para-Aminobenzoic Acid Is a Precursor in Coenzyme Q6 Biosynthesis in Saccharomyces cerevisiae*

Coenzyme Q (ubiquinone or Q) is a crucial mitochondrial lipid required for respiratory electron transport in eukaryotes. 4-Hydroxybenozoate (4HB) is an aromatic ring precursor that forms the benzoquinone ring of Q and is used extensively to examine Q biosynthesis. However, the direct precursor compounds and enzymatic steps for synthesis of 4HB in yeast are unknown. Here we show that para-aminobenzoic acid (pABA), a well known precursor of folate, also functions as a precursor for Q biosynthesis. A hexaprenylated form of pABA (prenyl-pABA) is normally present in wild-type yeast crude lipid extracts but is absent in yeast abz1 mutants starved for pABA. A stable 13C6-isotope of pABA (p- amino[aromatic-13C6]benzoic acid ([13C6]pABA)), is prenylated in either wild-type or abz1 mutant yeast to form prenyl-[13C6]pABA. We demonstrate by HPLC and mass spectrometry that yeast incubated with either [13C6]pABA or [13C6]4HB generate both 13C6-demethoxy-Q (DMQ), a late stage Q biosynthetic intermediate, as well as the final product 13C6-coenzyme Q. Pulse-labeling analyses show that formation of prenyl-pABA occurs within minutes and precedes the synthesis of Q. Yeast utilizing pABA as a ring precursor produce another nitrogen containing intermediate, 4-imino-DMQ6. This intermediate is produced in small quantities in wild-type yeast cultured in standard media and in abz1 mutants supplemented with pABA. We suggest a mechanism where Schiff base-mediated deimination forms DMQ6 quinone, thereby eliminating the nitrogen contributed by pABA. This scheme results in the convergence of the 4HB and pABA pathways in eukaryotic Q biosynthesis and has implications regarding the action of pABA-based antifolates.

Coenzyme Q (ubiquinone or Q) is a crucial mitochondrial lipid required for respiratory electron transport in eukaryotes. 4-Hydroxybenozoate (4HB) is an aromatic ring precursor that forms the benzoquinone ring of Q and is used extensively to examine Q biosynthesis. However, the direct precursor compounds and enzymatic steps for synthesis of 4HB in yeast are unknown. Here we show that para-aminobenzoic acid (pABA), a well known precursor of folate, also functions as a precursor for Q biosynthesis. A hexaprenylated form of pABA (prenyl-pABA) is normally present in wild-type yeast crude lipid extracts but is absent in yeast abz1 mutants starved for pABA. A stable 13 C 6 -isotope of pABA (pamino[aromatic-13 C 6 ]benzoic acid ([ 13 C 6 ]pABA)), is prenylated in either wild-type or abz1 mutant yeast to form prenyl-[ 13 C 6 ]pABA. We demonstrate by HPLC and mass spectrometry that yeast incubated with either [ 13 C 6 ]pABA or [ 13 C 6 ]4HB generate both 13 C 6demethoxy-Q (DMQ), a late stage Q biosynthetic intermediate, as well as the final product 13 C 6 -coenzyme Q. Pulse-labeling analyses show that formation of prenyl-pABA occurs within minutes and precedes the synthesis of Q. Yeast utilizing pABA as a ring precursor produce another nitrogen containing intermediate, 4-imino-DMQ 6 . This intermediate is produced in small quantities in wildtype yeast cultured in standard media and in abz1 mutants supplemented with pABA. We suggest a mechanism where Schiff base-mediated deimination forms DMQ 6 quinone, thereby eliminating the nitrogen contributed by pABA. This scheme results in the convergence of the 4HB and pABA pathways in eukaryotic Q biosynthesis and has implications regarding the action of pABAbased antifolates.
Coenzyme Q (Q) 2 is an essential polyprenylated benzoquinone lipid in cellular energy metabolism. The prenyl tail anchors Q in cellular membranes, whereas the redox chemistry of the benzoquinone ring plays a crucial role in respiratory electron transport, in catabolic and biosynthetic metabolism (1), as a co-antioxidant able to recycle vitamin E, and as a chain-terminating antioxidant (2). In these reactions the quinone ring of Q thus cycles between oxidized and reduced (QH 2 , or hydroquinone) states.
Cells rely on de novo synthesis for an adequate supply of Q. Studies in Escherichia coli, Schizosaccharomyces pombe, and Saccharomyces cerevisiae have made use of Q-deficient mutants to elucidate the biosynthetic pathway (3)(4)(5). In S. cerevisiae, nine COQ genes are required, and each of the yeast coq mutants (coq1 through coq9) lack Q 6 and are unable to grow on media containing non-fermentable carbon sources such as glycerol or ethanol. The dedicated precursors in the biosynthesis of Q are polyisoprene diphosphate, which provides the tail (S. cerevisiae synthesizes Q 6 , with a tail containing six isoprene units), and 4-hydroxybenzoic acid (4HB) (6,7). Studies in animal cells and in E. coli indicate that different metabolic pathways are used to produce 4HB. Animals (e.g. rats and humans) generate 4HB from the essential dietary amino acid tyrosine (6 -8). Phenylalanine also acts as a precursor for 4HB, however, the incorporation is thought to proceed primarily following its conversion to tyrosine via phenylalanine hydroxylase (8). The biosynthetic steps leading from 4-hydroxyphenylpyruvate to 4HB in animal cells are not yet characterized (see Fig. 1). E. coli relies on shikimate biosynthesis, the formation of chorismate, and chorismate pyruvate lyase (encoded by the ubiC gene) to synthesize 4HB (9,10). E. coli ubiC mutants lack Q unless 4HB is provided in the growth media (9). E. coli mutants lacking shikimate or chorismate also require exogenous 4HB to synthesize Q (11). Thus, E. coli cells are unable to convert tyrosine or phenylalanine to Q and rely exclusively on the de novo synthesis of 4HB from chorismate.
In contrast, S. cerevisiae may utilize either shikimate or tyrosine to synthesize the aromatic ring precursor of Q (6,12). Yeast preferentially utilize shikimate to produce Q, and tyrosine is utilized only when the synthesis of shikimate is blocked (12). Thus, yeast aro1C mutants (unable to synthesize shikimate), and yeast aro2 mutants (unable to synthesize chorismate) still synthesize Q de novo, because they are able to utilize exogenously added tyrosine (Fig. 1). The steps producing 4HB from tyrosine have not been identified, although the pathway may be similar to that described for the catabolism of p-coumarate to 4HB in Acinetobacter bayli (13). Although it has been assumed that yeast may generate 4HB via chorismate pyruvate lyase activity, S. cerevisiae lack a homolog of UbiC. This raises the question: how do yeast utilize chorismate to produce a ring precursor of Q?
Here we describe our surprising findings that para-aminobenzoic acid (pABA), a known precursor of folates, is also an aromatic precursor for Q biosynthesis, via the synthesis of 3-hexaprenyl-4-aminobenzoic acid (prenyl-pABA). These pathways are described in Fig. 1. The biosynthetic steps in yeast necessary for the production of pABA are catalyzed by the ABZ1 and ABZ2 gene products. Abz1 amidates chorismate to make the 4-aminodeoxychorismate intermediate (14,15), and the Abz2 lyase forms free pABA (16). Import of pABA into the mitochondria is necessary for further folate synthesis; the FOL1 gene product is required for this import and also performs multiple enzymatic functions in pteroglutamoyl synthesis (17). Immunogold particle labeling and a Fol1-GFP fusion localized the tri-functional polypeptide Fol1p in yeast to mitochondrial membranes (17).
We recently became aware of similar work identifying pABA and prenyl-pABA as Q biosynthetic precursors (18). These authors identified pABA as a Q precursor in their search for iron-mediated effects on the function of the Coq7 monooxygenase in Q biosynthesis. Our studies independently determined that pABA is a novel coenzyme Q precursor, and we show prenyl-pABA is an endogenously synthesized intermediate in the Q biosynthetic pathway. We further demonstrate the relative contributions of the 13 C 6 -isotope of 4HB and pABA under competition conditions with the alternative unlabeled ring precursor. In addition we identify 4-imino-DMQ 6 in wild-type yeast and in pABA-supplemented abz1 null mutants. Based on our identification of this intermediate, we suggest a possible mechanism for the removal of the nitrogen donated by pABA, and its replacement with an oxygen atom to form the 1,4-quinone moiety in DMQ via Schiff base chemistry.

EXPERIMENTAL PROCEDURES
Yeast Growth Analysis-Yeast strains used in this work are described in Table 1. The abz1 null mutant (W303⌬abz1) was generated as described (19). Yeast colonies from YPD (2% glucose, 1% yeast extract, 2% peptone, 2% agar) plates were inoculated into 18-ϫ 100-mm culture tubes containing 5 13 C 6 ]benzoic acid ([ 13 C 6 ]4HB) were obtained from Cambridge Isotope Laboratories (Andover, MA). The manufacturers' analyses of the pure 13 C-labeled aromatic ring compounds by GC-MS and NMR verified a better than 98% chemical purity with 99% isotopic enrichment. 13 C 6 -and 14 C-labeled aromatic ring precursors were also examined for purity by HPLC. Yeast cells were grown as described to deplete pABA and folate. Optical density (A) measurements (600 nm) were recorded for each sample at time of harvest, and the total A 600 nm values were used to normalize lipid content. For radioactive labeling studies, the cells were suspended in 300 -600 ml of fresh media as above (Dogal or Doge) and grown to an A 600 nm of 0.6, then collected by centrifugation (1000 ϫ g, 5 min) and resuspended in small volumes of fresh media (ϳ2 ml), containing 4 Ci, of the designated 14 C-labeled ring precursor. Cells were collected by centrifugation as before, and pellets were stored at Ϫ20°C. Stable isotopic labeling was essentially as above except that individual amounts of [ 13 C 6 ]pABA or [ 13 C 6 ]4HB were weighed and dissolved in DMSO (Sigma), and then added to a final concentration of 2-10 g/ml.
Competition and Pulse Experiments-Time course and competition experiments with normal and stable isotope forms of precursors were conducted in 18-ϫ 100-mm culture tubes. Cells were collected from a larger volume culture, and equivalent optical density amounts of yeast cells (between 20 and 100 A 600 nm ) were re-suspended in 5 ml of fresh media as described, which contained either the DMSO (vehicle control), [ 13 C 6 ]pABA, or [ 13 C 6 ]4HB at 10 g/ml, or each of these plus an additional 10 g/ml of the competing unlabeled precursor 4HB or pABA. These samples were prepared in duplicate or triplicate, and analyzed via mass spectrometry two or three times (n ϭ 4 -9). For pulse analyses cells were grown in large volume as described above, re-suspended in pre-warmed media (30°C) to a total volume of 16 ml in a 125-ml flask, and incubated with shaking (250 rpm, 30°C). Prior to addition of labeled ring precursors, two (1 ml) aliquots were removed, to represent a "nolabel" control. [ 13 C 6 ]pABA was added (final concentration, 10 g/ml), and further 1-ml aliquots were collected in duplicate at the time points as described. The time zero point was defined by removing two aliquots prior to addition of labeled precursors. The first time point (termed 2 min) represented the addition of label with mixing and immediate removal of two sequential samples, effectively allowing the label to be present for ϳ1 min and 3 min, respectively.
Lipid Extraction-Yeast cell pellets were thawed on ice, and solvents were added as follows: 100 l of H 2 O, 1 ml of methanol, and 2 ml of petroleum ether. Samples were vortexed for 30 s each. This was repeated, and then samples were centrifuged at 1000 ϫ g to separate layers. The upper layer was moved to a new tube, 2 ml of petroleum ether was added to the lower phase, and the sample was vortexed. This upper phase was added to the previous upper organic phase, and the solvent was evaporated under N 2 gas. Samples were routinely analyzed immediately after extraction. When Q or other intermediates were quantified, Q 4 (Sigma) was added in a known amount (expected final concentration, 1 pmol/l upon analysis) as internal standard to all samples, and to a simultaneously prepared and extracted calibration curve. Typical standard curve final concentrations were 0.2 fmol/l, 1 fmol/l, 25 fmol/l, 200 fmol/l, 1 pmol/l, and 5 pmol/l. The petroleum ether extracts were dried under nitrogen gas and resuspended in 200 l of ethanol (USP, Aaper Alcohol and Chemical Co., Shelbyville, KY), in sample vials compatible for use with HPLC. Lipid extractions for the pulse experiments were similar, except that the cells were collected onto glass microfiber filter disks (Whatman) placed on a vacuum apparatus, and the collected cells and disks were immersed in ice-cold methanol (2 ml), containing 125 l of 0.1% bromcresol green. Samples were stored in methanol at Ϫ20°C. Q 4 was added as an internal standard as described above, and samples were kept on ice during the extraction. Reextraction with petroleum ether (3 ml) was repeated two times. For all quantitative analyses, the standard curve was prepared and analyzed along side the samples, with freshly prepared internal standard and analytes. Routinely, the low end of the standard curve was monitored for recovery of the analyte by comparison of these concentrations a non-extracted standard curve. To ensure that analytes in low concentration gave a reproducible response, a signal to noise threshold ratio was set at 5. If measured analytes did not meet these criteria, or if the standard integrations did not identify the peak of interest accurately, the data were not used.

RP-HPLC and Detection by Scintillation
Counting-Detection of radioactive coenzyme Q and intermediates used a 〉-ram model 2 (IN/US Systems, Inc., Tampa, FL) with a 500-l flow cell. The column eluate was combined with Safety Solve (Research Products International Corp., Mount Prospect, IL) at a 2:1 ratio with a dwell setting of 1. Data from the device were collected into the Chem Station software supplied with the Agilent 1090 HPLC system. Simultaneous UV data were collected from the intrinsic diode array detector, 274 Ϯ 4 nm, and 250 Ϯ 4 nm. The system contained a Peltier cooled sample chamber maintained at 4°C and a column oven set to 40°C. A binary HPLC solvent delivery system was used with a phenyl-hexyl column (Luna 5, 100 ϫ 4.60 mm, 5 m, Phenomenex). The mobile phase consisted of Solvent A (methanol:isopropanol, 95:5, with 2.5 mM ammonium formate) to Solvent B (isopropanol, 2.5 mM ammonium formate) beginning at 100% Solvent A and over 8 min decreasing to 95%. The total run time was 11 min. Sample volumes for injection ranged from 40 to 70 l, and the software was set to respond at 1 mV ϭ 1 cpm.
RP-HPLC with Detection by MS-A 4000 QTRAP linear MS/MS spectrometer from Applied Biosystems (Foster City, CA) was used in combination with an Agilent Technologies 1200 HPLC system consisting of a PAL autosampler with thermostatted tray holders and Stack (LEAP Technologies, Carrboro, NC). Nitrogen was used for all gases in the mass spectrometer; the nitrogen gas was provided as boil-off from a bulk liquid nitrogen storage tank. Applied Biosystems software, Analyst version 1.4.2, was used for data acquisition and processing. Infusion experiments for tuning and optimization were performed with a model 11 plus syringe infusion pump (Harvard Apparatus, Inc., South Natick, MA). RP-HPLC separation was performed as described above. The 4000QT spectrometer was operated in Turbo electrospray positive mode. Q1 and Q3 were operated in unit resolution. For multiple reaction monitoring detection, the precursor-to-product ion transitions in multiple reaction monitoring mode were used to quantify Q and intermediates (m/z): 591.4/197 (Q 6 ), 455.2/197 (Q 4 ), 561.4/ 167.0 (DMQ 6 ), and 546.4/150 (prenyl-pABA). Optimum positive turbospray conditions for coenzyme Q compounds: nebulizer gas, 50 psig; turbo gas, 60 psig; curtain gas, 20 psig; collision gas set to "medium"; nebulizer current, 20; and temperature, 450°C. Optimal settings for compound-dependent parameters are in volts, and dwell is in milliseconds (data are declustering potential, entrance potential, collision energy, collision cell exit potential, and dwell): Q 4 (75, 10, 29, 12, and 125), Q 6 (111, 10, 37, 10, and 125), DMQ 6 (96, 9, 37, 11, and 125), and prenyl-pABA (96, 9, 33, 11, and 125). The same settings were used for 13 C-labeled forms. Settings as described are theoretical and based on differences required for the analyses of farnesylated standards (20), with a compensation for increased isoprene length. For ion trap detection, similar gases as above were used for Q and Q intermediates: nebulizer gas, 45 psig; turbo gas, 55 psig; curtain gas, 25 psig; collision gas, high; nebulizer current, 35; and temperature, 450°C. The enhanced product ion scan used default dynamic settings for trap filling and other parameters. The mass spectrometer detection conditions also included an enhanced resolution scan with standard parameters, between m/z 520 and m/z 620. The injection volume was 10 or 20 l. Stock solutions of the Q 4 and Q 6 (Sigma) were prepared in hexanes and stored under argon gas at Ϫ20°C. Aliquots added to ethanol and the concentrations were then determined spectrophotometrically with a molar extinction coefficient of E ϭ 14,900 at 275 nm (21). Integration of peak areas was performed with Analyst software, with a bunching factor of 1 and 3 smoothing events. Area ratios were constructed in Microsoft Excel for the calibration curve and experimental samples. The slope was calculated with a linear curve forced through zero. Standard deviations represent duplicate/ triplicate samples, independently extracted with duplicate/ triplicate injections (n ϭ 4 -6).

RESULTS
Prenyl-pABA Is a Naturally Occurring Lipid Component of Yeast Cells-Neutral lipid extracts prepared from wild-type yeast cells cultured in standard rich media contain a lipid that we have identified as prenyl-pABA. The identification of prenyl-pABA was based on the presence of an HPLC peak with an elution similar to that for HHB, a previously characterized yeast Q intermediate (22). The precursor ion [MϩH] ϩ of 546, and predominant tropylium [m/z ϭ 150] and chromenylium [m/z ϭ 190] product ions, detected in ion-trap analyses, were consistent with a ring amino replacing the ring hydroxyl present in HHB ( Fig. 2A). The tropylium-like ion is a transition ion generated from prenylated aromatic and benzoquinone rings and is formed under dissociation conditions by incorporation of a methylene remnant (produced by fragmentation of the prenyl tail after the first carbon) to form a 7-membered ring (23). The chromenylium-like ion is larger in mass by ϩ40 (C 3 H 4 ) under these electrospray ionization conditions and is derived by fragmentation and cyclization to include the first four prenyl tail carbons (23).
To confirm the identity of prenyl-pABA, wild-type yeast (W303-1A) were pre-cultured in Dogal media (minus pABA and folate) to deplete cellular stores of pABA (16). Either dextrose or galactose can be used as a fermentable carbon source in minimal media; galactose is used because it is non-repressing to aerobic respiration (24). Yeast cells were transferred to fresh Dogal medium plus folate, and then cultured in the presence of [ 13 C 6 ]pABA as described under "Experimental Procedures." Product ion analyses from the crude lipids of yeast grown in the presence of [ 13 C 6 ]pABA show that yeast incorporate the ring carbons into prenyl-pABA to generate prenyl-[ 13 C 6 ]pABA ( 13 C 6 12 C 31 H 55 NO 2 ) (Fig. 2B). Although prenyl-pABA is readily detectable in lipid extracts of wild-type yeast (W303-1A harvested at 1.8 A in YPGal media contain 89.0 Ϯ 5.6 fmol/A), it is much less abundant than Q 6 (84.6 Ϯ 4.3 pmol) under standard log phase growth conditions in rich media.
Yeast coq3 Mutants Cultured with 13 C 6 -4HB Produce 13 C 6labeled HHB-Previous work has shown that yeast coq mutants grown in the presence of 4HB produce HHB (25). However, this intermediate is unstable and difficult to detect without derivatization (22). For purposes of comparison, we wished to generate both the normal and 13 C-labeled form of HHB. To do this, a coq3 yeast mutant was cultured in the presence of [ 13 C 6 ]4HB. The E2-249 coq3 mutant is Q-deficient and is a member of the G31 complementation group defined by Dieckmann and Tzagoloff (26,27). As shown in Fig. 3 both the normal isotopic form of HHB and [ 13 C 6 ]HHB are detected in lipid extracts prepared from the coq3 mutant. Our product ion spectra match that described previously (28); of particular note is the fragmentation pattern showing the shift in mass from the precursor molecule ( Fig. 3A) with an analogous series of fragments from the same compound in the separated crude lipids of [ 13 C 6 ]4HBlabeled coq3 point mutant (Fig. 3B). Both HHB compounds elute with exactly the same retention time. These results demonstrate that the precursor and product ions of prenyl-pABA are each one mass unit less than for HHB, the pABA Is a Coenzyme Q Biosynthetic Precursor in Yeast adjusted to 50 Ci/mmol; 800 nM final concentration). Cells were incubated 24 h, lipid extracts were prepared and subjected to RP-HPLC, and the radioactivity was detected as described under "Experimental Procedures." Yeast lipid extracts prepared from wild-type cells contained two peaks of 14 C-radiolabeled material, consistent with the elution of Q 6 and Q 6 H 2 (the hydroquinone) (Fig. 4). Similarly, the yeast coq7-1 mutant NM101, when incubated with either [ 14 C]4HB or [ 14 C]pABA, produced radiolabeled material slightly more polar than Q 6 , previously identified as DMQ 6 (29). The synthesis of [ 14 C]Q 6 under both experimental conditions requires the yeast COQ2 gene product (Fig. 4). These results indicate that DMQ 6 and Q 6 may be synthesized in yeast from either pABA or 4HB aromatic ring precursors, and both ring precursors require Coq2p for prenyl tail attachment.
Yeast Cultured with [ 13 C 6 ]pABA produce 13 C 6 -labeled Prenyl-pABA, DMQ 6 , and Q 6 -The results obtained with the 14 C-labeled precursors strongly suggest that pABA functions as a ring precursor in yeast Q 6 biosynthesis. Metabolic labeling studies with stable isotopes provide a definitive test, because the ring carbons can be detected in both precursor and product ions by mass spectrometry. Wild-type yeast were first depleted of pABA and folate as described above and then cultured in the presence of [ 13 C 6 ]pABA as described under "Experimental Procedures." The identification of hexaprenylated compounds in the crude lipids of wild-type yeast is represented by the spectra in Fig.  2. The normal (unlabeled) precursor ions monitored included: prenyl-pABA ( Fig. 2A), demethoxy-Q 6 (Fig. 2C), and Q 6 (Fig.  2E). Product ion analyses from the crude lipids of yeast grown in the presence of [ 13 C 6 ]pABA show that the ring carbons of this compound are incorporated into demethoxy-Q 6 and Q 6 and alter the average isotopic masses of the tro-FIGURE 2. Detection of prenyl-pABA, DMQ 6 , and Q 6 in lipid extracts of wild-type yeast cultured in the absence or presence of 13 C 6 -pABA. Wild-type yeast (W303-1A) were pre-cultured in Dogal media (minus pABA and folate) as described under "Experimental Procedures." Cells were harvested during mid-log from large cultures (300 -600 ml of media; A 600 nm ϭ 0.6), resuspended in 2 ml of fresh media or fresh media containing 2-10 g/ml [ 13 C 6 ]pABA (final concentration), incubated at 30°C and harvested after 16 h. Lipid extracts were prepared and quinones and prenylated intermediates were subjected to RP-HPLC-ESI-MS/MS as described. . Q 6 (Sigma) and the lipid extract prepared from NM101 (coq7-1) yeast strains were used to establish the retention times of Q 6 and DMQ 6 , respectively (29). A, C, and E show the structures of the compounds detected and indicate the identity of the predicted tropylium-like ion of the base peak (tallest peak) formed under collision associated dissociative conditions. SEPTEMBER 3, 2010 • VOLUME 285 • NUMBER 36 pylium-like ion and molecular ion by 6 mass units. This results in the following 13 C 6 -labeled compounds in Fig. 2: prenyl-[ 13 C 6 ]pABA ( 13 C 6 12 C 31 H 55 NO 2 ) (Fig. 2B); 13 C 6 -demethoxy-Q 6 ( 13 C 6 12 C 32 H 56 O 3 ) (Fig. 2D); and [ 13 C 6 ]Q 6 ( 13 C 6 12 C 33 H 58 O 4 ) (Fig. 2F). The incorporation of [ 13 C 6 ]pABA into the penultimate Q intermediate, DMQ 6 , was readily detected in wild-type yeast cell lipid extracts (Fig.  2D). Wild-type yeast cultured in the presence of [ 13 C 6 ]4HB also generate the expected ϩ6 isotopically labeled forms of Q 6 and DMQ 6 , however, under these labeling conditions, it was difficult to detect the ϩ6 form of HHB (data not shown).

Prenyl-pABA Is Absent in Yeast abz1 Mutants Starved for pABA, Yet Content of Q Is Similar to That of
Wild-type Yeast-Yeast produce pABA from chorismate by a twostep process that requires the ABZ1 and ABZ2 gene products (14,16). Thus, it seemed likely that the production of prenyl-pABA would depend on the supply of pABA from this biosynthetic pathway, or from the pABA supplied in the media. To test this idea, wild-type yeast, or abz1 yeast mutants were serially cultured under conditions where the exogenous supply of pABA was eliminated (16). As shown in Fig. 5, abz1 mutants starved for pABA show dramatic decline in the content of prenyl-pABA, whereas the content of prenyl-pABA in wildtype cells remains unchanged or increased. These results indicate that a combination of the abz1 deletion and nutritional depletion of pABA results in the depletion of prenyl-pABA. Under these pABAdepleted conditions, the abz1 mutants are still able to produce DMQ 6 and Q 6 . This is consistent with the presence of at least two pathways in yeast able to supply aromatic ring precursors for Q biosynthesis (Fig. 1).
What Are the Relative Contributions of the pABA and 4HB Ring Precursors to Q Biosynthesis?-Once we recognized that [ 13 C 6 ]pABA could serve as a precursor to [ 13 C 6 ]Q 6 , we decided to investigate whether cells demonstrated a preference for pABA or 4HB as an aromatic ring precursor for Q. As expected, incubation of wild-type or abz1 mutant cells labeled for 3.5 h with a single designated precursor show that either [ 13 C 6 ]pABA or [ 13 C 6 ]4HB serve as ring precursors in biosynthesis of DMQ 6 (Fig. 6A) and Q 6 (Fig. 6B); the darkly shaded lower section of each column designates the amount of the [ 12 C]quinone, and the upper light gray bar shows the amount of [ 13 C 6 ]quinone. We then performed competition experiments to examine the ability of the normal isotopic 12 C form of the alternative precursor to diminish the incorporation of the 13 C-form into [ 13 C 6 ]DMQ 6 or [ 13 C 6 ]Q 6 (Fig. 6). For these competitor experi- The yeast coq3 mutant E2-249 was pre-cultured in Dogal media as described previously, harvested, and incubated in fresh media or fresh media with 10 g/ml [ 13 C 6 ]4HB as described in Fig. 2 ments, cells were incubated (3.5 h) in the presence of equivalent concentrations of the alternative ring precursor (10 g/ml). In wild-type cells, the presence of the competitor did not substan-tially affect the amount of [ 13 C 6 ]Q 6 content. Indeed, total levels of Q 6 were slightly higher in wild-type cells incubated with both aromatic ring precursors concurrently. In abz1 mutant cells, the total content of Q 6 was increased (relative to the no-addition control) by addition of any combination of aromatic ring precursors, and the presence of added 12 C competitor significantly decreased the amount of [ 13 C 6 ]Q 6 produced from the 13 C-aromatic ring precursor. It seems likely that the presence of both ring precursors better satisfies the requirement for both folate and Q.
Although the different ring precursors had only modest effects on the amount of Q 6 formed, they had dramatic effects on the content of DMQ 6 . For example, both wild-type yeast and abz1 mutants had increased content of DMQ 6 when incubated with pABA as compared with 4HB. In fact abz1 mutants incubated with [ 13 C 6 ]pABA produced a high content of DMQ 6 of which almost all is [ 13 C 6 ]DMQ 6 (Fig. 6A). We speculate that prenyl-pABA, or more likely a subsequent intermediate derived from prenyl-pABA, might act to inhibit the hydroxylation of DMQ.
Prenylation of pABA Precedes Biosynthesis of DMQ and Q from pABA-A pulse-labeling experiment was conducted to determine whether the incorporation of 13 C 6 -ring carbons into 13 C 6 -prenyl-pABA preceded the formation of [ 13 C 6 ]DMQ 6 and [ 13 C 6 ]Q 6 . Wild-type cells and abz1 yeast mutants were precultured as described in Fig. 6 and incubated with [ 13 C 6 ]pABA over a time course from 0 to 30 min, as described in Fig. 7 and "Experimental Procedures." Samples were removed from the incubation at stated times, collected by filtration, and quenched, and lipid extracts were analyzed to determine the amounts of prenyl-pABA, DMQ 6 , and Q 6 by HPLC-MS/MS and multiple reaction monitoring. In both wild-type and abz1 mutant yeast, prenyl-[ 13 C 6 ]pABA is detectable within a minute of label addition (see inset, Fig. 7), and its de novo synthesis precedes that of [ 13 C 6 ]DMQ 6 and [ 13 C 6 ]Q 6 , consistent with the notion of a precursor-product relationship.
pABA-replete Yeast Produce 4-Imino-DMQ 6 -In the crude lipid extracts of wild-type cells grown in media supplemented with [ 13 C 6 ]pABA, we identify what appears to be a nitrogen containing form of 4-imino-DMQ 6 (Fig. 8). The tropylium, chromenylium, and molecular ion are all shifted in accordance with Mϩ6 (m/z), as is detected for DMQ 6 (Fig. 2D);  . Nutritional and genetic depletion of pABA eliminate the formation of prenyl-pABA. BY4741 wild-type yeast and BY4741⌬abz1 yeast mutants were grown overnight in YPD, and then diluted 1:100 (v/v) into fresh Dogal media minus pABA and folate and incubated for 1 day. The day 1 culture was used to inoculate fresh Dogal media (minus pABA and folate), and the process was repeated to generate the day2 and day 3 cultures. The serial dilution into pABA minus media exhausts endogenous stores of pABA (16). Cells were harvested, and lipid extracts were examined for content of Q 6 (filled squares; left y axis), DMQ 6 (open squares; right y axis), and prenyl-pABA (triangles; right y axis). SEPTEMBER 3, 2010 • VOLUME 285 • NUMBER 36 however, the masses of these ions are reduced by 1 Da relative to the fragment ion masses of DMQ 6 , consistent with an imino nitrogen rather than a quinone oxygen atom in the para ring position. The chromenylium ion fragment of 4-imino-DMQ 6 is abundant, because it is likely to be more stable to dissociative conditions due to the interaction of the first four carbons of the prenyl tail with the imino nitrogen forming a second ring (23). The normal isotopic form of 4-imino-DMQ 6 is present in crude lipid extracts of strains grown in YPD (data not shown).

pABA Is a Coenzyme Q Biosynthetic Precursor in Yeast
The detection of 4-imino-DMQ 6 led us to test whether it might function as an intermediate in Q 6 biosynthesis. To simplify the precursor-product relationship between 4-imino-DMQ 6 and DMQ 6 , we examined the synthesis of both as a function of time in NM101 yeast, a strain known to be blocked in Q biosynthesis and to produce DMQ 6 due to a mutation in COQ7 (29). The pulse labeling with [ 13 C 6 ]pABA was performed over a time course of 0 to 60 min, and the content of 13 C-labeled intermediates (prenyl-[ 13 C 6 ]pABA, [ 13 C 6 ]imino-DMQ 6 , and [ 13 C 6 ]DMQ 6 ) were determined (Fig. 9). The rate of formation of prenyl-[ 13 C 6 ]pABA and [ 13 C 6 ]imino-DMQ 6 were very similar, suggesting that both serve as biosynthetic precursors in the synthesis of DMQ 6 . We note that DMQ 6 is relatively unstable in the coq7-1 mutant, as is evident from the rate of loss of the [ 12 C]DMQ 6 . This rate of loss seems to be compensated by the formation of [ 13 C 6 ]DMQ 6 . These results suggest that the nitrogen substituent of pABA may be retained up to the step generating 4-imino-DMQ 6 , and are consistent with a possible mechanism where oxygen from a water-based hydroxyl could replace the nitrogen imino via Schiff base chemistry (Fig. 10). FIGURE 6. 4HB and pABA are mutually incorporated into Q 6 , but differentially incorporated into DMQ 6 . Wild-type (W303-1A) or abz1 null (W303⌬ABZ1) cells were serially cultured to deplete them of pABA and folinic acid as described. The depleted cells were harvested and used to inoculate a 4-ml culture. This was divided to supply duplicate samples of cells to five different media conditions. The media contained either no addition or 10 g/ml of the following: [ 13 C 6 ]pABA, [ 13 C 6 ]4HB, or similar cultures in the presence of equivalent concentrations of the alternative unlabeled ( 12 C) ring precursor (10 g/ml). Independent replicates were made of each condition. The amount of the competing 12 C form of precursor (ratio of 13 C: 12 C) was 1:1. After incubation for 3.5 h, cells were harvested, and lipid extracts were subjected to analyses by RP-HPLC-ESI/MS-MS. A, DMQ 6 content; B, Q 6 content. The content of [ 12 C]DMQ 6 or [ 12 C]Q 6 is designated by the darkly shaded lower section of each bar (in A and B, respectively), and [ 13 C 6 ]DMQ 6 and [ 13 C 6 ]Q 6 by the upper light gray bar (in A and B, respectively). Each condition was cultured as an independent replicate, and each lipid extract was injected twice (n ϭ 4). This experiment is representative of three others, conducted independently. The brackets with letters denote the values being compared: Content of   1A) or B, abz1 null (W303⌬ABZ1) cells were pre-cultured as described to deplete endogenous pABA, resuspended in pre-warmed Dogal media (30°C) to a total volume of 16 ml in a 125-ml flask, and incubated with shaking (250 rpm, 30°C). Prior to addition of ring precursors, two (1 ml) aliquots were removed, to represent a "no-label or zero time" control. [ 13 C 6 ]pABA was added (final concentration, 10 g/ml), and further 1-ml aliquots were collected in duplicate at the time points indicated on the glass filter disks for lipid extraction: no label control (defines the "zero" time point), 2, 5, 10, 15, or 30 min. Lipid extracts were subjected to analyses by RP-HPLC-ESI/MS-MS as described, and the amount of de novo prenyl-[ 13 C 6 ]pABA (triangles), [ 13 C 6 ]DMQ 6 (squares), and [ 13 C 6 ]Q 6 (diamonds) is depicted. The left y axis represents the picomoles/A (for plots of [ 13 C 6 ]DMQ 6 and [ 13 C 6 ]Q 6 ), whereas the right y axis shows the content in picomoles/A for prenyl-[ 13 C 6 ]pABA. Values are plotted as average Ϯ S.D. (n ϭ 4, wild type; n ϭ 6 abz1 null). The insets in each panel depict early time points (in minutes) for all three components plotted on the same scale (pmol/A).

DISCUSSION
Schemes of Q biosynthesis in E. coli, yeast, and animals universally depict 4HB as the aromatic ring precursor. In each of these species, isoprenylation of 4HB is thought to represent a committed step in Q biosynthesis. Here we show that S. cerevi-siae can also utilize pABA as a ring precursor in Q biosynthesis. This is a surprising finding, because pABA is a crucial intermediate in folate biosynthesis. It is also surprising because the addition of pABA to either E. coli or human cells causes a concentration-dependent inhibition of Q biosynthesis (30 -33). In E. coli, rat, and human cells, the pABA ring competes with 4HB at the ring:polyprenyltransferase step (catalyzed by Coq2), and the product prenyl-pABA appears to be a dead-end product. Recently another aromatic ring inhibitor, 4-nitrobenzoic acid, was shown to inhibit Q biosynthesis in mammalian cells through its competition with 4HB for Coq2 (34). Thus it appears that several benzoic acid ring analogs function as competitive inhibitors of Q biosynthesis in mammalian cells (33,34).
In contrast, our studies identify prenyl-pABA, a normal metabolite present in lipid extracts of wild-type yeast, as a Q intermediate. The synthesis of prenyl-pABA depends on Coq2 (Fig. 4), and we show that prenyl-pABA is a normal metabolite present in lipid extracts of wild-type yeast cultured in standard yeast media (Fig. 2). An interesting independent confirmation of prenyl-pABA in yeast neutral lipid extracts was recently published, found in lipid extracts of yeast with defects in ferredoxin (YAH1) and ferredoxin reductase (ARH1) (18). These authors discovered the role of pABA in Q biosynthesis through their analyses of iron chaperones required for the activity of Coq7, the Coq diiron enzyme required for the last ring hydroxylation in Q biosynthesis (29,35).
Our work clearly explores the biochemical relationship of prenyl-pABA in yeast Q biosynthesis, demonstrating prenyl-pABA is a bona fide biosynthetic precursor. pABA can be prenylated immediately following its addition to cells. Pulse label- . This compound can be detected in wild-type strains grown in rich media and appears to be a normally accumulating form of demethoxy-Q 6 (see Fig. 9). [ 13 C 6 ]imino-DMQ 6   . Time course assays with 13 C 6 -pABA examine the formation novel Q intermediates in NM101 (coq7-1) yeast. NM101 yeast cells were pre-cultured and analyzed by pulse-labeling with [ 13 C 6 ]pABA as described in Fig. 7. Lipid extracts were subjected to analyses by RP-HPLC-ESI/MS-MS as described, and the amount of de novo prenyl-[ 13 C 6 ]pABA (green squares), [ 13 C 6 ]DMQ 6 (red circles), and [ 13 C 6 ]imino-DMQ 6 (purple cross) is depicted. Also depicted is the decay of non-isotopically labeled DMQ 6 (blue open circles). The right y axis represents the pmol/A (for plots of prenyl-[ 13 C 6 ]pABA and [ 13 C 6 ]imino-DMQ 6 ), whereas the left y axis shows the content in pmol/A for DMQ 6 and [ 13 C 6 ]DMQ 6 . Values are plotted as average Ϯ standard deviations (n ϭ 4). SEPTEMBER 3, 2010 • VOLUME 285 • NUMBER 36

pABA Is a Coenzyme Q Biosynthetic Precursor in Yeast
ing studies with [ 13 C 6 ]pABA show that pABA-deficient abz1 null yeast mutants synthesize prenyl-[ 13 C 6 ]pABA within 1 min of incubation (Fig. 7). This rapid incorporation of the isotope into prenyl-pABA precedes the formation of [ 13 C 6 ]DMQ 6 and [ 13 C 6 ]Q 6 and is consistent with a precursor-product relationship. Yeast abz1 mutants, known to be defective in pABA biosynthesis (15,16), become depleted in prenyl-pABA when cultured in pABA-free media (Fig. 5). However, under these culture conditions Q is still produced, reflecting use of an alternate ring precursor, 4HB (Fig. 1). Prenylation of 4HB by Coq2 is the "classic" arm of the pathway. We also identify for the first time the underivatized forms of HHB, detected in a coq3 yeast mutant.
Identification of 4-Imino-DMQ 6 and a Model for Nitrogen Loss-Both wild-type and abz1 yeast null mutants when precultured under conditions to deplete pABA, followed by growth with pABA supplementation, have a higher DMQ 6 content as compared with wild-type yeast or abz1 mutants supplemented with 4HB (Fig. 6). In isotopic labeling studies with the abz1 mutant Ͼ90% of the DMQ 6 detected contains the 13 C 6 -aromatic ring from pABA. The preferential incorporation of pABA into DMQ 6 is particularly intriguing in light of our finding that 4-imino-DMQ 6 , a lipid derived from prenyl-pABA, is present in the pABA-supplemented abz1 mutant and in normally cultured wild-type cells (Fig. 8). The rate of formation of 4-imino-DMQ 6 is similar to that of prenyl-pABA (Fig. 9), identifying it as a new Q intermediate and indicating that the nitrogen of pABA can potentially be retained up to the step preceding DMQ 6 .
If this is the case, what is the fate of 4-imino-DMQ 6 ? We speculate 4-imino-DMQ 6 is produced from the two-electron oxidation of 4-amino-DMQ 6 H 2 (Fig. 10). Once the imino-quinone is formed, the potential for loss of the ring nitrogen as ammonia and its replacement by oxygen from a water or hydroxide ion to form the quinone could reasonably occur by Schiff base chemistry. Although we designate the ring nitrogen loss step as immediately preceding DMQ 6 , it is possible the loss could occur earlier. However, we favor the depiction in Fig. 10, because DMQ 6 is a relatively abundant Q intermediate, detected even in wild-type cells (36,37), and is also a component of a Q biosynthetic complex (38). It seems possible that the enhanced accumulation of DMQ 6 in cells first starved for and then supplemented with pABA may be due to inhibition of the Coq7 monooxygenase step (Fig.  10). However, the effect may also be indirect, as the influx of pABA would also replete folate synthesis. We note that the redox state of mitochondria might very well affect the relative rates of oxidation and reduction of amino-DMQ 6 H 2 to imino-DMQ 6 , and DMQ 6 to DMQ 6 H 2 . Which Ring Precursor, 4HB or pABA, Is Normally Preferred?-Our labeling with 13 C-aromatic ring precursors is accomplished by lowering the supply of pABA and folate to cells with deficient media prior to isotopic incorporation. Thus, at present we cannot differentiate the relative importance of each ring precursor (4HB or pABA) for Q 6 production in wild-type cells, under physiological conditions. It remains possible that pABA could be converted to 4HB prior to lipidation in mitochondria. However, the competition experiments (Fig. 6) suggest that, when one 13 C-ring precursor is provided to yeast cells in a one to one molar ratio with the other unlabeled ring, pABA and 4HB are indistinguishable for [ 13 C]Q synthesis. In addition, if [ 13 C 6 ]pABA was converted into [ 13 C 6 ]4HB prior to its prenylation by mitochondrial Coq2, then the enhanced accumulation of [ 13 C 6 ]DMQ 6 from the [ 13 C 6 ]pABA labeling relative to that of [ 13 C 6 ]4HB (Fig. 6A) would be unlikely. Although the abz1 null avidly incorporates and synthesizes [ 13 C]demethoxy from [ 13 C]pABA, our data do not suggest that pABA is a better source for demethoxy Q 6 in normal yeast. Media conditions, for example carbon source and nitrogen source(s), may radically alter ring precursor preference. Finally, we note that our conclusions regarding the incorporation of 4HB and pABA into Q 6 are valid only if the uptake of 4HB and pABA are identical.
It is likely but uncertain that 4HB and pABA may share some mechanisms of uptake and transport of the free form into mitochondria. pABA and 4HB are weakly ionic compounds (pABA: pK a 4.9; 4HB: pK a 4.67 (39)), and their uptake and retention has a pH-dependent component. Uptake is favored at low pH, and the formation of the carboxylate anion once imported into the cell favors retention. Although the pH of our media is 6.0, which has been shown to relieve pABA growth inhibition in S. cerevisiae (40), other work has shown pABA uptake cannot be saturated (41). The studies of inhibition of yeast growth by pABA are intriguing (40), and are different from E. coli growth inhibi- FIGURE 10. Scheme for generation of 4-imino-DMQ 6 and loss of nitrogen generating DMQ 6 . The pulselabeling studies in Figs. 7 and 9 suggest that prenyl-pABA and 4-imino-DMQ 6 are bona fide precursors in yeast Q synthesis from pABA. A possible mechanism for replacement of the imino nitrogen with a hydroxy substituent via Schiff base chemistry is proposed. The new steps of oxidation, de-imination, and reduction allow the two pathways to converge at formation of DMQ 6 H 2 .
pABA Is a Coenzyme Q Biosynthetic Precursor in Yeast tion by pABA. E. coli have ubiC encoded chorismate lyase, which directly converts chorismate to 4HB (9) (Fig. 1), whereas yeast lack this homolog. E. coli can be growth inhibited by excess pABA, and relief is accomplished by large concentrations of 4HB (31). However, rescue of growth inhibition by pABA in S. cerevisiae requires aromatic amino acids, with phenylalanine and tyrosine in combination to best resume growth (40).
What Are the Pathway(s) to 4HB Production?-Radiolabeled cinnamic and coumaroyl acids produce radiolabeled Q when fed to bakers' yeast (42), with coumarate being the best precursor, and both these compounds are shown to precede 4HB. Other work describes the ability of radiolabeled aromatic amino acids to donate their carbons to Q in yeast (43), however the intermediary compounds have not been described. In other microorganisms the direct precursors for 4HB have been examined more recently and thoroughly. An alternative is described in plants (44), where phenylalanine is a probable precursor, although S. cerevisiae lacks an identifiable phenylalanine ammonia lyase homolog. The Acinetobactor baylyi system describes the production of 4HB precursors from the catabolism of primarily plant cell wall components into hydroxycinnamate precursors (13).
Do Anti-folates Target Both Folate and Q?-Studies of chorismate synthesis have been stimulated by investigations of drug resistance in microorganisms and pABA metabolism. Depletion of folate is difficult, because it is recycled as a cofactor and yeast carry reserves of both folate and pABA. Yeast cells must be serially cultured in media depleted of these nutrients to elicit a pABA or folate growth deficiency (16). A pABA deficiency is also accomplished by inclusion of sulfanilamide in the media, because sulfanilamide acts as a competitive pABA analog (14). The discovery of sulfanilamide antibiotics hinges on the ability of this class of compound to interfere with the condensation of pABA to pteroglutamoyl for the synthesis of folate (19). The enzyme dihydropteroate synthase is the target of the pABA analog sulfamethoxazole, within the "sulfa" drug class. Previous work by Macreadie's group on resistance to this drug showed resistance depended on expression of the dihydropteroate synthase homolog in yeast (FOL1), as long as pABA was supplied in the media (45).
This report describes a novel and physiologically relevant, lipidated form of pABA in yeast, establishing a possible linkage between Q biosynthesis and folate metabolism. Although direct relationships between coenzyme Q biosynthesis and folate metabolism have not been characterized until now, relationships are known to exist between folate synthesis and sulfa drug resistance. A storage form of pABA in microorganisms satisfies the logistics for folate synthesis, just as a glucosylated form of pABA functions in plants (41,45,46). It would be undesirable for a crucial metabolic intermediate such as pABA to fall into low supply, but large quantities of the free acid may act as uncouplers to the electrochemical gradient (47). Larger amounts of pABA may also act as a substrate inhibitor of dihydropteroate synthase, the enzyme that couples pABA to a pterin moiety during folate synthesis. Although inhibition occurs at micromolar levels of pABA for the bacterial dihydrop-teroate synthase homolog (48), its inhibition by high amounts is not complete and it retains much lower but steady activity.
In summary, our analyses document pABA and 4HB as two aromatic rings that serve as precursors for DMQ 6 and Q 6 . The observation of a normally produced imino form of DMQ found in the crude lipid extract of wild-type cells (Fig. 8), and its similar rate of formation as compared with prenyl-pABA in a coq7-1 point mutant (Fig. 9), suggest that both are novel Q intermediates. Based on the 4-imino-DMQ 6 intermediate, we suggest a mechanism for the loss of the pABA-derived nitrogen. Finally the findings presented here suggest an intimate relationship may exist between synthesis of folic acid, necessary for many cellular essential functions and cellular respiration in S. cerevisiae, coordinated through the overlapping substrate prenyl-pABA.