Open Access

Research Article

• Download Article XML
• Download Article PDF
• Download Citation
• E-mail this Article
• Order Reprints
• Print this Article

Introduction

Deinococcus geothermalis belongs to the Deinococcus-Thermus group, which is deeply branched in bacterial phylogenetic trees and has putative relationships with cyanobacteria [1], [2]. The extremely radiation-resistant family Deinococcaceae is comprised of greater than twenty distinct species [3] that can survive acute exposures to ionizing radiation (IR) (10 kGy), ultraviolet light (UV) (1 kJ/m²), and desiccation (years) [4], [5]; and can grow under chronic IR (60 Gy/hour) [6]. D. geothermalis was originally isolated from a hot pool at the Termi di Agnano, Naples, Italy [7], and subsequently identified at other locations poor in organic nutrients including industrial paper machine water [8], deep ocean subsurface environments [9], and subterranean hot springs in Iceland [10].

D. geothermalis is distinct from most members of the genus Deinococcus in that it is a moderate thermophile, with an optimal growth temperature (T_opt) of 50°C [7], is not dependent on an exogenous source of amino acids or nicotinamide for growth [11], [12], is capable of forming biofilms [8], and possesses membranes with very low levels of unsaturated fatty acids compared to the other species [7]. Based on the ability of wild-type and engineered D. geothermalis and D. radiodurans to reduce a variety of metals including U(VI), Cr(VI), Hg(II), Tc(VII), Fe(III) and Mn(III,IV) [11], [13], these two species have been proposed for bioremediation of radioactive waste sites maintained by the US Department of Energy (DOE) [11], [14], [15]. These characteristics were the impetus for whole-genome sequencing of D. geothermalis at DOE's Joint Genome Institute, and comparison with the mesophilic D. radiodurans (T_opt, 32°C), to date the only other extremely IR-resistant bacterium for which a whole-genome sequence has been acquired [16].

Chromosomal and plasmid DNAs in extremely resistant bacteria are as susceptible to IR-induced DNA double strand breaks (DSBs) as in sensitive bacteria [5], [17]–[19] and broad-based experimental and bioinformatic studies have converged on the conclusion that D. radiodurans uses a conventional set of DNA repair and protection functions, but with a far greater efficiency than IR-sensitive bacteria [17], [20], [21]. This apparent contradiction is exemplified by work which showed that the repair protein DNA polymerase I (PolA) of D. radiodurans supports exceptionally efficient DNA replication at the earliest stages of recovery from IR, and could account for the high fidelity of RecA-mediated DNA fragment assembly [22]. Paradoxically, however, IR-, UV-, and mitomycin-C (MMC)-sensitive D. radiodurans polA mutants are fully complemented by expression of the polA gene from the IR-sensitive Escherichia coli [4].

The reason why repair proteins, either native or cloned, in D. radiodurans function so much better after irradiation than in sensitive bacteria is unknown. The prevailing hypotheses of extreme IR resistance in D. radiodurans fall into three categories: (i) chromosome alignment, morphology and/or repeated sequences facilitate genome reassembly [5], [21], [23], [24]; (ii) a subset of uncharacterized genes encode functions that enhance the efficiency of DNA repair [20]; and (iii) non-enzymic Mn(II) complexes present in resistant bacteria protect proteins, but not DNA, from oxidation during irradiation, with the result that conventional enzyme systems involved in recovery survive and function with far greater efficiency than in sensitive bacteria [17], [23]. The extraordinary survival of Deinococcus bacteria following irradiation has also given rise to some rather whimsical descriptions of their derivation, including that they evolved on Mars [25]. On the basis of whole-genome comparisons between two Deinococcus genomes and two Thermus genomes, we present a reconstruction of evolutionary events that are inferred to have occurred both before and after the divergence of the D. radiodurans and D. geothermalis lineages. We revise down substantially the number of potential genetic determinants of radiation resistance, predict a Deinococcus radiation response regulon, and consider the implications of these comparative-genomic findings for different models of recovery.

Results and Discussion

Resistance to IR and UV

One approach to delineating a minimal set of genes involved in extreme resistance is to compare the whole-genome sequences of two phylogenetically related but distinct species that are equally resistant, whereby genes that are unique to both organisms are ruled out, whereas shared genes are pooled as candidates for involvement in resistance. We show that D. geothermalis (DSM 11300) and D. radiodurans (ATCC BAA-816) are equally resistant to IR (⁶⁰Co) (Figure 1A) and UV (254 nm) (Figure 1B) when pre-grown and recovered at their optimal growth temperatures, 50°C and 32°C, respectively. When recovered at 50°C, the survival of D. geothermalis exposed to 12 kGy was 1,000 times greater than at 32°C (Figure 1A) [7]. The extreme resistance to desiccation of D. geothermalis recovered at 50°C was demonstrated previously [5]. Thus, D. geothermalis and D. radiodurans are well-suited to defining a conserved set of genes responsible for extreme resistance.

Figure 1. Radiation resistance and genome structure of D. geothermalis and D. radiodurans. A, IR (⁶⁰Co, 5.5 kGy/h).

B, UV (254 nm) (3 J/m² s⁻¹). Open circle, D. radiodurans (32°C); open triangle, D. geothermalis (50°C); and open square, D. geothermalis (32°C). Values are from three independent trials with standard deviations shown. At near-optimal growth temperatures, the 10% survival values (D₁₀) following IR for D. radiodurans (32°C) and D. geothermalis (50°C) are 15 kGy; for E. coli, 0.7 kGy (37°C) [5]; and for T. thermophilus (HB27) 0.8 kGy (65°C) [27]. C, PFGE of genomic DNA prepared from irradiated (0.2 kGy) D. radiodurans (DR+IR) and D. geothermalis (DG+IR); and genomic DNA from non-irradiated D. geothermalis digested with SpeI (DG+SpeI). (M) PFGE DNA size markers. PFGE was as described previously [77].

doi:10.1371/journal.pone.0000955.g001

Genome Sequence and Structure: General Features

The random shotgun method [16] was used to acquire the complete sequence of the D. geothermalis (DSM 11300) genome, that is comprised of a main chromosome (2,467,205 base pairs (bp)), and two megaplasmids (574,127 bp and 205,686 bp). The general structure of the predicted D. geothermalis genome was tested by pulsed field gel electrophoresis (PFGE) of genomic DNA linearized in vivo by exposure to IR (0.2 kGy), and by restriction endonuclease (SpeI) cleavage (Figure 1C). The IR-treatment revealed the existence of a ~570 kb megaplasmid in D. geothermalis, and the SpeI-treatment yielded the expected number of chromosomal bands: 3 singlets (632 kb, 376 kb and 282 kb) and one doublet (574/579 kb); the plasmids do not contain a SpeI site. In comparison, IR-treated D. radiodurans (ATCC BAA-816) subjected to PFGE displayed the presence of the DR412 (412 kb) and DR177 (177 kb) megaplasmids, previously observed [26]. The approximately 206 kb D. geothermalis megaplasmid was not visualized by PFGE although its size lies between the two D. radiodurans megaplasmids, which were readily observed (Figure 1C). Consistently, the abundance of DNA clones for the 206 kb megaplasmid was significantly lower than the 574 kb megaplasmid during construction of the D. geothermalis genome-library used for sequencing (data not shown). Thus, the 574 kb megaplasmid of D. geothermalis exists at higher copy-number than the 206 kb megaplasmid.

Genome Comparison: General Features

Comparison of the general genome features of D. geothermalis and D. radiodurans revealed major differences in genome partitioning, and in the number of noncoding repeats (SNRs) (Table 1).

Table 1. General Characteristics

doi:10.1371/journal.pone.0000955.t001

Genome Partitioning.

We previously demonstrated homologous relationships between the DR412 megaplasmid of D. radiodurans and the sole 233 kb megaplasmid (pTT27) of T. thermophilus [27]. Based on the gene contents of DR412 and pTT27, we concluded that these megaplasmids evolved from a common ancestor (Figure S1), are essential to the survival of both species, and appear to serve as a sink for horizontally transferred genes [27]. In contrast, the 574 kb megaplasmid (DG574) of D. geothermalis is distinct from pTT27, and appears to have been derived from a fusion of DR412 and DR177 (Table S1), followed by numerous rearrangements. Levels of gene order conservation for the D. geothermalis and D. radiodurans chromosomes and megaplasmids were determined by genomic dot plots [28] (Figure S2). The dot plots of the chromosomes showed a clear pattern characteristic of chromosomes of relatively closely related bacteria that retain significant colinearity of the gene order. The X-shape pattern is thought to arise from inversions of a chromosomal segment around the origin of replication [28]. By contrast, DR412 and DR177 did not display any discernable colinearity (Figure S2B), indicating substantially greater levels of rearrangement in the megaplasmids.

Repeats and Prophages.

Dozens of small noncoding repeats (SNRs) of an unusual, mosaic structure have been identified in the D. radiodurans genome, suggesting a possible role in resistance [29]. In stark contrast, no mosaic-type SNRs were found in the D. geothermalis genome (Table 1), suggesting that SNRs are not involved in recovery from radiation or desiccation [26], [29], [30]. Further, there are about 20 DNA repeats in D. radiodurans that contain oligoG stretches (Figure S3). Such DNA sequences might adopt an ordered helical structure (G-quadruplex), predicted to form parallel four-stranded complexes capable of promoting chromosomal alignment [31]. However, the absence of such oligoG stretches in the G-rich sequence of D. geothermalis (G+C content, 66%) indicates that G-quartets are not essential for resistance. In contrast, the D. geothermalis genome contains CRISPR repeats [32], whereas D. radiodurans does not (Table 1). CRISPR repeats are part of a predicted RNA-interference-based system implicated in immunity to phages and integrative plasmids [33], [34]. Since no homologous prophages were identified in the two deinococci, and no CRISPR repeats are present in D. radiodurans, these sequences apparently have no role in determining levels of resistance either.

The 206 kb D. geothermalis megaplasmid (DG206), predicted by genome sequencing, is in lower copy-number than DG574 (Figure 1C). The presence of DG206 in genomic DNA preparations was confirmed in D. geothermalis (DSM 11300) DNA samples used for sequencing and from independent preparations by polymerase chain reaction (PCR) using DG206-specific primers that yielded DNA products of the predicted sizes (Figure S4). DG206 contains 205 predicted open reading frames (ORFs), of which 103 have significant similarity to genes in current databases; approximately 40 are identical to genes in either the D. geothermalis chromosome or DG574; and 28 have homologs in D. radiodurans, including 3 ORFs encoding highly diverged single-stranded DNA-binding proteins. Among other sequences of interest in DG206 are 22 transposon-related ORFs; 11 ORFs related to phage proteins; and 5 ORFs related to conjugal plasmid replication systems. In summary, DG206 is enriched for phage-, integrative plasmid- or transposon-related ORFs, but encodes no known metabolic enzymes and very few replication or repair proteins. Thus, DR206 seems to mimic the trend seen for ORFs in the smallest plasmid (46 kb) of D. radiodurans [16], [21], with no predicted genes implicated in resistance.

The Deinococcus-Thermus Group: Gene-Gain and Gene-Loss

Our previous analysis of the major events in the evolution of the Deinococcus-Thermus group was based on D. radiodurans (ATCC BAA-816) and T. thermophilus strain HB27 [27]. The current study includes additional comparisons with D. geothermalis (DSM 11300) and a second strain of T. thermophilus (HB8). Based on the standard approach of COGs (clusters of orthologous groups of proteins) [35], [36], COGs for Deinococcus and Thermus (tdCOGs) were constructed (Table S2). The tdCOGs were used as a framework for the whole-genome comparisons and evolutionary reconstructions (Figure 2). Using a weighted parsimony method and distantly related bacteria as outgroups, the evolutionary reconstructions revealed significant and independent expansion of the repertoire of genes in the Deinococcus and Thermus lineages following their divergence from a common ancestor. The expansion appears to have occurred through both lineage-specific duplications and gene acquisition via horizontal gene transfer (HGT). The high level of protein family expansion (paralogy), and the larger complement of species-specific genes acquired principally by HGT, could account for the existence of 600–900 more genes in Deinococcus than Thermus.

Figure 2. Whole genome evolutionary reconstructions for D. radiodurans, D. geothermalis, T. thermophilus (HB8) and T. thermophilus (HB27).

For each internal node of tree (open boxes), the inferred number of tdCOGs is shown. For each tree branch the inferred number of tdCOGs lost (minus sign) and gained (plus sign) is shown. For the deep ancestor of the Cyanobacteria, Actinobacteria and Deinococcus-Thermus group (shaded box), the inferred number of COGs is shown. For the extant species, the number of tdCOGs, the number of proteins in tdCOGs (in parentheses), and the number of “free” (not assigned to tdCOGs) proteins (plus sign) are shown.

doi:10.1371/journal.pone.0000955.g002

The Common Ancestor of the Deinococcus Lineage: Trends of Gene-Gain and Gene-Loss

Our previous comparative analysis of T. thermophilus and D. radiodurans identified several evolutionary trends that correlate with the distinct phenotypes of these bacterial lineages [27]. These trends were further refined through the analysis of the D. geothermalis sequence, and the unique features of the Deinococcus lineage were used to better define the pathways implicated in extreme radiation resistance (Table S2). One such trend in Deinococcus, in comparison to the inferred common ancestor of the Deinococcus-Thermus group, is the acquisition of a set of genes involved in transcriptional regulation and signal transduction. Examples of acquired transcriptional regulators include two proteins of the AsnC family, two proteins of the GntR family, and one protein of the IclR family. These families likely are involved in amino acid degradation and metabolism [37]–[39]. Further, the Deinococcus lineage acquired at least six TetR and MerR family regulators dedicated to diverse stress response pathways [40], [41]. Among the acquired signal transduction genes, the most notable examples are two-component regulators of the NarL family (four distinct tdCOGs) involved in the regulation of a variety of oxygen and nitrate-dependent pathways of Escherichia coli [42]; and the presence of several diguanylate cyclase (GGDEF) domain-containing proteins supports an increased role of cyclic diGMP in Deinococci. A second evolutionary trend in Deinococcus is the acquisition of genes encoding proteins involved in nucleotide metabolism, in particular, degradation and salvage [43]–[45]. For example, this group includes genes for xanthine dehydrogenase, urate oxidase, deoxynucleoside kinases, thymidine kinase, FlaR-like kinase, and two UshA family 5′-nucleotidases.

Other gene-gains in Deinococcus relative to Thermus include genes for enzymes of amino acid catabolism and the tricarboxylic acid (TCA) cycle (Table S2). Beyond the differences reported previously [11], [12], the new reconstructions indicate that several catabolic genes of Deinococcus were already present in the Deinococcus-Thermus common ancestor. Following their divergence, however, the Thermus lineage appears to have lost many of these systems, including all enzymes involved in histidine degradation. By contrast, the Deinococcus lineage not only retained a majority of the predicted ancestral catabolic functions, but acquired new pathways including ones involved in the degradation of tryptophan and lysine, and several peptidases (Table S2). A hallmark of the Deinococcus lineage is the presence of two predicted genes for malate synthase, an enzyme of the glyoxylate bypass which converts isocitrate into succinate and glyoxylate, allowing carbon that enters the TCA cycle to bypass the formation of α-ketoglutarate and succinyl-CoA [12]. It has been proposed that the strong upregulation of the glyoxylate bypass observed in D. radiodurans following irradiation facilitates recovery by limiting the production of metabolism-induced reactive oxygen species (ROS) [46]. Dgeo_2616/DRA0277 is the malate synthase ortholog present in the Thermus lineage, but the second predicted deinococcal malate synthetase (Dgeo_2611/DR1155) is unique and only distantly related to homologs in other bacteria. Although the two predicted deinococcal malate synthetases could have similar functions, the genomic context of Dgeo_2611/DR1155 indicates otherwise; Dgeo_2611/DR1155 are both located in a predicted operon with two cyclic amidases of unknown biochemical function.

In a broader context, the present reconstruction indicates that many expanded families of paralogous genes in D. radiodurans proliferated before the emergence of the common ancestor of the Deinococci, but the expansions were not present in the ancestor of the Deinococcus-Thermus group (Table 2). Such Deinococcus-specific expanded families include the Yfit/DinB family of proteins, acetyltransferases of the GNAT family, Nudix hydrolases, α/β superfamily hydrolases, calcineurin family phosphoesterases, and others. Many of these expansions are for predicted hydrolases, phosphatases in particular, but their substrate specificities are either unknown or the affinity of known substrates is extremely low [47]. It has been proposed, therefore, that the majority of these predicted enzymes perform cell-cleaning functions including degradation of damaged nucleic acids, proteins and lipids, and/or other stress-induced cytotoxins [47]. The global proliferation of these enzymes in the Deinococcus lineage (Table S3) supports the acquisition of chemical stress-resistance determinants early in its evolution; and the independent proliferation of determinants within these deinococcal species (e.g., calcinurin phosphatses, Figure S5) might represent secondary adaptations to specific stress environments. In summary, these findings indicate that the Deinococcus stress-resistance phenotypes evolved continuously, both by lineage-specific gene duplications and by HGT from various sources (Table S3, S4 and S5) [21].

Table 2. Ancestral expansions: paralogous gene families expanded in the Deinococcus lineage (DD) versus the Thermus lineage (TT) ancestors

doi:10.1371/journal.pone.0000955.t002

Individual Deinococcus Species: Gene-Gain and Gene-loss

The comparison of gene-gain and gene-loss events in the D. radiodurans and D. geothermalis lineages reveals numerous differences, many of which correlate with their distinct metabolic phenotypes (Figure 3).

Figure 3. Gene-gain and gene-loss for different functional groups for D. radiodurans and D. geothermalis.

Designations of functional groups (from the COG database): J–Translation, ribosomal structure and biogenesis; K–Transcription; L–DNA replication, recombination and repair; D–Cell division and chromosome partitioning; O–Posttranslational modification, protein turnover, chaperones; M–Cell envelope and outer membrane biogenesis; N–Cell motility and secretion; P–Inorganic ion transport and metabolism; T–Signal transduction mechanisms; C–Energy production and conversion; G–Carbohydrate transport and metabolism; E–Amino acid transport and metabolism; F–Nucleotide transport and metabolism; H–Coenzyme metabolism; I–Lipid metabolism; Q–Secondary metabolites biosynthesis, transport and catabolism; V–genes involved in stress response and microbial defense.

doi:10.1371/journal.pone.0000955.g003

D. geothermalis.

The most notable, distinctive feature of D. geothermalis is a greater abundance of genes for sugar metabolism enzymes, which could have been acquired after the divergence of the two Deinococci. The largest group within this set of genes is predicted to be involved in xylose utilization, needed for growth on plant material. D-xylose, which forms xylan polymers, is a major structural component of plant cell walls [48], and the presence of genes for aldopentose (xylose)-degradation explains why D. geothermalis is a persistent contaminant in paper mills [8]. Specifically, D. geothermalis contains genes encoding xylanases (Dgeo_2723; Dgeo_2722), an ABC-type xylose transport system (Dgeo_2699-Dgeo_2703), xylose isomerase (Dgeo_2375, Dgeo_2692, Dgeo_2693, Dgeo_2826), and xylose kinase (Dgeo_2691). Several of the genes that encode enzymes of xylose metabolism form paralogous families (Table S4), most of which form a cluster on the megaplasmid DG574 (Dgeo_2703-Dgeo_2687), which also contains two gene clusters predicted to be involved in carbohydrate utilization (Dgeo_2669-Dgeo_2693, Dgeo_2832-Dgeo_2812). By comparison, there are no large clusters of functionally related genes on the D. geothermalis chromosome; approximately 80 and 120 encoding proteins involved in sugar-metabolism were identified on DG574 and the chromosome, respectively. The putative xylose metabolism functions of D. geothermalis appear to represent an expansion of a pre-existing, broad and diverse set of functions underlying the saccharolytic phenotypes of all Deinococci [7], [11], [49], [50]. In contrast, D. radiodurans has a proteolytic lifestyle, where a loss of various amino acid biosynthetic pathways (Figure 3) [51] was accompanied by a gain of several predicted peptidases (DR0964, DR1070, DR2310, DR2503) and a urease system (DRA0311-DRA0319) [27]. Thus, the evolutionary processes underlying the emergence of extreme resistance in Deinococci appear not to be dependent on a particular set of genes for sugar- or nitrogen-metabolism. In summary, these findings support that DG574 is essential to the natural growth modes of D. geothermalis, which is a proficient saccharolytic organism [7], [49], [50] and strengthen the case that the megaplasmids in the Deinococcus-Thermus group are major receptacles of horizontally acquired genes, as proposed previously [27].

Further supporting the notion that a distinct set of metabolic genes is not a prerequisite for high levels of radioresistance, there are patent differences between sulfate and energy metabolism in D. geothermalis and D. radiodurans. In agreement with previously published results [7], [11], [51], the prototrophic D. geothermalis has orthologs of the nadABCD genes that are required for nicotinamide adenine dinucleotide (NAD) biosynthesis, whereas the auxotrophic D. radiodurans lacks these genes and is dependent on an exogenous source of this coenzyme [21], [51]. Another example illustrating the relationship in D. radiodurans between gene-loss and its growth requirements is that of cobalamine (vitamin B12). Whereas D. geothermalis and T. thermophilus are not dependent on B12 in minimal medium, D. radiodurans can utilize inorganic sulfate as the sole source of sulfur only when vitamin B12 is present [52]. Conversely, D. geothermalis has lost several genes for enzymes of protoheme biosynthesis (HemEZY) [53], which in D. geothermalis likely yields siroheme under the microaerophilic conditions which predominate at the T_opt of D. geothermalis; the solubility of dioxygen in water at 50°C is significantly lower than at 32°C, the T_opt of D. radiodurans.

There are also important differences between the systems for enzymes implicated in energy transformation in D. geothermalis and D. radiodurans. The D. geothermalis chromosome encodes two heme-copper cytochrome oxidases of types ba3 and caa3 [54]; and a cytochrome bd ubiquinol oxidase system (Dgeo_2707-Dgeo_2704), known to be expressed under oxygen-limiting conditions [55], is encoded by DG574. In contrast, D. radiodurans encodes only the caa3 oxidase system (DR2616-DR2620), which apparently was present in the Deinococcus-Thermus common ancestor. Furthermore, D. geothermalis encodes genes for proteins that comprise an assimilatory nitrite NAD(P)H reductase and a molybdopterin-cofactor-dependent nitrate reductase system (Dgeo2392-Dgeo_2389), which also is known to be expressed under anaerobic conditions [56], [57]; and D. geothermalis encodes several predicted multi-copper oxidases (Dgeo_2590, Dgeo_2559, Dgeo_2558) that are not present in D. radiodurans and are most similar to homologs from nitrogen-fixing bacteria. Since nitrogen fixation in D. geothermalis has not yet been studied, the possibility remains that these enzymes are involved in dissimilatory anaerobic reduction of nitrate or nitrite [58], [59]. D. geothermalis, but not D. radiodurans, also encodes a formate dehydrogenase, which is related to nitrate reductase and has a possible role in energy transfer under anaerobic conditions [60].

D. radiodurans.

In general, the evolutionary trends in D. radiodurans lineage appear to mimic closely those of the Deinococcus lineage, which is evident from the analysis of expanded families of paralogous genes (Table S5). In particular, proliferation of genes for the Yfit/DinB family, Nudix enzymes, acetyltransferases of the GNAT superfamily, and the α/β hydrolase superfamily was observed (Table 2). Plausible resistance-related functions readily can be proposed for these and other expanded families of deinococci. For example, hydrolases might degrade oxidized lipids; Yfit/DinB proteins might be involved in cell damage-related pathways [21]; subtilisin-like proteases might degrade proteins oxidized during irradiation [17], [61]; and the Nudix-related hydrolase, diadenosine polyphosphatase (ApnA), yields adenosine, a molecule that has been implicated in cytoprotection from oxidative stress and radiation [62], [63].

Several families expanded in D. radiodurans are predicted to possess functions potentially relevant to stress response, but are not conserved in D. geothermalis; most likely, non-conserved families can be disqualified as major contributors to the extreme IR and desiccation resistance phenotypes. Families that are specifically expanded in D. radiodurans include the TerZ family of proteins, which are predicted to confer resistance to various DNA damaging agents [64], [65]; secreted proteins of the PR1 family, whose homologs are involved in the response to pathogens in plants, and resistance to hydrophilic organic solvents in yeast [66], [67]; PadR-like regulators, which are implicated in the regulation of amino acid catabolism and cellular response to chemical stress agents and drugs [68]–[70]; TetR/AcrR transcriptional regulators, which are involved in antibiotic resistance regulation [40]; and KatE-like catalases, which would decompose hydrogen peroxide [71]–[73]. In contrast, there are family expansions which are shared by D. radiodurans and D. geothermalis, but have no obvious role in radiation or desiccation resistance. These include SAM-dependent metyltransferases (COG0500) and an uncharacterized family of predicted P-loop kinases (COG0645). In some bacteria, homologs of these kinases are fused to phosphotransferases that mediate resistance to aminoglycosides [74].

Since the IR-, UV- and desiccation-resistance profiles of D. radiodurans and D. geothermalis are identical (Figure 1) [5], the subset of stress response genes in D. radiodurans that are not unique, but exist in excess compared to D. geothermalis are unlikely to be required for extreme resistance either (Figure 3). This subset includes two Cu-Zn superoxide dismutases (SOD), a peroxidase, two HslJ-like heat shock proteins, and many genes implicated in antibiotic resistance (Table S5). Consistently, SodA and KatA of D. radiodurans can be disrupted with almost no loss in radiation resistance [75], and antibiotics have little effect on survival following irradiation provided corresponding antibiotic resistance genes are present [18], [76]–[79].

The Deinococcus lineage.

Considerable independent gene-gain was detected in both D. geothermalis and D. radiodurans lineages in several other functional categories including transcriptional regulation, signal transduction, membrane biogenesis, inorganic ions metabolism, and to a lesser extent DNA replication and repair (Figure 3). In general, regulatory functions mirror the metabolic and stress-response-related differentiation of these two species outlined above. For instance, among the 12 genes for predicted transcriptional regulators that apparently were acquired in the D. geothermalis lineage, five are similar to ones known to be involved in the regulation of sugar metabolism in other bacteria, two of the RpiR family and three of the AraC family [80], [81]. By contrast, D. radiodurans has at least 25 unique genes for transcriptional regulators: three of the ArcR family; 16 of the Xre family; one of the CopG/Arc/MetJ family; and five of a species-specific expanded family reported previously [61] that likely is responsible for stress-response control [82]^-[85]. Other potentially independent gains involve genes predicted to be involved in signal transduction systems. D. radiodurans, for example, encodes photochromic histidine kinase, a protein that has been extensively studied in D. radiodurans and plays a role in the regulation of pigment biosynthesis [86], [87], but is missing in D. geothermalis. Alternatively, D. geothermalis encodes a putative negative regulator of sigma E, a periplasmic protein of the RseE/MucE family (Dgeo_2271). So far, RseE/MucE-members have been detected only in proteobacteria, where it regulates the synthesis of alginate, an extracellular polysaccharide which plays a key role in the formation of biofilms [88]. D. geothermalis, however, likely does not produce alginate itself since it has no orthologs of the genes of the alignate pathway [89]. On the other hand, D. geothermalis has clusters of genes implicated in exopolysaccharide biosynthesis, with the most notable cluster located on DG574 (Dgeo_2671-Dgeo_2646). It seems likely that this cluster is involved in the biosynthesis of exopolysaccharides, which might facilitate biofilm formation in D. geothermalis, and the Dgeo_2271 protein could be a regulator of this process. Overall, D. radiodurans encodes approximately 470 unique, uncharacterized proteins, for which no function could be predicted, compared to approximately 286 such proteins in D. geothermalis. Thus, an additional 756 unique, uncharacterized genes of the Deinococcus lineage can be excluded from the pool of putative determinants of the extreme IR, UV and desiccation resistance phenotype.

Reassessment of the Genetic Determinants of Radiation Resistance

Evolutionary Provenance of the Genomic Features Previously Implicated in the Radiation Resistance of D. radiodurans.

Over the last two decades, extensive experimental and comparative-genomic analyses have been dedicated to the identification and evolutionary origin of the genetic determinants of radiation resistance in D. radiodurans. Early on, it became evident that the survival mechanisms underlying extreme radiation resistance in D. radiodurans probably were not unique. In 1994, for example, IR-sensitive D. radiodurans polA mutants were fully complemented by expression of the polA gene from the IR-sensitive E. coli [4]; and in 1996, UV-sensitve D. radiodurans uvrA mutants were complemented by uvrA from E. coli [90], suggesting that these recombination and excision repair genes are necessary but not sufficient to produce extreme DNA damage resistance. Following the whole-genome sequencing of D. radiodurans in 1999 [16], comparative-genomic analysis revealed many distinctive genomic features that subsequently became the focus of high throughput experiments, including the analysis of transcriptome and proteome dynamics of D. radiodurans recovering from IR [46], [91], [92]. Surprisingly, the cellular transcriptional response to IR in D. radiodurans appeared largely stochastic, and mutant analyses confirmed that many of the highly induced uncharacterized genes were unrelated to survival. So far, those correlative studies have failed to produce a coherent, comprehensive picture of the complex interactions between different genes and systems that have been thought to be important for the resistance phenotype.

The complete sets of orthologous genes in D. radiodurans and D. geothermalis are listed in Table S2. Within the subgroup of genes in D. radiodurans previously implicated in resistance by transcriptional induction following exposure to IR [46] (3 hours after irradiation and displaying more than a 2-fold induction), 45% have no othologs in D. geothermalis. This raises the possibility that many genes induced in irradiated D. radiodurans do not functionally participate in recovery, or that D. geothermalis carries a distinct set of resistance determinants. From the subgroup of putative resistance genes lacking counterparts in D. geothermalis, we constructed D. radiodurans knockouts of four representative genes: i) a ligase predicted to be involved in DNA repair (DRB0100) [46]; ii) a LEA76 desiccation resistance protein homolog (DR0105) [46]; iii) a predicted protein implicated in stress response (DR2221) [46]; and iv) a protein of unknown function (DR0140) [46]. Homozygous disruptions of each of these genes in D. radiodurans (Figure S6) had no significant effect on IR resistance (Figure 4).

Figure 4. IR resistance of wild-type (ATCC BAA-816) and D. radiodurans mutants lacking orthologs in D. geothermalis (DSM 11300).

Survival values following 9 kGy (⁶⁰Co) are from three independent trials with standard deviations shown. The structure of the homozygous mutants DRB0100, DR2221, DR105 and DR0140 are presented in Figure S6.

doi:10.1371/journal.pone.0000955.g004

By contrast, most of the genes whose mutants display radiation-sensitive phenotypes in D. radiodurans [4], [20], [46], [92], [93] are conserved in D. geothermalis. To date, 15 single-gene mutants of D. radiodurans have been reported to be moderately to highly radiation-sensitive; of these, 13 genes have orthologs in D. geothermalis (Table 3). The exceptions are DR0171 and DR1289, which encode the DNA helicase RecQ and a transcriptional regulator, respectively (Table 3). Remarkably, 10 of the 15 genes are conserved in other bacteria and are well-characterized components of DNA repair pathways. However, 5 of the 15 genes (DR0003, DR0070, DR0326, DR0423, DRA0346) are unique to the Deinococcus lineage, supporting the existence of at least a few novel resistance mechanisms.

Table 3. D. radiodurans genes implicated in radiation resistance

doi:10.1371/journal.pone.0000955.t003

Given that the two Deinococcus species are equally resistant to IR (Figure 1A), genes dedicated specifically to the extreme radiation/desiccation response are expected to belong to the set of tdCOGs. D. radiodurans and D. geothermalis share 231 tdCOGs that are relatively uncommon in other prokaryotes, and 63 of these are unique to the Deinococcus lineage. Using the most sensitive methods available to predict function, we reanalyzed these tdCOGs by using a remote sequence similarity search, and genomic context analysis [94]–[96]. Interpretation of such analyses, however, is constrained by the complexity and ambiguities inherent in the approach, and by the knowledge base. In contrast, many cytosolic proteins (e.g., RecA, PolA, SodA and KatA) are known to be intimately involved in resistance, so we present functional predictions for 50 genes (Table S6). Among the predictions for cytosolic proteins, several are new and potentially relevant to resistance. For example, DR0644 (Figure 5A) is predicted to be a distinct Cu/Zn superoxide dismutase that could defend against metabolism-induced oxidative stress during recovery (Table S7); and DR0449 (Figure 5B) is a divergent member of the RNAse H family that is fused to a novel domain, a combination that is currently unique to Deinococcus. Other functional insights were for DR0041/Dgeo_0188, that is a paralog of DR0432 (DdrA) (Figure 5C); and a member of the RAD22/Rad52 family (Figure 5C) of single-stranded annealing proteins [97], that yields a moderately sensitive phenotype in D. radiodurans upon disruption [98]. Interestingly, the radiation-sensitive T. thermophilus encodes a homolog of DdrA (TTC1923), indicating that this protein had an ancestral role that was not directly related to radiation resistance. Notably, we continue to find proteins in Deinococcus species which are only remotely similar to well-characterized enzymes in other organisms, and it is difficult to predict their role in the cell or radiation resistance. For example, we have identified a protein that is conserved in both D. geothermalis and D. radiodurans and is distantly related to enzymes of the QueF/FolE family, which are involved in queuosine/folate biosynthesis (Figure 5D), but their role in the Deinococci remains undefined. Collectively, these results support the conclusion that many genes that are significantly induced in irradiated D. radiodurans are not involved in recovery (Table 3). Thus, the genome of D. geothermalis is a resource of major importance in delineating a reliable minimal set of resistance determinants, by corroborating those that are conserved and ruling out those which are unique.

Figure 5. Multiple alignments of selected families conserved in two Deinococcus species.

The multiple alignments were constructed for selected representative sets of sequences by the MUSCLE program [154]. Where necessary, alignments were modified manually on the basis of PSI-BLAST outputs [94]. The positions of the first and the last residue of the aligned region in the corresponding protein are indicated for each sequence. The numbers within the alignment refer to the length of inserts that are poorly conserved between all the families. Secondary structure elements are denoted as follows: E-β-strand; and H-α-helix. The coloring scheme is as follows: predominantly hydrophobic residues are high-lighted in yellow; positions with small residues are in green; positions with turn-promoting residues are in cyan; positions with polar residues are in red; hydroxyl-group containing residues are in blue; aromatic residues are in magenta; and invariant, highly conserved groups are in boldface. A, DR0644-Dgeo_0284 conserved pair of orthologs belong to the copper/Zinc superoxide dismutase family; shaded letters refer to amino acids that play an important role in the Cu²⁺/Zn²⁺ coordination environment and in the active site region. The bottom line shows the correspondence between the most conserved regions corresponding to the β-stand structural core and conserved in most family members as denoted in Bordo et al [157]. B, Dgeo_0137-DR0449 are highly specific for the Deinococcus lineage proteins that have an RNAse H-related domain. Catalytic residues conserved in the RNAse H family are shaded. Secondary structure elements are shown for E. coli RNase HI (PDB:2rn2). C, DR0041-Dgeo_0188 is another conserved pair (DdrA-related) of proteins belonging to the Rad52 family of DNA single-strand annealing proteins [97]. Secondary structure elements are shown for human RAD52 (PDB:1KN0) [158]. sak is a phage gene described previously [159]; D, DR0381-Dgeo_0373 are diverged homologs of NADPH-dependent nitrile reductase (GTP cyclohydrolase I family) that might be involved in nucleotide metabolism. The conserved Cys and Glu found in the substrate binding pocket of both protein families and specific zinc-binding and catalytic residues in the FolE family are shaded. The QueF family motif is boxed. Other catalytic residues in FolE not found in QueF are in yellow. Genbank Identifier (gi) numbers are listed on the right.

doi:10.1371/journal.pone.0000955.g005

Delineation of the Deinococcus Radiation Response Regulon.

A potential radiation-desiccation response regulon and the corresponding regulator common to D. radiodurans and D. geothermalis were identified using the approach developed by Mironov et al [99], [100]. In the search for such a regulator, we used a training-set comprised of sequences flanking D. radiodurans genes that were strongly upregulated by IR, and for which the corresponding mutants were radiosensitive (Table 3) [92]. The upstream regions of several genes from the training set (DR0326, ddrD; DR0423, ddrA; DRA0346, pprA; DR0070, ddrB) revealed a strong palindromic motif, designated the radiation/desiccation response motif (RDRM). Using a positional weight matrix, the RDRM was used to generate the initial profile and to scan the entire D. radiodurans genome. This genome survey picked up a similar motif in the upstream regions of other genes upregulated after irradiation [92]. The upstream regions with the highest scores (DR0219, DR0906, DR1913 and DR0659) were then used to better define the RDRM, and the complete genomes of D. radiodurans and D. geothermalis were scanned with the updated motif. Using moderately relaxed parameters (Materials and Methods), approximately 120 genes in each of the Deinococcus genomes were selected by the screen. The final, most conservative prediction of the radiation/desiccation response (RDR) regulon consisted of two groups: (i) a set of orthologous genes present in both Deinococcus species that contain the RDRM; and (ii) a set of unique genes of D. radiodurans that contain the RDRM and are upregulated during the recovery from irradiation [46], [92]. Since microarray data for D. geothermalis are not available, it was not possible to predict a set of unique RDRM-dependent genes for this species. Table 4 lists the set of genes predicted to comprise the regulon together with the corresponding RDRM sites (Figure 6). Collectively, the RDR regulon is predicted to consist of a minimum of 29 genes in D. radiodurans and 25 genes in D. geothermalis, contained within 20 operons in each species.

Figure 6. Sequence signature of a predicted site of a radiation response regulator.

Four different nucleotides are shown by four letters (A, G, C, T) in different colors. The height of the letter is proportional to its contribution to the information content in the corresponding position of the multiple alignment used for “sequence logo” construction. The figure was constructed by the “sequence logo” program described previously [160].

doi:10.1371/journal.pone.0000955.g006

Table 4. The predicted radiation and desiccation resistance regulon of Deinococci

doi:10.1371/journal.pone.0000955.t004

The RDR regulon is dominated by DNA repair genes, including the recombinational repair proteins RecA and RecQ [101], [102]; the mismatch repair proteins MutS and MutL, that are located in one operon in D. geothermalis; and the UvrB and UvrC proteins, which are involved in nucleotide excision repair (Table 4). In addition, the predicted RDR regulon includes the transketolase gene. In bacteria, transketolase is a key enzyme of the pentose-phosphate pathway for carbohydrate metabolism and is known to be induced by a variety of stress conditions including cold shock, and mutagens that trigger the SOS response [103]. Moreover, the pentose-phosphate pathway in D. radiodurans is reported to facilitate DNA excision repair induced by UV irradiation and hydrogen peroxide (H₂O₂)