Main

The Rickettsia are α-proteobacteria that multiply in eukaryotic cells only. R. prowazekii is the agent of epidemic, louse-borne typhus in humans. Three features of this endocellular parasite deserve our attention. First, R. prowazekii is estimated to have infected 20–30 million humans in the wake of the First World War and killed another few million following the Second World War (ref. 1). Because it is the descendent of free-living organisms2,3,4, its genome provides insight into adaptations to the obligate intracellular lifestyle, with probable practical value. Second, phylogenetic analyses based on sequences of ribosomal RNA and heat-shock proteins indicate that mitochondria may be derived from the α-proteobacteria5,6. Indeed, the closest extant relatives of the ancestor to mitochondria seem to be the Rickettsia 7,8,9,10. That modern Rickettsia favour an intracellular lifestyle identifies these bacteria as the sort of organism that might have initiated the endosymbiotic scenario leading to modern mitochondria11. Finally, the genome of R. prowazekii is a small one, containing only 1,111,523 base pairs (bp). Its phylogenetic placement and many other characteristics identify it as a descendant of bacteria with substantially larger genomes2,3,4. Thus Rickettsia, like mitochondria, are good examples of highly derived genomes, the products of several types of reductive evolution.

The genome sequence of R. prowazekii indicates that these three features may be related. For example, prokaryotic genomes evolving within a cell dominated by a much larger, eukaryote genome and constrained by bottle-necked population dynamics will tend to lose genetic information12,13. Predictable sets of expendable genes will tend to disappear from the prokaryotic genome when they are made redundant by the activities of nuclear genes. Likewise, non-essential sequences and otherwise highly conserved gene clusters may be obliterated by deleterious mutations that are fixed in clonal parasite or organelle populations because they cannot be eliminated by selection. This process is ongoing in the Rickettsia genomes, as shown by the identification of sequences that have recently become pseudogenes. Also, a large fraction (25%) of non-coding sequences in this genome may be gene remnants that have been degraded by mutation and have not yet been removed from the genome. Finally, transfer of genes from a mitochondrial ancestor to the nucleus of the host would both reduce the mitochondrial genome size and stabilize the symbiotic relationship. Phylogenetic reconstructions that identify genes in the Rickettsia genome as sister clades to eukaryotic homologues found in the nucleus or the organelle support this interpretation. Rickettsia and mitochondria probably share an α-proteobacterial ancestor and a similar evolutionary history.

General features of the genome

The circular chromosome of R. prowazekii strain Madrid E has 1,111,523 bp and an average G+C content of 29.1% (Figs 1 , 2 (Note: For Figure 2, please refer to PDF File: 217k)). The genome contains 834 complete open reading frames with an average length of 1,005 bp. Protein-coding genes represent 75.4% of the genome and 0.6% of the genome encodes stable RNA. We have assigned biological roles to 62.7% of the identified genes and pseudogenes; 12.5% of the identified genes match hypothetical coding sequences of unknown function and the remaining 24.8% represent unusual genes with no similarities to genes in other organisms (Table 1 (Note: For Table 1, please refer to PDF File: 469k)). Multivariate statistical analysis has shown that there is no major variation in codon-usage patterns among genes that are expressed in different amounts, indicating that codon-usage patterns in R. prowazekii may be dominated mainly by mutational forces14. G+C-content values at the three codon positions average 40.4, 31.2 and 18.6%, and these values are similar at different positions in the genome. We classified the open reading frames with significant sequence-similarity scores to gene sequences in the public databases into functional categories (Table 1 (PDF File: 469k)) that allow comparisons with the metabolic profiles of other bacterial genomes15,16,17,18,19,20,21,22,23.

Figure 1: Overall structure of the R. prowazekii genome.
figure 1

The putative origin of replication is at 0 kb. The outer scale indicates the coordinates (in base pairs). The positions of pseudogenes are highlighted with death's heads. The distribution of genes is shown on the first two rings within the scale. The location and direction of transcription of rRNA are shown by pink arrows and of tRNA genes by black arrows. The next circle in shows GC-skew values measured over all bases in the genome. Red and purple colours denote positive and negative signs, respectively. The window size was 10,000 nucleotides and the step size was 1,000 nucleotides. The central circles shows GC-skew values calculated for third positions in the codon only. GC-skew values were calculated separately for genes located on the outer strand (green) and on the inner strand (blue). To allow easier visual inspection, the signs of the values calculated for genes located on the inner strand have been reversed.

Non-coding DNA. The coding content of previously sequenced bacterial genomes is, on average, 91%, ranging from 87% in Haemophilus influenzae to 94% in Aquifex aeolicum. In comparison, a large fraction of the R. prowazekii genome, 24%, represents non-coding DNA ( Fig. 3). A small fraction of this corresponds to pseudogenes (0.9% of the genome) and less than 0.2% of the genome is accounted for by non-coding repeats. The remaining 22.9% contains no open reading frames of significant length and it has the low G+C content (mean 23.7%) that is characteristic of spacer sequences in the R. prowazekii genome14. A region of 30 kilobases (kb) located at position 886–916 kb contains as much as 41.6% non-coding DNA and 11.5% pseudogenes. The non-coding DNA in this region has a small, but significantly higher, G+C content (mean 27.3%) than non-coding DNA in other areas of the genome (mean 23.7%) ( P < 0.001), indicating that it may correspond to inactivated genes that are being degraded by mutation (Fig. 3).

Figure 3: G+C content in intergenic regions longer than 20 bp in the R. prowazekii genome.
figure 2

The empty circles correspond to spacer sequences located at 886 to 916 kb, a region with an unusually large fraction of non-coding DNA and pseudogenes.

Origin of replication. The origin of replication has not been experimentally identified in the R. prowazekii genome, but we identified dnaA at 750 kb. However, the genes flanking the dnaA gene differ from the conserved motifs found in Escherichia coli and Bacillus subtilis (rnpA–rpmH–dnaA–dnaN–recF–gyrB ). In R. prowazekii, the genes rnpA and rpmH are located in the vicinity of dnaA, but in the reverse orientation compared to the consensus motif, and dnaN, recF and gyrB are located elsewhere.

The origin and end replication in microbial genomes are often associated with transitions in GC skew (G − C/G + C) values24. In R. prowazekii we observe transitions in the GC skew values at around 0 and 500–600 kb (Fig. 1). There is a weak asymmetry in the distribution of genes in the two strands, such that the first half of the genome has a 1.6-fold higher gene density on one strand and the second half of the genome has a 1.6-fold higher gene density on the other strand. The shift in coding-strand bias correlates with the shift in GC-skew values. As most genes are transcribed in the direction of replication in microbial genomes, the origin of replication may correspond to the shift in GC-skew values at the position that we have chosen as the start point for numbering. Indeed, several short sequence stretches that are characteristic of dnaA -binding motifs are found in the intergenic region of genes RP001 and RP885 at 0 kb, supporting this interpretation.

Stable RNA sequences and repeat elements. We identified 33 genes encoding transfer RNA, corresponding to 32 different isoacceptor-tRNA species. There is a single copy of each of the rRNA genes, with rrs located more than 500 kb away from the rrl–rrf gene cluster (Fig.1). Comparison of the sequences from ten different Rickettsia species indicates that the disruption of the rRNA gene operon preceded the divergence of the typhus group and spotted fever group Rickettisia (S.G.E.A. et al., unpublished observations). In addition, the genome contains a short sequence with similarity to a 213-nucleotide RNA molecule in Bradyrhizobium japonicum that may regulate transcription25.

There are unusually few repeat sequences in this genome. We identified four different types of repeat sequence: all of these are located in intergenic regions. There is a sequence of 80 bp that is repeated seven times downstream of rpmH and rnpA in the dnaA region. A repetitive sequence of 325 bp is found at two intergenic regions that are more than 80 kb apart, downstream of the genes ksgA and rnh, respectively. A 440-bp-long repetitive sequence has been identified at two intergenic sites, 140 kb apart; one of these sites is downstream of rrf and the others downstream of pdhA and pdhB. Finally, two similar sequences of 730 bp are located immediately next to each other at 850 kb.

Paralogous families. We have identified 54 paralogous gene families comprising 147 gene products. Of these, 125 have an assigned function. Most paralogues encode proteins with transport functions, such as the ABC transporters, the proline/betaine transporters and the ATP/ADP transporters. Five paralogous genes located next to each other at 115 kb encode putative integral membrane proteins with unknown functions.

Biosynthetic pathways

A striking feature of the R. prowazekii genome is the small proportion of biosynthetic genes compared with free-living proteobacterial relatives (such as Haemophilus influenzae, Helicobacter pylori and E. coli)15,19,20. This scarcity of biosynthetic functions is also seen in diverse endocellular and epicellular parasites16,17,18,23. This scarcity of biosynthetic functions is also seen in diverse endocellular and epicellular parasites16,17,18,23.

Amino-acid metabolism. As many as 43 and 69 genes required for amino-acid biosynthesis are found in Helicobacter pylori and Haemophilus influenzae , respectively. In contrast, Mycoplasma genitalium and Borrelia burgdorferi contain only glyA, which encodes serine hydroxymethyltransferase. This gene is also found in R. prowazekii (Table 1 (PDF File: 469k)). Serine hydroxymethyltransferase catalyses the conversion of serine and tetrahydrofolate into glycine and methylenetetrahydrofolate, respectively. A role in tetrahydrofolate metabolism may account for the ubiquity of glyA in bacteria.

Seven genes normally associated with lysine biosynthesis (lysC, asd, dapA, dapB, dapD, dapE and dapF) are also present in R. prowazekii. The biosynthetic pathways leading to lysine, methionine and threonine share the first two of these (lysC and asd). However, none of the downstream genes for threonine biosynthesis are found in R. prowazekii. Likewise, the lysine pathway is incomplete, and lysA, which encodes the enzyme that converts meso-diaminopimelate to lysine, is missing. The likely role of the upstream genes of this pathway in R. prowazekii is the biosynthesis of diaminopimelate, an essential envelope component. We have therefore classified these genes as ‘cell-envelope’ genes (Table 1 (PDF File: 469k)).

We have identified other genes that are superficially involved in the metabolism of amino acids, but which apparently function in deamination pathways that divert amino acids into the tricarboxylic acid (TCA) cycle. For example, there is aatA, encoding aspartate aminotransferase, which catalyses the degradation of aspartate to oxaloacetate and glutamate. tdcB encodes threonine deaminase, which converts threonine into α-ketobutyrate. Another gene (ilvE) encodes branched-chain-amino-acid aminotransferase, which converts leucine, isoleucine or valine into glutamate. pccA and pccB encode propionyl-CoA carboxylase, which converts propionyl-CoA, an intermediate in the breakdown of methionine, valine and isoleucine, into succinyl-CoA. The pccA and pccB gene products show greatest similarity to the eukaryotic proteins that are located in the mitochondrial matrix.

Nucleotide biosynthesis. No genes required for the de novo syntheses of nucleosides have been found in the R. prowazekii genome. However, four genes required for the conversion of nucleoside monophosphates into nucleoside diphosphates (adk, gmk, cmk and pyrH ) are present. There are also two genes encoding ribonucleotide reductase, which converts ribonucleoside diphosphates into deoxyribonucleoside diphosphates. Nucleoside diphosphate kinase (encoded by ndk), which converts NDPs and dNDPs to NTPs and dNTPs, is also present in R. prowazekii. Finally, there is a complete set of genes for the conversion of dCTP and dUTP into TTP, including thyA, which codes for thymidylate synthase. Thus, the R. prowazekii genome encodes all of the enzymes required for the interconversion of nucleoside monophosphates into all of the other required nucleotides. The nucleoside monophosphates are probably imported from the eukaryotic host.

Energy metabolism

Early in its infectious cycle, R. prowazekii uses the ATP of the host with the help of membrane-bound ATP/ADP translocases. However, R. prowazekii is also capable of generating ATP, which may compensate for the gradual depletion of cytosolic ATP later in the infection. R. prowazekii 's repertoire of genes involved in ATP production and transport include determinants for the TCA cycle, the respiratory-chain complexes, the ATP-synthase complexes and the ATP/ADP translocases (Table 1 (PDF File: 469k)). Genes to support anaerobic glycolysis are absent.

Pyruvate dehydrogenase. Pyruvate is imported into mitochondria directly from the cytoplasm and converted into acetyl-CoA by pyruvate dehydrogenase. The genes encoding three components (E1–E3) of the pyruvate dehydrogenase complex are found in R. prowazekii, indicating that it too uses cytosolic pyruvate. Pyruvate dehydrogenase (E1) consists of two subunits (α and β) in R. prowazekii, mitochondria and Gram-positive bacteria; the corresponding genes are clustered in the genome. In contrast, proteobacteria such as E. coli, Haemophilus influenzae and Helicobacter pylori have a single subunit for the E1 component and these have little similarity to the α and β subunits of the E1 component in R. prowazekii and mitochondria (data not shown).

Two paralagous genes code for the dihydrolipoamide dehydrogenase (E3) in R. prowazekii. One of these most resembles mitochondrial homologues, whereas the other is most similar to bacterial homologues (data not shown). The presence of several paralogous gene families for pyruvate dehydrogenases complicates attempts to reconstruct a genome phylogeny based on these genes.

ATP production. Genes encoding all enzymes in the TCA cycle are found in R. prowazekii. Proton translocation is mediated by NADH dehydrogenase (complex I), cytochrome reductase (complex III) and cytochrome oxidase (complex IV). Several clusters of genes code for components of the NADH dehydrogenase complex. Seven of these genes (nuoJKLM and nuoGHO) are located near to each other, but the order of genes is inverted relative to the order of this cluster in E. coli. An additional set of five genes is grouped in the order nuoABCDE, but the single genes nuoF and nuoN are distant from both of these clusters. Several proteins in the cytochrome bc 1 reductase complex, such as ubiquinol–cytochrome c reductase (encoded by petA), cytochrome b (encoded by cytb) and cytochrome c 1 (encoded by fbhC), are present, as are several subunits of the cytochrome oxidase complex.

The ATP-synthesizing complex is composed of the ATP synthase F1 component (comprising five polypeptides, α, β, γ, ε and δ) and the Fo component, a hydrophobic segment that spans the inner mitochondrial membrane. The genes encoding these components are normally clustered in one of the most highly conserved operon structures in microbial genomes. In R. prowazekii, however, the ATP-synthase genes encoding the α, β, γ, δ and ε subunits of the F1 complex (atpH, atpA, atpG, atpD and atpC) are clustered in the common order, but atpB, atpE and atpF, encoding the A, B and C chains of the Fo complex, are split from this cluster.

Replication, repair and recombination

R. prowazekii has a smaller set of genes involved in DNA replication than do free-living bacteria such as E. coli, Haemophilus influenzae and Heliocobacter pyrlori. Four genes have been identified that code for the core structure of DNA polymerase III, which includes the α (dnaE), ε (dnaQ), β (dnaN), γ and θ (dnaX) subunits. Extra subunits present in the E. coli DNA polymerase III are missing from R. prowazekii, as well as from M. genitalium and B. burgdorferi.

Genes encoding DNA-repair mechanisms are similar in the small genomes of the parasites R. prowazekii, M. genitalium and B.burgdorferi . Thus, genes involved in the repair of ultraviolet-induced DNA damage (uvrABCD) have been identified in all three genomes. In R. prowazekii , DNA-excision repair probably occurs by a pathway involving endonuclease III, polI and DNA ligase, as in B.burgdorferi.

The R. prowazekii genome has a limited capacity for mismatch repair. The DNA-mismatch-repair enzymes encoded by mutL and mutS are present, but mutH and mutY are not. There is a complete lack of mut genes in M. genitalium, but mutL and mutHLY have been identified in B. burgdorferi and Chlamydia trachomatis . The transcription-repair coupling factor (encoded by mfD) is found in R. prowazekii, B.burgdorferi and C. trachomatis but not in M. genitalium.

The R. prowazekii genome contains several genes involved in homologous recombination, such as recA, recF, recJ, recN and recR. A similar set of genes has been found in A. aeolicus 21. The rec genes are scattered in the other small genomes of parasites. The RecBCD complex is missing in R. prowazekii, M. genitalium and Helicobacter pylori but it has been identified in B. burgdorferi.

Transcription and translation

R. prowazekii has three subunits (α, β and β′) of the core RNA polymerase, as well as σ70 and one alternative σ factor, σ32, which controls transcription of the genes encoding heat-shock proteins in E. coli. Genes involved in transcription elongation and termination, nusA, nusB, nusG, greA and rho, are also present. The gene encoding σ32 is absent in most other small genomes, such as those of B.burgdorferi , Helicobacter pylori, M. genitalium and C. trachomatis , although genes for heat-shock proteins are present.

An unusually large number of genes involved in RNA degradation are found in R. prowazekii. Of these, only four appear to be common to the bacterial genomes analysed so far (those encoding polyribonucleotide nucleotidyltransferase and ribonucleases HII, III and P). Four more ribonucleases (D, E, HI and PH) are present in R. prowazekii, but in none of the other small parasites.

Of the 33 identified tRNA genes, which code for 32 different tRNA isoacceptor species, two code for tRNAPhe. There are two tRNA species for most of the amino acids that are encoded by four-codon boxes; the exceptions are the four-codon boxes for proline and valine, for which we have identified only one isoacceptor-tRNA species, with U in the first anticodon position. selC, which codes for tRNASec, and selABD are missing. R. prowazekii has a set of genes coding for tRNA modifications (tgt , queA, trmD, truA, truB and miaA) which resembles that of Helicobacter pylori, C. trachomatis and B. burgdorferi; M. genitalium has only trmD and truA.

In R. prowazekii, 21 genes encode 18 of the 20 aminoacyl-tRNA synthetases normally required for protein synthesis. There are two genes (gltX) encoding glutamyl-tRNA synthetase. As seen in several bacterial genomes25, the gene coding for glutaminyl-tRNA synthetase, glnS, is missing. Three genes encoding subunits of the glutamyl-tRNA amidotransferase are present, indicating that a glutamyl-tRNA charged with glutamic acid may be transamidated to generate Gln-tRNA. The gene coding for asparaginyl-tRNA synthetase, asnS, is also missing from the R. prowazekii genome as well as from Helicobacter pylori, C. trachomatis and A. aeolicus 26. A transamidation process to form Asn-tRNAAsn from Asp-tRNAAsn has been proposed for the archaeon Haloferax volcanii 27 and this reaction may also occur in R. prowazekii. The valyl-tRNA synthetase is 38.3% identical to its homologue in Methanococcus jannaschii, but only 27.6% identical to its most similar homologue in bacteria, which is found in Bacillus stearothermophilus, possibly indicating a horizontal transfer event. The lysyl-tRNA synthetase (encoded by lysS) in R. prowazekii is a class I enzyme with no resemblance to the conventional class II lysyl-tRNA synthetases. Class I type of lysyl-tRNA synthetases have been observed previously in only B.burgdorferi, Pyrococcus woesii, Methanococcus jannaschii and a few other methanogens26.

Regulatory systems

As in other genomes of small parasites, R. prowazekii has a reduced set of regulatory genes. There are a few members of two-component regulatory systems, such as the proteins encoded by barA, envZ, ntrY , ntrX, ompR and phoR. spoT, which is involved in the stringent response, has been identified in B. burgdorferi, Helicobacter pylori and M. genitalium. Only remnants of genes coding for amino-terminal fragments of proteins similar to those encoded by spoT and relA are identifiable in R. prowazekii. No fragments of spoT encoding the carboxy-terminal segments of the protein have been identified in the genome.

Cell division and protein secretion

Proteins involved in detoxification, such as superoxide dismutase, and those involved in thiophen and furan oxidation are present in R. prowazekii . Two genes encoding haemolysins have also been identified, and an R. typhi homologue of tlyC exhibits haemolytic activities when expressed in E. coli (S. Radulovic, J. M. Troyer, B.Noden, S.G.E.A. and A. Azad, unpublished observations).

The data indicate that the basic mechanisms of cell division and secretion in R. prowazekii are similar to those in free-living proteobacteria. There is a common set of bacterial chaperones (encoded by dnaK, dnaJ, hslU, hslV, groEL, groEL, groES and htpG) and genes involved in the secA -dependent secretory system (secABDEFGY, ffH and ftsY). R. prowazekii has a significantly larger set of genes involved in peptide secretion than does M. genitalium.

Membrane-protein analysis

Many studies of R. prowazekii have focused on outer-surface membrane proteins because of their potential importance in bacterial detection and vaccination. The superficial lipopolysaccharide (LPS) molecule is important in the pathogenesis of R. prowazekii. LPS consists of a polysaccharide that is covalently linked to lipid A, the biosynthesis of which is catalysed by products of lpxABCD, all of which are present in the R. prowazekii genome. These genes are clustered in E. coli, but lpxA and lpxD are separate from lpxB and lpxC in R. prowazekii . Three genes involved in the biosynthesis of the 3-deoxy-D-manno-octulosonic acid (KDO) residues reside in the R. prowazekii genome (kdsA, kdsB and kdtA). Only one gene (rfaJ) with a putative function in outer-core biosynthesis has been identified.

We have identified a set of genes involved in the biosynthesis of murein and diaminopimelate and a set involved in the biosynthesis of fatty acids. These includes: fabD, which is involved in the last step of the initiation phase of fatty-acid biosynthesis; four genes involved in the elongation cycle of fatty-acid biosynthesis (fabFGHI); and three genes involved in the first three steps of the synthesis of polar head groups (cdsA, pssA and pgsA). Finally, post-translational processing and addition of lipids to an N-terminal cysteine require the gene products prolipoprotein diacylglycerol transferase (lgt), prolipoprotein signal peptidase ( lspA) and apolipoprotein:phosholipid N -acyl transferase (lnt ). These are found in the genome with several genes involved in the degradation of fatty acids, such as fadA which encodes the 3-ketoacyl-CoA thiolase.

Virulence

The R. prowazekii genome contains several homologues of the VirB gene operon found in Agrobacterium tumefaciens. This gene family encodes proteins that direct the export of the T-DNA–protein complex across the bacterial envelope to the plant nuclei28. R. prowazekii has two homologues of VirB4 and one homologue each of VirB8, VirB9, VirB10, VirB11 and VirD4. The latter five genes are clustered with the gene trbG, which is involved in conjugation in Agrobactrium tumefaciens. Homologues of the single-stranded DNA-binding proteins VirD2 and VirE2 are missing. In Agrobacterium tumefaciens , these proteins are bound to the transferred T-DNA, indicating different functions for the homologues of the VirB genes in R. prowazekii . Indeed, VirB proteins are homologous to components of the E. coli transport system for plasmids, as well as to components of the Pt1 transport machinery in Bordetella pertussis, which exports pertussis toxin28. A set of genes coding for VirB4 and several other VirB proteins has been identified in the cag pathogenicity island of Helicobacter pylori. In this species, the VirB proteins facilitate export of a factor that induces interleukin-8 secretion in gastric epithelial cells28. Thus, R. prowazekii may encode components of a transport system for both conjugal DNA transfer and protein export.

The virulence of Staphylococcus aureus has been correlated with the production of capsular polysaccharides in phagocytic assays and mouse lethality assays29,30. A cluster of ten capsule genes (capA–M) is involved in capsule biosynthesis in S. aureus strain M31. We have identified three R. prowazekii genes with sequence similarities to S. aureus cap genes. Two of these (capD and capM) are separated by ten genes, most of which are unknown genes or genes involved in the biosynthesis of LPS or techoic acid. Thus, R. prowazekii may produce components of a microcapsular layer that is involved in virulence.

Reductive evolution

Genome sequences of organisms enjoying an endosymbiotic lifestyle are at risk. The activities of homologous nuclear genes may render genes of the endosymbiont expendable and as a consequence they become vulnerable to obliteration by mutation. Good candidates for such purged genes in Rickettsia and mitochondria are genes required for amino-acid biosynthesis, nucleoside biosynthesis and anaerobic glycolyis. These and other genes would have been deleted when an ancestral genome first lived in a nucleated cell. Once genes essential to a free-living mode are lost, the endosymbiont becomes an obligate resident of its host.

Likewise, small, bottle-necked populations of bacteria infecting a eukaryotic cell will tend to accumulate deleterious mutations because selection cannot remove them from such clonal populations13. The accumulation of such harmful but non-lethal mutations is referred to as ‘Muller's ratchet’32 or ‘near-neutral evolution’33,34. The consequence of accumulation of these mutations will be the inactivation and eventual deletion of non-essential genes.

The first mutation that inactivates an expendable gene is likely to initiate a sequence of events in which subsequent mutations freely transform it, by degrees, from a pseudogene, to unrecognizable sequence, to small fragments, to extinction. In this sequence, mutations are released from amino-acid-coding constraints. Thus nucleotide substitutions will reflect the mutation bias of the genome. This bias can be estimated roughly by frequencies of third-position bases in the codons. For R. prowazekii, the bias of the third-position bases is 18% G+C rather than the 29% G+C average for the genome. So, as sequences age in R. prowazekii, their composition should gradually approach the low G+C content of third codon positions. Nearly one-quarter of the R. prowazekii genome is composed of non-coding sequences, with a G+C content lower than that of coding sequences (25% G+C compared to 30%; P < 0.001). Thus, much of the non-coding sequence may be remnants of coding sequences that are in the process of being eliminated from the genome.

The gene encoding S -adenosylmethionine synthetase (metK), which catalyses the biosynthesis of S -adenosylmethionine (SAM), illustrates the initiation of this process. The metK sequence in the strain of R. prowazekii studied here has a termination codon within a region of the gene that is otherwise highly conserved among bacterial species35. However, a closely related strain does not have the termination codon. Many other defects, such as termination codons, insertions, and a preponderance of small deletions, have also been observed in the metK genes in several members of the spotted fever group Rickettsia (J.O.A. and S.G.E.A., unpublished observations). This random distribution of lethal mutations among some metK alleles from different Rickettsia species indicates that the gene may have just entered the extinction process. This distribution, and the identification of 11 more pseudogenes for carboxypeptidase (ypwA ), penicillin-binding protein (pbpC), succinyl CoA-transferase (scoB), transposase (tra3), resolvase (pin), conjugative transfer protein (taxB), a hypothetical protein (yfc 1) and four different fragmented open reading frames for (p)ppGpp 3′-pyrophosphohydrolase, indicates that the R. prowazekii genome continues to eliminate genes.

Genome sequences can be purged by a more abrupt mechanism. This consists of intrachromasomal recombination at duplicated sequences, which can result in the deletion of intervening sequences, the loss of a sequence duplication and the rearrangement of flanking sequences. Such a mechanism will account for the presence in R. prowazekii of one, unlinked copy of rrs and rrl, both of which are surrounded by new flanking sequences36. Likewise, R. prowazekii has one tuf gene and one fus gene in atypical clusters that seem to have been created by intrachromosomal recombination between the two tuf genes that are normally found in Gram-negative bacteria37. Indeed, rearranged gene operon structures encoding ribosomal proteins are characteristic of all members of the genus Rickettsia (H. Amiri, C.A. and S.G.E.A., unpublished observations).

Conserved operons that are found in free-living bacteria are often dispersed throughout the Rickettsia genome (see above). The R. prowazekii genome contains an unusually small fraction of repeat sequences (<10% of that observed in free-living bacteria). We suggest that the repeat sequences found in the ancestor to the Rickettsia have been ‘consumed’ by the intrachromosomal-recombination mechanism that generated some of the deletions and rearrangements seen in R. prowazekii. Such intrachromosomal recombinants arise at a substantial rate in bacteria growing in culture, but here they are eliminated from the populations by selection. That such remnants of intrachromosomal recombination are retained in R. prowazekii indicates that purifying selection has been attenuated in this organism.

Mitochondrial affinities

The reduction in genome size in mitochondria and Rickettsia is likely to have occurred independently in the two lineages. Most of the genes supporting mitochondrial activities are nuclear. Many of the 300 proteins encoded in the nucleus of the yeast Saccharomyces cerevisiae but destined for service within the mitochondrion are close homologues of their counterparts in R. prowazekii. Nearly one-quarter of these proteins are required for bioenergetic processes and another one-third of them are required for the expression of the genes encoded in the mitochondrial genome. In total, more than 150 nucleus-encoded mitochondrial proteins share significant sequence homology with R. prowazekii proteins (Fig. 4).

Figure 4: Histogram representation of the similarity of predicted R. prowazekii proteins to yeast proteins targeted to the mitochondria.
figure 3

Only protein pairs with per cent identity values greater than 25% are shown. Numbers in parentheses represent the total number of yeast mitochondrial proteins within each category. The yeast mitochondrial protein sequences have been taken from http://www.proteome.com.

Another group of 58 nucleus-encoded mitochondrial proteins represents components of the mitochondrial transport machinery and regulatory system ( Fig. 4). These include proteins found in the mitochondrial outer membrane and others involved in splicing reactions. Such proteins have probably been secondarily recruited to mitochondria from genomes not necessarily related to that of the α-proteobacterial ancestor.

The mitochondrial genome of the early diverging, freshwater protozoan Reclinomonas americana is more like that of a bacterium than any other mitochondrial genome sequenced so far38. This genome contains 67 protein-coding genes, most of which provide components of genetic processes and the bioenergetic system38. Several gene clusters in this mitochondrial genome are reminiscent of those in bacteria ( Figs 5a, 6a). Most similarities represent retained, ancestral traits present in the common ancestor of bacteria and mitochondria. For example, the genes rplKAJL and rpoBC are identically organized in R. prowazekii and the mitochondrial genome of Reclinomonas americana . Likewise, the genes encoding the S10, spc and the α-ribosomal protein operons are organized similarly in the two genomes. The immediate proximity of these two clusters in the Reclinomonas americana mitochondrial DNA is reminiscent of the arrangement in free-living bacteria, whereas the physical separation of the two clusters in the R. prowazekii genome is atypical. A further rearrangement event is indicated by the fact that the rpsLrpsGfus cluster is located upstream of the rplKAJLrpoBC cluster in R. prowazekii, rather than downstream as it is in the Reclinomonas americana mtDNA. Phylogenetic reconstructions based on ribosomal proteins within each of these two clusters indicate that there is a close evolutionary relationship between R. prowazekii and mitochondria ( Fig. 5b).

Figure 5: The organization and phylogenetic relationships of gene encoding ribosomal protein from R.prowazekii and the mitochondrial genome of Reclinomonas americana.
figure 4

a, The organization of ribosomal-protein genes. The S10, spc and α-operons are organized similarly in these two genomes, except that several ribosomal-protein genes38 have been deleted from the mitochondrial genome of Reclinomonas americana. b, The phylogenetic relationships of mitochondria and bacteria were derived from the combined amino-acid sequences of ribosomal proteins S2, S3, S7, S10, S11, S12, S13, S14, S19, L5, L6 and L16. Neighbour-joining and maximum-parsimony methods gave identical topologies. Branch lengths are proportional to those reconstructed by using the neighbour-joining method. Values at nodes are bootstrap values indicating the degree of support for individual clusters under each method (neighbour-joining, maximum parsimony). Only bootstrap values >90% are shown.

Figure 6: The organization and phylogenetic relationships of genes involved in ATP synthesis from R.prowazekii and the mitochondrial genome of Reclinomonas americana.
figure 5

a, The organization of bioenergetic genes. b, The phylogenetic relationships of mitochondria and bacteria were derived from the combined amino-acid sequences of NADH dehydrogenase I chains A, J, K, L, M and N, which are encoded by the genes nuoA, J, K, L, M, N. These genes are highlighted by asterisks in a. Neighbour-joining and maximum-parsimony methods gave identical topologies. Branch lengths are proportional to those reconstructed using the neighbour-joining method. Values at nodes are bootstrap values indicating the degree of support for individual clusters under each method (neighbour-joining, maximum parsimony). Only bootstrap values >90% are shown.

Mitochondria and R. prowazekii have a similar repertoire of proteins involved in ATP production and transport, including genes encoding components of the TCA cycle, the respiratory-chain complexes, the ATP-synthase complexes and the ATP/ADP translocases. There are some similarities in the gene orders of some functional clusters (Fig. 6a). There are also some rearrangements of clusters that are specific to Rickettsia. One example is the inversion of segments corresponding to nuoJKLM and nuoGHI. Another is the scattered displacement of genes involved in the biogenesis of cytochrome c. Nevertheless, phylogenetic reconstructions based on components of the NADH dehydrogenase complexes indicate that there is a close evolutionary relationship between R. prowazekii and mitochondria (Fig. 6b).

We have identified as many as five genes coding for ATP/ADP transporters, all of which are expressed (R.M.P. et al., unpublished observations). The Rickettsia ATP/ADP translocases are monomers with 12 transmembrane regions each, whereas the mitochondrial translocates are dimers with six transmembrane regions per dimer. We found no relationship between the primary structures of the mitochondrial and Rickettsia ATP/ADP translocases, indicating that these transport systems may have originated independently.

The study of the R. prowazekii genome sequence supports the idea that aerobic respiration in eukaryotes originated from an ancestor of the Rickettsia, as indicated previously by phylogenetic reconstructions based on the rRNA gene sequences7,9. Phylogenetic analyses of the petB and coxA genes indicate that the respiration systems of Rickettsia and mitochondria diverged 1,500–2,000 million years ago10, shortly after the amount of oxygen in the atmosphere began to increase. The finding that the ATP/ADP translocases in R. prowazekii and mitochondria are of different evolutionary origin is problematic (R.M.P. et al., unpublished observations). Free-living bacteria do not seem to have homologues of ATP/ADP translocases, which are found only in organelles and in two obligate intracellular parasites, Rickettsia and Chlamydia. Thus it is not known whether the original endosymbiont was capable of efficient exchange of adenosine nucleotides with its host cell. More detailed comparative analysis of the genomes of α-proteobacteria may refine our understanding of the origins of mitochondria.

Methods

Genome sequencing. We prepared genomic DNA from the Madrid E strain of R. prowazekii, which was originally isolated in Madrid from a patient who died in 1941 with epidemic typhus. We propagated R. prowazekii in the yolk sac of embryonated hen eggs and purified DNA according to standard procedures39. We sequenced the R. prowazekii genome by a whole-genome shotgun approach in combination with shotgun sequencing of a selected set of clones from a cosmid library (A.Z. et al., unpublished observations). Genomic and cosmid DNA was sheared by nebulization to an average size of 2 kb. The random fragments were cloned into a modified M13 vector using the double adaptor method40. We collected 19,078 sequence reads during the random sequencing phase using Applied Biosystems 377 DNA sequencers (Perkin-Elmer).

The sequences were assembled and the consensus sequence was edited using the STADEN program41. We verified the structure of the assembled sequence by end-sequencing of 3-kb-insert λ Zap II clones36, 10-kb λ clones and 30-kb cosmid clones. More than 97% of the genome was covered by clones from the three different libraries (A.Z. et al., unpublished observations). Gaps between contigs were closed by direct sequencing of clones from the three libraries or of polymerase chain reaction (PCR) products. The final four gaps were closed by direct sequencing of PCR products generated with the Long Range PCR system (Gene Amp). Regions of ambiguity were identified by visual inspection of the assembly and resequenced. The final assembly contains 20,000 sequences. The genome sequence has eightfold coverage on average and no single region has less than twofold coverage. We estimate the overall error frequency to be <1 × 10−5.

Informatics. Sequence analysis and annotation was managed by CapDB (T.S.-P. et al., unpublished observations). We identified open reading frames of more than 50 codons as genes on the basis of their characteristic patterns in nucleotide-frequency statistics14 using BioWish42. The identified genes were analysed using the program BLASTX43 to search for sequence similarities in EMBL, TREMBL, SwissProt and in-house databases. We identified tRNA genes with the program tRNA scan-SE44. Remaining frameshifts were considered to be authentic and annotated as pseudogenes. Families of paralogues were constructed using BLAST to search for sequence similarities within the R. prowazekii genome. Multiple alignments and phylogenetic trees for genes with significant sequence similarities to genes in the public databases were constructed automatically using CLUSTAL-W45, Phylo_win46 and GRS47. The final annotation was based on manual inspection of the phylogenetic placement of R. prowazekii in the resulting gene trees.