucose or 3% (w/v) glycerol, and the appropriate amino acids)ϪLeuϪUra medium.
Purification of IspB-To overexpress and purify IspB, vectors containing the 6-His or glutathione S-transferase (GST) tag fused to IspB were constructed. The amplified BamHI-HindIII fragment containing ispB from pKO56 was cloned into pQE31 (QIAGEN Inc.) to yield pQKO56. The amplified BamHI-XhoI fragment containing wild-type or A79Y mutant ispB was cloned into pGEX-1X in which an XhoI linker had been inserted to yield pGKO56 or pG79Y, respectively. The plasmids were transformed into E. coli JM109. Transformants were grown to stationary phase in LB medium containing 50 g/ml ampicillin, and 10 ml of culture was inoculated into 100 ml of the same medium. The culture was grown at 37°C for 3 h, and recombinant protein expression was induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside. The cells were collected by centrifugation at 2500 ϫ g for 10 min. To purify His-tagged IspB, cells were suspended in 50 mM sodium phosphate, 300 mM NaCl, and 10 mM imidazole and sonicated 10 times for 10 s at 10-s intervals with an ultrasonic disintegrator in an ice bath. Ruptured cells were centrifuged at 15,000 ϫ g for 20 min. The resulting supernatants were added to a Ni 2ϩ -nitrilotriacetic acid (NTA) slurry and mixed gently at 4°C for 60 min. This mixture was loaded onto a column and washed with 50 mM sodium phosphate, 300 mM NaCl, and 20 mM imidazole. The His-IspB protein was eluted with 50 mM sodium phosphate, 300 mM NaCl, and 250 mM imidazole. To purify the GST-IspB protein, cells were suspended in 50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol (sonication buffer). Cells were ruptured by sonication, and the lysate was mixed with glutathione-Sepharose 4B (Amersham Pharmacia Biotech) at 4°C for 60 min. This mixture was washed twice with 140 mM NaCl, 2.7 mM KCl, 10 mM sodium phosphate, and 1.8 mM potassium phosphate and then with sonication buffer. The GST-IspB protein was eluted with sonication buffer containing 10 mM reduced glutathione.
Coexpression of Wild-type and Mutant IspB Proteins-The EcoRI-HindIII fragment from pQKO56 was recloned into pSTVK28, which had been converted from expressing chloramphenicol resistance to kanamycin resistance, to yield pSTVKQKO56. KO229 cells harboring pBRA, which expresses IspB containing the mutation R321A, were transformed with pSTVKQKO56 and produced transformants that were resistant to ampicillin and kanamycin. Ampicillin-sensitive and kanamycin-resistant strains were selected following the method described above and transformed with pG79Y, and the strains harboring both plasmids were selected and named KO229/35-2.
Ubiquinone Extraction and Measurement-Ubiquinone extraction was performed by the method described previously (8,28). The crude extract of ubiquinone was analyzed by normal-phase TLC with authentic standard Q-10. Normal-phase TLC was carried out on a Kieselgel 60 F 254 plate with benzene/acetone (97:3, v/v). The band containing ubiquinone was collected from the TLC plate following UV visualization and extracted with chloroform/methanol (1:1, v/v). Samples were dried and redissolved in ethanol. The purified ubiquinone was further analyzed by HPLC with ethanol as the solvent.
Prenyl-diphosphate Synthase Assay of Mutant IspB and Product Analysis-Prenyl-diphosphate synthase activity was measured by the method described previously (8), in which incorporation of [1-14 C]IPP into reaction products is detected. E. coli DH10B or KO229 cells, harboring plasmids containing mutant ispB, were incubated to a late log phase in LB medium containing appropriate antibiotics at 37°C. Cells were harvested by centrifugation; suspended in buffer A (100 mM potassium phosphate (pH 7.4), 5 mM EDTA, and 1 mM 2-mercaptoethanol); and ruptured by six sonication treatments, each lasting 30 s with 30-s intervals, in an ice bath. After centrifugation of the homogenate, the supernatant was used as a crude enzyme extract. The assay reaction mixture contained 1.0 mM MgCl 2 , 0.1% (w/v) Triton X-100, 50 mM potassium phosphate buffer (pH 7.5), 10 M [1-14 C]IPP (specific activity of 0.92 TBq/mol), 5 M FPP, and 200 g of crude extract containing the enzyme in a final volume of 0.4 ml. Sample mixtures were incubated for 60 min at 30°C. Reaction products such as prenyl diphosphates were extracted with 1-butanol-saturated water and hydrolyzed with acid phosphatase (29). The products of hydrolysis were extracted with hex-ane and analyzed by reversed-phase TLC with acetone/water (19:1, v/v). Radioactivity on the plate was detected with a BAS1500-Mac imaging analyzer (Fuji Film Co.). The plate was exposed to iodine vapor to detect the spots of the marker prenols.

RESULTS
Site-directed Mutagenesis of IspB-It is known that the side chain length of ubiquinone is determined by the corresponding polyprenyl-diphosphate synthase (8), but it is not clear how polyprenyl-diphosphate synthase determines this length. To understand the nature of polyprenyl-diphosphate synthase, we analyzed the activity of octaprenyl-diphosphate synthase (IspB), which produces the side chain of Q-8 in E. coli. For this purpose, we constructed 15 IspB mutants by site-directed mutagenesis as shown in Fig. 1 and Table II (primers that were used). Site-directed mutagenesis was performed following the method of Ito et al. (26), and the substitutions in all mutant ispB genes were confirmed by sequence analysis.
Complementation of E. coli ispB-and S. cerevisiae COQ1defective Mutants with Mutant ispB Genes-E. coli DH10B was transformed with the plasmids containing mutant ispB genes, and the transformants were used in ubiquinone analysis. Because DH10B has the wild-type ispB gene in the form of genomic DNA, the main product is expected to be Q-8. Although most DH10B cells harboring the mutant ispB gene produced Q-8, a number of mutants produced Q-8 with small amounts of Q-6 and Q-7 (data not shown); and interestingly, DH10B harboring the A79Y mutant produced mainly Q-6 (see Fig. 6A). To detect the actual ubiquinone species produced by the product of the mutant ispB gene, E. coli KO229 (ispB::cat)/pKA3 (ispB), which is defective for the genomic ispB gene, but retains ispB in a plasmid, was transformed with the plasmids containing the mutant ispB genes. KO229, which harbors the mutant ispB genes and has lost the wild-type ispB plasmid (pKA3), was selected as described under "Experimental Procedures." L31V, I32V, Y38A, Y37A/Y38A, Y38A/R321V, Y61V, F75A, K235L, R321A, R321D, and R321V mutant KO229 strains were obtained; however, Y37A, A79Y, K170G, and K170A mutant KO229 strains could not be isolated. Since ispB is essential for growth of E. coli (11), the inability to replace wild-type ispB with these mutants suggested that the Y37A, A79Y, K170G, and K170A mutants do not retain functional activity. The mutants that could complement the loss of the wild-type gene were further analyzed by ubiquinone extraction and analysis (Fig. 2). In the Y38A mutant, Q-7 was mainly produced, with lesser amounts of Q-6 and Q-8 (Fig. 2D). In the Y37A/Y38A mutant, Q-7 and Q-6 were mainly produced, with a little Q-8 (Fig. 2E). In the Y38A/R321V and R321V mutants, Q-6 was mainly produced, with a small amount of Q-5 and Q-7 (Fig. 2, F and L, respectively); however, hardly any Q-8 was produced. In the K235L and R321A mutants, Q-8 was mainly produced; however, a minor product (Q-7) was produced at a level that was greater than that with wild-type IspB (Fig. 2, I  and J, respectively). These results indicate that Tyr-38, Lys-  Although ispB in E. coli is essential for growth, the chromosomal COQ1 gene (homolog of ispB) in S. cerevisiae can be deleted to produce a respiration-deficient phenotype. We took advantage of this COQ1 mutant phenotype for analysis of the function of ispB (8). To express ispB mutants and to analyze their ubiquinone production in YKK6 (COQ1::URA3), mutant ispB genes fused with 53 amino acids of the Coq1 mitochondrial import signal were constructed. YKK6 was transformed with various mutant ispB fusion plasmids, and transformants were replicated on Synthetic CompleteϪLeuϪUra plates containing glucose (Fig. 3A) or glycerol (Fig. 3B) as a non-fermentable carbon source. Although most of the strains grew on the glycerol plate, the Y37A, A79Y, K170G, and K170A mutant YKK6 strains did not grow, indicating that these mutants do not retain functional activity. These results are consistent with the complementation analysis of mutants in E. coli KO229/pKA3 (Fig. 2).
We next analyzed the ubiquinone species produced by YKK6 harboring mutant ispB plasmids (Fig. 4). In the Y38A mutant, Q-8 was mainly produced, along with a significant amount of Q-7 (Fig. 4C). In the Y37A/Y38A mutant, Q-8 and Q-7 were mainly produced, with a lesser amount of Q-6 and Q-5 (Fig.  4D). In the R321V mutant, three ubiquinone species (Q-7, Q-6, and Q-5) were mainly produced, with a small amount of Q-8 (Fig. 4I). In the Y38A/R321V mutant, Q-5 was the main product, with lesser amounts of Q-6 to Q-8 (Fig. 4E). These results again suggest that Tyr-38 and Arg-321 are associated with chain length determination. Most species of ubiquinone produced by mutant ispB in E. coli were similar to ones produced in S. cerevisiae, but some ubiquinone species were shorter in length in S. cerevisiae compared with the ones produced in E. coli.
Arg-321 Is Important for Thermostability of IspB-We tested all ispB mutants for temperature sensitivity and found two temperature-sensitive mutants. The KO229 strain harboring the Arg-321 mutant (R321A or R321D) grew on LB medium containing chloramphenicol and ampicillin at 30°C (Fig. 5A,  a), but stopped growing or grew only slowly at 43°C (Fig. 5A,  b), whereas KO229/pKA3 grew normally at 43°C. The growth curves of the mutants are shown in Fig. 5B. The R321A, R321D, and R321V mutant KO229 strains grew normally at 30°C; but the R321A and R321D mutant KO229 strains did not grow at all at 43°C, and the R321V mutant KO229 strain grew slowly and reached a plateau phase faster than KO229/pKA3 at 43°C. These results indicate that Arg-321 is important for thermostability of IspB, as well as reconfirm that ispB is essential for the growth of E. coli. Overexpression, Purification, and Characterization of IspB-As mentioned above, we constructed the 15 IspB mutants by site-directed mutagenesis and expressed them in E. coli and S. cerevisiae. Among these mutants, the A79Y mutant had an interesting property in that DH10B harboring the A79Y mutant produced Q-6 ( Fig. 6A), whereas the A79Y mutant in S. cerevisiae YKK6 (COQ1::URA3) did not retain functional activity (Fig. 3). To further analyze the A79Y mutant, we purified the wild-type and A79Y mutant IspB proteins. JM109 harboring pQKO56 produced His-IspB fusion proteins with a 6-His tag at the amino terminus of wild-type IspB. JM109 harboring pGKO56 and pG79Y produced GST-IspB and GST-IspB(A79Y) fusion proteins, respectively, with GST at the amino terminus. His-IspB, GST-IspB, and GST-IspB(A79Y) were purified, and prenyltransferase activity was examined as described under "Experimental Procedures." Purified His-IspB and GST-IspB retained functional activity of octaprenyl-diphosphate synthase, but GST-IspB(A79Y) had no detectable activity (data not shown). This result suggests that the substitution of tyrosine for alanine at position 79 abolishes the enzyme activity of IspB, which is consistent with the results of A79Y mutant activity in YKK6 (Fig. 3).
Wild-type IspB and IspB(A79Y) Form a Dimer Structure-We next tested, by Western blot analysis, whether wildtype IspB and IspB(A79Y) form dimers. Purified His-IspB and His-DdsA (as a negative control) were incubated with GST-IspB(A79Y) in buffer A and then subjected to Ni 2ϩ -NTA (Fig. 6C, lanes 2) or glutathione-Sepharose 4B (lane 5) column chromatography. The His-IspB protein was detected together with GST-IspB(A79Y) from purified products on the glutathione-Sepharose 4B column (Fig. 6C, lanes 5), but the His-DdsA protein was not detected in the GST-purified products (Fig. 6C,  panel a, lane 6). Conversely, GST-IspB(A79Y) was detected with His-IspB from purified proteins separated on the Ni 2ϩ -NTA column (Fig. 6C, panel b, lane 2). Homodimerization of wild-type IspB and the A79Y mutant itself was also observed. When purified His-IspB was incubated with GST-IspB(A79Y) in buffer A with disuccinimidyl suberate as a protein crosslinker and subjected to electrophoresis, the band corresponding to a heterodimer with a molecular mass of 97 kDa was detected (data not shown). These results indicate that wild-type IspB and IspB(A79Y) can form a heterodimer, but these proteins cannot form dimers with the similarly structured enzyme DdsA (18). We analyzed the chain lengths of products of heterodimeric IspB consisting of wild-type and A79Y subunits in vitro. Only octaprenyl diphosphate was detected in this assay. We believe that because the homodimer of His-IspB comprises the majority of active enzyme in vitro, we could not detect the activity of the heterodimer composed of IspB and IspB(A79Y). We then made a strain expressing both His-IspB and GST-IspB(A79Y). We constructed the pSTVKQKO56 plasmid, which expresses His-IspB, and co-transformed it with the pG79Y plasmid, which expresses GST-IspB(A79Y); and the resulting strain was named KO229/35-2. This strain produced Q-6 and Q-8 (Fig. 6B). Crude proteins were extracted from KO229/35-2  A and B, shown are the results from the analysis of ubiquinones produced in DH10B harboring the A79Y mutant ispB gene and in KO229 harboring pSTVKQKO56 and pG79Y, respectively, by HPLC. C, crude proteins were extracted from KO229 harboring pSTVKQKO56 and pG79Y (KO229/35-2 strain) and incubated in buffer A containing 1.0 mM MgCl 2 and 0.1% (w/v) Triton X-100 at 30°C for 1 h and then purified on a Ni 2ϩ -NTA column (lanes 1) or a glutathione-Sepharose 4B column (lanes 4). Purified His-IspB and GST-IspB(A79Y) proteins were used in incubations as described above and purified on a Ni 2ϩ -NTA column (lanes 2) or a glutathione-Sepharose 4B column (lanes 5). Purified His-DdsA and GST-IspB(A79Y) proteins were used for the same experiment and purified on a glutathione-Sepharose 4B column (lanes 6). and purified on a Ni 2ϩ -NTA column (Fig. 6C, lanes 1) or a glutathione-Sepharose 4B column (lanes 4). The His-IspB protein was detected in the GST-tagged purified proteins (Fig. 6C,  panel a, lane 4); and similarly, the GST-IspB(A79Y) protein was detected in the His-tagged purified proteins (panel b, lane 1). Thus, it was confirmed that a heterodimer of wild-type IspB and IspB(A79Y) formed in KO229/35-2. When enzyme activity analysis was performed again using crude extract from KO229/35-2, isoprenoid products shorter than octaprenyl diphosphate were detected (Fig. 7, lane 3), although in KO229 harboring only pSTVKQKO56, octaprenyl diphosphate alone was produced (lane 2). These results indicate that the IspB(A79Y) mutant can function as a component of heterodimeric IspB to produce shorter isoprenoid chains compared with the isoprenoid product of homodimeric IspB, although IspB(A79Y) itself does not retain enzyme activity. DISCUSSION In E. coli, the side chain length of Q-8 is determined by octaprenyl-diphosphate synthase (IspB) (8). To discover how IspB determines the chain length, we constructed 15 IspB mutants and expressed them in the E. coli ispB-defective mutant KO229 and in the S. cerevisiae COQ1-defective mutant YKK6 and then analyzed the ubiquinone species produced. In KO229 or YKK6 expressing the IspB mutants L31V, I32V, Y61V, and F75A, the resulting ubiquinone species were almost the same as those produced by the wild-type enzyme (Figs. 2 and 4). However, in KO229 cells expressing the IspB mutant Y38A, Q-7 was the major product; and in KO229 cells expressing the Y37A/Y38A double mutant, the levels of Q-6 were increased. Although the Tyr-37 mutant of IspB did not have enzyme activity, an additional mutation at Tyr-38 restored activity. We speculate that the abnormal protein structure in the Tyr-37 mutant is compensated by an additional Tyr-38 mutation. In E. coli KO229/pKA3 cells, complementation of the lost ispB gene that resided on a plasmid with the Y37A, A79Y, K170G, and K170A mutant ispB genes was unsuccessful. Furthermore, S. cerevisiae YKK6 transformed with the same genes could not grow on Synthetic CompleteϪLeuϪUra containing glycerol (Fig. 3). These results indicate that Tyr-37, Ala-79, and Lys-170 are important residues for activity. The pattern of ubiquinone species synthesized in E. coli or S. cerevisiae harboring the same mutant IspB was some different. This difference might be due to the difference in the intracellular conditions of two organisms because it was reported that metal ions and substrate concentration affected the produced chain length catalyzed by prenyl-diphosphate synthases (30,31). We intentionally mutated aromatic residues located before the first aspartate-rich motif because tyrosine or phenylalanine is important for chain length determination in FPP and GGPP synthases (16). From our results, we concluded that Tyr-61 and Phe-75 are not important for catalytic activity, whereas Tyr-37 and Tyr-38 play an important role in catalytic activity and chain length determination.
Lys-170 is conserved in all known polyprenyl-diphosphate synthases from various organisms. Mutation of the corresponding lysine to glutamic acid in farnesyl-diphosphate synthase from S. cerevisiae produced an enzyme that synthesized geranyl diphosphate rather than FPP because the glutamic acid substitution decreased substrate affinity (32). The substitution of this corresponding lysine in IspB with alanine or glycine strongly decreased enzyme activity.
KO229 harboring the Arg-321 mutant (R321A or R321D) grew at 30°C, but did not grow at 43°C (Fig. 5). The arginine located at the third position from the carboxyl terminus is conserved in all long-chain polyprenyl-diphosphate synthases (Fig. 1) and FPP synthases (14). However, in GGPP synthases, there is very little conservation in the C-terminal region; for example, the corresponding amino acid in mouse and human GGPP synthases is glutamate (15). Amino acid alignment of GGPP synthases showed that the length of the carboxyl terminus region was different between FPP synthases and longchain polyprenyl-diphosphate synthases. It has been reported that an FPP synthase mutant from B. stearothermophilus in which the arginine was replaced by valine had catalytic activity similar to that of the wild-type enzyme, indicating that the amino acids in the carboxyl terminus are not essential for catalytic function in FPP synthase (33). These results suggest that the role of the carboxyl terminus in FPP and GGPP synthases is different from that in long-chain polyprenyl-diphos-  phate synthases. In the IspB(R321V) mutant, the ubiquinone product had a shorter isoprenoid side chain, and KO229 cells carrying the ispB gene with this mutation did not grow at 43°C. These results suggest that Arg-321 is important for thermostability of IspB. This temperature-sensitive Arg-321 mutant will be very useful for future study to determine why ispB is an essential gene in E. coli.
Ala-79 of IspB is the fifth amino acid before the first aspartate-rich motif. This position is important for chain length determination by FPP synthases (16) and decaprenyl-diphosphate synthase (18). In FPP synthases, the corresponding residue is an aromatic amino acid, either tyrosine or phenylalanine. When this amino acid residue is substituted with glycine, an isoprenoid longer than the wild-type product is produced (16). Similarly, substitution of the corresponding amino acid (G70Y) in DdsA resulted in production of a small amount of undecaprenyl diphosphate that was longer than the original decaprenyl diphosphate (18); however, the A70Y mutant of DdsA had no activity. This is consistent with our observation that the IspB(A79Y) mutant did not have enzyme activity. Interestingly, wild-type E. coli DH10B harboring the A79Y mutant produced Q-6, although the A79Y mutant itself had no activity (Fig. 6A). We have shown that the nonfunctional IspB(A79Y) protein forms a dimer with wild-type IspB produced by the genomic ispB gene and synthesizes a ubiquinone product with an altered chain length. We have also shown that His-IspB and GST-IspB(A79Y) form a heterodimer and produce shorter isoprenoids than octaprenyl diphosphate. As shown in a model (Fig. 8), the homodimer of the A79Y mutant did not have catalytic activity, but the heterodimer of wild-type IspB and the A79Y mutant did have activity and produced a shorter isoprenoid than a homodimer of wild-type IspB. This result means in a sense that in the creation of IspB-IspB(A79Y), we have produced a hexaprenyl-diphosphate synthase. Furthermore, this IspB-IspB(A79Y) form can be thought of as an evolutional intermediate toward a heterodimer of the mediumchain polyprenyl-diphosphate synthases, such as hexaprenyland heptaprenyl-diphosphate synthases from M. luteus BP26, B. stearothermophilus, and B. subtilis (19,20). These heterodimers are composed of components I and II; component I does not resemble other proteins, whereas component II has conserved regions, similar to known polyprenyl-diphosphate synthases. Component I serves the cavity for a catalitic space of substrate and has an important role in chain length determination (34). IspB(A79Y) probably plays a role similar to that of component I in a IspB-IspB(A79Y) heterodimer. In support of this idea, it has been reported that the subunits of FPP synthase interact with each other to form a shared active site in the homodimer structure (35).
The formation of homodimers of long chain-producing prenyl-diphosphate synthase was demonstrated, for the first time, in this study. In addition, homodimerization of long chainproducing prenyl-diphosphate synthase was found to be important for enzyme activity and chain length determination. Longchain polyprenyl-diphosphate synthase (C 40 , C 45 , and C 50 ) genes have been isolated from E. coli (36), Synechocystis sp. strain PCC6803 (11), R. capsulatus (37), G. suboxydans (18) and S. pombe (2). Because these synthases, except for the one from S. pombe, could be expressed in E. coli and S. cerevisiae, we believe that they form homodimeric enzymes like IspB. However, because the enzyme from the eukaryote was not able to express a functional product in the other species, eukaryotic long-chain polyprenyl-diphosphate synthase might be regulated differently. Therefore, the long-chain polyprenyl-diphosphate synthase of eukaryotic origin will be an interesting subject for further analysis.