Overexpression of DNA Polymerase Zeta Reduces the Mitochondrial Mutability Caused by Pathological Mutations in DNA Polymerase Gamma in Yeast

In yeast, DNA polymerase zeta (Rev3 and Rev7) and Rev1, involved in the error-prone translesion synthesis during replication of nuclear DNA, localize also in mitochondria. We show that overexpression of Rev3 reduced the mtDNA extended mutability caused by a subclass of pathological mutations in Mip1, the yeast mitochondrial DNA polymerase orthologous to human Pol gamma. This beneficial effect was synergistic with the effect achieved by increasing the dNTPs pools. Since overexpression of Rev3 is detrimental for nuclear DNA mutability, we constructed a mutant Rev3 isoform unable to migrate into the nucleus: its overexpression reduced mtDNA mutability without increasing the nuclear one.


Introduction
In yeast, DNA polymerase zeta (Pol zeta) and Rev1 are involved in the error-prone translesion synthesis (TLS) during replication of nuclear DNA: mutations in any of these polymerases lead to a decrease of mutagenesis in nuclear genome either in normal conditions or in conditions known to increase DNA damage such as UV treatment [1]. The yeast Pol zeta is composed by a catalytic subunit encoded by REV3 and an accessory subunit encoded by REV7. Biochemically, Pol zeta is a poorly processive enzyme [2], lacks of a 39-59 exonucleolytic proofreading activity [3] and is able to extend terminally mismatched primers [4][5][6]. From these properties derives its propensity to generate mutations. On the contrary, Pol zeta is very inefficient in inserting nucleotides opposite lesions in vitro, suggesting that a distinct polymerase should act in this first step and that Pol zeta can efficiently extend from primer-terminal base pairs containing mismatches or lesions [6].
Rev1 encodes for a deoxycytidyl transferase that preferentially incorporates C opposite to an abasic site [7]. Also, Rev1 interacts in vitro with Pol zeta through a Rev3-Rev1 binding and this interaction strongly stimulates the activity of Pol zeta and, probably, contributes to its targeting to the replication fork [8].
Recently, it was found that Pol zeta and Rev1 localize in mitochondria and that they participate in the mitochondrial DNA mutagenesis, since disruption of REV3, REV7 or REV1 resulted in a strongly decreased frequency of 21 frameshift mutations in mtDNA [9]. On the contrary, it was shown that deletion of REV3, REV7 or REV1 resulted in increased frequencies of both spontaneous and UV-induced mitochondrial erythromycin resistance, due to base substitutions [10]. On the basis of their results, these authors speculated that, in absence of the TLS by Rev1 and/ or Pol zeta, the mitochondrial DNA polymerase gamma (Pol gamma), the mitochondrial replicase, could create mutations in efforts to bypass lesions that block the replication fork [10]. As a consequence, Pol zeta and Rev1, which in the nucleus are responsible with other DNA polymerases for the error-prone bypass of DNA lesions, in the mitochondrial compartment function in a less mutagenic pathway [10].
Pol gamma is a protein conserved in fungi and animals. Structurally, human Pol gamma is a heterotrimer composed by one catalytic subunit, encoded by POLG, and two accessory subunits, encoded by POLG2 [11]. The catalytic subunit contains an N-terminal 39-59 exonuclease domain (residues 170-440), a Cterminal polymerase domain (residues 440-475 and 785-1239) and a spacer region encompassing residues 475-785 [12]. To date, more than 150 pathological mutations in POLG have been identified in severe mitochondrial disorders (http://tools.niehs. nih.gov/polg/) and, among them, a few mutations have been recently described in stavudine induced toxicity [13,14], in valproate induced hepatotoxicity [15,16] and in breast tumorigenesis [17]. This places POLG as a major locus for mitochondrial diseases and disorders.
In yeast the DNA polymerase gamma, encoded by MIP1 gene, is composed by a sole catalytic subunit [18]. Thanks to the similarity between human Pol gamma and Mip1 (approximately 43%), yeast has been used to validate the role of human putative pathological mutations, to understand the biochemical consequences associated to these mutations, to study the pharmacogenetics of drugs such as valproate and stavudine, and to find mechanisms able to rescue the detrimental effects of Mip1 mutations [16,[19][20][21][22][23][24][25][26][27][28]. Since a therapy has not been yet developed for mitochondrial diseases caused by POLG mutations, a major goal of the research involving yeast/animal models is to find new strategies to rescue the pathological phenotypes associated to these mutations.
Yeast is a suitable model organism to study the effects of nuclear mutations affecting mtDNA stability, thanks to its ability to survive in absence of a functional mitochondrial genome and to the possibility to easily analyze mtDNA mutability. Yeast cells containing deletions-carrying mtDNA, called rho-, or cells which have completely lost mtDNA, called rho 0 , are respiratory deficient and produce colony of small size, called petite. rho 2 cells, which arise more frequently than rho 0 cells, retain only small fragments of wt mtDNA, from 30% to less than 1% (reviewed in [29]). Although the mechanisms thanks to which rho 2 mutations arise are not fully know, analysis of the regions flanking the deleted regions suggest that mtDNA of some of these rho 2 cells derives from recombination between direct repeats flanking the deleted region with consequent excision [29]. In addition, deletions do not occur randomly at all. As a matter of fact, some fragments, encompassing COB, COX2 and COX3 genes, are most frequently retained [29]. These large deletions make mtDNA irreversibly unfunctional. In S. cerevisiae, petite mutants arise spontaneously with high frequency (approximately 10 22 ) [29]. The frequency of petite mutants is a measure of mtDNA extended mutability and is an index of mtDNA instability. Extended mutability is indeed a series of large deletions (generally .70%) that arise spontaneously, and partially randomly, in the mtDNA, without involvement of point mutations. Mutations in a large group of nuclear genes, encoding for proteins directly or indirectly involved in mtDNA replication, recombination, stability and maintenance (reviewed in [30]) produce increase of petite mutability. mtDNA point mutations can be also easily measured as frequency of spontaneous mutants resistant to erythromycin (Ery R mutants), an antibiotic that inhibits mitochondrial but not cytoplasmic translation. In fact, resistance to erythromycin is acquired through specific transversions or transitions in the mitochondrial gene encoding the 21 S rRNA, in particular at 1950,1951,1952 and 3993 positions [31][32][33][34]. Ery R mutants arise spontaneously approximately at a frequency of 10 27 -10 28 . Ery R mutability is influenced by replication fidelity of DNA polymerase gamma as well as by the activity of some enzymes involved in mtDNA repair and/or recombination [34][35][36][37].
In an attempt to investigate about genetic conditions that might reduce the damages caused by mutations in POLG, we analyzed the effect of deletion or overexpression of TLS polymerases that, together with DNA polymerase gamma, are present in mitochondria, Pol zeta (Rev3-Rev7) and Rev1. Genetic tests based on gene overexpression are widely used in yeast to assess a role of a gene function in a biological process. Several researches are reported where one or more proteins were overexpressed with this aim, among which a study showing that overexpression of Rev3 led to increasing UV-induced mutagenesis [38]. In addition, gene overexpression is commonly used to rescue the phenotypic defects caused by mutations in another gene which is functionally associated. For example, in the case of MIP1, overexpression of RNR1, encoding the large subunit of the ribonucleotide reductase, lead to the discovery that increased levels of dNTP pools reduced the extended mutability caused by mutations in Mip1 [19,26,39]. This finding has had impacted also in studies on human DNA polymerase gamma pathological mutations [26], where, starting from analysis in yeast, it has been shown that human DNA polymerase gamma harboring the pathological mutation H932Y has a ,200-fold reduced affinity to the incoming dNTPs.
Here we found that increased expression of Pol zeta resulted in a significant reduction of extended mutability caused by mutations mapping in different domains of MIP1 and that overexpression of both Pol zeta and Rev1 led to a reduction in mtDNA point mutability.
Plasmids pFL38-MIP1 and pFL39 harboring wt or mutant mip1 alleles have been previously described [19]. YEplac195-REV3 was obtained by PCR-amplification of REV3, including the upstream and the downstream regions, from genomic DNA of strain DWM-5A with primers REV3-Fw (gcgcgggtaccggatccaagaatccctgtgg) and REV3-Rv (gcgccgaattccttagaggatacgaagattcc), digestion with KpnI and EcoRI of both the PCR fragment and YEplac195 and ligation. YEplac195-REV7 was obtained by PCR-amplification of REV7 from genomic DNA of strain DWM-5A with primers REV7-Fw (gcgccgtcgacgatgaatgggaaaagaagtgc) and REV7-Rv (gcgcgggtacccatggaaccagacagaagacc), digestion of both the PCR fragment and YEplac195 with SalI and KpnI and ligation. YEplac195-REV3/ REV7 was obtained by subcloning REV7 from YEplac195-REV7 to YEplac195-REV3 digested with SalI and KpnI. YEplac181-REV1 was obtained by PCR amplification of REV1 from genomic DNA of strain BY4742 with primers REV1-Fw (cggcgggtcgacgtgaagtgatcatgcacatcg) and REV1-Rv (cggccggtacccttacgcatttctactgttgg), digestion of both the PCR fragment and YEplac181 with SalI and KpnI and ligation. pUG35-REV3 was obtained by PCR amplification of REV3 ORF from genomic DNA of strain DWM-5A with primers REV3GFPFw (ccgccgaattcatgtcgagggagtcgaacgacac) and REV3GFPRv (ccgccgtcgacccaatcatttagagatattaatgc), digestion of both the PCR fragment and pUG35 with EcoRI and SalI and ligation in order to clone REV3 in frame with the EGFP gene. pTL85-REV3 was obtained by subcloning REV3 from pUG35-REV3 to pTL85 digested with XbaI and SalI. The REV3 DNA region corresponding to the amino acid stretch 1 to 534 was cloned in frame with the EGFP gene of pUG35 by amplification of the fragment from genomic DNA of strain BY4742 with primers REV3GFPFw and R3GFPRvS (ggcccgtcgaccagaatatcttgataaccg) and cloning it between the EcoRI and SalI sites. All the cloned fragments were sequenced to check for the absence of mutations.
REV3 mutagenesis 505-508 RKRKK stretch of Rev3 was mutagenized to AAAAA through the overlap extension technique [46]. In the first set of PCR reactions, the REV3 fragment encoding for the 1-534 Rev3 stretch was amplified with i) the primer REV3GFPFw and the mutagenic primer REV3NONSRV (cgtaatcccgcagccataaccGCc-GCtGCgGCcGCagacagcaacgatgttttattagc), and ii) the mutagenic primer REV3NONSFW, complementary to REV3NONSRV, and the R3GFPRVS primer. The final mutagenized product was obtained by using the overlapping PCR fragments as template and REV3NONSFW and R3GFPRVS as external primers. The product was then digested with EcoRI and SalI and cloned in EcoRI-SalI digested pUG35.
To mutagenize YEPlac195-REV3, a fragment of REV3 was amplified with the primer REV3MUTFW (cattgagacaccaaagatgtgc) and with the mutagenic primer REV3NONSRV, and with the mutagenic primer REV3NONSFW and the primer REV3MUTRV (cagaatggaaatctattcctatc). The final product was obtained as above and was cloned in NgoMIV-SacI digested YEplac195-REV3. The entire cloned fragment was sequenced to confirm the presence of the mutations.

Disruption of REV3, REV7 and REV1
REV3 was disrupted in strain DWM-5A by using the HphMX4 cassette amplified from plasmid pAG32 [34] by using primers DELREV3FW (gttcataaaacgctggaagtaaaaattagggcatcctttaaaaggC-CTTGACAGTCTTGACGTGC) and DELREV3RV (aagtaatc-gaatacatttttcaataatcttctgctgggctgggatCGCACTTAACTTCGCAT-CTG), each of which contains a 59 stretch complementary to a sequence internal to REV3 ORF and a 39 stretch, in uppercase, complementary to the extremities of the HphMX4 cassette. Disruptants were as resistant to 250 mg/ml hygromycin B. Correct disruption was confirmed by PCR.
REV7 and REV1 were disrupted in strain DWM-5A9 by using the KanMX4 cassette. The disrupted rev7::Kan R and rev1::Kan R loci were amplified from strain BY4742 rev7D and BY4742 rev1D, respectively, by using primers external to the loci: REV7-Fw and REV7-Rv, or REV1-Fw and REV1-Rv. Disruptants were selected on YP supplemented with 200 mg/ml geneticin and disruption was confirmed by PCR.

Determination of mitochondrial and nuclear mutability
Petite frequency was determined as previously reported [48]. To determine the effect of antioxidant agents on the extended mtDNA mutability, wild-type and mutant strains, were grown on liquid SCD medium supplemented with 40 mM dihydrolipoic acid (Sigma) or 10 mM MitoQ (a kind gift from Luigi Palmieri, University of Bari, Italy) for two 24-hour growth cycles. Dihydrolipoic acid was supplemented from a 40 mM stock solution in ethanol, while MitoQ was supplemented from a 5 mM stock solution in DMSO. Control experiment in which equal amounts of ethanol (0.1%) or DMSO (0.2%) were added to untreated cells was done in parallel. For each strain/condition at least three independent experiments were performed on three independent clones. Statistical analysis of petite frequency was performed by a two-tailed t-test.
Ery R mutant frequency was determined as previously reported [48]. Briefly, 15 independent colonies, isolated on SC medium supplemented with 2% glucose, were inoculated into 2 ml of SC liquid medium supplemented with 2% ethanol and grown to saturation (approximately 48 h). Cultures were then plated on YPAEG-Ery medium and dilutions were plated nonselectively on YP medium supplemented with 3% glycerol (300-500 colonies) for a total cell count. After 7 days incubation, Ery R mutant frequency was determined by the method of the median [49]. Each experiment was repeated at least three times on two independent clones for each strain. Statistical analysis of Ery R mutant frequency was performed by a one-tailed t-test.
Nuclear point mutability was evaluated by measuring the mutation rate of Can R mutants in strain W303-1A CAN S , bearing the wt copy of CAN1, as previously reported [50], except that colonies were grown on SC medium supplemented with 2% glucose. Mutation rates were calculated using median values according to Lea and Coulson [49] by using the program FALCOR [51]. Rates are given as mean of four independent experiments in each of which 10 independent colonies for each clone were tested. Unpaired two-tailed t-test was used to evaluate the statistical significance.
In every test, frequencies or rates were considered significantly different when p,0.05.

Northern blot and semiquantitative RT-PCR
Total RNA was extracted from 10 ml of culture at a concentration of 1 OD 600 /ml, according to [52]. For Northern blot analysis, 25 mg of RNA were blotted according to [53]. The probes for hybridization were obtained by purifying the 1 Kb EcoRV fragment of REV3 and by amplifying REV7 with primers REV7N-Fw (gggtagagaagtggctgagg) and REV7N-Rv (ctcgccctcttcatattgag). Probes were labeled using [a-32P]dCTP and the kit 'Rediprime II DNA Labelling System' (GE Healthcare). Signals were acquired after different time exposures with Kodak films or with PhosphorImager and quantified by ImageQuant software (Molecular Dynamics).
For RT-PCR, 1.5 mg of RNA were treated with DNase (Invitrogen), precipitated and converted to cDNA with Proto-ScriptH M-MuLV First Strand cDNA Synthesis Kit (NEB) by using oligod(T). Semiquantitative PCR was performed by amplifying a fragment of REV1 cDNA with primers REV1rtFW (gtatattccagccgatcaaattcc) and REV1rtRV (cgtttggaagcctgaatgatcc) and, as control, a fragment of TAF10 DNA with primers FWTAF10 (atattccaggatcaggtcttccgtagc) and RVTAF10 (gtagtcttctcattctgttgatgttgttgttg). PCR samples were taken after 20, 23, 26, 29, 32 and 35 cycles, and aliquots of equal volume was loaded on a 0.8% agarose gel. Images were acquired by Kodak Digital Science D120 zoom digital camera, converted to gray scale with Photo-shop; signals were quantified through Multianalyst software (Molecular Dynamics). To avoid saturation, signals were considered only if the signal deriving from the PCR amplification with 3cycle less was 6-8-fold lower. Both in the case of Northern analysis and in the case of RT-PCR, relative levels of mRNA were quantified by comparison with a house-keeping gene signal (ACT1 or TAF10). All experiments were performed three times.

Fluorescence microscopy
Cells were grown in solid SC medium without methionine and supplemented with 2% glucose. After 24 hours, cells were resuspended in Tris-HCl 50 mM pH 7.5, incubated with 1 mg/ ml DAPI (49,6-diamidino-2-phenylindole) for 15 min on ice and observed with a 1006 magnification using DAPI or FITC filters. Pictures were taken with constant parameters (100 ms exposure for DAPI and 250-500 ms exposure for GFP with Nikon ACT-1 acquisition software) and processed to optimize contrast and brightness with Adobe Photoshop 7.0.

Results
Pol zeta overexpression rescues the mtDNA mutability induced by specific mip1 mutations We have previously constructed a set of Saccharomyces cerevisiae mip1 mutant alleles, corresponding to POLG pathological mutations, and determined their effects on mtDNA extended and point mutability (Table 1) [19][20][21]23]. These mutant alleles localized in different Mip1 domains, displayed different in vivo and in vitro defects and behaved either as dominant or recessive (Table 1 and Figure 1).We analyzed the effect of overexpression and deletion of Pol zeta and Rev1 on the mtDNA mutability induced by these mip1 mutations, based on the recent finding that these polymerases were found in mitochondria, where Pol zeta and Mip1 belong to the same epistatic pathway [9].
REV3 and REV7 were cloned in the multicopy plasmid YEplac195 and introduced i) in the strain DWM-5A (mip1D) transformed with a number of recessive mip1 mutant alleles (haploid condition) and ii) in the strain W303-1B (MIP1) transformed with dominant mip1 mutant alleles (heteroallelic condition). Extended mtDNA mutability was first determined in these transformants by analyzing the frequency of spontaneous petite mutants.
In presence of overexpressed REV3, the level of petite mutants decreased; the rescue was mutation specific, being null for some mutations and significant for other ones ( Figure 2). The strongest effect was observed for haploid strains bearing mutations G807R (50% reduction) and A692T (25% reduction). The REV7 overexpression, on the contrary, had slighter effects, which however were significant for mutations A692T and G807R (Figure 2A). In heteroallelic strains harboring wt MIP1 and dominant mip1 E698G , mip1 K749R or mip1 Y757C mutant alleles ( Figure 2B) no rescue was observed in the presence of either REV3 or REV7.
We also evaluated the combined effects of REV3 and REV7 overexpression by cloning the two genes together in the YEplac195 plasmid. Northern blot analysis indicated that REV3 and REV7 were expressed at high level (10-12 fold increase of mRNA levels) when cloned either alone or together, demonstrating that their concomitant presence in the plasmid did not affect the level of their expression ( Figure 3A). Overexpression of whole Pol zeta had the same effect than overexpression of Rev3 (Figure 2), suggesting that Rev3 alone was responsible of the observed rescue or that the physiological levels of Rev7, which is more abundant than Rev3 [54], were sufficient to ensure the highest level of rescue. However the slight decrease in petite frequency obtained overexpressing Rev7 alone suggests, as demonstrated below, that also Rev7 participates to the rescue and that the petite mutants' reduction observed is most likely due to a slight increase of the whole Pol zeta.

Effect of REV3 deletion on extended and point mtDNA mutability
We next investigated the effect of REV3 deletion on mtDNA stability. No difference was observed between REV3 and rev3D regarding the frequency of petite mutants in strains harboring either wt MIP1 allele or mutant mip1 alleles, indicating that the REV3 deletion produced no effect on extended mutability (data not shown).
However, deletion of REV3 resulted in alteration of mtDNA point mutation, with 2.5 fold increase of Ery R mutant frequency in the MIP1 wt strain, in agreement with results reported by Kalifa and Sia [10] (Table 2). In addition, strains carrying mip1 C261R , mip1 A692T and mip1 G807R mutant alleles showed a further increase of Ery R mutation (an increase of 3.9-, 3.3-and 5.1-fold, respectively) ( Table 2), indicating that mutations in MIP1 and deletion of REV3 have a synergistic effect in rising point mutability. Furthermore, the ratios of Ery R mutant frequencies of rev3D strains to REV3 wt strains parallelized the mutator phenotype of the mip1 mutant allele, being the lowest in mip1 A692T strain showing a moderate increase of Ery R frequency compared to wt MIP1 strain (approximately 3.4-fold), and being the highest in the strain mip1 G807R showing an increase of 8.8 times the Ery R frequency of MIP1 strain (Table 2). Therefore, Pol zeta is not physiologically involved in the maintenance of full mtDNA but is required for the fidelity of the mtDNA replication.

Effects of REV1 overexpression
Rev3 subunit of Pol zeta binds to Rev1, and the complex Pol zeta-Rev1 increases the efficiency of translesion synthesis by Pol zeta [8]. We tested whether overexpression of REV1, together with overexpression of Pol zeta, could further reduce the petite frequency due to mip1 C261R , mip1 A692T and mip1 G807R mutant alleles. REV1 was cloned in the multicopy plasmid YEplac181 and introduced in strains overexpressing both REV3 and REV7 and carrying wt MIP1, mip1 C261R , mip1 A692T and mip1 G807R mutant alleles. RT-PCR analysis confirmed that REV1 was 9-13-fold overexpressed in transformant cells ( Figure 3B) and that REV3 and REV7 were overexpressed at the same levels (10-12 fold overexpression) as in strain not harboring YEplac181-REV1 (data not shown). Together with REV1, we overexpressed both REV3 and REV7, and not REV3 alone, because it is known that Rev1 can recruit Rev7 through a Rev3 independent Rev1-Rev7 binding thus depleting the levels of Rev7 [54]. REV1 overexpression did not alter the frequency of petite, both in presence and absence of overexpressed Pol zeta (Figure 4). This result indicates that either the physiological levels of REV1 are sufficient for the rescue by Pol zeta overexpressed or that Rev1 is not involved. Moreover, we observed that overexpression of both Pol zeta and Rev1, but not of either Pol zeta or Rev1 individually, resulted in a significant reduction of the Ery R mutant frequency, in particular for the strain carrying the mutation G807R ( Table 3). The results described here are symmetrical to those obtained previously by Kalifa and Sia [10], who showed an increase in Ery R mutant frequency when REV1, REV3 or REV7 were disrupted The mutation is slightly dominant in yeast, where the petite frequency in heterozygosis is approximately 1.5-fold compared to the wild type. In humans, the equivalent mutation has been associated either to sporadic PEO when it is in trans with mutation R579W, suggesting a recessive trait [80] and to PEO with tremor and peripheral neuropathy when it is in cis with E1143G SNP and in trans with wt POLG, suggesting a dominant trait [81]. doi:10.1371/journal.pone.0034322.t001

Rescue of extended mutability by Rev3 overexpression is Rev7 dependent and Rev1 independent
We tested whether the rescue by Rev3 overexpression was dependent on the physiological levels of Rev7 and Rev1. The deletion of REV1 or REV7 had no effect on petite mutability (data not shown). However, the rescue of petite mutability in MIP1 mutants, exerted by overexpressed Rev3, occurred only when Rev7 was present ( Table 4), indicating that the whole Pol zeta participated in the rescue. We furthermore observed that overexpression of REV7 in a rev3D strain did not reduce petite frequency both in wt MIP1 and in mip1 mutant strains (data not shown), corroborating this hypothesis. Unlike Rev7, Rev1 was dispensable (Table 4), thus supporting the hypothesis reported in the previous paragraph that Rev1 did not participate in the rescue of the extended mutability.

Combined action of REV3 overexpression and other chemical and genetic rescues
We have shown previously that a decrease of reactive oxygen species and/or an increase of the mitochondrial dNTP pool reduced the mtDNA extended mutagenesis induced by pathological Pol gamma mutations in yeast [19]. Several mutations in MIP1, but not all, were rescued, albeit at different extent, by treatment with dihydrolipoic acid or with the mitochondrial specific ROS scavenger MitoQ [21]. To know whether this effect could be increased by REV3 overexpression the petite frequency was measured either in haploid ( Figure 5A) or in heteroallelic strains ( Figure 5B) treated with these molecules. We found that there was a negative correlation between the rescue exerted by ROS scavengers and the rescue caused by REV3 overexpression (Figure 2 and Figure 5).
In fact, in haploid strains harboring A692T and G807R mutations, whose effect was reduced by overexpression of Rev3, no rescue was exerted by dihydrolipoic acid or MitoQ. In contrast, ROS scavengers strongly mitigated the effect of mip1 mutations E698G, K749R, Y757C, that are insensitive to Rev3 overexpression.
The only exception was mutation C261R, whose effect was slightly rescued both by treatment with ROS scavengers and by Rev3 overexpression ( Fig. 2A and Fig. 5A). In this case, the additive effect of Pol zeta overexpression and ROS scavengers (data not shown) suggested that two different mechanisms were involved in the reduction of petite mutability observed.
The second mechanism able to rescue the mip1 induced mtDNA extended mutability is the increase of the dNTP pools. In S. cerevisiae an increase of the dNTP pool was obtained either by overexpressing RNR1 gene, which encodes the large subunit of the ribonucleotide reductase or by deleting SML1 gene, which encodes an inhibitor of the latter activity [55]. It has been shown that overexpression of RNR1 or deletion of SML1, reduced the petite frequency in strains harboring specific MIP1 mutations [20,26,39,55].
To evaluate the relationship between the rescue exerted by REV3 overexpression and the rescue exerted by increasing the dNTP pools, we deleted SML1 in strains harboring mip1 C261R , mip1 A692T and mip1 G807R mutant alleles, in presence of overexpressed REV3. We observed that, for every mutation studied, the  two processes act synergistically ( Figure 6). For example, in strain carrying mip1 A692T mutant allele the overexpression of REV3 reduced the petite frequency of approximately 1.3-fold, while deletion of SML1 reduced the petite frequency of 4.4 fold: together, the reduction was 9.1-fold. The synergistic effects were even stronger in the case of strain harboring mip1 G807R mutation (1.9fold when REV3 was overexpressed, 4.1-fold when SML1 was deleted, 15.7-fold together). It is likely that the observed synergy is due to an increase of dNTP pools, which are the substrate for both Mip1 and Pol zeta.

Construction of a mutant Rev3 which rescues mtDNA mutability without affecting nuclear mutability
It was previously shown that in UV-treated cells, enhanced expression of REV3 led to increase of nuclear point mutability, measured as increase in the rate of arg4-17 reversion. Pol zeta mutagenic function is dependent on Rev1 protein [38]. We measured nuclear mutability in cells transformed with REV3 and/ or REV1 overexpressing plasmids to determine whether increased levels of Pol zeta/Rev1 exerted mutagenic effect as a consequence of a general mechanism associated to the translesion bypass. The overexpression of Rev3 or Rev1 caused an approximately two-fold increase of nuclear point mutability, measured as rate of Can R mutants. When both Pol zeta and Rev1 were overexpressed, the rate of Can R mutations increased approximately 3-fold (Figure 7). Thus, overexpression of Pol zeta is useful to reduce the mtDNA instability caused by specific mutations in MIP1 but is detrimental to spontaneous nuclear mutability.
A Rev3 mutant which could enter into mitochondria, but not in the nucleus is expected to reduce the mitochondrial mutability without increasing nuclear mutability. To this end, we thought of designing a REV3 gene whose product could not enter the nucleus. We searched inside Rev3 for the potential nuclear localization signal, by using the PredictNLS server (http://cubic.bioc. columbia.edu/services/predictNLS/). Rev3 protein is predicted to possess a unique bipartite nuclear localization signal encompassing amino acids 487-508 ( Figure 8). Both upstream and downstream basic amino acid stretches composing the bipartite signal are critical for nuclear import [56].
To evaluate the localization of Rev3, we first cloned the whole coding sequence either in a centromeric plasmid or in a multicopy plasmid in frame with the EGFP gene. In both cases no GFP signal was detected in transformant cells, as it was found previously in high throughput analysis [57]. To overcome this problem, a shorter Rev3 fragment encompassing residues 1-534, which include the bipartite signal, was cloned in frame with EGFP. In this case, in transformant cells we observed a fluorescent signal corresponding to Rev3, localized in the nucleus. This was confirmed by DAPI staining of mitochondrial and nuclear DNA ( Figure 6). Both in rho + strain, where DAPI stained both mtDNA and nuclear DNA, and in the rho 0 , where DAPI stained only nuclear DNA, the GFP signal co-localized with the nuclear DAPI signal.
To obtain a version of Rev3 unable to migrate in the nucleus, we changed the 504-508 RKRKK downstream stretch to AAAAA (Figure 8). When Rev3 fragment encompassing amino acid 1-534 and harboring the 504-508 AAAAA stretch was fused  with EGFP, the fluorescent signal corresponding to the fusion protein localized in the cytoplasm only ( Figure 9). The AAAAA mutation was then introduced in the full REV3 gene, cloned in the multicopy plasmid YEplac195 and its effect was evaluated in strains harboring mip1 A692T or mip1 G807R . In the presence of overexpressed mutant Rev3, the mtDNA extended mutability decreased at the same extent as achieved in the presence of overexpressed wild type Rev3, indicating that the alteration of the nuclear signal did not affect the ability of Rev3 of operating in mitochondria (Table 5). If the mutated version of the enzyme did not enter the nucleus, then it was expected that its overexpression did not cause an increase in nuclear mutability. Indeed, in the presence of overexpressed mutant Rev3 the nuclear point mutability, measured as rate of Can R mutants accumulation, did not increase (Figure 7).

Discussion
In this study we showed that increased levels of Rev3, encoding the catalytic subunit of Pol zeta, reduces both the mtDNA extended mutability and the point mutability caused by specific mutations in Mip1, the S. cerevisiae mtDNA polymerase. The rescue of mtDNA extended mutability requires overexpression of the catalytic subunit Rev3, but not that of the accessory subunit encoded by REV7. However, a basal level of Rev7, which is known to be expressed more than Rev3 [54], is necessary, because in a rev7D strain the rescue does not occur.
Rev1, which interacts with Pol zeta for efficient bypass and extension past the DNA lesion, is not required to rescue mtDNA extended mutability, suggesting that Pol zeta can function independently of Rev1 in mitochondrion. In contrast, Rev1 overexpression is necessary for the rescue of point mutability mediated by Rev3. These observations suggest that the rescues mediated by Rev3 overexpression are driven by two different mechanisms: i. the rescue of mtDNA extended mutability involving only Pol zeta and ii. the rescue of mtDNA point mutability for which both Pol zeta and Rev1 are required. These results are consistent with previous findings which indicate that the instability of mitochondrial DNA in general is caused by mechanisms different from those producing single point mutations [19].
Deletion of REV3 did not influence the mtDNA extended mutability (petite frequency), both in the wt strain and in mip1 mutant strains. In contrast, REV3 deletion caused an increase of point mutability. Therefore Pol zeta, in mitochondria, functions in a pathway less mutagenic than Mip1, supporting the hypothesis of Kalifa and Sia [10]. Moreover, strains carrying mip1 mutant alleles were themselves mutators, i.e. showed an Ery R mutant frequency higher than that of the wt strain. The observation that the higher the mtDNA point mutability of the mutant strain, the higher the effect of REV3 deletion, indicates that mutator Mip1 isoforms are more sensitive to the absence of Rev3 than the wild type one and is coherent with that hypothesis. Petite frequency was measured as reported in Figure 2. For each mip1 mutant rev1D or rev7D strain, *indicates that the frequency is significantly different (p,0.05) compared to that of strain transformed with the YEplac195 empty plasmid. doi:10.1371/journal.pone.0034322.t004 Until now, it has not been yet elucidated the physiological role of Pol zeta in mitochondria. Rev3, that is expressed at basal levels during the whole vegetative cell cycle and only moderately (,2-fold) overexpressed when cells are treated with UV, is expressed at levels comparable or higher (,18-fold) than those obtained in this study during the late phase of sporulation [58]. Interestingly, Figure 6. Petite frequency of mip1 C261R , mip1 A692T and mip1 G807R strains disrupted in SML1 and overexpressing REV3. In each graph the following petite frequencies are reported: mip1 strains transformed with YEplac195 empty plasmid (wt) or with YEPlac195-REV3 (REV3); mip1 sml1D strains transformed with YEplac195 empty plasmid (sml1D) or with YEplac195-REV3 (sml1D/REV3 observed). In addition, the penultimate bar represents the petite frequency of sml1D strain transformed with YEplac195-REV3 that would be expected if the effects of the overexpression of REV3 and of the deletion of SML1 were additive (sml1D/REV3 expected). * indicates that the observed frequency is significant lower than the expected one. NS: not significant. doi:10.1371/journal.pone.0034322.g006 Figure 7. Rate of Can R mutants. Rate of Can R mutants (A) in wt strains overexpressing REV3, REV7 and/or REV1 or (B) wt REV3 and mutant REV3, encoding for the Rev3 protein harboring the 504-508 AAAAA stretch. Rate of Can R mutants was measured as indicated in Materials and Methods on three independent clones. If no stated differently, * indicates that the rate is significantly different from that of strain transformed with the empty plasmid(s). NS: not significant. doi:10.1371/journal.pone.0034322.g007 maintenance of a full non-mutated mtDNA genome is fundamental in meiosis, since respiratory-deficient diploid strains are unable to sporulate or show increased levels of dead spores [59,60]. A possible interpretation is that increased levels of Pol zeta could have importance in the maintenance of wt mtDNA during sporulation.
Regarding extended mutability, some MIP1 mutations equivalent to human pathological mutations were rescued by overex-pression of Rev3 while others were not. Y757C, E698G and K749R, dominant mutations located in the polymerase domain, together with the recessive mutation in the exonuclease domain G259R, belong to this second group. These mutations are rescued by treatment with antioxidant molecules such as dihydrolipoic acid and MitoQ. Biochemical properties are available only for human Y955C mutation, equivalent to yeast Y757C, that strongly reduces the polymerase activity (,0.03% compared to the wild type), the processivity and the nucleotide binding affinity, but does not affect the DNA binding affinity [61,62]. It was previously reported also that human Pol gamma harboring the Y955C mutation increases the in vitro incorporation of 8-oxo-dGTP, the most abundant oxidized nucleotide in mitochondria, and the translesion synthesis opposite to an 8-oxo-dG residue [63]. The last observation   suggests that, although most of the mtDNA extended mutability by Y757C mutation is caused by stalled replication due to strongly reduced polymerase activity [61,62,64,65], the mtDNA damage can be partially induced also by the accumulation of oxidized bases. In addition, for heteroallelic strains containing dominant mutations, it is possible that replication stalling by mutant Mip1 itself causes oxidative damage on mtDNA, which can partially block replication by wt Mip1. Independently of the mechanisms, it is known that there is a correlation between mtDNA extended mutability and oxidative stress: petite frequency is higher in strains with a reduced antioxidant potential [66], in strains treated with hydrogen peroxide [67] and in strains deleted both in MSH1 and in OGG1, encoding the mitochondrial glycosylase/lyase that specifically excises 8-oxo-dG residues in mtDNA [68]. Differently from Y757C, E698G, K749R and G259R, mutations C261R, R467W, A692T and G807R are rescued by Rev3 overexpression. These mutations are known to reduce the polymerase activity and/or the processivity of Mip1 [20,27]. Yeast C261, equivalent to human S305, is part of a loop-helix orienter module (amino acids 304-316) that creates steric constraints leading to a destabilization of the template strand and a modification in the arrangement of the primer strand [27]: this subdomain in the 39-59 exonuclease plays a crucial role in the DNA polymerase function with no major effect on the exonucleolytic activity. yG807, equivalent to hG1051, is part of a partitioning loop (amino acids 1050-1095) [12] that modulates the partitioning of the primer strand between the polymerase and the exonuclease active sites by forming stable contacts with correctly base-paired primer-template DNA and destabilizing primertemplate DNA that contains mispairs [69] This subdomain is adjacent to the orienter module, whose role is to position correctly the partitioning loop. Mutations in the orienter loop and in the partitioning loop both result in reduced DNA-binding affinity, reduced polymerase activity, increased point mutability and increased extended mutability [20,26,27]. yA692T, equivalent to hA889, is part of a b-sheet which surrounds the catalytic residues in the palm [12]. Mutations in this b-sheet are predicted to reduce polymerase activity [69], and this effect was observed for A692T mutation in yeast [26]. yR467, equivalent to hR574, is part of a subdomain responsible for the intrinsic processivity of Pol gamma [12]. As a matter of fact, mutation yR467W results in a decreased processivity [27].
We showed that these recessive mutations with reduced polymerase activity or processivity are insensitive to antioxidants treatment. Thus, we speculate that different Pol gamma mutations exert their negative effect according to two distinct ways. The first one, rescued by the treatment with antioxidants, can be related to the increased incorporation of 8-oxo-dG: in presence of antioxidant molecules the concentration of oxidized nucleotides and, as a consequence, their incorporation in the mtDNA is reduced. The inability of Pol zeta to rescue these Mip1 mutations may depend partially on the low efficiency (approximately 12%) of Pol zeta to incorporate nucleosides opposite to 8-oxo-dG [70,71]. The second way, rescued by Rev3 overexpression, could be related to reduced polymerase activity, processivity and/or TLS synthesis of mutated mtDNA polymerase as in the case of C261R, R467W, A692T and G807R mutations. Two modes of action may account for ability of Rev3 to rescue this kind of mip1 mutations. i) Pol zeta could participate directly to the replication of undamaged mtDNA at the replicative fork partially playing the role of Mip1, as it was recently demonstrated in the case of defective nuclear replisome due to mutations in the replicative polymerases [72]. Interestingly, the rescue is observed for mutations which cause reduced DNA binding affinity (C261R) or for which reduced DNA binding affinity is predicted (R467W and G807R) [27,69]. ii) Alternately, Pol zeta could partially substitute mutant Mip1 defective of TLS, either during the incorporation opposite to a lesion or the extension from mismatched terminally nucleotides. Reports of translesion synthesis by Pol gamma are limited. In vitro studies showed that Pol gamma from higher eukaryotes is an enzyme with a low TLS activity: it can bypass 8-oxo-dG incorporating dCMP (,73%) or dAMP (,27%), but stalls the majority of the time opposite abasic sites and opposite adducts containing benzo[a]pyrene and benzo[c]phenantrene [73,74]. Moreover, it has been recently reported that human Pol gamma barely bypasses cyclobutane T-T dimers incorporating dAMP (,80%) or dCTP (,20%), whereas it extends very inefficiently past the dimer [75]. These observations suggest that DNA polymerase gamma can promote mutagenesis during TLS.
It is worthwhile mentioning that the Rev3 rescue was observed only in the absence of wt Mip1. It could be possible that in heteroallelic strains wt Mip1 out-competes even a 10-fold overexpression of Pol zeta for the stalled replication that is caused by mutant Mip1. As a consequence, putative beneficial effects of overexpression of Pol zeta in heteroallelic strain could be hidden by the presence of wt Mip1.
It has been demonstrated previously that mtDNA instability caused by specific mutations in MIP1 is reduced by increasing the dNTPs in the cell, either by overexpressing the RNR1 gene, which encodes the large subunit of the ribonucleotide reductase, or by deleting the SML1 gene, which encodes an inhibitor of the latter activity [55]. Increased concentration of dNTPs is beneficial since the basal levels of mitochondrial dNTP pools are a limiting factor to maintain mtDNA copy number in yeast [76,77].
Here we describe that overexpression of Rev3 and increased activity of ribonucleotide reductase have a synergistic effect on reducing the frequency of petites. The synergistic effect is probably due to the fact that the activity of Pol zeta but also of Mip1 have improved with the increase of the dNTP pools, which are the substrate of both DNA polymerases. In addition, the REV3 overexpressing sml1D strains harboring mip1 mutations A692T and G807R reached petite frequency values which are slightly higher or similar, respectively, to those of wt strain, thus indicating that through genetic modifications it is possible to erase completely the detrimental effects of such mip1 mutations on extended mutability.
It is worth mentioning that the human Pol zeta protein does not localize to the mitochondria [9]. From the data here described, we suggest that in mammalian cells carrying mutant Pol gamma a beneficial effect could be obtained by overexpressing the yeast Rev3 mutant isoform unable to enter the nucleus but capable of rescuing mtDNA mutability, or as an alternative a mitochondrial localizing version of human Rev3. It has been reported that expression of a yeast mitochondrial repair/recombination enzyme, Cce1, the cruciform-cutting endonuclease, in human cells harboring partially duplicated mtDNA induced the appearance of wt mtDNA, thus demonstrating that expression of a yeast protein could be beneficial for mtDNA stability in mammalian cells [78]. Such a genetic rescue could be associated to a treatment with dNTPs, which seems to be beneficial for the detrimental effects caused by specific mutations in human POLG: the processivity defect of human DNA polymerase gamma harboring Y955C mutation was overcome by increasing the dATP or dTTP concentration during in vitro synthesis of mtDNA fragments [65] and the treatment with dAMP and dGMP leads to a slight increase in the levels of mtDNA in myoblasts deriving from an Alpers' patient harboring mutations A467T and K1191N in POLG [79].