Background

TheRickettsia genus includes 25 validated species, 17 of which are proven human pathogens. Among these, the pathogenicity varies greatly, from the highly virulentR. prowazekii, which causes epidemic typhus and kills its arthropod host, to the mild pathogenR. africae, the agent of African tick-bite fever, which does not affect the fitness of its tick vector.

Results

We evaluated the clonality ofR. africae in 70 patients and 155 ticks, and determined its genome sequence, which comprises a circular chromosome of 1,278,540 bp including atra operon and an unstable 12,377-bp plasmid. To study the genetic characteristics associated with virulence, we compared this species toR. prowazekii,R. rickettsii andR. conorii.R. africae andR. prowazekii have, respectively, the less and most decayed genomes. Eighteen genes are present only inR. africae including one with a putative protease domain upregulated at 37°C.

Conclusion

Based on these data, we speculate that a loss of regulatory genes causes an increase of virulence of rickettsial species in ticks and mammals. We also speculate that inRickettsia species virulence is mostly associated with gene loss.

The genome sequence was deposited in GenBank under accession number [GenBank: NZ_AAUY01000001].

Rickettsiae are obligate intracellular Gram-negative bacteria mostly associated to arthropods, some of which causing mild to severe diseases in humans. Pathogenic species are classified into two groups based on phylogenetic analyses [1]. The typhus group (TG) includes twoRickettsia prowazekii (R. prowazekii) andR. typhi, and the spotted fever group (SFG) includes 15 pathogenic species and numerous species of unknown pathogenicity [2,3]. Two additional validated species,R. bellii andR. canadensis, and a variety of unvalidated species from insects or leeches are organized into the most outer outgroups of the genusRickettsia [3–5]. The relatively low rate of lateral gene transfer, the continuous gene loss and the colinearity of most of their genomes makeRickettsia species an outstanding model for comparative genomics [4,6,7]. Indeed, genome reduction [8] paradoxically results in higher virulence inR. prowazekii.

The pathogenic mechanisms of rickettsiae are unclear. Within ticks, rickettsiae remain quiescent during the starvation of their vector but undergo a reversion to the virulent state, termed reactivation, following incubation at 37°C or blood meal [9]. This phenomenon is marked inR. rickettsii by morphological changes in the microcapsular and slime layers [9]. The precise molecular mechanisms of this change, however, are only poorly understood. During human infection, attachment to and invasion of host cells were suggested to involve the outer membrane proteins rOmpA and rOmpB and the adhesins Adr1 and Adr2 [10,11]. A phospholipase D activity was proposed to play a role in escape from phagosomes [8,12], and intracellular motility was demonstrated to rely on actin polymerization [13,14]. None of these factors nor the presence of a type IV secretion system [15], however, explain the virulence differences observed amongRickettsia species [6].

Over the last ten years,R. africae has emerged as the causative agent of African tick-bite fever [2], the most common SFG rickettsiosis both in terms of seroprevalence [16] and incidence [17–20]. Such an epidemiologic success is due to various factors, including the increase of tourism to wildlife parks in sub-Saharan Africa, the attack host-seeking behavior of its vector ticks,Amblyomma sp., and the elevated prevalence ofR. africae in these ticks, with infection rates of up to 100% [21]. In addition, the bacterium has been identified in other areas with warm climates, such as the West Indies, where it was found in Guadeloupe, Martinique, St Kitts and Nevis, and Antigua islands [2]. Such a distribution, as well as the presence ofR. africae in Reunion island, is likely to result from the transfer from Africa of cattle bearing infected ticks [2]. Tick-associated rickettsiae may infect ticks feeding on infected hosts or may be passed from one generation to the next transovarially.R. africae is transmitted transovarially and appears to be the most successful rickettsia in its adaptation to its vector tick, as the prevalence of tick infection is higher than that of any other rickettsia [22]. In addition, infection does not appear to alter tick fitness (P. Parola, unpublished data). These data highlight the fact thatR. africae is an extremely successful and fit bacterium.

By comparison withR. conorii, the second most prevalent SFG rickettsia in Africa, whose genome has previously been sequenced [23],R. africae exhibits a higher prevalence in ticks [2], a lower virulence in humans [17], and a greater genetic homogeneity [24]. The genetic factors underlying these characteristics are, however, unknown. We assumed that theR. africae genome sequence might help understand the characteristics of this species and the genetic mechanisms associated with the difference in virulence. Here, we present the sequence of theR. africae genome and additional data that suggest that this species has emerged recently. In support of this hypothesis, we show thatR. africae is a clonal population. We also present data that support the assumption that rickettsial virulence increases following gene inactivation.

General Features of the Genome

The genome ofR. africae consists of two replicons: a circular chromosome of 1,278,540 base pairs (bp) (Figure1) and a 12,377 bp circular plasmid (Table1, Figure2[25,26]). We acknowledge the fact that the ESF-5 strain, first isolated in 1966 [27], may have undergone loss or rearrangement of plasmid or chromosomal genes during multiple passages in cell culture. Sequences were deposited in GenBank under accession number [GenBank: NZ_AAUY01000001]. The chromosome has a G + C content of 32.4%, in the range of other SFG rickettsial genomes (32.3 – 32.5%), whereas the plasmid has a G + C content of 33.4%, similar to those ofR. felis (33.2 and 33.6%) [28] but higher than that ofR. massiliae plasmids (31.4%). The predicted total complement of 1,271 open reading frames (ORFs), 1,260 chromosomal (78.26% coding sequence), and 11 plasmidic (81.3% coding sequence) ORFs [see Additional file1], is in the range of genomes from SFG rickettsiae with the exception ofR. felis, which exhibits a larger genome (Table1). Of these, 1,117 (87.9%) exhibited homologs in the non-redundant database, and 1,024 (80.5%) were assigned putative functions [see Additional file2]. Overall, the 1,260 chromosomal ORFs encoded 1,112 protein-coding genes, with 87 of these being split into 2 to 10 ORFs by the presence of one to several stop codons. By comparison with other SFG genomes,R. africae had fewer split genes than any other species with the exception ofR. felis (Table1). In addition,R. africae exhibited a single rRNA operon, with non-contiguous 16S and 23S rRNA genes as in other rickettsial genomes, 33 tRNAs and another three RNAs. TheR. africae chromosome exhibited an almost perfect colinearity with theR. conorii genome [23], with the exception of a 88,459-bp inversion [see Additional file3]. At both extremities of the inversion, there were repeats of theRickettsia palindromic element – 6 (RPE-6) familly. In this inverted fragment,R. africae exhibited 20 ORFs and 10 RPEs that were absent fromR. conorii. Among these 20 ORFs, a cluster of 11 consecutive ORFs had orthologs in the 3'-extremity of the Tra cluster previously identified in theR. massiliae genome [29]. These 11 ORFs includedtraD F (ORF0650), a transposase (ORF0651),spoT 15 (ORF0652), a splitspoT 13 (ORF0653/ORF0654), a splitspoT 6 (ORF0655/ORF0656), a split signal transduction histidine kinase (ORF0657/ORF0658),dam 2, a site-specific DNA adenine methylase (ORF0659), and ORF0660 of unknown function (Figure3). In addition to the orthologs inR. massiliae, these genes had orthologs in similar clusters inR. felis,R. bellii,R. canadensis andO. tsutsugamushi but were absent from all other species. As inR. massiliae,R. bellii andR. canadensis, theR. africae cluster was bounded at its 3'-end by a tRNA-Val, but, in contrast with these three species, neither an integrase with itsattI site nor a tRNA-Val fragment marker of integration was present at the 5' end (Figure3). The presence of a similar gene cluster inserted at the same position in severalRickettsia species, with a GC content different from that of the genome (29.78%vs 32.4%, respectively, inR. africae) suggests that it was acquired horizontally from a common ancestor and then transmitted vertically. InR. africae, anattC site, specific to integron-inserted gene cassettes, located at the 3'-end (coordinates: 687890–688018) of thespoT 15 gene (ORF652), supports the role of integration in the insertion of this gene cluster.AttC sites were also identified inR. massiliae (coordinates: 743029–743145),R. felis (coordinates: 407889–408017), andR. bellii (coordinates 468143–468211). Nevertheless, the presence of transposases in all species and the fact that, inR. felis, nine of these genes are located in the pRF plasmid support the role of several genetic mechanisms at the origin of this cluster, possibly involving plasmids, integrons and transposons. In comparison with other species containing this gene cluster,R. africae had the smallest number of genes. In particular, it lacked most of the Tra cluster, with the exception oftraDF, but retained threespoT genes, including two degraded to pseudogenes. InR. bellii andR. massiliae,tra genes were described as encoding components of a type IV secretion system (T4SS) for conjugal DNA transfer [15,29]. In terms of gene content, theR. africae cluster was more similar to those ofR. felis andR. canadensis, with the loss of the Tra cluster, the conservation ofspoT genes and the presence of pseudogenes, than to those ofR. massiliae andR. bellii, in which the Tra cluster was intact butspoT genes were partially degraded. Such findings suggest that species-specific evolution of this gene cluster occurred, which likely resulted from gene excisions inR. africae,R. felis andR. canadensis, or gene expansion by transposase duplication inR. massiliae.

Circular representation of the genomes ofR. africae, R. conorii, andR. prowazekiibased on data from GenBank entries [GenBank: NZ_AAUY01000001], [GenBank: NC_003103]and [GenBank: NC_000963],respectively. Protein coding genes common to all species are in blue; genes common toR. africae andR. conorii are in green; genes common toR. africae andR. prowazekii are in red; genes common toR. conorii andR. prowazekii are in pink and specific genes in each genomes are in yellow. Common genes are identified using best BLAST match. The region of rearrangement of the genome betweenR. africae andR. conorii is colored in purple; the regions of rearrangment betweenR. prowazekii andR. conorii are colored in orange, light green, yellow and light blue. Also represented are transfer RNAs (red arrows), ribosomal RNAs (dark arrows) and other RNAs.

Full size image

Circular representation ofRickettsiaplasmids. A) The pRA plasmid: circles indicate (from the outside to the inside, on the reverse and forward strands) the GC skew, GC content, and ORFs; B)Rickettsia plasmids sequenced to date: circles indicate (from the outside to the inside, on the reverse and forward strands)R. felis pRF plasmid (red),R. felis pRFδ plasmid (blue),R. monacensis pRM plasmid (green),R. massiliae pRMa plasmid (grey), andR. africae pRA plasmid (black).

Full size image

Presence of thetragene cluster inRickettsiaspecies. Raf:R. africae; Rma:R. massiliae; Rfe:R. felis; Rbe:R. bellii; Rco:R. conorii; Rsi:R. sibirica; Rri:R. rickettsii; Rak:R. akari; Rpr:R. prowazekii; Rty:R. typhi.

Full size image

In addition to thetraD_Fgene described above, theR. africae chromosome retained many of the components of the type IV secretion system (T4SS) involved in both DNA transfer and effector translocation in other bacteria [30], includingvirB1,virB2 (ORF0232),virB3 (ORF0128),virB4 (ORF0129, ORF1109),virB6 (ORF0130, ORF0131, ORF0132, ORF0133, ORF0134, ORF0135),virB8 (ORF0359, ORF0361),virB9 (ORF0358, ORF0362),virB10 (ORF0363),virB11(ORF0364), andvirD4 (ORF0365). In addition,R. africae possessed atraX (ORF0816) and a splitfimD (ORF0592/ORF0593/ORF0594) gene but lacked other Tra cluster genes found inR. massiliae,R. felis,R. bellii andO. tsutsugamushi, such astraC andtraG_F[15,28,29,31]. Therefore, the Tra cluster was mostly eliminated from theR. africae, and, following a "use it or lose it" scheme, this species probably did not need atra gene-linked conjugation system. In addition, the pRA plasmid did not contain genes encoding proteins involved in conjugation.

Six transposase-encoding genes were identified in the chromosome, including one split into two ORFs (ORF0955/ORF0956) and one present as a remnant and two in the pRA plasmid, including one present as a fragment. This contrasts with the large expansion of transposases caused by gene duplications previously detected inR. felis andR. bellii [15,28].

Common rickettsial gene set and phylogeny

When compared to eight other available rickettsial genomes, a total of 645 genes and 39 RNA-encoding genes ofR. africae had orthologs in all genomes. In addition, another 32R. africae genes had orthologs only in SFG rickettsiae and were either absent or remnant in TG rickettsiae. Consequently, we identified 645 genes as constituting the core gene set of all available rickettsial genomes and 700 ORFs as the core gene set of SFG rickettsiae. Following concatenation of the 645 core genes, a reliable phylogenetic organization (Figure4) was obtained using three analysis methods that was consistent with previous phylogenetic studies ofRickettsia species [4,32–36].

Phylogenetic tree inferred from the comparison of 645 concatenatedRickettsiacore protein-coding genes. Similar organizations were obtained using both the maximum parsimony and neighbor joining methods. Bootstrap values are indicated at branch nodes.

Full size image

In comparison with otherRickettsia genomes,R. africae had 242, 238 and 69 fewer genes thanR. bellii,R. felis andR massiliae, respectively, but 279, 260, 52, 23, 17, and 15 more genes thanR. typhi,R. prowazekii,R. akari, R. rickettsii,R. sibirica, andR. conorii, respectively. When comparing the numbers of degraded genes (split + remnants),R. africae, with 127 degraded genes, had a significantly less degraded genome (P < 10^-2) than that of other spotted fever group rickettsiae includingR. akari (176),R. conorii (196),R. massiliae (212),R. rickettsii (198) andR. sibirica (199) (Table1). It had, however, significantly more degraded genes thanR. felis (86,P < 10^-2).

Transcription of genes conserved inR. africae but absent from highly pathogenic species

R. africae had 18 intact genes that were either absent or degraded in all three virulent speciesR. conorii,R. rickettsii andR. prowazekii. Of these, 12 encoded proteins of unknown functions (raf_ORF0036, raf_ORF0064, raf_ORF0391, raf_ORF0412, raf_ORF0414, raf_ORF0415, raf_ORF0445, raf_ORF0660, raf_ORF0758, raf_ORF0793, raf_ORF0876, and raf_ORF0884) (Figure5) [see Additional file4]. The remaining six genes encoded a plasmid maintenance system antidote protein (raf_ORF0424), thespoT15 gene (raf_ORF0652), a site-specific DNA adenine methylase (Dam2) (raf_ORF0659), an ankyrin repeat (raf_ORF0782), a putative integral membrane protein (raf_ORF0973), and a protein (RIG1002) exhibiting a high degree of amino acid sequence identity (>50%) with proteins of γ-proteobacteria classified within the COG3943 as putative virulence proteins. When investigating the transcription of these 18 genes inR. africae grown at 28, 32 and 37°C, we observed a significantly higher transcription level at 37°C than at lower temperatures for two genes, raf_ORF414 and raf_ORF660. The former gene contained a putative protease domain site, but the latter had no known function.

Schematic representation of the genes conserved inR. africaebut lost by highly pathogenic rickettsiae. Genes highlighted in yellow are upregulated at 37°C. The state of a gene is represented by a small box colored in green (full-length), blue (pseudogene), red (fragment), orange (remnant) or black (absent). Gene numbers are indicated in the left column.

Full size image

TheR. africae plasmid

TheR. africae plasmid (Figure2) is a new example of a plasmid inRickettsia species, following those inR. felis [28],R. massiliae [29],R. monacensis [37],R. helvetica, R. peacockii, R. amblyommii andR. hoogstraalii [38]. This plasmid, named pRA, is smaller (12,377 bp) than those ofR. felis (62,829 bp and 39,263 bp long, for pRF and pRFδ, respectively),R. monacensis (23,486 bp), andR. massiliae (15,286 bp). The pRA plasmid is predicted to contain 11 genes, 6 of which (54%) have homologs in public databases and are associated with functional attributes. These six genes encode for a chromosomal replication initiator DnaA-like protein (ORF1260), a site-specific recombinase (ORF1262), two contiguous transposases exhibiting 100% sequence similarity (ORF1263 and 1264) but with one (ORF1263) shorter than the other, the auto-transporter protein SCA12 (ORF1268), and a ParA-like plasmid stability protein (ORF1270). Five genes (ORFs 1260, 1263, 1264, 1269 and 1270) have orthologs in theR. massiliae plasmid, six have orthologs in theR. felis plasmids (ORF1260, 1263, 1264, 1268, 1269 and 1270), and three have orthologs in theR. monacensis plasmid (ORF1260, ORF1268, and ORF1270). The presence of two genes (ORF1260 and 1270) conserved in plasmids from four species suggests that these plasmids have a common origin. The presence of two almost identical successive transposases inR. africae matching a single gene inR. massiliae andR. felis suggests a duplication event in the former species. The pRA plasmid lacks heat shock protein-encoding genes found in other rickettsial plasmids. In contrast, ORF1262, a site-specific recombinase, is absent from other species. Its closest phylogenetic neighbour is a site-specific recombinase fromMagnetospirillum magnetotacticum, a high G-C content α-proteobacterium living in aquatic environments [39]. The sca12 gene (ORF) found intact inR. africae pRA was absent from theR. massiliae andR. monacensis plasmids and present but fragmented withinR. felis pRF, but it was absent from pRFδ as well all otherRickettsia species.

As outlined by Baldridgeet al. [38], the plasmid content of aRickettsia species may vary according to the passage history of rickettsial strains. When estimating the prevalence of the plasmid amongR. africae strains, we detected it in the 22 tested isolates from South Africa and in the 48 eschar biopsies from patients with ATBF contracted in the same country and in 20/32R. africae- positiveAmblyomma ticks [see Additional files5 and6]. Therefore, it appears from these results that, depending on the geographic location, the plasmid ofR. africae may be unstable. Whether the plasmid has been lost by PCR-negative strains or cannot be amplified with the primers we used is as yet unknown. Such inter-strain differences in plasmid content were also observed inR. felis (Unpublished data).

Stress response

Rickettsiae live intracellularly in both arthropod and mammal hosts. This implies that periods of tick starvation and feeding cause bacterial dormancy and multiplication following reactivation [40]. As a consequence, and despite their obligate intracellular location, rickettsiae may face, and thus have to adapt to, highly variable and extreme environmental conditions. Known as the stringent response, this bacterial adaptation to nutritional stress has been described to be mediated by the accumulation of guanosine nucleotides pppGpp (guanosine 3'-diphosphate 5'-triphosphate) and ppGpp (guanosine 3'-diphosphate 5'-diphosphate) [41]. Accordingly, the transcriptional analysis ofR. conorii exposed to a nutrient deprivation was characterized by the up-regulation ofgmk and of genes from thespoT family, suggesting a role for these nucleotides as effectors of the stringent response [42,43]. TheR. africae genome exhibited eightspoT genes phylogenetically classified within two major clades [see Additional file7]. The largest clade includedspoT genes with hydrolase activity (1–10, 14, 15, 17–21), while the second included those with a synthetase domain. With eight genes,R. africae had morespoT genes thanR. rickettsii (5 genes),R. conorii (4),R. sibirica (4),R. akari (7),R. canadensis (5),R. typhi (4) andR. prowazekii (1) but fewer genes thanR. felis (14) andR. bellii (10) [see Additional file8]. Altogether, our data suggest thatR. africae is more regulated than more pathogenic species.

Infection of mammal hosts

TheR. africae genome encoded rOmpA (or Sca0) and rOmpB (or Sca5), two surface-exposed and immunodominant proteins belonging to the paralogous "surface cell antigen" (SCA) family and known inRickettsia species to be responsible for antigenic differences between species [1] and to elicit an immune response in patients [44]. Experimental studies suggested that these two auto-transporter proteins could function as adhesins [10,11,45,46]. In addition, another eight SCA-encoding genes were found in the genome. These 10 genes were represented by 22 ORFs due to partial degradation of some of the paralogs [see Additional file8]. Among the 17 SCA-encoding genes detected inRickettsia species [47],R. africae had similar sets of conserved (sca 0 – 2, 4 and 5), degraded (sca 3, 8 – 10 and 13) and absent (sca 6, 7, 11, 14 – 17)sca genes asR. conorii andR. rickettsii. In addition to these 10 SCA-encoding genes,R. africae exhibited a degradedsca 9 gene and a completesca 12 gene carried by the pRA plasmid, only shared withR. felis, where it was also found partially degraded on the pRF plasmid. Thesca 12 genes from both species were grouped into a distinct cluster close to thesca 1, 2 and 6 genes [see Additional file9]. This result further supports a common origin of the pRA and pRF plasmids.

A proteomic approach recently allowed the identification of two paralogous proteins encoded by the genes RC1281-RC1282 and RP827-RP828, as putative adhesins Adr1 and Adr2. These proteins may be key actors for entry and infection in bothR. conorii andR. prowazekii [11]. Both proteins are ubiquitously present within theRickettsia genus [4]. Their presence within theR. africae genome (ORF1174 + ORF1175) [see Additional file10] reinforces their suspected key role in rickettsial life.

Bothpld andtlyC, encoding phospholipase D [8] and hemolysin C [12], respectively, which play a role in phagosomal escape [13,48], were conserved in theR. africae genome (ORF1161 and ORF1039, respectively). This bacterium also exhibited genes encoding other proteins with membranolytic activity, includingtlyA (hemolysin A) andpat 1 (patatin-like phospholipase) [12,49]. As expected, the genome ofR. africae has arickA gene (ORF0824) orthologous to all rickettsialrickA genes and coding a protein activating the Arp2/3 complex, whose nucleation triggers actin polymerisation [50] [see Additional file11]. The Rick A protein inR. africae is slightly different from those of other species, with a phenylalanine instead of a serine within the G-actin-binding site, an ENNIP [PS] motif repeated twice instead of four times in the central proline-rich region of the protein [see Additional file11], and an aspartate and an isoleucine instead of an asparagine and an alanine or valine, respectively, in the carboxy-terminal region. Despite these differences, the RickA protein ofR. africae appeared to be functional as demonstrated by its ability to polymerize actin and multiply intranuclearly (Figure6).

Intracellular motility ofR. africae. A) Actin tail formation byR. africae. L-929 cells were infected withR. africae, fixed and stained with fluorescent phalloidin (green) and a polyclonal antibody againstR. africae and visualized using anti-rabbit-Alexa549 as a secondary antibody (red). The white arrows show actin tails. B)R. africae in the cytoplasm and nucleus of L-929 cells. C = cytoplasm; black arrow = nucleus; white arrows =R. africae bacilli. Transmission electron microscopy. Scale bar = 5 μm.

Full size image

Sixteenvir gene paralogs were found in theR. africae genome. Virulence genes of thevir family belong to the type IV secretion machinery, a system that allows the delivery of virulence factors from bacterial and eukaryotic host membranes to the cytoplasm of the host cell [51]. All 16 genes were found to be intact and common to allRickettsia genomes with the exception ofvirB 6-2 inR. africae andvirB 6-5 inR. massiliae [see Additional file8]. In both species, these genes were split into two ORFs. Phylogenetic analysis of thevirB 6-2 gene distinguished clearly the SFG and TG and showed that theR. africae VirB6-2 protein is phylogenetically closer to that ofR. sibirica [see Additional file12].

Clonality ofR. africae

Of the 155Amblyomma ticks tested, 139 (89.6%) were PCR-positive forR. africae [see Additional file5]. Therefore, infection rates ofAmblyomma ticks withR. africae may be higher than previously described [21,22,52], which suggests an extreme fitness of this rickettsia for its vector. In addition, such infection rates are the highest amongRickettsia species [see Additional file13].

Using MST, PCR products of the expected sizes were obtained from thedksA-xerC,mppA-purC andrpmE- tRNA^fMet intergenic spacers from all tested specimens. Sequences obtained from these amplicons were in all cases identical to those previously obtained forR. africae [GenBank: DQ008280], [GenBank: DQ008301], and [GenBank: DQ008246], for thedksA-xerC,mppA-purC andrpmE- tRNA^fMet spacers, respectively). This is the first rickettsia demonstrated to be clonal. Other testedRickettsia species, includingR. conorii (31 MST genotypes out of 39 strains tested [53]),R. massiliae (2/7 [24]),R. sibirica (3/3 [24]), andR. felis (3/6 [24]), were significantly more genetically variable thanR. africae (p < 10^-2 in all cases).

Using a comparative study of rickettsial genomes, we found that virulence inRickettsia species is not correlated with acquisition of foreign DNA but may rather result from a reduction in regulation due to genome decay [6,23]. Comparative genomics sheds light on a much wider spectrum of virulence acquisition mechanisms in bacteria than initially thought [54]. Based on the examples of enterobacteria and staphylococci, gain in pathogenicity in bacteria was mainly thought to result from horizontal gene transfer, either directly or through mobile genetic elements [55,56]. However, a recent study ofRickettsia species associated with arthropods, insects, leeches and protists clearly demonstrated that horizontal gene transfer was a rare event within this genus [5]. In addition, genomic studies demonstrated that rickettsiae are undergoing genome decay, affecting in priority horizontally-acquired genes [57], and that there is no association between pathogenicity and acquisition of virulence markers [6]. In fact, the genome of the most virulent species,R. prowazekii [58], is a subset of the less pathogenic speciesR. conorii [23], thus highlighting a paradoxical relationship between smaller genome size and higher pathogenicity. Careful comparison of theR. prowazekii andR. typhi genomes also demonstrated that the former species, more pathogenic than the latter, had a more decayed genome despite a 12-kb insertion that likely resulted from a single genetic event [59].

When investigating the genomic characteristics associated with the milder virulence ofR. africae, we first ruled out a potential role of the plasmid by the fact that it is unstable in this species. Then, we compared the gene contents ofR. africae withR. conorii,R. rickettsii, andR. prowazekii, which exhibit a higher pathogenicity in humans and their arthropod hosts. We observed thatR. africae showed no gene loss but had 18 genes fully conserved that were either absent or degraded in the other species (Figure5). We speculated that, becauseR. africae had more intact genes than more virulent species, some of these genes may be involved in maintaining a low virulence level. Such a behavior may not be unique to rickettsiae. It was found that gene knockout resulted in increased virulence inMycoplasma,Streptococcus pyogenes, andVibrio cholerae [60–62]. InM. ulcerans, genome reduction was also linked to gain in virulence [63]. It emerges as a concept that virulence may be increased by gene loss [54]. We assume that a similar phenomenon may happen in rickettsiae, and that inactivation of some genes may deregulate the control of bacterial multiplication, in particular during the reactivation phenomenon following warming, thus enhancing pathogenesis.

Among the 18 putative candidate genes unique toR. africae, we identified only two genes (raf_ORF414 and raf_ORF660) that were significantly more transcribed at 37°C than at lower temperatures. Of these, one (raf_ORF414) encoded a protein that had a putative protease domain. A protease was previously shown inVibrio cholerae to be a virulence repressor [60]. However, whether this differentially-transcribed protease plays a role in virulence repression inR. africae is as yet unknown. In contrast, thespoT 15 gene (raf_ORF652) unique toR. africae was not upregulated, and this species retained another twospoT pseudogenes (raf_ORF653–654 and raf_ORF655–656) that were completely lost by other species.SpoT genes, effectors of the stringent response, were shown to play a major role in adaptation to stress inR. conorii, in particular when subjected to abrupt temperature variations similar to those occurring during a tick blood meal [42].R. africae, however, has morespoT genes thanR. conorii orR. rickettsii and does not show any modification of expression of its specificspoT 15 gene during temperature variations. We speculate that higher regulation ability inR. africae is linked to lower pathogenicity.

In addition, when compared to other tick-borneRickettsia species,R. africae exhibited several unique characteristics. First, this species is extremely successful and fit: it is highly adapted and harmless to its tick host, being efficiently transmitted both transtadially and transovarially inAmblyomma sp. ticks, which consequently act as efficient reservoirs [64]. In contrast,R. rickettsii [65,66] andR. conorii [67] have a negative effect on their tick vectors in experimental models. As a result, the prevalence ofR. africae in its host ticks is higher than that of most other rickettsiae. Similarly,R. africae is less pathogenic for humans than other SFG species such asR. conorii andR. rickettsii, in particular because the infection is never lethal [17]. This observation was later supported by the demonstration that inoculation eschars in ATBF were histologically different from those in MSF [68]. In particular, in contrast with other SFG rickettsioses where eschars are characterized by perivascular infiltration of T cells and macrophages, with some B lymphocytes and few polymorphonuclears, the vasculitis in ATBF is made of a large infiltrate of neutrophils causing an extensive cutaneous inflammation and necrosis [see Additional file14] [68]. Such a local reaction, in addition to the fewR. africae cells detected in eschars [68], suggests that the bacterium replicated poorly in human tissues. Second,R. africae has significantly fewer degraded genes than other SFG species (p < 10^-2), exceptR. felis. Specifically, this characteristic suggests thatR. africae is undergoing a slower degradation process than other rickettsiae. Third, the identification of a single MST genotype among 102 strains suggested thatR. africae was clonal [24,69]. This contrasted with the variable plasmid content of this species. Originally thought to be absent inRickettsia species, plasmids have been detected in eight species to date [28,29,37,38], and their plasmid content may exhibit intraspecies variability. InR. felis, two plasmid forms have been sequenced [28], and Baldridgeet al. found two plasmids in bothR. peacockii andR. amblyommii [38]. In addition, these authors showed thatR. peacockii lost its plasmids during long-term serial passages in cell culture [38]. InR. africae, the pRA plasmid may also be unstable, as shown by the absence of plasmid detection in 12/32Amblyomma ticks tested. This plasmid encodes 11 ORFs, two of which are common toR. felis,R. massiliae andR. monacensis plasmids [see Additional file1], which strongly suggests a common source for these mobile elements. We suspect that rickettsial plasmids and Tra clusters are vertically inherited but are apparently unstable and are currently degrading.

Based on its genome and lifestyle, we suspect that the clonalR. africae is more regulated and more specifically adapted to its host and warm environment than other tick-associated rickettsiae. We speculate that losing this regulation, as observed in several intracellular pathogens, is a critical cause of virulence [6]. Further transcriptomic analysis ofR. africae and otherRickettsia species grown at various temperatures is currently ongoing to identify putative other candidate genes involved in stress response.

Genome Sequencing

Bacterial purification and DNA extraction

In this study, we usedR. africae ESF-5 strain, CSUR R15 (Collection de souches de l'Unité des Rickettsies, Marseille, France), which was isolated in anAmblyomma variegatum tick collected from cattle in the Shulu province of Ethiopia in 1966 [27].R. africae was cultivated in Vero cells growing in MEM with 4% fetal bovine serum supplemented with 5 mM L-glutamine. Bacterial purification, DNA extraction and pulsed-filed gel electrophoresis were performed as described in Additional file15 [see Additional file15].

Shotgun sequencing ofR. africae genome

Three shotgun genomic libraries were made by mechanical shearing of the DNA using a Hydroshear device (GeneMachine, San Carlos, CA, USA). Sequence blunt ends, to which the BstXI adaptator was linked, were obtained using the T4 DNA polymerase (New England Biolabs). Fragments of 3, 5, and 10 kb were separated on a preparative agarose gel (FMC, Rockland, ME, USA), extracted using the Qiaquick kit (Qiagen, Hilden, Germany), and ligated into a high copy-number vector pCDNA2.1 (Invitrogen, Carlsbad, CA, USA) for the two smaller inserts and into the low copy-number vector pCNS [28] for the largest inserts. Further details are available in Additional file15 [see Additional file15].

Annotation

We predicted protein-coding genes (ORFs) using SelfID as previously described [15]. Functional assignments for the ORFs were based on database searches using BLAST [70] against UniProt [71], NCBI/CDD [72], and SMART [73] databases. In most cases, we applied an E-value threshold of 0.001 for the database searches to retrieve homologues. Detailed analyses using multiple sequence alignments and phylogenetic reconstructions were carried out to assign putative functions to the ORFs, when needed. Orthologous gene relationships betweenR. africae and otherRickettsia species were approximated using the best reciprocal BLAST match criterion. The numbers of transposases, ankyrin/tetratricopeptide repeat-containing genes, and integrases were computed using RPS-BLAST with NCBI/CDD entries related to those domains with a 10^-5 E-value threshold. tRNA genes were identified using tRNAscan-SE [74]. To identifyRickettsia palindromic elements, we used hidden Markov models [75] based on the previously identifiedRickettsia palindromic element sequences. ClustalW [76], T-coffee [77], and MUSCLE [78] were used to construct multiple sequence alignments. Toxin-antitoxin genes were identified using the Rasta-Bacteria softwarehttp://genoweb.univ-rennes1.fr/duals/RASTA-Bacteria.

Phylogenetic analysis

We based our analysis on the 645 complete orthologous genes found by Blast programmes in allRickettsia genomes [70]. Subsequently, the amino acid sequences of these 645 proteins were concatenated for each genome and multiple alignment was performed using the Mafft software [79]. Gapped positions were removed. The maximum parsimony and neighbor joining trees were constructed using the MEGA 3.1 software [80].

Clonal origin ofR. africae

We examinedR. africae within 155Amblyomma sp. ticks and eggs from various geographical origins [see Additional file5]. These included 80 adults (40 male and 40 female), 40 larvae, 15 nymphs and 20 eggs. PCR amplification of thetraD gene was performed using theR. africae-specific primer pair traD-F (5'-caatgcttgatctatttggtag-3') and traD-R (5'-cttccttttctctaagctatgc-3') and the probe traD-probe (5'-FAM-ttatggtgctaactccatgcgtgatg-TAMRA-3'). The presence of the plasmid was estimated using the primer pair 1267F (5'-ccagccattaccgtaatcac-3') and 1267R (5'-tagtgccttatactcaagttc-3') and the probe 1267-probe (5'-FAM-gcagaaagtgattaaggcgatcagctg-TAMRA-3') that is able to detect ORF 1267 encoding a protein of unknown function specific to the plasmid. The presence of the plasmid was examined in 22 strains obtained from patients who contracted the disease in South Africa and maintained in the CSUR [see Additional file6], in PCR-positive eschar biopsies from another 48 patients who developed ATBF following a trip to South Africa, and in 32Amblyomma sp. ticks found positive forR. africae, using the above-described PCR assay [see Additional file5]. To evaluate the genetic diversity ofR. africae, we used the multi-spacer typing (MST) method as previously described [53]. This method has been described as the most discriminatory genotyping tool at the intraspecies level inRickettsia sp. [53]. We applied this method to the aforementioned 22 humanR. africae strains, 48 eschar biopsies, and 32Amblyomma sp. ticks from Sudan (3), Madagascar (3), Mali (3), Niger (6), Central African Republic (6), Ivory Coast (3), Guadeloupe (4), Martinique (2), and S^t Kitts and Nevis (2) [see Additional file5]. The obtained sequences were compared to those available in GenBank, and the MST genotypes were determined as previously described [53].

Transcription of genes conserved inR. africae but absent from highly pathogenic species

To evaluate the transcription of the 18 genes conserved byR. africae and degraded in highly pathogenic species, we designed specific primer pairs and probes for each gene and tested the transcription of these genes by RT-PCR on RNA extracted fromR. africae- infected Vero cells cultivated at 32 and then at 37°C and in XTC cells at 28 and 32°C. Experimental protocols are detailed in Additional file15 [see Additional file15].

This work was funded by the Network of Excellence "EuroPathoGenomics".

Authors and Affiliations

1. Unité des rickettsies, IFR 48 CNRS UMR 6020, Faculté de médecine, Université de la Méditerranée, 27 Boulevard Jean Moulin, 13385, Marseille cedex 05, France

   Pierre-Edouard Fournier, Khalid El Karkouri, Quentin Leroy, Catherine Robert, Bernadette Giumelli, Patricia Renesto, Cristina Socolovschi, Philippe Parola & Didier Raoult

2. Information Génomique et Structurale, CNRS UPR2589, Institut de Biologie structurale et Microbiologie, Marseille, France

   Stéphane Audic

Corresponding author

Correspondence toDidier Raoult.

Authors' contributions

PEF and DR designed the study, drafted the manuscript, and gave final approval of the submitted version; KE, QL, CR, BG, PR, CR, PP, and SA performed experiments, drafted the manuscript and gave final approval of the submitted version.

Open Access This article is published under license to BioMed Central Ltd. This is an Open Access article is distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions