Coq3 and Coq4 define a polypeptide complex in yeast mitochondria for the biosynthesis of coenzyme Q.

Coenzyme Q (Q) is a redox active lipid essential for aerobic respiration in eukaryotes. In Saccharomyces cerevisiae at least eight mitochondrial polypeptides, designated Coq1-Coq8, are required for Q biosynthesis. Here we present physical evidence for a coenzyme Q-biosynthetic polypeptide complex in isolated mitochondria. Separation of digitonin-solubilized mitochondrial extracts in one- and two-dimensional Blue Native PAGE analyses shows that Coq3 and Coq4 polypeptides co-migrate as high molecular mass complexes. Similarly, gel filtration chromatography shows that Coq1p, Coq3p, Coq4p, Coq5p, and Coq6p elute in fractions higher than expected for their respective subunit molecular masses. Coq3p, Coq4p, and Coq6p coelute with an apparent molecular mass exceeding 700 kDa. Coq3 O-methyltransferase activity, a surrogate for Q biosynthesis and complex activity, also elutes at this high molecular mass. We have determined the quinone content in lipid extracts of gel filtration fractions by liquid chromatography-tandem mass spectrometry and find that demethoxy-Q(6) is enriched in fractions with Coq3p. Co-precipitation of biotinylated-Coq3 and Coq4 polypeptide from digitonin-solubilized mitochondrial extracts shows their physical association. This study identifies Coq3p and Coq4p as defining members of a Q-biosynthetic Coq polypeptide complex.

and Saccharomyces cerevisiae Q 6 . Within the inner mitochondrial membrane Q functions in respiratory electron transport, where it transfers two electrons from either complex I or complex II to complex III (1). Q is present in almost all organisms, with the quantity roughly matching the respiratory capability of the tissue from which it is isolated (2). In addition to this role, Q has been found to act as a chain-breaking antioxidant (3) and to be involved in the electron transport chains of plasma and lysosomal membranes (4,5). Q supplementation in humans slows the functional decline in patients with Parkinson disease (6), is useful for treating patients with respiratory chain defects (7), and has promise as a treatment in other neurodegenerative diseases (8).
As observed with other redox active compounds, Q has the potential to act as a source of oxidative stress. In certain species, the amount of reactive oxygen species generated by mitochondria has been related to Q content (9,10). These described dual roles as antioxidant and prooxidant associate Q with both extended and shortened life spans. C. elegans fed E. coli diets lacking Q 8 , and C. elegans clk-1 mutants with defects in Q biosynthesis show increased life spans, and we have proposed that the decreased levels of Q are associated with a decreased level of the Q semiquinone radical (Q . ) and a decreased propensity to produce superoxide and other reactive oxygen species (11,12). However, supplementation of the E. coli Q 8 -replete diet with Q 10 leads to enhanced life span in C. elegans (13). The mechanisms producing these different life span outcomes are unknown.
Many of the gene products required for Q biosynthesis have been identified in both prokaryotes and eukaryotes (14,15). Yeast mutants with defects in Q biosynthesis (the coq mutants) have provided a valuable system for understanding eukaryotic Q biosynthesis (16,17). The functions of the Coq polypeptides identified so far include synthesis of the polyprenyl diphosphate tail (Coq1) and its transfer to the benzoquinone precursor 4-hydroxybenzoic acid (Coq2), and subsequent ring modifications (Coq3, Coq5, Coq6, and Coq7/Clk-1) (Fig. 1). Two Coq polypeptides, Coq4 and Coq8, are essential for Q biosynthesis but function in an unknown manner (18,19). Yeast strains harboring null mutations in coq3, coq4, coq5, coq6, coq7, or coq8 accumulate the first lipid-associated intermediate 3-hexaprenyl-4-hydroxybenzoic acid (HHB) (Fig. 1) (20,21). This lipid intermediate is detected after metabolic labeling with the dedicated precursor 4-[U- 14 C]hydroxybenzoic acid. When the orthologous genes are mutated in E. coli, more distinct intermediates characteristic of each blocked step accumulate (22). One possible explanation for the difference in the accumulation of Q intermediates between E. coli and yeast is that in yeast the Coq polypeptides assemble into a multisubunit complex necessary to perform each of the steps subsequent to formation of HHB. This prediction is sup-ported by the presence of DMQ 6 (2-hexaprenyl-6-methoxy-3methyl-1,4-benzoquinone, Fig. 1) (23,24) in yeast coq7 mutants harboring certain amino acid substitution mutations.
Recently, we have observed a co-dependence between the COQ gene products and the stability of Coq3p, Coq4p, and Coq6p (25). Each of the coq -null mutants showed decreased steady-state levels of Coq3p, Coq4p, and Coq6p; however, levels of Coq1p and Coq5p were not affected. Hsu et al. (26) showed that O-methyltransferase activity was decreased in the coqnull mutants relative to respiratory deficient control mutants (atp2, cor1). COQ3 mRNA levels were constant, suggesting post-transcriptional regulation. The decrease in Coq3p, Coq4p, and Coq6p levels observed in the coq mutants may thus be caused by their degradation secondary to a destabilized Coq multisubunit complex(es).
Of the eight proteins involved in the biosynthesis of Q in S. cerevisiae, five have been identified as mitochondrial matrix proteins peripherally associated with the inner membrane (18,25,(27)(28)(29). Submitochondrial localization of a hemagglutinintagged Coq7 fusion protein identified it as an integral membrane protein of the mitochondrial inner membrane (30), and it is hypothesized to be an interfacial membrane protein (31). At the present time the submitochondrial localization of Coq2p has not been characterized, but it has been proposed to contain four transmembrane domains based on hydropathy predictions (32) and is imported to the mitochondria (33).
Here we characterize the yeast Q-biosynthetic complex. Coq3p is a key polypeptide component of the proposed complex because it functions in both an early step and the last step of Q biosynthesis (27). The enzymatic assay for Coq3p methyltransferase (EC 2.1.1.114), converting DMeQ 3 H 2 (2-farnesyl-5-hydroxy-6-methoxy-3-methyl-1,4-benzoquinol) to Q 3 H 2 (Fig. 1), provides a surrogate for Q biosynthesis and complex activity. In this study, detergent solubilization of mitochondria, Blue Native (BN)-PAGE analyses, size exclusion chromatography, and avidin capture of biotinylated Coq3 provide the first physical evidence of interactions between Coq polypeptides and identify Coq3p and Coq4p as defining members of a Q-biosynthetic Coq polypeptide complex.
Mitochondrial Isolation and Solubilization-Yeast cultures were grown to an A 600 nm of 3-4, and mitochondria were isolated and purified on nycodenz gradients as described previously (35), except that mitochondria were collected from the interface of the 14%/20% step gradient. Mitochondria were stored frozen as 1-2-mg aliquots as described but without bovine serum albumin. Nycodenz-purified mitochondria were solubilized as described previously (36). In brief, 2 mg of mitochondria was solubilized in 200 l of 2% digitonin, 1 mM dithiothreitol, 10% glycerol, 50 mM NaCl, 20 mM HEPES, pH 7.4, 1.6 mM Pefabloc/ AEBSF (Sigma), and 0.39 mg/ml protease inhibitor mixture (EDTAfree, Roche Applied Science) for 30 min at 4°C. Alternatively, the digitonin/protein ratio was adjusted as described. Insoluble material was pelleted by centrifugation in a Beckman Airfuge (100,000 ϫ g, 10 min, chilled centrifuge tubes and rotor), and the supernatant was collected. Proteins were assayed by BCA assay (Pierce), with standards and samples in 0.2% Triton X-100 as a diluent.
BN-PAGE-BN-PAGE was performed as described previously (36) with 5-13.5% or 6 -16% gradient gels, except that gel volumes were decreased to use the Bio-Rad mini gel format. First dimension gels were 1 mm thick. Special combs (polystyrene, 1.5 mm thick, Orange County Industrial Plastics, Anaheim, CA) were made by the UCLA physics machine shop for casting the preparative well and included three separate lanes for molecular mass standards and control yeast extracts in the second dimension analysis. First dimension gel slices were soaked in 500 l of 65°C 2ϫ SDS sample buffer for 10 min before loading onto pre-cast 10 -13% SDS-polyacrylamide gels (38). Proteins were routinely blotted onto polyvinylidene difluoride membranes, and the dyes removed by serial immersion in 100% methanol prior to blocking in 1% nonfat milk, phosphate-buffered saline, 0.1% Tween 20. Detection of proteins was by ECL using Super Signal West Pico (Pierce) and luminescence detected by VersaDoc image processing software (Bio-Rad).
2-Hydroxy-3,4-dimethoxy-6-methylacetophenone (1)-Under an argon atmosphere, aluminum chloride (11 g, 82 mmol) was dissolved in Et 2 O (32 ml). Acetyl chloride (2.1 ml, 30 mmol) was added followed by the addition of 3,4,5-trimethoxytoluene (4.6 ml, 27 mmol). The reaction was allowed to proceed for 30 min and then quenched with H 2 O. The resulting mixture was brought to pH 1 with concentrated HCl and extracted with Et 2 O. The organic layers were combined and extracted with 1 M NaOH. The aqueous layers were then combined and adjusted to pH 1 as before, extracted with Et 2 O, dried over Na 2 SO 4 , and concentrated in vacuo to obtain compound 1 (3.8 g, 65% yield). prior to the reaction. Compound 1 (5.00 g, 23.8 mmol) and NaOH (19.0 g, 47.6 mmol) were dissolved in 75 ml of H 2 O. A solution of 6.0 ml of 30% H 2 O 2 and 90 ml of H 2 O was transferred to the reaction flask, and the reaction was allowed to proceed for 15 min. Concentrated HCl was added to quench the reaction until pH 3 was reached. This solution was extracted with EtOAc, dried over Na 2 SO 4 , and concentrated in vacuo. Flash chromatography (8:2 Hex/EtOAc) using silica gel provided compound 2 (3.7 g, 85%).
2-Farnesyl-5-hydroxy-3-methoxy-6-methyl-1,4-benzoquinone (4) (DMeQ 3 )-All steps were performed under argon in the absence of light. Granular lithium metal (278 mg, 40.0 mmol) was suspended in 8 ml of tetrahydrofuran, and the mixture was cooled to Ϫ70°C. Trimethyltin chloride (1.0 ml, 8.0 mmol) was added dropwise, and the mixture was allowed to warm to room temperature over 30 min. The resulting (Me) 3 SnLi solution was transferred via syringe to a new vessel and cooled to Ϫ78°C. Farnesyl chloride (2.3 ml, 8.8 mmol) was added to the solution, and the reaction was allowed to proceed for 1 h before warming to room temperature. The reaction was quenched with brine, extracted with Et 2 O, dried over Na 2 SO 4 , and concentrated in vacuo to give farnesyl trimethylstannane.
Fumagatin (250 mg, 1.50 mmol) was dissolved in 30 ml of CH 2 Cl 2 and cooled to Ϫ78°C before adding BF 3 ⅐OEt 2 (0.56 ml, 4.5 mmol). Farnesyl trimethylstannane (604 mg, 1.60 mmol) was then dissolved in 1 ml of CH 2 Cl 2 and added to the mixture. The reaction was allowed to proceed for 1 h and then worked up as before. The crude DMQ 3 H 2 was dissolved in 20 ml of Et 2 O before adding Ag 2 O (344 mg, 0.32 mmol). After 1 h, the reaction mixture was filtered and then worked up as before. Flash chromatography (6.5:3.5 Hex/EtOAc) was performed using silica gel to obtain DMeQ 3 (53 mg, 21%).
Methyltransferase Assays-Coq3 methyltransferase assays were performed as described (26)  To stop the reaction, each tube was transferred to dry ice, and samples were extracted twice with 1 ml of heptane. Extracts were dried under N 2 gas and stored at Ϫ20°C. Reverse phase HPLC (Betabasic C18, 150 ϫ 4.6 mm, 5 m, Thermo Electron Corporation) was used to separate the 3 H-labeled product, Q 3 . Initial conditions were: Solvent A (2 mM borate) 30% and Solvent B (95% ACN, 5% 2 mM borate) 70%. After injection the buffer composition and flow rate were linearly changed (time min/%B/flow rate (ml/min): 0/70/1, 2/70/1, 5/100/1.4). After 14 min the mobile phase and flow rate reverted to the initial state. HPLC fractions (1 min) were collected, added to 10 ml of Safety Solvent (Research Products International), and subjected to scintillation counting. Methyltransferase assays of gel filtration fractions were performed on the same day the gel filtration fractions were collected.
Construction of Plasmids Containing Biotin-tagged Coq3-The gene for biotinylated Coq3p was constructed by an in-frame fusion of a PCR-amplified fragment containing the COQ3 open reading frame and 650 bp of 5Ј-flanking sequence to the vector YEp352-Bio6 (42). The amplification was carried out with pRS12A2-2.5Sma as template DNA (43) and a forward primer pBC3-1 (5Ј-TAAATTTCTGAGCTCGC-CCCCGGGTATTTCATTTG-3Ј) and a reverse primer pBC3-2 (5Ј-CGCGGGATCCATTCAGTCTCTGAATAGCCA-3Ј). The PCR product was digested with SacI and BamHI and ligated to the SacI and BamHI sites of YEp352-Bio6 to generate pBT3-1. To construct a single copy vector containing the biotinylation site, pRS316 (44) was digested with BamHI and SacI, treated with Klenow to generate blunt ends, and the vector re-ligated. This modified vector was then digested with HindIII and EcoRI and ligated with the 250-bp HindIII/EcoRI fragment obtained from YEp352-Bio6 to generate pSBT1. The COQ3 amplicon described above was then inserted into the SacI and BamHI sites of pSBT1 to generate pSBT3-1. The Coq3-biotinylated fusion protein (Coq3bt) retains activity as assayed by the ability of either the single copy or the multicopy plasmid construct to rescue the coq3 null yeast strain for growth on a nonfermentable carbon source (YPG).
Avidin Capture-Mitochondria (500 g) were lysed in 125 l of digitonin buffer as above at a digitonin/protein ratio of 2:1. 10% of a high speed supernatant was reserved as a gel loading control, and the remaining supernatant was incubated 2 h at 4°C with 100 l of avidinagarose (Pierce). Digitonin conditioning and wash buffer was 20 mM HEPES, pH 7.6, 50 mM NaCl, 1.6 mM AEBSF, 1 mM dithiothreitol, and 0.1% digitonin. The agarose was collected by a brief centrifugation (500 ϫ g, 30 s) the supernatant reserved, and the agarose beads were washed twice with wash buffer (1-ml volume, with and without 500 mM NaCl). The original supernatant and each wash were precipitated with trichloroacetic acid (final concentration 10%). The remaining agarose pellet was resuspended in an equivalent volume of 2ϫ SDS sample buffer at 74°C. After gel electrophoresis, proteins were transferred to a polyvinylidene difluoride membrane and blotted as described in figure legends. NeutraAvidin-horseradish peroxidase antisera were obtained from Pierce and used as suggested.

Reverse Phase HPLC-APCI-MS/MS Multiple Reaction Monitoring
Analysis-An aliquot of each gel filtration fraction (150 l) was placed into 5-ml borosilicate glass conical tubes with Teflon-lined caps and stored at Ϫ20°C until extraction. Samples were thawed at 4°C with gentle shaking, 968 pmol of internal standard was added (Q 4 , Sigma), followed by 1 ml of hexane/2-propanol (3:2, v/v) (45). After shaking at 200 rpm for 10 min the samples were vortexed vigorously for 10 s and the phases allowed to separate (room temperature). The upper phase was transferred to a 2-ml amber glass vial. The lower phase was reextracted with heptane (1 ml) and the pooled organic phases concentrated by vacuum centrifugation and stored dry at Ϫ20°C. Standard samples were prepared simultaneously with each batch of gel filtration samples. The standard samples contained the same amount of Q 4 internal standard and increasing amounts of Q 6 (0.7 fmol/l to 7.3 pmol/l); one set was extracted simultaneously with the gel filtration fractions, and the other set was not extracted; both sets were taken to dryness by vacuum centrifugation.
Dried samples were redissolved in acetonitrile (500 -1,500 l) and injected (50 l/injection) onto a reverse phase column (Thermo Hypersil Multiple reaction monitoring peak areas were measured by instrument manufacturer-supplied software (MacSpec version 3.3, PE Sciex, Ontario, Canada). The calibration curve from the extracted set of standards (ordinate, Q 6 /Q 4 peak area ratio; abscissa, Q 6 amount) had correlation coefficients (r 2 ) around 0.996 with slopes around 2.2 (intercept forced to 0). Concentrations of Q 6 and DMQ 6 (in Q 6 equivalents) present in the extracts were interpolated directly from the calibration curve, and quantities in each gel filtration fraction were calculated after correction for dilution and amount assayed. Lipid extractions and analyses were completed on aliquots from two separate gel filtration separations and were essentially identical.

Detergent Solubilization of Coq3p and O-Methyltransferase
Activity-Based on the hypothesis that Coq3p is a key component of a coenzyme Q-biosynthetic complex, the assay for detection of such a complex focused upon the enzymatic activity of Coq3p, which methylates a farnesylated analog of the penultimate intermediate DMeQ 3 H 2 to form Q 3 H 2 ( Fig. 1) (27). Although Coq3 also methylates an earlier Q intermediate, the farnesylated analog 3,4-dihydroxy-5-farnesyl-benzoic acid (46), the highest level of [ 3 H-methyl]AdoMet-dependent yeast Coq3 activity was associated with Q 3 H 2 formation from DMeQ 3 H 2 (27). The yeast Coq3 polypeptide is peripherally associated with the mitochondrial inner membrane on the matrix side (27). Attempts to generate soluble active Coq3p by subjecting either mitochondria or mitoplasts to sonication, or treatment with Triton X-100, were not successful (data not shown). Therefore, de-tergent solubilization studies were performed with digitonin and dodecyl maltoside at several different ratios of detergent to protein. Mitochondria from the yeast strain W303-1B were treated with detergent, separated into 100,000 ϫ g supernatant and pellet fractions, and Coq3 methyltransferase activity (Fig. 3A) and recovery of the Coq3 and Coq4 polypeptides (Fig. 3B) determined. In general, more Coq3 methyltransferase activity was recovered in the supernatant after digitonin solubilization compared with dodecyl maltoside (Fig. 3A), although dodecyl maltoside was more effective at releasing Coq3p and Coq4p from the 100,000 ϫ g pellet (Fig. 3B). Interestingly, the two detergent treatments resulted in profound differences in the distribution of Coq3p and Coq4p as revealed by BN-PAGE followed by SDS-PAGE in the second dimension (Fig. 3, C and D). This analysis showed that the majority of Coq3p and essentially all of Coq4p migrate as a high molecular mass complex in the digitoninsolubilized mitochondria. However, Coq3p appears less punctate, Wild-type mitochondria (80 g) were solubilized with dodecyl maltoside (Ddm) or digitonin (Dig) at the indicated ratios of detergent to protein (g/g). Equal amounts of each supernatant (4 l) were added to methyltransferase assays containing 50 M DMeQ 3 H 2 as substrate. The formation of the methylated product Q 3 was determined as described under "Experimental Procedures" and is depicted as the pmol of transferred CH 3 groups/mg of protein/h. B, detection of Coq3p and Coq4p in supernatant (S) and pellet (P) fractions of detergent-extracted mitochondria. Right, dodecyl maltoside; left, digitonin, as shown in A. The pellet proteins were resuspended in an equivalent volume of extraction buffer, and 4-l aliquots were subjected to SDS-PAGE (10 -13% gradient gel) and immunoblot analyses with antisera to Coq3p or Coq4p. C, two-dimensional BN-PAGE analysis of Coq3p in digitonin-solubilized mitochondria (1:1, g/g). 200 g of mitochondria was solubilized by digitonin as described above and subjected to BN-PAGE (6 -16%) in the first dimension and SDS-PAGE (13%) in the second dimension, and immunoblot analyses were performed as described. The point of origin and size separation for the first BN dimension is designated by an arrow with the position of the indicated molecular mass standards indicated in kDa. Aliquots of mitochondria from wild-type (wt) or coq3-null mutants (⌬3) were subjected only to the second dimension (SDS-PAGE) separation. D, two-dimensional BN-PAGE analysis of Coq3p in dodecyl maltoside-solubilized mitochondria (1:1, g/g). 200 g of mitochondria was solubilized with dodecyl maltoside and subjected to the same analysis as in C. and Coq4p is shifted predominantly to lower molecular mass species in mitochondria solubilized with dodecyl maltoside. Because solubilization with digitonin better preserved O-methyltransferase activity and discrete high molecular mass complexes of Coq3p and Coq4p, we chose to carry out further analyses with supernatants obtained from digitonin-solubilized mitochondria.
Coq3p and O-Methyltransferase Activity Coelute with Coq4p and Coq6p as a High Molecular Mass Complex-Gel filtration analyses were performed to study the oligomeric state of Coq3p and Coq4p in the putative Coq biosynthetic complex. Mitochondria were solubilized with digitonin (2:1, g/g) and the 100,000 ϫ g supernatant obtained from 2 mg of mitochondria was applied to a Superose 6 column. The column eluant was monitored for protein content (A 280 nm ), and fractions were collected and assayed for Coq3 O-methyltransferase activity, Coq polypeptides, and analyzed for the content of Q 6 and DMQ 6 . Coq3 O-methyltransferase activity (Fig. 4A) eluted predominantly in fractions 17-21, peaking in fraction 19, just larger than the migration point of thyroglobulin (669 kDa). A second region of Coq3 O-methyltransferase activity eluted in fractions 26 -27 in the size range of 150 -240 kDa. SDS-PAGE and immunoblot analyses of aliquots of the same gel filtration fractions show that Coq3p is detected in two distinct peaks that correspond to the regions of Coq3p activity (Fig. 4B). The elution profiles of mitochondrial proteins of known size (Rip1p/Isp2p and IDH) were also detected in the gel filtration fractions by immunoblot. These oligomeric mitochondrial protein complexes provide an endogenous calibration for the gel filtration analysis. Rip1p/ Isp2p is the Rieske iron sulfur protein of the homodimeric bc 1 complex (47) of about 500 kDa (48). Digitonin solubilization of yeast mitochondria produces a bc 1 -cytochrome c oxidase supercomplex (III 2 -IV 2 ) of about 1,000 kDa (37). IDH is an octamer of 304 kDa (49). Most of the O-methyltransferase activity is found in fractions that contain Coq3p, Coq4p, and Coq6p, with some cohort of Coq5 (Fig. 4B) suggesting a functional association among these polypeptides. Indeed, the majority of Coq4p and Coq6p coelute with Coq3p and O-methyltransferase activity, suggesting that these three polypeptides may form a stable complex. However, only a small portion of Coq5p coelutes with Coq3p, and Coq1p elutes as an apparently distinct 300-kDa complex. Thus, by gel filtration, Coq3p resides in two multimeric protein complexes, a large complex involving Coq4p and Coq6p, and a smaller complex that coelutes with Coq5p. The relationship between these two states of physical association is not clear.
The Q Biosynthetic Intermediate DMQ 6 Is Associated with the Coq3-Coq4-Coq6 Complex-The co-migration of Coq3 activity and of Coq3, Coq4, and Coq6 polypeptides in gel filtration fractions provides physical evidence for a high molecular mass Q-biosynthetic complex. Previous work with the yeast coq mutants has implicated Q, or a Q-biosynthetic intermediate, as a lipid component potentially involved in maintaining steady levels of the Coq3, Coq4, and Coq6 polypeptides (25,26). To understand the relationship between Q and Q intermediates and Coq polypeptide levels, we have applied an assay for the detection of Q 6 and DMQ 6 to the lipid extracts of gel filtration fractions. This is a novel application of a very sensitive LC-MS/MS method, probing whether a lipid intermediate (or product) is associated with the enzyme complex responsible for its synthesis (Fig. 4C). The smallest amount of standard for quantification from the calibration curve is 700 fmol/l (50-l injection volume). The lipids present in the fractions were quantified by LC-MS/MS, and representative data for fraction 19 are shown in Fig. 5. Fig. 5A shows the product ions produced by collisional dissociation of two compounds targeted for analysis, DMQ 6 , and Q 6 , whose protonated molecular ions are found at mass/charge ratio (m/z) of 561.4 and 591.4, respectively (the molecular ion is designated). DMQ 6 produces a predominant tropylium ion at m/z 167.2, whereas Q 6 produces a tropylium ion of m/z 197.0. The benzylium or tropylium ion (50) is the characteristic base peak of the product ion spectra of benzoquinone intermediates and contains the quinone head group fragment with a single methylene remnant of the prenyl tail. Multiple reaction monitoring, where the precursor ions are instrument-selected to produce a specific fragment to be quantified, is a sensitive, specific, and highly quantitative procedure. When used in tandem with HPLC, a defined retention time can further raise the specificity (Fig. 5B). As can be seen, the intermediate DMQ 6 m/z transition 561.4 -167 has a slightly earlier elution than that of Q 6 .
Although Q 6 elutes in the earliest gel filtration fractions (Fig.  4C), perhaps associated with high molecular mass electron transport chain components (Fig. 4B; Rip1p), most Q 6 elutes in later fractions, with little discrimination between fractions 23 through 28. In contrast, DMQ 6 content is highest in the gel FIG. 4. Coq3, Coq4, and Coq6 coelute as a high molecular mass complex with Coq3 O-methyltransferase activity and DMQ 6 . A, gel filtration analysis of Coq3 O-methyltransferase activity. Wild-type mitochondria (2 mg) were solubilized with digitonin (2:1, g/g, detergent/ protein) and the 100,000 ϫ g supernatant subjected to gel filtration chromatography. For details of the calibration standards, see "Experimental Procedures." The A 280 nm and O-methyltransferase activity (pmol of methyl groups/fraction/h) are depicted. B, gel filtration analysis of Coq polypeptides. The proteins Coq3, Coq4, Coq1, Coq5, Coq6, Rip1, and IDH were detected in eluate fractions by immunoblot analysis. C, analysis of Q 6 and DMQ 6 content in gel filtration fractions. Reverse phase HPLC-APCI-MS/MS multiple reaction monitoring was used as described under "Experimental Procedures" and in Fig. 5 to detect specifically ions produced from precursor ions DMQ 6 and Q 6 . The Q 6 content (pmol/fxn) is designated by bars; the content of DMQ 6 (pmol/fxn) by (ࡗ). All analyses shown were done in duplicate from two gel filtrations done on the same day and are similar to previous analyses. filtration fractions containing Coq3p. DMQ 6 is a stable intermediate and the product of the Coq5 C-methyltransferase step (Fig. 1). The distribution of Q 6 is more homogeneous, perhaps reflecting its participation in mitochondrial electron transport complexes and dehydrogenases, whereas DMQ 6 distribution may reflect a more specific residence with the Coq polypeptides that produced it.
Coq3p and Coq4p Co-migrate at High Molecular Mass in One-dimensional BN-PAGE Analyses-To characterize better the size of the Coq3p and Coq4p complexes, digitonin-solubilized mitochondria were separated on 5-10% gradient BN gels along with protein standards in the high molecular mass range. Immunoblots were prepared and probed sequentially with antibodies to Coq4 (Fig. 6A), Coq3 (Fig. 6B), and then Rip1 (Fig.  6C). The high molecular mass complexes containing Coq3 and Coq4 polypeptides are absent in mitochondria isolated from a coq3 null mutant control strain. Coq3 and Coq4 co-migrate under these separation conditions that allow for enhanced resolution of high molecular mass species. The migration of Coq3p and Coq4p coincides with the highest molecular mass complexes containing the Rip1 polypeptide. Rip1 localization has generally been attributed to respiratory supercomplexes containing the bc 1 complex at predicted masses of 750 and 1,000 kDa (51). High molecular mass complexes containing Rip1 are not detected in either a coq3-null mutant or a cor1-null mutant (Fig. 6C). The absence of the bc 1 complex in the cor1-null mutant is predicted because the Cor1 polypeptide, an extramembranous subunit that is attached to the transmembrane domains (48), is required for assembly of the bc 1 complex, and without it many of the bc 1 complex polypeptides are unstable (52,53). The effect of the coq3-null mutation on Rip1 high molecular mass complex suggests that the bc 1 complex is also dysfunctional in the absence of Coq3p. In contrast, the high molecular mass complexes containing Coq3 and Coq4 are maintained in the cor1-null mutant, suggesting that components of the bc 1 complex are not required for formation of the Coq3-Coq4 complex. Multiple analyses in both one-and twodimensional BN-polyacrylamide gels have confirmed that Coq3p and Coq4p co-migrate in this very high molecular mass region. BN-PAGE analyses have established physical interactions between electron transport chain complexes (51). The Rip1p immunolocalization in Fig. 6C suggests that the true molecular mass of the complex with which the Coq3 and Coq4 polypeptides associate is above 1 mDa.
A Biotin-tagged Coq3p Interacts with Coq4p-To determine directly whether the similar fractionation profiles of Coq3p and Coq4p involved physical interactions between the two proteins, a Coq3p construct was generated which contains an amino acid motif that is biotinylated by the endogenous biotin ligase. This approach has been utilized previously to isolate complex IV of the respiratory chain (54). Importantly, the biotinylated form of Coq3p (Coq3bt) rescued the coq3⌬ strain with respect to growth on glycerol with similar rates on solid media, and roughly similar amounts of protein were present in analyses of whole cell lysates from the rescued strain and wild-type controls (data not shown). Additionally, the Coq3-biotin protein (Coq3bt) is localized to mitochondria (Fig. 7A) when expressed from either single or multicopy plasmid vectors and migrates similarly to wild-type Coq3p in BN-PAGE analyses (Fig. 7B).
To determine whether Coq3p associates with Coq4p, mitochondria were solubilized with digitonin and Coq3bt captured from the derived lysate with avidin-agarose beads. Subsequent immunoblot analysis (Fig. 7C) reveals that Coq3 is precipitated only when expressed as a biotin-tagged form and that Coq4p co-precipitates with Coq3bt. Thus, Coq3p and Coq4p are physically associated with one another. DISCUSSION Experimental data here support the hypothesis that a multisubunit Coq polypeptide complex is involved in coenzyme Q biosynthesis. Detergent conditions were determined which solubilize Coq3 methyltransferase activity from yeast mitochondria. The co-migration of Coq3 activity and of Coq3, Coq4, and Coq6 polypeptides in gel filtration fractions provide physical evidence for a high molecular mass (greater than 669 kDa) Q-biosynthetic complex. In addition, these fractions contain the bulk of DMQ 6 , a stable lipid precursor of the Q-biosynthetic pathway. Further evidence that Coq3p and Coq4p coexist in a protein complex was provided by BN-PAGE analyses, demon-  6. Coq3 and Coq4 polypeptides are associated in a high molecular mass complex larger than 1,000 kDa. A, BN-PAGE analysis of digitonin-solubilized mitochondria from wild-type (wt), coq3null (⌬3), or cor1-null (⌬c) yeast strains. A 5-13% polyacrylamide gradient gel was used. The indicated polypeptides were detected by immunoblotting with anti-Coq4. B, the blot in A was stripped and re-probed with anti-Coq3p. C, the blot in B was stripped and re-probed with anti-Rip1p. Molecular mass markers are designated: thyroglobulin, 669 kDa; ferritin, 440 kDa; catalase, 232 kDa. The top the gels, as transferred to the blot, is indicated. strating that the Coq3 and Coq4 polypeptides, with respective subunit molecular masses of 32.7 and 35.5 kDa, co-migrate at a molecular mass higher than 1,000 kDa. Co-precipitation of Coq4p and biotinylated Coq3p from digitonin-solubilized mitochondrial extracts indicates their physical association. Taken together, the data demonstrate that Coq3p and Coq4p define a complex of polypeptides that synthesize Q.
The gel filtration elution profile shows that Coq3p and Omethyltransferase activity elute in a second smaller molecular mass fraction (150 -240 kDa) that contains the bulk of Coq5p. This smaller Coq3 complex may represent an interaction with Coq5p. A small amount of Coq5p elutes in the high molecular mass fractions containing Coq3, Coq4, and Coq6 polypeptides. It is possible that only a small fraction of Coq5p is associated with the larger complex, or perhaps Coq5p forms a necessary but labile interaction with this other cohort of Coq proteins. This may suggest some subunit movement between these two predominant Coq complexes or minimally some degree of precursor product exchange between them. Genetic evidence supports this idea because steady-state levels of Coq3, Coq4, and Coq6 proteins are decreased in a coq5-null mutant (25). Additionally, the analysis of a panel of coq5 point mutants showed that a subset of the mutants that are rescued by expression of E. coli ubiE, an ortholog of Coq5, retain Coq3 O-methyltransferase activity and normal steady-state levels of Coq3p and Coq4p (28).
The gel filtration data suggest that Coq1p is unlikely to participate in the Coq3-Coq4-Coq6 complex and this protein is probably not required for O-methyltransferase activity. The elution position of Coq1p is similar to that of IDH, an octamer of 304 kDa. Other polyprenyl-diphosphate synthases have been postulated to form dimers and tetramers (55,56). Coq1p elutes independently of both the large and small Coq3 complexes, and no Coq1p was detected in the high molecular mass regions by BN-PAGE (data not shown). Although the steady-state levels of Coq3p, Coq4p, and Coq6p are decreased in a coq1-null mutant, their levels can be partially restored by polyprenyl diphosphate synthases from diverse prokaryotes, including some species that do not produce Q (25). These findings suggest that a lipid product derived from polyprenyl diphosphate, rather than Coq1p itself, is important for the stability of Coq3, Coq4, and Coq6 polypeptides.
A candidate for the presumptive lipid intermediate in the Q-biosynthetic complex is DMQ 6 , the product of the Coq5 Cmethyltransferase, and the substrate for the Coq7 monooxygenase. DMQ 6 is a relatively stable intermediate present in lipid extracts of select coq7 mutants as well as wild-type yeast (23,24). We describe here a novel application of a sensitive method to detect Q 6 and DMQ 6 in lipid extracts of gel filtration fractions. This assay indicated that the majority of DMQ 6 elutes in fractions enriched in the Coq3-Coq4-Coq6 complex. The enrichment of DMQ 6 in this high molecular mass region suggests that DMQ 6 may be functionally associated with the Q-biosynthetic complex. Although Q 6 may also be associated with a Q-biosynthetic complex, it elutes over a broad range, reflecting its participation in mitochondrial electron transport complexes and possibly with other dehydrogenases.
Multienzyme complexes can serve to enhance catalytic efficiency, effect substrate channeling, and sequester reactive intermediates. In Q biosynthesis, a potentially important function is to limit the concentration of catechol and unsubstituted quinone intermediates, compounds that are notorious for their reactivity and participation in oxidative damage (57). A multienzyme complex also provides a mechanism for regulating the flux of Q and Q intermediates through this pathway and potentially impacts performance of the respiratory electron transport chain. In fact, the BN-PAGE analyses presented here suggest that assembly of the high molecular mass bc 1 complex depends upon Coq3p, or its product, QH 2 . This suggests that Coq3p, the catalytic subunit producing QH 2 , is necessary for proper bc 1 complex assembly.
The data presented here identify Coq3p and Coq4p as interacting partners in a complex estimated at about 1,000 kDa by BN-PAGE analysis. It is likely that Coq6p is also a member of this complex based on its the coelution with Coq3p and Coq4p by gel filtration. Although prokaryotic orthologs have been identified for Coq3, Coq6, and other Coq polypeptides, homologs of Coq4 are not found in prokaryotes. Whether Coq4 serves as an enzyme in Q biosynthesis is unknown, but it is now clear that it serves as an important structural polypeptide component of the Q-biosynthetic complex. The relative stoichiometry of these polypeptides in this complex has yet to be determined. Given the size of the complex, it is quite likely that other polypeptides are also present. What anchors the peripherally associated Coq subunits to the membrane? It is tempting to speculate that one or both of the two Coq polypeptides predicted to be integrally associated with the inner membrane (Coq2p) and interfacially associated with the membrane (Coq7p) may serve as membrane protein anchors in this complex. However, the nature of the association of Coq1, Coq3, Coq4, Coq5, and Coq6 polypeptides with the membrane or with other peripherally associated membrane proteins is not clear. A better understanding of the Q-biosynthetic complex will require identifying each of the protein components, determining the stoichiometry of the polypeptide and lipid components, and elucidating whether the complexes observed represent stable distinct entities or are dynamically related.