Proposal to create subspecies of Rickettsia conorii based on multi-locus sequence typing and an emended description of Rickettsia conorii

Background Rickettsiae closely related to the Malish strain, the reference Rickettsia conorii strain, include Indian tick typhus rickettsia (ITTR), Israeli spotted fever rickettsia (ISFR), and Astrakhan fever rickettsia (AFR). Although closely related genotypically, they are distinct serotypically. Using multilocus sequence typing (MLST), we have recently found that distinct serotypes may not always represent distinct species within the Rickettsia genus. We investigated the possibility of classifying rickettsiae closely related to R. conorii as R. conorii subspecies as proposed by the ad hoc committee on reconciliation of approaches to bacterial systematics. For this, we first estimated their genotypic variability by using MLST including the sequencing of 5 genes, of 31 rickettsial isolates closely related to R. conorii strain Malish, 1 ITTR isolate, 2 isolates and 3 tick amplicons of AFR, and 2 ISFR isolates. Then, we selected a representative of each MLST genotype and used multi-spacer typing (MST) and mouse serotyping to estimate their degree of taxonomic relatedness. Results Among the 39 isolates or tick amplicons studied, four MLST genotypes were identified: i) the Malish type; ii) the ITTR type; iii) the AFR type; and iv) the ISFR type. Among these four MLST genotypes, the pairwise similarity in nucleotide sequence varied from 99.8 to 100%, 99.4 to 100%, 98.2 to 99.8%, 98.4 to 99.8%, and 99.2 to 99.9% for 16S rDNA, gltA, ompA, ompB, and sca4 genes, respectively. Representatives of the 4 MLST types were also classified within four types using MST genotyping as well as mouse serotyping. Conclusion Although homogeneous genotypically, strains within the R. conorii species show MST genotypic, serotypic, and epidemio-clinical dissimilarities. We, therefore, propose to modify the nomenclature of the R. conorii species through the creation of subspecies. We propose the names R. conorii subsp. conorii subsp. nov. (type strain = Malish, ATCC VR-613), R. conorii subspecies indica subsp. nov. (type strain = ATCC VR-597), R. conorii subspecies caspia subsp. nov. (type strain = A-167), and R. conorii subspecies israelensis subsp. nov. (type strain = ISTT CDC1). The description of R. conorii is emended to accomodate the four subspecies.


Background
Taxonomic classification of members of the order Rickettsiales was originally based on relatively few phenotypic criteria [1]. Over the past twenty years, the taxonomic classification of the Order Rickettsiales has undergone many changes. On the basis of 16S rDNA sequences, Coxiella burnetii was reclassified within the Legionellaceae [2], Eperythrozoon sp. within the Mycoplasmataceae [3], and Rochalimaea sp., Grahamella sp., and Bartonella sp. within the Bartonellaceae, closely related to Brucella sp [4,5]. Among the remaining members of the order Rickettsiales, the Anaplasma-Ehrlichia clade was recently reorganized into 4 genera [6]; and R. tsutsugamushi was reclassified into a new genus, Orientia [7]. Within the genus Rickettsia, since the pioneering work of Philip in 1978 [8], two rickettsial strains were considered as having different serotypes if they exhibited a specificity difference of ≥ 3 and some authors have considered this to be a useful guide to speciation [9,10]. However, opinions divide as to whether this holds true for rickettsial strains related to Rickettsia conorii (R. conorii) [11]. These include Indian tick typhus rickettsia (ITTR) [12], Israeli spotted fever rickettsia (ISFR) [13][14][15] and Astrakhan fever rickettsia (AFR) [16][17][18][19]. Phylogenetically, these rickettsiae constitute a homogeneous cluster supported by significant bootstrap values and are distinct from other Rickettsia species (Figure 1) [20][21][22]. Some authors have considered these rickettsial strains as different species and proposed the names R. sharonii and R. caspii for ISFR and AFR, respectively [23,24]. Other authors believe that they belong to a "R. conorii complex" [11,14,15,17,19].
These rickettsiae closely related to R. conorii have been described as the causative agents of Mediterranean spotted fever, Astrakhan fever, Israeli spotted fever, and Indian tick typhus in the Mediterranean basin and Africa, Southern Russia, Middle East, and India, respectively. These rickettsioses are transmitted to humans by Rhipicephalus ticks [1]. Furthermore, most of their clinical features overlap and are characterized by a febrile illness and a generalized maculopapular rash. However, an inoculation eschar is seldom present in Astrakhan fever and Israeli spotted fever but common in Mediterranean spotted fever (Table 1).
Recently, we have proposed gene sequence-based criteria for the identification of Rickettsia isolates at the genus, group and species levels and proposed criteria to define Unrooted phylogenetic tree derived from the comparison of sequences of the complete gltA gene from rickettsiae closely related to R. conorii and other validated Rickettsia species Figure 1 Unrooted phylogenetic tree derived from the comparison of sequences of the complete gltA gene from rickettsiae closely related to R. conorii and other validated Rickettsia species. The analysis used the Neighbor Joining method and the Kimura 2 parameter as described in the Methods section. Bootstrap values are indicated at the nodes of the phylogenetic tree. GenBank accession numbers are indicated in parentheses for each taxon.  species [9]. Using these criteria, we have shown that ITTR, ISFR and AFR could not be considered as distinct species. Instead, they belonged to the R. conorii species [9]. However, these rickettsiae exhibit differentiable serotypes and cause diseases with distinct clinical features in defined geographic locations. The ad hoc committee on reconciliation of approaches to bacterial systematics [25] has proposed that even if related genetically, bacterial isolates within a given species could be considered as distinct subspecies if they differed phenotypically.
In this study, we undertook to examine whether phenotypic differences existed among various R. conorii isolates that would enable to classify them as subspecies. We estimated the degree of genotypic variabilities among 31 isolates of R. conorii, 1 isolate of ITTR, 2 isolates and 3 tick amplicons of AFR, and 2 isolates of ISFR using multi-locus sequence typing (MLST). We incorporated 16S rDNA and gltA genes, as well as 3 membrane-exposed protein-encoding genes ompA, ompB and sca4 (formerly gene D) [26,27] in MLST. To further characterize the specificities of distinct MLST types, we incorporated a prototype isolate from each of these into a multi-spacer typing (MST) assay, which we have previously demonstrated to be more discriminant than MLST at the strain level for R. conorii [28]. Furthermore, mouse serotypes were obtained for each of these MLST types.

MLST-gene sequencing
All five genes studied were amplified from all 36 studied isolates and from 3 of 57 Rhipicephalus pumilio ticks tested. All negative controls remained negative. Four MLST types were identified. Within the first type, which we named Malish type, all R. conorii isolates exhibited identical sequences for each of the five genes tested except isolates Moroccan and M1. These latter isolates showed deletions of 24 and 156 bp, respectively, at the 5'-end of the ompA gene. The second type, the ITTR type, was made of ITTR. This exhibited unique nucleotide substitutions in ompA, ompB and sca4 genes. Within the AFR type, all isolates and tick amplicons tested showed identical sequences of the five genes studied except the Chad isolate which showed subtle nucleotide substitutions in each of them. In this isolate, the pairwise nucleotide similarity with AFR for 16S rDNA, gltA, ompA, ompB and sca4 genes was 99.7, 99.5, 99.5, 99.2, and 99.8 %, respectively. The fourth type, ISFR type, which included both ISFR isolates, showed the same nucleotide substitutions in each of the five genes tested. This differentiated them from other rickettsiae. As representative strains for each MLST genotype, we have chosen R. conorii isolate Malish, ITTR isolate Indian, AFR isolate A-167, and ISFR isolate ISTT CDC1.
Among representatives of the four MLST genotypes, the degree of pairwise nucleotide similarity ranged from 99.8 % between AFR and ISFR to 100 % between R. conorii and ITTR for the 16S rDNA gene (Table 3). For gltA, such similarity ranged from 99.5 % between Malish (or ITTR) and The dendrograms obtained using the three different tree building analysis methods showed similar organization for the five rickettsiae studied. We identified two monophyletic groups with the trees so constructed. One included R. conorii and ITTR and the other, ISFR and AFR ( Figure 2A). The clustering of these rickettsiae was supported by high bootstrap values. Both AFR isolates were grouped together for each of the five genes tested.

MST genotyping
ITTR showed unique 232-bp dksA / xerC (type P, GenBank accession number AY836513) and 153-bp mppA / purC (type F, AY836517) spacer sequences. However, it also showed a 200-bp rpmE / tRNA-fMet spacer sequence identical to that of R. conorii type B (AY345092); ISFR had unique 232-bp dksA / xerC (type Q, AY836510), 153-bp mppA / purC (type G, AY836514), and 200-bp rpmE / tRNA-fMet (type C, AY836518) spacer sequences; AFR iso-lates A-167 and Chad showed unique dksA / xerC of 232-(type R, AY836511) and 167-bp (type S, AY836512), respectively. They also showed identical 153-bp mppA / purC (type H, AY836516) and 200-bp rpmE / tRNA-fMet (type D, AY836519) spacer sequences that were different from those of other strains. When combining the genotypes obtained from each of the three spacers into a MST genotype, each of the four rickettsiae showed a distinct MST genotype. The MST genotype so obtained was different from that of R. conorii. However, on pairwise comparisons of the four strains, both AFR strains showed similar spacer types. The phylogenetic tree inferred from the concatenated spacer sequences showed an organization similar to that obtained from MLST when Malish strain was taken as R. conorii type strain ( Figure 2B). However, the node where ISFR branched was not supported by a significant bootstrap value.

Microimmunofluorescence (MIF) serotyping
When injected into mice, all five rickettsiae tested provoked antibody responses that were detectable by MIF.
The endpoint titers and specificity differences (SPDs) were calculated as described in the Methods section. These values are presented in Table 4. The antibody titres obtained in homologous sera were at least one-fold higher than those observed in heterologous sera. The SPDs for heterologous sera varied between 1 and 5. Four serotypes were identified. These included R. conorii, ITTR, both AFR isolates, and ISFR. The dendrogram inferred from the matrix based on SPDs ( Figure 2C) showed that ITTR occupied a position that slightly differed from its phylogenetic organization. Instead of forming a cluster with R. conorii, it occupied an intermediate position between R. conorii and ISFR (Figures 2A and 2B).

Discussion
We propose to create subspecies within the R. conorii species to accommodate both the genotypic homogeneity and the serotypic, MST genotypic, geographic and pathogenic diversities of rickettsia strains within this species.
Defining a species within the Rickettsia genus has long been difficult but rickettsiologists have agreed to use mouse serotyping as one criterion [8]. This classification scheme served as a basis for the development of genotypic criteria at the species level [9]. However, confusion still exists about the taxonomic status of rickettsiae. This is especially the case for rickettsiae which are closely related to classic R. conorii isolates. Philip et al. [8], using mouse serotyping, have demonstrated that R. conorii isolates Malish, Moroccan and Kenya were closely related antigenically to ITTR and that they belonged to a single serotype. Using complement fixation, Bozeman et al. [29] were unable to distinguish ITTR from R. conorii isolates. In contrast, others, using mouse polyclonal antibodies [14], and monoclonal antibodies [11], have demonstrated substantial differences between ITTR and R. conorii isolates. ISFR and AFR could not be distinguished using PCR-RFLP [17] until ompA was used as target gene [30]. In 1993, we have demonstrated that AFR differed from both ISFR and R. conorii in terms of SDS-PAGE mobility and PFGE profiles [19]. Although they are difficult to distinguish serologi-cally, the cross-reacting epitopes they share are located on antigenic proteins that differ in SDS-PAGE mobility and hence in molecular weight [19]. In 1995, Walker et al., using serotyping, Western bloting, monoclonal antibody reactivity and PCR amplification of the tandem repeats within ompA, concluded that ISFR belonged to the R. conorii species [15]. In 1998, Dasch and Jackson differen- tiated R. conorii from ISFR using PCR-RFLP [23]. Recently, using a combination of genotypic criteria, we have demonstrated that the genotypic differences among ITTR, AFR and ISFR were too subtle to classify them as new species [9]. Instead, they belonged to the R. conorii species.
In the present study, we have demonstrated, using MLST, that rickettsiae closely related to R. conorii isolate Malish are very homogeneous. Except isolates Moroccan and M1 which showed one ompA deletion of 24 and 156 bp, respectively, all studied R. conorii isolates had identical sequences for the five genes studied and thus constituted a reliable taxon. The ITTR isolate showed unique sequences of ompA, ompB and sca4 genes. Among isolates and tick amplicons of AFR, the degree of MLST sequence similarity was of 100% except for the Chad isolate which is genotypically slightly different from AFR isolate A-167 [31]. In another study, we have found identical ompA sequences in four tick amplicons from Kosovo and AFR isolate A-167 [32]. We have also observed serologic, antigenic and genetic homogeneities among two tick and one human AFR isolates from Astrakhan [19]. Thus, except the Chad isolate which was slightly different genotypically, AFR, with three strains and ten tick amplicons studied, appeared to form a homogeneous taxon. The two isolates of ISFR tested showed 100% pairwise nucleotide similarity in all five genes studied. As we have compared only two isolates, one may argue that this may not be representative of the diversity of ompA in ISFR. However, Walker et al. found four human ISFR isolates from Israel to be highly homogeneous antigenically as well as genetically [15]. Bacellar et al. described three human isolates of ISFR from Portugal that showed 100% similarity in 16S rDNA, gltA and ompA sequence with those of ISFR isolate ISTTCDC1 [33]. Giammanco et al. identified a tick rickettsial isolate in Italy that had 100% ompA sequence similarity with ISFR isolate ISTTCDC1 [34]. Therefore, it appeared that our ISFR isolates were representatives of the same homogeneous taxon.
In the first part of the study, we have identified four MLST genotypes which were subjected to further evaluation. Among representatives of these MLST genotypes, the minimal pairwise similarity in the nucleotide sequences of 16S rDNA, gltA, ompA, ompB, and sca4 genes between the rickettsial isolates tested and R. conorii, was 99.8%, 99.4%, 98.3%, 98.5%, and 99.3%, respectively. Thus, All four isolates tested fulfilled the criteria to belong to the R. conorii species [9].
Phylogenetically, the members of the R. conorii complex have been demonstrated to form a stable group distinct from other Rickettsia species (Figure 1) [20][21][22]. We found that within this cluster they consistently formed two phylogenetic clusters (Figure 2A). One of these clusters included R. conorii and ITTR, and the other, AFR isolates and ISFR. Using MST, we identified five genotypes. However, using pairwise comparison of spacer sequences, we identified only four genotypes similar to those obtained by MLST. MST-based phylogeny also demonstrated a similar pattern ( Figure 2B). Serotypically, we could identify four serotypes which, according to the definition of species proposed by Philip [8], would classify them as different species: 1) R. conorii isolate Malish; 2) ITTR; 3) AFR; and 4) ISFR. These four serotypes also showed distinct MST genotypes.
We demonstrated a discrepancy between the genotypic homogeneity of these rickettsiae, which classified them within the R. conorii species [9], and their MST genotypic, serotypic, geographic, and pathogenic heterogeneity. For ITTR, an additional discrepancy was observed between its phylogenetic and serotypic classifications (Figures 2A and  2B versus 2C). Such findings are in accordance with those of Walker et al. who had shown substantial differences between ITTR and R. conorii isolates [11]. We also agree with these authors who hypothesized that all members of the "so-called" R. conorii complex belonged to the R. conorii species [15].

Conclusion
We have demonstrated that R. conorii and closely related strains exhibit a high level of conservation in nucleotide sequence despite genotypic specificities, and phenotypic specificities including serotyping, epidemiological and clinical characteristics. Therefore, according to the accepted criteria of the ad hoc committee on reconciliation of approaches to bacterial systematics which allow the creation of subspecies for genetically-close isolates which express phenotypic differences [25], we are justified in creating new subspecies for these organisms.

Emendation of the description of Rickettsia conorii
Rickettsia conorii (Brumpt 1932 The characteristics are the same as those of the species. Transmitted to humans through the bite of Rhipicephalus sanguineus and R. pumilio ticks. Grows in Vero cells at 32°C in antibiotic-free Minimal Essential Medium supplemented with 2 % fetal calf serum and 2 mg/ml Lglutamine.

Rickettsial strains
All rickettsial strains tested and their original sources are listed in Table 2. Rickettsial isolates for which sequences were not available were grown in Vero cell monolayers at 32°C in 150 cm 2 -cell culture flasks as previously described [37]. When Gimenez staining demonstrated the cells have been infected heavily, they were harvested using sterile glass beads and centrifuged at 12,000 g for 10 min, resuspended in fresh medium and stored at -70°C until purification of nucleic acid was carried out. All rickettsial stocks were tested and shown to be free of Mycoplasma sp using a 16S rDNA-based PCR assay as previously described [38].

Nucleic acid purification, PCR amplification and sequencing
When available, we used gene sequences present in Gen-Bank. The accession numbers for these sequences are reported at the end of the manuscript. Missing sequences were obtained in this study. Genomic DNA was extracted from 200 µl of rickettsial suspension and from 57 Rhipicephalus pumilio ticks collected in Astrakhan [19] using the QIAamp Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. For MLST, we used the previously described primers and PCR conditions to amplify 1,424-bp, 1,134-bp, 590-bp, 4,890-bp, and 3,028-bp fragments of the 16S rDNA [36], gltA [37], ompA [30], ompB [22] and sca4 [20]  For each gene, the pairwise sequence similarity among studied isolates and tick amplicons was estimated using the multisequence alignment program CLUSTAL W [39]. Each isolate or tick amplicon exhibiting specific nucleotide substitutions within one or more of the five genes studied was considered a different MLST genotype. Then, we selected a representative isolate of each MLST genotype for further tests. When a MLST genotype incorporated more than one isolate, we selected the type strain or the first strain described in the literature as representative strain. In addition to AFR isolate A-167, we incorporated AFR isolate Chad in the phylogenetic study and the mouse immunization and MST assays because MLST sequence differences between both isolates were due to nucleotide substitutions as well as deletions.
Distance matrices generated by DNADIST were determined under the assumptions of Kimura and were used to infer dendrograms by the neighbor-joining method [40] using the PHYLIP version 3.4 software package [41]. Dendrograms were also constructed by data processing with the maximum-likehood and parsimony methods (DNAML and DNAPARS, respectively, in PHYLIP). A bootstrap analysis based on 100 randomly generated trees using SEQBOOT and CONSENSE in PHYLIP was performed to estimate the node reliability of the trees obtained by three phylogenetic methods [42].

Multi-spacer typing (MST)
We amplified and sequenced the dksA / xerC, mppA / purC, and rpmE / tRNA-fMet intergenic spacers for each of R. conorii isolate Malish, ITTR, ISFR, and AFR isolates A-167 and Chad. Primers and methods were those described previously [28]. Subsequently, spacer sequences were concatenated and used to infer the phylogenetic relationships among strains tested using the neighbor-joining method [40] as described above.

Mouse immunization
For each R. conorii isolate Malish, ITTR, ISFR, and AFR isolates A-167 and Chad, five BALB/c mice were inoculated intravenously (tail vein) with purified bacterial suspension (0.1 ml containing 10 4 bacteria). The concentration of rickettsial suspensions was estimated using a plaque assay as previously described [43]. On day 7, mice were boosted with a similar inoculum. On day 10, mice were anesthetized and exsanguinated by cardiac puncture [8]. Sera from each group of five mice were pooled and stored at -20°C until used for microimmunofluorescence (MIF) serotyping.

MIF serotyping
The MIF test was performed and the SPD was calculated using the method of Philip et al. [8]. Each of the five rickettsiae tested was compared with each other. The highest serum dilutions giving positive reactions were recorded as endpoint titer: SPD = (Aa + Bb) -(Ab + Ba), where Aa or Bb is -log2 of the endpoint titer between serum A or B and homologous antigen a or b and Ab or Ba is -log2 of the endpoint titer of serum A or B against heterologous antigen b or a. If the SPD was less than 3, the two isolates were assumed to belong to the same serotype. If the SPD was ≥ 3, the isolates were assumed to belong to different serotypes. The SPDs were used to construct a matrix, and then a dendrogram was constructed from the matrix by the unweighted pair group method with arithmetic mean (UPGMA) available in the MEGA2 software package [44].