Characterization of Francisella-LE haplotypes
We amplified and sequenced the Francisella-LE 16S rRNA, rpoB, groEL, ftsZ, and gyrB gene sequences from the 51 DNA templates belonging to 14 tick species (Table 1). We completed these data with additional sequences from 15 other tick species that include genomic and multi-locus typing of Francisella-LE datasets available on GenBank (listed in Table 1). Overall, the complete multi-locus dataset included Francisella-LE sequences from 29 tick species.
On the basis of 16S rRNA, rpoB, groEL, ftsZ, and gyrB gene sequences, we characterized 24 to 32 distinct alleles depending on the gene, leading to the identification of 38 genetically different Francisella-LE haplotypes in the 29 tick species (Table 1). For 22 tick species, we only observed one Francisella-LE haplotype per tick species. For the seven other tick species, we detected sequence variation at one to five Francisella-LE genes between conspecific specimens, and up to four distinct Francisella-LE haplotypes could be present in the same tick species as observed in Hyalomma excavatum (Table 1). Of the 38 Francisella-LE haplotypes, 37 are specific to their respective tick species and are not shared by two or more tick species. Only the Francisella-LE haplotype #6 was detected in three tick species (A. dissimile, A. geayi, and A. latepunctatum) (Table 1).
Phylogenetic and statistical analyses
ML analyses based on 16S rRNA, rpoB, groEL, ftsZ, and gyrB gene sequences were conducted to examine the Francisella-LE phylogeny (Fig. S1–S5). We observed no sign of recombination in the dataset (all p> 0.17) and we thus further conducted a new ML analysis based on the 16S rRNA, rpoB, groEL, ftsZ, and gyrB concatenated dataset (Fig. 1). All but one phylogenetic reconstructions showed that the Francisella-LE, including F. persica, delineate a robust monophyletic clade within the Francisella genus (Fig. 1, Fig. S2–S5). Only the topology of the16S rRNA gene tree is poorly resolved due to insufficient sequence polymorphism (Fig. S1). The closest relative of Francisella-LE is an opportunistic Francisella pathogen (F. opportunistica [41]), as well as other Francisella pathogens, including the agent of tularemia, F. tularensis (Fig. 1, Fig. S1–S5).
Phylogenetic reconstructions showed that the different Francisella-LE haplotypes found in the same tick species always cluster together (Fig. 1). Indeed, the four Francisella-LE haplotypes of Hy. excavatum are more closely related to each other than to any other Francisella-LE haplotype. A similar pattern was observed for the six other tick species hosting more than one Francisella-LE haplotype.
The cophylogeny analysis returns a significant cophylogenetic signal between Francisella-LE and ticks (global sum of squared residuals: =0.002, p<0.001, n=10 000) (Fig. 2). This supports the hypothesis that the Francisella-LE phylogeny tracks the tick phylogeny, and that the two partners have undergone coupled evolutionary change. Indeed, Francisella-LE of some tick species belonging to the same genus can cluster together along the phylogeny, as shown with the Francisella-LE of A. varium, A. goeldi, and A. humerale (Fig. 1, Fig. 2). Further comparisons of pairwise phylogenetic distances showed evidence on Francisella-LE specificity in two tick genera, Hyalomma (Ixodidae) and Ornithodoros (Argasidae) (Fig. 3). Francisella-LE haplotypes of the Hyalomma genus are more closely related to each other (intrageneric pairwise phylogenetic distances, mean ± SE: 0.0094 ± 0.0008, n=36 pairwise comparisons) than to Francisella-LE haplotypes of other tick genera (intergeneric pairwise phylogenetic distances, 0.0152 ± 0.0002, n=667) (Wilcoxon test, W=4710, p<10-9). A similar pattern was observed in the Ornithodoros genus with intrageneric pairwise phylogenetic distances (0.0030 ± 0.0010, n=3) significantly lower than intergeneric pairwise phylogenetic distances (0.0150 ± 0.0002, n=700) (W=59, p= 0.0048). No evidence of Francisella-LE specificity was observed in other tick genera (Amblyomma: W=42146, p=0.97; Dermacentor: W=1529, p=0.26; Ixodes: W=330, p=0.92; Argas and Rhipicephalus: not applicable because only one haplotype was observed in each of these genera) (Fig. 3). However, it is noteworthy that some Francisella-LE of several Amblyomma species (e.g., A. pacae, A. latum, A. oblongoguttatum, A. maculatum, A. longirostre, A. dissimile, A. geayi, A. latepunctatum, and A. rotundatum) cluster together (Fig. 1, Fig. 2): It suggests that some degree of Francisella-LE specificity in these Amblyomma species may also exist.
Worthy of note is that the tick phylogeny is incompletely resolved: While relationships between tick genera is fully resolved, all congeneric tick species were arbitrarily considered here as equally distant because of the lack of data for some tick species (see Materials and Methods). It implies that the cophylogenetic signal is certainly significant between Francisella-LE haplotypes and tick genera, but not necessarily with tick species. Furthermore, the diagram of the interaction network shows some major phylogenetic incongruences (Fig. 2). As can be seen, no tick genus harbors a specific and monophyletic Francisella-LE subclade: The Francisella-LE of Amblyomma are scattered among Francisella-LE of other tick genera, as best shown with the Francisella-LE of A. sculptum and A. paulopunctatum that are more closely related to the Francisella-LE of the soft ticks O. moubata and O. porcinus than to the Francisella-LE of other Amblyomma species (Fig. 1). Hence, the Francisella-LE of Amblyomma belonged to a minimum of three distinct phylogenetic clusters (Fig. 1, Fig. 2). Similarly, the Francisella-LE of Dermacentor tick species, as well as for Ixodes species, were each scattered among two different Francisella-LE branches (Fig. 1, Fig. 2). In addition, we found a non-significant clustering signal for tick families (Argasidae and Ixodidae) on the phylogeny of Francisella-LE haplotypes (D=-0.43): Their distribution on the tree is significantly random (p(D<1)= 0.04, non-significant after sequential Bonferroni correction), and statistically distinguishable from a clustered distribution expected by Brownian motion (p(D>0)=0.67) (Fig. 2). Overall, these patterns are the signatures of repeated horizontal transfer events, revealing the ability of Francisella-LE to extensively move among tick species.
We found no signal of phylogenetic clustering for Francisella-LE endosymbiosis types (D=1.81): The distribution of obligate and facultative Francisella-LE on the tree is random (p(D<1)=0.92) and facultative Francisella-LE are scattered along the phylogeny among obligate Francisella-LE (Fig. 2). The Francisella-LE haplotype #6 illustrates this pattern well, as it was associated with obligate endosymbiosis in A. dissimile but with facultative endosymbiosis in A. geayi and A. latepunctatum (Table S3).
The geographic origin of Francisella-LE haplotypes showed a significant phylogenetic signal (D=-0.08, p(D<1)=0.008, p(D>0)=0.59) with a clear non-random distribution of Francisella-LE haplotypes between ticks from the Old and New World (Fig. 2). The best examples include Francisella-LE haplotypes of American Dermacentor and Ixodes species that cluster together within the same subclade, while Francisella-LE haplotypes of European Dermacentor and Ixodes species cluster together in another subclade. There are only a few exceptions to this geographical pattern as shown with the Francisella-LE haplotype of an African Amblyomma species, A. latum, which clusters within a clade otherwise only composed of American Amblyomma species (Fig. 2).
We also found a significant signal of phylogenetic clustering for certain vertebrates on which ticks feed (Fig. 2) : Francisella-LE haplotypes cluster with tick species feeding on birds (D=0.14, p(D<1)= 0.02, p(D>0)=0.39), but not with tick species feeding on mammals (D=0.78, p(D<1)=0.36, p(D>0)=0.10) or on reptiles (D=0.99, p(D<1)=0.51, p(D>0)=0.06). However, although globally non-significant, some tick species that exclusively feed on mammals (e.g., O. moubata, O porcinus, A. sculptum, and A. paulopunctatum) harbor closely related Francisella-LE haplotypes (Fig. 2). Similarly, A. rotundatum and A. dissimile, which both feed on reptiles, harbor closely related Francisella-LE haplotypes. However, exceptions are also observed on the haplotype-based tree as shown with A. longirostre: This species feeds on birds, but harbors a Francisella-LE haplotype more closely related to haplotypes of Amblyomma species feeding on mammals and reptiles than to haplotypes of other tick species feeding on birds (Fig. 2).