Feather mites play a role in cleaning host feathers: New insights from DNA metabarcoding and microscopy

Abstract Parasites and other symbionts are crucial components of ecosystems, regulating host populations and supporting food webs. However, most symbiont systems, especially those involving commensals and mutualists, are relatively poorly understood. In this study, we have investigated the nature of the symbiotic relationship between birds and their most abundant and diverse ectosymbionts: the vane‐dwelling feather mites. For this purpose, we studied the diet of feather mites using two complementary methods. First, we used light microscopy to examine the gut contents of 1,300 individual feather mites representing 100 mite genera (18 families) from 190 bird species belonging to 72 families and 19 orders. Second, we used high‐throughput sequencing (HTS) and DNA metabarcoding to determine gut contents from 1,833 individual mites of 18 species inhabiting 18 bird species. Results showed fungi and potentially bacteria as the main food resources for feather mites (apart from potential bird uropygial gland oil). Diatoms and plant matter appeared as rare food resources for feather mites. Importantly, we did not find any evidence of feather mites feeding upon bird resources (e.g., blood, skin) other than potentially uropygial gland oil. In addition, we found a high prevalence of both keratinophilic and pathogenic fungal taxa in the feather mite species examined. Altogether, our results shed light on the long‐standing question of the nature of the relationship between birds and their vane‐dwelling feather mites, supporting previous evidence for a commensalistic–mutualistic role of feather mites, which are revealed as likely fungivore–microbivore–detritivore symbionts of bird feathers.

Host-symbiont interactions rarely involve a simple one-symbiont: one-host interaction. Rather, even without considering the interaction of the host species with other free-living species, any host-symbiont interaction typically involves several other species (Hopkins, Wojdak, & Belden, 2017;Poulin, 2010). In addition, whether a particular symbiont species acts as a parasite, commensal or mutualist can be highly context-dependent (i.e., the mutualism-parasitism continuum framework; for example, Brown, Creed, Skelton, Rollins, & Farrell, 2012;Cheney & Côt e 2005;Newton, Fitt, Atkins, Walters, & Daniell, 2010;Jovani et al., 2017). Thus, the study of symbionts as a whole, and not separately according to the presumed nature of their relationships with their hosts, is needed (Jovani, 2003;Jovani et al., 2017).
Defensive mutualisms (i.e., those in which symbionts protect their hosts from natural enemies, which have been often perceived as biological curiosities) have been reviewed recently following this approach and placed into this framework (Hopkins et al., 2017).
Accordingly, defensive mutualisms, instead of being anecdotal hostsymbiont associations, have been revealed as diverse and common associations in a wide range of plants and animal hosts from nearly all habitats on the planet. Nonetheless, with a few exceptions, most of the diversity of host-symbiont associations remains unexplored or largely unstudied.
A long-standing question in understanding the interaction between feather mites and birds is whether these mites feed on bird tissues (e.g., feathers, skin, blood) or upon resources found on the bird's surface (e.g., algae, fungi). If they feed on bird tissues, they are more likely to be classified as parasites (Harper, 1999;Poulin, 1991;Thompson, Hillgarth, Leu, & McClure, 1997), while if they do not, feather mites would more likely be commensals or even mutualists (Blanco, Tella, & Potti, 1997;Blanco, Tella, Potti, & Baz, 2001;Galv an et al., 2012). Previous evidence has suggested that feather mites could feed mainly on the uropygial gland oil of birds (Dubinin, 1951;Proctor, 2003;Walter & Proctor, 2013c). However, this oil is a nitrogen-deficient source (Jacob & Ziswiler 1982;Proctor, 2003), and previous evidence has shown that feather mites complement their diet with fungi, pollen and algal particles (Blanco et al., 2001;Dubinin, 1951;Proctor, 2003;Walter & Proctor, 2013c). Examining thousands of slidemounted feather mites from 26 mite species, Dubinin (1951) found that almost all mite species had fungal spores in their guts, most from Cladosporium, Alternaria and rust fungi. Moreover, Blanco et al. (2001) found fungal mycelia and spores in the guts of 53% of Pterodectes rutilus (Robin) (Proctophyllodidae) and 38% of Scutulanyssus nuntiaventris (Berlese) (Pteronyssidae) mites from two species of swallows (Hirundinidae). Likely because of this potential mixture of feather mite diet, a recent isotopic study (Stefan et al., 2015) of the diet of two feather mite species produced inconclusive results. Interestingly, however, this study showed a strong correlation between the isotopic carbon signatures among mites inhabiting the same individual host, and between the carbon signature (but not the nitrogen signature) of feather mites and the blood of their individual bird host, thus suggesting that diet could be mainly based on shared host-associated resources, arguably preen gland oil (Stefan et al., 2015). Thus, it remains an open question to what extent feather mites feed on uropygial oil or also upon other bird tissues, whether exogenous resources, such as fungi and bacteria, constitute an important food resource for these mites, and which specific taxa are eaten by feather mites.
In this study, we investigated the diet of feather mites using two complementary methods. First, we used light microscopy to examine feather mite gut contents under the microscope from a large sample of feather mites from~200 bird species. Light microscopy allows detection of feather fragments, fungi, plant material and algae that are refractory to the clearing and mounting media (see Materials and methods). In a second approach, for a smaller number of vane-dwelling mite species, we studied gut contents using high-throughput sequencing (HTS) and DNA metabarcoding. This molecular approach complemented the light microscope analysis for certain potential food resources that would not be easily recognized in the slide-mounted specimens (e.g., bacteria, soft bird tissues) and also allowed for a detailed analysis of fungi, bacteria and plant taxa in the mites' diet.

| Gut content assessment via light microscopy
For the microscopy analysis, we used previously slide-mounted mites from the Proctor Lab collection of feather mites from around the world.
Selection of mites to examine was based on taxonomic diversity of mites and host birds, and ecological breadth of hosts (e.g., birds from terrestrial, marine and freshwater habitats, including predators, granivores, nectarivores, etc.). We initially examined several thousand mites using a Leica DMLB compound microscope with DIC lighting. Mites with visible gut contents were photographed at various magnifications (200,400 and 8009) depending on size of material in the gut. For each host bird species included in the study, our goal was to photograph a minimum of five individual mites from each mite genus present on the bird species.
In some cases, if there were fewer than five mites with gut contents available for a mite genus and/or bird species, then all the available mites that contained gut contents were photographed. Under ideal circumstances, we would have focused on mite species rather than genera, but particularly for tropical areas, feather mite alpha-taxonomy is in an early state and many species have yet to be described. Also, for many taxa, only adult males can be readily ascribed to species, and we wished to include nymphal and female mites in our assessment. Mites were identified to genus using Gaud and Atyeo (1996) with additional literature for more recently described genera (e.g., Valim & Hernandes, 2010). In total, 1,300 individual mites representing 100 genera (18 families) from 190 host bird species (72 families; 19 orders) were photographed.
Each morphologically unique type of gut content was given a code, and for every individual mite, all the types of gut content present were recorded, as well as the approximate amount of each type of gut content. Aided by illustrations in Lacey and West (2006) and consultation with a mycologist (T. Spribille, University of Alberta), we then classified all unique types of gut contents as fungi, diatoms, plant spores, "unidentifiable" and oily globules (possibly uropygial gland oil or digestive by-products in peritrophic membranes). Unidentifiable objects were mainly extremely small fragments or flecks of material <5 lm long (some of which could have potentially been tiny remnants of feather barbules) (e.g., Figure S10). Oil globules were not included in the analyses, as we consider that our ability to consistently identify this material was much lower than for other types of gut content (see an example of potential oil globs in Figure S11).

| Sample collection and sterilization for DNA metabarcoding
For the DNA metabarcoding study, 1,833 individual mites of 18 mite species from 18 passerine bird host species (34 individual birds or infrapopulations) were sampled from birds captured with mist nets in Andalusia (Spain) during the spring of 2015 (see Table S1, for sampling details). An effort was made to collect all mites found on the wing flight feathers from each sampled bird, using a sterile swab impregnated with ethanol. Mites were preserved at À20°C in tubes with 96% ethanol. In those cases in which more than one mite species was found on an individual bird, one different sterile swab was used for collecting each tentative mite species (according to Doña et al., 2016 based on genus-specific location on bird feathers) into different tubes.
Mites were sterilized in AllGenetics & Biology, SL (A Coruña, Spain) with three ethanol washes following Andrews (2013). Each time, tubes containing mites were agitated manually. Then, all ethanol was collected with the pipette using a thin pipette tip, with careful visual checks to avoid removing any mites. Tubes were then refilled with ethanol. Washed mites were then used for further analyses (hereafter mite samples) and the ethanol extracted from the first wash was used as the environmental control sample (hereafter, external sample).

| DNA extraction, amplification, library construction and sequencing
DNA isolation, amplification and library preparation were carried out at AllGenetics & Biology, SL (A Coruña, Spain). Genomic DNA was extracted from each mite sample using the HotSHOT method (Truett et al., 2000). Briefly, the ethanol from the last mite wash was evaporated and a 1-M NaOH solution was added to the dried wells, incubated at 95°C and neutralized with equivalent amounts of Tris-Cl.
The final extraction volume was 30 ll. A negative control that contained no sample was included in every extraction round to check for contamination during the experiments. This procedure preserves exoskeletons for morphological identifications (see Doña, Diaz-Real, et al., 2015). However, in contrast to more aggressive isolation methods, DNA from Gram-positive bacteria, undigested diatoms and intact fungal spores may not have been amplified. After DNA extraction, the remaining exoskeletons were separated from the buffer and stored in 80% ethanol. External samples were extracted as follows.
The ethanol phase from the first mite wash was pipetted onto a nitrocellulose filter (ca. 9 cm² with a pore size of 22 lm), and then, DNA was isolated using the PowerSoil DNA isolation kit (Mobio) following manufacturer's instructions. The final elution volume was 50 ll.   Kerr, Lijtmaer, Barreira, Hebert, & Tubaro, 2009). In addition, we amplified the COI gene of feather mites (bcdF05/bcdR04, Dabert, Ehrnsberger, & Dabert, 2008) to molecularly confirm the mite species identity (Doña, Diaz-Real, et al., 2015). Only bacterial and fungal regions were amplified from the external samples.
Libraries were built following the recommended protocol by Illumina for bacterial 16S metabarcoding, with some modifications. Similar protocols have been used by other authors (e.g., Lange et al., 2014;Vierna, Doña, Vizca ıno, Serrano, & Jovani, 2017). Briefly, the libraries were constructed in a two-step PCR (hereafter, PCR1 and PCR2 was carried out using 2.5 ll of the amplified DNA from PCR1 as a template and was performed under the same conditions as PCR1, but only running five cycles at 60°C as the optimal annealing temperature. A total of 31 different index combinations were used, and 40 PCR cycles were performed (Vierna et al., 2017). The resulting products were purified following the SPRI method as indicated above.
Likewise, the purified products were loaded in a 1% agarose gel stained with GreenSafe (NZYTech) and visualized under UV light.
All products (a total of 238 libraries) were pooled together in 21 sets of differentially indexed samples. All pools were quantified with Qubit ™ fluorometer (Invitrogen). We did not obtain bird DNA in any sample and plant DNA only from two samples (see Results below). Accordingly, all except one plant pool (i.e., the one containing the only two samples successfully amplified, see Results below) were not sequenced as they did not reach the minimum amount of DNA for HTS.
All pools were sequenced by Novogene (Beijing, China) on Illumina HiSeq 4000 using the PE 250 strategy (see Supporting Information for coverage information; Table S2). Quality controls were carried out using company in-house Perl scripts to remove contaminated adaptors and low-quality sequences.
Mite identity was molecularly confirmed in all cases using a similar pipeline to that used in Doña, Moreno-Garc ıa, Criscione, Serrano, and . In brief, we used Geneious R10 (http://www.ge neious.com, Kearse et al., 2012) plugin Sequence classifier, over a concatenated file containing the forward and reverse reads (quality trimmed as described above for plant libraries and with a minimum length of 200 bp). Then, we used the recommended threshold and a reference DNA barcode library (Doña, Diaz-Real, et al., 2015).

| Statistical analysis
Differences in prevalence and morphological diversity of diet resources (the maximum diversity retrieved for each mite sample, that is, each mite infrapopulation; see above) from microscopy assessments were analysed using generalized linear mixed models (GLMM) (GLMER function from package LME4 1.1-12, Jovani & Tella, 2006;Bates, M€ achler, Bolker, & Walker, 2015). For assessing differences in prevalence, we ran a binomial GLMM considering prevalence (1: presence, 0: absence) as the response variable, the type of food resource as the predictor variable and the bird infrapopulation nested into bird species plus mite genera as random factors. For assessing differences in morphotype diversity of fungi and diatoms, we ran a Poisson GLMM considering morphotype diversity as the response variable, and the same structure of predictor and random factors. We confirmed assumptions underlying GLMMs by exploring regression residuals for normality against Q-Q plots.
Fungal and bacterial OTUs were imported to R and manipulated using PHYLOSEQ R package (McMurdie & Holmes, 2013). In particular, we studied the variance in bacterial and fungal assemblage composition among infrapopulations using a permutational multivariate analysis of variance on Bray-Curtis and Jaccard distance matrices (PERMANOVA; adonis function from the VEGAN v2.4.1 R package, Oksanen et al., 2017). The null hypothesis was that the centroid does not differ between host species and/or mite species (Anderson & Walsh, 2013). This test is highly sensitive to data dispersion (Anderson, 2001), and thus, we tested it with the multivariate homogeneity PERMDISP2 procedure (Anderson, 2006; betadisper function from VEGAN, Anderson & Walsh, 2013) with 999 permutations. Additionally, following previous approaches to overcome this statistical issue (e.g., Brice, Pellerin, & Poulin, 2017), we explored the community clustering with ordination analyses (principal coordinates analyses, PCoA) and stacked bar plots at the infrapopulation level.

| Composition and morphological diversity of feather mites' diets assessed by microscopy
From a total of 481 infrapopulations (1,300 individual mites) belonging to 190 bird species and 100 mite genera, fungal material (spores and hyphae) was the most prevalent type of gut content (GLMM: v² = 168.73, df = 2, p < .001; Figure 1) and the most morphologically diverse (GLMM: v² = 442.5, df = 2, p < .001; Figure 1). In addition, diatoms and plant material were also found, but in a much lower frequency and morphotype diversity than fungi ( Figure 1). Highly similar results were found when only analysing passerines ( Figure S1 and S2), the avian order in which bird species were also studied using DNA metabarcoding (see below). The overall predominance of fungi was widespread across the avian phylogeny ( Figure 2) and feather mite taxonomy (Table 1).

| DNA metabarcoding of feather mites' diets
Metabarcoding results of the mite species from the genera Proctophyllodes Robin, 1877, Trouessartia Canestrini, 1899, Dolichodectes Park & Atyeo, 1971, and Scutulanysuss Mironov, 1985 highly congruent results with the microscopic analyses in terms of the prevalence and diversity of food resources, while complementing them with bacterial detection and providing taxonomic detail of the organisms involved. We found bacterial DNA in all samples (Table S2). The bacterial genera identified primarily belonged to the phyla Proteobacteria, Actinobacteria and Bacteroidetes, with Proteobacteria being the most frequently represented ( Figure S5).
Within these phyla, we retrieved a high diversity of bacterial genera (Figures 3, S7 and S8). Genera commonly found in soil and as environmental "background noise" such as Sphingomonas, Acinetobacter and Pseudomonas were the most prevalent genera ( Table 2,  between bird species (F = 11.84, p = .001). In addition, ordination and profile plots did not show clustering by either mite or bird species in bacterial OTUs and genera ( Figures S8 and S9).
We found fungal DNA in all infrapopulations except one (

B u l w e r i a b u l w e r i i C a l o n e c t r i s d i o m e d e a P u f f in u s p u f f in u s P h a la c r o c o r a x a u r it u s G a v ia p a c if ic a G a v ia im m e r P h a p it r e r o n le u c o t is P ti li n o p u s s u p e r b u s P e r d ix p e r d ix P h a s ia n u s c o lc h ic u s B o n a s a u m b e ll u s L a g o p u s la g o p u s D e n d ra g a p u s c a n a d e n s is A le c tu ra la th a m i A n a s d is c o rs A n a s cl yp e a ta A na s am er ic an a An as pl at yr hy nc ho s Ox yu ra jam aic en sis An ser alb ifro ns Cyg nus colu mbi anu s Nothu ra macu losa Rhynchotu s rufescens
F I G U R E 2 50% majority-rule consensus phylogenetic tree depicting the distribution of food resources retrieved by microscopic analysis of feather mite gut contents across the phylogeny of birds. In brief, 1,000 trees were obtained from BirdTree ( Chaumont, 1999;Gunderson, 2008;Marchisio, Curetti, Cassinelli, & Bordese, 1991;Nwadiaro, Ogbonna, Wuyep, & Adekojo, 2015) in feather mite guts. Whether the quantities of bacteria and fungi eaten by feather mites are enough to increase host fitness requires further study. Altogether, our results support previous evidence on the commensalistic-mutualistic role of vane-dwelling feather mites (Blanco et al., 1997(Blanco et al., , 2001Galv an et al., 2012;Proctor, 2003;Walter & Proctor, 2013a,b,c). Thus, vane-dwelling feather mites probably should no longer be considered to be parasites of birds (e.g., Harper, 1999) but rather commensalists-mutualists. This does not apply to the few taxa of quill-dwelling feather mites that clearly feed on feather pith (e.g., Ascouracaridae) or those that live on or in the epidermis of the host (e.g., Dermationidae, Epidermoptidae) (Gaud & Atyeo, 1996;Proctor, 1999 Relative abundance of bacterial genera F I G U R E 3 Stacked bar plots of the bacterial genera retrieved in the molecular analyses of mite species. Low abundance taxa (<2%) were not shown for illustrative purposes [Colour figure can be viewed at wileyonlinelibrary.com] for a rapid integration of this knowledge into bird-related practices, such as those in wild bird conservation programmes. Also, our results suggest that further studies of birds in farms, zoos and the pet trade are needed, where traditionally feather mites were viewed as parasites, with birds provided with treatment using acaricides (e.g., Alekseev, 1998;Salisch, 1989). This practice not only has the downside of monetary expense because of the use of acaricides, but could also result in the loss of the potential services provided by feather-cleaning mites, as our results suggest.
Analyses of the bacterial and fungal DNA found in the guts of feather mites revealed a high diversity of taxa that were not structured by host or by mite species (Figures 4, 6 and S6-S9). This suggests trophic opportunism of mites (da Silva, Dorrestein, & Quinn, 2015; Kent & Burtt, 2016), which may graze upon whatever food resources might be available at the time. This opportunistic "feathercleaning" feeding behaviour is also supported by the large amount of unidentifiable items we found in the guts and by the higher abundance and diversity of fungi found in the mite samples in comparison with the external samples (e.g., Figures S3 and S10). Overall, many other species of sarcoptiform mites, including many free-living Astigmata, are functionally defined as fungivore-microbivore-detritivores T A B L E 2 Prevalence and abundance (mean; minimum-maximum) statistics from the 30 most prevalent fungal and bacterial genera retrieved by DNA metabarcording. The three genera which were, on average, most abundant for each taxon, are asterisked and highlighted in bold. Relative abundance was calculated as the % of sequences of the given genus in those samples where the genus was found Fungi Prevalence (% of samples) | 211 evidence has been found for the utilization of bacteria as a food source in free-living astigmatan species (Erban & Hubert, 2008, 2010Hubert, Nesvorna, Kopeck y, S agov a-Mare ckov a, & Poltronieri, 2014;Hubert et al., 2016). In these studies, microbiomes composed of highly diverse taxa in low abundance have been interpreted as evidence for microbivory. In contrast, microbiome profiles showing a low diversity of highly abundant taxa are interpreted as evidence of symbiotic or pathogenic bacterial species (Hammer, Janzen, Hallwachs, Jaffe, & Fierer, 2017;Hubert et al., 2016). In this way, the prevalence-abundance patterns of the bacteria found here (Table 2) suggest a combination of bacteria used as food resource (mostly environmental-associated genera, which were more prevalent but less abundant, for example, Sphingomonas and Acinetobacter;  Table 2).

Relative abundance (% sequences within samples)
Lack of a stable "microbiome" across different individuals of a given species has been found in other organisms with a nutritionally broad diet (Shapira, 2016). In contrast, species with highly biased diets, such as lice feeding on bird feathers (mainly keratin) or termites feeding on dead wood (mainly cellulose), typically have permanent and relatively stable endosymbiotic bacteria which provide them essential vitamins or other nutritional supplements (Puchta, 1955;Ohkuma, 2008;Perotti, Kirkness, Reed, & Braig, 2009;Boyd et al., 2016; but see Hammer et al., 2017). Thus, our results suggesting the lack of a stable microbiome at the mite species level add support to the hypothesis of a generalist fungivore-microbivore-  (2014) and Denef, Fujimoto, Berry, and Schmidt (2016) [Colour figure can be viewed at wileyonlinelibrary.com] detritivore diet for the feather mites reported here, instead of these resources being taken as a by-product of a diet based mostly on uropygial oil (Engel & Moran, 2013;Sanders et al., 2017;Shapira, 2016). In fact, in 42% of the mites in which we detected any food resource, we did not see any oil globules (but see Materials and Methods) also suggesting that resource intake does not depend on oil ingestion.
A further understanding of the multilayered hologenome (i.e., to distinguish between stable-unstable, adapted-unadapted bacterial taxa, Shapira, 2016) through large-scale microbiome-oriented studies will help in disentangling the role of these potentially symbiotic or pathogenic bacteria of feather mites. Furthermore, whether feather mites select among available food resources (fungal preferences have been found in free-living fungivorous Astigmata, Hubert et al., 2003;Hubert, Jarosık, Mourek, Kubatova, & Zdarkova, 2004) or do not need to rely on bacterial symbionts requires further experimental study. Lastly, a hypothesis of an "external-rumen" mode of feeding, in which mites ingest predigested food (by bacteria), has been also supported in free-living astigmatan mites (Hubert et al., 2014(Hubert et al., , 2016 and would be also compatible with our results. Feather mite species are relatively host-specific and (presumably) of switching to new host species (Doña, Proctor, et al., 2017;Doña, Sweet, et al., 2017;Gaud 1992;Klimov, Mironov, & O'Connor, 2017;Matthews et al., 2018). These switches mostly involve closely related hosts, but major-host switches (e.g., between bird orders) have been revealed as a major driver of their diversification (Doña, Proctor, et al., 2017). As for many other host-symbiont systems (Clayton, Bush, & Johnson, 2016;Nylin et al., 2017), understanding the (co)eco-evolutionary scenario of host-switching in this host-sym- F I G U R E 6 Principal coordinates analysis (PCoA) of fungal communities of feather mite infrapopulations: First row, samples coloured by mite species and (a) based on Bray-Curtis and (b) Jaccard distances, respectively; second row, samples coloured by bird species and (c) based on Bray-Curtis and (d) Jaccard distances, respectively. OTUs counts were scaled to that of the smallest library following McMurdie and Holmes (2014) and Denef et al. (2016) [Colour figure can be viewed at wileyonlinelibrary.com] context-dependent (possibly even occasionally parasitic) relationship between vane-dwelling feather mites and birds (Blanco et al., 2001).
In particular, future studies should investigate the following. (i) Using appropriate and sensitive methods such as HPLC, test whether uropygial gland oil is part of the diet of feather mites. A comparative exploration of the diet of feather mites inhabiting birds with vestigial uropygial gland that produce powder down would be also useful. If uropygial oil is a large component of vane-dwelling feather mites, it would be then important to test whether removal of the oil affects bird fitness. (ii) Investigate whether the diet of feather mites differs along the annual cycle of birds (e.g., migration, moult). (iii) Examine the potential aerodynamic costs of harbouring different quantities of feather mites. (iv) Determine effects of feather mites on host fitness as mediated by other ectosymbionts (e.g., feather lice). (v) Test whether an experimental increase in feather mites' abundance increases, decreases or has no overall effect on host fitness. Lastly, (vi) examine whether experimental variation in feather mites abundance has a context-dependent (e.g., under different environmental conditions) effect on host fitness over time.