CLD1 Reverses the Ubiquinone Insufficiency of Mutant cat5/coq7 in a Saccharomyces cerevisiae Model System

Ubiquinone (Qn) functions as a mobile electron carrier in mitochondria. In humans, Q biosynthetic pathway mutations lead to Q10 deficiency, a life threatening disorder. We have used a Saccharomyces cerevisiae model of Q6 deficiency to screen for new modulators of ubiquinone biosynthesis. We generated several hypomorphic alleles of coq7/cat5 (clk-1 in Caenorhabditis elegans) encoding the penultimate enzyme in Q biosynthesis which converts 5-demethoxy Q6 (DMQ6) to 5-demethyl Q6, and screened for genes that, when overexpressed, suppressed their inability to grow on non-fermentable ethanol—implying recovery of lost mitochondrial function. Through this approach we identified Cardiolipin-specific Deacylase 1 (CLD1), a gene encoding a phospholipase A2 required for cardiolipin acyl remodeling. Interestingly, not all coq7 mutants were suppressed by Cld1p overexpression, and molecular modeling of the mutant Coq7p proteins that were suppressed showed they all contained disruptions in a hydrophobic α-helix that is predicted to mediate membrane-binding. CLD1 overexpression in the suppressible coq7 mutants restored the ratio of DMQ6 to Q6 toward wild type levels, suggesting recovery of lost Coq7p function. Identification of a spontaneous Cld1p loss-of-function mutation illustrated that Cld1p activity was required for coq7 suppression. This observation was further supported by HPLC-ESI-MS/MS profiling of monolysocardiolipin, the product of Cld1p. In summary, our results present a novel example of a lipid remodeling enzyme reversing a mitochondrial ubiquinone insufficiency by facilitating recovery of hypomorphic enzymatic function.

Introduction from their E. coli food supply. Nonetheless, mutant clk-1 worms cultured for several generations on a bacterial food source that is unable to manufacture Q 8 (GD1 E. coli) remain phenotypically long-lived [27]. Under these conditions, otherwise wild type worms also display lifeextension. Expression profiling reveals that these animals elicit a transcriptional response similar to the retrograde response activated following mitochondrial ETC disruption in petite yeast, which are also long lived [28]. Moreover, clk-1 mutants reprogram their metabolism, similar to other long-lived mitochondrial electron transport chain mutants in C. elegans, including nuo-6(qm200) and isp-1(qm150)-which disrupt complex I and III, respectively [29,30]. At least part of the clk-1 longevity response is now known to be due to a nuclear-targeted form of CLK-1 that unexpectedly binds chromatin [31].
Given the essential nature of COQ7 in humans, its centrality to Q production in cells, and its unexpected role in lifespan control of C. elegans, we sought to identify new genetic loci that could suppress the disruption of this gene in S. cerevisiae. To this end we identify the cardiolipin remodeling enzyme CardioLipin-specific Deacylase 1 (Cld1p) as a novel modulator of Coq7p activity.

Results
In an effort to identify genes that are able to overcome the essential requirement of Coq7p in yeast cultured on a non-fermentable carbon source, we transformed a coq7 null mutant (Δcoq7) with a library of genomic DNA fragments isolated from wild type yeast and contained on a high copy number vector. Transformants were directly selected for growth on ethanol (YEPE 3% ). We obtained no suppressors, which was unexpected since PCR confirmed that our library contained multiple copies of the wild type COQ7 locus. If Δcoq7 cells containing a bona fide copy of COQ7 were allowed to exhaust their supply of glucose prior to selection on ethanol, however, then could growth be rescued. This observation suggests that initiation of the diauxic shift in yeast is required to de-repress a genetic network permissive for COQ7 expression and activation. This idea is consistent with previous reports showing (i) Coq7p is dephosphorylated upon entry into the diauxic shift [18,32], (ii) Q 6 biosynthesis proceeds only to the level of DMQ 6 prior to the diauxic shift [14], (iii) DMQ 6 does not support respiration in yeast [33], and (iv) Coq7p is part of a ubiquinone-biosynthesis megacomplex, the stability of which is dependent upon full length Coq7p [14,15]. We reasoned that many prior studies that searched for coq7 loss-of-function suppressor genes using a similar library-screening approach may have missed targets either because cultures had not been at the diauxic shift prior to transformation and selection, or because yeast cells have an absolute requirement for small amounts of Coq7p, or both. In an effort to identify novel coq7 suppressors, and to by-pass both of these potential limitations, we first generated a novel coq7 allelic series, comprised of multiple hypomorphic mutants, and then used it to undertake high-copy genomic DNA suppressor screens. We also made use of selection media comprised of non-fermentable ethanol supplemented with a small amount of dextrose (YEPE 3% + 0.1% DEX. ) in order to facilitate isolation of potential suppressors; our rational being that small amounts of glucose would provide a smooth transition into the diauxic shift after transformation.

Generation of a coq7 allelic series
Mutagenic PCR was used to generate over seventy coq7 alleles that were hypomorphic for growth on YEPE 3% (Fig 1A and 1B). All mutant alleles generated using this approach carried a Q48R point mutation that was present in the parent allele before mutagenesis. coq7-2(Q48R) yeast grow indistinguishably from wild type SEY6210 yeast between the biologically relevant temperatures of 15-35°C. We have named this allele coq7-2 to distinguish it from the COQ7 Strategy for generating mutant coq7 alleles: coq7-19(Δcoq7) yeast were transformed with a mix of PCR mutagenized coq7 constructs and then hypomorphic alleles identified by slow growth on 3% ethanol (YEPE 3% ). (B) Mutagenized clones were patched in triplicate, then replica plated onto YEPE 3% , and monitored for growth at 15°C, 25°C and 30°C (top left panel shows patch numbering) relative to wild type SEY6210 yeast (patches #37-39). Alleles of interest are described in full in S1 Table and include: coq7-6 (patches # 13-15); coq7-11 (# 19-21); coq7-13 (# 25-27); coq7-5 (# 28-30); coq7-16 (# sequence of SEY6210. Alleles that were selected for further study are summarized in Table 1 and described in full in S1 Table. Most relevant are the single point mutations encoded by [31][32][33] and coq7-2(Q48R) (# [34][35][36]. (C) The coq7-11 allele is temperature-sensitive and inviable at 35°C when integrated back into the wild type COQ7 locus. Shown are four independent re-integrants. (D) Addition of a HA or myc epitope tag to the C-terminus of wild type Coq7p results in hypomorphic growth on YEPE 3% . Shown also is the effect of COQ7 copy number (CEN-low and 2μhigh) and promoter identity (ADH1 vs. native) on growth at 30°C YEPE 3% (10 days). P ADH1 -COQ7-HA(CEN) was used for the library screen described in Fig 2. coq7-2(Q48R) grows indistinguishably from wild type at 30°C. (E) Relative growth rate of mutant coq7 alleles versus wild type yeast (SEY6210) cultured at four different temperatures-15°C, 25°C, 30°C and 35°C (refer to Table 1 for allele identification, and S1 Table for raw data). Data is normalized to SEY6210 cell density at late log phase, for each respective temperature. Data for coq7-5, coq7-15 and coq7-16 at 35°C was not collected. (F) Location of relevant amino acid disruptions caused by various hypomorphic coq7 alleles (labeled). Changes affect highly conserved residues (red) and have been mapped onto a model of monomeric rat CLK-1/Coq7p [21]. Shown is the predicted position of the conserved C-terminus submerged in the mitochondrial inner membrane (dotted box), the di-ironcontaining active site (green) and the DMQ 6 substrate loaded into the active site (saffron). Coq7 dimerizes [35] and we have previously provided a model of dimeric rat CLK-1 [21].

Allele
Key Mutations* Other Mutations*** coq7-5(H153L) and coq7-11(W120R)-both of which severely limit growth on YEPE 3% ( Fig  1B) without affecting growth on 2% dextrose (YEPD 2% ). We re-integrated select alleles into the original COQ7 locus of SEY6210 (underlined in Table 1, and designated hereafter by subscript "i"), and found that coq7-11 i (W120R) mutants displayed temperature-sensitive, hypomorphic growth ( Fig 1C). During the course of our analyses we also observed that many of the epitopetagged COQ7 alleles that have been used in prior studies [22], are also hypomorphic for growth on YEPE 3% (Fig 1D). We have assigned allelic designations to some of these genes as well, since they vary in their degree of hypomorphism (Table 1 and S1 Table). Relevant to the current study is coq7-22(P ADH -COQ7-HA) which, when carried on a low copy centromeric (CEN) plasmid, conferred only residual growth to coq7-19(Δcoq7) null mutants cultured on YEPE 3% ( Fig  1D). We also generated two new epitope-tagged coq7 alleles, coq7-9(myc tag) and coq7-10(HA tag), using the promoter, open reading frame (ORF) and terminator sequences from SEY6210 wild type COQ7, and we confirmed that the presence of a C-terminal tag alone results in hypomorphic growth on YEPE 3% (Fig 1E), without disrupting growth on YEPD 2% . Finally, we mapped all newly identified allelic mutations onto a previously published structural model of Coq7p [21], (Fig 1F), the reliability of which has since been supported by spectroscopic and kinetic analyses [34,35]. coq7-5(H153L) disrupts the strictly conserved His153 residue and replaces it with a leucine. His153 chelates one of two iron atoms required for catalysis in the Coq7p active site. coq7-11(W120R) replaces the highly conserved Trp120 residue with positively charged arginine. Trp120 resides in a loop predicted to insert directly into the mitochondrial inner membrane. Trp120 also juxtaposes against the C-terminus and we have previously noted that Coq7p orthologs across multiple species all end abruptly with a bulky hydrophobic amino acid (F, I, or L) [21]. Presumably coq-7-11(W120R), as well as coq7-9(myc tag) and coq7-10(HA tag), disrupt this juxtaposition.

Identification of coq7 suppressor genes
To identify new regulators of Coq7p function, we undertook a high copy suppressor screen using the severely hypomorphic coq7-22(P ADH1 -COQ7-HA) allele. Mutant yeast were transformed with a library of SEY6210 genomic fragments carried on a high copy number plasmid. Growth suppressors were isolated following selection on YEPE 3% + 0.1% DEX. (Fig 2A). Using this approach we recovered multiple genomic fragments containing a wild type copy of COQ7 -validating our screening strategy. In addition, a single library suppressor clone (sup+) was identified that, when isolated and transformed into other coq7 mutant backgrounds, conferred suppression to coq7-9[P COQ7 -COQ7-Myc (CEN)], coq7-10[P COQ7 -COQ7-HA(CEN)] and coq7-11 i (W120R), but not coq7-5 i (H183L) (Fig 2B). These findings confirm that suppression by this clone did not depend upon the HA epitope tag, the presence of a Q48R point mutation, or the ADH1 promoter, and was not simply due to alterations in the copy number of the plasmid upon which hypomorphic P ADH1 -COQ7-HA(CEN) resided. Intriguingly, this genomic suppressor restored activity to Coq7p in an allele-specific manner-specifically, only coq7 mutations predicted to affect the membrane association of Coq7p were rescued ( Fig 1F).
CLD1 suppression occurs independent of the pet56 mutation Several yeast laboratory strains contain the his3-Δ200 mutation, including SEY6210-the parental background for our studies. The his3-Δ200 allele contains a 1036 base pair deletion that removes the entire HIS3 coding region and part of the neighboring promoter region of the PET56 gene [39]. As a consequence, PET56 expression is decreased by~80% [40]. PET56 is responsible for formation of 2'-O-methylguanosine at position G2251 in the mitochondrial large ribosomal RNA (21S rRNA). This nucleotide sits in the peptidyl transferase center of the mitochondrial ribosome and it has been previously reported that his3-Δ200 mutants are respiration-deficient when grown at 18°C, and sluggish at 30°C [39]. TA405 yeast contain the his3-11,15 loss-of-function allele which contains two single base pair deletions and does not affect PET56 expression. We exchanged the COQ7 locus in TA405 with the coq7-11(W120R) mutant allele and observed severely retarded growth on YEPE 3% which was suppressible by CLD1 overexpression ( Fig 2G). These findings show that CLD1 is a bona fide suppressor of hypomorphic coq7 mutants.
Cld1p activity is required for coq7 suppression While examining the role of CLD1 in the recovery of coq7 function, we isolated a cld1 mutant that was completely blocked in its ability to rescue the growth of any hypomorphic coq7 allele ('clone 2C' in Fig 2E and 2F). Sequence analysis revealed a PCR-induced point mutation that converted Trp170 to arginine (W170R). It has been previously reported that Cld1p contains an α/β hydrolase fold [38]. As for other hydrolases belonging to this family, the fold acts as a scaffold to correctly position a catalytic triad of residues situated on the end of three distal loops. The α/β hydrolase family has evolved to accommodate a wide range of substrate specificities because only the triad and shape of the fold are essential to catalysis [41]. We generated a model of Cld1p based on the crystal structure of epoxide hydrolase 1 from Solanum tuberosum (Fig 3A-3C). We identified the active site residues as Ser230, Asp392 and His424. This model is in good agreement with the model constructed by Baile and colleagues [37] using the α/β hydrolase domain of CumD. These authors experimentally verified that Ser230, Asp392 and His424 were each essential for Cld1p activity. Surprisingly, Trp170 sits distal from the catalytic triad ( Fig 3A), suggesting W170R must disrupt Cld1p function indirectly-possibly through destabilization of the α/β hydrolase fold (Fig 3B), or by disrupting membrane binding ( Fig 3C).
amplicons. (D) Growth suppression requires residual COQ7 activity to mediate enhanced growth on 3% ethanol (30°C). Strain genotype (top) and transformed plasmid identity (see panel C) are indicated. (Prom.-promoter). (E) The CLD1 locus is sufficient to confer growth enhancement on 3% ethanol to coq7-11 (30°C, left panel) and coq7-10 (35°C, right panel). Strain genotype (top) and transformed plasmid identity (refer to panel C) are indicated. (F) Quantification of coq7-11 i (W120R) growth inhibition in 3% ethanol (30°C) and its suppression by overexpression of CLD1. Neither the CLD1 promoter nor the cld1 lossof-function 2C amplicon are able to suppress the slow growth phenotype of coq7-11 i (W120R). (wt, wild type). Part of the promoter of CLD1 was removed when constructing the 1D amplicon to avoid incorporation of the Ty1 LTR into the clone and we presume this is the reason we did not observe the same degree of suppression as the original library clone (sup+). Disruption of just CLD1 within the sup+ clone background did not confer any suppression (see later, main text). (G) The ability of CLD1 to suppress hypomorphic coq7-11i (W120R) is independent of strain background. TA405 [COQ7/coq7-11 i (W120R)] heterozygous diploids were sporulated and four independent haploid coq7-11 i (W120R) isolates retained (left In regards to the latter, Cld1p is a membrane protein that faces the mitochondrial matrix [37]. Our model shows that Cld1p is surrounded by a skirt of surface-exposed hydrophobic patches, consistent with its function as a membrane lipid-modifying enzyme, and W170R sits directly below one of these regions ( Fig 3C).
Overexpressing Cld1p re-models acyl composition of MLCL in coq7-11 i (W120R) mutants In yeast, Cld1p deacylates cardiolipin (CL) to generate monolysocardiolipin (MLCL) and this enzyme shows a preference for removing palmitic acid (C16:0) [38]. Once formed, MLCL can be re-acylated with longer (C18) unsaturated fatty acids by the Tafazzin ortholog, Taz1p [36]. When Taz1p is removed from cells MLCL accumulates to detectable levels, suggesting Cld1p is constitutively active [42]. Cld1p and Taz1p therefore work together to dynamically control cardiolipin remodeling in the mitochondria [38]. It has been previously reported that wild type cells overexpressing Cld1p display no detectable alteration in their relative ratio of MLCL to CL [38], suggesting Taz1p, and not Cld1p, is rate limiting for CL production in yeast. In contrast, when Cld1p was overexpressed in coq7-11 i (W120R) cells we saw a significant increase in the relative ratio of MLCL to CL in whole-cell lipid extracts (Fig 4A and 4B, Multiple Regression Analysis, p<0.007). Cardiolipin contains four acyl chains, the composition of which influences CL functionality [43]. To examine the enzymatic activity of Cld1p further, we investigated how overexpressing Cld1p changed the acyl composition of CL in coq7-11 i (W120R) cells. We used HPLC-ESI-MS/MS and focused specifically on the product of the Cld1p reaction, MLCL. Four dominant MLCL species have been previously described in S. cerevisiae [44]. We observed a marked shift in the acyl composition of the MLCL pool in coq7-11 i (W120R) cells overexpressing CLD1 relative to control lines (Fig 4C and 4D). Notably, MLCL species containing C18:1 became significantly more abundant (t-test, p < 0.05). Interestingly, the appearance of MLCL(18:1) 3 in these cells suggests Cld1p may have both phospholipase A 2 and A 1 activity, and that this enzyme might be more appropriately described as a phospholipase B [45]. In contrast, coq7-11 i (W120R) cells overexpressing the cld1(W170R) mutation ('clone 2C') showed that the protein encoded by this allele was effectively enzymatically dead, which again emphasized the necessity of functional Cld1p for phenotypic rescue. Overall, our findings showing direct changes in the abundance of the MLCL product of the Cld1p reaction are in agreement with an earlier study that showed removal of CLD1 decreased the abundance of CL species containing unsaturated acyl chains [36].
The Q 6 /DMQ 6 ratio is normalized in coq7 mutants following CLD1 overexpression We next sought to determine how CLD1 overexpression suppressed the hypomorphic growth of mutant coq7 yeast. We observed by western analysis that hypomorphic coq7-22(P ADH1 -COQ7-HA) mutants overexpressing a genomic fragment containing wild type COQ7 produced less COQ7-HA protein. Overexpression of CLD1 induced the same effect, while overexpression of mutant cld1(W170R) did not (Fig 5). Q 6 is known to feedback and regulate the stability of its biosynthetic complex [14], of which Coq7p is a component, so our findings suggest that CLD1 ; the W170R point mutation encoded by the non-suppressing cld1 '2C amplicon' (red); (in B only) structural residues of the α/β hydrolase fold immediately surrounding Trp170 (pink); and (in C only) surface-exposed hydrophobic patches (mesh). doi:10.1371/journal.pone.0162165.g003 Reversal of a Ubiquinone Insufficiency in Yeast overexpression may be leading to direct recovery of Coq7p activity. We therefore quantified quinone amounts in CLD1 overexpressing cells during growth on YEPE 3% + 0.1% DEX. (Fig 6A). We analyzed several strains and observed that SEY6210 (wild type) and coq7-2 i (Q48R) cells both showed a burst of Q 6 production shortly after entering log phase growth on ethanol ( Fig  6B and 6C). We observed only small amounts of DMQ 6 accumulate in both strains under these growth conditions. coq7-11 i (W120R) cells exhibited a similar trend in Q 6 production at the onset of log phase growth, but it was much more exaggerated in magnitude. Notably, coq7-11 i (W120R) cells took 8 days to reach early log phase (25 times longer than wild type), and moreover, in contrast to the control lines, by this phase of growth DMQ 6 levels were also overtly elevated in this strain ( Fig 6D). Introduction of the original genomic DNA library suppressor clone (sup + ) into coq7-11 i (W120R) cells resulted in reversion of both the Q6 and growth phenotype toward that of the control coq7-2 i (Q48R) line ( Fig 6E and S3 Table). Introduction of just the CLD1 open reading frame showed partial rescue of both phenotypes ( Fig  6F), and lack of full suppression was presumably because the flanking DNA control regions were not present in their entirety in this clone (Fig 2C and [ 37]). Together, these data strongly suggest that CLD1 overexpression overrides the predicted membrane interacting defects of hypomorphic Coq7p proteins to allow them access to their DMQ 6 lipid substrate.
CLD1 overexpression lengthens the replicative lifespan of coq7-11 i (W120R) mutants Mutations that reduce or remove clk-1/coq7 function in C. elegans result in life extension [25]. It has not been possible to determine the effect of complete coq7 removal on yeast cell survival because Coq7p is essential for growth on YEPE 3% . With the identification of hypomorphic coq7 alleles we can now examine the effect of reduced coq7 function on yeast survival. We measured replicative lifespan, which monitors the ability of individual 'mother' cells to generate buds, and is generally considered the functional equivalent of lifespan studies in higher organisms [46]. coq7-11 i (W120R) mutants showed no significant alteration in replicative lifespan when compared with coq7-2 i (Q48R) cells, their relevant control ( Fig 7A). Interestingly, coq7-2 i (Q48R) cells, generated significantly fewer buds than SEY6210 yeast cells (Fig 7A and S4  Table). This is unexpected since, by all prior measures these two strains behaved indistinguishably (Figs 1E, 2F and 6). We next examined the effect of CLD1 overexpression on the replicative capacity of coq7-11 i (W120R) cells. Under the specific assay conditions we employed Whole-cell lipid extracts were prepared from the indicated yeast strains following growth at 25°C in liquid YEPE 3% to the indicated optical density (OD 600 ). Lipids were analyzed by thin layer chromatography (TLC). In (A), average MLCL/CL ratios (+/-range) are plotted for duplicate independent experiments, with triplicate technical replicates assayed at each OD 600, in each experiment. Multiple regression analysis (S2 Table)  Reversal of a Ubiquinone Insufficiency in Yeast (where cells were maintained on auxotrophic, plasmid-selection media immediately prior to analysis on 3% ethanol), the replicative lifespan of coq7-11 i (W120R) cells was extended between 2-and 3-fold when overexpressing CLD1 (Fig 7B). These findings underscore the major finding of this work-that some forms of ubiquinone insufficiency can be fully reversed by cardiolipin remodeling in yeast.

Discussion
In this study, we sought to identify genes that, when overexpressed, could suppress the ubiquinone insufficiency caused by hypomorphic disruption of the Q biosynthetic enzyme Coq7p. Using a newly-generated panel of coq7 mutants in the yeast S. cerevisiae, we discovered that ] mutant containing DMQ 6 (right peak); middle trace: extract from NM101 co-injected with Q 6 standard; top trace: purified Q 6 and Q 9 standards. Q 9 was added as an internal standard for all yeast quinone extractions. Peaks were detected using λ 275nm . (B-F) UV-HPLC quantitation of Q 6 (dashed line) and DMQ 6 (solid line) in the indicated strains following growth in liquid YEPE 3% . Quinone content is shown relative to culture density (OD 600, dotted line). Broken axes highlight the temporal difference in growth rate between strains (abscissa). (E) and (F) represent coq7-11 i (W120R) yeast containing the indicated construct (see Fig 2C for details). Each data point is the average of three independent biological replicates and represents total (reduced + oxidized) Q 6 and DMQ 6 . Error bars indicate standard deviation (See S3 overexpression of CLD1, encoding a cardiolipin-specific phospholipase A 2 [38], was able to fully rescue the slow growing phenotype of several different coq7 mutants. We provide evidence that rescue by CLD1 overexpression is dependent on the enzymatic function of Cld1p, and requires residual Coq7 activity. We have found through structural modeling that the mutant Coq7p proteins that were suppressed by Cld1p all had mutations positioned in, or adjacent to, a predicted membrane-binding region. We therefore cautiously speculate that in the absence of Cld1p overexpression, these mutant Coq7p proteins were unable to insert into the mitochondrial inner membrane efficiently or, if they were, they were unable to adopt the correct depth or orientation in the membrane to permit DMQ 6 access to the otherwise intact diiron active site [21,34,35]. Consistent with this interpretation, Cld1p overexpression was incapable of suppressing the hypomorphic coq7-5(H153L) mutation which encodes a H153L lesion that directly disrupts the Coq7p catalytic site. The most important finding of our present study is the identification of a functionally significant interaction between a cardiolipin remodeling enzyme and an enzyme of the Q biosynthetic machinery. We showed that enzymatically active Cld1p, when overexpressed, could normalize Q 6 and DMQ 6 levels in mutant coq7-11i(W120R) yeast. The extent and kinetics of this suppression is apparent in Fig 6: In control yeast grown in non-fermentable media (Fig 6B  and 6C), Q 6 levels reproducibly exhibit a spike upon entry into log-phase growth (refer to OD 600 axis). A second, minor peak is also observed in SEY6210 cells just before cells reach stationary phase. This latter peak may reflect activation of a stress response as nutrients become limiting, or it may reflect a drop in Q 6 levels below some critical threshold, in which case Q 6 synthesis presumably becomes reactivated. In any case, in contrast to Q 6 levels, DMQ 6 is maintained at almost undetectable levels during the entire growth phase. The situation for hypomorphic coq7-11i(W120R) mutants is, however, very different. We recorded a remarkable elevation of both Q 6 and DMQ 6 in these cells as they entered log-phase growth. At first this appears counterintuitive for an allele that is obviously hypomorphic for growth. However, it is clear from Fig 6D that DMQ 6 levels in this mutant almost always remain higher than those of Q 6 . Due to technical restrictions we were unable to collect sufficient cells to monitor earlier growth points than those shown, but we speculate that hypomorphic coq7-11i(W120R) mutants first accumulate massive amounts of DMQ 6 . At first, small amounts of Q 6 would be synthesized by the mutant enzyme, however, once Q 6 levels reach a critical threshold, cells enter log-phase and a much more rapid expansion of the mitochondrial network could then ensue. We speculate that either additional Coq7p is then synthesized, which would accelerate the conversion of the remaining DMQ 6 to Q 6 , as observed, or perhaps Cld1p is naturally activated as part of the mitochondrial expansion, given it is a cardiolipin modifying enzyme and cardiolipin is the most abundant lipid in the mitochondria. Regardless, in the context of coq7-11i(W120R) mutants, both options would result in growth enhancement. In support of this idea we note there is a second spike in the quinone profile of coq7-11i(W120R) mutants, toward the beginning of stationary phase. Here, DMQ 6 and Q 6 are present in much greater amounts relative to the equivalent growth phase of controls cells; and DMQ 6 levels are almost twice those of Q 6 . Again, this is consistent with a delay in DMQ 6 processing. The original library clone containing CLD1 in its genomic context conveyed near perfect suppression of the coq7-11i(W120R) mutant in regard to its quinone profile (Fig 6E). Overexpression of the Cld1p open reading frame was also effective at suppressing the slow growth defect of coq7-11i (W120R) mutants, though with delayed temporal efficacy relative to the genomic clone ( Fig  6F), likely because part of the CLD1 promoter region was removed (Fig 2 and [37]). Comparison of quinone levels between strains at various growth stages (compare OD 600 axes) clearly reveals the extent to which CLD1 overexpression is able to normalize DMQ 6 and Q 6 levels in coq7-11i(W120R) mutants. In summary, while natural elevation of Cld1p during growth could provide a simple explanation for why CLD1 overexpression was identified in our screen, less easy to explain is how Cld1p acts to suppress hypermorphic Coq7p in order to recover enzyme functionality.
Our finding that overexpression of Cld1p did not change the MLCL/CL ratio (Fig 4A and  4B), but instead increased the average length and degree of unsaturation of the acyl chains in its enzymatic product, monolysocardiolipin (Fig 4C and 4D), suggests that a shift in the inner membrane lipid composition likely plays an important role in the rescue of hypomorphic coq7 mutants. To date, there has been no report of Coq7p directly binding either cardiolipin or MLCL. In S. cerevisiae, cardiolipin species fall into five prevalent sub-types: CL(16:1) 4 ; CL (16:1) 3 (18:1) 1 ; CL(18:1) 2 (16:1) 2 ; CL(18:1) 3 (16:1) 1 and CL (18:1) 4 [44]. Even so, this amounts to 13 possible CL isomers, more when minor acyl species are included, and even more when 13 C isotopomers are included; understandably the precise molecular profile of CL in yeast remains poorly uncharacterized. It has been reported, however, that by late log phase the relative abundance of C16:0, C16:1, C18:0 and C18:1 acyl groups is 30%, 30%, 5% and 35%, respectively [38]. Our MS data clearly showed that CLD1 overexpression increased the abundance of C18:1 in MLCL. We do not know if alterations to MLCL per se, or to CL were required for hypomorphic Coq7p functional recovery, but we believe the most simple interpretation of our data is that changes in the mitochondrial lipid environment allowed previously hypomorphic variants of Coq7p to assume full catalytic functionality. Phospholipids are known to provide a conducive environment for the activity of many membrane proteins by dictating their folding and assembly [47,48]. The activity of many integral proteins is also dependent upon lipid bilayer thickness, which in turn is dependent upon acyl chain length [49,50]. Lipid-specific binding sites on integral membrane proteins are also known [47,51], and in this regard computational studies indicate that the side chains of hydrophobic residues such as isoleucine, leucine, valine and phenylalanine often interact with the acyl chains of phospholipids, while the more polar residues interact with lipid polar head groups and their glycerol backbones [47]. We found that the mutation in coq7-11i(W120R) replaces the highly hydrophobic tryptophan 120 residue with arginine. Since tryptophan does not generally tolerate substitution because of its structural uniqueness, in the absence of cardiolipin data, we cautiously speculate that arginine changed the hydrophobic surface of Coq7p enough to impair its interaction with the acyl chains of cardiolipin, and that Cld1p caused a compensatory change in the cardiolipin acyl profile that led to recovery of Coq7p activity in the mutant. Finally, it is possible that MLCL/CL substitution with C18:1 simply affected the thermal stability of mutant Coq7p proteins, a previous example of which is found in a site directed mutagenesis study involving interaction between a diacidic molecule of cardiolipin and the purple bacterial reaction center [48].
While alterations to the mitochondrial lipid environment that in turn directly affect Coq7p's ability to insert into the mitochondrial inner membrane is our favored hypothesis for how CLD1 overexpression functions to suppress hypomorphic Coq7p mutants with predicted membrane-binding defects, alternate explanations do exist. All are less-strongly supported by our data. For example, one possibility is that mutant Coq7p proteins with reduced membranebinding capacity are toxic to yeast cells and Cld1p overexpression somehow reduces their level. Although we showed that Cld1p overexpression decreased the level of mutant Coq7p protein in coq7-22(HA-tag) yeast, ubiquinone levels went up, not down, arguing against this hypothesis. Moreover, coq7-11(W120R) mutants, which are suppressible by CLD1 overexpression, exhibited no evidence of toxicity with respect to replicative lifespan when compared to control coq7-2 i (Q48R) cells, further arguing against this idea. Another alternate explanation is that the shift in the acyl signature of MLCL (and presumably cardiolipin by extension) toward C18:1, led to enhanced mitochondrial electron transport chain activity in hypomorphic coq7 mutants, and this was the reason they recovered their ability to grow on ethanol. Consistent with this idea, it is well established that mitochondrial respiratory chain complexes require cardiolipin for their structural and functional integrity (reviewed in [43]). Against this hypothesis, however, is our observation that the catalytically-compromised coq7-5(H153L) hypomorph was not equally rescued by CLD1 overexpression. Also, we observed a marked recovery in Q production in suppressed cells, again suggesting that the suppressive effect of Cld1p overexpression was specific to the Coq7p protein and not due to supplemental enhancement of ETC activity.
Recently, Busso and colleagues [52] also made use of random PCR mutagenesis in an effort to probe the structure/function of Coq7p. They identified only a single and weakly hypomorphic allele of coq7, (D53G), which contrasts with the 78 that we identified in this study, highlighting the strength of our new screening approach. These authors turned to an automated server-generated structural model of Coq7p and, in conjunction with site-directed mutagenesis, tested the function of several amino acid changes in the vicinity of D53G. Using this approach, they discovered a novel coq7(S114E) mutant which accumulated DMQ 6 at the expense of Q 6 , and which was severely hypomorphic for growth on ethanol/glycerol at 30°C, and lethal at 37°C. Overexpression of Coq8p partially rescued the growth defect. When coq7 (S114E) cells were placed at the non-permissive temperature of 37°C, the stability of Coq3p and Coq4p halved. This was not observed at 30°C, even though these mutants remained severely hypomorphic for growth on ethanol/glycerol. This finding suggests that formation of the 700kDa Q biosynthetic pre-complex was not limiting in coq7(S114E) mutants at lower temperatures and that some other function of Coq7p was disrupted. S114E maps to the same region of Coq7p where all our Cld1p-suppressible mutations localize. As mentioned above, this region is predicted to be surface-exposed, hydrophobic, membrane-binding, interfacial with the Coq7p C-terminus, and sit directly underneath the active site where the DMQ 6 tail extends [21].
Finally, Freyer and colleagues [6] recently reported the first known COQ7 mutation in a patient with primary Q 10 deficiency. The Freyer study demonstrated that administration of 2,4-dihydroxybenzoic acid (2,4DHB), a quinone analogue containing the ring modification normally catalyzed by COQ7, rescued the biochemical defect in patient-derived fibroblasts. This same compound was shown in an earlier study to essentially cure a mouse model of Coq7 deficiency [53]. Both studies had their genesis in yeast work using coq6 mutants and related quinone analogues [54,55]. Together, these findings highlight the non-linearity of Q biosynthesis and portend what will hopefully be a brighter future for patients with Q 10 deficiency. These findings also underscore the exciting conclusion from our study, that not only is there potentially another way to treat Q 10 deficiency in humans, albeit not as elegant as the solution above, but more importantly that we have broadened the field towards identifying alternate mechanisms for potentially reversing other deficiencies that result from alterations in the activity of critical membrane-bound enzymes.

Generation of coq7 allelic series
A 1371 base pair (bp) fragment containing the COQ7 gene was PCR amplified from SEY6210 and cloned into the BamH1 and Sal1 sites of pRS315(CEN) using forward (5'-CGCGGATCC GCTAGATGATGGATCTAAC-3') and reverse (5'-GGACGCGTCGACGTTCATTATCTTC GTTCGGCATTTCC-3') primers to form pRS315/COQ7. Mutagenic PCR, using Taq DNA polymerase, 40 μM Mn 2+ , 20 rounds of thermal cycling and the same primer pair listed above, was then used to randomly introduce~1 error/kbp along the entire 1371 bp COQ7 fragment. The resulting ensemble of PCR products was then directly transformed into strain SHM1 [SEY6210 Δcat5/coq7-19::GFP;HIS3] [56] along with an equal molar amount of purified pRS315/COQ7 previously cut with SnaB1 and HindIII. Both restriction sites reside within the COQ7 locus-approximately 50 bp from the 5' and 3' end, respectively. Homologous recombination occurred with high enough frequency to make this a viable alternative to cloning the PCR products manually. This and all subsequent transformations were undertaken using lithium acetate. Transformants containing the recombined plasmid were selected on glucose-containing minimal media lacking leucine (SD-L + 2% glucose) at either 25°C (2300 recombinants) or 30°C (1300 recombinants). Plasmid alone recombined with a frequency of <0.2% of the total number of positives. After 3 days, colonies were doubly replica plated onto YEPE 3% + 0.1% DEX. and grown at 25°C and 37°C, or 30°C and 37°C, respectively, and hypomorphic or temperature-sensitive coq7 alleles isolated. Seventy-eight mutants in total were identified. Plasmids were recovered from a sub-collection of these yeast and their entire COQ7 insert sequenced bi-directionally. Several alleles were re-integrated back into the coq7Δ locus of SHM1 (S1 Table). To do this, mutant coq7 alleles were excised from their pRS315 vector backbone using BamH1 and SalI and cloned by homologous recombination into the SHM1 coq7Δ locus. Integrants were selected on YEPE 3% at a temperature permissive for growth (S1 Table), and insertion at the coq7Δ locus was confirmed by PCR. The HA-epitope-tagged coq7 alleles in psHA71 and pmHA71 have been described previously [22]. Both alleles are under the expression of the ADH1 promoter. Sequence analysis revealed both constructs also differ from the COQ7 sequence present in SEY6210 at position 177 (M to I). This is a naturally-arising Coq7p polymorphism that confers resistance to petite formation [57]. NM101 (MATa coq7-1 leu2-3,112 ura3-52 his3 -), which contains the previously described coq7-1(G65D) allele [15], was the original source of genomic DNA used for PCR generation of psHA71 and pmHA71.

Total membrane lipid extraction
Total membrane lipids were isolated using a modification of the Bligh and Dyer method [61]. Briefly, relevant yeast strains were grown to an OD 600 of 3.7, 10.5 and 13.2, then harvested by centrifugation (10 min at 624 × g, 4°C). A standardized cell-pellet wet weight of 1 gram was employed for all subsequent analyses. For spheroplasting, cell pellets were resuspended in ice cold water (20 ml), washed twice in the same volume of water, then resuspended in 1.4 ml/wet weight (g) pellet in pre-warmed spheroplast buffer (100 mM Tris-HCl (pH 9.3), 5 mM EDTA and 1% β-mercaptoethanol or 10 mM DTT), for 15 mins at 30°C. The cell suspension was then centrifuged at 4°C, washed thrice with 20 mM potassium phosphate buffer (pH 7.65), then the cell pellet resuspended in 20 mM potassium phosphate (pH 7.65) containing 1.2M sorbitol 4 ml/wet weight (g) pellet. Cells were then warmed to 40°C for 3-4 mins and Zymolyase (Fisher Scientific) added to a final concentration of 2.5 mg/g. The cell suspension was incubated at 30°C for 3 hours. At the end of the incubation, spheroplasts were pelleted by centrifugation, washed three times in 1.2M sorbitol (2 ml /wet weight (g)), then resuspended in 1 ml of the same solution. For each 1 ml sample, 3.75 ml of 1:2 (v/v) chloroform:methanol was added and samples vortexed for 2 minutes. Subsequently, 1.25 ml of chloroform and 1.25 ml of distilled water was also added. The bottom phase was recovered following centrifuging at 1500 x g for 5 mins. The extracted solvent was evaporated under air flow and then the lipid extract was re-dissolved in 50 μl of a 2:1 mix of chloroform/methanol (v/v). The lipids were then stored in glass vials with butylated hydroxytoluene (BHT) to prevent lipid oxidation, ready for separation by Thin Layer Chromatography (TLC).

Thin layer chromatography
Analysis of cardiolipin and monolysocardiolipin in lipid extracts from whole-cell lysates and mitochondrial fractions was undertaken using TLC. Briefly, lipids were separated using Whatman 1 Partisil™ LK5 TLC plates, following the methodology of Vaden and colleagues [62]. TLC plates were first wetted with 1.8% boric acid in ethanol, then dried at 110°C for 15 minutes to activate the silica. Next, equal volumes (10 μL) of test samples were spotted onto the relevant lane of a TLC plate then separated using a solvent phase consisting of 30:35:7:35 (v/v/v/v) chloroform/ethanol/water/triethylamine. When separation was complete, TLC plates were vacuum dried and spots developed using sulfuric acid charring. Spots were identified using purified lipid standards that included cardiolipin (Avanti polar lipids, # 840012P), phosphatidylserine (Avanti polar lipids, # 870336P), and a defined mixture of abundant eukaryotic lipids extracted from Soy (Avanti Polar Lipids, # 690050P). Developed spots were immediately photographed and then spot area and density quantified using Image J (http://imagej.nih.gov/ij/). For statistical testing, MLCL to CL ratios were calculated for samples measured in triplicate technical replicates, collected from duplicate independent experiments, at three separate OD 600 values. Significance testing for strain-and growth phase-related differences was undertaken using multiple regression analysis (Real Statistics Resource Pack software, Release 4.3). Details provided in S2 Table).

Q 6 and DMQ 6 quantification
Yeast quinone levels were quantified using a modification of the procedure described by Schultz and colleagues [63]. Strains of interest were cultured in liquid YEPE 3% at 25°C to the following optical densities (OD 600 , 1 cm pathlength): 1.7, 3.7, 6.7, 8.7, 10.5, 13.2. In total, three independent experimental replicates were collected for each OD 600 . Cells were collected by centrifugation (10 min at 624 × g, 4°C), washed once with H 2 O (10 ml), transferred to 50-ml pre-weighed glass tubes, then re-pelleted and again resuspended in H 2 O (12 ml). Three 1 ml aliquots were taken from each sample and cell pellet wet-weights determined. Cell pellets were dried for 1 hr. at 56°C and then their dry-weights determined. The average of the dry-to wet-weight ratio for the three 1 ml aliquots was then used to calculate the dry weight of the remaining sample. The remaining cell suspension was centrifuged, and the total wet weight of the cell pellet was determined. Glass beads (10 times the cell pellet wet weight), and 250 μl of Q 9 internal standard (20 μg/ml) were added to the pellet and then the tubes were immediately flooded with nitrogen, capped, covered with foil, and kept on ice to prevent oxidation. Cells were lysed by vortexing for 2 min. Lipids were extracted by adding water/petroleum ether/methanol (1:4:6) and vortexing for an additional 30 sec. In some instances the water component was replace by 1M NaCl to enhance phase separation. Phases were separated by centrifugation (10 min at 624 × g and 4°C). The upper petroleum ether layer was then transferred to a 10-ml glass tube. 4 ml of petroleum ether was added to the glass bead-aqueous phase, and the samples were re-vortexed for 30 sec. The petroleum ether layers from a total of three extractions were pooled and dried under nitrogen. Lipids were resuspended in a final volume of 1 ml isopropanol. Quinones (Q 6 , DMQ 6 and Q 9 ) were separated by reversedphase high pressure liquid chromatography using a C18 column (Alltech Econosphere 5-μm, 4.6 × 250-mm, isocratic mode, mobile phase 72:20:8 (v/v/v) methanol/ethanol/propanol, 1 ml/ min, 40°C) and quantitated using a Waters 600E UV detector at a wavelength of 275 nm. Peaks representing Q 6 (Avanti polar lipids, # 900150O) and Q 9 (Sigma, # 27597) were identified using purified standards. The peak representing DMQ 6 was identified using a quinone extract from the yeast strain NM101, which is unable to manufacture Q 6 and instead accumulates this intermediate [15]. Peak areas were normalized based on starting dry weight and then significance testing for quinone-, strain-and growth phase-related differences was undertaken using multiple regression analysis (Real Statistics Resource Pack software, Release 4.3). Details are provided in S3 Table. Replicative lifespan analysis Replicative lifespan assays were conducted at 25°C following the procedure of Steffen and colleagues [64], except YEPE 3% + 2% agar was employed. Strains were coded to remove scoring bias. A total of 40 virgin cells (over replicate experiments) for each strain was analyzed using microdissection and the number of daughter cells that budded off from individual cells was recorded. Mother cells were monitored until no new buds were produced. Data was analyzed using a log rank test with significance (p<0.05) set against a chi square distribution using one degree of freedom (χ 2 >3.84). Mother cells that were either damaged or could not be distinguished from daughter cells were censored from the analysis. Censored daughter cells were excluded from the calculation of mean RLS in Fig 7. High performance liquid chromatography-electrospray ionizationtandem mass spectrometry (HPLC-ESI-MS/MS) Four or five cultures of relevant yeast strains were cultured at 25°C in YEPE 3% to mid log phase (OD 600 6-10). Sample OD 600 values were adjusted to 7.0 and a 35 ml volume was then processed for HPLC-ESI-MS/MS as follows. Cells were collected by centrifugation (10 min at 624 × g, 4°C), washed with MS-grade H 2 O (3 x 10 ml), then cardiolipins (CLs) and monolysocardiolipins (MLCL) extracted from yeast pellets with methanol, chloroform, and 0.1N HCl (1:1:0.9) following the procedure of Bazan and colleagues [44]. Briefly, cell pellets were homogenized with a Mini-Beadbeater-24 Homogenizer (BioSpec Products). CL (14:0) 4 [1',3'-bis [1,2-dimyristoyl-sn-glycero-3-phospho]-sn-glycerol] (Avanti Polar Lipids, #710332P) was added as the stable isotope-labeled standard prior to extraction. MS analyses were conducted on a Thermo Fisher Q Exactive (San Jose, CA) with on-line separation using a Thermo Fisher/ Dionex RSLC nano HPLC system and Waters Atlantis dC18 column (150 μm x 105 mm; 3 μm) (Waters Corporation, Massachusetts). The gradient was started at 10% B and run from 10% B to 99% B over 40 min with the flow rate of 6 μl/min-where mobile phase A was acetonitrile/water (40:60) containing 10 mM ammonium acetate and mobile phase B was acetonitrile/isopropanol (10:90) containing 10 mM ammonium acetate. Data-dependent analyses were conducted using one full MS scan (70,000 resolution) followed by six tandem-MS scans with electrospray negative ion detection. Quantitative results were obtained by referencing experimental MLCL peak areas against a standard curve of CL(18:1) 4 [1',3'-bis[1,2-dioleoyl-snglycero-3-phospho]-sn-glycerol] (Avanti Polar Lipids, #710335P), following CL(14:0) 4 normalization. Significance testing for differences between strains was undertaken using Student's t-test (p < 0.05), without correction for multiple testing.
Supporting Information S1

Author Contributions
Conceptualization: AK SLR.
Formal analysis: AK HB XG SLR.
Investigation: AK HB ML MBB XG SLR.