Yeast Cells Expressing the Human Mitochondrial DNA Polymerase Reveal Correlations between Polymerase Fidelity and Human Disease Progression*

Background: Mutations in the human mitochondrial DNA polymerase (Pol-γ) have been linked to diseases with varying severity and age of onset. Results: Yeast cells expressing human Pol-γ reveal a correlation of Pol-γ fidelity with human disease onset. Conclusion: Humanized yeast provides an efficient system to correlate biochemical defects in Pol-γ with physiological consequences. Significance: The Pol-γ-associated diseases may be caused by the low accuracy of Pol-γ mutants, not low rates of replication. Mutations in the human mitochondrial polymerase (polymerase-γ (Pol-γ)) are associated with various mitochondrial disorders, including mitochondrial DNA (mtDNA) depletion syndrome, Alpers syndrome, and progressive external opthamalplegia. To correlate biochemically quantifiable defects resulting from point mutations in Pol-γ with their physiological consequences, we created “humanized” yeast, replacing the yeast mtDNA polymerase (MIP1) with human Pol-γ. Despite differences in the replication and repair mechanism, we show that the human polymerase efficiently complements the yeast mip1 knockouts, suggesting common fundamental mechanisms of replication and conserved interactions between the human polymerase and other components of the replisome. We also examined the effects of four disease-related point mutations (S305R, H932Y, Y951N, and Y955C) and an exonuclease-deficient mutant (D198A/E200A). In haploid cells, each mutant results in rapid mtDNA depletion, increased mutation frequency, and mitochondrial dysfunction. Mutation frequencies measured in vivo equal those measured with purified enzyme in vitro. In heterozygous diploid cells, wild-type Pol-γ suppresses mutation-associated growth defects, but continuous growth eventually leads to aerobic respiration defects, reduced mtDNA content, and depolarized mitochondrial membranes. The severity of the Pol-γ mutant phenotype in heterozygous diploid humanized yeast correlates with the approximate age of disease onset and the severity of symptoms observed in humans.

Human mitochondria contain a 16.5-kb circular doublestranded DNA genome that encodes 13 of the proteins involved in oxidative phosphorylation. In humans, there are thousands of mitochondrial DNA (mtDNA) 3 copies per cell, so that mitochondrial genetics are complicated by heteroplasmy, the mixed population of mutant versus wild-type alleles. Mitochondrial DNA is replicated and repaired by the nuclearly encoded DNA polymerase (polymerase-␥ (Pol-␥)), which is a heterotrimeric complex that can be reconstituted in vitro (1) from the 140-kDa catalytic subunit (Pol-␥A) and a 55-kDa dimeric accessory subunit (Pol-␥B). The catalytic subunit contains both DNA polymerase and exonuclease proofreading activities (2), whereas the accessory subunit facilitates processive DNA synthesis (1) but may also play other roles during replication. The Pol-␥ holoenzyme functions in conjunction with the mitochondrial DNA helicase (3) and single-stranded DNA-binding protein (mtSSB), forming the minimal replisome (4,5).
Mutations in POLG lead to clinical symptoms of varying severity, including progressive external ophthalmoplegia (PEO), Alpers syndrome, parkinsonism, and other encephalomyopathies associated with mutations, deletions, or depletions of mtDNA (6). More than 200 point mutations in the polymerase have been correlated clinically with various mitochondrial diseases (7). The effects of a small subset of these mutations on the kinetics of DNA replication have been quantified, but different models have been proposed for how changes in enzyme kinetics lead to the accumulation of mtDNA defects, ranging from low fidelity to slow rates of replication or stalling of the replisome (8 -12). Moreover, many Pol-␥-associated mitochondrial disorders exhibit slow onset and a broad clinical spectrum, and the presence of the wild-type allele further complicates our ability to extrapolate from the observed biochemical defects of a mutant enzyme to a projected physiological effect (13).
Kinetic analysis of the effects of several point mutations on Pol-␥ fidelity and efficiency of replication, combined with the structure of human Pol-␥ (14), has provided molecular details regarding the primary effects of these disease-causing point mutations (8 -12). For example, the mutation H932Y at the polymerase active site is associated with an autosomal recessive form of PEO and SANDO (sensory ataxic neuropathy, dysarthria, and opthalmoparesis) (15). Pre-steady state kinetic studies (8) showed that the H932Y mutation reduces the efficiency (k cat /K m ) governing correct nucleotide incorporation 150-fold. Another active site mutation in Pol-␥, Y955C, was found in patients presenting "early onset autosomal dominant PEO" in their twenties (16). Kinetic analysis showed a 1300-fold reduction in k cat /K m while incorporating a dATP base opposite a template T (9,11).
A recent study suggested that the severity of the clinical symptoms correlates with reduced rates of DNA replication by the mutant forms of the enzyme (17). However, predicting the physiological consequences of biochemical defects measured in vitro is complicated by the interplay of wild-type and mutant alleles and unknown adaptive control pathways and repair processes in vivo. Therefore, it is important to study the effects of mutations in the context of mtDNA replication in heterozygous diploid cells in order to elucidate the molecular basis of the diseases and to distinguish the consequences of various biochemical defects observed in vitro.
Yeast has been proposed as a suitable in vivo model to provide relevant information on the mechanism of human mtDNA replication because of the significant homologies in the enzymes involved (for reviews, see Refs. 18 -21). Phylogenetic analysis has identified significant similarities between the DNA replication machinery of the T-odd bacteriophage and that for mtDNA replication in humans, plants, and yeast (22). Thus, evolutionary conservation and fundamental genetic similarities of mitochondria in yeast and humans make yeast a valuable tool for the investigation of mitochondrial function in normal and disease states (11,(23)(24)(25)(26)(27)(28).
Previous studies (24,27,28) using yeast as the model to mimic human mitochondrial disease were based on the homology of yeast mitochondrial DNA polymerase (MIP1) and human Pol-␥A (ϳ65% in polymerase and exonuclease domains). These studies relied on the simple viability screen to provide a simple yes/no answer as to whether a given mutation could produce a defective phenotype consistent with the predictions from human genetic analysis (27). However, mutations in MIP1 have not been characterized biochemically, and therefore correlations rely upon untested assumptions regarding the effects of mutations on enzyme activity. Indeed, mutations in MIP1 may not cause the same changes in DNA replication rate and fidelity as their human counterparts due to subtle changes in structure and kinetics in comparing the yeast and human enzymes (1,2,14,29). Therefore, the significance of studies on the physiological consequence of mutations in MIP1 may by questionable. In addition, not all important amino acid residues in human Pol-␥ have yeast orthologs. In particular, yeast does not encode the homolog of accessory subunit of Pol-␥ (Pol-␥B), and the accessory interaction domain is not conserved in yeast. Several pathogenic point mutations are found in these domains (30,31) in humans and cannot be tested using MIP1. Thus, despite evolutionary relationships between yeast and humans, one might question whether the human mtDNA polymerase could provide functional substitution for the yeast polymerase.
The "humanized" yeast system, with the yeast genes replaced by their human orthologs, has been developed to rapidly examine the pathology of human genetic variation, thanks to the powerful genetic tools available for yeast (32)(33)(34)(35). Here, we begin addressing important questions by developing "humanized" yeast expressing human mitochondrial DNA polymerase genes with wild-type and mutant alleles. We show that the human Pol-␥ can efficiently complement the mip1 knockout in yeast. We take advantage of the ability to grow yeast in both haploid and diploid states to resolve the primary effects of single point mutations in Pol-␥ and to investigate the complex interplay of mutant and wild-type genes within diploid cells.
The results for a series of mutations suggest that the observed clinical symptoms in humans may primarily arise from higher mtDNA mutation frequency due to the low enzyme fidelity, rather than from the slower rate of replication as suggested in prior studies (12,17). Although more extensive studies are needed to examine this preliminary correlation, the present study establishes the utility of the humanized yeast system to rapidly assess the possible physiological consequences of point mutations thought to be correlated with human disease.

EXPERIMENTAL PROCEDURES
Yeast Strains and Media-Medium preparation, techniques for culturing, and genetic manipulations of yeast were carried out according to standard procedures (36). Synthetic dextrose (SD), synthetic glycerol (SG), rich dextrose (YPD), and rich glycerol (YPG) media were used as indicated. Unless otherwise indicated, experiments were carried out with the strain BY4741 (MATa his3⌬0 leu2 ⌬0 met15 ⌬0 ura3 ⌬0). The MAT␣ type of haploid strain was generated by transformation with a centromeric plasmid pAG-GAL-HO expressing the HO endonuclease to initiate the mating type conversion (37). The mating type of strains was confirmed by its ability to grow on the minimal medium plate without any supplements after mating with the MATa haploid test strain (MATa HIS1 LEU2 MET15 URA5). The heterozygous diploid strains (mip1::Pol-␥A ⌬25 wild-type/ mip1::hPol-␥A ⌬25) for all mutants studied were obtained by isolation of zygotes after 4 h of mating between the MAT␣ type of haploid strain (mip1::hPol-␥A ⌬25 wild-type) with the MATa type of haploid strains (mip1::hPol-␥A ⌬25 mutants). The heterozygous diploid strains were verified by their inabilities to mate either with MATa or MAT␣ haploid strain.
Humanized Yeast Strain Construction-Our method for replacement of the MIP1 gene with the human Pol-␥A (Pol-␥A) gene is modified from the method described previously (38). Briefly, in order to generate the mip1 knock-out haploid strain, the wild-type BY4741 strain was transformed with a PCR-amplified Kluyveromyces lactis I-SceI-URA3 counterselectable marker cassette flanked by 60 bp of homology (primers P1 and P2; supplemental Table S1) to the yeast genomic region to be disrupted ( Fig. 1 and supplemental Table S1). The disruption of the MIP1 gene was done so that the native yeast mitochondrial localization signal sequence was retained at the N terminus of the gene along with the promoter. Primers (P3 and P4; supplemental Table S1) binding to the yeast genome ϳ200 bp outside of the homology region were used to confirm the knockout of the mip1 gene and were also used for verification by DNA sequencing. The ⌬mip1 strain was then transformed with plasmid pGAL1-I-SceI-G418 expressing an I-SceI endonuclease under the Gal promoter. To create a double-stranded break at the target site, overnight-grown cells in YPD medium were diluted into YEP-raffinose to 0.5 ϫ 10 6 cells/ml and allowed to grow for 6 h at 30°C to derepress the GAL1 promoter. Galactose was added at 1.5% final concentration in the growth medium to induce the expression of I-SceI for 1 h. Cells were pelleted for high efficiency transformation with a linear DNA amplified by PCR encoding the human Pol-␥A gene with 60-bp homology as chosen earlier for disruption of the MIP1 locus. The PCR (primers P5 and P6; supplemental Table S1)-amplified POLGA fragment has a 25-amino acid N-terminal deletion encoding a human mitochondrial targeting sequence (Pol-␥A ⌬25). The transformants were first plated on YPD-agar and allowed to grow overnight followed by replica-plating onto 5-fluoroorotic acid-agar plates to select for the loss of the URA cassette. PCR (primers P1 and P2) was carried out to check whether the URA Ϫ transformants were due to precise replacement of the URA marker cassette with the human gene or only to removal of the URA cassette (Fig. 1).
Plasmid Construction-Site-directed mutagenesis of Pol-␥A was performed using the QuikChange site-directed mutagenesis kit (Invitrogen) cloned in a pUC19 plasmid. The mutations were verified by DNA sequencing. The POLGB encoding the accessory subunit of human Pol-␥ was obtained the human ORFeome collection in the form of an entry clone, pDONR223-POLGB. PCR using primers (P7, P8, and P9; supplemental Table S1) was used to replace the human mitochondrial localization sequence (MLS) of POLGB with the yeast MLS. The engineered POLGB carrying the yeast MLS was then TOPOcloned into an entry clone vector, pENTR, followed by integrating the gene into the yeast destination vector pAG415Gal-ccdB under the Gal promoter or into the vector pAG416GPD-ccdB under the constitutive glyceraldehyde-3-phosphate dehydrogenase (GPD) promoter using the LR reaction (Gateway Technology).
Mitochondria Isolation, Co-immunoprecipitation, and Western Blots-Mitochondrial and non-mitochondrial fractions were collected with the yeast mitochondrial isolation kit MITOISO3 (Sigma-Aldrich) following the standard procedures provided by the manufacturer. The protein content was measured by Bradford assay. A total of 100 g of the protein was loaded into wells of a 10% polyacrylamide gel and transferred into a nitrocellulose membrane. Western blotting was carried out using a 1:2000 dilution of horseradish peroxide-conjugated anti-His 6 tag as the primary antibody. Tetramethylbenzidine substrate was used for visualization of antibody-bound protein bands.
For Ni-NTA-agarose co-immunoprecipitation experiments, 2 mg of mitochondria were solubilized in buffer containing 20 mM Tris, pH 7.5, 100 mM potassium acetate, 10% (v/v) glycerol, and 2% digitonin (w/v). The supernatants were incubated with 100 l of Ni-NTA-agarose resins for 1 h at 4°C, beads were washed, and bound proteins were eluted with elution buffer containing 250 mM imidazole. Samples were analyzed by SDS-PAGE and immunoblotting.
Growth Curve Measurement and Analysis-The cell growth was monitored at an absorption of 600 nm on the microplate reader (BioTek Instruments, Winooski, VT). Unless otherwise indicated, cells were seeded to 96-well plates at 1 ϫ 10 5 cells/ ml/well in the 150-l culture and grown at 30°C with continuous shaking. For cultures grown with glucose as the sole carbon source, doubling time during the exponential phase of growth before or after diauxic shift was calculated from the slope of the linear increase of the natural log of cell numbers. The lag time was estimated from the intersection of two linear fits of the time course. For cultures grown in conditions other than glucose as the carbon source, the growth curve was modeled as the initial lag phase followed by the exponential phase before the cells enter the stationary phase. To mimic chemostat conditions, cells were maintained in early exponential phase by successive dilution with fresh medium. Cells were allowed to grow until an A 600 of 0.2 and were diluted 10-fold into fresh medium. This process was repeated at least 6 -8 times.
Measurement of Petite Frequency, Mutation Frequency, and mtDNA Contents-Measurement of petite frequency and erythromycin-resistant (Ery R ) mutant frequency of strains harboring Pol-␥ or Pol-␥ mutant alleles was carried out based on the method described previously (27). The mtDNA contents were quantified relative to nuclear DNA contents using realtime quantitative PCR. Primers and probes were designed to specifically amplify mitochondrial encoded COX3 gene (P10 and P11; supplemental Table S1) and nuclearly encoded GAL4 gene (P12 and P13; supplemental Table S1). PCRs and data analysis were performed as described previously (39).
Confocal Microscopy-Confocal microscopy imaging was performed on a Zeiss LSM 510 laser-scanning microscope (Carl Zeiss Inc., Thornwood, NY). To visualize mitochondria, cells harvested from different conditions (Fig. 7A) were washed with water and resuspended in PBS buffer with 1 mM Ca 2ϩ and 0.5 mM Mg 2ϩ containing 100 nM MitoTracker-Red (Invitrogen) and 1 g/ml DAPI (Invitrogen). Cells were stained at room temperature with shaking for 30 min and then washed three times with PBS. Stained cells were mounted directly on microscope slides and examined by confocal microscopy.
Chronological Life Span Measurement-The cell viability assay was modified from the method described previously (40). Briefly, cells of each strain were grown in liquid SD medium in flasks at 30°C with shaking (250 rpm) to ensure aeration. Cells entered stationary phase after ϳ48 h, when the majority of cells stop dividing. Every 2 days, aliquots from the cultures were diluted accordingly and plated (200 -300 cells) onto solid SD medium plates. The plates were incubated at 30°C for 2 days, and the viability of each strain in the flask was measured as cfu. Viability at day 3 was considered to be the initial survival (100%), and day 3 was set to be time 0.
Measurement of ⌬m-To measure the membrane potential change of yeast cells, we used the fluorescent carbocyanine dye JC-1 (Invitrogen) to label the mitochondria membrane following the staining procedure provided by the manufacturer. Briefly, cells were harvested at times Fig. 7C legends, resuspended in 1 ml of PBS containing 5 M JC-1 probe, and incubated for 30 min at room temperature in the dark. Cells were then washed once in PBS and analyzed immediately by flow cytometry. Control cells with depolarized mitochondrial membranes were treated with 5 M carbonyl cyanide m-chlorophenyl hydrazone (Sigma-Aldrich) for 30 min at room temperature before JC-1 labeling. Photomultiplier settings were adjusted to detect red fluorescence ( em ϭ 590 nm) of JC-1 aggregates at channels FL-1 and green fluorescence ( em ϭ 529 nm) of JC-1 monomer at channels FL-2.
Protein Purification-Wild-type and mutant Pol-␥A were expressed in insect cells and purified to homogeneity as described previously (8). Pol-␥B was expressed in Escherichia coli and purified to homogeneity as described (1).
Nucleotide Incorporation Assays-Single nucleotide incorporation assays were performed with a RQF-3 rapid quench flow apparatus (KinTek Corp.) as described (9). In brief, 100 nM Pol-␥ exo ϩ enzyme was preincubated with 50 nM 5Ј-32 P-labeled DNA template (25-mer/45-mer duplex) on ice for 10 min, and the complex was rapidly mixed with 12.5 mM Mg 2ϩ and 50 M dATP for variable time at 37°C before the reaction was quenched by mixing with 250 mM EDTA. The products were resolved on a 15% denaturing polyacrylamide sequencing gel, and then the dried gel was exposed to a phosphor screen (Molecular Dynamics). Radioactive bands were detected with a Typhoon 9400 scanner (GE Healthcare) and quantified with ImageQuant software (GE Healthcare). The data have been fit globally using the KinTek Explore program (Kin-Tek Corp.) to Scheme 1, and kinetic parameters k cat and k cat /K m were derived for Scheme 1 by the simple relationship k cat ϭ k pol and K m ϭ K d,app . where ED 25 represents a preformed complex of enzyme with 25/45 duplex DNA. Ground state nucleotide (N) binding is represented by the term K d,app , whereas the chemistry step to form the complex of enzyme with 26/45 duplex (ED 26 ) and pyrophosphate (PP i ) is irreversible and described by k pol .
Excision Reactions-Enzyme exo ϩ (100 nM) was preincubated with 75 nM DNA. The reaction was initiated by mixing it with Mg 2ϩ for variable time at 37°C and was quenched by mixing with 250 mM EDTA. The loss of full-length substrate primer due to exonuclease digestion was plotted against time and fit to a single or double exponential indicated in the Fig.  4 legend.

RESULTS
The Human Pol-␥A Complements Loss of Saccharomyces cerevisiae MIP1-The yeast nuclear MIP1 gene encodes the mtDNA polymerase. A ⌬mip1 deletion mutant is respirationdeficient but is able to obtain energy to sustain growth from glycolysis when provided an appropriate carbon source. We took advantage of this conditional phenotype to enable replacement of the MIP1 gene with a Ura3 ϩ marker cassette containing an I-Sce1 endonuclease site while still retaining the upstream sequence encoding the MLS (Fig. 1) (see "Experimen-tal Procedures"). The marker cassette was then replaced chromosomally by the human POLGA wild-type (or mutant) gene fused to the yeast MLS encoding gene fragment.
To examine whether the human POLGA could functionally substitute MIP1 in S. cerevisiae, we first assayed the localization of the human Pol-␥A protein by immunofluorescence using anti-His 6 antibodies. The Pol-␥A protein was detected in the purified mitochondrial fraction ( Fig. 2A), indicating that the human Pol-␥A protein was expressed and successfully targeted to mitochondria by the yeast MLS. We then compared growth curves of the humanized and wild-type strains. In haploid cells, a deletion of mip1 resulted in an inviable strain when grown in synthetic medium with 2% galactose as the carbon source ( Fig. 2B) and formed 100% petite colonies when grown in glucose medium ( Table 1 and Fig. 2C). The human gene rescued the petite phenotype of the ⌬mip1 strain (Table 1 and Fig. 2C). These strains can metabolize galactose to sustain growth with a doubling time of 3.2 Ϯ 0.02 h, which is only slightly less than that of the wild-type strain (2 Ϯ 0.3 h). However, growth of the POLGA strain showed a longer lag time of 18 Ϯ 1.1 h, as opposed to the wild-type strain (7.2 Ϯ 0.2 h) in recovering from stationary phase.
Pol-␥B Is Required for Complete Complementation-We sought to investigate whether the inefficient complementation was due to the absence of the human accessory protein Pol-␥B. In vitro studies have shown that Pol-␥B forms a stable protein complex in its dimeric state with Pol-␥A (41), stimulates DNA binding to Pol-␥ by 3.5-fold, and increases the polymerization rate by 4.5fold (1). We expressed human Pol-␥B fused to the MIP1 MLS in yeast cells using a CEN plasmid, pAG415GAL-POLGB, under the control of a GAL-inducible promoter. The empty plasmid was used as the negative control. As shown in Fig. 2B, expression of Pol-␥B in the Pol-␥A-expressing strain dramatically shortened the lag time from 18 h to 6.9 h and slightly increased the growth rate. Surprisingly, in the presence of human Pol-␥B, the wild-type strain expressing yeast MIP1 grows at a rate that is comparable with that seen with Pol-␥A ( Fig. 2B). Thus, when one takes into account a slight inhibition of grown caused by expressing Pol-␥B, the human Pol-␥A appears to efficiently complement the mip1 knockouts.
The interaction of the Pol-␥A with Pol-␥B in vivo was analyzed with mitochondrial extracts from cells expressing Pol-␥A with His 6 tag and Pol-␥B (Fig. 2D). When mitochondrial extracts containing Pol-␥A-His 6 were pulled down by Ni-NTA, we detected a large portion of Pol-␥B (52 kDa band) co-immunoprecipitated with Pol-␥A-His 6 . Control experiments using mitochondrial extracts from cells expressing Pol-␥B only showed that Pol-␥B was not retained by Ni-NTA. Likewise, the 52-kDa protein was absent in the mitochondrial extracts from cells expressing Pol-␥A-His 6 only. The results demonstrate the specificity of the interaction between Pol-␥A with Pol-␥B in the cell.
To further examine the functional complementation of mip1 deletion by human POLG, we measured the mtDNA content and mtDNA mutation frequency. The relative mtDNA content was measured using qPCR by comparing the ratio of copy numbers of the mitochondrially encoded COX3 gene to the nuclearly encoded GAL4 gene, as described previously (39). Compared with wild-type strains, the POLGA haploid strain contained about 30% FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9 of the mtDNA during the exponential growth phase; this value increased to about 52% when Pol-␥B was also present (Fig. 3C).

Humanized Yeast Cells Model Human Disease
The frequency of mtDNA mutation was assayed by selecting for resistance to erythromycin (Ery R ), which is conferred by point mutations in the mtDNA-encoded 21 S rRNA gene (42). There are no reported nuclear mutations in yeast that result in Ery R ; therefore, the frequency of Ery R can be used to estimate the frequency of mtDNA mutation. The POLG haploid cells showed no significant increase in mutation frequency compared with the MIP1(WT) haploid cells (Table 1). We next directly visualized mitochondria using confocal microscopy by staining nuclear and mitochondrial DNA by DAPI and mitochondria using MitoTracker Red. As shown in Fig. 3E, both mtDNA and functional mitochondria were retained in haploid cells expressing human Pol-␥, further supporting the functional complementation of mip1 by the human POLG gene.
Mutations in Pol-␥ Cause Oxidative Growth Defects in Haploid Yeast Cells-Using the humanized yeast model, we can now evaluate the effects of specific disease-associated point mutations in human Pol-␥ in vivo. The mutations considered in this study and their associated diseases are listed in Table 1. We assayed several features for each mutant: oxidative growth phenotype, mtDNA content, mitochondrial function, petite phenotype percentage, and mtDNA mutation frequency (Ery R phenotype). As shown in Fig. 3A, all six mutants except the ⌬mip1 strain showed similar doubling times in the initial exponential growth phase on glucose, where ATP is primarily generated by glycolysis. After the diauxic shift, when cells are required to obtain energy from aerobic respiration, no apparent growth was observed for any of the mutant strains compared with the wild-type POLG exo ϩ strain (which showed a doubling time of 37 h).
To provide a direct test of mitochondrial function, cells were grown on glycerol, a non-fermentable carbon source. Of the six mutant haploid strains, five strains (S305R, H932Y, Y951N, Y955C, and ⌬mip1) were unable to grow on glycerol (Fig. 3B), whereas the exonuclease-deficient (exo Ϫ ) strain grew poorly on glycerol ( Fig. 3B) with doubling times of 21 h, compared with 4.5 h for wild-type Pol-␥ exo ϩ . The maximum cell densities obtained with exo Ϫ mutant, determined by A 600 , were dramatically lower than that of the wild-type exo ϩ strain (Fig. 3B). Moreover, all mutant haploid showed significantly increased frequencies of petite formation compared with the Pol-␥ exo ϩ wild-type strain (Table 1).

Oxidative Growth Deficiency Is Associated with mtDNA Depletion and Mutation in Haploid Cells Harboring Mutant
Pol-␥-Because mtDNA depletion is a common feature of Pol-␥-related diseases, we hypothesized that Pol-␥ mutant strains would lead to decreased mtDNA content. To measure mtDNA copy number, we performed real-time qPCR to quantify mito- chondrial genes in total DNA extractions from haploid mutant strains. In all strains, large increases in petite frequency were correlated with significant decrease or total depletion of mtDNA (Fig. 3C).
We also examined the effects of mutations in Pol-␥ on haploid strains by confocal fluorescence microscopy following DAPI staining of both nuclear and mtDNA and MitoTracker Red staining of mitochondria. In haploid strains, mtDNA was detected in the wild-type Pol-␥ and exo Ϫ cells (Fig. 3E). However, no apparent mtDNA staining could be observed in the cells harboring S305R, H932Y, Y951N, or Y955C mutations in Pol-␥. Consistent with mtDNA depletion, no functional mitochondria were observed in these cells. Although we have not tested the effects of POLG mutations on nuclear DNA replication as measured by GAL4 gene levels, the data clearly show a selective reduction in mtDNA replication relative to nuclear DNA.
The haploid cells carrying exonuclease deficient Pol-␥ (exo Ϫ ) showed the largest increase (Table 1) in the Ery R mutation frequency, supporting the significant role of proofreading activity of Pol-␥ in mtDNA maintenance. DNA sequencing analysis of the cloned 21 S rRNA gene (primers P14 and P15; supplemental Table S1) from Ery R resistant   FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9 exo Ϫ cells mapped mutations at nucleotide position 1951 (A to T, A to G, A to C, and a single G insertion between 1950 and 1951). The Pol-␥ showed an average discrimination factor of 3 ϫ 10 5 against mismatched nucleotides, which equals a mutation rate of 3.3 ϫ 10 Ϫ6 /nucleotide/genome replication in the absence of the proofreading activity (8,9). In accordance with in vitro measurements, the mutation rate of exo Ϫ cells after ϳ4 -5 generations is 1.7 ϫ 10 Ϫ5 (Table 1). The mutation rate increased over time by maintaining the cells in exponential growth phase (Fig. 3D), and accumulated mtDNA mutation may lead to the observed decrease in mtDNA contents (Fig. 3D).

Humanized Yeast Cells Model Human Disease
Oxidative Growth Deficiency of Haploid Cells Is Correlated with the Low Fidelity of Pol-␥ Mutants-The haploid strains carrying H932Y, Y951N, or Y955C mutations form 100% petite cells (Table 1). Kinetic analysis of purified Pol-␥ mutants showed a ϳ150-, 11-, and 1300-fold decrease in the rate of dATP incorporation for H932Y, Y951N, and Y955C, respec-tively, compared with the wild-type enzyme (1,8,9). 4 Strikingly, the fidelity of Pol-␥ mutants as calculated from the ratio of k cat /K m values for correct (dATP:dT) versus mismatched (TTP: dT) base pairs dropped ϳ100-fold for Y955C and ϳ3-fold for H932Y (8,9). Moreover, unlike wild-type enzyme, both H932Y and Y955C mutants can quickly incorporate multiple mismatched base pairs in vitro, suggesting the importance of enzyme fidelity and the ability of the enzyme to sense and remove mismatched base pairs (8,9).
The S305R haploid cells are not viable in glycerol medium, which requires oxidative metabolism for use as a carbon source.  The cells were continuously grown in SD liquid medium (Ϫura) and maintained in exponential growth phase by successive dilutions to fresh medium (see "Experimental Procedures" for details). Aliquots were taken at the indicated time to measure the relative mtDNA contents using qPCR as described above or to determine the Ery R mutant frequency (red symbols) following methods described previously (27). The ratio of COX3/GAL4 was normalized to 1 for wild-type Pol-␥ cells. E, representative confocal microscopy images of haploid cells with mitochondria stained with MitoTracker Red and DNA stained with DAPI. The DAPI-stained mtDNA was revealed by a network of punctuate dots located at the cell periphery for wild-type MIP1, wild-type Pol-␥-exo ϩ , and Pol-␥-exo Ϫ cells. The green spots present in the cytoplasm are nuclear DNA. For Pol-␥-S305R, Pol-␥-H932Y, and Pol-␥-Y955C cells, MitoTracker Red diffused in the cytoplasm, suggesting loss of functional mitochondria.
(k cat /K m ϭ 40 M Ϫ1 s Ϫ1 ). It is unlikely that this small reduction of the replication rate would lead to the complete loss of functional mitochondria. Because the Ser-305 residue is located near the partitioning loop between the polymerization active site and exonuclease active site, we next examined the proofreading activity by S305R Pol-␥. Compared with the wild-type enzyme, the S305R mutant showed a 10-fold reduction in the rate of excision of single-stranded DNA and a ϳ3-fold decrease in the rate of excision of mismatched base pairs (Fig. 4, B and C). Therefore, the deficiency of proofreading activity coupled with the slow replication rate by S305R Pol-␥ could account for the oxidative growth defect of S305R haploid cells.
Collectively, each mutation that we examined presents a severe defect in mtDNA replication in haploid yeast cells, consistent with the observation that these mutations are never observed in the homozygous state in humans, presumably due to the lethality of the mutations. We next examined the effect of these mutations in the diploid heterozygous cells expressing one wild-type and one mutant gene.
Heterozygous Diploid Cells Mimic the Compound Heterozygotes of Human Pol-␥ Mutations-We generated heterozygous diploid strains of yeast by mating ancestor haploid cells harboring the wild-type POLG gene with ones harboring mutant POLG gene. The Pol-␥B is always co-expressed in all heterozygous diploid cells under the constitutive GPD promoter. The doubling times of cells grown on glucose are quite similar for all heterozygous mutants in their initial exponential growth phase (Fig. 5A), although the mutants differed in growth rates after the diauxic shift.
To directly monitor the respiratory function of the heterozygous diploid mutant strains, cells were grown on glycerol. As shown in Fig. 5B, the mutant strains started out with the same growth rate as the wild-type strains until the OD reached ϳ0.4, when most mutants began to grow much more slowly than the wild-type strain. Interestingly, the heterozygous strain (POLG WT/mip1) grew with the same doubling time as the wild-type cells (POLG WT/POLG WT) although it showed a longer lag time (Fig. 5B), suggesting a possible underlying mechanism to control mtDNA copy number.
In the heterozygous diploid strains, qPCR analysis of the mtDNA content indicated that mtDNA was retained to different extents, but all mutant strains grown in the early exponential phase still retained Ͼ50% relative to the POLG(WT)/POLG(WT) diploid strain (Fig. 5C). Consistent with the qPCR results, we also detected the presence of DAPI-stained mtDNA in all heterozygous mutant cells (Fig. 5D), indicating that the wildtype Pol-␥ is able to efficiently replicate the mtDNA to maintain the functional mitochondria initially. Thus, deficiencies in rates of polymerization by the mutant forms of Pol-␥ may not contribute significantly to the observed mtDNA depletion. Rather, the wild-type enzyme appears to compensate for deficiencies in the rates of replication catalyzed by the mutant forms of the enzyme in heterozygous cells.

Long Term Growth in Chemostatic Conditions Leads to Accumulation of Mitochondrial Damage in Heterozygous Mutant
Diploid Cells-All mutant heterozygous diploid cells showed a growth deficiency after the cultures reached an A 600 greater than 0.4. We propose that this phenomenon may indicate that the mtDNA replication is eventually impaired by the accumulation of mutations deriving from the mutant Pol-␥. However, interpreting the changes in growth rate is complicated by the complex nutrient changes occurring in the medium as the cultures approach stationary phase. Therefore, to test our hypothesis, we performed a continuous growth experiment by maintaining the cells in exponential growth phase for several days (see "Experimental Procedures" for details). This was accomplished by successive dilutions of cultures into fresh medium; when each culture reached an A 600 of Ͼ0.2, it was diluted by the addition of fresh medium to an A 600 of 0.02 to begin a new  Table S1), and Mg 2ϩ was added to initiate the excision reaction. The percentage of 25-mer remaining was plotted against time and fit to a double exponential to yield a fast excision rate of 15.4 Ϯ 1.1 s Ϫ1 for WT enzyme (E) and 5.5 Ϯ 0.7 s Ϫ1 for S305R enzyme (F). FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9 growth phase (Fig. 6A). We reasoned that by avoiding shifts in metabolism due to changes in the medium, the observed growth defects should reflect intrinsic imparities in mtDNA replication. As shown in Fig. 6A, after 40 h of continuous growth (one dilution cycle) and ϳ10 generations from the ancestor, the Y955C and Y951N heterozygous diploid strains began to grow notably more slowly than the wild-type cells. After ϳ30 generations, all heterozygous diploid mutant cells showed growth deficiencies on glycerol. The doubling times of all mutants for each round of growth are summarized in Fig. 6B, demonstrating the early onset of growth defect for Y995C and Y951N heterozygous diploid strains.

Humanized Yeast Cells Model Human Disease
To understand the molecular basis for this oxidative growth phenotype, we measured the mtDNA content of cells from different generations (Fig. 6C). For heterozygous diploid cells (exo Ϫ ,  H932Y, Y955C, and Y951N), mtDNA content reduced with successive generations; nonetheless, levels of mtDNA remained above 30% relative to the wild-type Pol-␥ strain even after 30 generations, indicating that mtDNA depletion may not be the only factor leading to the slower growth rate.
Onset of Growth Defects Is Associated with Mitochondrial Membrane Depolarization-For Y955C mutants, we observed dramatically decreased fluorescence from MitoTracker Redlabeled mitochondria with successive generations (Fig. 7A), possibly suggesting mitochondrial membrane depolarization because MitoTracker Red accumulation depends upon a functional mitochondrial membrane potential. To test this idea, we labeled the cells with JC-1, which undergoes an emission spectrum change from 590 nm (JC-1 aggregates) to 530 nm (JC-1 monomer) upon membrane depolarization. Ratiometric measurements of the red (monitored at FL-1 channel) to green (monitored at FL-2 channel) JC-1 fluorescence indicate ⌬m (44). Flow cytometry (Fig. 7B) revealed a progressive population shift of cells with high fluorescence intensity in the FL-1 channel toward low fluorescence intensity as a function of the cell generations for Y955C heterozygous diploid strains, consistent with our observations by confocal microscopy (Fig. 7A). We performed similar experiments for other heterozygous diploid mutants and have plotted the relative change of mean FL-1 fluorescence for each heterozygous diploid mutant at different generations, as shown in Fig. 7C. The Pol-␥ Y955C, Y951N, and exo Ϫ mutants showed more rapid decline in ⌬m compared with other mutants, which is correlated with a higher mutation frequency of these mutants and precedes the reduction in mtDNA content (Fig. 6C).
Mutation phase and has been proposed as a model of aging in postmitotic cells of higher organisms (45). We measured the viability of the aged yeast cells growing on glucose (Fig. 8A). As expected, the CLS values of mutant strains (exo Ϫ , H932Y, and Y955C) were significantly shorter than that of wild-type Pol-␥ exo ϩ strain. For the H932Y heterozygous diploid strain, roughly 50% of cells died 3 days after entering stationary phase. Interestingly, the mip1/POLG exo ϩ strain showed the same CLS as wild type, and S305R mutation caused only minor effects on CLS. To further explore the role of mitochondrial function on aging, we measured the mtDNA content of cells as a function of time in stationary phase. As shown in Fig. 8B, the significant age-dependent mtDNA content reduction was detected for all mutants. The measured mtDNA damage (mutation and depletion) is expected to in turn affect mitochondrial membrane function. We then calculated the relative ⌬m using the ratio of mean intensity of FL-1 over the mean intensity of FL-2 and plotted it against the aging time. As shown in Fig. 8C, the depolarization of mitochondria for mutants shared a similar trend with the viability assay (Fig. 8A), indicating that mutant heterozygous diploid cells showed a progressive decline in function manifesting in loss of ⌬m, which appeared to precede the loss of mtDNA content.
Human Disease Onset Is Correlated with Time of Onset of Growth Defects in Yeast-Mitochondrial disorders stemming from mutations in the Pol-␥ show widely varying severity of symptoms and differing delayed onset, leading to the postulate that mitochondrial disorders result from the accumulation of DNA damage (13). Here we have tested this theory using diploid strains of humanized yeast harboring mutations associated with heritable human diseases. We have sought to test whether the age of onset in humans is attributable more to reduced rates of polymerization or increased mutation frequency. Although accumulation of deletions may also play a role, we have not fully quantified those effects in this study.
As shown in Fig. 9A, for the mutants characterized here, the yeast mean survival time is negatively correlated with mutation frequency, measured from the rates of emergence of erythromycin resistance. There is no correlation between the reduced k cat /K m values measured for purified enzymes harboring different mutations and the yeast mean survival time or the age of disease onset observed in humans (Fig. 9, B and C). However, there is an apparent correlation between the onset of growth defects in yeast and the age of disease onset in humans (Fig. 9D) for those mutations with good evidence supporting a cause/effect relationship based upon human genetic studies.

DISCUSSION
In this paper, we sought to provide a link between the measurable biochemical defects caused by specific point mutations in Pol-␥ with their associated physiological consequences and resulting phenotypes described clinically. Studies from our laboratory and others have quantified the effects of Pol-␥ point mutations on efficiency and fidelity of DNA replication in vitro, suggesting that accurate biochemical measurements can provide a basis to rationalize clinical observations (8 -11). However, there are significant complexities in extrapolating from quantitative measurements of the effects of point mutations on DNA replication rate and fidelity in vitro to the final physiological consequences in humans. In particular, the interplay of wild-type and mutant alleles (recessive/dominant mutation) raises many questions that cannot be answered without studying the effects of mutations in the context of mitochondrial replication within living cells. Here we show indeed that the wild-type allele largely compensates for defects in the mutant forms of Pol-␥, especially those defects leading to slower rates of polymerization.
Using the humanized yeast as a model, we have provided a novel analysis to correlate the physiological consequences of point mutations in the human mitochondrial DNA polymerase with measurements of activity in vitro. Our results demonstrate that human Pol-␥ can efficiently complement mip1 deletion mutants in yeast and that complementation can be improved by the addition of the Pol-␥B accessory protein. The slight growth deficiency of humanized yeast as compared with the wild type (Fig. 2B) may be in part due to the decrease in the polymerization rate of human Pol-␥ at 30°C compared with 37°C. Nonetheless, humanized yeast haploid cells showed a growth rate and spontaneous mtDNA mutation rate comparable with that of MIP1 wild-type cells (Fig. 2B and Table 1), indicating that human Pol-␥ can function together with the endogenous yeast mitochondrial DNA replicative components at the replication fork for accurate DNA synthesis.
Notably, the helicase involved in mtDNA replication in yeast has yet to be identified, and the yeast genome does not encode a homologue of the human mitochondrial DNA helicase. Although two DNA helicases (Pif1p and Hmip) are imported into yeast mitochondria, they are primarily linked with mtDNA recombination and maintenance rather than in DNA replication (46,47). Thus, future study is needed to identify the endogenous helicase that functions with human Pol-␥ during mtDNA replication in humanized yeast. Despite differences in mtDNA structure and replication models between human and yeast, common fundamental mechanisms of replication are retained sufficiently to support replication. Our data provide tangible support for the functional conservation of enzymes required for mitochondrial DNA replication.
The complementation by Pol-␥ enables us to study the effects of Pol-␥ mutations. Three of the mutations chosen for study (H932Y, Y951N, and Y955C) affect the Pol-␥A polymerase active site. The H932Y mutation reduces the specificity constant governing correct nucleotide incorporation 150-fold but does not significantly affect the maximal rate of polymerization (8). The Y955C mutant of Pol-␥ showed an ϳ8-fold decrease in maximal rate of polymerization and a 155-fold increase in the K m for correct incorporation, thus resulting in a ϳ1300-fold decrease in efficiency as measured by k cat /K m (9). Mutation Y951N in Pol-␥ caused an 11-fold reduction in the rate of replication. 4 In accordance with the in vitro data, haploid cells harboring either the H932Y, Y951N, or Y955C mutation in Pol-␥ were unable to respire using glycerol as the sole carbon source (Fig. 3B). The suspicion that these cells had lost their mitochondria was confirmed by the absence of DAPI and MitoTracker Red staining by confocal microscopy (Fig. 3E) and also by qPCR (Fig. 3C). We were unable to examine the mtDNA mutation frequency for these three mutations in haploid cells due to their inabilities to grow on glycerol. Nevertheless, in vitro measurements of H932Y and Y955C Pol-␥ mutants show that they readily extended multiple mismatches to decrease the overall fidelity by suppressing the contribution of the proofreading exonuclease (8,9). In heterozygous diploid cells, mtDNA mutation frequency (measured after ϳ5-6 genera- tions) increased to ϳ70-fold and ϳ20-fold for mutants carrying H932Y and Y955C mutation, respectively. Thus, decreased replication fidelity may be a major consequence of these point mutations.
To further demonstrate the significant role of enzyme fidelity in mtDNA maintenance, we studied mutations D198A/E200A and S305R, which are located in the exonuclease and linker domains of the polymerase, respectively. Exonuclease activitydeficient (D198A/E200A) haploid cells exhibited a ϳ70-fold increase in Ery R frequencies (Table 1), although cells could grow poorly in glycerol (Fig. 3B). Continuous growth in glycerol over ϳ40 generations eventually results in 100% petite cells, due to the accumulation of mtDNA mutations and deletions, possibly compounded by the accelerated oxidative stress. The biochemical characterization of S305R mutation in Pol-␥ revealed a 3-10-fold decrease in the proofreading exonuclease excision rate of mismatched base pairs and a ϳ3-fold reduction in the replication rate (Fig. 4). The S305R mutation in haploid yeast was quite detrimental, giving rise to nearly 100% petite cells and a complete loss of functional mitochondria (Table 1 and Fig. 3E).
Our data are consistent with studies in humans, which have shown that mtDNA mutations accumulate with aging in several tissues of humans (48,49) and subsequently cause mtDNA deletion and respiration chain deficiency. The significant role of high mtDNA mutation rates in mitochondrial dysfunction is also supported by the mouse model carrying homozygous Pol-␥ exo Ϫ mutations (50, 51), which resulted in a premature aging phenotype.
It is reasonable to postulate that both the replication rate and the enzyme fidelity may lead to the observed phenotype in haploid cells. Nevertheless, the role of the replication rate appears to be minor, due to the compensation by the wild-type enzyme in heterozygous diploid cells (Fig. 5), as reflected by the functional mitochondria and retention of over 50% of mtDNA content relative to wild type. Most strikingly, all mutant heterozygous diploid cells grown in glycerol showed a lag time and growth rate similar to the wild-type cells at the beginning of growth (Fig. 5B), in contrast to the much longer lag time of mip1/POLG(WT) cells. It has been reported that genes involved in mitochondrial function and protection from oxidative stress are significantly up-regulated for the entry into and exit from the stationary phase (52). The mip1/POLG(WT) cells contain ϳ50% mtDNA relative to the POLG(WT)/POLG(WT) cells (Fig. 5C) and thus may require a longer time to up-regulate POLG expression and to synthesize mtDNA to pass the transition. For heterozygous diploid mutant cells, both wild-type and mutant copy are likely to be actively involved in mtDNA replication. However, mtDNA mutations will accumulate over time and may reach a threshold sufficient to generate a dominant (negative) physiological effect (Fig. 5B).
Our working hypothesis is further supported by the continuous growth experiments performed by successive dilutions into fresh medium (Fig. 6A) to circumvent complex changes in metabolism as cells enter stationary phase. Cells harboring the Y951N mutation exhibited much earlier onset of growth defects than for the S305R mutation despite their similar replication rates at physiological nucleotide concentrations ( 6A). 4 This again argues for the minor role of the replication rate, suggesting that the interplay between wild-type and mutant enzyme determines the overall observed rate and accuracy of mtDNA replication. In particular, the S305R mutant protein showed a ϳ20-fold decrease of the apparent DNA binding affinity compared with wild type. Therefore, the detrimental effects of the S305R mutant enzyme could be suppressed by the presence of wild-type protein in vivo due to its weaker DNA binding affinity. In contrast, Y951N mutant may compete with wild type on the replication origin site due to a comparable DNA binding affinity, 4 resulting in early onset of the growth defect. Similar dominant effects were also observed for Y955C mutants (Fig. 5, A and B) (9,11). Our analysis suggests that the primary effect leading to early onset of growth defects is likely to be the increased mutation frequency, not an decreased polymerization rate. In Fig. 9A, we show a direct correlation between mutation frequency and the yeast mean survival time. To approximate this dependence, we have fit the correlation curve to a hyperbolic function, but other functions may be more appropriate. The correlation is not expected to be linear, but it does predict a maximum survival time in the limit where the mutation frequency approaches zero, suggesting other physiological constraints on survival. At the opposite limit, one can reasonably expect that survival time will asymptotically approach zero as the mutation frequency increases. Most importantly, we observed a clear correlation of the onset of the phenotypes for heterozygous diploid cells with the disease onset age (Fig. 9D), which is independent of the enzyme's replication efficiency as defined by the k cat /K m value (Fig. 9C). Note that we have excluded S305R from the analysis in Fig. 9D because there is no clear link between the single point mutation and the heritable disease; rather, it appears that other mutations in trans contribute to the observed clinical phenotypes (13,53). Our results imply that the single mutation S305R, heterozygous with wild-type POLG, is unlikely to be significantly detrimental because the wild-type allele sufficiently complements defects in replication. Thus, our methods can help to sort out the often complex human genetic analysis in looking for correlations between point mutations and human disease. Further work is planned to examine the effects of multiple mutations.
POLG mutations exhibit a broad clinical spectrum of mitochondrial disorders. There is no apparent correlation between observed yeast mtDNA content with the disease severity or the age of disease onset in humans (13). In accord with the clinical data, the mtDNA content reduction did not seem to completely correlate with the progressive growth deficiency in heterozygous diploid cells (Fig. 6, A and C), suggesting that other factors are also causing the mitochondrial damage (i.e. mutations and mtDNA deletion). Our data suggest that reductions in mtDNA content may lag behind the initial accumulation of mutations and deletions that are likely to be the primary cause of disease phenotypes. The mitochondrial membrane potential is another way of evaluating the mitochondrial function because it depends on an active electron transport chain. Accordingly, we observed that mitochondrial depolarization was clearly associated with the oxidative growth defects (Fig. 7C) in continuous growth experiments and preceded the observed reduction in mtDNA content. The mitochondrial depolarization may be due to mutations in ATP synthase as described in a previous study (54), although our results could not differentiate whether the mitochondrial depolarization is a marker of the progressive oxidative growth defects or the mitochondrial depolarization is causing the latter phenotype.
The mutations linked to PEO mostly affect the postmitotic tissues and are correlated with the mtDNA deletions and depletions (55,56). Measurements of CLS in heterozygous diploid cells can provide information on the effects of these mutations in non-dividing cells. The heterozygous diploid cells harboring PEO-associated mutations (H932Y or Y955C) significantly curtailed CLS (Fig. 6A). Postmitotic cells have much lower levels of dNTP pools, and the negative effects of H932Y and Y955C mutation in mtDNA synthesis may begin to dominate due to the 150-or 160-fold increase in the K d for correct nucleotide binding, respectively. The slower rate of nucleotide incorporation may cause polymerase stalling that can increase deletions or mutagenesis (Table 1), which could result in dysfunctional, depolarized mitochondria that would be targeted for degradation by autophagosomes (43).
As with many model systems, in this case, one must ponder how observations of cell growth in yeast over the period of a few days can be predictive of the progression of diseases in humans that develop over the course of years. In part, the faster onset of the phenotype in yeast could be attributable to growth on a single carbon source requiring oxidative metabolism, which could add additional oxidative stress and demands on mtDNA replication. Alternatively, there could be repair mechanisms or other adaptive responses in human tissues that extend the lag period leading to the onset of symptoms. In either case, our studies provide good evidence in support of the utility of the yeast system, and further studies using this system should afford a novel method to unravel the complex series of events from biochemical defect to physiological symptoms.