Tripartite associations between Afrotropical bats, eukaryotic parasites, and microbial symbionts
Lutz HL1,2,3, Gilbert JA1,2 , Dick CW,3,4
1 Department of Pediatrics, University of California San Diego, La Jolla, California, USA
2 Scripps Institution of Oceanography, University of California San Diego, La Jolla, California, USA
3 Integrative Research Center, Field Museum of Natural History, Chicago, Illinois, USA
4 Department of Biology, Western Kentucky University, Bowling Green, Kentucky, USA
*Corresponding Author
Holly L. Lutz
Department of Pediatrics
University of California San Diego
9500 Gilman Drive
La Jolla, CA 92161
hllutz@health.ucsd.edu
ABSTRACT
Skin is the largest mammalian organ and the first defensive barrier against the external environment. The skin and fur of mammals can host a wide variety of ectoparasites, many of which are phylogenetically diverse, specialized, and specifically adapted to their hosts. Among hematophagous dipteran parasites, volatile organic compounds (VOCs) are known to serve as important attractants, leading parasites to compatible sources of blood meals. VOCs have been hypothesized to be mediated by host-associated bacteria, which may thereby indirectly influence parasitism. Host-associated bacteria may also influence parasitism directly, as has been observed in interactions between animal gut microbiota and malarial parasites. Hypotheses relating bacterial symbionts and eukaryotic parasitism have rarely been tested among humans and domestic animals, and have to our knowledge never been tested in wild vertebrates. In this study, we use Afrotropical bats, hematophagous ectoparasitic bat flies, and haemosporidian (malarial) parasites vectored by bat flies as a model to test the hypothesis that the vertebrate host microbiome is linked to parasitism in a wild system. We identify significant correlations between bacterial community composition of the skin and dipteran ectoparasite prevalence across four major bat lineages, as well as striking differences in skin microbial network characteristics between ectoparasitized and non-ectoparasitized bats. We also identify links between the oral microbiome and presence of malarial parasites among miniopterid bats. Our results support the hypothesis that microbial symbionts may serve as indirect mediators of parasitism among eukaryotic hosts and parasites.
Keywords: microbiome, bats, Chiroptera, Hippoboscoidea, malaria, Haemosporidia, Afrotropics
1. INTRODUCTION
Animals are capable of hosting myriad biologically interdependent symbionts including viruses, bacteria, archaea, and eukarya. Many associations between eukaryotic parasites and hosts have ancient origins (1, 2), and mounting evidence suggests that bacterial symbionts may be responsible for mediating host-parasite interactions in ways that could ultimately shape host evolution (3). For example, studies of human and anthropophilic mosquito interactions have found that the human skin microbiome can influence mosquito feeding preference, thereby affecting transmission patterns of vector-borne pathogens (e.g. WNV, yellow fever, dengue, malaria, etc.), and potentially imposing selective pressures on human populations (4). Conversely, parasites may influence the relative abundance of host-associated microbes in ways that facilitate parasite transmission, as has been observed in rodent models of Plasmodium transmission in which parasitism positively correlates with abundance of skin-associated microbes that produce volatile organic compound attractive to arthropod vectors ofPlasmodium (5-7). Despite the potential evolutionary significance of interactions between animal hosts, microbial symbionts, and eukaryotic parasites, such interactions have not been well studied in wild vertebrates.
Bats (Mammalia: Chiroptera) are an ideal system for examining the interactions between microbial symbionts, eukaryotic parasites, and hosts. Bats are among the most speciose orders of mammals (second only to the Rodentia), providing a diverse comparative phylogenetic framework for hypothesis testing, and they harbor a great diversity of eukaryotic parasites such as dipteran insects, haemosporidia, and helminths (8-11). Bats are also associated with microbial pathogens of importance for human health (e.g. Bartonella , Pasteurella , SARS coronaviruses, filaviruses, rabies, Hendra and Nipah viruses (12-15)), and serological surveys have supported the role of Afrotropical bats as reservoirs for a number of viruses (16-18), making our understanding of factors regulating parasite and pathogen transmission all the more relevant. Inter- and intraspecific transmission of pathogens among bats and other animals is an area of increasing concern in light of recent zoonotic pandemics. Bat flies (Arthropoda: Hippoboscoidea), which are obligate blood-feeding parasites of bats, are known to transmit several pathogens of human relevance (19, 20) and also are the primary hosts for bat-specific malarial parasites (Apicomplexa: Haemosporida). Bat flies are typically host-specific (21, 22) and prevalence can vary widely within a single species across different geographic locations (23). Malarial parasites vectored by bat flies are also typically host-specific but are observed less frequently in many bats relative to ectoparasitic bat flies.
Bat flies are nutritionally dependent on their hosts and maintain contact throughout most of their lives, living in fur and on skin membranes and leaving their hosts only for brief reproductive periods (24). How these parasites maintain host-specificity with bats over evolutionary time is unknown, but blood protein and immunocompatibility between host and parasite are thought to play a role (21, 25). The proximal mechanism by which parasites locate “preferred” hosts is also unknown, but as has been observed in other hematophagous parasites (e.g. mosquitoes (26-28), tse-tse flies (29)), may involve host-associated chemical cues. Chemical cues can be produced in a number of ways, including via metabolic processes involving bacteria in the gut, oral cavity, or on the skin (30, 31).
Here, we build on previous broad-scale studies of Afrotropical bat-associated microbes (32, 33) to characterize parasitism by bat flies and haemosporidia in four widespread lineages of bats. Using these data, we test the hypothesis that the bat microbiome (skin, oral, and gut) is linked to parasitism by two obligate, host-specific eukaryotic parasites.
2. METHODS & MATERIALS
2.1 Host and parasite sampling
Sampling was conducted at sixteen field sites in Kenya and Uganda from August to October 2016, using a combination of mist-netting and hand-netting. Samples were taken from bats collected as voucher specimens for biodiversity inventories, allowing for extensive post-mortem sampling of skin and fur from multiple points on the body. These included three biopsies from wing membrane, one from tail membrane, one from the ear, one from the interscapular region of the back, and one from the interclavicular region of the chest (Figure 1) using 3mm sterile disposable biopsy punches (IntegraTMMiltex®). Biopsy samples from each individual were combined and stored in sterile 95% ethanol. Whole tongues were collected (from apex to root) and stored in 95% ethanol for oral microbiome analysis. We collected ~2-4mL of blood from euthanized bats via cardiac puncture. Blood was placed on Whatman FTA cards for nucleic acid extraction, and 2-3 blood films per individual were prepared for microscopic analyses. The remainder of each blood sample was stored in cryovials placed in lN2. Following euthanasia, skin, and blood sampling, bats were fumigated in ethyl acetate for 15 minutes and then examined for ectoparasites. Presence or absence of dipteran parasites was noted, and parasites were collected into 95% ethanol for taxonomic identification. Bat flies were identified morphologically by examination under magnification using a Leica MZ16 stereozoom microscope.  Collected specimens were compared to relevant taxonomic keys, to descriptions in the alpha-literature, and to reference collections of the Field Museum of Natural History, Chicago and the Bernice P. Bishop Museum, Honolulu.
Malarial parasite presence and taxonomic identity were determined as described in Lutz et al. (2015). In brief, DNA was extracted from whole blood using Qiagen DNeasy (Qiagen, Valencua, CA) and screened for the presence of malarial parasites using triplicate PCR and Sanger sequencing confirmation, followed by BLASTn to confirm parasite taxonomy. All sampling was conducted in accordance with the Field Museum of Natural History IACUC. Host and parasite vouchers are accessioned at the Field Museum of Natural History (Chicago, IL, USA) (Table S1;S2).
2.2 Microbial DNA extraction, library preparation, and data generation
DNA was extracted using the MoBio PowerSoil 96 well soil DNA isolation kit (catalog no. 12955-4; MoBio, Carlsbad, CA, USA) following the standard Earth Microbiome Project protocol (http://www.earthmicrobiome.org/). PCR amplification and sequencing were performed as previously described in Lutz et al. (2019). Amplicon sequence variants (ASVs) were identified using Deblur following standard demultiplexing and quality filtering using the Quantitative Insights into Microbial Ecology pipeline (QIIME2) (34). Skin, oral, and gut libraries were rarefied to an even read depth of 5000 reads, 1000 reads, and 1000 reads per library, respectively, based on rarefaction curve estimates. All 16S rRNA sequence data are publicly available via the QIITA platform (https://qiita.ucsd.edu) under the study identifier (ID) 11815 and the European Bioinformatics Institute (EBI) under accession number PRJEB32520; additional sequence library information is provided in Table S2. Code for sequence processing and analyses can be viewed at https://github.com/hollylutz/BatMP.
2.3 Statistical analyses
Alphadiversity and betadiversity analyses were performed using the programming language R (35) and packages vegan2.4-2 (36), phyloseq (37), dplyr (38), and ggplot2 (39). Differences in mean alphadiversity measures (observed richness and Shannon index metrics) between ectoparasitized and non-parasitized bats within families were assessed using the Kruskal-Wallis test. PERMANOVA tests for differences in betadiversity were performed using the adonis2 function (R package vegan2.4-2 (36), with 1000 permutations.
We evaluated differences among skin associated microbial ASVs between parasitized and non-parasitized bats grouped at the host family level by ranking multinomial regression coefficients (hereafter referred to as ranked differentials). This approach, implemented using the program Songbird (40), relies on estimated centered log ratios of features between sample groupings and thereby surpasses the need for absolute measures of feature differentiation between groupings. Songbird multinomial regressions were run for 100,000 epochs with a batch size of 3, minimum feature count of 5, a learning rate of 1e-5, and a differential prior of 0.50. Ranked differentials were visualized using the program Qurro (41).
To examine whether skin microbial communities differ in stability and structure between parasitized and non-parasitized bats, we reconstructed skin microbial networks using the R package Sparse Inverse Covariance Estimation for Ecological Association Inference (SPIEC-EASI) (42). All network datasets were filtered to contain only ASVs that appeared in at least three individuals within each respective dataset and consisted of skin microbial libraries grouped by host family and ectoparasite status. Network results produced with SPIEC-EASI were summarized using the R packages CAVnet (43) and igraph (44). Network stability was assessed by sequentially removing network nodes (ordered by betweenness centrality and degree) and observing natural connectivity (i.e. eigenvalue of the graph adjacency matrix) as nodes are removed.
3. RESULTS
3.1 Microbiome, ectoparasite, and malarial parasite sampling and detection.
We sampled 283 individuals representing eight species from four chiropteran families (Hipposideridae, Miniopteridae, Rhinolophidae, and Pteropodidae). Rarefaction and quality filtering of 16S rRNA libraries resulted in the retention of 237 skin samples (29,270 ASVs, rarefied to 5000 reads), 202 oral samples (3,361 ASVs, rarefied to 1000 reads), and 230 gut samples (5,771 ASVs, rarefied to 1000 reads) for microbiota profiling (Table 1). Hippoboscoid ectoparasites were recovered from all host taxa, with an average prevalence of 51% (SD±13%). Malarial parasitism was restricted almost entirely to the family Miniopteridae, within which prevalence ranged from 47-65% (mean 53%±11%). All malarial parasites observed in miniopterids belonged to the haemosporidian genus Polychromophilus , and shared 99-100% sequence similarity to Cytochrome b lineages previously identified in Kenyan and Uganda miniopterid bats (11). We observed only two non-miniopterid individuals of the species Rhinolophus clivosus acrotis (Rhinolophidae) to be positive for malarial parasites, which belonged to the genus Nycteria and exhibited 98% sequence similarity to a Cytochrome b lineage previously identified in Uganda. All other bats were negative for haemosporidia by molecular and microscopic analyses (Table S1).
3.2 Associations between the bat microbiota and ectoparasitism.
No differences were observed between skin, oral, or gut microbial alphadiversity and ectoparasitized or non-ectoparasitized bats at the host family or species levels (p > 0.05, Kruskal-Wallis) (Figure 2). However, subtle but significant differences in skin-associated bacterial betadiversity were observed between parasitized and non-parasitized bats for both weighted and unweighted UniFrac metrics (p < 0.005, PERMANOVA). Differences were only observed among the gut and oral microbiota using unweighted but not weighted UniFrac metric (p < 0.002, PERMANOVA) (Table 2), suggesting that differences were driven by the compositional variance of rare taxa.
Multinomial regression analysis of the skin microbiome using Songbird found models that included ectoparasite status of hosts significantly outperformed null models (Miniopteridae pseudo Q2 = 0.12, Hipposideridae pseudo Q2 = 0.28, Rhinolophidae pseudo Q2 = 0.17, Pteropodidae pseudo Q2 = 0.26), allowing us to identify a number of ASVs potentially associated with ectoparasitism (Figure 3). The bacterial order Actinomycetales exhibited the greatest number of ASVs that were differentially abundant between parasitized and non-parasitized bats. Indeed, eight of fourteen ASVs found to be consistently associated with the presence or absence of ectoparasites in all four host families belonged to the order Actinomycetales, with the remaining six belonging to the order Bacillales (phylum Firmicutes) and orders Burkholderiales, Pseudomonadales, Rhizobiales, and Sphingomonadales (phylum Proteobacteria) (Table 3).
Network analyses revealed striking differences in the topology and stability of the skin microbiome in parasitized versus non-parasitized bats, revealing a significant decrease in cluster size (p< 0.05, Mann-Whitney-Wilcoxon rank sum test) and median node degree (p < 0.05, t test), as well as a significant reduction in network connectivity (p < 0.05, t test) for parasitized bats from three of the four bat families examined (pteropodids being the exception) (Figure 4).
3.3 Host microbiome and haemosporidian parasitism
Of the bat taxa sampled, only species belonging to the family Miniopteridae exhibited malarial parasitism adequate for statistical analysis. We observed no differences in alphadiversity of the skin, oral, and gut microbiota of bats based on malarial infection status but identified significant differences between unweighted UniFrac betadiversity of the oral microbiota between malarial and non-malarial bats (p < 0.002, PERMANOVA) (Table 4). Multinomial regression analyses identified a number of bacterial ASVs associated with malarial parasitism. Two ASVs exhibiting the greatest proportional increase in malaria positive bats belonged to the species Pantoea agglomerans and the genus Acinetobacter . ASVs most strongly associated with absence of malarial parasites belonged to the family Pasteurellaceae (Figure 5).