Introduction

Mismatch repair (MMR) systems are evolutionarily conserved and play a primary role in mutation avoidance by correcting replication errors resulting from nucleotide misincorporation and polymerase slippage (for reviews see Modrich and Lahue 1996; Schofield and Hsieh 2003). In addition, MMR has an important role in ensuring the fidelity of homologous recombination. Indeed, MMR proteins are able to recognize mismatches in recombination intermediates and thus to inhibit recombination between moderately divergent (homeologous) sequences (Rayssiguier et al. 1989; Worth et al. 1994; Matic et al. 1995).

In Escherichia coli, the mismatch repair system requires the proteins MutS, MutL, MutH and MutU (formerly called UvrD), whose functions are quite well understood (Schofield and Hsieh 2003). The MutS protein recognizes and binds mismatches in DNA and recruits MutL. Together, the two proteins are thought to stimulate the nicking activity of MutH. MutH cleaves transiently unmethylated GATC sites in the nascent DNA strand, and provides entry sites for the MutU/UvrD helicase, which melts the nicked strand over a stretch of several kilobases to allow repair synthesis (Modrich and Lahue 1996).

The highly radiation-resistant bacterium Deinococcus radiodurans can reconstitute a functional genome from hundreds of radiation-induced chromosomal fragments. This process occurs mainly through homologous recombination and requires RecA function (for reviews see Battista 1997, 1999). The D. radiodurans genome contains numerous repeated sequences, including expanded families of transposons (Makarova et al. 2001). Since extensive intra- and inter-chromosomal recombination between these sequences can be potentially highly deleterious, D. radiodurans must possess regulatory mechanisms that negatively control these recombination events, in particular, during the reconstruction of its genome after DNA damage. One such mechanism might involve a particular organization of its genome which serves to limit ectopic recombination events (Minton and Daly 1995; Levin-Zaidman et al. 2003). Another, and non mutually exclusive mechanism, may involve the operation of an efficient mismatch repair system (Myung et al. 2001).

Here we analyze the functional status of mismatch repair in this organism. Only the MutSL core of the mismatch repair system appears to have been conserved throughout evolution (for review see Kolodner 1996). A recent phylogenetic analysis of MutS homologs shows that this family can be divided into two major subfamilies, MutS1 and MutS2 (Eisen 1998) which can be aligned only in the central portion of their sequences containing the nucleotide-binding domain (Moreira and Philippe 1999). Only the members of the MutS1 subfamily have been implicated in mismatch repair (Eisen and Hanawalt 1999). D. radiodurans possesses homologs of mutL ( DR1696), mutS1 ( DR1039), mutS2 ( DR1976), and uvrD ( DR1775). However, the sequence data (http://www.tigr.org/tigr-scripts/CMR2/GenePage.spl?locus=DR1039) indicate that the putative product of the mutS1 gene is a truncated protein that is likely to be inactive in mismatch repair. Indeed, the mutS1 homolog from D. radiodurans requires two translational frameshifts to encode a full length protein that aligns with the other members of the MutS1 subfamily.

Here, we show that mismatch repair is functional in the reference wild-type strain of D. radiodurans, ATCC 13939. First, we re-examined the sequence of the DR1039 gene ( mutS1). We found that it contains two additional nucleotides relative to the published sequence, which restore a full-length MutS1 ORF. Second, we show that cells devoid of MutS1 or MutL have a mutator phenotype. These cells also display increased efficiency of integration, after bacterial transformation, of a donor DNA fragment containing a point mutation in the rpoB gene. In contrast, lack of the MutS2 protein had no effect on the levels of spontaneous mutagenesis or recombination. However, mismatch repair-deficient D. radiodurans cells were as resistant to γ-rays, mitomycin C or UV-light as wild-type bacteria, suggesting that the MMR system is not essential for the reconstitution of a functional genome after DNA damage.

Materials and methods

Bacterial strains, cultures, media, and transformation

Bacterial strains are listed in Table 1. The Escherichia coli strains used were DH5α and GM2163. Strain GM2163 was used to propagate plasmids prior to transformation of D. radiodurans, since this process is sensitive to the methylation status of the transforming DNA (Meima et al. 2001). The D. radiodurans R1 strain used in this work was ATCC 13939. Strains GY11713 (Δ mutS1:: tet), GY11712 (Δ mutS2:: tet) and GY11708 (ΔmutL:: tet) were constructed by transformation of wild-type D. radiodurans R1 cells with a PCR fragment containing the mutated allele amplified from plasmid pGY11635, pGY11624 or pGY11625, followed by selection for TetR. Gene replacement in the TetR transformants, and their genetic purity, were verified by PCR using appropriate primer pairs to demonstrate the presence of the two predicted junction fragments and the absence of an internal fragment of the wild-type allele (Supplementary Fig. S1, Supplementary Table S1).

Table 1 Bacterial strains
Full size table

Strains GY11724 (Δ mutS2:: kan) and GY11725 (Δ mutS2:: kan Δ mutS1:: tet) were constructed as follows. A kan cassette (Bonacossa de Almeida et al. 2002) and two 500-bp genomic fragments from upstream and downstream of the coding region of mutS2 were amplified by PCR using primer pairs that introduced Bam HI and Xba I restriction sites. The three fragments were ligated together so that the kan cassette was flanked by the two genomic fragments, and the tripartite ligation product was used to transform D. radiodurans R1 to replace the wild-type allele by homologous recombination. This gave rise to strain GY11724. The DNA from this strain was then used to transform strain GY11713 containing the Δ mutS1:: tet allele, to generate strain GY11725. The genetic structure of GY11725 was checked by PCR as described previously (see Supplementary Table S1 for the diagnostic primer pairs used).

Growth of D. radiodurans and E. coli was as described in Bonacossa de Almeida et al. (2002). When necessary, media were supplemented with the appropriate antibiotics used at the following final concentrations: ampicillin, 100 μg/ml for E. coli; chloramphenicol, 20 μg/ml for E. coli and 3 μg/ml for D. radiodurans; tetracycline, 10 μg/ml for E. coli and 2.5 μg/ml for D. radiodurans; rifampicin, 25 μg/ml for D. radiodurans; kanamycin 6 μg/ml for D. radiodurans; spectinomycin, 40 μg/ml for E. coli and 75 μg/ml for D. radiodurans.

Transformation of D. radiodurans with genomic DNA, PCR products, or plasmid DNA was performed as described by Bonacossa de Almeida et al. (2002).

PCR, oligonucleotide primers and plasmids

PCR amplification of plasmid or genomic DNA was performed with DyNAzyme EXT DNA polymerase (Finnzyme). Oligonucleotide primers used for cloning and diagnostic PCR experiments are listed in Supplementary Table S1.

Plasmids are listed in Table 2. Plasmids pGY11618, pGY11619 and pGY11617, containing the Δ mutS1:: tet, Δ mutS2:: tet, and ΔmuL:: tet alleles used to inactivate the corresponding chromosomal genes, were constructed as follows. First, D. radiodurans genomic fragments spanning mutS1, mutS2 and mutL, respectively, were cloned into the vector pBCSKP to generate the plasmids pGY11618, pGY11619 and pGY11617. Second, the genes were inactivated in vitro by replacing an internal fragment (a Sex AI-ApaI fragment for mutS1, a Spl I-BstEII fragment for mutS2 and a Pvu II fragment for mutL) with a tet cassette. Plasmid pGY11615 containing a tet cassette, which confers tetracycline resistance on D. radiodurans, was constructed by inserting, upstream of the tet gene, between the Pst I and Eco RI sites of plasmid pBR322, a 213-bp PCR fragment spanning the Deinococcal groESL promoter amplified from plasmid pRADZ3.

Table 2 Plasmids
Full size table

Plasmids pGY11637, pGY11633 and pGY11634, used in complementation experiments, were constructed by inserting, between the Bam HI and HindIII sites of the shuttle vector pI8, the Deinococcal mutS1, mutS2 , or mutL genes, respectively.

Plasmid pGY11559, used to assess bacterial competence, was constructed by replacing the cat gene of the shuttle vector pGY11530 with the spec gene from plasmid pGB2 in order to express spectinomycin resistance in D. radiodurans.

Sequencing

To sequence the DR0139 gene of D. radiodurans, a segment (coordinates 1051459 to 1054168 on chromosome I) encompassing the DR0139 gene was amplified from genomic DNA purified from eight independent cultures of strain ATCC 13939. To sequence the rpoB gene for RifR mutations, a segment (coordinates 924608 to 925464 on chromosome I) encompassing a region of the rpoB gene in which most of the RifR mutations occur (Kim et al. 2004), was amplified by PCR from genomic DNA purified from independent RifR mutants. The primers used are described in Supplementary Table S1. Sequencing of the PCR fragments was performed by Genome Express (Grenoble, France).

Determination of the frequency of spontaneous RifR mutants

Single colonies from each strain were inoculated into 3 ml of TGYX2 and grown to an A650=3. The cells were then harvested by centrifugation, resuspended in 300 μl of 20 mM potassium phosphate buffer (pH 7.0) and plated on TGY plates containing 25 μg/ml rifampicin to determine the number of RifR mutants, and on TGY plates without rifampicin to determine the total number of viable cells. The frequencies of the RifR mutants per viable cell from at least 11 independent measurements (from 11 to 47) were used to calculate a median value.

Analysis of mutation rates

The mutation rate (μ) per replication was calculated as the ratio m/N, where m is the number of mutational events per culture and N the total number of cells in the culture. The m value was calculated by the maximal likelihood method described by Lea and Coulson (1949) using the statistical fluctuation analysis program Salvador (Zheng 2002). This program takes into account a theoretical value of ρ (ρ being the ratio of the averaged growth rate of the mutants to the growth rate of the wild type) that is calculated from the distribution of the numbers of mutants in cultures of the same size using the Mandelbrot law. The ρ parameter might be of particular importance in evaluating the μ value in D. radiodurans, since an important segregational lag due to the multiple genome copies present in this organism, as well as a possibly reduced growth rate of the mutants, may lead to an underestimation of the mutational events. However, an evaluation of the rate of spontaneous RifR mutagenesis in wild-type D. radiodurans from a set of 47 independent experiments using the Salvador program gave a μ value (μ=3.78) that was of the same order of magnitude as that obtained (μ=3.54) from the same set of data by the method of Drake (1991) using the formula μ=f /ln(N μ) where f is the median RifR frequency and N is the average number of cells in a culture. Note that the ρ parameter is especially important for mutational events arising early in a culture. If the frequency of mutation is low, it is expected that mutational events essentially occur late in the culture, when the size of the culture is high enough. Under these conditions, the effect of ρ on the μ value could be very weak. This effect could be more important in mismatch repair-deficient strains in which the frequency of mutations is higher than in the wild-type strain.

Results and discussion

The DR1039 gene of the D. radiodurans ATCC 13939 reference strain encodes a full length MutS1 protein without the need for translational frameshifts

An analysis of the annotated D. radiodurans genome sequence (White et al. 1999; http://www.tigr.org/tigr-scripts/CMR2/GenomePage3.spl?database=gdr) reveals that this organism possesses key genes pertaining to the core of the mismatch repair system, mutS ( DR1039 and DR1976), mutL ( DR1696), and mutU / uvrD ( DR1775). Of the two Deinococcal mutS genes, DR1976 encodes a putative protein belonging to the MutS2 subfamily, members of which are found in some bacterial species but are not currently known to play a role in mismatch repair (Eisen 1998). The other, DR1039 , requires two translational frameshifts to encode a putative MutS protein that can be aligned with the other proteins of the MutS1 subfamily known to play a key role in mismatch recognition (Eisen 1998). In the absence of these frameshifts, DR1039 is predicted to encode a truncated, 237-amino acid MutS protein, which is probably inactive.

The data just mentioned would imply that mismatch repair does not operate in wild-type D. radiodurans unless DR1039 is subjected to a particular type of translational regulation. Alternatively, the product of the mutS2 gene, DR1976 , might be unusually active in mismatch repair in D. radiodurans. On the other hand, the recent finding that disruption of the uvrD / mutU ( DR1775) gene results in a mutator phenotype (Kim et al. 2004) suggests that the mismatch repair system may indeed function in D. radiodurans. Given these contrasting data, we sequenced the DR1039 gene in the reference D. radiodurans strain ATCC 13939.

We found that the sequence of DR1039 in strain ATCC 13939 differs at two positions from the published sequence (http://www.tigr.org/tigr-scripts/CMR2/GenePage.spl?locus=DR1039): an additional cytosine is present at position 711 and an additional guanine is found at position 887. These two changes make it possible to construct a full-length alignment of the predicted DR1039 product with MutS1 proteins from Thermus thermophilus and E. coli, with which it shares 49% and 34% sequence identity, respectively (Fig. 1). Accordingly, we named the DR1039 gene mutS1, following the nomenclature of Eisen (1998).

Fig. 1
figure 1

Alignment of the D. radiodurans MutS1 protein with the corresponding E. coli and T. thermophilus MutS proteins. The alignment was generated using the CLUSTALW program. Shading was based on amino acid identity ( black boxes) or similarity ( grey boxes) using the BioEdit program

Full size image

The mutS1 and mutL genes are required for efficient mismatch repair in D. radiodurans

To determine the phenotype of cells lacking the mutS1, mutS2 or mutL gene, and to identify the cellular function of the proteins encoded by these genes, we constructed deletion mutants. For this purpose, the central parts of mutS1, mutS2 or mutL were replaced by a tet cassette in vitro (Supplementary Fig. S1). The mutated alleles were then used to transform D. radiodurans, selecting for tetracycline resistance in order to inactivate the corresponding chromosomal alleles via homologous recombination. The genetic structure of the inactivated genes was verified by PCR (Supplementary Fig. S1).

To determine the impact of the inactivation of mutS1, mutS2 and mutL genes on the rate of spontaneous mutagenesis, we used the RifR forward mutation assay. This assay detects essentially base substitutions that occur at specific positions in the chromosomal rpoB gene which alter the β subunit of RNA polymerase, rendering the enzyme resistant to rifampicin (Campbell et al. 2001). The RifR/ rpoB system has recently been used to evaluate the frequency of spontaneous mutagenesis in D. radiodurans (Kim et al. 2004).

A comprehensive analysis of spontaneous mutation frequencies and rates in the wild type and in strains in which the putative mismatch repair genes have been inactivated is shown in Table 3. In addition to the Δ mutS1, Δ mutS2 and ΔmutL mutants constructed here, we also included in the analysis, as an internal standard, the uvrD::TnDrCat mutant previously shown to possess a mutator phenotype (Kim et al. 2004).

Table 3 Effect of mismatch repair genes on spontaneous mutagenesis
Full size table

We observed a seven-fold increase in the mutation frequencies, and a five-fold increase in the mutation rates, in cells lacking MutS1, MutL or UvrD proteins as compared to the wild type. Furthermore, when the mutated proteins were restored in Δ mutS1 or ΔmutL bacteria by transformation with the wild-type genes cloned in the shuttle vector pI8, the mutation frequencies and rates shifted to approximately wild-type levels. Since both mutS1 and mutL are the first genes of putative operons, this result eliminates the possibility that the mutator phenotype observed in Δ mutS1 or ΔmutL bacteria might be due to a polar effect on the expression of downstream genes.

In contrast to the above results, we did not observe any increase in the level of spontaneous mutagenesis upon deletion of the mutS2 gene (Table 3). However, if MutS1 and MutS2 play a somewhat redundant function in mismatch repair, the phenotype of Δ mutS2 bacteria might be masked by the activity of the MutS1 protein. We thus constructed a mutant lacking both MutS proteins. The mutation frequency in these Δ mutS1 Δ mutS2 bacteria was of the same order of magnitude as that observed in bacteria deleted for the mutS1 gene alone (Table 3).

We conclude that mismatch repair is functional in D. radiodurans. Both MutS1 and MutL proteins, together with the associated UvrD helicase, are required for this process, but MutS2 appears to play no major role.

MutS1 and MutL proteins inhibit recombination between DNAs that differ in only one base pair

To investigate the action of the mismatch repair proteins in ensuring the fidelity of homologous recombination in D. radiodurans, we took advantage of the natural competence of this organism to use genetic transformation as a recombination assay. Although the mechanism of genetic transformation in D. radiodurans has not yet been elucidated, it seems likely to involve processing of donor DNA into single-stranded DNA fragments that are subsequently assimilated into homologous double-stranded DNA through the action of the RecA protein. This process directly produces heteroduplex DNA containing one strand from each parent. In many transformable species, the presence of even a single mismatch within otherwise identical sequences can inhibit transformation (Mejean and Claverys 1984; Claverys and Lacks 1986).

To determine whether a single base pair difference between donor and recipient DNAs also inhibits genetic transformation in D. radiodurans, we used three donor DNAs isolated from spontaneous RifR mutants, each containing a single point mutation in rpoB: an AT→GC transition at position 1259 (mutant rif1), an AT→GC transition at position 1304 (mutant rif2), and a GC→AT transition at position 1319 (mutant rif3). The DNA purified from another spontaneous RifR mutant bearing an in frame 9-bp deletion (Δ1250–1258) in rpoB (mutant rifΔ9) was chosen as the reference donor DNA. In making this choice, we assumed, by analogy with the recognition specificities of the MutS protein from E. coli (Parker and Marinus 1992; Joshi and Rao 2001), that the Deinococcal mismatch repair proteins do not recognize a large 9-nt DNA loop.

When samples containing 2 or 20 ng of each genomic DNA were used to transform wild-type D. radiodurans, the frequencies of the RifR transformants differed greatly, depending on whether the donor DNA contained a point mutation or the Δ9 deletion. Indeed, as can be seen in Fig. 2a, all three point-mutation DNAs were much less efficient in transforming a wild-type host than the DNA containing the 9-bp deletion. The same difference was observed when PCR fragments (850 bp) encompassing each mutation were used for transformation. In this case, using 300 ng of donor DNA in the transformation mixture, the frequencies of RifR transformants were 1.6×10−5, 0.6×10−5, 2.4×10−5, and 20.5×10−5 for the rif1, rif2, rif3 point mutants and the rifΔ9 deletion mutant DNA, respectively.

Fig. 2a–d
figure 2

Effects of MutS and MutL on the inhibition of transformation by a single point mutation in the donor DNA. GY9613 (wild type, panel a), GY11713 (Δ mutS1:: tet, panel b), GY11712 (Δ mutS2:: tet, panel c) and GY11708 (ΔmutL:: tet, panel d) bacteria were transformed with 2 ng and 20 ng of genomic DNA purified from D. radiodurans RifR mutants (rif1, diamonds; rif2, squares; rif3, triangles; rifΔ9, circles) and appropriate dilutions were plated on TGY plates to measure numbers of viable cells and on TGY plates containing 25 μg / ml rifampicin to select RifR transformants. Transformation frequencies were expressed as the number of transformants divided by the total number of viable cells in the transformation mixture

Full size image

To relate the observed marker effect to the activity of the mismatch repair proteins, the same set of genomic DNAs was used to transform Δ mutS1, Δ mutS2, or ΔmutL recipients. As shown in Fig. 2, the three DNAs with point mutations (rif1, rif2, and rif3) transformed recipient bacteria devoid of MutS1 or MutL with approximately the same elevated frequency as the deletion mutant DNA (rifΔ9) (Fig. 2b and d), while they showed a reduced transformation frequency in recipient bacteria devoid of MutS2, which appear to behave in a manner identical to wild-type bacteria (Fig. 2a and c).

We interpret the above data as indicating that an efficient mismatch repair system can abort recombination intermediates containing a single mismatched base, while it tolerates intermediates containing a 9-bp DNA loop. In this process, MutS1 and MutL play a fundamental role, while MutS2 appears to have no major effect. Thus, the requirement for these mismatch repair proteins in the control of recombination fidelity parallels that observed in the process of error avoidance during DNA replication.

Two types of models could account for the MutS1 and MutL-mediated inhibition of recombination between divergent DNA sequences: (1) preferential correction of the donor strand could occur through a strand targeting mechanism involving the presence of single-strand breaks in the initial integration product (Claverys and Lacks 1986), or (2) binding of MutS1 and MutL proteins to the nascent mismatched heteroduplex recombination intermediate directly could prevent further steps in the recombination process, possibly by reducing the rate and the extent of heteroduplex formation, as shown in vitro for the MutS and MutL proteins from E. coli (Worth et al. 1994).

To discriminate between these two models, we analyzed the effect of inactivating the UvrD helicase on the inhibition of recombination between divergent sequences. For this purpose, we measured the efficiencies of transformation of a uvrD::TnDrCat mutant by the previously described set of DNAs containing a point mutation or a 9-bp deletion in rpoB.

Curiously, the uvrD::TnDrCat strain displayed a general reduction in transformation efficiency as compared to the other strains (note the changes in the abscissa and the ordinate between Fig. 2 and Fig. 3). Since the uvrD::TnDrCat strain also displayed reduced efficiency in plasmid transformation (data not shown), it may have a defect in DNA uptake or processing. Whatever the exact reasons for the reduced transformation efficiency in the uvrD::TnDrCat strain, the relative frequency of transformation by the point mutations (rif1, rif2 or rif3) and the deletion (rifΔ9) in the uvrD host was similar to that observed in the wild-type strain. For instance, the ratio of transformation frequencies of the rif1 to rifΔ9 transformants was 0.02 in the wild type (Fig. 2a, 20 ng of input DNA) and 0.02 in the uvrD::TnDrCat recipient (Fig. 3, 200 ng of input DNA).

Fig. 3
figure 3

UvrD is not involved in transformation inhibition by a single point mutation in donor DNA. GY12307 ( uvrD::TnDrCat) bacteria were transformed with 20 ng or 200 ng of genomic DNA purified from D. radiodurans RifR mutants (rif1, diamonds; rif2, squares; rif3, triangles; rifΔ9, circles) and transformation frequencies were measured as described in the legend to Fig. 2

Full size image

The above results suggest that the interaction of the MutS1 and MutL proteins with a mismatched heteroduplex is sufficient to abort recombination in the absence of the UvrD helicase, and that the inhibition of transformation by DNA containing a single point mutation is not due to a preferential mismatch correction in the donor strand.

Cells devoid of MutS1 or MutL proteins are resistant to DNA-damaging agents

To analyze the possible involvement of the D. radiodurans mismatch repair proteins in other repair pathways, we measured the survival of mismatch repair-deficient bacteria after γ-ray irradiation, treatment with mitomycin C or UV irradiation.

As can be seen in Fig. 4, Δ mutS1 and ΔmutL bacteria displayed wild-type levels of resistance to the three DNA-damaging treatments, indicating that the MMR system is not required for overall cell survival after DNA damage. Bacteria devoid of MutS2 protein were also damage resistant, yielding no clues to the biological function, if any, of this Deinococcal protein.

Fig. 4a–c
figure 4

mutS1 -, mutS2 - or mutL -deficient bacteria are resistant to DNA-damaging agents. GY9613 (wild type, diamonds), GY11713 (Δ mutS1:: tet, squares), GY11712 (Δ mutS2:: tet, triangles), GY11708 (ΔmutL:: tet, crosses) and GY12307 ( uvrD::TnDrCat, circles) bacteria were exposed to γ-rays (panel a), mitomycin C ( b) or UV-irradiation ( c) at the doses indicated on the abscissa. a Bacteria were grown in TGY2X to an A650 value of 2–3. The cultures were concentrated ten times in TGY2X and irradiated on ice with a 137Cs irradiation system (Institut Curie, Orsay, France) at a dose rate of 56.6 Gy/min. Following irradiation, diluted samples were plated on TGY plates and incubated for 3–4 days at 30°C before the colonies were counted. b Bacterial cultures (A650=2–3) were serially diluted in TGY2X and plated on TGY plates supplemented with increasing concentrations of mitomycin C. c Bacterial cultures (A650=2–3) were serially diluted in TGY medium, and 10 μl drops of each dilution were deposited on TGY plates and exposed to UV light at a dose rate of 3.5 J/m2/s

Full size image

In contrast, the UvrD-deficient bacteria were moderately sensitive to γ-rays, extremely sensitive to mitomycin C, and resistant to UV irradiation (Fig. 4). In E. coli, the UvrD helicase, in addition to participating in mismatch repair, is also a component of the UvrABC-dependent nucleotide excision repair pathway. The Deinococcal UvrD enzyme may have a similarly broad range of function. The UV resistance of the uvrD mutant (Fig. 4c) can be accounted for by the fact that D. radiodurans possesses a functionally redundant pathway dependent upon the UV-specific endonuclease UvsE, that compensates for the loss of nucleotide excision repair to restore resistance to UV irradiation (Minton 1994; Earl et al. 2002). Our results thus suggest that the UvrD protein does not participate in the UvsE pathway and that, in the absence of UvrD, the UvsE pathway might be more efficient in eliminating UV-induced DNA damage.

Finally, it should be noted that, although prokaryotic and eukaryotic MMR proteins are known to recognize numerous types of DNA damage (for reviews see Schofield and Hsieh 2003; Harfe and Jinks-Robertson 2000), the damage-resistant phenotype of bacteria lacking MutS1 or MutL sheds no light on the capacity of these Deinococcal proteins for lesion recognition. Indeed, the genetic outcomes of an interaction of the MMR proteins with damaged DNA might be largely dependent on the relative affinity for a given lesion of the MMR proteins versus specific DNA repair proteins, and on the capacity of the MMR system to initiate or to interfere with other repair pathways. For example, inactivation of the MMR system in E. coli results in a slight sensitization of the cells to UV-light, but also to increased tolerance to many drugs, since the mismatch repair system stimulates the transcription-coupled repair pathway for UV lesions while it interferes with the repair of drug-induced lesions (for reviews see Harfe and Jinks-Robertson 2000; Schofield and Hsieh 2003). These MMR-related effects on DNA repair may fail to take place or go undetected in D. radiodurans since this organism may possess regulatory mechanisms that favor pre-replicative processing of DNA lesions (Battista 1997), while the MMR proteins mainly recognize mismatches that result from misincorporation opposite DNA damage (Mu et al. 1997). Moreover, D. radiodurans possesses highly redundant repair activities (Makarova et al. 2001) that may mask a possible involvement of the mismatch repair system in a given repair pathway.

Conclusions

Here, we have characterized the mismatch repair system of the D. radiodurans type strain ATCC 13939 and demonstrated that it participates in ensuring the fidelity of DNA replication and recombination. We also showed that the mutS1 gene in this strain contains two additional nucleotides relative to the published sequence, thus yielding a full length and functional MutS1 protein. In a recent report, the groups that published the D. radiodurans genome sequence (White et al. 1999) acknowledged that they had actually sequenced a D. radiodurans strain from the Minton laboratory collection and not the ATCC 13939 type strain, as initially indicated (see Science vol 303, p 766; 2004). If the published sequence data are correct, the sequenced strain is likely to be a mutator strain, as its mutS1 gene encodes a truncated protein. This could explain why it accumulated many mutations, as suggested by the many basepair changes and nucleotide deletions/insertions found between the ATCC 13939 strain and the sequenced strain in several recent reports (this paper; Science vol 303, p 766, 2004; Eggington et al. 2004; Southworth and Perler 2002).

Inactivation of MutS1 or MutL proteins in D. radiodurans resulted in a seven-fold increase in spontaneous RifR mutagenesis, whereas the data of Glickman and Radman (1980), who used the same mutagenesis assay, showed that inactivation of the corresponding proteins in E. coli resulted in an 80- to 100-fold increase. Since the mutation rate per cell per generation in wild-type D. radiodurans was 15-fold higher than that reported for E. coli (Glickman and Radman 1980), the modest mutator effect of inactivating the MutS1 and MutL proteins in D. radiodurans might be related to an inefficient mismatch repair system, rather than to a highly accurate replication apparatus. D. radiodurans does not possess homologs of the MutH and Dam proteins that ensure methylation-directed targeting of mismatch repair to the newly synthesized DNA strand in E. coli. However, the ends present in the leading strand, or the nicks located between Okazaki fragments in the lagging strand of a replication fork, could provide a strand discrimination mechanism in D. radiodurans in which the mismatch repair system may be coupled to DNA replication. Thus, a possible cause of the low efficiency of the mismatch repair system in this organism may be the short time window during which the system can act. Nevertheless, it should be noted that other bacterial species, such as Bacillus subtilis and Streptococcus pneumoniae, which also lack a methylation-directed strand discrimination mechanism, possess mismatch repair systems as effective as that of E. coli (Tiraby and Fox 1973; Rossolillo and Albertini 2001).

Inactivation of the MutS1 or MutL protein in D. radiodurans also resulted in a 10-fold increase in recombination between DNAs that differed by a single base pair, indicating that the two Deinococcal proteins are required to inhibit recombination between non-identical (homeologous) DNA sequences. However, a comparison of the efficiency of the Deinococcal mismatch repair system in inhibition of homeologous recombination with that of other species is complicated by the fact that the recombination efficiencies can vary widely, depending upon the genetic assay used to measure recombination, the type of mismatches present in heteroduplex recombination intermediates, and the degree of divergence between the recombining DNAs (Rayssiguier et al. 1989; Petit et al. 1991; Humbert et al. 1995; Abdulkarim and Hughes 1996; Zahrt et al. 1994; Majewski and Cohan 1998).

The MMR system has been implicated in the prevention of potential lethal rearrangements mediated by dispersed repetitive elements (for review see Harfe and Jinks-Robertson 2000) and in the control of damage-inducible chromosomal duplications (Hill and Combriato 1973; Petit et al. 1991). The D. radiodurans genome contains numerous repeated sequences, including expanded families of transposons (Makarova et al. 2001) that may constitute a source of genome instability upon recombination. By its capacity to inhibit recombination between partially divergent DNA sequences, the Deinococcal MMR system might contribute to the maintenance of genome stability. However, further elucidation of this issue awaits the construction of adequate tester strains to detect damage-induced rearrangements that may occur at a frequency that is too low to significantly affect overall cell survival after DNA damage.