Introduction

It is generally accepted that the accumulation of free oxygen during the Precambrian era, both in the atmosphere and in local environments, must have exerted a severe selection pressure on early microbial species that had to cope with the resulting novel stress conditions. It has been argued that oxygen-induced damage must have led to a number of protection and detoxification mechanisms, some of which may have led to oxygen utilization as in aerobic respiration (Margulis 1993; Falkowski 2006). Although the significance of a number of molecular mechanisms that protect against oxygen-induced damage such as those involving superoxide dismutases and catalases has been acknowledged, little or no attention has been given to the evolutionary significance of the repair of oxidized methionine.

The oxidation of methionine residues in proteins leads to a diastereomeric mixture of methionine S-sulfoxide and methionine R-sulfoxide (Sharov et al. 1999; Sharov and Schoneich 2000). The reduction of the diastereoisomers of oxidized methionine can be explained by sulfenic acid chemistry (Boschi-Muller et al. 2005) catalyzed by two different monomeric methionine sulfoxide reductases (Kumar et al. 2002). The activity of methionine sulfoxide reductase A (MsrA) is restricted to methionine S-sulfoxide residues in proteins, while methionine sulfoxide reductase B (MsrB) reduces methionine R-sulfoxide residues.

The two methionine sulfoxide reductases are broadly distributed nonhomologous enzymes (Table 1) that have been known for a long time (cf. Grimaud et al. 2001). MsrA is a well-studied protein that is an essential regulator of longevity in a wide range of microbial and animal species (Ruan et al. 2002). Its role as a physiological virulence determinant in Mycoplasma genitalium (Dhandayuthapani et al. 2001) and the plant pathogen Erwinia chrysanthemi (El Hassouni et al. 1999) has been established, and the enzyme is known to play a role in protecting eye lens cells against cataracts (Kantorow et al. 2004). Sequence comparisons have indicated that MsrB, which are absent in some species that are endowed with MsrA (Table 1), can be divided into two families distinguished by the presence or absence of zinc (Kryukov et al. 2002; Kumar et al. 2002).

Table 1 Phylogenetic distribution of MsrA and MsrB sequences in fully sequenced cellular genomes (December 2004)
Full size table

There is increasing recognition of the medical and biochemical significance of MsrA and MsrB. However, as of today, no attempt has been made to analyze these enzymes from the viewpoint of microbial evolution during the Precambrian or to discuss their significance as a response to the accumulation of oxygen in the terrestrial environment. The current availability of a large number of completely sequenced genomes now allows us to discuss the relationship between organism lifestyles and the presence or absence of MsrA and MsrB, as well as the phylogenetic patterns exhibited by these enzymes. Although the early evolutionary history of these enzymes has been partially obscured by a number of lateral gene transfer events, the results presented here may be interpreted as suggesting that the sequences encoding MsrA and MsrB originated in the bacterial domain and spread from it into the Archaea and the Eukarya.

Materials and Methods

MsrA (b4219) and MsrB (b1778) sequences from Escherichia coli were downloaded from the Kyoto Encyclopedia of Genes and Genomes, KEGG database (http://www.genome.jp/kegg/) (Kanehisa and Goto 2000). Homologues to these two sequences were searched in a database of 159 completely sequenced genomes from the three domains of life available as of December 2004, using BLAST search (Altschul et al. 1997). Results are shown in Table 1 (http://bacteria.fciencias.unam.mx/Msr/Msr.html). The lifestyles recorded in Table 1 follow the description given in the original publication of each genome sequence. Sequences were aligned using ClustalX (Thompson et al. 1997), and the result was visually edited using Bioedit software (Hall 1999). In order to perform phylogenetic analysis, minimum evolution trees using MEGA3 software version 3.1 (Kumar et al. 2004) were calculated for MsrA and MsrB separately (100 bootstrap replications each). Uniform rate variation was assumed along all sites. A Poisson correction was employed to calculate distances between the sequences compared in this analysis. All trees were rooted using the midpoint value using the MEGA3.1 software. The complete deletion option available in the MEGA3.1 program was employed to avoid considering positions exhibiting gaps. In the resulting trees, only clades with >70% bootstrap support were depicted, and branches harboring species from the same taxonomic group were pulled together.

Figure 1 shows cladograms of naturally (Fig. 1a) and artificially (Fig. 1b) concatenated MsrA and MsrB sequences. The tree in Figure 1b was constructed by concatenating in silico the MsrA and MsrB sequences from those species whose genomes are endowed with only a single copy of each gene. The hypothesis underlying this approach is that concatenated sequences with significant bootstrap values may have been inherited jointly (Fig. 1b). Concatenation of sequences has been recognized as enhancing phylogenetic signals (Brown et al. 2001; Rokas et al. 2003). Although leaving out of the sample under consideration all the genomes endowed with paralogous sequences reduces the number of species from 159 to 72, a significant gain in phylogenetic signals is achieved, as indicated by the bootstrap values. Zn-bearing MsrB sequences were identified on the basis of the conserved Cys81XXCys84...Cys101XXCys104 motif described by Kumar et al (2002) for the Drosophila melanogaster enzyme. Species endowed with Zn-bearing MsrB sequences are indicated by a black rhombus in the trees presented here.

Fig. 1
figure 1figure 1

a Minimum evolution phylogeny of naturally concatenated MsrA and MsrB proteins, calculated on the basis of 100 bootstrap replications (259 amino acid sites and 26 sequences). Only branches with more than 70% bootstrap are shown in this tree. Root is placed on midpoint. Naturally fused MsrA-MsrB sequences are indicated by a black-white rectangle. Species codes: Proteobacteria γ Pasteurellales—Haemophilus influenzae Rd (hin); Proteobacteria γ Vibrionales—Vibrio cholerae El Tor N16961 (serotype O1) (vch), Vibrio vulnificus CMCP6 (vvu), Vibrio vulnificus yj016 (vvy), Vibrio parahaemolyticus RIMD 2210633 (vpa); Proteobacteria β—Neisseria meningitidis MC58 (serogroup B) (nme), Neisseria meningitidis Z2491 (serogroup A) (nma), Nitrosomonas europaea ATCC 19718 (neu); Proteobacteria ε—Helicobacter pylori 26695 (hpy), Helicobacter pylori J99 (hpj); Firmicutes Bacillales—Bacillus anthracis Ames (ban), Bacillus cereus ATCC 14579 (bce), Oceanobacillus iheyensis HTE831 (oih); Firmicutes Lactobacillales—Streptococcus pyogenes SF370 (serotype M1) (spy), Streptococcus pyogenes MGAS8232 (serotype M18) (spm), Streptococcus pyogenes MGAS315 (serotype M3) (spg), Streptococcus pyogenes SSI-1 (serotype M3) (sps), Streptococcus pneumoniae TIGR4 (spn), Streptococcus pneumoniae R6 (spr); Actinobacteria—Bifidobacterium longum NCC2705 (blo); Fusobacteria—Fusobacterium nucleatum ATCC 25586 (fnu); Bacteroid—Bacteroides thetaiotaomicron VPI-5482 (bth), Porphyromonas gingivalis W83 (pgi). b Minimum evolution phylogeny of in silico concatenated MsrA and MsrB proteins from those species whose genomes are endowed with only one single copy of each gene. 100 bootstrap replications (225 amino acid sites and 67 sequences). Only branches with more than 50% bootstrap are shown in this tree. Black rhombus, MsrB sequence with zinc motif; white rhombus, MsrB sequence without zinc motif. Root is placed on midpoint. Species codes: Eukarya—Caenorhabditis elegans (cel), Dictyostelium discoideum (ddi), Saccharomyces cerevisiae (sce), Schizosaccharomyces pombe (spo); Proteobacteria γ Enterobacteriales—Escherichia coli K-12 MG1655 (eco), Escherichia coli K-12 W3110 (ecj), Escherichia coli O157:H7 EDL933 (ece), Escherichia coli O157:H7 Sakai (ecs), Escherichia coli CFT073 (ecc), Salmonella typhi CT18 (sty), Salmonella typhi Ty2 (stt), Salmonella typhimurium LT2 (stm), Yersinia pestis CO92 (ype), Yersinia pestis KIM (ypk), Shigella flexneri 301 (serotype 2a) (sfl), Shigella flexneri 2457T (serotype 2a) (sfx), Photorhabdus luminescens laumondii TTO1 (plu); Proteobacteria γ Pasteurellales—Pasteurella multocida PM70 (pmu); Proteobacteria γ Xanthomonadales—Xylella fastidiosa 9a5c (xfa), Xylella fastidiosa Temecula1 (xft), Xanthomonas campestris pv. campestris ATCC 33913 (xcc), Xanthomonas axonopodis pv. citri 306 (xac); Proteobacteria γ Pseudomonadales—Pseudomonas aeruginosa PA01 (pae), Pseudomonas putida KT2440 (ppu), Pseudomonas syringae pv. tomato DC3000 (pst); Proteobacteria δ proteobacteria—Geobacter sulfurreducens PCA (gsu); Proteobacteria β proteobacteria—Ralstonia solanacearum GMI1000 (rso), Bordetella bronchiseptica (bbr), Bordetella parapertussis 12822 (bpa), Bordetella pertussis Tohama I (bpe), Chromobacterium violaceum ATCC 12472 (cvi); Proteobacteria ε—Campylobacter jejuni NCTC11168 (cje), Helicobacter hepaticus ATCC 51449 (hhe); Proteobacteria α—Agrobacterium tumefaciens C58 (UWash/Dupont) (atu), Agrobacterium tumefaciens C58 (Cereon) (atc), Brucella melitensis 16M (bme), Brucella suis 1330 (bms); Firmicutes Bacillales—Bacillus subtilis 168 (bsu), Listeria monocytogenes EGD-e (lmo), Listeria innocua CLIP 11262 (lin); Firmicutes Lactobacillales—Streptococcus pyogenes SF370 (serotype M1) (spy), Streptococcus pyogenes MGAS8232 (serotype M18) (spm), Streptococcus pyogenes MGAS315 (serotype M3) (spg), Streptococcus pyogenes SSI-1 (serotype M3) (sps), Streptococcus agalactiae 2603 (serotype V) (sag), Streptococcus agalactiae NEM316 (san), Enterococcus faecalis V583 (efa); Firmicutes Clostridia—Clostridium acetobutylicum ATCC824 (cac), Clostridium perfringens 13 (cpe); Firmicutes Mollicutes—Mycoplasma genitalium G-37 (mge), Mycoplasma pneumoniae M129 (mpn), Mycoplasma pulmonis (mpu), Mycoplasma gallisepticum R (mga); Actinobacteria—Mycobacterium tuberculosis H37Rv (lab strain) (mtu), Mycobacterium tuberculosis CDC1551 (mtc), Corynebacterium glutamicum ATCC 13032 (cgl), Corynebacterium efficiens YS-314 (cef), Streptomyces coelicolor A3(2) (sco), Streptomyces avermitilis (sma), Corynebacterium diphtheriae NCTC 13129 (cdi), Mycobacterium bovis AF2122/97 (mbo); Cyanobacteria—Thermosynechococcus elongatus BP-1 (tel); radioresistant bacteria—Deinococcus radiodurans R1 (dra); Euryarchaeota—Methanosarcina acetivorans C2A (mac), Methanosarcina mazei Goe1 (mma), Methanobacterium thermoautotrophicum deltaH (mth), Halobacterium sp. NRC-1 (hal).

Full size image

Quantitative estimates of the levels of agreement between the MsrA and the MsrB phylogenies were performed by comparing the number of symmetric differences that exist among the 16/18SrRNA, MsrA, MsrB, and concatenated MsrA-MsrB trees, with a randomly generated 16/18SrRNA tree that included all species under consideration, as well as with a tree built with randomly concatenated MsrA-MsrB sequences. The latter tree was constructed using a random number-based algorithm that generated chance concatenated MsrA-MsrB sequences from our database. The number of symmetric differences was calculated using the treedist algorithm from the PHYLIP 3.65 package (Felsenstein 1989). The statistical significance of the symmetric differences among these various trees was estimated by comparison with 100,000 randomly generated trees (Supplementary Table 3) built using a perl script (available upon request).

The three-dimensional structure classification of MsrA and MsrB used here follows the Structural Classification of Proteins, SCOP database (Murzin et al. 1995). Sequence alignments and all other supplementary information are available at http://bacteria.fciencias.unam.mx/Msr/Msr.html.

Results

The phylogenetic distributions of MsrA and MsrB are listed in Table 1, and extend and confirm previous reports by Kryukov et al. (2002) and Ezraty et al. (2005). With the exception of well-characterized endosymbionts (Buchnera, Tropheryma, Wigglesworthia, and Blochmannia) and some endoparasites (Phytoplasma, Rickettsia, and Chlamydia), almost all of the reported bacterial genomes, including the mycoplasma, are endowed with at least one copy of MsrA and one of MsrB. Two exceptions are Mycoplasma penetrans and Ureaplasma urealyticum (serovar 3), which have one MsrA copy each but lack MsrB. The only free-living bacterial species in the sample studied here that lack both sequences are the anaerobic pathogen Wolinella succinogenes, the hyperthermophilic microaerophile Aquifex aeolicus, and the anaerobe Thermotoga maritima.

Several bacterial species present more than one copy of MsrA and/or MsrB (Table 1). Phylogenetic analysis of this set of possible paralogous genes (Supplementary Fig. 4) does not indicate any obvious evolutionary trend. Although it could be argued that some of this redundancy is due to lateral gene transfer, the bootstrap values of these two trees (Figs. 2a and b) do not allow any major evolutionary inference. Fused MsrAB or MsrBA sets appear to be the outcome of polyphyletic arrangements. However, in the tree shown in Fig. 2b the phylogenetic distribution of Zn-bearing MsrB sequences is not random, and it is interesting to note that no firmicutes or naturally fused MsrAB pairs are endowed with the Zn-binding motif. This suggests that MsrB that took part in the ancestral fusion that led to the MsrAB arrangements lacked Zn-binding abilities.

Fig. 2
figure 2figure 2figure 2figure 2figure 2figure 2figure 2figure 2

a Minimum evolution MsrA phylogeny, 100 bootstrap replications (134 amino acid sites and 106 sequences). Only branches with more than 70% bootstrap are shown in this tree. Clades with lesser values have been collapsed. Root is placed on midpoint. b Minimum evolution MsrB phylogeny, 100 bootstrap replications (103 amino acid sites and 125 sequences). Only branches with more than 70% bootstrap value are shown. Root is placed on midpoint. Zinc-bearing MsrB sequences are indicated with a black rhombus. Naturally MsrA-MsrB fused sequences are indicated by a black-white rectangle, while naturally fused MsrB-MsrA sequences are indicated by a white-black rectangle. Species codes: Proteobacteria γ Enterobacteriales—Escherichia coli K-12 MG1655 (eco), Escherichia coli K-12 W3110 (ecj), Escherichia coli O157:H7 EDL933 (ece), Escherichia coli O157:H7 Sakai (ecs), Escherichia coli CFT073 (ecc), Salmonella typhi CT18 (sty), Salmonella typhi Ty2 (stt), Salmonella typhimurium LT2 (stm), Yersinia pestis CO92 (ype), Yersinia pestis KIM (ypk), Shigella flexneri 301 (serotype 2a) (sfl), Shigella flexneri 2457T (serotype 2a) (sfx), Photorhabdus luminescens laumondii TTO1 (plu); Proteobacteria γ Pasteurellales—Pasteurella multocida PM70(pmu), Haemophilus influenzae Rd (hin), Haemophilus ducreyi 35000HP (hdu); Proteobacteria γ Xanthomonadales—Xylella fastidiosa 9a5c (xfa), Xylella fastidiosa Temecula1 (xft), Xanthomonas campestris pv. campestris ATCC 33913 (xcc), Xanthomonas axonopodis pv. citri 306 (xac); Proteobacteria γ Vibrionales—Vibrio cholerae El Tor N16961 (serotype O1) (vch), Vibrio vulnificus CMCP6 (vvu), Vibrio vulnificus yj016 (vvy), Vibrio parahaemolyticus RIMD 2210633 (vpa); Proteobacteria γ Pseudomonadales—Pseudomonas aeruginosa PA01 (pae), Pseudomonas putida KT2440 (ppu), Pseudomonas syringae pv. tomato DC3000 (pst); Proteobacteria γ Alteromonadaceae—Shewanella oneidensis MR-1 (son); Proteobacteria γ Legionellales—Coxiella burnetii SA 493 (cbu); Proteobacteria δ proteobacteria—Geobacter sulfurreducens PCA (gsu); Proteobacteria β proteobacteria—Ralstonia solanacearum GMI1000 (rso), Nitrosomonas europaea ATCC 19718 (neu), Bordetella bronchiseptica (bbr), Bordetella parapertussis 12822 (bpa), Bordetella pertussis Tohama I (bpe), Chromobacterium violaceum ATCC 12472 (cvi), Nitrosomonas europaea ATCC 19718 (neu); Proteobacteria ε—Helicobacter pylori 26695 (hpy), Helicobacter pylori J99 (hpj), Campylobacter jejuni NCTC11168 (cje), Helicobacter hepaticus ATCC 51449 (hhe); Proteobacteria α—Mesorhizobium loti MAFF303099 (mlo), Sinorhizobium meliloti 1021 (sme), Agrobacterium tumefaciens C58 (UWash/Dupont) (atu), Agrobacterium tumefaciens C58 (Cereon) (atc), Brucella melitensis 16M (bme), Brucella suis 1330 (bms), Bradyrhizobium japonicum USDA110 (bja), Caulobacter crescentus (ccr), Rhodopseudomonas palustris CGA009 (rpa); Firmicutes Bacillales—Bacillus subtilis 168 (bsu), Bacillus halodurans C-125 (bha), Bacillus anthracis Ames (ban), Bacillus cereus ATCC 14579 (bce), Oceanobacillus iheyensis HTE831 (oih), Staphylococcus aureus N315 (meticillin-resistant) (sau), Staphylococcus aureus Mu50 (vancomycin-resistant) (sav), Staphylococcus aureus MW2 (sam), Staphylococcus epidermidis ATCC 12228 (sep), Listeria monocytogenes EGD-e (lmo), Listeria innocua CLIP 11262 (lin); Firmicutes Lactobacillales—Lactococcus lactis IL1403 (lla), Streptococcus pyogenes SF370 (serotype M1) (spy), Streptococcus pyogenes MGAS8232 (serotype M18) (spm), Streptococcus pyogenes MGAS315 (serotype M3) (spg), Streptococcus pyogenes SSI-1 (serotype M3) (sps), Streptococcus pneumoniae TIGR4 (spn), Streptococcus pneumoniae R6 (spr), Streptococcus agalactiae 2603 (serotype V) (sag), Streptococcus agalactiae NEM316 (san), Streptococcus mutans UA159 (serotype C) (smu), Lactobacillus plantarum WCFS1 (lpl), Enterococcus faecalis V583 (efa); Firmicutes Clostridia—Clostridium acetobutylicum ATCC824 (cac), Clostridium perfringens 13 (cpe), Clostridium tetani E88 (ctc); Firmicutes Mollicutes—Mycoplasma genitalium G-37 (mge), Mycoplasma pneumoniae M129 (mpn), Mycoplasma pulmonis (mpu), Mycoplasma penetrans HF-2 (mpe), Ureaplasma urealyticum (serovar 3) (uur), Mycoplasma gallisepticum R (mga); Actinobacteria—Mycobacterium tuberculosis H37Rv (lab strain) (mtu), Mycobacterium tuberculosis CDC1551 (mtc), Mycobacterium leprae TN (mle), Corynebacterium glutamicum ATCC 13032 (cgl), Corynebacterium efficiens YS-314 (cef), Streptomyces coelicolor A3(2) (sco), Streptomyces avermitilis (sma), Bifidobacterium longum NCC2705 (blo), Corynebacterium diphtheriae NCTC 13129 (cdi), Mycobacterium bovis AF2122/97 (mbo); Fusobacteria—Fusobacterium nucleatum ATCC 25586 (fnu); Spirochete—Treponema pallidum Nichols (tpa), Leptospira interrogans 56601 (serovar lai) (lil); Bacteroid—Bacteroides thetaiotaomicron VPI-5482 (bth), Porphyromonas gingivalis W83 (pgi); Planctomyces—Rhodopirellula baltica (Pirellula sp.) (rba); Cyanobacteria—Synechocystis sp. PCC6803 (syn), Thermosynechococcus elongatus BP-1 (tel), Anabaena sp. PCC7120 (Nostoc sp. PCC7120) (ana), Gloeobacter violaceus PCC7421 (gvi), Prochlorococcus marinus SS120 (CCMP1375) (pma), Prochlorococcus marinus MED4 (CCMP1378) (pmm), Prochlorococcus marinus MIT9313 (pmt), Synechococcus sp. WH8102 (syw); green sulfur bacteria—Chlorobium tepidum TLS (cte); radioresistant bacteria—Deinococcus radiodurans R1 (dra); Eukarya—Homo sapiens (hsa), Mus musculus (mmu), Drosophila melanogaster (dme), Caenorhabditis elegans (cel), Arabidopsis thaliana (ath), Dictyostelium discoideum (ddi), Saccharomyces cerevisiae (sce), Schizosaccharomyces pombe (spo); Euryarchaeota—Methanosarcina acetivorans C2A (mac), Methanosarcina mazei Goe1 (mma), Methanobacterium thermoautotrophicum deltaH (mth), Halobacterium sp. NRC-1 (hal); Crenarchaeota—Sulfolobus solfataricus P2 (sso).

Full size image

MsrA and MsrB homologs are absent in all crenarchaeotal species included in this study (Pyrobaculum aerophilum, Aeropyrum pernix, and Sulfolobus tokodaii), with the exception of S. solfataricus, which has an ORF sequence homologous to the E. coli MsrA (Table 1). The parasitic microsporidium Encephalitozoon cuniculi lacks both MsrA and MsrB, probably due to a secondary loss, but all the other eukaryotic genomes included in Table 1 exhibit at least one copy each of MsrA and MsrB. All eukaryotic MsrB sequences are endowed with the Zn-binding motif, with the sole exception of a highly divergent copy of MsrB present in Arabidopsis thaliana (Fig. 2b).

MsrA and MsrB have an ample biological distribution, but their phylogenies (Figs. 2a and b) resemble only in part the 16/18S rRNA-based cladograms with three well-defined monophyletic major domains (Woese et al. 1990). In some cases high bootstrap values are observed, but with few exceptions, discussed below, the clades have little if any biological significance. Some recently diverged bacterial groups, such as the γ-proteobacterial enterobacterial groups (the different E. coli strains, Shigella, Salmonella), exhibit similar topology in both 16/18S rRNA trees and MsrA (Fig. 2a) and MsrB (Fig. 2b) cladograms. MsrA and MsrB trees are characterized by very low bootstrap values, and a mixed distribution of actinobacteria, cyanobacteria, firmicutes, most proteobacteria, and archaeal and eukaryal groups is observed. However, quantitative estimates of the similarity between the MsrA and the Msr B cladograms based on orthologue sequences show that despite the low preservation of phylogenetic signals, these two trees are more similar to each other, and to a 16/18S rRNA tree, than to randomly generated cladograms constructed as described above (Table 2).

Table 2 Number of symmetric differences among the following trees: (1) 16/18S rRNA, (2) MsrA, (3) MsrB, (4) in silico concatenated MsrAB, (5) randomly concatenated MsrAB sequences, and (6) randomly generated 16/18SrRNA tree
Full size table

Based on the similarities between the MsrA and the MsrB orthologue trees (Table 2), an attempt to recover phylogenetic signals that may be susceptible to evolutionary analysis was perfomed by an in silico artificial joining of MsrA and MsrB sequences, as described under Materials and Methods. No naturally concatenated archaeal or eucaryal MsrA and MsrB sequences are present in our database. Minimum evolution trees for naturally (Fig. 1a) and artificially (Fig. 1b) concatenated MsrA and MsrB sequences have been calculated. In the minimum evolution tree constructed with naturally concatenated sequences shown in Fig. 1a, several groups form clades whose branching order is comparable to those observed in rRNA phylogenies. Although the firmicutes are clearly separated in two well-defined monophyletic clades formed by the bacillales and lactobacillales groups, their internal branching orders are quite similar to that observed in trees based on 16/18S rRNA and other canonical markers. The same is true for the the γ-vibrionales clade, which has significant bootstrap values, although other proteobacterial branches are dispersed throughout the tree. The presence of the sequences of Nitrosomonas europaea in a separate branch can be explained as a long-branch attraction phenomenon, but because of the low bootstrap values, little can be said about the distribution of other species in this sample.

A significant reduction of the differences between the concatenated MsrA-MsrB tree and the canonical 16/18S rRNA tree is observed when the two are compared with the corresponding random trees (Table 2). The tree shown in Fig. 1b was constructed by concatenating in silico the sequences from those species whose genomes are endowed with only a single copy of the gene encoding MsrA and a single copy of the gene encoding MsrB. Concatenation of sequences has been recognized as a way of enhancing phylogenetic signals (Brown et al. 2001; Rokas et al. 2003). Although leaving out of the sample under consideration all the genomes endowed with paralogous sequences reduces the number of species from 159 to 72, a significant gain in phylogenetic signals is achieved, as indicated by the higher bootstrap values. The rationale behind the in silico concatenation is based on the possibility that naturally concatenated sequences may have been inherited jointly (Fig. 1a). In order to evaluate the significance of the in silico concatenation used here, a phylogeny based on randomly concatenated sequences was constructed. The resulting tree (not shown) exhibits very low bootstrap values.

The minimum evolution tree based on in silico concatenated MsrA and MsrB sequences shown in Fig. 1b fails to resolve clearly the three domains of life (Bacteria, Archaea, and Eukarya) and exhibits many unresolved basal branches. Nonetheless, it represents a significant improvement over the phylogenies shown in Fig. 2 and has good bootstrap values for branches depicting recent speciation events, which are in good agreement with trees based on canonical phylogenetic markers including 16/18S rRNA cladograms. Although in this tree the proteobacteria do not group into a single clade, their different divisions are well defined and include branches with high bootstrap values (>70%) that clearly differentiate (a) separate branches of α-, β-, and ε-proteobacteria, as well as three clades for γ-proteobacteria (enterobacteriales, xanthomonadales, pseudomonadales); (b) the four firmicutes subgroups (mollicutes, clostridia, bacillales, and lactobacillales), although the Deinococcus radiodurans branch separates early from the bacillales; and (c) the actinobacteria. All naturally concatenated sequences group within the Firmicutes clade, and their incorporation as a branch of the in silico concatenated sequence cladogram does not modify its overall topology (Supplementary Fig. 3). Quite interestingly, the euryarcheaota species included here cluster with Geobacter sulfurreducens (gsu), a δ-proteobacterium that shares ecological niches with some of the euryarchaeotal species, suggesting a possible case of lateral gene transfer in related environments. With the exception of the branch corresponding to the C. elegans MsrAB sequences, all other eukarya included in this sample are grouped together and branch with ε-proteobacteria. Despite the large number of unresolved branches, the Zn-bearing motif does not exhibit a random biological distribution.

Discussion and Conclusions

Analysis of the three-dimensional structure of MsrA and MsrB has demonstrated the independent origin of these two enzymes (Lowther et al. 2002); their active sites exhibit approximate mirror symmetry. However, comparisons of their crystal structures have shown that MsrB is formed by several coiled β-sheets, while MsrA is an α/β protein that has a β-sheet core surrounded by α-helices (Gladyshev 2002). Comparison of MsrA and B structures with available databases shows that these two enzymes do not have structural similarity to other known proteins (Gladyshev 2002; Lowther et al. 2002). Although both MsrA and thioredoxin are thiol oxidoreductases with similar reaction mechanisms and comparable architectures (Gladyshev 2002), no sequence similarity or common structural folds are shared between them. MsrA has been included in the SCOP database classification within the ferredoxin-like fold group together with 48 other superfamilies, but this does not allow inferences about their evolutionary relationships, and depends on the likelihood of homology among superfamilies defined solely by one fold.

It is not always easy to rationalize the presence or absence of the methionine sulfoxide reductases in different species. Their presence in anaerobic species (including Methanobacterium thermoautotrophicum, Clostridium acetobutylicum, C. perfringens, and Chlorobium tepidum) may reflect their role in repairing proteins during transient exposures to oxygen. The lack of genes encoding MsrA and MsrB in endosymbionts and endoparasites can perhaps be understood as the outcome of a secondary adaptation that makes them unnecessary under the anaerobic conditions of the intracellular cytoplasm of their hosts. This is consistent with the many other losses that the genomes of these organisms have undergone.

There are two possible explanations of the lack of detectable MsrA and MsrB genes in the bacteria Aquifex aeolicus and Thermotoga maritima, and in the archaea Thermoplasma acidophilum, T. volcanii, Aeropyrum pernix, Pyrobaculum aerophilum, and Sulfolobus tokodaii. In these species the corresponding genes may have been lost secondarily, or their absence may indeed be an ancestral character. Given the low levels of free oxygen in which these species live (Jannasch and Mottl, 1985; Brock 1986; David E. Graham, personal communication), which may be comparable to those of the primitive environment, both explanations are feasible. The absence of MsrA and MsrB in these species could perhaps be explained by the early development of other protective mechanisms, including antioxidant ferritin-like di-iron-carboxylate proteins (Wiedenheft et al. 2005). The presence of the MsrA sequence in S. solfataricus is probably due to a horizontal gene transfer event.

Although MsrA and MsrB have an ample biological distribution, they are small proteins that exhibit low sequence conservation. The inconsistencies of MsrA and MsrB trees (Figs. 2a and b) with the 16/18S rRNA topology demonstrate the limited value of these enzymes as phylogenetic markers, due to a combination of their small sizes, their unequal rates of substitution, a number of paralogous duplications (as in the case of cyanobacteria), and lateral transfer events. However, several clades in the trees depicting naturally and artificially concatenated sequences (Figs. 1a and b) suggest that the phylogenies of MsrA and MsrB still record recent speciation events. This conclusion is supported by the nonrandom distribution of the Zn-bearing MsrB sequences in all trees depicted here. The similarity between the 16/18S rRNA tree and the concatenated MsrA-MsrB phylogeny suggests that, with the few exceptions described below, these two genes have undergone joint vertical inheritance, although this does not imply that both sequences first evolved in the same species.

It is likely that the dispersal of eukaryotic MsrA and MsrB sequences within proteobacterial branches reflects the ultimate bacterial origin of the sequences observed in nucleated cells. This is consistent with the idea that nucleated cells may have acquired the enzyme with the endosymbiotic bacteria that led to mitochondria. Since the mitochondria are closely related to the γ-proteobacteria, the eukaryal MsrA and MsrB genes should be more closely related to this group. However, the observed low bootstrap values make it difficult to support this conclusion.

The comparison of the complete genomes of Haemophilus influenzae and Mycoplasma genitalium led to an inventory of 256 genes that included MsrA, which was hypothesized to correspond to the genetic complement of the ancestor of the Gram-positive and Gram-negative bacteria and, perhaps, of even older entities (Mushegian and Koonin 1996). However, other searches have failed to confirm the presence of the genes encoding MsrA in sets of highly conserved orthologues (Daubin et al. 2002; Harris et al. 2003; Yang et al. 2005; Delaye et al. 2005).

It has been suggested that cellular stress response mechanisms, including methionine sulfoxide reductase activities, may have already been present in the last common ancestor of all extant living beings (Kültz 2005). Methionine sulfoxide reductase activity is probably very ancient (Hansel et al. 2005), but it is unlikely that MsrA and MsrB were broadly distributed before the evolution of oxygenic photosynthesis. Methionine oxidation is mediated by many highly reactive chemical species, including hydrogen peroxide, hydroxyl radicals, ozone, peroxynitrite, hypochlorite, and metal catalyzed oxidation systems (Moskovitz et al. 1997). It could argued that the fact that these two enzymes lack obvious homologues is evidence that they were present prior to the ancient separation among the three major domains, and that their uneven distribution is the outcome of a complex history of secondary losses. However, the ubiquity of methionine sulfoxide reductase activity is probably best understood as the evolutionary response to the high levels of atmospheric oxygen that accumulated during prokaryotic evolution. Although the broad dissemination of the genes encoding the MsrA and MsrB proteins was probably the result of the increasing levels of free oxygen in the atmosphere, their origin probably took place locally at an earlier time, for example, in cyanobacteria or their neighbors. Unfortunately, recognition of the prokaryotic group in which these proteins originated is clouded not only by the poor preservation of phylogenetic signal of these two sequences, but also by the complex history of duplications and lateral gene transfer events that MsrA and MsrB have undergone.