INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

In transformation of Streptococcus pneumoniae, double-stranded DNA binds to the membrane and is randomly cleaved (see reference 23 for a review). Then, single-stranded segments enter the cell from a 3' end while the complementary strands are degraded to oligonucleotides with the opposite polarity (32). About half of the entering segments integrate into the chromosome by homologous recombination (21). The recombination process is RecA dependent (37), exchanges strands from 5' to 3' relative to the donor (43), and forms a donor-recipient structure which is heteroduplex when the donor and the recipient sequences are not identical (11). Both strands of the donor DNA have the same probability of entering a cell so that two complementary heteroduplexes are generated in equal frequency among the recipient bacteria (7). Pneumococcal transformation therefore allows study of the in vivo processing of heteroduplexes such as base-base mismatches. In particular, the variability observed in the transformation efficiencies of point mutations led to the discovery of a mismatch repair system (10, 20, 55). This system, called Hex, recognizes the different mismatches with varying efficiencies and induces the complete excision of the donor strand (31). Base mismatches are ranked as a function of decreasing repair efficiency by Hex as follows: G/T = A/C = G/G > C/T > A/A > T/T > A/G > C/C (6). The deletions and the additions of one or two nucleotides lead to mismatches that are very efficiently recognized by Hex (13, 14). The heterologies longer than 2 bases lead to heteroduplexes which are poorly recognized by Hex, and for those longer than 5 bases, there is no repair at all by Hex (13, 22) nor by any bacterial repair or conversion system (42).

A Hex-independent repair, specific for A/G mismatches, was found in S. pneumoniae. The existence of such a system was signalled by the hyperrecombinationi.e., the abnormally high frequency of wild-type transformantsshown by the mutation ami36 when involved in two point crosses (24). The mutation ami36 results from a CG-to-AT transversion and, consequently, forms upon transformation the mismatches A³⁶/G⁺ and C⁺/T³⁶. It was shown that only A/G triggers hyperrecombination (51), leading to an excess of wild-type but not double-mutant transformants (38). In addition, we know that this hyperrecombination requires the pneumococcal polA product, which is a functional homolog of the Escherichia coli PolI protein, and that upon transformation, about 50% of the A³⁶/G⁺ mismatches lead to CG independently of the replication (41). These results have suggested the existence of a repair specifically changing A/G mismatches to CG pairs in S. pneumoniae. In contrast to Hex, this repair seems specific for A/G mismatches, changes A/G into CG irrespective of the recipient strand, and is closely localized around the mismatch (12).

Shortly after the proposal of an A/G-to-CG conversion in S. pneumoniae, a similar repair system was found in E. coli, depending on the gene mutY (2, 26). In vitro analyses have identified MutY as an adenine glycosylase specific for A/G mispairs (3). The complete A/G-to-CG repair requires a short patch resynthesis by the DNA polymerase I (47, 56). Further studies have suggested that the major in vivo substrate for MutY is the adenine from A/7,8-dihydro-8-oxoguanine (8-OxoG) mismatches (34). This finding indicates that MutY, as the 8-OxoG glycosylase MutM and the 8-OxodGTPase MutT, should antagonize the mutagenicity of 8-OxoG, a major product of oxidative damage of DNA (see references 17 and 35 for reviews). The gene mutY is partly homologous to the gene nth of E. coli, which encodes the DNA glycosylase endonuclease III (EndoIII) involved in the removal of oxidatively damaged pyrimidines (36). In particular, MutY and EndoIII display a four-cysteine domain which ligate a [4Fe-4S]²⁺ cluster presumably involved in specific DNA recognition (16, 44).

Based on the conserved domains shared by MutY, EndoIII, and a third related protein (40), we designed degenerate oligonucleotides to probe by Southern and PCR analysis the pneumococcal genome to identify a mutY homologue. Such investigations being unsuccessful we left this question, until most of the pneumococcal genome sequence became available, revealing the presence of a putative mutY homologue. However, this gene lacks the conserved cysteine domain, which may account for our previous failure to identify it by DNA probing. The aim of this work was to characterize the pneumococcal mutY homologue, mainly to analyze its involvement in the hyperrecombination induced by ami36 and its functional relatedness with the MutY protein of E. coli.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains, plasmids, media, transformation procedures, and transformation efficiencies. Strains and plasmids are listed in Table 1. The pneumococcal strains were cultured at 37°C in CAT medium (46). The selection of ami⁺ bacteria was carried out on synthetic medium containing excess isoleucine (50). S. pneumoniae was transformed as described (43). The transformation efficiency of a chromosomal marker is the number of transformants for this marker divided by the number of transformants for a reference marker also carried on the chromosome, in order to correct for fluctuations in competence. Usually the point mutation str41 which confers resistance to streptomycin is the reference marker.

DNA techniques. Plasmids were extracted from E. coli by the alkaline lysis method (48). Chromosomal extraction of S. pneumoniae was performed as described (5). PCR amplifications were done with the Hot Tub DNA polymerase (Amersham). The oligonucleotides used for the amplification of the pneumococcal rpoB region involved in the resistance to rifampin were upstream, 5'-CGCTTCTTTGACCCACGTCG, and downstream, 5'-CCGTCAGCGATGAAATCGCC. Sequencing of the Rif^r mutations was performed directly on the PCR products, with the CircumVent Thermal Cycle DNA Sequencing Kit (New England Biolabs) and using the following internal primer: 5'-GACAATGAAGTCTTGACACC. Sequencing the cloned mutY gene in the region of the expected cysteines was performed on the plasmid pAMY2 using a CEQ 2000 apparatus and a CEQ dye terminator cycle sequencing kit (Beckman). The oligonucleotide used to prime this sequence was: 5'-TGCGGGTCTTGGCGCGTCTG.

Construction of strains disrupted for the mutY homologue. A 575-bp fragment, internal to the pneumococcal mutY homologue, was amplified from chromosomal DNA of the strain R800, using the following primers: upstream, 5'-TTGCCTTGGAGGAGAAGTAA, and downstream, 5'-ATTCTGATATGCCGCACTAA. The PCR product was digested with HindIII and HincII, generating an HindIII-HincII fragment of 227 bp and two flanking fragments. The 229-bp fragment was cloned in pR350 digested with HindIII and HincII. The resulting recombinant plasmid pRT51 was used to transform the S. pneumoniae strains R800, R801, and 553, in order to inactivate by homology-directed insertion the mutY-like gene (30). The transformants resistant to spectinomycin were selected.

Cloning of the pneumococcal mutY homologue. A DNA segment of 1,561 bp including the open reading frame (ORF) containing the mutY homologue was obtained by PCR amplification, using chromosomal DNA of the strain R800 as a template, and the following oligonucleotides: up, 5'-CAGCTTCACCTTGCCGTAGG, and down, 5'-TGCTCTAGCGCTTCACGACC. Further cloning is described in Fig. 1.

View larger version (11K): [in this window] [in a new window]   FIG. 1.   The 1,561-bp fragment (A) containing the pneumococcal mutY was cut with Ssp1, leaving 17 bp before the 35 region of the putative promoter. The resulting 1,402-bp fragment was directly ligated into pAM238 (B) linearized with the blunt cutting enzyme HincII. The ligation mix was used to transform the E. coli strain DH5. Spectinomycin-resistant transformants leading to white colonies on IPTG-X-gal-containing medium were cultured, and their plasmids were extracted and analyzed by restriction digests, EcoRI, PstI, EcoRI/PstI, and HindIII. One recombinant clone, called pAMY2 (C), displaying the 1,402-bp fragment with the mutY gene oriented downstream of the lac promoter, was kept for further analysis.

Homology searches, genome analysis, and multiple sequence alignments. Finding sequences sharing homology with the proteins MutY, EndoIII, and MutM of E. coli was performed with the TBLASTN search on microbial genomes, finished and unfinished (http://www.ncbi.nlm.nih.gov/BLAST /unfinishedgenome.html). Finding the complete sequence of the mutY gene of S. pneumoniae was possible due to the generosity of The Institute for Genomic Research (TIGR) allowing us to access the unfinished pneumococcal genome. Alignments of the protein sequences were performed using the Multalin program (http://www.toulouse.inra.fr/multalin.html).

Nucleotide sequence accession number. The sequence of the internal region of the mutY gene of the pneumococcal strain R800 has been assigned EMBL accession no. AJ271596.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Presence of an atypical mutY homologue in the pneumococcal genome. The hyperrecombination induced by ami36 in pneumococcus relies on an A/G-to-CG repair (51). Such a repair might be mediated by a protein related to MutY, the adenine glycosylase of E. coli which removes A from A/G mispairs, allowing their repair to CG pairs. Inasmuch as more than 90% of the pneumococcal genome sequence was available, we have looked for a mutY-like gene in this sequence. A BLAST search revealed the presence of a putative ORF whose deduced translation leads to a protein containing 391 amino acids and displaying 30% identity with MutY of E. coli (Fig. 2). This ORF lies in the contiguous region Sp80 of the pneumococcal genome sequenced by TIGR. The presence of putative promoter and terminator sequences upstream and downstream of the ORF suggests that it can be transcribed as a single gene. As shown (Fig. 2), the deduced amino acids sequence of the putative pneumococcal mutY gene does not display four cysteines, characteristic of the MutY/EndoIII family, and which are involved in the formation of a [4Fe-4S]²⁺ cluster. We have sequenced the internal region of the cloned mutY gene of our wild-type strain R800 and confirmed this atypical feature.

View larger version (52K): [in this window] [in a new window]   FIG. 2.   Alignment of the MutY amino acid sequence of E. coli (E.col) (GI accession number 1789331) and the 391-amino-acid sequence deduced from a putative ORF found in the S. pneumoniae (S.pne) genome sequence. Shading highlights identical amino acids in both sequences. The asterisks indicate the positions of the four cysteines present in the E. coli protein but lacking in the ORF of S. pneumoniae.

The mutY gene controls hyperrecombination. In crosses involving two linked ami mutations, one donor and one recipient, breaks or recombinations occur between the two sites involved in the cross. Such breaks generate transformants which are either double ami mutants or ami⁺. The mutations of the ami locus confer resistance to methotrexate, but double mutants cannot be distinguished by this procedure, while ami⁺ bacteria are selectable in synthetic medium containing excess isoleucine (Ile^r) (50). The number of breaks and recombination events occurring between two closely linked sites, thus the number of Ile^r transformants, is proportional to the physical distance separating the two sites (4). With regard to this rule of proportionality, ami36 generates an abnormally high frequency of ami⁺ transformants. For example, crosses between ami36 and ami6, 27 bp apart, produce 20 to 25% ami⁺ bacteria while about 1% is predicted by the distance (24). The pneumococcal strains 553 (hexB ami36 mutY⁺) and 553T51 (hexB ami36 mutY::pRT51), were transformed with a chromosomal DNA carrying the mutations ami6 and str41 in order to test whether the hyperrecombination induced by ami36 was affected by the inactivation of mutY. Hex-deficient strains were used as the recipients to avoid interferences between the Hex system and the hyperrecombination process (24). Hyperrecombination was abolished in the strain 553T51 (Table 2). The proportion of double transformants, ami⁺ str41, which lead to bigger colonies than single str41 transformants, was 20 times lower in the strain 553T51, confirming that hyperrecombination was suppressed.

Disruption of the pneumococcal mutY confers a mutator phenotype. Hyperrecombination depends on the repair to CG of an A/G mismatch formed upon transformation. If this repair is also active on the mismatches which form spontaneously within the chromosome, the mutants affected for this function might display a high mutation frequency. The mutation frequencies of resistance to streptomycin (Sm^r), rifampin (Rif^r), and methotrexate (Mtx^r) of the pneumococcal strains R800 and 800T51 were investigated (Tables 3 and 4). Single colonies were picked and used to inoculate liquid cultures that were then grown to stationary phase and plated on selective media (Table 3). Alternatively, the colonies were picked and directly streaked on solid selective media to score the mutants appeared within the colonies (Table 4). While the rate of Sm^r mutants remained unchanged (not shown), both methods indicate that the proportions of Mtx^r and Rif^r mutants were enhanced in 800T51 compared with R800. Streaking the colonies on selective media and scoring the mutants grown on the streaks appears to be an easy way to estimate accurately the mutation frequency.

The pneumococcal mutY gene specifically prevents CG-to-AT transversions. Mtx^r mutations, which inactivate the ami operon, may arise throughout within the 6,000 bp of the ami locus. The sequence determination of the mutations appearing in such a long locus requires at first the localization of these mutations. By contrast, the Rif^r mutations are usually located in the gene rpoB, which encodes the -subunit of the RNA polymerase, and map mostly in a region of about 300 bases, called cluster I in E. coli (19). The determination of the DNA changes which arise in independent Rif^r mutants should reveal the mutator specificity of the strain 800T51. Searching the pneumococcal genome sequence (with the TIGR database) indicates that cluster I is almost identical in pneumococcus and other bacteria, as confirmed recently by Enright et al. (9). A 1,500-bp segment including cluster I was amplified in independent Rif^r mutants. All the amplified fragments carried the Rif^r mutation as verified by their ability to transform a Rif^s strain to the Rif^r phenotype. A third oligonucleotide located 50 bp upstream of cluster I was used to prime the sequencing reactions directly on the PCR products. Out of 23 independent Rif^r mutants, the mutations were found on three codons of cluster I, among which was the codon for serine 495 (Table 5). Rif^r mutations in this codon were previously reported in E. coli but not in S. pneumoniae (9, 19). We have detected GC-to-TA (and CG-to-AT) transversions in three types of Rif^r mutants out of five types characterized. Such transversions represent 3 mutations out of 9 sequenced in R800, and 13 mutations out of 14 sequenced in 800T51. CG-to-AT transversions are therefore specifically enhanced in the strain disrupted for the mutY homologue.

Transformation efficiencies of CG-to-AT transversions. Crosses between ami36 and ami6 produce about 20% ami⁺ transformants instead of the 1% predicted on the basis of the 27 bp separating the two mutations. The correction of A³⁶/G⁺ mismatches to C⁺G⁺ pairs by MutY is responsible for this excess of ami⁺ transformants. In transformations of an ami⁺ strain with ami36 DNA, the MutY correction must occur similarly, changing the same amount of A/G mismatches to CG pairs. In a mutY⁺ strain, the transformation efficiency of ami36 should display a decrease of about 20% compared with the transformation efficiency in a mutY strain. If we assume that in a hexB mutY background, i.e., without any potential repair, the transformation efficiency of ami36 is 1, then a 20% decrease in a hexB mutY⁺ should lead to an efficiency of 0.8. Despite the normal fluctuation occurring in the measurements of the transformation efficiencies, an efficiency of 0.8 should be distinct from an efficiency of 1. We have constructed a hexB mutY strain by disrupting the mutY gene of the hexB strain R801, leading to the strain 801T51. Chromosomal DNA containing the mutations ami36 and str41 was used to transform R801 (hexB mutY⁺ ami⁺) and 801T51 (hexB mutY ami⁺). The same transformations were carried out with chromosomal DNA containing the mutations ami6 and str41. ami6 is a spontaneous mutation corresponding to a GC-to-AT transition (6), which generates upon transformation the mismatches A/C and G/T. On average (Table 6) and in a reproducible way, ami36 displays a transformation efficiency reduced by 0.2 in the mutY⁺ background, while ami6 remains mostly unaffected.

Complementation of a mutY strain of E. coli. To investigate whether the mutY gene of S. pneumoniae could complement a mutY mutant of E. coli, we cloned it into pAM238, a low-copy-number plasmid. In the recombinant plasmid pAMY2, the pneumococcal gene, which nevertheless keeps its own putative promoter, is oriented so that transcription could also occur from the lacZ promoter. The control plasmid pAMY derives from pAMY2 by BamHI-BglII digestion (Fig. 1), leading to a 599-bp deletion, including the promoter and the 545 adjacent base pairs of the pneumococcal mutY gene. The strains AB1157 and AB1157-Y11 containing pAMY2 or pAMY were grown on Luria-Bertani agar plates. Independent colonies were streaked on rifampin-containing medium to analyze the mutator phenotype (Table 7). Suppression of the mutY mutator phenotype was observed with pAMY2, not with pAMY, suggesting that the pneumococcal mutY gene encodes a protein functionally similar to the adenine glycosylase MutY of E. coli. The complementation was confirmed from liquid cultures and from colonies grown on minimal medium allowing or not allowing gene induction from the lacZ promoter (not shown). The repression of the lacZ promoter had no influence on the complementation by pAMY2, suggesting that the cloned gene is transcribed from its own promoter.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The hyperrecombination induced by the mutation ami36 is triggered by an A/G-to-CG correction system requiring the DNA polymerase I of pneumococcus. However, the genetic control of this repair was still unknown. As the sequence flanking ami36 looked important for this hyperrecombination (12, 24), it was actually unclear whether this pathway was more related to MutY, which repairs A/G to CG, or to the VSP pathway, which repairs G/T to GC in defined sequence environments (reviewed in reference 25). As we show here that the disruption of a pneumococcal mutY homologue abolishes hyperrecombination, we can assume that the main protein controlling the A/G-to-CG repair involved in this hyperrecombination is a MutY-like adenine glycosylase. We have found that mutY strains are mutators, suggesting that this pathway is active not only on the mismatches occurring upon transformation, but also on the spontaneous ones. The Rif^r mutation pattern reveals an enhancement of CG-to-AT transversions in the pneumococcal mutY background, which is expected if MutY repairs A/G mismatches to CG pairs. Also confirming the specialized action of mutY is the fact that no increase in the Sm^r mutant frequency was detected in a mutY background. Indeed, only AT-to-CG transversions in the str gene, which encodes the ribosomal protein S12, have been found to confer the Sm^r phenotype (33).

A comparison between the spontaneous mutation frequencies and the transformation efficiencies of the Rif^r mutations characterized in this work is relevant. In a mutY background, we have observed at least a fivefold increase of the frequency of Rif^r mutants (Table 4). These mutants are mostly due to CG-to-AT mutations (Table 5), and we can estimate at least a 10-fold increase of these transversions. Such an increase is observed in a mutY-deficient strain, suggesting that, in a mutY⁺ background, at least 90% of the A/G mismatches are corrected to CG pairs if the pneumococcal MutY is specific for A/G mispairs. Transforming Rif^r bacteria with a DNA mutated to Rif^r by a CG-to-AT change generates either A/G or C/T mismatches in the recipient cells. Assuming that MutY repairs at least 90% of the A/G mismatches to CG pairs and has no effect on C/T mismatches, the transformation efficiency of such Rif^r mutations should be reduced by at least 45% in a mutY⁺ strain compared with a mutY strain. Such a reduction was not observed. The best decrease we have detected is a 20% decrease with Rif^r type 5. The transformation efficiencies of the Rif^r mutations rather suggest that the MutY system has a weak effect on A/G mismatches created upon transformation. To render this observation compatible with the high and specific mutagenicity of mutY strains, we propose that the A/G mismatches which form upon transformation might be structurally different from those which form spontaneously, in particular because upon transformation a three-stranded DNA could be formed.

Alternatively, the best substrate for the pneumococcal system might not be A/G but A/8-OxoG mismatches. In E. coli, A/8-OxoG is thought to be the primary in vivo substrate for the adenine glycosylase MutY and the major cause of CG to-AT-transversions observed in the mutY strains (34). As the pneumococcal wild-type gene cloned on a low-copy-number plasmid suppresses the mutator phenotype of an E. coli mutY mutant, we can assume that the pneumococcal function is similar to that of the E. coli MutY glycosylase and must be able to trigger the change A/8-OxoG to C/8-OxoG. In addition, it is well established in E. coli that MutY cooperates with the proteins MutT and MutM to form an antimutator system specialized against the mutagenic potential of 8-OxoG, which relies mainly on its capability to pair with adenine (53). MutT, first identified as a dGTPase, is mainly involved in the hydrolysis of oxodGTP to oxodGMP (27), and MutM is a DNA glycosylase specialized for the removal of 8-OxoG from C/8-OxoG mispairs (54). Interestingly, the pneumococcal mutX gene was found to be functionally similar to mutT (33), and searching the pneumococcal genome reveals a putative mutM-like gene. The presence of three genes in S. pneumoniae, homologous to mutY, mutT, and mutM, strongly suggests that the same antimutator system exists in S. pneumoniae. The oxidative stress and the mutagenicity it generates must be actual problems for this nonrespiratory bacterium.

The deduced amino acid sequence of the pneumococcal MutY does not display the highly conserved stretch of four cysteines, Cys-X₆-Cys-X₂-Cys-X₅-Cys, which coordinates a [4Fe-4S]²⁺ cluster loop (36), called FCL. Searching the genome sequences reveals no eukaryotic MutY lacking this domain (not shown) and just two bacterial MutY proteins lacking the cysteine domain, those of Treponema pallidum and Streptococcus pyogenes (Fig. 3). The MutY of S. pyogenes is similar to the pneumococcal one, in particular in the region where the cysteines were expected (Fig. 3), suggesting that the FCL domain was absent before divergence of the two species. In E. coli this domain is reported to be crucial for the specific DNA substrate recognition by MutY (16, 44). As the pneumococcal mutY displays a CG-to-AT antimutator action which complements the MutY deficiency in E. coli, the absence of an FCL domain in the pneumococcal protein is intriguing. Structural differences in the protein of S. pneumoniae might compensate for the absence of an iron-sulfur cluster. The C-terminal domain of the pneumococcal protein, which is reported in E. coli to be more specifically involved in 8-OxoG recognition (15, 39), is more divergent than the rest of the protein (Fig. 2) and could lead to structural differences. Nevertheless, despite its antimutator effect complementing in vivo the MutY deficiency of E. coli, the pneumococcal glycosylase might display a weaker activity than the E. coli glycosylase especially toward the A/G mismatch. In E. coli, although A/G is not the main in vivo substrate of MutY, in vitro studies indicate that the relative glycosylase activity of MutY toward A/G is not weaker than toward A/8-Oxo-G (34), although the kinetic and the mechanism of the reaction are not identical for A/G and A/8-oxoG (45). In S. pneumoniae, the correction by MutY of A/G mismatches is undetected upon transformation of Rif^r mutations types 1 and 3 (Table 6). It is detected but weak upon transformation of ami36 and, less significantly, of Rif^r mutation type 5 (Table 6). Previous studies have indicated that the sequence adjacent to the A/G mismatch is important for hyperrecombination and should be related to the structure generated by ami36: 5'ATTAAT-3'TAAGTA (12). The heteroduplex structures generated by the Rif^r mutations are as follows, for type 1, 3, and 5, respectively: 5'CCAAGT-3'GGTGCA, 5'TCTAAC-3'AGAGTG, and 5'TTTATG-3'AAAGAC. Out of the three tested Rif^r mutations, type 5 only might have undergone a detectable repair by MutY. Interestingly, the heteroduplex structure generated by Rif^r mutation type 5 is quite related to the one generated by ami36. In particular three AT pairs are localized 5' to the mismatched adenine in both ami36 and Rif^r type 5. An attractive hypothesis is that the absence of an FCL domain on the pneumococcal MutY is specifically detrimental to the repair of A/G mismatches, which consequently might occur in a few sequence contexts only. Although the biological importance of A/G mismatches arising in vivo is unknown, they may represent a minor source of CG-to-AT transversions. The relatively high AT content of the pneumococcal chromosome, 60% compared with 50% in E. coli, might originate in part in the inefficient repair of A/G mismatches by a handicapped MutY protein.

View larger version (39K): [in this window] [in a new window]   FIG. 3.   Alignment of the cysteine-containing region in some bacterial MutY homologues. Asterisks indicate the position of the cysteines. Shading highlights amino acid identity in at least 6 sequences out of 10. Origins of the proteins and their GI accession numbers (in parentheses) are as follows: H.pyl, Helicobacter pylori (4154640); B.sub, Bacillus subtilis (2633186); H.inf, Haemophilus influenzae (1573768); E.col, E. coli (1789331); N.men, Neisseria meningitidis (3860539); S.pne, S. pneumoniae; S.pyo, Streptococcus pyogenes; M.tub, Mycobacterium tuberculosis (2950412); S.coe, Streptomyces coelicolor (4585587); T.pal., T. pallidum (3322622).

The unusual lack of the FCL domain in the pneumococcal MutY made us wonder whether similar features are also absent from related proteins in S. pneumoniae. We have detected in the pneumococcal genome a putative MutM homologue and a putative EndoIII homologue. MutM is an oxoG glycosylase specific for A/8-OxoG mismatches and displays a zinc finger motif coordinated by four cysteines (54). The same structure was found in the putative MutM homologue of S. pneumoniae. EndoIII has a glycosylase activity specific for damaged pyrimidines and usually displays a four-cysteine domain identical to the MutY one, also ligating a [4Fe-4S]²⁺ cluster. Interestingly, the putative EndoIII found in S. pneumoniae displays only the two first cysteines of a putative four-cysteine domain (not shown). To support this observation, homologues of the pneumococcal EndoIII protein were searched for in the microbial genomes. An amino acid sequence sharing the same half-cysteine domain was found in Streptococcus mutans only, a streptococcal species strongly related to S. pneumoniae. This observation suggests that the EndoIII of some streptococcal species has lost the carboxy-terminal part of an ancestral cysteine domain and should lack the iron-sulfur cluster. That an iron-sulfur cluster is lacking in EndoIII should not indicate that the activity is lost. Two EndoIII activities have been characterized in Saccharomyces cerevisiae, one of which lacks the [4Fe-4S]²⁺ cluster. Both are active but display some differences in the substrate specificities (1, 49). Concerning S. pneumoniae, the absence of a [4Fe-4S]²⁺ ligating domain in both MutY and the putative EndoIII suggests that both proteins have evolved so that the iron-sulfur cluster is not required. It is tempting to speculate that the scarcity of iron in the natural environment of S. pneumoniae and its importance for bacterial growth and virulence (18, 52), have partly oriented such an evolution.