|Home | About | Journals | Submit | Contact Us | Français|
7,8-Dihydro-8-oxoguanine (8-oxoG) is an abundant and mutagenic lesion produced in DNA exposed to free radicals and reactive oxygen species. In Saccharomyces cerevisiae, the OGG1 gene encodes the 8-oxoG DNA N-glycosylase/AP lyase (Ogg1), which is the functional homologue of the bacterial Fpg. Ogg1-deficient strains are spontaneous mutators that accumulate GC to TA transversions due to unrepaired 8-oxoG in DNA. In yeast, DNA mismatch repair (MMR) and translesion synthesis (TLS) by DNA polymerase η also play a role in the prevention of the mutagenic effect of 8-oxoG. In the present study, we show the RAD18 and RAD6 genes that are required to initiate post-replication repair (PRR) are also involved in the prevention of mutations by 8-oxoG. Consistently, a synergistic increase in spontaneous CanR and Lys+ mutation rates is observed in the absence of Rad6 or Rad18 proteins in ogg1 mutant strains. Spectra of CanR mutations in ogg1 rad18 and ogg1 rad6 double mutants show a strong bias in the favor of GC to TA transversions, which are 137- and 189-fold higher than in the wild-type, respectively. The results also show that Polη (RAD30 gene product) plays a critical role on the prevention of mutations at 8-oxoG, whereas Polζ (REV3 gene product) does not. Our current model suggests that the Rad6–Rad18 complex targets Polη at DNA gaps that result from the MMR-mediated excision of adenine mispaired with 8-oxoG, allowing error-free dCMP incorporation opposite to this lesion.
The integrity of DNA in the cell is under constant threat from physical and chemical agents of endogenous and exogenous origin (1–4). Reactive oxygen species (ROS) that escape the cellular metabolism are an important source of DNA damage, which has been involved in pathological processes such as carcinogenesis or aging (1–6). ROS can attack base and sugar moieties in DNA yielding a variety of lesions such as damaged bases, apurinic and apyrimidinic (AP) sites or DNA strand breaks (7). An oxidized guanine, 7,8-dihydro-8-oxoguanine (8-oxoG), is an abundant and potentially mutagenic lesion in DNA (8–12). Several in vitro studies show that 8-oxoG can be bypassed by eukaryotic DNA polymerases or RNA polymerases (9,13–18). However, the ability to incorporate nucleotides opposite 8-oxoG greatly depends upon the enzyme used and the assay conditions (13–18). These studies point to the incorporation of dCMP or dAMP opposite 8-oxoG in the template strand (13–18). Therefore, 8-oxoG is thought to be mutagenic, yielding GC to TA transversions after two rounds of replication (8–12). To counteract the deleterious effects of 8-oxoG, living organisms have evolved robust DNA repair mechanisms. In Escherichia coli, the repair of 8-oxoG mostly relies on two DNA glycosylases, Fpg and MutY, whose simultaneous inactivation results in a strong and specific (GC to TA) spontaneous mutator phenotype (8–10). In Saccharomyces cerevisiae, the main defense against the mutagenic effect of 8-oxoG is also the base excision repair (BER) pathway that is initiated by the Ogg1 protein (12). Yeast Ogg1 is a DNA glycosylase/AP lyase that excises 8-oxoG and 2,6-diamino-4-hydroxy-5-formamidopyrimidine (FapyG) from γ-irradiated DNA (19–22). In S.cerevisiae, the inactivation of the Ogg1 protein results in a specific GC to TA spontaneous mutator phenotype (23). Indeed, the yeast Ogg1 is the functional homologue of the bacterial Fpg, although they do not belong to the same structural family (24,25).
In S.cerevisiae, general DNA mismatch repair (MMR) and error-free translesion DNA synthesis (TLS) cooperate with Ogg1 to prevent the mutagenic effect of 8-oxoG (26,27). The data suggest that MMR acts as a functional homologue of the MutY protein, catalyzing the excision of adenine residues incorporated by DNA polymerase δ opposite to 8-oxoG (26). It should be noted that MutY is absent in S.cerevisiae but present in mammals (12). On the other hand, DNA polymerase η (the RAD30 gene product) is thought to efficiently promote the error-free incorporation of cytosine opposite 8-oxoG (27). It has been proposed (26,27) that 8-oxoG residues escaping repair by Ogg1 are bypassed by DNA Polδ, which inserts an adenine across from the lesion yielding 8-oxoG/A. Afterwards, 8-oxoG/A is recognized by MMR that excises the DNA fragment containing adenine opposite the lesion. The resulting DNA gap is filled in by DNA Polη, which inserts a cytosine opposite 8-oxoG thus regenerating the 8-oxoG/C. Together, these data point to a complex cellular network that prevents the mutagenic effect of 8-oxoG in S.cerevisiae. From this point of view, the spontaneous mutator phenotype of Rad18-deficient strains was challenging. Indeed, rad18 mutant strains exhibit a spectrum of mutations that indicates an accumulation of GC to TA transversions, like Ogg1-deficient cells (28,29). In addition, genetic studies place RAD30 and RAD18 in the RAD6 group of epistasis (30). These observations led us to hypothesize a role of Rad18 and Rad6 in the cellular network that prevents the mutagenic effect of 8-oxoG.
In S.cerevisiae, the RAD18 and RAD6 genes play a prominent role in the post-replication repair (PRR) process thought to come into play when the replication machinery stalls at DNA damage (31–33). Yeast strains lacking Rad6 or Rad18 proteins are highly sensitive to a wide variety of DNA damaging agents such as γ-radiation, UVC-light and alkylating agents (31–36). Biochemical studies show that Rad6 and Rad18 form a heterodimer that binds single-stranded DNA (37,38), and promotes the PRR process (32,33). The Rad6–Rad18 complex is an ubiquitin-conjugating enzyme that catalyzes the mono-ubiquitination of proliferating cell nuclear antigen (PCNA) at Lys-164, allowing the recruitment of DNA polymerases involved in TLS such as Rev1, Polζ or Polη (39–42).
To explore the impact of the RAD18 and RAD6 genes and PRR on the processing of 8-oxoG in S.cerevisiae, we have constructed and characterized ogg1 rad18 and ogg1 rad6 double mutants. The results show a synergistic increase in the rates of spontaneous CanR and Lys+ mutations in the ogg1 rad18 and ogg1 rad6 double mutants, compared to those in the ogg1, rad18 or rad6 single mutants. The analyses of spontaneous CanR mutation spectra reveal that ogg1 rad18 and ogg1 rad6 double mutants accumulate GC to TA transversions, which is the hallmark of 8-oxoG. Our genetic data indicate a role for the RAD18 and RAD6 genes in the cellular ‘replication–repair network’ that prevents mutations at 8-oxoG. Our current model suggests that the role of Rad18–Rad6 relies on the specific recruitment of Polη to perform incorporation of dCMP opposite 8-oxoG after the MMR-dependent removal of adenine from 8-oxoG/A.
Yeast strains were grown at 30°C in YPD medium (1% yeast extract, 1% bactopeptone, 2% glucose, with 2% agar for plates) or yeast nitrogen base dextrose (YNBD) medium (0.17% yeast nitrogen base without amino acids, 2% glucose, with 2% agar for plates) supplemented with appropriate amino acids and bases. All media, including agar, were purchased from Difco. Supplemented YNBD medium lacking arginine but containing canavanine (Sigma) at 60 mg/l was used for the selective growth of canavanine-resistant (CanR) mutants. Pre-sporulation and sporulation media have been described previously (43).
Saccharomyces cerevisiae strains used in the present study are listed in Table Table1.1. All yeast strains are isogenic to wild-type strain FF18733 (MATa, leu2-3,112, trp1-289, his7-2, ura3-52, lys1-1). The RAD6, RAD18, RAD30 and MSH6 gene deletions were performed by a PCR-mediated one-step replacement technique using plasmid pFA6a-kanMX6 (44,45). Mutants were selected on YPD medium containing geneticin (Gibco-BRL) at 200 mg/l. All disruptions were confirmed by PCR on genomic DNA. The primer sequences used for all disruptions are available upon request. Pre-sporulation and sporulation procedures were performed as described previously (43). Micromanipulation and dissection of asci were performed using a Singer MSM System (46). The genotypes of strains were inferred from the segregation patterns and PCR analysis when two genes possessed the same marker. Plasmid pYSB210 contains the OGG1 gene placed under the control of the TPI promoter in the pYX212 plasmid (47). Plasmid pYCP11 contains the RAD18 gene placed under the control of its own promoter in the pYCP50 plasmid, and is a gift by Dr F. Fabre (Fontenay aux Roses, France).
Yeast strains were grown in 2 ml of YPD or YNBD medium with the appropriate amino acids and bases for 2–3 days at 30°C. For each strain, 10 independent cultures were inoculated with about 102 cells and grown at 30°C for 2–3 days. Cell density was measured by plating dilutions on YPD or YNBD agar plates and counting the colonies after 3 days at 30°C. The quantification of canavanine-resistant mutants (CanR) was determined after plating on selective medium (YNBD containing canavanine) (48). The quantification of Lys+ revertants was determined after plating on selective medium (YNBD without lysine). This assay measured the reversion of the suppressible ‘ochre’ nonsense allele lys1-1 (49). Colonies were counted after 4–5 days at 30°C. All experiments were carried out independently 3–5 times. Mutation rates were determined from the number of CanR and Lys+ colonies by the method of the median (50).
For each strain tested, at least 30 independent cultures were inoculated with 102 cells, grown at 30°C in YPD and spread onto YNBD canavanine-containing plates. From each culture, a single mutant colony was selected and streaked onto selective canavanine-containing plates and grown for 3 days at 30°C. Genomic DNA was extracted from saturated 1.5 ml YPD cultures obtained from individual colonies isolated from the canavanine-containing plates using ZymolyaseR (ICN) and the DNeasy Tissue Kit (Qiagen). CAN1 gene was amplified by PCR from genomic DNA with 5CAN-40 (5′-CAGACTTCTTAACTCCTG-3′) and 3CAN1880 (5′-GAAATGTGATCAAAGGTAATAAAACG-3′). Sequencing of the CAN1 gene (1773 bp) was performed using three primers: 3CAN577 (5′-GGAACTTTGTACGTCCAAAATTG-3′), 5CAN650 (5′-GGAACTTAGTGTAGTTGG-3′) and 3CAN1880 (5′-GAAATGTGATCAAAGGTAATAAAACG-3′). Sequence alignment and analysis were performed using the DNA-STRIDER program.
To investigate the capacity of the RAD18 and RAD6 genes and more generally of the PRR pathway in reducing the mutagenic effect of 8-oxoG in S.cerevisiae, we constructed the ogg1 rad18 and ogg1 rad6 double mutants and the ogg1 rad6 rad18 triple mutant, and examined spontaneous mutation rates in these strains. Table Table22 shows that in terms of forward CanR mutation rates, ogg1, rad18 and rad6 single mutants are moderate spontaneous mutator strains: the rank order being ogg1 > rad18 > rad6. Table Table22 also shows that reversion rates of the suppressible ochre nonsense allele lys1-1 are also higher in ogg1 and rad18 single mutants than in the wild-type strain. Importantly, the ogg1 rad18 and ogg1 rad6 double mutants exhibit a synergistic (greater than additive) increase in CanR and Lys+ mutation rates, compared with those of the ogg1, rad18 or rad6 single mutants (Table (Table2).2). Indeed, the ogg1 rad18 and ogg1 rad6 double mutants exhibit 30- and 35-fold enhanced CanR mutation rates, respectively, whereas additive effects would result in 9- and 7-fold increases in CanR mutation rates, respectively (Table (Table2).2). It should be noted that the ogg1 rad6 and ogg1 rad18 double mutants present about the same spontaneous CanR and Lys+ mutation rates (Table (Table2).2). Furthermore, the rate of CanR mutations in the ogg1 rad6 rad18 triple mutant is very similar to those of the ogg1 rad6 and ogg1 rad18 double mutants (Table (Table22).
Since, ogg1, rad6 and rad18 single mutants are spontaneous mutators, the enhanced mutagenesis observed in ogg1 rad6 or ogg1 rad18 double mutants could be due to secondary mutations in unidentified genes. To exclude this possibility, we have measured CanR mutation rates in ogg1 and rad18 single mutants and ogg1 rad18 double mutant harboring plasmids that express either Ogg1 or Rad18 proteins. The results show that Ogg1 and Rad18 suppress the spontaneous mutator phenotype of ogg1 and rad18 single mutants, respectively (Figure (Figure1).1). Figure Figure11 also shows that both Ogg1 and Rad18 greatly reduce the CanR mutation rates in the ogg1 rad18 double mutant. Therefore, the synergistic effect on spontaneous mutagenesis observed in the ogg1 rad18 double mutant strain is most likely due to the inactivation of OGG1 and RAD18 genes. Taken together, these results strongly suggest that RAD18 and RAD6 cooperate with OGG1 to prevent spontaneous mutations in yeast. The critical role played by OGG1 in this process points to endogenous 8-oxoG at the origin of the mutations.
If 8-oxoG was at the origin of the spontaneous mutator phenotype of ogg1 rad6 and ogg1 rad18 double mutants, the spectrum of spontaneous CanR mutations in these strains should be strongly biased in favor of the GC to TA transversions. To test this hypothesis, sequence analysis of a number of independent CanR mutations was performed in wild-type and mutant yeast strains (Table (Table3).3). The results show that the spectrum of CanR mutations in the wild-type strain is mostly composed of base pair substitutions (BPSs) (84%) with an excess of mutations at G/C base pairs (76% of BPS) (Table (Table3).3). It should be noted that the spectrum of CanR mutations in the wild-type strain used in the present study was very similar to that obtained with another wild-type strain (26). Furthermore, spectra of CanR mutations in ogg1, rad6 and rad18 single mutants are consistent with previous studies (26,28,29,51). Interestingly, the CanR mutation spectra derived in rad6 and rad18 single mutants are significantly different, since the level of BPS is reduced in rad6 (64%) compared to rad18 (94%). Relevant to this study, rad18 and ogg1 single mutants present 11- and 24-fold increase in GC to TA transversions compared to the wild-type strain, respectively (Table (Table3).3). Moreover, CanR mutation spectra in ogg1 rad6 and ogg1 rad18 double mutants exhibit a strong bias in favor of GC to TA transversion events that represent by far the major class of mutations (Table (Table3).3). Indeed, GC to TA events are 189- and 137-fold higher in ogg1 rad6 and ogg1 rad18 double mutants, respectively, than in the wild-type strain (Table (Table3).3). Therefore, the synergistic increase in spontaneous CanR mutations in ogg1 rad6 and ogg1 rad18 double mutants is mostly due to high levels of GC to TA tranversions, which convincingly points to 8-oxoG as the origin of the mutations.
Genetic and biochemical studies argue against a direct role of Rad18 and Rad6 in the removal of 8-oxoG in DNA and rather suggest an indirect intervention (32,33). In the current model of PRR, Rad6 and Rad18 form a complex that binds lesion-containing single-stranded DNA at stalled replication forks, Rad18 being the DNA binding protein (32,33). The Rad6–Rad18 complex catalyzes the mono-ubiquitination of target proteins such as PCNA, and promotes TLS, a process that involves DNA polymerases such as Polη (RAD30) or Polζ (REV3) (32,33). Therefore, the role of Rad18 and Rad6 in the prevention of spontaneous mutations at 8-oxoG could rely on the recruitment of DNA polymerase(s). To explore this possibility, we have constructed the rev3 and rad30 single mutants, the ogg1 rev3 and ogg1 rad30 double mutants and the ogg1 rev3 rad18 and ogg1 rad30 rad18 triple mutants. Table Table44 shows that the CanR and Lys+ mutation rates in the ogg1 rev3 double mutant are very similar to those in the ogg1 single mutant. Therefore, the inactivation of Polζ does not affect the CanR or Lys+ mutation rates in Ogg1-deficient strains. Furthermore, the ogg1 rad18 double mutant and the ogg1 rad18 rev3 triple mutant exhibit similar CanR mutation rates (Table (Table4).4). These results strongly suggest that Polζ is not required either to generate, or to prevent mutations at 8-oxoG. Consequently, the role of Rad18 and Rad6 in the prevention of 8-oxoG-induced mutations is presumably not to recruit Polζ at the site of this lesion.
In contrast, the ogg1 rad30 double mutant exhibits a synergistic increase in CanR and Lys+ mutation rates, compared to those in the rad30 and ogg1 single mutants, which confirms the role for Polη in the prevention of mutations at 8-oxoG [Table [Table44 and (27)]. It should be noted that the rad30 single mutant does not exhibit a spontaneous mutator phenotype (Table (Table4).4). Furthermore, the results show that the ogg1 rad18 and ogg1 rad30 double mutants exhibit similar rates of CanR and Lys+ mutation rates (Table (Table4).4). Finally, CanR and Lys+ mutation rates in the ogg1 rad18 rad30 triple mutant are only slightly (~1.5-fold) higher than those in the ogg1 rad18 double mutant (Table (Table4).4). Spectrum of CanR mutations was determined in the rad30 single mutant, the ogg1 rad30 double mutant and the ogg1 rad18 rad30 triple mutant (Table (Table3).3). Spectrum of CanR mutations in the rad30 single mutant and in the wild-type are not significantly different (Table (Table3).3). Furthermore, spectrum of CanR mutations in the ogg1 rad30 double mutant exhibits a strong bias in favor of GC to TA events, which are 148-fold higher than in the wild-type (Table (Table3).3). Spectrum of CanR mutations in the triple mutant is nearly exclusively composed of GC to TA transversions, which are 282-fold higher than in the wild-type (Table (Table3).3). These results suggest that Rad18, Rad6 and Polη act in the same pathway to prevent mutations at 8-oxoG. However, the action of Polη is probably not strictly dependent upon the functional Rad18–Rad6 complex, which could explain the high mutations rates in the ogg1 rad18 rad30 triple mutant.
Several lines of evidence indicate that Msh2–Msh6-dependent long patch MMR is the major pathway by which adenine incorporated opposite 8-oxoG by DNA polymerase δ is corrected in S.cerevisiae (26). In this study, we show that the RAD6 and RAD18 genes also play a critical role in the prevention of mutations at 8-oxoG. To further explore the role of the Rad18–Rad6 complex, we compared the CanR mutation rates in yeast strains whose OGG1, RAD18 or MSH6 genes have been deleted. The results show a synergistic increase in CanR mutations in the ogg1 msh6 double mutant, compared to those of ogg1 and msh6 single mutants [Table [Table55 and (26)]. The results also show that the ogg1 msh6 double mutant and the ogg1 rad18 msh6 triple mutant present about the same CanR mutation rate, which is the highest one observed in this study and similar to that of the ogg1 rad18 rad30 triple mutant (Tables (Tables44 and and5).5). These results indicate that the msh6 mutation is epistatic to the rad18 mutation. Therefore, Rad6 and Rad18 presumably act after the MMR-dependent excision of the DNA fragment that contained adenine mispaired with 8-oxoG.
Recent studies point to the evolution of an efficient and complex ‘Replication–repair’ cellular network that prevents the mutagenic action of endogenous 8-oxoG in bacteria and eukaryotes (8–12). In S.cerevisiae, this network involves at least three components: the 8-oxoG DNA glycosylase Ogg1, the Msh2–Msh6-dependent MMR and the DNA polymerase η. It should be noted that Ogg1 is the essential player, because it is the only one able to remove 8-oxoG from DNA (12). As a consequence, the role of the other members of the network is unambiguously revealed only in Ogg1-deficient cells. Based on genetic and biochemical evidences, a model of repair of 8-oxoG has been proposed. It suggests that DNA Polδ bypasses 8-oxoG during the replication process, mostly incorporating an adenine opposite this lesion yielding 8-oxoG/A. This is in agreement with biochemical studies showing that 8-oxoG is not a strong block for RNA and DNA polymerases (13–18). Although yeast Polδ replicates through 8-oxoG inefficiently in vitro, additional factors such as PCNA may promote bypass of this lesion as described for the human enzyme (13). Afterwards, MMR promotes the removal of the adenine mispaired with 8-oxoG. Finally, Polη incorporates a cytosine across from the lesion (12,26,27).
In the present study, we show that Rad6 and Rad18 are also involved in the ‘replication–repair’ network that prevents mutation at 8-oxoG. How can Rad18 and Rad6 prevent the mutagenic action of endogenous 8-oxoG in DNA? The antimutagenic action of RAD18 and RAD6 could be due to a higher rate of the formation of 8-oxoG in Rad18- or Rad6-deficient cells. Several data argue against this possibility. The over-expression of Ogg1 in the rad18 single mutant does not suppress its spontaneous mutator phenotype, which is not consistent with an enhanced level of 8-oxoG in DNA saturating the wild-type level of Ogg1 (Figure (Figure1).1). In fact, the over-expression of Ogg1 in the rad18 mutant does not significantly reduce its mutator phenotype, which also argues against the idea that 8-oxoG is at the origin of GC to TA transversions in rad18 single mutant strains (28). The results also show that the msh6 mutation is epistatic to rad18 for the spontaneous mutator phenotype in Ogg1-deficient strains, which is not expected if the role of Rad18 and Rad6 was to prevent the formation of 8-oxoG in DNA. Alternatively, Rad6 and Rad18 could be involved in the removal of 8-oxoG, like Ogg1 or in the removal of adenine incorporated opposite 8-oxoG, like MMR or MutY. Two facts argue against this hypothesis. First, the 8-oxoG DNA glycosylase activity in cell-free extracts of the rad18 and rad6 simple mutants is similar to that observed in wild-type extracts, whereas it is abolished in ogg1 extracts (data not shown). Second, the msh6 mutation is epistatic to the rad18 mutation, which strongly argues against a direct involvement of Rad6 and Rad18 in the removal of 8-oxoG in 8-oxoG/C or that of adenine in 8-oxoG/A. Finally, these hypotheses do not fit well with the known properties of the Rad18–Rad6 complex, which primarily acts as a mono-ubiquitin-conjugating enzyme (32,33). Therefore, we favor an indirect role of Rad18 and Rad6 downstream of MMR-dependent excision of adenine in 8-oxoG/A.
Our data are consistent with models where Rad6 and Rad18 prevent the mutagenic effect of 8-oxoG via the modulation of the activity of a DNA polymerase such as Polη. Figure Figure22 summarizes our interpretation of the antimutator role of Rad6 and Rad18 in S.cerevisiae. The novelty of the model relies on the targeting of Polη by the Rad6–Rad18 complex at the 3′-OH primer end of single-stranded DNA gaps that result from the excision of DNA fragments containing adenine opposite 8-oxoG by MMR. The 3′-OH end can be immediately upstream of 8-oxoG in the template strand, allowing error-free dCMP incorporation opposite this lesion and minimal incorporation at undamaged DNA by Polη. This model is consistent with the apparent lack of contribution of Polζ to the mutagenic process at 8-oxoG and the very limited capacity of Polζ to incorporate opposite 8-oxoG (15). In addition, Polζ is not required to extend the primer following Polη-mediated nucleotide incorporation opposite 8-oxoG (27). The model is also consistent with the fact that msh6 or msh2 mutations are epistatic to rad18 or rad30 mutations, for spontaneous mutagenesis in Ogg1-deficient cells. This model allows some speculation at the molecular level concerning the enzymatic properties of the Rad6–Rad18 complex. Rad6–Rad18 could mono-ubiquitinates PCNA at Lys-164 promoting the recruitment of Polη (39–41). It also implies that Rad18 and Rad6 are not only recruited at stalled replication forks but also at single-stranded DNA gaps resulting from the action of the long-patch MMR. The binding of Rad18 at single-stranded DNA gaps may or may not require the presence of DNA damage in the template strand.
The complexity of the network orchestrated by Ogg1 points to 8-oxoG as a major cellular threat. Indeed, Ogg1 is highly conserved in the course of evolution and it has been characterized in S.cerevisiae, Arabidopsis thaliana, Drosophila melanogaster and mammals (52). In mammals, the inactivation of two major components of the network that prevents mutation at 8-oxoG, namely Ogg1 and Myh1, results in a high incidence of lung cancer (53). In contrast, the ogg1−/− and mhy1−/− single mutant mice are viable, and they do not show gross pathological changes compared to wild-type mice (53–55). Dysfunction of the human Rad18 results in hypersensitivity to several carcinogens, like in the yeast Rad18-deficient cells (56). Our results suggest there could be enhanced instability and cancer predisposition in ogg1−/− rad18−/− double mutant mice. Intriguingly, human RAD18 and OGG1 map on chromosome 3p24–25 (56) and 3p25–26 (57), respectively, where deletions are often found in many types of cancer in human (58). Therefore, many of the deletions found in cancers would encompass both OGG1 and RAD18. This could lead to hypersensitivity following the mutagenic and carcinogenic action of the endogenous oxidative stress.
We thank Dr Francis Fabre for yeast strains, plasmids and helpful discussions. We also thank Dr Evelyne Sage and Dr J. Pablo Radicella for their interest in this work. Our thanks are also due to the Centre National de la Recherche Scientifique (CNRS) and the Commissariat à l'Energie Atomique (CEA) for their support. M.P. was supported by CAPES (Post-doctoral fellowship No. 0438-01-4).