Background

Wolbachia, frequently present in invertebrate microbiomes, provide a unique example of diversity and phenotypic flexibility found within a single monophyletic group of bacterial symbionts. Originally described as a causative agent of cytoplasmic incompatibility in Culex pipiens mosquitoes, the genus is today known to be widely distributed component of many arthropod and some nematode microbiomes [1]. The diversity of Wolbachia lifestyles spans from parasites to obligate mutualists. Phylogenetically, the genus forms several distinct clusters, usually called supergroups [2] which have developed their own characteristic features and tendencies. For example, while most Wolbachia supergroups seem to be specific to arthropods, few are reported exclusively from filarial nematodes. A particularly interesting group is supergroup F, the only supergroup known to infect both arthropods and nematodes [2,3,4]. Moreover, some members of this supergroup are highly adapted strains with degraded genomes, which maintain a mutualistic relationship with their hosts [5, 6].

The rapidly growing number of Wolbachia genome assemblies now allows for evolutionary and functional comparisons and identification of the characteristics underlying different life strategies [4, 7]. However, the distribution of the available data across the supergroups and host taxa is extremely uneven and biased. The recent meta-analysis performed by Scholz et al. [1] included an impressive number of 1166 Wolbachia genomes or genome drafts, but the majority of them (1018) originated from dipterans, almost exclusively from Drosophila (1011). Similarly, regarding the taxonomic diversity of Wolbachia, 1055 of the genomes represented supergroup A, while only 11 belonged to supergroup F and originated from three different hosts. This uneven distribution of genomes is likely to reflect the difference in attention paid to model and non-model organisms, rather than the real diversity of Wolbachia. Occasional screenings suggest that the supergroups underrepresented by genomic data may encompass a high diversity of Wolbachia strains. As an example, supergroup F, currently represented by four genomes and genome drafts from two nematodes and two arthropods [1, 8,9,10], seems to contain a wide variety of Wolbachia from different hosts when screened for specific Wolbachia genes [2, 11,12,13].

Phthiraptera belong to the insect taxa which have been screened specifically for the presence of Wolbachia symbionts and seem to be frequently infected. Kyei-Poku et al. [14] performed a PCR-based screening of 19 species, encompassing both sucking lice of the suborder Anoplura as well as the chewing lice of the suborders Amblycera and Ischnocera. They showed that all the tested samples produced specific Wolbachia markers, in some cases suggesting the occurrence of multiple strains. Since the screening was based on specific phylogenetic markers, genomic data is not available for these symbionts. It is therefore difficult to hypothesize on the nature of these symbiotic relationships and role of these Wolbachia in the host microbiomes. This is in sharp contrast to symbionts which originated from other bacterial taxa, particularly within gammaproteobacteria, where several complete genomes are available. From Anoplura that feed exclusively on vertebrate blood, obligate symbionts of different phylogenetic origins have been characterized, and for some, their role in provisioning B vitamins has been demonstrated [15,16,17,18,19,20,21]. Of the sucking lice included in the Kyei-Poku et al. [14] screening, this is the case for Riesia in Pediculus and Phthirus [17, 22], and Legionella polyplacis in Polyplax serrata [20, 21]. Since the mutualistic role of these symbionts is well established, it is likely that Wolbachia do not play a nutritional role in these lice. They may rather be accompanying commensals or even parasites, as shown in many other insects. In chewing lice, the situation is less clear. These ectoparasitic insects are likely not a monophyletic group [23] and their feeding strategies are more diverse than in Anoplura. While all chewing lice are ectoparasites living in the fur or feathers of their hosts, the source of food varies among the groups, and in several cases, their diet may also include host’s blood [24, 25]. Currently, a single genome is available for a chewing louse symbiont [26]. This symbiont, described from slender pigeon louse Columbicola wolffhuegeli, belongs to gammaproteobacteria and is phylogenetically related to the genus Sodalis within Gammaproteobacteria. Its genomic characterization revealed features resembling other obligate symbionts in insects, namely strong size reduction and shift of GC content (797,418 bp, 31.4% of GC), but did not provide any clear evidence for its function in the host [26].

In this study, we screened the microbiomes of Menacanthus eurysternus sampled across four continents and investigated the nature of Wolbachia associations. While the study encompasses several other chewing louse species, the primary focus lies in the widely distributed Menacanthus eurysternus [27] from a broad spectrum of passeriform and several piciform bird species [25]. This allows us to test the obligate nature of M. eurysternus-associated Wolbachia across a broad range of samples. To assess the genomic characteristics and capacity of the symbiont, we assemble metagenomic data from M. eurysternus and reconstruct the complete genome of its Wolbachia symbiont. We use fluorescent in situ hybridization to demonstrate the symbiont’s localization in the host’s specialized cells. Finally, for comparative reasons, we assemble additional genome drafts from several chewing lice samples with metagenomic data available in the SRA database [28].

Methods

Material and DNA extraction

Samples of Menacanthus eurysternus were collected across a large geographic distribution (Supplementary data 1) from 2000 to 2016. For 16S rRNA gene amplicon analysis, DNA templates were extracted from 54 individuals using QIAamp DNA Micro Kit (Qiagen) (Supplementary data 1). DNA template for metagenomics was isolated from pool of 8 individuals collected from one specimen of Fringilla coelebs morelatti GA72. To avoid environmental DNA contamination, lice were washed with pure ethanol (3 × for 30 min) in Mini-rotator (Bio RS-24) and DNA was extracted with QIAamp DNA Micro Kit (Qiagen). Concentration of the isolate was quantified with a Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA) and the integrity of DNA was verified on agarose gel electrophoresis (1,5%). NEBNext® Microbiome DNA Enrichment Kit (New England BioLabs) was used for increasing the proportion of bacterial DNA (via the procedure of selective binding and disposing of methylated host DNA). Final DNA concentration was quantified with a Qubit 2.0 Fluorometer using High Sensitivity reagents.

16S rRNA gene amplicon sequencing and analysis

The diversity and distribution of microbial associates in Menacanthus eurysternus samples were assessed using a 16S rRNA gene amplicon sequencing protocol developed by our group [29]. Briefly, multiplexing was based on a double barcoding strategy using fused primers with 12-bp Golay barcodes in forward primer 515F, and 5-bp barcodes within the reverse primer 926R [30, 31]. An 18S rRNA gene-blocking primer (Brown et al. [2]) was involved in all PCR reactions to ensure sufficient yields of 16S rRNA gene amplicons from the metagenomic templates.

M. eurysternus samples were part of a highly multiplexed library containing 384 samples altogether. In order to control for amplification bias and contamination, two positive controls (commercially purchased mock communities ATCC® MSA-1000™ and ATCC® MSA-1001™), and two negative controls for PCR amplification were processed along with M. eurysternus samples (complete metadata including barcodes are available in Supplementary data 1). The purified library was sequenced on Illumina Miseq using V2 chemistry with 500 cycles (Norwegian High Throughput Sequencing Centre, Department of Medical Genetics, Oslo University Hospital).

The raw fastq data were processed into the OTU table with an in-house workflow combining Usearch [32] and Qiime 1.9 [33] scripts as described previously [29]. Taxonomic classification was assigned to individual OTUs using BLAST searches of representative sequences against the SILVA 138 database (as of February 2021). Non-bacterial OTUs and potential contaminants found in the negative controls were cleaned from the data via a series of decontamination processes using different levels of stringency to evaluate the overall pattern of Wolbachia dominance and ubiquity in M. eurysternus microbiomes. While the less stringent decontamination involved eliminating 12 OTUs shared by both negative controls, the strict decontamination removed every OTU found the negative controls (35 OTUs altogether). The details on the control profiles and eliminated OTUs can be found in Supplementary data 1. The decontaminated datasets were rarefied in 5 iterations at a level of 1000 and 2000 reads and imported into RStudio [34] using the phyloseq package [35]. Compositional heat maps were produced for the 20 most abundant OTUs and ordered to reflect the phylogenetic relationship among analyzed M. eurysternus samples (complete COI phylogeny available in Supplementary Fig. 1).

Metagenomic sequencing and assembly (Menacanthus eurysternus)

The shotgun genomic libraries were prepared from the enriched gDNA of M. eurysternus GA72 sample using the Hyper Library construction kit from Kapa Biosystems (Roche). The library was sequenced in a multiplexed mode on one SP lane of NovaSeq 6000 for 251 cycles from both ends of the fragments. Fastq files were generated and demultiplexed with the bcl2fastq v2.20 Conversion Software (Illumina). The quality of 145,910,396 paired reads was checked with FASTQC and the data were trimmed by the Bbduk tool (https://sourceforge.net/projects/bbmap/) to a minimal phred score of 20. Spades with the option –meta was used to assemble the metagenome. Initially, Wolbachia contigs were identified by blastn searches [36] using the complete set of genes from the Cimex lectularius symbiont wCle (ACC) as a query. Wolbachia origin of the preselected contigs was verified by blastn searches against the NCBI nt database. The amplicon analyses indicated the presence of two different Wolbachia strains. Based on the considerable length difference between the first contig and the rest of Wolbachia contigs (732,857 bp vs 26,534 and less) and different GC contents (28% compared to 33%), we hypothesized that the first contig may be an almost complete genome of one strain, while the others represent a second strain. To test this possibility by closing the genome of the first Wolbachia strain, we used two approaches. First, we extended the longest contig by aTRAM 2.0 (an assembler capable of extending existing contigs using the original set of the pair-end reads; [37]) and closed it into a 733,850-bp long circular sequence. Second, using specific primers designed based on the longest contig, we sequenced the missing part and completed the genome into a circular sequence identical with the result from aTRAM-based approach. The remaining 189 contigs were considered as parts of the second strain of Wolbachia.

Fluorescent in situ hybridization (FISH)

In order to reveal the localization of Wolbachia within M. eurysternus body, we collected 15 individuals from Sturnus vulgaris hosts captured around Mysnik pond, Czech Republic (49°14′16.8″N 16°06′12.6″E). Prior to the FISH procedure the population was screened for the presence of the two Wolbachia strains using conventional PCR. The specific primers for each strain (W551F: 5′-GTA AGT TAA AAG TGA AAT CCC AGA GC-3′ and wMeur1_1388R: 5′-TTG CGG TTA GGT TAT TAG TTTT GAG-3′; wMeur2_145F: 5′-AAT AAT TGT TGG AAA CGG CAA C-3′ and W495R: 5′-GCA CGG AGT TAG CCA GGA-3′) were designed based on the full-length 16S rRNA gene sequences retrieved from our metagenomic assemblies and their specificity was validated using selected DNA templates with known microbiome profile. For the lice samples intended for FISH, the cuticle was teared with fine forceps to allow the 4% paraformaldehyde fixative to penetrate the tissues and incubated in 4 °C for 24 h. To lower the autofluorescence, fixed samples were treated with 2% hydrogen peroxide ethanol solution for 2 days and further incubated for 2 weeks in − 4 °C in 6% hydrogen peroxide ethanol solution that was repeatedly (every 2 days) exchanged for a fresh one. The FISH procedure followed previously published protocol [21] involving regular rehydration in an ethanol series, washing, and prehybridization. The hybridization step (46 °C, hybridization buffer containing 0.9 M NaCl, 20 mM Tris–HCl of 7.2pH, 0.1% SD, and 0.25uM probes) was adjusted to 48 h as recommended for less accessible parts of 16S rRNA molecule [38]. Since the sampled population only harbored the Wolbachia strain with the reduced genome (see above wMeur1 primer designation), we used a combination of Wolbachia general (three fluorescein-labeled probes) and Cy3-labeled long probe specifically hybridizing with the reduced Wolbachia strain. The probes included Flc-labeled Wolbachia W2 (5′-CTTCTGTGAGTACCGTCATTATC-3′; [39]) wCle modified probe TsWol944R (5′-AAC CGA CCC TAT TTC TTC A-3′; [6]) and modified TsWol1187R (5′-CTC ACG ACT TCG CAG CCT A-3′; [6]), and the newly designed Cy3-labeled 255_278Meur1 (5′-TAGTCTTGGTAGGCCATTACCCCAC-3′; five mismatches between the sequenced strains are highlighted, the numbers designate the positions in E. coli reference sequence used in [38]). As the negative control, a combination of Flc-labeled Wolbachia W2 probe (0.25 μM) and unlabeled W2comp (5′-GATAATGACGGTACTCACAGAAG-3′) oligonucleotides (2.5 μM) were used for hybridization to suppress the fluorescent signals (results not shown). The hybridized samples were washed and mounted on slides with Vectashield (Vector Laboratories) containing 4′,6-diamidino-2-phenylindole (DAPI) and kept in dark at 4 °C until observed on laser scanning confocal microscope Olympus FV3000 (Olympus).

Screening and assembly of chewing lice SRA

To check for the presence of Wolbachia in other chewing lice, we screened the available metagenomic data in SRA (Leinonen et al. [28]) (Supplementary Table 1). Assembling of the reads and detection of Wolbachia were done in the same way as for M. eurysternus (described above). To identify candidate Wolbachia contigs, we used two genomes as queries, wCle and the newly assembled complete genome from M. eurysternus. Of five assemblies in which Wolbachia contigs were identified, two contained only a few short contigs with low coverage and were not included in the subsequent analyses (Supplementary Table 1). For the remaining three assemblies we extracted Wolbachia genome drafts with different degree of fragmentations (from 9 contigs in Meromenopon meropis to 386 in Alcedoecus sp.). Since their sizes did not deviate from the common size of the other Wolbachia genomes and the completeness assessed by BUSCO version v5.2.1 [40] was also comparable to other Wolbachia (Supplementary Table 2; see below for BUSCO analyzes), we considered these sets of contigs as representative genome drafts.

Completeness assessment and annotations were done for both M. eurysternus strains and the three SRA-derived strains by the same procedure. Completeness was assessed in BUSCO with two different references, rickettsiales_odb10.2019–04-24 and proteobacteria_odb10.2019–04-24. Functional annotations of the genes were obtained by RAST [41]. The presence of phage-related sequences was further checked by PHASTER [42]. Potential pseudogenes were identified by Pseudofinder [43] based on annotation obtained from Prokka [44]. Possible horizontal gene transfers (HGT) were identified by diamond blastx [45] against NCBI nr database with a complete set of annotated genes as queries, e-value set to 10, and number of hits to five. Assignment of the genes to clusters of orthologous groups (COGs) was done in web-based eggNOG-Mapper [46]. To visualize sharing of the genes across the supergroup F strains, we plotted the results of the Orthofinder [47] analysis by UpSetR package of R [48].

Phylogeny

To determine the position of the new strains, we designed two different matrices. First, since our preliminary analyses suggested that the new strains belonged to supergroup F, the “fbpA_coxA” nucleotide matrix was built to represent this supergroup. The matrix contained 48 F strains, and 23 additional strains representing other supergroups. The genes were retrieved from NCBI (https://www.ncbi.nlm.nih.gov/), pubMLST [49], and our new assemblies (Supplementary Table 3). The alignment was done in MUSCLE [50]. The web-based IQ-TREE tool [51] was used to select the best models (TN + F + I + G4 and TVM + F + G4 models for fbpA and coxA respectively) and to perform the phylogenic analysis. To verify the position of the new strains within supergroup F, we designed an amino acid “multigene matrix,” restricted to the strains for which genomic data are available. This set contained all available genomes for supergroup F and several additional genomes representing other major supergroups (Supplementary Table 3). For the included genomes, we identified 101 shared single-copy orthologs by Orthofinder, aligned them in MUSCLE, and removed the unreliably aligned sites by Gblocks [52]. The final concatenated matrix containing 23,218 positions was analyzed by two different approaches. Maximum likelihood analysis was done in IQ-TREE with HIVb + F + I + G4 selected as best model. Since our data contained several long-branch sequences we used in addition PhylobayesMPI [53] with the CAT-GTR model to minimize possible artifacts [54]. This analysis was run for 50,000 generations under two different coding systems, first coding for each amino acid and second with amino acids recoded by the Dayhoff6 system.

Genomic and metabolic comparisons

Genomic analyses and comparisons were done for the nine strains of supergroup F for which complete genomes or drafts are available (Supplementary Table 3). Average nucleotide diversity (ANI) was calculated using a web-based ANI calculator [55]. Synteny of the genomes was analyzed in Mauve [56] as implemented in Geneious [57]. Assessment of metabolic capacities was done using the web-based tools Blastkoala and KEEG mapper [58]. To obtain a metabolic overview comparable with other Wolbachia supergroups, we adopted the scheme used by Lefoulon et al. [10] and extended its content with a comparison of amino acid synthesis.

Auxiliary verification of the wMeur1 genome

Since the analyses summarized above revealed a unique nature of the wMeur1 genome, particularly its strong reduction, we decided to verify its size by Oxford-Nanopore technology. To avoid risk of highly fragmented DNA of the ethanol-preserved samples used for the Illumina sequencing, we collected fresh samples and stored them in liquid nitrogen prior processing. Lice were sent to Roy J. Carver Biotechnology Center (University of Illinois, Urbana, USA) for extraction of HMW DNA, construction of the UltraLow library, and Oxford-Nanopore sequencing. 6.4 ng HMW gDNA from a single louse specimen was sequenced on a GridION flowcell, producing 22 Gb with read lengths between 4 and 28 kb (mean of 7676 bp). The basecalling was performed with Guppy 6.1.5 (http://staff.aist.go.jp/yutaka.ueno/guppy/). The reads were trimmed with Filtlong (version 0.2.1; https://github.com/rrwick/Filtlong) to remove sequences shorter than 4000 bp and to preserve reads with a phred score of at least 20. The program Flye version 2.9 [59] was used to assemble de novo metagenome assembly with the expected genome size of 200 Mbp. The resulting assembly was polished once with racon [60] and twice with medaka (version 1.6.1; https://nanoporetech.github.io/medaka). The polished assembly was checked for quality with BUSCO.

Results

Amplicon screening of M. eurysternus

On average, 16S rRNA gene sequencing yielded 6686 reads per sample under less stringent decontamination and 5911 reads under the strict decontamination (see “Methods” section). The mock communities yielded, on average, 20,876 reads for the equally composed samples and 48,356 for the staggered communities. We were able to recover the expected profiles for both equal and staggered DNA template, including overrepresentation of Staphylococcus epidermidis (ATCC 12,228) and vast underrepresentation of Rhodobacter sphaeroides (ATCC 17,029) reported previously by the manufacturer (Supplementary data 1). Within the staggered communities, we retrieved all three extremely low abundant taxa (0.04%, Supplementary data 1). The presence of an eleventh OTU of the genus Granulicatella however pointed out marginal (tens of reads) well-to-well contamination between positive and negative controls (details in Supplementary data 1). In all the datasets decontaminated under different stringencies (see “Methods” section, and Supplementary Fig. 2), Wolbachia OTUs clearly dominate M. eurysternus microbiomes. While the analyzed individuals generally associate with a single Wolbachia (OTU2), some show a dual Wolbachia infection. However, absence of any correlation pattern indicates that the number of Wolbachia OTUs is not determined either by geographic or by the phylogenetic origin of the analyzed hosts (Fig. 1). Although few individuals in Fig. 1 seemingly lack Wolbachia, these samples did not meet the rarefaction thresholds of 1000 and 2000 reads in the strictly decontaminated dataset (Supplementary Fig. 2) thus confirming a robust pattern of Wolbachia ubiquity across diversified populations of M. eurysternus.

Fig. 1
figure 1

Compositional heat map produced for the 20 most abundant OTUs. The samples are ordered to reflect the phylogenetic relationships among analyzed M. eurysternus samples (complete COI phylogeny available in Supplementary Fig. 1), and color coded according to their geographical origin

Full size image

Metagenomic assemblies and genomes characterization

The meta-assembly of the M. eurysternus data (sample GA72) contained two different Wolbachia strains (Table 1). One strain (designated as wMeur1) was assembled into a single 732,857 bp long contig and closed into a complete circular genome using aTRAM extension to 733,850 bp. An identical sequence was obtained using PCR with specific primers. The size was also confirmed independently by the Oxford-Nanopore sequence technology. The assembly produced by the Flye assembler contained a single 734,125 bp long contig, 275 bp longer than the Illumina-derived assembly (Supplementary Fig. 3A). The two assemblies shared 98% identity and produced one continuous colinear block in Mauve (Supplementary Fig. 3B). The Nanopore dataset also contained individual long reads (up to 19 kb long) which bridged connection between 5′ and 3′ ends of the Illumina-derived assembly (Supplementary Fig. 3C–E).

Table 1 Comparison of the main genomic characteristics of the nine Wolbachia strains from supergroup F
Full size table

The second strain, wMeur2, was fragmented into 189 contigs. Screening of the 36 chewing lice metagenomic data available in the SRA database revealed additional three strains of Wolbachia, designated hereafter as wAlce, wPaur, and wMmer from Alcedoecus sp., Penenirmus auritus, and Meromenopon meropis, respectively (Table 1). These genomes could only be assembled as drafts, composed of 9 (wMmer) to 386 (wAlce) contigs. The BUSCO assessments indicate that these fragmented genomes are complete or almost complete (SupplementaryTable2). When assessed against Rickettsiales database, the average completeness was 95% (85.7–98.7%). The assessment against Proteobacteria, performed to provide direct comparison with Lefoulon et al. [10], produced considerably lower average of 78.9% (68.5–83.5%) corresponding to the results of Lefoulon et al. [10]. For the wAlce strain, BUSCO predicted a higher degree of possible duplications, indicating that this assembly may not contain a single Wolbachia strain but could rather be a mixture of two closely related strains. wMeur1 and wMmer display unusual, derived features. Particularly, wMeur1 is only 733,850 bp long with GC content of 28%. Based on RAST annotation it contains 592 protein-coding genes, 3 rRNA genes, and 35 tRNAs. wMmer genome is considerably longer (1,005,754 bp when concatenated) but with similarly low GC content (28.6%). Both genomes form long branches in phylogenetic trees (Fig. 2, Supplementary Figs. 4 and 5) and possess characteristics typical for obligate symbionts.

Fig. 2
figure 2

A Phylogenetic relationships within the supergroup F derived for the available genomes (“multigene matrix” analyzed by PhyloBayes). The new strains printed in bold. Posterior probabilities of the nodes are indicated by the colored dots. Super Supergroups are designated by the capital letters at the branches or clusters. B Average nucleotide identity among the supergroup F genomes

Full size image

Besides the low GC content, they do lack cinA-cinB operon, any transposase sequences, phage-related sequences, and mobile elements. Ankyrin repeats are not present in wMeur1 and only one instance was detected in wMmer. The genomes of the remaining three strains resembled the previously reported genomes of supergroup F, containing transposase sequences, phage-related sequences, mobile elements, and ankyrin repeats (Table 1).

Orthofinder placed most of the protein-coding genes from the five new strains into orthogroups shared with the other included strains from supergroup F. Overlap comparison between the genomes showed a high proportion of genes shared by all or most strains (Fig. 3; Supplementary data 2). Most of the Wolbachia MLST markers were present in the new genomes but in several cases not recognized and annotated by Prokka (Supplementary data 3). The comparison also revealed that wMeur1 and wMmer do not share any unique genes in the exclusion of other genomes, despite their close phylogenetic relationship. On the other hand, while sharing a high proportion of genes, the strains displayed a very limited degree of synteny (Supplementary Fig. 6). For example, Mauve analysis of closely related wMeur1 and wMmer produced 159 local collinear blocks, the longest spanning 18 kb.

Fig. 3
figure 3

Orthogroups shared by the supergroup F Wolbachia genomes. Data for the new strains from chewing lice are printed in light blue

Full size image

Horizontal transfer of genes for pantothenate synthesis

Orthofinder also identified a set of genes, each unique for a single strain, mostly annotated as hypothetical proteins (Supplementary Table 4). A particularly interesting case is presented by three pantothenate synthesis-related genes (panB, panG, and panC) found exclusively in wMeur1. The only Wolbachia homologues found by diamond blastx were genes from wClefT (NZ_CP051156), a strain infecting the cat flea Ctenocephalides felis (Driscoll et al. [8]). The other closest relatives originated from phylogenetically distant bacteria, including several symbiotic forms. In the phylogenetic analysis of the closest homologues retrieved by blast from NCBI, all three wMeur1 pantothenate genes clustered as sister taxa to their wCfeT homologues (Supplementary Fig. 7).

In wMeur1, the three pantothenate genes were the only instances of apparent horizontal transfers. In the other four chewing lice strains of Wolbachia, the search for HGT did not reveal horizontally transferred genes with recognized metabolic function. Majority of the genes provided blast hits from other Wolbachia strains, while the few instances of the non-Wolbachia origin included mostly ankyrins, transposases, and hypothetical proteins (SupplementaryData3).

Comparison of metabolic capacities

Reconstruction of metabolic capacities shows high consistency across supergroup F genomes (Supplementary Table 5). The main differences are found in the wMeur1 genome and, to a lesser extent, in the wMmer genome. Both genomes lost a high portion of recombination genes and ABC transports. The smaller wMeur1 also lost genes for two B vitamin pathways retained in other strains (pyridoxine and folate) but acquired three genes for pantothenate synthesis by horizontal transfer (see above). All genomes show very limited capacity for amino acid synthesis. They only retain a near-complete pathway for lysine (leading probably to the peptidoglycan pathway rather than lysine) and the glyA enzyme allowing for interconversion of serine and glycine.

Phylogenetic relationships of the Wolbachia symbionts

All phylogenetic analyses of both matrices placed the newly described Wolbachia strains invariantly within the supergroup F (Fig. 2, Supplementary Figs. 4 and 5). The two most derived strains wMeur1 and wMmer formed extremely long branches within the supergroup, comparable only to those of the nematode-associated mutualists from the supergroups J. In ML analysis of the “multigene” matrix these two long branches clustered at the base of the supergroup F (Supplementary Fig. 5), while in PhyloBayes analysis they were nested among the other supergroup F taxa (Fig. 2). The remaining three strains, wMeur2, wAlce, and wPaur, were placed on considerably shorter branches, comparable to the rest of Wolbachia included in the analysis.

Wolbachia localization in M. eurysternus

Using wMeur1 specific Cy3 labeled probe, we have localized Wolbachia symbionts in M. erysternus larvae and adults (Fig. 4 and Supplementary Fig. 8). In the larvae, symbionts reside in host cells (bacteriocytes) forming paired aggregates adjacent to the crop that do not resemble any known structure in the body cavity of Menoponidae lice (Fig. 4). A similar localization was repeatedly recorded for wMeur1 cells in all analyzed M. eurysternus female individuals (Supplementary Fig. 8A).

Fig. 4
figure 4

Localization of Wolbachia wMeur1 symbionts in M. eurysternus larvae. The whole mount FISH on a larva with apparent bacteriocyte clusters adjacent to the crop (merged picture A; yellow is produced by the overlapping signal of the used probes: wMeur1-specific Cy3-labeled probe signals in red, Flc-labeled Wolbachia general probes (see the “Methods” section) in green are shown on the side panels). The detail of the bacteriocyte clusters visualizes from a break-open larva (B). Blue signals are produced by DAPI-stained nuclei

Full size image

In addition, wMeur1 was visualized in the reproductive system of both males and females (Supplementary Fig. 8B, C). In adult females, wMeur1 cells are being transmitted vertically to a developing egg where they are concentrated at the anterior pole (Supplementary Fig. 8B, C). For the whole mount individuals, we have only achieved partial quenching of autofluorescence produced by the chewing louse cuticle and the ingested keratin/chitin-rich diet.

Discussion

New highly derived members of supergroup F

The new Wolbachia supergroup F strains described here from four species of chewing lice represent two remarkably different types of symbiont genomes. Three of them (wMeur2, wPaur, wAlce) resemble the other Wolbachia strains of supergroup F in their genomic characteristics (size above 1 Mb, average GC content between 35 and 37%, presence of phage-related sequences, mobile elements, transposases, ankyrin repeats, etc.). This close similarity is also reflected in the short branches they form in the phylogenetic trees (Fig. 2, Supplementary Figs. 4 and 5). However, two additional strains (wMeur1, wMmer) show very specific and derived traits, unique within the context of the whole Wolbachia diversity. Particularly, wMeur1 represents the first highly reduced, insect-associated Wolbachia strain with characteristics typical for obligate mutualists known from other bacterial groups [20, 61, 62]. It also provides another candidate example of transition towards mutualism by horizontal gene transfer [20, 63]. This modifies the view that the arthropod-associated strains of Wolbachia generally possess larger genomes, richer with transposable elements, prophage-related genes or repeat-motif proteins than their nematode-associated relatives [4, 64]. With its 733,850 bp length wMeur1 is currently the smallest known genome among Wolbachia, almost 100 kb shorter than its nematode-associated “predecessors” wCtub and wDcau [10]. The second derived strain, wMmer, possesses a genome larger than wMeur1, but shows similar signs of strong degeneration: both strains lack phage-related sequences, mobile elements, transposases and ankyrin repeats (with exception of one ankyrin repeat found in wMmer).

Phylogenetic relationships

The highly derived nature of wMeur1 and wMmer genomes is also apparent from the low ANI values, when compared to their close relatives, and the long branches they form in phylogenetic trees (Fig. 2). In principle, such long branches may distort phylogenetic analyses. This phenomenon is particularly dangerous when analyzed taxa differ considerably in their nucleotide composition, e.g., when symbiotic genomes with extremely low GC content are included. In our study, the placement of these two symbionts within supergroup F is supported by several indications. First, this placement is not likely to be affected by long branches since no other members of supergroup F possess long branches which would attract the wMeur1 + wMmer pair. Second, the position of the two genomes as sister taxa has been retrieved with high support from all analyses, including the PhyloBayes analysis which is particularly resistant to this type of artifact [54].

Phylogenetic relationships within supergroup F suggest that at least the short-branched Wolbachia strains were acquired independently by their chewing lice hosts. For example, the two strains from philopterid lice, wAlce and wPaur, do not cluster as sister taxa in any of the analyses. While the tree resolution is rather poor and does not provide clear evidence, horizontal transfers within the F supergroup have been deduced previously, e.g., between isopods and termites [13]. This apparent lack of a coevolutionary signal is also concurrent with the absence of Wolbachia in all other screened SRA data for chewing lice (discussed below), and with the presence of a phylogenetically distant gammaproteobacterial symbiont in Columbicola wolffhuegeli [26, 65].

Distribution of the Wolbachia in chewing lice species

In our study, we found Wolbachia contigs in five out of the 36 assemblies of the SRA datasets. This is in sharp contrast with the results reported by Kyei-Poku et al. [14]. In their screening, focused on Wolbachia in lice, these authors showed the presence of these bacteria in all 19 tested species (interestingly, none of them from supergroup F, all falling into A and B). However, their approach was based on PCR amplification of selected genes using specific primers. It is likely that this method can detect Wolbachia even if present in extremely low numbers. In contrast, the WGS-based approach will only produce data for the dominant bacteria in the microbiome. This is also reflected in the amplicon analyses, which revealed considerably more bacterial taxa in each individual M. eurysternus sample (Fig. 1). This ambiguity raises the question, which of the Wolbachia previously detected in lice are parasites (known to be broadly distributed across arthropod taxa) and which may possibly represent the comparatively rarer instances of obligate mutualists.

Highly reduced wMeur1: transition to mutualism by HGT of pantothenate synthesis genes?

The genomic comparisons of the new strains are entirely consistent with their phylogenetic patterns. While the short-branched strains are closely similar in their genome characteristics, the two most derived genomes (wMeur1 and wMmer) differ in many aspects. Their GC content is significantly lower than in other analyzed strains (Table 1) and the majority of the other known Wolbachia (this characteristic being recognized as one of the typical features in the highly derived genomes; [66]). They both underwent considerable deterioration of the recombination and transport systems and unlike the other strains, they lack the elements related to the genomic dynamics and fluidity (phages, transposases, mobile elements). Such evolutionary trends accompanying the reduction in genome size are well known from many obligate insect symbionts from gammaproteobacteria [67] but are much less common in obligate Wolbachia, where the reduced genomes retain various mobile elements and phage-related genes. For example, several Wolbachia strains have been suggested to establish mutualistic relationships with their hosts after acquiring complete biotin operon [8, 63]. Within insect hosts, these systems include wCle in the bedbug Cimex lectularius, wCfeT in the flea Ctenocephalides felis [8], or two Wolbachia strains from Nomada bees [68]. In the latter, Wolbachia phylogeny even suggests co-divergence across several host species, typical for mutualistic obligate symbionts. However, in all these insects, Wolbachia symbionts retain genomes that exceed 1 Mb and contain many mobile elements. As pointed out by Driscoll et al. [8], wCle and wCfeT even possess extremely high numbers of pseudogenes (see Table 1 for wCle). The wMeur1 strain presented here differs from these insect symbionts by a dramatic reduction and “cleansing” of its genome. It is not only the smallest known genome among Wolbachia but also the first insect Wolbachia with genomic characteristics typical for obligate mutualists.

These features bring into question the role of the wMeur1 strain in its host. Unlike the above examples of presumably mutualistic Wolbachia, the wMeur1 strain does not possess genes required for biotin synthesis. Of the other vitamin B pathways, often considered to be of potential importance in nutritional symbionts, it only retains the production of riboflavin, a pathway conserved across many Wolbachia strains [69]. However, a striking metabolic difference between wMeur1 and all other supergroup F strains is the presence of three genes required for the synthesis of pantothenate, most likely acquired by horizontal gene transfer (HGT). The blast-based HGT analysis retrieved the most closely similar homologue from wCfeT, a Wolbachia strain from cat flea Ctenocephalides felis, while the other retrieved homologues belonged to other, often phylogenetically distant bacterial groups (Supplementary data 4). In the phylogenetic analysis, all three genes from wMeur1 and wCfeT formed closely related sister taxa (Supplementary Fig. 7). Remarkably, the same triad of genes is also present in the genome of a Sodalis-related symbiont from another chewing louse, C. wolffhuegeli (Fig. 5; NCBI BioProject PRJNA692390). When characterizing the metabolic capacity of this bacterium, Alickovic et al. [26] mainly addressed the issue of keratin digestion and concluded that no clear metabolic role can be deduced from the symbiont’s genome content. Similarly, in our new strains, we failed to detect the production of enzymes with keratinase activity, and we observed almost complete loss of capacity for amino acid synthesis (Supplementary Table 5). However, the retention of the three pantothenate-related genes in C. wolffhuegeli symbiont and their HGT acquisition by wMeur1 suggests that production of this vitamin might be at least part of the metabolic function in these putatively obligate mutualists.

Fig. 5
figure 5

Distribution of the pantothenate synthesis-related genes in the Wolbachia genomes. A The genes mapped on a schematic phylogeny tree. B Overview of the pantothenate synthesis pathway in Pediculus humanus and its symbiont Riesia pediculicola (reconstruction based on KEGG database)

Full size image

While the triad panB, panG, and panC form a core of pantothenate synthesis, full functionality of the pathway requires two additional genes, ilvE and panD (Fig. 5). These genes are missing in the wMeur1 genome but are very likely present in the genome of the host. The phylogenetically closest system with fully characterized genomic capacities is the symbiosis between human louse Pediculus humanus and its symbiont Riesia pediculicola [70]. In this blood feeding insect, Riesia provides the same three genes (panB, panG, and panC located on the plasmid) while ilvE and panD are present in the host genome (in Fig. 5 shown on simplified reconstruction based on the KEGG database). Our blast screening of the M. eurysternus assembly shows that these genes are also present in the wMeur1 host and suggests the complete pantothenate pathway is potentially functional.

According to the hypothetical scenario introduced by Lo et al. [71], symbiogenesis involves two bouts of dramatic genomic changes. The first occurs during the transition from free-living bacterium to a facultative symbiont and the second one with a transition to an obligate symbiont. During these processes, bacterial genomes first undergo a dramatic decrease of coding density due to large-scale inactivation of the genes. This step is followed by the removal of the inactivated genes and restoration of the coding density. Our set of new Wolbachia strains fit well into this scenario. As shown in Fig. 6 most of the supergroup F genomes display relatively high coding density between 78 and 84%. These genomes carry transposases and mobile elements, and with exception of wCle, also have a relatively rich repertoire of recombination/repair genes (Supplementary data 3). In contrast, the wMmer genome shows a considerable drop in coding density to 65%. While its size is comparable to the former strains, it contains a significantly lower gene number, indicating large-scale deactivation. Finally, the wMeur1 genome restores the coding density to 76%, apparently due to removal of its deactivated regions. Its position in this evolutionary spectrum, loss of most of the recombination/repair systems, presumed metabolic role in pantothenate provision, and bacteriocyte localization provide strong evidence that wMeur1 is the first known insect-associated Wolbachia strain which completed the transition to a putatively obligate mutualist.

Fig. 6
figure 6

Comparison of the supergroup F Wolbachia genomes in respect to coding density and numbers of genes in different COGs. A – RNA processing and modification, C – energy production and conversion, D – cell cycle control and mitosis, E – amino acid metabolism and transport, F – ucleotide metabolism and transport, G – carbohydrate metabolism and transport, H – coenzyme metabolism, I – lipid metabolism, J – translation, K – transcription, L – replication, recombination and repair, M – cell wall/membrane/envelop biogenesis, N – cell motility, O – post-translational modification, protein turnover, chaperone functions, P – inorganic ion transport and metabolism, Q – secondary structure, T – signal transduction, U – intracellular trafficking and secretion, V – defense mechanisms, S – function unknown, ambig – assigned to more than one category, n.a. – not assigned, co density – curve showing coding density. Dashed line highlights differences in the L category related to replication, recombination, and repair

Full size image

Conclusions

Screening of 37 chewing lice species revealed a relatively frequent presence of Wolbachia in their microbiomes. Genomic traits of the five new F supergroup strains correspond to symbionts in different stages of evolution, and well-illustrate hypothetical evolutionary trajectory towards the extremely reduced mutualist, represented here by the strain wMeur1. Finding of this strain with the smallest known Wolbachia genome, fixed worldwide in microbiomes of Menacanthus eurysternus across four continents, shows that despite the large number and great diversity of the known Wolbachia strains, the evolutionary potential of these bacteria still remains underexplored. Considering the diversity of the screened chewing lice microbiomes and the five reconstructed Wolbachia genomes, we suggest that this vast parasitic group may provide a suitable model for further investigations.