INTRODUCTION

Top Abstract Introduction Materials and methods Results Discussion References

DNA repair and recombination are fundamental molecular processes that were most likely present in the earliest life. Supporting this view is the observation that Bacteria, Archaea, and Eucarya all have phylogenetically related DNA repair and recombination genes that encode a crucial protein involved in synapsing two parental DNA molecules. The first and archetypal member of this protein family was identified by mutations in the recA gene of the bacterium Escherichia coli (5); therefore, the protein it encodes is called RecA. Homologues of the E. coli recA gene have now been found in all bacterial divisions in which they have been sought (7, 12, 20). Two budding-yeast genes, RAD51 and DMC1, have been recognized to be homologues of bacterial recA genes (22, 26). Since then, homologues of these two genes have been found in all Eucarya species tested (24). Finally, Sandler et al. (21) identified genes from three archaeal genera whose putative proteins are similar to RecA proteins but are even more similar to the eucaryal Rad51 and Dmc1 proteins. These genes and proteins are called radA and RadA, respectively.

There were two objectives to this research: a protein structure-function objective and a phylogenetic objective. The protein structure-function objective was to determine how consistent the differences are between the RecA group of proteins and the RadA-Rad51-Dmc1 group observed by Sandler et al. (21) and Brendel et al. (3). These authors noted that the archaeal RadA proteins provided additional information that allowed a better alignment of the core regions of eucaryl Rad51 and Dmc1 proteins with the bacterial RecA proteins. Existence of a homologous core region in all of the RecA, Rad51, and Dmc1 proteins was already known (22). Also known was the fact that, relative to the core, the Rad51 and Dmc1 proteins have longer amino-terminal ends and the bacterial RecA proteins have longer carboxy-terminal ends (22). Unclear, however, was the alignment of amino acids within the core. Sandler et al. (21) and Brendel et al. (3) contended that the cores of RadA-Rad51-Dmc1 proteins have four highly conserved insertions and one deletion relative to the cores of the RecA proteins. Furthermore Sandler et al. (21) stated that there might be functional significance in the locations of the insertions. They located these insertions in the context of the X-ray crystal structure of E. coli RecA protein complexed with ADP and found them to be on the outside of the protein, away from the putative DNA binding site. This is consistent with the insertions not compromising the synaptase function of RadA while adding features that potentially might interact with accessory proteins.

Phylogenetic relationships among RecA proteins from greater than 65 different species of bacteria have been the focus of several studies (7, 12, 13, 17). The trees are robust and correlate well with the trees formed by 16S rRNAs from the corresponding bacteria, leading to the conclusion that RecA is a useful genetic marker for reconstructing bacterial phylogeny. Others have found that the eucaryal Dmc1 and Rad51 proteins are not good phylogenetic markers because they appear to have unequal rates of evolution (24). Thus, we wanted to determine if the RadA proteins would be useful in deciphering the phylogeny of the Archaea. In addition, we wanted to test the hypothesis of Sandler et al. (21) that there might be taxonomic significance in the number of amino acids present in one or more of the insertions.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and methods Results Discussion References

Archaeal genomic DNA. DNA from cultured archaea was obtained for the strains listed in Table 1. Environmental DNAs were obtained from three different hot springs in Yellowstone National ParkObsidian Pool, hot pool N10 in the Norris Geyser Basin, and hot pool O1 in the White Creek areaas previously described (11).

PCR, cloning, and sequencing. The primers used to amplify portions of genomic DNA in this work are listed in Table 2. For most PCR amplifications, we followed the procedure of Sandler et al. (21), which specifies an initial three cycles in which hybridization is carried out at temperatures increasing at 1°C per 10 s from 37 to 72°C. These cycles are followed by 26 cycles in which hybridization is carried out at 43°C. Six DNA preparations, however, did not yield amplification products by this procedure. For them (see Table 1) we used an alternative called "touchdown PCR" (9). In this procedure, hybridization is performed during the first cycle at 50°C for 1.5 min. After that cycle, the hybridization temperature is decreased by 0.5°C to 40°C in each of 19 successive cycles. An additional 20 cycles in which the hybridization temperature was 43°C were performed. In every cycle there was a denaturation step of 1 min at 94°C preceding and an extension step of 1 min at 71°C following hybridization. Lastly, the reactions were incubated at 71°C for 10 min before storage at 4°C. An MJ Research thermocycler, PTC-150, was used for PCR.

Sequence alignment and phylogenetic inference. The RadA alignment used in this study is based on our previous alignment (21). One hundred thirty-two unambiguously alignable amino acid positions, excluding the N and C termini, or 264 first- and second-codon positions of the corresponding nucleotide alignment were used in all RadA analyses. A total of 1,206 unambiguously alignable nucleotide positions were used in 16S rRNA analyses.

Nucleotide sequence accession numbers. GenBank accession numbers for the nucleotide sequences determined in this study are listed in Table 1.

	RESULTS

Top Abstract Introduction Materials and methods Results Discussion References

Design of primers. Sandler et al. (21) used one set of primers to isolate a fragment of the radA gene of Sulfolobus solfataricus and another set to isolate fragments of radA genes of two other genera. Only 18 amino acids of the E. coli RecA protein lay between the equivalent positions of the primers in the E. coli recA gene. Our goal in this study required that we maximize the number of amino acids between the conserved regions used to design primers so that meaningful phylogenetic data could be extracted from the fragments isolated. Consequently, we chose two regions separated by 120 E. coli RecA amino acids for our primer design. The upstream region consists of portions of beta strand b1 and the Walker phosphate binding hole motif. The downstream region consists of portions of beta strand b5 and loop L2. The region in between comprises an integral domain of the E. coli RecA protein (27).

Amplification of radA gene fragments and alignment of their putative protein products. Three sets of primers were designed to be used in conjunction with DNAs containing high, low, and medium levels of G+C, as shown in Table 2. We used these primers to clone fragments of radA genes from 13 cultured species of Archaea. Alignments of the amino acids that are encoded by the fragments are shown in Fig. 1A and C. Evident from these panels is the occurrence of inserted or deleted amino acids in the RadA, Rad51, and Dmc1 sequences in comparison to the RecA sequences. Inserted amino acids occur between residues equivalent to residues 84 and 85, 102 and 103, 110 and 111, and 137 and 138 of E. coli RecA; these are designated "Indel sites" 1 to 4, respectively. Deleted are amino acids equivalent to residues 152 to 155 of E. coli RecA; this is called Indel site 5. The numbers of amino acids inserted at Indel site 2 and deleted at Indel site 5 are the same in all of the RadA, Rad51, and Dmc1 sequences shown in Fig. 1A and C. The numbers of amino acids at the other Indel sites differ, which indicates that multiple events may have occurred. Four sequences show 15- or 16-amino-acid insertions between E. coli RecA residues 84 and 85 rather than the 6 amino acids characteristic of the others (Indel site 1); we call these four sequences EL1 insertions (for extra length at site 1). These four are all from members of the genus Methanococcus. Four RadA sequences have between 11 and 33 amino acids between E. coli RecA residues 110 and 111 (Indel 3) rather than the 4 amino acids characteristic of the others; we call these sequences EL3 insertions. Three of these are from extreme halophiles. The same four sequences also have eight amino acid residues at Indel site 4 (EL4), thus differing from the three to five amino acids present in the others, except the sequence for Archaeoglobus fulgidus. Further information on the way these patterns of insertions and deletions are distributed phylogenetically is presented below.

View larger version (83K): [in this window] [in a new window]   View larger version (96K): [in this window] [in a new window]   FIG. 1.   Alignment of amino acids encoded by PCR-amplified fragments of radA genes. (A) Residues 1 to 90 of radA fragments from 15 cultivated and 1 enriched archaeal species. Thirteen fragments were isolated by the PCR technique and cloned. Those from H. volcanii (Hf. vol), M. jannaschii (Mc. jan), and S. solfataricus (Sul. sol) were taken from cognate parts of the sequences published by Sandler et al. (21). Cognate parts of Rad51, Dmc1, and RecA have previously been published (2, 10). (B) Residues 1 to 90 of radA fragments from environmental samples. These residues are shown in their proper alignment to the residues of the RecA, Rad51, and Dmc1 fragments in panel A. (C) Residues 91 to 180 of radA fragments from 15 cultivated and 1 enriched archaeal species. See the legend to panel A for further information. (D) Residues 91 to 180 of radA fragments from environmental samples shown in appropriate alignment to residues of the RecA, Rad51, and Dmc1 fragments in panel C. Abbreviations are defined in Table 1.

Sequences of complete RadA genes. Using cloned radA fragments (21) as hybridization probes, we cloned the entire radA genes of Halobacterium halobium and C. symbiosum. The protein sequences are presented in Fig. 2. Portions of the sequences, called "domains" by Brendel et al. (3), are indicated. A distinctive feature of the C. symbiosum sequence is that it does not have a complete domain A sequence. This distinguishes it from the H. halobium sequence and all other RadA, Rad51, and Dmc1 sequences so far examined (Fig. 2) (3). Another distinctive feature is the existence of a carboxy-terminal sequence similar in length, but not in sequence, to that of E. coli RecA protein and not possessed by any other RadA, Rad51, or Dmc1 sequence so far studied (3, 24).

View larger version (54K): [in this window] [in a new window]   FIG. 2.   Sequences of two complete putative RadA proteins (from H. halobium [Hb. hal] and C. symbiosum [Ca. sym]) aligned with E. coli RecA, S. cerevisiae Rad51, and S. cerevisiae Dmc1 proteins.

Phylogeny of the Sequences: RadA versus 16S rDNA. The sequences of the RadA fragments shown in Fig. 1A and C were analyzed phylogenetically, and the phylogeny obtained was compared to that obtained with cognate 16S ribosomal DNA (rDNA) sequences (Fig. 3). In general there is a high degree of consistency between the two phylogenies. The shaded regions in this figure show that the two molecules reveal similar clades of halophiles, Methanococcus spp. and Sulfolobus spp., with minor branching-order discrepancies. In performing this analysis, we excluded the extra amino acids at Indel sites 1 to 5 (Fig. 1). Nonetheless, certain features of the inserts are consistent with the clade structure based on the RadA fragments. In Fig. 3, the numbers in circles refer to corresponding extra-length (EL) numbers and show the common phylogenetic ancestry of these extra-long insertions. The most notable inconsistency between the RadA and 16S rDNA phylogenies is the position of C. symbiosum in the Crenarchaeota by 16S phylogeny but in an independent lineage by RadA phylogeny. The implications of this are discussed below.

View larger version (30K): [in this window] [in a new window]   FIG. 3.   Comparison of 16S rRNA and radA trees generated by rate-corrected ML analyses of nucleotide sequences (first and second codon positions only) for radA encoding the amino acid sequences shown in Fig. 1A and C. One additional unpublished sequence, that of P. aerophilium, was generously provided by S. Fitz-Gibbon (see reference 28). Branch points supported (bootstrap values, 75%) by most or all phylogenetic analyses (see Materials and Methods) are indicated by filled circles. Open circles indicate branch points supported by some analyses but marginally supported (bootstrap, 50 to 74%) or unsupported (bootstrap, <50%) by others (see Results and Discussion). Strongly supported monophyletic groups consistent between the two molecules are indicated by shading. Three bacterial outgroup sequences (from E. coli, Deinococcus radiodurans and Thermotoga maritima) were used for all 16S rRNA analyses, and four eucaryal outgroup sequences (Homo sapiens RAD51 and DMC1 and S. cerevisiae RAD51 and DMC1) were used for all radA analyses. Three secondary structural features of radA (see Fig. 1) are indicated at branch points; that is, all sequences to the right of the indicated branch point share the secondary structural feature. 1, 3, and 4 represent extra-long inserts EL1, EL3, and EL4 at Indel sites 1, 3, and 4 respectively (see Fig. 1 and text). Abbreviations are defined in Table 1.

Phylogeny of environmental samples. We determined phylogenetic relationships of the radA nucleotide sequences extracted from environmental samples (Fig. 1B and D) with the radA nucleotide sequences from known archaeal species (Fig. 4). The results show that 14 of the 16 sequences cluster unambiguously with Crenarchaeota sequences. Two of the 14 (NGB#6 and NGB#13) are probably from representatives of the genus Sulfolobus because they occur within the radiation of reference Sulfolobus species in Fig. 4. Nine others are monophyletic with the Sulfolobales representatives, but at present there are insufficient reference RadA sequences to identify their generic affiliations. However, based on a proposed generic lower limit of 78% identity (data not shown), clones OP#1, NGB#14, and OP#9 may represent Sulfolobus species, as indicated by the dashed line in Fig. 4. Three others (OP#3, OP#4, and OP#6) are monophyletic with Pyrobacculum aerophilium, but again there are insufficient reference sequences at this time to conclude their generic affiliations.

View larger version (31K): [in this window] [in a new window]   FIG. 4.   Phylogenetic relationships of the 17 radA sequences from cultivated archaeal species and 16 environmental radA sequences determined by using rate-corrected ML analyses of first- and second-position nucleotide sequences. The radA nucleotide sequences used were those encoding the amino acid sequences in Fig. 1 with the exception of the P. aerophilium sequence (see the legend to Fig. 3). Eucaryal outgroup sequences were as described for Fig. 3. Filled and open circles indicating phylogenetic integrity of branch points are as in Fig. 3. Branch points without circles were not resolved (bootstrap, <50%) as specific groups in different analyses. Abbreviations are defined in Table 1.

Signature amino acid codons. Signature amino acids derived from aligned RecA family sequences are useful for deducing the subfamily to which each protein belongs without resort to formal phylogenetic analysis. Comparative analysis of the sequences in Fig. 1 reveals that nine amino acids can be used to distinguish five groups of RecA-like proteins. The first set of three amino acids (Table 3, group A) separates the family into three groups consisting of RecA in one group, RadA, Rad51, and Dmc1 in another, and RadB in a third. The next set of two amino acids (Table 3, group B) separates RadA, Rad51, and Dmc1 from each other and increases the definition of RecA and RadB from each other and the other three family members. Finally, there are four additional amino acids that enhance the definitions still further.

	DISCUSSION

Top Abstract Introduction Materials and methods Results Discussion References

Sandler et al. (21) proposed that three putative archaeal RadA proteins, although about 20% identical to bacterial RecA proteins, possess certain primary structural features that set them apart. We have found evidence to support this proposal by sequencing 2 complete and 11 partial radA genes from different archaeal species. In particular, we found that all archaeal RadA species have four regions of inserted and one region of deleted amino acids relative to bacterial RecA sequences (Fig. 1A and C). In addition, we found that the lengths of three of the inserts have taxonomic significance and distinguish particular phylogenetic clades determined by either RadA or 16S rRNA sequences (Fig. 3). Furthermore, all 13 new sequences share these diagnostic features with Rad51 and Dmc1 sequences from representative eukaryotes. This strengthens the proposal of Sandler et al. (21) that RadA is orthologous to the common ancestor of Rad51 and Dmc1.

Close correspondence of the phylogenies obtained for archaeal species with either the RadA or 16S rRNA sequences (Fig. 3) is also noteworthy because it suggests that RadA sequences can be used as an independent measure of archaeal diversity. Indeed, RadA clearly resolves the phylogenetic position of A. fulgidus as monophyletic with Halobacteriales and Methanomicrobiales. Furthermore, a secondary structural feature (the length of the insertion of Indel site 4) corroborates this association. This analysis independently supports the conclusion of Woese et al. (29), who based their finding on compensation for G+C bias in the 16S rDNA sequences. However, the failure of the C. symbiosum fragment to be monophyletic with the Crenarchaeota may indicate that distant phylogenetic relationships cannot be resolved by using this small RadA fragment. An alternative explanation is that we have sequenced a RadA paralog from C. symbiosum. We can test this alternative by screening more completely the C. symbiosum genome for additional RadA-like sequences.

As another test of the phylogenetic utility of the RadA molecule, we examined DNAs obtained from four different natural sources: three ecologically distinct hot pools in Yellowstone National Park (11) and the Antarctic Ocean. Fourteen of the 15 hot pool sequences cluster with members of the Crenarcheota. Six of these were from Obsidian Pool, whose analysis by 16S rDNA phylogeny had also shown a dominance of crenarchaeotan sequences (1). The other eight were from a pool in the Norris Geyser Basin for which there is no equivalent 16S rDNA analysis. However, we would predict that such an analysis would reveal a majority of crenarchaeotan sequences.

Two of the sequences, for C. symbiosum and Ant#17, stand out by their unique characteristics. Figure 4 demonstrates the novelty of these sequences by their independent branching in the archaeal tree. Although they may represent proteins paralogous to RadA, we are encouraged to anticipate the discovery of more diversity in RecA family sequences, with others possibly showing mixtures of the features of RecA and RadA proteins and perhaps representing evolutionary intermediates.

The RecA family actually contains two different member classes in the Archaea: radA and radB. In this study, we have been concerned mainly with radA. radB, however, was originally found in the genome sequence of Methanococcus janaschii (4), where it was identified as a Rad51 relative. Other members of the radB subfamily have been found in A. fulgidus (14), Pyrococcus sp. KOD1 (19), Pyrococcus furiosus (6), and Methanobacterium thermoautotrophicum (23). Putative RadB proteins differ from RadA in two major features. First, they are only about 70% as big, having neither an N- nor a C-terminal extension. Second, their sequences differ appreciably, and we have shown in Table 3 the signature sequences that distinguish RadA from RadB. The sequences are different enough that special primers would have to be designed to amplify them by the PCR technique. The primers that we used to amplify radA sequences (Table 1) would have had a very low probability of amplifying radB sequences. It seems worthwhile in the future to search for radB sequences, given the similarity of their proteins in sequence and possibly in function to the Rad55 and Rad57 proteins of Saccharomyces cerevisiae (17a).