Microbial composition of enigmatic bird parasites: Wolbachia and Spiroplasma are the most important bacterial associates of quill mites (Acariformes: Syringophilidae)

Abstract Background The microbiome is an integral component of many animal species, potentially affecting behavior, physiology, and other biological properties. Despite this importance, bacterial communities remain vastly understudied in many groups of invertebrates, including mites. Quill mites (Acariformes: Syringophilidae) are a poorly known group of permanent bird ectoparasites that occupy quills of feathers and feed on bird subcutaneous tissue and fluids. Most of the known species have strongly female‐biased sex ratio, and it was hypothesized that this is caused by endosymbiotic bacteria. Previously, Anaplasma phagocytophilum (Foggie) and a high diversity of Wolbachia strains were detected in quill mites via targeted PCR screens. Here, we use an unbiased 16S rRNA gene amplicon sequencing approach to determine other bacteria that potentially impact quill mite biology. Results We performed 16S rRNA gene amplicon sequencing of 126 quill mite individuals from eleven species parasitizing twelve species (four families) of passeriform birds. In addition to Wolbachia, we found Spiroplasma as potential symbiont of quill mites. Consistently, high Spiroplasma titers were only found in individuals of two mite species associated with finches of the genus Carduelis, suggesting a history of horizontal transfers of Spiroplasma via the bird host. Furthermore, there was evidence for Spiroplasma negatively affecting Wolbachia titers. We found no evidence for the previously reported Anaplasma in quill mites, but detected sequences of high similarity to the potential pathogens Brucella and Bartonella at low abundances. Other amplicon sequence variants (ASVs) could be assigned to a diverse number of bacterial taxa, including several that were previously isolated from bird skin. Further, many frequently found ASVs were assigned to taxa that show a very broad distribution with no strong prior evidence for symbiotic association with animals. We interpret these findings as evidence for a scarcity of resident microbial associates (other than inherited symbionts) in quill mites.


| INTRODUC TI ON
There is abundant evidence that microbial taxa are an essential component of many animal species (McFall-Ngai et al., 2013). Bacteriaencoded traits may significantly impact host phenotypes, for example, through providing essential nutrients (Duron et al., 2018;Hosokawa, Koga, Kikuchi, Meng, & Fukatsu, 2010), defending against pathogens (Ballinger & Perlman, 2017;King et al., 2016), but also affecting ecological features of their hosts, such as mate choice (Sharon et al., 2010) and life history traits (Laughton, Fan, & Gerardo, 2013). Because of their potential importance in understanding the biology of many organisms, the number of microbiome studies has been soaring (Hird, 2017).
In the present study, we focus on quill mites (Acariformes: Syringophilidae). These obligatory bird ectoparasites live and reproduce inside the quills of feathers where they feed on subcutaneous fluids (lymph, blood). Quill mite dispersion has been observed on the same individual (from infected to uninfected feathers), between individuals of the same species (e.g., from parents to hatchings) and occasionally by transfer between gregarious bird species (Casto, 1974a(Casto, ,1974bKethley, 1970Kethley, , 1971. This mode of feeding and dispersion makes quill mites potential vectors for bacterial pathogens, similar to ticks or lice (Azad & Beard, 1998). However, only two bacterial taxa were recorded in quill mites so far: (a) Anaplasma phagocytophilum (Foggie) (Alphaproteobacteria, Rickettsiales) was detected in two quill mite species from three bird species (Skoracki et al., 2006); (b) multiple genetically distinct lineages of Wolbachia (Alphaproteobacteria, Rickettsiales) were found in five species of quill mites (Glowska, Dragun-Damian, Dabert, & Gerth, 2015). As these studies were targeted PCR screens, it remains unclear what other bacteria populate quill mites. Furthermore, the importance of quill mites for bird pathogen dynamics is not known.
To address these questions, we here assess the bacterial composition of 126 quill mite individuals encompassing eleven species with a more unbiased 16S rRNA gene amplicon sequencing approach. We find that the symbionts Wolbachia and Spiroplasma are among the most commonly associated taxa with quill mites. Other taxa include bacteria that were previously found in association with arthropods and bacteria with a very broad distribution. Strikingly, neither quill mite taxonomy nor bird host taxonomy significantly influences bacterial composition in quill mites. Furthermore, we find that despite the detection of Bartonella and Brucella, quill mites do not seem to be major pathogen vectors in birds.

| Animal collection and DNA extraction
A summary of collected quill mite species and their bird hosts can be found in Table 1. All quill mites used in this study were collected in Kopan, Poland, during spring migration of birds monitored by the Bird Migration Research Station, University of Gdansk, April 2009. One secondary flight feather was analyzed from each bird specimen and dissected under a stereo microscope (Olympus ZS30). Individual mites were washed twice and preserved in 96% ethanol, and total genomic DNA was extracted from single specimens using DNeasy Blood & Tissue Kit (Qiagen GmbH), as described previously (Dabert, Ehrnsberger, & Dabert, 2008). This procedure left the exoskeletons intact, and the specimens were subsequently mounted on microscopic slides in Faure medium and determined using the key from Skoracki, Spicer, and Oconnor (2016). All morphological observations were carried out with an Olympus BH2 microscope with differential interference contrast (DIC) optics and a camera lucida. All DNA samples and corresponding voucher specimens are deposited in the collection of the Department of Animal Morphology, Faculty of Biology, Adam Mickiewicz University in Poznan, Poland. To identify potential contaminants, in addition to sequencing a negative control alongside all samples, we further extracted DNA from reagents and materials commonly used in the laboratory this work was carried out in. One library each was created from extraction buffer (ALT), millipore water, microscope swabs, pipette swabs, and swabs of other equipment (pincettes, scalpels, benches, etc). These five libraries were processed and sequenced separately from the other samples, but by using identical procedures.

| Read processing and statistical analyses
Reads were trimmed of adaptors and primer sites by using cutadapt version 1.16 (Martin, 2011). The remaining reads were dereplicated, denoised, and chimeras eliminated using the DADA2 package version 1.8 (Callahan et al., 2016) within the R statistical programming environment (R Core Team 2015). Taxonomic assignment of the ASVs (amplicon sequence variants), to species level where possible, was also performed within DADA2 using the SILVA database version 132 (Quast et al., 2013). Next, contaminant taxa were identified from the sequenced extraction control using the "prevalence" algorithm implemented in the R package decontam (Davis, Proctor, Holmes, Relman, & Callahan, 2018). Further potential contaminants were identified by processing the five libraries derived from reagents and materials as described, and then excluding all ASVs that were found in any of these control libraries from subsequent analyses.
To reduce the impact of ASVs with very low abundance, we removed all ASVs that were present in only a single sample and also discarded ASVs from bacterial phyla that only occurred once in total.
To account for potential biases between samples with uneven sequencing depth, all read counts from the remaining samples were rarefied to the read depth of the sample with the lowest read number. An overview of how our filtering steps affected ASV counts can be found in Table 3. All subsequent statistical analyses were done on log-transformed read counts. Because the symbionts Wolbachia and Spiroplasma were dominant in some of the samples, we excluded all ASVs corresponding to these taxa prior to statistical comparisons between groups. First, we plotted the abundance of the most frequently found bacterial families using the R packages phyloseq and ggplot2 (McMurdie & Holmes, 2013;Wickham, 2009). Next, ordination analyses were performed with phyloseq using Bray distances and non-metric multidimensional scaling (NMDS). Differences in abundances of particular taxa between groups (quill mite species, bird host species, developmental stage, Wolbachia positive and negative samples) were determined with Kruskal-Wallis rank sum tests, and p-values were adjusted to these multiple comparisons to control for the false discovery rate (Benjamini & Hochberg, 1995). These tests were done separately for differences in abundance of bacterial phyla, orders, families, and genera. Furthermore, we calculated Jensen-Shannon distances between the aforementioned groups and used adonis tests (analysis of variance using distance matrices) implemented in the package vegan (Dixon, 2003) to test if they differed significantly. A phyloseq object file containing all data used in the described analyses is available at http://doi.org/10.5281/zenodo.3475151.

| RE SULTS
We have investigated the microbial composition of 126 individuals belonging to eleven quill mite species that parasitize twelve bird host species of passeriform birds. Amplicon sequencing of the v4  Figure A1). However, when trying to identify differential abundance patterns of microbial composition between groups using analysis of variances, we found that bacterial composition was more similar between samples from the same quill mite species or genus and bird host species or genus than expected by chance (p < .01). Furthermore, six bacterial families were found to be differentially abundant between quill mite species with a Kruskal-Wallis test (p < .01, Figure 3), one of which (Xanthobacteraceae) was also found to differ between samples of different bird host genera.
Out of 912 detected ASVs, the 10 most abundantly encountered genera were Micrococcus, Corynebacterium, Acinetobacter, TA B L E 3 Overview on the impact of filtering and decontamination on the number of retained ASVs and samples in this study. For details on each of the steps, please refer to the materials and methods section Step #ASVs #samples    samples ( Figure 1b). Other notable findings were the pathogens Brucella which was detected in 20 samples with an average abundance of 1.3%, and Bartonella which was found in two samples at 1.8% and 0.7% relative abundance, respectively.
As opposed to the general trend in the microbiome composition data, there was strong evidence for differential abundance of the symbionts Wolbachia and Spiroplasma between the bird hosts from which the mites were collected. For example, high Spiroplasma titers were only observed in two mite species collected from the host genus Carduelis (Figures 1a, 4a, Table 1). Further, although Wolbachia was present in mites sampled from all bird hosts, it was especially prevalent in mites collected from birds of the genera Turdus, Erithacus, and Fringilla. On contrast, it was absent or at very low titers in mites parasitizing Luscinia sp. (Figure 4a, Table 1). On average, the abundance of Wolbachia was lower in samples that also contained Spiroplasma

| Origin of microbial DNA in quill mites
We have sequenced microbial taxa from quill mites, an enigmatic Further, we document a very high abundance of these taxa in at least some of the investigated samples (Figure 4), which is in line with the assumptions above. In a previous study, Wolbachia strains of quill mites were investigated with a multi-locus approach and it has been shown that quill mite associated strains are genetically very different to any other Wolbachia strains described so far (Glowska et al., 2015). Here, we have found eight different ASVs annotated as Wolbachia, each of which is 100% identical to at least one Wolbachia sequence previously isolated from quill mites.
For Spiroplasma, we found a single ASV that is only 92% identical to the next closest match in the Silva database. This implies that Spiroplasma in quill mites might be genetically distinct from Spiroplasma of other arthropods, as is the case for Wolbachia.
However, sequencing data of more loci are needed to establish the phylogenetic placement of Spiroplasma from quill mites. Despite these similarities, and some statistical support for bird hosts shaping the microbiome community in our study, the lack of clustering in ordination analysis indicates that environment is not the major determining factor of quill mite microbiome composition.
It is important to consider that contaminants from reagents and kits may significantly impact microbiome composition estimates, especially when using low biomass samples such as quill mites (de Goffau et al., 2018;Łukasik et al., 2017;Salter et al., 2014). This is problematic in any microbiome study and is very difficult to exclude with certainty. Here, we removed contaminants statistically in silico based on the microbial composition of the sequenced extraction control (Davis et al., 2018). Further, we removed all ASVs present in independently sequenced controls derived from reagents and equipment commonly used in the laboratory where this study was performed (see methods for details). However, a number of ASVs we recovered correspond to common kit contaminants in 16S rRNA microbiome studies (e.g., Ralstonia, Kocuria), human skin bacteria (Corynebacterium), or ubiquitous taxa with no strong evidence for symbiotic associations with arthropods (Pseudomonas, Acinetobacter). These taxa might constitute true associates, but we cannot exclude the possibility that they originate from contaminating sources.
In summary, we found a diverse range of bacteria associated with quill mites. The lack of differentiation between different mite species or between species collected from different bird hosts leads us to conclude that there are no strong associations with typical gut bacteria as observed in other arthropods. However, we cannot exclude that we missed such potential associates due to the limited amount of DNA that can be extracted from the minute hosts.

| Exchange of bacteria via bird hosts
Due to their ectoparasitic life style with occasional host switching, quill mites have the potential to transmit bacteria between their hosts. Here, we detected two pathogenic microbes that might be important in that respect: Brucella and Bartonella. Brucella is the agent of brucellosis, which is considered to be the most widespread zoonotic infection (Pappas, Papadimitriou, Akritidis, Christou, & Tsianos, 2006). Several Brucella species are a human health threat, and people typically become infected through contact with domesticated Brucella infected animals, such as goats, sheep, or swine (Young, 2006). However, several blood-sucking arthropods, such as ticks and lice, are regarded as possible vectors for Brucella (Neglia et al., 2013;Pritulin, 1954;Rementzova, 1966). To our knowledge, there is no data indicating that Acari other than ticks are natural Brucella carriers. It was hypothesized that birds and other wild animals act as natural reservoirs for Brucella (Zheludkov & Tsirelson, 2010), which is in line with our finding of this bacterium in bird ectoparasites. The importance of quill mites in spreading Brucella between bird species remains to be assessed, but its prevalence (21/126 investigated individuals, 8 different mite species) suggests its finding is of potential importance in understanding this pathogen's dynamics.
Bartonella are gram-negative bacteria that are typically transmitted by blood-sucking arthropods and are infectious in mammalian hosts (Billeter, Levy, Chomel, & Breitschwerdt, 2008;Klangthong et al., 2015;Reeves, Nelder, Cobb, & Dasch, 2006). There are also reports on Bartonella incidence in birds (Ebani, Bertelloni, & Mani, 2016;Mascarelli, McQuillan, Harms, Harms, & Breitschwerdt, 2014), and it is conceivable that the bacteria originate from the birds, rather than from the mites. That would suggest that the host range for Bartonella spp. is broader than previously reported and here we expand the list of potential sources for this zoonotic infection.   honey bees and ants (Segers, Kešnerová, Kosoy, & Engel, 2017, Bisch et al. 2018. Further, Bartonella-like symbionts were recently found in a number astigmatid mites (Kopecký, Nesvorná, & Hubert, 2014), indicating that the Bartonella detected here might be quill mite symbionts, rather than pathogens. With our data, it is not possible to rule out either possibility.
Finally, we found the symbionts Spiroplasma and Wolbachia in quill mites. Both of these are common across a range of arthropod species (Anbutsu & Fukatsu, 2011;Zug & Hammerstein, 2012), are typically transmitted intraovarially, and may cause sex ratio distorting phenotypes (Haselkorn, 2010;Werren, 1997). Whereas Spiroplasma was so far not reported from quill mites, Wolbachia was previously detected and our findings confirm that this is a common symbiont of quill mites (Glowska et al., 2015). The observed presence and abundance of both taxa are not uniform across the sampled taxa (Figures 1, 4a). For example, Wolbachia is most abundant in mites parasitizing birds of the genera Turdus, Erithacus, and Fringilla, whereas Spiroplasma is most strongly associated with mites parasitizing Carduelis. One reason for this may be that some taxa are more susceptible than others for endosymbiosis with certain bacteria, and this phylogenetic effect has been reported for other host taxa as well (Gerth, Saeed, White, & Bleidorn, 2015;Russell, 2012). Strikingly, very high Spiroplasma abundances were only found in two investigated mite species that are specialized parasites of two bird species of the genus Carduelis (Figure 4a.,  (Huigens et al., 2000;Jaenike, Polak, Fiskin, Helou, & Minhas, 2007). Although the potential mechanism of horizontal symbiont transmission via feather quills is unclear, our data suggest that the bird-parasite interactions may be important for endosymbiont transmission dynamics in quill mites.
Interestingly, we found that Spiroplasma presence leads to reduced Wolbachia titers, although this is based on a small sample size for samples that are both Wolbachia and Spiroplasma positive (N = 11, Figure 4b). Furthermore, in these eleven samples, Spiroplasma and Wolbachia titers seem to be positively correlated ( Figure 4c). It is conceivable that sharing of hosts leads to competition for finite resources the host can provide (Vautrin & Vavre, 2009), and thus the growth of one symbiont might limit that of another. In Drosophila, for example, Spiroplasma seem to limit the proliferation of Wolbachia (Goto, Anbutsu, & Fukatsu, 2006) and in aphids, competition between co-occurring secondary symbionts appears to be harmful to the host (Oliver et al. 2006). Such negative fitness impacts can also be expected when both symbiont titers are very high, as found here in quill mites. Although purely speculative, this may be the reason why we only observed simultaneously high Spiroplasma and Wolbachia titers in very few of the 126 investigated quill mites (Figure 4c).

| SUMMARY
We found a diverse, but relatively uniform set of bacterial taxa within quill mites that includes arthropod endosymbionts, pathogens, and bird-associated bacteria. The importance of most of these microbes for quill mite biology is unclear, but abundances and distribution patterns suggest that Spiroplasma and Wolbachia are the most important quill mite associates.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
EG, ZF, and MG involved in study conception and design. EG involved in collecting and identification of the mite material, funding acquisition, project administration, supervision, writing review and editing and contributed resources. ZF and MD involved in acquisition of data.
MG analyzed the data. EG and MG interpreted the data and wrote the paper.

E TH I C A L S TATEM ENT
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available under NCBI BioProject accession number PRJNA482380 and in Zenodo repository at http://doi.org/10.5281/zenodo.3475151.