Comparative tissue-specific microbiome analyses identify keystone endosymbionts shaping pathogen interactions in dromedary camel ticks

Abstract

Background

Ticks are blood-feeding arthropods that carry diverse pathogenic and nonpathogenic microorganisms, yet their dynamics within tick tissues remain poorly understood. We compared the microbial communities in the haemolymph, saliva, ovaries, midgut, and salivary glands of individual Amblyomma gemma and Hyalomma rufipes sampled from dromedary camel using V1-V2 16S rRNA gene metabarcoding.

Results

The haemolymph exhibited the highest bacterial diversity, followed by the saliva and ovaries, while the midgut and salivary glands had lower diversities. Keystone taxa exhibited varied connectivity across different tissues and played a crucial role in structuring microbial communities in both tick species. We exclusively detected Rickettsia africae in Am. gemma and Rickettsia aeschlimannii in Hy. rufipes. Both Rickettsia species showed negative correlations with dominant endosymbionts and environmental bacteria across multiple tissues. Coxiella endosymbionts were detected solely in Am. gemma and were most abundant in the salivary glands, while Francisella endosymbionts were dominant in Hy. rufipes’ salivary glands. Francisella endosymbionts were less abundant when R. aeschlimannii was high. In Am. gemma’s haemolymph and saliva, Wolbachia endosymbionts were prevalent, where they were inversely associated with R. africae at the individual tick level. In Hy. rufipes, Wolbachia endosymbionts were present in the ovaries, but not in the midgut. Candidatus Midichloria mitochondrii was found in all tissues of Hy. rufipes and predominantly in saliva, but was only detected in the ovaries, midgut, and salivary glands of Am. gemma. Positive interactions were found between Ca. Midichloria mitochondrii and R. aeschlimannii in Hy. rufipes’ saliva and ovaries, and Ca. Midichloria mitochondrii and Francisella endosymbionts in Am. gemma’s ovaries. Rickettsiella spp. were found in the saliva and haemolymph but not the midgut in both tick species. Environmental bacteria, such as Pseudomonas, Acinetobacter, Staphylococcus, and Corynebacterium, whilst abundant in the haemolymph and saliva were scarce in other tissues.

Conclusions

Tick tissue significantly influences bacterial composition, with tissue-specific interactions between pathogens and endosymbionts suggesting bacterial functional specialization. The complex interactions observed between key endosymbionts and Rickettsia pathogens across tissues, including negative correlations with Coxiella, Francisella, and Wolbachia endosymbionts, and positive correlations between Ca. Midichloria mitochondrii and R. aeschlimannii, require further investigation. Identified keystone taxa may serve as targets for anti-microbiota vaccines.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12866-026-04864-5.

Background

Ticks are significant blood-sucking vectors, competent at spreading a wide range of pathogens that are harmful to both humans and animals [1]. Both their geographic range as well as the pathogens they carry are increasing [2]. The microbiomes of ticks are varied and comprise commensal, symbiotic, and pathogenic microorganisms, including bacteria, eukaryotes, and viruses [3, 4]. Ticks can carry and transmit multiple agents simultaneously [2], and their microbiomes can play fascinating roles in modulating the vector’s life cycle and the transmission of pathogens [5, 6]. Several variables shape the bacterial microbiome in ticks, including tick species, geographical location, developmental stage, feeding status, and tick tissues [7, 8]. This complex microbial community seems essential to the biology of ticks and pathogen transmission. For instance, as obligatory blood feeders, ticks rely on their bacterial endosymbionts such as Coxiella and Francisella for essential B vitamins, and several amino acids, which are lacking in the blood [9–11]. The interaction between the microbiome and tick immune systems may support or inhibit tick-borne pathogen infections and transmission [6, 11]. Manipulating tick microbiomes has been shown to reduce vector competence [12], suggesting that the identification of the key microbes associated with specific tick tissues could inform targeted anti-microbiota vaccine development. Such vaccines could potentially block pathogen development within these tissues [13]. However, to fully understand the role of individual bacterial species, it is crucial to consider them within their ecological context within ticks [14]. A good example is the Ornithodoros spp., which exhibit distinct microbiomes within salivary glands and midgut [13]. The identification of keystone taxa and their interaction across various tissues indicates that certain microbes could serve as targets for anti-microbiota vaccines, by impacting tick physiology and pathogen transmission [13]. Ticks acquire pathogens during feeding. The pathogens transit through the midgut before entering the haemolymph, which serves as a transport medium to other tick tissues. Pathogens may migrate to the ovaries, facilitating vertical transmission to offspring or invade the salivary glands, enabling spread to new hosts via the saliva during subsequent feeding [15, 16]. Thus, the potential interactions between tick vectors, pathogens, and microbiomes are an ecologically complex relationship [6].

Pastoral tribes in northern Kenya are estimated to rear over 4.7 million camels, about 80% of the nation’s total camel count [17]. This makes Kenya a major camel-rearing nation, accounting for 6% of all African camels [18, 19]. In northern Kenya, where most of the camel herds are reared, the animals are heavily infested with Amblyomma gemma and Hyalomma rufipes tick species, which transmit diverse tick-borne pathogens, posing a risk of exposure to humans and other livestock [20–22]. Spotted fever group rickettsioses caused by Rickettsia africae and Rickettsia aeschlimannii are of zoonotic relevance and have been previously reported in camels [23–25]. Additionally, Ehrlichia ruminantium, the causative agent of heartwater in ruminants, which is transmitted by Am. gemma, has been reported in humans [26]. In this study, we identified and compared the bacterial communities in five tick tissue/fluid types (saliva, haemolymph, salivary glands, midgut, and ovaries) of individual Am. gemma and Hy. rufipes ticks that were collected from camels in two distinct agroecological zones in northern Kenya. By using a V1-V2 16S rRNA gene metabarcoding sequencing approach, we investigated the diversity of bacterial microbiomes, their localisation within the tissues and interactions with pathogens, and identified keystone taxa, which may hold potential for anti-microbiota-based control strategies. Building on our earlier whole-tissue profiling of these tick species [8], here we apply co-occurrence network analysis at the individual tick level across five tissue types, including saliva and haemolymph, to identify keystone taxa and their interactions with Rickettsia pathogens, which may hold potential for anti-microbiota-based control strategies.

Methods

Study area

Marsabit County, situated in northern Kenya, is approximately 550 km north of Nairobi [27]. It is the largest County in Kenya, spanning an area of ~ 66,923 km2 between longitudes 37°57’ and 39°21’E and latitudes 02°45’ and 04°27’ N [28]. Predominantly inhabited by nomadic pastoralists, the County’s economy is based on nomadic livestock production, with camels, cattle, sheep, and goats being the main livestock types. We collected Am. gemma and Hy. rufipes ticks from dromedary camels (Camelus dromedarius) in camel ranches within two subcounties of Marsabit: Laisamis and Moyale (Fig. 1).

Fig. 1.

Study area and tick species. Map showing sampling sites in Marsabit County, Laisamis and Moyale Subcounties, northern Kenya. Dorsal views of Amblyomma gemma and Hyalomma rufipes collected from dromedary camels. Tick species were identified using standard taxonomic keys [29, 30]. Tick images reproduced from Khogali et al. [21] under CC BY 4.0 license

Ethical approval

Ethical approval for this study was obtained from the Pwani University Ethics Review Committee (ERC/EXT/002/2020E) and a license from the National Commission for Science Technology and Innovation (NACOSTI/P/22/16467). This study did not involve human subjects as participants; however, verbal informed consent was obtained from all camel owners prior to tick collection from their animals. Given the low literacy rates among the camel farmers, the procedure for obtaining verbal informed consent before sample collection was reviewed and approved by the Ethics Review Committee as part of the overall study protocol. The sampling process was conducted with the assistance of a veterinarian and local field assistants, who provided explanations of the study objectives and facilitated communication with the livestock owners. Although this study did not constitute medical research on human subjects, the ethical principles of the Declaration of Helsinki were upheld throughout, particularly with respect to informed consent and ethical review.

Tick collection and transportation

We collected 834 partially engorged adult ticks from 192 one-humped camels (Camelus dromedarius) during September and October 2023 in Marsabit County, northern Kenya. Ticks were stored in Falcon tubes covered with a cotton lid and wrapped in moistened cotton to maintain high relative humidity. The tubes were then placed in a Styrofoam box containing ice packs to preserve a low temperature and high relative humidity. All ticks were transported alive to the Martin Lüscher Emergence Infectious Diseases (ML-EID) laboratory at the International Centre of Insect Physiology and Ecology (icipe) in Nairobi.

Tick identification, dissection, and tissue collection

Ticks were morphologically identified to the species level using standard taxonomic keys [29, 30] under a Stemi 2000-C light microscope (Zeiss, Oberkochen, Germany). Representative images of ticks sampled from camels were previously published in Khogali et al. (2024) [21]. Haemolymph (HL) and saliva (SL) collection and tick dissection were carried out according to the described protocol [31] in the ML-EID laboratory at icipe, Nairobi, Kenya, using a Stemi 2000-C microscope (Zeiss, Oberkochen, Germany). Prior to dissection, ticks were sterilized by immersion in 1% bleach solution three times, followed by a final rinse with nuclease-free water [32]. Subsequently, salivary glands (SGs), midgut (MG), and ovaries (OVs) were isolated using the method outlined by Edwards et al. (2009) [33]. HL contamination was reduced by washing SGs, MG, and OVs with five sequential droplets of the phosphate-buffered saline. After that, we homogenized the SGs, MG, and OVs in sterile 1.5-mL Eppendorf tubes containing 750 mg of 2.0-mm stabilized zirconium oxide beads (Biospec, USA) using a Mini-Beadbeater-16 (BioSpec, Bartlesville, OK) for 1 min to disrupt the tissue mechanically.

DNA extraction, PCR, and library preparation

Total DNA was extracted from tick tissue homogenates and the SL and HL using the HighPrep™ Viral DNA/RNA kits (Magbio, Gaithersburg, USA) as per the manufacturer’s instructions. We amplified the V1-V2 region of the 16S ribosomal RNA (rRNA) gene from each sample type (SL, HL, SGs, MG, and OVs) to identify bacterial communities. The amplification process utilized the specific primers targeting the 16S rRNA gene V1-V2 (27F: 5’-GAGTTTGATCMTGGCTCAG-3’; 388R: 5’-GCTGCCTCCCGTAGGAGT-3’) region [8]. Negative controls were included throughout the workflow to monitor potential contamination. Two extraction blank controls were processed alongside samples during DNA extraction which were subjected to sequencing, and negative PCR controls were included in all amplification runs.

DNA amplicons were shipped to Macrogen Europe (Netherlands) for sequencing on an Illumina MiSeq platform. Library preparation adhered to Illumina’s standard protocol, incorporating overhang adapters to primers using 2x KAPA HiFi HotStartReadyMix. Briefly, in the first PCR, the DNA template was amplified using forward and reverse primers at a concentration of 1 µM. The thermal cycling conditions included initial denaturation at 95 °C for 3 min, then 25 cycles consisting of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s; and a final extension at 72 °C for 5 min. Amplicons were purified using AMPure XP beads to remove residual primers and primer dimers. In the second PCR, known as the index PCR, dual indices and Illumina sequencing adapters were attached to the amplified DNA using the Nextera XT Index Kit v2, following the manufacturer’s instructions. The thermal cycling conditions for this step included an initial denaturation at 95 °C for 3 min; followed by 8 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s; and a final extension at 72 °C for 5 min. Paired-end sequencing (300 bp) was conducted using MiSeq V3 chemistry.

Bioinformatics analyses

Amplicon sequence processing and microbial diversity analyses

The raw amplicon sequences generated by the Illumina MiSeq platform underwent preprocessing using the nf-core / ampliseq pipeline (v2.6.1). Raw sequence quality was assessed using FastQC, and primer sequences were identified, trimmed, and processed using Cutadapt (v4.1). The bioinformatics analysis pipeline was executed using Nextflow (v21.10.3) and Singularity (v3.6.3) to optimize processing efficiency [38]. The Divisive Amplicon Denoising Algorithm 2 (DADA2, v1.26.0) was employed to process the amplicon sequence data, resulting in the generation of an amplicon sequence variant (ASV) abundance table and corresponding taxonomic classifications. The amplicon sequence data were processed using the DADA2 pipeline [39]. The analysis steps included filtering, trimming, denoising, chimera removal, dereplication, read merging, ASV inference, and taxonomic assignment. Taxonomic classification of ASVs was performed using the Silva 138 database which was trained on the V1-V2 region using the 27F and 388R primers, and the Basic Rapid Ribosomal RNA Predictor (Barrnap, v0.9) [40]. Non-target and non-bacterial features were removed prior to downstream analyses. Particularly, ASVs annotated as chloroplasts (Order Chloroplast), mitochondria (Family Mitochondria), and Archaea (Kingdom Archaea) were excluded. ASVs assigned to the phylum Chloroflexi, which predominantly comprises environmental soil and water bacteria, were also excluded as likely non-target contaminants in these low-biomass tick tissue samples. ASVs lacking genus-level taxonomic assignment were also removed to focus downstream analyses on identifiable taxa. Finally, low-abundance ASVs were filtered out by retaining only taxa with a total count > 10 across all samples. Sequence processing and downstream analyses were conducted in R v 4.4.1 [41].

Microbial compositional and diversity analyses

We compared the microbial composition of five tick tissues (HL, SL, MG, OVs, and SGs) and the tissue sample diversities within individual ticks of two tick species (Am. gemma and Hy. rufipes) using heatmaps to measure abundance. The data comprised of adjusted mean counts converted using the centered-log ratio (CLR). Z-scores were utilized to standardize the CLR-transformed numbers, minimizing range fluctuation and improving visualization. Alpha diversity metrics, including observed richness and Shannon indices were calculated to assess microbial diversity. Rarefaction analysis based on ASV read counts was conducted using the Microbiota Process function (v1.9.3).

Beta diversity was assessed using Bray–Curtis dissimilarities and visualized by Principal Coordinates Analysis (PCoA). The effects of tick species and tissue type on bacterial community composition were tested using PERMANOVA (adonis2) with 9,999 permutations. To ensure that PERMANOVA results reflected true compositional differences rather than within-group variability, homogeneity of multivariate dispersions was evaluated using PERMDISP (betadisper) with 9,999 permutations prior to interpretation. Collectively, these analyses provided a comprehensive understanding of the complex microbial dynamics within the tick microbiome. All sequences were submitted to GenBank nr database of the NCBI (http://www.ncbi.nlm.nih.gov) under BioProject accession number: PRJNA1134173 (http://www.ncbi.nlm.nih.gov/bioproject/1134173).

Co-occurrence network analyses and identification of keystone taxa

The microbial taxonomic data were analysed in R using the following packages: Phyloseq for data manipulation, SpiecEasi for network construction [42], and igraph and ggraph for network properties and visualization. For keystone taxa identification, we calculated centrality measures including degree, betweenness, closeness, and eigenvector centrality. Bacteria with the highest degree of centrality have the most direct connections with other bacteria. High betweenness centrality suggests that specific bacteria may serve to connect distinct microbial subcommunities. Moreover, eigenvector centrality indicates bacterial connections to other influential nodes within the network, while closeness reflects proximity to other nodes, facilitating more influential interactions. Nodes ranked within the top 10% were classified as potential keystone taxa. Centrality values and corresponding ranks were used to identify potential keystone taxa.

Phylogenetic analysis

All samples were screened for Rickettsia spp. by targeting the outer surface membrane protein B (ompB) gene to generate longer sequences of 856 bp using the set of forward and reverse primers OmpB 120–2788 F: AAACAATAATCAAGGTACTGT and OmpB 120–3599 R: TACTTCCGGTTACAGCAAAGT, and thermal cycling conditions described by Getange et al. [20] (2021). The PCR products were sent for sequencing at Macrogen Europe (Amsterdam, The Netherlands). The sequences obtained were edited, trimmed, and MAFFT-aligned using Geneious Prime software v. 2020.2.2 (Biomatters, Auckland, New Zealand) [34]. Nucleotide sequences were queried against known annotated Rickettsia spp. sequences in the GenBank nr database of NCBI (http://www.ncbi.nlm.nih.gov) using BLAST nucleotide searches [35]. Maximum likelihood phylogenetic trees were generated for each gene in PhyML v. 3.0 using the best-fit model of sequence evolution identified under Akaike information criterion for automatic model selection [36]. The resulting phylogenetic trees were visualized using FigTree 1.4.4 [37].

Results

We aseptically harvested haemolymph (HL), saliva (SL), ovaries (OVs), midgut (MG), and salivary glands (SGs) from ten Am. gemma (n = 10; HL = 9, SL = 9, SGs = 10, OVs = 10, and MG = 10) and nine Hy. rufipes (n = 9; HL = 9, SL = 9, SGs = 9, OVs = 9, and MG = 9) by adhering to efficient dissection and tick tissue preservation protocols. DNA from one SL sample and one HL sample, both from Am. gemma, were not successfully sequenced due to low DNA quality.

Sequence analysis

After preprocessing 9,108,991 V1-V2 16S rRNA gene amplicon reads, we obtained 7,989,616 quality reads from the two species, yielding 6,025 amplicon sequence variants (ASVs). Of these, 15 chloroplast ASVs (612 reads), 86 Chloroflexi ASVs (2,224 reads), and 7 mitochondrial ASVs (132 reads) were removed, while no archaeal ASVs were detected. An additional 744 ASVs lacking genus-level assignment (38,901 reads) were excluded. Subsequent filtering of 1,256 low-abundance ASVs (7,605 reads) resulted in a final dataset of 3,917 ASVs (7,940,142 reads) retained for downstream analyses.

A total of 18 phyla, 33 classes, 48 orders, 176 families, and 483 genera were recovered from the 16S rRNA gene ASVs. Moreover, all ASV sequences classified as Coxiella spp., Francisella spp., and Wolbachia spp. corresponded to Coxiella endosymbionts, Francisella endosymbionts, and Wolbachia endosymbionts (WEs). Candidatus Midichloria spp., and Proteus spp. were identified as Candidatus Midichloria mitochondrii and Proteus mirabilis, respectively. BLAST analysis of representative sequences against the NCBI nucleotide database confirmed > 98% sequence similarity and identity to reference sequences of these taxa.

As shown in the OmpB gene phylogenetic tree, the Rickettsia sequences obtained in this study clustered with the reference sequences from R. africae and R. aeschlimannii within the broader Rickettsia genus. All R. africae sequences were exclusively detected in Am. gemma, while R. aeschlimannii sequences were only identified in Hy. rufipes (Supp. Figure 1). Samples positive for Rickettsia spp. OmpB gene, were consistent with V1-V2 16S rRNA gene sequencing results.

Microbial compositional and diversity analysis

Using heatmaps, we analysed the top 25 bacterial genera of Am. gemma and Hy. rufipes. Specific bacteria, including Rickettsia, Wolbachia, Coxiella, Francisella, Ca. Midichloria, Sphingomonas, Rickettsiella, Pseudomonas, Proteus, Paracoccus, Corynebacterium, and Acinetobacter were selected for further comparison across the five tick tissues and individual samples. We also performed alpha diversity indices (observed richness and Shannon index) and beta analysis (Principal coordinate analysis, PCoA) to infer the bacterial community diversity.

Tick species

Among the top 25 bacterial taxa Wolbachia spp., Coxiella spp., Rickettsia spp. and Corynebacterium spp. were more abundant in Am. gemma (Figs. 1B & C and 2A, Supp. Table 1). However, Proteus spp. and Francisella spp. were highly abundant in Hy. rufipes (Figs. 1D & E and 2A). Alpha diversity showed a significantly greater number of bacterial taxa in Am. gemma compared to Hy. rufipes (P < 0.001) (Fig. 2B and C). PERMANOVA revealed a significant effect of tick species on bacterial community composition (F = 6.84, R² = 0.13, p = 0.0001). However, a test for homogeneity of multivariate dispersion indicated significant differences in within-group variability among tick species (F = 7.69, p = 0.0006), suggesting that PERMANOVA results should be interpreted with caution as they may be partially influenced by dispersion effects. (Fig. 2D).

Fig. 2.

A Heatmap of CLR-transformed abundances of two tick species Amblyomma gemma and Hyalomma rufipes. Z-score has been used to identify colour scale range. Boxplot showing alpha diversity of the tick species using two diversity metrics, (B) Observed richness and (C) Shannon index. D Beta diversity of bacterial communities in the two tick species using principal coordinate analysis (PCoA)

Tick tissues

Within Am. gemma, Coxiella endosymbionts were highly abundant in tick organs, particularly SGs, followed by MGs and OVs, and less abundant in the HL and SL. However, Francisella endosymbionts were less abundant in Am. gemma compared to Coxiella endosymbionts, with the highest concentrations being in the SGs, followed by the OVs. Rickettsia africae, was less abundant in the HL and SL, but highly concentrated in the OVs and SGs. Candidatus Midichloria mitochondrii exhibited overall lower abundance, with the highest concentration in the OVs. Wolbachia endosymbionts and Rickettsiella spp. were highly concentrated in the HL, whereas Pseudomonas spp., Staphylococcus spp., Corynebacterium spp., and Sphingomonas spp. were highly abundant in the SL and HL (Fig. 3A, Supp. Table 2).

Fig. 3.

Heatmaps of CLR transformed abundances of (A) Amblyomma. gemma’s and (B) Hyalomma rufipes’ tissues (HL; haemolymph, SL; saliva, MG; midgut, OVs; ovaries, and SGs; salivary glands). Z-score has been used to identify colour scale range

In Hy. rufipes, Coxiella endosymbionts were entirely absent. However, Francisella endosymbionts and Ca. Midichloria mitochondrii were highly abundant in the SGs, followed by MG and OVs. Pseudomonas spp., Corynebacterium spp., Acinetobacter spp., and Staphylococcus spp. were highly prevalent in the HL and SL and less prevalent in the other tissues/fluid types. Pseudomonas spp. was distributed throughout all tick tissues, with higher concentrations observed in the HL, SL, and OVs. Proteus spp. was detected in all tick tissues with the highest abundance in MG, OVs, and SL. Rickettsia aeschlimannii was concentrated in the SGs and MG and less concentrated in the other tick tissues. Wolbachia endosymbionts were highly abundant in the OVs, while Rickettsiella spp. were most dominant in the HL (Fig. 3B, Supp. Table 2).

Alpha diversity analysis in Am. gemma revealed significantly higher bacterial compositions in the HL and SL compared to the MG (P < 0.01), and SGs (P < 0.01), while OVs exhibited significantly greater bacterial composition than the MG (P < 0.05) (Fig. 4A, Supp. Table 3). Furthermore, Am. gemma HL and SL had significantly greater bacterial diversities than their MG, OVs, and SGs (Fig. 4B, Supp. Table 3). In Hy. rufipes the analysis revealed that both HL and SL harbored significantly more bacterial taxa than the SGs and MG (P < 0.01), while OVs also contained significantly higher bacterial richness than the MG (P < 0.01) (Fig. 4C, Supp. Table 3). Moreover, HL demonstrated greater diversity than the MG (P < 0.05) and SGs (P < 0.001). The bacterial diversity of the SL was higher than that of the MGs (P < 0.01), OVs (P < 0.01) and SGs (P < 0.001), while the SGs had the lower diversity than the MGs (P < 0.01), and OVs (P < 0.001) (Fig. 4D, Supp. Table 3).

Fig. 4.

Boxplots showing alpha diversities of five tick tissue/fluid types using the observed richness and Shannon index of Amblyomma gemma (A and B), and Hyalomma rufipes (C and D). HL: haemolymph, MG: midgut, NC: negative control, OVs: ovaries, SGs: salivary glands, SL: saliva

The principal coordinate analysis (PCoA) showed a significant dissimilarity in bacterial composition among the five tissue/fluid types of (A) Am. gemma (PERMANOVA, P < 0.05) and (B) Hy. rufipes (PERMANOVA, P < 0.001) (Fig. 5).

Fig. 5.

Beta diversity of bacterial communities in the five tissue/fluid type using principal coordinate analysis (PCoA) (A) Amblyomma gemma, (B) Hyalomma rufipes). NC: Negative control

PERMANOVA based on Bray–Curtis dissimilarities revealed significant differences in bacterial community composition among the tissue types of Am. gemma (R² = 0.17, F = 1.71, P = 0.018) and Hy. rufipes (R² = 0.47, F = 7.14, P < 0.001). Tests for homogeneity of multivariate dispersions showed no significant differences in within-group variability of tissues of Am. gemma (PERMDISP; F = 1.03, P = 0.401) and Hy. rufipes (PERMDISP; F = 2.46, P = 0.055), indicating that PERMANOVA results were not influenced by dispersion effects.

Co-occurrence network analysis and keystone taxa identification

We performed co-occurrence network analysis and identified keystone taxa for the tick tissues (HL, SL, MG, Ovs, and SGs) of Am. gemma and Hy. rufipes to elucidate the complex relationships within the tick microbial communities. Considering the number of nodes, the microbial co-occurrence network showed the highest bacterial association in the HL, followed by SL and OVs in both tick species (Supp. Figure 2). However, there were more bacterial interactions in Am. gemma than Hy. rufipes. In Am. gemma, lower interactions between bacteria were observed in the SGs followed by the MG. Notably, no bacterial interactions were detected in the SGs of Hy. rufipes.

Generally, we observed more positive mutualistic co-occurrences between bacteria than the negative competitive ones within all tick tissues. We identified more keystone taxa in the HL followed by SL, OVs, MG, and SGs, in both tick species. However, we identified more taxa in Am. gemma tissues than in Hy. rufipes (Supp. Figure 2).

In Am. gemma, based on eigenvector centrality measures, the keystone taxa identified in the HL were Reyranella spp., as the most influential bacteria based on eigenvector centrality measures, while Nesterenkonia spp. were identified according to degree measures.

Additionally, Dietzia spp. were identified as key taxa based on both betweenness and closeness centrality measures. Rickettsia africae was negatively correlated with Sphingomonas spp. and positively correlated with Aeromonas spp. and Corynebacterium spp. (Fig. 6A, Supp. Table 4). We detected a negative correlation between Wolbachia endosymbionts and Afipia spp. Coxiella endosymbionts showed positive interaction with Enterococcus spp. and Novosphingobium spp. (Fig. 6A, Supp. Table 5).

Fig. 6.

Co-occurrence network analysis of bacteria identified based on 16S rRNA gene sequencing in the (A) haemolymph and B saliva of Amblyomma gemma ticks. Nodes represent individual bacterial genera, with their size indicating the degree of network. Node colours correspond to bacterial phyla. Edges between nodes indicate significant co-occurrence relationships, with edge thickness reflecting the strength of association, being either positive (blue) or negative (red). Keystone taxa are indicated in bold

In the SL, Bacillus spp., with the highest degree, closeness, and betweenness centralities (Fig. 6B, Supp. Table 6), showed a negative correlation with Pseudomonas spp. The highest eigenvector centrality was shown by Flavobacterium spp. (Fig. 6B, Supp. Table 7). No co-occurrence was detected for R. africae, and Coxiella or Francisella endosymbionts. However, Wolbachia endosymbionts showed a strong positive correlation to Hymenobacter spp., and Rickettsiella spp. positively correlated with Enterococcus spp. and Streptococcus spp. and negatively with Altererythrobacter spp.

The bacterial network occurrence was less in the MG and SGs. In the MG, no betweenness centrality was recorded. Five bacteria interacted and identified as keystone taxa with Brevibacterium spp. and Brachybacterium spp. are the most influential bacteria based on Eigenvector centrality (Supp. Figure 3A, Supp. Table 8, Supp. Table 9). However, in the SGs, more interactions were observed with two negative and three positives being recorded, and Bacillus spp. and Nocardioides spp. being most influential on bacterial composition (Supp. Figure 3B, Supp. Table 10, Supp. Table 11).

Acinetobacter spp. were the most influential bacteria in the OVs of Am. gemma due to the highest centrality measures recorded (Supp. Figure 4A, Supp. Table 12). Coxiella endosymbionts showed negative correlation with Sphingomonas spp. Meanwhile, Francisella endosymbionts negatively interacted with Afipia spp. and positively with Ca. Midichloria mitochondrii. The latter was positively correlated with Citrobacter spp. (Supp. Figure 4A, Supp. Table 13). No network with R. africae was detected.

In Hy. rufipes, due to the high centrality measures observed, Paenibacillus spp. was the most influential bacterium in the HL, while Staphylococcus spp. showed the highest closeness measure (Fig. 7A, Supp. Table 14). Paenibacillus spp. were negatively correlated with Wolbachia endosymbionts. Francisella endosymbionts were positively correlated with Microbacterium spp., and Proteus spp. were negatively correlated with Bacillus spp. (Fig. 7A, Supp. Table 15). No connection was detected with R. aeschlimannii, Ca. Midichloria mitochondrii, and Rickettsiella spp. In the SL, Gallicola spp., Cutibacterium spp., Helcococcus spp., and Stenotrophomonas spp. had the highest eigenvector, degree, betweenness and closeness centrality measures, respectively (Fig. 7B, Supp. Table 16). Gallicola spp. were negatively connected to Enhydrobacter spp. and positively connected to Ezakiella spp. Rickettsia aeschlimannii was positively connected to Ca. Midichloria mitochondrii (Fig. 7B, Supp. Table 17). No network was detected with Francisella endosymbionts.

Fig. 7.

Co-occurrence network analysis of bacteria identified by16S rRNA gene sequencing in the (A) haemolymph and B saliva of Hyalomma rufipes ticks. Nodes represent individual bacterial genera, with their size indicating the degree of network. Node colours correspond to bacterial phyla. Edges between nodes indicate significant co-occurrence relationships, with edge thickness reflecting the strength of association, being either positive (blue) or negative (red) correlation. Keystone taxa are indicated in bold

The highest centrality measures in the OVs were observed with Nocardioides spp. (Supp. Figure 4B, Supp. Table 18), which was positively connected to Paracoccus spp. and Staphylococcus spp. Rickettsia aeschlimannii showed the highest closeness measure, with a strong positive connection to Ca. Midichloria mitochondrii. Wolbachia endosymbionts were negatively connected to Peptoniphilus spp., while Proteus spp. was negatively connected to Corynebacterium spp. (Supp. Figure 4B, Supp. Table 19). No network connections were observed to Francisella endosymbionts except for one positive interaction with Ca. Midichloria mitochondrii. Those two bacteria were identified as keystone taxa. No interaction was detected in the SGs of Hy. rufipes whereas Francisella endosymbionts were considered keystone taxa in SGs.

Specific pathogen-symbiont interactions at the individual tick level

To better understand the specific tissue localisation and interactions between pathogens and endosymbionts, we compared the abundance of the 13 bacteria using heatmaps for each tick tissue/fluid type (HL, SL, OVs, MG, and SGs) of individual tick samples.

In Am. gemma, Coxiella endosymbionts were predominantly concentrated in the SGs and MG, followed by the OVs. Lower concentrations of Coxiella endosymbionts were observed in three HL samples and one SL sample. Francisella endosymbionts were mainly concentrated in the SGs. Rickettsia africae was found in all tissues, less abundant in the SL and predominant in the OVs and exhibited negative correlations with Coxiella endosymbionts in the OVs, SGs, and MG, with Francisella endosymbionts in the HL, and with Sphingomonas spp. in the HL, SL, and OVs. No Wolbachia endosymbionts were detected in the MG. Interestingly, we found Wolbachia endosymbionts mostly in the HL and SL, where no R. africae was detected. Rickettsiella spp. was mainly found in SL and HL. Corynebacterium spp., Staphylococcus spp., and Acinetobacter spp. were highly concentrated in the HL and SL, followed by the OVs. We observed that Pseudomonas spp. were highly concentrated in all tick tissues, particularly in the SL, HL, and OVs. A negative correlation of Pseudomonas spp. with R. africae was observed. Ca. Midichloria mitochondrii was mainly found in the OVs (Fig. 8A).

Fig. 8.

Heatmap of centred-log ratio transformed abundances of 13 bacterial species across the five tissue/fluid types (HL; haemolymph, SL; saliva, MG; midgut, OVs; ovaries, and SGs; salivary glands) in samples of (A) Amblyomma gemma and (B) Hyalomma rufipes. Z-scores are colour-coded as per the range scale

In Hy. rufipes, Francisella endosymbionts and Ca. Midichloria mitochondrii were the predominant endosymbionts spread across all tick tissues, and most concentrated in the SGs. Rickettsia aeschlimannii was primarily detected in the MG and SGs, followed by the OVs, and less abundant in the HL and SL. Ehrlichia ruminantium was detected only in one MG sample. Wolbachia endosymbionts were most abundant in the OVs, followed by the HL and SL, and absent in the MG. Proteus spp. were found in all tissues but less abundant in the SGs. Pseudomonas spp. were highly abundant in HL, SL, and OVs, followed by MG and least abundant in the SGs. Notably, Pseudomonas spp. was associated with low abundance of R. aeschlimannii in all tick tissues. Rickettsiella spp. was detected in the HL and SL, with high concentrations in two OV samples, but they were not found in any of SGs or MG samples. Staphylococcus spp. and Corynebacterium spp. were mainly concentrated in the HL and SL, followed by the OVs, and absent from the SGs (Fig. 8B).

Discussion

By comparing tissue-specific microbiota within individual ticks rather than whole-tick homogenates, we uncovered distinct microbial distributions and co-occurrence patterns in these vectors of camel diseases in northern Kenya. Co-occurrence network analyses revealed microbial taxa with high centrality measures indicative of keystone taxa, whose disruption may impair tick survival or pathogen persistence potentially informing novel management strategies such as anti-microbiota vaccines.

Each tissue harbored a distinct bacterial community [5, 6, 11]. Overall, haemolymph (HL), followed by saliva (SL), exhibited the highest bacterial diversity, likely acting as hubs for transient colonizers [43], while ovaries (OVs), midgut (MG) and salivary glands (SGs) had more specialized profiles, potentially reflecting their direct roles in blood digestion, nutrient absorption, and pathogen transmission [15, 16]. Such compartmentalization suggests a combination of extracellular colonization (in fluids like the HL or SL) and intracellular residence (within specialized epithelial cells of the SGs, OVs, or MG).

Microbial diversity was higher in Am. gemma than in Hy. rufipes, potentially reflecting environmental conditions in Laisamis versus Moyale, such as temperature and relative humidity [44, 45], and species-specific genetic factors. Targeting these keystone taxa identified in Am. gemma and Hy. rufipes through anti-microbiota vaccines could have detrimental consequences on tick survival or fecundity [46, 47], as shown for Pseudomonas spp. and Lactobacillus spp. in Ornithodoros moubata [12].

The positive bacterial interactions, especially in tissues with greater diversity, aligned with findings from other tick species [12, 48, 49]. Several factors, including the presence, role, and microbial load of the pathogenic microorganisms, can influence these interactions [47, 50]. The highest bacterial load in the HL and SL may result from a high quantity of extracellular bacteria in these tissues. In general, symbiotic bacteria in arthropods demonstrate tissue tropism, ranging from extracellular bacteria in the body cavities to intracellular bacteria localised within specialized organ cells [21, 51]. Many obligatory endosymbionts are intracellular [52], essential for host survival and ubiquitous throughout the host population. Secondary facultative symbionts are extracellular [53] and could improve host fitness, however, they are not vital for host survival [6]. Further studies are required to investigate how microorganisms influence the survival and fitness of ticks across different tick tissues.

Notably, R. africae and R. aeschlimannii confirm host specificity to Am. gemma and Hy. rufipes, respectively. The consistency between OmpB and V1-V2 16S rRNA gene sequences in identifying both rickettsial species further validates the molecular identification of these pathogens and underscores the reliability of our detection approach [8].

Their distribution varied across tissues and correlated negatively with dominant endosymbionts. For example, R. africae inversely associated with Coxiella in SGs and MG, and with Wolbachia in the HL and SL, suggesting antagonism or immune exclusion [54, 21]. Similar dynamics were observed between Francisella and R. aeschlimannii in Hy. rufipes. Additionally, we observed a negative correlation between R. africae and Sphingomonas in the HL, SL, and OVs, hinting at receptor-level competition or an antagonistic effect. Franken Sphingomonas engineered through a frankenbacteriosis approach to express MSP4-like proteins can reduce Anaplasma phagocytophilum infection in Ixodes scapularis [56]. A comparable mechanism may influence R. africae–Sphingomonas dynamics in Am. gemma. Conversely, we recorded a positive correlation between R. africae and Aeromonas in the HL. Notably, Aeromonas hydrophila infection has been previously linked to tick bites in Haemaphysalis longicornis harboring Rickettsia spp [57]. Whether such associations promote, or hinder pathogen establishment is unclear, emphasizing the need for functional investigations that identify how symbionts shape the tick immune landscape.

Obligate Coxiella and Francisella endosymbionts were localised in multiple tissues particularly the SGs and OVs, aligning with their known roles in B vitamin synthesis [9, 58]. In Am. gemma, Coxiella endosymbionts predominated, while Francisella endosymbionts were dominant in Hy. rufipes. Genomic evidence indicates that both endosymbionts possess complete pathways for the synthesis of biotin (vitamin B7), folate (vitamin B9), and riboflavin (vitamin B2), supporting their overlapping nutritional roles [58], suggesting functional redundancy where different symbionts fulfill essential nutritional roles and contribute to tick fitness and reproduction [9, 15, 59–62]. Their high abundance in the OVs may potentially support transovarian transmission [58, 63, 64]. Negative correlations between Coxiella or Francisella endosymbionts and Rickettsia spp. were consistent across tissues and support earlier findings of endosymbiont-mediated interference [8, 54, 55, 65, 66]. However causal mechanisms, whether competition for resources, immune priming, or niche exclusion [67], require in vivo or in vitro validation. Additionally, the inverse relationship between Francisella endosymbionts and certain keystone taxa (e.g., Reynella, Afipia) in the HL of Am. gemma further underscores the complexity of these microbiota networks and the need for deeper mechanistic research.

Although obligatory intracellular Wolbachia endosymbionts are commonly known for reproductive manipulation in insects [68, 69], their detection in the HL, SL, and OVs of both tick species indicates a broader tissue and tick species distribution. In ticks, Wolbachia is also linked to parasitoid wasps such as Ixodiphagus hookeri, which parasitises Ixodes ricinus and has been reported in Am. variegatum in Kenya [70–72]. In our study, Wolbachia was more abundant in the HL and SL in Am. gemma, and in the OVs of Hy. rufipes, suggesting both horizontal transmission and vertical inheritance [58, 69].

Certain Wolbachia endosymbionts can traverse cell membranes, exist extracellularly, or invade other somatic cells. This capability may play a significant role in facilitating horizontal transfer between species [73, 74]. Horizontal transmission of Wolbachia endosymbionts is also supported by evidence of their localisation in the mouth parts of the parasitic wasps and within Drosophila spp. Stocks [75]. The low abundance in tick SGs may reflect tissue-specific tropism. The presence of Wolbachia in somatic tissues and fluids aligns with previous observations in Drosophila and mosquitoes, where extracellular transmission and immune modulation have been described [74–78].

In Am. gemma, we observed coinfection with R. africae in the OVs, but not in the HL and SL, where R. africae was entirely absent when Wolbachia was present. Conversely, in Hy. rufipes, coinfection with R. aeschlimannii was detected in the HL, SL, OVs, and SGs. The interaction of Wolbachia and Rickettsia spp. has been reported in the reproductive tissues of the parasitic wasp Spalangia endius [79], possibly explaining the ability of R. aeschlimannii to evade immune defenses enhanced by Wolbachia presence and the potential of Hy. rufipes to act as a competent vector. Further studies are needed to (i) investigate the genetic diversity of Wolbachia strains across tick tissues; (ii) elucidate Wolbachia localisation and potential transmission to and from parasitic wasps; and (iii) explore Wolbachia-Rickettsia spp. interactions in tick hosts.

Candidatus Midichloria mitochondrii, an intracellular symbiont known to colonise tick mitochondria [80], showed broad tissue distribution in Hy. rufipes, but was restricted to OVs, MG, and SGs in Am. gemma. Prior studies [55, 81] and our data suggest that Ca. Midichloria mitochondrii not only exploits mitochondria in reproductive tissues, but may also support pathogen survival or modulate energy metabolism during blood feeding [82]. We observed positive correlations between Ca. Midichloria mitochondrii and Francisella in Am. gemma OVs and Hy. rufipes MG, and between Ca. Midichloria mitochondrii and R. aeschlimannii in the SL and OVs of Hy. rufipes. A similar synergy has been reported with Rickettsia parkeri in Am. Maculatum [55], suggesting that Ca. Midichloria mitochondrii may facilitate R. aeschlimannii transmission. It may also contribute to the transmission of other tick-borne pathogens, including rickettsial species [83]. The observed tissue-specific variation suggests the presence of functionally distinct Ca. Midichloria mitochondrii subpopulations. Further research should investigate their localisation and specialization across tick tissues and their specific roles in tick pathogenesis [84].

Rickettsiella spp. are endosymbiotic bacteria commonly found in arthropods, where they can act as either obligate or facultative symbionts depending on the host [58, 85]. These bacteria can exhibit both detrimental and beneficial effects on their hosts; they induce cytoplasmic incompatibility in the agricultural spider, Mermessus fradeorum [86], but provide B vitamins to the poultry red mite, Dermanyssus gallinae [87]. We detected Rickettsiella spp. mainly in the HL and SL of both tick species, suggesting possible horizontal transmission via co-feeding or bodily fluids exchange [85]. Their occasional detection in the OVs may also indicate vertical transmission routes [58]. Despite these findings, the ecological roles of Rickettsiella spp. in ticks remain poorly understood. Our results offer new insight into their tissue localisation, but further research is needed to clarify their functional relevance in tick physiology and pathogen interactions.

Although surface sterilization procedures were applied to minimize external contamination [32], the possibility that some environmental bacteria, Pseudomonas, Acinetobacter, Staphylococcus, and Corynebacterium, represent residual contaminants cannot be entirely excluded [16]. These taxa were abundant in the HL and SL, and may play detoxification or ancillary roles within ticks [50, 88]. While some are opportunistic vertebrate pathogens, they may also function as facultative symbionts in ticks [8, 43]. Entry into the tick may occur via transovarial, oral, or cuticular routes [6], and many are considered as secondary facultative endosymbionts that lack functional specialization [89]. Consistent with prior observation [8], we found an inverse abundance pattern between Pseudomonas spp. and R. africae, suggesting possible antagonism that warrants further investigation. Potential antagonism between microbiome members, particularly symbionts and pathogens, is an emerging area of interest in tick biology. Further genomic and in vitro studies on endosymbiont genomes are needed to understand the mechanisms by which these bacteria help prevent pathogen transmission. For example, in vitro experiments have demonstrated that the endosymbiont Rickettsia buchneri in Ixodes scapularis may produce antibiotic-like compounds that significantly suppress pathogenic Rickettsia species [90]. To further elucidate these interactions, future studies could include co-culture assays to directly observe inhibitory effects between symbionts and pathogens, metabolomic analyses to identify antimicrobial compounds produced by symbionts, and gene expression studies to explore potential immune-modulating pathways. Experimental infection models in live ticks, combined with symbiont manipulation (e.g., antibiotic treatments or symbiont knockdowns), would also help clarify the role of symbionts in pathogen exclusion and vector competence.

Notably, Acinetobacter spp. emerged as a keystone taxon, highlighting the potential influence of environmental bacteria on vector competence and disease ecology [91]. In Hy. rufipes, the high abundance of Proteus mirabilis, a haemolytic bacterium, may be linked to tick blood digestion in the MG [8, 92]. While this study offers novel insights, it was limited to a relatively small sample of fed female ticks. Future work should expand the sample size to include different life stages, feeding states, considering species-specific factors and environmental factors to better capture the microbiome dynamics. High-resolution imaging techniques such as fluorescence in situ hybridization (FISH) could clarify the intra- versus extracellular localisation of symbionts [60]. Moreover, functional assays, such as antibiotic or RNAi knockdown and coinfection experiments, would help confirm whether specific symbionts inhibit or facilitate pathogen transmission. For instance, in RNAi knockdown experiments, double-stranded RNA molecules are used to silence symbiont genes, thereby reducing the expression of target genes and consequently the abundance or activity of the symbiont. This approach allows to examine how the depletion of particular symbionts affects pathogen colonization, replication, and transmission, providing direct evidence of the functional interactions between microbial partners and the tick host.

Conclusions

Tissue type plays a major role in shaping microbiome composition in Am. gemma and Hy. rufipes ticks. The complex interactions observed between key endosymbionts and Rickettsia pathogens across tissues, including negative correlations with Coxiella, Francisella, and Wolbachia endosymbionts, and positive correlations between Ca. Midichloria mitochondrii and R. aeschlimannii, require further investigation to determine their influence on pathogen establishment. Keystone taxa identified through co-occurrence networks present promising targets for anti-microbiota vaccines or other microbial interventions to curb tick-borne diseases. Understanding these microbial relationships will be key to developing sustainable strategies that reduce pathogen transmission to livestock and humans.

Abbreviations

ASVs

Amplicon Sequence Variants

BLAST

Basic Local Alignment Search Tool

bp

Base pair

DADA2

Divisive Amplicon Denoising Algorithm 2

DNA

Deoxyribonucleic Acid

icipe

International Centre of Insect Physiology and Ecology

MAFFT

Multiple Alignment using Fast Fourier Transform

MBBU

Molecular Biology and Bioinformatic Unit

MG

Midgut

ML-EID

Martin Lüscher Emerging Infectious Diseases

MSP4

Major Surface Protein 4

OmpB

Outer membrane protein B

OVs

Ovaries

PCoA

Principal Coordinates Analysis

PCR

Polymerase Chain Reaction

PERMANOVA

Permutational Multivariate Analysis of Variance

PhyML

Phylogenetic Maximum Likelihood

rRNA

ribosomal RNA

SGs

Salivary Glands

SL

Saliva

TBPs

Tick-Borne Pathogens

Authors’ contributions

**RK: ** Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. **JV: ** Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – original draft, Writing – review & editing. **AB, DM** : Supervision, Investigation, Writing – review & editing. **FMK: ** Supervision, Investigation, Writing – review & editing. **DG** : Data curation, Formal analysis, Investigation, Methodology, Writing – review & editing. **EY** : Data curation, Investigation, Methodology, Writing – review & editing. **JK: ** Investigation, Methodology, Writing – review & editing.

Funding

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement no. 101000365 (PREPARE4VBD), and icipe institutional support from the Swedish International Development Cooperation Agency (Sida); the Swiss Agency for Development and Cooperation (SDC); the Australian Centre for International Agricultural Research (ACIAR); the Government of Norway; the German Federal Ministry for Economic Cooperation and Development (BMZ); and the Government of the Republic of Kenya. RK was supported by a German Academic Exchange Service (DAAD) through an icipe ARPPIS-DAAD scholarship and through a UP post-graduate bursary. The views expressed herein do not necessarily reflect the official opinion of the donors.

Data availability

All sequences presented in this study were submitted to GenBank nr database of the NCBI (http://www.ncbi.nlm.nih.gov) under BioProject accession number: PRJNA1134173 (http://www.ncbi.nlm.nih.gov/bioproject/1134173).

Declarations

Ethical approval and consent to participate

Ethical approval for this study was obtained from the Pwani University Ethics Review Committee (ERC/EXT/002/2020E) and a license from the National Commission for Science Technology and Innovation (NACOSTI/P/22/16467). This study did not involve human subjects as participants; however, verbal informed consent was obtained from all camel owners prior to tick collection from their animals. Given the low literacy rates among the camel farmers, the procedure for obtaining verbal informed consent was reviewed and approved by the Ethics Review Committee as part of the overall study protocol.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.