COQ2 Is a Candidate for the Structural Gene Encoding puru-Hydroxybenzoate:Polyprenyltransferase*

Coenzyme Q functions as a lipid-soluble electron car- rier in eukaryotes. In Saccharomyces cerevisiae, the enzymes responsible for the assembly of the polyiso-prenoid side chain and subsequent transfer to para- hydroxybenzoate (PHB) are encoded by the nuclear genes COQl and COQ2, respectively. Yeast mutants defective in coenzyme Q biosynthesis are respiratory defective and provide a useful tool to study this non-sterol branch of the isoprenoid biosynthetic pathway. We isolated a 5.5-kilobase genomic DNA fragment that was able to functionally complement a coq2 strain. Additional complementation analyses located the COQ2 gene within a 2.1-kilobase HindIII-BgZII restriction fragment. Sequence analyses revealed the presence of a 1,116-base pair open reading frame coding for a predicted protein of 372 amino acids and a molecular mass of 41,001 daltons. The amino acid sequence exhibits a typical amino-terminal mitochon- drial leader sequence and six potential membrane- spanning domains. Primer extension and Northern analyses indicate the gene is transcriptionally active. Transformation of a coq2 strain with the 2.1-kilobase HindIII-BgZII genomic restriction fragment on a mul-ticopy plasmid restores PHB:polyprenyltransferase activity to wild-type

Coenzyme Q functions as a lipid-soluble electron carrier in eukaryotes. In Saccharomyces cerevisiae, the enzymes responsible for the assembly of the polyisoprenoid side chain and subsequent transfer to parahydroxybenzoate (PHB) are encoded by the nuclear genes COQl and COQ2, respectively. Yeast mutants defective in coenzyme Q biosynthesis are respiratory defective and provide a useful tool to study this nonsterol branch of the isoprenoid biosynthetic pathway. We isolated a 5.5-kilobase genomic DNA fragment that was able to functionally complement a coq2 strain. Additional complementation analyses located the COQ2 gene within a 2.1-kilobase HindIII-BgZII restriction fragment. Sequence analyses revealed the presence of a 1,116-base pair open reading frame coding for a predicted protein of 372 amino acids and a molecular mass of 41,001 daltons. The amino acid sequence exhibits a typical amino-terminal mitochondrial leader sequence and six potential membranespanning domains. Primer extension and Northern analyses indicate the gene is transcriptionally active. Transformation of a coq2 strain with the 2.1-kilobase HindIII-BgZII genomic restriction fragment on a multicopy plasmid restores PHB:polyprenyltransferase activity to wild-type levels. Disruption of the chromosomal COQ2 gene indicates the gene is not essential for viability, yet is required for PHB:polyprenyltransferase activity and respiratory function. In addition, the deduced amino acid sequence of PHB:polyprenyltransferase contains a putative allylic polyprenyl diphosphate-binding site. The presence of this aspartaterich domain in a number of functionally distinct proteins which utilize polyprenyl diphosphate substrates is reported.
Coenzyme Q (ubiquinone) functions as a lipid-soluble electron transporter between NADH dehydrogenase (Complex I), succinate dehydrogenase (Complex 11), and the bcl complex * This work was supported in part by Training Grant HL 07386 (to M. N. A.) and Grants HL 30568 and HL 22174 from the United States Public Health Service, and by the Laubisch Fund. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence (s)  11 To whom correspondence should be addressed.
(Complex 111) on the inner mitochondrial membrane (Green, 1966;Kroger and Klingenberg, 1973). In addition to the mitochondrial inner membrane, coenzyme Q has been found in the nucleus, plasma membrane, Golgi vesicles, and lysosomes (reviewed by Ramasarmsa, 1985). The function of coenzyme Q outside the mitochondria is uncertain; however, involvement in antioxidation (Stocker et al., 1990) and in transplasma membrane electron transport (Sun et al., 1990) has been suggested. The lipid solubility of coenzyme Q is the result of an all-trans polyprenyl side chain derived from isopentenyl diphosphate (Momose and Rudney, 1972).
The majority of coenzyme Q biosynthesis takes place in the mitochondria (reviewed in Olson and Rudney, 1983); however, coenzyme Q synthesis in rat liver endoplasmic reticulum-Golgi subcellular fractions has been observed (Kalen et al., 1987(Kalen et al., , 1990. The biosynthesis of sterols, dolichol, and coenzyme Q share a common synthetic pathway up to the formation of IPP.' The first steps unique to coenzyme Q biosynthesis involve the synthesis from IPP of an all-trans polyprenyl diphosphate where the chain length is a species-specific phenomenon (Ramasarma, 1985). Subsequently, the polyprenyl group is transferred to PHB to form 3-polyprenyl-4hydroxybenzoate. Winrow and Rudney (1969) first described the synthesis of 3-polyprenyl-4-hydroxybenzoate from PHB and preformed polyprenyl diphosphates in a cell-free rat tissue preparation. The authors designated the enzyme that catalyzes this reaction PHB:polyprenyltransferase. Mitochondria from rat liver (Momose and Rudney, 1972) and yeast (Casey and ThrelfaK, 1978) contain all the enzymes necessary for the synthesis of polyprenyl hydroxybenzoate from IPP and PHB. These include IPP isomerase, farnesyl-diphosphate synthetase, polyprenyl-diphosphate synthetase, and PHB: polyprenyltransferase.
The availability of yeast respiratory mutants has greatly facilitated the study of coenzyme Q biosynthesis and function (Tzagoloff et al., 1975;De Kok and Slater, 1975;reviewed in Tzagoloff and Dieckmann, 1990). The COQ3 gene encoding the coenzyme Q biosynthetic enzyme 3,4-dihydroxy-5-hexaprenylbenzoate methyltransferase has recently been isolated and characterized (Clarke et al., 1991). Identification of the defect in two yeast coenzyme Q mutants designated coql-1 and coq2-1 has been described in addition to the DNA sequence of COQl . Cells possessing coql-1 or coq2-1 mutant alleles are defective in hexaprenyldiphosphate synthetase or PHB:polyprenyltransferase activities, respectively. In this report we describe the molecular 16385) acquired from Alltech. Restriction enzymes and other enzymes employed in subcloning were purchased from Bethesda Research Laboratories. Yeast Strains and Growth Media-The strains of Saccharomyces cereuisiae used in this study are described in Table I. The coq2 mutants E3-71 and C33 are part of a large number of pet mutants derived from the respiratory competent haploid yeast strain D273-10B and D273-10B/Al, respectively, by mutagenesis with ethylmethane sulfonate (Tzagoloff et al., 1975). The following media were used for growth of yeast: YPD (2% glucose, 2% peptone, 1% yeast extract); YEPG (3% glycerol, 2% ethanol, 2% peptone, 1% yeast extract); YPGal (2% galactose, 2% peptone, 1% yeast extract); WO (2% glucose, 0.67% yeast nitrogen base without amino acids from Difco). Where required media were supplemented with auxotrophic requirements at 20 pg/ml. Solid media contained 2% agar.
Cloning of the COQ2 Gene-The COQ2 gene was cloned by transformation of aE3-71/Ul (a ura3-1 coq2-I) with a genomic library consisting of partial Sau3A fragments (5-15 kb) of yeast nuclear DNA ligated to the BamHI site of YEp24 (Botstein and Davis, 1982). This library was kindly provided by Dr. Marian Carlson, Department of Human Genetics, Columbia University. Approximately 1 X 10' cells were transformed with 10 pg of DNA by the procedure of Beggs (1978). Transformants were selected on minimal glycerol medium for complementation of both uracil auxotrophy and respiratory deficiency. The transformed phenotype of one such clone was verified to be due to an autonomously replicating plasmid, designated pGlO/Tl. The pGlO/Tl plasmid was subsequently used to subclone and characterize the COQ2 gene.
Preparation of Mitochondrial Membrane-rich Fractwns-Mitochondrial membrane fractions were prepared as described (Casey and Threlfall, 1978) except the buffer was replaced with buffer A (0.27 M sorbitol, 0.33 M mannitol, and 13.3 mM Tris-C1, pH 7.4). Yeast cells were grown to late log phase in YPD media and harvested by centrifugation at 1,500 X g for 10 min. Cells were washed once with water, sedimented as above, and washed again in buffer A. The washed pellets were resuspended in buffer A and ruptured by a single passage through a French pressure cell at 10,000 p.s.i. The homogenate was centrifuged at 1,500 X g for 10 min. The resulting supernatant was centrifuged at 10,000 X g for 15 min to sediment mitochondrial membranes. The pellet was resuspended in buffer A and centrifuged again at 10,000 X g for 15 min. This regimen of sedimenting the membranes and resuspending in buffer A was repeated three times. Finally, the mitochondrial membrane-rich pellet was resuspended in buffer A. The protein concentration was determined with a Bradford protein assay reagent (Bio-Rad).
Enzyme Assays-Hexaprenyl-diphosphate synthetase and PHB: polyprenyltransferase activities were measured as described (Casey and Threlfall, 1978;. The assay measures the formation of prenylated and thus, chloroform-soluble, hydroxybenzoate derivatives. Briefly, 1.0 mg of mitochondrial membrane-rich protein was incubated in the presence of 4 ptd [U-14C]PHB (50 Ci/ mol), 100 mM NaH2P04, pH 7.5, 150 p~ tri-lithium isopentenyl diphosphate, 120 p~ geranyl diphosphate, and 33 mM MgC12 in a total volume of 200 pl. The reaction was carried out at 30 "C for 2 h. Reaction products were solvent extracted overnight at room temperature in an equal volume of water-saturated chloroform. The total amount of prenylated hydroxybenzoate derivatives was determined by liquid scintillation counting an aliquot of the chloroform extract. Alternatively, the individual polyprenyl hydroxybenzoate homologues were separated by adsorbent thin layer chromatography. In this case, the chloroform phase was blown to dryness under a stream of nitrogen. The products were dissolved in 10 pl of chloroform and applied to silica thin layer chromatograms. The plates were developed in acetone/petroleum ether (3:7) and subsequently scanned for radioactivity with a Packard chromatogram Scanner model 7230.
DNA Sequencing-The 5.5-kb genomic HindIII-BglII fragment was sequenced by the method of Maxam and Gilbert (1980) and by the dideoxynucleotide chain termination method (Sanger, 1977). The latter method utilized single-stranded DNA, obtained after subcloning fragments into pUC18 and then sequencing with either universal primers (Bethesda Research Laboratories) or a 17-mer oligonucleotide corresponding to nucleotides 470-486.
Primer Extension-A synthetic oligonucleotide corresponding to the antisense strand (nucleotide 917-883) was gel purified on a 20% polyacrylamide, 8 M urea gel. The probe was then 32P-end-labeled with T, polynucleotide kinase to a specific activity of 9 X 10' cpmlpg DNA. Polyadenylated (A+) RNA was obtained from the strain D273-10B (see Table I). Primer extensions were performed by hybridizing 4 ng of the 32P-labeled oligonucleotide to 8 pg of poly(A+) RNA and incubating with reverse transcriptase as described by Teruya et al. (1990).
Computer Analyses-Nucleotide and deduced amino acid sequences were analyzed with computer programs from the University of Wisconsin Computer Group (Devereux et al., 1984). Hydrophobicity calculations were performed with the program PEPPLOT which is based on the algorithm of Kyte and Doolittle (1982). Predictions of a-helices and 0-sheets were determined as described (Chou and Fasman, 1978). Potential transmembrane regions were found by the method of Eisenberg (1984). This involves calculating the mean hydrophobic moment over a moving 11-amino acid window through regions of the protein sequence which exceed a predetermined average hydrophobicity. The various mean hydrophobic maxima are then plotted against the mean hydrophobicity for the same window to assess transmembrane potential.

Molecular
Cloning of COQB-Nine separate complementation groups of yeast coenzyme Q mutants have been previously isolated that are respiratory deficient (Tzagoloff et al., 1975;Tzagoloff and Dieckmann, 1990). Two of these mutants were shown to be defective in the first two committed steps of coenzyme Q biosynthesis . The two mutants, C296 (coql) and C33 (coq2), are defective in hexaprenyl-diphosphate synthetase and PHB:polyprenyltransferase activities, respectively. Yeast cells that are respiratory defective are unable to utilize the nonfermentable carbon source glycerol. To facilitate studies on the COQ2 product, the gene was cloned by complementing the respiratory deficient phenotype of aE3-71/Ul. Transformation of this mutant with a yeast genomic library yielded several respiratory competent clones that were also complemented for uracil auxotrophy. Vegetative growth of the transformants on nonselective rich glucose medium followed by replica plating onto glycerol or minimal glucose media indicated that the Gly' and Ura+ phenotype cosegregated. This established that respiratory competence and uracil prototrophy were functions of the presence of an autonomously replicating plasmid. Four independent Gly', Ura+ transformants carried a plasmid containing identical genomic inserts. Plasmid DNA (pGlO/Tl) prepared from one of the clones when back-transformed into aE3-71/Ul rescues the respiratory defect of the mutant.
The plasmid pGlO/Tl contains a genomic DNA insert approximately 5.5 kb in length. To locate the COQ2 gene various regions of the 5.5-kb genomic fragment were subcloned into YEp352 (Hill et al., 1986) and were tested for their ability to complement aE3-71/U1 and restore growth on glycerol-containing media (Fig. 1). These experiments indicate the COQ2 gene must be located within a HindIII- 1. Strategy employed to clone and sequence COQ2. The plasmid pGlO/Tl consists of an approximately 5.5-kb yeast genomic DNA inserted into the shuttle vector YEp24. Plasmids were constructed that contain various portions of the genomic DNA as described under "Experimental Procedures" and are designated. The ability of these plasmids to complement the coq2-1 strain C33 and allow growth on glycerol containing media are indicated by + or -.
Abbreviations for restriction sites are as follows: B, BamHI; Sph, SphI; G, BglII; H, HindIII; X , XbaI; P, PstI; S, SalI. E , the 2,195-bp HindIII-BglII fragment was sequenced the entire length in both directions by the method of Maxam and Gilbert (1980, thick arrows) or by the dideoxy chain termination (Sanger et al., 1977, thin arrows). Boxed areas on the sequencing map indicate open reading frames.
The unidentified ORF is depicted by the cross-hatched box and is oriented (5'-3') from right to left. The COQ2 ORF is designated by the open box and is oriented (5'-3') from left to right.
BglII restriction fragment. The entire 2,195-bp HindIII-BglII fragment was sequenced in both directions by the strategy shown in Fig. 1. The nucleotide sequence of this region is shown in Fig. 2. Two open reading frames (ORFs) are present. The first unidentified ORF is oriented from right to left in Fig. 1 and by the double-stranded DNA sequence in Fig. 2. This unidentified ORF begins with an ATG at nucleotide 583 and continues to the HindIII site (nucleotide 1) without encountering a termination codon. The second ORF is oriented from left to right shown schematically in Fig. 1. This ORF begins with an ATG at nucleotide 819 and ends with a termination codon at nucleotide 1,937 (Fig. 2). The predicted translation product of this ORF is a protein of 372 amino acids with a molecular mass of 41,001 daltons.
Construction of a COQ2 Disruption Allele-To confirm that the COQ2 gene had been correctly identified, the chromosomal copy of the gene was disrupted by the one-step gene replacement procedure of Rothstein (1983) in the respiratory competent haploid strain W303-1A. The construction of the disrupted gene is shown in the upper part of Fig. 3. The plasmid ST4 containing the 3-kb BglII insert was digested with XbaI to yield a 2.2-kb fragment containing the entire COQ2 coding region. One of the XbaI sites is inside the BglII fragment and the other XbuI site is in the multiple cloning region of the YEp352 vector. This DNA fragment was transferred to the XbaI site of YEp352X (this plasmid is similar to YEp352 except that the multiple cloning region of the latter plasmid has been replaced with a unique XbaI site). The resultant plasmid (pGlO/ST9) was opened at the unique PstI site and ligated to a 1,125-bp PstI fragment containing the yeast HIS3 gene to yield the disrupted allele, coqZ::HIS3.
A linear 3.3-kb XbaI fragment containing the disrupted gene with 5'-and 3"flanking sequences was isolated from pGlO/ST9 and used to transform W303-1A. Clones were selected on minimal glucose medium supplemented with all the auxotrophic requirements of W303-1A except histidine.
Most of the histidine-independent transformants were respiratory defective and were complemented by a p-tester but not by C33, an independent coq2 mutant. The results of these crosses imply a genetic linkage of the disrupted coq2::HIS3 allele to the coq2 mutation. As verification of this allelism, linkage analyses were performed on the diploid product of the mating of the cross between C33 and W303VCOQ2 (data not shown). To facilitate sporulation the C33/W303VCOQ2 diploid strain was transformed with the COQ2-containing plasmid pGlO/ST6. Twenty-two independent tetrads were dissected. After propagation on nonselective media (YPD), the segregants were replica plated onto media containing 5-fluoroorotic acid (Boeke et al., 1987) to select for uracil auxotrophy and, hence, loss of pGlO/ST6. None of the resulting haploid strains exhibited respiratory competence, as judged by growth on YEPG, confirming the allelism between the cloned COQ2 gene and an authentic coq2 mutation. Southern hybridization analysis of the genomic DNA from one of the Gly-, His' transformants (W303VCOQ2) confirmed the successful substitution of the wild-type gene by the disrupted allele. In this analysis, wild-type and mutant DNAs were digested with a combination of EcoRI and KpnI and hybridized to a nick-translated 2.2-kb XbaI-BglII fragment containing the entire COQ2 gene and flanking sequences. The probe detects a 3-kb band in the genomic digest of wild-type DNA and two coincident bands of approximately 2.1 kb in the mutant DNA digest (Fig. 3). The sizes of these fragments are consistent with the restriction maps of the wild-type COQ2 and the coq2::HIS3 allele.
Primer Extension and DNA Sequence Analyses-Both the

100
FIG. 2. Nucleotide and deduced amino acid sequence of the C092 gene. The nucleotide sequence shown is of the 2,195-bp HindIII-BglII genomic DNA restriction fragment. The complementary DNA strand corresponding to the unidentified ORF is shown with the presumptive initiator ATG underlined and bold. The amino acid sequence encoded by the COQ2 ORF is shown below the sense DNA strand. Nucleotides are numbered at the right and left margins and the amino acids are numbered below the appropriate residues. Two start-sites of transcription of COQ2 are depicted by asterisks.

T L V C M G V L S L L P A Q C W W L G L A S L P I V F T Y P -
180 190  gene disruption and linkage analysis confirm that the fulllength gene in the BglII-Hind111 fragment corresponds to COQ2. To map the start site of transcription of COQ2, a 35base oligonucleotide primer corresponding to nucleotide 917-883 of the antisense strand was hybridized to poly(A)+ RNA from the strain D273-10B and extended with reverse transcriptase. As shown in Fig. 4, two major products of this reaction are 534 and 234 bases in length. These products indicate that transcription is initiated a t nucleotide 383 and at nucleotide 683. The two start sites are located 436 and 136 nucleotides 5' of the ATG codon, respectively. Another less abundant extension product was observed 98 bases in length (data not shown). This indicates a minor start site is present 17 base pairs 5' of the initiator methionine codon. The COQ2 Gene Restores PHB:Polyprenyltransferase Actiuity to a Coenzyme Q Mutant-The pet mutant C33 has been shown to lack detectable PHB:polyprenyltransferase activity ). Therefore, we tested the ability of the plasmid ST4 to restore PHB:polyprenyltransferase activity. Mitochondrial membrane-enriched fractions were isolated from wild-type, from the coq2 mutant C33/U1, and from the transformant C33/Ul/ST4. Hexaprenyl-diphosphate (HPP) synthetase and PHB:polyprenyltransferase activities were assayed as described by Casey and Threlfall (1978). In this assay the allylic isoprenoid geranyl diphosphate is elongated to hexaprenyl diphosphate through the action of the COQl gene product, hexaprenyl-diphosphate synthetase. Subsequently, PHB:polyprenyltransferase prenylates C4 of PHB with the 30-carbon HPP. At the magnesium concentration used in these assays (33 mM), PHB:polyprenyltransferase is activated. In this case the enzyme will accept other allylic isoprenyl diphosphate substrates such as geranyl diphosphate. Thus, the production of 3-diprenyl-4-hydroxybenzoate requires the activity of only PHB:polyprenyltransferase, while the synthesis of 3-hexaprenyl-4-hydroxybenzoate requires the activity of both HPP synthetase as well as PHB:polyprenyltransferase. The assay measures the prenylation of [14C]PHB to form chloroform-soluble polyprenyl hydroxybenzoate products. Once extracted, the polyprenyl hydroxybenzoate species are separated by thin layer chromatography (see "Experimental Procedures"). Fig. 5A shows that mitochondrial membranes from the wild-type strain D273-10B can synthesize  shown (A, C, G, 7'). The arrows indicate the major primer extension products, and the length of these extensions in bases is indicated.

310 2071 A T A G G C T C T t C C A G A C G T A T G T~~~~G T A A T A T G A~T G T T C A A T A~T A T T T T C T G C~~C T G T T G~G C A G T
both diprenyl-and hexaprenyl-hydroxybenzoate species. In contrast, a similar protein fraction from C33 is unable to form any polyprenyl hydroxybenzoate products reflecting the absence of any detectable PHB:polyprenyltransferase activity (Fig. 5 B ) . Transformation of C33 with the plasmid ST4 results in the restoration of respiratory competence, as judged by growth on YEPG (Fig. 1). Additionally, mitochondrial membranes from the transformed cells have the restored ability to synthesize diprenyl-and hexaprenyl-hydroxybenzoate coenzyme Q intermediates (Fig. 5 C ) . These results indicate the COQ2 gene is able to restore respiratory compe-

FIG. 5. In vitro prenylation of PHB. Mitochondrial membrane-enriched fractions
were isolated and assayed for PHB:polyprenyltransferase activity as described under "Experimental Procedures." 3-Polyprenyl-4-hydroxybenzoate products of the assays were analyzed by thin layer chromatography. Radioactivity was detected by scanning the thin layer chromatography plates with a radiochromatogram scanner. Protein samples from the yeast strain D273-10B ( A ) , C33-Ul (R), or C33-U1/ST4 (C) were incubated under standard assay conditions. D, protein from D273-10B was incubated under standard assay conditions including 1 mM L-tyrosine. The origin and solvent front for each thin layer chromatography (run from left to right) scan are indicated by vertical arrows. The structures of the prenylated hydroxybenzoates are shown above the respective peaks. tence and PHB:polyprenyltransferase activity to the coenzyme Q mutant C33.
The substrate PHB of PHB:polyprenyltransferase is biosynthetically derived from and structurally closely resembles tyrosine. Because of the biological significance of protein prenylation, we examined whether tyrosine might also serve as a substrate for PHB:polyprenyltransferase. To address this question we performed a competition experiment with mitochondrial membranes from D273-10B exactly as in Fig. 5A except that 1 mM L-tyrosine was included in the incubation. As can be seen in Fig. 5D, even a t a 500-fold molar excess tyrosine does not affect the activity of PHB:polyprenyl transferase and, therefore, is not likely a substrate of this enzyme.
COQ2 Amino Acid Sequence Analyses-The translation product of COQ2 is a protein 372 amino acids in length with a molecular mass of 41,001 daltons (Fig. 6A). The protein possesses a typical mitochondrial leader sequence as judged by a preponderance of basic and hydroxylated residues and a predicted tendency to form a-helices (Hart1 et al., 1989). Moreover, the amino terminus contains a three-amino-acid proteolytic recognition sequence found in mitochondrial proteins that are processed in two steps (Hendrick et al., 1989;Fig. 6A). This consensus sequence is defined by an arginine a t position -10 (relative to the mature amino terminus), a hydrophobic residue (Leu, Val, Ile, or Phe) at position -8, and a serine, threonine, or glycine a t position -5. The pres-

A M F I W Q R K S I L L G~S~L G~G R V T V A G I I G S S R K R Y T S S S
1 10 20 30

M T V Y T A S V T A P V N I A T L K Y W G K R D T K L N L P T N S S I S V T 1 10 20 30 L S Q D D L R T L T S A A T A P E F E R D T L W L N G E P H S I D N E R T Q 40 50 60
7 0

N C L R D L R Q L R K E M E S K D A S L P T L S Q W K L H I V S E N N F P T
8 0 90 1 0 0

A A G L A S S A A G F A A L V S A I A K L Y Q L P Q S T S E I S R I A R K G 120 130
140 150 ence of this motif suggests the amino-terminal leader sequence of PHB:polyprenyltransferase protein is proteolysed first between amino acids 14-15 and then again between amino acids 22-23. These sites are represented as vertical arrows in Fig. 6A. The net result of this processing would be the removal of 22 amino acids from the mature protein.

S G S A C R S L F G G Y V A W E M G K A E . D G H D S M A V Q 1 A D S S D W P
Adjacent to the potential mitochondrial leader sequence lies a serine-rich region between amino acids 29-49. The PHB:polyprenyltransferase is generally thought .to be located at or within the inner mitochondrial membrane (Momose and Rudney, 1972). We employed the method of Eisenberg et al. (1984) to determine if the COQ2 gene product contains potential transmembrane sequences. The mean hydrophobic moment of an 11-amino-acid window, when plotted against the mean hydrophobicity can predict the tendency of a particular sequence to become transmembrane or globular, depending upon where the value plots (Eisenberg, 1984). Based on these criteria, the PHB:polyprenyltransferase amino acid sequence may possess as many as six membrane-spanning regions (Fig. 6A). The predicted amino acid sequence of the unidentified open reading frame is shown in Fig. 6B. Entry of this amino acid sequence into several protein databases did not retrieve any known sequences with significant similarities.
A Chou-Fasman analysis (1978) of the amino acid sequence predicted by COQ2 is shown in Fig. 7A. This analysis reveals the tendency of a protein to form a-helices or @-strands and indicates the protein consists, to a large extent, of @-strands. In addition, a hydrophobicity profile (Kyte and Doolittle, 1982) is shown in Fig. 7B. The profile reveals four extended regions of hydrophobicity. The first two hydrophobic regions each encompass two potential membrane spanning domains while the third and fourth hydrophobic regions each contain single potential transmembrane domains. The potential membrane spanning regions are indicated by the solid bar in Fig.  7B.
Previous alignments of the amino acid sequences of rat, human, and yeast FPP synthetase with yeast HPP synthetase amino acid sequence revealed only limited global similarities; however, three highly conserved domains were identified . Two of these domains (I, 11) possess a direct repeat of the consensus sequence (I, L, or V)xDDxxD and were proposed to represent substrate binding sites. Fig. 8 shows an alignment of amino acid sequences from several different proteins (from multiple species, in some cases) that share the common property of employing (po1y)isoprenyl diphosphate substrates. At least nine functionally distinct enzymes have been arranged into six groups (A-F) based upon functional similarities. The consensus sequence in Fig. 8 was derived from amino acids that were present in three or more functionally distinct enzymes.

DISCUSSION
PHB:polyprenyltransferase catalyzes the prenylation of para-hydroxybenzoate with an all-trans polyprenyl group, the second committed step in the coenzyme Q biosynthetic pathway (Winrow and Rudney, 1969). The respiratory deficient strain of S. cereuisiae C33, previously assigned to complementation group G10, is defective in PHB:polyprenyltransferase activity Fig. 5). Mitochondria isolated from C33 as well as other mutants from this complementation group have a normal composition of cytochromes and dehydrogenases but are unable to oxidize NADH unless supplemented with coenzyme Q (Tzagoloff et al., 1975). Taken together, the two observations suggest that the genetic lesion responsible for the respiratory defect of this group of mutants is either in the gene coding for the transferase or in a factor regulating the expression of the gene. This gene has been designated COQ2 to indicate a deficiency in coenzyme Q.
In this report we describe the molecular cloning of a yeast genomic DNA fragment containing the wild-type COQ2 gene by complementation of a coq2 mutant. Sequence analysis of the smallest subclone capable of complementing coq2 mutants and restoring their PHB:polyprenyltransferase activity disclosed the presence of two ORFs. One of the ORFs codes for the amino-terminal region of a protein whose derived primary sequence is not homologous to any entries in the current data bases. The second ORF is full-length and codes for a predicted primary translation product with a molecular weight of 41,001. This reading frame was confirmed by in situ disruption and allelism tests to be the COQ2 gene. Disruption of COQ2 elicits  N.c., Neurospora crassa; S.C., S. cereuisiae; H.s., Homo sapiens; R.r., Ratus ratus; E.c., E . coli; C.P., Cyanophora paradoxa; R.c, Rhodobacter capsulatus; E.h., Erwinia herbicola; L.e., Lycopersicon esculentum; A.t., Agrobacterium tumefaciens; P.s., Pseudomonas sauastanoi. respiratory incompetence and loss of detectable PHB:polyprenyltransferase activity. The mutant construct with the disrupted copy of COQ2 is complemented by a ptester strain but not by coq2 mutants. Furthermore, only respiratory deficient meiotic progeny were produced from a cross of the disruption strain to a coq2 mutant.
Even though definitive proof that COQ2 is the structural gene for PHB:polyprenyltransferase will require biochemical characterization of the purified enzyme, all the available data are most consistent with this conclusion. In addition to the already mentioned dependence of transferase activity on the presence of the wild-type COQ2 allele, the primary sequence itself suggests a mitochondrial membrane protein with a putative polyprenyl diphosphate-binding domain similar to those found in other enzymes known to bind isoprenoid substrates. The predicted amino acid sequence of the COQZ product displays a typical mitochondrial leader sequence with the three-amino-acid consensus found in precursors that are processed by two consecutive proteolytic cleavages (Hendrick et al., 1989). The protein also has a hydrophobic character with six potential membrane-spanning domains suggestive of a membrane localization. Both attributes would be expected of the transferase which has been shown to be associated with the membrane fraction of yeast (Casey and Threlfall, 1978) and rat liver (Momose and Rudney, 1972) mitochondria. It is interesting that recent evidence points to a second localization of PHB:polyprenyltransferase in endoplasmic reticulum. Kalen et al. (1987) reported a greater extent of labeling of microsomal than mitochondrial coenzyme Q when rat liver slices were incubated in the presence [3H]mevalonate or ["C] tyrosine. In another study, fractionation of rat liver tissue indicated substantial transferase activity in the endoplasmic reticulum-Golgi fraction (Kalen et al., 1990). Our own data suggest that a similar situation may prevail in yeast. Transferase assays revealed nearly identical distribution of the enzyme in the microsomal (100,000 X g pellet) and mitochondrial fractions of yeast (data not shown). Moreover, the PHB:polyprenyltransferase mutant C33 does not contain de-tectable transferase activity in either fraction. These data need to be interpreted with caution, however, since the procedure employed to disrupt the cells may fragment mitochondria and produce smaller membrane fragments that could cofractionate with endoplasmic reticulum and plasma membrane vesicles. It is significant, however, that under similar conditions of cell fractionation all of the HPP synthetase activity is recovered in the mitochondrial fraction. Conceivably, in yeast and mammalian cells the pathway associated with the endoplasmic reticulum employs polyprenyl diphosphates synthesized in mitochondria implying the existence of a mechanism for transporting such precursors between the two organelles. Gupta et al. (1984) have partially characterized a cytosolic protein factor that could fulfill this role. The protein was shown to bind all-trans polyprenyl diphosphate and stimulate its transport across the mitochondrial membrane.
Comparisons of the amino acid sequences of various prenyl synthetases have been useful in identifying four conserved regions Fujisaki et al., 1990;Carattoli et al., 1991). Two of the regions, referred to as Domains I and 11, are characterized by an aspartate-rich repeat. The third domain is similar to an active site peptide isolated from avian FPP synthetase (Brems et al., 1981). The hallmark of the fourth region, located at the extreme carboxyl terminus, is a preponderance of basic residues. Based on the alignments of enzymes that utilize allylic and/or homoallylic prenyl diphosphate substrates, we proposed that Domain I1 comprises the allylic binding site . Preliminary site-directed mutagenesis of aspartate residues within Domain I1 of rat FPP synthetase indicate is critical for enzyme activity and mutants may have a reduced affinity for the homoallylic substrate IPP.' Analysis of the protein encoded by COQ2 suggests that a sequence similar to the consensus of Domain I1 is also present in this transferase. The sequence in question is located between residues 134-156 (Fig. 8). Based on the number of identical amino acids, this * P. F. Marrero and P. A. Edwards, in preparation.

4135
region of PHB:polyprenyltransferase is most similar to the corresponding region in several enzymes that catalyze the head-to-head condensation of allylic polyprenyl diphosphate substrates such as squalene synthetase.
The amino acid sequence alignment in Fig. 8 reveals a significant amount of sequence identity within functionally similar groups; however, some interesting similarities exist between distinct groups. For instance, FPP synthetase (Escherichia coli, Fig. 8 A ) and CrtE (Rhodobacter capsulutus, Fig.   8C) share 10 of 23 identical amino acids. Additionally, CrtE (Erwiniu herbicolu, Fig. 8C) and MOD5 (S. cereuisiae, Fig.  8F) share 9 of 23 identical amino acids. ORF323 is encoded in cyanelle DNA from the alga Cyanophora paradoxa (Fig.   8B). The identity of the ORF323 gene product is unknown although the protein shows 29% sequence identity with the CrtE gene product from Rhodobacter (Michalowski et al., 1991). ORF323 actually shows considerably more sequence identity with the polyprenyl synthetase enzymes in group A. Alignment of ORF323 with HPP synthetase, for example, reveals 45% identity over 163 amino acids (not shown). Additionally, ORF323 is 70% (16/23) identical with Domain I of HPP synthetase (not shown). Domain I is another highly conserved region among polyprenyl synthetase enzymes .
The enzymes depicted in Fig. 8E catalyze the head-to-head condensation of polyprenyl diphosphate substrates. Consistent with this reaction mechanism CrtB and squalene synthetase both possess a second region with some identity to Domain I1 that may represent a binding site for the second polyprenyl diphosphate substrate (data not shown).3 The identity of the pTOM5 protein is unkown. The gene encoding pTOM5 from tomato was isolated on the basis of its induced expression during fruit ripening and may represent the plant homolog of CrtB. A significant amount of information exists about the interaction of farnesyl diphosphate synthetase with its substrates. Holloway and Popjak (1967) found that the binding of the homoallylic and allylic substrates to the enzyme is ordered with the allylic substrate binding first. Studies with purified avian liver farnesyl diphosphate synthetase indicated that, in the presence of magnesium, each subunit possesses a single binding site for the homoallylic substrate isopentenyl diphosphate and a single binding site for the allylic polyprenyl substrates dimethylallyl diphosphate, geranyl diphosphate, or farnesyl diphosphate (Reed and Rilling, 1976). These observations have been extended by King and Rilling (1977) who found that the presence of 1 mM MnC12 or MgCl, greatly enhances the binding of geranyl diphosphate and isopentenyl diphosphate to the enzyme. In addition, binding of Mn2+ to farnesyl diphosphate synthetase is apparently nonspecific only in the absence of substrate, whereas in the presence of substrate Mn2+ binds to two specific sites in the enzyme. These data imply that homoallylic and allylic substrates bind as their magnesium salts. Gotoh et al. (1988) suggested both substrates bind as magnesium salts, but, once bound, magnesium is captured by the enzyme from isopentenyl diphosphate. The magnesium bound to the allylic substrate would be expected to render the pyrophosphate a better leaving group. This mechanism predicts that once condensation has taken place magnesium captured by the enzyme is released together with the newly formed allylic diphosphate product.
Popjhk and co-workers (1969) suggested that interaction of prenyltransferases with isoprenoid diphosphates is likely to Location of second conserved region: Squalene synthetase (S. cereuisiae) amino acids 75-97; CrtB (R. capsulatus, E. herbicola) amino acids 39-61. be articulated through the highly charged phosphates. Therefore, assuming Domains I and I1 represent substrate binding sites, interaction of the substrates with the enzyme could be facilitated by salt bridges between the carboxyl groups of the aspartic acid side chains and magnesium . There is precedence for this type of interaction. For example, x-ray crystallographic studies have shown that the carboxylate group of Asp8' in EF-Tu was shown to form a magnesium salt bridge with the &phosphate group of GDP (Jurnak, 1985). Likewise, Aspb7 of ras (p21) was shown to interact with the P-phosphate of GDP in an analogous manner (McCormick et al., 1985).
Of significance in evaluating the possible role of magnesium in PHB:polyprenyltransferase function is the observation of Casey and Threlfall (1978) as well as that of our own laboratory ) that the specificity of the yeast polyprenyltransferase can be manipulated by altering magnesium concentrations in the assay. At low magnesium concentration (3 mM), PHB:polyprenyltransferase utilizes predominantly hexaprenyl diphosphate, its natural substrate. When the concentration of magnesium is raised to 33 mM the enzyme becomes activated and loses its specificity as evidenced by its utilization of any available allylic polyprenyl diphosphate as a substrate.
Similarly, preliminary experiments with the DPRl-dependent yeast protein: farnesyltransferase have shown that the divalent cation requirements are substantially different depending on whether this enzyme employs farnesyl diphosphate or geranylgeranyl diphosphate as a ~ubstrate.~