Primary Coenzyme Q Deficiency in Pdss2 Mutant Mice Causes Isolated Renal Disease

Coenzyme Q (CoQ) is an essential electron carrier in the respiratory chain whose deficiency has been implicated in a wide variety of human mitochondrial disease manifestations. Its multi-step biosynthesis involves production of polyisoprenoid diphosphate in a reaction that requires the enzymes be encoded by PDSS1 and PDSS2. Homozygous mutations in either of these genes, in humans, lead to severe neuromuscular disease, with nephrotic syndrome seen in PDSS2 deficiency. We now show that a presumed autoimmune kidney disease in mice with the missense Pdss2kd/kd genotype can be attributed to a mitochondrial CoQ biosynthetic defect. Levels of CoQ9 and CoQ10 in kidney homogenates from B6.Pdss2kd/kd mutants were significantly lower than those in B6 control mice. Disease manifestations originate specifically in glomerular podocytes, as renal disease is seen in Podocin/cre,Pdss2loxP/loxP knockout mice but not in conditional knockouts targeted to renal tubular epithelium, monocytes, or hepatocytes. Liver-conditional B6.Alb/cre,Pdss2loxP/loxP knockout mice have no overt disease despite demonstration that their livers have undetectable CoQ9 levels, impaired respiratory capacity, and significantly altered intermediary metabolism as evidenced by transcriptional profiling and amino acid quantitation. These data suggest that disease manifestations of CoQ deficiency relate to tissue-specific respiratory capacity thresholds, with glomerular podocytes displaying the greatest sensitivity to Pdss2 impairment.


Introduction
Coenzyme Q (CoQ) is a benzoquinone molecule with a polyisoprenylated side chain that ranges from 6 to 10 isoprenyl units in length. It functions as an electron carrier in the mitochondrial respiratory chain, where it transports electrons from complexes I or II to complex III. The polyisoprenyl diphosphate synthases which form the isoprenyl side chain of CoQ in mice and humans are each heterotetramers of two protein subunits [1]. The genes that encode these subunits are now designated Pdss1 and Pdss2 in mice, and PDSS1 and PDSS2 in humans. Although its identity was not understood at the time, the first known mutation in the Pdss2 gene arose spontaneously in the CBA/CaH colony of Dr. Mary Lyon and was designated kidney disease (kd). Homozygotes for the kd allele develop a lethal disease characterized by tubulointerstitial nephritis, dilated tubules, and proteinuria [2]. Mutant kd/kd homozygotes appear healthy for at least the first 8 weeks of life, but histological examination of the kidneys beginning at about 12 weeks of life reveals a mononuclear cell infiltrate and tubular dilatation with proteinaceous casts in cortical areas. Over time this extends to involve the entire kidney with resultant renal failure [3,4,5].
Renal disease pathogenesis was initially thought to be immune mediated, rather than resulting from a structural or developmental defect [3,6]. However, we have since shown that the same renal disease, including leukocytic infiltration of macrophages and natural killer cells, develops spontaneously in kd/kd;Rag-1 2/2 double homozygotes lacking functional T and B lymphocytes [7]. Furthermore, mutant kd/kd mice are now recognized to have features of collapsing glomerulopathy (CG), a unique glomerular morphology in which hyperplastic and hypertrophic podocytes overlie collapsed capillary loops. While interstitial nephritis is often present in CG, no single definable pathogenic trigger for this disease has emerged [8]. Dysregulation of podocyte terminal differentiation in kd/kd mice was demonstrated by de novo expression of cyclin D1 (marking cell-cycle engagement) and Ki-67 (indicating podoctye cell-cycle progression), with loss of expression of differentiation markers WT-1 and synaptopodin [9]. These results suggested that there may indeed be an intrinsic structural defect in the podocytes of kd/kd mice, with the inflammatory reaction playing only a secondary role.
A positional cloning approach demonstrated that the kd allele is a missense mutation in a prenyltransferase-like mitochondrial protein [10], now designated Pdss2. This enzyme forms a heterotetramer with another enzyme encoded by Pdss1 to generate nonaprenyl diphosphate in mice or decaprenyl diphosphate in humans. The polyisoprenoid diphosphate precursor is added to 4hydroxybenzoic acid, forming polyisoprenyl-hydroxybenzoic acid, an early intermediate in CoQ biosynthesis [1]. Compound heterozygous PDSS2 mutations were recently identified in a child with severe Leigh syndrome and nephrotic syndrome [11]. Primary CoQ deficiency in humans is a rare condition that typically presents with neurologic or myopathic disease that is often ameliorated with CoQ supplementation [12]. It has now been associated with mutations in three of the nine nuclear genes encoding component enzymes of the CoQ biosynthesis pathway including COQ2, PDSS1, and PDSS2 [11,13,14,15] as well as with electron-transferring-flavoprotein dehydrogenase (ETFDH) mutations affecting the transfer of electrons to CoQ [16]. Nephrotic syndrome has been seen in individuals with mutations in PDSS2 [11], as well as COQ2 [13,14,17]. Mechanisms underlying tissuespecific disease manifestations remain unclear.

Results
Pdss2 kd/kd Phenotype Is Recapitulated in B6.Podocin/ cre,Pdss2 loxP/loxP Mice The Pdss2 loxP construct was prepared as shown in Figure 1. The total knockout (B6.Zp3/Cre,Pdss2 loxP/loxP ) was embryonically lethal , with no homozygous embryos surviving beyond 10.5 days of gestation (data not shown). This is in agreement with the findings of Levavasseur et al. [18] and Nakai et al. [19], who reported that mouse embryos deficient in CoQ synthesis as a result of clk-1/coq7 (demethoxyubiquinone hydroxylase) mutations also arrest development at mid-gestation. Tissue-specific conditional knockouts were therefore generated by crossing B6.Pdss2 loxP/loxP mice with several Cre-expressing strains. As shown in Figure 2, the Pdss2 gene was successfully targeted in the glomeruli but not the collecting tubules of B6.Podocin/cre,Pdss2 loxP/loxP mice, whereas the collecting tubules and hepatocytes but not the glomeruli were targeted in B6.PEPCK/cre,Pdss2 loxP/loxP mice.
B6.Podocin/cre,Pdss2 loxP/loxP but not B6.PEPCK/cre,Pdss2 loxP/loxP mice had the same kidney disease phenotype as B6.Pdss2 kd/kd missense mice, as judged by albuminuria and histological evidence of nephritis. Sections of kidneys from B6.Podocin/cre,Pdss2 loxP/loxP mice, B6.PEPCK/cre,Pdss2 loxP/loxP mice, and controls are shown in Figure 3. The phenotype of B6.Podocin/cre,Pdss2 loxP/loxP is histologically indistinguishable from that of a B6.Pdss2 kd/kd missense mutant, with both showing greatly dilated tubules and extensive interstitial infiltration. In contrast, neither feature is seen in the B6.PEPCK/cre,Pdss2 loxP/loxP or the B6.Pdss2 loxP/loxP control. Renal tissue from conditional knockouts was also examined by electron microscopy, with results shown in Figure 4. No significant ultrastructural differences were observed between B6 and B6.Pdss2 loxP/loxP control kidneys ( Figures 4A and 4C), but the abnormalities present in the B6.Podocin/cre,Pdss2 loxP/loxP knockout ( Figure 4D) were the most extensive of any Pdss2 mutant studied [10]. Measurements of 24-hour urine albumin and semiquantitative histological scores from Pdss2 missense, all conditional knockout mutants, as well as controls, are shown in Table 1. Only the B6.Podocin/cre,Pdss2 loxP/loxP knockouts had a phenotype that resembled that of the B6.Pdss2 kd/kd mice, as measured by either albuminuria or histologically-scored nephritis.
Another phenotypic abnormality seen in B6.Pdss2 kd/kd homozygotes is elevated serum cholesterol and triglycerides [20]. In an effort to determine whether this alteration may in part result from the underlying biochemical defect in CoQ biosynthesis rather than wholly as a consequence of nephrotic syndrome, we measured cholesterol and triglyceride levels in B6.Podocin/cre,Pdss2 loxP/loxP and B6.Alb/cre,Pdss2 loxP/loxP mice. B6.Podocin/cre,Pdss2 loxP/loxP mice developed renal disease associated with significantly elevated plasma cholesterol levels ( Table 2), suggesting hypercholesterolemia is largely a consequence of nephrotic syndrome. However, significantly elevated plasma cholesterol was also seen in B6.Alb/cre,Pdss2 loxP/loxP (94 +/2 7.1 mg/dl) when compared with B6.Pdss2 loxP/loxP controls (61 +/2 5.5 mg/dl), suggesting that a Pdss2 defect in the hepatocytes does contribute to hypercholesterolemia.

Pdss2 Mutations Diminish CoQ Tissue Levels
CoQ content was determined in lipid extracts of tissue homogenates from livers and kidneys dissected from mutant and control mice. As shown in Figure 5 (Panel A), there was a significant reduction in CoQ 9 and CoQ 10 levels in the kidneys of B6.Pdss2 kd/kd mice compared to age-matched B6 controls. There were no significant differences in kidney CoQ 9 content between the B6, B6.Pdss2 loxP/+ , or B6.Pdss2 loxP/loxP control mice (Panels A and B). Neither the B6.Podocin/cre,Pdss2 loxP/loxP nor the B6.PEPCK/

Author Summary
Coenzyme Q is a critical component of the mitochondrial respiratory chain, the process by which cells make energy. Coenzyme Q deficiency in humans causes a wide range of disease manifestations affecting the nervous system, muscles, and kidneys. Here, we show that the failure to make Coenzyme Q due to a Pdss2 mutation is the cause of a lethal kidney disease in mice that was previously thought to result from an autoimmune process. Studying both a spontaneously occurring missense mutant and a series of mutants generated to have the Coenzyme Q deficiency targeted solely to liver, kidney, or macrophages, we show that the specific cell type in which the kidney disease arises is the glomerular podocyte. No other manifestations of disease are evident in these animals. However, our analysis of livers from these mice reveals that they have significant depletion of Coenzyme Q, impairment of mitochondrial respiratory chain function, and disturbance of many other basic metabolic processes. Similar microarray patterns of cellular alterations to primary mitochondrial dysfunction were seen both in these mice and in a previously reported nematode model, suggesting that a common cellular profile of primary respiratory chain function may exist across evolution.  cre,Pdss2 loxP/loxP mice had a significant reduction in the CoQ 9 levels of total liver or kidney homogenates (Panels B and C), which is consistent with the fact that only a small subset of cells were affected by these targeted disruptions. However, the B6.Alb/cre,Pdss2 loxP/loxP mice had less than 30 pmol CoQ 9 per mg liver protein, which would be expected if most hepatic cells were affected by the albumin promoter-driven Cre expression.

Pdss2 Mutations Impair Mitochondrial Respiratory Chain Function
B6.Alb/cre,Pdss2 loxP/loxP mice had no evidence of disease through 8 months of life, but isolated liver mitochondria respiratory chain capacity in 6 to 8 month old animals was impaired to a similar extent as seen in B6.Pdss2 kd/kd missense mice. Specifically, polarography of freshly isolated liver mitochondria showed significantly decreased complex I-and complex II-dependent integrated respiratory chain capacity in both B6.Pdss2 kd/kd and B6.Alb/cre,Pdss2 loxP/loxP mutants compared with controls ( Figure 6, panels A and B). Significantly increased complex IV-dependent respiratory capacity was also observed in both the B6.Pdss2 kd/kd and B6.Alb/cre,Pdss2 loxP/loxP mutants ( Figure 6, panel C). This suggests that the degree of CoQ deficiency in liver is sufficient to cause secondary upregulation of respiratory chain components (complex IV) distal to the genetic deficiency (CoQ). These functional alterations are supported by spectrophotometric electron transport chain enzyme activity analyses performed on frozen liver mitochondria from these same animals ( Figure 6, Panel D). Frozen mitochondria isolated from single whole kidney homogenates were also studied from five B6 and three B6.Pdss2 loxP/loxP controls, as well as three animals each of genotypes B6.Pdss2 kd/kd , B6.Podocin/cre,Pdss2 loxP/loxP , B6.PEPCK/ cre, Pdss2 loxP/loxP , and B6.Alb/cre,Pdss2 loxP/loxP . No significant differences were detected in activities of enzyme complex I-III, II-III, III, or IV normalized to citrate synthase activity for any of the mutants in comparison with controls, although a trend toward increase was observed for complex IV enzyme activity in B6.Pdss2 kd/kd kidney mitochondria (data not shown).

Metabolic Effects of Pdss2 Liver Conditional Knockout
A significantly altered metabolic phenotype at the level of concordant biochemical pathway expression changes was clearly present in B6.Alb/cre,Pdss2 loxP/loxP mutants by 6 months of life, as summarized in Figure 7 (left column). Specific metabolic consequences of Pdss2-based CoQ deficiency include significant upregulation at a nominal p-value,0.05 in 43 of 95 essential KEGG biochemical pathways analyzed by Gene Set Enrichment Analysis including oxidative phosphorylation itself, the tricarboxylic acid (TCA) cycle, and multiple metabolic pathways that provide substrates necessary for energy production including fatty acid metabolism and 8 different amino acid metabolic pathways. Many of these same pathways were similarly upregulated upon expression profiling of C. elegans mitochondrial mutants in nuclearencoded respiratory chain subunits of complexes I, II, and III ( Figure 7, middle and right columns) [21]. Unique to the CoQ biosynthetic defect, B6.Alb/cre,Pdss2 loxP/loxP mutants have upregulation of CoQ biosynthesis, likely in compensation for their underlying CoQ deficiency. Similarly, the substrates immediately prior to the Pdss2-encoded prenyldiphosphate synthase enzymatic block (such as farnesyl pyrophosphate) appear to get funneled toward alternative biochemical pathways, as above. Interestingly, regulation of autophagy is uniquely upregulated in the B6.Alb/cre,Pdss2 loxP/loxP mutants, supporting previously reported findings of mitophagy occurring in these animals as characterized by endoplasmic reticulum engulfing liver mitochondria [10]. Consistent with CoQ function in both prooxidant and antioxidant roles, global expression analysis demonstrates the upregulation of multiple cellular defense pathways including the metabolism of P450 and glutathione, ascorbate metabolism, as well as the degradation of bisphenol A, limonene, pinene, and gamma-hexachlorocyclohexane. These mutants also appear to have alterations in basic DNA metabolism, as evidenced by upregulation of the metabolism of purines, pyrimidines, nucleotide sugars, ribosomes, aminoacyl-tRNAs, and RNA polymerase. Finally, four KEGG pathways appear on initial inspection to be discordantly expressed between mouse Pdss2 mutants and C. elegans complex I subunit mutants (Figure 7, left and middle columns, green highlight); however, none of these pathways upregulated in B6.Alb/cre,Pdss2 loxP/loxP were validated as having altered expression in the independent C. elegans dataset comprised of eight different complex I, II, and III respiratory chain subunit mutants (Figure 7, right column).
Altered amino acid metabolism is further supported by significant concentration differences among multiple amino acids quantified in livers of B6.Alb/cre,Pdss2 loxP/loxP mutants (Figure 8, Panels C and D). Hepatic glutamate was lowered to less than half the control value (p,0.01; Figure 8, Panel D). Glutamate depletion could account for decreased levels of glutamine and alanine ( Figure 8, Panel D) that derive from glutamate via glutamine synthetase and alanine aminotransferase, respectively. Similarly, a loss of glutamate could lower hepatic concentrations of phenyalanine, methionine, leucine, isoleucine and valine ( Figure 8, Panel C), each of which forms from glutamate by a specific transaminase. The total absence of citrulline ( Figure 8, Panel C) suggests ureagenesis is compromised in Pdss2 mutant mice. Similarly, the sole amino acid to be significantly increased was the urea cycle precursor, aspartate ( Figure 8, Panel C), with a sharp reduction in the glutamate:aspartate ratio (1:4 vs 4:1) of mutant vs control liver.

Pdss2 Missense Mutants Have No Clinical Manifestations of Extra-Renal Disease
Histological sections of the brain, retina, liver, and skeletal muscle of B6.Pdss2 kd/kd mutants demonstrated no obvious structural abnormalities. Skeletal muscle sections stained with Gomori/trichrome and antibodies to muscle enzymes (ATPase, NADH, SDH) showed no evidence of subsarcolemmal mitochondrial aggregates or other abnormalities. In addition, no differences were observed in bone mineral density or concentration, lean body mass, complete blood count, or liver transaminase (ALT) (data not shown). Several significant differences were noted in the liver amino acid levels of these animals (Figure 8, Panels A and B), but these changes were not as intense or widespread as observed in livers of B6.Alb/cre,Pdss2 loxP/loxP mutants (Figure 8 Panels C and D). The fact that most of these changes affected essential amino acids -threonine, phenylalanine and isoleucine, for example -could reflect albuminuria with concomitant nitrogen wasting and relative protein deficiency in these animals.

Discussion
The occurrence of a kidney disease phenotype in B6.Podocin/ cre,Pdss2 loxP/loxP but not B6.PEPCK/cre,Pdss2 loxP/loxP mice resolves a long-standing question regarding the cellular localization of the primary defect in B6.Pdss2 kd/kd mice [3,20]. Table 1 shows that a primary defect in podocytes recapitulates the kd/kd disease phenotype and is responsible for the development of nephrotic syndrome, which was not the case for mice expressing the defect primarily in renal tubular epithelium and hepatocytes (B6.PEPCK/ cre,Pdss2 loxP/loxP ). Hence, interstitial nephritis occurs as a consequence of a Pdss2 defect in glomerular podocytes. Knocking out the Pdss2 gene in podocytes resulted in a more severe phenotype than that observed in mice homozygous for the Pdss2 kd mutation (Figure 4), as would be expected if the product of the missense allele has at least some residual activity. Detection of CoQ 9 and CoQ 10 in the kidneys of B6.Pdss2 kd/kd mice, while almost ten-fold lower than age-matched B6 controls, suggests that the V117M amino acid substitution mutation in the Pdss2 kd gene product retains partial activity ( Figure 5).
That tubular dilatation and interstitial nephritis are downstream consequences of a CoQ deficiency in podocytes is illustrated by Figure 3. Our results suggest that CoQ deficiency negatively impacts mitochondrial respiratory chain function, although loss of CoQ antioxidant function may also be involved in pathogenesis, which leads to podocyte death with important consequences for kidney function. The slit diaphragm, which is dependent on podocyte foot processes, is essentially destroyed, causing leakage of various proteins normally maintained within the circulation, such as serum albumin. Some of these molecules have deleterious effects on the collecting tubules, which leads to tubular dilatation and dysfunction, release of inflammatory mediators, and development of interstitial nephritis. The results in Table 2 suggest that the elevation in serum cholesterol that is characteristic of Pdss2 kd/kd homozygotes [20] is not entirely attributable to kidney disease, but likely results from shunting of farnesyl diphosphate (one of the precursors in the CoQ biosynthetic pathway immediately proximal to the Pdss2 enzymatic block) to alternative pathways, primarily involving cholesterol biosynthesis. This is supported by transcriptional profiling of B6.Alb/cre,Pdss2 loxP/loxP liver which demonstrated upregulation of two pathways for which cholesterol is the starting substrate, namely bile acid biosynthesis and steroid biosynthesis (Figure 7).
In light of the severe, multi-system phenotype of the child with primary CoQ deficiency due to PDSS2 mutations [11], we undertook an extensive phenotypic evaluation of B6.Pdss2 kd/kd spontaneous missense homozygous mutants. All studies were performed in animals more than 120 days old, the time by which renal dysfunction is present. No overt non-renal disease manifestations could be detected, although significant deficiencies were observed in their respiratory function and activity of isolated liver mitochondria (Figure 6), as well as quantitative amino acid profiles in liver (Figure 8). It remains possible that additional manifestations of CoQ deficiency would develop with time but do not because of the high mortality of their renal disease. This possibility is being further evaluated by prolonged observation of tissuespecific conditional knockouts without renal disease.
Interestingly, B6.Alb/cre,Pdss2 loxP/loxP liver-conditional knockout mutants manifest no overt symptoms or signs of liver disease through at least eight months of life. Nonetheless, focused evaluations of the livers of these animals demonstrated substantial (97%) knockdown of Pdss2 expression; pronounced CoQ 9 deficiency ( Figure 5) and significant impairment of complex Iand complex II-dependent integrated respiratory capacity as well as complex I-III and II-III enzyme activities with compensatory increase in complex IV-dependent respiratory capacity in isolated mitochondria ( Figure 6). The relatively robust respiratory electron transport chain activities are quite remarkable, given that CoQ 9 was not detected over background in these liver lipid extracts. The low levels of CoQ 9 present in these animals must support significant activities of complexes I, II, and III. Very small amounts of CoQ support significant respiratory chain activity in Figure 6. CoQ Deficiency Causes Mitochondrial Dysfunction in Pdss2 Mutant Mice. Impact of coenzyme Q deficiency on mitochondrial function in B6.Pdss2 kd/kd and B6.Alb/cre,Pdss2 loxP/loxP mice compared to concurrent age-and strain-matched controls (B6 and B6.Pdss2 loxP/loxP , respectively). Panels A, B, and C demonstrate polarographic results of freshly isolated liver mitochondria with substrates which specifically interrogate complex I-dependent (malate+glutamate), complex II-dependent (succinate), and complex IV-dependent (TMPD+ascorbate) OXPHOS capacity. B6.Pdss2 kd/kd missense mutants have impaired complex I-and complex II-dependent OXPHOS capacity by 23% and 33%, respectively, with a 19% increase in complex IV-dependent OXPHOS capacity. B6.Alb/cre,Pdss2 loxP/loxP mutants have impaired complex I-and complex II-dependent OXPHOS capacity by 20% and 40% respectively, with a 43% increase in complex IV-dependent OXPHOS capacity. Values represent state 3 (near maximal) oxygen consumption mean rate and standard error. Panel D summarizes electron transport chain activity assays for complexes I-III, II, III, and IV performed on remaining aliquots of liver mitochondria frozen following completion of polarography. Similar to polarographic results, I-III enzyme activity is significantly decreased by 19% in B6.Pdss2 kd/kd missense mutants and by 41% in B6.Alb/cre,Pdss2 loxP/loxP mutants when compared to pooled controls. Complex IV enzyme activity demonstrates a trend toward increase by 11% in B6.Pdss2 kd/kd missense mutants and by 23% in B6.Alb/ cre,Pdss2 loxP/loxP mutants when compared to pooled controls, but does not reach the level of statistical significance, perhaps because of small sample size (n = 2 for remaining mitochondria from B6.Alb/cre,Pdss2 loxP/loxP mutants). As expected, no change is seen in complex II or III activity for either mutant. Values represent mean enzyme activity and standard error. B6 and B6.Pdss2 loxP/loxP ETC activity values were pooled to increase control sample size following demonstration of no significant differences between these strains. Statistical analyses comparing mutants and controls were performed by one-way ANOVA, where * indicates p,0.05 and ** indicates p,0.01. doi:10.1371/journal.pgen.1000061.g006 other models (yeast, nematode, isolated mitochondria). Liu et al. [22] showed that hepatocytes in clk-1/+ heterozygous mice lose heterozygosity (become clk-1 2/2), and these cells undergo clonal expansion within the liver. These findings suggest that liver cells perform quite adequately with low CoQ content.
A significantly altered cellular metabolism phenotype in liver of B6.Alb/cre,Pdss2 loxP/loxP mice was evidenced by concordant upregulation of 43 biochemical pathways (Figure 7) with significant deficiency of 9 amino acids (Figure 8). Taken together, these alterations confirm that Pdss2 dysfunction is a nuclear gene-based mitochondrial respiratory chain defect to which cellular adaptation is occurring. Indeed, 15 key biochemical pathways were concordantly upregulated on transcriptional profiling in this murine model of primary mitochondrial dysfunction and in previously reported C. elegans models of primary respiratory chain dysfunction due to mutations in nuclear gene-encoded subunits of complexes I, II, and III (Figure 7). These similarities spanning evolution provides support that a transcriptional ''signature'' of mutations in the respiratory chain appears to be stimulation of the constituent components of oxidative phosphorylation, the enzymes of the TCA cycle, and many pathways (e.g., glycolysis, amino acid metabolism, fatty acid metabolism) that furnish substrate to it, as well as stimulation of key cellular defense pathways such as glutathione and P450 metabolism [21].
Global amino acid alterations in Pdss2 mutant liver further suggest that significant impairment may be occurring in flux Global genome expression profiling in B6.Alb/cre,Pdss2 loxP/loxP mouse liver identifies concordant transcriptional alterations interpretable at the level of multiple metabolic pathways, which suggest significantly altered intermediary metabolism occurs despite an apparent absence of symptomatic disease. Extensive evolutionary concordance in upregulation of key biochemical pathways is seen in primary mitochondrial dysfunction, both in this mammalian Pdss2 liver-conditional knockout model of coenzyme Q deficiency and in a previously reported C. elegans gas-1(fc21) missense mutant model of primary complex I dysfunction [21]. Biochemical pathways as curated from the KEGG online database (http://genome.jp.kegg) are indicated with the # of genes in each pathway (cluster size), normalized enrichment score (NES), statistical significance of altered pathway expression between mutant and wildtype controls (p-value), and false positive percentage in the form of a false discovery rate (FDR q-value) according to GSEA. Pathways are ranked by descending NES in the Pdss2 mutant (left data column). Comparison to previously reported complex I gas-1(fc21) missense C. elegans mutant dataset (middle data column) and a ''validation'' C. elegans dataset of 8 different complex I, II, and III missense and RNAi-interference generated mutants (right data column) is indicated by differential highlights [21]. Font color denotes a pathway as relatively upregulated (red), downregulated (green), or unchanged (black). doi:10.1371/journal.pgen.1000061.g007 through key biochemical pathways, including glycolysis, the pyruvate dehydrogenase reaction, the TCA cyle, and ureagenesis. Changes in the hepatic amino acid profile were not qualitatively dissimilar in B6.Alb/cre,Pdss2 loxP/loxP mutants compared with those of the B6.Pdss2 kd/kd missense mutants, but alterations in the liverconditional knockout were far more widespread and intense ( Figure 8). The characteristic pattern in both mutants was a sharp reduction of hepatic amino acid concentrations, which appear to be related to a relative depletion in the hepatic concentration of glutamate. Indeed, glutamate is the ''pivot'' of nearly all amino acid metabolism as it furnishes amino groups to ketoacids to support the concentrations of alanine, leucine, isoleucine, valine, and phenylalanine, as well as to glutamine via glutamine synthetase. In addition, glutamate rapidly enters the TCA cycle as a-ketoglutarate either via transamination or the glutamate dehydrogenase pathway. We propose that diminished glutamate may occur in Pdss2 mutants due to their enhanced utilization of glutamate as a metabolic fuel in response to a relative diminution in their glucose consumption rate via glycolysis. It is recognized that in some tissues -particularly kidney and brain -glutamate can become an important metabolic substrate [23]. The significant reduction observed in alanine concentration (Figure 8, Panel C) is consistent with this formulation, since alanine carbon forms from pyruvate that is produced in glycolysis. Similarly, the sharp reduction in the glutamate:aspartate ratio in mutant liver implies an increase of hepatic oxaloacetate, a TCA cycle intermediate which is the obligatory precursor both to aspartate carbon as well as to the TCA cycle citrate synthetase reaction. In other words, if glycolytic flux is relatively diminished in the Pdss2 mutants, less pyruvate will be produced and flux through pyruvate dehydrogenase will be attenuated, thereby limiting availability of acetyl-CoA to the citrate synthase reaction, which would augment the hepatic pool of oxaloacaetate and impair TCA cycle flux. Diminished glutamate would also impede production of N-acetylglutamate, an obligatory effector of the carbamylphosphate synthetase reaction sequence in the mitochondrial segment of the urea cycle that leads from ammonia to citrulline, the total absence of which ( Figure 8, Panel C) points to a compromise of ureagenesis in the liver of B6.Alb/cre,Pdss2 loxP/loxP mice. It is of interest in this regard that the sole amino acid to be significantly increased is aspartic acid (Figure 8, Panel C), since a failure of citrulline synthesis would sharply curtail flux through cytosolic argininosuccinate synthetase, a step in the urea cycle that utilizes both citrulline and aspartate as precursors. The diminished glutamate:aspartate ratio highlighted above also denotes compromised flux through the hepatic urea cycle, to which aspartate is a major nitrogen donor. Additional investigations are necessary and underway to verify the extent to which metabolic flux is impaired in Pdss2 CoQ biosynthetic mutants.
We conclude that the Pdss2 kd allele causes a milder phenotype in homozygous mice than that observed in the homologous human case [11]. What is especially puzzling about the Pdss2 kd/kd mutant mouse is why an intermediate level of activity in oxidative phosphorylation should uniquely affect podocytes, a cell type not previously characterized by unusually high levels of respiratory activity. However, it is clear that CoQ concentrations and Pdss2 mutants have altered amino acid profiles in liver. Quantitative liver amino acid analysis detects significant differences in B6.Pdss2 kd/kd missense mutants compared with B6 controls (Panels A and B), as well as in B6.Alb/ cre,Pdss2 loxP/loxP mutants compared with B6.Pdss2 loxP/loxP controls (Panels C and D). To better demonstrate differences in all amino acids, results for three amino acids in highest abundance are shown separately (Panels B and D) with a greater scale compared to that used for the remainder of the amino acids present in relatively lower abundance (Panels A and C). Values represent mean +/2 SEM. Statistical analyses comparing mutants and controls were performed by Student's t-test, where * indicates p,0.05 and ** indicates p,0.01. doi:10.1371/journal.pgen.1000061.g008 intermediary metabolism are significantly altered in the missense mutants. It is also interesting that the Pdss2 kd/kd mouse is a mammalian model of mitophagy, a process recently described in yeast to involve the preferential degradation of impaired mitochondria before initiating cell death [24]. The occurrence of a subsequent intense inflammatory response suggests that autoantigens are exposed by autophagy to initiate either an innate or adaptive immune response. The inflammatory response is not the primary cause of the renal disease, but appears to be the mechanism used to collect and dispose of cellular debris. Availability of B6.Pdss2 loxP/loxP mice will now permit future investigations into the specific sites of CoQ deficiency and tissue damage, particularly in the central nervous system and skeletal muscle, which lead to the heterogeneous but severe consequences of primary mitochondrial disease as observed in the human PDSS2 patient.

Construction of Vectors
The targeting construct and Pdss2 loxP/loxP construct were made as follows. The 11.8 kb linearized DNA containing the Pdss2 exon 2 was from a bacterial artificial chromosome (BAC #RP23-256E1) with digestion of BamH 1 and Dra III, and was subcloned into pBlueScript vector. The cassette containing loxP-neo-loxP was inserted into an Aat II site, and a loxP was introduced into an Xho I site. The Pdss loxP/loxP targeting construct is depicted in Figure 1 and contains approximately 7 kb of homology regions. This linear targeting construct was electroporated into R1 ES cells. Stably transfected cells were isolated after selection in G418 (350 ul/ml, Gibco), and 384 clones were screened for the desired homologous recombination event with PCR. Chimeric mice were generated by aggregation of ES cells with morulae of B6 mice, and the modified allele was passed through the germline by breeding chimeras to B6 mice. The genotypes of all offspring were analyzed by polymerase chain reaction using DNA from the ear of mice or the yolk sac of embryos. The k/oF and k/oR primers for determination of Exon 2 (PCR product of 440 bp) were 59-GGAGGTTGAGTCCCTGTGTC-39 and 59-GCAGGAAAT-CAGTGGGACTC-39 respectively. PCRs were carried out for 30 cycles (94uC for 20 s, 60uC for 30 s, 72uC for 1 min) in a buffer containing 1.5 mM MgCl 2 .

Mice
Mice homozygous for the floxed gene (B6.Pdss2 loxP/loxP ) were crossed with partners that expressed cre under the control of a zona pelucida glycoprotein 3 promoter (C57BL/6-Tg(Zps-cre)93 Knw/ J; Zp3/cre), obtained from The Jackson Laboratory (Bar Harbor, ME), which caused the floxed gene to be deleted in all tissues. Podocyte-specific knockouts (B6.Podocin/cre,Pdss2 loxP/loxP ), were obtained by crossing mice homozygous for the floxed gene with partners that expressed cre under the control of the Podocin promoter [25]. The mutation was targeted to the renal proximal tubular epithelial cells (as well as a subset of hepatocytes) by mating B6.Pdss2 loxP/loxP mice with partners that express a PEPCK/ cre transgene [26]; to hepatocytes by utilizing mice with an albumin/cre promoter (B6.Cg-Tg(Alb-cre)21 Mgn/J (Alb/cre)), obtained from The Jackson Lab; or cells of the myeloid lineage with mice that express cre under the control of the lysozyme M promoter (B6.129P2-Lyzs tm1(cre)Ifo /J; LysM/cre [27] obtained from The Jackson Laboratory. Mice with the Pdss2 kd/kd missense mutation on the B6 background have been previously reported [7,10]. All procedures were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania.

Albumin Assay
A mouse albumin ELISA kit was obtained from Bethyl Laboratories Inc. Briefly, ELISA plates (Nunc, No: 442404) were coated overnight with goat anti-mouse albumin antibody (10 ml / ml) in bicarbonate buffer pH 9.6. Wells were washed with distilled water, blocked with PBS-1% BSA, washed with PBS-0.05% Tween 20, and 100 ml of serially diluted samples or standards added for 1 hour at room temperature. Plates were then washed using PBS-0.05% Tween 20, and incubated with 100 ml of HRPconjugated goat anti-mouse albumin antibody at room temperature for 1 hour. After further washing with PBS-0.05% Tween 20, wells were incubated with 100 ml ABTS solution at room temperature for 20 min, then with 100 ml 2 M H 2 SO 4 stopping solution and the plate was read with an ELISA reader at 405 nm.

Histology
Kidneys from mutant and control mice were fixed in formalin, paraffin-embedded, and 4 mm sections through the longitudinal axis of each kidney were prepared and stained with hematoxylin and eosin. The sections were examined blindly and scored as follows: 0 = no tubular dilatation and no mononuclear cell infiltrates; 1 = small focal areas of cellular infiltration and tubular dilatation involving less than 10% of the cortex; 2 = involvement of up to 25% of the cortex; 3 = involvement of up to 50% of the cortex; 4 = extensive damage involving more than 75% of the cortex. Coronal sections of the brain were stained by H & E. To examine the retina, anesthetized mice were perfused through the heart with freshly prepared 2% PFA+2% glutaraldehyde in 0.2 M sodium cacadylate buffer (pH 7.4). Eyes were removed and eyecups were fixed in the same buffer for 4 hours at room temperature. Tissues were then dehydrated in a graded ethanol series, infiltrated, and embedded in EMbed812 (Electron Microscopy Sciences). Sections of 1-2 m in thickness were cut and stained with toluidine blue. Liver sections were examined after H & E and Oil Red O staining. Skeletal muscle was stained by Gomori/ trichrome stain, and with antibodies to ATPase, SDH, NADH, and acid phosphatase.

Electron Microscopy
As previously described [28], kidney tissue samples were fixed in 2.5% glutaraldehyde and 2.0% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.4, overnight at 4uC. After three cacodylate buffer washes, the samples were post-fixed with 2.0% osmium tetroxide in 0.1 M cacodylate buffer for one hour at 4uC. After two additional sodium cacodylate washes and a wash in dH 2 O, samples were stained with 2% aqueous uranyl acetate for 30 minutes at room temperature. Samples were rinsed again in H 2 O subsequent to dehydration in graded ethanols, and infiltration and embedding in Embed-812 (Electron Microscopy Sciences, Fort Washington, PA). All sections were examined in a JEOL 1010 electron microscope, and digital images were recorded with a Hamamatsu camera system.

CoQ Measurements in Mouse Kidney and Liver Homogenates
Whole kidneys and livers were dissected from sacrificed mice and were stored at 280uC until homogenization. Individual kidneys or lobes of livers were placed in 5 mL of 1X PBS, pH 7.4 (137 mM NaCl, 8 mM Na 2 HPO 4 , 2.7 mM KCl, 1.5 mM KH 2 PO 4 ) 4uC, and tissues were homogenized a total of 10 strokes with a tight-fitting Teflon pestle rotating at maximal speed with a Fisher Scientific Lab Stirrer LR400A. The homogenate was centrifuged at 10006g for 5 minutes. The supernatants were removed to fresh vials, and protein concentrations were measured by the bicinchoninic acid assay (Pierce, Rockford, IL). Aliquots of each homogenate were transferred to 50 ml glass tubes and stored at 280uC until extracted. Lipid extractions were performed on CoQ 9 standards, CoQ 10 standards, and mouse tissue homogenates by adding the same amount of coenzyme CoQ 6 internal standard along with 0.5 ml water, 9 ml methanol, and 6 ml petroleum ether to each. Mixtures were vortexed for 1 min, centrifuged at 9106g, and the top layer of petroleum ether was removed and saved in a separate vial. Fresh petroleum ether (6 ml) was added to each vial containing the aqueous phase and vortexed for 1 min. The vials were subjected to centrifugation as before and the second petroleum ether layer removed. The process was repeated once more, and the three pooled petroleum ether fractions were dried under nitrogen and resuspended in 200 ml methanol. The quinones were then separated and quantified by HPLC connected to an electrochemical detector as described [29], with the following exceptions: the precolumn electrode was the only electrode used and was set at +650 mV to oxidize all hydroquinones, and a Gilson 118 UV/Vis detector was utilized to detect quinones (275 nm) as they eluted from the column. The amount of CoQ 9 and CoQ 10 in the standards and samples was normalized to the amount of CoQ 6 recovered in the individual lipid extracts.

Integrated Respiratory Capacity Analysis of Freshly Isolated Liver Mitochondria
Mice were sacrificed by cervical dislocation without general anesthetic, since anesthetics may exacerbate mitochondrial dysfunction in primary mitochondrial disease [30]. Mouse liver mitochondria were isolated in a similar manner as described previously for rat liver mitochondria [31]. Briefly, livers were immediately dissected and collected in MSM buffer (220 mM mannitol, 70 mM sucrose and 5 mM MOPS, pH: 7.4). All subsequent mitochondrial isolation procedures were done on ice. Livers were washed twice with MSM, finely minced, and then homogenized in MSM-E (MSM containing 2 mM EDTA) with 2 slow strokes using a tight-fitting Teflon pestle in a glass Potter/ Elvehjem tissue grinder (DuPont, Wilmington, Delaware) at 400 rpm using an IKA RW20 digital homogenizer (Cole-Palmer, Illinois). Following initial centrifugation at 300 g610 minutes at 4uC, the supernatant was collected and centrifuged at 7000 g610 minutes at 4uC (Eppendorf 5810R centrifuge, rotor F34-6-38, Hamburg, Germany). The pellet containing mitochondria was then washed twice in MSM with high speed centrifugation at 7000 g610 minutes at 4uC. 2% protease inhibitor cocktail was added to the final washed pellet preparation (SIGMA, St. Louis, MO). Protein concentration was determined by Lowry assay [32]. Mitochondrial oxygen uptake by polarography using a clark-type electrode (Oxytherm, Hansatech Instruments, UK) was performed on freshly isolated mitochondria at 30uC, as previously described [30]. Substrates used were malate+glutamate, succinate, and TMPD+ascorbate to assess complex I-, II-, and IV-dependent integrated respiratory capacity, respectively. State 3 oxidative phosphorylation (OXPHOS) respiratory rates were measured in the presence of ADP to approximate maximal mitochondrial capacity. Mean state 3 rates with standard errors were calculated from at least duplicate tracings for each electron donor substrate for liver mitochondria isolated from 6 B6.Pdss2 kd/kd missense mutants, 3 B6. Alb/ cre,Pdss2 loxP/loxP conditional knockout mutants, and a total of 9 concurrent age-and strain-matched controls (66B6 and 36B6.Pdss2 loxP/loxP , respectively). As no statistically significant differences were detected in state 3 rates for any substrate between B6 and B6.Pdss2 loxP/loxP controls, all control data (B6+B6.Pdss2 loxP/ loxP ) were used for comparative analysis with each mutant strain. Comparison of means was done by one-way ANOVA assuming equal variance (SPSS v.12.0, Chicago, IL).

Liver and Kidney Mitochondrial Enzyme Activity Analyses
Following polarographic analysis of isolated liver mitochondria, as described above, remaining mitochondrial aliquots were promptly frozen in 280uC and subsequently available for study on 4 B6.Pdss2 kd/kd missense mutants, 2 B6.Alb/cre,Pdss2 loxP/loxP conditional knockout mutants, and a total of 6 concurrent ageand strain-matched controls (46B6 and 26B6.Pdss2 loxP/loxP , respectively). Duplicate frozen aliquots of each strain were studied, when available. Mitochondria from single whole kidney homogenates were isolated in a similar fashion as described above for liver mitochondria with the exception that the renal capsule was first manually removed by squeezing. As yields of mitochondria were insufficient to permit polarography with all substrates as described above, kidney mitochondria quality was assessed for 3 representative strains by polarography with malate+glutamate and succinate to demonstrate robust state 3 rates, respiratory control ratios . 3 and . 2, respectively, and ADP/O malate+glutamate . 3.5.
Enzyme assays for citrate synthase and the partial reactions of electron transport were performed at 37uC in a total reaction volume of 1 mL using standard methods [33]. Absorbance changes were continuously monitored using the dual-beam mode of an OLIS-converted DW2a spectrophotometer. Sensitivity to enzymatic inhibitors was used to confirm assay specificity. Complex I+III was measured at 550 nm minus 540 nm (extinction coefficient 19.0 mM 21 cm 21 ) as rotenone-sensitive NADH-cytochrome c oxidoreductase. Complex II+III was measured at 550 nm minus 540 nm 19.0 mM 21 cm 21 ) as antimycin Asensitive succinate-cytochrome c oxidoreductase [7,8]. Aliquots of mitochondria were preincubated with succinate for 10 minutes before assay of complex II+III. Complex III was measured at 550 nm minus 540 nm (19.0 mM 21 cm 21 ) using decylubiquinol as antimycin A-sensitive ubiquinol-cytochrome c oxidoreductase. Complex IV was measured at 550 nm minus 540 nm (19.0 mM 21 cm 21 ) as azide-sensitive ferrocytochrome c oxidase. Citrate synthase was measured at 412 nm minus 360 nm (13.6 mM 21 cm 21 ) using 5,5-dithio-bis(2-nitrobenzoic acid) to detect free sulfhydryl groups in coenzyme A. Individual enzyme activities are reported in nmol/min/mg protein. As no statistically significant differences were detected in activities for any substrate between B6 and B6.Pdss2 loxP/loxP liver mitochondria controls, all control data (B6+B6.Pdss2 loxP/loxP ) were used for comparative analysis with each mutant strain. Comparison of means was done by paired t-test.