Arsenophonus and Sodalis Symbionts in Louse Flies: an Analogy to the Wigglesworthia and Sodalis System in Tsetse Flies

Samples of M. ovinus were obtained from a licensed family sheep farm at Krásetín, Czech Republic (48°53′4.549″N, 14°18′51.894″E). Bacteriomes and milk glands were dissected in phosphate-buffered saline (PBS) under an Olympus SZ51 dissecting microscope. The bacteriome tissue samples were used for both DNA extraction and fluorescent in situ hybridization (FISH). Transmission electron microscopy (TEM) of milk glands was carried out as reported previously for bacteriomes (19).

The tissue samples including gut and reproductive tract with surrounding milk glands were fixed overnight in 4% paraformaldehyde and then incubated in 2% H₂O₂ for up to 48 h to quench tissue autofluorescence. Afterwards, tissues were embedded in paraffin, and 6-μm-thick sections were placed on silane-coated slides. The slides were kept dry at a room temperature until used in fluorescent in situ hybridization (FISH). The probes were designed to match specifically the 16S rRNA of each associated bacteria to avoid any cross-hybridization. The labeled probe sequences are as follows: “Ca. Arsenophonus,” Flc-GGCCTTACGGTCCCTCA (where Flc is fluorescein); “Ca. Sodalis,” Cy3-CATCGCCTTCCTCCCCGCTG; B. melophagi, TxRd-CGTCATTATCTTCACCGGTG (where TxRd is Texas Red); Wolbachia sp., Cy5-CTTCTGTGAGTACCGTCATTATC and Cy3-AACGCTAGCCCTCTCCGTA; these allow various probe combinations in a single hybridization. The sections were dewaxed in xylene for 10 min, followed by 10 min in a xylene-ethanol equal-ratio solution. After rehydration in a graded ethanol series, the samples were treated with a drop of 70% acetic acid and incubated at 45°C until dry. The samples were then dehydrated through a graded ethanol series and underwent deproteinization with proteinase K solution (Qiagen) for 8 min at 45°C. The slides were rinsed in PBS, dehydrated, and air dried prior to the prehybridization step, which was carried out for 30 min at 45°C using hybridization buffer (0.9 M NaCl, 20 mM Tris-HCl, pH 7.2, 0.1% SDS). Hybridization was performed for 3 h at 45°C directly on slides coated with 200 μl of hybridization buffer with 2 ng/μl of specific, labeled probe for each of the bacterial associates. Afterwards, the slides were rinsed twice for 5 min in the hybridization buffer and briefly rinsed in PBS and water. The slides were mounted in Vectashield medium (Vector Laboratories) containing 4′,6-diamidino-2-phenylindole (DAPI) and kept in the dark at 4°C until observed on an Olympus FV1000 confocal microscope. Negative controls (using RNase-treated slides) and controls for probe specificity (employing a general eubacterial probe EUB338, Flc-GCTGCCTCCCGTAGGA) (23) were performed along with each hybridization (data not shown).

Total genomic DNA (gDNA) for construction of an Illumina paired-end (PE) library was obtained from a single bacteriome. DNA was extracted using a QIAamp DNA microkit (Qiagen) according to the standard kit protocol. DNA concentration was determined using Qubit Fluorometric Quantification (Life Technologies), and its quality was assessed in 1% agarose gel using standard electrophoresis. In total, 3 μg of gDNA was used for PE library construction.

DNA isolates acquired from whole individuals were examined for “Ca. Sodalis,” B. melophagi, and Wolbachia presence with diagnostic PCR and further used for Wolbachia multilocus sequence typing (MLST). MLST genes from Wolbachia were sequenced according to a Wolbachia MLST protocol (24). Diagnostic PCR with specific primers was performed on 82 flies (55 males and 27 females). Details on primer sequences, either newly designed or retrieved from literature (25, 26), and amplicon lengths are presented in Table S5 in the supplemental material. In order to standardize the PCR conditions, ready-to-use PPT master mix (Top Bio) was employed, and the sensitivity of PCR was evaluated on five different DNA concentrations ranging from 5 to 60 ng/μl. Products of several diagnostic PCRs were Sanger sequenced to ensure primer specificity.

“Ca. Sodalis” gDNA for Illumina 3-kb mate-pair (MP) library construction was extracted from a single colony cultivation of “Ca. Sodalis” spread on Mitsuhashi and Maramorosch insect (MMI) medium blood plate (19) using the same procedure. In total, 10 μg of gDNA was used for MP library construction.

Bacteriome gDNA was sequenced using an Illumina PE library with an average insert size of 180 bp. The PE and MP libraries were sequenced bidirectionally on one lane of a HiSeq 2000 system on 100-bp and 75-bp runs, respectively (Keck Microarray Resource, West Haven, CT, USA). In total, over 80 million reads were obtained from the PE library. Reads underwent adaptor and quality trimming, and those shorter than 25 bp were discarded. Processed reads were assembled as metadata using two different approaches. First, de novo assembly of all the processed reads into contigs was achieved on a CLC Genomic Workbench (CLC bio A/S) with default parameter settings. Second, a subset of roughly 50 million reads consisting of paired reads only was randomly reduced in order to adjust the average coverage for Arsenophonus data to 100-fold and assembled using Velvet (27). De novo assembly in Velvet was performed with a k-mer size set to 55 bp. Contigs from each assembly were binned separately based on their average coverage and blastx (28) hits against a nonredundant (nr) database. Two sets of almost identical Arsenophonus binned data were then compared and combined into the draft genome of 40 Arsenophonus contigs. The contigs were submitted to an automatic annotation pipeline of the Joint Genome Institute, Integrated Microbial Genomes/Expert Review (IMG/ER) (29), using Prodigal (30) as the preferred gene calling method.

A draft genome of “Ca. Sodalis” was assembled from a combination of PE and MP Illumina data by the Velvet assembler (27), with all parameters determined by VelvetOptimiser (http://www.vicbioinformatics.com/software.velvetoptimiser.shtml) and default scaffolding. Automatic annotation for comparative purposes was executed in RAST (http://rast.nmpdr.org). A B. melophagi draft genome was assembled from PE data by following the same procedure. A k-mer size of 71 was used for B. melophagi, and a k-mer size of 57 was used for “Ca. Sodalis” because a 75-bp size was used for MP reads. The “Ca. Sodalis” genome was further scaffolded by SSPACE (31). Gaps in B. melophagi scaffolds were closed by PCR. The completeness of gammaproteobacterial draft genomes was evaluated according to the number of single-copy (SICO) genes (32) present in the assemblies assessed by blastp and tblastn searches (28).

Initial genome comparisons between “Ca. Arsenophonus” and Wigglesworthia glossinidia and between “Ca. Sodalis” and S. glossinidius were based on the elementary genome characteristics, e.g., genome size, GC content, number of coding sequences (CDS), and tRNA genes. For the more detailed analysis, orthologous gene clusters were used as the comparative data. This comparison, performed in a broader phylogenetic spectrum, was based on gene numbers within functional ortholog assignments as defined by the Cluster of Orthologous Genes (COG) database and implemented in the web project IMG (29). The data set was composed of gene numbers for 42 bacterial genomes retrieved by the abundance profile function in IMG/ER (29) and supplemented with the data from the genomes of the Wolbachia endosymbiont of Cimex lectularius, Wolbachia wCle, and “Ca. Sodalis” missing in IMG/ER (see Table S1 in the supplemental material). For the two latter genomes, the COG assignment was performed using the COG database (ftp://ftp.ncbi.nih.gov/pub/mmdb/cdd) and reverse position-specific (RPS)-BLAST, which allows for comparison of a query against a library of position-specific scoring matrices (PSSMs) (33). Altogether, 2,473 orthologous clusters were included in the comparison. The analysis was based on two different measures, one expressed as the number of the genes present in the genomes and the other as a percentage of these genes calculated from the sum of all COG clusters. A dissimilarity matrix was calculated from these data using Bray-Curtis distance in R (http://www.R-project.org). Nonmetric multidimensional scaling ordination (NMDS) was used to show similarity in genome composition of selected symbiotic bacteria. Differences in overall genome composition, based on COG clusters, among different host feeding strategies, and between the character of symbiotic relationships (primary versus secondary) were assessed using permutational multivariate test implemented in the Adonis function in R (http://www.R-project.org).

Matrices for individual bacterial genera and their outgroups were compiled from the data obtained in this study and the data retrieved from public databases. Complete data sets were aligned using the Mafft algorithm (34) implemented in SeaView (35) software and concatenated into multigene matrices. These contained sequences for the ftsK, fbaA, yaeT, and spoT genes of Arsenophonus, the gltA, groEL, ribC, and rpoB of Bartonella, and the gatB, coxA, hcpA, ftsZ, and fbpA of Wolbachia. Individual concatenates were processed in the GBlocks application (36) in order to remove unreliably aligned positions. For the genus Arsenophonus, two matrixes were generated: the multigene concatenate described above and, in order to include “Candidatus Riesia pediculicola” into the analyses, an incomplete matrix containing the fbaA gene present in the extremely reduced genome of this bacterium. The resulting alignments were analyzed using maximum likelihood (ML) and Bayesian inference (BI). The best evolutionary models fitting the data were selected using the program jModelTest (37). ML-based analyses and 100 nonparametric bootstrap replicates were performed in the PhyML program (38). Bayesian analyses were performed with correspondent evolutionary models implemented in MrBayes, version 3.1.2 (39).

Several independent approaches were employed to reliably reconstruct B-vitamin and cofactor pathways of all major bacteria present in M. ovinus-“Ca. Arsenophonus,” “Ca. Sodalis,” and B. melophagi. First, annotated draft genome assemblies were uploaded into Pathway Tools software using its pathological module (40), and individual pathways were compared with species comparisons and manually examined. Second, B-vitamin and cofactor genes from Escherichia coli K-12 MG1655 were searched by blastp (E value of 1e−5) against a custom database of all protein coding genes from all three bacteria. Third, putatively missing genes were checked by blasting homologues from E. coli, S. glossinidius, Arsenophonus nasoniae, and Bartonella schoenbuchensis species (blastx; E value, 1e−5) against a custom database of all contigs/scaffolds from bacteriome and “Ca. Sodalis” assemblies. B-vitamin and cofactor pathways were reconstructed with EcoCyc and KEGG databases as guides (41, 42).

The “Ca. Arsenophonus” draft genome is provided in GenBank format (available at http://users.prf.jcu.cz/novake01/). The “Ca. Sodalis” and B. melophagi draft genomes are provided as fasta files with all scaffolds/contigs (http://users.prf.jcu.cz/novake01/). All genomes are currently in the process of gap closure and will be submitted to NCBI when closed into a single circular molecule (or several, in case of plasmids). Raw Illumina reads are available in the ENA database under accession number PRJEB9958.

Screening of the shotgun metagenomic data revealed five microorganisms in the sheep ked microbiota. Phylogenetic analyses showed that four of them belong to the well-known bacterial genera Arsenophonus, Sodalis, Bartonella, and Wolbachia (Fig. 1) and that one is the unicellular eukaryote Trypanosoma melophagium (see Fig. S1 in the supplemental material). The distribution of the bacteria in the host digestive and reproductive tracts, including overall arrangement of the bacteriome and likely transmission routes the symbionts undertake when passed to the progeny, is presented in Fig. 2 and 3.

The Arsenophonus bacterium possesses a number of features also found for some other obligate symbionts. The bacterium was present in all of the examined adults, housed in specialized cells of the host intestine wall (bacteriocytes), forming a compact symbiotic organ (bacteriome) (Fig. 2a to g). Under TEM, the bacteriocytes were tightly packed with pleomorphic populations of the symbiont cells (Fig. 2b). Apart from this typical location, in adult females, these bacteria also occur extracellularly in the lumen of the milk glands (Fig. 3a). This suggests that, similar to Wigglesworthia in tsetse flies, this Arsenophonus member also uses milk glands for its vertical transmission. Correspondingly, it was never observed in oocytes and early embryos, while its presence was confirmed in the late stages of embryogenesis (see Fig. S2 in the supplemental material). Based on the biological, phylogenetic, and evolutionary traits described in this study, we propose the designation “Candidatus Arsenophonus melophagi” (me-lo-phag-i; melophagi, of the genus Melophagus).

In contrast to the universally present “Ca. Arsenophonus,” “Candidatus Sodalis melophagi” does not occur in all adult hosts (diagnostic PCR performed on 82 individuals from a single population indicates approximately a 50% prevalence, with 59% of females and 42% of males positive). In the gut, the bacterium resides predominantly outside the bacteriome. However, in contrast to S. glossinidius, which has never been described from the bacteriome, “Ca. Sodalis” may enter the symbiotic organ. Here, it occupies intercellular space surrounding individual bacteriocytes (Fig. 2e). Similar to “Ca. Arsenophonus,” “Ca. Sodalis” was also detected in the milk glands. However, in addition to the extracellular location, it was also found intracellularly inside the secretory cells (Fig. 3a). Rarely, both bacterial symbionts co-occur in the gland lumen (see Fig. S2 in the supplemental material). No “Ca. Sodalis” cells were detected in oocytes or early embryos.

Wolbachia found in M. ovinus branches as a sister lineage to the F supergroup (Fig. 1) and is highly prevalent in the screened population (100% of females and 94% of males positive) (see Materials and Methods for more detail on diagnostic PCR). Wolbachia was observed intracellularly in various tissues, including fat bodies (adipocytes) (Fig. 3b), secretory cells of the milk glands (Fig. 3a), and the gut tissue (Fig. 2f and i). Within the bacteriome, Wolbachia was scattered in small numbers among bacteriocytes (Fig. 2f). Although present in the secretory cells, Wolbachia was never observed in the milk gland lumen. In contrast to “Ca. Arsenophonus” and “Ca. Sodalis” symbionts, it infected embryos prior to intrauterine development (Fig. 3d). The findings described above distinguish the Melophagus-associated Wolbachia from the strains found in Glossina.

Bartonella melophagi was found in all of the screened individuals and is the only one of the detected bacteria located mostly extracellularly. It was found along the entire gut lumen (Fig. 2g and h) closely attached to the microvilli (see Fig. S1 in the supplemental material). Its abundance is particularly high along the gut wall outside the bacteriome section (Fig. 2h), whereas the cell numbers decrease within the bacteriome (Fig. 2g). Except for the gut, all the other tissues were found to be Bartonella free.

The shotgun metagenomic data combined with genomic data from “Ca. Sodalis” culture (see Materials and Methods) allowed for draft assembly of three bacterial genomes, i.e., “Ca. Arsenophonus,” Bartonella melophagi, and “Ca. Sodalis,” each with coverage above 100-fold (Table 1). The only bacterium with the coverage too low for further analysis was the symbiont of the genus Wolbachia. The sequencing and the assembly strategy we used resulted in genome assemblies truncated in the number of contigs/scaffolds by multiple rRNA operons (“Ca. Arsenophonus”) and a number of duplicated genes and mobile genetic elements in the genomes of Bartonella melophagi and “Ca. Sodalis” (Table 1). The high coverage and the genome properties, particularly the number of tRNA genes and retained single-copy (SICO) genes in gammaproteobacterial genomes (see Materials and Methods), imply that the drafts represent the complete gene set. The Bartonella genome assembly fully reflects the gene content of the U.S. isolate of B. melophagi deposited in the NCBI Assembly database (GCA_000278255), sharing complete gene synteny and ∼99% sequence identity in single-copy genome regions. Altogether, the basic genome characteristics listed in Table 2 underpin our microscopic findings, resembling the tsetse fly symbiosis. While the low-GC content and genome size of “Ca. Arsenophonus” are in line with its obligate relationship to the host, the considerably larger genome of “Ca. Sodalis” and its unbiased base composition suggest a less intimate association.

Compared to other Arsenophonus species and the bacterial symbionts from other taxa (see Table S1 in the supplemental material), “Ca. Arsenophonus” displays a remarkable degree of similarity with the tsetse fly symbiont W. glossinidia in terms of its genome content (Table 2; see also Table S2 in the supplemental material). Both genomes contain similar gene numbers and show similar patterns of present/absent orthologous genes. The ordination plot of overall similarities among 44 symbiotic bacteria, based on the NMDS analysis (Fig. 4), shows remarkably close distance between “Ca. Arsenophonus” and both Wigglesworthia genomes. As expected, the statistical evaluation of the clustering pattern also showed significant differences (P = 0.001) between genome composition of primary and secondary symbionts and among symbionts of hosts with different feeding strategies (i.e., blood-feeding, sap-sucking, and omnivorous insects) (Fig. 4).

Using the assignment of the orthologous clusters to individual COG categories (as defined by the KEGG database), the ratios of shared/present/absent orthologs was found to differ considerably across functional categories (see Tables S2 and S3 in the supplemental material). While in some categories (e.g., translation, ribosomal structure, and biogenesis; intracellular trafficking, secretion, and vesicular transport; and nucleotide transport and metabolism) the three genomes share similar subsets of genes, in others they differ considerably. Such a case is best illustrated by the categories replication, recombination, and repair, energy production and conversion, and cell motility. The “Ca. Arsenophonus” genome prevails in the first two categories, whereas the genome of W. glossinidia possesses a higher number of the genes in the last category. The cell motility category provides an extreme example where W. glossinidia possesses 28 genes, but the entire category is missing in “Ca. Arsenophonus.” A further assignment of the genes within each COG category to particular metabolic pathways shows that the differences between the two symbionts in the gene counts are often due to absence/presence of a metabolic path or its part, rather than to random loss of individual genes (see Table S4 in the supplemental material).

Because the crucial role of B vitamins produced by symbiotic bacteria in hematophagous systems has been generally accepted and because other possible functions of symbiotic bacteria in blood-feeding arthropods (such as immune system maturation) are impossible to infer based on genome data only, we streamlined the analyses toward B-vitamin biosynthetic pathways. Compared to the Glossina-associated symbiotic system, two of the B-vitamin pathways, those for thiamine and pantothenate, were arranged in notably different ways. The genes for the thiamine transporter are present in the obligate symbiont of M. ovinus, “Ca. Arsenophonus,” while the facultative “Ca. Sodalis” symbiont codes for its complete biosynthetic pathway. B. melophagi is also capable of thiamine biosynthesis but uses an alternative pathway starting from glycine (Fig. 5). In correspondence with the absence of pantothenate biosynthesis genes (panBCDE) in the known Arsenophonus chromosomes (BioSample accession numbers SAMN02440620, SAMN02603451, and SAMN03078484), “Ca. Arsenophonus” also lacks these genes and carries only the genes necessary for synthesis of coenzyme A (CoA) from pantothenate (Fig. 5). On the other hand, the genome reduction in the obligate symbionts of blood-sucking insects led to losses of several identical enzymes in B-vitamin pathways, e.g., phoA in folate and nadABC in nicotinate pathways (see Fig. S3 in the supplemental material).

The symbiotic system of Melophagus ovinus shows a remarkable overall analogy to that described for the related hippoboscoid group, the genus Glossina. Compared in more detail, these two systems possess an interesting combination of common features and unique characteristics. While the two obligate mutualists W. glossinidia and “Ca. Arsenophonus” originated independently from two phylogenetically distant groups, they share several important traits. The most apparent trait is their location in adult hosts within the bacteriomes formed by enlarged cells of the intestinal wall (Fig. 2). Considering their nutritional role, this seems to be a suitable location and has been reported for several other hosts, but it is certainly not the only and inevitable arrangement of symbionts in blood feeders, as documented by a variety of bacteriome structures and locations in sucking lice, bat flies, or bed bugs (43–45). Even more striking is the close resemblance of the transmission modes. Both bacteria, “Ca. Arsenophonus” and W. glossinidia, constitute two separate subpopulations in host adult females, one within the bacteriome and the other in the lumen of the milk glands (46–48). A similar arrangement was also reported for the Arsenophonus-related symbionts “Candidatus Aschnera chinzeii” in the bat fly Penicillidia jenynsii (44). It indicates that in these insects the vertical transmission takes place via the accessory organs rather than oocytes, and this mechanism has originated in or been preserved for different groups of bacterial symbionts.

The same transmission path to the developing embryo is also used by the facultative symbionts of the genus Sodalis. However, according to our microscope observations, its transmission is not synchronized with that of “Ca. Arsenophonus.” Rather, “Ca. Arsenophonus” seems to colonize the developing embryo prior to “Ca. Sodalis.” Similarly, the common presence in the host milk gland lumen was only rarely reported for W. glossinidia and S. glossinidius in Glossinidae (48). This suggests that different mechanisms initiate the vertical transmission of facultative and obligate symbionts in M. ovinus or that there is possible competition between them, as proposed previously for the Glossina symbionts (48).

The presence of B. melophagi along microvilli of the M. ovinus midgut indicates possible scavenging of nutrients (e.g., biotin) (from blood digestion and perhaps from the other symbionts (Fig. 5). Although this bacterium, originally cultivated from sheep keds by Noller (49) and described as Rickettsia melophagi, was later also cultivated from commercial batches of sheep blood (50), its life cycle is still ambiguous. Halos et al. (51) reported that despite its presence in all Melophagus samples collected from a sheep, B. melophagi could not be isolated from the sheep itself and explained this by its symbiotic association without transmission to ruminants. Our genomic data do not provide any strong evidence to confirm or rule out this hypothesis. Additional functional data will be needed to infer the exact role of B. melophagi in the system.

While evolving from different ancestors, the two obligate symbionts “Ca. Arsenophonus” and W. glossinidia have been subject to similar host-mediated influences, reflecting phylogenetic relationships of the two insect groups and their nutritional reliance on vertebrate blood. They thus provide a unique system, potentially useful for distinguishing specific adaptive modifications of the genome from the general degenerative processes. For example, apart from genome reduction, which by itself is hardly surprising as it is a trait generally shared by the obligate symbionts, the two symbionts display high similarity in the pattern of the present/absent orthologous genes (Fig. 4). However, despite this overall similarity, the degree of genomic changes differs across the COG categories (see Tables S2 and S3 in the supplemental material), indicating a possible role of adaptive processes. Two patterns may be of particular interest. The first of them is exemplified by the COG category coenzyme transport and metabolism, for which the two hippoboscoid symbionts possess the highest number of genes among all included obligate symbionts. It is interesting that the high gene content of this category is also found in otherwise strongly reduced genomes of “Ca. Riesia,” a symbiont from sucking lice, and mutualistic Wolbachia from bed bugs. This makes the category a promising candidate for the traits possibly important in blood-feeding hosts. A second important pattern is provided by categories where the two symbionts differ considerably from each other in the numbers of present genes. These categories are useful in discriminating the traits that are not associated with the common lifestyle. From this perspective, the difference in the COG category motility is of particular interest. It has been speculated previously that the flagellar system present in the Wigglesworthia genome is necessary for transmission of the symbiont via the host milk glands or to enable entry into the larval or pupal bacteriocytes (52). However, in “Ca. Arsenophonus,” which uses the same transmission route (Fig. 3a and c; see also Fig. S2 in the supplemental material), this COG category is entirely absent. While this finding does not exclude use of the flagella for the transmission in Wigglesworthia, it at least indicates that this motility apparatus is not necessary for such a transmission mode. Another striking example for the “Ca. Arsenophonus”-W. glossinidia comparison is the category energy production and conversion, where the difference of 18 genes is largely due to the presence of an Rnf/Nqr complex and ferredoxin in the “Ca. Arsenophonus” genome and its absence in W. glossinidia. Similar differences between the two symbionts are also found in some other COG categories, e.g., defense mechanisms, signal transduction mechanisms, and replication, recombination, and repair (Tables S2 and S3 in the supplemental material). The number of genes in the last category is much higher in the “Ca. Arsenophonus” genome (as well as the set of genes with unknown functions). This is in good agreement with the perception of “Ca. Arsenophonus” as a less degenerate symbiont.

To interpret these observations on a rigorous comparative basis, it will be important to distinguish the variability due to different physiological demands (i.e., adaptive processes) from that reflecting evolutionary age of a given symbiont and/or stochastic processes. This can only been achieved in a broader phylogenetic framework of the hippoboscid-Arsenophonus associations using large data sets and proper statistical analysis. The polyphyly of the Hippoboscidae-Arsenophonus symbiosis, revealed by 16S rRNA gene- and MLST-based phylogenies (21, 53), suggests its multiple independent origins. This could explain why the genome of “Ca. Arsenophonus” is slightly larger than that of W. glossinidia if “Ca. Arsenophonus” is considered a younger symbiont which originated more recently from a free-living bacterium (possibly replacing other, more ancient symbiont of hippoboscids). In theory the higher number of “Ca. Arsenophonus” genes could also be due, at least partially, to gene acquisitions via horizontal gene transmission and/or gene duplications. A simple check showed that a majority of the 725 “Ca. Arsenophonus” protein-coding genes returned the best or second-best BLAST hits corresponding to the genus Arsenophonus (664) or the closely related genera Providencia, Proteus, Photorhabdus, and Xenorhabdus (16). Eighteen hypothetical proteins did not return any hit; for 27 of the remaining genes phylogenetic affinity is less clear and will require more detailed phylogenetic analysis.

Biosynthesis of B vitamins is traditionally considered a core aspect of the associations between bacteria and blood-sucking insects (13, 43, 54, 55). In our comparison, the arrangement of thiamine biosynthesis and transport presents one of the unexpected metabolic differences between the tsetse fly and louse fly symbiotic systems. In tsetse flies, the genome of W. glossinidia codes for the thiamine biosynthetic pathway, and S. glossinidius possesses only its ABC transporter (52, 56, 57). In M. ovinus, it is completely reversed. Our current data are not sufficient to infer the exact role of thiamine in the system; rather they allow for several not mutually exclusive hypotheses: (i) thiamine is provided to M. ovinus and/or “Ca. Arsenophonus”by “Ca. Sodalis” and/or B. melophagi; (ii) “Ca. Sodalis” and B. melophagi use thiamine solely for their own metabolism, but it is also available from blood for “Ca. Arsenophonus”; (iii) thiamine supplemented by the symbionts is beneficial but not necessary for host survival (during the whole life cycle or only in certain life stages). Similar hypothetical scenarios may apply to pantothenate biosynthesis: either it is provided by “Ca. Sodalis,” or it is available from other sources, i.e., sheep blood. This arrangement of the B-vitamin biosynthesis genes and the implied hypotheses indicate that the nutritional interactions between bacteria and blood-feeding insects may be more complex (less determined) than usually supposed. From this perspective, it is interesting that the only other obligate symbiont known to code for thiamine transporter is another member of the Arsenophonus clade, “Ca. Riesia pediculicola” (58). Since “Ca. Riesia” is an exclusive symbiont in human lice, thiamine is likely available from blood and needed by the symbionts for their own metabolism. This uncertainty further highlights our limited knowledge on symbiotic systems of blood feeders. Unlike the symbiotic associations in sap-sucking insects (where symbionts provide their hosts with well-known sets of essential amino acids [59]), it is still quite unclear which B vitamins are provided to blood-sucking hosts and which are needed only for symbionts, with the exception of pyridoxine and thiamine provisioning in the tsetse fly model (11, 14, 15).

The system presented here, i.e., the bacteria regularly associated with M. ovinus, is an example of the diverse symbiotic associations known from other hippoboscoids. As noted above, the presence of Arsenophonus bacteria in different Hippoboscidae members is likely due to multiple origins of the symbiosis or perhaps to recurrent replacements of the obligate symbionts by facultative Arsenophonus bacteria (21). The affinity to Arsenophonus symbionts is also well demonstrated in a related hippoboscoid group, the bat flies of the family Nycteribiidae (44, 60). Similarly, two other bacterial genera reported here from M. ovinus, namely, Bartonella and Wolbachia, have also been previously identified in other Hippoboscoidea (61–65). Melophagus ovinus is, however, the only host known so far with fixed infection of Bartonella (51; also this study). Considering this diversity and the abundance of the hippoboscoid-associated symbionts, together with the results presented here for a single hippoboscoid species, this group seems to be a valuable model for further comparative analyses of the insect-bacterium evolution.