Synergistic Effects of the in cis T251I and P587L Mitochondrial DNA Polymerase γ Disease Mutations*

Human mitochondrial DNA (mtDNA) polymerase γ (Pol γ) is the only polymerase known to replicate the mitochondrial genome. The Pol γ holoenzyme consists of the p140 catalytic subunit (POLG) and the p55 homodimeric accessory subunit (POLG2), which enhances binding of Pol γ to DNA and promotes processivity of the holoenzyme. Mutations within POLG impede maintenance of mtDNA and cause mitochondrial diseases. Two common POLG mutations usually found in cis in patients primarily with progressive external ophthalmoplegia generate T251I and P587L amino acid substitutions. To determine whether T251I or P587L is the primary pathogenic allele or whether both substitutions are required to cause disease, we overproduced and purified WT, T251I, P587L, and T251I + P587L double variant forms of recombinant Pol γ. Biochemical characterization of these variants revealed impaired DNA binding affinity, reduced thermostability, diminished exonuclease activity, defective catalytic activity, and compromised DNA processivity, even in the presence of the p55 accessory subunit. However, physical association with p55 was unperturbed, suggesting intersubunit affinities similar to WT. Notably, although the single mutants were similarly impaired, a dramatic synergistic effect was found for the double mutant across all parameters. In conclusion, our analyses suggest that individually both T251I and P587L substitutions functionally impair Pol γ, with greater pathogenicity predicted for the single P587L variant. Combining T251I and P587L induces extreme thermal lability and leads to synergistic nucleotide and DNA binding defects, which severely impair catalytic activity and correlate with presentation of disease in patients.

Human mitochondrial DNA (mtDNA) 2 is replicated and repaired by DNA polymerase ␥ (Pol ␥), which consists of a 140-kDa catalytic subunit (encoded by POLG at nuclear chromosomal locus 15q25) and a 55-kDa accessory subunit that forms a dimer (encoded by POLG2 at nuclear chromosomal locus 17q24.1) (1-3). The catalytic subunit contains the N-terminal exonuclease domain connected by a linker region to the C-terminal polymerase domain. The catalytic subunit has DNA polymerase, 3Ј 3 5Ј exonuclease, and 5Ј-deoxyribose phosphate lyase activities (4,5). The accessory subunit functions to enhance polymerase processivity by increasing affinity of the catalytic subunit for DNA (6 -8).
Deletions and point mutations in mtDNA as well as depletion of mtDNA are associated with several mitochondrial disorders and aging (9 -11). Many of these deleterious effects on mtDNA are caused by defects in the nuclear genes encoding mtDNA replication proteins, such as POLG. To date, over 300 disease-associated mutations in POLG are listed in the Human DNA Polymerase Gamma Mutation Database. Mitochondrial diseases broadly vary and include progressive external ophthalmoplegia (PEO), Alpers syndrome, ataxia neuropathy syndrome (ANS), myocerebrohepatopathy spectrum disorders, myoclonus epilepsy myopathy sensory ataxia, parkinsonism, and male infertility (12)(13)(14)(15). In general, all tissues/organs are susceptible, but tissues with a large oxygen consumption, such as skeletal muscle, heart, and brain, are particularly vulnerable. Further, the severity of phenotypic expression and the age at which the disease first presents itself can be unpredictable.
Among unrelated families harboring two mutant POLG alleles, the most common mutations are A467T (ϳ31%), G848S (ϳ10%), and W748S (ϳ8%), followed by in cis T251I ϩ P587L (ϳ6%) (15). With the exception of T251I ϩ P587L, all have been previously characterized biochemically to provide insight into the consequence and mechanism of these Pol ␥ mutations. Located in the linker region between the exonuclease and polymerase domains of Pol ␥, the A467T mutation is associated with PEO, Alpers, and ANS. The purified recombinant A467T Pol ␥ protein possesses diminished DNA polymerase activity (4% of the wild type) and disrupted interaction with the p55 accessory subunit (16). The G848S mutation changes a highly conserved residue in the polymerase thumb subdomain that causes Alpers syndrome. In vitro biochemical studies show reduced DNA polymerase activity (Ͻ1% of WT) and a 5-fold reduction in DNA binding affinity as compared with the wildtype (WT) protein (17). Linked to Alpers syndrome and ANS is the W748S mutation in the linker region of Pol ␥. The recombinant W748S Pol ␥ exhibits low DNA polymerase activity, processivity, and affinity for DNA (18). These defects can be modulated by the common E1143G single nucle-otide polymorphism (SNP), which is almost always found in cis with W748S (18). This outcome strongly suggests that the presence of other in cis mutations could alter the function of Pol ␥ for better or worse.
Scuderi et al. (19) recently compiled a comprehensive list of clinical phenotypes and genetic characteristics of the approximately 50 cases of T251I ϩ P587L mentioned in the literature. The main clinical presentation is PEO, with or without ptosis, and secondary clinical features include ataxia, myopathy, epilepsy, neuropathy, and hepatic diseases. Age of onset is variable, and disease is equally distributed between the sexes. Alpers and myocerebrohepatopathy spectrum disorders have been diagnosed in infancy, although infrequently (20).
The T251I and P587L substitutions are located in the exonuclease domain and linker region of the Pol ␥ gene, respectively (Fig. 1A). Each exists at the same frequency in any given database, ranging from 0.30 to 0.52% of the worldwide population. Even more dramatically, the in cis T251I ϩ P587L mutation pair was reported in heterozygosity with a wild-type allele in ϳ1% of Italian controls (21,22). In patients, the in cis T251I ϩ P587L mutation pair is found either on both alleles (22)(23)(24) or, more frequently, as a compound heterozygous mutation pair in trans with another putative pathogenic mutation (15, 20, 21, 24 -28). Three previous studies have reported T251I as a compound heterozygote without P587L (29 -31). However, resequencing of Pol ␥ in the first two cases revealed that T251I was actually in cis with P587L (32). Conversely, P587L as a compound heterozygote without T251I has been documented in three patients with PEO (33)(34)(35). PolyPhen-2, a publicly available computer algorithm that predicts the impact of an amino acid substitution on the structure and function of a protein, projected P587L as probably damaging and T251I as benign. This may be due in part to the higher phylogenetic conservation of P587L as compared with T251I (Fig. 1B) and the presence of two known pathogenic mutations, G588D and P589L, near P587L (36). Given these data, it has been predicted that P587L is the pathogenic allele (20,22). Nonetheless, because both mutations are found in almost all clinical cases, definitive assignment of pathogenicity to T251I or P587L and/or possible synergistic effects has remained uncertain until now. In this work, we biochemically characterized the individual (WT, T251I, and P587L) and combination (T251I ϩ P587L) human Pol ␥ variants along several parameters, including intrinsic affinity for double-stranded DNA, thermostability, steady-state kinetics analysis of polymerase and exonuclease activities, physical association with the p55 accessory subunit, and processivity of DNA synthesis. Our results demonstrate not only functional impairment of each of the individual variants on Pol ␥, but a dramatic synergistic effect, thus exposing the underlying molecular mechanism.

Results
Production of Pol ␥ Variants-To study the biochemical properties of the individual T251I and P587L Pol ␥ variants and the combination variant in comparison with WT, we created each substitution employing site-directed mutagenesis, overexpressed the recombinant proteins in baculovirus-infected Sf9 cells, and purified each to homogeneity following established purification processes (37)(38)(39). Immunoblot analysis confirmed the identity of each full-length variant. The overall yields of WT and each individual variant were equivalent but were less than previously reported for the exonuclease-deficient WT Pol ␥ (37). However, the yield of the T251I ϩ P587L combination variant was significantly less, although predictable chromatographic behavior during purification and SDS-PAGE analysis showed purity and structural integrity of all enzymes (data not shown).
DNA Binding Affinity Is Impaired for the Pol ␥ Variants-Mutations in the Pol ␥ polymerase domain and linker region have resulted in impaired binding to DNA (17,18,38). Hence, the DNA binding affinity of the WT and Pol ␥ variants was measured by EMSA. Various concentrations of each enzyme were incubated with a fluorescently labeled 54-bp, doublestranded, forked oligonucleotide substrate, and the mixtures were resolved by native PAGE to separate the protein-DNA complex from free DNA (Fig. 2, A-D). Of note are the two distinct shifted species for bound DNA in all variants with the exception of T251I ϩ P587L. Each species is dependent on the concentration of the enzyme and may reflect two areas of DNA binding. The apparent dissociation constant, K d(DNA) , for each variant was calculated by fitting the fraction of DNA bound as a function of enzyme concentration to a quadratic equation by non-linear regression analysis with compensation for ligand depletion (Fig. 2E) (40). Both WT and T251I had strong affinities for DNA at 72.5 Ϯ 10.7 and 60.5 Ϯ 12.2 nM, respectively (Fig. 2F). P587L affinity for DNA was moderately, yet significantly, compromised with a 2-fold reduction as compared with WT (138.9 Ϯ 10.1 nM). Similarly, the binding affinity of T251I ϩ P587L for DNA was even more impaired with a ϳ3-fold reduction from WT (200.5 Ϯ 16.3 nM) (Fig. 2F).
Thermostability of the Pol ␥ Variants Is Reduced-Pol ␥ activities are extremely sensitive to heat inactivation, particularly in the absence of a DNA substrate (16,41). Although intrinsic protein stability of the ␣-helices within Pol ␥ variants has been gauged previously by circular dichroism spectroscopy (17,18,39), heat inactivation of the holoenzyme (p140 ϩ p55) followed by its functional assessment in the presence of a primer-template limits determination of protein stability to only active molecules in the protein population. Consequently, we heatinactivated each holoenzyme variant (1 nM p140 ϩ 2 nM p55) by incubation at 37°C at several time points between 0 and 45 min and then measured reverse transcriptase activity in 10-min reactions using poly(rA)⅐oligo(dT) 12-18 as the primer-template (see "Experimental Procedures"). WT was inactivated with a half-life of 7.5 Ϯ 0.8 min (Fig. 3, A and B). T251I, P587L, and T251I ϩ P587L each had significantly lower half-lives at 5.7 Ϯ 0.4, 3.6 Ϯ 0.2, and 2.1 Ϯ 0.2 min, respectively (Fig. 3, A and B). The significant reduction in half-life for the Pol ␥ variants indicates that these amino acid substitutions promote a higher probability of losing activity through heat-induced conformational change that causes unrecoverable loss of activity, and this effect is additive for the combination mutant.
Exonuclease Activities of the Pol ␥ Variants Are Diminished-Because one of the mutations resides within the exonuclease domain, exonuclease activity was determined by monitoring degradation of the 5Ј-end-labeled 35-mer single-stranded M13 DNA primer by 12% urea-PAGE denaturing gel (see "Experimental Procedures"). The rate of excising 3Ј-primer termini for each enzyme was determined by plotting the loss of substrate versus time at points along a 3-min reaction. Final enzyme concentrations were experimentally determined to be within the linear range of the reaction for each variant. A representative gel revealing nucleotide excision events with the plot of exonuclease product formed as a function of time for the Pol ␥ variants is shown (Fig. 4, A-E). The turnover number (k exo ) for WT exonuclease activity was 20.9 Ϯ 1.0 min Ϫ1 (Fig. 4F). The T251I substitution in the exonuclease domain of Pol ␥ resulted in a 7.2-fold reduction (2.9 Ϯ 0.3 min Ϫ1 ) in exonuclease activity. Although P587L is in the linker region, k exo for P587L was reduced 3.4-fold (6.2 Ϯ 0.4 min Ϫ1 ). Excision by the double mutant form was severely impaired by 13.1-fold (1.6 Ϯ 0.4 min Ϫ1 ), suggesting synergistic defects for the T251I ϩ P587L variant (Fig. 4F).

Kinetic Parameters of the Pol ␥ Variants Reveal Catalytic
Defects-Steady-state kinetic measurements for the Pol ␥ variants in the presence of p55 were determined in a standard DNA synthesis assay utilizing the homopolymeric primer-template substrate poly(dA)⅐oligo(dT) [12][13][14][15][16][17][18] with Mg 2ϩ as cofactor and with varying concentrations of dTTP. The rate of dTTP incorporated over time was used to calculate K m(dTTP) and k cat by fitting the data to the steady-state Michaelis-Menten model (Fig. 5A). The WT enzyme had a K m(dTTP) of 1.1 Ϯ 0.1 M and a k cat of 8.0 Ϯ 0.6 min Ϫ1 (Table 1), in accordance with values that have been reported previously (18,42). The efficacy of each enzyme can be estimated by the specificity constant,  k cat /K m(dTTP) , which is equivalent to the pre-steady-state indicator of enzymatic efficiency, k pol /K d (17,43,44). For ease of comparison, specificity constants for mutant enzymes were also expressed as a fraction of the WT value. For example, the catalytic efficiencies for the single variants T251I and P587L were similar at 29 and 32% of WT activity, respectively (Table 1). We were unable to attain kinetic parameters for the T251I ϩ P587L variant without using higher dTTP concentrations (Fig. 5B). Despite this, the double mutant demonstrated severe catalytic dysfunction and retained only 5% of WT activity.
Physical Association with the p55 Accessory Subunit Is Unperturbed in the Pol ␥ Variants-Because failure of the Pol ␥ holoenzyme to assemble properly could emulate catalytic defi-ciency in vitro, we evaluated physical association between the accessory subunit and each catalytic subunit under stringent conditions in vitro. Polymerase activity was measured on poly(rA)⅐oligo(dT) 12-18 with 25 M dTTP, 220 mM NaCl, the indicated p140 subunit at 2.5 nM, and varying concentrations of p55. Because the isolated catalytic subunit is inactive at 220 mM NaCl, only activity of p140-p55 complexes is measured under these conditions (45). Binding isotherms were constructed from plots of polymerase activity measured at each concentration of p55, permitting calculation of a subunit dissociation constant, apparent K d(p55) , for each variant (Fig. 6, A-D). The   and k cat were determined for the holoenzymes with poly(dA)-oligo (dT) 12-18 as the primer-template under 100 mM NaCl as described under "Experimental Procedures." All kinetic parameters were determined by nonlinear regression analysis of the data presented in Fig. 5. The average of three independent experiments is shown with error expressed as S.D. and significance determined by Student's t test.  (Fig. 6E). Interestingly, despite critical disruptions in thermostability, exonuclease activity, and polymerase activity, the double mutant retained high binding affinity for p55 (apparent K d(p55) of 0.08 Ϯ 0.08 nM). To control for the nucleotide binding defect of T251I ϩ P587L (Table 1), the assay was repeated with 10-fold higher dTTP, and physical association with p55 dimers did not change (data not shown). High affinity binding of Pol ␥ subunits suggests that the biochemical defects associated with these p140 variants are caused by subtle structural changes instead of gross structural alterations that would impair Pol ␥ holoenzyme assembly.
Primer Extension of the Pol ␥ Variants Is Impaired Even in the Presence of p55-To determine whether the strong physical interaction between the catalytic and accessory subunits translated into proper function of Pol ␥ while copying a natural DNA template, we compared activity of each Pol ␥ variant alone or reconstituted with p55 in gel-based primer extension assays under conditions that permit multiple binding events in vitro. WT Pol ␥ reactions utilized a bacteriophage M13-based natural ssDNA template under physiological NaCl and dNTP concen- trations in the presence and absence of p55 without use of a DNA trap. Without p55 present, the WT catalytic subunit can bind the primed M13 ssDNA primer-template and can synthesize 50 -100 nucleotides (nt) before dissociating from the DNA (37). The presence of p55 enhances the DNA binding affinity of the holozyme complex, and under a physiological salt concentration, p55 conveys a salt tolerance and stimulates processivity as much as 50-fold (6). WT functioned as expected in this assay with virtually complete extension of the 35-mer DNA primer in the presence of the accessory subunit (Fig. 7, compare lanes 2  and 3). Distinct pausing at known template secondary structures (ϳ85 and ϳ125 nt on the gel) was evident, as depicted previously (16,18,37,38). The abilities of T251I (Fig. 7, compare lanes 4 and 5) and P587L (Fig. 7, compare lanes 6 and 7) to extend the DNA primer in the absence of p55 were moderately inhibited as compared with WT. Similarly, the processivity of both T251I and P587L in the presence of p55 was less than WT, with slightly less activity for P587L (Fig. 7, compare lane 3 with  lanes 5 and 7). Extension of the DNA primer by T251I ϩ P587L was undetectable in the absence of p55 (Fig. 7, lane 8). Even in the presence of the accessory subunit, the double mutant did not produce long fragments, although a small quantity of primer extension products accumulated at the first major pause site at ϳ85 nt (Fig. 7, lane 9). In general, the lengths of the primer extension products for all Pol ␥ variants were proportional to polymerase activity, DNA binding affinity, thermostability, and kinetic parameters described earlier.

Discussion
After A467T, G848S, and W748S mutations, the in cis T251I ϩ P587L double mutation is the fourth most common mutant allele in human POLG (15). Despite the high prevalence of this allele in mitochondrial disease, particularly in PEO, a kinetic and biochemical analysis of the in cis T251I ϩ P587L Pol ␥ had not been reported until this study. We specifically wanted to determine whether an individual amino acid substitution was sufficient to cause dysfunction in vitro or whether the double mutation was necessary, and we wished to correlate dysfunction in vitro with pathogenicity.
T251I Pol ␥ Mutation-The scientific literature contains one case report of mitochondrial disease in a 45-year-old female PEO patient bearing a POLG allele that was wild type at codon 587 but bore a T251I substitution. The patient was compound heterozygous at POLG and carried an in trans G848S substitution in the other POLG allele (31). G848S is a recessive mutation in which the presentation of symptoms and progression of disease appear to depend on the mutation in the other allele (17), which taken together suggests that an isolated T251I substitution may be a disease allele (31). Low phylogenetic conservation of Thr-251 and infrequent reports of isolated T251I without an in cis P587L mutation imply that phenotypic effects of T251I substitution are very rare, benign, or well tolerated. Our biochemical analysis found T251I to possess reduced exonuclease activity, lowered thermostability, and catalytic efficiency that was only 29% of WT, despite a strong DNA binding affinity and proper physical association with the p55 accessory subunit that were similar to WT.
Reduced exonuclease activity for an exonuclease domain mutation was unsurprising. In vitro, other exonuclease domain mutations in POLG have either increased exonuclease activity, as in the case of R232G/H, or decreased proofreading ability, as with S305R, G303R, and L304R (46,47). Skeletal muscle mtDNA from PEO patients with exonuclease mutations (A189G, T408A, and T414G) had an increased frequency of random point mutations as compared with controls, which was attributed to reduced proofreading exonuclease activity (48). However, disease mutations in conserved exonuclease domain codons (L211P, Q264H, and R265L) in the MIP1 gene, which encodes the Saccharomyces cerevisiae ortholog of human Pol ␥, did not affect exonuclease activity (49,50). This may be because there is no accessory subunit in yeast (51). Similarly, exonuclease activity of human R232G/H substitutions was not increased in the absence of p55 (46). Additionally, the decreased thermostability and catalytic efficiency of T251I may be reflective of structural instability of the variant. However, protein structure appeared intact, as ascertained from predictable chromatographic behavior during purification and unhindered ability to bind DNA and to interact with the accessory subunit.
Structural insight also sheds light upon the proficiencies and incapacities of the T251I variant. Structures for the Pol ␥ holoenzyme (Fig. 8A) and Pol ␥-DNA complex (Fig. 8B) were solved by the Yin laboratory (52,53). Pol ␥ undergoes intra-and intersubunit conformational changes upon binding to primertemplate DNA (53). The position of Thr-251 is readily apparent in the holoenzyme structure (Fig. 8, A and C). However, in the Pol ␥-DNA complex structure, Thr-251 resides in a disordered portion and is not visible. To highlight its approximate position, the Val-249 and Gln-262 residues that flank the flexible loop containing T251I are labeled and provide a general idea of how Thr-251 moves upon binding DNA (Fig. 8, B and C). For example, the distance between Thr-251 in the unbound structure and residues Val-249 and Gln-262 in the DNA complex form is 20.8 and 24.5 Å, respectively. In the unbound form, the putative Thr-251 residue is not found in the DNA binding pocket and moves even further away from the DNA primer-template in the Pol ␥-DNA complex (PyMOL Molecular Graphics System, version 1.8, Schrödinger, LLC, New York) (Fig. 8C). Further, Thr-251 is not located at the interface of p140 and p55. Therefore, the T251I mutation would not be predicted to interfere with DNA binding or interaction with p55. Additionally, the isoleucine substitution of threonine replaces one C ␤ -branched amino acid with another, which results in comparable bulkiness near the protein backbone with parallel limitations to conformations that the main chain can adopt. However, the amino acid substitution changes the side chain from uncharged polar to nonpolar, which may discourage movement in the exonuclease domain of Pol ␥ should the residue encounter other nonpolar side chains during DNA binding. The DNA-bound structure also places the Thr-251 residue in a surface-exposed position, which favors an uncharged polar residue but not a nonpolar residue. Thus, a decrease in movement and an unfavorable surface-exposed position could partly contribute to a decrease in exonuclease activity, reduced catalytic efficiency, and protein instability with this variant.
P587L Pol ␥ Mutation-The scientific literature describes three cases of PEO in which the P587L POLG substitution is not in cis with T251I (29 -31). In each case, another mutation occurs in trans with P587L that could be contributing to or causing pathology. However, the high degree of phylogenetic conservation of P587L in vertebrates strongly supports the notion that P587L is a disease allele. Our biochemical analysis revealed that P587L has deficiencies in DNA binding, thermostability, exonuclease activity, kinetics parameters, and primer extension activity. Like T251I, P587L does not affect physical association between the catalytic and accessory subunit.
Again, examination of Pol ␥ structures helps to explain these biochemical shortcomings. The shift between the C ␣ -C ␣ measurement for Pro-587 for the holoenzyme and the Pol ␥-DNA complex is 17.2 Å (Fig. 8, A, B, and D). In the holoenzyme, the Pro-587 residue is located within the DNA binding pocket. Therefore, Pro-587 needs to shift to accommodate DNA binding and in the Pol ␥-DNA complex moves significantly closer to the p55 interface. When mutated, the substitution replaces the more rigid, ringed, rotomer-restricted proline with the more flexible, longer side chain of leucine. The rigidity of proline is frequently required to maintain the structural characteristics of a protein. Given the large movement that Pro-587 undergoes upon DNA binding, the loss of proline may thermodynamically favor or accommodate conformations less favorable for turnover. Indeed, this may cause instability and could lead to the observed changes in DNA binding affinity, primer extension activity, protein stability, and exonuclease activity. A K d(p55) equivalent to WT, however, suggests no obstruction of binding to p55.
T251I ϩ P587L Pol ␥ Mutation-Our biochemical analysis revealed that the individual T251I and P587L mutations negatively affected the function of Pol ␥ in vitro, with P587L causing more deleterious effects. Together, the mutations act synergistically along all measured parameters and cause more severe dysfunction than either mutation alone. For instance, the DNA binding affinity of T251I ϩ P587L was 3-fold weaker than WT, whereas binding by P587L was 2-fold lower than WT, and T251I had no impaired DNA binding. The reduced ability of T251I ϩ P587L to bind DNA, in combination with the intrinsic deficiency in T251I exonuclease activity, probably leads to poor proofreading and repair abilities. T251I ϩ P587L was also found to be remarkably sensitive to heat inactivation. Early indications of protein instability were inferred during protein purification, when only low quantities of protein were recovered during elution from the MonoQ column as compared with the yields for the WT and the single mutant variants. Functionally, these weaknesses, in conjunction with a profound nucleotide binding deficiency, caused remarkably inefficient catalysis (ϳ5% of WT) and DNA processivity for T251I ϩ P587L, although none of these failings were due to diminished intersubunit affinity for p55.
Our results reveal that isolated T251I and P587L substitutions in POLG individually hamper the in vitro functioning of the enzyme. We infer that individuals bearing either mutation alone have not been identified because they may not present with any pathology. Conversely, T251I ϩ P587L is a profoundly impaired enzyme in vitro, suggesting synergistic dysfunction when the mutations present in cis. As a recessive mutation in vivo, the presence of T251I ϩ P587L on one allele does not result in disease because the WT allele provides sufficient polymerase function for survival. Therefore, we suggest that pathogenicity only becomes evident in those patients carrying this recessive combination mutation in trans with deleterious mutations on the other POLG allele. For instance, two female infants that were compound heterozygous for T251I ϩ P587L and R232G presented with a severe phenotype in infancy and died before 16 months of age (21,35). The R232G substitution causes decreased polymerase and increased exonuclease activities with decreased selectivity for mismatches (46). Conversely, no pattern emerges with G848S, a POLG mutation in the highly conserved thumb subdomain with extreme deficiencies in polymerase activity and DNA binding in vitro (17). Individuals with the T251I ϩ P587L in trans with G848S include a 75-yearold man with severe PEO and myopathy who first presented clinically at the age of 55 (54); an 80-year-old man with ptosis, SANDO, and myopathy who was first seen at the age of 73 (55); and a 6-month-old infant with Alpers syndrome (20). Because of the highly dysfunctional nature of T251I ϩ P587L, patients homozygous for the combination mutation would be predicted to present with a severe phenotype and/or early age of onset. However, identified patients have a midlife age of onset with a mild phenotype (22,24,32). Taken together, we suggest that factors other than multiple POLG mutations contribute to the severity and age of onset of mitochondrial disease. Possible contributing factors include disease-causing mutations in other nuclear genes; altered interactions with other mtDNA replication proteins, such as the mitochondrial Twinkle helicase or the single-stranded DNA-binding protein; the inherited level of somatic mtDNA heteroplasmy; epigenetic factors; and geneenvironment interactions, coined ecogenetics (18,22,56,57).
Gene-environment interactions are not unprecedented. Valproate, a first-line anticonvulsant, caused hepatotoxicity in a 2-year-old boy with POLG mutations (58). A sequential study of subjects enrolled in the Drug-Induced Liver Injury Network (DILIN) from 2004 to 2008 found that heterozygous genetic variation in POLG was strongly associated with valproate-induced hepatotoxicity (59). Also, chain-terminating nucleoside reverse transcriptase inhibitors used to combat HIV infection inhibit Pol ␥ during mtDNA replication (60). Genetic polymorphisms in POLG also explain the variation in mitochondrial toxicity in HIV-infected patients (61,62). Other viruses, such as human herpesvirus 6, caused encephalitis in two patients with mutations in POLG and exacerbated the Alpers phenotype, contributing to a more rapid clinical deterioration (63).
Although not tied to specific POLG mutations, mitochondria are susceptible to environmental toxicants, such as heavy metals and pesticides (64,65), which leads to questioning whether exposure to environmental contaminants could contribute to altering the age of onset and severity of POLG diseases. Indeed, we utilized S. cerevisiae as a model system to demonstrate that methyl methanesulfonate-induced mutagenesis of mtDNA is increased 30-fold in certain MIP1 disease mutants relative to WT (66). It is critical that future studies continue to identify mechanisms by which mutated forms of human POLG interact with environmental stressors to alter severity of the phenotype and age of onset of mitochondrial diseases.

Experimental Procedures
Construction, Expression, and Purification of Pol ␥ Protein Variants and p55-WT POLG cDNA with a His 6 affinity tag and without a mitochondrial targeting sequence was cloned into the pVL1393 baculovirus transfer vector, which served as the PCR template (43). The T251I, P587L, and T251I ϩ P587L mutations were generated using the QuikChange site-directed